Evidence of Latency and Reactivation of Both Herpes Simplex Virus... and HSV-2 in the Genital Region

Evidence of Latency and Reactivation of Both Herpes Simplex Virus (HSV)-1
and HSV-2 in the Genital Region
Gina Sucato, Anna Wald, Eiko Wakabayashi,
Jeffrey Vieira, and Lawrence Corey
Departments of Pediatrics, Medicine, and Laboratory Medicine,
University of Washington, and Fred Hutchinson Cancer Research
Center, Program in Infectious Diseases, Seattle, Washington
While superinfection with different herpes simplex virus (HSV) types has been demonstrated in
animals, the ability of the two HSV types to colonize and reactivate in the same anatomic region
in humans has not been well demonstrated. In 6 patients, both HSV-1 and HSV-2 was recovered
from genital lesions. In 4 of them, who initially acquired genital HSV-1 infection, subsequent HSV2 infection presented as a prolonged episode of genital lesions and a marked increase in the frequency
of genital recurrences. While most of the subsequent clinical reactivations were HSV-2, in 2 patients
the recurrence rate of genital HSV-1 increased after the acquisition of HSV-2. These data demonstrate the ability of a second HSV type to infect the same anatomic region and illustrate the
difference in reactivation frequency of the two types in the same person. Typing of HSV isolates
may be useful in persons with recent alteration in recurrence rates of genital HSV.
In genital infection, herpes simplex virus (HSV) maintains
latency in sacral ganglia and reactivates in response to various
ill-defined triggers. As such, most recurrent genital herpes is
thought to result from reactivation of latent disease [1, 2].
In animal models, experimental exogenous reinfection with
different types or subtypes of HSV has been reported. However,
the subsequent natural history of such dual infection of sacral
or trigeminal nerve root ganglia is difficult to assess in animal
models. Although reports of genital reinfections with a second
type or another strain of HSV have been described in people
[3 – 7], the ability of these second infections to result in subsequent clinical recurrences is unclear. Herein we describe 6
patients who sequentially acquired both HSV-1 and HSV-2 in
their genital tracts.
of recurrences, and culture results. Serologic samples were typed
with the HSV Western blot [7]. Virus isolates were confirmed and
typed by monoclonal antibody. Restriction endonuclease analysis
of HSV DNA was done as described [8], with the following modification. HSV-2–infected cells were washed once with PBS, pelleted in a 1.5-mL microfuge tube, resuspended in 0.3 mL of a
buffer containing 10 mM Tris (pH 8.0) and 10 mM EDTA, and
then lysed by the addition of 0.3 mL of a solution containing 10
mM Tris (pH 8.0), 10 mM EDTA, and 1.2% SDS. Viral DNA was
digested overnight with SalI or XhoI, and the DNA was analyzed
by electrophoresis with a 0.6% agarose gel. After electrophoresis,
the gel was stained with SYBR Green (FMC BioProducts, Rockland, ME) and photographed under UV illumination.
The patients were drawn from a cohort of 468 patients with firstepisode and 1212 patients with recurrent genital herpes followed at
the University of Washington Virology Research Clinic between
1981 and 1994. We selected patients who had both HSV-1 and
HSV-2 isolated from the genital tract at some time during their
follow-up. We reviewed the medical records of these patients and
extracted data on clinical manifestations of genital herpes, pattern
Received 17 June 1996; revised 29 October 1997.
Financial support: NIH (AI-30731; MH-1890 to G.S.; AI-31448 to A.W.);
Benjamin and Mary Siddons Measey Foundation (to G.S.); American Social
Health Association (to A.W.).
Reprints or correspondence: Dr. Anna Wald, University of Washington
Virology Research Clinic, 1001 Broadway, Suite 320, Seattle, WA 98122
([email protected]).
The Journal of Infectious Diseases 1998;177:1069–72
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
Initial genital HSV infection. We identified 6 patients who
had both HSV-1 and HSV-2 isolated from their genital tract;
5 of the 6 developed genital HSV-2 after acquisition of HSV-1
infection (table 1). All patients were healthy and received no
immunosuppressive medications. Among the 5 patients with
genital HSV-1, 4 were evaluated in clinic at their initial acquisition of HSV-1. All 4 had primary genital HSV-1 documented
by isolation of HSV-1 from genital lesions and seroconversion
to HSV-1 only. Follow-up of the 4 patients with primary genital
HSV-1 revealed symptomatic recurrences in 2 of them. Consistent with previous findings [9], the recurrence rate was low,
ranging from 0 to 0.27/month. Two patients experienced no
symptomatic recurrences of genital HSV-1 until the acquisition
of genital HSV-2, 1 with ú24 months of follow-up.
Patient 5 was HSV-1 – seropositive and had a history of genital herpes for 10 years with infrequent recurrences. Patient 6
was seropositive for HSV-1 and -2 at entry and reported a
7-year history of frequent recurrences of genital herpes. During
UC: J Infect
Concise Communications
JID 1998;177 (April)
Table 1. Serologic, clinical, and virologic pattern of HSV infection in 6 patients with both genital HSV-1 and HSV-2 infection.
HSV isolate from
genital lesion
HSV Serology
At entry and
after first episode
Clinical course of
initial HSV infection
to HSV-1
HSV-1, then
HSV-1 &
HSV-1 &
to HSV-1
HSV-1 &
HSV-1 &
to HSV-1
HSV-1 &
HSV-1 &
in evolution to
HSV-1 &
HSV-1 &
HSV-1 &
HSV-1 & HSV-2
HSV-1 &
Primary genital HSV1; 40 penile lesions;
symptoms for 15
days; no recurrences
in subsequent 2
Primary oral and
perianal HSV-1; 25
perianal, 1 oral
lesion; local and
systemic symptoms
for 21 days; 4 (1
oral, 3 anal)
recurrences in 15
Primary genital HSV1; 29 bilateral
vulvar and perianal
lesions with local
and systemic
symptoms for 13
days; no recurrences
in 2 months
Primary genital HSV1; 17 bilateral
vulvar lesions;
symptoms for 9
days; 1 recurrence
in 20 months
10-year history of
genital HSV-1; 15
recurrences in 11.3
7-year history of
genital HSV; 6
recurrences in 8
months; HSV-2
isolated once
first and
presentation of
second genital
HSV infection
Clinical course of dual
2 years
6 unilateral penile
lesions; pain
for 25 days;
HSV-2 cultured
12 recurrences in 16
months; 5 culture-positive
recurrences (HSV-1, 1;
HSV-2, 3; HSV-1 & -2,
3.5 years
3 perianal lesions;
symptoms for
23 days; HSV2 cultured
12 recurrences (all perianal)
in 3.5 years; 10 culturepositive recurrences
(HSV-1, 1; HSV-2, 9)
2 months
6 unilateral
vulvar lesions;
symptoms for
20 days; HSV2 cultured
5 recurrences in 40 months;
4 culture-positive
recurrences (HSV-1, 2;
HSV-2, 2)
1.7 years
3 bilateral vulvar
symptoms for 4
days; HSV-2
Patient lost to follow-up
11.3 years
4 vulvar lesions;
HSV-2 cultured
7.7 years
6 unilateral
lesions on
vulva; pain for
6 days; HSV-1
14 recurrences in 27
months; 8 culture-positive
recurrences (HSV-2, 6;
HSV-1 & 2, 2)
12 recurrences in 22
months; 1 culture-positive
recurrence (HSV-2)
NOTE. M, man; W, woman.
the course of follow-up, episodes of reactivation of genital
HSV-1 and genital HSV-2 were documented.
Acquisition of second genital HSV infection. The time
from acquisition of the first to the second genital HSV infection ranged from 2 to 136 months. In all 5 patients with
prior genital HSV-1, the episode of the new HSV-2 infection
was observed in the clinic and was confirmed by virus isolation and subsequent seroconversion to HSV-2. No patient
reported systemic symptoms with the initial episode of genital HSV-2. Among the 4 patients who were evaluated in
/ 9d43$$ap48
03-04-98 19:03:33
clinic during both the first episode of genital infection with
HSV-1 and the subsequent first episode with HSV-2, fewer
lesions were present during the infection with the second
serotype than during the initial infection (median, 4.5 lesions
vs. 28.5 lesions). However, in 3 of these patients, the first
genital HSV-2 episode lasted ú20 days. Five patients reported new sex partners before presenting with infection
with a second type of HSV.
Four of the 5 patients for whom follow-up data were available had subsequent recurrences during which the second
UC: J Infect
JID 1998;177 (April)
Concise Communications
infecting virus was isolated. In 2 patients whose frequency
of recurrences was documented at least 2 years before and
2 years after the second HSV infection, the number of recurrences increased substantially after acquisition of genital
HSV-2. For example, patient 1 had no recurrences with
HSV-1 in 24 months but experienced 12 recurrences in 16
months after acquiring HSV-2. Of the 5 recurrences that
yielded positive cultures, 3 were positive for HSV-2, 1 for
HSV-1 and -2, and 1 for HSV-1 only. Patient 5 had reported
15 recurrences of HSV-1 in ú11 years. After acquiring
HSV-2, she had 14 recurrences in 27 months. HSV was
isolated during 8 recurrences: HSV-2 in 6 episodes and HSV1 and -2 in 2 episodes.
Patient 6, who was HSV-1 – and -2 – seropositive at the
beginning of follow-up, had a recurrence rate of 0.75/month
with isolation of HSV-2 on one occasion in the 5 months
before isolation of genital HSV-1. The initial episode of
genital HSV-1 appeared clinically similar to recurrent HSV2, and we cannot tell whether it represented genital acquisition of HSV-1 in a person with prior oral herpes or reactivation of previously acquired genital HSV-1 infection. The
patient continued to have frequent recurrences (recurrence
rate, 0.55/month) in the subsequent 22 months.
To differentiate whether the infection with HSV-2 resulted
from reactivation of the initial strain or reinfection with new
strains, we performed restriction endonuclease analyses on
strains from patients 1, 3, and 5. In patient 3, all 5 isolates
analyzed revealed identical restriction enzyme pattern. Patient 1 had 2 strains of HSV-2 identified. One strain was
shed during 3 separate recurrences 10 months apart; another
strain with different restriction endonuclease pattern was
recovered during an intercurrent reactivation. Patient 3 had
restriction endonuclease analysis done on 11 HSV-2 isolates
obtained over a 17-month period, and 3 different HSV-2
strains were noted. Figure 1 shows the SalI restriction endonuclease digestion pattern of 6 HSV-2 isolates from patient
3, illustrating the 3 different strains, designated A, B, and
Figure 1. Restriction endonuclease analysis of HSV-2 isolates
from patient 3. Molecular weight
markers are indicated at left (in
kb). Arrows at right indicate positions of some DNA fragments distinguishing different isolates. Collection dates of isolates: lane 1, 8
May 1990; lane 2, 23 October
1990; lane 3, 11 November 1991;
lane 4, 27 March 1991; lane 5, 12
July 1991; lane 6, 25 October
1991. Isolates in lanes 1, 3, and 6
are designated strain A; lanes 2 and
5, strain B; lane 4, strain C.
/ 9d43$$ap48
03-04-98 19:03:33
C. The 5 other HSV-2 isolates examined from patient 3 were
all type A. Both strains A and B were reactivated: strain A
was detected in 5 separate episodes and strain B in 2 separate
During a primary infection, HSV infects nerve endings, establishes latency in the sacral nerve root ganglia, and subsequently reactivates in response to a variety of triggers to cause
recurrent clinical disease. Animal studies have indicated that
colonization of a ganglion with HSV protects from establishment of infection by subsequent HSVs [10]. In people, the best
evidence for this is the infrequent finding of ú1 HSV-2 strain
among isolates from infected persons. The ability of a second
herpesvirus to establish a latent neural infection has been addressed in animal models. Some data suggest that despite the
appearance of symptoms or viral shedding, genital reinfection
does not lead to neural reinfection [10, 11]. However, other
authors have shown that a reinfecting strain or type of HSV
can establish a second latent infection in a ganglion [12, 13].
Landry and Zibello [14] infected guinea pigs with HSV-1 or
-2 and then superinfected the animals at the same site with
the heterologous serotype of HSV; cocultivation subsequently
demonstrated that the superinfecting virus frequently established latency in the ganglia.
Several case reports describe genital reinfection with a second distinct HSV type [3 – 6]. However, only one of the case
reports has demonstrated that the second infecting virus established latency [6]. Infection with HSV-1 offers partial protection from HSV-2 acquisition; it is possible that prior genital
HSV-1 is more protective than prior oral HSV-1 infection. In
this report we describe 5 patients initially infected genitally
with HSV-1 who subsequently acquired genital HSV-2 and
experienced clinically recurrent HSV-2 infection. Patient 6 also
recurred clinically with both subtypes, but we could not establish the order of acquisition of the two viruses.
We were surprised to find ú1 strain, as distinguished by
restriction endonuclease analysis, in 2 of 3 patients with available isolates. This clearly differs from published reports, including our own previous studies, that show that infection with
multiple strains is rare [1 – 3, 15]. Our patients may represent
unusually susceptible hosts, despite appearing clinically immunocompetent. Alternatively, they may have acquired a second
HSV-2 infection in rapid succession, before complete development of immunity.
Perhaps our most interesting finding was the increased reactivation rate among persons who acquired HSV-2 subsequent to HSV1 infection. Our patients who were reinfected with HSV-2 (for
whom follow-up data are available) subsequently established patterns of recurrence typical of HSV-2 infection [9]. In all patients
with both genital HSV-1 and -2, HSV-2 was subsequently isolated
more frequently from genital lesions, as expected given that genital HSV-2 is likely to recur more frequently than genital HSV-1.
UC: J Infect
Concise Communications
Moreover, we found 2 patients who had no symptomatic recurrences of HSV-1 before infection with HSV-2 but who had recurrences with HSV-1 after acquisition of HSV-2. Thomas et al.
[12], using a murine model, found that reinfection with HSV-2
would reactivate a latent HSV-1 infection. It is interesting that
the reactivation rates of the two viruses in the same ganglia in
the same host differ, suggesting that the virologic behavior associated with reactivation or the host responses to the two HSV types
do differ.
These cases illustrate the importance of determining the serotype of a genital HSV infection. As HSV-2 recurs more frequently than HSV-1, serotyping has prognostic value and informs patients of expected frequency of future recurrences.
Moreover, typing may be useful in persons with genital HSV1 who present with a prolonged episode of genital herpes or
who markedly alter their recurrence rates, as that may indicate
superinfection with HSV-2. Many patients with genital herpes
do not practice safer sex if their partner also has genital herpes;
our data suggest that they should if the infections are caused
by different herpesvirus serotypes. Whether infection with ú1
HSV-2 strain occurs frequently in a subset of patients requires
further study.
1. Schmidt OW, Fife KH, Corey L. Reinfection is an uncommon occurrence
in patients with symptomatic recurrent genital herpes. J Infect Dis 1984;
149:645 – 6.
2. Lakeman AD, Nahmias AJ, Whitley RJ. Analysis of DNA from recurrent
genital herpes simplex virus isolates by restriction endonuclease digestion. Sex Transm Dis 1986; 13:61 – 6.
3. Buchman TG, Roizman B, Nahmias AJ. Demonstration of exogenous
genital reinfection with herpes simplex virus type 2 by restriction endonuclease fingerprinting of viral DNA. J Infect Dis 1979; 140:295 – 304.
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
4. Gerson M, Portnoy J, Hamelin C. Consecutive infections with herpes
simplex virus types 1 and 2 within a three-week period. J Infect Dis
1984; 149:655.
5. Kit S, Trkula D, Qavi H, et al. Sequential genital infections by herpes
simplex viruses types 1 and 2: restriction nuclease analyses of viruses
from recurrent infections. Sex Transm Dis 1983; 10:67 – 71.
6. Samarai AM, Shareef AA, Kinghorn GR, Potter CW. Sequential genital
infections with HSV-1 and 2. Genitourin Med 1989; 65:39 – 41.
7. Ashley RL, Militoni J, Lee F, Nahmias A, Corey L. Comparison of Western blot (immunoblot) and glycoprotein G – specific immunodot enzyme
assay for detecting antibodies to herpes simplex virus types 1 and 2 in
human sera. J Clin Microbiol 1988; 26:662 – 7.
8. Kuzushima K, Kudo T, Kimure H, et al. Prophylactic oral acyclovir in
outbreaks of primary herpes simplex virus type 1 in a closed community.
Pediatrics 1992; 89:379 – 83.
9. Lafferty WE, Coombs RW, Benedetti J, Critchlow C, Corey L. Recurrences after oral and genital herpes simplex virus infection. N Engl J
Med 1987; 316:1444 – 9.
10. Centifanto-Fitzgerald YM, Varnell ED, Kaufman HE. Initial herpes simplex virus type 1 infection prevents ganglionic superinfection by other
strains. Infect Immun 1982; 35:1125 – 32.
11. Stanberry LR, Bernstein DI, Kit S, Myers MG. Genital reinfection after
recovery from initial genital infection with herpes simplex virus type 2
in guinea pigs. J Infect Dis 1986; 153:1055 – 61.
12. Thomas E, Lycke E, Vahlne A. Retrieval of latent herpes simplex virus
type 1 genetic information from murine trigeminal ganglia by superinfection with heterotypic virus in vivo. J Gen Virol 1985; 66:1763 – 70.
13. Meigner B, Norrild B, Roizman B. Colonization of murine ganglia by a
superinfecting strain of herpes simplex virus. Infect Immun 1983; 41:
702 – 8.
14. Landry ML, Zibello TA. Ability of herpes simplex virus (HSV) types 1
and 2 to induce clinical disease and establish latency following previous
genital infection with the heterologous HSV type. J Infect Dis 1988;
158:1220 – 6.
15. Sakaoka H, Aomori T, Gouro T, Kumamoto Y. Demonstration of either
endogenous recurrence or exogenous reinfection by restriction endonuclease cleavage analysis of herpes simplex virus from patients with
recrudescent genital herpes. J Med Virol 1995; 46:387 – 96.
UC: J Infect
Isolation of a Second Recombinant Human Respiratory Syncytial Virus
Monoclonal Antibody Fragment (Fab RSVF2-5) that Exhibits Therapeutic
Efficacy In Vivo
James E. Crowe, Jr., Page S. Gilmour, Brian R. Murphy,
Robert M. Chanock, Lingxun Duan, Roger J. Pomerantz,
and Glenn R. Pilkington
Laboratory of Infectious Diseases, National Institute of Allergy and
Infectious Diseases, National Institutes of Health, Bethesda, Maryland;
Dorrance Hamilton Laboratories, Center for Human Virology,
Division of Infectious Diseases, Department of Medicine,
Jefferson Medical College, Philadelphia, Pennsylvania;
Intracel Corporation, Issaquah, Washington
A second human respiratory syncytial virus (RSV) – neutralizing monoclonal antibody was isolated
and its binding site was identified. Fab F2-5 is a broadly reactive fusion (F) protein – specific recombinant Fab generated by antigen selection from a random combinatorial library displayed on the
surface of filamentous phage. In an in vitro plaque-reduction test, the Fab RSVF2-5 neutralized the
infectivity of a variety of field isolates representing viruses of both RSV subgroups A and B. The
Fab recognized an antigenic determinant that differed from the only other human anti-F monoclonal
antibody (RSV Fab 19) described thus far. A single dose of 4.0 mg of Fab RSVF2-5/kg of body
weight administered by inhalation was sufficient to achieve a 2000-fold reduction in pulmonary
virus titer in RSV-infected mice. The antigen-binding domain of Fab RSVF2-5 offers promise as
part of a prophylactic regimen for RSV infection in humans.
Respiratory syncytial virus (RSV) is the major cause of serious viral lower respiratory tract illness in infants and young
children worldwide, especially in children with underlying cardiopulmonary disease, such as those with bronchopulmonary
dysplasia [1]. We are pursuing the development of a panel of
human antibodies of different specificities for use as a cocktail
for treatment or prophylaxis of high-risk persons. The binding
site of a new antibody, Fab RSVF2-5, was sought in the hopes
of obtaining a second broadly reactive human neutralizing antibody directed to a different site on the RSV F glycoprotein,
providing the basis for further development of a prophylactic
RSV human antibody cocktail for use in high-risk patients.
Materials and Methods
Polymerase chain reaction amplification of antibody genes, library construction, and cloning and expression of Fab molecules.
A phage library displaying randomly combined antibody heavy
and light chains obtained from peripheral blood mononuclear cells
of an adult donor was constructed as described [2]. Phage displaying surface Fab binding to RSV (Long strain)–infected cell
Received 28 March 1997; revised 21 October 1997.
All animal experiments were performed according to National Institute of
Allergy and Infectious Diseases Guidelines.
DNA sequences are lodged with GenBank (accession nos. L41061 and
Reprints or correspondence (present address): Dr. James E. Crowe, Jr.,
Vanderbilt University Medical Center, D-7235 MCN, 1161 21st Ave. S., Nashville, TN 37232-2581 ([email protected]).
The Journal of Infectious Diseases 1998;177:1073–6
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
lysate (abV Immune Response, Derry, NH) was enriched from a
packaged phagemid library of 107 clones by four rounds of panning
against proteins of the RSV-infected cell lysate bound to 96-well
microtiter plates under conditions as described [2]. Amplification
of the eluted phage between each round and the final preparation of concentrated purified Fab preparations was done as described [2].
Determination of protein specificity and Fab binding constant.
Antibodies in purified Fab preparations binding to RSV F or G
glycoproteins were quantitated in a previously described ELISA
procedure [3] by use of F or G glycoprotein that had been immunoaffinity-purified from RSV subgroup A (Long strain)–infected
cell lysates as described [4, 5]. Binding constant determinations
were done by use of RSV-infected cell lysate bound to ELISA
plates at 0.1 mg of protein/mL and blocked with 3% bovine serum
albumin (BSA)–PBS. Serial 2-fold dilutions of purified human
Fab in 1% BSA-PBS were added to wells (50 mL/well), incubated,
and washed. Goat anti-human F(ab*)2 –alkaline phosphatase conjugate (Pierce, Rockford, IL) and p-nitrophenyl phosphate solution
were then added. The binding constants were determined as the
Fab concentration (grams per liter) at 50% binding divided by the
approximate molecular weight of the Fab (5 1 1004 g).
Nucleotide sequence determination of the gene encoding Fab
RSVF2-5. Sequence analysis of double-stranded plasmid DNA
purified from large-volume overnight bacterial cultures with a
commercial plasmid maxiprep kit (Qiagen, Chatsworth, CA) was
done on an automated DNA sequencer (373A; Applied Biosystems, Foster City, CA), with a Taq fluorescent dideoxy-terminator
cycle-sequencing kit (Applied Biosystems) and previously described primers [2].
Assays of RSV-neutralizing antibodies. Neutralizing activity
of the soluble Fab preparations was measured by plaque reduction
[6], using HEp-2 cell cultures and the A2 strain of RSV or 20
different RSV field isolates representing viruses of both subgroups
A and B collected over 30 years from infected humans on several
UC: J Infect
Concise Communications
continents [7]. Plaques were detected by an immunoperoxidase
labeling procedure, as described [3]. The specific activity of the
Fab was calculated as the concentration (micrograms per milliliter)
required to effect 60% reduction of plaque number. In addition,
the Fab was assayed with the plaque-reduction test for activity
against a well-characterized collection of monoclonal antibody–
resistant mutants generated by use of murine RSV-neutralizing
antibodies [8, 9]. The C4848 mutant (also designated v324) was
derived previously from the A2 strain of RSV by use of a murine
monoclonal antibody designated MAb 19 [9]. This mutant and a
similarly passaged RSV A2 virus stock were kindly supplied by
Geraldine Taylor (AFRC Institute for Animal Health, Berkshire,
Therapeutic studies in mice. BALB/c mice (42 weeks old,
female, respiratory pathogen–free) were obtained from the National Cancer Institute (Frederick, MD). These animals weighed
on average 25 g each at the time of therapy. Mice were inoculated
intranasally with 107.0 pfu of RSV strain A2 suspended in 100 mL
of tissue culture medium. Inoculation was done while mice were
under light anesthesia with inhaled methoxyflurane (Metofane; Pitman-Moore, Mundelein, IL). Under these conditions, liquid materials inoculated intranasally are delivered predominantly into the
lungs by direct aspiration. On day 4 after infection, 100 mL of Fab
suspension was instilled as a single dose intranasally under light
methoxyflurane anesthesia, as above. Eighteen hours after treatment with Fab (on day 5 after virus inoculation), mice from each
treatment or control group were sacrificed by carbon dioxide inhalation. Nasal turbinates (NT) and lungs were harvested separately,
homogenized in 3 mL of Hanks’ balanced salt solution with sucrose, clarified, and frozen as described [3, 10]. NT and lung
homogenates were titrated for RSV by plaque assay on HEp-2
cells as described [11]. Plaques were visualized with RSV antibody
staining by the immunoperoxidase procedure as above. Plaque titer
was calculated as log10 pfu/g of tissue, with a minimum level of
detection of 1.7 in the lungs and 2.0 in the NT.
Isolation of RSV-specific human Fab RSVF2-5. Fabs were
selected from an IgG1 l/k random combinatorial library displayed on the surface of filamentous DNA phage [2]. The
library was derived from an adult donor who had a 1:640 serum
ELISA titer against the RSV-infected cell lysate. After the
library of 107 clones was panned against RSV-infected cell
lysate four times, eight clones were isolated that produced
soluble Fab binding to the RSV-infected cell lysate. One of
these Fab clones, designated RSVF2-5, was specific for the
RSV lysate and did not react in ELISA with BSA, recombinant
HIV-1 gp120, or recombinant hepatitis B virus (HBV) surface
protein (end-point titer, §1000-fold lower against these other
proteins [data not shown]). The protein expression level for
the Fab RSVF2-5 varied from 500 to 700 mg of purified protein/
L of culture. DNA sequences were determined for the Fd and
light chain and submitted to GenBank (accession nos. L41061
and L41062). Analysis of translated DNA sequences, DNA
insert size, and molecular weights of purified expressed pro-
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
teins on SDS-PAGE gels (data not shown) revealed RSVF2-5
to consist of an Fd fragment (VH3 gene family) and a light
chain (VL6 gene family).
Characterization of Fab RSVF2-5 in vitro. The approximate binding constant for the human anti-RSV Fab RSVF2-5
was determined to be 8.7 1 1009 M from ELISA titration of
purified Fab. Each of several purified preparations of Fab
RSVF2-5 neutralized the RSV A2 wild type strain in vitro,
with an average specific activity of 0.05 mg/mL, to effect 60%
plaque reduction. An identically prepared purified preparation
of the human HBV Fab c-41 (directed against HBV core antigen) had no effect on RSV plaque formation (specific activity,
ú115 mg/mL). Fab RSVF2-5 neutralized with high efficiency
each of 10 subgroup A and 10 subgroup B RSV strains isolated
over a 30-year interval [7], indicating that the Fab is directed
against a highly conserved site on a viral protein of RSV strains
of antigenic subgroups A and B. The range of specific activity
against subgroup A strains was 0.2 to 3 mg/mL and against
subgroup B strains was 0.3 to 2.6 mg/mL.
Determination of antigen-binding site. The specificity of
Fab RSVF2-5 for the RSV F glycoprotein was demonstrated
by ELISA binding with purified RSV F or G glycoprotein. A
1-mg/mL purified preparation of Fab RSVF2-5 exhibited a titer
of ú1:16,384 to F glycoprotein, while a similarly prepared
suspension of an HBV Fab (HBV Fab c-41, also 1 mg/mL,
with a titer of 1:64 against HBV core antigen) had a titer of
1:2 in the same test. This preparation of Fab RSVF2-5 did not
bind purified G glycoprotein, as evidenced by an ELISA titer
of õ1:2, while an RSV-positive adult control serum exhibited
a titer of 1:4096 in the same test. A more detailed analysis of
the binding site of Fab RSVF2-5 was achieved by testing for its
ability to neutralize a well-characterized collection of antibodyresistant mutants generated using murine RSV F glycoprotein –
neutralizing monoclonal antibodies [8, 9] (Crowe J, et al., unpublished data). The antibody fragment neutralized all mutants
tested, except for the C4848 mutant (table 1), which differs
from its A2 parent by a single amino acid at position 429 [9].
The only other previously described human monoclonal RSV
antibody (RSV Fab 19) neutralizes the C4848 mutant with high
efficiency, indicating that its binding site differs from that of
Fab RSVF2-5 (Crowe J, unpublished data).
Therapeutic efficacy of Fab RSVF2-5. Fab RSVF2-5 and
the HBV Fab c-41 control antibody were tested in mice for
therapeutic efficacy 4 days after infection, which is the height
of RSV replication in the lungs under these experimental conditions [12]. As little as 0.0625 mg of Fab RSVF2-5/kg of body
weight was active therapeutically in mice, and 4.0 mg of Fab
RSVF2-5/kg of body weight had the greatest therapeutic effect:
The titer of RSV in the lungs of these mice was reduced by a
factor of 2 1 103 (table 2). The identically prepared control
Fab HBV c-41 failed to reduce the titer of RSV in the lungs
or NT of infected mice.
We evaluated the possibility that the reduction in titer of
RSV in the lungs observed in these studies was due to neutral-
UC: J Infect
JID 1998;177 (April)
Concise Communications
ization in vitro, when the lung homogenates were prepared or
assayed for infectivity in cell culture. Individual lung or NT
homogenates of 3 untreated, infected mice were prepared on
day 5 after infection, and each was mixed separately with an
equal volume of homogenate derived from a different uninfected mouse that had received 4.0 mg of Fab RSVF2-5/kg the
day prior. The difference in geometric mean titer of the 2
groups was only 0.3 log10 pfu/mL for the NT and 0.2 log10
pfu/mL for the lungs (consistent with the 1:2 dilution of the
infected homogenate with the uninfected homogenate). This
indicated that the therapeutic effect observed in Fab RSVF25 recipients was due to an effect of the Fab in vivo and not
due to an in vitro artifact. Rebound in virus titer following
initial reduction due to single-dose Fab treatment of RSV infection in BALB/c mice was demonstrated previously for the RSV
Fab 19 but was not tested for in these experiments.
Table 2. Intranasal instillation of Fab RSVF2-5 at height of RSV
infection in BALB/c mice causes significant decrease in amount of
virus in respiratory tract.
* Groups of 6 animals were infected intranasally with 7.0 log10 pfu of RSV
strain A2 in 100-mL inoculum under light methoxyflurane anesthesia on day
0, then treated intranasally on day 4 after infection (height of infection) with
indicated dose of Fab in 100 mL of sterile PBS. Tissue homogenates were
obtained for virus quantitation on day 5 (18 h after treatment), frozen, then
titered on HEp-2 cell monolayers.
P õ .001 vs. log10 pfu/g of nose in PBS group.
Virus was recovered from 1 of 6 animals at titer of 2.6 log10 pfu/g (detectable level, 1.7 log10 pfu/g). Means were calculated using 1.7 for 5 animals
without detectable virus.
The high level of in vitro specific activity of Fab RSVF2-5
against all field isolates tested, and its high level of therapeutic
efficacy in vivo, suggest that the antigen-binding domain of
this antibody fragment holds promise for use in humans. A
full-length antibody molecule incorporating Fab RSVF2-5 is
currently under construction. The isolation of Fab RSVF2-5
allows the evaluation of the efficacy of an antibody cocktail
containing several broadly reactive antibodies of more than one
specificity. Fab RSVF2-5 failed to neutralize the monoclonal
Table 1. Site on RSV F glycoprotein required for binding of Fab
RSVF2-5 is defined by its failure to neutralize monoclonal antibody –
resistant mutant.
A2 wild type
Specific activity
(mg/mL required for
60% plaque reduction)
AA change at
* All mutants were derived from RSV wild type strain A2.
As defined by competition binding assays [8].
Per Crowe J, et al., unpublished data.
C4848 is mutant resistant to RSV murine monoclonal antibody (MAb19),
which differs from RSV A2 wild type parent virus at single amino acid on F1
subunit of F glycoprotein. It is also designated v324 [9].
Per [9].
/ 9d43$$ap48
03-04-98 19:03:33
Material used to treat,*
dose (mg of Fab/kg of
body weight)
Fab RSVF2-5
HBV Fab c-41
Titer of virus recovered from tissue
homogenate (mean log10 pfu/g
of tissue { SE)
Nasal turbinates
4.7 { 0.14
4.7 { 0.10
5.3 { 0.11
5.3 { 0.11
antibody escape mutant C4848, which differs from its neutralizable RSV A2 parent by a single amino acid change (Arg to
Ser) at amino acid position 429 of the RSV F glycoprotein [9].
This site probably represents the epitope to which the murine
MAb 19 and the human Fab RSVF2-5 both bind, although it
is possible that the site of mutation in the C4848 escape mutant
represents a change at another site that alters the conformation
of the epitope to which MAb 19 or Fab RSVF2-5 binds. The
importance of the residue 429 for binding of both mouse and
human neutralizing antibodies indicates that random pairings
of the heavy and light chains generated using combinatorial
libraries can produce antibodies that are similar to naturally
occurring antibodies. The shared site also supports the concept
of the evolutionary conservation of the ability to recognize
virus-neutralizing epitopes. The epitope recognized by the only
other previously described human RSV F antibody, RSV Fab
19 [7, 12], differs from that of Fab F2-5 because the human
Fab 19 neutralizes the C4848 RSV mutant that Fab RSVF2-5
does not neutralize, and the Fab 19 selects for antibody escape
mutants that possess a mutation at amino acid 266 on the RSV
F protein (Crowe J, et al., unpublished observation).
1. Collins PL, McIntosh K, Chanock RM. Respiratory syncytial virus. In:
Fields BN, Knipe DM, Howley PM, eds. Fields virology. 3rd ed. New
York: Lippincott-Raven Press, 1996:1313 – 51.
UC: J Infect
Concise Communications
2. Pilkington GR, Duan L, Zhu M, Keil W, Pomerantz R. Recombinant
human Fab antibody fragments to HIV-1 Rev and Tat regulatory proteins: direct selection from a combinatorial phage display library. Mol
Immunol 1996; 33:439 – 50.
3. Murphy BR, Sotnikov AV, Lawrence LA, Banks SM, Prince GA. Enhanced pulmonary histopathology is observed in cotton rats immunized
with formalin-inactivated respiratory syncytial virus (RSV) or purified F
glycoprotein and challenged with RSV 3 – 6 months after immunization.
Vaccine 1990; 8:497 – 502.
4. Walsh EE, Brandriss MW, Schlesinger JJ. Purification and characterization of
the respiratory syncytial virus fusion protein. J Gen Virol 1985;66:409–15.
5. Walsh EE, Schlesinger JJ, Brandriss MW. Purification and characterization
of GP90, one of the envelope glycoproteins of respiratory syncytial
virus. J Gen Virol 1984; 65:761 – 7.
6. Coates HV, Alling DW, Chanock RM. An antigenic analysis of respiratory
syncytial virus isolates by a plaque reduction neutralization test. Am J
Epidemiol 1966; 83:299 – 313.
7. Barbas CF, Crowe JE Jr, Cababa D, et al. Human monoclonal Fab fragments derived from a combinatorial library bind to respiratory syncytial
virus F glycoprotein and neutralize infectivity. Proc Natl Acad Sci USA
1992; 89:10164 – 8.
JID 1998;177 (April)
8. Beeler JA, van Wyke Coelingh K. Neutralization epitopes of the F glycoprotein of respiratory syncytial virus: effect of mutation upon fusion
function. J Virol 1989; 63:2941 – 50.
9. Arbiza J, Taylor G, Lopez JA, et al. Characterization of two antigenic
sites recognized by neutralizing monoclonal antibodies directed against
the fusion glycoprotein of human respiratory syncytial virus. J Gen
Virol 1992; 73:2225 – 34.
10. Prince GA, Jenson AB, Horswood RL, Camargo E, Chanock RM. The
pathogenesis of respiratory syncytial virus infection in cotton rats. Am
J Pathol 1978; 93:771 – 91.
11. Crowe JE Jr, Collins PL, London WT, Chanock RM, Murphy BR. A
comparison in chimpanzees of the immunogenicity and efficacy
of live attenuated respiratory syncytial virus (RSV) temperaturesensitive mutant vaccines and vaccinia virus recombinants that express the surface glycoproteins of RSV. Vaccine 1993; 11:
1395 – 404.
12. Crowe JE Jr, Murphy BR, Chanock RM, Williamson RA, Barbas CF III,
Burton DR. Recombinant human RSV monoclonal antibody Fab is
effective therapeutically when introduced directly into the lungs of respiratory syncytial virus – infected mice. Proc Natl Acad Sci USA 1994;
91:1386 – 90.
Bronchoalveolar Interferon-a, Tumor Necrosis Factor-a, Interleukin-1, and
Inflammation during Acute Influenza in Pigs: A Possible Model for Humans?
Kristien Van Reeth, Hans Nauwynck,
and Maurice Pensaert
Laboratory of Veterinary Virology, Faculty of Veterinary Medicine,
University of Ghent, Merelbeke, Belgium
Biologically active interferon-a, tumor necrosis factor-a (TNF-a), and interleukin-1 (IL-1) were
detected in bronchoalveolar lavage (BAL) fluids of 3-week-old cesarian-derived colostrum-deprived
pigs inoculated with H1N1 influenza virus. Cytokine titers and lung virus titers were significantly
higher 18 – 24 h after inoculation than at 48 – 72 h after inoculation in all 4 litters of pigs examined.
All three cytokines were positively correlated with a 3- to 4-fold increase in BAL cell numbers (P
õ .036) and with a drastic neutrophil infiltration (24% – 77% of BAL cells vs. 0 – 1.5% in controls)
(P õ .001). In addition, cytokine production coincided with the onset of general and respiratory
symptoms of influenza and with the development of a necrotizing bronchopneumonia. This study
is the first demonstration of TNF-a and IL-1 in BAL fluids of a natural influenza virus host.
It documents that pigs may be a highly valuable experimental model in human influenza virus
Swine influenza (SI) is a highly important respiratory disease
of pigs. The causative viruses are type A influenza viruses of
H1N1 and H3N2 subtype, which are antigenically related to
human influenza viruses. Typical outbreaks involve an abrupt
Received 21 July 1997; revised 17 October 1997.
Presented in part: 4th International Congress of Veterinary Virology, Edinburgh, Scotland, August 1997.
Financial support: Ministry of Agriculture, Brussels.
Reprints or correspondence: Dr. Kristien Van Reeth, Laboratory of Veterinary Virology, Faculty of Veterinary Medicine, Salisburylaan 133, B-9820
Merelbeke, Belgium ([email protected]).
The Journal of Infectious Diseases 1998;177:1076–9
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
onset of fever, anorexia, tachypnea, dyspnea, and coughing.
Considerable economic losses result from growth retardation
or weight loss. Experimental viral infections of pigs have documented massive viral replication in lung epithelial cells, accompanied by polymorphonuclear leukocyte infiltration and epithelial degeneration [1]. However, the exact mechanisms by which
SI virus produces lung pathology and disease have not been
During respiratory virus infections, interferon-a (IFN-a),
tumor necrosis factor-a (TNF-a), and interleukin-1 (IL-1) are
among the first cytokines to be produced. These three cytokines
have been demonstrated in bronchoalveolar washings from influenza virus – infected mice [2 – 4]. TNF-a and IL-1 play preeminent roles in neutrophil migration to sites of infection and
in neutrophil activation. Both cytokines can induce fever, an-
UC: J Infect
JID 1998;177 (April)
Concise Communications
phragmatic lung were collected for standard histopathology, influenza virus titrations, and fluorescent antibody stainings [7]. The
right lung was lavaged with 60 mL of PBS. Bronchoalveolar lavage
(BAL) cells were separated by centrifugation at 400 g and counted,
and cytospin preparations were stained with DiffQuik (Baxter,
Du¨dingen, Switzerland) to determine percentage of neutrophils.
Cytokine bioassays. Cell-free BAL fluids were concentrated
10 times by dialysis against 20% polyethylene glycol (molecular
weight, 20,000) and cleared of residual virus by ultracentrifugation
at 100,000 g. Bioassays were done with 2-fold dilutions of BAL
fluid samples in 96-well microtiter plates. IFN-a activity was determined in a cytopathic effect reduction assay by use of MDBK
cells and vesicular stomatitis virus [8]. Percentage of cytopathic
effect was determined spectrophotometrically after staining with
neutral red. The number of units of IFN-a activity per milliliter
was defined as the reciprocal of the dilution producing 50% inhibition of cytopathic effect. IFN-a specificity was demonstrated by
neutralization of samples with rabbit antiserum against recombinant porcine IFN-a (gift from C. La Bonnardie`re, Jouy en Josas,
France). TNF-a was assayed as cytotoxic activity in PK(15) subclone 15 cells (gift from G. Bertoni, Bern, Switzerland) in the
presence of actinomycin D [9]. The plates were stained with crystal
violet and read spectrophotometrically. The number of units of
TNF-a per milliliter was defined as the reciprocal of the dilution
producing 50% cytotoxicity. TNF-a specificity was established
by neutralization with rabbit anti-human TNF-a (Innogenetics,
Zwijndrecht, Belgium). IL-1 was measured as proliferative activity
in D10(N4)M cells in the presence of concanavalin A and recombinant human IL-2 [10]. Percentage of proliferation was quantitated
by the MTT method, and optical densities were measured. The
orexia, and weight loss, particularly if produced at higher levels
[5]. Evidence for a role of TNF-a and IL-1 in influenza virus
pathogenesis and disease is growing [4, 6].
The pathogenesis of human influenza has been studied almost exclusively in volunteers and in small laboratory animals.
Pigs, unlike mice and guinea pigs, are natural influenza virus
hosts, and their lungs resemble those of humans in many physiologic aspects. They may thus represent a highly valuable experimental model.
Here we studied the production of IFN-a, TNF-a, and IL-1
in the lungs of pigs and its relation with disease or inflammation
during the acute stage of influenza virus infection. Furthermore,
we assessed the usefulness of pigs as a biomedical model in
human influenza virus research.
Materials and Methods
Experimental design, virologic examinations, and lung inflammatory parameters. Four litters (18 total) of 3-week-old cesarianderived colostrum-deprived (CDCD) pigs were used. They were
inoculated intratracheally with 107.0 EID50 of influenza virus
(H1N1 A/Sw/Belgium/1/83), third passage in embryonated eggs.
Control pigs were left either uninoculated or inoculated with PBS
or with sterile allantoic fluid (table 1). Inocula contained õ1.25
endotoxin units/mL in the limulus assay. Pigs were euthanized at
18, 24, 48, or 72 h after inoculation with virus or at 24 h after
inoculation with PBS or allantoic fluid. Gross lung lesions were
scored by visual inspection. Samples from the left apical and dia-
Table 1. Clinical response and lung virus titers, bronchoalveolar lavage (BAL) cellular changes, and cytokine titers of individual pigs after
H1N1 influenza virus infection.
Virus titer
(log 10 EID50/g of lung)
BAL cells
BAL cytokine titers
(biologic units/mL)
(h after inoculation)
Clinical symptoms
(h after inoculation)*
Allantoic fluid
Allantoic fluid
17 –
24 –
24 –
18 –
22 –
24 –
20 –
22 –
22 – 36
18 –
NOTE. Individual data from pigs belonging to 4 different litters (groups) are given. — , no cytokine response detected; neg, negative.
* Where range has no ending value, symptoms persisted until euthanasia.
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
Concise Communications
number of units of IL-1 per milliliter was defined as the reciprocal
of the dilution producing 50% of the maximal proliferation level
of D10 cells. To demonstrate IL-1 specificity, monoclonal rat antimouse IL-1 receptor type 1 antibodies (Genzyme, Cambridge, MA)
were included in the assay. Laboratory standards were run in each
bioassay and used for correction of titers of samples. Samples were
tested in three (IFN-a, TNF-a) or two (IL-1; duplicates of samples)
independent bioassays, and geometric mean titers of the corrected
values were calculated.
Statistical analysis. When comparing cytokine profiles at set
times across all groups, Kruskal-Wallis rank sum tests were used
on individual cytokine bioassay data. When comparing 2 groups,
standard two-sample Mann-Whitney tests were used. Spearman
rank correlation tests were used to compare geometric mean cytokine responses per animal and other parameters. Finally, standard
two-sample t tests were used to consider virus titers. P õ .05 was
considered significant.
Clinical responses, influenza virus titers, BAL cell numbers,
percentage of neutrophils, and cytokine titers of individual pigs
are summarized in table 1.
Clinical symptoms. Control pigs remained healthy. In influenza virus – inoculated pigs, lethargy, shivering, anorexia,
tachypnea, and labored abdominal respiration developed between 18 and 24 h after inoculation. Recovery started between
48 and 72 h after inoculation. Pigs in group 1 were most
severely affected.
Influenza virus replication. Control pigs tested negative for
virus. Virus titers and immunofluorescence scores were similar
in the 4 influenza virus – inoculated groups (P ú .577). Titers
in apical and diaphragmatic lung lobes were significantly higher
18 – 24 h after inoculation than at 48 – 72 h after inoculation (P
õ .016 and .008, respectively). Fluorescence was evident in
all sections examined. Eighteen and 24 h after inoculation,
bronchi/bronchioli and alveoli had, respectively, 90% and 30%
of their epithelial cells fluorescing. By 48 – 72 h, more of the
alveolar tissue became involved.
Lung inflammatory changes. Control pigs did not have
macroscopic or microscopic lung pathology. BAL cell numbers
were between 35 and 60 1 106. Fewer than 1.5% of cells were
neutrophils; ú95% had macrophage morphology.
After influenza virus inoculation, gross lung lesions appeared
between 48 and 72 h after inoculation. At that time, Ç85%,
18%, 18%, and 8% of lung tissue was affected in groups 1, 2,
3, and 4, respectively. On histopathology, bronchi/bronchioli
and, to a lesser degree, alveoli showed epithelial necrosis and
massive neutrophil infiltration at 18 – 24 h after inoculation.
Forty-eight and 72 h after inoculation, bronchioli and alveoli
were filled with exudate containing necrotic debris and macrophages and only few neutrophils. Histologic changes were most
dramatic in group 1. BAL cell numbers were between 112 and
160 1 106 at 18 – 24 h after inoculation and consisted of a
maximum of 56% – 77% neutrophils in groups 1, 2, and 3. In
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
group 4, cell numbers did not exceed 43 1 106 and neutrophil
infiltration was only 24%.
Cytokine production. While controls were negative, IFNa, TNF-a, and IL-1 were detected in the 4 virus-infected
IFN-a was detected from 18 through 72 h after inoculation,
with maximal levels between 18 and 24 h. Peak IFN titers were
176,600, 140,400, and 337,400 U/mL in groups 1, 2, and 3,
respectively. In group 4, IFN titers peaked at only 27,800 U.
TNF-a was detected at 18 or 24 h (or both) after inoculation
in all 4 groups. By 48 and 72 h after inoculation, TNF-a was
found only in pigs from group 1. Peak TNF titers were 370,
225, and 175 U/mL in groups 1, 2, and 3, respectively, but
only 25 U/mL in group 4.
IL-1 activity was found 18 or 24 h (or both) after inoculation
in all groups. Maximum IL-1 titers were 535, 245, 520, and
150 U/mL in the 4 respective groups. Only in group 1 pigs
was IL-1 also found at 72 h after inoculation.
Levels of the three cytokines were significantly higher 18 –
24 h after inoculation than at 48 – 72 h after inoculation (P õ
.016 for all three cytokines). Significant correlations were noted
between production of all three cytokines and BAL cell numbers (P õ .036 for each) and percentage of neutrophils (P õ
.001 for each). Cytokine production coincided with the onset
of clinical disease and development of bronchopneumonia.
At the start of this study, it was unknown whether the lungs
of 3-week-old CDCD pigs are fully capable of TNF-a and
IL-1 production. Thus far, TNF-a and IL-1 have only been
demonstrated in the lungs of pigs 6 – 8 weeks old and older,
either conventional [11] or gnotobiotic [12]. In our study, the
use of CDCD pigs was required, since we regularly detect
TNF-a and/or IL-1 in BAL fluids from conventional pigs in
the absence of experimental viral infections. The age of 3 weeks
was selected for practical reasons. Because sanitary status and
age may influence cytokine production, we performed a preliminary experiment in 3-week-old CDCD pigs. Two pigs were
inoculated intratracheally with 17 mg of Escherichia coli O111:
B4 lipopolysaccharide, a known potent TNF-a and IL-1 inducer. BAL fluids collected 6 and 12 h after inoculation revealed TNF titers of 184 and 128 U/mL and IL-1 titers of 564
and 596 U/mL in the respective bioassays. Consequently, 3week-old CDCD pigs were found suitable for further influenza
virus – cytokine studies.
To our knowledge, this is the first demonstration of influenza
virus – induced TNF-a and IL-1 in BAL fluids of a natural virus
host. Interestingly, influenza virus was an equally effective
inducer of TNF-a and IL-1 as was E. coli endotoxin. TNF-a
and IL-1 have remarkably overlapping and synergistic effects,
several of which are consistent with clinicopathologic manifestations of influenza [5]. Our findings in pigs are in agreement
with previous reports in mice [3, 4] and further substantiate the
UC: J Infect
JID 1998;177 (April)
Concise Communications
hypothesis that TNF-a and IL-1 contribute to clinicopathologic
effects of influenza. First of all, both cytokines were positively
correlated (r ú .879) with neutrophil recruitment to the lungs,
and peak cytokine production coincided with the onset of clinical disease. Furthermore, there was a clear association between
individual TNF levels on the one hand and the extent of neutrophil infiltration and severity of lung pathology on the other.
Group 1 pigs, which had consistently higher values of TNF
than the other groups, showed most severe disease and lesions.
Group 4 pigs, with barely detectable TNF levels, also had lower
neutrophil infiltration than the other groups.
Interferon-a was detected at extremely high titers in some
pigs. As for TNF-a and IL-1, IFN-a production was tightly
correlated (r ú .879) with neutrophil infiltration to the lungs,
and peak IFN production coincided with the appearance of
illness. These findings support the idea that interferons contribute to constitutional and other effects during influenza. Indeed,
IFN-a is intrinsically pyrogenic [13] and can mediate neutrophil migration and stimulation. Besides, it has been shown in
vitro that IFN-a may enhance the neutrophil respiratory burst
to influenza virus [14]. Given the excessive amounts of IFNa in pig BAL fluids, IFN could considerably add to the pyrogenic and inflammatory effects of TNF-a and IL-1.
In group 4, overall cytokine production was lower than in
the other groups. Most striking was the virtual lack of TNF
production. Differences in cytokine production between this
and previous groups cannot be attributed to differences in viral
multiplication. One possibility is that genetic components contribute to decreased cytokine production, since group 4 pigs
belonged to a genetically different line. It would be worthwhile
to examine a larger number of genetically different pigs for
their cytokine responses on influenza infection and clinical
Although influenza viruses in humans primarily infect the
upper respiratory tract, influenza pneumonia yearly causes high
mortality in patients with cardiopulmonary disease and in the
elderly. Experimental research thus remains of high priority,
and commonly used animal models have some limitations. In
ferrets, for example, upper respiratory tract infection predominates, and alveolar infection and pneumonitis are less pronounced [15]. Mice and guinea pigs, on the other hand, are
not natural hosts, and influenza virus strains have to be adapted
to them. Also, mice show a fall in body temperature instead of
fever and, with more virulent strains, the infection is invariably
lethal. In this study, pigs — as natural hosts — have shown to
be valuable animal models for research on human influenza.
/ 9d43$$ap48
03-04-98 19:03:33
We thank R. Ducatelle for help with histology and D. J. Shaw
for statistics.
1. Haesebrouck F, Pensaert MB. Effect of intratracheal challenge of fattening
pigs previously immunised with an inactivated influenza H1N1 vaccine.
Vet Microbiol 1986; 11:239 – 49.
2. Wyde PR, Wilson MR, Cate TR. Interferon production by leukocytes
infiltrating the lungs of mice during primary influenza virus infection.
Infect Immun 1982; 38:1249 – 55.
3. Vacheron F, Rudent A, Perin S, Labarre C, Quero AM, Guenounou M.
Production of interleukin 1 and tumour necrosis factor activities in
bronchoalveolar washings following infection of mice by influenza virus. J Gen Virol 1990; 71:477 – 9.
4. Peper RL, Van Campen H. Tumor necrosis factor as a mediator of inflammation in influenza A viral pneumonia. Microb Pathog 1995; 19:175 –
5. Le J, Vilcek J. Tumour necrosis factor and interleukin 1: cytokines with
multiple overlapping biological activities. Lab Invest 1987; 56:234 – 48.
6. Kozak W, Zheng H, Conn CA, Soszynski D, Van der Ploeg LHT, Kluger
MJ. Thermal and behavioral effects of lipopolysaccharide and influenza
in interleukin-1b – deficient mice. Am J Physiol 1995; 269:969 – 77.
7. Van Reeth K, Pensaert MB. Porcine respiratory coronavirus – mediated
interference against influenza virus replication in the respiratory tract
of feeder pigs. Am J Vet Res 1994; 55:1275 – 81.
8. La Bonnardie`re C, Laude H. High interferon titer in newborn pig intestine
during experimentally induced viral enteritis. Infect Immun 1981; 32:
28 – 31.
9. Bertoni G, Kuhnert P, Peterhans E, Pauli U. Improved bioassay for the
detection of porcine tumor necrosis factor using a homologous cell line:
PK(15). J Immunol Methods 1993; 160:267 – 71.
10. Hopkins SJ, Humphreys M. Simple, sensitive and specific bioassay of
interleukin-1. J Immunol Methods 1989; 120:271 – 6.
11. Baarsch MJ, Scamurra RW, Burger K, Foss DL, Maheswaran SK, Murtaugh MP. Inflammatory cytokine expression in swine experimentally
infected with Actinobacillus pleuropneumoniae. Infect Immun 1995;
63:3587 – 94.
12. Asai T, Okada M, Ono M, et al. Increased levels of tumor necrosis factor
and interleukin 1 in bronchoalveolar lavage fluids from pigs infected
with Mycoplasma hyopneumoniae. Vet Immunol Immunopathol 1993;
38:253 – 60.
13. Dinarello CA, Cannon JG, Wolff SM, et al. Tumor necrosis factor
(cachectin) is an endogenous pyrogen and induces production of interleukin 1. J Exp Med 1986; 163:1433 – 50.
14. Little R, White MR, Hartshorn KL. Interferon-a enhances neutrophil respiratory burst responses to stimulation with influenza A virus and FMLP.
J Infect Dis 1994; 170:802 – 10.
15. Smith H, Sweet C. Lessons for human influenza from pathogenicity studies
with ferrets. Rev Infect Dis 1988; 10:56 – 74.
UC: J Infect
Long-Lasting Remission of Cytomegalovirus Retinitis without Maintenance
Therapy in Human Immunodeficiency Virus–Infected Patients
C. Tural, J. Romeu, G. Sirera, D. Andreu, M. Conejero,
S. Ruiz, A. Jou, A. Bonjoch, L. Ruiz, A. Arno´,
and B. Clotet
AIDS Care Unit, Ophthalmologic Department, and ‘‘irsiCaixa’’
Retrovirology Laboratory, ‘‘Germans Trias i Pujol’’ University
Hospital, Badalona, Spain
Seven AIDS patients who were receiving suppressive therapy for previously diagnosed cytomegalovirus (CMV) retinitis were offered treatment with protease inhibitors (PIs). Secondary prophylaxis
for CMV was discontinued after 3 months of therapy with PIs if patients had ú150 CD4 cells/mm3
and a human immunodeficiency virus (HIV) load of õ200 copies/mL and if they were negative for
CMV as determined by qualitative CMV polymerase chain reaction (PCR). Ophthalmologic exams
were done periodically. After a median follow-up of 9 months (range, 9 – 12), no new episodes of
CMV retinitis were observed. CD4 cell counts were ú150 cells/mm3 in all cases, HIV loads were
õ200 copies/mL, and results for qualitative CMV PCRs remained negative. These observations
suggest that for selected patients with healed CMV retinitis who have immunologic and virologic
evidence of a clinical response to potent combination antiretroviral therapy, temporary discontinuation of a chronic anti-CMV suppressive therapy may not result in further retinal necrosis. However,
the long-term immunologic benefit of PIs and hence the safety of prolonged withdrawal of antiCMV therapy is unknown.
Cytomegalovirus (CMV) is a common opportunistic infection among AIDS patients, with its incidence increasing with
the severity of immunodeficiency. At least 20% of patients
with CD4 lymphocyte counts of õ100 cells/mm3 will develop
CMV-related disease over a 2-year period [1]. Retinitis is the
most common manifestation of CMV disease in human immunodeficiency virus (HIV) – infected patients, and intravenous or
oral maintenance therapy is required to avoid or delay progression to retinal destruction and complete vision loss [1]. However, maintenance therapy is associated with a poor quality of
life; morbidity, mainly related to catheter complications; and
a high prevalence of adverse side effects [2, 3].
Data from observational studies suggest that newer methods
for detection of CMV viremia, such as polymerase chain reaction (PCR) and the detection of pp56 antigenemia in peripheral
blood polymorphonuclear leukocytes, could have a role for the
early diagnosis of CMV disease [4, 5]. Although it has not
been validated in prospective trials, CMV PCR appears to be
the best technique for selecting HIV-positive patients suitable
for primary prophylactic treatment (or preemptive therapy) and
also for monitoring maintenance therapy with oral or intravenous ganciclovir and foscarnet [6-8].
Several clinical end-point studies with protease inhibitors
(PIs) demonstrated a significant delay in the short-term devel-
Received 21 July 1997; revised 10 October 1997.
All patients gave written informed consent to allow us to interrupt maintenance therapy.
Reprints or correspondence: Dr. B. Clotet, AIDS Care Unit and ‘‘irsiCaixa’’
Retrovirology Laboratory, Hospital Universitari ‘‘Germans Trias i Pujol,’’
Ctra. Canyet s/n, 08916 Badalona, Spain ([email protected]).
The Journal of Infectious Diseases 1998;177:1080–3
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
opment of further clinical events and death in a group of patients with advanced HIV infection [9, 10]. In addition, other
studies [11] have shown the immunologic and virologic benefit
of antiretroviral therapy (ARVT) that includes PIs. However,
important issues about the functional impact of the increase in
CD4 cell counts are now being investigated. Connors et al.
[12] have demonstrated that there is a decline in naive CD4/
T cells during the course of HIV infection and that highly
active antiretroviral therapy (HAART) cannot always restore
the number of naive CD4/ T cells even when a dramatic increase in the absolute CD4/ T cell count is achieved. On the
other hand, Autran et al. [13] have shown that immunologic
reconstitution may be achieved after prolonged viral suppression. Whitcup et al. [14] reported a series of 4 patients who
were followed a median of 6 months (range, 4 – 12) without
recurrence of CMV after initiating ARVT with PIs.
The improved knowledge about the natural history of CMV
disease and about the immunologic and virologic events following ARVT that includes PIs prompted us to discontinue
secondary prophylaxis for CMV retinitis after having initiated
ARVT combined with PIs in 7 patients with advanced AIDS.
Patients and Methods
A cohort of 16 AIDS patients on CMV secondary prophylaxis
because of previously diagnosed CMV retinitis were offered
HAART. Three of the 16 patients could not adhere to the new
therapeutic regimen. After 3 months of follow-up, 3 of the remaining 13 patients did not achieve an immunologic (CD4 cell
count ú150/mm3) or virologic (plasma HIV-1 RNA levels õ200
copies/mL) response. Three of the remaining 10 patients who fulfilled the HAART response criteria and who had negative qualitative plasma CMV PCR results refused to discontinue secondary
UC: J Infect
JID 1998;177 (April)
Concise Communications
This study focusses on the remaining 7 patients who consented
to discontinue secondary prophylaxis. The patients included 6 men
and 1 woman with a median age of 33 years (range, 30–43).
Their risk behaviors for HIV infection were as follows: 5 were
homosexual, 1 had been an injection drug user, and 1 was a blood
recipient. All 7 patients presented a first AIDS-defining event prior
to the diagnosis of CMV retinitis, with a median interval of 28
months between the two events. The characteristics of the 7 patients at the time of the diagnosis of CMV retinitis and during
maintenance therapy are summarized in table 1. All of the patients
had prior ARVT experience with reverse transcriptase inhibitors
(nucleoside analogues), and all of them continued their ARVT
regimen until the introduction of PIs.
During the maintenance period, catheter-related bacteremia involving Enterobacter species and Bacillus cereus was observed;
however, the infections resolved with appropriate antimicrobial
therapy. Three patients presented with neutropenia that required
periodic administration of granulocyte colony-stimulating factor.
Patient number 3 developed biopsy-confirmed CMV colitis 3
months after the diagnosis of CMV retinitis. No new AIDS-defining events were diagnosed during the maintenance period prior to
discontinuation of secondary prophylaxis for CMV. After the initial diagnosis, all patients received induction and maintenance therapy with ganciclovir (induction therapy: 5 mg/kg twice daily for
2–3 weeks; maintenance therapy: 6 mg/kg daily, 5 days a week)
or foscarnet (induction therapy: 100 mg/kg twice daily for 2–3
weeks; maintenance therapy: 120 mg/kg daily, 5 days a week).
After the induction phase was completed, ophthalmologic evaluations were conducted every 4 weeks during the maintenance therapy.
Once PIs were made available for compassionate use in Spain,
they were added to the ARVT regimen for each of the 7 patients.
All but 1 patient had been receiving a nucleoside analogue combination since the diagnosis of CMV retinitis (median of 8 months;
range, 3–12). Because most of the patients had been treated with
combination therapy that included zidovudine plus zalcitabine, didanosine, or lamivudine (3TC), we offered treatment with the fol-
lowing combinations: 3TC-stavudine (d4T)-ritonavir (4 patients),
3TC-d4T-ritonavir-saquinavir (2 patients), and 3TC-d4T-indinavir
(1 patient).
At the onset of CMV retinitis and every 3 months after the
change in ARVT, we determined the HIV load for each patient
by use of a reverse transcription–PCR (Amplicor Roche, Madrid)
with a level of detection of 200 copies/mL. In addition, qualitative
plasma CMV PCR was done as described [5] at the diagnosis of
CMV retinitis, once the patient achieved a CD4 cell count ú150/
mm3 after changing ARVT, and at 3, 6, and 9 months after discontinuation of secondary prophylaxis.
Secondary prophylaxis was stopped once the patient had a CD4
cell count ú150/mm3, an undetectable plasma HIV load, and a negative plasma PCR for CMV. Secondary prophylaxis was planned to
be reintroduced if CD4 cell counts declined to õ150/mm3 during
follow-up. All patients achieved CD4 cell counts of ú150 cells/mm3
3 months after the initiation of PIs. After secondary prophylaxis was
discontinued, ophthalmologic evaluations were done weekly for the
first month and monthly thereafter.
Patients were followed for 9 – 12 months (median, 9) after
the secondary prophylaxis was stopped, during which time, no
new episodes of CMV retinitis and no new AIDS-defining
events were recorded. Table 2 shows the most relevant laboratory data for each patient after CMV maintenance therapy was
discontinued. Between the diagnosis of CMV retinitis and after
9 months of discontinued CMV prophylaxis, the median increase in CD4 cell counts was 245 cells/mm3 (range, 124 –
975). All patients had a CD4 cell count of ú180 cells/mm3 3
months after starting treatment with PIs. At 6 months, all patients had CD4 cell counts of ú230/mm3, and at 9 months, all
patients had ú150 cells/mm3. At all time points during followup, all patients had a CD4 cell percent ú10%, an HIV virus
load of õ200 copies/mL, and a negative qualitative CMV PCR.
Table 1. Characteristics of 7 AIDS patients at the time of CMV retinitis diagnosis and during the CMV maintenance therapy period.
since AIDS
median, 28)
Time on
median, 30)
CD4 cell
median, 35)
HIV virus load
Time on anti-CMV
therapy (months;
median, 11)
No. of relapses
(median, 1)
NOTE. ARVT Å antiretroviral therapy; PCR Å polymerase chain reaction.
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
Other complications
1 episode of
2 episodes of
bacteremia, CMV
Concise Communications
JID 1998;177 (April)
Table 2. CD4 cell counts for 7 AIDS patients after the discontinuance of CMV maintenance therapy.
No. of CD4 cells/mm3 (%)
Patient, HAART regimen
At time CMV
prophylaxis stopped*
(median Å 230)
9 months after CMV
prophylaxis stopped
(median Å 300)
No. of months of
(median Å 9)
NOTE. At both time points, all patients had HIV loads of õ200 copies/mL and negative results for qualitative
polymerase chain reaction. During follow-up, no new episodes of CMV or new AIDS-defining events occurred.
HAART Å highly active antiretroviral therapy, d4T Å stavudine, 3TC Å lamivudine, IDV Å indinavir, RTV Å
ritonavir, SQV Å saquinavir.
* 3 months after initiation of treatment with protease inhibitors.
CMV involvement of the retina appears to occur mainly in
persons with a CD4 cell count of õ100 cells/mm3 [1]. However
a recent report [15] suggests that CMV retinitis may appear
shortly (õ2 months) after the introduction of PIs even if the
patient has a CD4 cell count of ú150 cells/mm3.
Our data differ from those of Jacobson et al. [15]; however,
this is probably due to the fact that our patients had CD4 cell
counts of ú150 cells/mm3 longer than the other patients, and
they might also have recovered their immunologic repertoire
in a more consistent manner. Regarding the restoration of the
immune system, Connors et al. [12] have examined the changes
in CD4 T cell surface marker phenotype and antigen receptor
following HIV ARVT. They found that ARVT induced increases in CD4 cell counts that led to only minor changes in
previously disrupted repertoires. According to Connors et al.
[12], CMV can be regarded as a latently infecting agent in HIVinfected patients. For this reason, the frequency of memory T
cells with the appropriate receptor exceeds the critical number
necessary to provide protection as long as CD4 cell counts
remain ú100 cells/mm3. Due to this high precursor frequency,
we can hypothesize that after a sustained increase in total CD4
cell counts, an improvement in the repertoire needed for CMV
prevention could occur.
Our observations suggest that for selected patients with
healed CMV retinitis who have immunologic and virologic
evidence of a clinical response to HAART, temporary discontinuation of chronic, CMV-suppressive therapy may not result
in further retinal necrosis. However, the long-term immunologic benefit of PIs and hence the safety of prolonged withdrawal of anti-CMV therapy is unknown.
1. Pertel P, Hirschtick R, Phari J, Chmiel J, Poggense L, Murphy R. Risk of
developing cytomegalovirus retinitis in persons infected with the human
immunodeficiency virus. J Acquir Immune Defic Syndr 1992;5:1069–74.
/ 9d43$$ap48
03-04-98 19:03:33
2. Drew WL, Ives D, Lalezari JP, et al. Oral ganciclovir as maintenance
treatment for cytomegalovirus retinitis in patients with AIDS. N Engl
J Med 1995; 333:615 – 20.
3. Jacobson MA, Kramer F, Bassiakos Y, et al. Randomized phase I trial of
two different combination foscarnet and ganciclovir chronic maintenance therapy regimens for AIDS patients with cytomegalovirus retinitis: AIDS clinical trials group protocol 151. J Infect Dis 1994; 170:
189 – 93.
4. Hansen KK, Richsten A, Hofmann B, Norrild B, Olofsson F, Mahiessen
L. Detection of cytomegalovirus DNA in serum correlates with clinical
cytomegalovirus retinitis in AIDS. J Infect Dis 1994; 170:1271 – 4.
5. Spector SA, Merrill R, Wolf D, Danker WM. Detection of human cytomegalovirus in plasma of AIDS patients during acute visceral disease by
DNA amplification. J Clin Microbiol 1992; 30:2359 – 65.
6. Shinkai M, Bozzette SA, Powderly W, Frame P, Spector SA. Utility of
urine and leukocyte cultures and plasma DNA polymerase chain reaction
for identification of AIDS patients at risk for developing human cytomegalovirus disease. J Infect Dis 1997; 175:302 – 8.
7. Bowen EF, Wilson P, Cope A, et al. Cytomegalovirus retinitis in AIDS
patients. Influence of cytomegaloviral load on response to ganciclovir,
time to recurrence and survival. AIDS 1996; 10:1515 – 20.
8. Dodt KK, Jacobsen PH, Hofmann B, et al. Development of cytomegalovirus
(CMV) disease may be predicted in HIV-infected patients by CMV polymerase chain reaction and the antigenemia test. AIDS 1997;11:F21–8.
9. Cameron B, Heath-Chiozzi M, Kravick S, Mills R, Pothoff A, Henry D.
Prolongation of life and prevention of AIDS in advanced HIV immunodeficiency with ritonavir [abstract LB6a]. In: Program and abstracts: 3rd
Conference on Retroviruses and Opportunistic Infections (Washington,
DC). Alexandria, VA: Infectious Diseases Society of America, 1996.
10. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of two
nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cell counts of 200 per cubic millimeter
or less. N Engl J Med 1997; 337:725 – 33.
11. Carpenter CC, Fischl MA, Hammer SM, et al. Antiretroviral therapy for
HIV infection in 1996. recommendations of an international panel.
JAMA 1996; 276:146 – 54.
12. Connors M, Kovacs JA, Krevat S, et al. HIV infection induces changes
in CD4/ T-cell phenotype and depletions within the CD4/ T-cell repertoire that are not immediately restored by antiviral or immune-based
therapies. Nat Med 1997; 3:533 – 9.
13. Autran B, Carcelain G, Li TS, et al. Positive effects of combined antiretroviral therapy on CD4/ T cell homeostasis and function in advanced
HIV disease. Science 1997; 277:112 – 6.
UC: J Infect
JID 1998;177 (April)
Concise Communications
14. Whitcup SM, Fortin E, Nussenblatt RB, Polis MA, Muccioli C. Therapeutic effect of combination antiretroviral therapy on cytomegalovirus retinitis [letter]. JAMA 1997; 227:1519 – 20.
15. Jacobson MA, Zegans M, Pavan PR, et al. Cytomegalovirus retinitis after
initiation of highly active antiretroviral therapy. Lancet 1997; 349:
1443 – 5.
Major Expansions of Select CD8/ Subsets in Acute Epstein-Barr Virus Infection:
Comparison with Chronic Human Immunodeficiency Virus Disease
Jonathan E. Lynne, Ingrid Schmid, Jose L. Matud,
Karim Hirji, Scott Buessow,* Deborah M. Shlian,
and Janis V. Giorgi
Departments of Medicine and Biomathematics, UCLA School of
Medicine, Los Angeles, and SmithKline Beecham Clinical Laboratories,
Van Nuys, California
CD8/ lymphocyte phenotypes were characterized during acute Epstein-Barr virus (EBV) infection, and a comparison was made to previous studies of human immunodeficiency virus (HIV).
This was of interest because CD8/ cells contribute to immunologic control of both infections, but
the usual outcome of EBV infection is benign, whereas untreated HIV infection is fatal. During
acute EBV infection, CD8/ cells expressed elevated levels of the activation antigens CD38 and
HLA-DR, similar to that during chronic HIV infection. Within 16 weeks, when EBV latency is
established, CD8/ cell activation had resolved. In contrast, activation persists in HIV infection.
Expression of CD38 and HLA-DR on CD8/ cells could be a marker for ongoing viral replication
in both infections. Other CD8/ cell alterations observed in this study of acute EBV infection included
increases in both CD62L0 and CD62L/ CD8/ cells and unique kinetics in the expansion of the
CD57/CD8/ cell subset.
Epstein-Barr virus (EBV), a human g-type herpesvirus, is
commonly recognized as the agent that causes acute infectious
mononucleosis (AIM) [1]. Initially, EBV replicates in pharyngeal epithelium but within a few weeks establishes a latent
infection in B cells. The CD8/ cell immune response, including
CD8/ lymphocytosis during AIM, is believed to be responsible
for controlling acute EBV infection and maintaining viral latency. Despite the presence of cytotoxic T lymphocytes (CTL),
a low level of EBV replication in pharyngeal epithelium can
be intermittently detected years after AIM resolves.
In contrast to EBV, which rapidly establishes a viable hostpathogen relationship in most people, human immunodeficiency virus (HIV) replicates continuously in the lymphoid
system of the infected host [2]. Replication occurs at such high
levels that it ultimately leads to deterioration of the immune
system despite the maintenance for years of anti-HIV – specific
CTL. Like acute EBV infection, chronic HIV infection is ac-
Received 23 July 1997; revised 22 October 1997.
All subjects provided informed consent, and guidelines for the US Department of Health and Human Services were followed in the conduct of this
Financial support: NIH (CA-16042, AI-32883, AI-35040 and AI-37613).
Reprints or correspondence: Dr. Janis V. Giorgi, UCLA School of Medicine,
Dept. of Medicine/Cellular Immunology and Cytometry, 12-236 Factor Building, Los Angeles, CA 90095-1745 ([email protected]).
* Present affiliation: SmithKline Beecham Laboratories, Dallas.
The Journal of Infectious Diseases 1998;177:1083–7
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
companied by CD8/ cell lymphocytosis. We and others have
described selective expansions of certain CD8/ subsets in
chronic HIV infection [3]. Of particular importance, expansions
of CD8/ cells with elevated expression of the CD38 activation
antigen provide strong prognostic value for predicting progression of HIV infection to AIDS and death [4]. In the current
study, we examined the temporal pattern of CD8/ cell alterations in EBV infection and the similarities and differences
between the CD8/ subsets that contribute to EBV- and HIVinduced CD8/ lymphocytosis.
Materials and Methods
Study populations. Thirteen acute EBV mononucleosis patients (12, 18–22 years old; 1, 37 years old; 6 men and 7 women)
were followed at the UCLA Student Health Service. All had atypical lymphocytosis at presentation. Acute EBV mononucleosis was
confirmed by quantitative immunofluorescent antibody test for the
titers of specific antibodies against EBV antigens [1] done at
SmithKline Beecham Clinical Laboratories. Symptoms of acute
mononucleosis had resolved completely in all subjects by 16
weeks. Most patients contributed blood at 2, 4, 6, and 8 weeks,
and all contributed at 0 and 16 weeks.
Data on the HIV-infected cohort (n Å 98) and the heterosexual
control group (n Å 50, men; 32 { 7 years old) have been published
previously [3], and relevant data are summarized in table 1 for
comparisons with EBV infection. The HIV-infected homosexual
men were AIDS-free participants in the Los Angeles Multicenter
AIDS Cohort Study (MACS) and had been infected for 1 to an
estimated 8 years; their average absolute CD4/ cell number was
UC: J Infect
Concise Communications
JID 1998;177 (April)
Table 1. Similar to chronic HIV disease, lymphocytosis during acute EBV infection is characterized
by expansions in CD38/, HLA-DR/, CD62L0, CD57/, and CD570 subsets of CD8/ cells but also
includes an increase in CD62L/ cells.
Presentation (week 0)
Week 16
Infection, cell subset
Acute EBV infection
82 – 8284
113 – 477
0 – 7693
245 – 2809
341 – 6627
68 – 4902
27 – 1788
354 – 8639
128 – 264
168 – 677
0 – 161
238 – 708
98 – 613
84 – 369
28 – 387
204 – 710
Chronic HIV infection
Heterosexual control
5th – 95th
5th – 95th
88 – 888
148 – 861
49 – 702
207 – 1297
116 – 435
167 – 1192
107 – 623
230 – 976
29 – 204
156 – 729
0 – 33
216 – 944
127 – 660
82 – 399
27 – 229
183 – 746
NOTE. Values for medians and ranges or 5th – 95th percentiles are absolute number of CD8/ cells/mm3 with
indicated phenotype. Results for EBV infection represent values for same 13 subjects at weeks 0 (presentation) and
16 weeks later; P values are for Wilcoxon signed rank test for comparisons between paired week 0 and week 16
values. Values for HIV-infected (n Å 98) and heterosexual control (n Å 50) groups are from [3] and were compared
here by Wilcoxon rank test for independent samples. NS, not statistically significant at P Å .01 level.
Immunofluorescent staining and flow cytometry. Lymphocyte
subset whole blood analysis was done on a flow cytometer (EPICS
C; Coulter, Miami) within 6 h of the blood draw as described previously [3]. CD3 fluorescein isothiocyanate (FITC), CD4 phycoerythrin (PE), CD8 FITC, CD8 PE, CD38 PE, anti–HLA-DR PE,
CD62L FITC, CD57 FITC, and isotypic control antibodies were from
Becton Dickinson Immunocytometry Systems (San Jose, CA).
Statistical analysis. The Wilcoxon rank sum test was used to
compare the lymphocyte subset values of recovered (week 16)
mononucleosis patients (n Å 13) and healthy heterosexual controls
(n Å 50) and to compare chronically HIV-infected MACS participants (n Å 98) and healthy heterosexual controls. The Wilcoxon
signed rank test was used to compare the lymphocyte subset values
of the EBV mononucleosis subjects at presentation (week 0) and
recovery (16 weeks after presentation). A cutoff significance level
of .01 rather than .05 was used to allow for more conservative
hypothesis testing.
CD8/ lymphocytosis in acute EBV infection was severe at
presentation and resolved within 16 weeks. The extent of the
CD8/ lymphocytosis was much greater in acute EBV infection
/ 9d43$$ap48
03-04-98 19:03:33
at week 0 (median, 2456/mm3; range, 409 – 9231) than in
chronic HIV disease (median, 807/mm3; 5th – 95th percentile,
330 – 1597). A rapid drop in the circulating CD8/ cell level
occurred by week 2 (median, 797/mm3; range, 325 – 1647),
when 7 of the EBV patients already had CD8/ cell numbers
within the normal range. The decrease in CD8/ cell numbers
continued from week 2 through week 16, with median values
at weeks 4 and 8 of 733 and 605/mm3, respectively. At week
16, when the median CD8/ value was 536/mm3 (range, 252 –
903), all 13 subjects had CD8/ cell values within the normal
range defined by the heterosexual control group (median, 490/
mm3; 5th – 95th percentile, 238 – 959; P Å .90 for comparison
between week 16 EBV patients and heterosexual controls),
and their clinical symptoms of acute EBV mononucleosis had
resolved. The week 16 specimen was therefore considered the
‘‘baseline’’ value for each subject in the analyses.
CD8/ lymphocytosis during acute EBV infection was caused
by a selective increase in activated CD8/ cell subsets. To
determine which subsets are responsible for the CD8/ cell
lymphocytosis during AIM, a statistical analysis was done comparing the absolute number of the CD38-, HLA-DR-, CD62L-,
and CD57-positive and -negative subsets at week 0 versus week
UC: J Infect
JID 1998;177 (April)
Concise Communications
Figure 1. Selective elevations in CD8/ cell subsets during course of acute EBV mononucleosis. Data are % of CD8/ cells that were CD38/,
HLA-DR/, CD62L0, or CD57/ at presentation (day 0) through week 16. Each line represents every result available for 1 of 13 acute EBV
infection subjects.
16 of the EBV-infected subjects (table 1). Virtually all of the
expanded CD8/ cell population at week 0 was CD38/HLADR/, a cell phenotype that is rare in healthy controls. The
CD380 and HLA-DR0 subsets of CD8/ cells, which are considered resting cells, showed no increases due to EBV infection.
At presentation with AIM, both the CD62L/ and CD62L0
populations were also significantly elevated, with about half
of the cells, on average, in each population. The increases in
the CD570 and CD57/ cell populations were both significant,
with Ç6-fold and 2-fold expansions, respectively.
To determine the relative contribution to the CD8/ cell lymphocytosis of the CD8/ subsets that are usually referred to as
‘‘activated,’’ the percentage of CD8/ cells that were CD38/,
HLA-DR/, CD62L0, and CD57/ was determined at each time
point (week 0 through week 16) for each of the 13 subjects,
as shown in figure 1. At week 0, the CD38/CD8/ percentage
was at its peak in all 13 patients, and the HLA-DR/CD8/
percentage was at its peak in 12 of these. Expression of these
activation antigens gradually resolved at a variable rate in each
person. In aggregate, the percentage of CD8/ cells that were
CD62L0 did not change markedly over the 16-week period, as
/ 9d43$$ap48
03-04-98 19:03:33
reflected by the relatively stable median percentage of
CD62L0CD8/ cells at each visit. The percentage of
CD57/CD8/ cells was at its lowest at week 0 in 12 of the 13
subjects and was higher at one or more time points thereafter
(figure 1).
CD8/ lymphocyte subset alteration patterns in acute EBV
mononucleosis resembled the pattern in chronic HIV disease
but differed in the CD8/CD62L/ subset. We next evaluated whether the pattern of CD8/ cell alterations observed
at presentation with AIM (table 1) was also observed during
chronic HIV infection (table 1). For this comparison, we
examined the absolute values for the CD8/ T cell levels in
chronically HIV-infected persons versus a group of uninfected heterosexual controls. Similar to the pattern in AIM
patients, significant increases were observed in the CD38/,
HLA-DR/, CD62L0, CD570, and CD57/ subsets of CD8/
cells in the HIV-infected persons compared with the same
subsets in heterosexual controls (P õ .01 for all comparisons). An elevation in the CD62L/CD8/ lymphocyte subset
was apparent in AIM but absent in chronic HIV infection.
With this exception, and the greater overall increase in CD8/
UC: J Infect
Concise Communications
cell numbers during acute EBV infection, the features of
CD8/ cell subset alterations during acute EBV infection that
were observed using CD38, CD57, CD62L, and anti – HLADR monoclonal antibodies were similar to those observed
during chronic HIV infection.
In the current study, we examined CD8/ T lymphocyte subset alterations in acute EBV infection. The same reagents and
techniques used in our previous study of HIV infection [3]
were used here, allowing a direct comparison of the CD8/ cell
alterations in the two viral infections (table 1). Our regular
follow-up of the AIM patients provides information on the
kinetics of activation of CD8/ subsets during acute EBV infection (figure 1).
The resolution of CD8/ cell activation in acute EBV infection and its persistence in chronic HIV infection suggest that
CD8/ cell activation could be an immune response to active
viral replication. We did not measure the viral kinetics for AIM
patients in the current study, but declining infectious virus titers
and subsequent establishment of viral latency within a few
weeks after EBV infection have been observed by other investigators [1]. Our data (figure 1) indicate that CD8/ cell activation, measured as elevated CD38 and HLA-DR expression,
was highest at presentation with AIM and resolved thereafter.
Expression of these antigens is also markedly elevated during
primary HIV infection [5, 6]. Although the increase usually
resolves to some degree during the first year of HIV infection,
values remain above the normal range in most subjects and
show a strong correlation with virus load [7, 8].
An important question is what fraction of the CD8/ cell
immune response during viral infections is directed against the
virus and whether ‘‘bystander’’ CD8/ activation also occurs.
Bystander activation may contribute to some of the pathologic
features of acute viral illnesses as well as to maintenance of
CD8/ cell memory responses. Several reports indicate that
CD8/ cell lymphocytosis during viral infections includes bystander CD8/ cell activation [9 – 11]. However, recent studies
on T cell receptor usage during acute human EBV and HIV
infections argue that a majority (ú50%) of the CD8/ cells in
some patients may be the result of clonal or oligoclonal responses to the invading pathogen [12, 13]. Another recent report involving murine viral infections also argues against a
substantial role of bystander activation [14].
One hypothesis that might be pursued is that the expansion
of the CD62L/ population that we observed during AIM could
reflect bystander activation while that in the CD62L0 population might be antigen-specific. Antigen-specific triggering of
the T cell receptor results in immediate down-modulation of
CD62L [15]. Future studies could use cell sorting to separate
the CD8/ cells on the basis of CD62L expression to examine
antiviral CTL activity and oligoclonality in the sorted subsets.
The expectation would be that the CD62L0 and CD62L/ sub-
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
sets might differ for these parameters, with greater evidence
of antigen-specific responses present in the CD62L0CD8/ fraction.
Why the CD8/ cells fail to clear or completely control viral
replication in either EBV or HIV infections, despite a vigorous
immune response triggered in both infections, remains speculative and is a critical question to answer. CD8/ CTL are believed
to contribute long-term to reducing viral replication in both
EBV and HIV infections [1, 2]. However, persons who had
resolved acute EBV infection did not maintain high levels of
activated T cells (table 1). Similarly, our previously published
results demonstrate that some HIV-seropositive long-term nonprogressors do not manifest extensive CD8/ cell activation [7].
Thus, maintaining high levels of activated CD8/ T cells is
not a requisite for low virus burden but rather could reflect
persistence of viral replication.
We thank Roger Detels for helpful suggestions, Eve Schenker
and Karen Ferran for coordinating the study, and Diane Smith
and John Fahey (UCLA Center for Interdisciplinary Research on
Immunologic Diseases) for enumeration of atypical lymphocytosis.
1. Rickinson AB, Kieff E. Epstein-Barr virus. In: Fields BN, Knipe DM,
Howley PM, eds. Fields virology. 3rd ed. Philadelphia: LippincottRaven Publishers, 1996:2397 – 446.
2. Pantaleo G, Fauci AS. New concepts in the immunopathogenesis of HIV
infection. Annu Rev Immunol 1995; 13:487 – 512.
3. Giorgi JV, Detels R. T-cell subset alterations in HIV-infected homosexual
men: NIAID Multicenter AIDS Cohort Study. Clin Immunol Immunopathol 1989; 52:10 – 8.
4. Giorgi JV, Liu Z, Hultin LE, Cumberland WG, Hennessey K, Detels R.
Elevated levels of CD38/CD8/ T cells in HIV infection add to the
prognostic value of low CD4/ T cell levels: results of 6 years of followup. J Acquir Immune Defic Syndr 1993; 6:904 – 12.
5. Cossarizza A, Ortolani C, Mussini C, et al. Massive activation of immune
cells with an intact T cell repertoire in acute human immunodeficiency
virus syndrome. J Infect Dis 1995; 172:105 – 12.
6. Ferbas J, Daar ES, Grovit-Ferbas K, et al. Rapid evolution of human
immunodeficiency virus strains with increased replicative capacity during the seronegative window of primary infection. J Virol 1996; 70:
7285 – 9.
7. Ferbas J, Kaplan AH, Hausner MA, et al. Virus burden in long-term
survivors of human immunodeficiency virus (HIV) infection is a determinant of anti-HIV CD8/ lymphocyte activity. J Infect Dis 1995; 172:
329 – 39.
8. Bouscarat F, Levacher-Clergeot M, Dazza MC, et al. Correlation of CD8
lymphocyte activation with cellular viremia and plasma HIV RNA levels
in asymptomatic patient infected by human immunodeficiency virus
type 1. AIDS Res Hum Retroviruses 1996; 12:17 – 24.
9. Doherty PC, Hou S, Tripp RA. CD8/ T-cell memory to viruses. Curr
Opin Immunol 1994; 6:545 – 52.
10. Yang H, Dundon PL, Nahill SR, Welsh RM. Virus-induced polyclonal
cytotoxic T lymphocyte stimulation. J Immunol 1989; 142:1710 – 8.
UC: J Infect
JID 1998;177 (April)
Concise Communications
11. Tough DF, Borrow P, Sprent J. Induction of bystander T cell proliferation
by viruses and type I interferon in vivo. Science 1996; 272:1947 – 50.
12. Pantaleo G, Demarest JF, Soudeyns H, et al. Major expansion of CD8/
T cells with a predominant Vb usage during the primary immune response to HIV. Nature 1994; 370:463 – 7.
13. Callan MF, Steven N, Krausa P, et al. Large clonal expansions of CD8/
T cells in acute infectious mononucleosis. Nature Med 1996; 2:906 – 11.
14. Zarozinski CC, Welsh RM. Minimal bystander activation of CD8 T cells
during the virus-induced polyclonal T cell response. J Exp Med 1997;
185:1629 – 39.
15. Jung TM, Dailey MO. Rapid modulation of homing receptors (gp90 MEL14) induced by activators of protein kinase C: receptor shedding due to
accelerated proteolytic cleavage at the cell surface. J Immunol 1990;
144:3130 – 6.
Natural History of Epstein-Barr Virus Infection in a Prospective Pediatric
Cohort Born to Human Immunodeficiency Virus–Infected Mothers
Louise Pedneault, Normand Lapointe, Caroline Alfieri,
Parviz Ghadirian, Linda Carpentier, Johanne Samson,
and Jean Joncas
Department of Microbiology and Infectious Diseases, Research Center,
and Centre Maternel et Infantile sur le SIDA, Hoˆpital Sainte-Justine;
Departments of Microbiology and Immunology and of Pediatrics,
Universite´ de Montre´al; Epidemiology Research Unit, Research Center,
Hoˆpital Hoˆtel-Dieu de Montre´al, Montre´al, Canada
To determine whether Epstein-Barr virus (EBV) constitutes a contributing factor in AIDS and,
conversely, whether the human immunodeficiency virus (HIV) alters the course of primary EBV
infection in a pediatric population, 62 children born to HIV-infected mothers and prospectively
followed were evaluated. EBV infection was documented by EBV-specific serology and polymerase
chain reaction and by clinical history. HIV infection status was determined according to the Centers
for Disease Control and Prevention pediatric classification system. Demographics from HIV-infected
and HIV-uninfected children were comparable. The data suggest that HIV-infected children may
acquire primary EBV infection earlier in life. The incidence of accompanying splenomegaly or
hepatomegaly (or both) around the time of EBV seroconversion was higher among HIV-infected
children than among HIV-uninfected children. In contrast, HIV disease progression and HIV-1
RNA load did not seem to be influenced by primary EBV infection.
A number of reports have indicated that Epstein-Barr virus
(EBV) and human immunodeficiency virus (HIV) are capable
of mutually influencing the other’s potential for infectivity and
replication [1 – 3]. Virologic and serologic findings also suggest
a role for EBV in AIDS; higher levels of EBV replication in
oropharyngeal secretions and serologic evidence of persistent
or reactivated EBV infection seem more common in HIVinfected subjects [4].
Clinically, EBV infection has been associated with the development of several lesions in AIDS patients, such as oral hairy
Received 28 April 1997; revised 6 November 1997.
Presented in part: First National Conference on Human Retroviruses and
Related Infections, December 1993, Washington, DC (abstract 397); 34th Interscience Conference on Antimicrobial Agents and Chemotherapy, October
1994, Orlando, Florida (abstract 362).
Informed consent was obtained from the patients’ parents. The project and
the consent form had been approved by the ethics committee of Ste-Justine
Financial support: National Health Research and Development Program,
Health and Welfare Canada (6605-3953-AIDS); Fonds de la Recherche en
Sante´ du Que´bec (to L.P.); J.A. de Se`ve Foundation (to C.A.).
Reprints or correspondence (present address): Dr. Louise Pedneault, Glaxo
Wellcome, Inc., Antiviral Clinical Research, Five Moore Dr., PO Box 13398,
Research Triangle Park, NC 27709 ([email protected]).
The Journal of Infectious Diseases 1998;177:1087–90
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
leukoplakia in adults [5], lymphocytic interstitial pneumonitis
in children [6], central nervous system lymphoid neoplasms
both in adults and children [6, 7], and, more recently, leiomyosarcomas in children [8]. However, it may be extremely difficult to differentiate a typical EBV mononucleosis syndrome
from the clinical category A for children with HIV infection,
as described in the Centers for Disease Control and Prevention
(CDC) classification system for HIV infection in children [9].
Only limited information is available on the role played by
EBV during the course of HIV infection, and this is especially
true in pediatrics. Consequently, the main objectives pursued
in this study were to determine the incidence of primary EBV
infection and its possible occurrence at an earlier age, if not
at birth, in HIV-infected infants and children born to HIVinfected mothers, compared with their uninfected counterparts
also born to HIV-infected mothers; to determine whether or
not the natural history of EBV infection is worsened in those
same HIV-infected children compared with uninfected ones;
and to determine if the rate of progression of the HIV infection
in regard to the primary EBV infection is accelerated.
Materials and Methods
Study population and specimens. The study group included 65
children (18 HIV-infected and 47 -uninfected) born to 62 HIV-
UC: J Infect
Concise Communications
infected mothers, followed prospectively since birth at the Centre
Maternel et Infantile sur le SIDA (CMIS) of Sainte-Justine Hospital. HIV infection was documented by HIV culture and/or HIV
polymerase chain reaction (PCR). Follow-up of patients included
determination of CD4 cell counts, HIV serology, HIV p24 antigenemia, and HIV-1 RNA levels. HIV infection status was determined
according to the CDC revised pediatric classification system [9].
Clinical and laboratory information available for each child who
seroconverted for EBV was retrospectively matched at the time of
EBV seroconversion, Ç3 months prior to and 3 months after its
documentation. Data examined at these three time points were the
HIV classification of the child; HIV-1 RNA levels; CD4 cell
counts; signs, symptoms, and laboratory markers compatible with
a primary EBV infection; and administration of transfusions and/
or immunoglobulins.
A total of 594 frozen serial serum samples collected at birth
and generally every 3 months thereafter from the pediatric study
group, and 1 serum specimen each from their mothers, were readily
available from the CMIS and analyzed for the presence of antibodies to EBV. For each EBV seroconverter, serial frozen peripheral
blood mononuclear cells (PBMC) and serial frozen serum specimens collected around the time of EBV seroconversion were also
evaluated for the detection of EBV DNA by PCR and for HIV-1
RNA levels, respectively. Furthermore, to substantiate the possibility that EBV infection might occur in immunosuppressed HIVinfected children, without the development of EBV-specific antibodies, we also used PCR to detect the presence of EBV DNA on
serial PBMC specimens from the HIV-infected but EBV-seronegative children and from matched HIV- and EBV-seronegative children from the same group. Frozen PBMC specimens collected
every 3 months between the ages of 7 and 25 months were tested
for each child. This time frame was chosen to entirely cover the
age period for EBV seroconversion in the HIV-infected children
from our group who seroconverted for EBV (range, 9–23 months).
EBV serology and serologic diagnosis. EBV serology, including standard indirect immunofluorescence microscopy for the detection of viral capsid antigen (VCA) and early antigen (EA) IgG
antibodies [10] ([11] with minor modifications) and anti-complement immunofluorescence for the determination of Epstein-Barr
virus nuclear antigen titers [12], was done on all sera. VCA IgM
antibody titers were measured by a commercially available EIA
(Gull Laboratories, Salt Lake City) in suspected cases of EBV
Patients were considered to have primary infection with EBV
if their VCA IgM titer was positive or, in the absence of blood
transfusions within 2–3 months, if they showed seroconversion of
their VCA IgG. Chronic activation of EBV infection was defined
by persistence of EA titers §1:80 for at least 6 months after
documentation of primary EBV infection.
PCR and detection of EBV DNA. PCR was done on PBMC by
use of oligoprimers from the BMLF1 region of the EBV genome,
according to previously described methods [13].
HIV-1 RNA quantitation. The NucliSens HIV-1 QT, a nucleic
acid sequence–based amplification assay, was used for the determination of HIV-1 RNA levels in serum (100–200 mL/specimen),
according to manufacturer’s instructions (Organon Teknika, Scarborough, Canada).
Statistical analysis. Subject groups (15 HIV-infected and 47
HIV-uninfected children) were compared for their EBV infection-
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
free interval by use of Kaplan-Meier survival analysis and the
Breslow statistic (BMDP/pc/ version 7.01 software; BMDP Statistical Software, Los Angeles). Qualitative variables were analyzed by the x2 test or, in the case of correlated proportions,
the McNemar test. After appropriate transformations, differences
between HIV-infected and HIV-uninfected children and between
the EBV seroconverters and the EBV-seronegative subjects regarding the various continuous parameters (e.g., EA and VCA IgG
titers, CD4 cell counts) were investigated by use of Student’s t
test applied to log-transformed data.
Sixty-five children born to 62 HIV-infected mothers between
November 1987 and May 1993 constituted the study cohort.
HIV infection was documented in 18 (28%); 47 (72%) were
uninfected. Three of the 18 HIV-infected children had received
transfusions or immunoglobulins and were therefore excluded
from all analyses. Demographic data (except for sex distribution), socioeconomic levels, and ethnic backgrounds were comparable in the 2 groups of children (table 1). Seven mothers
received zidovudine during pregnancy because of CD4 cell
counts õ200/mm3; none of the newborns was HIV-infected.
One African-Canadian mother was infected with HIV-2; the
infant was uninfected.
The mean duration of follow-up from birth for all children
was 30.0 months (range, 7 – 75), comparable for both groups
(33.6 and 28.9 months for the HIV-infected and -uninfected
children, respectively). Eleven (73%) of the 15 HIV-infected
children and 26 (55%) of the 47 HIV-uninfected children devel-
Table 1. Characteristics of children born to HIV-infected mothers
(n Å 62).
HIV-infected HIV-uninfected
Family income (Canadian dollars)*
10,000 to 19,999
20,000 to 39,999
Household no. of people (mean)
Maternal risk factor for HIV infection
White (non-Hispanic)
Intravenous drug abuser
White (non-Hispanic)
Transfusion (Haitian-Canadian)
6 (40.0)
9 (60.0)
33 (70.2)
14 (29.8)
5 (35.7)
7 (50.0)
1 (7.1)
1 (7.1)
11 (25.0)
23 (52.3)
4 (9.1)
6 (13.6)
1 (6.7)
1 (6.7)
NOTE. Data are no. (%).
* Information was available for 58 children.
UC: J Infect
10 (21.3)
1 (2.1)
JID 1998;177 (April)
Concise Communications
oped primary EBV infection, as defined by their EBV serology
(P Å .22). Median EBV infection-free intervals were 18 months
and 24 months for HIV-infected and HIV-uninfected children,
respectively (P Å .13). Sex distribution, demographic data,
socioeconomic levels, and ethnic backgrounds (as detailed previously) were compared among EBV seroconverters and EBVuninfected children, and no significant difference was observed
(data not shown).
VCA IgM antibodies were detected in sera from 2 HIVinfected and 3 HIV-uninfected children (sensitivity, 13.5%).
PCR allowed detection of EBV DNA in PBMC specimens
from 35 of the 37 EBV seroconverters (sensitivity, 94.6%),
including the 5 children who mounted a detectable VCA IgM
response. When serial PBMC samples from 6 EBV-seronegative (3 HIV-infected and 3 HIV-uninfected) children were evaluated for the presence of EBV DNA by PCR, none was found
to be infected with EBV.
At the time of EBV seroconversion, the geometric mean
titers of IgG antibodies to VCA were not statistically different
among HIV-infected children compared with their HIV-uninfected counterparts (321 vs. 347; P Å .91). Chronic activation
of EBV infection, defined as EA titers §1:80 for at least 6
months after primary EBV infection, was documented in 3
(27.3%) of 11 HIV-infected children and 3 (11.5%) of 26 HIVuninfected children (P Å .335). While both HIV-infected and
-uninfected children seemed to develop antibodies to EBNA
on EBV seroconversion, the EBNA1/EBNA2 profile differed
between groups [12].
All 62 HIV-infected mothers were also found to be EBVseropositive. Maternal EA titers §1:80 were not correlated
with a higher rate of HIV transmission; 1 (6.7%) of 15 HIVinfected and 7 (14.9%) of 47 HIV-uninfected children were
born to mothers with EA titers §1:80.
Among all children experiencing primary EBV infection,
significantly more HIV-infected children developed splenomegaly (n Å 4) and hepatomegaly (n Å 3) de novo around
the time of EBV seroconversion or shortly thereafter than did
uninfected children (n Å 0) (P Å .005 and P Å .02). More
than 75% of HIV-uninfected children were asymptomatic at
the time of their EBV seroconversion, in contrast to 45.5% of
the HIV-infected ones (P Å .12). One case of lymphocytic
interstitial pneumonitis was diagnosed 7 months after the primary EBV infection in an HIV-infected child; his HIV classification remained unchanged. No lymphoma was reported, even
during extended follow-up.
In the short term, primary EBV infection did not seem to
affect HIV disease progression or HIV-1 RNA load. Mean
HIV-1 RNA levels measured before, during, and after EBV
seroconversion were 5.03 log10, 5.06 log10, and 4.95 log10,
respectively. No change in the HIV classification status of the
HIV-infected children was observed around the time of their
EBV seroconversion. Three deaths occurred among HIV-infected children. One infant experienced growth failure at the
age of 6 months, severe neurologic impairment believed to be
/ 9d43$$ap48
03-04-98 19:03:33
HIV-related and elevated liver transaminase levels at 8 months
of age, and bacterial septicemia, pneumonia, and EBV seroconversion when he was 11 months old. The patient died at
the age of 12.5 months. The two other deaths observed among
HIV-infected children occurred ú21 months after primary
EBV infection. Four HIV-infected children remained EBVseronegative; they all developed symptomatic HIV infection at
a young age (1 patient each was classified as C1 at age 5
months, C2 at 10 months, C3 at 16 months, and A2 at 9 months
of age, respectively). In contrast, 4 of 11 HIV-infected children
who seroconverted for EBV were still HIV-asymptomatic at
about the same age (range, 9 – 22 months).
No differences in CD4 cell counts (absolute values and percentage) were observed around the time of EBV seroconversion
in HIV-infected and HIV-uninfected children (data not shown).
In our study, both groups (HIV-infected and -uninfected
children) had comparable socioeconomic conditions of living,
ethnic backgrounds, and maternal risk factors for HIV infection, therefore minimizing bias introduced by differences in
such factors. This is particularly important, since a higher prevalence of EBV seropositivity has been described for persons
from lower socioeconomic and hygienic levels.
On the basis of serologic data, the incidence of primary
EBV infection tended to be slightly higher in the HIV-infected
children and it seemed to occur at a slightly earlier age in
the same population compared with these parameters in HIVuninfected children. However, the difference was not statistically significant.
Not all immunosuppressed patients develop VCA IgM antibodies in the course of a primary EBV infection [14]. However,
VCA IgM could not be detected in both some uninfected and
some HIV-infected children in our study. The IgM response
to EBV VCA is transient and lasts only 1 – 2 months, or even
less in very young children [15]. Therefore, the failure to detect
VCA IgM antibodies in most of our patients was likely a
function of the long intervals (in general 3 months) between
serial serum sampling.
There is evidence in the literature that in the context of
immunosuppression, EBV infection might occur without the
development of EBV-specific antibodies, particularly of antibodies to EBNA [14]. However, as described elsewhere, we
noted that HIV-infected children did develop EBNA antibodies
following primary EBV infection, except that the antibody response pattern was different from that observed in HIV-uninfected children, having ruled out maternal IgG antibodies [12].
Since no other case of EBV infection was documented when
PCR was done on PBMC specimens from EBV-seronegative
children, it appears that all EBV-infected children mounted
appropriate quantitative IgG responses against VCA and
EBNA, independent of their immune status.
UC: J Infect
Concise Communications
Persistent antibody to EA is believed to be evidence of
chronic active EBV infection in immunosuppressed patients
[14]. Following primary EBV infection, if they appear at all,
antibodies to EA should become undetectable within 3 – 6
months or persist at low titers. In our study, a slightly greater
proportion of HIV-infected children had EA titers §1:80 for
at least 6 months after primary EBV infection, although the
difference with EA titers found in HIV-uninfected children was
not significant. Last, as evidenced by EA titers within normal
ranges, mothers who transmitted HIV infection to their newborns did not seem to be experiencing EBV reactivation at a
higher rate than did mothers who gave birth to a child uninfected with HIV.
While EBV infection in normal children is generally asymptomatic, primary infection occurring in adolescents or adults
may give rise to the signs and symptoms of infectious mononucleosis. In our study, HIV-infected children were more often
symptomatic at the time of EBV seroconversion or shortly
thereafter and experienced splenomegaly and/or hepatomegaly
significantly more often than did their HIV-uninfected counterparts. However, hepatosplenomegaly is also a described feature
of HIV infection in children.
Interestingly, at least in the short term, HIV-1 RNA load
and HIV disease progression did not seem to be influenced by
primary EBV infection. The 3 deaths reported among HIVinfected children were, in this limited study, more likely the
consequence of HIV disease progression and less likely related
to primary infection with EBV.
The interpretation of our results may be somewhat limited
by the sample size of the study group. However, in an era in
which maternal-infant HIV transmission can be substantially
reduced by treating HIV-infected pregnant women with effective antiretroviral therapy, prospective natural history studies
in perinatally infected newborns may be a thing of the past,
giving a unique value to our findings.
We thank Marc Dumont for statistical analysis of the data and
Organon Teknika for providing the NucliSens HIV-1 QT kit.
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
1. Laurence J, Astrin SM. Human immunodeficiency virus induction of malignant transformation in human B lymphocytes. Proc Natl Acad Sci
USA 1991; 88:7635 – 9.
2. Mallon R, Borkowski J, Albin R, Pepitoni S, Schwartz J, Kieff E. The
Epstein-Barr virus BZLF1 gene product activates the human immunodeficiency virus type 1 5* long terminal repeat. J Virol 1990; 64:6282 – 5.
3. Birx DL, Redfield RR, Tosato G. Defective regulation of Epstein-Barr
virus infection in patients with acquired immunodeficiency syndrome
(AIDS) or AIDS-related disorders. N Engl J Med 1986; 314:874 – 9.
4. Sumaya CV, Boswell RN, Ench Y, et al. Enhanced serological and virological findings of Epstein-Barr virus in patients with AIDS and AIDSrelated complex. J Infect Dis 1986; 154:864 – 70.
5. Greenspan D, Greenspan JS. Significance of oral hairy leukoplakia. Oral
Surg Oral Med Oral Pathol 1992; 73:151 – 4.
6. Andiman WA, Eastman R, Martin K, et al. Opportunistic lymphoproliferations associated with Epstein-Barr viral DNA in infants and children
with AIDS. Lancet 1985; 2:1390 – 3.
7. MacMahon EME, Glass JD, Hayward SD, et al. Epstein-Barr virus in
AIDS-related primary central nervous system lymphoma. Lancet 1991;
338:969 – 73.
8. McClain KL, Leach CT, Jenson HB, et al. Association of Epstein-Barr
virus with leiomyosarcomas in young people with AIDS. N Engl J Med
1995; 332:12 – 8.
9. Centers for Disease Control and Prevention. 1994 revised classification
system for human immunodeficiency virus infection in children less
than 13 years of age. MMWR Morb Mortal Wkly Rep 1994; 43:1 – 10.
10. Henle G, Henle W. Immunofluorescence in cells derived from Burkitt’s
lymphoma. J Bacteriol 1966; 91:1248 – 56.
11. Henle W, Henle G, Niederman JC, Klemola E, Haltia K. Antibodies to
early antigens induced by Epstein-Barr virus in infectious mononucleosis. J Infect Dis 1971; 124:58 – 67.
12. Pedneault L, Lapointe N, Alfieri C, et al. Antibody responses to two
Epstein-Barr virus (EBV) nuclear antigens (EBNA-1 and EBNA-2) during EBV primary infection in children born to mothers infected with
human immunodeficiency virus. Clin Infect Dis 1996; 23:806 – 8.
13. Savoie A, Perpeˆte C, Carpentier L, Joncas J, Alfieri C. Direct correlation
between the load of Epstein-Barr virus – infected lymphocytes in the
peripheral blood of pediatric transplant patients and risk of lymphoproliferative disease. Blood 1994; 83:2715 – 22.
14. Ho M, Miller G, Atchison RW, et al. Epstein-Barr virus infections and
DNA hybridization studies in posttransplantation lymphoma and lymphoproliferative lesions: the role of primary infection. J Infect Dis 1985;
152:876 – 86.
15. Sumaya CV. Epstein-Barr virus infections in children. Curr Probl Pediatr
1987; 17:677 – 722.
UC: J Infect
Circulating Levels of RANTES in Human Immunodeficiency Virus Type 1
Infection: Effect of Potent Antiretroviral Therapy
Pa˚l Aukrust, Fredrik Mu¨ller, and Stig S. Frøland
Section of Clinical Immunology and Infectious Diseases and Research
Institute for Internal Medicine, Medical Department A, University of
Oslo, Rikshospitalet, Oslo, Norway
RANTES has been found to suppress human immunodeficiency virus type 1 (HIV-1) replication.
To further elucidate the role of this chemokine in HIV-1 infection, RANTES levels were analyzed
in serum and platelet-free plasma (PFP) in 53 HIV-1 – infected patients and 20 controls. RANTES
levels were significantly elevated in both serum and PFP in all clinical stages of HIV-1 infection,
with the highest levels in CDC groups A and B. In longitudinal testing, the progressors were
characterized by a pronounced decline in serum levels over time; the nonprogressors, however, had
only a slight reduction or an increase in RANTES levels. During 16 weeks of indinavir therapy,
there was an increase in circulating RANTES levels and enhanced release of RANTES from stimulated CD8/ lymphocytes. The decline in RANTES levels along with disease progression is compatible
with RANTES having a beneficial role in HIV-1 – infected patients. The increase in RANTES levels
during protease inhibitor – containing regimens may represent a previously unrecognized immunologic effect of such therapy.
RANTES, a member of the C-C chemokine family, is a
potent chemoattractant of memory T lymphocytes, monocytes,
eosinophils, and basophils [1] and may also activate specific
effector cell populations of these cells [1, 2]. Recently,
RANTES in combination with two other members of the C-C
chemokine family was found to suppress the replication of
macrophagetropic human immunodeficiency virus type 1 (HIV1) strains in vitro [3]. Of these chemokines, RANTES was
reported to be the most effective inhibitor of HIV-1 replication
[3]. However, unidentified soluble factors other than these three
C-C chemokines may also be involved in the CD8/ lymphocyte – mediated suppression of HIV-1 replication [4]. Furthermore, RANTES and other chemokines may augment inflammatory and nonbeneficial processes in HIV-1 – infected patients
[2], and the role of RANTES in the pathogenesis of HIV-1
infection remains to be established. There are some studies
reporting circulating levels of RANTES in HIV-1 – infected
patients, but the results are somewhat conflicting [5, 6]. To
better understand the possible role of RANTES in HIV-1 infection, we analyzed levels of RANTES in both serum and platelet-free plasma (PFP) in different clinical and immunologic
stages of HIV-1 infection, in both cross-sectional and longitudi-
Received 21 July 1997; revised 31 October 1997.
Informed consent for blood sampling was obtained from all patients and
Grant support: Norwegian Cancer Society, Research Council of Norway,
Anders Jahre’s Foundation, Medinnova Foundation, and Odd Ka˚re Rabben’s
Memorial Fund for AIDS Research.
Reprints or correspondence: Dr. Pa˚l Aukrust, Section of Clinical Immunology and Infectious Diseases, Medical Dept. A, Rikshospitalet, N-0027 Oslo,
Norway ([email protected]).
The Journal of Infectious Diseases 1998;177:1091–6
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
nal testing. We also examined the effect of potent antiretroviral
therapy on RANTES levels in HIV-1 – infected patients.
Materials and Methods
Patients and controls. Fifty-three HIV-1–seropositive patients
were included in the study (40 men, 13 women; median age, 40
years [range, 25–59]). The patients were clinically classified according to the 1993 revised criteria from the Centers for Disease
Control and Prevention (CDC). Patients with ongoing acute infection or exacerbation of chronic infection at the time of blood
collection were excluded. Forty-one patients received antiretroviral
therapy with nucleoside analogue(s), but none had begun therapy
within the previous 5 months. Controls were 20 healthy HIV-1–
seronegative blood donors (15 men, 5 women; median age, 37
years [range, 22–61]). Sampling of serum and PFP (EDTAtreated) were performed as previously described [7, 8].
Isolation and stimulation of CD8/ lymphocytes. Positive selection of CD8/ lymphocytes by monodisperse immunomagnetic
beads was done as described elsewhere [7]. The isolated CD8/
lymphocytes were detached from beads by incubation with goat
anti–mouse Fab antiserum (DetachAbead, CD4/CD8; Dynal,
Oslo) for 1 h at room temperature. Thereafter, CD8/ lymphocytes
(purity: 90%–95%, as determined by flow cytometry; final concentration, 106/mL) were incubated in flat-bottomed, 96-well microtiter trays (200 mL/well; Costar, Cambridge, MA) in medium alone
(RPMI 1640 with 2 mmol/L L-glutamine and 25 mmol/L HEPES
buffer [Gibco, Paisley, UK] and 5% fetal calf serum) or stimulated
with anti–CD3 monoclonal antibody (MAb; final concentration,
1.2 ng/mL; clone SpvT3b) in combination with anti–CD28 MAb
(final concentration, 50 ng/mL; clone 15E8 [402], CLB, Amsterdam). The cell surface markers were cross-linked using monodispersed immunomagnetic beads coated with sheep anti–mouse IgG
(Dynal) at a cell-to-bead ratio of 1:1. After a 72-h culture, supernatants were harvested and stored at 0807C until analysis.
UC: J Infect
Concise Communications
EIA for detection of RANTES. RANTES levels in serum, PFP,
and cell supernatants were measured by EIA (R&D Systems, Minneapolis) as previously described [9].
Quantification of HIV RNA copy numbers in plasma. HIV
RNA levels were measured in plasma by quantitative reverse polymerase chain reaction (Amplicor HIV Monitor; Roche Diagnostic
Systems, Branchburg, NJ). The limit of detection was 200 copies/
mL plasma.
Statistical analysis. For comparison of 2 groups, the MannWhitney U test (two-tailed) was used. When ú2 groups were
compared, the Kruskal-Wallis test was used as previously described [7]. For comparison of parameters within an individual,
the Wilcoxon signed rank test for paired data (two-tailed) was
used. Coefficients of correlation (r) were calculated by the Spearman rank test. Data are given as medians and 25th–75th percentiles
if not stated otherwise. P values are two-sided and considered
significant when P õ .05.
Circulating levels of RANTES in different clinical stages of
HIV-1 infection. HIV-1 – infected patients in all clinical stages
had significantly higher serum levels of RANTES than did
healthy controls, with the highest levels in CDC groups A and
B (figure 1A). In fact, patients in CDC group C had significantly
lower serum levels of RANTES than did patients in both CDC
groups A and B (figure 1A). Since RANTES may be released
from platelets during coagulation [9], we also measured
RANTES levels in parallel samples of PFP in 27 patients and
11 controls. We again found a marked increase in RANTES
levels in all clinical stages of HIV-1 infection, with the highest
levels in CDC groups A and B (figure 1B).
Relationship between RANTES levels and immunologic and
virologic parameters. Serum levels of RANTES were positively
correlated with both CD8/ (r Å .42, P õ .003) and CD4/ (r Å
.37, P õ .007) lymphocyte counts and negatively correlated with
plasma virus load (r Å 0.39, P õ .005). Similar correlations
were also found for RANTES levels in PFP (data not shown).
Longitudinal testing. Our findings in cross-sectional testing
suggest declining RANTES level with disease progression,
with the lowest levels in the AIDS group. We therefore conducted a longitudinal analysis of serum levels of RANTES in
14 HIV-1 – infected patients over a 5-year period. For the purpose of the present study, these patients were defined as nonprogressors (n Å 6) and progressors (n Å 8; see legend to
figure 2). At the first serum sampling, RANTES levels were not
significantly different between progressors and nonprogressors
(figure 2). However, while the progressors were characterized by a decline in serum levels over time (at least 50%),
the nonprogressors had only a slight reduction (õ15%)
or an increase in RANTES levels during the study period
(figure 2).
The effect of indinavir therapy on RANTES level. We next
analyzed serum levels of RANTES before and during therapy
with the HIV-1 protease inhibitor indinavir (800 mg three times
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
daily) combined with zidovudine (250 mg twice daily) and
lamivudine (150 mg twice daily) in 14 patients. As described
by others [10, 11], there was a marked fall in HIV-1 RNA
copies in plasma (median decrease, 2.5 log10 [1.6 – 3.2 log10])
and a marked rise in numbers of CD4/ (median increase, 62 1
106/L [35 – 115 1 106/L]: Ç180% increase) and CD8/ (median
increase, 290 1 106/L [110 – 400 1 106/L]: Ç130% increase)
lymphocytes after initiating potent antiretroviral therapy. Concomitantly, there was a rapid and marked increase in serum
levels of RANTES in all but 1 patient, although all had elevated
RANTES levels at baseline compared with controls (figure
3A). In 5 of the patients, parallel samples of PFP were obtained,
and a similar pattern was again found, with a significant rise
in RANTES levels after therapy was initiated (data not shown).
There was a significant positive correlation between the maximum increase in RANTES level and the maximum increase
in CD8/ lymphocyte counts (r Å .53, P õ .05), but not with
a maximum increase in CD4/ lymphocyte counts (data not
shown). Also, the reduction in virus load tended to be correlated
with the rise in RANTES level (r Å 0.48, P Å .07).
We also analyzed the release of RANTES in supernatants from
unstimulated and anti–CD3/anti–CD28-stimulated CD8/ lymphocytes during indinavir therapy in 6 of the patients and found
a marked increase in stimulated release of RANTES from CD8/
lymphocytes after 4 weeks of therapy (figure 3B). No differences
were found in unstimulated release of RANTES (data not shown).
The present study demonstrates for the first time markedly
elevated levels of RANTES in HIV-1 – infected patients in both
serum and PFP, which decline in accordance with disease progression (demonstrated on both cross-sectional and longitudinal
testing). Furthermore, our results indicating increased
RANTES levels with enhanced release from CD8/ lymphocytes during HIV-1 protease inhibitor therapy represent a previously unrecognized immunologic effect of such therapy.
Previous studies of serum levels of RANTES in HIV-1 infection have reported normal [5] or increased levels, with the
highest levels in the AIDS group [6]. This apparent discrepancy
with our findings may have several explanations. The term
AIDS denotes a heterogeneous group of patients both clinically
and immunologically, and discrepant results of RANTES levels
in AIDS patients may partly reflect this heterogeneity. In the
present study, most of the AIDS patients were severely immunosuppressed (CD4/ lymphocytes, õ50 1 106/L) with advanced clinical disease, and it is possible that the decline in
RANTES levels during HIV-1 infection is restricted to this
subgroup of AIDS patients. In fact, our findings in longitudinal
testing suggest a decline in RANTES levels with disease progression even within the AIDS group.
The release of RANTES from platelets during coagulation
may also influence the measured RANTES level in serum samples [9]. For example, if serum is collected before coagulation
UC: J Infect
JID 1998;177 (April)
Concise Communications
Figure 1. Circulating levels of RANTES in HIV-1 infection. Levels of RANTES in (A) serum and (B) platelet-free plasma (PFP) in 3
clinical groups of HIV-1 – infected patients and healthy controls. Patients in CDC group C had significantly lower RANTES levels in both
serum and PFP than did other HIV-1 – infected patients; P õ .05 vs. both CDC groups A (asymptomatic HIV-1 – infected patients) and B
(symptomatic non-AIDS HIV-1 – infected patients); CDC group C, AIDS patients. Bars represent median values.
is completed, lower RANTES levels will be found (Aukrust P,
Mu¨ller F, Frøland SS, unpublished observations). Confounding
effects of ex vivo release of RANTES from platelets may be
eliminated if analyzing in PFP. In the present study, however,
simultaneous analyses of RANTES levels in serum and PFP
gave similar patterns of results.
We found a significant positive correlation between
RANTES levels and CD8/ lymphocyte counts in peripheral
blood, but this does not necessarily mean that serum levels
merely reflect the release of RANTES from CD8/ lymphocytes. RANTES may be produced by both CD4/ lymphocytes
and monocytes [1], and CD8- and CD4-depleted peripheral
blood mononuclear cells from HIV-1 – infected patients were
found to produce comparable levels of RANTES under most
/ 9d43$$ap48
03-04-98 19:03:33
conditions [12]. Furthermore, as discussed above, platelets may
be an important cellular source of RANTES [1, 9], and we
have demonstrated enhanced release of RANTES upon stimulation in platelets from asymptomatic HIV-1 – infected patients
[9]. Thus, the high circulating RANTES levels in HIV-1 – infected patients may reflect altered release from other cellular
sources than CD8/ lymphocytes.
A striking finding in the present study was the marked increase
in RANTES levels after indinavir therapy was initiated. Bisset et
al. [13] found a slight increase in RANTES levels during indinavir
therapy in HIV-1–infected patients. However, they measured
RANTES levels in plasma without removing platelets, a procedure
which the manufacturer of the EIA does not recommend, as such
results may be unreliable. In the present study, increasing
UC: J Infect
Concise Communications
JID 1998;177 (April)
Figure 2. RANTES levels in serum in 2 groups of HIV-1 – infected patients. Six nonprogressors (A) remained clinically (CDC group A or
B), immunologically (CD4/ lymphocyte counts of 250 – 500 1 106/L), and virologically (õ0.5 log10 increase in HIV RNA copies in plasma
and always õ50 1 103/mL) stable throughout study. Eight progressors (B) either died of Mycobacterium avium complex infection within 3
years (n Å 6, all in AIDS group at baseline) or demonstrated marked deterioration clinically (development of AIDS from CDC group A
classification), immunologically (CD4/ lymphocyte decline, from ú200 1 106/L to õ10 1 106/L), and virologically (ú1 log10 increase in
HIV RNA copies in plasma to ú106/mL) during study (n Å 2). Hatched region represents ranges of RANTES levels in healthy controls.
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
JID 1998;177 (April)
Concise Communications
Figure 3. RANTES levels during potent antiretroviral therapy with indinavir in combination with zidovudine and lamivudine. A, Serum
levels of RANTES during 16 weeks of treatment in 14 HIV-1 – infected patients. Hatched region represents range of RANTES levels in healthy
controls. * P õ .05, ** P õ .01 vs. baseline. B, Release of RANTES from CD8/ lymphocytes stimulated with anti – CD3 and anti – CD28
antibodies (see Materials and Methods) in 6 HIV-1 – infected patients before and after 4 weeks of treatment.
RANTES levels were found both in serum and PFP. Moreover,
the increase in RANTES levels was seen, although all patients
had RANTES levels elevated over those of healthy controls at
baseline. Thus, rather than normalization of RANTES levels, a
further enhancement is seen with potent antiretroviral therapy. It
is interesting that there was a significant correlation between the
magnitude of the CD8/ lymphocyte response and the RANTES
response. Furthermore, we found a marked increase in
CD3/CD28/-stimulated release of RANTES per CD8/ lymphocyte after initiating indinavir therapy, possibly at least partly reflecting increased percentage of CD28/CD8/ lymphocytes. AntiCD3/CD28/ stimulation of CD8/ lymphocytes has recently been
reported to be a potent inducer of RANTES release from these
cells, possibly resulting in enhanced antiviral activity [14]. Thus,
increased circulating RANTES levels and enhanced secretion of
this chemokine from CD8/ lymphocytes might result in improved
suppression of HIV-1 replication by the immune system, representing an important beneficial effect of HIV-1 protease inhibitor–
containing regimens.
Decreased antiretroviral effects of C-C chemokines in advanced HIV-1 infection have been reported, associated with
acquisition of the ability to use CXCR-4 as a coreceptor for
infection in advanced disease [15]. In the present study, we
found an inverse correlation between circulating RANTES levels and virus load in HIV-1 – infected patients and, more importantly, a decrease in RANTES levels and an increase in virus
load along with disease progression in longitudinal testing. It
is tempting to hypothesize that in addition to the increase in
/ 9d43$$ap48
03-04-98 19:03:33
use of CXCR-4 as a coreceptor, the decrease in RANTES levels
may also favor enhanced HIV-1 replication in patients with
advanced disease. However, the association between high virus
load and a decline in RANTES levels may also reflect suppressive effects of overwhelming HIV-1 replication on
RANTES secretion from various cells in these patients.
Our findings in this study of a decline in circulating RANTES
levels along with disease progression in HIV-1–infected persons,
both in cross-sectional and longitudinal testing, and a rise in
RANTES levels during HIV-1 protease inhibitor therapy are compatible with a beneficial role for RANTES in HIV-1–infected
patients. However, further studies are needed to clarify the immunologic and virologic roles of dysregulated RANTES levels in
the pathogenesis of HIV-1 infection.
We thank Bodil Lunden, Lisbeth Wikeby, and Vigdis Bjerkeli
for excellent technical assistance.
1. Baggiolini M, Dewald B, Moser B. Interleukin-8 and related chemotactic
cytokines — CXC and CC chemokine. Adv Immunol 1994; 55:97 – 179.
2. Strieter RM, Standiford TJ, Huffnagle GB, Colletti LM, Lukacs NW,
Kunkel SL. The good, the bad, and the ugly. The role of chemokines
in human diseases. J Immunol 1996; 356:3583 – 6.
3. Cocchi F, DeVico AL, Garzino-Demo A, Arya SK, Gallo RC, Lusso P.
UC: J Infect
Concise Communications
Identification of RANTES, MIP-1a, and MIP-1b as the major HIVsuppressive factors produced by CD8/ T cells. Science 1995; 270:
1811 – 5.
Levy JA, Mackewicz CE, Barker E. Controlling HIV pathogenesis: the
role of the noncytotoxic anti-HIV response of CD8/ T cells. Immnol
Today 1996; 17:217 – 24.
McKenzie SW, Dallalio G, North M, Frame P, Means RT Jr. Serum
chemokine levels in patients with non-progressive HIV infection. AIDS
1996; 10:F29 – F33.
Zanussi S, D’Andera M, Simonelli C, Tirelli U, De Paoli P. Serum levels
of RANTES and MIP-1a in HIV-positive long-term survivors and progressors patients. AIDS 1996; 10:1431 – 2.
Aukrust P, Svardal AM, Mu¨ ller F, Lunden B, Berge RK, Ueland
PM, Frøland SS. Increased levels of oxidized glutathione in CD4 /
lymphocytes associated with disturbed intracellular redox balance
in human immunodeficiency virus type 1 infection. Blood 1995; 86:
258 – 67.
Aukrust P, Hestdal K, Lien E, et al. Effects of intravenous immunoglobulin
in vivo on abnormally increased tumor necrosis factor-a activity in
patients with human immunodeficiency virus type 1 infection. J Infect
Dis 1997; 176:913 – 23.
Holme PA, Mu¨ller F, Solum NO, Brosstad F, Frøland SS, Aukrust P.
Enhanced activation of platelets with abnormal release of RANTES in
HIV-1 infection. FASEB J 1998 (in press).
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
10. Kelleher AD, Carr A, Zaunders J, Cooper DA. Alterations in the immuneresponse of the human immunodeficiency virus (HIV) – infected subject
treated with a HIV-specific protease inhibitor, ritonavir. J Infect Dis
1996; 123:321 – 6.
11. Hammer SH, Squires KE, Hughes MD, et al. A controlled trial of two
nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cell counts of 200 per cubic millimeter
or less. N Engl J Med 1997; 337:725 – 33.
12. Kinter AL, Ostrowski M, Goletti D, et al. HIV replication in CD4/ T cells
of HIV-infected individuals is regulated by a balance between the viral
suppressive effects of endogenous b-chemokines and the viral inductive
effects of other endogenous cytokines. Proc Natl Acad Sci USA 1996;
93:14076 – 81.
13. Bisset LR, Rothen M, JollerJemelka HI, Dubs RW, Grob PJ, Opravil M.
Change in circulating levels of the chemokines macrophage inflammatory proteins 1a and 1b, RANTES, monocyte chemotactic protein-1 and
interleukin-16 following treatment of severely immunodeficient HIVinfected individuals with indinavir. AIDS 1997; 11:485 – 91.
14. Moriuchi H, Moriuchi M, Fauchi AS. Nuclear factor-kB potently upregulates the promotor activity of RANTES, a chemokine that blocks
HIV infection. J Immunol 1997; 158:3483 – 91.
15. Connor RI, Sheridan KE, Ceradini D, Choe S, Landau NR. Change in
coreceptor use correlates with disease progression in HIV-1 – infected
individuals. J Exp Med 1987; 185:621 – 8.
UC: J Infect
Intrapartum Mucosal Exposure to Human Immunodeficiency Virus Type 1
(HIV-1) of Infants Born to HIV-1–Infected Mothers Correlates with Maternal
Plasma Virus Burden
Mounir Ait-Khaled, E. G. Hermione Lyall, Chris Stainsby,
Graham P. Taylor, Anne Wright, Jonathan N. Weber,
Myra O. McClure, and Gareth Tudor-Williams
Department of Genito-Urinary Medicine and Communicable Diseases,
Jefferiss Research Trust Laboratories; Department of Paediatrics,
Imperial College School of Medicine at St. Mary’s, London,
United Kingdom
The majority of vertical infections with human immunodeficiency virus type 1 (HIV-1) occur at
or near delivery, strongly suggesting a mucosal route of transmission. The frequency and level of
intrapartum mucosal exposure to HIV-1 of 22 infants born to infected mothers was investigated.
Maternal plasma HIV-1 RNA and CD4 cell count were measured at delivery. Infant oropharyngeal
aspirates obtained at birth were examined for HIV-1 RNA by reverse transcription – polymerase
chain reaction and qualitative nucleic acid sequence – based amplification. Nine infants (41%) had
detectable levels of HIV-1 RNA, 3 of which were quantifiable (mean, 3000 copies/mL). This mucosal
exposure to HIV-1 during delivery did not lead to infection of any infant. Cesarian delivery did not
reduce mucosal exposure to HIV-1. Mucosal exposure did not correlate with maternal CD4 cell
count but did correlate with maternal plasma virus load and was reduced by antiretroviral therapy.
Human immunodeficiency virus type 1 (HIV-1) vertical
transmission may occur in utero, during delivery, or through
breast-feeding. In a non – breast-feeding population, transmission occurs predominately (70%) during delivery. This and
other lines of evidence point to the importance of oral mucosal
exposure in HIV-1 vertical transmission [1], the source of virus
being maternal blood, cervicovaginal secretions (CVS), or both.
By using polymerase chain reaction (PCR) – based techniques,
HIV RNA can be detected in CVS of 40% – 75% of HIVinfected women [2]. The duration of infant exposure to the
birth canal as a risk factor for HIV transmission has been
clearly demonstrated in the Women and Infants Study [3] and
the New York Study [4], which showed an almost 2-fold increase in transmission when the fetal membranes were ruptured
for ú4 h before delivery. Moreover, the presence of HIV in
the gastric aspirates of newborns could be a risk factor for
perinatal transmission [5]. Oral infection can also occur postnatally and is well documented in breast-fed neonates [6].
In rhesus monkeys, oral administration of cell-free simian
immunodeficiency virus (SIV) to newborn animals led to infec-
tion, demonstrating virus transmission via the oral mucosa,
although it is important to note the viral infectious dose was
300 times that required for intravenous transmission [7]. This
may reflect the physical and immunologic barrier characteristics of the intact upper gastrointestinal mucosa. More evidence
for the requirement of a high infectious dose for mucosal transmission comes from studies of HIV-1 – infected discordant
twins, in whom the first born is exposed to maternal fluids
(CVS, blood) for longer and carries the higher risk of infection,
regardless of the mode of delivery [8].
To date, a quantitative assessment of intrapartum mucosal exposure of neonates to HIV-1 using sensitive molecular techniques
has not been done. In the present study, oropharyngeal aspirates
(OPA) were collected immediately after birth and assayed for
HIV-1 RNA. The study aimed to examine the frequency and level
of intrapartum exposure to HIV-1, as defined by the detection of
HIV-1 RNA in infants’ OPA, and the relationship of those factors
to maternal stage of disease and mode of delivery. Such investigations may shed light on the parameters that influence mucosal
exposure to HIV-1 during delivery and help in the design and
monitoring of birth canal interventions.
Received 6 August 1997; revised 21 October 1997.
Presented: Conference on Global Strategies for the Prevention of HIV Transmission from Mothers to Infants, Washington, DC, September 1997.
Informed written consent was obtained from all mothers enrolled in the
study. The study was reviewed by and acceptable to the Ethical Committee of
St. Mary’s Hospital, London.
Grant support: Pediatric AIDS Foundation (50620-20-PG) and St. Mary’s
Special Trustees (Project E).
Reprints or correspondence: Dr. E. G. Hermione Lyall, Dept. of Paediatrics,
Imperial College School of Medicine at St Mary’s, Norfolk Place, London,
W2 1PG, UK ([email protected]).
Materials and Methods
The Journal of Infectious Diseases 1998;177:1097–100
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
Study design. The level of HIV-1 RNA in OPA of neonates
born to HIV-1–infected women receiving prenatal care at St. Mary’s
Hospital, London, was analyzed prospectively. Mothers were recruited into the study antenatally and their clinical, laboratory, and
therapeutic profiles collected. Twenty of the 23 mothers were of
African origin and the median age of the cohort was 28 years (range,
20–35). All infants were followed prospectively from birth, and HIV1 infection status was determined by viral culture and PCR.
OPA. OPA were collected by vacuum suction of the neonates’
mouths immediately after birth. The volumes of OPA collected
UC: J Infect
Concise Communications
ranged from 200 mL to 2 mL. Within 2 h of collection, OPA were
equally divided into 100-mL aliquots and centrifuged at 12,000 g
for 5 min at room temperature. Electron microscopy studies have
shown that HIV-1 particles form aggregates in adult human salivary mucins [9]. OPA were therefore centrifuged after collection,
and HIV-1 RNA was assessed in both supernatants and pellets.
The supernatants and pellets were either stored at 0707C until
used or immediately processed for RNA extraction using the nucleic acid sequence-based amplification (NASBA; Organon Teknika, Boxtel, The Netherlands) method according to the manufacturer’s instructions. For both NASBA and reverse transcription
(RT) PCR amplification, low positive (100 HIV-1 RNA copies/100
mL) and negative materials extracted in parallel with the clinical
specimens were used as controls.
HIV-1 RNA detection. HIV-1 RNA amplification was carried
out using both an in-house RT-PCR and the NASBA assay. All
OPA supernatants and pellets positive in the NASBA HIV-1 RNA
qualitative assay and all maternal plasma samples collected at
delivery were subsequently quantified using the quantitative
NASBA assay (calibrators diluted 1/10). This assay has a threshold
detection limit of 1300 RNA copies/mL.
The RT-PCR was carried out using nested primer sets based on
conserved regions of the HIV-1 pol gene and are as follows (relative
(nt 4165–4186); outer anti-sense, 5*-TCTACTTGTCCATGCATGGCTTC (nt 4386–4408); inner sense, 5*-AAAGGAATTGGAGGAAATGTGAACAAGTAG (nt 4182–4209); and inner anti-sense,
5*-CACTAGCCATTGCTCTCCAATTAC (nt 4294–4317) (137-bp
Extracted RNA (5 mL) was subjected to cDNA synthesis for 1
h at 427C in a reaction volume of 20 mL containing 6 mM outer
anti-sense primer, 11 first-strand buffer (Gibco-BRL, Gaithersburg,
MD), 20 U of RNasin (Promega, Madison, WI), 10 mM dithiothreitol (Gibco-BRL), each dNTP at 500 mM (Pharmacia, Milton Keynes,
UK), and 100 U of reverse transcriptase (MMLV-RT Superscript;
Gibco-BRL). cDNA (5 mL) was used for the nested PCR, which
was carried out in quadruplicate. The first-round PCR mixture contained 11 buffer II (Perkin Elmer, Norwalk, CT), 1.5 mM MgCl2,
each dNTP at 30 mM, 10 pmol of each outer primer, and 2 U of
Taq polymerase (Perkin Elmer). First-round amplification product
(1.5 mL) was nested in a second-round PCR (20 mL) carried out
under the same conditions as the outer PCR but containing 5 pmol
of each inner primer. The cycling conditions for the first round were
947C for 30 s, 457C for 42 s, 727C for 42 s (30 cycles), and a final
extension at 727C for 10 min. For the second round, the annealing
temperature was increased to 527C. Amplification reactions were
carried out on a thermocycler (model 480; Perkin Elmer). The nested
amplification products (5 mL) were visualized on a 2% agarose gel
by ethidium bromide staining.
Lower detection limit of HIV-1 RNA by amplification using OPA.
All RNA isolation was carried out using the NASBA assay, which
includes an RNA internal control in each extraction tube. Amplification of this control acts as an indicator of the extraction efficiency
and a marker for the presence of potential inhibitors in the clinical
HIV-1–negative OPA of different consistencies that fell into three
categories (blood-stained, meconium-stained, and clear) were spiked
with a fixed number of HIV-1 RNA molecules and used in control
experiments to define the lower detection limits of both the qualitative
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
NASBA and RT-PCR assays. Extractions and amplifications of the
three categories of OPA with each dilution (100, 50, and 25 copies/
100 mL) were carried out in triplicate to determine whether the quality
of the OPA affects the sensitivity of the assays.
Statistical analysis. Medians were compared by Mann-Whitney U test, and the proportions of patients in distinct groups were
compared using Fisher’s exact test.
Study population. The route of maternal infection was presumed to be heterosexual, with the exception of 2 mothers of
Caucasian origin who had used intravenous drugs. Only 1
mother had AIDS, the remainder being asymptomatic. Seven
mothers previously had HIV-infected children and 6 mothers
were primiparous. Twelve mothers had received zidovudine
monotherapy according to the AIDS Clinical Trials Group 076
protocol [10], and 4 mothers received zidovudine and lamivudine. One mother-infant pair was excluded from further analysis
because the maternal viral variant could not be detected by
NASBA or RT-PCR. Subtyping of this HIV-1 strain by heteroduplex mobility assay [11] showed this variant to be of subtype
G. None of the 22 infants were HIV-infected.
HIV-1 RNA detection in OPA. For OPA, the lower detection limit of the NASBA is 25 – 50 HIV-1 RNA copies/100 mL
compared with 50 – 100 copies/100 mL in RT-PCR. In NASBA
of samples spiked with 50 and 25 copies, RNA was detected
in 100% and 55% of cases, respectively. In RT-PCR of samples
spiked with 100, 50, and 25 copies, RNA was detected in
100%, 78%, and 22%, respectively, regardless of the nature or
appearance of the OPA.
HIV-1 RNA was detected in the OPA of 9 (41%) of the 22
infants. Two OPA were positive for HIV-1 RNA only in the
pelleted fraction, 4 were positive only in the supernatant, and
3 were positive in both fractions. Three of the 9 positive OPA
were negative by RT-PCR, perhaps due to its lower sensitivity.
Four of the 22 pellets tested were inhibitory to the NASBA
assay, as the internal controls could not be detected; this is
probably due to excess protein overloading the silica matrix
during the RNA isolation. The discrepant results between supernatants and pellets might reflect the variable constitution
of clinical specimens such as OPA, with more viscous and
proteinaceous OPA confining virus particles to the pelleted
fraction by aggregation and entrapment.
Three OPA (2 supernatants, 1 pellet) were quantifiable by
NASBA with loads of 4100, 2730, and 2333 HIV RNA copies/
mL, respectively. Two of these quantifiable OPA corresponded
to the mothers with high virus load at delivery (130,000 and
57,000 HIV RNA copies/mL). It is important to note that 7 of
9 infants with detectable HIV-1 RNA in OPA were delivered
by elective cesarian section, which suggests that the source of
virus was maternal blood rather than maternal cervicovaginal
secretions, as these infants did not come into contact with the
birth canal. At these levels of mucosal exposure to the maternal
virus, none of the 9 infants proved to be infected.
UC: J Infect
JID 1998;177 (April)
Concise Communications
Table 1. Detection of HIV-1 RNA in oropharyngeal aspirates (OPA) of neonates and maternal
HIV-1 RNA in infant OPA
Maternal factors
CD4 cell count 1 106/L*
Spontaneous vaginal delivery
Delivery with labor§
Perinatal antiretroviral therapy
Plasma virus load, log10 RNA
Positive (n Å 9)
Negative (n Å 13)
275 (153 – 380)
390 (20 – 700)
4 (2.9 – 5.1)
3.1 (2.9 – 4.7)
NOTE. Nos. shown for CD4 cell count and plasma virus load from maternal samples obtained at delivery are
median (range).
* One CD4 cell count from HIV-1 – positive OPA group was not available.
Mann-Whitney U test.
Fisher’s exact test.
Spontaneous vaginal deliveries and emergency cesarian sections.
Maternal and obstetric factors. The median maternal
plasma virus load at delivery was significantly higher for the
group in whom infant OPA was HIV RNA – positive (4.0 log10
RNA copies/mL) than for the group in whom neonate OPA
was HIV RNA – negative (3.1 log10 RNA copies/mL) (MannWhitney U test, P õ .05). Antiretroviral therapy significantly
reduced the incidence of mucosal exposure; 4 of the 9 mothers
of infants with positive OPA had received antiretrovirals, compared with 12 of 13 mothers whose infants had negative OPA
(Fisher’s exact test, P Å .023). There was no correlation between OPA positivity and mode of delivery or median maternal
CD4 cell count at delivery. Table 1 summarizes the relationships between an HIV-1 – positive OPA and maternal factors.
The presence in 22 neonates of HIV-1 RNA in OPA obtained
immediately at delivery was used as a marker of mucosal exposure to HIV-1 during delivery. We detected HIV-1 RNA in the
OPA of 41% (n Å 9) of infants born to HIV-1 – infected women.
The virus load was low, with 6 of the 9 having levels below
the threshold of the NASBA quantitative assay (1300 HIV
RNA copies/mL) and a mean load of 3000 HIV RNA copies/
mL in the remaining 3. This level of exposure did not result
in infection of the infants. This observation is in accordance
with the SIV model where only high virus titers induce infection of newborn monkeys via the oral route [7].
The fact that none of the 22 infants was found to be infected
may be due to the relatively low maternal virus burden of this
cohort and the widespread use of perinatal antiretroviral therapy
(16/22). Detection of HIV-1 RNA in the mouth of neonates
immediately after delivery correlated with maternal virus load.
Previous studies have highlighted the lack of direct correlation
between peripheral virus load and CVS viral shedding [4, 12],
so correlation analysis could be confounded by the absence of
/ 9d43$$ap48
03-04-98 19:03:33
virus load measurement in the CVS of these women. However,
in our study, 7 of 9 infants with HIV-positive OPA were delivered by elective cesarian section, which suggests that in those
cases, the source of virus was maternal blood.
Our pilot study shows that virus can be found in ú40% of
neonates’ OPA obtained at delivery. Intrapartum exposure to
a detectable level correlated with maternal plasma HIV burden,
and perinatal antiretroviral therapy was significantly associated
with a reduced incidence of oral exposure to HIV-1. For countries where antiretroviral therapy is not affordable, development
of birth canal preventative interventions is important. Biggar
et al. [13] undertook such a study in Malawi, using vaginal
cleansing with chlorhexidine solution every 4 h during labor.
Overall, there was no difference in the rate of HIV transmission
between the intervention and control groups (transmission rate
with chlorhexidine, 27%; without chlorhexidine, 28%). However, when the fetal membranes were ruptured for ú4 h, vaginal cleansing appeared to have an effect (transmission rate with
chlorhexidine, 25%; without chlorhexidine, 39%; P Å .02).
Although further studies will be necessary to correlate the detection of HIV-1 RNA in OPA with a higher risk of transmission, our technique could be used to assess the effect of birth
canal interventions designed to prevent perinatal transmission.
We thank the mothers and their infants who took part in this
study; St. Mary’s Hospital Trust midwives, S. Dick and M. Horby,
for their assistance with the management of the study; and R.
Pitman for providing the primers for the HIV-1 pol RT-PCR.
1. Lyall EGH, Tudor-Williams G. Perinatal transmission of HIV. Curr Opin
Infect Dis 1997; 10:239 – 45.
UC: J Infect
Concise Communications
2. O’Shea S, de Ruiter A, Mullen J, et al. Quantitation of HIV-1 RNA in
cervicovaginal secretions: an improved method for sample collection.
AIDS 1997; 11:1056 – 8.
3. Landesman SH, Kalish LA, Burns DN, et al. Obstetric factors and the
transmission of HIV-1 from mother to child. N Engl J Med 1996; 334:
1617 – 23.
4. Kuhn L, Abrams E, Matheson P, Thomas P, New York City Perinatal
Transmission Collaborative Study Group. Timing of maternal-infant HIV transmission: associations between intrapartum factors
and early polymerase chain reaction results. AIDS 1997; 11:
429 – 35.
5. Nielsen K, Boyer P, Dillon M, et al. Presence of human immunodeficiency
virus (HIV) type 1 and HIV-1 – specific antibodies in cervical secretions
of infected mothers and in the gastric aspirates of their infants. J infect
Dis 1996; 173:1001 – 4.
6. Dunn D, Newell ML, Ades A, Peckham C. Risk of HIV-1 Transmission
through breast feeding. Lancet 1992; 340:585 – 8.
7. Baba TW, Koch J, Mittler ES, et al. Mucosal infection of neonatal
rhesus monkeys with cell free SIV. AIDS Res Hum Retroviruses
1994; 10:351 – 7.
JID 1998;177 (April)
8. Duliege AM, Amos CI, Felton S, Biggar RJ, Goedert JJ. Birth order,
delivery route, and concordance in the transmission of HIV-1 from
mothers to twins. J Pediatr 1995; 126:625 – 32.
9. Bergey EJ, Cho MI, Blumberg BM, et al. Interaction of HIV-1 and human
salivary mucins. J Acquir Immune Defic Syndr 1994; 7:995 – 1002.
10. Connor EM, Sperling RS, Gelber R, et al. Reduction of maternal-infant
transmission of HIV-1 with zidovudine treatment. N Engl J Med 1994;
331:1173 – 80.
11. Bachmann MH, Delwart EL, Shpaer EG, Lingenfelter P, Singal R, Mullins
JI. Rapid genetic characterization of HIV type 1 strains from four World
Health Organization – sponsored vaccine evaluation sites using a heteroduplex mobility assay. WHO Network on HIV-isolation and Characterization. AIDS Res Hum Retroviruses 1994; 10:1345 – 53.
12. Rasheed S, Li Z, Xu D, Kovacs A. Presence of cell free human immunodeficiency virus in cervicovaginal secretions is independent of viral load
in the blood of human immunodeficiency virus – infected women. Am
J Obstet Gynecol 1996; 175:122 – 130.
13. Biggar RJ, Miotti PG, Taha TE, et al. Perinatal intervention trial in Africa:
effect of a birth canal cleansing intervention to prevent HIV transmission. Lancet 1996; 347:1647 – 50.
Human Immunodeficiency Virus Type 1 RNA Shedding in the Female Genital
Claudia Goulston,* William McFarland,
and David Katzenstein
Division of Infectious Diseases and Geographic Medicine, Stanford
University, Stanford, and Center for AIDS Prevention Studies,
University of California, San Francisco, California
Cervical and plasma samples obtained twice, at 2-week intervals, from 49 human immunodeficiency virus type 1 (HIV-1) – positive women were assayed for HIV-1 RNA. More than 100 copies
of HIV-1 RNA were detected in cervical swab supernatants (CSS) from 24 (49%) of 49 women.
HIV-1 RNA in CSS was detected in younger women with higher levels of plasma HIV-1 RNA
(median, 31,984 vs. 2880 copies/mL; P Å .0004), lower CD4 cell counts (median, 190 vs. 390 per
mm3; P Å .012), and lower CD4 cell percents (median, 16% vs. 25%; P Å .03). In multiple logistic
regression analysis, only plasma HIV-1 RNA was significantly associated with CSS HIV-1 RNA,
with an odds ratio of 4.79/log10 increase in plasma HIV-1 RNA (95% confidence interval, 1.4 – 16;
P Å .01). Detection of HIV-1 RNA in cervical secretions is primarily associated with increased
plasma HIV-1 RNA.
The contributions of human immunodeficiency virus type 1
(HIV-1) – infected cells and cell-free virus in the genital tract to
sexual and vertical transmission of HIV-1 are not fully known.
Studies of female genital tract secretions have identified the
Received 21 May 1997; revised 21 August 1998.
Informed consent was obtained from all subjects participating in the study.
Conduct of the study conformed to human experimentation guidelines of the
US Department of Health and Human Services.
Grant support: NIH (AI-35538, AI-35539).
Reprints or correspondence: Dr. David Katzenstein, Division of Infectious
Diseases and Geographic Medicine, S156 Grant Bldg., Stanford University,
Stanford, CA 94305-5107 ([email protected]).
* Current affiliation: Peace Health Medical Group, Eugene, Oregon.
The Journal of Infectious Diseases 1998;177:1100–3
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
presence of HIV-1 by culture and polymerase chain reaction
(PCR) amplification of HIV-1 proviral DNA [1 – 8]. Detection
of cell-associated HIV-1 DNA in the genital tract has been
associated with oral contraceptive use, cervical mucopus and
inflammation, and genital ulceration [4 – 6]. Cell-free virus has
been detected in genital tract secretions in 20% – 50% of women
by culture of cell-free fluids [1 – 3]. Amplification of HIV-1
RNA in plasma and tissues provides important data on viral
replication and disease progression in HIV-1 infection [9 – 12].
Reverse transcription (RT) – PCR quantification of HIV-1 RNA
from plasma may be more sensitive than culture in the detection
and quantification of virus [11 – 13].The present study assessed
HIV-1 RNA in the plasma and in cervical secretions of HIV1 – positive women by use of an internally controlled RT-PCR
assay to quantify HIV-1 RNA [14]. Women were studied at
two time points to determine if there were differences in plasma
UC: J Infect
JID 1998;177 (April)
Concise Communications
or genital HIV-1 RNA in samples obtained early and late in
the menstrual cycle.
Materials and Methods
Study subjects. Fifty HIV-1–positive women were recruited
at Stanford University Medical Center and the San Mateo County
HIV Clinic from March to October 1995. After consenting to
participate, women completed a self-administered questionnaire,
which included questions regarding demographics, current medications, and medical history. Examinations were scheduled early and
late in each woman’s menstrual cycle, except for 1 woman who had
only one examination late in her menstrual cycle. For 4 women, the
timing of the menstrual cycle could not be determined; 3 had
not menstruated in the last 3 months, and 1 woman had had a
hysterectomy and was excluded from analysis.
Specimens. Subjects underwent two pelvic examinations by a
single examiner (C.G.) 2 weeks apart and were assessed for lesions,
discharge, inflammation, and blood. Cervical HIV-1 RNA specimens were collected by gently inserting a Dacron swab (Baxter
Healthcare, McGaw Park, Il) 1 cm into the cervical os, rotating
the swab 3607, and then placing it in a sterile tube containing 2
mL of RPMI with 1% L-glutamine, 20% fetal bovine serum, and
1% penicillin plus streptomycin. Cervical swabs were processed
by vortexing the tubes and rotating the swabs against the inside
of the tubes to express the remaining supernatant. At the first
examination only, Pap smear and screening for sexually transmitted diseases were performed after collection of the cervical
HIV-1 RNA specimen. Pap smear results were interpreted using
the Bethesda system for cytologic diagnosis [14]. Cervical swabs
were tested for Chlamydia trachomatis DNA (Amplicor PCR Diagnostics Chlamydia; Roche Diagnostic Systems, Branchburg, NJ)
and Neisseria gonorrhea by culture on Thayer-Martin medium.
Blood for plasma HIV-1 RNA quantification was collected in acidcitrate-dextran at each examination. CD4 and CD8 cell counts
were done at the first examination only. All specimens were processed within 4 h of collection.
HIV-1 RNA assays. Two hundred microliters of cervical swab
supernatant (CSS) and plasma was centrifuged for 10 min at 800
g and stored at 0707C. Clarified plasma and CSS were assayed
according to the manufacturer’s instructions, using the Amplicor
HIV-1 monitor kit (provided by Beverly Dale, Roche Diagnostic
Systems) [15]. A 142-bp sequence of HIV-1 gag gene was amplified with an internal quantification standard to detect inhibition in
the RT and amplification steps. Samples with calculated HIV-1
RNA copy numbers ú100 were considered positive when the
optical density of the final ELISA step was at least three times
greater than the negative (no RNA) control. Negative and positive
samples (provided by the manufacturer) were included in each
assay. In 10 consecutive assays, no nucleic acid was detected in
the negative controls, and the positive standard had a mean of 4.71
log10 copies/mL ({SD, 0.15).
Statistical methods. The Wilcoxon matched-pairs signed rank
test was used to assess whether HIV-1 RNA differed within individuals when sampled early versus late in the menstrual cycle and
when visible blood was present or not. Differences in CSS HIV1 RNA detection by age, plasma HIV-1 RNA, CD4 cell count,
CD4 cell percent, and CD4:CD8 ratio were assessed by use of the
/ 9d43$$ap48
03-04-98 19:03:33
Wilcoxon rank sum test, where ‘‘detection’’ was defined as HIV1 RNA titers ú100 copies/mL. Univariate and multiple logistic
regression analyses were also used to identify correlates of increased likelihood of CSS HIV-1 RNA detection. Last, a linear
regression analysis was done to characterize the relationship between the quantity of CSS HIV-1 RNA and the quantity of plasma
HIV-1 RNA for the subjects with detectable levels of both (n
Å 24). After we excluded subjects with õ100 copies/mL and
transformed the CSS and plasma HIV-1 RNA to log10 values, the
data were consistent with a normal distribution according to the
Shapiro-Wilk W test.
The mean of the two CSS HIV-1 RNA measurements was used
for analysis of measurements where the variable under consideration was unlikely to change greatly in 2 weeks (i.e., mean plasma
HIV-1 RNA and CD4 and CD8 cell counts). Associations between
CSS, visible blood, cervical inflammation, and dysplasia used only
the measurement of CSS HIV-1 RNA obtained at the same examination.
Subject characteristics. The 49 women ranged in age from
25 to 51 years (median, 37.5). Twenty-eight (57%) were white,
19 (39%) were African American, 1 (2%) was Hispanic, and
1 was Native American. Twenty-nine women were taking one
or more antiretroviral drugs, including zidovudine (20 subjects), didanosine (6), zalcitabine (3), stavudine (3), and lamivudine (3). CD4 cell counts ranged from õ5 to 1230 cells/
mm3 (median, 250).
Plasma and cervical swab supernatant HIV-1 RNA levels.
In 95 (98%) of 97 plasma samples taken from the 49 HIV-1 –
positive women, HIV-1 RNA was ú100 copies/mL (median,
10,267; range, 280 – 736,832). The mean of the two plasma
HIV-1 RNA measures was ú100 copies/mL for 47 women;
plasma HIV-1 RNA in the woman with one measurement was
7901 copies/mL. Forty-eight women (98%) were considered
to have detectable levels of plasma HIV-1 RNA.
In 47 (48%) of the 97 CSS samples, HIV-1 RNA was ú100
copies/mL (median, 761; range, 147 – 32,559). CSS HIV-1
RNA was ú100 copies/mL in 1 sample from 30 (61%) of 49
women and in both samples from 18 (38%) of the 48 women
with 2 samples. For 23 of 48 women with two measurements,
the mean of the two measures was ú100 copies/mL. CSS HIV1 RNA was 175 copies/mL for the woman with one measure.
Twenty-four women (49%) were considered to have detectable
levels of HIV-1 RNA in cervical secretions.
Correlates of cervical HIV-1 RNA detection. Among the
45 women for whom early and late menstrual cycle specimens
were obtained, there was no significant difference in CSS HIV1 RNA between the 2 samples (P Å .38). Eleven women had
visible blood present on one pelvic exam (8 during menstruation). Specimens collected at these examinations tended to have
higher CSS HIV-1 RNA levels compared with samples collected from the same women at a 2-week interval when visible
blood was not present (P Å .066). There were no significant
UC: J Infect
Concise Communications
differences in CD4 cell counts, plasma HIV-1 RNA, or detection of cervical HIV-1 RNA in samples from 9 women with
cervical atypia or dysplasia and 15 women with inflammation
on Pap smear, compared with 25 women without these findings.
None of the women tested positive for C. trachomatis or N.
Characteristics of the 24 women with and 25 women without
detectable HIV-1 RNA in CSS are compared in table 1. In
nonparametric analysis, women with ú100 copies/mL CSS
were significantly younger (34 vs. 40 years; P Å .005), had
higher plasma HIV-1 RNA levels (31,984 vs. 2889; P Å .0004),
lower numbers of CD4 cells (190 vs. 390; P Å .012), lower
CD4 cell percents (16% vs. 25%; P Å .032), and marginally
lower CD4:CD8 ratios (0.27 vs. 0.41; P Å .093).
Univariate associations using logistic regression analysis are
presented as odds ratios in table 2. Younger age, higher level
of plasma HIV-1 RNA, lower number of CD4 cells, lower
CD4 cell percent, and lower CD4:CD8 ratio were significantly
associated with increased likelihood of CSS HIV-1 RNA detection. However, in multiple logistic regression analysis, only
plasma HIV-1 RNA was significantly associated with detection
of cervical HIV-1 RNA, whether included with other covariates
individually or collectively (adjusted odds ratio in full model,
4.79/log10 increase in plasma HIV-1 RNA; 95% confidence
interval, 1.43 – 15.96). No significant differences in CSS HIV1 RNA were found between women currently taking one or
more antiretroviral medications and those taking no antiretrovirals in univariate (P Å .298) or multivariate analysis. Women
on antiretroviral medications did have significantly lower CD4
cell counts (median, 200 vs. 350/mm3; P Å .034).
Linear regression analysis was used to characterize the relationship between the quantity of plasma and cervical HIV-1
RNA in 24 women with detectable levels of CSS HIV-1 RNA
(figure 1). For these subjects, plasma and cervical RNA values
were consistent with a normal distribution after values were
transformed to log10 units. The significance of the regression
coefficient equation (b Å 0.43; P Å .006) suggests a linear
relationship between cervical RNA shedding and plasma levels
of HIV-1 RNA.
JID 1998;177 (April)
Table 2. Correlates of cervical swab supernatant (CSS) detection
of HIV RNA among 49 HIV-positive women by logistic regression
Age (per year)
Plasma HIV-1 RNA (per log10 copies/mL)
CD4 cells (per 100/mm3 decrease)
% CD4 cells (per %)
CD4:CD8 ratio (per unit)
Currently receiving §1 antiretroviral
Age (per year)
Plasma HIV-1 RNA (per log10 copies/mL)
CD4 cells (per 100/mm3 decrease)
% CD4 cells (per %)
CD4:CD8 ratio (per unit)
Currently receiving §1 antiretroviral
0.78 – 0.97
1.86 – 13.80
1.04 – 250
0.90 – 0.99
0.02 – 0.95
0.58 – 5.86
0.77 – 1.02
1.43 – 15.96
0.005 – 33.3
0.78 – 1.23
õ0.01 – 1609
0.55 – 15.33
A sensitive RT-PCR test was used to detect and quantify
HIV-1 RNA in plasma and in supernatant from cervical swabs.
The collection of cervical secretions with a swab eluted into a
ú10-fold volume of collection medium does not permit a precise quantification of HIV-1 RNA and likely underestimates
viral shedding from the cervix. Nonetheless, 24 of 49 women
had detectable CSS HIV-1 RNA, and 15 women had mean
CSS HIV-1 RNA ú1000 copies/mL.
Detection of cervical HIV-1 shedding was associated with
younger age, increased plasma HIV-1 RNA, decreased number
of CD4 cells, and decreased CD4 cell percent. These findings
suggest that cervical viral shedding is more frequent in women
with a higher plasma virus load and more advanced immunodeficiency. Repeated sampling of women early and late in their
menstrual cycles showed no difference in plasma or cervical
Table 1. Characteristics (median values) of women with (n Å 24)
and without (n Å 25) detectable HIV RNA in cervical swab supernatant (CSS) by Wilcoxon rank sum test.
HIV plasma RNA
No. of CD4 cells
% CD4 cells
CD4:CD8 ratio
Women with
§100 copies of
Women with
õ100 copies of
34 years
31,984 copies/mL
40 years
2880 copies/mL
/ 9d43$$ap48
03-04-98 19:03:33
Figure 1. Relationship between cervical swab supernatant and
plasma HIV RNA in 24 HIV-positive women.
UC: J Infect
JID 1998;177 (April)
Concise Communications
HIV-1 RNA levels. As expected, there was an increase in CSS
HIV-1 RNA when blood was visible on the cervical swab.
Cervical secretions were collected from the first Dacron swab
rotated in the cervical os, followed by swabs for Pap smear,
N. gonorrhea, and Chlamydia testing. The order of collection
of swabs may reduce the sensitivity of subsequent sampling.
Recent reports of a significant effect of zidovudine treatment
on vertical transmission has focused interest on the impact of
antiretroviral therapy on HIV-1 RNA in plasma and the genital
tract [16]. In this cross-sectional study, antiretroviral therapy
was not associated with differences in plasma or cervical HIV1 RNA levels, although women taking antiretroviral drugs had
lower CD4 cell counts. These findings may be explained by
the diversity of agents used and the duration of exposure to
individual agents. The effects of antiretroviral administration
were not measured at initiation of treatment, when significant
responses in plasma HIV-1 RNA might be expected. It is likely
that women using antiretroviral medication would have had
more plasma HIV-1 RNA and may have had increased genital
tract shedding had they not been treated. Measurement of HIV1 RNA in plasma and cervical secretions in longitudinal studies
of women initiating antiretroviral therapy, particularly during
pregnancy, may be useful in defining the effect of antiretroviral
treatment on cervical shedding.
Plasma HIV-1 RNA levels and CD4 cell counts are each
associated with HIV-1 disease progression [10 – 13]. Evidence
presented here suggests that these measures are also predictors
of cervical HIV-1 RNA shedding. Measurement of cervical
HIV-1 RNA shedding may provide information about risks for
sexual and mother-to-child HIV-1 transmission and serve as
a useful tool in the evaluation of systemic and intravaginal
antiretroviral agents to interrupt these modes of transmission.
1. Vogt MW, Craven DE, Crawford DF, et al. Isolation of HTLV-III/LAV
from cervical secretions of women at risk for AIDS. Lancet 1986; 1:
525 – 7.
2. Wofsy CB, Hauer LB, Michaelis BA, et al. Isolation of AIDS-associated
retrovirus from genital secretions of women with antibodies to the virus.
Lancet 1986; 1:527 – 9.
/ 9d43$$ap48
03-04-98 19:03:33
3. Henin Y, Mandelbrot L, Henrion R, Pradinaud R, Coulaud JP, Montagnier
L. Virus excretion in the cervicovaginal secretions of pregnant and
nonpregnant HIV infected women. J Acquir Immune Defic Syndr 1993;
6:72 – 5.
4. Clemetson DBA, Moss GB, Willerford DM, et al. Detection of HIV DNA
in cervical and vaginal secretions: prevalence and correlates among
women in Nairobi, Kenya. JAMA 1993; 269:2860 – 4.
5. Kreiss J, Willerford DM, Hensel M, et al. Association between cervical
inflammation and cervical shedding of human immunodeficiency virus
DNA. J Infect Dis 1994; 170:1597 – 601.
6. Kreiss JK, Coombs R, Plummer F, et al. Isolation of human immunodeficiency virus from genital ulcers in Nairobi prostitutes. J Infect Dis 1989;
160:380 – 4.
7. Nielsen K, Boyer P, Dillon M, et al. Presence of human immunodeficiency
virus (HIV) type 1 and HIV-1 – specific antibodies in cervicovaginal
secretions of infected mothers and in gastric aspirate of their infants. J
Infect Dis 1996; 173:1001 – 4.
8. Poss M, Martin HL, Kreiss JK, et al. Diversity in virus populations from
genital secretions and peripheral blood from women recently infected
with human immunodeficiency virus type 1. J Virol 1995; 69:8118 – 22.
9. Mellors JW, Kingsley LA, Rinaldo CR, et al. Quantification of HIV-1 in
plasma predicts outcome after seroconversion. Ann Intern Med 1995;
122:573 – 9.
10. Galetto-Lacour A, Yerly S, Perneger T, et al. Prognostic value of viremia
in patients with long-standing human immunodeficiency virus infection.
J Infect Dis 1996; 173:1388 – 93.
11. Sei S, Kleiner DE, Kopp JB, et al. Quantitative analysis of viral burden
in tissues from adults and children with symptomatic human immunodeficiency virus type 1 infection assessed by polymerase chain reaction.
J Infect Dis 1994; 170:325 – 33.
12. Winters MA, Tan LB, Katzenstein DA, Merigan TC. Biological variation
and quality control of plasma human immunodeficiency virus type 1
RNA quantitation by reverse transcriptase polymerase chain reaction. J
Clin Microbiol 1993; 31:2960 – 6.
13. Goulston C, Stevens E, Gallo D, Mullins JI, Hansen CV, Katzenstein DA.
Human immunodeficiency virus in plasma and genital secretions during
the menstrual cycle. J Infect Dis 1996; 174; 858 – 61.
14. Kurman R, Solomon D. The Bethesda system for reporting cervical/vaginal
cytologic diagnoses. New York: Springer-Verlag, 1992.
15. Mulder J, McKinney N, Christopherson C, Snisky J, Greenfield L, Kwok
S. Rapid and simple PCR assay for quantitation of human immunodeficiency virus type 1 RNA in plasma: application to acute retroviral
infection. J Clin Microbiol 1994; 32:292 – 300.
16. Weiser B, Nachman S, Tropper P, et al. Quantitation of human immunodeficiency virus type 1 during pregnancy: relationship of viral titer to
mother-to-child transmission and stability of viral load. Proc Natl Acad
Sci USA 1994 91:8037 – 41.
UC: J Infect
Dissemination of Mycobacterium tuberculosis across the San Francisco Bay Area
Williamson Z. Bradford, Jane Koehler, Hiyam El-Hajj,
Philip C. Hopewell, Arthur L. Reingold,
Cristina B. Agasino, M. Donald Cave, Sangeeta Rane,
Zhenhua Yang, Charles M. Crane, and Peter M. Small
Medical Service, San Francisco General Hospital, and University of
California, San Francisco, and Francis J. Curry National Tuberculosis
Center, San Francisco, and Department of Medicine, Stanford Medical
School, Palo Alto, and Contra Costa County Health Services
Department, Martinez, and Division of Public Health Biology and
Epidemiology, School of Public Health, University of California,
Berkeley, California; Epidemiology and Prevention Branch, Division of
Public Health, Georgia Department of Human Resources; Departments
of Anatomy, Medicine, and Pathology, University of Arkansas for
Medical Sciences and JL McClellan Memorial Veterans Hospital, Little
Rock, Arkansas
The propensity of Mycobacterium tuberculosis genotypes to spread across geographic boundaries
was investigated by comparing the IS6110 and polymorphic GC-rich sequence patterns of M.
tuberculosis isolates from San Francisco and the East Bay, two distinct regions separated by San
Francisco Bay. Of 724 isolates from incident tuberculosis patients during 1992 and 1993, only 53
(7.3%) had patterns matching §1 isolates from the other region. In the multivariable analysis of
patient risk factors, an AIDS diagnosis (odds ratio [OR], 1.89; 95% confidence interval [CI], 1.00 –
3.57) and non-Asian race (OR, 3.43; 95% CI, 1.59 – 7.42) were associated with having an isolate
with a matching pattern. Of 375 unique IS6110 patterns among San Francisco isolates, only 9 (2.4%)
matched patterns of East Bay isolates. These population-based data suggest that in the San Francisco
Bay Area, M. tuberculosis does not rapidly spread across geographic boundaries, and tuberculosis
control efforts should focus on transmission within defined areas.
The modern age of transportation has brought unprecedented
movement of large numbers of people across geographic
boundaries, heightening concerns about the global dissemination of infectious pathogens. This increasing mobility of human
populations has resulted in instances of widespread dissemination of a number of pathogens, including serogroup A Neisseria
meningitidis, Vibrio cholerae O1, and Mycobacterium tuberculosis [1 – 5].
The dissemination of M. tuberculosis is of particular concern
given its airborne mode of transmission, the virulence of the
organism, and the significant morbidity and costs associated
with tuberculosis. Furthermore, disparities in the prevalence
of tuberculosis infection and disease, the frequency of drug
resistance, and the effectiveness of tuberculosis control practices between different communities may favor the dissemination of M. tuberculosis from higher to lower prevalence areas
[6, 7]. The propensity of M. tuberculosis to spread between
communities, therefore, has important public health implications. Although spread of M. tuberculosis between widely sepa-
Received 20 May 1997; revised 22 October 1997.
The study protocol was reviewed and approved by the UCSF Committee
on Human Research prior to initiation of the study.
Grant support: NIH (AI-34238) and CDC (U52-CCU 900454).
Reprints or correspondence: Dr. Williamson Z. Bradford, Division of Infectious Diseases, San Francisco General Hospital, 1001 Potrero Ave., Room
5K1, San Francisco, CA 94110 ([email protected]).
The Journal of Infectious Diseases 1998;177:1104–7
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
rated countries and spread of multidrug-resistant organisms
between distant domestic sites have been demonstrated, spread
of M. tuberculosis between geographically proximate regions
has not been systematically examined [3, 5].
Advances in M. tuberculosis genotyping techniques have
made it possible to determine the genetic relatedness of individual isolates and to study the geographic spread of specific
genotypes [3 – 5, 8 – 11]. In this report, we describe a study in
which IS6110 genotyping was used to investigate the occurrence of tuberculosis caused by the same genotype of M. tuberculosis in geographically distinct but proximate areas, San
Francisco and the East Bay. These areas are separated by San
Francisco Bay, but an estimated 371,403 individual trips are
made daily across the bay by either bridge or rail [12, 13].
Materials and Methods
Cases of culture-positive tuberculosis diagnosed during 1992
and 1993 in San Francisco County and the two East Bay counties
of Alameda and Contra Costa were identified prospectively by
active laboratory-based surveillance for cultures positive for M.
tuberculosis from any anatomic site and review of county tuberculosis registries. Isolates were genotyped using standard IS6110based methods; the San Francisco and the East Bay isolates were
genotyped at Stanford University and the University of Arkansas,
respectively [14].
Preliminary matching of the San Francisco and East Bay isolates
was performed by comparing the IS6110 pattern of each East
Bay isolate to the IS6110 patterns of all San Francisco isolates
(BioImage Whole Band Analyzer, version 3.3; Millipore, Ann
UC: J Infect
JID 1998;177 (April)
Concise Communications
Arbor, MI). Because different gel conditions and molecular weight
standards were used in the two laboratories, at least 1 representative isolate of every distinct IS6110 matching pattern from both
San Francisco and the East Bay was rerun on the same gel to
confirm an identical pattern. Isolates with discrepant IS6110 patterns on repeat testing were considered not to be a match.
Polymorphic GC-rich sequence (PGRS) genotyping was then
performed on all available isolates with confirmed matching
IS6110 patterns but having õ6 copies of IS6110. PGRS genotyping was done in both laboratories using standard methods, and
identical matches were determined by visual inspection of gels
[15]. A ‘‘cluster’’ was defined as a group of isolates with identical
IS6110 patterns and for isolates with õ6 IS6110 copies, identical
PGRS patterns as well, with at least 1 isolate coming from San
Francisco and 1 from the East Bay. Isolates not available for
confirmatory IS6110 or PGRS testing were excluded from further
All demographic and clinical data, except for human immunodeficiency virus (HIV) infection status, were obtained from a review
of medical and Department of Public Health records. For San
Francisco cases, AIDS was defined as being listed on the San
Francisco County AIDS Registry as of 1 March 1996, whereas in
the East Bay AIDS was defined as having medical record documentation of HIV infection.
Data were managed and analyzed using FoxPro (Microsoft, Redmond, WA) and Stata (Stata Corporation, College Station, TX)
software. The univariate analysis was performed using a t test
for continuous variables; x2 or Fisher’s exact test was used for
categorical variables. The multivariable analysis was done using
logistic regression models; the modeling strategy was to discern
etiologic risk factors. All variables with a univariate P £ .20
were included in the model, and a backwards stepwise elimination
procedure was used. The models included only a main-effect term
for each potential risk factor; no interaction terms were included.
Table 1. Size and locations of clusters including M. tuberculosis
isolates from both San Francisco and the East Bay.
No. of IS6110
No. in area/
total (%)
39/495 (7.9)
East Bay
14/229 (6.1)
53/724 (7.3)
care facility in San Francisco [10]. This cluster included only
a single isolate from the East Bay. Of 375 distinct IS6110
patterns observed in isolates from San Francisco during 1992
and 1993, only 9 (2.4%) patterns were found to match with
that of §1 isolates from the East Bay.
Univariate analysis of patient risk factors for having a clustered isolate showed that being US-born, non-Asian, African
American, and having an AIDS diagnosis were all associated
at a statistically significant level (table 2). In the final multivariable model, an AIDS diagnosis (OR, 1.89; 95% CI, 1.00 – 3.57)
and non-Asian race (OR, 3.43; 95% CI, 1.59 – 7.42) were the
only variables independently associated with having a clustered
Of the culture-positive tuberculosis patients identified in
1992 and 1993, isolates were available for initial genotyping
for 511 (80.6%) of 634 patients in San Francisco and 244
(54.1%) of 451 patients in the East Bay. In an analysis of the
factors (including age, sex, race, HIV infection status, substance abuse, place of birth, and history of tuberculosis) associated with being a culture-positive tuberculosis patient from the
East Bay and not having an available isolate for genotyping,
being African American was the only factor associated at a
statistically significant level. Of 755 patients with initially
available M. tuberculosis isolates, 31 did not have isolates
available for confirmation of IS6110 matches or for PGRS
typing; therefore, 724 patients were included in the study.
Ten clusters were identified; table 1 shows the number of
patients in each cluster and their locations. Altogether, 39
(7.9%) of 495 San Francisco isolates and 14 (6.1%) of 229
East Bay isolates were found to be in clusters. The largest
cluster, which accounted for 16 of the 53 clustered isolates,
was part of a well-described outbreak in an HIV residential
/ 9d43$$ap48
03-04-98 19:03:33
This study demonstrates that in the San Francisco Bay Area,
spread of M. tuberculosis between proximate geographic regions occurs infrequently and contributes little to overall tuberculosis case rates. This finding is striking given the urban environment, the highly mobile population, and the high prevalence
of HIV infection in the San Francisco Bay Area, which should
facilitate the dissemination of genotypes. Furthermore, other
population-based studies of tuberculosis cases using similar but
not identical methods and matching definitions in San Francisco County and in the East Bay during 1992 and 1993, revealed 36.6% and 29.8% of isolates were in clusters, respectively (Small P, Cave D, unpublished data). The results of these
studies, while not directly comparable, suggest that in the San
Francisco Bay Area, M. tuberculosis spreads efficiently within
discrete geographic regions but not between these regions.
An AIDS diagnosis and non-Asian race were the only risk
factors identified in the multivariable analysis for tuberculosis
patients having clustered isolates. Finding that an AIDS diagnosis was a predictor of clustering was not surprising, given the
propensity for HIV-infected patients to develop M. tuberculosis
UC: J Infect
Concise Communications
Table 2. Univariate analysis of patient risk factors for having a
clustered M. tuberculosis isolate.
Risk factor
Age (mean)
African American
AIDS diagnosis
Unstable housing
Parenteral drug use
Nonparenteral drug use
Alcohol abuse
History of tuberculosis
* Determined using a t test for continuous variables and a x 2 or Fisher’s
exact test for categorical variables.
infection with rapid progression to active disease, a factor that
should accelerate the dissemination of M. tuberculosis genotypes [10]. The significance of non-Asian race likely stems
from the relatively higher proportion of Asian patients developing tuberculosis as a result of reactivation of M. tuberculosis
infections acquired at distant times and places.
There are a number of possible explanations for finding little
spread of M. tuberculosis between geographic regions within
the San Francisco Bay Area. One possibility is that infectious
tuberculosis patients are not particularly mobile due either to
illness or to the behavioral characteristics of this patient population, resulting in minimal contact between infectious patients
in one population and susceptible individuals in another population. In support of this hypothesis, intensive interviewing of
incident tuberculosis cases in San Francisco during 1994 and
1995 revealed that only 10 (2.4%) of 413 patients reported
either working or living in the East Bay in the 2 years prior
to diagnosis.
Alternatively, though contact may occur between infectious
and susceptible individuals in these two geographically separated populations, the nature of the contact may not be conducive to the spread of tuberculosis. Casual contact of a short
duration, low intensity, or in a ventilated space might not result
in efficient transmission of M. tuberculosis. While this study
did not directly address these issues of patient mobility and
the low likelihood of transmission with casual contact, it is
possible both these factors contributed to the observed low
frequency of clustering. Another factor that may have limited
our efforts to detect the dissemination of genotypes is the relatively slow pace of the tuberculosis epidemic. Tuberculosis
transmission during the study period without rapid progression
to active disease would not have been identified in our study,
which was conducted over only a 2-year period. The relatively
effective tuberculosis control programs in the San Francisco
/ 9d43$$ap48
03-04-98 19:03:33
JID 1998;177 (April)
Bay Area may also be an important factor in limiting the spread
of M. tuberculosis between the different regions.
The ability to make precise inferences regarding rates of
transmission of M. tuberculosis between these geographically
separated populations is limited by the lack of detailed epidemiologic information. While the low magnitude of clustering described in this investigation supports the inference of infrequent
transmission, not all patients within a given cluster are necessarily related in a common chain of transmission. The reported
magnitude of clustering may therefore overestimate the true
rate of transmission between these populations.
The primary limitation of this study is the number of culturepositive tuberculosis patients excluded from the study because
their isolates were not available for initial or confirmatory genotyping. This sampling problem might have led us to underestimate the magnitude of M. tuberculosis dissemination by excluding isolates from patients relatively more likely to be
involved in recent transmission. However, the small number
of patients with clustered isolates despite the inclusion in the
study of the majority of cases from these population-based
cohorts still argues against significant spread of M. tuberculosis
between these regions. Failure to identify matches in the preliminary analysis, in which IS6110 patterns performed in different laboratories were compared, is also a potential source for
underestimating the magnitude of clustering.
These findings suggest M. tuberculosis does not spread
across geographic regions with the rapidity and devastation of
some other bacterial pathogens, even in the setting of a highly
mobile urban population. This underscores the importance of
transmission dynamics and patient characteristics in addition
to organism virulence and mode of transmission when assessing
the propensity of an infectious pathogen to disseminate. In the
San Francisco Bay Area, tuberculosis control efforts should
focus on the transmission of M. tuberculosis within defined
geographic regions, and not between these regions.
1. Mahon BE, Mintz ED, Greene KD, Wells JG, Tauxe RV. Reported cholera
in the United States, 1992 – 1994: a reflection of global changes in
cholera epidemiology. JAMA 1996; 276:307 – 12.
2. Moore PS, Broome CV. Cerebrospinal meningitis epidemics. Sci Am
1994; 271:38 – 45.
3. Hermans PW, Messadi F, Guebrexabher H, et al. Analysis of the population
structure of Mycobacterium tuberculosis in Ethiopia, Tunisia, and The
Netherlands: usefulness of DNA typing for global tuberculosis epidemiology. J Infect Dis 1995; 171:1504 – 13.
4. Casper C, Singh SP, Rave S, et al. The transcontinental transmission of
tuberculosis: a molecular epidemiological assessment. Am J Public
Health 1996; 86:551 – 3.
5. Bifani PJ, Plikaytis BB, Kapur V, et al. Origin and interstate spread of a
New York City multidrug-resistant Mycobacterium tuberculosis clone
family. JAMA 1996; 275:452 – 7.
6. Brudney K, Dobkin J. Resurgent tuberculosis in New York City. Human
immunodeficiency virus, homelessness, and the decline of tuberculosis
control programs. Am Rev Respir Dis 1991; 144:745 – 9.
UC: J Infect
JID 1998;177 (April)
Concise Communications
7. Bloch AB, Cauthen GM, Onorato IM, et al. Nationwide survey of drugresistant tuberculosis in the United States. JAMA 1994; 271:665 – 71.
8. Small PM, Hopewell PC, Singh SP, et al. The epidemiology of tuberculosis
in San Francisco. A population-based study using conventional and
molecular methods. N Engl J Med 1994; 330:1703 – 9.
9. Alland D, Kalkut GE, Moss AR, et al. Transmission of tuberculosis in
New York City. An analysis by DNA fingerprinting and conventional
epidemiologic methods. N Engl J Med 1994; 330:1710 – 6.
10. Daley CL, Small PM, Schecter GF, et al. An outbreak of tuberculosis
with accelerated progression among persons infected with the human
immunodeficiency virus. An analysis using restriction-fragment-length
polymorphisms. N Engl J Med 1992; 326:231 – 5.
11. Dwyer B, Jackson K, Raios K, Sievers A, Wilshire E, Ross B. DNA
restriction fragment analysis to define an extended cluster of tuberculosis
in homeless men and their associates. J Infect Dis 1993; 167:490 – 4.
12. Bay Area Rapid Transit. BART trips for fiscal year 1996. Oakland, CA:
BART, 1996.
13. Caltrans. 1995 traffic volumes. Oakland, CA: Caltrans, 1995.
14. van Embden JD, Cave MD, Crawford JT, et al. Strain identification of
Mycobacterium tuberculosis by DNA fingerprinting: recommendations
for a standardized methodology. J Clin Microbiol 1993; 31:406 – 9.
15. Chaves F, Yang Z, el Hajj H, et al. Usefulness of the secondary probe
pTBN12 in DNA fingerprinting of Mycobacterium tuberculosis. J Clin
Microbiol 1996; 34:1118 – 23.
Stability of Mycobacterium tuberculosis DNA Genotypes
Robert W. Yeh, Alfredo Ponce de Leon,*
Cristina B. Agasino, Judith A. Hahn, Charles L. Daley,
Philip C. Hopewell, and Peter M. Small
Division of Infectious Diseases and Geographic Medicine, Department
of Medicine, Stanford University School of Medicine, Stanford;
Department of Epidemiology and Division of Pulmonary and Critical
Care Medicine, Department of Medicine, University of California,
San Francisco
To assess genotype stability in Mycobacterium tuberculosis, DNA genotypes were compared in
sequential isolates from 49 patients who had sputum cultures separated by at least 90 days that
grew M. tuberculosis. By use of IS6110 and the polymorphic GC – rich sequence (PGRS) as markers,
it was found that paired isolates from 14 (29%) of 49 patients showed changes in their DNA
genotypes between isolates (12 in IS6110 genotypes and 2 in PGRS genotypes). Changed IS6110
genotypes were confined to strains with 8 – 14 bands and were not related to the bacterial drug
susceptibility, the patients’ human immunodeficiency virus serostatus, or adherence to therapy.
Although this rate of change complicates the interpretation of molecular epidemiologic studies, it
can be exploited to gain additional insight into disease transmission. Furthermore, IS6110-related
mutations may be a major source of genetic plasticity in M. tuberculosis and provide insights into
the organism’s evolution and virulence.
Restriction fragment length polymorphism (RFLP) analysis
based on the insertion sequence IS6110 has been widely used
to study the epidemiology of Mycobacterium tuberculosis [1,
2]. This application is based on the assumption that persons
Received 24 July 1997; revised 29 October 1997.
Presented in part: ‘‘Molecular Epidemiology and Evolutionary Genetics of
Pathogenic Microorganisms,’’ Montpellier, France, 26 – 28 May 1997; ‘‘Molecular Epidemiology and Control of Tuberculosis’’ meeting, Cordoba, Spain,
15 – 16 June 1997.
Informed consent was obtained from patients or their parents or guardians,
and human experimentation guidelines of the US Department of Health and
Human Services and those of Stanford University and the University of California, San Francisco, were followed in the conduct of clinical research.
Grant support: NIH (AI-34238, AI-35969).
Reprints or correspondence: Dr. Peter M. Small, Division of Infectious
Diseases and Geographic Medicine, S156, Stanford University School of Medicine, 300 Pasteur Dr., Stanford, CA 94305 ([email protected]).
* Present affiliation: Laboratorio de MicrobiologıB a ClıB nica, Departamento
de InfectologıB a, Instituto Nacional de la Nutricio´n Salvador Zubira´n, Mexico
City, Mexico.
The Journal of Infectious Diseases 1998;177:1107–11
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
infected with strains of M. tuberculosis that have identical
genotypes ( ‘‘fingerprints’’ ) are epidemiologically linked,
whereas those with different genotypes are unrelated. This assumption relies on two seemingly paradoxical assertions: that
the DNA genotype of a given strain remains constant and that
changes in genotypes over time generate considerable genotypic diversity within a population.
If IS6110-based genotypes change rapidly, it would obscure
epidemiologic links and underestimate transmission. To accommodate this possibility, some studies have considered cases
with similar but nonidentical patterns as epidemiologically related [2]. On the other hand, if genotypes change too slowly,
then IS6110-based RFLP analysis would link cases that are
only distantly related, overestimating transmission. For this
reason, some studies have required that cases match with an
additional genetic marker to be considered epidemiologically
linked [3 – 5]. The most commonly used alternative marker is
the polymorphic GC – rich sequence (PGRS).
While previous studies have quantified genotype instability
in small numbers of uncharacterized M. tuberculosis isolates
[6 – 8], the bacterial and patient characteristics that affect
IS6110 genotypic instability are unknown, and thus the poten-
UC: J Infect
Concise Communications
tial impact of sampling biases unclear, and it is not certain that
these results can be generalized to population-based studies.
Furthermore, there has been no assessment of the relative stability of IS6110 and PGRS-based genotypes. To gain a better
understanding of genotypic instability of M. tuberculosis, we
genotyped sequential samples from 49 patients from whom M.
tuberculosis was repeatedly isolated.
Materials and Methods
Between 1991 and 1995, 1335 cases of active tuberculosis were
reported in San Francisco. The population identified for the study
consisted of patients from whom §2 cultures of M. tuberculosis
were obtained separated by ú90 days. There were 132 eligible
patients. Eighty-three of them had only 1 viable isolate available
for analysis and thus were excluded from the study. Bacterial
isolates from the 49 remaining patients for whom §2 cultures
were available were examined. To assess the possibility of sampling bias, the 49 subjects included in the study were compared
with the 83 excluded patients with respect to the number of bands
that hybridized to IS6110, the patients’ human immunodeficiency
virus (HIV) serostatus, age, sex, and ethnicity.
Each pair of samples for the 49 patients was run on a 20-cm
agarose gel side by side to produce the most accurate comparison
between DNA genotypes. Standard IS6110-based RFLP analysis
was done with PvuII digestion of genomic DNA and hybridization
to bp 631–875, located between the element’s PvuII restriction
site and the 5* terminus (referred to as the right side), and hybridization to bp 30–450, located between the PvuII restriction site
and the 3* terminus (the left side) [9]. PGRS-based genotyping
entailed SmaI digestion and probing for a 32-bp oligonucleotide
[3]. Visual comparisons between the two RFLP patterns from each
patient were done independently by three readers. DNA genotypes
of 2 isolates obtained within 5 days of each other from 51 other
patients were analyzed to determine if genotype changes were
simply a consequence of repeated sampling.
Two-tailed Fisher’s exact tests were done to determine if changing DNA genotypes were related to HIV seropositivity, extrapulmonary tuberculosis disease, drug-resistant isolates, and nonadherence to treatment regimen.
JID 1998;177 (April)
was significantly greater than the frequency of changed PGRS
patterns (P Å .019).
Of the 12 changed IS6110 right-side genotypes, 11 had differences of only 1 band between the first and second patterns.
In some instances, quantitative examination of band intensities
suggests that there was a mixed population, with only some
bacilli having the new genotype. These included 5 strains that
showed an extra band in the second genotype, 4 strains that
lost 1 band in the second genotype, and 2 strains in which 1
band shifted position (figure 1). The genotypes of the isolates
from the 12th patient gained 2 bands and had 2 additional
molecular weight band shifts. The PGRS genotype for this
patient was one of the two PGRS patterns that changed. Of
the patients with changes in IS6110 right-sided genotypes, 7
had concomitant changes and 5 had no changes in the left-side
genotypes. One additional patient only had changes in the leftside genotype.
Statistical analyses showed no association between changed
patterns and increasing time interval between which isolates
were taken in the 49 patients (figure 2). All 12 of the strains
with changed IS6110 right-sided patterns could be characterized as having an intermediate number of bands (8 – 14) (figure
2). Comparison of groups of isolates with low (1 – 7 bands),
intermediate (8 – 14 bands), and high numbers of bands (15 –
21 bands) revealed a statistical association between middle
band number and changed patterns (P Å .026).
Altered drug susceptibility of the organism was not associated with a changed genotype. In addition, no statistical associations were found between changed patterns and HIV positivity, nonadherence to therapy, or drug resistance of the isolate.
After matching the RFLP patterns of the 49 patients to the
existing San Francisco database, 26 of the patients (53%) were
found to be involved in clusters between 1991 and 1995. Of
the 12 patients with changed IS6110 right-side patterns, there
were four instances in which the initial and final pattern each
matched a different cluster in San Francisco. Two other patients
were found to be matched to clusters by only their first RFLP
Of the 51 patients who had samples separated by £5 days,
only 1 pair of isolates had a change in the IS6110 right-side
pattern. No changes were seen in PGRS patterns.
Characteristics of the 49 patients with positive cultures separated by §90 days were not statistically different from those
of the 83 excluded patients among all variables examined.
Patient isolates were separated by a range of 90 days to almost
3 years.
Of the 49 patients, 14 (29%) had patterns that changed between the first and second isolates by at least one technique.
There were 12 changes in IS6110 right side, 8 in IS6110 left
side, and 2 in PGRS genotypes. In 11 cases, changes in IS6110
right-side patterns were not accompanied by changes in PGRS
patterns. The frequency of changed IS6110 right-side patterns
/ 9d43$$ap48
03-04-98 19:03:33
This study provides a comprehensive population-based analysis of the genotype stability of M. tuberculosis. The finding
that 14 (29%) of 49 patients whose cultures spanned at least
90 days had DNA genotypes that changed is both surprising and
important. Previous studies examining sequential genotypes of
M. tuberculosis isolates from persons with repeatedly positive
cultures have not found this degree of instability. Cave et al.
[10] examined sequential IS6110 genotypes of M. tuberculosis
in 18 patients and found that genotypes were identical for 17
of them. Other studies have examined genotypes in a small
UC: J Infect
JID 1998;177 (April)
Concise Communications
Figure 1. Examples of changed M. tuberculosis DNA genotypes. Lanes contain
genomic bacterial DNA from paired isolates separated by §90 days taken from
each of 4 patients (A – D), digested with
PvuII and probed for IS6110 (left) or with
SmaI and probed for polymorphic GC – rich
sequence (PGRS; right). Although all 4 patients showed changed patterns for IS6110,
only those from isolates of patient D
showed concomitant change in PGRS genotype.
number of patients who persistently excreted M. tuberculosis;
none found changes in IS6110 genotypes [6 – 8]. Those studies,
however, examined substantially fewer patients from a non –
population-based sample. The demographic similarity between
the study sample and all serial secretors in San Francisco led
us to believe that the 29% proportion of genotype instability
can be generalized to the population-based molecular epidemiologic studies, such as those conducted in San Francisco.
The primary conclusion that can be drawn from these results
is that DNA genotypes of M. tuberculosis change at a relatively
rapid rate. It is extremely unlikely that our observations were
the result of exogenous reinfection with a different strain, since
the changes were minor and occurred in a patient population
with a relatively low incidence of tuberculosis. The high rate
of change suggests that strains with identical DNA genotypes
are likely to be epidemiologically linked and supports the use
of RFLP analysis for tracking transmission. It has been postulated that IS6110-based RFLP analysis overestimates recent
transmission by grouping distantly related strains with genotypes that have remained identical for large amounts time [4].
However, the degree of instability demonstrated here suggests
/ 9d43$$ap48
03-04-98 19:03:33
that requiring an identical match may in fact underestimate the
frequency of epidemiologic links among patients. It is possible
that such rapid changes in genotypes could result in a convergence of patterns and inappropriately link unrelated cases.
However, this phenomenon has never been observed. In sum,
the data presented here suggest that it may be appropriate to
include 1-band differences in IS6110 genotype defined clusters.
Isolation of bacteria with different IS6110 genotypes could
be a consequence of patients being simultaneously infected
with bacteria with a diversity of similar but nonidentical patterns and that our results are simply an artifact of repeated
sampling. Alternatively, it is possible that cultures obtained
from different sites in the patient’s body could contain different
genotypes of bacilli. Both of these possibilities were not supported by the results of this study, in which 98% (50/51) of
the patients with isolates spanning õ5 days had identical genotypes, more than half of whom had isolates from different body
An additional conclusion of the study is that PGRS has a
significantly lower rate of change than IS6110 (2/49 for PGRS
vs. 12/49 for IS6110; P Å .019). IS6110 appears to provide
UC: J Infect
Concise Communications
JID 1998;177 (April)
Figure 2. Changes in IS6110-based DNA genotypes as function of time interval between isolates
(A) and number of IS6110 hybridizing bands (B).
Solid regions of bars represent patterns that
changed; open regions represent unchanged patterns.
greater resolution in detecting genotypic differences between
strains of M. tuberculosis and, thus, should be considered a
more stringent test for epidemiologic links between cases. It is
important to note that PGRS and IS6110 instability are largely
independent of one another, explaining previous findings that
IS6110 clusters can sometimes be subdivided by PGRS.
The data also support the potential for the generation of
phylogenetic trees describing degrees of relatedness among
strains of M. tuberculosis on the basis of similarities in RFLP
patterns. If 1-band differences can be thought of as related, it
follows that 2-band differences can be considered related, albeit
slightly less so, and so on. Phylogenetic trees may provide
valuable information about the evolution of M. tuberculosis
within a population and the migration of strains between distinct geographic regions [11].
From a microbiologic standpoint, it is interesting that all of the
changed patterns occurred in strains with an intermediate number
of bands (8–14). Lack of transposition in strains with only a few
copies of IS6110 may be a consequence of there simply being
fewer opportunities for transposition. Alternatively, it may be that
/ 9d43$$ap48
03-04-98 19:03:33
these strains have copies in relatively stable positions in the genome and thus have not spawned progeny with greater numbers
of copies. If specific isolates have intrinsically high (or low) rates
of mutations, it may be difficult to deduce general rules governing
genotype stability. It is also plausible that increasing numbers of
IS6110 impose a burden on the strain and that transposition in
strains that already have a high number of copies frequently results
in nonviable organisms. An understanding of the dynamics of
IS6110 transposition in M. tuberculosis may provide insight into
the mechanism by which this organism generates genomic variability.
For DNA genotypes to change, a mutation or transposition
must occur in the genome of at least one mycobacterial cell,
and that cell must progressively outcompete the remaining unchanged cells within the host. It is interesting to postulate that
the 12 new IS6110 patterns reflected the emergence of more
fit strains within the patients. Tracking these strains as they are
transmitted between patients in San Francisco over coming
years to see if they differ in fitness may provide valuable
information concerning the virulence of M. tuberculosis.
UC: J Infect
JID 1998;177 (April)
Concise Communications
We are indebted to the personnel of the San Francisco Department of Public Health Division of Tuberculosis Control for continued support in molecular epidemiologic studies; to H. Salamon,
M. Javanillo, M. Farid-Moamar for assistance in the implementation of the study design and analysis of the data; and to M. Feldman
and his group for consultation on issues pertaining to population
1. Small PM, Hopewell PC, Singh SP, et al. The epidemiology of tuberculosis
in San Francisco. A population-based study using conventional and
molecular methods. N Engl J Med 1994; 330:1703 – 9.
2. Yang ZH, de Haas PE, van Soolingen D, van Embden JD, Andersen AB.
Restriction fragment length polymorphism analysis of Mycobacterium
tuberculosis strains isolated from Greenland during 1992: evidence of
tuberculosis transmission between Greenland and Denmark. J Clin Microbiol 1994; 32:3018 – 25.
3. Chaves F, Yang Z, el Hajj H, et al. Usefulness of the secondary probe
pTBN12 in DNA fingerprinting of Mycobacterium tuberculosis. J Clin
Microbiol 1996; 34:1118 – 23.
/ 9d43$$ap48
03-04-98 19:03:33
4. Warren R, Richardson M, Sampson S, et al. Genotyping of Mycobacterium
tuberculosis with additional markers enhances accuracy in epidemiologic studies. J Clin Microbiol 1996; 34:2219 – 24.
5. Ross BC, Raios K, Jackson K, Bwyer B. Molecular cloning of highly
repeated DNA element from Mycobacterium tuberculosis and its use
as an epidemiological tool. J Clin Microbiol 1992; 30:942 – 6.
6. Lemaitre N, Sougakoff W, Truffot C, Grosset J, Jarlier V. Analyse du
polymorphisme de longueur des fragments de restriction (RFLP) de
souches de Mycobacterium tuberculosis isolees de malades ayant fait
plusieurs episodes de tuberculose. Pathol Biol (Paris) 1996; 44:452 – 5.
7. Stra¨ssle A, Putnik J, Weber R, Fehr-Merhof A, Wu¨st J, Pfyffer GE. Molecular epidemiology of Mycobacterium tuberculosis strains isolated from
patients in a human immunodeficiency virus cohort in Switzerland. J
Clin Microbiol 1997; 35:374 – 8.
8. van Soolingen D, de Haas PE, Hermans PW, Groenen PM, van Embden
JD. Comparison of various repetitive DNA elements as genetic markers
for strain differentiation and epidemiology of Mycobacterium tuberculosis. J Clin Microbiol 1993; 31:1987 – 95.
9. van Embden JD, Cave MD, Crawford JT, et al. Strain identification of
Mycobacterium tuberculosis by DNA fingerprinting: recommendations
for a standardized methodology. J Clin Microbiol 1993; 31:406 – 9.
10. Cave MD, Eisenach KD, Templeton G, et al. Stability of DNA genotype
pattern produced with IS6110 in strains of Mycobacterium tuberculosis.
J Clin Microbiol 1994; 32:262 – 6.
11. Hermans PWM, Messad F, Guesbrexabher H, et al. Analysis of the population structure of Mycobacterium tuberculosis in Ethiopia, Tunisia, and
the Netherlands: usefulness of DNA typing for global tuberculosis epidemiology. J Infect Dis 1995; 171:1504 – 13.
UC: J Infect
Antibody Avidity as a Surrogate Marker of Successful Priming by Haemophilus
influenzae Type b Conjugate Vaccines following Infant Immunization
D. Goldblatt, A. R. J. P. M. Pinto Vaz, and E. Miller
Immunobiology Unit, Institute of Child Health, and Immunisation
Division, Public Health Laboratory Service, Communicable Disease
Surveillance Centre, London, United Kingdom
Evaluation of the new generation of conjugate vaccines is hampered by the absence of reliable
surrogate markers of immunologic memory. Memory responses are characterized by rapid production of relatively high-avidity antibody; thus, a solid-phase ELISA was adapted for the measurement
of anti – Haemophilus influenzae type b (Hib) IgG avidity. In a cohort of infants vaccinated at 2, 3,
and 4 months of age with Hib conjugate vaccines, avidity increased in the period following vaccination, while antibody titer fell. After a booster dose at 1 year of age, both antibody titer and avidity
increased. In a cohort with anti-Hib IgG õ1.0 mg/mL following primary immunization, antibody
avidity after booster was low, indicating an absence of priming. Antibody avidity may help distinguish, in persons with low antibody titers, between those who are primed for memory and those
who are not.
New-generation bacterial conjugate vaccines are currently
in accelerated development following the dramatic success of
the Haemophilus influenzae type b (Hib) conjugate vaccines in
reducing the incidence of invasive Hib disease [1]. The success
of such vaccines is due to the chemical conjugation of the
capsular polysaccharide to a carrier protein, thereby rendering
the carbohydrate component immunogenic, even in very young
children. In addition to their immunogenicity, the vaccines also
appear to prime for memory responses, which is crucial if the
vaccine is to provide long-lasting immunity. Evaluation of Hib
conjugate vaccine immunogenicity relies on the measurement
of serum antibody titers to Hib capsular polysaccharide (polyribosylribitol phosphate [PRP]) and the interpretation of the titer
in the light of the accepted levels for short-term (ú0.15 mg/
mL) or long-term (ú1.0 mg/mL) protection [2]. These levels,
however, were derived by passive immunization or immunization with pure polysaccharides; thus, their relevance for protection in children who have been primed for memory responses
remains unclear. Ideally, therefore, the evaluation of vaccine
immunogenicity should incorporate a surrogate marker of priming.
Received 27 June 1997; revised 21 October 1997.
Presented in part: 37th Interscience Conference on Antimicrobial Agents
and Chemotherapy, Toronto, September 1997 (abstract G-103).
Written informed consent was obtained from the parents or guardians of all
study participants, and the study protocol was approved by the ethics committees of the Gloucestershire and East and North Hertfordshire Health Authorities.
Financial support: Medical Research Council (grant G9431974). D.G. is
funded by the Wellcome Trust.
Reprints or correspondence: Dr. D. Goldblatt, Immunobiology Unit, Institute
of Child Health, 30 Guilford St., London WC1N 1EH, UK ([email protected]
The Journal of Infectious Diseases 1998;177:1112–5
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
Antibody avidity, the strength with which an antibody binds
to an antigen, increases over time following encounter with an
antigen. Memory responses are characterized by the production
of high-avidity antibody; thus, avidity could be considered a
surrogate of successful priming. Recent advances in the measurement of antibody avidity permit the application of such
measurements to large numbers of sera, such as those obtained
during a vaccine trial [3].
We have developed an assay for the measurement of antiPRP antibody avidity and have evaluated sera obtained from
infants who have undergone primary immunization with Hib
conjugate vaccines and received a booster dose at 1 year of
age. We have measured both titer and avidity of Hib-specific
IgG after primary immunization, before booster, and after
booster immunization to assess whether antibody avidity may
provide a surrogate marker of successful priming following
primary immunization.
This study is a follow-up of a previously reported study in which
we assessed the interchangeability of two different Hib conjugate
vaccines given according to the United Kingdom accelerated immunization schedule [4]. We enrolled infants aged 8–12 weeks
of age and randomly allocated them to receive one of six different
sequences of the two Hib conjugate vaccines, which were administered at the same time as the routine immunization with diphtheriatetanus toxoids–pertussis (DTP) and oral poliovirus vaccine at 2,
3, and 4 months of age. A booster dose of either of the two
conjugate vaccines was given at 1 year of age. This was an open
randomized study. The two Hib conjugate vaccines studied were
PRP-T (ActHIB; Pasteur-Me´rieux-MSD, Lyon, France) and HbOC
(HibTITER; Cyanamid-Lederle-Praxis Biologicals, Pearl River,
NY). All infants also received the Wellcome DTP whole cell vaccine in 0.6 mg of Al(OH)3 (Biotech, Beckenham, UK) and oral
poliovirus vaccine. DTP was mixed with PRP-T or HbOC in the
UC: J Infect
JID 1998;177 (April)
Concise Communications
same syringe [5] and administered intramuscularly in the anterolateral aspect of the buttock, thigh, or arm.
Serum was obtained by venipuncture or heel prick from each
infant before the first dose, 4 weeks after the third dose, and before
and 4 weeks after the booster dose and was stored at 0707C
until analyzed. IgG antibody to PRP was assayed by ELISA as
previously described [6] by investigators blind to the randomization groups. Antibody avidity was measured by an elution ELISA,
previously described by us [3, 7] and modified for the Hib assay.
Briefly, sera were diluted in PBS–0.05% Tween 20–1% bovine
serum albumin (PBS-T–BSA) to a final concentration of Ç0.5
mg/mL (minimum dilution, 1:4) and were then allowed to bind for
2 h at room temperature to an antigen-coated plate. After washing,
ammonium thiocyanate diluted in PBS-T–BSA was added in duplicate at concentrations of 0–1 M. After 15 min at room temperature, the wells were washed, and antibody binding was then detected by the addition of a mouse anti-human IgG (R10;
Recognition Sciences, Birmingham, UK) followed by a sheep antimouse–horseradish peroxidase conjugate. The assay was then developed by use of the chromogenic substrate o-phenylenediamine
in citric acid–phosphate buffer with hydrogen peroxide. The reaction was stopped with 4 N H2SO4 and the absorbance at 490 nm
was read by use of an automatic microtiter plate reader (Dynatech,
Chantilly, VA). Results were then expressed as the log percentage
of reduction in absorbance in the presence of ammonium thiocyanate plotted against the concentration of ammonium thiocyanate.
Antibody avidity was displayed as an avidity index corresponding
to the molar concentration of ammonium thiocyanate required to
produce a 50% reduction in absorbance [3].
Statistical analyses were done by use of the statistical software
Minitab (Minitab, State College, PA). Anti-PRP titers and avidity
were positively skewed and thus log-transformed before analysis.
Geometric mean titers were compared between the groups by twosample Student’s t tests.
titers were high enough for antibody avidity to be analyzed
before boosting at 1 year of age, the GMAI had increased from
0.28 (95% CI, 0.24 – 0.31) after primary immunization to 0.55
(95% CI, 0.46 – 0.64) before boosting. After boosting, the
GMAI increased further to 0.60 (95% CI, 0.5 – 0.73) (figure 1).
Postbooster avidity correlated with the prebooster avidity for
subjects in this group (r Å .71).
When the postbooster avidity indices of individual vaccinees
in whom data on antibody titer were available at all time points
were stratified according to their antibody titer after primary
immunization, differences were noted. Subjects with an antibody titer õ1.0 mg/mL at 5 months of age had significantly
lower GMAIs after boosting than did vaccinees who mounted
a long-term protective response after primary immunization
(table 1). In addition, the antibody response of this group after
boosting was significantly lower than that of the subjects who
mounted an IgG response ú1.0 mg/mL after primary immunization. Interestingly, the GMAI for this group after boosting
(0.28 [95% CI, 0.23 – 0.35]) was the same as that seen for all
vaccinees after the primary immunization series. For all of
the vaccinees, however, the correlation between either post –
primary immunization avidity or titer and postbooster avidity
or titer was relatively weak (0.3 and 0.4, respectively). Thus,
the effect seen when the subjects are stratified according to
their initial response is a threshold effect.
Responses to the two conjugate formulations used for boosting were compared. The GMT of anti-PRP IgG for the group
receiving PRP-T as a booster was 139.86 (95% CI, 109.2 –
179.1) compared with a GMT of 85.70 (95% CI, 68.9 – 106.5;
P Å .004) for those boosted with HbOC. Despite the difference
As previously reported, 512 infants were analyzed before
and after primary immunization. The geometric mean titer
(GMT) of anti-PRP IgG for the entire cohort 1 month following
primary immunization was 6.23 mg/mL (95% confidence interval [CI], 6.15 – 6.30), with 92.4% achieving an antibody titer
ú1.0 mg/mL [4]. Paired sera from 376 infants were available
for analysis before and after boosting. At 1 year of age, the
GMT of anti-PRP IgG had fallen to 0.40 (95% CI, 0.34 – 0.47)
for all vaccinees; 38.8% had antibody titers below the shortterm protective titer of 0.15 mg/mL and only 31.9% had antibody titers ú1.0 mg/mL. After booster immunization at 1 year
of age, antibody titer increased significantly for all vaccinees
(GMT, 113.92; 95% CI, 97.4 – 133.2).
The geometric mean avidity index (GMAI) for all vaccinees
at 5 months of age following three doses of Hib conjugate
vaccine was 0.28 (95% CI, 0.27 – 0.3). After boosting, avidity
increased significantly compared with post – primary immunization avidity to a GMAI of 0.52 (95% CI, 0.49 – 0.56; P õ
.001). In a small subgroup of infants (n Å 40) in whom antibody
/ 9d43$$ap48
03-04-98 19:03:33
Figure 1. Geometric mean titer (GMT) of anti-PRP (Hib capsular
polysaccharide) IgG before immunization and GMT and geometric
mean avidity index after primary immunization and before and after
boosting with Hib conjugate vaccine in subgroup of the main cohort
(n Å 40). Bars indicate 95% confidence intervals.
UC: J Infect
Concise Communications
JID 1998;177 (April)
Table 1. Geometric mean titer (GMT) of anti-PRP IgG before and after booster dose at 1 year of
age, and GM avidity index after booster in children stratified according to adequacy of their response
to primary immunization series.
Post – primary immunization
Prebooster GMT
Postbooster GMT
Postbooster GM
avidity index
õ1 mg/mL (n Å 26)
ú1 mg/mL (n Å 290)
0.10 (0.07 – 0.13)
0.43 (0.36 – 0.51)
62.6 (39.72 – 98.8)
144.8 (84.71 – 201.5)
0.28 (0.23 – 0.35)
0.55 (0.51 – 0.59)
NOTE. Data are mg/mL for GMTs. Data in parentheses are 95% confidence interval. PRP, Hib capsular polysaccharide. Suboptimal responses were defined as anti-PRP IgG titers õ1.0 mg/mL 1 month after third dose of Hib
conjugate vaccine. Differences in antibody levels were compared by 2-sample Student’s t test.
in antibody titer for recipients of the two conjugates, no difference was noted in the GMAI after boosting. Recipients of a
PRP-T booster had a GMAI of 0.52 (95% CI, 0.46 – 0.57),
compared with a GMAI of 0.53 (95% CI, 0.49 – 0.58) for the
recipients of an HbOC booster (P Å .60).
Surrogate markers of priming are relevant to an understanding of the immunologic basis of protection afforded by the newgeneration bacterial conjugate vaccines. Theoretically, persons
with low or undetectable specific IgG titers will be protected
from infection if they are primed for memory responses, as
they are able to rapidly produce specific IgG of relatively high
avidity on encounter with the relevant antigen. Thus, in persons
with low titers of antibody, measures of antibody avidity may
be able to discriminate between those successfully primed in
infancy, who may be considered to be protected, and those
with low-avidity antibody, who may technically be at risk of
invasive Hib disease. Evidence of the superiority of high- compared with low-avidity antibody in its ability to kill Haemophilus influenzae type b in vitro has been shown [8].
The strict requirements for the measurement of ‘‘true’’ antibody affinity (monoclonal sources of antibody, highly purified
haptenic antigens) have precluded the widespread use of such
assays in the assessment of vaccine responses. Our approach
has been to adapt solid-phase assays developed for the measurement of antibody titer to measure antibody avidity. Avidity,
defined as the bivalent interaction of antibody with complex
antigens, is, we believe, more closely analogous to the complex
interaction encountered in vivo and thus a more relevant biologic measurement than is affinity. We, and others, have validated the thiocyanate elution assays by comparison with equilibrium dialysis as well as with biospecific interaction, analysis
[9, 10].
In this study, we have shown that antibody avidity is relatively low following primary immunization and significantly
higher following boosting. The majority of the increase in avidity was seen in the months after primary immunization, and
boosting only slightly increased the avidity further. It is possi-
/ 9d43$$ap48
03-04-98 19:03:33
ble that further increases in avidity may take place in the
months after the booster dose.
In a small subgroup of subjects with suboptimal primary
responses, the avidity after boosting was low and the same as
that seen after primary immunization. The suggestion that these
subjects had not been adequately primed and thus were mounting a primary response at 1 year of age was reinforced by the
lower GMT of anti-PRP IgG achieved after boosting, a titer
comparable to that achieved by older children after a single
dose of Hib conjugate [11]. It will be interesting to see whether
these subjects show increase in antibody avidity over time after
this booster dose. With increased sensitivity of the assay, it
will be possible to measure the antibody avidity of low-titer
serum. This will permit studies of naturally occurring anti-PRP
IgG antibodies and an analysis of changes in antibody avidity
with increasing age as well as the study of vaccine failure.
Previous reports of higher antibodies being induced by
booster doses of PRP-T compared with other Hib conjugate
formulations [12, 13] were confirmed in our study. Despite the
higher anti-PRP IgG titers after boosting with PRP-T, the
GMAI was similar for the two vaccines. These data support
those of Agbarakwe et al. [14], who analyzed antibody avidity
by means of a urea elution assay and found no difference in
the avidity of antibodies induced by primary immunization of
infants with PRP-T, HbOC, or PRP conjugated to outer membrane proteins (PRP-OMP). In contrast, in a study using a
radioantigen-binding technique for the measurement of avidity,
antibodies induced by primary immunization with HbOC were
shown to be of higher avidity than those induced by PRP-T
and PRP-OMP [15].
Currently, antibody titer alone is the sole measure of vaccine
immunogenicity, and while such measurements may adequately reflect the immediate response to a vaccine, they cannot
be relied on to reflect immune status months or years after the
immunization with a vaccine that induces memory. Current
concern surrounding the reduced immunogenicity of combination vaccines containing acellular pertussis and Hib conjugate
in the same formulation [16] highlight the urgent need for an
improved understanding of the relationship between primary
antibody responses and long-term protection. Reduced anti-
UC: J Infect
JID 1998;177 (April)
Concise Communications
body titers immediately after immunization in the face of successful priming may not necessarily compromise postimmunization protection from disease. In the light of our demonstration
of an increase in avidity in the months after vaccination, immunogenicity studies incorporating avidity measurements would
optimally require the analysis of serum at time point(s) additional to 4 weeks after vaccination. Furthermore, validation of
avidity as a surrogate for priming will require studies in which
antibody titers induced by a booster dose of pure PRP months
or years after primary immunization are compared with avidity.
A high proportion (38.8%) of subjects in our study had
antibody titers below the minimum protective titer at 1 year of
age, yet, despite this, the effectiveness of the Hib conjugate
vaccine in the United Kingdom, where no booster dose of Hib
conjugate vaccines is given, remains high [17]. This reinforces
our view that antibody titers below the minimum protective
level in primed persons do not necessarily reflect a susceptibility to infection, and the search for adequate markers of such
protection continue. Antibody avidity may play a part in this
We thank Ronwen Morris for help with coordinating the study,
Joan Vurdien and Marie Rush for administrative help, Pauline
Waight for data handling, Paddy Farrington for statistical advice,
Natalie McCloskey for technical help, and Kit Fairley and Keith
Cartwright and the study nurses of Gloucester and East and North
1. Adams WG, Deaver KA, Cochi SL, et al. Decline of childhood Haemophilus influenzae type b (Hib) disease in the Hib vaccine era. JAMA 1993;
269:221 – 6.
2. Ka¨yhty H, Peltola H, Karanko V, Ma¨kela¨ P. The protective level of serum
antibodies to capsular polysaccharide of Haemophilus influenzae type
b. J Infect Dis 1983; 147:1100 – 3.
3. Goldblatt D. Simple solid phase assays of avidity. In: Turner MW, Johnstone AP, eds. Immunochemistry 2: a practical approach. Oxford: IRL
Press at Oxford University Press, 1997:31 – 51.
/ 9d43$$ap48
03-04-98 19:03:33
4. Goldblatt D, Fairley CK, Cartwright K, Miller E. Interchangeability of
conjugated Haemophilus influenzae type b vaccines during the primary
immunisation of infants at 2, 3, and 4 months. Br Med J 1996; 312:
817 – 8.
5. Begg NT, Miller E, Fairley CK, et al. Antibody responses and symptoms
after DTP and either tetanus or diphtheria Haemophilus influenzae type
b conjugate vaccines given for primary immunisation by separate or
mixed syringe. Vaccine 1995; 13:1547 – 50.
6. Phipps DC, West J, Eby R, Koster M, Madore DV, Quataert SA. An
ELISA employing a Haemophilus influenzae type b oligosaccharide –
human serum albumin conjugate correlates with the radioantigen binding assay. J Immunol Methods 1990; 135:121 – 8.
7. Goldblatt D, van Etten L, van Milligen FJ, Aalberse RC, Turner MW.
The role of pH in modified ELISA procedures used for the estimation
of functional antibody affinity. J Immunol Methods 1993; 166:281 – 5.
8. Schlesinger Y, Granoff DM. Avidity and bactericidal activity of antibody
elicited by different Haemophilus influenzae type b conjugate vaccines.
JAMA 1992; 267:1489 – 94.
9. MacDonald RA, Hosking CS, Jones CL. The measurement of relative
antibody affinity by ELISA using thiocyanate elution. J Immunol Methods 1988; 106:191 – 4.
10. McCloskey N, Turner MW, Goldblatt D. Correlation between the avidity
of mouse-human chimeric IgG subclass monoclonal antibodies measured by solid phase elution ELISA and biospecific interaction analysis
(BIA). J Immunol Methods 1997; 205:67 – 72.
11. Goldblatt D, Johnson M, Evans J. Antibody responses to Haemophilus
influenzae type b conjugate vaccine in children with sickle cell disease.
Arch Dis Child 1996; 75:159 – 61.
12. Scheifele D, Law B, Mitchell L, Ochnio J. Study of booster doses of
two Haemophilus influenzae type b conjugate vaccines including their
interchangeability. Vaccine 1996; 14:1399 – 406.
13. Kurikka S, Ka¨yhty H, Saarinen L, Ronnberg PR, Eskola J, Ma¨kela¨ PH.
Immunologic priming by one dose of Haemophilus influenzae type b
conjugate vaccine in infancy. J Infect Dis 1995; 172:1268 – 72.
14. Agbarakwe AE, Griffiths H, Begg N, Chapel HM. Avidity of specific IgG
antibodies elicited by immunisation against Haemophilus influenzae
type b. J Clin Pathol 1995; 48:206 – 9.
15. Schlesinger Y, Granoff DM, Vaccine Study Group. Avidity and bactericidal activity of antibody elicited by different Haemophilus influenzae
type b conjugate vaccines. JAMA 1992; 267:1489 – 94.
16. Eskola J, Olander R, Hovi T, Litmanen L, Peltola S, Ka¨yhty H. Randomised trial of the effect of co-administration with acellular pertussis DTP
vaccine on immunogenicity of Haemophilus influenzae type b conjugate
vaccine. Lancet 1996; 348:1688 – 92.
17. Booy R, Heath PT, Slack MPE, Begg N, Moxon ER. Vaccine failures after
primary immunisation with Haemophilus influenzae type-b conjugate
vaccine without booster. Lancet 1997; 349:1197 – 202.
UC: J Infect
Identification of a Highly Encapsulated, Genetically Related Group of Invasive
Type III Group B Streptococci
Shinji Takahashi, Elisabeth E. Adderson, Yukiko Nagano,
Noriyuki Nagano, Mark R. Briesacher,
and John F. Bohnsack
Department of Microbiology, Joshi-Eiyoh University, Sakado,
Department of Microbiology, School of Allied Health Sciences, Kitasata
University, Sagamihara, and Medical Microbiology Laboratory,
Funabashi Medical Center, Funabashi, Japan; Department of
Pediatrics, University of Utah Health Sciences Center, Salt Lake City
Type III group B streptococci (GBS) isolated from Tokyo and Salt Lake City were classified
according to the similarity of HindIII and Sse83871 restriction digest patterns (RDPs) of bacterial
DNA. The bacteria were clustered into three RDP types, with excellent correlation between subtyping
based on the two enzymes. The majority (91%) of invasive isolates obtained from neonates were
RDP type III-3. The mean sialic acid content of the III-3 strains was higher than that of other type
III strains. Closely related isolates were concordant for expression of the bacterial enzyme C5a-ase,
but invasive strains were no more likely to be C5a-ase positive than were strains isolated from the
genitourinary tract of pregnant women. These data indicate that a group of genetically related
organisms with increased capsule production causes the majority of invasive type III GBS disease.
Group B streptococci (GBS) are an important cause of serious bacterial disease in neonates, pregnant women, and patients
with underlying illnesses [1]. GBS are subclassified into serotypes according to the immunologic reactivity of the polysaccharide capsule. Of the nine serotypes, types I, II, III, and,
more recently, type V and VIII GBS cause the majority of
neonatal human GBS disease [1 – 3]. Serotype III GBS are
particularly important because type III cause a significant percentage of early-onset disease (within the first week of life)
and the majority of late-onset disease (after the first week of
life) in human neonates and also cause the vast majority of
neonatal GBS meningitis [1].
GBS can be further subclassified into restriction digest pattern (RDP) types by electrophoretic analysis of fragments produced by restriction enzyme digestion of bacterial DNA. RDP
typing of clinical isolates in Japan by use of the restriction
enzyme HindIII showed that a single RDP type (RDP type III3) causes most serotype III neonatal sepsis, suggesting the
existence of a genetically related subgroup of serotype III GBS
that are intrinsically more virulent than other serotype III strains
[4]. The purpose of these studies was to determine whether
invasive serotype III isolates from Salt Lake City, as well as
more recently isolated clinical isolates from Japan, are also
RDP III-3. We also validated the HindIII typing by typing the
Received 30 May 1997; revised 24 October 1997.
Presented in part: annual meeting of the Society for Pediatric Research, May
1996, Washington, DC (abstract 1101).
Grant support: Primary Children’s Research Foundation; US Public Health
Service (AI-13150).
Reprints or correspondence: Dr. John F. Bohnsack, Dept. of Pediatrics,
Room 2A152, University of Utah Health Sciences Center, 50 N. Medical Dr.,
Salt Lake City, UT 84132 ([email protected]).
The Journal of Infectious Diseases 1998;177:1116–9
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
GBS with a second restriction endonuclease, Sse83871, and
determined whether other putative virulence factors, including
capsular sialic acid content, R protein, and C5a-ase expression,
correlated with RDP type.
Materials and Methods
Bacteria. GBS were isolated from routine vaginal swabs from
women seen at the University of Utah Hospital during 1993–1995.
GBS isolated from blood or cerebrospinal fluid or from a sterile
body site at autopsy were obtained from the Microbiology Laboratories at the University Medical Center and at Primary Children’s
Medical Center, a regional children’s referral hospital, from 1993
to 1995. Most Japanese isolates were collected from patients cared
for at the Funabashi Medical Center, as well as other hospitals in
the Tokyo area, between 1988 and 1995. GBS were serotyped with
commercially available antisera (Denka Seiken, Tokyo). Colonies
of GBS were isolated from blood-agar plates, grown in ToddHewitt broth, and stored at 0707C until analysis.
RDP typing. DNA was extracted from GBS suspended in agarose
gel plugs (InCert; FMC BioProducts, Rockland, ME) according to
the manufacturer’s instructions except that mutanolysin and SDS containing proteinase K were used for digestion of the bacteria. DNA in
the agarose gel plugs was digested with HindIII, extracted from the
agarose with phenol, and redigested with HindIII. The DNA sample
was then subjected to electrophoresis in a conventional ethidium
bromide–agarose gel. The similarity between densitometric RDPs
from individual strains was expressed as a Pearson product moment
correlation coefficient (PPMCC) and clustered by the unweighted
pair group method average as previously described [4]. RDPs from
a single isolate always resemble each other with a PPMCC §0.99,
while RDPs of GBS isolates of the same serotype always resemble
each other with a PPMCC ú0.93 (data not shown). Therefore, strains
with RDPs that resemble each other with PPMCCs §0.99 are considered identical, and isolates that resemble each other with PPMCCs
ú0.93 are clustered into subtypes.
Sse83871 RDP typing was done similarly to HindIII typing
except that Sse83871-digested DNA fragments were separated by
UC: J Infect
JID 1998;177 (April)
Concise Communications
pulsed-field gel electrophoresis. Similarity coefficients between all
possible pairs of RDPs were calculated as follows: similarity Å
(number of shared bands 1 2)/total number of bands. The RDPs
were then clustered by unweighted pair group method average.
RDPs from strains of the same serotype resemble each other with
a similarity coefficient ú0.45 (data not shown). Thus, isolates
were clustered into subtypes in which the isolates resembled each
other with a similarity coefficient §0.45.
Detection of plasmid DNA. DNA was extracted from agarose
plugs, alkaline-denatured, and subjected to conventional agarose
gel electrophoresis to detect plasmids.
Sialic acid content. Bacteria were grown in Todd-Hewitt
broth, harvested at mid-log phase to maximize capsular and sialic
acid content [5], and washed by centrifugation. The capsule was
extracted by hydrolysis with 0.1 N HCl at 847C for 20 min, and
the sialic acid content of the capsular extract was determined by
the thiobarbituric acid method [6].
C5a-ase activity. Functional C5a-ase activity was determined
as previously described by use of a quantitative neutrophil adhesion
assay [7].
The 62 isolates were divided into three HindIII RDP types,
as in the previous study [4]. The majority of the isolates (41)
were found to be RDP type III-3, while 18 and 3 of the remaining isolates were RDP type III-2 and III-1, respectively
(figure 1). Isolates clustered into five Sse83871 RDP types,
with the exception of isolate 59 (figure 1). Bacteria within these
5 groups of bacteria, designated III-1, III-2a, III-2b, III-3a, and
III-3b, were found in the corresponding HindIII RDP types
(figure 1). Furthermore, bacteria that clustered into Sse83871
RDP type III-3a (strains 1 – 27) and III-3b (strains 28 – 41) also
clustered into these subtypes within the HindIII dendrogram
(figure 1). Strains 42 – 54 in the HindIII-2 dendrogram were
III-2a except strain 45, which was III-2b, while strains 55 – 58
were all III-2b, except strain 58, which was III-2a. Thus, the
two methods of RDP typing clustered the type III isolates into
virtually identical groups, strongly supporting the validity of
these two methods for subtyping type III GBS. Isolates that
resembled each other with similarity coefficients ú0.97 by
HindIII typing, or that had identical RDPs by Sse83871 typing,
were always isolated from the same city (not shown). No plasmids were detected in any of the isolates.
Overall, 35 of the GBS were isolated from a normally sterile
body fluid (blood or cerebrospinal fluid). The overwhelming
majority of these strains (91%; 32 strains) were RDP type III3, whereas only 9, or 33%, of the vaginal isolates were III-3.
In contrast, 59% (16) of the vaginal isolates were RDP type
III-2 strains, while only 6% (2) of the invasive strains were
III-2. RDP type III-3 strains were significantly more likely to
be invasive isolates than to be vaginal isolates, while type III2 strains were significantly more likely to be vaginal isolates
than to be invasive (P õ .01, x2, Yates’s modification).
For 27 isolates, the patient’s age at the onset of disease was
available. Two of the 27 isolates were from older children (2
/ 9d43$$ap48
03-04-98 19:03:33
and 18 years of age). The remaining 25 were from neonates
with either early- (n Å 12) or late-onset (n Å 12) disease. Both
of the III-2 isolates were from infants with late-onset disease.
The III-3 strains were still significantly more likely to be invasive than to be vaginal isolates (P õ .01, x2, Yates’s modification), even when the isolates from older children and from
those patients for which clinical information was lacking were
The mean sialic acid content of the type III-3 strains was
significantly greater than the sialic acid content of either the
III-2 or III-1 strains (P õ .05, Student’s t test; table 1), although
2 of the III-2 strains had sialic acid contents greater than the
mean sialic acid content of the III-3 strains. The sialic acid
content of the invasive III-3 isolates was not significantly
greater than the sialic acid content of colonizing III-3 strains
(6.19 { 1.14 vs. 5.83 { 1.65 mg of sialic acid/mg of cell dry
weight), nor was the sialic acid content of the invasive III-2
or III-1 isolates elevated compared with that of the isolates of
the same RDP type.
While 100% of the III-2 and III-1 isolates were C5a-ase –
positive, only 63% of the III-3 strains were C5a-ase – positive
(table 1). All of the III-3a strains were C5a-ase – positive except
1 (isolate 9), and all of the III-3b strains were C5a-ase – negative. Overall, 76% of the type III GBS expressed functional
C5a-ase activity, but there was no correlation between C5aase activity and whether the organism was invasive or colonizing: 69% (24/35) of the invasive strains expressed C5a-ase,
while 85% (23/27) of the vaginal isolates expressed C5a-ase.
Consistent with a previous report [8], all of the strains expressed R protein, except for the III-1 strains and 1 of the III3 isolates.
In the studies reported here, we demonstrate that type III
GBS from both Salt Lake City and Tokyo can be clustered
into three subtypes on the basis of HindIII RDPs. The validity
of the HindIII subtyping is strongly supported by the Sse83871
typing and is further corroborated by correlation with several
different phenotypic determinations. First, the III-3 isolates had
a significantly higher sialic acid content than did the III-2 or
III-1 isolates. Second, the III-3a, III-2, and III-1 isolates all
expressed C5a-ase activity (with 1 exception), while the III3b isolates did not. Third, the 3 III-1 strains in this study were
uniformly R protein negative, as were the 4 III-1 strains studied
in our previous study [4]. In contrast, the overwhelming majority of III-2 and III-3 strains were found to express R protein
both in this study and in our previous report [4]. This suggests
that lack of R protein expression is a unique phenotypic characteristic of III-1 strains, although the small number of strains
identified as III-1 limits the power of this observation. Additional support for the validity of the RDP typing is derived from
the fact that isolates identified as closely related or identical by
RDP typing always originated from the same city.
UC: J Infect
Concise Communications
JID 1998;177 (April)
Figure 1. Dendrogram of type III GBS isolates based on HindIII and Sse83871 restriction digest patterns. Nos. in center of dendrogram are
isolate numbers. GBS from different serotypes always resemble each other with Pearson product moment correlation coefficients lower than
that indicated by dashed lines.
Studies from several laboratories have suggested that most
invasive type III GBS disease is caused by related strains of
bacteria [9, 4, 10], although a distinct group of type III GBS
with increased virulence could not be identified in one study
[11]. Our data indicating that the vast majority of invasive
isolates in this sample were RDP type III-3 suggest that RDP
type III-3 strains are the most common type III GBS to cause
invasive disease. These studies therefore support the existence
of a subtype of type III GBS with increased pathogenic potential.
/ 9d43$$ap48
03-04-98 19:03:33
We previously proposed that the bacterial enzyme C5a-ase
contributes to the pathogenic potential of GBS by its ability to
rapidly inactivate the potent complement-derived neutrophil
agonist C5a [12, 13]. Data presented here, however, suggest
that expression of C5a-ase is not the basis for increased virulence of the III-3 strains, since there was no correlation between
C5a-ase activity and invasiveness. These studies do not address
the possibility that expression of the gene encoding C5a-ase is
down-regulated under the in vitro conditions used in these
experiments, leaving open the possibility that C5a-ase contri-
UC: J Infect
JID 1998;177 (April)
Concise Communications
Table 1. Sialic acid content, C5a-ase activity, and R protein expression of type III GBS.
RDP type
No. of
Sialic acid
C5a-ase –
positive (%)
R protein –
positive (%)
3 (100%)
18 (100%)
18 (100%)
26 (63%)
26 (96%)
40 (98%)
26 (96%)
14 (100%)
* Mean sialic acid content { SD, expressed as mg/mg of cell dry weight,
of all isolates in indicated restriction digest pattern type.
Significantly less than sialic acid content of III-3 strains (P õ .05, Student’s
t test).
III-2 isolate 59 is neither III-2a nor III-2b.
butes to the pathogenesis of all type III GBS in vivo. Further
investigations are underway to determine the molecular basis
for the lack of C5a-ase expression in the III-3b strains.
Our observation that the average sialic acid content of the
III-3 strains is higher than that of the III-2 and III-1 strains
indicates that III-3 strains are more encapsulated than are III2 and III-1 strains and suggests that there is a common genetic
basis for the greater encapsulation of the III-3 strains. Resistance to opsonization by complement probably contributes to
the increased virulence of the III-3 strains, because resistance
of type III GBS to opsonization is largely due to the sialic acid
content of the capsule and is the major virulence factor known
in type III GBS [14, 15].
While the sialylated polysaccharide capsule is critical for
type III GBS to evade host defenses, it seems likely that additional bacterial factors contribute to the virulence of type III
GBS in human neonates. Our ability to identify more virulent,
genetically related strains of GBS by RDP typing should facilitate identification of such virulence factors.
We thank Andrew Pavia for helpful comments and assistance
with statistical analysis, Karen Carroll and Judy Daly for supplying
bacterial samples, and Harry Hill for discussion.
/ 9d43$$ap48
03-04-98 19:03:33
1. Baker CJ, Edwards MS. Group B streptococcal infections. In: Remington
JS, Klein JO, eds. Infectious disease of the fetus and newborn infant.
4th ed. Philadelphia: WB Saunders, 1995; 980 – 1054.
2. Blumberg HM, Stephens DS, Modansky M, et al. Invasive group B streptococcal disease: the emergence of serotype V. J Infect Dis 1996; 173:
365 – 73.
3. Kogan G, Uhrin D, Brisson JR, et al. Structural and immunochemical
characterization of the type VIII group B Streptococcus capsular polysaccharide. J Biol Chem 1996; 271:8786 – 90.
4. Nagano Y, Nagano N, Takahashi S, et al. Restriction endonuclease digest
patterns of chromosomal DNA from group B b-haemolytic streptococci.
J Med Microbiol 1991; 35:297 – 303.
5. Paoletti LC, Ross RA, Johnson KD. Cell growth rate regulates expression
of group B Streptococcus type III capsular polysaccharide. Infect Immun
1996; 64:1220 – 6.
6. Aminoff D. Methods for the quantitative estimation of N-acetylneuraminic
acid and their application to hydrolysates of sialomucoids. Biochem J
1961; 81:384 – 92.
7. Bohnsack JF, Mollison KW, Buko AM, Ashworth JW, Hill HR. Group B
streptococci inactivate complement component C5a by enzymic cleavage at the C-terminus. Biochem J 1991; 273:635 – 40.
8. Linden V, Christensen KK, Christensen P. The occurrence of R-protein
among isolates of group B streptococci from human sources. Acta Pathol
Microbiol Scand [B] 1983; 91:153 – 6.
9. Musser JM, Mattingly SJ, Quentin R, Goudeau A, Selander RK. Identification of a high-virulence clone of type III Streptococcus agalactiae
(group B Streptococcus) causing invasive neonatal disease. Proc Natl
Acad Sci USA 1989; 86:4731 – 5.
10. Helmig R, Uldgjerg N, Boris J, Kilian M. Clonal analysis of Streptococcus
agalactiae isolated from infants with neonatal sepsis or meningitis and
their mothers and from healthy pregnant women. J Infect Dis 1993;
168:904 – 9.
11. Hauge M, Jespersgaard C, Poulsen K, Kilian M. Population structure of
Streptococcus agalactiae reveals an association between specific evolutionary lineages and putative virulence factors but not disease. Infect
Immun 1996; 64:919 – 25.
12. Takahashi S, Nagano Y, Nagano N, Hayashi O, Taguchi F, Okuwaki Y.
Role of C5a-ase in group B streptococcal resistance to opsonophagocytic
killing. Infect Immun 1995; 63:4764 – 9.
13. Bohnsack JF, Widjaja K, Ghazizadeh S, et al. A role for C5 in the acute
neutrophil response to group B streptococcal infections. J Infect Dis
1997; 175:847 – 55.
14. Edwards ME, Kasper DL, Jennings HJ, Baker CJ, Nicholson-Weller A.
Capsular sialic acid prevents activation of the alternative pathway by
type III, group B streptococci. J Immunol 1982; 128:1278 – 83.
15. Marques MB, Kasper DL, Pangburn MK, Wessels MR. Prevention of C3
deposition by capsular polysaccharide is a virulence mechanism of type
III group B streptococci. Infect Immun 1992; 60:3986 – 93.
UC: J Infect
Colonization with and Acquisition of Uropathogenic Escherichia coli as Revealed
by Polymerase Chain Reaction–Based Detection
James R. Johnson, Jennifer J. Brown, Ulrike B. Carlino,
and Thomas A. Russo
Department of Medicine, University of Minnesota, and Medical Service,
VA Medical Center, Minneapolis, Minnesota; Department of Medicine,
State University of New York at Buffalo
The prevalence of colonization with uropathogenic Escherichia coli and their reservoirs and routes
of acquisition are incompletely defined. To help clarify these issues, polymerase chain reaction
(PCR) – based strain typing assays were used to evaluate the fecal and vaginal E. coli flora of 11
volunteers. PCR detected the virulence genes papG, aer, and cnf significantly more frequently in
mixed intestinal samples than in the corresponding predominant strains, evidence that traditional
methods are suboptimal for detecting colonization with uropathogens. For strain typing, repetitiveelement PCR was as discriminating as pulsed-field gel electrophoresis and O:H serotyping but more
convenient. Molecular epidemiologic analysis of subjects’ E. coli suggested emergence of occult
uropathogenic strains from within the host’s own intestinal flora, strain sharing between household
members, and de novo acquisition of (unshared) uropathogenic strains. These methods should
facilitate the studies needed to clarify the relative contributions of these three pathways to the
pathogenesis of urinary tract infection.
The Escherichia coli strains that cause urinary tract infection
(UTI) in otherwise healthy hosts often differ from commensal
intestinal E. coli strains by having specialized virulence factors
(VFs) that promote extraintestinal infection [1]. When they
cause UTI, such strains usually emerge from the host’s own
intestinal (and, in women, vaginal) flora [2] and at the time
often are the predominant strain at these sites [3]. However,
little is known regarding the external reservoir(s) for such
strains, the mechanisms whereby they come to colonize UTIprone hosts, or how commonly they are present at an occult
level within the complex human intestinal flora.
Traditional methods for assessing intestinal or vaginal colonization with urovirulent E. coli, which involve testing several
colonies from a fecal culture, usually capture the quantitatively
predominant intestinal strain [4]. However, because the intestinal E. coli population is highly complex [5, 6], this ‘‘predominant strain’’ approach may fail to identify many hosts who in
fact are colonized with a potential uropathogen.
Newly developed polymerase chain reaction (PCR) – based
methods permit rapid assessment of the identity and VF repertoire of E. coli strains [7, 8]. PCR’s tremendous sensitivity
Received 21 August 1997; revised 3 November 1997.
Informed consent was obtained from subjects. Human experimentation
guidelines of the US Department of Health and Human Services and those of
the University of Minnesota Institutional Review Board were followed in the
conduct of clinical research.
Grant support: NIH (DK-47504 to J.R.J); Research for Health in Erie County
(T.A.R.); VA Merit Review (J.R.J.).
Reprints or correspondence: Dr. James R. Johnson, Infectious Diseases Section (111F), Minneapolis VA Medical Center, 1 Veterans Dr., Minneapolis,
MN 55417 ([email protected]).
The Journal of Infectious Diseases 1998;177:1120–4
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
conceivably could help detect occult VF/ strains within a predominantly VF0 mixed population. We undertook this study
to determine whether, with new PCR-based techniques, we
could more sensitively detect urovirulent colonizing strains and
better understand their origins. We also evaluated PCR-based
strain typing in comparison with conventional methods for assessing E. coli colonization patterns.
Subjects and protocol. Study subjects were the 11 members of
three households (table 1). On one or more occasions, all subjects
provided fecal swabs. These were collected (at the collector’s discretion) by inserting the swab into the anus, by wiping the anal
opening, or by touching feces on toilet tissue, provided swabs
showed visible fecal staining. Women also collected swabs from
the vaginal introitus, which were inserted and rubbed circumferentially inside the vaginal opening several times. One subject with
acute UTI (subject AA) provided clean-catch midstream urine
specimens. Serial samplings were at 3-week intervals.
Fecal and vaginal swabs were inoculated on one quadrant of a
MacConkey’s agar plate, then streaked for isolation in the remaining three quadrants with a sterile loop or swab and incubated
overnight at 377C. Lactose- and indole-positive gram-negative bacilli with a consistent colonial morphology were defined as E. coli.
From each plate, 3 isolated colonies of E. coli were saved, as was
a mixed sample from an area of confluent growth in the inoculum
area. Individual colonies and mixed-growth samples were passed
separately once in broth, then frozen at 0707C. Urine cultures
were done by the University of Minnesota Clinical Microbiology
Laboratory. A single colony of any E. coli isolate was saved.
PCR assays. Target DNA was prepared by a whole cell
method [9]. Broth cultures were inoculated from frozen stocks of
individual colonies or mixed samples, incubated overnight, then
washed and boiled [9]. Strain typing was done by repetitive element (rep)–PCR using the Enterobacterial Repeat Intergenic Con-
UC: J Infect
JID 1998;177 (April)
Concise Communications
Table 1. Results of strain typing and virulence factor PCR assays for mixed intestinal bacteria and individual E. coli strains.
Gender (M/
F), role, age
F, SP, 34
Virulence factor PCR
UTI status
Pyelo; hx RUTI
M, SP, 32
(see above)
(see above)
Pyelo; hx RUTI
F, mother, 42
M, father, 43
M, son, 4 1/2
F, daughter, 2
M, son 1 1/2
M, father, 39
F, mother, 39
F, daughter, 3
F, daughter, 3
(see above)
(see above)
(see above)
(see above)
(see above)
Ont:H7, H24
O16, O142:Hnt
Target DNA
Mixed sample
Predom. strain
Predom. strain
Occult strain
Sole strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Sole strain
Predom. strain*
Sole strain
Mixed sample
Predom. strain
Predom. strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Predom. strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Predom. strain
Occult strain
Mixed sample
Predom. strain
Predom. strain
Occult strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Predom. strain
Mixed sample
Predom. strain
Predom. strain
Mixed sample
Predom. strain
Mixed sample
Predom. strain
Predom. strain
Mixed sample
Predom. strain
Predom. strain
Occult strain
O:H serotype
NOTE. Underlined rep-PCR (repetitive element – PCR) genotypes are shared with another member of the household. Predom. Å predominant; M/F Å male/
female; SP Å sex partner; pyelo Å acute pyelonephritis; ABU Å asymptomatic bacteriuria; UTI Å urinary tract infection; none Å no prior symptomatic UTI;
hx UTI Å prior symptomatic UTI; hx RUTI Å frequent prior UTIs; Ont and Hnt Å O or H nontypeable; Omult and Hmult Å reactivity with multiple O or H
* Mixed sample not available for virulence factor PCR assays.
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
Concise Communications
sensus (ERIC) sequence primers ERIC1R and ERIC2 (ERIC PCR)
or the BOX A1R primer (BOX PCR) [7]. Genomic patterns differing by £2 bands were defined as the same genotype and those
differing by ú2 bands as different genotypes. The three papG
alleles, aer (encoding aerobactin biosynthesis), and cnf1 (encoding
cytotoxic necrotizing factor 1) were detected with primers and
amplification conditions as previously described [10].
Conventional strain typing was done by pulsed-field gel electrophoresis (PFGE) analysis of XbaI-digested total DNA, as previously described [11], and by O:H serotyping, which was done
by the E. coli Reference Laboratory (College Park, PA).
Hemagglutination (HA) assays were as previously described [12].
Retrieval of occult pap/ strains. Mixed-sample bacteria from
frozen stocks were grown on agar plates overnight at 377C. Dense
bacterial suspensions were prepared in PBS (pH 7.4) with 5% Dmannose (PBS-mannose) as for HA assays [12], then combined
with an equal volume of 3% human A1P1 erythrocytes in PBSmannose and incubated at 47C for 1 h. Erythrocytes were washed
with cold (47C) PBS-mannose to remove nonadherent bacteria and
plated in serial dilutions on 5% sheep blood agar. Any hemolytic
colonies noted after incubation overnight at 377C were picked for
HA and other testing. In preliminary studies involving mixtures
of laboratory strain HB101 (nonhemolytic, HA0) [12] and wild
type strain IA2 (hemolytic, HA/) [12], a single cold-hemadsorption enrichment step resulted in an Ç100-fold enrichment with the
HA/ strain relative to the HA0 strain.
Statistical methods. Comparisons of proportions were tested
by Fisher’s exact test.
VF PCR with individual isolates. Sixty E. coli colonies
were initially picked for analysis by rep-PCR (e.g., figure 1).
JID 1998;177 (April)
Elimination of duplicates of the same genomic pattern from
individual specimens yielded 23 unique genomic patterns (A –
N, P, and R – Y), or 31 separate predominant strains (i.e., different specimen and/or genomic pattern) (table 1). Of the 23
unique strains, only 2 (9%) were positive for papG, 3 (13%)
for aer, and 3 (13%) for cnf.
Mixed-sample VF PCR. PCR analysis of mixed samples
(which was done for the 2 vaginal samples that contained E.
coli and for 16 of the 17 intestinal samples, all of which grew
E. coli) was significantly more sensitive in detecting VF positivity than was PCR analysis of individual isolates from the
corresponding specimens. Mixed-sample PCR identified 5 additional intestinal specimens as papG/, 1 as aer/, and 2 as
cnf / (table 1). Together, a total of 6 specimens had at least 1 VF
gene detected in the mixed sample but not in the corresponding
predominant strain(s) (table 1). The added VF detection included pap / cnf (2 specimens), pap only (3), and aer only
(1). In only one instance was mixed-sample PCR negative in
the presence of a VF/ predominant strain (P Å .045). Thus,
among the 16 intestinal specimens that were analyzed by both
methods, when presence of a VF gene as detected by either or
both methods was used as the denominator, mixed-sample testing detected a significantly greater proportion of detectable VF
genes than did testing of only the predominant strains, both
for papG (7/7 vs. 2/7; P Å .02) and for papG, aer, and cnf
combined (16/17 vs. 9/17; P Å .017).
Retrieval of occult VF/ strains from mixed samples. Mannose-resistant hemadsorption was used to extract occult
papG/ strains from within the 4 mixed intestinal samples in
which no papG/ predominant strain was identified, if another
Figure 1. Repetitive element (rep) – polymerase chain reaction genomic patterns (enterobacterial repeat intergenic consensus primers) of
selected subjects’ intestinal (i), vaginal (v), and urine (u) E. coli isolates. A, Family 1: lanes b – k, first sampling; lane l, second sampling; lanes
m – p, third sampling. Note similarity of patterns in lanes e – p (strain C), compared with patterns in lanes b – c (strain A) and d (strain B). B,
Family 3, first sampling: intestinal isolates from son GG (lanes b – d), father HH (lanes e – g), mother II (lanes h – j), daughter JJ (lanes k – m),
and daughter KK (lanes n – p). Note similarity of patterns in lanes l and p for subjects JJ and KK (strain N). Lanes a, molecular weight markers.
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
JID 1998;177 (April)
Concise Communications
household member simultaneously had a papG/ specimen
(subjects AA, sampling 1; JJ, samplings 1 and 2; and KK,
sampling 1) (table 1). Multiple hemolytic colonies were obtained from all 4 mixed samples following a single enrichment step. All hemolytic colonies tested exhibited HA and
within each specimen exhibited a uniform rep-PCR genomic
pattern. The 4 new occult papG/ strains represented previously encountered rep-PCR genotype C (subject AA), plus
three unique genotypes (patterns O, Q, and Z: subjects JJ
and KK). VF PCR of each strain yielded papG, aer, and cnf
results consistent with the mixed sample from which the
strain derived (table 1).
Comparison of typing methods. PFGE yielded satisfactory results for 33 (94%) of the 35 discrete isolates, resolving
24 unique genomic patterns that in each instance corresponded with the results of rep-PCR, including duplicates
of the same strain (table 1). However, 2 isolates that were
confirmed as unique strains by both rep-PCR (patterns H
and U) and O:H serotyping consistently sheared, so were
not definitively resolved by PFGE (table 1). O:H serotyping
generally agreed with, but was less discriminating than, repPCR and PFGE (table 1). Three distinct rep-PCR genotypes
(P, S, and W) were all serotyped as O-nontypeable:H-nontypeable. Conversely, 1 multiply encountered rep-PCR genotype (C) was variously serotyped, falsely suggesting multiple
strains. Although the remaining 22 genotypes exhibited
unique serotypes, only 13 had unambiguous designation of
both O and H antigens (table 1).
We found that PCR analysis of mixed intestinal bacterial
samples was significantly more sensitive for identifying intestinal colonization with VF/ strains than was the traditional predominant-strain approach. This suggests that colonization with
urovirulent E. coli may be substantially more frequent than
suggested by available prevalence data for VF positivity among
fecal isolates from healthy hosts (e.g., P fimbriae in 19% [1]
or 11% [13]), since these data generally have been derived
through predominant-strain sampling [4]. Given the complexity
of the intestinal E. coli flora [5, 6], sampling only the predominant strain predictably would be insensitive for detecting particular subtypes of E. coli. Its improved detection of VF/ strains
could make mixed-sample PCR extremely useful for future
epidemiologic studies of colonization.
Since the VF PCR assays we used probably are not maximally
sensitive, our data provide a minimum estimate of the true prevalence of occult intestinal colonization with uropathogenic E. coli
in our study population. The sensitivity of PCR for detecting
VF/ strains in mixed bacterial samples probably could be increased by technical refinements in the assay and by including
additional VFs, such as afa, sfa, kps, and hly [1, 14].
By ‘‘fishing out’’ VF/ E. coli strains from mixed samples
in which the predominant E. coli strains were VF0, we showed
/ 9d43$$ap48
03-04-98 19:03:33
that the positive PCR results we obtained from such samples
were due not to PCR contamination or to non – E. coli species
containing homologous sequences but rather to the presence
within the samples of occult VF/ E. coli. This also allowed us
to type the strains, for comparison with other strains from the
same subject and from other household members, to clarify
colonization and transmission patterns.
Rep-PCR strain typing provided discriminating power comparable to PFGE or serotyping and superseded these methods
by resolving strains with absent or nontypeable surface antigens
or that sheared during PFGE (table 1). Rep-PCR was technically easier and faster than PFGE and, in contrast to serotyping,
required only a small number of (commercially available) reagents. Rep-PCR’s simplicity, its availability to any laboratory
with PCR capability, and its resolving capability at the subspecies level recommends it for future epidemiologic investigations involving uropathogenic E. coli.
Our findings support three possible mechanisms through
which a host may acquire a urovirulent E. coli strain as the
predominant intestinal strain, including emergence of an occult
strain from the host’s own intestinal flora (subject AA), possible acquisition from a household member (AA and BB, JJ and
KK [sampling 1]), and acquisition from some extra-household
reservoir (JJ and KK [sampling 2]). Subject AA provided novel
evidence that an individual can occultly harbor within the intestinal flora potential uropathogens that can subsequently emerge
as seemingly ‘‘new’’ strains. Her partner BB provided novel
evidence of intestinal colonization of the male partner of a
female UTI patient with the woman’s urine organism [15], a
situation that would be expected to put the woman at risk for
recolonization with the strain once cured of it. The finding in
subjects JJ and KK at sampling 2 of new papG/ strains not
found at sampling 1 in any household member even by mixedsample PCR, together with the papG PCR assay’s ability to
detect 1 positive genome among 103 – 107 negative genomes
[8], strongly suggests de novo introduction of these strains
into the household from an outside reservoir. Studies of larger
populations with sensitive methods such as described here will
be needed to determine the frequency and relative importance
to UTI pathogenesis of these three pathways.
Ann Emery, Mary Hayes, and Diana Owensby helped prepare
the manuscript; Parvia Ahmed and Teresa Barela assisted in the
laboratory; Dave Prentiss helped prepare the figure; the University
of Minnesota Clinical Microbiology Laboratory did the urine cultures; and James R. Lupski and Tearith Koeuth helped with the
rep-PCR assays. We thank the study subjects for their participation.
1. Johnson JR. Virulence factors in Escherichia coli urinary tract infection.
Clin Microbiol Rev 1991; 4:80 – 128.
UC: J Infect
Concise Communications
2. Hooton TM, Stamm WE. The vaginal flora and urinary tract infections.
In: Mobley HLT, Warren JW. Urinary tract infections: molecular pathogenesis and clinical management. Washington, DC: American Society
for Microbiology, 1996:67 – 94.
3. Gru¨neberg RN. Relationship of infecting urinary organism to the faecal
flora in patients with symptomatic urinary infection. Lancet 1969; 1:
766 – 8.
4. Lidin-Janson G, Kaijser B, Lincoln K, Olling S, Wedel H. The homogeneity of the faecal coliform flora of normal school-girls, characterized by
serological and biochemical properties. Med Microbiol Immunol 1978;
164:247 – 53.
5. Bettelheim KA, Faiers M, Shooter RA. Serotypes of Escherichia coli in
normal stools. Lancet 1972; 1:1224 – 6.
6. Caugant DA, Levin BR, Selander RK. Genetic diversity and temporal variation
in the E. coli population of a human host. Genetics 1981;98:467–90.
7. Versalovic J, Schneid M, de Bruijn FJ, Lupski JR. Genomic fingerprinting
of bacteria using repetitive sequence – based polymerase chain reaction.
Methods Mol Cell Biol 1994; 5:25 – 40.
8. Johnson JR, Brown JJ. A novel multiply primed polymerase chain reaction
assay for identification of variant papG genes encoding the Gal(a14)Gal – binding PapG adhesins of Escherichia coli. J Infect Dis 1996;
173:920 – 6.
9. Woods CR, Versalovic J, Koeuth T, Lupski J. Whole-cell repetitive element sequence – based polymerase chain reaction allows rapid assess-
JID 1998;177 (April)
ment of clonal relationships of bacterial isolates. J Clin Microbiol 1993;
31:1927 – 31.
Johnson JR, Stapleton AE, Russo TA, Scheutz FS, Brown JJ, Maslow JN.
Characteristics and prevalence within serogroup O4 of a ‘‘J96-like’’
clonal group of uropathogenic Escherichia coli O4:H5 containing the
‘‘class I’’ and ‘‘class III’’ alleles of papG. Infect Immun 1997; 65:
2153 – 9.
Russo TA, Stapleton A, Wenderoth S, Hooton TM, Stamm WE. Chromosomal restriction fragment length polymorphism analysis of Escherichia
coli strains causing recurrent urinary tract infections in young women.
J Infect Dis 1995; 172:440 – 5.
Johnson JR, Swanson JL, Barela TJ, Brown JJ. Receptor specificities of
variant Gal(a1-4)Gal – binding papG adhesins of uropathogenic Escherichia coli as assessed by hemagglutination phenotypes. J Infect Dis
1997; 175:373 – 81.
Ika¨heimo R, Siitonen A, Heiskanen T, et al. Recurrence of urinary tract
infection in a primary care setting: analysis of a 1-year follow-up of
179 patients. Clin Infect Dis 1996; 22:91 – 9.
Johnson JR, Russo TA, Scheutz F, et al. Discovery of disseminated J96like strains of uropathogenic Escherichia coli O4:H5 containing genes
for both PapGJ96 (‘‘class I’’) and PrsGJ96 (‘‘class III’’) Gal(a1-4)Gal –
binding adhesins. J Infect Dis 1997; 175:983 – 8.
Foxman B, Zhang L, Tallman P, et al. Transmission of uropathogens
between sex partners. J Infect Dis 1996; 175:989 – 92.
The Individual and Joint Contributions of Helicobacter pylori Infection and
Family History to the Risk for Peptic Ulcer Disease
Hermann Brenner, Dietrich Rothenbacher, Gu¨nter Bode,
and Guido Adler
Departments of Epidemiology and Internal Medicine I,
University of Ulm, Ulm, Germany
Family history of peptic ulcer and infection with Helicobacter pylori have been identified as
major risk factors for peptic ulcer disease. It is unclear, however, to what degree their impacts are
independent of each other. This question was addressed in a cross-sectional study among 299
consecutive out-patients (25 – 54 years old) of a general practitioner. Adjusted odds ratios (95%
confidence intervals) for gastroscopically verified peptic ulcer disease were 3.8 (1.4 – 10.1) for persons
with H. pylori infection, 8.4 (2.9 – 24.1) for persons with a family history of ulcer, and 29.5 (6.1 –
143.9) for persons with both risk factors compared with persons without these risk factors. These
results suggest strong, multiplicative contributions of both factors to the risk for peptic ulcer disease.
It has long been noted, that gastric and duodenal ulcers
cluster within families. Studies of twins suggest that the observed familial aggregation is largely due to genetic factors,
but common environmental factors also seem to play a major
role [1].
Other well-known risk factors for peptic ulcer include smoking and use of nonsteroidal antiinflammatory drugs [2]. In re-
Received 20 June 1997; revised 22 October 1997.
Participation in the study was voluntary, and informed consent was obtained
from each participant. The study was approved by the Ethics Board of the
University of Ulm.
Reprints or correspondence: Dr. Hermann Brenner, University of Ulm, Dept.
of Epidemiology, Albert-Einstein-Allee 43, D-89081 Ulm, Germany.
The Journal of Infectious Diseases 1998;177:1124–7
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
cent years, infection with Helicobacter pylori has been identified as another strong risk factor [2 – 4]. Like peptic ulcer
disease, H. pylori infection, which is typically acquired in early
childhood, has been found to cluster within families [5 – 6].
Familial aggregation of infection may again be due to various
pathways, including common genetic or environmental factors
that increase susceptibility to infection [7 – 8], but also to intrafamilial transmission of the agent [9]. It is unclear, however,
to what degree H. pylori infection may explain intrafamilial
clustering of peptic ulcer disease. Furthermore, the joint contributions of H. pylori infection and family history of ulcer to
the risk for peptic ulcer disease are unknown.
Herein, we assess the independent and joint contributions of
H. pylori infection and family history of ulcer to the risk for
peptic ulcer disease in an unselected sample of out-patients of
a general practitioner in southern Germany.
UC: J Infect
JID 1998;177 (April)
Concise Communications
Material and Methods
Study design and study population. A cross-sectional study
was done among consecutive 15- to 79-year-old outpatients who
visited the office of a general practitioner in Blaustein, a community of Ç15,000 inhabitants located in southern Germany. Patients
were recruited 3 days a week during usual office hours between
June and September 1996. Patients were recruited without regard
to the reason for their visit.
Of 531 eligible subjects, 501 (94.4%) agreed to participate in
the study. The study was designed to address several questions,
including the prevalence and risk factors for H. pylori infection,
which are reported elsewhere [10], and the consequences of H.
pylori infection. The current analysis concerning the impact of H.
pylori infection and of family history on the risk for peptic ulcer
is restricted to 299 adults who were 25–54 years of age. Patients
õ25 years old (n Å 60) were excluded because none of these
patients had a gastroscopically verified gastric or duodenal ulcer
(GVU). Patients ú55 years old (n Å 142) were excluded to minimize eventual bias due to incomplete or inaccurate recall or ascertainment of family history of ulcer. In particular, parents of these
patients would typically have been born between 1873 and 1923,
and it is likely that diagnostic verification of peptic ulcers has been
very poor throughout most of their lifetime.
Data collection. The field work was done by trained research
assistants in a separate room in the general practice. Active infection with H. pylori was determined by use of the 13C-urea breath
test. First, an initial breath sample was collected in a plastic bag.
The patients then received 200 mL of apple juice (pH 2.2–2.4),
which contained 75 mg of 13C-urea (Mass Trace, Woburn, MA).
Thirty minutes later, a second breath sample was collected. The
breath samples were analyzed using an isotope-selective, nondispersive infrared spectrometer (Wagner-Analytical Systems, Worpswede, Germany). A change of the 13CO2/12CO2 ratio over baseline
of ú5‰ was considered to indicate active infection. Sensitivities
and specificities of the 13C-urea breath test close to 100% have
consistently been reported, suggesting the test to be the reference
standard in patients in whom endoscopy is not indicated [11, 12].
Although urea breath test results among H. pylori –positive patients
are affected to some extent by food intake, the distinction between
infected and uninfected patients is highly accurate even under
nonfasting conditions [13].
Patients filled out a standardized questionnaire between collection of the first and second breath samples. They were asked if
they ever had a GVU and whether their mother or father had a
history of a physician-diagnosed gastric or duodenal ulcer. The
questionnaire also contained questions on sociodemographic factors and on other potential risk factors for peptic ulcer, such as
cigarette smoking.
Statistical analysis. We compared the frequency of GVU
among persons with a currently or previously treated H. pylori
infection with the frequency among other persons, and we compared the frequency of GVU among persons with a family history
of ulcer with the frequency among other persons. The individual
and joint relationships of H. pylori infection and family history of
ulcer with subjects’ lifetime history of GVU were quantified by
crude and adjusted odds ratios (OR) and their 95% confidence
intervals (CI). Adjustment was made for age, sex, and smoking
by conditional multiple logistic regression. All analyses were done
with the SAS statistical software package [14].
The study sample included 191 women and 108 men with
a mean age of 38.1 years. About one-third of the subjects were
current smokers (33.6%), and about one-quarter were former
smokers (25.2%).
Nineteen patients (6.4%) reported a lifetime history of GVU.
Seventy-nine patients (26.4%) were either infected with H.
pylori at the time of the examination (n Å 70) or had been
successfully treated for H. pylori infection in the past (n Å 9).
Forty-one patients (13.7%) reported that their father (n Å 26),
their mother (n Å 12), or both (n Å 3) had had a physiciandiagnosed gastric or duodenal ulcer.
The individual relationships of infection with H. pylori and
of family history of ulcer with subjects’ lifetime history of
GVU are shown in table 1. H. pylori infection was associated
with an almost 4-fold increase in risk for GVU (adjusted OR,
3.8; 95% CI, 1.4 – 10.1). An even stronger relationship was
seen between family history of peptic ulcer and subjects’ lifetime history of GVU (adjusted OR, 8.4; 95% CI, 2.9 – 24.1).
Table 2 shows the joint relationship of infection with H.
pylori and family history of ulcer with subjects’ lifetime history
of GVU. Lifetime prevalence of GVU ranged from 2.6%
among persons without H. pylori infection and without family
history of ulcer to 31.3% among persons with both risk factors.
Intermediate levels (7.9% and 16.0%, respectively) were observed for individuals with only one of the two risk factors.
Effects of both risk factors were almost multiplicative. After
adjustment for age, sex, and cigarette smoking in multivariable
analysis, OR were 3.7 (95% CI, 1.0 – 14.0) for H. pylori – positive persons with a negative family history, 8.4 (95% CI, 1.9 –
36.4) for H. pylori – negative persons with a positive family
history, and 29.5 (95% CI, 6.1 – 143.9) for H. pylori – positive
persons with a positive family history, compared with persons
with neither of the two risk factors.
/ 9d43$$ap48
03-04-98 19:03:33
Table 1. Individual relationships of infection with H. pylori and of
family history of ulcer with gastroscopically verified ulcer (GVU).
Odds ratio (95% confidence
Infection with H. pylori
No (n Å 220)
Yes‡ (n Å 79)
Family history of ulcer
No (n Å 258)
Yes (n Å 41)
GVU, no. (%)
9 (4.1)
10§ (12.7)
3.4 (1.3 – 8.6)
3.8 (1.4 – 10.1)
10 (3.9)
9 (22.0)
6.9 (2.6 – 18.2)
8.4 (2.9 – 24.1)
* Adjusted for sex, age (in years), and smoking (ever vs. never) by multiple
logistic regression.
Reference category.
Currently infected or previously treated for H. pylori infection.
Includes 9 patients with GVU among 70 currently infected patients and 1
patient with GVU among 9 who were currently uninfected but who were
previously treated for H. pylori infection.
UC: J Infect
Concise Communications
JID 1998;177 (April)
Table 2. Joint relationship of infection with H. pylori and of family history of ulcer with gastroscopically verified ulcer (GVU).
Odds ratio (95% confidence interval)
Infection with
H. pylori
Family history
of ulcer
GVU, no. (%)
3.3 (0.9 – 11.7)
7.2 (1.8 – 28.7)
17.0 (4.3 – 67.0)
3.7 (1.0 – 14.0)
8.4 (1.9 – 36.4)
29.5 (6.1 – 143.9)
* Adjusted for sex, age (in years), and smoking (ever vs. never) by multiple logistic regression.
Reference category.
Currently infected or previously treated for H. pylori infection.
The differences between the adjusted and the crude OR were
mainly due to removal of confounding by sex, which was
strongly related to peptic ulcer in this sample (adjusted OR for
women vs. men, 0.19; 95% CI, 0.06 – 0.59). After sex had been
controlled for, additional control for age and smoking did not
materially affect the results.
This study demonstrates that H. pylori infection and family
history of peptic ulcer disease are strong, independent risk
factors for peptic ulcer. If both factors are present, the risk for
gastric or duodenal ulcer is increased to very high levels in an
almost multiplicative manner.
Our result concerning the individual effect of infection with
H. pylori on the risk for peptic ulcer is in close agreement with
findings in previous studies [2]. Our estimate of the impact of
family history of ulcer is somewhat higher than that found in
most previous studies [1]. This may be due to several reasons.
First, previous studies have not controlled for potential confounders. In our study, we controlled for covariates by multiple
logistic regression, which strengthened the association between
family history of ulcer and peptic ulcer among study participants. Second, previous studies have included older patients,
among whom ascertainment of parental ulcer is likely to be
far from complete. To minimize this problem, we excluded
patients ú55 years of age from the analysis. We feel confident
about this decision, given that only 2.8% of patients ú55 years
old, compared with 13.7% of patients 25 – 54 years old, reported
a physician-diagnosed gastric or duodenal ulcer among their
To address the impact of excluding patients ú55 years old,
we carried out additional analyses in which these patients were
included. These analyses showed the same pattern of multiplicative, independent associations of both H. pylori infection and
family history with patients’ lifetime history of ulcer, but the
estimate of the impact of family history was somewhat lower.
This pattern would be consistent with underestimation of the
impact of family history due to its incomplete ascertainment
in studies including older subjects [15].
/ 9d43$$ap48
03-04-98 19:03:33
To our knowledge, this is the first study that addresses the
joint effects of H. pylori infection and family history of ulcer.
This issue is of particular interest, given that the mechanisms
by which family history of ulcer increases disease risk are
not fully understood. It would appear plausible that familial
clustering of peptic ulcer disease may at least partly be mediated by H. pylori infection. In this study, however, family
history of ulcer was a very strong risk factor for GVU even in
the absence of H. pylori infection. These results suggest that
H. pylori infection accounts for, at best, a rather minor share
of the strong familial clustering of peptic ulcer disease. Our
results also illustrate that H. pylori infection is a major risk
factor for peptic ulcer whether familial predisposition does or
does not exist.
In our study, the individual relationship of family history of
ulcer with peptic ulcer disease was about twice as strong as
the individual relationship of H. pylori infection with peptic
ulcer disease. However, because H. pylori infection was about
twice as common as family history of ulcer, it accounts for as
large a share of peptic ulcers in this population as does a family
history of the disease. This is of important clinical relevance
since H. pylori infection, in contrast to family history of ulcer,
offers possibilities for preventive or therapeutic intervention.
In particular, the very high risk for peptic ulcer among H.
pylori – positive persons with a family history of ulcer suggests
that treatment of H. pylori infection may be worthwhile in this
group even before the development of peptic ulcer. In this
context, family history of ulcer may serve as a useful screening
tool in deciding who to test and eventually treat for H. pylori
We appreciate the help of A. Behr and his staff (general medical
practice, Blaustein) in the conduct of this study.
1. Rotter JI, Shohat T, Petersen GM. Peptic ulcer disease. In: King RA,
Rotter JI, Motulsky AG. The genetic basis of common diseases. New
York, Oxford: Oxford University Press, 1992:240 – 78.
UC: J Infect
JID 1998;177 (April)
Concise Communications
2. Kurata JH, Nogawa AN. Meta-analysis of risk factors for peptic ulcer. J
Clin Gastroenterol 1997; 24:2 – 17.
3. Graham DY, Lew GM, Klein PD, et al. Effect of treatment of Helicobacter pylori infection on the long-term recurrence of gastric or
duodenal ulcer. A randomized, controlled study. Ann Intern Med
1992; 116:705 – 8.
4. Nomura A, Stemmermann GN, Chyou P, Perez-Perez GI, Blaser MJ.
Helicobacter pylori infection and the risk for duodenal and gastric
ulceration. Ann Intern Med 1994; 120:977 – 81.
5. Drumm B, Perez-Perez GI, Blaser MJ, Sherman PM. Intrafamilial clustering of Helicobacter pylori infection. N Engl J Med 1990; 322:359 – 63.
6. Wang JT, Sheu JC, Lin JT, Wang TH, Wu MS. Direct DNA amplification
and restriction pattern analysis of Helicobacter pylori in patients with
duodenal ulcer and their families. J Infect Dis 1993; 168:1544 – 8.
7. Malaty HM, Engstrand L, Pedersen NL, Graham DY. Helicobacter pylori
infection: genetic and environmental influences. a study of twins. Ann
Intern Med 1994; 120:982 – 6.
8. Go MF, Graham DY. How does Helicobacter pylori cause duodenal ulcer
disease: the bug, the host, or both? J Gastroenterol 1994; 9:S8 – S12.
/ 9d43$$ap48
03-04-98 19:03:33
9. Riccardi VM, Rotter JI. Familial Helicobacter pylori infection. Ann Intern
Med 1994; 120:1043 – 5.
10. Rothenbacher D, Bode G, Winz T, Berg G, Adler G, Brenner H. H. pylori
in out-patients of a general practitioner: prevalence and determinants
of current infection. Epidemiol Infect 1997; 119:151 – 7.
11. Braden B, Scha¨fer F, Caspary WF, Lembcke B. Nondispersive isotopeselective infrared spectroscopy: a new analytical method for 13C-urea
breath tests. Scand J Gastroenterol 1996; 31:442 – 5.
12. Thijs JC, van Zwet AA, Thijs WJ, et al. Diagnostic tests for Helicobacter
pylori: a prospective evaluation of their accuracy, without selecting a
single test as the gold standard. Am J Gastroenterol 1996; 91:2125 – 9.
13. Braden B, Duan LP, Caspary WF, Lembcke B. More convenient 13C-urea
breath test modifications still meet the criteria for valid diagnosis of
Helicobacter pylori infection. Z Gastroenterol 1994; 32:198 – 202.
14. SAS Institute. SAS Language: reference. Version 6, 1st ed. Cary, NC:
SAS, 1990.
15. Copeland KT, Checkoway H, McMichael AJ, Holbrook RH. Bias due to
misclassification in the estimation of relative risk. Am J Epidemiol
1977; 105:488 – 95.
UC: J Infect
Selection and Characterization of Toxoplasma gondii Mutants Resistant to
Randolph L. Berens, Edward C. Krug,* Paul B. Nash,*
and Tyler J. Curiel
Division of Infectious Diseases, University of Colorado Health Sciences
Center, Denver, Colorado
Toxoplasma gondii infection, like malaria, is sensitive to inhibition by artemisinin (ART). Mechanisms of action for ART in malaria treatment have been proposed, but little is known about its
effects in T. gondii infection. To better understand its inhibitory effects on T. gondii, mutants resistant
to ART were selected by progressive culture in permissive levels of the drug. Five clonal isolates
were established and characterized. The isolates were Ç65-fold less sensitive to ART than is the
parental RH and showed cross-resistance to the ART derivatives dihydroartemisinin and artemether.
In addition to ART resistance, 1 clone (C9) formed morphologically unusual parasitophorous vacuoles and another (A2) was avirulent for mice and protected mice from challenge with the wild type.
These clonal T. gondii mutant isolates will be useful for the study of not only the mechanism of
action of ART but also parasite vacuole biology and virulence factors.
Infection with Toxoplasma gondii is a significant cause of
morbidity and mortality in immunocompromised hosts, including those infected with the human immunodeficiency virus.
Current anti-Toxoplasma drugs have significant toxicities,
which may compromise anti-Toxoplasma or antiretroviral therapy. Thus, there is great need for alternative agents.
Studies from this [1] and other laboratories [2] have
shown that T. gondii tachyzoites are sensitive to growth
inhibition by artemisinin (qinghaosu, ART) or its derivatives. ART is a sesquiterpene lactone with an unusual endoperoxide bond [3] that is, along with several derivatives,
very effective against chloroquine-resistant strains of Plasmodium falciparum [3 – 5]. Although a mechanism of action
for ART in malaria treatment has been proposed [6], its
mode of action in T. gondii infection is unknown. Mutants
of T. gondii have been used previously to study the mechanism of action of several anti-Toxoplasma agents [7, 8].
Using similar methods, we derived mutant lines of T. gondii
resistant to ART. Reported here are the results of the selection, isolation, and partial characterization of several ARTresistant clonal lines.
Received 6 June 1997; revised 24 October 1997.
Presented in part: IX International Congress of Parasitology, Sydney, Australia, July 1997.
Animal studies were conducted according to appropriate guidelines of the
University of Colorado Health Sciences Center Animal Care and Use Committee under the supervision of veterinary staff. Animal facilities are accredited
by the American Association for the Accreditation of Laboratory Animal Care.
Grant support: NIH (AI-39379 to T.J.C.) and Colorado Advanced Technology Institute (to R.L.B.).
Reprints or correspondence: Dr. Randolph L. Berens, Division of Infectious
Diseases, University of Colorado Health Sciences Center, 4200 E. 9th Ave.,
Denver, CO 80262 (berens [email protected]).
* Present affiliations: Denver Public Schools, Aurora, CO (E.C.K.); Heska
Corporation, Fort Collins, CO (P.B.N.).
The Journal of Infectious Diseases 1998;177:1128–31
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
Materials and Methods
Materials. ART, dihydroartemisinin (DHART), deoxyartemisinin (DESART), and artemether (MEART) were gifts of Hasuer,
Inc. (Boulder, CO). All other reagents were from Sigma (St. Louis)
except bovine serum (Gemini, Calabasas, CA), plasticware (Corning, Corning, NY), rabbit anti-Toxoplasma polyclonal antibody
(Dako, Carpinteria, CA), and female ICR mice (Harlan SpragueDawley, Indianapolis).
Parasite growth. The RH isolate of T. gondii was maintained
in and harvested from human fibroblasts as previously described
[1, 9]. Human fibroblasts (HF) were grown in VA-l3 [1] medium
supplemented with l0% adult bovine serum for growth of noninfected fibroblasts or 0.3% bovine (Cohn fraction V) serum albumin
for infected fibroblasts. Harvested tachyzoites were enumerated
using a hemacytometer. Parasites were frozen, stored, and thawed
as described [10].
Drug stocks. ART and its derivatives were made as 20 mg/
mL DMSO (dimethysulfoxide) stocks, aliquotted, and stored at
0207C. Fresh aliquots were used for each study. Pyrimethamine
was made as a 5-mg/mL DMSO stock. In all studies, the DMSO
concentration was õ1%.
Plaque assay. Confluent HF monolayers were infected with
200 fresh tachyzoites/well in 24-well plates and incubated in duplicate for 3 h before addition of test compounds or with pyrimethamine (10 mg/mL) as a positive drug control. Plates were incubated
for 120 h and then methanol-fixed. Plaques were scored using a
‘‘4/’’ scale as described [1]. For mutant growth rate estimations,
plates were infected as above and incubated until the plaque size
was that of RH (4/) at 120 h. Differences in growth rates were
estimated by comparing the time for the mutant and RH to reach
4/. Plaque-forming units and plating efficiency were determined
as described [8].
ELISA quantitation. ELISA was done as described [11] in 96well culture plates with confluent HF monolayers. Test compound
was added 3 h after infection with 2000 tachyzoites/well. At 72
h after initial infection, plates were examined by phase-contrast
microscopy to estimate approximate concentrations for 0%, 50%,
and 100% growth inhibition and then methanol-fixed. The plates
were washed and developed by sequential addition of rabbit anti-
UC: J Infect
JID 1998;177 (April)
Concise Communications
Toxoplasma antibody diluted 1:20,000, horseradish peroxidase–
conjugated goat anti–rabbit IgG, and o-phenylenediamine dihydrochloride as described [11]. The optical density was read at 450
nm. Results were averaged, and percent inhibition was calculated
and plotted versus compound concentration. The concentrations
resulting in 50% (IC50) and 90% inhibition (IC90) were determined
Selection of mutants. Two different procedures were used. For
chemical mutagenesis, infected HF were exposed to either Nmethyl-N*-nitro-N-nitrosoguanidine (MNNG) or ethylmethane sulfonate as described [7]. For selection of mutants by progressive
culture in ART, cultures were started at an ART concentration of
0.2 mg/mL. The ART concentration was increased each time parasites displayed reproducible growth and infectivity after three subpassages at that drug concentration. Clonal isolates were derived
by limiting dilution in HF cells in 96-well plates [12].
Pathogenicity studies. Groups (3 each) of female ICR mice
weighing 19–23 g were used. All mice were infected by intraperitoneal injection of 2000 freshly isolated tachyzoites [13]. Impression smears were made of the peritoneal fluid of dying mice and
examined by phase-contrast microscopy for tachyzoites. Mice that
survived 45 days after test inoculation were then challenged with
RH tachyzoites and observed for 45 additional days.
Attempts were made to establish resistant mutants using
ethylmethane sulfonate or MNNG, but neither mutagen resulted in establishment of resistance using an ART selection
range of 0.4 – 4 mg/mL. However, exposure to either mutagen
resulted in pyrimethamine-resistant mutants capable of
growth in 25 mg/mL pyrimethamine, 100 times the RH IC90
(data not shown).
By progressive culture in ART, however, a culture growing in 35 mg/mL ART was established over a period of 9
months. At this concentration, growth of the mutants was
Ç5% that of RH, and ART was significantly toxic for HF.
Since the mutant population growing in 25 mg/mL ART had
an acceptable doubling time and because there were no signs
of HF toxicity, 25 mg/mL was chosen as the concentration
for further work.
Based on plaque assay, the uncloned mutant line growing
in 25 mg/mL ART was Ç50-fold less sensitive to ART than
RH and grew approximately half as fast. No significant abnormalities were seen in mutant tachyzoite morphology by phasecontrast microscopy or hematoxylin-eosin staining.
Twenty-three clonal lines of the resistant culture were established by limiting dilution, 5 of which were chosen for further
study. The plating efficiency of the clones was Ç50%, equal
to that of RH [8]. Four clones, designated A2, H9, A4, and
G23, had normal morphology and growth rates relative to RH
of 40%, 50%, 60%, and 60%, respectively.
One clone (C9) had a growth rate of 60% but formed
enlarged, atypical parasitophorous vacuoles. Normally there
is little space between the tachyzoites and the vacuole membrane, but in C9 there was a large separation between it
Figure 1. Artemisinin inhibition of wild type (WT) T. gondii and resistant clones (A2, A4, C9, G23, and H9).
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
Concise Communications
JID 1998;177 (April)
Figure 2. Comparative inhibition of artemisinin (ART) and artemisinin derivatives dihydroartemisinin (DHART) and artemether (MEART)
in wild type (WT) and A2 clone.
and the parasites. Within 24 h of infection, the resulting
parasitophorous vacuole increased to a size comparable to
the HF nucleus, while containing as few as 2 tachyzoites.
The ART resistance of each clone was compared with RH
(figure 1). All clones were Ç65-fold more ART-resistant
than was the parental RH. The average IC50 and IC90 values
for the 5 clones were 6.6 mg/mL (23.4 mM) and 25.2 mg/mL
(89.2 mM), respectively, compared with IC50 and IC90 values
of 0.10 mg/mL (0.36 mM) and 0.38 mg/mL (1.3 mM) for
RH. Each clone was also tested for resistance to the ART
derivatives DHART and MEART, which have significant
anti-Toxoplasma activity [1]. The respective IC50 and IC90
values for DHART for A2 (figure 2) were 3.8 mg/mL and
10.1 mg/mL versus 0.11 mg/mL and 0.23 mg/mL for RH.
MEART values were 1.6 mg/mL and 4.0 mg/mL for the A2
compared with 0.04 mg/mL and 0.09 mg/mL for RH. The
other clones showed similar resistance profiles, with IC50
and IC90 values essentially identical to those found for A2
(data not shown).
Both RH and the clones were tested for sensitivity to
DESART [14]. RH, the clones, and HF were not inhibited by
50 mg/mL DESART. All clones showed parental sensitivity to
pyrimethamine (data not shown).
The C9 clone showed the same atypical parasitophorous
vacuole in the presence of DHART or MEART as for ART.
/ 9d43$$ap48
03-04-98 19:03:33
Exposure to DESART resulted in normal parasitophorous vacuoles.
The effect of verapamil on ART-resistant clones was tested
because it increases chloroquine efficacy in multidrug-resistant
P. falciparum malaria [15]. However, incubation in the presence of 20 mM verapamil (a subinhibitory concentration [2])
did not affect the activity of ART (data not shown).
Each clone was tested for virulence in mice, and all except
for A2 caused fatal toxoplasmosis in 8 – 11 days, with a clinical
course identical to that of RH. Tachyzoites were present in
peritoneal fluid of all dying mice. Mice infected with A2
showed no signs of toxoplasmosis for up to 45 days after
infection, when they were challenged with 2000 fresh RH
tachyzoites. These mice showed no signs of toxoplasmosis for
up to 45 additional days after challenge with RH. These tests
were repeated three times with identical results. However, 2000
A2 tachyzoites injected peritoneally into SCID mice killed
them in £8 days.
The 5 clones were cultured in the absence of ART for 30
days and then placed back in 25 mg/mL ART. After a growth
lag of Ç48 h, all clones resumed growth in the presence of
drug identical to that of cultures continuously on drug and had
virulence in mice as was described above. Of interest, the C9
clone formed normal parasitophorous vacuoles when cultured
in the absence of ART but reformed enlarged vacuoles when
UC: J Infect
JID 1998;177 (April)
Concise Communications
placed back in drug. All the clones were stable to freezing and
storage in liquid nitrogen for at least 6 months.
We thank Matthew B. Purner for excellent technical assistance.
ART and certain of its derivatives appear to be potential
new anti-Toxoplasma agents. As a system for future research,
ART-resistant mutants of T. gondii were obtained by exposure
to progressively higher concentrations of drug. Chemical mutagenesis, which was successfully used to produce resistance to
pyrimethamine, did not yield ART resistance. The uncloned
ART-resistant line was Ç50-fold less sensitive to ART than
was the parental RH. The mutants cloned from this population
were Ç65-fold less sensitive to ART than was RH and showed
cross-resistance to the ART derivatives DHART and MEART.
Neither RH nor the clones were sensitive to DESART (ART
with a reduced endoperoxide bond), suggesting that, as for
Plasmodium species, the presence of the endoperoxide bond is
required for the inhibitory effects of ART and derivatives on
Toxoplasma species [6, 14].
None of the clones showed cross-resistance to the antifolate
pyrimethamine, and incubation of the clones with ART in the
presence of verapamil did not affect the activity of ART, suggesting that the resistance of these clones is specific for ART
and its derivatives and does not involve increased drug efflux.
These clones will be useful in understanding the mechanism
of action of ART in Toxoplasma species.
Ten RH tachyzoites are lethal for mice [13], whereas 2000
A2 tachyzoites were not, demonstrating the avirulence of A2.
Inoculation with A2 protected mice against RH challenge, suggesting that A2 elicits protective immunity. Study of the differences between A2 and RH may shed light on T. gondii virulence.
The C9 replicated in a normal parasitophorous vacuole when
cultured in the absence of ART but formed the enlarged atypical vacuole when placed back on drug. The genetic changes
resulting in ART resistance in the C9 appear to be linked with
the physiology of the parasitophorous vacuole. This linkage
may aid in vacuole studies.
/ 9d43$$ap48
03-04-98 19:03:33
1. Ke OY, Krug EC, Marr JJ, Berens RL. Inhibition of growth of Toxoplasma
gondii by qinghaosu and derivatives. Antimicrob Agents Chemother
1990; 34:1961 – 5.
2. Holfels E, McAuley J, Mack D, Milhous WK, McLeod R. In vitro effects
of artemisinin ether, cycloguanil hydrochloride (alone and in combination with sulfadiazine), quinine sulfate, mefloquine, primaquine phosphate, trifluoperazine hydrochloride, and verapamil on Toxoplasma gondii. Antimicrob Agents Chemother 1994; 38:1392 – 6.
3. China Cooperative Research Group on Qinghaosu and Its Derivatives as
Antimalarials. Chemical studies on qinghaosu (artemisinin). J Tradit
Chin Med 1982; 2:3 – 8.
4. China Cooperative Research Group on Qinghaosu and Its Derivatives as
Antimalarials. Clinical studies on the treatment of malaria with qinghaosu and its derivatives. J Tradit Chin Med 1982; 2:45 – 50.
5. Karbwang J, Na-Bangchang K, Thanavibul A, Ditta-in M, Harinasuta T.
A comparative clinical trial of two different regimens of artemether
plus mefloquine in multidrug resistant falciparum malaria. Trans R Soc
Trop Med Hyg 1995; 89:296 – 8.
6. Meshnick SR. The mode of action of antimalarial endoperoxides. Trans
R Soc Trop Med Hyg 1994; 88(suppl):S31 – 2.
7. Pfefferkorn ER, Pfefferkorn LC. Arabinosyl nucleosides inhibit Toxoplasma gondii and allow the selection of resistant mutants. J Parasitol
1976; 62:993 – 9.
8. Pfefferkorn ER, Pfefferkorn LC. Toxoplasma gondii: characterization of
a mutant resistant to 5-luorodeoxyuridine. Exp Parasitol 1977; 42:44 –
9. Krug EC, Marr JJ, Berens RL. Purine metabolism in Toxoplasma gondii.
J Biol Chem 1989; 264:10,601 – 7.
10. Daggett PM, Nerad TA. Cryopreservation of Toxoplasma gondii. In: Lee
JJ, Soldo AT, eds. Protocols in protozoology. Lawrence, KS: Society
of Protozoology, 1992:A – 85.
11. Derouin F, Chastang C. Enzyme immunoassay to assess effect of antimicrobial agents on Toxoplasma gondii in tissue culture. Antimicrob
Agents Chemother 1988; 32:303 – 7.
12. Baum KF, Berens RL, Jones RH, Marr JJ. A new method for cloning
Giardia lamblia, with a discussion of the statistical considerations of
limiting dilution. J Parasitol 1988; 74:267 – 9.
13. Luft BJ. Potent in vivo activity of arprinocid, a purine analogue, against
murine toxoplasmosis. J Infect Dis 1986; 154:692 – 4.
14. Brossi A, Venugopalan B, Dominguez Gerpe L, Yeh HJ, et al. Arteether,
a new antimalarial drug: synthesis and antimalarial properties. J Med
Chem 1988; 31:645 – 50.
15. Martiney JA, Cerami A, Slater AF. Verapamil reversal of chloroquine
resistance in the malaria parasite Plasmodium falciparum is specific for
resistant parasites and independent of the weak base effect. J Biol Chem
1995; 270:22,393 – 8.
UC: J Infect
Persistence of Humoral Response against Sporozoite and Blood-Stage Malaria
Antigens 7 Years after a Brief Exposure to Plasmodium vivax
E´rika M. Braga, Cor J. F. Fontes,
and Antoniana U. Krettli
Departamento de Parasitologia, Universidade Federal de Minas Gerais
and Centro de Pesquisas Rene´ Rachou/FIOCRUZ, Belo Horizonte, and
Universidade Federal de Mato Grosso, Cuiaba´, Brazil
The persistence of malarial antibodies was evaluated in subjects living in a rural community (in
Minas Gerais State, Brazil) briefly exposed to a Plasmodium vivax malaria outbreak outside of the
area in which malaria was endemic. Transmission was interrupted by treatment of all patients and
their relatives and/or neighbors, although the latter had neither symptoms nor blood parasites.
Antibodies to P. vivax antigens (recombinant proteins from sporozoites [rPvCS] and from blood
stages [rPv200]) were measured in parallel by ELISA with sera collected at two time points after
transmission. Anti-rPvCS IgG antibodies were positive in Ç40% and 20% of the subjects 8 months
and 7 years after exposure, respectively. Anti-rPv200 IgG was first detected in 61% of the subjects
who had had malarial symptoms and remained positive in 47% after 7 years. Among the prophylactically treated group, anti-rPv200 IgG was detected in only 28% after 8 months. The levels of both
antibodies decreased with time in all positive subjects.
Thirty years ago, malaria was endemic in Ç80% of Brazil.
The Brazilian Ministry of Health (BMH) has virtually eliminated transmission from the areas of greatest socioeconomic
importance in the country, including Minas Gerais State [1].
However, malaria has always remained endemic in the Amazon
Basin, which in the last decades has undergone a process of
settlement and exploitation that contributes in many ways to
the increase in disease prevalence. According to the BMH,
Plasmodium vivax is now responsible for 64% of the halfmillion malaria cases presently reported; Plasmodium falciparum accounts for the rest of the cases. Moreover, human migration from the Amazon has initiated new foci of transmission
in areas previously free of the disease, including the present
study area in Minas Gerais.
In 1988, in a small rural community near Mantena, Minas
Gerais, a P. vivax malaria outbreak occurred for 50 days. It
was promptly identified and controlled by the BMH. In a previous study on this outbreak, we observed that 42% of 62 subjects
studied had seroreactivity against recombinant sporozoite antigens 8 months after transmission [2]. Considering that the population was briefly exposed to infected vectors, this rate was
considered high. Now we study whether the antibodies against
Received 25 March 1997; revised 13 October 1997.
Informed consent was obtained from patients according to the Fundac¸a˜o
Oswaldo Cruz Ethics Comittee (Brazilian Ministry of Health), 26 November
Financial support: Brazilian National Research Council (grant 523231/946); Fundac¸a˜o de Amparo a` Pesquisa do Estado de Minas Gerais (grant CBS1224/95).
Reprints or correspondence: Dr. A. U. Krettli, Centro de Pesquisas Rene´
Rachou, Av. Augusto de Lima, 1715, Belo Horizonte, Brazil 30190-002
([email protected]).
The Journal of Infectious Diseases 1998;177:1132–5
q 1998 by The University of Chicago. All rights reserved.
/ 9d43$$ap48
03-04-98 19:03:33
P. vivax antigens persist in the absence of reinfection or reexposure to the parasite. Immunologic investigations of P. vivax
have long been neglected. This is partly because in vitro P.
vivax continuous cultures are not yet available, and parasitemias in naturally infected persons are frequently low and
unsuitable as a source of antigen. Studies on the mechanisms
of immunity to P. falciparum suggest that protection shown
by adults in areas in which the organism is hyperendemic
requires regular contact with the parasite to persist [3]. The
nonsterile and short-lived character of such immunity, which
is species-, stage-, and strain-specific, is well documented, although not sufficiently understood [4]. Protection against severe disease decreases rapidly in the absence of antigenic stimulation [5]. However, antibodies to defined P. falciparum
antigens persist in some adults who once lived in areas in
which malaria is hyperendemic but have long since moved
from there [6, 7].
The absence of further malaria transmission in the area studied here and the availability of frozen sera collected from exposed persons 8 months after the outbreak [2] allowed the
present study. The persistence of antibodies to P. vivax antigens
was studied 7 years later in the same area by use of freshly
isolated plasma obtained from the previously exposed persons.
The P. vivax circumsporozoite protein (CS) [8] and the bloodmerozoite surface antigen (MSP-1) [9] were used in parallel
experiments. Data on the persistence of antibodies to P. vivax
antigens could be useful for understanding naturally acquired
immunity and for evaluation of future P. vivax vaccines.
Materials and Methods
The malaria outbreak. P. vivax transmission lasted for Ç50
days, in 1988, in a rural community (Mantena, Minas Gerais) as
we previously described [2]. Malaria has never been reported in
this area before, according to the BMH. The study area is located
UC: J Infect
JID 1998;177 (April)
Concise Communications
Ç1100 miles from the nearest area reporting endemic transmission
of P. vivax at the time, Mato Grosso State. Control measures
initially included active search for acute malaria by collecting thick
blood films from all febrile and symptomatic persons (n Å 45)
plus their relatives and neighbors without symptoms who were
considered to be exposed to the risk of infection (n Å 140). The
latter group had neither symptoms nor blood parasites by direct
examination of Giemsa-stained thick smears (100 high-power
fields before declaring a slide negative). Nevertheless, chloroquine
and primaquine were administered orally to both groups [2]. The
BMH has maintained a rigid and active surveillance of the area.
In the first year, thick blood smears were prepared and examined
three times to ensure the absence of parasites from all previously
symptomatic persons and from their prophylactically treated relatives and neighbors. All persons complaining of fever living in or
around the areas of the outbreak had their blood examined during
the 2 years after the interrupted transmission. In addition, indoor
insecticides were used twice a year for the first 2 years. No further
malaria transmission nor relapses have been reported in the area
since 1988.
All subjects studied answered a questionnaire that included past
malaria experience, administered 8 months and 7 years after the
outbreak. Their answers revealed that all persons of our sample
were born and raised in the outbreak area and reported no travel
to regions in which malaria was endemic.
Sera samples. Sera were collected at two time points: 8 months
and 7 years after the malaria outbreak. At first, we worked in the
area with 31 subjects (median age, 19 years) who had had positive
thick blood smears and 18 of their relatives and/or neighbors (median age, 21 years) similarly treated despite negative thick blood
smears [2]. Venous blood was drawn in vacuum tubes, allowed to
clot for 3–5 h at room temperature, and centrifuged (400 g, 20
min); then serum was collected and stored at 0207C until use.
Seven years later, 17 of the 31 malaria patients and 9 of the 18
relatives and neighbors were relocated in the same area for the
present study. About 50% of the sample loss was due to difficulties
of access in the extensive rural area during the rainy season. Venous blood samples were drawn into heparinized tubes, and plasma
samples were separated as above. All persons studied at the two
time points were parasite-free at the time of blood collection. Their
thick blood smears, prepared by us, were negative. Control subjects
included 40 healthy adult volunteers who had never experienced
malaria nor been exposed to malaria transmission.
Antigens. The antigens used were two recombinant P. vivax
proteins expressed in Saccharomyces cerevisiae. The CS protein
(rPvCS) Bele´m isolate [8] was produced and provided by Chiron
(Emeryville, CA). This clone contains Ç70% of the entire CS
protein, including the repeats and parts of the flanking N- and Cterminal sequences. The major blood-stage surface antigen from
the Sal-1 P. vivax isolate (rPv200) includes amino acids 1357–
1729 of the carboxyl-terminal domains [9] (provided by David
Kaslow, NIH, Bethesda, MD).
Antibody measurement. The ELISA detected total IgG antibodies and was done as described [10]. The serum dilution used
was 1:40. For the recombinant antigens rPvCS and rPv200, the
final optical density (OD) at 405 nm was calculated by subtracting
the OD obtained with the yeast extract (antigen control). The
threshold of positivity (cutoff) was an OD of 0.1 based on the
mean plus 2 SD of the sera reactivity from the 40 healthy controls
(0.02 { 0.04 for rPvCS and 0.04 { 0.03 for rPv200).
Antibodies to rPvCS and rPv200. Anti-rPvCS IgG antibodies were detected in 45% of the subjects who had had malarial
symptoms 8 months before (table 1). Of their prophylactically
treated relatives and/or neighbors, 39% were also positive for
anti-rPvCS IgG. These values are similar for both groups (P
ú .05). Seven years later, anti-rPvCS antibodies were still
detectable in Ç20% of both groups.
Anti-rPv200 antibodies were highly prevalent (61%) in persons who had had malarial symptoms 8 months before; however, among the prophylactically treated subjects, 28% were
anti-rPv200 – positive, a significantly lower result (P õ .05)
(table 1). Seven years later, anti-rPv200 antibodies persisted
in 47% (8/17) of those who had had symptomatic malaria. All
prophylactically treated subjects were negative at this time
(table 1). None of the 40 controls had antibodies to the antigens
Decrease in levels of malarial antibodies with time. Tests
to measure antibody levels to P. vivax antigens were done in
parallel in all samples from the two different time points (figure
1). The mean OD of the sera from all 26 subjects retested 7
years later decreased with time. Some subjects who had had
clinical malaria still remained clearly positive for antibodies
against both antigens tested. Two prophylactically treated sub-
Table 1. Prevalence of anti-rPvCS and anti-rPv200 antibodies in persons exposed during single malaria outbreak, 8 months and 7 years
previously, in an area in which malaria was not endemic and specifically treated regardless of presence of symptoms.
8 months
7 years
With symptoms
No symptoms
x2; P
With symptoms
No symptoms
x2; P
14/31 (45)
19/31 (61)
7/18 (39)
5/18 (28)
.02; P Å .89
3.86; P Å .04
4/17 (24)
8/17 (47)
2/9 (22)
.01; P Å 1.0
4.11; P Å .02
NOTE. rPvCS, recombinant protein from sporozoite; rPv200, recombinant protein from blood stage. Data are no. ELISA-positive/total (%). Control subjects
(n Å 40) living in Belo Horizonte, outside area in which malaria was endemic, were all negative. Statistical analysis compared groups with and without symptoms
at each time point.
* Of those, only 2 had been studied for antibodies at 8 months after malaria outbreak.
/ 9d43$$ap48
03-04-98 19:03:33
UC: J Infect
Concise Communications
JID 1998;177 (April)
Figure 1. Levels of antibodies to P. vivax antigens from sporozoites (rPvCS) (A) or blood-stage
forms (rPv200) (B) measured by optical density
(absorbance at 405 nm) by ELISA with sera from
persons exposed to P. vivax outbreak, 8 months
and 7 years before, who had or did not have malaria symptoms.
jects, initially positive, were negative for blood-stage antibodies 7 years later.
A previously characterized P. vivax outbreak outside of the
Amazon region, in which malaria is endemic, offered an excellent opportunity to explore whether a malarial humoral immune
response persists in the absence of reinfection and reexposure.
To determine whether immunologic memory lasts is an important criterion for understanding immunity to an organism
and for planning future vaccine intervention. To address this,
we restudied subjects who had been studied before, after they
had had their first, brief, and only exposure to malaria [2].
/ 9d43$$ap48
03-04-98 19:03:33
Although malaria transmission lasted only 50 days in the
outbreak area, the levels and prevalence rates of anti-CS IgG
(Ç40%) were considered high 8 months later. The results presented here confirm previous data [2]. Seven years after the
transmission, anti-CS antibodies persisted in Ç20% of the subjects previously exposed to transmission, regardless of malaria
Kinetic studies of antibodies to recombinant P. vivax CS
protein in soldiers with limited exposure to malaria transmission in a high-risk area (Southeast Asia) demonstrated decreased levels with time [11]. The authors have also estimated
a half-life of Ç1 month for those antibodies when using a
protein derived from the central repeated region of the P. vivax
CS protein. These data contrast with our results, most likely
because we used a P. vivax CS protein that corresponds to 70%
UC: J Infect
JID 1998;177 (April)
Concise Communications
of the entire CS, including the repeats and parts of N- and Cterminal sequences. Moreover, long-term studies of antibodies
against whole sporozoites after malaria patients had departed
from African areas in which malaria is hyperendemic [6] also
show that these antibodies persist for 10 years in the absence
of reinfection.
The presence of anti-CS antibodies in naturally exposed populations is believed to reflect the rates of exposure to malaria
[12]. However, they are not ideal for measuring recent exposure, since they may persist in the absence of reinfections.
The prevalence of anti-rPv200 antibodies was high and persisted among the persons who had had malarial parasites and
clinical symptoms. They were positive 8 months (61%) and
7 years (47%) after the transmission. However, among the
asymptomatic prophylactically treated subjects in the area,
none were positive 7 years later, although 28% had been positive at 8 months. We propose that these prophylactically treated
subjects who had antibodies against rPv200 contracted the infection but were treated before parasites were detectable. The
possibility of a subpatent asymptomatic malaria among the
subjects with persistent anti-rPv200 antibodies seems unlikely,
since no new malaria cases or relapses were observed in the
area by active and careful search through the years by the local
health service.
It seems that an active symptomatic infection is required for
the long-term maintenance of blood-stage antibodies. Since P.
vivax is less pathogenic, persons may have been infected for
longer periods before diagnosis was established. The elapsed
time between acute malaria and treatment might have been
longer for such positive persons, thus contributing to the persistence of blood-stage antibodies. The absence of anti-rPv200
antibodies among the asymptomatic subjects probably results
from the effect of early prophylactic treatment decreasing the
circulating blood-stage antigens. However, it is not possible to
correlate the persistence of antibodies against rPv200 to higher
levels of parasitemia at the time of diagnosis, because parasitemia was only qualitatively evaluated at that time.
B cell memory seems to depend on the persistence of stimulating antigen [13] maintained over long periods by immune
complexes on the membrane of follicular dendritic cells in
lymphoid tissues [14]. It is also possible that B cell memory
is maintained by exposure to cross-reactive epitopes present in
other organisms to which the person is normally exposed. If
antigenic persistence is responsible for B cell memory to P.
vivax antigens, then live persisting vectors carrying critical
epitopes may be required for long-term vaccine-induced immunity.
In conclusion, the recombinant proteins, rPvCS and rPv200,
include B cell epitopes recognized by a significant proportion
/ 9d43$$ap48
03-04-98 19:03:33
of subjects who had been exposed to a brief P. vivax malaria
outbreak 7 years ago. Whether the persistent anti-malarial antibodies in such subjects will be sufficient to protect them in
the case of a new episode of malaria transmission remains
We thank Lu´cia Fraga (Universidade Vale Rio Doce) for providing facilities to work in the area, Jessica Kissinger for English
corrections, and Luzia Carvalho for suggestions and helpful discussions.
1. Cruz Marques A. Human migration and the spread of malaria in Brazil.
Parasitol Today 1987; 3:166 – 70.
2. Fontes CJF, Bathurst I, Krettli AU. Plasmodium vivax sporozoite antibodies in individuals exposed during a single malaria outbreak in a nonendemic area. Am J Trop Med Hyg 1991; 44:28 – 33.
3. Nardin EH, Nussenzweig RS, McGregor IA, Bryan JH. Antibodies to
sporozoites: their frequent occurrence in individuals living in an area
of hyperendemic malaria. Science 1979; 206:597 – 9.
4. McGregor IA. Malarial Immunity: current trends and prospects. Ann Trop
Med Parasitol 1987; 81:647 – 56.
5. Druilhe P, Pe´rigon JL. Mechanisms of defense against P. falciparum asexual blood stages in humans. Immunol Lett 1994; 41:115 – 20.
6. Druilhe P, Pradier O, Marc JP, Miltgen F, Mazier D, Parent G. Levels of
antibodies to Plasmodium falciparum sporozoite surface antigens reflect
malaria transmission rates and are persistent in the absence of reinfection. Infect Immun 1986; 53:393 – 7.
7. Chougnet C, Deloron P, Savel J. Persistence of cellular and humoral
response to synthetic peptides from defined Plasmodium falciparum
antigens. Ann Trop Med Parasitol 1991; 85:357 – 63.
8. Barr PJ, Gibson HL, Enea V, Arnot DE, Hollingdale MR, Nussenzweig
V. Expression in yeast of Plasmodium vivax antigen of potential use in
a human malaria vaccine. J Exp Med 1987; 165:1160 – 71.
9. Gibson HL, Tucker JE, Kaslow DC, et al. Structure and expression of the
gene for Pv200, a major blood-stage surface antigen of Plasmodium
vivax. Mol Biochem Parasitol 1992; 50:325 – 34.
10. Carvalho LH, Fontes CJF, Fernandes AAM, Marinuzzi HC, Krettli AU.
Cross-reactive cellular immune response to the circumsporozoite proteins of Plasmodium vivax and P. falciparum in malaria-exposed individuals. Parasite Immunol 1997; 19:47 – 59.
11. Brown AE, Webster K, Krinchai K, et al. Characterization of natural
antibody responses to the circumsporozoite protein of Plasmodium vivax. Am J Trop Med Hyg 1991; 44:21 – 7.
12. Webster HK, Grincrich JB, Wongsrichanalai C, et al. Circumsporozoite
antibody as a serologic marker of Plasmodium falciparum transmission.
Am J Trop Med Hyg 1992; 47:489 – 97.
13. Gray D, Skarvall H. B-cell memory is short lived in the absence of antigen.
Nature 1988; 336:70 – 3.
14. Bachmann MF, Ku¨ndig TM, Hengartner H, Zinkernagel RM. Regulation
of IgG antibody titers by the amount persisting of immune-complexed
antigen. Eur J Immunol 1994; 24:2567 – 9.
UC: J Infect