Novel adeno-associated viruses from rhesus monkeys
as vectors for human gene therapy
Guang-Ping Gao, Mauricio R. Alvira, Lili Wang, Roberto Calcedo, Julie Johnston, and James M. Wilson*
Institute for Human Gene Therapy and Department of Medicine, University of Pennsylvania, Philadelphia, PA 19104; and Wistar Institute, 3601 Spruce
Street, Philadelphia, PA 19104
Communicated by Thomas E. Shenk, Princeton University, Princeton, NJ, July 10, 2002 (received for review March 21, 2002)
Tissues from rhesus monkeys were screened by PCR for the presence
of sequences homologous to known adeno-associated virus (AAV)
serotypes 1–6. DNA spanning entire rep-cap ORFs from two novel
AAVs, called AAV7 and AAV8, were isolated. Sequence comparisons
among these and previously described AAVs revealed the greatest
divergence in capsid proteins. AAV7 and AAV8 were not neutralized
by heterologous antisera raised to the other serotypes. Neutralizing
antibodies to AAV7 and AAV8 were rare in human serum and, when
(ITRs) from AAV2 and were compared with similarly constructed
of skeletal muscle and liver-directed gene transfer were used to
evaluate relative vector performance. AAV7 vectors demonstrated
efficiencies of transgene expression in skeletal muscle equivalent to
that observed with AAV1, the most efficient known serotype for this
application. In liver, transgene expression was 10- to 100-fold higher
with AAV8 than observed with other serotypes. This improved
efficiency correlated with increased persistence of vector DNA and
higher number of transduced hepatocytes. The efficiency of AAV8
vector for liver-directed gene transfer of factor IX was not impacted
by preimmunization with the other AAV serotypes. Vectors based on
these novel, nonhuman primate AAVs should be considered for
human gene therapy because of low reactivity to antibodies directed
to human AAVs and because gene transfer efficiency in muscle was
similar to that obtained with the best known serotype, whereas,
in liver, gene transfer was substantially higher than previously
Parvoviridae family and require helper viruses such as adenovirus
to replicate. Six primate AAVs have been isolated, and five have
been determined to be distinct serotypes based on antibody cross-
reactivity studies (2–8). AAV6 appears to be a recombinant
between AAV1 and AAV2 (9). All primate AAVs were isolated
initially as contaminants in preparations of adenoviruses except for
AAV5, which was recovered from a human condylomatous wart
(2–8). Seroepidemiologic studies indicate that AAV serotypes 2, 3,
and 5 are endemic to humans whereas AAV4 primarily infects
nonhuman primates (2–8, 10). The reservoir for AAV1 (and the
associated AAV6 species) is unclear because it has not been
primarily isolated from tissues and reactive antibodies exist in both
humans and nonhuman primates (10–13).
of in vivo gene therapy. Several themes have emerged from these
studies. In tissues such as liver, muscle, retina and the central
nervous system, AAV2-mediated gene transfer confers extremely
stable transgene expression (15). Recipient animals do not elicit T
cell responses to most AAV-encoded transgene products, even if
these proteins contain foreign epitopes. This phenomenon is be-
cells (16). The AAV genome appears to persist in a number of
deno-associated viruses (AAV) have been isolated from a
number of species, including primates (1). They belong to the
different integrated and nonintegrated forms after in vivo gene
Despite the impressive longevity of transgene expression
obtained with AAV2, its application has been limited because of
low levels of transgene expression. Blocks at the level of vector
entry and post entry processing contribute to these inefficiencies
(20, 21). Progress in overcoming these barriers has been made
through the development of vectors based on other serotypes
that enter the cell via receptors distinct from those that recog-
nize AAV2. We showed that AAV1 vectors very efficiently
transduce skeletal muscle (9) and retina (22) whereas others
demonstrated high-level transduction of the central nervous
system and lung with AAV5 vectors (23, 24).
In this report, we describe the isolation of two novel AAVs,
AAV7 and AAV8, and their use as vectors for somatic gene
Isolation of AAV7 and AAV8 Sequences from Nonhuman Primate
Tissues. In an attempt to isolate novel AAV sequences from
nonhuman primate tissues, published AAV sequences including
primate AAV1–AAV6 and AAVs from duck and goose origins
were aligned for comparison by using the CLUSTAL W program. A
stretch of AAV sequence spanning 2,886 to 3,143 bp of AAV1 and
corresponding sequences in AAV2–AAV6 and AAVs from duck
and goose origins was selected as a PCR amplicon. This region is
about 255 bp in length in which both 5? and 3? sequences are highly
conserved, but the middle sequence is variable and unique to each
known AAV serotype (called the ‘‘signature region’’). A pair of
universal primers that can anneal to the 5? and 3? ends of this
signature region were designed for PCR to amplify from total
DNAs isolated from various tissues of rhesus monkeys. Sequences
for the primers are 5?-GGTAATGCCTCAGGAAATTGGCA-
TT-3? (19s) and 5?-GACTCATCAACAACAATTGGGGATTC-3?
(18as), respectively. The first novel AAV signature sequence iso-
lated, named AAV serotype 7, was isolated by PCR amplification
of heart DNA of a rhesus monkey (98E044) that was previously
treated with an adenovirus vector for study of innate immune
response (25). A second novel AAV sequence was obtained from
heart DNA of another rhesus monkey (98E056) in the same
adenovirus study, and was designated AAV serotype 8. Identical
signature sequences were isolated from several tissues of rhesus
monkeys from the facility at the University of Pennsylvania as well
as animals derived from the Primate Center at Tulane University.
To extend cap gene sequences of AAV7 and AAV8, two other
highly conserved regions were identified for use in PCR ampli-
fication. One region is ?1.7 kb upstream of the signature region
whereas the other is located 1.5 kb downstream. A primer within
Abbreviations: AAV, adeno-associated virus; ITR, inverted terminal repeat; A1AT, ?-1
CMV, cytomegalovirus early promoter; X-Gal, 5-bromo-4-chloro-3-indolyl ?-D-galactoside;
TBG, thyroid hormone binding globulin gene promoter; FIX, factor IX.
Database deposition: The sequences reported in this paper have been deposited in the
GenBank database [accession nos. AF513851 (AAV7) and AF513852 (AAV8)].
*To whom reprint requests should be addressed. E-mail: email@example.com.
September 3, 2002 ?
vol. 99 ?
the upstream conserved region was selected (AV1Ns: 5?-
gctgcgycaactggaccaatgagaac-3?) in combination with the 3?
primer (18as) in the signature region to extend AAV7 and
AAV8 sequence amplification to 5? of the cap gene and the
junction of rep?cap genes, followed by Topo cloning of 1.7-kb
fragments. Another primer located in a conserved region at the
end of the cap gene (AV2Cas: 5?-cgcagagaccaaagttcaactgaaacga-
3?) was used as 3? primer to pair with the 5? primer of the
signature region (19s) to reamplify heart DNAs of the two
monkeys, resulting in isolation of 1.5-kb fragments. Using the
overlapping 255-bp signature sequence as an anchor, the 5?
fragment of 1.7-kb fragment was fused with the 3? fragment of
1.5 kb at a DraIII site to form the fusion plasmids containing the
complete cap genes of AAV7 and AAV8. The fusion plasmids
were designated as pCRAAV7-Fusion and pCRAAV8-Fusion.
A similar strategy was used to isolate the corresponding rep
genes. The full sequences for rep and cap of AAV7 and AAV8
have been submitted to GenBank (accession numbers: AAV7,
AF513851; and AAV8, AF513852).
Directly amplifying, cloning, and sequencing full-length cap
sequences (i.e., 2.8-kb fragments) from the same monkey further
for confirmation of AAV7 are 7VRF-1s, 5?-ccttcgagcaccagcagc-
cgtt-3? and 7VRF-2as, aggctcagagtaaacaccctggctgtca, whereas
the primers for the AAV8 amplicon are 8VRF-2s, 5?-ccgcatcta-
ccgcatcctcgct-3? and 8VRF-2as, gttccaatttgaggagccgtgttttgct-3?.
Production of AAV Vectors. A pseudotyping strategy was used to
produce AAV vectors packaged with AAV7 and AAV8 capsid
proteins (26). Recombinant AAV genomes equipped with
AAV2 inverted terminal repeats (ITRs) were packaged by triple
transfection of 293 cells with cis-plasmid, adenovirus helper
plasmid and a chimeric packaging construct where the AAV2
rep gene is fused with cap genes of novel AAV serotypes. To
create the chimeric packaging constructs, the XhoI site of p5E18
plasmid at 3,169 bp was ablated, and the modified plasmid was
restricted with XbaI and XhoI in a complete digestion to remove
the AAV2 cap gene and replace it with a 2,267-bp SpeI?XhoI
fragment containing either AAV7 or AAV8 cap gene (9). A
similar cloning strategy was used for creation of chimeric
packaging plasmids of AAV2?1 and AAV2?5. All recombinant
vectors were purified by the standard CsCl2 sedimentation
method except for AAV2?2, which was purified by single step
Genome copy (GC) titers of AAV vectors were determined by
targeting SV40 poly(A) region as described previously (27).
Serologic Analysis. C57BL?6 mice were injected with vectors of
different serotypes intramuscularly (5 ? 1011GC), and serum
samples were collected 34 days later and analyzed for the presence
transduction of 84-31 cells by reporter viruses (AAVCMVEGFP)
of different serotypes (28). Serum samples were also kindly pro-
vided from J. Samulski (University of North Carolina at Chapel
Hill) from animals that received two i.p. injections of vectors of
different serotypes. Specifically, the reporter virus AAVCM-
VEGFP of each serotype (at multiplicity of infection equal to 105
genome copies?cell) was preincubated with heat-inactivated serum
from animals that were immunized against different serotypes of
added to 84-31 cells in 96-well plates for 48 or 72 h, depending on
the virus serotype. The number of green fluorescent protein-
expressing cells was assessed by fluorescent microscopy. Neutral-
In Vivo Evaluation of Different Serotypes of AAV Vectors. In this
study, seven recombinant AAV genomes, AAV2CBhA1AT,
AAV2TBGcFIX, AAV2CMVLacZ, and AAV2TBGLacZ were
packaged with capsid proteins of different serotypes. In all seven
constructs, minigene cassettes were flanked with AAV2 ITRs.
cDNAs of human ?-antitrypsin (A1AT; ref. 9), ?-subunit of
rhesus monkey choriogonadotropic hormone (CG; ref. 29),
canine factor IX (30), and bacterial ?-galactosidase (i.e., LacZ)
genes were used as reporter genes. For liver-directed gene
transfer, either mouse albumin gene promoter (Alb; ref. 9) or
human thyroid hormone binding globulin gene promoter (TBG;
ref. 30) was used to drive liver specific expression of reporter
genes. In muscle-directed gene transfer experiments, either
cytomegalovirus early promoter (CMV) or chicken ?-actin
promoter with CMV enhancer (CB) was used to direct expres-
sion of reporters.
For muscle-directed gene transfer, vectors were injected into
the right tibialis anterior of 4- to 6-wk-old NCR nude or
C57BL?6 mice (Taconic Farms). In liver-directed gene transfer
studies, vectors were infused intraportally into 7- to 9-wk-old
NCR nude or C57BL?6 mice (Taconic Farms). Serum samples
were collected retroorbitally at different time points after vector
administration. Muscle and liver tissues were harvested at dif-
ferent time points for cryosectioning and 5-bromo-4-chloro-3-
indolyl ?-D-galactoside (X-Gal) histochemical staining from
animals that received the LacZ vectors. For the readministration
experiment, C56BL?6 mice initially received AAV2?1, 2?2, 2?5,
2?7, and 2?8 vectors expressing hA1AT driven by the CB
promoter intramuscularly and followed for A1AT gene expres-
intraportally and studied for cFIX gene expression.
ELISA-based assays were performed to quantify serum levels
of hA1AT, rhesus monkey CG (rhCG), and cFIX proteins as
described previously (28–30). The experiments were completed
when animals were killed for harvest of muscle and liver tissues
for DNA extraction and quantitative analysis of genome copies
of vectors present in target tissues by TaqMan using the same set
of primers and probe as in titration of vector preparations (31).
Isolation of AAV7 and AAV8 from Nonhuman Primate Tissues. Our
strategy was to use molecular techniques to isolate novel AAV
genomes from tissues of latently infected nonhuman primates. A
DNA sequence spanning 255 bp of the cap gene was selected as
a PCR amplicon for the recovery of novel AAV genomes. This
region has highly conserved 5? and 3? sequences and a middle
variable sequence that is unique to each known serotype. The
middle region was designated the signature region.
Rhesus monkey tissues were harvested at necropsy, and DNA
was prepared and subjected to PCR amplification by using
oligonucleotides spanning the conserved sequences of the sig-
nature region. Fragments amplified by using this strategy were
cloned and sequenced. DNA representing two novel AAV
genomes, called AAV7 and AAV8, were found in multiple
tissues of several animals obtained from two different sources.
AAV7 and AAV8 genomes containing contiguous sequence
across rep and cap ORFs were obtained by amplification of
overlapping fragments using oligonucleotides to conserved re-
gions within rep and cap. This sequence has been deposited into
GenBank [accession nos. AF513851 (AAV7) and AF513852
(AAV8)]. The predicted amino acid sequences for capsid pro-
teins of AAV7 and AAV8 are compared with the other known
serotypes in Fig. 1. Comparisons of nucleotide and amino acid
sequences between AAVs 1–8 revealed divergence primarily in
capsid (Table 1). Comparisons of the capsid for AAV7 to the
other serotypes (except serotype 5, which is highly divergent)
revealed 63% to 85% identity for amino acid sequence and 68%
Gao et al.PNAS ?
September 3, 2002 ?
vol. 99 ?
no. 18 ?
to 84% identity in nucleotide sequence. Similar results were
obtained for AAV8.
Creation of Vectors Based on AAV7 and AAV8 and Serologic Analyses.
for other alternative serotypes as follows. The AAV2 capsid was
exchanged with capsid encoding AAV7 or AAV8 in an AAV2
rep-cap-packaging plasmid. This hybrid construct was cotrans-
transgene, together with a plasmid containing the necessary
adenovirus helper genes. This procedure results in the produc-
tion of chimeric virions, called AAV2?7 and AAV2?8 vectors, in
to -4 and AAV6 to -8 were aligned for comparison by using the CLUSTAL W program. AAV5 sequence was excluded from the alignment because of its strong
divergence from other serotypes. The dots in the alignment represent the amino acids that are identical to those of AAV1 VP1. The dashes indicate the amino
acids that are missing at the positions in the alignment.
Predicted amino acid sequences for capsid protein VP1 of AAV1, -2, -3A, -3B, -4, -6, -7, and -8. Amino acid sequences of capsid protein VP1 from AAV1
Table 1. Percentage of sequence similarity between VP1 coding regions of AAV serotypes 1, 2, 3, 4, 6, 7, and 8
Multiple sequence alignments of nucleic acid and amino acid of VP1 of different serotype AAVs were performed using the VECTOR NTI 6.0 (Informax, Bethesda,
MD) program. AAV5 sequence was excluded from the alignment because of its strong divergence from other serotypes. Bold numbers represent similarity in
amino acid sequences whereas light face numbers indicate similarity in nucleic acid sequences. The underlined numbers represent similarity in both.
www.pnas.org?cgi?doi?10.1073?pnas.182412299 Gao et al.
which AAV2 vector genomes are packaged into heterologous
capsids. Similar constructs were created for AAV1 (2?1), AAV5
(2?5), and AAV2 (2?2).
Recombinant virions were recovered by cesium chloride sed-
imentation in all cases except AAV2?2, which was purified by
heparin chromatography. Vectors were constructed for each
serotype for a number of in vitro and in vivo studies. Eight
different transgene cassettes were incorporated into the vectors,
and recombinant virions were produced for each serotype. The
recovery of virus, based on genome copies, is summarized in
Table 2. The yields of vector were high for each serotype, with
no consistent differences between serotypes. Western blot anal-
ysis indicated that the ratio of capsid protein to genome copies
was constant for all vector serotypes (data not shown). The
relative transduction efficiency of the different serotypes on
293-based cells expressing E4 of Ad5 measured at GC?LacZ
forming units varied as follows: AAV2 (6 ? 102), AAV2?1
(1 ? 104), AAV2?5 (3 ? 104), AAV2?7 (2 ? 104), and AAV2?8
(3 ? 104).
The availability of green fluorescent protein-expressing vec-
tors simplified the development of an assay for neutralizing
antibodies that was based on inhibition of transduction in a
permissive cell line (i.e., 293 cells stably expressing E4 from
Ad5). Antisera to all AAV serotypes (i.e., AAV1 to -8) were
generated by intramuscular and?or i.p. injection of the recom-
binant viruses. The only antiserum that neutralized AAV7 was
generated to AAV7, and AAV8 antiserum was the only one to
neutralize AAV8 (Table 3). This result supports the proposal
to designate AAV7 and AAV8 as new serotypes.
Human sera from 52 normal subjects were screened for
neutralization against selected serotypes. A low level of neutral-
izing activity (i.e., ?1:20) was detected in 3 of 52 subjects to
AAV7 and 3 of 52 to AAV8, whereas substantially higher titer
were observed in up to 20% of human subjects for AAV2. High
titer neutralizing antibody to these serotypes was detected in
rhesus monkey colonies that in the aggregate was 30% for AAV7
and 45% for AAV8.
In Vivo Performance of AAV Vectors. The performance of vectors
based on the new serotypes were evaluated in murine models of
muscle and liver-directed gene transfer and compared with
vectors based on the known serotypes AAV1, AAV2, and
AAV5. Vectors expressing secreted proteins (A1AT, Fig. 2; and
CG, Table 4) were used to quantitate relative transduction
efficiencies between different serotypes through ELISA analysis
histochemistry (Fig. 3).
The performance of AAV vectors in skeletal muscle was
analyzed after direct injection into the tibialis anterior muscles.
Vectors contained the same AAV2-based genome, with the
immediate early gene of CMV or a CMV-enhanced ?-actin
promoter driving expression of the transgene. A time course of
expression of A1AT in nu?nu NCR and C57BL?6 mice is shown
in Fig. 2. Previous studies indicated that immune competent
C57BL?6 mice elicit limited humoral responses to the human
A1AT protein when expressed from AAV vectors (9). In each
strain, AAV2?1 vector produced the highest levels of A1AT and
AAV2?2 vector the lowest, with AAV2?7 and AAV2?8 vectors
showing intermediate levels of expression. Peak levels of CG at
28 days after injection of nu?nu NCR mice showed the highest
levels from AAV2?7 and the lowest from AAV2?2 with
AAV2?8 and AAV2?1 in between. Injection of AAV2?1 and
AAV2?7 LacZ vectors yielded gene expression at the injection
sites in all muscle fibers with substantially fewer LacZ positive
fibers observed with AAV2?2 and AAV 2?8 vectors (Fig. 3).
These data indicate that the efficiency of transduction with
AAV2?7 vectors in skeletal muscle is similar to that obtained
with AAV2?1, which is the most efficient in skeletal muscle of
the previously described serotypes (9, 32, 33).
Similar murine models were used to evaluate liver-directed
gene transfer. Identical doses of vector based on genome copies
murine skeletal muscle and liver. Immune deficient (NCR nudes, A and C) and
immune competent mice (C57BL?6, B and D) were treated with different
serotypes of AAVCBA1AT (A and B) or AAVAlbA1AT (C and D) through
intramuscular or intraportal injections, respectively, at a dose of 1 ? 1011GC
per animal. Serum samples were collected retroorbitally at different time
points post vector administration and assayed for A1AT concentration. The y
axis represents A1AT concentration in the unit of nanogram per milliliter of
serum. The x axis represents time points of different bleeds. The data were
A1AT level in the control animals that received PBS was under the detection
limit of the assay.
Table 2. Production of recombinant vectors
AAV2?1AAV2?2 AAV2?5AAV2?7 AAV2?8
7.30 ? 4.33 (n ? 9)
6.43 ? 2.42 (n ? 2)
4.18 (n ? 1)
4.67 ? 0.75 (n ? 2)
0.567 (n ? 1)
8.78 ? 2.37 (n ? 7)
8.51 ? 6.65 (n ? 6)
1.24 ? 1.29 (n ? 3)
4.49 ? 2.89 (n ? 6)
3.39 ? 2.42 (n ? 2)
0.23 (n ? 1)
4.77 (n ? 1)
0.438 (n ? 1)
1.43 ? 1.18 (n ? 2)
3.47 ? 2.09 (n ? 5)
0.63 ? 0.394 (n ? 6)
5.19 ? 5.19 (n ? 8)
5.55 ? 6.49 (n ? 4)
0.704 ? 0.43 (n ? 2)
4.09 (n ? 1)
2.82 (n ? 1)
1.63 ? 1.15 (n ? 3)
5.26 ? 3.85 (n ? 4)
3.74 ? 2.48 (n ? 7)
3.42 (n ? 1)0.87 (n ? 1)
2.98 ? 2.66 (n ? 2)
2.16 (n ? 1)
5.04 (n ? 1)
2.78 (n ? 1)
2.04 ? 0.67 (n ? 3)
6.52 ? 3.08 (n ? 4)
4.05 (n ? 1)
3.74 ? 3.88 (n ? 2)
0.532 (n ? 1)
2.02 (n ? 1)
0.816 ? 0.679 (n ? 2)
1.32 ? 0.87 (n ? 3)
1.83 ? 0.98 (n ? 5)
15.8 ? 15.0 (n ? 5)
Data presented in the table are average genome copy yields with standard deviation ? 1013of multiple production lots of 50 plate (150 mm) transfections.
All recombinant vectors were purified by the standard CsCl2sedimentation method except for AAV2?2, which was purified by single-step heparin chromatog-
raphy. EGFP, enhanced green fluorescent protein; Alb, albumin gene promoter; CB, chicken ?-actin promoter.
Gao et al.PNAS ?
September 3, 2002 ?
vol. 99 ?
no. 18 ?
were infused into the portal veins of mice that were analyzed
subsequently for expression of the transgene. Each vector con-
tained an AAV2-based genome using previously described liver-
specific promoters (i.e., albumin or thyroid hormone binding
globulin) to drive expression of the transgene. The impact of
in nu?nu and C57BL?6 mice and CG vectors in C57BL?6 mice
was consistent (Fig. 3 and Table 4). In all cases, AAV2?8 vectors
yielded the highest levels of transgene expression that ranged
from 16 to 110 times greater than what was obtained with
AAV2?2 vectors; expression from AAV2?5 and AAV2?7 vec-
tors was intermediate, with AAV2?7 higher than AAV2?5.
Analysis of X-Gal-stained liver sections of animals that received
the corresponding LacZ vectors showed a correlation between
the number of transduced cells and overall levels of transgene
expression. DNAs extracted from livers of C57BL?6 mice who
received the A1AT vectors were analyzed for abundance of
vector DNA by using real time PCR technology. The amount of
vector DNA found in liver 56 days after injection correlated with
the levels of transgene expression (Table 5). These studies
indicate that AAV8 is the most efficient vector for liver-directed
gene transfer because of increased numbers of transduced
The serologic data described above suggest that AAV2?8 vector
should not be neutralized in vivo after immunization with the other
serotypes. C57BL?6 mice received intraportal injections of
AAV2?8 vector expressing canine factor IX (1011genome copies)
56 days after they received intramuscular injections of A1AT
vectors of different serotypes. High levels of factor IX expression
(17 ? 2 ?g?ml, n ? 4), which were not significantly different from
what was observed in animals immunized with AAV2?1 (31 ? 23
?g?ml, n ? 4), AAV2?2 (16 ?g?ml, n ? 2), and AAV2?7 (12
?g?ml, n ? 2). This result contrasts with what was observed in
AAV2?8-immunized animals that were infused with the AAV2?8
factor IX vector in which no detectable factor IX was observed
(?0.1 ?g?ml, n ? 4).
The goal of this study was to isolate new AAV serotypes from
a species other than humans and develop them as vectors for
human gene therapy. The hope was that they would have
improved efficiencies of gene transfer and would not be recog-
nized by antibodies generated to AAV infections in humans. We
recently pursued a similar strategy for the development of
adenovirus-based vaccines where we were able to create vectors
Table 3. Neutralization of AAV2?7 and AAV2?8 with antisera
directed to all serotypes of AAV
Antisera was generated to all known serotypes in the mouse as described
in Methods. The ability of these antisera to neutralize AAV2?7 and AAV2?8
vectors is presented in the table: antisera were generated at the University of
Pennsylvania for serotypes 1, 2, 5, 7, and 8 and at the University of North
Carolina for serotypes 1, 2, 3, 4, 5, and 6.
Table 4. Expression of ?-unit of rhCG in mouse muscle and liver
4.5 ? 2.1
0.5 ? 0.1
14.2 ? 2.4
4.0 ? 0.7
1.6 ? 1.0
0.7 ? 0.3
4.8 ? 0.8
8.2 ? 4.3
76.0 ? 22.8
CMVCG and TBGCG minigene cassettes were used for muscle and liver-
directed gene transfer, respectively. Levels of rhCG were defined as relative
units (rU ? 103). The data were from assaying serum samples collected at day
28, post vector administration (four animals per group).
*ND, Not determined in this experiment.
expressing vectors. (Left, Liver) Representative tissue sections from NCR nude
mice that received AAV vectors expressing LacZ driven by human thyroid
hormone binding globulin gene promoter (TBG), a liver-specific promoter, at
gene transfer for cryosections and X-Gal staining. (Right) Panels show LacZ
after vector administration, and injected muscles were harvested for cryosec-
tions and X-Gal staining.
X-Gal histochemistry of muscle and liver after injection of LacZ-
Table 5. Real-time PCR analysis for abundance of AAV vectors in
nu?nu mouse liver after injection of 1 ? 1011genome copies
AAV vectors?dose Genome copies per cell
0.6 ? 0.36
0.003 ? 0.001
0.83 ? 0.64
2.2 ? 1.7
18 ? 11
A set of probe and primers targeting the SV40 poly(A) region of the vector
genome was used for TaqMan PCR. Values shown are means of three individ-
ual animals with standard deviations. The animals were sacrificed at day 56 to
harvest liver tissues for DNA extraction.
www.pnas.org?cgi?doi?10.1073?pnas.182412299Gao et al.
based on adenoviruses from chimpanzees that were shown to
activate T and B cells in model systems but were not neutralized
by antibodies generated to naturally acquired adenovirus infec-
tions in humans (34–36).
Our strategy for isolating unique AAVs differed from the
that potentially contain AAV genomes are infected with helper
viruses, such as adenovirus, to rescue latent AAV virus. We
attempted to retrieve novel AAV sequences by PCR amplifica-
tion of latent genomes by using primers to conserved regions of
AAV that flank an area of divergence. Tissues from nonhuman
primates were targeted for AAV isolation for a number of
reasons. We believed that nonhuman primate AAVs should be
sufficiently similar to human AAVs such that they retain tropism
for human target cells and could be propagated with human
adenoviruses and expanded as vectors by using standard tech-
niques. However, it was hoped that there would be sufficient
divergence from human AAVs that antibodies generated as a
result of naturally acquired AAV infections in humans would not
neutralize infection with the nonhuman primate AAVs.
Oligonucleotides to conserved regions of the cap gene did
amplify sequences from rhesus monkeys that represented unique
AAVs. Identical cap signature sequences were found in multiple
tissues from rhesus monkeys derived from at least two different
colonies. Full-length rep and cap ORFs were isolated and se-
were necessary to evaluate their potential as vectors because we
were able to generate vectors with the AAV7 or AAV8 capsids by
using the ITRs and rep from AAV2. This strategy also simplified
the comparison of different vectors because the actual vector
genome is identical between different vector serotypes. In fact, the
yields of recombinant vectors generated by using this approach did
not differ between serotypes.
The objectives of our study were indeed realized. Vectors
based on AAV7 and AAV8 do appear to be immunologically
distinct (i.e., they are not neutralized by antibodies generated
against other serotypes). Furthermore, sera from humans dem-
onstrate very little neutralizing activity to AAV7 and AAV8
vectors, which is an advantage over the human-derived AAVs
currently under development, for which a larger proportion of
the human population has preexisting immunity that is high level
and neutralizing (37).
The tropism of each new vector is indeed favorable for in vivo
applications. AAV2?7 vectors appear to transduce skeletal muscle
as efficiently as AAV2?1, which is the serotype that confers the
highest level of transduction in skeletal muscle of the primate
AAVs tested to date (9, 32, 33). Importantly, AAV2?8 provides a
of gene transfer to liver that until now has been relatively disap-
pointing in terms of the numbers of hepatocytes stably transduced.
AAV2?8 consistently achieved a 10- to 100-fold improvement in
gene transfer efficiency as compared with the other vectors. The
basis for the improved efficiency of AAV2?8 is unclear, although
it presumably is due to uptake via a different receptor that is more
active on the basolateral surface of hepatocytes. This improved
efficiency will be quite useful in the development of liver-directed
gene transfer where the number of transduced cells is critical, such
as in urea cycle disorders and familial hypercholesterolemia. We
described a novel approach for isolating new AAVs based on PCR
retrieval of genomic sequences. The amplified sequences were
easily incorporated into vectors and tested in animals. The relative
they should be considered for further development as vectors in
human gene therapy.
The support of the Vector, Quality Control, and Cell Morphology Cores
of the University of Pennsylvania was appreciated. TaqMan analysis
provided by Julio Sanmiguel. Production of novel AAV vectors per-
formed by Phoi Tran. We thank Dr. Jude Samulski for providing antisera
to all of the known AAV serotypes. J.M.W. holds equity in Targeted
Genetics, Inc. Support for this work was provided by the National
Institutes of Health (National Institute of Diabetes and Digestive and
Kidney Diseases Grant P30 DK 47757-09 and National Heart, Lung, and
Foundation, the Cystic Fibrosis Foundation, and GlaxoSmithKline
P. M. (Lippincott Williams & Wilkins, Philadelphia), Vol. 2, pp. 2327–2359.
2. Atchison, R. W., Casto, B. C. & Hammon, W. M. (1965) Science 194, 754–756.
3. Melnick, J. L., Mayor, H. D., Smith, K. O. & Rapp, F. (1965) J. Bacteriol. 90,
4. Hoggan, M. D., Blacklow, N. R. & Rowe, W. P. (1966) Proc. Natl. Acad. Sci.
USA 55, 1467–1474.
5. Bantel-Schaal, U. & zur Hausen, H. (1984) Virology 134, 52–63.
6. Georg-Fries, B., Biederlack, S., Wolf, J. & zur Hausen, H. (1984) Virology 134,
7. Rutledge, E. A., Halbert, C. L. & Russell, D. W. (1998) J. Virol. 72, 309–319.
8. Bantel-Schaal, U., Delius, H., Schmidt, R. & zur Hausen, H. (1999) J. Virol. 73,
9. Xiao, W., Chirmule, N., Berta, S. C., McCullough, B., Gao, G. & Wilson, J. M.
(1999) J. Virol. 73, 3994–4003.
10. Parks, W. P., Boucher, D. W., Melnick, J. L., Taber, L. H. & Yow, M. D. (1970)
Infect. Immun. 2, 716–722.
11. Blacklow, N. R., Hoggan, M. D. & Rowe, W. P. (1968) J. Natl. Cancer Inst. 40,
12. Rapoza, N. P. & Atchinson, R. W. (1967) Nature (London) 215, 1186–1187.
13. Boucher, D. W., Parks, W. P. & Melnick, J. L. (1970) J. Immunol. 104, 555–559.
15. Rabinowitz, J. E. & Samulski, J. (1998) Curr. Opin. Biotechnol. 9, 470–475.
16. Jooss, K., Yang, Y., Fisher, K. J. & Wilson, J. M. (1998) J. Virol. 72, 4212–4223.
17. Nakai, H., Iwaki, Y., Kay, M. A. & Couto, L. B. (1999) J. Virol. 73, 5438–5447.
18. Duan, D., Sharma, P., Yang, J., Yue, Y., Dudus, L., Zhang, Y., Fisher, K. J. &
Engelhardt, J. F. (1998) J. Virol. 72, 8568–8577.
19. Chen, Z. Y., Yant, S. R., He, C. Y., Meuse, L., Shen, S. & Kay, M. A. (2001)
Mol. Ther. 3, 403–410.
20. Bals, R., Xiao, W., Sang, N., Weiner, D. J., Meegalla, R. L. & Wilson, J. M.
(1999) J. Virol. 73, 6085–6088.
21. Yan, Z., Zak, R., Luxton, G. W., Ritchie, T. C., Bantel-Schaal, U. &
Engelhardt, J. F. (2002) J. Virol. 76, 2043–2053.
22. Auricchio, A., Kobinger, G., Anand, V., Hildinger, M., O’Connor, E., Maguire,
A. M., Wilson, J. M. & Bennett, J. (2001) Hum. Mol. Genet. 10, 3075–3081.
23. Davidson, B. L., Stein, C. S., Heth, J. A., Martins, I., Kotin, R. M., Derksen,
T. A., Zabner, J., Ghodsi, A. & Chiorini, J. A. (2000) Proc. Natl. Acad. Sci. USA
24. Zabner, J., Seiler, M., Walters, R., Kotin, R. M., Fulgeras, W., Davidson, B. L.
& Chiorini, J. A. (2000) J. Virol. 74, 3852–3858.
25. Schnell, M. A., Zhang, Y., Tazelaar, J., Gao, G. P., Yu, Q. C., Qian, R., Chen,
S. J., Varnavski, A. N., LeClair, C., Raper, S. E. & Wilson, J. M. (2001) Mol.
Ther. 3, 708–722.
26. Hildinger, M., Auricchio, A., Gao, G., Wang, L., Chirmule, N. & Wilson, J. M.
(2001) J. Virol. 75, 6199–6203.
27. Gao, G., Qu, G., Burnham, M. S., Huang, J., Chirmule, N., Joshi, B., Yu, Q. C.,
Marsh, J. A., Conceicao, C. M. & Wilson, J. M. (2000) Hum. Gene Ther. 11,
28. Gao, G. P., Yang, Y. & Wilson, J. M. (1996) J. Virol. 70, 8934–8943.
29. Zoltick, P. W. & Wilson, J. M. (2000) Mol. Ther. 2, 657–659.
30. Wang, L., Zoppe, M., Hackeng, T. M., Griffin, J. H., Lee, K. F. & Verma, I. M.
(1997) Proc. Natl. Acad. Sci. USA 94, 11563–11566.
31. Zhang, Y., Chirmule, N., Gao, G. P., Qian, R., Croyle, M., Joshi, B., Tazelaar,
J. & Wilson, J. M. (2001) Mol. Ther. 3, 697–707.
32. Chao, H., Monahan, P. E., Liu, Y., Samulski, R. J. & Walsh, C. E. (2001) Mol.
Ther. 4, 217–222.
33. Chao, H., Liu, Y., Rabinowitz, J., Li, C., Samulski, R. J. & Walsh, C. E. (2000)
Mol. Ther. 2, 619–623.
34. Farina, S. F., Gao, G. P., Xiang, Z. Q., Rux, J. J., Burnett, R. M., Alvira, M. R.,
Marsh, J., Ertl, H. C. & Wilson, J. M. (2001) J. Virol. 75, 11603–11613.
35. Cohen, C. J., Xiang, Z. Q., Gao, G. P., Ertl, H. C., Wilson, J. M. & Bergelson,
J. M. (2002) J. Gen. Virol. 83, 151–155.
36. Xiang, Z., Gao, G., Reyes-Sandoval, A., Cohen, C. J., Li, Y., Bergelson, J. M.,
Wilson, J. M. & Ertl, H. C. (2002) J. Virol. 76, 2667–2675.
37. Chirmule, N., Propert, K., Magosin, S., Qian, Y., Qian, R. & Wilson, J. (1999)
Gene Ther. 6, 1574–1583.
Gao et al.PNAS ?
September 3, 2002 ?
vol. 99 ?
no. 18 ?