Phase 1 gene therapy for Duchenne muscular dystrophy using a translational optimized AAV vector.

Abstract

Efficient and widespread gene transfer is required for successful treatment of Duchenne muscular dystrophy (DMD). Here, we performed the first clinical trial using a chimeric adeno-associated virus (AAV) capsid variant (designated AAV2.5) derived from a rational design strategy. AAV2.5 was generated from the AAV2 capsid with five mutations from AAV1. The novel chimeric vector combines the improved muscle transduction capacity of AAV1 with reduced antigenic crossreactivity against both parental serotypes, while keeping the AAV2 receptor binding. In a randomized double-blind placebo-controlled phase I clinical study in DMD boys, AAV2.5 vector was injected into the bicep muscle in one arm, with saline control in the contralateral arm. A subset of patients received AAV empty capsid instead of saline in an effort to distinguish an immune response to vector versus minidystrophin transgene. Recombinant AAV genomes were detected in all patients with up to 2.56 vector copies per diploid genome. There was no cellular immune response to AAV2.5 capsid. This trial established that rationally designed AAV2.5 vector was safe and well tolerated, lays the foundation of customizing AAV vectors that best suit the clinical objective (e.g., limb infusion gene delivery) and should usher in the next generation of viral delivery systems for human gene transfer.

Full text

Page 1

original article
© The American Society of Gene & Cell Therapy
Molecular Therapy
1
Efficient and widespread gene transfer is required for
successful treatment of Duchenne muscular dystrophy
(DMD). Here, we performed the first clinical trial using
a chimeric adeno-associated virus (AAV) capsid variant
(designated AAV2.5) derived from a rational design strat-
egy. AAV2.5 was generated from the AAV2 capsid with
five mutations from AAV1. The novel chimeric vector
combines the improved muscle transduction capacity
of AAV1 with reduced antigenic crossreactivity against
both parental serotypes, while keeping the AAV2 recep-
tor binding. In a randomized double-blind placebo-
controlled phase I clinical study in DMD boys, AAV2.5
vector was injected into the bicep muscle in one arm,
with saline control in the contralateral arm. A subset of
patients received AAV empty capsid instead of saline
in an effort to distinguish an immune response to vec-
tor versus minidystrophin transgene. Recombinant AAV
genomes were detected in all patients with up to 2.56
vector copies per diploid genome. There was no cellular
immune response to AAV2.5 capsid. This trial established
that rationally designed AAV2.5 vector was safe and well
tolerated, lays the foundation of customizing AAV vectors
that best suit the clinical objective (e.g., limb infusion
gene delivery) and should usher in the next generation
of viral delivery systems for human gene transfer.
Received 17 June 2011; accepted 5 October 2011; advance online
publication 8 November 2011. doi:10.1038/mt.2011.237
IntroductIon
Duchenne muscular dystrophy (DMD) is the most common
severe, life-threatening form of muscular dystrophy in child-
hood. DMD is associated with progressive muscle degeneration,
weakness, and mortality. DMD is X-linked and is genetically
inherited, caused by mutations in dystrophin, a large (427 kDa)
cytoskeletal protein that is normally expressed in skeletal and car-
diac muscle, as well as smooth muscle, brain, and retina in various
isoforms.1 The incidence of DMD is 1 in 3,500–5,000 newborn
males.2 The gene encoding dystrophin is the largest identified to
date,1 and the risk of spontaneous mutation is high (1/10,000 germ
cells).
Because DMD is caused by recessive and monogenic muta-
tions in the dystrophin gene, this disease is thought to be amenable
to correction or improvement by gene therapy. Gene replacement
therapy for DMD is a promising strategy because in theory it
could benefit all DMD patients regardless of the nature of their
genetic mutations e.g., deletions and point mutations. A number
of additional drugs are in development and include alternative
gene therapy approaches, cell therapy, antisense oligonucleotides,
and small molecule drug therapies.3 However, DMD presents with
a multitude of challenges for the development of effective treat-
ments including the largest disease gene that requires miniaturi-
zation to be compatible with gene replacement vectors, the very
large mass of target tissue that is widely distributed throughout
the body with layers of biological barriers, and unknown changes
at the cell membrane level that may impact on vector binding and
uptake. Furthermore, the immunological components of DMD
pathophysiology may also impact on the development and deliv-
ery of novel therapeutics and therefore assays of immune system
recognition and response must be carefully integrated into clinical
trial designs.
Recombinant adeno-associated virus (rAAV) vectors car-
rying a miniaturized functional dystrophin gene (designated
minidystrophin) have the potential to arrest or reverse muscle
failure.4–6 Comprehensive proof of concept and preclinical testing
has evaluated the effectiveness of AAV-minidystrophin gene
delivery in animal models of DMD including the mdx mice and
dystrophin/utrophin double knockout mice. Minidystrophin
The first three authors contributed equally to this work.
Correspondence: R Jude Samulski, Gene Therapy Center, University of North Carolina at Chapel Hill, 7119 Thurston Bowles CB 7352 Chapel Hill,
North Carolina 27599-7352, USA. E-mail: rjs@med.unc.edu or Xiao Xiao, University of North Carolina at Chapel Hill, Eshelman School of Pharmacy,
CB # 7362, Genetic Medicine Building, Chapel Hill, North Carolina 27599-7362, USA. E-mail: xxiao@email.unc.edu
Phase 1 Gene Therapy for Duchenne Muscular
Dystrophy Using a Translational Optimized AAV
Vector
Dawn E Bowles1, Scott WJ McPhee2, Chengwen Li3, Steven J Gray3, Jade J Samulski2,
Angelique S Camp4, Juan Li5, Bing Wang5, Paul E Monahan3, Joseph E Rabinowitz6, Joshua C Grieger4,
Lakshmanan Govindasamy7, Mavis Agbandje-McKenna7, Xiao Xiao5 and R Jude Samulski3
1Department of Surgery, Division of Surgical Sciences, Duke University Medical Center, Durham, North Carolina, USA; 2Asklepios BioPharmaceutical Inc.,
Chapel Hill, North Carolina, USA; 3Gene Therapy Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; 4Joint Vector Core,
University of North Carolina, Chapel Hill, North Carolina, USA; 5Eshelman School of Pharmacy, University of North Carolina, Chapel Hill, North Carolina,
USA; 6Center for Translational Medicine, Department of Medicine, Thomas Jefferson University Philadelphia, Philadelphia, Pennsylvania, USA;
7Department of Biochemistry and Molecular Biology, Center for Structural Biology, McKnight Brain Institute, University of Florida, Gainesville, Florida, USA

Page 2

2
www.moleculartherapy.org
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
expressed after AAV delivery to dystrophin deficient models has
been shown to: correctly localize to the sarcolemma, restore the
missing dystrophin-associated protein complex to the cell mem-
brane, ameliorate dystrophic pathology in mdx muscle, normalize
myofiber morphology, normalize cell membrane integrity, restore
missing dystrobrevin complex, partially restore α-syntrophin
association with the cell membrane, partially restore nitric oxide
synthase activity, reduce muscle fibrosis, reduce myofiber central
nucleation, improve whole-body endurance and muscle force
transduction, reduce kyphosis and limb deformation, and increase
general health and lifespan.5–8
Genetic strategies that target the muscular dystrophies will
ultimately require widespread delivery to a large volume of
skeletal musculature and/or cardiac tissue. Strategies to improve
transgene expression to the musculature have included the use
of AAV serotypes other than AAV2 and efforts to evolve tissue
specificity variants by directed capsid evolution as well as mosaic
vector with a mixture of capsid from different serotypes.9,10 It
is clear that no single natural AAV serotype will be useful for
every clinical application, nor will directed evolution evolve all
characteristics desirable for a clinical scenario simultaneously.
Any given serotype may contain biological characteristics both
beneficial and detrimental to the given clinical application.
Instead of attempting to “fit” a known AAV serotype to a disease
process or coevolve multiple traits in a single capsid, we chose
to use a rational approach to identify capsid regions on alter-
native AAV serotypes responsible for enhanced skeletal muscle
transduction and to combine these modifications into the AAV2
capsid which offers the benefits of a well-defined safety profile
coupled with purification ease. The availability of capsid protein
sequences from several AAV serotypes, muscle transduction pro-
files with different serotypes of AAV vector and antigenic epitope
information for AAV2, combined with the three-dimensional
structure of the AAV2 capsid11 provided us valuable information
to rationally design efficient vectors for clinical trials. Through
mutagenesis with insertion and substitution, a chimeric AAV2-
AAV1 vector, dubbed AAV2.5, was designed to contain desir-
able biological properties from both parent viruses. Compared
A263
T265
III
I
V
VII
II
N
N717
N717
N717-5f
A709
A263
T265
T265-5f
A706
A706-5f
A709-5f
A263-5f
c
b
a
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
AAV2
AAV3b
AAV1
AAV7
AAV8
AAV9
120140 160 180200 220
240260**
* *
280
300
320
340
360
460480
580600
700 720740
620640660680
500520540 560
380
400
420
440
120 4060 80 100
A706
A709IX
VI
VIII
IV
Figure 1 Amino acid candidates responsible for efficient skeletal muscle transduction. (a) Capsid amino acids of low skeletal muscle transduc-
ing serotypes (AAV2, AAV3) versus high skeletal muscle transducers (AAV1, AAV6, AAV7, AAV8, AAV9) were aligned using the Vector NTI program
(Invitrogen). Alignments were examined for distinct amino acids of AAV2 from the others. See text for additional modeling criteria. Amino acids
boxed or marked with arrows were deemed to be of interest. AAV2.5 is composed of the five amino acids indicated by * and @. (b) Location of the
five amino acids on a single VP subunit which were modified in the AAV2.5 variant. Notice that the five amino acids are located on opposite posi-
tions of one subunit. (c) Location of the same five amino acids (circles and arrows) in the context of an assembled AAV capsid pentamer. Notice that
the five amino acids are now in close proximity when two subunits are assembled. The five amino acid changes are located near the twofold axis of
symmetry. AAV, adeno-associated virus.

Page 3

Molecular Therapy
3
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
to AAV2, AAV2.5 has similar transduction efficiency in several
cell lines and binds to heparin sulfate in vitro. However, AAV2.5
induces stronger transduction in skeletal muscles than AAV2 and
demonstrated lower crossreactivity to AAV2 neutralizing anti-
bodies (Nab).
The novel AAV2.5 capsid offering improved skeletal muscle
gene transfer efficiency and potentially reduced immunogenicity
compared with naturally occurring serotypes was next evaluated
in a clinical trial for DMD utilizing the minidystrophin transgene
cassette. The initial development of AAV2.5-minidystrophin
clinical trial capitalized on the fact that AAV2 was the only sero-
type approved for clinical use, and AAV1 was the only other
AAV serotype under serious consideration for clinical studies.
To this end, a subset of amino acids (5aa) in AAV type 1 were
constructed into type 2 capsid backbone. The engineering and
testing of this chimeric capsid in clinical setting has provided
a paradigm where as the investigator is no longer obligated to
natural viral isolates for gene delivery and can therefore address
additional clinical concerns (e.g., capsid immune response) that
lie outside of primary objective of measuring therapeutic trans-
gene expression.
Therefore the novel nature of the proposed therapeutic strategy
and study population led to the inclusion of careful monitoring of
both humoral and cell-mediated immune responses in this phase I
clinical trial. We have recently reported elsewhere on unexpected
findings related to dystrophin immunity observed in this trial.12
A subset of patients were shown post-hoc to have had pre-existing
immunity to dystrophin epitopes believed to be expressed by rever-
tant myofibers, and cellular immune responses to minidystrophin
epitopes were also observed. We detail herein the immune response
to the novel AAV2.5 capsid as well as other study endpoints, such
as: (i) successful vector transgene delivery to all patients at each
dose based on PCR analysis of biopsy sectioning, (ii) no difference
in immune infiltration when comparing placebo to vector treated
arms, (iii) lack of detectable immune response in empty vector only
tissues, (iv) no CTL response to chimeric capsid at any dose. All
and all, this trial established that rationally designed AAV capsid
was safe, well tolerated and lays a foundation of customizing AAV
vectors that ideally suit the clinical objective (e.g., heart tropic, liver
detargeted, neuroselective, etc.).
results
rational design of AAV chimeric vectors
Several of the AAV serotypes characterized to date trans-
duce mouse skeletal muscle with greater efficiency than AAV2,
e.g., AAV1, AAV7, AAV8, and AAV9.13–17 In an effort to identify
candidate amino acids responsible for enhanced transduction of
skeletal muscle the VP1 amino acid sequences of these serotypes
with high muscular tropism were compared to AAV2 using an
alignment (Figure 1a). The criteria for selection of amino acid
candidates are: (i) they must differ from AAV2 and be similar to
the high muscle tropism serotypes, (ii) they must be located in
a structurally variable region (VR) on the capsid surface,18,19 or,
(iii) they must be located in an AAV2 antigenic region that is rec-
ognized by an antibody. Shown boxed in Figure 1a are the amino
acids which met all criteria with the exception of one amino acid
depicted by @ (residue 716, AAV2 Vp1 numbering) due to its
close proximity to four other amino acids.
Three variants were initially generated in which AAV2 was
modified to resemble AAV1: AAV2.5, AAV2-Q325T/T329V, and
AAV2-T450N/Q457N. In the AAV2.5 variant four residues were
substituted with AAV1 amino acids (Q263A, N705A, V708A,
T716N, AAV2 numbering) and one AAV1 amino acid (T265,
AAV1 numbering) was inserted into the AAV2 capsid (amino
acids indicated by the asterisks (*) and @ in Figure 1a and the
3D model shown in Figure 1b,c). These mutations are all on the
VRs of the virion surface (VR I and VR IX, Figure 1b). Based
on the AAV atomic structure, AAV2.5 was designed because resi-
dues 263 and 265 intercommunicate with residues 706, 708 and
717 (Figure 1c). The locations of all five amino acid mutations in
AAV2.5 are close to the A20 antigenic epitope.18,20,21 The AAV2-
Q325T/T329V variant (AAV2 numbering; depicted by arrows on
Figure 1a)) partially met the set criteria, the amino acids are con-
served in AAV8 but differ in AAV1, AAV7, and AAV9, and are not
close to a mapped AAV2 antigenic site. The AAV2-T450N/Q457N
(boxed only in Figure 1a) contains amino acids located in VR V.
The amino acid mutations in the three AAV2 variants did
not influence the ability to generate recombinant AAV vectors
compared to unmodified AAV1 and AAV2 capsids containing
the same transgene cassette as judged by physical particle titers
(Table 1). The AAV2.5 variant resembled AAV2 with respect to
its ability to bind heparin in affinity columns used for purification
(Table 1). Furthermore, the transduction profiles of these variants
in cultured HeLa, Cos, and 293 cells were more like that of AAV2
than AAV1 (Table 1).
Genetic modification of the AAV2 capsid can improve
muscle transduction
To determine whether amino acids responsible for in vivo
transduction could be accurately predicted, the variants with a pack-
aged luciferase gene (AAV2.5-luciferase, AAV2-Q325T/T329V-
luciferase, and AAV2-T450N/Q457N) were evaluated for their
table 1 the characteristics of chimeric variants
Virus
Physical titer
(vg/µl)
Heparin
binding
Heparin
competition293Helacos1
Muscle
transduction in vivo
A20 neutralizing
activity
Parental
serotype
AAV21.3–8.5 × 108
++ ++++++++++++++++++
AAV1
AAV2.5
Q425T/T329V
T450N/Q457N
1.3–15 × 108
5.0–9.2 × 108
6.9–13 × 108
4.4–9.3 × 108
−−+++ ++++
+++
+
++
−
−++
N/D
N/D
++
N/D
N/D
++++
++++
++++
++++
N/D
N/D
++++
++++
++++
N/D
N/D
Abbreviations: AAV, adeno-associated virus; N/D, not done; vg, vector genome.

Page 4

4
www.moleculartherapy.org
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
ability to transduce skeletal muscle following injection of equiva-
lent genome-containing particles into the gastrocnemius muscle of
BALB/c mice. Luciferase expression was evaluated over time using
biophotonic in vivo imaging and compared to the parental AAV1 and
AAV2 (for AAV2.5) or AAV2 (for AAV2-Q325T/T329V) viruses
(Figure 2a and b, respectively). In skeletal muscle, the AAV2.5 vari-
ant consistently produced higher transgene expression than AAV2
at all time points tested albeit not to the identical level as observed
with AAV1 (Figure 2a). AAV1 exhibited 5–12.5-fold higher levels
of light emission than AAV2; whereas AAV2.5 exhibited 1.8–5.5-
fold higher light emission than AAV2. Mice injected with AAV2.5
exhibited high levels of expression up to 8.4 months postinjection.
The expression level of the AAV2-Q325T/T329V variant was indis-
tinguishable from that of AAV2 (Table 1 and Figure 2b) and was not
tested further. The third variant, AAV2-T450N/Q457N exhibited a
3.5-fold enhancement in transgene expression over AAV2 (Table 1),
but only at a later time point postinjection (day 42 postinjection)
and was not tested any further.
unique antigenic properties of the AAV2.5 capsid
The antigenic profile of the AAV2.5 vector was explored first by
examining its recognition by the well characterized A20 anti-AAV2
monoclonal antibody.21 These studies revealed that the ability of
the A20 antibody to recognize the AAV2.5 vector is extinguished
compared to its strong recognition of AAV2 (Table 1). Thus this
data suggest that AAV2.5 may have an immune profile distinct
from AAV2.
Minimal crossreactive neutralizing antibodies exist between
AAV2 and other serotypes such as AAV122 (C. Li and RJ. Samulski,
unpublished results). To determine whether the neutralizing
antibody profile of AAV2.5 was more like AAV2 or AAV1, sera
from mice treated with AAV1-luciferase, AAV2-luciferase, or
AAV2.5-luciferase vectors were analyzed for neutralizing anti-
body crossreactivity. Shown in Figure 3a are the dilutions of the
sera from AAV2, AAV2.5, and AAV1 injected mice needed for
50% inhibition of infectivity of AAV2, AAV1, or AAV2.5 in 293
cells. The sera from AAV2-treated mice neutralized infectivity of
AAV2 more efficiently than the infectivity of AAV2.5 by fivefold.
A similar finding was observed in the examination of the ability
of AAV1 sera to inhibit the infectivity of AAV1 and AAV2.5. The
converse was found to be true for sera from AAV2.5 injected
animals which was observed to be four- and eightfold more
efficient at neutralizing AAV2.5 compared to AAV2 or AAV1,
respectively. Therefore AAV2.5 with minimal change of 5 aa has
antigenic properties that are distinct from those of the paren-
tal viruses suggesting that the engineered AAV2.5 capsid may
eliminate the AAV2 or AAV1 antibody recognized epitopes or
change virion three-dimension structure and such that it is less
372128
RLU/RPI
42
714
Days postinjection
1.00x108
1.00x107
1.00x106
600x106
2.5
a
b
AAV1
AAV2
AAV2
325 329
500x106
400x106
300x106
200x106
100x106
00x100
1.00x105
1.00x104
2842
Figure 2 evaluation of skeletal muscle transduction of AAV2 mutants.
(a) Skeletal muscle transduction of AAV1, AAV2, and AAV2.5 examined
over time (days 3, 7, 21, 28, 42) using in vivo biophotonic imaging.
The relative light units per region of interest in each injected mouse (n =
6) are graphed over time. (b) Graphical representation of quantity of
emitted light from transduction of AAV2 and AAV2-Q325T/T329V. The
relative light units per region of interest in each injected mouse leg (n =
6) are graphed over time. AAV, adeno-associated virus.
0
50
40
30
20
10
0
AAV2
5 × 109
No2–<10 10–<100100–
1 × 1010
5 × 1010
5 × 1010
AAV1
Nab titer
Nab titer
Percentage
Sera from mice
a
b
c
AAV2.5
AAV2
AAV1
AAV2.5
AAV2
AAV2.5
400
800
1,200
Figure 3 neutralizing antibody analysis to AAV2.5. (a) Crossreactive
Nab between AAV1, AAV2, and AAV2.5. C57 mice were immunized with
1 × 1010 particles of AAV/luc vectors via muscular injection. Thirty days
later, sera from three mice was collected for Nab analysis. (b) The effect
of AAV2 Nab on AAV2.5-induced transgene expression in vivo. Mice were
immunized with AAV2/AAT viruses (left three panels) or not immunized
(right panel), 2 months later, AAV2.5/luciferase (mouse right leg) and
AAV2/luciferase (left leg) vectors with different dosages were applied in
the same mice intramuscularly (5 × 109 particles, 1 × 1010 particles, or
5 × 1010 particles), imaging was taken 6 weeks later post-luciferase vector
injection. (c) Neutralizing antibody assay for human sera. The sera from
36 human subjects were detected for Nab against AAV2 and AAV2.5.
AAV, adeno-associated virus.

Page 5

Molecular Therapy
5
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
likely to be neutralized by the sera of animals pretreated with
AAV2 and AAV1.
Genetic modification of the AAV2 capsid enables
repeat administration of variant vectors
The serological data described above suggested that AAV2.5 was
antigenically distinct from the parental viruses. To test whether
this phenotype would make this variant refractory to pre-exist-
ing antibodies and allow readministration in subjects previously
treated with AAV2 vectors, in vivo studies were performed in
which mice previously injected with an AAV2-AAT vector were
subsequently injected intramuscularly (i.m.) with increasing
doses of either AAV2-luciferase or AAV2.5-luciferase vectors.
For direct comparison, each mouse received the same dose of the
two vectors in separate leg muscles. In vivo biophotonic imag-
ing performed at 6 weeks postinjection of the luciferase express-
ing virus revealed no detectable expression from either vector
at low doses (5 × 109 particles, Figure 3b). However, at higher
doses (1 × 1010 particles (Figure 3b) and 5 × 1010 particles per
leg (Figure 3b)) the AAV2.5-luciferase administered legs con-
sistently exhibited elevated transgene expression (~10–20-fold)
over its AAV2-luciferase injected counterpart (Figure 3b). The
enhanced transduction by AAV2.5 is interpreted as being due
to both its inherently higher skeletal muscle transduction com-
pared to AAV2, as evidenced by its ~2–5.5-fold higher transgene
expression over AAV2 observed in control mice with no previous
exposure to AAV2 (Figures 2 and 3b, right panel) as well as to its
ability to overcome pre-existing anti-AAV2 neutralizing antibod-
ies. However, although it is clear that the high dose of AAV2.5
vector could escape AAV2 neutralizing antibody partially in vivo,
we could not rule out the possibility that AAV2 neutralizing anti-
bodies completely block AAV2.5 transduction after muscular
injection in AAV2 pretreated mice when a much lower dose of
AAV2.5 was used.
summary of late preclinical studies
Sera from 36 individuals were screened for neutralizing antibodies
and pre-existing neutralizing antibody titers of >1:2 to AAV2 were
seen in 75% of individuals while titers of >1:2 to AAV2.5 were
seen in only 56% of individuals (Figure 3c). Of those individu-
als that have AAV2 Nab, a very high Nab antibody titer (≥100) is
found in 25% of these individuals; whereas, only 19.5 % of indi-
viduals with Nab to AAV2.5 exhibit high Nab titer. Generally the
titer against AAV2.5 is 2–20-fold lower than that against AAV2 in
the Nab positive population. The data from this human sera neu-
tralizing antibody assay again support the conclusion that AAV2
and AAV2.5 have different immune profiles.
A penultimate preclinical study was conducted to assess possi-
ble adverse interactions between the commonly prescribed corti-
costeroid prednisone and AAV-minidystrophin in C57/BL10 mice.
No discernable adverse interaction and influences on transgene
expression were observed (data not shown, see Supplementary
Materials and Methods—Preclinical AAV2.5-Minidystrophin
Studies).
The AAV-minidystrophin expression cassette including
the cytomegalovirus (CMV) promoter and the polyadenyla-
tion signal site in the vector plasmid was fully sequenced on
both strands except the inverted terminal repeats, which were
difficult to sequence. A pivotal good-laboratory-practice toxic-
ity and biodistribution study evaluated AAV-minidystrophin
delivery in C57/BL10 mice with timepoints up to 36 weeks post
vector administration and doses up to 1 × 1012 vector genomes/
kg, which was equivalent to 10 times the highest dose proposed
in the clinical trial. Overall, the AAV-minidystrophin vector
caused no significant toxicity in C57BL mice, with all animals
surviving until scheduled sacrifice. Effects on clinical chemistry
parameters in the toxicity study were within the normal refer-
ence ranges, and there were no differential effects observed by
histopathological analyses. There were no related effects on clini-
cal hematology parameters or gross necropsy observations. PCR
biodistribution evaluations indicated that the vector remained
concentrated in the injected muscle (data not shown, see
Supplementary Materials and Methods—Preclinical AAV2.5-
Minidystrophin Studies).
regulatory approval process
The clinical protocol and Appendix M documents were reviewed
by the Recombinant DNA Advisory Committee under the aus-
pices of the Office of Biotechnology Activities at the National
Institutes of Health (NIH). The Nationwide Children’s Hospital
institutional review board approved the study protocol. A
Pre-IND discussion was held with FDA before initiation of the
pivotal toxicology and biodistribution study, and upon conclu-
sion of preclinical studies the IND was submitted to CBER/FDA
(BB-IND 12936). The trial is registered with ClinicalTrials.gov
(Reference no. NCT00428935).
clinical study design
The experimental design was a randomized double-blind placebo-
controlled study where vector was administered into one bicep
and saline control in the contralateral arm. In a subset of patients,
we substituted AAV empty capsid for saline in an effort to distin-
guish an immune response to vector versus minidystrophin trans-
gene. The active phase of the study included a 2-week baseline
screening period, a 2-day inpatient period for vector injection and
acute toxicity monitoring, and a 2-year outpatient follow-up and
toxicity-monitoring period. Ongoing long-term follow-up will
continue out to 15 years post vector injection. (Supplementary
Figure S1). Blood and urine analyses were conducted in conjunc-
tion with outpatient clinics on post administration days 8, 15, 30,
43, 53, 60, 90, and 120 and months 6, 9, 12, 18, and 24.
The study enrolled six DMD boys ranging in age from 5
to 11 years and ranging in mass from 16 to 57 kg, each with
unique and defined dystrophin mutations. Four boys had been
on corticosteroid medication schedules before vector injection
(Table 2). All six patients were administered methylprednisolone
(2 mg/kg, limited to <1 g total) four hours prior to vector admin-
istration, with repeat doses on the following two mornings. A
MyoJect hypodermic needle (Oxford Instruments, Hawthorne,
NY) was used to deliver 1.2 ml of vector in three equivalent boluses
spaced 0.5 cm apart along an injection tract that was placed in a
longitudinal trajectory relative to the biceps muscle orientation.
Administration of AAV2.5-Minidystrophin to each subject was
staggered to allow for at least 6 weeks of follow-up before the

Page 6

6
www.moleculartherapy.org
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
next subject receiving the vector. There were two-dose cohorts of
three subjects each that received unilateral i.m. delivery of either
6 × 1011 vector genomes or 3 × 1012 vector genomes. Dosing was
defined by locally administered dose as the vector was mediat-
ing production of a structural protein and not a secreted protein,
and vector genomes remained localized to the region of the injec-
tion site. Internalized AAV capsid peptide competition for MHC
presentation would also be more heavily influenced by the local-
ized dosing. As a result of the inclusion of an empty capsid control
vector administered to the last two subjects, there was a full log
difference in the capsid dose between subjects 1–3 and 5 and 6.
Biopsies were taken at 1.5 month in the majority of patients and
out to 3 months in two patients (Table 2).
clinical observations
Physical examinations were unremarkable with no symptoms of
fever, lymphadenopathy, organomegaly, and no signs of inflam-
mation at the injection site. During the 2-year-long active phase
of the trial monitoring, no serious or mild adverse events have
been observed in any subject. A few minor adverse events were
observed including sore throats, rashes and nausea, but they were
not considered related to the vector administration as they are
commonly seen in this age group of subjects (detailed in ref. 12).
Hematology and chemistry panels that included an assessment of
liver function also indicated that the gene vector was well toler-
ated in all subjects (Supplementary Figure S2). There were no
abnormal elevations in the levels of creatine kinase, alkaline phos-
phatase levels, or lymphocyte counts in all blood samples tested.
AAV2.5-mediated gene delivery in patients
Biopsies were undertaken in four subjects (subjects 1, 3, 4, and 6) at
d43 post administration, and in two subjects (subjects 2 and 5) at
d90 post administration. The biopsy procedure utilized ultrasound
imaging to guide the retrieval of the injected tissue. A battery of
assays were undertaken on the biopsied muscle tissue and included:
PCR analysis of vector genomes, immunolabeling of minidystro-
phin, tissue histochemistry including CD8 T-cell infiltration.
Quantitative Pcr of the vector genome in human
muscle biopsies
The presence of the vector genomes was assessed in muscle biop-
sies from the left and right arms 1.5 and 3 months after injection.
The transgene DNA was detected in 6 out of 6 patients, in only one
arm with no detectable spread to the contralateral arm. The high-
est vector genome copy number per nucleus (diploid cell genome)
detected in each patient is shown in Table 2.
detection of minidystrophin protein in muscle tissue
biopsy
Cryothin-sections of the muscle biopsies were immunofluorescently
stained with antibodies recognizing the dystrophin N-terminal
region (present in both minidystrophin and endogenous revertant
myofibers) or the C-terminal region (absent in minidystrophin but
present in revertant myofibers). Limited minidystrophin expression
was observed with appropriate localization to the sarcolemma in
two of six subjects. In patients 3 and 6, only a few minidystrophin
positive myofibers were detected by anti-N-terminus antibody but
stained negative by anti-C-terminus antibody. However, in patients
1, 2, 4 and 5, no minidystrophin positive myofibers were detected,
suggesting that transgene expression was either very poor or extin-
guished in those patients (detailed in ref. 12).
Immunological studies
Post administration hematological testing indicated no abnormal
changes in white blood cell counts, and there were no changes in
markers of liver toxicity (Supplementary Figure S2). However, clin-
ical AAV2 administration has previously been associated immune
recognition of AAV capsid peptides,23 and therefore we evaluated
cellular and humoral immunological recognition of AAV2.5.
cell-mediated immunity to AAV2.5
Peripheral blood mononuclear cells (PBMCs) of the subjects were
tested for recognition of potential AAV capsid and minidystrophin
peptide epitopes by Elispot assay.23 Antigens used to stimulate these
responses were pools of overlapping peptides spanning the AAV
table 2 the clinical data in patients with AAV2.5/minidystrophin muscular delivery
subject
1
Age
8
deletion
45
Mass
(kg)
22.6
steroid
use
Y
Pre-nab
titer
<1:2
Vector
dosea
6.0 × 1011
Placebo
control
Saline
capsid doseb
Low
6.6 × 1011
Low
6.6 × 1011
Low
6.6 × 1011
Intermediate
3.3 × 1012
High
6.6 × 1012
High
6.6 × 1012
time of
biopsy
D43
Vector
genomec
0.75
Gene
expression
ND
revertant
fibers
30–125
29 5028.5Y1:8006.0 × 1011
SalineD900.01ND0–9
3946–5035.5N<1:26.0 × 1011
SalineD43 0.61weak 0–9
4549–5415.8N <1:23.0 × 1012
Saline D432.56ND0–11
511 3–17 57.1Y 1:1003.0 × 1012
Empty capsidD90 0.08NDND
6946–5228.7Y1:23.0 × 1012
Empty capsidD431.42weak1–25
Abbreviations: AAV, adeno-associated virus; ND, none detected; qPCR, quantitative PCR.
aAAV2.5 minidystrophin vector genome dose (vector genomes/patient). bTotal capsid dose (minidystrophin + empty capsid in subject’s 5 and 6) capsid particles/
patient. cVector genome copy number isolated per nucleus as determined by qPCR (skeletal muscle cells are multinucleated).

Page 7

Molecular Therapy
7
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
capsid VP1 protein. Peptides were 20 amino acids in length over-
lapping by 10 residues and thus would cover all epitopes of VP3 and
stimulate CD4+ helper and CD8+ cytotoxic T cells. PBMCs from
the subjects were cultured with the AAV capsid peptide pools for
48 hours and then developed for interferon (IFN)-γ spot forma-
tion. Subject 1 showed AAV2.5 specific responses that substantially
exceeded the threshold of 50 spot-forming cells (SFC) per million
PBMC collected at two timepoints, d30 and d90, but not at other
timepoints including d43, d53, and d60 post vector injection. In
further analyses, the d30 time-point was confirmed as positive after
stimulation with peptide pools spanning the capsid protein. Cell
lines generated from this sample could not be expanded to support
additional characterization. We did not observe significant T cell
responses to AAV2.5 capsid antigens seen at any time-point before
or after vector administration in any other subjects (Figure 4), how-
ever, some samples from subjects 2, 3, and 4 were weakly positive at
various time points (sporadically >50 SFC/106 PBMC).
lack of cd8 infiltration into injected muscle
Cellular infiltrates were quantified in a blinded examination
of bicep muscle biopsies. Evaluation of muscle biopsy sections
immunofluorescent stained with anti-CD8 antibody showed
no statistic difference in the counts of CD8+ cells between
AAV-minidystrophin injected and placebo (saline and empty
capsids) injected biopsies (52.1 ± 35.2 and 41.9 ± 23.2 cells/mm3
cross-section area, respectively). These cell counts were within the
range commonly observed in patients with DMD.
Humoral immunity to AAV2.5
The AAV capsid itself is a completely foreign protein and there-
fore a humoral immune response to vector administration was
anticipated. Pre-existing humoral immunity to AAV was not an
exclusion criterion in this study with the rationale that the rel-
evance of in vitro assays of vector neutralization before i.m.
administration is unknown due to a lack of previous clinical data.
In order to characterize the humoral response to AAV2.5, titers
of AAV neutralizing antibodies in serum were also evaluated in
an assay involving in vitro transduction in the presence of sera
from patients. Subject’s sera were screened for neutralizing anti-
bodies to both seroprevalent wild-type AAV2 and the synthetic
AAV2.5 capsids. As detailed in Table 2, subjects 2 and 5 were
observed to have significant pre-existing neutralizing antibody
200
150
Subject 1
Subject 3
Subject 5
Subject 2
AAV capsid peptide pool 1
AAV capsid peptide pool 2
Subject 4
Subject 6
100
IFN-γ SFC/106 PBMC
50
0
200
150
100
IFN-γ SFC/106 PBMC
50
0
200
150
100
IFN-γ SFC/106 PBMC
50
0
Base8 1530 4353 60 90120 180 365 Base8 1530 43536090 120 180 365
Figure 4 temporal t cell response to AAV2.5 capsid. Two capsid peptide pools were comprised of peptides spanning the AAV2.5 capsid sequence.
Elispot assays measured interferon (IFN)-γ release upon peptide exposure, with the threshold for peripheral blood mononuclear cell (PBMC) recogni-
tion of an epitope within the peptide pool being 50 spots/106 PBMCs. Temporal responses are shown for subjects 001–006 in separate panels for the
two peptide pools. AAV, adeno-associated virus.

Page 8

8
www.moleculartherapy.org
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
titers to both AAV2 and 2.5, with baseline titers of 1:800 and 1:100
to both vectors, respectively (titers are the inverse of the highest
serum dilution mediating 50% neutralization to either capsids).
Baseline titers were <1:2 for both AAV2 and AAV2.5 for subjects
1, 3, and 4, and Subject 6 had a baseline titer of 1:4 for AAV2
and 1:2 for AAV2.5. Classical humoral responses were observed
after AAV2.5-minidystrophin administration, with increases in
Nabs that peaked from weeks 2 to 6 and ranged from 50× to
1,000× baseline titers (Figure 5). Subject 2 who had high base-
line titers was the only subject who did not exhibit a classical post
administration strong increase in neutralizing antibody titers,
with titers at all time points staying within the range of 1:800 to
1:2,000, approximately a 2.5× increase.
dIscussIon
Our study represents the first clinical trial of a synthetic rationally
designed AAV vector, AAV2.5, and provides a preliminary insight
to the clinical tolerability of this first-in-class gene vector. In this
study, we have utilized the atomic structure of the AAV2 virion
and homologous models generated for muscle tropism serotypes
and VP1 sequence alignments as a guide to rationally design an
AAV2 variant (AAV2.5) with altered transduction and antigenic
profile from AAV2. We have successfully identified amino acids
that differ between AAV2 and muscle tropism serotypes and
located in a structurally VR. Most strikingly, the genetically modi-
fied AAV2.5 vector demonstrated the enhanced muscle tropism
and different antigenic properties from parents (AAV2 and
AAV1). Furthermore, this study showed that i.m. administration
of a laboratory-derived novel AAV2.5 gene vector encoding
minidystrophin at capsid doses up to 6.6 × 1012/kg capsid particles
was well tolerated by all subjects, with no vector related adverse
events observed.
AAV rational design
The Q263A substitution and T265 insertion in AAV2.5 are located
in one of the most VRs in the common VP3 capsid protein (struc-
turally and at the amino acid sequence level) when AAVs are com-
pared. They are located in VR I. A structural comparison of the
loop containing these amino acids on the capsid surface of the
atomic structures of AAV1 and AAV2 shows a conformational
difference in the main chain (Figure 6). This surface loop region
is also predicted to be different in the homologous model gen-
erated for AAV2.5 from the atomic coordinates of AAV1 (data
not shown) and AAV211 (Figure 6). This is one of the three most
structurally divergent surface loops in main chain conformation
when the structure of AAV8, another muscle tropism serotype,
was compared to that of AAV2.24 The Q263A substitution and
T265 insertion were also among the nine residues identified as
contributing to AAV1 skeletal muscle transduction by the AAV1/
AAV2 domain swap study by Hauck et al.25 Like the AAV2.5 vec-
tor, their AAV-221-IV vector approached but did not quite reach
the transduction profile of AAV1, strongly suggesting other capsid
motif may be required.
Gene expression and immunological recognition
AAV2.5-mediated minidystrophin delivery to skeletal muscle was
confirmed by vector DNA quantitative PCR (qPCR). Retrieved
genome copy numbers (up to 2.56 copies/diploid genome in sub-
ject 4) were within the range of that observed in previous trials of
i.m. administration of AAV vectors (Supplementary Table S1).
However, the overall minidystrophin transgene expression
detected in this clinical trial was low. Limited minidystrophin
positive myofibers were detected in patients 3 and 6, but not
in other patients. However, patients 2 and 5 had elevated T
cell Elispot responses to dystrophin epitopes from endogenous
revertants and/or transgene product after vector delivery, sug-
gesting a transient transgene expression.12 As an alternative to
the hypothesis that minidystrophin-stimulated transient rec-
ognition by T cell PBMCs resulted in loss of expression before
biopsy, the low levels of transgene expression could be partially
explained by numerous factors, e.g., pre-existing anti-AAV neu-
tralizing antibodies (seen in patients 2 and 5), pre-existing T
cell responses to endogenous revertant dystrophin epitopes that
are also present on minidystrophin (seen in patients 2 and 4),12
silencing of the CMV promoter in human dystrophic muscles,
disease-associated inflammation, and low tropism of AAV2.5 to
human muscle tissue.
With respect to these possibilities, we observed significantly
higher vector genome copy numbers in muscle from patients who
both lacked pre-existing Nab and who were also biopsied at ear-
lier time-point. Subjects 4 and 6, who received the high vector
dose, also had the highest vector genome copy numbers detected
(2.56 and 1.42 vg/dg, respectively), consistent with animal stud-
ies. Although subjects 4 and 6 retained the highest genome copy
number, they showed no detectable (subject 4) or very low levels
100,000
10,000
1,000
100
10
1
−50510
Timepoint relative to AAV2.5 administration (weeks)
AAV2.5 neutralizing antibody response profile
Inverse of highest NAB dilution
Subject 1
Subject 4
Subject 5Subject 6
Subject 2Subject 3(Low-capsid dose group)
(Medium-capsid dose)
(High-capsid dose group)
1520 30 40 50 60
Figure 5 temporal neutralizing antibody response to AAV2.5.
The temporal profile of the highest dilution of serum that neutralizes
AAV2.5 transduction in vitro is shown for all six subjects. Subjects 1–3
received the same dose of AAV2.5 capsid, 6.6 × 1011 capsid particles.
Subject 4 received 3.3 × 1012 capsid particles. Subjects 5 and 6 received
6.6 × 1012 capsid particles due to administration of empty capsid pla-
cebo (see table 2). AAV, adeno-associated virus.

Page 9

Molecular Therapy
9
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
of minidystrophin expression (subject 6), a strong indication of
CMV promoter silencing in the injected muscle. Lower levels of
vector DNA were detected in subjects 1 and 3 (0.75 and 0.61 vg/dg,
respectively), but only subject was found to have minidystrophin
positive myofibers. Subjects 2 and 5 had high levels of pre-existing
neutralizing antibodies (1:800 and 1:100) to AAV2.5. They also
had elevated T cells recognizing dystrophin. The humoral and
T-cell immunity could have jointly contributed to the very low
vector genome copy numbers (0.01 and 0.08 vg/dg, respectively)
and the lack of detectable minidystrophin expression, again as
predicted from animal studies (see Figure 3). Unfortunately
due to the small size of this safety study, the variability in sub-
ject characteristics and the two biopsy timepoints, it is premature
to draw definitive conclusions regarding the association between
pre-existing immunities and gene transfer and expression effi-
ciency, but mark a key observation that has been noted by more
recent AAV studies (>50 exclusion of patients with pre-existing
Nab in AAV1 CUPID trial).
With the caveat that many experimental differences exist
between trials, it remains interesting to compare the current
results with previous clinical studies. While minidystrophin
expression was observed only at low levels in one subject from
each dose cohort, there are similarities with clinical observations
in other trials using similar doses of AAV via i.m. injections in
the AAV2-FIX, AAV2-AAT, and AAV1-AAT skeletal muscle
clinical trials (Supplementary Table S2). In this present study, we
observed T cell recognition of nonself-epitopes encoded by the
minidystrophin in subject 5.12 Unpredicted T cell responses against
self-epitopes encoded by the minidystrophin and the endogenous
revertant dystrophin were also observed in subjects 2 and 4, appar-
ently as a result of priming by rare revertant myofibers. In contrast
to this and previous studies, in the recently reported LGMD 2D
trial with AAV1 α-sarcoglycan,26,27 no T-cell immunity against
either the endogenous or the transgene-encoded α-sarcoglycan
was detected. In fact, 5 out of 6 patients had efficient and long-term
transgene expression. The starkly different outcomes between the
DMD and LGMD 2D clinical trials could be attributed to differ-
ent complexity and underlying pathology of the diseases, choos-
ing of patients (deletion versus point mutations) and the use of
different AAV serotype vectors (AAV2.5 versus AAV1), promot-
ers (nonspecific CMV versus muscle-specific tMCK) and the
transgenes (minidystrophin versus α-sarcoglycan). The ongoing
oligo exon skipping studies for DMD should provide supportive
evidence for “revertant myofiber” hypothesis or simply note that
peripheral T-cell recognition is not indicative of cytotoxic T-cell
response.
In the present study, cell-mediated responses to the AAV
capsid peptides were absent at baseline in all subjects. This
absence of T cell response to AAV capsid epitopes in subjects 2
and 5 who had high levels of pre-existing AAV Nab was somewhat
unexpected. Interestingly, we did not observe any T cell response
to AAV capsid in subjects 5 and 6 who received a 1 log higher
AAV2.5 capsid dose than Cohort 1 due to the fivefold dose escala-
tion and the additional equal dose of empty capsid vector in the
contralateral arm. In a minority of subjects, Elispot IFN-γ release
approached and sometimes exceeded the threshold values con-
sidered positive, but there was no consistent pattern or obvious
relationship to the timing of vector administration.
significance of immunological data
This trial was the first to use a laboratory-derived novel AAV
capsid to mediate gene delivery. The vector was well tolerated
and there were no vector-related adverse events. These data are
similar to a meta-analysis of all adverse events reported in 12
AAV trials enrolling >400 subjects overall, in which no AAV vec-
tor related adverse event pattern was observed. The safety pro-
file observed in this trial supports further clinical evaluation of
the AAV-minidystrophin. However, T cell-mediated immune
AAV2.5 (light blue), AAV1 (purple);
AAV2 (blue)
Balls:
A263, T265:RED
A706, A709, N717: magenta
c
a
b
Figure 6 topology of AAV2.5 virion. Surface topology view of the five VP monomers (in different colors) immediately surrounding the reference
monomer in light blue. The symmetry operation that brings the monomer into contact is given in the labels. The positions of the AAV2.5 residue
mutations are colored as in Figure 1b,c. (b) Top panel is a close up of the boxed region in the a. The bottom panel shows the same image rotated
by ~75° and shows that the region containing the 263/265 amino acids are raised on the capsid surface. (c) Image showing a surface view (same as
in Figure 1) of the loop containing the 263/265 region, VR III and VR IX (close to the 706, 709, 717 mutations) for AAV1 (purple), AAV2 (blue), and
AAV2.5 (light blue). The positions of the AAV2.5 mutants are shown in the balls and labeled. AAV, adeno-associated virus.

Page 10

10
www.moleculartherapy.org
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
recognition of vector derived neoantigens must continue to be
monitored in future trials to determine whether there are dose
thresholds or specific response stimuli that can be avoided.
Notwithstanding, the relationship remains elusive between T cell
recognition and a response against transduced cells. Manno et al.
reported contemporaneous increases in PBMCs that recognized
the AAV2 capsid with an increase in transaminase levels and
reduction in circulating FIX concentrations.23 In a recent study
of AAV1-α-antitrypsin (AAT) gene delivery to muscle,28,29 T cell
recognition of AAV capsid epitopes was also observed. In con-
trast to the FIX trial, quantitative examination of blood AAT levels
suggested that the PBMC T cell recognition of AAV capsid did
not have an impact on gene expression as it was stable over an
extended duration.28,29 A recent publication by the Walker group
suggests that in some scenarios peripheral PBMCs may become
apoptotic in muscle tissue and a functional immune response
targeting an antigen may be muted or curtailed.30 The impact
of immunological recognition is likely to be multifactorial. For
example, Vandenberghe described the generation of high levels
of T cells against capsid of serotypes capable of binding heparin
or heparan sulfate proteoglycans.31 Although AAV2.5 contains
heparin binding capabilities we observed weak or minimal T cell
activation to the AAV2.5 capsid using ELISPOT, suggesting that
besides vector serotype and tropism, the route of administration,
capsid particle dose, choice of promoter, and underlying disease
pathology are likely to be critical determinants of any immune
effector response.
Future directions
Pre-existing humoral immunity will continue to be a critical fac-
tor to be considered in the selection of study populations and
in the determination of enrollment criteria. The general popu-
lation is exposed predominantly to AAV2 in the first decades
of life.32 Patients with pre-existing high-titer Nab should be
avoided in future trial designs. Effort is needed to continue the
development of novel chimeric AAV vectors that evade the com-
mon pre-existing Nab.33,34 Diminished seroprevalence of pre-ex-
isting crossreactive neutralizing antibodies is a goal of novel capsid
development. The preliminary evaluation of AAV2.5 in vivo and
preliminary study of humoral responses showed reduced cross-
reactivity to AAV2.5 in animal and human sera. The pre-exist-
ing humoral immunity to AAV2.5, we observed is attributable
to AAV2 crossreactivity, with lower titers of anti-AAV2.5 Nab
being observed compared with titers of anti-AAV2. These results
support the predictive utility of the assays and methods used to
develop the AAV2.5 capsid. However, we did observe less of a dif-
ference in crossreactivity between AAV2 and AAV2.5 Nab titers
in the current study than in the preclinical evaluations, which
could be due to the small sample sizes of both studies. As was
seen in our studies, this ability to develop a new generation of
more tissue-selective gene vectors must be tempered with our
limited abilities to predict human immunological recognition/
responses with the available preclinical models. Recent find-
ings suggest that pre-existing anti-AAV8 neutralizing titers of
1:10 or greater may result in altered biodistribution patterns in
nonhuman primates.35 This observation is of limited relevance
to the current study that involved direct i.m. administration.
However, future trials involving delivery via the vasculature may
be impacted by this finding.
Immunological recognition of both capsid and transgene-en-
coded epitopes should be carefully monitored in any ongoing or
future clinical study that targets DMD, in particular, potentially
neoantigens not previously presented to the immune system, as
well as antigens generated by revertant myofibers (see ref. 12).
The route of administration is known to impact upon the host
immune response,36,37 and future trials of isolated limb delivery
(Fan et al., this issue) and eventual systemic delivery will need to
continue to monitor for immune responses. Intravascular AAV
delivery results in some level of gene transfer to the liver, thymus,
and spleen, which has been associated with transgene product tol-
erance in preclinical studies.38 Our preliminary preclinical limb
perfusion study in the DMD dog model using AAV2i8, AAV8, or
AAV9 mediated long-term minidystrophin expression (data not
shown). Although the immunological mechanisms behind these
observations are unclear, either immune tolerance or T-cell apopto-
sis or both could favor blood-vessel-mediated AAV vector delivery
to the muscle, which has a more uniform vector and antigen dis-
tribution than local i.m. injection. Given that regional limb deliv-
ery and eventual body wide delivery of the AAV vectors require
proportionally higher vector doses of up to 1015 vg per patient, this
will be associated with unprecedented foreign antigen load. The
success of future clinical development of AAV based gene transfer
for DMD will continue to require careful evaluation of the relation
between vector doses, cell-mediated immune recognition and T
cell immune response at all levels of preclinical and clinical devel-
opment. To this end we have initiated the next generation AAV
capsid for efficient muscle delivery via vascular route that detargets
liver but efficiently transduces muscle and cardiac tissue.39,40
MAterIAls And MetHods
Plasmid and DNA mutagenesis. The starting plasmid for these experi-
ments was the packaging plasmid pXR2.41 All plasmid mutagenesis was
performed using either the Quick Change Multi Site Mutagenesis or the
Quick Change Site Directed Mutagenesis kits (both from Stratagene, Santa
Clara, CA). When possible, mutagenesis of multiple amino acids was done
concurrently. When concurrent mutagenesis was not possible because dis-
tance between nucleotides was not optimal, mutagenesis was performed
sequentially within inclusion of previous nucleotide changes within the
new primer to prevent reversion. Accuracy of the nucleotide changes was
verified by DNA sequence analyses followed by DNA subcloning to elimi-
nate any unwanted artifacts generated by the mutagenesis.
Recombinant virus production (preclinical studies). All recombinant
AAV viruses were generated using the standard triple transfection method
using the XX6-80 adenoviral helper plasmid with a packaging plasmid
(either AAV1, AAV2 or a modified packaging plasmids) and an inverted
terminal repeat plasmid [containing either green fluorescent protein,
luciferase, or hAAT].42
rAAV was purified using standard methodology and the physical
titer of the different viral preparations was evaluated using dot blot
hybridization.42 For direct comparison of in vivo transduction efficiency,
the vectors being compared within one experiment were evaluated on the
same dot blot.
In vitro vector characterization. Batch binding of rAAV to heparin agarose
was performed as described previously.43 The ability of the virus encoding
the firefly luciferase complementary DNA to transduce cultured cells was
also evaluated. Briefly, 293, Cos1, or HeLa cells were infected with 1,000

Page 11

Molecular Therapy
11
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
particles of rAAV/cell and Ad dl309 at a multiplicity of infection of 10. Cells
lysates were then assessed via luminometry 24 hours post-transduction.
Animals. C57BL and BALB/c mice were purchased from Jackson
Laboratories (Bar Harbor, ME). Animals were maintained and treated in
accordance with the Animal Care and Use Committee of UNC Chapel Hill.
All care and procedures were in accordance with the Guide for the Care
and Use of Laboratory Animals (DHHS Publication No. [NIH] 85-23), and
all procedures received prior approval by the University of North Carolina
Institutional Animal Care and Usage Committee.
In vivo bioluminescence imaging of luciferase. 1 × 1010 viral
genome-containing particles (vg) were injected into the gastrocnemius
male BALB/c mice. A total of six limbs were injected for each vector type
using 25 µl of virus. The mice were anesthetized by intraperitoneal admin-
istration of 2.5% averdin. A solution of the luciferase substrate luciferin
(150 mg/kg; Invitrogen, Carlsbad, CA) was injected intraperitoneally. A
grey scale reference image of animals was generated and mice were imaged
for 5 minutes using a NightOwl cabinet with charge couple device camera
(Berthold Technologies, Bad Wildbad, Germany). Total photon emission
from selected and defined areas within the images of each mouse was quan-
tified with the WinLight32 software (Berthold Technologies). The photon
signal was presented as a pseudocolor image representing light intensity
(red most and blue least intense). This image was superimposed on the
reference image for orientation.
Detection of neutralizing anti-AAV antibodies. For comparison of
humoral immune response among AAV capsids in mice, 100 µl phosphate-
buffered solution containing 1 × 1010 particles of rAAV/GFP virus for each
capsid type was intraperitoneally injected into 6–10-week-old mice at day
0, and boosted at day 14. Blood sera were collected from mice via retro-
orbital plexus at indicated time points or from human via peripheral vein.
Nab titer was assayed using methodology described by Moskalenko et al.
with slight modifications.44 Briefly, 293 cells were seeded in a 48-well plate
at a density of 105 cells/well in 200 µl DMEM containing 10% fetal bovine
serum. The cells were cultured for 3–4 hours at 37 °C and allowed to adhere
to the well. AAV-GFP (1 × 108 particles) was incubated with mice sera at
serial dilution with phosphate-buffered solution for 2 hours at 4 °C in a
total volume of 25 µl. The mixture was added to cells in a final volume of
200 µl which contained 4 × 10 6 particles of adenovirus dl309 and incu-
bated 24 hours or 48 hours at 37 °C. GFP-expressing cells were counted
under a fluorescent microscope. The neutralizing antibody titer was cal-
culated using the highest dilution where the percentage of GFP expressed
cells were 50% less than control without sera.
Readministration. For readministration application of AAV vectors
in vivo, mice were immunized with AAV/AAT vectors by muscular injec-
tion, and then challenged with AAV/luciferase vectors 2 months later. The
luciferase transgene expression was measured 6 weeks following last vector
injection.
Human subjects (preclinical). The sera from adult volunteers were
screened for neutralizing antibodies to AAV. Individuals were research-
ers at the University of North Carolina at Chapel Hill (Chapel Hill, NC).
The study was approved by the institutional review board of the University
of North Carolina at Chapel Hill.
Homologous model building for AAV2.5. The available atomic coordinates
for AAV2 (ref. 11; PDB accession No.1LP3) and AAV1 (L. Govindasamy
and M. Agbandje-McKenna, unpublished results) were used to generate a
homologous model for the AAV2.5 variant to enable comparison with the
parental viruses. The model was generated by submitting the VP3 sequence
of AAV2.5 to the SWISS-MOLEL model building program with the coor-
dinates of structures of the AAV1 and AAV2 VP3 supplied as template.45
The model of AAV2.5 VP3 was then superimposed onto the AAV1 and
AAV2 structures using the SSM option of the COOT program for com-
parative analysis.46,47 To visualize the location of the mutated residues in the
context of an assembled capsid in the predicted structure of AAV2.5 ico-
sahedral symmetry operators were applied to the VP3 model coordinates
by matrix multiplication using the program O and the amino acids were
highlighted in the context of the symmetry related VPs.48 Figures show-
ing ribbon drawings and surface representations were generated using the
program PyMOL.
Regulatory approval process. The clinical protocol and Appendix M
documents were reviewed by the Recombinant DNA Advisory Committee
under the auspices of the Office of Biotechnology Activities at the NIH.
The Nationwide Children’s Hospital institutional review board approved
the study protocol. A Pre-IND discussion was held with FDA prior to
initiation of the pivotal toxicology and biodistribution study, and upon
conclusion of preclinical studies the IND was submitted to CBER/FDA
(BB-IND 12936). The trial is registered with ClinicalTrials.gov (Reference
no. NCT00428935).
Study design. The experimental design was a randomized double-blind
placebo-controlled study where vector was administered into one bicep
and saline control in the contralateral arm. In a subset of patients, we sub-
stituted AAV empty capsid for saline in an effort to distinguish immune
response to vector versus minidystrophin transgene. The active phase of
the study included a 2-week baseline screening period, a 2-day inpatient
period for vector injection and acute toxicity monitoring, and a 2-year out-
patient follow-up and toxicity-monitoring period. Ongoing long-term fol-
low-up will continue out to 15 years post vector injection (Supplementary
Figure S1). Blood and urine analyses were conducted in conjunction with
outpatient clinics on postadministration days 8, 15, 30, 43, 53, 60, 90, and
120 and months 6, 9, 12, 18, and 24.
The trial was comprised by two-dose cohorts, with each patient
receiving bilateral injections, with double-blinded randomization
of AAV2.5-minidystrophin vector to one bicep and placebo to the
contralateral bicep. Four patients received a saline placebo injection
into the contralateral bicep, and two patients in dose cohort II received
an empty capsids placebo injection into the contralateral bicep with an
identical dose as of the minidystrophin-containing full capsids.
qPCR analysis of AAV2.5 vector genomes. DNA was isolated from frozen
muscle biopsies using a Qiagen DNeasy Tissue kit. 0.5% of the recovered
DNA from each sample was used as template in duplicate qPCR using
three primer sets targeting the minidystrophin transgene and two primers
targeting human genomic DNA reference sites at the laminB2 and globin
loci (Supplementary Table S3). The minidystrophin primers crossed exon
splicing junctions, so they specifically amplify the complementary DNA
transgene but not the endogenous dystrophin gene. qPCR was performed.
Minidystrophin vector. The vector genome packaged in AAV2.5 encoded
the aminoterminal actin-binding domain, five rod repeat domains (R1,
R2, R22, R23, and R24), three hinge domains (H1, H3, and H4), and the
cysteine-rich domain of the human dystrophin gene. The CMV immediate
early promoter regulated transgene expression, along with a bovine growth
hormone derived polyadenylation signal.
Both the test vector and the empty capsid reagents were generated by
transient transfection of mammalian HEK293 cell cultures, undertaken at
the Human Applications Laboratory, University of North Carolina.
With Dot–Blot tittering, the total lot yield was calculated as 1.72 × 1013
vector genomes at a concentration of 4.4 × 1012 vector genomes/ml. This
final product was then further diluted in sterile 1× phosphate-buffered
solution 5% sorbitol to generate the clinical dose concentrations (see
Table 2) for additional characterization and lot release testing. The
clinical dose is determined by the vector genome dose, which is 6 × 1011
vector genomes/subject in the first low dose cohort and 3 × 1012 vector
genomes/subject in the second high dose cohort.

Page 12

12
www.moleculartherapy.org
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
Apart from two critical differences, the manufacturing procedures
used in the production of the empty capsid lot were identical to those
used in the production of rAAV2.5 minidystrophin. The first difference
is merely the absence of the AAV-minidystrophin-containing plasmid
in the cell transfection stage. The second difference is the use of a dot
blot western to titer the empty capsids (in units of capsid particles) in the
absence of a single-stranded DNA insert. The AAV2.5 CMV-empty capsid
lot underwent nearly identical release testing (titering assays differed) to
that performed for AAV2.5 CMV-3978, and in addition was released with
a CoA using the same procedures.
Estimation of capsid particle titer. Transmission electron microscopy
was undertaken to evaluate the ratio of genome-containing particles to
the total number of genome-containing (full) or empty capsid particles,
with the ratio estimated as being 0.907. The total dose of AAV2.5 capsid
particles was then estimated from this ratio and the genomic particle titer
generated by the dot blot assay.
Evaluation of T-cell reactivity to capsid derived peptides. Peripheral
blood T cell responses to the novel AAV capsid were quantified by IFN-γ
ELISpot assay, as described before.12 Briefly PBMC isolated on Ficoll
hypaque gradients were cultured with synthetic peptides (20 amino acids
in length, overlapping by 10 residues) that spanned the VP1 capsid protein.
To identify individual peptides within a pool that elicited IFN-γ activity,
so that each peptide was present in two of the intersecting mapping sub-
pools. After incubation at 37 °C for 36 hours. IFN-γ SFC were counted.
Fewer than 10 SFC/well were observed with peptides from a control pool
(enhanced GFP). Responses were considered positive when SFC exceeded
50/106 PBMC in duplicate wells.
suPPleMentArY MAterIAl
Figure S1. Trial procedures timeline.
Figure S2. Temporal profiles of blood chemistry, hematology, and
urinalyses.
Table S1. Genome detection in subject biopsies.
Table S2. Comparison of AAV.Minidys dose–response-related expression.
Table S3. Quantitative PCR primers.
Materials and Methods.
AcKnoWledGMents
The participation and cooperation of the trial participants and their guard-
ians was greatly appreciated. The authors would like to acknowledge
the following individuals for providing expert assistance on this project,
including trial planning and conduct, data collection and analysis, and
reviewing this manuscript: Jerry Mendell, Chris Walker, Chris Shilling,
Kristine Campbell, L. Radino-Klapac, Z. Sahenk, S. Lewis, G. Galloway,
V. Malik, B. Coley, R. Clark, D. McCarty, S. Hesterlee, J. Larkindale,
S. Moore, V. Alekseeva, Aravind Asokan. D.E.B., R.J.S., and X.X. hold
stock in Asklepios BioPharmaceutical Inc. C.L., S.J.G., R.J.S., and X.X.
have received research funding from Asklepios BioPharmaceutical Inc.
This work was funded by a Translational Corporate Grant to Asklepios
BioPharmaceutical Inc. from the Muscular Dystrophy Association USA.
This study was also supported by NIH research grants 1R01AI080726
and 5R01DK084033 (to C.L. and R.J.S.), 5U54AR056953 (to R.J.S.) and
R01 AI072176 (R.J.S. and M.A.-M).
reFerences
1. Koenig, M, Hoffman, EP, Bertelson, CJ, Monaco, AP, Feener, C and Kunkel, LM (1987).
Complete cloning of the Duchenne muscular dystrophy (DMD) cDNA and preliminary
genomic organization of the DMD gene in normal and affected individuals. Cell 50:
509–517.
2. Emery, AE (2002). The muscular dystrophies. Lancet 359: 687–695.
3. Pichavant, C, Aartsma-Rus, A, Clemens, PR, Davies, KE, Dickson, G, Takeda, S et al.
(2011). Current status of pharmaceutical and genetic therapeutic approaches to treat
DMD. Mol Ther 19: 830–840.
4. Gregorevic, P, Allen, JM, Minami, E, Blankinship, MJ, Haraguchi, M, Meuse, L et al.
(2006). rAAV6-microdystrophin preserves muscle function and extends lifespan in
severely dystrophic mice. Nat Med 12: 787–789.
5. Wang, B, Li, J, Fu, FH and Xiao, X (2009). Systemic human minidystrophin gene
transfer improves functions and life span of dystrophin and dystrophin/utrophin-
deficient mice. J Orthop Res 27: 421–426.
Wang, B, Li, J and Xiao, X (2000). Adeno-associated virus vector carrying human
minidystrophin genes effectively ameliorates muscular dystrophy in mdx mouse
model. Proc Natl Acad Sci USA 97: 13714–13719.
Wang, B, Li, J, Fu, FH, Chen, C, Zhu, X, Zhou, L et al. (2008). Construction and analysis
of compact muscle-specific promoters for AAV vectors. Gene Ther 15: 1489–1499.
Wang, B, Li, J, Qiao, C, Chen, C, Hu, P, Zhu, X et al. (2008). A canine minidystrophin
is functional and therapeutic in mdx mice. Gene Ther 15: 1099–1106.
Koerber, JT, Jang, JH and Schaffer, DV (2008). DNA shuffling of adeno-associated virus
yields functionally diverse viral progeny. Mol Ther 16: 1703–1709.
10. Li, W, Zhang, L, Johnson, JS, Zhijian, W, Grieger, JC, Ping-Jie, X et al. (2009).
Generation of novel AAV variants by directed evolution for improved CFTR delivery to
human ciliated airway epithelium. Mol Ther 17: 2067–2077.
11. Xie, Q, Bu, W, Bhatia, S, Hare, J, Somasundaram, T, Azzi, A et al. (2002).
The atomic structure of adeno-associated virus (AAV-2), a vector for human
gene therapy. Proc Natl Acad Sci USA 99: 10405–10410.
12. Mendell, JR, Campbell, K, Rodino-Klapac, L, Sahenk, Z, Shilling, C, Lewis, S et al.
(2010). Dystrophin immunity in Duchenne’s muscular dystrophy. N Engl J Med 363:
1429–1437.
13. Arruda, VR, Schuettrumpf, J, Herzog, RW, Nichols, TC, Robinson, N, Lotfi, Y et al.
(2004). Safety and efficacy of factor IX gene transfer to skeletal muscle in murine and
canine hemophilia B models by adeno-associated viral vector serotype 1. Blood 103:
85–92.
14. Chao, H, Monahan, PE, Liu, Y, Samulski, RJ and Walsh, CE (2001). Sustained and
complete phenotype correction of hemophilia B mice following intramuscular
injection of AAV1 serotype vectors. Mol Ther 4: 217–222.
15. Wu, Z, Asokan, A and Samulski, RJ (2006). Adeno-associated virus serotypes: vector
toolkit for human gene therapy. Mol Ther 14: 316–327.
16. Louboutin, JP, Wang, L and Wilson, JM (2005). Gene transfer into skeletal muscle
using novel AAV serotypes. J Gene Med 7: 442–451.
17. Xiao, W, Chirmule, N, Berta, SC, McCullough, B, Gao, G and Wilson, JM (1999).
Gene therapy vectors based on adeno-associated virus type 1. J Virol 73: 3994–4003.
18. Govindasamy, L, Padron, E, McKenna, R, Muzyczka, N, Kaludov, N, Chiorini, JA et
al. (2006). Structurally mapping the diverse phenotype of adeno-associated virus
serotype 4. J Virol 80: 11556–11570.
19. Padron, E, Bowman, V, Kaludov, N, Govindasamy, L, Levy, H, Nick, P et al. (2005).
Structure of adeno-associated virus type 4. J Virol 79: 5047–5058.
20. Lochrie, MA, Tatsuno, GP, Christie, B, McDonnell, JW, Zhou, S, Surosky, R et al.
(2006). Mutations on the external surfaces of adeno-associated virus type 2 capsids
that affect transduction and neutralization. J Virol 80: 821–834.
21. Wobus, CE, Hügle-Dörr, B, Girod, A, Petersen, G, Hallek, M and Kleinschmidt, JA
(2000). Monoclonal antibodies against the adeno-associated virus type 2 (AAV-2)
capsid: epitope mapping and identification of capsid domains involved in AAV-2-cell
interaction and neutralization of AAV-2 infection. J Virol 74: 9281–9293.
22. Gao, G, Vandenberghe, LH, Alvira, MR, Lu, Y, Calcedo, R, Zhou, X et al. (2004).
Clades of Adeno-associated viruses are widely disseminated in human tissues.
J Virol 78: 6381–6388.
23. Manno, CS, Pierce, GF, Arruda, VR, Glader, B, Ragni, M, Rasko, JJ et al. (2006).
Successful transduction of liver in hemophilia by AAV-Factor IX and limitations
imposed by the host immune response. Nat Med 12: 342–347.
24. Nam, HJ, Lane, MD, Padron, E, Gurda, B, McKenna, R, Kohlbrenner, E et al. (2007).
Structure of adeno-associated virus serotype 8, a gene therapy vector. J Virol 81:
12260–12271.
25. Hauck, B, Xu, RR, Xie, J, Wu, W, Ding, Q, Sipler, M et al. (2006). Efficient AAV1-AAV2
hybrid vector for gene therapy of hemophilia. Hum Gene Ther 17: 46–54.
26. Mendell, JR, Rodino-Klapac, LR, Rosales, XQ, Coley, BD, Galloway, G, Lewis, S et al.
(2010). Sustained α-sarcoglycan gene expression after gene transfer in limb-girdle
muscular dystrophy, type 2D. Ann Neurol 68: 629–638.
27. Mendell, JR, Rodino-Klapac, LR, Rosales-Quintero, X, Kota, J, Coley, BD, Galloway,
G et al. (2009). Limb-girdle muscular dystrophy type 2D gene therapy restores
α-sarcoglycan and associated proteins. Ann Neurol 66: 290–297.
28. Brantly, ML, Chulay, JD, Wang, L, Mueller, C, Humphries, M, Spencer, LT et al. (2009).
Sustained transgene expression despite T lymphocyte responses in a clinical trial of
rAAV1-AAT gene therapy. Proc Natl Acad Sci USA 106: 16363–16368.
29. Flotte, TR, Trapnell, BC, Humphries, M, Carey, B, Calcedo, R, Rouhani, F et al. (2011)
Phase 2 clinical trial of a recombinant adeno-associated viral vector expressing
alpha(1)-Antitrypsin: interim results. Hum Gene Ther (epub ahead of print).
30. Velazquez, VM, Bowen, DG and Walker, CM (2009). Silencing of T lymphocytes by
antigen-driven programmed death in recombinant adeno-associated virus vector-
mediated gene therapy. Blood 113: 538–545.
31. Vandenberghe, LH, Wang, L, Somanathan, S, Zhi, Y, Figueredo, J, Calcedo, R et al.
(2006). Heparin binding directs activation of T cells against adeno-associated virus
serotype 2 capsid. Nat Med 12: 967–971.
32. Calcedo, R, Vandenberghe, LH, Gao, G, Lin, J and Wilson, JM (2009). Worldwide
epidemiology of neutralizing antibodies to adeno-associated viruses. J Infect Dis
199: 381–390.
33. Li, W, Asokan, A, Wu, Z, Van Dyke, T, DiPrimio, N, Johnson, JS et al. (2008).
Engineering and selection of shuffled AAV genomes: a new strategy for producing
targeted biological nanoparticles. Mol Ther 16: 1252–1260.
34. Maheshri, N, Koerber, JT, Kaspar, BK and Schaffer, DV (2006). Directed evolution
of adeno-associated virus yields enhanced gene delivery vectors. Nat Biotechnol
24: 198–204.
35. Wang, L, Calcedo, R, Bell, P, Lin, J, Grant, RL, Siegel, DL et al. (2011). Impact of
Pre-Existing Immunity on Gene Transfer to Nonhuman Primate Liver with Adeno-
Associated Virus 8 Vectors. Hum Gene Ther (in press).
36. Toromanoff, A, Adjali, O, Larcher, T, Hill, M, Guigand, L, Chenuaud, P et al. (2010).
Lack of immunotoxicity after regional intravenous (RI) delivery of rAAV to nonhuman
primate skeletal muscle. Mol Ther 18: 151–160.
6.
7.
8.
9.

Page 13

Molecular Therapy
13
© The American Society of Gene & Cell Therapy
Clinical Evaluation of Custom-designed AAV Capsid in DMD Patients
37. Toromanoff, A, Chérel, Y, Guilbaud, M, Penaud-Budloo, M, Snyder, RO, Haskins, ME
et al. (2008). Safety and efficacy of regional intravenous (r.i.) versus intramuscular
(i.m.) delivery of rAAV1 and rAAV8 to nonhuman primate skeletal muscle. Mol Ther
16: 1291–1299.
38. Herzog, RW (2010). Hepatic AAV gene transfer and the immune system: friends or
foes? Mol Ther 18: 1063–1066.
39. Asokan, A, Conway, JC, Phillips, JL, Li, C, Hegge, J, Sinnott, R et al. (2010).
Reengineering a receptor footprint of adeno-associated virus enables selective and
systemic gene transfer to muscle. Nat Biotechnol 28: 79–82.
40. Pulicherla, N, Shen, S, Yadav, S, Debbink, K, Govindasamy, L, Agbandje-McKenna,
M et al. (2011). Engineering Liver-detargeted AAV9 Vectors for Cardiac and
Musculoskeletal Gene Transfer. Mol Ther 19: 1070–1078.
41. Rabinowitz, JE, Rolling, F, Li, C, Conrath, H, Xiao, W, Xiao, X et al. (2002). Cross-
packaging of a single adeno-associated virus (AAV) type 2 vector genome into multiple
AAV serotypes enables transduction with broad specificity. J Virol 76: 791–801.
42. Grieger, JC, Choi, VW and Samulski, RJ (2006). Production and characterization of
adeno-associated viral vectors. Nat Protoc 1: 1412–1428.
43. Rabinowitz, JE, Bowles, DE, Faust, SM, Ledford, JG, Cunningham, SE and Samulski,
RJ (2004). Cross-dressing the virion: the transcapsidation of adeno-associated virus
serotypes functionally defines subgroups. J Virol 78: 4421–4432.
44. Moskalenko, M, Chen, L, van Roey, M, Donahue, BA, Snyder, RO, McArthur, JG et al.
(2000). Epitope mapping of human anti-adeno-associated virus type 2 neutralizing
antibodies: implications for gene therapy and virus structure. J Virol 74: 1761–1766.
45. Arnold, K, Bordoli, L, Kopp, J and Schwede, T (2006). The SWISS-MODEL workspace:
a web-based environment for protein structure homology modelling. Bioinformatics
22: 195–201.
46. Krissinel, E and Henrick, K (2004). Secondary-structure matching (SSM), a new
tool for fast protein structure alignment in three dimensions. Acta Crystallogr D Biol
Crystallogr 60(Pt 12 Pt 1): 2256–2268.
47. Emsley, P and Cowtan, K (2004). Coot: model-building tools for molecular graphics.
Acta Crystallogr D Biol Crystallogr 60(Pt 12 Pt 1): 2126–2132.
48. Jones, TA, Zou, JY, Cowan, SW and Kjeldgaard, M (1991). Improved methods for
building protein models in electron density maps and the location of errors in these
models. Acta Crystallogr, A, Found Crystallogr 47 (Pt 2): 110–119.