Directed evolution of adeno-associated virus for enhanced gene delivery and gene targeting in human pluripotent stem cells.

Abstract

Efficient approaches for the precise genetic engineering of human pluripotent stem cells (hPSCs) can enhance both basic and applied stem cell research. Adeno- associated virus (AAV) vectors are of particular interest for their capacity to mediate efficient gene delivery to and gene targeting in various cells. However, natural AAV serotypes offer only modest transduction of human embryonic and induced pluripotent stem cells (hESCs and hiPSCs), which limits their utility for efficiently manipulating the hPSC genome. Directed evolution is a powerful means to generate viral vectors with novel capabilities, and we have applied this approach to create a novel AAV variant with high gene delivery efficiencies (~50%) to hPSCs, which are importantly accompanied by a considerable increase in gene-targeting frequencies, up to 0.12%. While this level is likely sufficient for numerous applications, we also show that the gene-targeting efficiency mediated by an evolved AAV variant can be further enhanced (>1%) in the presence of targeted double- stranded breaks (DSBs) generated by the co-delivery of artificial zinc finger nucleases (ZFNs). Thus, this study demonstrates that under appropriate selective pressures, AAV vectors can be created to mediate efficient gene targeting in hPSCs, alone or in the presence of ZFN- mediated double-stranded DNA breaks.

Full text

Page 1

original article
© The American Society of Gene & Cell Therapy
Molecular Therapy vol. 20 no. 2, 329–338 feb. 2012
329
Efficient approaches for the precise genetic engineering
of human pluripotent stem cells (hPSCs) can enhance
both basic and applied stem cell research. Adeno-
associated virus (AAV) vectors are of particular interest for
their capacity to mediate efficient gene delivery to and
gene targeting in various cells. However, natural AAV sero-
types offer only modest transduction of human embry-
onic and induced pluripotent stem cells (hESCs and
hiPSCs), which limits their utility for efficiently manipu-
lating the hPSC genome. Directed evolution is a powerful
means to generate viral vectors with novel capabilities,
and we have applied this approach to create a novel AAV
variant with high gene delivery efficiencies (~50%) to
hPSCs, which are importantly accompanied by a con-
siderable increase in gene-targeting frequencies, up to
0.12%. While this level is likely sufficient for numerous
applications, we also show that the gene-targeting effi-
ciency mediated by an evolved AAV variant can be further
enhanced (>1%) in the presence of targeted double-
stranded breaks (DSBs) generated by the co-delivery of
artificial zinc finger nucleases (ZFNs). Thus, this study
demonstrates that under appropriate selective pressures,
AAV vectors can be created to mediate efficient gene
targeting in hPSCs, alone or in the presence of ZFN-
mediated double-stranded DNA breaks.
Received 26 May 2011; accepted 27 October 2011; published online
22 November 2011. doi:10.1038/mt.2011.255
INTRODUCTION
Te capacity to mediate high efciency gene delivery to human
pluripotent stem cells (hPSCs) has numerous applications, rang-
ing from the study of specifc genes in stem cell self-renewal and
diferentiation to the directed diferentiation of stem cells into
specifc lineages for therapeutic application. Furthermore, pre-
cise manipulation of the human stem cell genome using gene-
targeting techniques that exploit the natural ability of cells to
perform homologous recombination (HR) has broad applications
and implications, including safe harbor integration of genes for
basic or therapeutic application, creating in vitro models for investi-
gating human development and disease, and high-throughput drug
discovery and toxicity studies.1,2 However, while gene delivery and
gene targeting are well established for various mammalian somatic
cells,3 a readily generalized approach for efcient gene expression
and gene targeting in hPSCs requires further development.
Current methods to deliver genes to hPSCs range from viral
vectors to plasmid-based transient gene expression. Lentiviral
vectors—which are highly efcient and result in long-term gene
expression—have been extensively employed in numerous studies
in human stem cells.4 However, transient expression is desirable in
some cases, such as for the temporary overexpression of regulatory
signals to manipulate stem cell fate decisions.5,6 Also, vector integra-
tion into the genome can risk insertional mutagenesis, a potential
concern for downstream clinical application.4 As an alternative,
electroporation can be used to achieve transient gene expression,
though this method can sufer from low transfection efciencies and
moderate toxicity in human stem cells.7 In addition to gene deliv-
ery, there is a need to develop efcient gene-targeting methods that
rely on HR to introduce permanent and sequence-specifc genome
modifcations in hPSCs. Several impressive studies have demon-
strated successful gene targeting in hPSCs, though the initial rates
reported using conventional methodologies are low (10−7–10−5 cor-
rectly targeted cells for every original cell in the population), which
necessitates the use of positive and negative selection to improve the
overall efciency and specifcity of the process.8,9
Recently, several approaches have been utlized to improve
gene targeting in hPSCs, including the introduction of double-
stranded breaks (DSBs) into the cellular genome by engineered
nucleases.10–14 Such breaks stimulate the cellular DNA repair
machinery and thereby greatly enhance the rate of homologous
recombination with a donor DNA.15 For example, Zou et al. found
that cotransfection of plasmids containing the donor DNA and
zinc fnger nucleases (ZFNs) signifcantly increased gene tar-
geting in hPSCs up to a frequency of 0.24% (~1 targeted cell in
The first two authors contributed equally to this work.
Correspondence: David V Schaffer, Department of Chemical and Biomolecular Engineering, University of California, Berkeley, 274 Stanley Hall, Berkeley,
California 94720-1462, USA. E-mail: schaffer@berkeley.edu
Directed Evolution of Adeno-associated Virus
for Enhanced Gene Delivery and Gene Targeting
in Human Pluripotent Stem Cells
Prashanth Asuri1–3, Melissa A Bartel1–3, Tandis Vazin1–3, Jae-Hyung Jang1–4, Tiffany B Wong5 and
David V Schaffer1–3
1Department of Chemical and Biomolecular Engineering, University of California, Berkeley, California, USA; 2Department of Bioengineering,
University of California, Berkeley, California, USA; 3The Helen Wills Neuroscience Institute, University of California, Berkeley, California, USA;
4Department of Chemical and Biomolecular Engineering, Yonsei University, Seoul, Korea; 5Department of Molecular and Cell Biology,
University of California, Berkeley, California, USA

Page 2

330
www.moleculartherapy.org vol. 20 no. 2 feb. 2012
© The American Society of Gene & Cell Therapy
AAV for Enhanced Gene Delivery and Targeting
415 original cells), as compared to less than 1 targeted cell per 106
original cells in the absence of the ZFNs.13 More recently, a new
approach has been described to engineer DNA-binding specifci-
ties based on transcription activator–like efector (TALE) proteins
from Xanthomonas plant pathogens and artifcial restriction
enzymes generated by fusing TALEs to the catalytic domain of
FokI was used to generate discrete edits or small deletions within
endogenous human genes at efciencies of up to 25%.16 However,
while the utilization of ZFNs and transcription activator–like
efector nucleases (TALENs) to enhance gene-targeting efcien-
cies in hPSCs is highly promising, the approach entails custom
engineering of a new nuclease for each new target locus, which
requires labor, time, and resources.
Adeno-associated virus (AAV) is a nonpathogenic, nonenvel-
oped virus containing a 4.7 kb single-stranded DNA genome that
encodes the structural proteins of the viral capsid (encoded by the
cap gene) and the nonstructural proteins necessary for viral rep-
lication and assembly (encoded by the rep gene), fanked by short
inverted terminal repeats.17 Recombinant versions of AAV can
be created by inserting a sequence of interest in place of rep and
cap, and the resulting recombinant vectors can efciently deliver a
transgene and safely mediate long-term gene expression in divid-
ing and nondividing cells of numerous tissues.17 Such AAV-based
vectors have proven safe, efcient, and recently very efective for
clinical application.18,19
An interesting property of AAV is that, as demonstrated
by Russell and colleagues, AAV vector genomes carrying gene-
targeting constructs can mediate HR with target loci in a cel-
lular genome at efciencies 103–104-fold higher than plasmid
constructs.3 Te AAV genome’s inverted terminal repeats, which
apparently mediate its entry into the RAD51/RAD54 component
of the cellular HR pathway, play a central role in this property.20
AAV-mediated gene targeting has been successfully applied to cells
that AAV can efectively transduce, but naturally occurring AAV
variants are typically inefcient at infecting a number of stem cell
types, particularly human embryonioc stem cells (hESCs).21,22
We have implemented directed evolution—a rapid, high-
throughput selection approach to create and isolate novel mutants
from millions of genetic variants—to rapidly engineer AAV vari-
ants with desired gene delivery properties in the absence of the
extensive mechanistic information typically required for rational
biomolecular design. Recent work highlights the ability to apply
directed evolution to create AAV mutants with altered receptor
binding, neutralizing antibody-evasion properties, and altered cell
tropism in vitro and in vivo.23–29 Likewise, we have recently demon-
strated that directed evolution can yield novel AAV variants that
enhance gene delivery and gene targeting in neural stem cells.21
Here, we implement this approach to create new AAV mutants
with the enhanced capacity to infect and, subsequently, increase
the efciency of AAV-mediated gene targeting in hPSCs, both
in the presence and absence of ZFN-mediated double-stranded
DNA breaks.
RESULTS
Evaluation of wild-type serotypes
Tere are numerous naturally occurring AAV variants and
serotypes, each of which has diferent protein capsids and thus
somewhat diferent gene delivery properties for various cell types
and tissues.30 AAV serotypes 1–6 and 8 were tested to assess their
potential for hESC gene delivery. Initially, HSF-6 hESCs were
infected with AAV vectors carrying green fuorescent protein
(GFP) complementary DNA at a multiplicity of infection (MOI)
of 10,000, and 48 hours postinfection, fow cytometry was used to
determine the percentage of cells that express GFP and were thus
infected by AAV. Te most efcient was AAV6 (Figure 1); how-
ever, its low delivery efciency (14.75 ? 3.08% GFP positive cells)
places signifcant limitations on the fraction of cells that express a
transgene or could undergo gene targeting. Tese results are con-
sistent with earlier reports showing that natural AAV serotypes
are typically inefcient in gene delivery to human stem cells,31,32
thus demonstrating the need to fnd, or evolve, an AAV variant
capable of more efcient transduction.
AAV library generation and selection through
directed evolution
Te AAV capsid proteins, encoded by the cap gene, determine the
virus’ ability to infect cells through their initial binding to vari-
ous cell-surface receptors, intracellular trafcking, and entry into
the nucleus. Directed evolution is a high-throughout approach
that involves the creation and functional selection of libraries of
genetic mutants to isolate novel variants with desirable proper-
ties. We implemented a directed evolution strategy to create AAV
variants capable of mediating efcient gene delivery to hESCs
(Figure 2a). Briefy, hESCs were infected with pools of virus
variants created using several diferent techniques: a library in
which AAV2 and AAV6 cap were subjected to error-prone PCR,
a “shufed” library composed of random cap chimeras of seven
parent AAV serotypes,24 an AAV2 cap library with random 7-mer
inserts,33 and AAV2 cap with substituted loop regions.25 Following
infection with AAV libraries (Figure 2a, step 4) and amplifca-
tion of the infectious AAV variants through adenovirus super-
infection—as AAV requires a helper virus such as adenovirus to
induce replication—(step 5), the resulting titre of AAV rescued
from each library condition was quantifed and compared to titres
of recovered wild-type AAV2 as a metric for relative success of
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
% GFP positive cells
AAV1 AAV2AAV3 AAV4AAV5AAV6AAV8
Figure 1 Gene expression in hESCs mediated by various AAV sero-
types. HSF-6 cells were infected with numerous natural AAV serotypes at
a MOI of 10,000. Transduction efficiency was assessed as the percentage
of GFP positive cells measured by flow cytometry 48 hours postinfection.
Error bars indicate standard deviation (n = 3). AAV, adeno-associated
virus; GFP, green fluorescent protein; hESC, human embryonic stem cell;
MOI, multiplicity of infection.

Page 3

Molecular Terapy vol. 20 no. 2 feb. 2012
331
© The American Society of Gene & Cell Therapy
AAV for Enhanced Gene Delivery and Targeting
the selection. For each selection step, viral pools from the library
produced higher viral titres than wild-type AAV2 at MOIs of both
10 and 100 (data not shown). Tese viral pools were then used as
the starting point for the subsequent selection step (step 6). Afer
three such selection steps, the successful viral cap genes were iso-
lated (step 7) and tested individually to identify variants with the
most efcient gene delivery. In addition, the cap genes isolated
afer the third selection step were subjected to an additional round
of evolution, i.e., additional mutagenesis (step 8) and three selec-
tion steps, to further increase the ftness of the pool.
Increased transduction efficiency of the novel
evolved AAV variant in hESCs
Nine cap variants from each library isolated afer the third selection
step of the frst round of evolution were fully sequenced, and the
AAV capsid protein variations and the frequency with which each
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
100.0
10.0
1.0
0.1
AAV2
EP 1.1
EP 1.3
EP 1.5
EP 1.8
EP 1.9
% GFP positive cells
% GFP positive cells
50.0
40.0
30.0
20.0
10.0
0.0
% GFP positive cells
1,000
Multiplicity of infection
10,000 100,000
HSF6
ESCs
MSC
iPSCs
hFIb2
iPSCs
H1
ESCs
AAV2
AAV1.9
AAV2
AAV6
AAV1.9
*
*
*
*
*
*
*
CAP
CAP
CAP
CAP
1
2
3
6
8
7
5
4
R459G
VP1 VP2 VP3
(IU:vg) ? 104
CHOHEK293T HSF6
*
*
*
a
c
fg
de
b
Figure 2 Directed evolution of AAV and hPSC transduction by AAV 1.9. (a) Schematic representation of directed evolution. 1) A viral library
is created by mutating the cap gene. 2) Viruses are packaged in HEK293T cells using plasmid transfection, such that each particle is composed of
mutant capsid surrounding the cap gene encoding that protein capsid. 3) Viruses are harvested from 293 cells and purified. 4) The viral library is
introduced to the HSF-6 hESCs in vitro. 5) Successful viruses are amplified and recovered using adenovirus rescue. 6) Successful clones are enriched
through repeated selections. 7) Isolated viral DNA reveals selected cap genes. 8) Selected cap genes are mutated again to serve as a new starting
point for selection. (b) Molecular model of the full AAV2 capsid, based on the solved structure,51 shows the location of the R459G mutation (blue) on
the surface of the capsid (VP3 region), near the threefold axis of symmetry and residues known to be important for heparin and FGF receptor bind-
ing (pink). (c) AAV-mediated gene expression in hESCs. HSF-6 cells were infected with selected mutants at a MOI of 10,000. Transgene expression
was assessed as the percentage of GFP positive cells measured by flow cytometry 48 hours postinfection. Error bars indicate the standard deviation
(n = 3), *P < 0.01. (d) Elevating the MOI increases transduction. HSF-6 cells were infected with AAV1.9 at MOIs of 1,000, 10,000, and 100,000.
Transgene expression was assessed as the percentage of GFP positive cells measured by flow cytometery 48 hours postinfection. Error bars indicate
standard deviation (n = 3), *P < 0.01. (e) AAV1.9-mediated gene expression in hPSCs. HSF-6 hESCs, human dermal fibroblast-derived hiPSCs, human
MSC-derived hiPSCs, and H1 hESCs were infected at a MOI of 10,000. Transgene expression was assessed as the percentage of GFP positive cells
measured by flow cytometry 48 hours postinfection. Error bars indicate the standard deviation (n = 3), *P < 0.01. (f) In vitro analysis of AAV1.9
tropism. Variant AAV1.9 (white), selected on hESCs, is less infectious on AAV-permissive cell types (HEK293T, CHO) than recombinant AAV2 (black)
and AAV6 (gray), but more infectious on HSF-6 hESCs. Error bars indicate standard deviation (n = 3), *P < 0.01. (g) HSF-6 cells infected by AAV1.9
maintain pluripotency marker expression. HSF-6 cells infected with AAV1.9 encoding GFP at an MOI of 100,000 were fixed and immunostained
1 week after infection for the presence of GFP (green) and the pluripotency marker Nanog (red). AAV, adeno-associated virus; FGF, fibroblast growth
factor; GFP, green fluorescent protein; hESC, human embryonic stem cell; hiPSC, human induced pluripotent stem cell; MOI, multiplicity of infection;
MSC, mesenchymal stem cell.

Page 4

332
www.moleculartherapy.org vol. 20 no. 2 feb. 2012
© The American Society of Gene & Cell Therapy
AAV for Enhanced Gene Delivery and Targeting
clone was detected in the error-prone PCR-mutagenized AAV2
and AAV6 library are shown in Table 1. Te cap genes of these
variants were also used to package recombinant AAV carrying
the GFP gene, under the control of the cytomegalovirus (CMV)
promoter. Tese recombinant AAV variants were then used to
infect hESCs, and the gene delivery efciency of each virus was
measured via fow cytometry. Of the mutants isolated from the
error-prone PCR-mutagenized AAV2 and AAV6 library, the vari-
ant AAV EP 1.9 (or AAV1.9), which carries a single-point muta-
tion (R459G) (Figure 2b), showed the highest infection efciency
at 28.25 ? 1.68% GFP positive cells, representing an approximately
threefold increase over AAV2 (its parental serotype) and a two-
fold increase over AAV6 (the best serotype) at an MOI of 10,000
(Figure 2c). Variants isolated from the shufed, 7-mer insertion,
and substituted loop region libraries were sequenced and analyzed
as well. Variants from these libraries showed increased infection
efciency compared to AAV2, but not AAV6 (data not shown). At
an MOI of 100,000, the variant AAV1.9 achieved a strong infec-
tion efciency of 48.21 ? 12.92% GFP positive cells (Figure 2d).
It should be noted that hESCs grow as colonies that are typically
multilayered, making accessing and infecting all of the cells in a
colony difcult. However, AAV1.9 provides an increase in gene
delivery efciency, by an increase in the number of cells infected
(Figure 2c), an increase in the relative fuorescence intensity
of infected cells (approximately threefold higher than AAV6),
and a signifcantly increased number of donor DNA molecules
delivered to each cell (Supplementary Materials and Methods
and Supplementary Figure S1). Interestingly, sequencing of cap
variants isolated afer an additional round of mutagenesis to the
cap library and three additional selection steps all matched the
sequence of AAV1.9 (data not shown).
AAV1.9 was isolated through multiple rounds of selection
using the HSF-6 hESC line; however, the increase in gene deliv-
ery efciency also extended to the broadly utilized H1 hESC line,
as well as induced pluripotent cell lines including mesenchymal
stem cell (MSC)-derived, and dermal fbroblast-derived hiPSC
lines (Figure 2e).34 In addition, while AAV1.9 is capable of higher
gene delivery efciency to hESCs and hiPSCs, it is less infectious
towards several cell types typically permissive to AAV infec-
tion (Figure 2f), indicating that in this case directed evolution
yielded a variant that efciently and somewhat selectively trans-
duces an ordinarily nonpermissive cell. Furthermore, the hESCs
maintained pluripotency marker expression following infection,
assessed via Nanog immunostaining (Figure 2g).
Te R459G mutation in AAV1.9 lies close to but not within
the heparin or fbroblast growth factor (FGF) receptor binding
domains on the AAV capsid surface (Figure 2b), though it could
conceivably modulate binding to the cell surface. To determine
whether the amino acid 459 mutation alters the heparin-binding
properties of the variant, AAV2, AAV6, and AAV1.9 were ana-
lyzed using heparin column chromatography, which can provide a
measure of the heparin-binding afnity of AAV variants. Despite
its mutation residing outside the reported heparin-binding
domain,35 and despite the loss of a positively charged residue,
AAV1.9 displays a higher afnity for heparin compared to the
parental serotype AAV2, as indicated by the higher NaCl concen-
tration needed to elute the majority of AAV1.9 from the heparin
column. (Figure 3a). However, in vitro transduction of CHO and
pgsA (a mutant CHO cell line lacking surface glycosaminoglycans
(GAGs)) cells showed that AAV1.9 had less dependence on cell-
surface heparan sulfate proteoglycans (AAV2’s primary receptor)36
for cell transduction (Figure 3b). Furthermore, while it exhibited
higher transduction levels, AAV1.9 bound to hESCs at levels simi-
lar to AAV2 (Figure 3c). Collectively, these data indicate that the
60
50
40
30
20
10
0
% Viral load
90
80
70
60
50
40
30
20
10
0
% Virus bound
Elution solution (mmol/l NaCI)
load
150200 250 300350 400450500 550600 650 700 750
1,000
AAV2
AAV6
AAV1.9
CHO
pgsA
Amount transduced
Amount bound
100.0
10.0
1.0
0.1
(IU:vg) ? 104
(IU:vg) ? 104
AAV2AAV6 AAV1.9
AAV2 AAV6 AAV1.9
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
*
*
*
abc
Figure 3 Mechanistic analysis of transduction by variant AAV. (a) AAV1.9 has a higher affinity for heparin than AAV2 and AAV6. The heparin affin-
ity column chromatogram of recombinant AAV2 (black), AAV6 (gray), and AAV1.9 (white) is shown, where virus was eluted from the column using
increasing concentrations of NaCl. Virus was quantified using qPCR. (b) In vitro characterization of AAV1.9 HSPG dependence. CHO (black) and pgsA
(white) cells were transduced to demonstrate the decrease in HSPG dependence of AAV1.9 compared to AAV2. Error bars indicate standard deviation
(n = 3), *P < 0.01. AAV, adeno-associated virus; hESC, human embryonic stem cell, qPCR, quantitative PCR. (c) In vitro characterization of AAV1.9
binding affinity. AAV1.9 and its parental serotype AAV2 bind hESCs to similar extents, yet AAV1.9 is capable of significantly higher transduction of
hESCs. Error bars indicate standard deviation (n = 3), *P < 0.01.
Table 1 Summary of variants isolated from three rounds of selection
against hESCs
Clone name MutationsFrequency
EP 1.1
EP 1.3
EP 1.5
EP 1.8
EP 1.9
S85G, R459G
R459G, V600A
S85G, R447S, R459G
R447S, R459G
R459G
3/9
1/9
1/9
2/9
2/9
Protein sequence features and the frequency of appearance during analysis are
listed for each clone isolated from the EP AAV2/AAV6 library.
AAV, adeno-associated virus; hESC, human embryonic stem cell.

Page 5

Molecular Terapy vol. 20 no. 2 feb. 2012
333
© The American Society of Gene & Cell Therapy
AAV for Enhanced Gene Delivery and Targeting
R459G mutation may exert its efects afer initial docking to the
cell surface, though future studies will be needed to elucidate its
mechanism of enhanced hESC infection.
Enhanced gene-targeting efficiency of the novel
evolved AAV variant in hESCs
We assessed the ability of the novel variant to mediate gene tar-
geting in hPSCs. To do so, we employed a mutated GFP-based
reporter system, previously described by Zou et al.13 Briefy, GFP
complementary DNA containing a 35 bp insertion harboring a
stop codon (GFP∆35) was introduced into HSF-6 hESCs using a
lentiviral vector (see Materials and Methods and Figure 4a). Te
donor construct (t37GFP) consisted of a promoter-less, nonfunc-
tional GFP with a 37 bp 5? truncation, as well as 290 nucleotides
of homology upstream and 1,619 nucleotides downstream of the
35 bp insertion in GFP∆35. Te donor plasmid was packaged
into recombinant AAV vectors (AAV2, AAV6, and AAV1.9), and
a hESC line carrying the mutated GFP was infected with these
gene-targeting constructs to test their capacity to mediate gene
correction and thereby restore GFP fuorescence. Te levels of
gene correction achieved using recombinant AAV2 and AAV6
were below the limits of detection of fow cytometry in this assay
(Figure 4b), whereas the variant AAV1.9 was capable of tar-
geted gene correction resulting in fuorescent hESCs (0.12%).
Furthermore, consistent with previous reports,3 we also show that
at similar efciencies of gene delivery, AAV vectors can medi-
ate targeted gene correction at rates much higher than plasmid
constructs (Figure 4b). Moreover, as discussed earlier, AAV1.9 is
LTRUB GFP?35
t37GFP ITR
ITR
c
CMV
ZFN1 ITR
ITR
CMV
ZFN2
ITR
UB GFP
IRES
IRES
IRES
PuroR
PuroR
PuroR
LTR
ITR
35 bp insert
37 bp deletion
ZFNs site
Lentiviral vector, GFP?35
Donor AAV vector, t37GFP
Corrected GFP sequence
AAV ZFN vectors
1.75
1.50
1.25
1.00
0.75
0.50
0.25
0.00
% GFP positive cells
AAV1.9 AAV2 AAV6 MXcell
*
ND NDND
With ZFNs
Without ZFNs
a
b
No digest
1212
Xhol digest
CAAGGACGACGGCAACTACAAGACCCGCGCCGAGGTGAAGTTCGA
GGGCGACACC
GTTATCCCTAAGCTTCGCGCCGAGGTGAAGTTCGAGGGCGACACC
CAAGGACGACGGCAACTACAAGACCTAAGCTCTCGAGATTACCCT
d
fg
10
1
0.1
0.01
% GFP positive cells
0 10 20 30
Days
With ZFNs
Without ZFNs
e
Figure 4 AAV1.9-mediated correction of a nonfunctional GFP expressed in hESCs. (a) Schematic overview depicting the targeting strategy for
the GFP gene correction. The defective GFP gene—GFP?35, mutated by the insertion of a 35 bp fragment containing a translational stop codon13—
was integrated into HSF-6 cells using a lentiviral vector. A targeting vector (t37GFP) containing a 5? truncated GFP coding sequence was packaged
into a recombinant AAV vector lacking a promoter. Homologous recombination between donor vector and integrated defective GFP?35 gene would
correct the 35 bp mutation and result in fluorescent hESCs. (b) Gene targeting frequencies assessed as the percentage of GFP positive cells mea-
sured via flow cytometry 72 hours postinfection. Error bars indicate standard deviation (n = 3), *P < 0.01. ND, not detectable by flow cytometry.
(c,d) Representative analyses of the targeted correction of GFP?35 gene. Genomic DNA from the “corrected” and “uncorrected” HSF-6 cells was
amplified using PCR. The 35 bp mutation within the GFP?35 gene contains an Xho I site, and therefore only the PCR products from the “uncorrected”
cells (sample 2) can be digested using Xho I, whereas the PCR products from the corrected cells (sample 1) show the restoration of the functional GFP
gene without the Xho I site. DNA sequencing analysis also shows AAV1.9-mediated correction of the 35 bp mutation in HSF-6 cells originally express-
ing the mutated GFP?35 gene. (e) Time course of AAV-mediated GFP?35 correction in HSF-6 cells. Error bars indicate standard deviation (n = 3).
(f) Maintenance of an undifferentiated state of gene-corrected, GFP-expressing cells for 30 days postinfection. GFP positive HSF-6 cells, isolated via
FACS, were cultured for 30 days, then fixed and probed for the presence of GFP (green), Oct4 (red), and DAPI (blue). (g) Pluripotency of the HSF-6
cells carrying the corrected GFP gene was further confirmed through embryoid body-mediated in vitro differentiation to generate derivatives of all
three germ layers, as indicated by the expression of the ectodermal marker ?-III tubulin, the mesodermal marker ?-smooth muscle actin (?-SMA),
and the endodermal marker hepatocyte nuclear factor 3 ? (HNF3?) after 24 days of differentiation (Bar = 100 µm). AAV, adeno-associated virus; CMV,
cytomegalovirus; DAPI, 4?,6-diamidino-2-phenylindole; FACS, fluorescent-activated cell sorting; GFP, green fluorescent protein; hESC, human embry-
onic stem cell; IRES, internal ribosome entry site; ITR, inverted terminal repeat; LTR, long terminal repeat.