Document 993

Biologic Properties of a Bispecific Single-Chain Antibody
Directed Against 17-1A (EpCAM) and CD3
Tumor Cell-Dependent T Cell Stimulation and Cytotoxic Activity’
Matthias Mack, Rudi Cruber, Sabine Schmidt, Cert Riethmiiller,2 and Peter Kufer
ispecific Abs have been used to redirect T lymphocytes
against defined Ags on tumor cells (1-4). This approach
relies on binding of one arm of the bispecific Ab to a
tumor-associated Ag, while the other arm, by recognizing the CD3
rnolecule on T cells, triggers their cytotoxic activity (5-6). Crosslinkage of the TCR-associated CD? molecule by Abs immobilized
to solid surfaces or cell membranes has been found to be a particularly strong activation signal.
First clinical trials with conventionally generated bispecific Abs
(F(ab‘), fragments from hybrid-hybridomas), however, have revealed considerable toxic effects. While these Abs preincubated
with lymphocytes in vitro and injected directly into or near the
tumor have shown little side effects (7-9), their systemic administration resulted in activation of T cells with subsequent release of
cytokines such as TNF-a and IFN-y (10, 11). The systemic release
of these cytokines has been made responsible for the dose-dependent and dose-limiting toxicity (fever, chills, and dyspnea) of current bispecific Ab preparations. The rapid onset of symptoms, accompanied by high plasma levels of cytokines and negligible
toxicity against normal tissue, also expressing the detected tumorassociated Ag, points to an activation of T cells independent from
the target cells. The target cell-independent activation of T cells
may be attributed to contaminations of bispecific F(ab’), fragInstitute of Immunology, Munich, Germany
Recewed for publlcation August 8, 1996. Accepted for publicatlon January
14, 1997.
The costs of publication of this article were defrayed in part by the payment of
pagecharges. This article must therefore be hereby marked advertisement in
accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
I This work was supported by grants irom Deutsche Krebshilfe, Bonn, Genzen
trum Munchen (BMFT), and Curt-Bohnewand-Fond.
Address correspondence and reprint requests to Dr. Gert RiethmOller, Institute
of Immunology, Goethestrasse ? l , 80?36 Munchen, Germany.
Copyright 0 1997 by The American Association of lrnmunologlsts
ments with intact Abs present after incomplete proteolytic digestion. This gives rise to binding of CD3 Abs to the surface of FcRpositive cells, and thus entails the activation of T cells. A study
investigating different preparations of OKT3-F(ab’), fragments
showed a marked reduction in T cell activation and cytokine release with meticulously purified OKT3-F(abf), fragments, and
therefore strongly supports the application of bispecific Abs that
are free of contaminating Fc parts ( 1 2).
Recently, we have described the construction and expression of
bispecific single-chain Abs in eukaryotic cells (1 3). The design of
this molecules is shown in Figure 1. Four variable domains of two
different Abs are joined by flexible linkers to appear on one single
One Ag
protein chain in the order V,(A)-V,(A)-V,(B)-V,(B).
binding site of this construct is directed against 17-1A; the other
against the CD3 surface molecule on T cells. The construct could
be expressed in Chinese hamster ovary cells as fully functional
molecule in large amounts, and was easy to purify via its C-term i d histidine tail. In contrast to bispecific Abs produced with the
hybrid-hybridoma technology (14, IS) or chemical cross-linkage.
the bispecific single-chain Ab preparation is devoid of any Fc parts
or anti-CD3 homodimers probably responsible for the severe side
effects seen after i.v. application in cancer patients.
In this study, we demonstrate that the bispecific single-chain Ab
does not induce an activation of PBMCs in the absence of target
cells, whereas addition of target cells gives rise to a strong T cell
activation measured by surface activation markers and cytokine
release. This observation is consistent with plenty of data showing
that monovalent nonimmobilized CD3 Abs do not stimulate T cells
(16-1 8). Therefore, bispecific Abs produced in the described single-chain approach should be ideal candidates for a systemic administration in cancer patients. The systemicadministration clearly
is a key element of Ab-mediated tumor therapy, as a large primary
or metastatic tumor mass is not a suitable target in contrast to
Downloaded from on May 10, 2010
Anti-CD3 X anti-tumor-bispecific Abs have been usedto redirect cytotoxicT cells to tumor cells in an MHC-unrestricted fashion
and to induce their rejection i n vivo. We have recently described a recombinant bispecific single-chain Ab that combines four
different V regions of two Abs, anti-7 7-1A and anti-CD3, on one polypeptide chain. It folds correctly to a 60-kDa globular
protein and is secreted in fully functional form by a high producer Chinese hamster ovary cell line. I n this work, we report that
its remarkable cytotoxicity against 17-1A+ tumor cells i s exerted via T cells without an apparent engagement of a detectable
costimulatory pathway. T cells are activated only by the bispecific Ab when coincubated with 17-1A+ target cells. In a chromium release assay, CD8+ T cells reach maximal tumor cell cytotoxicity within 4 h, while CD4+ T cells need about 20 h to reach
similar levels of cytotoxicity. Addition of costimulatory CD28 Abs did notlead to a further increase in cytotoxicity. Its remarkable stabilityat 37O in serum, the ease of production, and purification by affinity chromatography via polyhistidine tail make this
smaller version of a bispecific Ab a promising candidate for a therapeutic trial in patients with solid tumor. Because adjuvant
therapy with an intact, much less cytotoxic lgG2a Ab against the 17-1A target had already increased the 7-yr survival of
colorectal cancer patients by 30%, the presented small bispecific construct lacking the immunogenic murine Fc region as well
as autochthonous-T lymphocyte Stimulatory activity warrants a therapeutic trial in patients with minimal residual 17-1A+
cancer. The Journal of Immunology, 1997, 158: 3965-3970.
Transfectants were cultured in RPMI 1640 (Sigma Chemical Co., St.
Louis, MO) with 10%FCS (Life Technologies), 2 mM L-glutamine, and 1
mg/ml (3418. Several clones were screened for expression of 17-IA by
FACS analysis. The clone used in the experiments expressed 17-1A to a
lesser degree than Kat0 cells.
Kat0 is a 17-1A-positive human gastric cancer cell line and was obtained from American Type Culture Collection (Rockville, MD).
Cytotoxicity assay
FIGURE 1. Scheme of the17-1NCU3-bispecificsingle-chain
Two single-chain Fv fragments, one directed against 17-1A, the other
against CD3, were joined by a flexible linker consisting of five amino
for easy detection, and a
acids. An N-terminal flag-tag was added
C-terminal histidine tail (His) for protein purification.
Materials and Methods
Production o f the bispecific single-chain Ab
The bispecific 17-1A/CD3 single-chain Ab was produced as described previously (13). Briefly, two single-chain Fv fragments were joined through a
flexible Gly-Ser linker. The single-chain Fv fragment directed against
17-1A is derived from the M79 hybridoma (22); the other directed against
CD3 is derived from the TR66 hybridoma (23). The construct was expressed in Chinese hamster ovary cells as fully functional molecule and
was purified via its C-terminal histidine tail on a Ni-NTA (Nitrilotriacetic
acid) afinity column (Qiagen, Germany).
Preparation of effectorcells
Human PBMCs were isolated from the fresh buffy coat of healthy donors
by Ficoll density gradient centrifugation (Pharmacia, Piscataway, NJ) and
a subsequent 100 X g centrifugation step to remove thrombocytes. Where
indicated, CD4+ or CDSt cells were depleted with magnetic beads (Dynabeads; Dynal, Great Neck, NY), according to manufacturer's instructions.
The depleted cellpopulations were analyzed by flow cytometry (FACScan;
Becton Dickinson, Mountain View, CA), which showed a 99% depletion of
CD8+ or CD4' cells. For the preparation of purified CD4+ or CD8+ T
cells, monocytes were first depleted by adherence to plastic for 3 h, and
then magnetic beads (Dynabeads) were used for positive selection ofCD4+
or CDS+ T cells,according to the procedure described by Halvorsen et al.
(24). Analysis by flow cytometry showed a purity of 98% for CD4+ Tcells
and 98% for CD8+ cells (91% CD8+ T cells and 7% CD8+ NK cells).
Exposition of the bsc Ab to PBS and serum
The bsc Ab was stored for 6 mo in PBS at 4°Cand compared with a freshly
prepared sample of the protein. To detect thestability in human serum, 20
pl of the bsc Ab and 1 0 0 pl of freshly prepared serum from a healthy donor
were incubated at 37°C for different periods of time. Experiments have
been repeated two times with serum of different donors.
Flow cytometric analysis and cytokine ELlSA
A cytotoxicity assay was performed similar to the 5'Cr release assay described above. Effector cells (150,000) were added to each well of a Ratbottom microtiter plate and incubated overnight at 37°C. Further additions
were either PBS only, target cells only, bsc Ab only, OKT3 only, PHA
only, or both target cells and bsc Ab. Target cells were added in an E:T
ratio of 1:1 for good cell-cell contact, the bscAb was used in different
concentrations, and OKT3 mAb and PHA in final concentrations of 20
ng/ml and 2 pg/ml, respectively. Three identical wells were prepared for
each test. After an incubation time of 4 h and 20 h,the supernatant and the
cells were collected separately. The supernatants were frozen immediately
at -70°C and stored for 24 h at most. The cells were incubated for 30 min
at 4°C with different Ab mixtures in the proper dilution. One Ab mixture
CD25 phycoerythrin (Becton
was CD3 FlTC (Coulter Corp., Hialeah, a),
Dickinson), and CD4 Tricolor (Medac, Hamburg, Germany); one was CD3
FITC (Coulter Corp.), CD25 phycoerythrin (Becton Dickinson), and CD8
Tricolor (Medac); andanother the appropriate isotype controls. After washing, the samples were analyzed byflow cytometry (Becton Dickinson).
Quantitative analysis of CD25 and CD3 expression was conducted by gating the lymphocytes by virtue of their light scatter properties, with careful
inclusion of activated larger lymphoblasts. For analysis of the CD3 and
CD25 expression in the CD4' and CD8' subset, a second gate was set on
the CD4+ andCD8' cells, respectively, and the mean channel of CD3 and
CD25 expression was determined. Experiments have been repeated at least
four times with effector cells from different donors, and one representative
experiment is shown.
Cell lines
Murine X63 myeloma cells were stable transfected with 17-1A cDNA
using avector with the @-actinpromotor and the neomycin resistance gene.
Abbreviations used in this paper: bsc Ab, 17-1NCD3-bispecific single-chain
antibody; aa, amino acid.
Downloaded from on May 10, 2010
disseminated tumor cells that are lodging frequently in interstitial
tissue, easily accessible to macromolecules and effector cells (19).
For an in vivo application, the presence of sufficient numbers of
effector cells certainly is a critical point. Therefore, we investigated whether subpopulations of PBMCs, such as purified CD4+
and CD8+ T cells, which do not contain accessory cells in significant numbers, can be redirected efficiently against different tumor
cell lines with the bispecific Ab alone. The infiuence of a costimulation of T cells with CD28 Abs is also investigated. Furthermore,
an in vivo application of bispecific single-chain Abs would lead to
a prolonged exposure of these moIecules to human plasma and to
long term interactions between target cells and redirected T cells.
Therefore, we investigated the stability of the construct, when exposed to fresh human serum, and also performed long term assays
with different effector cell populations and tumor cell lines.
A recent clinical trial has shown that patients with a surgically
completely resected colorectal carcinoma havea significant benefit
concerning the overall survival and the frequency of distant metastasis, when treated i.v. with a mAb directed against 17-1A (20),
now known as epithelial cell adhesion molecule (21). The use of
bispecific single-chain Abs, which are severalfold more effective
than mAbs (2), would be an important improvement of this
For the "Cr release assay, different numbers of effector cells were added
in a volume of 1 0 0 4 RPMI medium (Sigma Chemical Co.) -t 10% FCS
(Life Technologies) to each well of a flat-bottom microtiter plate (Costar,
Cambridge, MA) and incubated overnight at 37°C. Target cells were labeled for 2 h with 5'Cr, and 25,000 of them were added to each well in a
) ~ were
volume of 100 4 . Different concentrations of the bispecific ( b ~ c Ab
added in a volume of 25 pl together with an extra solution of 25 pl,
containing either medium or CD28 Ab (mAb 9.3) in different concentrations or an Ab mixture for inhibition of cytokines. This later Ab mixture
consisted of TNF-a, TNF-p, and IFN-y Abs in final concentrations of 20,
IO, and 2.5 pg/ml, respectively. The incubation time was 4 or 20 h at37°C.
Maximal release was determined by lysis of target cells with lysis buffer
(2% SDS, 0.37% EDTA, and 0.53% Na,CO,), and spontaneous 5'Cr release was determined with target cells only. The 17-IA-negative parental
X63 cells were completely refactory to lysis in 5'Cr release assays (13).
Unlinked sc-Fv fragments of the 17-IA and CD3 Ab were ineffective. As
a further control,we used a mutated derivative of the bsc Ab, in which the
17-1A binding was abolished after site-directed mutagenesis of aa,,, in V,
and aa, in V, of the anti-17-1A part. This constmct proved to be completely inactive in the lytic assays. Incubation of target cells only with bsc
Ab at the highest concentration did not result in measurable lysis. Specific
lysis was calculated as: (cpm, experimental release - cpm spontaneous
release)/(cpm, maximal release - cpm, spontaneous release). Triplicate
samples were measured in each assay, and SD was always less than 6% of
the mean. Experiments were repeated at least four times with effector cells
from different donors, and one representative experiment is shown.
The Journal of Immunology
FIGURE 3. IL-2R up-regulation and CD3 modulation induced by the
bispecific Ab on T cells in contact with target cells. The same experiment as shown in Figure 2 was performed with equal lettering of the
x-axis. The expression of CD25 (IL-2R) and CD3 was measured by
FACS analysis after 20 h, and is shown on the y-axes in the diagrams.
Activation of T cells by bispecific single-chain Abs
F I GU R E 2. Target cellinduced release ofcytokinesfrom
PBMCs (150,000) were incubated for4 h and 20 h with medium alone
(E), with target cells (150,000 Kato) only (E+T), with the bispecific
single-chain Ab at 3 yghl or 61 n g h l only (E+bsc-Ab), with target
cells and the bispecific single-chain Ab at 3 pg/ml or 61 n d m l or 1.2
ng/ml (E+T+bsc-Ab), with PHA at 1 pCLg/rnl only sPHA, or with a CD3
rnAb at 20 ng/ml only (antiLCD3 mAb).
ELISA for TNF-a, IFN-y, and IL-6
All ELISAs were obtained commercially (GenzymeCorp.,Cambridge,
MA) and utilized according to manufacturer's instructions. The frozen supernatants from two identical wells of each experiment were analyzed separately, and mean values were calculated. ELISA experiments were performed twice with PBMCs from different donors.
3H-/abeled thymidine uptake
PBMCs were added in triplicates to a 96-well U-bottom plate at a concentration of 500,000 cells/well. Bsc Ab and anti-CD3 mAb were added in
different concentrations to a final volume of 250 pVwell. Addition of medium alone served as negative control. After an incubation of 48 h at 3 7 T ,
'H-labeled thymidine was added at a final concentration of 10 pCi/ml for
another 18-h incubation. Cells were harvested on a semiautomated cell
harvester, and incorporation of 'H-labeled thymidine was determined on a
scintillation counter (Beckmann, Fullerton, CA). The stimulation index
was calculated as log (counts (experimenta1)kounts (negative control)).
Experiments were repeated three times with PBMCs from different donors.
According to the therapeutic intention, T cells should only be activated when they are attached to tumor cells via the Ab bridge, but
not when they are exposed to bispecific Abs alone in the peripheral
blood or lymphoid tissue. To test the activation of T cells by bispecific Abs, we performed short and long term assays (4 h, 20 h) in
the absence or presence of target cells, and used PBMCs as effector
cells to imitate the in vivo condition. We explored early events in
T cell activation for both the CD4+ and CD8' T cell subset, and
also studied the proliferation of PBMCs by a 'H-labeled thymidine
uptake test. Early events in T cell activation are the expression of
IL-2R (CD25), release of IFN-y, TNF-a, and IL-6, as well as
down-regulation of CD3. As shown in Figures 2 and 3, we could
demonstrate that incubation of PBMCs with the bispecific Ab
alone did lead neither to the expression of CD25, measurable release of cytokines, nor down-regulation of CD3. Only after addition of target cells (Kato) did T cells become activated to a similar
degree as when stimulated with PHA or whole anti-CD3 Abs. Activation could be induced in both CD4+ and CD8+ subsets. In the
presence of the murine cell line X63-17-1A, transfected with
17-1A cDNA, activation of T cells was diminished significantly
(Fig. 4). We also used immobilized recombinant 17-1A coated to
the culture well bottom as target structure for PBMCs, and found
that under these conditions the bispecific single-chain Ab activated
CD4+ or CD8+T cells to about the same level as the murine target
cells (data not shown). With purified CD4' and CD8' T cells, we
found in all experiments similar results as with PBMCs concerning
the expression of CD25 and CD3 (data not shown).
Even after prolonged incubation, the bispecific single-chain Ab,
in concentrations up to 3 pg/ml, failed to induce proliferation of
PBMCs, as measured by incorporation of 3H-labeled thymidine,
while an intact CD3 mAb induced a strong proliferation with a
stimulation index > I O 0 (data not shown).
Downloaded from on May 10, 2010
- 450
8 200
Retargeting ofPBMCs against17-1A-transfectedmouse
tumor cells (X63-17-IA).Activation of CD4+ and CD8+ T cells, measured by IL-2R up-regulation, was determined with and without addition of CD28 Abs (1 Fglml).
The lytic capacity of CD4+ and CD8+ T cells
As bispecific (anti-tumor/anti-CD3) Abs can target both CD4+ and
CD8+ T cells to tumor cells, resulting in a strong activation ofboth
T cell subsets, we investigated whether this activation also led to
the lysis of target cells. A 'lCr release cytotoxicity assay was
performed with CD8-depleted PBMCs (containing 100,000 CD4+
T cells), CD4-depleted PBMCs (containing 100,000 CD8+ T
cells), and nondepleted PBMCs (containing 66,000 CD4+ and
33,000 CD8+ T cells). It is noteworthy that monocytes, B cells,
and NK cells were still present in these negatively depleted effector
cell preparations. While in the 4-h assay (Fig. 5, top) CD8-depleted PBMCs did not cause significant lysis of target cells, the
CD4-depleted cell preparations or nondepleted PBMCs showed a
clearly measurable specific lysis already after 4 h. However, after
an incubation of 20 h (Fig. 5, bottom), the CD8-depleted PBMCs
clearly induced a significant specific lysis of almost 70%.
Addition of Fas Abs (clone ZB4; Immunotech, Marseille,
France), inhibiting the human Fas on Kato cells, caused a reduction of lysis (Fig. 6) mediated by PBMCs, while there was no effect
on lysis of murine X63-17-1A cells not recognized by the Fas Ab,
precluding unspecific effects of the Fas Ab on the function of effector cells. Addition of an Ab mixture, consisting of TNF-aP and
IFN-y, did not reduce the rate of specific lysis (data not shown).
To assess the effect of accessory cells such as monocytes in the
effector cell population, we compared the lytic activity of purified,
positively selected CD4" and CD8+ T cells with the unselected
PBMCs. As shown in Figure 7, the bispecific Ab alone suffices to
induce cytotoxic activity of purified CD8+ and CD4+ T cells to
lyse Kat0 tumor cells as well as X63-17-1A transfectants. In comparison with the human cell line Kato, the specific lysis of the
murine cell line X63-17-1A was diminished significantly, especially if only CD4+ T cells were used.
Costimulatory signals delivered via CD28 do not enhance
the lytic capacity of PBMCs, CD4+, or CD8+ T cells
redirected against different tumor cells
Many authors have shown that T cells, redirected against tumor
cells with bispecific anti-tumodanti-CD3 Abs, can be costimulated
effectively with CD28 Abs to increase the cytolytic capacity of T
cells (4,25-28). We.investigated whether addition of CD28 Abs in
bsc-Ab concentration (nglml)
FIGURE 5. The lvsis of human Kat0 tumor
different effector ceil populations in a 5'Cr release. EitherPBMCs(Containing 66,000 CD4+ T cells and 33,000 CD8+ T cells, white bars), or
PBMCs depleted of CD8+ cells (containing 100,000 CD4+ T cells,
black bars), or PBMCs depleted of CD4+ cells (containing 100,000
CD8+ T cells, gray bars) were used. Incubation times:4 h in the upper
panel, and 2o i n the lower panel,
bsc-Ab concentration (ng/ml)
FIGURE 6. Partial inhibition of lysis by FasAbs. A 20-h 5'Cr release
assay was performed with 500,000 PBMCs redirected against 25,000
Kat0 cells with the bispecific single-chain Ab. Fas Abs were used in a
concentration of 1 pg/rnl. SD was consistently below5% of the mean.
concentrations of 5 pg/ml and 30 ng/ml to the effector cell population could increase the lysis of target cells. As shown in Figure
8, CD28 Abs do not increase the lytic capacity of redirected purified CD4+ and CD8+ T cells. The same was seen with PBMCs
(data not shown). In parallel, we investigated the activation of T
cells by monitoring the expression of CD25. Addition of CD28
Abs to Kat0 cells increased the activation of retargeted CD4+ and
CD8+ T cells only slightly (about 2%) (data not shown). Addition
of CD28 Abs to the murine X63-17-1A cells, however, induced a
significantincrease in the activation of the retargeted CD4+ and
CD8+ T cells almost to the same level as seen with Kat0cells (Fig. 4).
With immobilized 17-1A as target structure for redirected T
cells, results were comparable with the murine target cells X6317-1A (data not shown).
Downloaded from on May 10, 2010
These data show that under all tested conditions, the bispecific
single-chain Ab does not activate CD3+ T cells in the absence of
17-1Af target cells, even when abundant FcR-carrying accessory
cells are present and high concentration of bispecific Ab constructs
are used.
The Journal of Immunology
bsc-Ab concentration (nglrnl)
Kat0 cells
(solid lines) and murine X63-17-1A cells (broken lines), in a 20-h *'Cr
release assay. The effector cell population was either 500,000 PBMCs
(containing 200,000 CD4+ and 100,000 CD8+ T cells), orpurified
CD4+ and CD8+ T cells (200,000 each), asindicated with circles,
triangles, and squares, respectively.
FIGURE 7. Comparison of twodifferenttargets,human
Flow cytometry has shown that the tumor cell line Kat0 does not
express B7-1 or B7-2 (data not shown).
The stability of the bispecific single-chain Abs
Bispecific Abs that redirect T lymphocytes to tumor cells have
been found distinctly more efficient in target cell killing or elimination than intact unmodified Abs of the same specificity (2).
However, unlike intact Igs, bispecific Abs endowed with one of
their both specificities for the TCR-associated CD3 molecule displayed severe clinical toxicity ascribed to CD3-triggered systemic
cytokine release (10, 11). As this remarkable toxicity is most likely
caused by Fc fragments contaminating the conventionally produced bispecific Ab preparations, we have resorted to rDNA-based
protein engineering to generate bispecific single-chain Abs devoid
of any Fc fragments. We isolated the V, and V, genes of the Ab
M79 directed against the 17-1A Ag for onearm, while for theother
arm the gene of a single-chain Fv fragment of a CD3 Ab was
selected. The choice of a 17-1A Ab was warranted, since an unmodified murine mAb against the 17-1A epithelial Ag had been
found previously to reduce the5-yr (20) andalso the 7-yrmortality
(29) of patients with minimal residual disease. Furthermore, since
minimal residual cancer is intended as therapeutic indication in patients that are virtually tumor free after curative regional surgery, the
avoidance of toxicity appearsto be essential for any form of therapy.
The data presented in this work show that the bispecific 17"
CD3 single-chain Ab does not activate T lymphocytes in the absence of tumor targets. In addition, in the presence of potent costimulatory cells that are abundant in preparation of PBMCs, T
cells were not activated by the addition of bispecific Abs alone.
Only when 17-1A-positive tumor cells were introduced into the
system were T cells induced to express the IL-2R and to secrete
cytokines, such as TNF-a, IFN-y, and IL-6. In addition, the murine
X63 plasmocytoma cells transfected with the 17-1A target Ag could
trigger activation of T cells, albeit to a distinctly lower degree.
20 h
FIGURE 8. Effectof a costimulation with CD28 Abs on the lytic activity
of T cells redirected with the bispecific single-chain Ab. A 5'Cr release
assay was performed with purifiedCD4+ or CD8+T cells (200,000each),
of 4 or 20 h, and different
Kat0 orX63-I 7-1 A target cells, incubation times
concentrations of CD28 Abs. In each experiment, specific lysis was determined with five different concentrations
of the bsc Ab (as in Fig. 5), and
mean values werecalculated. Relative lytic activityin percentageof mean
values obtained in experiments without CD28 Abs.
As T lymphocytes require a costimulatory signal for activation,
the question was whether the activation observed in this study via
bispecific Abs bound to tumor cells depended on the main costimulatory pathway CD28&37. On human Kat0 cells, we could
neither detect B7-1/2 by sensitive FACS analysis, nor did the addition of stimulatory CD28 Ab result in a measurable increased
activation by bispecific Abs. In the presence of 17-1A-transfected
murine target cells (X63-17-1A), however, CD28 Abs costimulated T cells significantly. This difference between the human and
murine target cell may be partially due to a lower concentration of
17-1A onX63-17-1A cells, as well as to other as yet unrecognized
costimulatory molecules on the human cell not present or fully
active on the xenogenic murine target.
As both major subsets of T cells, Le., CD4+ and CD8+ cells,
were equally well stimulated when bound to the targets via the
bispecific bridge, it was of interest to see whether activation correlated with cytotoxicity. While purified CD4+ and CD8+ T cells
induced a similar degree of cytotoxicity of Kat0 targets in a 20-h
51Cr release assay, in a 4-h assay only CD8+ cells proved to be
cytotoxic. This difference may be explained by the preformed lytic
molecules of CD8+ cells that are readily released from storage
granules by an activation signal, whereas CD4+ T cells acquire the
cytotoxic phenotype only after prolonged activation (30). The significantly lower T cell activation induced by the murine target cells
correlated well to the much smaller lysis of these cells. Consistently, murine X63-17-1A cells hardly induced lytic activity of
CD4' cells, also not after prolonged incubation. Although costimulatory CD28 Abs increased the activation of redirected T
cells, their cytotoxic capability was not increased (Fig. 8). Thus,
we suggest that activation parameters (CD25) do not necessarily
predict the cytotoxic activity of T cells.
As the Fas ligand can be induced by prolonged activation in
both CD8+ and CD4+ T cells (31), Fas-mediated apoptosis, the
cytotoxic pathway used predominantly by CD4 T cells (32, 33),
may contribute to the cytotoxicity observed after 20 h of incubation. The partial inhibition of tumor cell lysis (Fig. 6) by anti-Fas
Abs indicates that also this form of lytic mechanism can be mobilized by bispecific Abs. Interestingly, the most pronounced inhibition is seen at lower concentrations of the bsc Ab. As killing by
T cells is mediated by more than one mechanism, the inhibitory
effect of anti-Fas Abs may be seen only with moderately activated
T cells, Le., at lower bsc Ab concentrations (see also Fig. 3), while
Downloaded from on May 10, 2010
In general, bispecific single-chain Abs appeared rather labile, as
the four variable domains, although joined by flexible linkers, are
kept in the right quaternary structure only by noncovalent interactions between the light and heavy variable domains. We investigated the stability of the 17-1NCD3-bispecific Ab, when stored in
PBS at 4°C or in freshhuman serum at37°C for prolonged periods
of time. Storage inPBS at a concentration of 200 pg/ml over 6 mo
did not reduce the amount of active molecules nor the affinity of
the Ab significantly. Incubation of the bsc Ab in serum at 37°C
causes a slow decrease in biologic activity to 87% over 56 h, as
measured in a 51Cr release assay (data not shown).
We thank E. Kopp for protein purification and S . Dittrich for excellent
technical assistance. This study includes
work performed by M. Mack in
partial fulfillment of requirements for a medical thesis of the Faculty of
Medicine, University of Munich.
I. Titus, J. A,, M. A. Garrido, T. T. Hecht, D. F. WinMer, J. R. Wunderlich, and
D. M. Segal. 1987. Human T-cells targeted with antiGT3 cross-linked to antitumor
antibody prevent tumor growth in nude mice. J. Immunol. 138:4018.
2. Weiner, G. J., and J. R. Hillstrom. 1991. Bispecific anti-idiotype/anti-CD3 antibody therapy of murine B cell lymphoma. J. Immunol. 147:4035.
3. Demanet, C.. J. Brlssinck, 0. Leo, M. Moser, and K. Thielemans. 1994. Role of
T-cell subsets in the bispecitic antibody (anti-idiotype X anti-CD3) treatment of
the BCLl lymphoma. Cancer Res. 54:2Y73.
4. Renner, C., W. Jung, U. Sahin, R. Denfeld, C. Pohl, L. Trumper. F. Hartmann,
V. Diehl, R. van Lier. and M. Pfreundschuh. 1994. Cure of xenografted human
tumors by bispecific monoclonal antibodies and human T-cells. Science 264:833.
5 . Staerz, U. D.. 0. Kanagawa, and M. J. Bevan. 1985. Hybrid antibodies can target
sltes for attack by T-cells. Nature 314:628.
6. Perez, P , R. W. Hoffman, S. Shaw, J. A. Bluestone, and D. M. Segal. 1985.
Specific targeting of cytotoxic T-cells by antiLT3 linked to anti-target cell antibody. Narure 316,354.
7. Kroesen, B. J., A. ter Haar, H. Spakman, P. Willemse, D. T. Sleijfer, E. G. de
Vries, N. H. Mulder, H. H. Berendsen, P. C. Limburg. T. H. The, and L. de Leij.
1993. Local antitumor treatment in carcinoma patients with bispecific monoclonal antibody-redirected T-cells. Cancer Immunol. Immunother. 37:400.
8. Bolhuis, R. L.,C. H. Lamers, S. H. Goey, A. M. Eggermont, J. B. Trimbos,
G. Stoter, A. Lanzavecchia, E. di Re, S. Miotti, F. Raspagliesi, L. Rivoltini, and
M. 1. Colnaghi. 1992. Adoptive immunotherapy of ovarian carcinoma with bsmAb-txgeted lymphocytes: a multicenter study. lnt. J. Cuncer Suppl. 7:78.
9 Nitta, T., K. Sato, H. Yaglta, K. Okumura, and S. Ishii. 1990. Preliminary trial of
specific targeting therapy against malignant glioma. Lancer 335:368.
I O Tibben, J. G., 0. C. Boerman, R. A. Claessens, F. H. Corstens, M. van Deuren,
P. H. de Mulder, J. W. van der Meer, K. G. Keljser, and L. F. Massuger. 1993.
Cytokine release in an ovarian carcinoma patient following intravenous administration of bispecific antibody OCiTR F(ab'),. J. Narl. Cuncer Insr. 85:1003.
1 1 Kroesen, B. J., J. Buter, D. T. Sleijfer. R. A. Janssen. W. T. van der Graaf, T. H.
The, L. de Leij, and N. H. Mulder. 1994. Phase I study of intravenously applied
bispecific antibody in renal cell cancer patients receiving subcutaneous interleukin 2. Br. J. Cancer 70:652.
12. Woodle, E. S., J. R. Thistlethwaite, I. A. Ghohrial, L. K. Jolliffe, F. P. Stuart, and
J. A. Bluestone. 1991. OKT3 F(ab'), fragments: retention of the immunosuppressive properties of whole antibody with marked reduction in T-cell activation
and lymphokine release. Transpluntafion 52:354.
13. Mack, M., G. Riethmuller, and P. Kufer. 1995. A small bispecific antibody construct expressed as a functional single-chain molecule with high tumor cell cytotoxicity. Proc. Narl. Acad. Sci. USA Y2:702l.
14. Staerz, U. D., and M. J. Bevan. 1986. Hybrid-hybridoma producing a bispecific
monoclonal antibody that can focus effector T-cell activity. Proc. Narl. Acud. Sci.
USA 83:1453.
15. Lanzavecchia, A,, and D. Scheidegger. 1987. The use of hybrid-hybridomas to
target human cytotoxic T lymphocytes. Eur. J . Immunol. 17:105.
16. Ledbetter, A. J., C. H. June. P. J. Manin. C. E. Spooner, J. A. Hansen, and K. E.
Meler. 1986. Valency of CD3 binding and internalization of the CD3 cell-surface
complex control T-cell responses to second signals: distinction between effects on
protein kinase C, cytoplasmatic free calcium and prolifemtion.J. Immunol. 146:3945.
17. Qian, J. H., J. A. Titus, S. M. Andrew, D. Mezzanzanica. M. A. Garrido, I. R.
Wunderlich, and D. M. Segal. 1991. Human peripheral blood lymphocytes targeted with bispecific antibodies release cytokines that are essential for inhibiting
tumor growth. J. Immunol. 146:3250.
18. Schwab, R., M. K. Crow, C. Russo, and M. E. Weksler. 1985. Requirements for
T-cell activation by OKT3 monoclonal antibody: role of modulation ofT3 molecules
and interleukin I . J. Immunol. 135:1714.
19. Riethmuller, G., E. Schneider-Gadicke, and J. P. Johnson. 1993. Monoclonal
antibodies in cancer therapy. Curr. Opin. Immunol. 5:732.
20. Riethmuller, G., E. Schneider-Gadicke. G. Schlimok, W. Schmiegel, R. Raab, K.
Hotfken, R.Gruber, H. Pichelmaier, H. Hirche, R. Pichlmayr, P. Buggisch,
J. Witte, and the German Cancer Aid 17-IA Study Group. 1994. Randomized
trial of monoclonal antibody for adjuvant therapy of resected colorectal carcinoma. Lancet 343:1177.
21. Litvinov, S. V., M. P. Velders. H. A. M. Bakker, G.J. Fleuren, and S. 0.
Warnaar. 1994. Ep-CAM: a human epithelial antigen is a homophilic cell-cell
adhesion molecule. J. Cell Biol. 125:437.
22. Gottlinger, H. G., 1. Funke, J . P. Johnson, J. G. Gockel, and G. Riethmiiller. 1986.
The epithelial cell surface antigen 17-IA, a target for antibody-mediated tumor
therapy: its biochemical nature. tissue distribution and recognition by different
monoclonal antibodies. Irrt. J. Cuncer 38:47.
23. Traunecker, A,, A. Lanzavecchia, and K. Karjalainen. 1991. Bispecific single
chain molecules (Janusins) target cytotoxic lymphocytes on HIV infected cells.
EMBO J. l0:3655.
24. Halvorsen, R., G. Gaudemack, T. Leivestad, F. Vartdal, and E. Thorsby. 1987. Activation of resting, pure CD4''. and CD8+ cells viaCD3. S c a d . J. Immunol. 26:197.
25. Jung, G.. U. Freimann, Z . Von Marschall, R. A. Reisfeld, and W. Wilmanns.
1991. Target cell-induced T-cell activation with bi- and trispecitic antibody fragments. Eur. J. Immunol. 21:2431.
26. Jung, G . , J. A. Ledbetter, and H. J. Muller-Eberhdrd. 1987. Induction of cytotoxicity in resting human T lymphocytes bound to tumor cells by antibody heteroconjugates. P r w . Nutl. Acud. Sci. USA 84:461l.
27. Jung, G., D.E. Martin, and H. J. Muller-Eberhard. 1987. Induction of cytotoxicity
in human peripheral blood mononuclear cells by monoclonal antibody OKT3.
J . Immunol. 139:63Y.
28. Bohlen.H., T. Hopff. 0. Manzke, A. Engert, D. K u k , P. D. Wickramanayake,
V. Diehl, and H Tesch. 1993, Lysis of malignant B cells from patientswithBchronic lymphocytic leukemia by autologous T-cells activated with CD3 X CD19
bispecific antibodies in combination with bivalent CD28 antibodies. Blood 82:1803.
29. Riethmiiller, G., E. Holz, G. Schlimok, W. Schmiegel, R. Raab. K. Hoftken,
R. Gruber, H. Pichlmaier, H. Hirche. R. Pichlmayr, and G:A. S . Group. 1996.
Monoclonal antibody (mAb) adjuvant therapy of Dukes C colorectal carcinoma:
7-year update of a prospective randomized trial. In Proceedings offhe American
Socieh of Clinical Oncoloxy, Vol. 15. p. 444.
30. Tani, M., H. Tanimura, H. Yamaue, S. Mizobata, M. Iwahashi, T.Tsunoda,
K. Noguchi, T. Hotta, H. Terasawa. and K. Arii. 1995. Generafion of CD4+
cytotoxic T lymphocytes stimulated by immobilized anti-CD3 monoclonal antibody and interleukin-2 in cancer patients. h i . J. Cuncer 60:X02.
31 Suda, T. 0..Y. Takahiro. T. Naito, N. Yokota, S. Arai, K. Ozaki, K. Nakao, and
S. Nagata. 1995. Expression of the Fas ligand in cells of T-cell lineage. J. lrnmunol. 154:3806.
32. Stalder, T., S. Hahn, and P. Erb. 1994. Fas antigen is the major target molecule
for CD4+ T cell-mediated cytotoxicity. J. Immunol. 152:1127.
33 Hanabuchi, S., M. Koyanagi, A. Kawasaki. N. Shinohara, A. Matsuzawa,
Y . Nishimura, Y. Kobayashi, S. Yonehara. H. Yagita, and K. Okumura. 1994. Fas
and Its ligand in a general mechanism of T-cell-mediated cytotoxicity. Proc. N a f l .
Acali. Sci. USA YI:4Y30.
34 Yokota, T., D. E. Milenic. M. Whitlow, and J. Schlom. 1992. Rapid tumor penetration of single-chain Fvand comparison with other immunoglobulin forms.
Cunrer Res. 52:3402.
35 Shen, J., B. Atkinson. H. Koprowski, and H. F. Sears. 1984. Binding of human
immunoglobulin to human tissues after immunotherapy with anticolorectal carcinoma monoclonal antibody. Inr. J. Cancer 33:465.
Downloaded from on May 10, 2010
at maximum T cell activation other lytic mechanisms may override
the Fas blockade.
Taken together, the described bispecific Ab appears to be a promising candidate for a therapeutic trial in minimal residual disease. Its
secretion from theproducer cell line in a fully functional form and the
ease of its purification via aftinity chromatography on Ni-NTA columns will greatlyfacilitate the procurement of clinical grade material.
Another major advantage is clearly its small molecular mass of 60
kDa, since size is important for tumor and tissue penetration(34). The
remarkable stability of the construct in protease-rich human serum at
37°C is a further asset. Moreover, as the bispecific single-chain Ab
does not modulate the CD3 molecule on nonengaged T cells, and as
the 17-1A Ag is not endocytosed by tumor cells (35), the Ab should
be resident for sufficient periods of time on both effectors and targets.
The data with purifiedCD4+ and CD8+ T cells show that large numbers of accessory cells within the tumor may not be necessary for
efficient activation of retargeted T cells and lysis of tumor targets.
However, the strongest argument for its clinical testing may be derived from its high lytic activity and the lack of autochthonous stimulatory activity on T cells in the absence of tumor targets.
The peculiar characteristics of the single-chain construct should
allow the administration of higher doses required for tissue penetration, as well as for optimal tumor cell cytotoxicity. A caveat
needs to be added because of the wide distribution of the 17-1A
target Ag on normal simple epithelia. However, since the unmodified murine 17-1A Ab (20) and a conventionally produced bispecific 17- lA/CD3 Ab derived from hybrid-hybridoma (1 1) did not
exhibit major tissue-specific toxicity, one can therefore expect that
the presented bispecific single-chain Ab is also well tolerated. The
immunogenicity of the construct could be reduced further by humanization to prevent the formation of human anti-mouse Abs, which
could lead to tumor-independent T cell activation via their Fc parts.
Particularly for patients with minimal residual disease, who are
essentially asymptomatic afterresection of theirprimary tumor, a
nomutagenic therapywith no orlittle side effectsoffers a great