Document 2533

9 Oct 2004 0:15
Annu. Rev. Genet. 2004. 38:819–45
doi: 10.1146/annurev.genet.37.110801.143717
c 2004 by Annual Reviews. All rights reserved
Copyright First published online as a Review in Advance on August 30, 2004
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Douglas M. McCarty,1,2 Samuel M. Young Jr.,3
and R. Jude Samulski4,2
School of Pharmacy, University of North Carolina, Chapel Hill, North Carolina 27599;
Gene Therapy Center, University of North Carolina, Chapel Hill, North Carolina 27599;
Salk Institute, San Diego, California; 4Department of Pharmacology, University of North
Carolina, Chapel Hill, North Carolina 27599; email: [email protected],
[email protected], [email protected]
Key Words AAV, integration, vector, site-specific, gene therapy
■ Abstract The driving interest in adeno-associated virus (AAV) has been its potential as a gene delivery vector. The early observation that AAV can establish a latent
infection by integrating into the host chromosome has been central to this interest.
However, chromosomal integration is a two-edged sword, imparting on one hand the
ability to maintain the therapeutic gene in progeny cells, and on the other hand, the
risk of mutations that are deleterious to the host. A clearer understanding of the mechanism and efficiency of AAV integration, in terms of contributing viral and host-cell
factors and circumstances, will provide a context in which to evaluate these potential
benefits and risks. Research to date suggests that AAV integration in any context is
inefficient, and that the persistence of AAV gene delivery vectors in tissues is largely
attributable to episomal genomes.
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
SITE-SPECIFIC AAV INTEGRATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
AAV Genome Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
AAV As a Latent Integrating Virus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Site-Specific Integration in Human Chromosome 19 . . . . . . . . . . . . . . . . . . . . . . . .
Mechanism of Rep-Mediated Integration in Chromosome 19 . . . . . . . . . . . . . . . . .
Unique Nature of the AAVS1 Integration Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Elements Within the AAV Genome that Contribute to Specific
Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Structure of AAV Proviral Junctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Models for AAV Targeted Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Efficiency of AAV Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hybrid Gene Delivery Systems Taking Advantage of Rep-Mediated
Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Chromosome 19 Disruptions and Consequences for the Cell . . . . . . . . . . . . . . . . . .
Evolutionary Considerations of AAV Integration . . . . . . . . . . . . . . . . . . . . . . . . . . .
INTEGRATION OF AAV VECTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Integration Efficiency of rAAV Vectors In Vitro . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Episomal Conformation of rAAV Genomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Efficiency of rAAV Integration In Vivo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contribution of TRs to Persistence and Integration . . . . . . . . . . . . . . . . . . . . . . . . .
Structure of rAAV Integration Junctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Role of Double-Strand DNA Breaks (DSB) in rAAV Integration . . . . . . . . . . . . . .
Integration into Active Genes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Safety of rAAV Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The advent of viral gene therapy has brought adeno-associated virus (AAV) into
the limelight of virus research. The previous obscurity of this parvovirus had been
rooted in two important features of its biology. First, AAV does not appear to
cause any disease in humans, and none of its close zootropic relatives cause any
known diseases in animals. Second, it tends to remain quiescent in the absence of
a helper virus, most commonly, adenovirus (Ad). These are the same two features
that initially attracted the attention of gene therapy researchers, and have driven the
idea that AAV can be used for safe and stable gene delivery. However, the stealthy
nature of AAV has also hindered the understanding of its biology in terms of
epidemiology, persistence, and potential occult effects on the host cell, all essential
elements in a safe and effective gene delivery vector. Although the ability of AAV
to establish a latent infection in the absence of a helper virus was recognized early
on, the nature of the provirus was elusive, and has only relatively recently been
characterized. The recognition that one of the viral proteins, the replication protein
(Rep), was a key component in establishing the latent integrated state meant that
subsequent research in AAV persistence would diverge in two different directions.
One line of inquiry would follow the mechanism by which the AAV Rep protein
mediates integration of the viral genome into a specific region of the human host
chromosome. The second would elucidate the mechanism of persistence in the
absence of viral proteins, which better reflects the situation envisioned for the
AAV-derived gene therapy vectors.
The role of integration in the biology of AAV, whether Rep-mediated targeted
integration or Rep-independent recombinant AAV (rAAV) vector integration, is
highly relevant to the future use of this virus as a gene delivery tool. Whereas
most rAAV vectors will not include Rep and will not integrate specifically, the
potential for mutation and oncogenesis due to random integration may still exist.
The evaluation of the frequency of rAAV vector integration and its propensity
for targeting transcriptionally active regions of the genome is therefore an area
of research being pursed with some sense of urgency (60, 63, 85). On the other
hand, many novel gene delivery systems are being devised to take advantage of the
9 Oct 2004 0:15
targeted integration properties of the AAV Rep protein. These include hybrid viral
vectors as well as nonviral systems for delivering Rep protein and naked plasmids.
The ultimate success of these systems may largely depend on the intrinsic efficiency
of the AAV integration mechanism, how it is affected by cell type or replication
status, and the long- and short-term consequences for the targeted cell. All of these
issues are certain to have an impact on the safety and efficacy of rAAV-mediated
gene therapy.
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
AAV Genome Structure
Like all parvoviruses, AAV packages a single-stranded DNA genome, of approximately 4700 nucleotides, with palindromic regions at each end. In the case of
AAV, the palindromes are arranged such that each terminus can fold into a Tshaped DNA secondary structure through base-pairing between palindromic subregions of the terminal repeats (TR). Though the sequence content at each end
of the genome is essentially identical, the palindromic region of each TR can be
in either of two orientations. The four subregions of the AAV TR are denoted
A, B, C, and D, wherein B and C are asymmetric small internal palindromes
forming the arms of the T structure. The symmetric A palindromes flank B and
C, and form the stem of the T structure when folded and base-paired. The D
sequence is present only in one copy at each end of the genome, and therefore
remains single-stranded when the TR is in its hairpin configuration. By folding
into a hairpin, the TR at the 3 end of the genome serves as a primer for host-cell–
mediated DNA synthesis. The TRs are therefore essential for conversion of the
single-stranded virion DNA to a double-strand DNA template for transcription and
The AAV TRs also serve as origins for subsequent DNA replication through
interaction with the large AAV Rep proteins (Rep78 and Rep68). The minimal
Rep binding element (RBE) within the A palindrome is a tetranucleotide repeat
region (GAGC)3,but additional specific sequences within the arm of the T-structure
also contribute to the stability of the Rep-TR complex (50, 81, 98). After binding to the RBE, Rep mediates an ATP-dependent isomerization of the A-region
exposing the terminal resolution site (trs) at the junction between the A- and Dregions. A sequence-specific, strand-specific endonuclease activity intrinsic to the
Rep protein nicks the trs on one strand, becoming covalently attached to the newly
generated 5 end (88). The 3 end then serves as primer for replication of a new
TR end. The meticulous characterization of the steps involved in Rep-mediated
resolution of the AAV TR provided the essential background for understanding
the ability of AAV to integrate site specifically, which is unique among mammalian viruses. As discussed below, it is the recapitulation of these steps on a sequence located on human chromosome 19 (Ch19) that leads to AAV site-specific
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
AAV As a Latent Integrating Virus
The ability of AAV to establish a latent proviral state in the absence of helper
virus was recognized soon after its discovery as a contaminant in Ad stocks (26,
27). The latent AAV could be rescued from infected cells after 100 passages by
superinfecting with Ad helper virus (28). Although this demonstrated that the
virus could persist in a latent state, it did not establish its ability to integrate into
chromosomal DNA. This ability was made clear, however, when Cheung et al.
analyzed a single clone of latently infected Detroit 6 cells at early and late passage
(10). The AAV genome was found in a tandem array, joined to cellular sequences
at the viral terminal repeats. Thus, the latent state involved stable integration of the
intact viral genome into host cellular DNA, rather than an autonomously replicating
episome. The head-to-tail organization of the AAV tandem array provided the first
suggestion that the genome underwent limited replication before, during, or after
the process of integration, perhaps involving a circular intermediate. These headto-tail tandem arrays have been a common feature of AAV proviral structures in
many different contexts.
In the initial experiments to create and characterize rAAV transduction vectors,
the rep gene was retained in the recombinant vector while the neoR gene was
substituted for the capsid coding sequences (24, 94). Like the wtAAV, these vectors
integrated into the host chromosome and transduced cells to geneticin resistance.
The ability to co-opt this important property of AAV, the establishment of latency
by integration in the host chromosome, represented the first step in creating a
usable gene delivery vector. Subsequent comparisons were made between vector
transduction in the presence and absence of the rep gene, either in cis or in trans.
The results showed that the rep gene inhibited integration in some experiments
and slightly enhanced integration in others (52–54, 82). Generally, integration was
enhanced in HeLa cells and inhibited in HEK 293 and other cell lines. In retrospect,
some of these effects could be attributed to the toxic effects of Rep expression,
which is relatively low in HeLa cells but high in HEK 293 cells, due to the effect
of the Ad E1a gene product on the AAV p5 promoter (8, 9).
Site-Specific Integration in Human Chromosome 19
The initial characterizations of AAV proviral structures revealed a great deal of
heterogeneity. This property was also observed in cells carrying integrated copies of
early recombinant AAV (rAAV) gene delivery vectors (52). Restriction fragments
carrying AAV sequences had molecular weights (Mr) that did not readily correlate
with the observed number of AAV copies. In the absence of cloned proviral-cellular
junctions, this led to the interpretation that AAV integrated randomly within the
genome. This view changed, however, as AAV proviral-cellular junction sequences
became available.
The isolation of AAV proviral sequences, either by direct cloning or by enrichment for rAAV sequences containing a specific protein binding site, provided
the key to revealing the specific integration properties of AAV and AAV vectors
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
containing the rep gene (40, 42, 83). When the associated cellular sequences from
these provirus isolates were used to probe DNA from independently isolated, latently infected cell lines, it became clear that most of the provirus was integrated
within the same sequence context at a specific region of human chromosome 19
(Ch19) (19q13.3-qter). Further, examination of cells with integrated rAAV revealed that vectors containing the AAV Rep gene were mostly integrated within
this specific region of Ch19, while those lacking Rep were integrated elsewhere,
apparently at random. The Southern blot results were corroborated in studies using PCR and fluorescence in situ hybridization (37, 71). Metaphase spreads of IB3
cells revealed that AAV sequences associated with Ch19 accounted for up to 94%
of the detected wtAAV provirus, but none of the rAAV sequences lacking the rep
Mechanism of Rep-Mediated Integration
in Chromosome 19
Sequencing of the Ch19 preintegration site, termed AAVS1, and comparison with
the sequence of the cloned AAV provirus, revealed a great deal about the integration
process (41). No large regions of homology were found between the AAVS1 and
the virus, suggesting that integration occurred through a nonhomologous recombination pathway. Small (4–5-bp) homologies at the junctions between host cell
and viral DNA were consistent with illegitimate recombination products. Partial
deletion of sequences within the AAV TRs, as well as large-scale rearrangements
of the host sequences around the integration site, suggested that the process was
both complex and imprecise.
Concurrent with advances in the understanding of the biochemical activities and
DNA-binding properties of AAV Rep, Weitzman et al. found a sequence within
the AAVS1 region that could also bind specifically to the Rep protein (98). This
comprised the same tetranucleotide repeat (4 copies) that mediated Rep binding in
the AAV terminal repeat. Further, Rep protein was able to simultaneously bind the
AAVS1 RBE and the AAV TR sequence. This immediately suggested a mechanism
for AAV site-specific integration wherein the Rep protein tethers the AAV genome
to the AAVS1 chromosomal sequence.
The biochemical interactions between AAV Rep and the TR that led to resolution
and replication of the genome ends had previously been characterized. These
included secondary structure and sequence-specific DNA binding, ATP-dependent
DNA helicase, and site-specific, strand-specific endonuclease (32, 33, 90). Each
of the steps leading to terminal resolution had been recapitulated in vitro using
purified Rep protein. Indeed, the entire AAV rolling hairpin replication process
could be reconstituted by mixing Rep protein with HeLa cell nuclear extracts (64).
The recognition that the AAVS1 fragment contained an active RBE suggested
the possibility of a human chromosomal version of the AAV replication origin,
which could participate in the specific integration process. When a cloned subfragment of the AAVS1 sequence, containing the RBE, was incubated in nuclear
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
extracts containing Rep protein, specific DNA replication products were readily
detected (95). Not surprisingly, given that Rep nicks specifically on one strand
of the AAV TR, the replication observed from the plasmid carrying the AAVS1
was specific for only one strand. This observation of asymmetric replication was
consistent with the known activities of the Rep protein, in that AAV replication is
mediated entirely by leading-strand DNA synthesis.
The initiation site for asymmetric replication from the AAVS1 origin mapped
to a sequence that conserved 5 of 6 bases relative to the trs associated with the
AAV TR, and included a central dithymidine (TT), which is the sequence cleaved
in the AAV TR site. Although this putative chromosomal nicking site was spaced 5
bases closer to the RBE than the TR site, and the spacing can affect the efficiency
of trs endonuclease activity in the context of the AAV TR (7, 89), the AAVS1
nicking site homologue was cleaved predominantly at the analogous site, between
the two thymidine residues. Further, the products of the cleavage reaction were a
free DNA 3 OH end and a covalent 5 DNA-Rep complex, as is produced from
the endoclease reaction with the AAV terminal repeat.
Taken together, the characterization of the AAVS1 site on human chromosome
19 revealed a functional homologue to the replication origin in the AAV terminal
repeat. The presence of this AAV Rep-dependent replication origin near the integration site implied that integration was associated with limited DNA replication of
cellular sequences. This was consistent with the chromosomal rearrangements, including duplications and inversions, which were present at the host-virus junctions
from the latently infected cells (41).
Deeper understanding of the AAV site-specific integration process was achieved
through the use of episomal cognates of the chromosomal integration site. Giraud
et al. cloned the AAVS1 fragment into Epstein-Barr virus-based vectors, which
could be rescued from mammalian cells and grown in bacteria (19, 20). This system
allowed both quantification of AAV integration events and characterization of the
products. At 48 h postinfection, approximately 1.5%–3.0% of the rescued episomes
contained AAV sequences. Consistent with the hypothesis that the RBE and nicking
site were involved in the process, a subcloned 510-bp fragment containing these
elements was both necessary and sufficient to direct targeted integration of the
AAV genome into the episome. The system was later refined to show that the
two cis-acting elements, the RBE and the trs homolgue, contained within a 33-bp
sequence, were the mediators of integration through interaction with the AAV Rep
protein (47, 48).
Unique Nature of the AAVS1 Integration Site
Although the RBE in AAVS1 and the AAV TR is a (GAGC)3 repeat sequence, Rep
protein is somewhat permissive in its binding specificity, and substitutions within
the core sequence are tolerated (11, 81). Substantial deviations from the consensus
repeat are still specifically bound and have biological function. The AAV p5 promoter binds Rep to mediate transactivation and repression of viral gene expression,
but contains only one complete (GAGC) and one degenerate (GAGT) copy of the
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
repeat (43, 49, 67). The sequence of the human inmunodeficiency virus (HIV) long
terminal repeat (LTR) contains a degenerate sequence similar to the AAV TR Rep
binding site, but with only one intact GAGC motif, yet the transcriptional activity
of the HIV LTR is strongly repressed by the AAV Rep protein (5, 39, 65, 76).
Additionally, a spurious sequence within the plasmid pBR322, as well as most
derivatives of this bacterial vector, also binds specifically to the Rep protein (49).
The relatively nonstringent requirements for Rep binding and the simplicity of
its consensus binding site suggest that there are a great many potential interactions
within the human genome. At least 18 Rep binding sequences have been located
within or flanking human genes (99). Based on a consensus GAGYGAGC sequence, there may be up to 2 × 105 Rep binding sites in the human genome (107).
However, no secondary preferred integration sites have been observed. Clearly, the
close proximity of the essential nicking site plays a major role in this specificity.
Other factors that may contribute are elements of chromosomal context including a nearby CpG island, insulator sequences, and numerous transcription factor
binding sites (41, 66). The AAVS1 also contains a DNase hypersensitive region,
suggesting that it is situated within an open chromatin structure (46).
The AAVS1 sequence has been found only in humans and higher primates.
Integration of AAV and rAAV vectors carrying the rep gene is apparently random
in normal rodent cells. However, in transgenic mice and rats carrying the human
AAVS1, on an 8.2-kb, 3.5-kb, or 2.7-kb fragment, AAV integration is specific for
the exogenous sequence (3, 77, 107). This suggests that all of the essential cis
elements for specific integration are located within a 1.6-kb region common to all
of the exogenous fragments used to make these transgenic animals. The DNase
hypersensitve region of the exogenous AAVS1 is also maintained in the open
chromatin conformation in the animal models, suggesting that cis-acting signals
associated with the transgenic fragment direct specific chromatin modeling. This
further suggests that open chromatin is likely to be a prerequisite for Rep access
to the RBE and nicking site.
Elements Within the AAV Genome that Contribute
to Specific Integration
As stated above, the Rep protein bound to the AAVS1 RBE can form a complex
with the AAV RBE located within the terminal repeat. This suggests a mechanism
for bringing the AAV genome into close proximity to the chromosomal integration
site. While the presence of the nicking site in chromosomal or episomal AAVS1
is essential for specific integration, this does not appear to be the case for the
trs of the viral sequences. Young et al. were able to achieve specific integration
using an AAV construct with mutant TRs, which contained an insertion between
the RBE and trs (108). These TRs had previously been shown to be defective
for Rep-mediated endonuclease activity (89). Other investigators have observed
Rep-dependent specific integration of plasmids lacking the AAV TR, though at
an efficiency at least tenfold lower than plasmids containing at least one copy
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
of the terminal repeat (93). This may have been mediated through the spurious
Rep binding site within the plasmid bacterial sequences, as mentioned above.
Alternatively, the DNA damage induced by Rep-mediated nicking of the AAVS1,
and its subsequent repair, may create a hotspot for illegitimate recombination.
In addition to the RBE contained within the AAV TR, sequences within the
AAV p5 promoter region also appear to have a role in mediating or enhancing
Rep-mediated specific integration. Junction breakpoints between viral and cellular
sequences frequently cluster around the p5 promoter in both the EBV episomal
model system and in proviral structures isolated from latent cell lines (20, 105).
Additionally, the presence of p5 on a rAAV construct serves to enhance the rate
of site-specific integration by 10–100-fold in the presence of complementing Rep
protein (68, 69). The sequence responsible for this enhancement, termed the p5
integration enhancer element (p5IEE), is contained within a 138-bp fragment that
also encompasses the functional p5 promoter region, including the RBE (9, 49).
Further, in plasmid constructs, this sequence alone can mediate specific integration
more efficiently than constructs containing both the p5IEE and the terminal repeats.
Philpott et al. suggest that the TRs can introduce boundaries to the exogenous
integrated sequence, which may lead to integration of only half of a circular plasmid
containing two TRs rather than the whole plasmid (69). Thus, the p5-associated
RBE, which is essential for the regulation of p5 promoter activity as well as
the activities of the two downstream AAV promoters, is likely to be the primary
mediator of targeting in these constructs, through its interaction with the Rep
protein.Other transcription factors bound within the promoter region may also
Structure of AAV Proviral Junctions
The organization of the AAV proviral structures, either rescued from EBV shuttle
vectors or cloned from cell lines, is complex, with large-scale deletions, rearrangements, and duplications. The EBV-rescued sequences present the greatest sampling
of proviral junctions, which generally incorporate the following features.
First, most of the crossover points between AAV and cellular sequences were
within 100 bp from the AAVS1 nicking site. This is somewhat different from
junctions cloned from latent cell lines, which are more typically farther away,
but within 1000 bp of the RBE and nicking site (41, 83, 105). In either case, the
junction between viral and cellular DNA was always downstream from the nicking
site, consistent with nicking and initiation of DNA synthesis on the chromosome
prior to incorporation of AAV sequences. This limited DNA synthesis has the effect
of moving the crossover point to a location downstream from the initial nicking
Second, all of the integrated AAV sequences had one end associated with the
AAVS1 sequence. The other end, however, was linked to unrelated sequences
within the shuttle vector. In several of the EBV clones, however, it was later found
that an insertion of a human Line-1 element into the EBV episome had occurred
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
independently of AAV integration (15). This EBV-Line-1 recombinant had then
propagated and become the target for several AAV integration events, and some of
these had formed junctions between AAV and the Line-1 sequence. Other junctions
were associated with EBV vector sequence. Taken together, this demonstrates that
AAV integration is not a simple insertion process, and that large segments of target
sequence can be deleted. This was also noted in a careful analysis of a proviral
structure from a latent cell line, along with duplications and inversions of the
AAVS1 sequence (41).
Third, the crossover points within the AAV sequences were usually near a Rep
binding site, either within the TR or the p5 promoter. In contrast, crossover points
at the AAV trs were not seen, suggesting that nicking of the AAV molecule by
Rep was not an important feature for integration, consistent with the observation
discussed above, that AAV genomes with defective nicking sites, or lacking TRs,
are still integrated into the AAVS1 site.
Lastly, the AAV sequences frequently included a head-to-tail viral junction,
suggesting that multiple copies in a tandem array had integrated. This was also
common to many of the proviral junctions previously characterized in cell lines.
This suggested that a circular form of the viral genome was a precursor to the
integrated form and that some degree of DNA replication of the viral sequence
had occurred before or during integration.
Models for AAV Targeted Integration
The interpretation of early events of AAV integration, including binding of both the
AAVS1 sequence and the integrating genome by Rep protein, and the consequent
formation of protein-DNA complexes tethering both structures together, are well
supported by observations both in vivo and in vitro. This complex formation is
likely to be followed by nicking of the AAVS1 site, but not necessarily the AAV
sequence. Again, the necessity of this nicking event in the target sequence has
been confirmed in vitro, and in the EBV episomal system using mutant target
sites. Exactly what sequence of events takes place after the nicking step is not
entirely clear. However, the general complexity of the junction sequences, and the
rearrangements in both target and viral DNA, suggest that the mechanism is not
precise and has little in common with cut-and-paste type integration systems such
as retrovirus or lambda phage.
Most of the observations regarding AAV proviral junctions have been incorporated into a general model of AAV integration that is dependent on initiation of
DNA replication, followed by several rounds of template switching by the host
polymerase complex. Some features of the AAV proviral junctions that point to a
DNA replication-dependent mechanism are the locations of the breakpoints at a
distance downstream from the AAVS1 nicking site, and the duplications of AAVS1
sequences. Whereas these imply replication of host chromosomal DNA, the frequent presence of head-to-tail junctions of tandem AAV sequences, joined via TR
structures, suggest that the AAV genome has also been replicated. Linden et al.
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
have proposed that integration begins with the formation of a Rep-DNA complex
involving the AAVS1 site and the AAV genome, either through binding at the TR
or through the RBE associated with the p5 promoter (47, 48). The presence of
head-to-tail AAV junctions within the proviral sequence suggests that the genome
is circularized prior to this complex formation. The Rep protein then creates a nick
at the chromosomal site, with concomitant covalent linkage to the 5 end of the
nicked chromosomal DNA strand. Ensuing replication from the 3 end of the nick
displaces the original chromosomal DNA strand, along with the attached Rep protein complex and circular viral DNA. At some point after limited replication (up
to approximately 1000 bp), the host DNA polymerase complex transfers from the
chromosomal template strand to the displaced strand, now returning back toward
the Rep complex. Again, the host DNA polymerase complex switches templates,
jumping from the host DNA to the associated viral genome. After one or more
times around the viral genome, creating the head-to-tail concatemers, the polymerase complex switches templates once more to return to the host chromosomal
Clearly, given the multiple template switching required to create the observed
structures, such a mechanism would also entail a great deal of gap filling and repair
activity subsequent to the synthesis of the leading strand. Whether this activity is
the cause of the multiple duplications, rearrangements, and insertions frequently
observed following AAV integration into the AAVS1 site, or whether they are
created through the meanderings of a DNA polymerase complex as it is tripped-up
by the AAV Rep protein, remains to be determined.
Efficiency of AAV Integration
There are several different perspectives from which to evaluate the efficiency of
AAV integration into Ch19. The initial estimates were based on the number of
rescue-competent, latent infections that could be derived from a pool of infected
cells (6). This is similar to measurements of the percentage of transduced cells using
rep+ rAAV vectors, and is always dependent on the multiplicity of infection of the
virus or vector (52, 53, 82). In these studies, integration efficiencies of 20%–80%
of infected cells were typically achieved. In contrast, the EBV episomal target
integration studies discussed above measured the percent of rescued plasmids
containing AAV sequences, and these were in the range of 0.01%–0.05%. Again,
the multiplicity of infection was a critical factor in these studies, as was the copy
number of EBV episomes. To account for these variables, and to best describe the
behavior of the AAV virus, the measurement of integration efficiency in terms of
integrations per infectious unit of AAV probably has the greatest relevance.
The evaluation of many of the earlier studies on AAV integration in the presence
of Rep protein in terms of integrations per infectious unit generally yielded frequencies in the range of 0.1%–0.5% (reviewed in 51). More recent studies have supported this conclusion. Huser et al. developed a rapid and efficient assay for Ch19specific integration based on polymerase chain reaction (PCR) amplification using
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
primers specific for the AAV and chromosomal sequences (30). The frequency of
integration was consistent with earlier results at approximately 0.1% of infecting
viral genomes. The time course of integration was also evaluated in this study and
found to plateau at approximately 24–48 h postinfection. This was also consistent
with the EBV episomal studies, where most integration occurred between 24–48 h
postinfection, and few integration events were observed during the first 10 h (19).
The PCR-based assay also provided a large sampling of integration events that led
to the conclusion that there was little or no bias between the left and right ends of
the AAV genome in terms of its orientation in the AAVS1 site (29). It is not clear
how this observation relates to the effect of the p5 promoter-associated RBE and
its effect on integration efficiency, as noted above. The RBEs are possibly favored
sites for template switching, whether a polymerase complex is entering the AAV
DNA sequence, or leaving it to reassociate with the chromosomal DNA sequence.
The integration efficiency of wtAAV vectors in vivo has not been well characterized. Because rodents do not contain the Ch19 integration sequence, these
studies must be performed in primates or in the transgenic rodent models. To date,
only one such study has been reported, using rhesus macaques (25). Chromosome
19 integration-specific PCR signals were detected from the nasal tissue of one
of two nasally infected animals, and weakly from the liver tissue of both of two
intravenously infected animals. This suggested that the integration frequency for
wt AAV in vivo was low.
Hybrid Gene Delivery Systems Taking Advantage
of Rep-Mediated Integration
Although most AAV-based vectors are not envisioned to include the Rep gene,
and thus are likely to integrate randomly and at a low frequency, the ability to take
control of the integration mechanism remains attractive. Such vectors resemble
the first of the constructs used in AAV gene transfer, which included the rep gene
and the neomycin phosphotransferase gene (52). More recently, many researchers
have created hybrid systems that include the Rep gene and the AAV TRs but are
delivered using either nonviral or chimeric virus constructs. These possibilities
were first explored by cotransfecting plasmids containing AAV rep, and the transgene plus TRs on separate molecules (4, 31, 38, 70, 93, 101). While successful
in directing targeted integration, this method retains an active episomal rep gene
in the transfected cells, which could lead to chromosomal instability or mobilization of the transgene. These problems have been addressed using methods such as
direct transfection of the Rep protein rather than the rep gene (45), or using a conditionally inactivated rep gene containing Lox-Cre recombination elements (84).
Another strategy has been to fuse the rep gene to a hormone-dependent ligandbinding domain such that it is only transported to the nucleus in the presence of a
hormone analogue (75).
Other hybrid systems for Rep-mediated Ch19-specific integration have utilized
chimeric virus constructs to deliver both the transgene sequences and the rep gene.
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Recchia et al. created helper-dependent adenovirus constructs, one containing the
rep gene, the other containing the transgene and TR sequences (74). Coinfection
with the two virus constructs allowed efficient targeting of the transgene to the
AAVS1 site. An important advantage of this system is the large transgene capacity
of the helper-dependent Ad system, though it is not clear whether the imprecise
nature of AAV integration can maintain the integrity of very large transgenes.
Chimeric herpes simplex virus (HSV) vectors utilizing Rep-mediated targeted
integration have also been created and tested in several different contexts (36, 97).
In each case, the presence of the rep gene in the chimeric vector resulted in efficient
integration of the construct into AAVS1 as well as increased stability of transgene
expression. In contrast, similar vectors used to infect glioma cell lines resulted in
decreased stability of transgene expression over time, presumably due to toxicity
of the Rep protein (44). These studies have more recently been extended to a
transgenic animal model, wherein the exogenous human AAVS1 sequence served
as a target for integration of the HSV/AAV hybrid vector (3).
There is still a great deal of potential for the use of these hybrid systems for gene
delivery, whether in a research or a clinical setting. Future developments are likely
to include more efficient means of direct protein delivery using nonviral vectors,
nonviral delivery of rep RNA transcripts, or RNA virus-mediated delivery of the
rep gene. In any of these cases, rep expression would be transient, thus allowing
for long-term stability of the integrated transgene and minimal toxicity from Rep
Chromosome 19 Disruptions and Consequences for the Cell
The propensity for rep-containing AAV vectors, and hybrid vector systems derived
from them, to integrate into human Ch19 has potential benefits for gene therapy
applications, but potential side effects as well. In addition to the deletions, insertions, and duplications of the AAVS1 locus upon AAV integration, as noted above,
the nonintegrated AAVS1 allele can also suffer amplifications and rearrangements
(58, 108). These rearrangements are dependent only upon the expression of the
Rep protein, and do not require delivery of a construct containing the AAV terminal repeats. While no specific disease syndrome has been linked to AAV infection,
the loss of both AAVS1 alleles would probably be a relatively rare circumstance,
making it difficult to evaluate a putative detrimental effect epidemiologically.
Early characterization of the AAVS1 sequence suggested that it was part of
an actively transcribed region (41). In addition to the presence of a CpG island
and multiple transcription factor binding sites, part of the AAVS1 sequence was
represented in a cDNA library. The DNase hypersensitive region associated with
this region in human cells, and its maintenance in transgenic animals, was also
consistent with an overlapping transcription unit (46). Dutheil et al. have characterized the transcript from this region as part of a muscle cell-specific gene called
the slow skeletal troponin T gene (TNNT1), which also maps to 19q13.4 in humans
(14). Because AAV is generally considered to be an upper respiratory virus, based
9 Oct 2004 0:15
on the tissue tropism of its helper Ad, it is not clear whether disruption of both
alleles of TNNT1 would have any potential pathogenic effect. However, rAAV
vectors can efficiently infect muscle cells when directly injected, and a phenotype
from disruption of AAVS1 in this tissue might be more readily detectable. A clear
understanding of the role of the TNNT1 gene product will be essential before
AAV-derived targeting vectors can be applied in humans.
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Evolutionary Considerations of AAV Integration
As stated above, the AAVS1 sequence, with active RBE and nicking site, is found
only in humans and higher primates. Whether AAV benefits from the ability to
integrate into this site is still a matter of speculation. However, establishing a
stable proviral state in the absence of a helper virus would allow for long-term
maintenance of the AAV sequence in dividing cells, until conditions were favorable
for replication and packaging of progeny virus.
In favor of the hypothesis that AAV has adapted the ability to integrate specifically is the observation of what appears to be coevolution of the AAVS1 site and
AAV serotype 4 from monkeys. Amiss et al. noted that the AAV-4 RBE contains an
extra copy of the GAGC repeat unit (2). Upon sequencing the AAVS1 homologue
from simian CV-1 cells, it was also found to contain an additional copy of the
repeat unit. Subsequent experiments revealed that binding of the AAV-4 Rep to the
simian sequence had a higher affinity than that of AAV-2 or AAV-4 to the human
sequence. Further, the frequency of integration into the AAVS1 homologue was
also greater using AAV-4 in simian cells. This suggests an evolving relationship
between the virus and the preferred host, based on interactions at the AAVS1 site,
which would be difficult to explain unless there is some benefit to the virus.
On the other hand, AAV integration is relatively inefficient compared with obligate integrating viruses, such as retrovirus. With an observed frequency of less
than 0.5%, it is not clear how integration would be biologically relevant. Additionally, the integration mechanism is imprecise, often leading to rearrangements
of the proviral sequences as well as the target sequence. It would appear that AAV
integration is largely a consequence of host-cell–mediated illegitimate recombination. The fact that the recombination can be initiated by the action of AAV Rep
protein at a serendipitous juxtaposition of an RBE and a functional nicking site
may be essentially an accident of nature. However, evolution is driven by such
accidents and their consequences, and AAV integration might well be regarded as
an evolutionary work in progress.
The concept of using AAV as a gene therapy vector was built largely on the idea
that it was an integrating virus, an immensely desirable property in that infected
cells would pass the transgene to daughter cells, and readministration of the vector
would unnecessary. While the advantages of vector integration may still apply to
9 Oct 2004 0:15
rapidly dividing cell types, such as bone marrow–derived cells, they are not so clear
for most tissues in the body, which divide slowly if at all. The maturing consensus
in gene therapy now includes the recognition that the risks inherent in an integrating
vector, oncogenic transformation and gene inactivation, outweigh the benefits of
permanent transduction in many tissues. Fortunately for adherents of rAAV gene
therapy, the intervening years have also seen a change in our assessment of the
integration properties of rAAV vectors.
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Integration Efficiency of rAAV Vectors In Vitro
Some of the earliest experiments with rAAV demonstrated that vectors lacking
the Rep protein were still competent for chromosomal integration (52, 54, 82).
Although these vectors were initially thought to integrate more efficiently than
those retaining rep, this does not appear to be the case (see above). Most of the
early studies, as well as many current reports, focused on the percentage of cells
that could be stably transformed in the presence of a relatively high multiplicity
of infection with vector. Such experiments frequently resulted in transduction
frequencies over 80%; i.e., up to 80% of the cells in the culture expressed the
transgene. When a selectable marker gene such as neo was used, this translated to
an 80% integration frequency. While informative, a more generally useful measure
of integration efficiency is the number of transformation events per infectious unit
of vector. This could be calculated from the data in many studies by estimating
the infectious dose based on typical particle to infectious unit ratios, which range
from approximately 10-100:1 in cultured cells (reviewed in 51). These calculations
yield numbers in the range of 0.1–0.5 integrations per infectious unit, either with
or without selection. Although this is far from the efficiency typically associated
with obligate integrating virus vectors such as retrovirus, it is still higher than
has been observed from a nonintegrating virus vectors such as adenovirus, which
exhibits integration frequencies ranging from 10-3 to 10-5 per cell at a multiplicity
of infection of 10 PFU per cell (23). This translates to approximately 10-4 to 10-6
per infectious vector genome, compared with 0.2–1.0 × 10-3 for rAAV vectors.
Further, unlike retroviruses or retrovirus vectors, which must integrate to express
their genes, rAAV vectors clearly do not require integration for gene expression
The efficiency of Rep-independent rAAV vector integration is within the same
range as Rep-mediated integration. Although Rep clearly has no role in the integration of these vector genomes, some of the same host-cell factors that mediate
Rep-independent integration may also participate in the specific integration mediated by Rep protein. Very little is known about host-cell factors involved in
Rep-independent integration of rAAV vectors, or whether any factors specifically
recognize and operate through interaction with the AAV terminal repeats. Indeed,
the view that the AAV TRs mediate integration at all has recently come into question. However, a body of evidence has accumulated over the years, supporting the
idea that the presence of the AAV TRs enhances the rate of integration, at least under the experimental conditions applied. A further consideration is the possibility
9 Oct 2004 0:15
that the AAV TRs enhance the persistence of the vector DNA episomally, by resisting degradation, promoting the formation of circular and concatemeric molecules,
or by providing a replication origin. An increased probability of chromsomal integration would then be a secondary consequence of this episomal persistence. For
this reason, it is important to consider the behavior of episomal rAAV genomes,
as these are the likely precursors for integration.
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Episomal Conformation of rAAV Genomes
Many recent studies have focused on the fate of rAAV genomes after transport
to the nucleus, uncoating, and second-strand synthesis. These have often been
performed in an in vivo setting, partly because of the relevance to gene therapy,
but also because the episomes are maintained for much longer periods of time in
the nondividing cells of animal tissues than they are in cell cultures. Generally,
these studies point to processing of the rAAV genomes by intramolecular and intermolecular recombination to form circles and concatemers, respectively. Several
studies using split-gene vector coinfections were contrived to increase the gene
delivery capacity of rAAV vectors (92, 104). In these constructs, half the gene is
in one vector and the other half in a second vector, such that reconstitution of the
reporter gene requires the joining of two different vectors in the correct orientation.
These and earlier experiments demonstrating rescue of mixed vector concatemers
in bacteria clearly demonstrate that some, if not all, concatemers result from intermolecular recombination (106). Concatemers created by amplification of circular
episomes would not lead to gene expression from these constructs.
The circularization of rAAV genomes has generally been evaluated by either
rescue of the circularized molecules after transfection into bacteria or visualization of circular forms in Southern blots. These studies have demonstrated that
monomeric linear rAAV DNA is converted to circular forms over time, and that
these circular forms are then slowly converted into large Mr concatamers (>12
kb) (12, 13, 55, 96). Whereas some concatemers result from the joining of two
different vector molecules early after infection, rolling circle replication from the
circular monmomeric genomes may also contribute to the high Mr forms. Rolling
circle replication from plasmids containing the AAV TR in the absence of Rep
protein has been reported in cultured cells, and in cell-free extracts, after subjecting the cells to genotoxic stress (102, 103). This would be consistent with the
predominantly head-to-tail arrangement of genomes in these structures.
In order to determine whether single-stranded rAAV DNA had a role in recombination between AAV genomes, Yue et al. performed sequential infections
with split-gene vectors in mouse muscle tissue at times separated by several weeks
(109). The second vector was added after the DNA of the first vector had been
either converted to double-strand or cleared. Indeed, most of the vector DNA had
been converted to monomeric circles and large concatemers before the second vector was added. Reconstitution of gene expression after infection with the second
vector demonstrated that single-strand DNA was not essential for the formation of
heteroconcatemers. Further, it implied that the second vector interacted with DNA
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
that had already circularized or joined into a higher Mr structure. However, it is not
clear whether the circular forms recombine directly or through a linear intermediate. Conversion of circular plasmids containing a modified AAV TR sequence,
resembling that often found at the junction between joined AAV ends, has been
observed in cultured cells (101). This raises the possibility that episomal forms of
rAAV exist in a dynamic balance between circular and linear forms, with circularization predominating. Transient linearization would create an opportunity for
recombination with a second linear genome, or a larger concatemer. Such a mechanism of transient availability of rAAV ends could also contribute to integration
into the host chromosome.
Efficiency of rAAV Integration In Vivo
The safe and efficacious use of rAAV vectors for gene therapy will rely on a thorough understanding of its propensity for integration into the host chromosome
and the potential consequences of integration. Early experiments with rAAV vectors in cell culture, with or without selection, had established that integration was
relatively inefficient, at approximately 0.1%–0.5% of infectious vector genomes.
However, it was not clear that this would be the case in the nondividing or slowly
dividing cells of animal tissues. Several recent studies have addressed the question
of integration efficiency in vivo, with results that are somewhat inconsistent.
The first studies designed to estimate the fraction of rAAV vector DNA integrated into the host chromosome were performed in transduced mouse liver tissue.
It had already been noted that the rAAV vector genomes are converted from low
Mr episomes to higher Mr forms in this and other tissues over the course of several
weeks (16, 91, 100). In a subsequent study, Miao et al. assessed the integration
status of these high Mr forms using both pulsed-field electrophoreses and in situ
hybridization (56). Southern blots of the pulsed-field gels suggested that all of the
high Mrvector DNA was associated with chromosomal DNA, or at least with DNA
segments greater than 1 megabase in size. Based on their previous estimates of
an average rAAV vector content in transduced mouse liver of 3.5 copies per cell,
and their observation that only 5% of liver cells were expressing the transgene, the
authors suggested that most vector DNA was within cells containing an average
70 vector genome copies. If these were arrayed in concatemers, the average size
would be in the range of only 275 kilobases. The authors then looked at interphase
and metaphase nuclei for rAAV sequences associated with chromosomes. Consistent with their transgene expression observations, they found vector in 4.8% and
5.5% of interphase and metaphase nuclei, respectively. The vector sequences were
detected on sister chromatids in the metaphase nuclei, supporting the conclusion
that they were integrated.
These results were put in doubt, however, when subsequent studies revealed a
population of extrachromosomal vector DNA within the transduced hepatocytes,
in the form of monomeric circular and linear molecules (62). Further evidence that
the preponderance of rAAV DNA remained episomal came from studies employing
two thirds partial hepatectomy in transduced mice to induce liver tissue regenera-
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
tion (63). After one to two cycles of hepatocyte cell division, the transgene expression would decrease if the vector DNA were episomal, and remain stable if it were
integrated. Mice subjected to partial hepatectomy at 12 weeks or 12 months postinfection displayed decreases in vector gene expression of approximately 85%–95%,
and similar decreases in the average number of vector genomes per cell. There was
no difference between mice hepatectomized at 12 weeks and 12 months postinjection, implying that the balance between integrated and episomal vector DNA was
relatively stable. Because the limited number of cell divisions did not ensure that
all of the extrachromosomal vector DNA was lost, this suggests that the observed
5%–15% integrated vector represented a maximum value in mouse hepatocytes.
Chen et al. also observed integration of rAAV in liver tissue using an in vivo
selection model (92). In this case, a population of transduced hepatocytes representing 0.1%–0.5% of the liver could be expanded to approximately 50%–90%
of the liver after selection was applied. While this does not provide quantitative
information about the frequency of vector integration, it does corroborate the presence of integrated rAAV genomes in liver tissue. Whether these genomes were
integrated before or after the induction of hepatocyte cell division is not known.
The evaluation of rAAV integration frequency in mouse muscle tissue has
yielded far more conservative estimates. Using three different assays, Schnepp
et al. were unable to detect integrated vector DNA to a sensitivity of <0.5% of
total vector DNA (85). First, genomic DNA from transduced muscle tissue was
size selected and cloned into phage. Screening of 1.5 × 106 clones, representing
twice the cellular genome content, revealed no vector-specific sequence. A second method of quantitation relied on a carefully calibrated PCR assay to amplify
integrated rAAV sequences. At a sensitivity of 75 integrated vector genomes per
1.5 × 104 total vector genomes, no integrated rAAV sequences could be detected.
In a third assay, a DNase that was specific for linear double-strand, and linear
or circular single-strand DNA, but not double-strand circular DNA, was used to
digest DNA from transduced mouse muscle tissue. The DNase treatment did not
significantly reduce the amount of vector sequence in these samples, suggesting
that all of the vector was in the form of double-stranded circular DNA molecules.
Contribution of TRs to Persistence and Integration
Recombinant AAV genomes have long been viewed as special substrates for chromosomal integration because of the presence of the hairpin TR structures. However,
the role of these hairpins in the recombination mechanism that leads to integration
has been difficult to verify or characterize. Direct comparisons between rAAV
genomes and plain DNA are difficult due to the qualitative and quantitative differences in delivery mechanisms. One study that attempted to normalize for these
problems compared naked linear DNA, either with or without the AAV TR sequences, in mouse liver after in vivo hydrodynamic delivery (61). Although the
TRs of the transfected DNA were not in the hairpin configuration at the outset,
it remained possible that isomerization of the TRs within the cell would create
a molecule resembling the rAAV genome, and potentially provide a primer for
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
at least one round of replication. The integration status of the DNA was then
evaluated by measuring transgene expression after a two thirds partial hepatectamy. The authors found that induction of liver regeneration reduced expression
from linear molecules lacking the AAV TR sequences by approximately tenfold,
whereas expression from linear molecules containing the TR sequences dropped
by only fivefold. Expression from circular molecules containing one copy of the
TR dropped by approximately 20-fold. This suggested that the TRs did have a role
in integration, albeit small, and further, that the linear form was more likely to
integrate than the circular form. Isolation and characterization of some of the hostchromosome junctions with vector DNA showed that the recombination usually
occurred within the TR sequence, again implying a role for the hairpin, at least to
the extent of providing a free end as a substrate for recombination.
Structure of rAAV Integration Junctions
Like the Rep-mediated integration described above, a great deal may be learned
about rAAV vector integration from the structures of the vector-host cell DNA
junctions. Examples of these structures have been characterized from both cell
culture and animal tissues (59, 80, 105). Each study revealed similar junction
structures, whether the source was liver tissue in animals, rapidly dividing cells
in culture, or recombination reactions carried out in cell-free lysates. They were
also similar to junctions recovered from Rep-mediated integration, except that
they were not located near the AAVS1 site. The vector-cellular DNA breakpoints
were predominantly found within the AAV TR, though all had some degree of
deletion of TR sequences. Small, 2–5-bp microhomologies between vector and
host sequence were generally found at the crossover points, but no large regions
of homology were observed. This suggested that rAAV vector integration was
mediated by nonhomologous recombination, either specifically enhanced by the
TR sequence or secondary structure, or utilizing a free DNA end formed by the
hairpin structure.
More recent studies of rAAV vector integration have benefited from the availability of the human genome sequence in characterizing the chromosomal integration site and the changes that took place at the junctions (58). In addition to
deletions within the TRs, small deletions (9–71 bp) and small insertions (1–13 bp)
were found within the cellular sequences. Larger deletions, up to 2 kb, were also
noted in a separate study (60). A segment of the plasmid vector pBR322 was also
found at one junction, implying that it had been packaged and transduced with
the vector. An unexpectedly large proportion (4 of 9) of the junctions were found
within genes. Also, four of the junctions were located within a highly transcribed,
22-Mb region of Ch19, though at great distances from the AAVS1 site. These observations suggested that rAAV preferentially integrates into actively transcribed
regions. One junction spanned a chromosome 9 to chromosome 17 translocation.
The association with actively transcribed regions, which are hot spots for DNA
damage, and instances of chromosomal rearrangement at the junctions suggested
the possibility that the rAAV genome was interacting with previously formed
9 Oct 2004 0:15
double-strand DNA breaks. This further suggested a model for rAAV integration
based on nonhomologous end-joining between the hairpin TR and the broken
chromosome ends (78).
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Role of Double-Strand DNA Breaks (DSB) in rAAV Integration
The possibility that rAAV integrates through the cellular nonhomologous endjoining DNA repair pathway was further explored using an induced DSB model
(57). The authors used a retrovirus vector to create a cell line with a unique I-SceI
restriction site within a bacterial shuttle construct. A second retrovirus construct
was used to introduce the I-SceI enzyme gene, such that a fraction (approximately
5%) of the cells would have a DSB at the specific site. The cells were then infected
with an rAAV vector containing a selectable marker, and integration at the induced
DSB was assayed in integration-positive cells. One of the 12 positive clones contained the rAAV at the induced DSB site. When the DSB site was more specifically
examined by rescue in bacteria, 8 of 190 clones (4.2%) contained the rAAV within
the site, suggesting that the DSB created a target for rAAV integration.
A converse experiment, assaying the percentage of rAAV integrations occurring
at this specific DSB, utilized an rAAV shuttle vector for replication in bacteria.
This resulted in the finding that 7.4% of the integrated rAAV had targeted the
induced double-strand break. Selection with the target site marker gene revealed
that 0.59% of the I-SceI sites contained the rAAV vector.
The authors examined rAAV integration at DSBs induced by treatment of cells
with etoposide or γ -irradiation. Both treatments resulted in an increased frequency
of rAAV integration. The structures of the chromosomal integration sites from cells
with induced DSBs were then compared with sites rescued from uninduced cells.
Both types of junctions showed a preference for integration into transcribed regions, and both showed similar aberrations of chromosomal sequence at the junction sites. Together, these results suggest that rAAV integrates into pre-existing
chromosome breaks by nonhomologous end-joining, and that the aberrations associated with rAAV junctions may be the result of inaccurate repair of previously
generated deletions and insertions.
A distinctly different role for DSB was characterized in a study of rAAV vectormediated gene repair by homologous recombination. Previous reports had shown
that rAAV could mediate gene correction by homologous recombination driven
by flanking homologous sequences in the vector genome, and that the singlestranded rAAV genome was likely to participate directly in this interaction (34,
35, 79). Porteus et al. created an I-SceI endonuclease model for gene correction at an
induced DSB site (73). An rAAV vector with sequences flanking the inserted I-SceI
site could correct a GFP gene by deletion of the intervening sequences. Induction of
a DSB at this site increased the frequency of gene repair by >100-fold. Importantly,
the enhancement of homologous recombination by induced DSBs can also be
mediated by plasmid (72). This suggests that while rAAV integrates randomly by
interacting with nonhomologous end-joining repair pathways, it can also interact
with repair pathways involving homologous recombination. This points to the
9 Oct 2004 0:15
possibility that the rAAV genome integrates into the host chromosome as a passive
bystander rather than an initiator of recombination.
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Integration into Active Genes
The study discussed above on the effects of rAAV integration into host chromosomal DNA also implied that rAAV preferentially integrates into actively transcribed
genes in cultured cells (58). These results have been confirmed and extended in
a recent report demonstrating preferential integration into active genes in mouse
liver tissue (60). Of the integration sites examined, 21 of 29 (72%) were in genes,
and all of these genes were actively expressed in hepatocytes. These events were
accompanied by deletions of up to 2 kb, though smaller deletions predominated.
Safety of rAAV Vectors
The possibility that rAAV integration preferentially targets transcribed genes raises
implications for its use as a clinical gene therapy vector, particularly in light
of instances of oncogenic transformation in a recent clinical trial as a result of
retrovirus insertional activation (22). However, these concerns should be tempered
by considering the underlying mechanisms of this phenomenon and factors that
contribute to it (78). First, the preferential integration into active regions of the
chromosome and transcribed genes is not unique to rAAV vectors. Vectors derived
from HIV are also prone to integrate into active genes, though it is not clear
to what extent this is due to the local chromatin structure or to interaction with
transcription factors (86). A preference toward transcribed regions was also noted
in the characterization of adenovirus integration, even though this is generally
considered to be a nonintegrating virus (87). Thus, exogenous DNA delivered by
any means may carry a similar risk.
A second factor that must be considered in evaluating the safety of rAAV with
respect to disruption of active genes is the absolute integration frequency of the
vector in the target tissue. As discussed above, this frequency is low in liver tissue
and not yet detected in muscle tissue. Further, the instances of integration in liver
tissue were all recovered from tissue that had been induced to regenerate through
injury, which may affect the interaction between vector and host genomes.
Finally, the relationship between actively transcribed regions and DNA damage
repair is well established, and up to one third of DSBs in transcribed regions result
in mutations (1). Thus, if the suggestion that rAAV integrates as a bystander to
DNA damage repair is correct, it is not clear that the integration event would cause
significantly more damage than would have occurred in that cell in any event, at
least in terms of gene inactivation. On the other hand, the possibility of insertional
activation of an oncogene, as seen with retroviruses, must also be considered.
Unlike retroviruses, however, neither rAAV nor its parent virus has ever been
associated with malignancies in humans or animals, or with transformation of
cells. Insertional activation from retrovirus is associated with the strong promoter
activity of its downstream long terminal repeat. Although there is a weak promoter
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
activity associated with the rAAV TR, due to its position at the A/D sequence
junction, it is likely limited to transcription inward through vector sequence rather
than outward into chromosomal DNA (17, 21).
Although the safety of rAAV gene therapy will ultimately be determined in
controlled clinical trials, the theoretical risks and their underlying mechanisms
will continue to be the subject of active research. Foremost among the questions
to be answered will be the actual integration frequency in any given tissue or
cell type, under growth conditions that reflect the clinical situation. Additionally,
an assessment of the consequences of rAAV integration-mediated mutagenesis in
vivo, in terms of cell or tissue damage or transformation, will provide a useful
framework for the evaluation of rAAV vector safety.
The Annual Review of Genetics is online at
1. Allen C, Miller CA, Nickoloff JA. 2003.
The mutagenic potential of a single DNA
double-strand break in a mammalian chromosome is not influenced by transcription. DNA Repair 2:1147–56
2. Amiss TJ, McCarty DM, Skulimowski
A, Samulski RJ. 2003. Identification and
characterization of an adeno-associated
virus integration site in CV-1 cells from
the African green monkey. J. Virol.
3. Bakowska JC, Di Maria MV, Camp SM,
Wang Y, Allen PD, Breakefield XO. 2003.
Targeted transgene integration into transgenic mouse fibroblasts carrying the fulllength human AAVS1 locus mediated by
HSV/AAV rep(+) hybrid amplicon vector. Gene Ther. 10:1691–702
4. Balague C, Kalla M, Zhang WW. 1997.
Adeno-associated virus Rep78 protein
and terminal repeats enhance integration of DNA sequences into the cellular
genome. J. Virol. 71:3299–306
5. Batchu RB, Hermonat PL. 1995. The
trans-inhibitory Rep78 protein of adenoassociated virus binds to TAR region DNA
of the human immunodeficiency virus
type 1 long terminal repeat. FEBS Lett.
6. Berns KI, Cheung A, Ostrove J, Lewis M.
1982. Adeno-associated virus latent infection. In Virus Persistance, ed. BWJ Mahy,
AC Minson, GK Darby, pp. 249–65. Cambridge: Cambridge Univ. Press
Brister JR, Muzyczka N. 2000. Mechanism of Rep-mediated adeno-associated
virus origin nicking. J. Virol. 74:7762–71
Chang L-S, Shenk T. 1990. The adenovirus DNA binding protein stimulates
the rate of transcription directed by adenovirus and adeno-associated virus promoters. J. Virol. 64:2103–9
Chang L-S, Shi Y, Shenk T. 1989. Adenoassociated virus p5 promoter contains an
adenovirus EIA inducible element and a
binding site for the major late transcription factor. J. Virol. 63:3479–88
Cheung AK, Hoggan MD, Hauswirth
WW, Berns KI. 1980. Integration of the
adeno-associated virus genome into cellular DNA in latently infected human Detroit 6 cells. J. Virol. 33:739–48
Chiorini JA, Yang L, Safer B, Kotin RM.
1995. Determination of adeno-associated
virus Rep68 and Rep78 binding sites by
random sequence oligonucleotide selection. J. Virol. 69:7334–38
Duan D, Sharma P, Yang J, Yue Y, Dudus
L, et al. 1998. Circular intermediates of
recombinant adeno-associated virus have
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
defined structural characteristics responsible for long-term episomal persistence
in muscle tissue. J. Virol. 72:8568–77. Erratum. 1999. J. Virol. 73(1):861
Duan D, Yan Z, Yue Y, Engelhardt
JF. 1999. Structural analysis of adenoassociated virus transduction circular intermediates. Virology 261:8–14
Dutheil N, Shi F, Dupressoir T, Linden RM. 2000. Adeno-associated virus
site-specifically integrates into a musclespecific DNA region. Proc. Natl. Acad.
Sci. USA 97:4862–66
Dyall J, Szabo P, Berns KI. 1999. Adenoassociated virus (AAV) site-specific integration: formation of AAV-AAVS1 junctions in an in vitro system. Proc. Natl.
Acad. Sci. USA 96:12849–54
Fisher KJ, Jooss K, Alston J, Yang Y,
Haecker SE, et al. 1997. Recombinant
adeno-associated virus for muscle directed gene therapy. Nat. Med. 3:306–12
Flotte TR, Afione SA, Solow R, Drumm
ML, Markakis D, et al. 1993. Expression
of the cystic fibrosis transmembrane conductance regulator from a novel adenoassociated virus promoter. J. Biol. Chem.
Flotte TR, Afione SA, Zeitlin PL. 1994.
Adeno-associated virus vector gene expression occurs in nondividing cells in the
absence of vector DNA integration. Am. J.
Respir. Cell Mol. Biol. 11:517–21
Giraud C, Winocour E, Berns KI.
1994. Site-specific integration by adenoassociated virus is directed by a cellular DNA sequence. Proc. Natl. Acad. Sci.
USA 91:10039–43
Giraud C, Winocour E, Berns KI. 1995.
Recombinant junctions formed by sitespecific integration of adeno-associated
virus into an episome. J. Virol. 69:6917–
Haberman RP, McCown TJ, Samulski
RJ. 2000. Novel transcriptional regulatory
signals in the adeno-associated virus terminal repeat A/D junction element. J. Virol. 74:8732–39
22. Hacein-Bey-Abina S, Von Kalle C,
Schmidt M, McCormack MP, Wulffraat
N, et al. 2003. LMO2-associated clonal T
cell proliferation in two patients after gene
therapy for SCID-X1. Science 302:415–
23. Harui A, Suzuki S, Kochanek S, Mitani K.
1999. Frequency and stability of chromosomal integration of adenovirus vectors.
J. Virol. 73:6141–46
24. Hermonat PL, Muzyczka N. 1984. Use
of adeno-associated virus as a mammalian DNA cloning vector: transduction
of neomycin resistance into mammalian
tissue culture cells. Proc. Natl. Acad. Sci.
USA 81:6466–70
25. Hernandez YJ, Wang J, Kearns WG,
Loiler S, Poirier A, Flotte TR. 1999. Latent adeno-associated virus infection elicits humoral but not cell-mediated immune
responses in a nonhuman primate model.
J. Virol. 73:8549–58
26. Hoggan MD. 1970. Adeno-associated
viruses. Prog. Med. Virol. 12:211–39
27. Hoggan MD, Blacklow NR, Rowe WP.
1966. Studies of small DNA viruses found
in various adenovirus preparations: physical, biological, and immunological characteristics. Proc. Natl. Acad. Sci. USA
28. Hoggan MD, Thomas GF, Thomas FB,
Johnson FB. 1972. Continuous carriage
of adenovirus associated virus genome in
cell culture in the absence of helper adenovirus. Presented at Proc. Lepetite Colloquium, 4th, Cocoyac, Mexico
29. Huser D, Heilbronn R. 2003. Adenoassociated virus integrates site-specifically into human chromosome 19 in either orientation and with equal kinetics
and frequency. J. Gen. Virol. 84:133–37
30. Huser D, Weger S, Heilbronn R. 2002. Kinetics and frequency of adeno-associated
virus site-specific integration into human
chromosome 19 monitored by quantitative real-time PCR. J. Virol. 76:7554–59
31. Huttner NA, Girod A, Schnittger S,
Schoch C, Hallek M, Buning H. 2003.
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
Analysis of site-specific transgene integration following cotransduction with recombinant adeno-associated virus and a
rep encoding plasmid. J. Gene Med. 5:
Im D-S, Muzyczka N. 1989. Factors that
bind to the AAV terminal repeats. J. Virol.
Im D-S, Muzyczka N. 1990. The AAV
origin binding protein Rep68 is an ATPdependent site-specific endonuclease with
DNA helicase activity. Cell 61:447–57
Inoue N, Dong R, Hirata RK, Russell DW.
2001. Introduction of single base substitutions at homologous chromosomal sequences by adeno-associated virus vectors. Mol. Ther. 3:526–30
Inoue N, Hirata RK, Russell DW. 1999.
High-fidelity correction of mutations
at multiple chromosomal positions by
adeno-associated virus vectors. J. Virol.
Johnston KM, Jacoby D, Pechan PA, Fraefel C, Borghesani P, et al. 1997. HSV/AAV
hybrid amplicon vectors extend transgene
expression in human glioma cells. Hum.
Gene Ther. 8:359–70
Kearns WG, Afione SA, Fulmer SB, Pang
MC, Erikson D, et al. 1996. Recombinant
adeno-associated virus (AAV-CFTR) vectors do not integrate in a site-specific fashion in an immortalized epithelial cell line.
Gene Ther. 3:748–55
Kogure K, Urabe M, Mizukami H, Kume
A, Sato Y, et al. 2001. Targeted integration of foreign DNA into a defined locus
on chromosome 19 in K562 cells using
AAV-derived components. Int. J. Hematol. 73:469–75
Kokorina NA, Santin AD, Li C, Hermonat PL. 1998. Involvement of proteinDNA interaction in adeno-associated
virus Rep78-mediated inhibition of HIV1. J. Hum. Virol. 1:441–50
Kotin RM, Berns KI. 1989. Organization of adeno-associated virus DNA in
latently infected Detroit 6 cells. Virology
41. Kotin RM, Linden RM, Berns KI. 1992.
Characterization of a preferred site on
human chromosome 19q for integration
of adeno-associated virus DNA by nonhomologous recombination. EMBO J. 11:
42. Kotin RM, Siniscalco M, Samulski RJ,
Zhu XD, Hunter L, et al. 1990. Sitespecific integration by adeno-associated
virus. Proc. Natl. Acad. Sci. USA 87:
43. Lackner DF, Muzyczka N. 2002. Studies
of the mechanism of transactivation of the
adeno-associated virus p19 promoter by
Rep protein. J. Virol. 76:8225–35
44. Lam P, Hui KM, Wang Y, Allen PD, Louis
DN, et al. 2002. Dynamics of transgene
expression in human glioblastoma cells
mediated by herpes simplex virus/adenoassociated virus amplicon vectors. Hum.
Gene Ther. 13:2147–59
45. Lamartina S, Roscilli G, Rinaudo D, Delmastro P, Toniatti C. 1998. Lipofection
of purified adeno-associated virus Rep68
protein: toward a chromosome-targeting
nonviral particle. J. Virol. 72:7653–58
46. Lamartina S, Sporeno E, Fattori E, Toniatti C. 2000. Characteristics of the adenoassociated virus preintegration site in human chromosome 19: open chromatin
conformation and transcription-competent environment. J. Virol. 74:7671–77
47. Linden RM, Ward P, Giraud C, Winocour
E, Berns KI. 1996. Site-specific integration by adeno-associated virus. Proc. Natl.
Acad. Sci. USA 93:11288–94
48. Linden RM, Winocour E, Berns KI. 1996.
The recombination signals for adenoassociated virus site-specific integration.
Proc. Natl. Acad. Sci. USA 93:7966–72
49. McCarty DM, Pereira DJ, Zolotukhin I,
Zhou X, Ryan JH, Muzyczka N. 1994.
Identification of linear DNA sequences
that specifically bind the adeno-associated
virus Rep protein. J. Virol. 68:4988–97
50. McCarty DM, Ryan JH, Zolotukhin S,
Zhou X, Muzyczka N. 1994. Interaction of the adeno-associated virus Rep
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
protein with a sequence within the A
palindrome of the viral terminal repeat.
J. Virol. 68:4998–5006
McCarty DM, Samulski RJ. 1997. Adenoassociated viral vectors. In Concepts in
Gene Therapy, ed. M Strauss, JA Barranger, pp. 61–78. Berlin: Walter de
McLaughlin SK, Collis P, Hermonat PL,
Muzyczka N. 1988. Adeno-associated
virus general transduction vectors: analysis of proviral structures. J. Virol. 62:
Mendelson E, Miller IL, Carter BJ. 1988.
Expression and rescue of a nonselected
marker from an integrated AAV vector.
Virology 166:154–65
Mendelson E, Smith MG, Miller IL,
Carter BJ. 1988. Effect of a viral rep
gene on transformation of cells by an
adeno-associated virus vector. Virology
Miao CH, Nakai H, Thompson AR,
Storm TA, Chiu W, et al. 2000. Nonrandom transduction of recombinant adenoassociated virus vectors in mouse hepatocytes in vivo: Cell cycling does not influence hepatocyte transduction. J. Virol. 74:
Miao CH, Snyder RO, Schowalter DB,
Patijn GA, Donahue B, et al. 1998. The
kinetics of rAAV integration in the liver
[letter]. Nat. Genet. 19:13–15
Miller DG, Petek LM, Russell DW.
2004. Adeno-associated virus vectors integrate at chromosome breakage sites.
Nat. Genet. 36:767–73
Miller DG, Rutledge EA, Russell DW.
2002. Chromosomal effects of adenoassociated virus vector integration. Nat.
Genet. 30:147–48
Nakai H, Iwaki Y, Kay MA, Couto LB.
1999. Isolation of recombinant adenoassociated virus vector-cellular DNA
junctions from mouse liver. J. Virol. 73:
Nakai H, Montini E, Fuess S, Storm TA,
Grompe M, Kay MA. 2003. AAV serotype
2 vectors preferentially integrate into active genes in mice. Nat. Genet. 34:297–
Nakai H, Montini E, Fuess S, Storm TA,
Meuse L, et al. 2003. Helper-independent
and AAV-ITR-independent chromosomal integration of double-stranded linear
DNA vectors in mice. Mol. Ther. 7:101–
Nakai H, Storm TA, Kay MA. 2000. Recruitment of single-stranded recombinant
adeno-associated virus vector genomes
and intermolecular recombination are responsible for stable transduction of liver
in vivo. J. Virol. 74:9451–63
Nakai H, Yant SR, Storm TA, Fuess S,
Meuse L, Kay MA. 2001. Extrachromosomal recombinant adeno-associated virus
vector genomes are primarily responsible
for stable liver transduction in vivo. J. Virol. 75:6969–76
Ni TH, Zhou X, McCarty DM, Zolotukhin
I, Muzyczka N. 1994. In vitro replication
of adeno-associated virus DNA. J. Virol.
Oelze I, Rittner K, Sczakiel G. 1994.
Adeno-associated virus type 2 rep genemediated inhibition of basal gene expression of human immunodeficiency virus
type 1 involves its negative regulatory
functions. J. Virol. 68:1229–33
Ogata T, Kozuka T, Kanda T. 2003. Identification of an insulator in AAVS1, a preferred region for integration of adenoassociated virus DNA. J. Virol. 77:9000–
Pereira DJ, McCarty DM, Muzyczka N.
1997. The adeno-associated virus (AAV)
Rep protein acts as both a repressor and
an activator to regulate AAV transcription during a productive infection. J. Virol.
Philpott NJ, Giraud-Wali C, Dupuis C,
Gomos J, Hamilton H, et al. 2002. Efficient integration of recombinant adenoassociated virus DNA vectors requires a
p5-rep sequence in cis. J. Virol. 76:5411–
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
69. Philpott NJ, Gomos J, Berns KI, FalckPedersen E. 2002. A p5 integration efficiency element mediates Rep-dependent
integration into AAVS1 at chromosome
19. Proc. Natl. Acad. Sci. USA 99:12381–
70. Pieroni L, Fipaldini C, Monciotti A, Cimini D, Sgura A, et al. 1998. Targeted integration of adeno-associated virus-derived
plasmids in transfected human cells. Virology 249:249–59
71. Ponnazhagan S, Erikson D, Kearns WG,
Zhou SZ, Nahreini P, et al. 1997. Lack
of site-specific integration of the recombinant adeno-associated virus 2 genomes
in human cells. Hum. Gene Ther. 8:275–
72. Porteus MH, Baltimore D. 2003. Chimeric nucleases stimulate gene targeting in
human cells. Science 300:763
73. Porteus MH, Cathomen T, Weitzman MD,
Baltimore D. 2003. Efficient gene targeting mediated by adeno-associated virus
and DNA double-strand breaks. Mol. Cell
Biol. 23:3558–65
74. Recchia A, Parks RJ, Lamartina S, Toniatti C, Pieroni L, et al. 1999. Sitespecific integration mediated by a hybrid
adenovirus/adeno-associated virus vector.
Proc. Natl. Acad. Sci. USA 96:2615–20
75. Rinaudo D, Lamartina S, Roscilli G,
Ciliberto G, Toniatti C. 2000. Conditional
site-specific integration into human chromosome 19 by using a ligand-dependent
chimeric adeno-associated virus/Rep protein. J. Virol. 74:281–94
76. Rittner K, Heilbronn R, Kleinschmidt
JA, Sczakiel G. 1992. Adeno-associated
virus type 2-mediated inhibition of
human immunodeficiency virus type
1 (HIV-1) replication: involvement of
p78rep/p68rep and the HIV-1 long terminal repeat. J. Gen. Virol. 73:2977–81
77. Rizzuto G, Gorgoni B, Cappelletti M,
Lazzaro D, Gloaguen I, et al. 1999. Development of animal models for adenoassociated virus site-specific integration.
J. Virol. 73:2517–26
78. Russell DW. 2003. AAV loves an active
genome. Nat. Genet. 34:241–42
79. Russell DW, Hirata RK. 1998. Human
gene targeting by viral vectors. Nat.
Genet. 18:325–30
80. Rutledge EA, Russell DW. 1997. Adenoassociated virus vector integration junctions. J. Virol. 71:8429–36
81. Ryan JH, Zolotukhin S, Muzyczka N.
1996. Sequence requirements for binding
of Rep68 to the adeno-associated virus
terminal repeats. J. Virol. 70:1542–53
82. Samulski RJ, Chang LS, Shenk T. 1989.
Helper-free stocks of recombinant adenoassociated viruses: Normal integration
does not require viral gene expression. J.
Virol. 63:3822–28
83. Samulski RJ, Zhu X, Xiao X, Brook JD,
Housman DE, et al. 1991. Targeted integration of adeno-associated virus (AAV)
into human chromosome 19. EMBO J.
10:3941–50. Erratum. 1992. EMBO J.
84. Satoh W, Hirai Y, Tamayose K, Shimada T. 2000. Site-specific integration of
an adeno-associated virus vector plasmid
mediated by regulated expression of rep
based on Cre-loxP recombination. J. Virol. 74:10631–38
85. Schnepp BC, Clark KR, Klemanski DL,
Pacak CA, Johnson PR. 2003. Genetic fate
of recombinant adeno-associated virus
vector genomes in muscle. J. Virol. 77:
86. Schroder AR, Shinn P, Chen H, Berry C,
Ecker JR, Bushman F. 2002. HIV-1 integration in the human genome favors active
genes and local hotspots. Cell 110:521–29
87. Schulz M, Freisem-Rabien U, Jessberger
R, Doerfler W. 1987. Transcriptional activities of mammalian genomes at sites of
recombination with foreign DNA. J. Virol.
88. Snyder RO, Im D-S, Muzyczka N. 1990.
Evidence for covalent attachment of the
adeno-associated virus Rep protein to the
ends of the AAV genome. J. Virol. 64:
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
89. Snyder RO, Im DS, Ni T, Xiao X, Samulski RJ, Muzyczka N. 1993. Features of the
adeno-associated virus origin involved in
substrate recognition by the viral Rep protein. J. Virol. 67:6096–104
90. Snyder RO, Samulski RJ, Muzyczka
N. 1990. In vitro resolution of covalently joined AAV chromosome ends. Cell
91. Snyder RO, Spratt SK, Lagarde C, Bohl D,
Kaspar B, et al. 1997. Efficient and stable
adeno-associated virus-mediated transduction in the skeletal muscle of adult immunocompetent mice. Hum. Gene Ther.
92. Sun L, Li J, Xiao X. 2000. Overcoming
adeno-associated virus vector size limitation through viral DNA heterodimerization. Nat. Med. 6:599–602
93. Surosky RT, Urabe M, Godwin SG, McQuiston SA, Kurtzman GJ, et al. 1997.
Adeno-associated virus Rep proteins target DNA sequences to a unique locus in
the human genome. J. Virol. 71:7951–
94. Tratschin JD, Miller IL, Smith MG, Carter
BJ. 1985. Adeno-associated virus vector for high-frequency integration, expression, and rescue of genes in mammalian
cells. Mol. Cell Biol. 5:3251–60
95. Urcelay E, Ward P, Wiener Sm, Safer B,
Kotin Rm . 1995. Asymmetric replication
in vitro from a human sequence element is
dependent on adeno-associated virus Rep
protein. J. Virol. 69:2038–46
96. Vincent-Lacaze N, Snyder RO, Gluzman
R, Bohl D, Lagarde C, Danos O. 1999.
Structure of adeno-associated virus vector
DNA following transduction of the skeletal muscle. J. Virol. 73:1949–55
97. Wang Y, Camp SM, Niwano M, Shen X,
Bakowska JC, et al. 2002. Herpes simplex virus type 1/adeno-associated virus
rep(+) hybrid amplicon vector improves
the stability of transgene expression in human cells by site-specific integration. J.
Virol. 76:7150–62
98. Weitzman MD, Kyostio SR, Kotin RM,
Owens RA. 1994. Adeno-associated virus
(AAV) Rep proteins mediate complex formation between AAV DNA and its integration site in human DNA. Proc. Natl.
Acad. Sci. USA 91:5808–12
Wonderling RS, Owens RA. 1997. Binding sites for adeno-associated virus Rep
proteins within the human genome. J. Virol. 71:2528–34
Xiao X, Li J, Samulski RJ. 1996. Efficient
long-term gene transfer into muscle tissue of immunocompetent mice by adenoassociated virus vector. J. Virol. 70:8098–
Xiao X, Xiao W, Li J, Samulski RJ. 1997.
A novel 165-base-pair terminal repeat sequence is the sole cis requirement for the
adeno-associated virus life cycle. J. Virol.
Yalkinoglu AO, Heilbronn R, Burkle A,
Schlehofer JR, zur-Hausen H. 1988. DNA
amplification of adeno-associated virus as
a response to cellular genotoxic stress.
Cancer Res. 48:3123–32
Yalkinoglu AO, Zentgraf H, Hubscher U.
1991. Origin of adeno-associated virus
DNA replication is a target of carcinogeninducible DNA amplification. J. Virol.
Yan Z, Zhang Y, Duan D, Engelhardt JF.
2000. Trans-splicing vectors expand the
utility of adeno-associated virus for gene
therapy. Proc. Natl. Acad. Sci. USA 97:
Yang CC, Xiao X, Zhu X, Ansardi DC,
Epstein ND, et al. 1997. Cellular recombination pathways and viral terminal repeat
hairpin structures are sufficient for adenoassociated virus integration in vivo and in
vitro. J. Virol. 71:9231–47
Yang J, Zhou W, Zhang Y, Zidon T,
Ritchie T, Engelhardt JF. 1999. Concatamerization of adeno-associated virus
circular genomes occurs through intermolecular recombination. J. Virol. 73:
Young SM Jr, McCarty DM, Degtyareva N, Samulski RJ. 2000. Roles of
9 Oct 2004 0:15
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
adeno-associated virus Rep protein and
human chromosome 19 in site-specific recombination. J. Virol. 74:3953–66
108. Young SM Jr, Samulski RJ. 2001. Adenoassociated virus (AAV) site-specific recombination does not require a Repdependent origin of replication within the
AAV terminal repeat. Proc. Natl. Acad.
Sci. USA 98:13525–30
109. Yue Y, Duan D. 2003. Double strand
interaction is the predominant pathway for intermolecular recombination of
adeno-associated viral genomes. Virology
October 18, 2004
Annual Reviews
Annual Review of Genetics
Volume 38, 2004
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
MOBILE GROUP II INTRONS, Alan M. Lambowitz and Steven Zimmerly
CELLS, Stephen P. Goff
Joanne Chory, and Christian Fankhauser
Arthur R. Grossman, Martin Lohr, and Chung Soon Im
THE GENETICS OF GEOCHEMISTRY, Laura R. Croal, Jeffrey A. Gralnick,
Davin Malasarn, and Dianne K. Newman
ITS REGULATION, Frank Stegmeier and Angelika Amon
and Lorraine S. Symington
John Tower
Angela Taddei, Florence Hediger, Frank R. Neumann,
and Susan M. Gasser
and Jonathan J. Ewbank
NETWORK, Eirı́kur Steingrı́msson, Neal G. Copeland,
and Nancy A. Jenkins
AND TRITHORAX GROUP PROTEINS, Leonie Ringrose and Renato Paro
and Tomas Lindahl
MITOCHONDRIA OF PROTISTS, Michael W. Gray, B. Franz Lang,
and Gertraud Burger
October 18, 2004
Annual Reviews
Christian S. Riesenfeld, Patrick D. Schloss, and Jo Handelsman
David Haig
Viola Willemsen and Ben Scheres
Annu. Rev. Genet. 2004.38:819-845. Downloaded from
by Medical Library of the Chinese PLA on 02/25/08. For personal use only.
AND OLD IDEAS, John S. Taylor and Jeroen Raes
Hendrik Poinar, David Serre, Viviane Jaenicke-Despres, Juliane Hebler,
Nadin Rohland, Melanie Kuch, Johannes Krause, Linda Vigilant,
and Michael Hofreiter
Reed B. Wickner, Herman K. Edskes, Eric D. Ross, Michael M. Pierce,
Ulrich Baxa, Andreas Brachmann, and Frank Shewmaker
Tim Clausen
and Hiten D. Madhani
Tamas Gaal, and Richard L. Gourse
Stephen D. Bentley and Julian Parkhill
Robert Swanson, Anna F. Edlund, and Daphne Preuss
Samuel M. Young Jr., and Richard J. Samulski
Subject Index
An online log of corrections to Annual Review of Genetics chapters
may be found at