Therapeutic expression of human clotting factors IX and × following adeno‐associated viral vector‐mediated intrauterine gene transfer in early‐gestation fetal macaques

Adeno‐associated viral vectors (AAVs) achieve stable therapeutic expression without long‐term toxicity in adults with hemophilia. To avert irreversible complications in congenital disorders producing early pathogenesis, safety and efficacy of AAV‐intrauterine gene transfer (IUGT) requires assessment We therefore performed IUGT of AAV5 or ‐8 with liver‐specific promoter‐1 encoding either human coagulation factors IX (hPIX) or × (hFX) into Macaca fascicularis fetuses at ∼0.4 gestation. The initial cohort received 1 × 1012 vector genomes (vgs) of AAV5‐hFIX (n = 5; 0.45 × 1013 vg/kg birth weight), resulting in ∼3.0% hPIX at birth and 0.6–6.8% over 19–51 mo. The next cohort received 0.2‐1 × 1013 vg boluses. AAV5‐hPX animals (n = 3; 3.57 × 1013 vg/kg) expressed <1% at birth and 9.4–27.9% up to 42 mo. AAV8‐hFIX recipients (n = 3; 2.56 × 1013 vg/kg) established 4.2–41.3% expression perinatally and 9.8–25.3% over 46 mo. Expression with AAV8‐hFX (n = 6,3.12 × 1013 vg/kg) increased from <1% perinatally to 9.8–13.4% >35 mo. Low expressers (<1%, n = 3) were postnatally challenged with 2 × 1011 vg/kg AAV5 resulting in 2.4–13.2% expression and demonstrating acquired tolerance. Linear amplification‐mediated‐PCR analysis demonstrated random integration of 57–88% of AAV sequences retrieved from hepatocytes with no events occurring in or near oncogenesis‐associated genes. Thus, early‐IUGT in macaques produces sustained curative expression related significantly to integrated AAV in the absence of clinical toxicity, supporting its therapeutic potential for early‐onset monogenic disorders.—Chan, J. K. Y., Gil‐Farina I., Johana, N., Rosales, G., Tan, Y. W., Ceiler, J., Mcintosh, J., Ogden, B., Waddington, S. N., Schmidt, M., Biswas, A., Choolani, M., Nathwani, A. G., Mattar, C. N. Z. Therapeutic expression of human clotting factors IX and × following adeno‐associated viral vector‐mediated intrauterine gene transfer in early‐gestation fetal macaques. FASEB J. 33, 3954–3967 (2019). www.fasebj.org

in adults limits phenotype correction despite efficient transduction (3,4). Early AAV-GT preceding permanent tissue damage is thus an attractive prospect for children with monogenic conditions (3)(4)(5)(6). Infants with spinal muscular atrophy treated with self-complementary AAV (scAAV) 9.CB.hSMN demonstrated significant dose-dependent neurologic improvement and higher unventilated survival when compared with controls (7). The most beneficial application of GT may be during early fetal development, particularly for syndromes that manifest in utero and in which postnatal therapeutic potential is limited by irreversible tissue damage or inefficient protein replacement (8)(9)(10), examples of which include neuronopathic Gaucher disease and other neurometabolic syndromes (11)(12)(13) and congenital factor X (hFX) deficiency, which causes perinatally lethal intracranial hemorrhage (14)(15)(16). In such conditions, better outcomes are demonstrated when genetic correction is attempted in utero rather than in later childhood, and improving accuracy of early prenatal testing is likely to drive the demand for fetal therapies (6,(17)(18)(19)(20)(21)(22).
Clinical outcomes depend on fetal maturity status at intervention (23). Increased fetal stem cell receptivity to vector transduction (24)(25)(26), minimal pre-existing tissue damage (27,28), and immune naïveté facilitating vector tolerance (20,(29)(30)(31) increase the likelihood of long-term phenotype correction. It is, however, critical to interrogate organ-and genotoxicity in addition to protein deficiency correction in GT applications-issues that have arisen in the successful gene transfer of young children with severe congenital immunodeficiency and spinal muscular atrophy (7,32). Most intrauterine gene transfer (IUGT) animal models do not demonstrate the degree of transgenic protein production the we have described in nonhuman primates (NHPs), here and previously (29,33). Murine IUGT with AAV1-human factor IV (hFIX) or AAV2-hFIX at 10 12 -10 13 vector genomes (vgs)/kg showed persistent low transgene levels, requiring AAV readministration for therapeutic expression (31). Ovine AAV-IUGT produced only low expression up to 6 mo after peak expression at ;3 wk after injection (34). NHP-IUGT with scAAV5-liverspecific promoter (LP)-1-hFVII maintained ;20% expression at birth, successfully boosted, after loss of expression, by the alternate AAV8 serotype, which thereafter supported therapeutic FVII levels without a sustained immune response (35). Despite the expanding applicability of recombinant AAV pseudotypes engineered for improved safety via specific organ targeting and reduced insertional oncogenesis, a major limitation is its predominantly episomal nature (36)(37)(38)(39)(40)(41)(42). These animal studies demonstrate the progressive loss of expression expected after fetal or neonatal treatment as the recipient grows (43). Introduction of GT vectors during fetal development to induce immune tolerance and facilitate repeat AAV administration after birth (33,(44)(45)(46) can be a useful strategy for stabilizing waning protein expression in the hemophilias and neurometabolic disorders (11,47). Despite the encouraging outcomes of clinical trials, high-dose AAV has resulted in complications associated with integration, direct cellular toxicity, and systemic inflammation, resulting in hepatocellular carcinoma in rodents, neurodegeneration in piglets and liver failure in juvenile NHPs (48,49). Data regarding AAV-IUGT safety and efficacy can thus originate only from a high-fidelity NHP model in which longitudinal surveillance is clinically relevant (50). We have described long-term outcomes of liver-directed late-gestation IUGT in NHP by using the clinical vector scAAV-LP1-human codonoptimized human FIX transgene (FIXco) (5) in which a single injection at 0.9 gestation demonstrated therapeutic hFIX expression in most animals (29,33). Low-expressing animals were safely boosted with only transient immune responses (33). Using hFX as a marker transgene in the same manner as hFIX, we interrogated the expression, immune response, and vector biodistribution after single-dose early-IUGT at ;0.35-0.4 gestation in an established NHP model in which we used AAV5 and -8 vectors expressing either hFIX or hFX codon-optimized transgenes (50). This study is the first that we know of to describe in extensive detail the clinically relevant outcomes of early IUGT in a large number of NHPs. We report unexpectedly prolonged and stable postnatal transgenic protein production with only transient transaminitis, a high level of vector integration with no hotspots, and evidence of immune tolerance.

MATERIALS AND METHODS
Vector production scAAV-LP1-hFIXco was produced as described by Nathwani et al. (44). scAAV-LP1-hFXco was produced similarly, replacing the hFIXco with the codon-optimized hFX transgene (hFXco), from pAV-LP1-hFX with a truncated poly-A. AAV-inverted terminal repeats with the LP1 promoter were cut with HindIII/ MfeI, ligated with T4 DNA ligase (New England Biolabs, Ipswich, MA, USA), and heat shocked into competent Escherichia coli, which were grown at 37°C overnight (51). Correct clones were identified after DNA extraction by digestion with HindIII/PstI, AhdI, and BglI. Plasmid expression of hFX expression was assessed by ELISA after transient transfection of HepG2 cells with Fugene (Roche, Basel, Switzerland), and clones were amplified with Megaprep (Qiagen, Limberg, The Netherlands). scAAV-LP1-hFIXco and scAAV-LP1-hFXco plasmids were used in adenovirus-free transient transfection of 293T cells to generate scAAV5 and scAAV8 pseudotypes, as described by Nakai et al. (45). Titers were determined by quantitative PCR with linearized plasmids used as standards.
IUGT, fetal blood sampling, delivery and monitoring All procedures were performed in Macaca fascicularis strictly adhering to recommendations from the Institutional Animal Care and Use Committee at the National University of Singapore and Singapore Health Services Pte, Ltd. (IACUC 2012/SHS/692). The in vivo work was conducted at the SingHealth Experimental Medicine Centre (SEMC, Singapore), accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. Timed mating and pregnancy monitoring are described in refs. 29 and 50. Vector was administered to macaques under ketamine sedation via intraperitoneal, intracardiac, or intravascular injection with an 18-gauge Quincke spinal needle (BD Biosciences, Franklin Lakes, NJ, USA) with ultrasound (US) guidance ( Table 1). Injected volumes were restricted to 500 ml, and fetal heart rate was monitored for 5 min after injection. Routine maternal and fetal surveillance, fetal blood sampling (FBS) at ;0.6 G, caesarean delivery, and infant hand-rearing were performed (50). From 1 yr of age, hepatic and other tissue biopsies were performed at 6-mo intervals to assess vector biodistribution and integration (29).

Postnatal vector challenge
Three infants with postnatal transgenic protein production of ,1% were prepared for and serially monitored after vector challenge (VC), as previously described (33). Each received 2 3 10 11 vg/kg body weight of AAV-hFIX or AAV-hFX and were monitored for immediate adverse reactions (Fig. 1).
Treated subjects underwent open or percutaneous US-guided liver biopsies (Tru-Cut Needle; CareFusion, Basingstoke, United Kingdom) 3 mo after VC to assess vector load, hepatotoxicity, and vector integration. Wedge biopsies were collected via upper abdominal midline incision with hemostasis (50). All procedures were performed under ketamine sedation (IUGT, FBS, and VC) or general anesthesia (caesarean delivery, liver biopsy).

Transgenic protein production
hFIX was analyzed by quantitative ELISA, as previously described (29). hFX was analyzed by using the same protocol with FX-EIA-C capture antibodies and FX-EIA-D detecting antibodies (Matched-Pair Antibody Set for ELISA of human factor X antigen, Affinity Biologicals, Ancaster, ON, Canada). Serum samples from injected offspring were diluted at 1:50 and 1:100 in HEPES-buffered saline-bovine serum albumin (BSA)-Tween 20 sample diluent for analysis. Human FX protein (ab62549; Abcam, Cambridge, United Kingdom) reconstituted in naive NHP serum and diluted serially from 0.075 to 10 mg/ml (0.75-100%) in HEPES-buffered salinebovine serum albumin (BSA)-Tween 20 served as reference standards. Colorimetric change was developed and analyzed.

Liver-specific expression
Liver-specific expression of hFIX and hFX was determined by RT-PCR, as previously described (29). The resultant cDNA (2 ml) was used to amplify a 617-bp region of the hFIX transgene with the forward 59-TTTCCTGATGTGGACTATGT-39 and reverse 59-TCATGGAAGCCAGCACAGAACATG-39 primers. The hFX transgene was amplified with the human-specific forward 59-GAGAGGGGCGACAACAACCTCAC-39 and reverse 59-AG-GCATCCTCCTGCTTGGTGTCG-39 primers, producing a 568-bp amplicon. cDNA integrity was determined by amplifying a 604-bp region of the macaque b-actin gene by using forward 59-TGACGGGGTCACCCACACTGTGCCCATCTA-39 and reverse 59-CTAGAAGCATTTGCGGTGGACGATGGAGGG-39 primers. Cycling conditions were: 95°C for 5 min, 40 cycles of 95°C for 30 s, 57°C for 30 s (hFIX), or 60°C for 30 s (hFX, b-actin) and 72°C for 60 s, with final extension of 72°C for 7 min, with the product resolved on 2% agarose gel.

Viremia and vector biodistribution
Assessment of viremia and vector biodistribution were performed as previously described for both hFIX and hFX-treated animals (29). In brief, 5 ml of cell-free maternal plasma or 15 mg genomic DNA extracted from tissue was amplified for an 84-bp region of the transgene encompassing the 39 end of the codon-optimized transgene and the 59 region of the SV40 late poly A sequence with 100 nM each of forward 59-GGGCAAGTATGGCATCTACA-39 and reverse 59-AAAGCATCGAGTCAGGTCAG-39 primers. Viral load was expressed as copies per microliter maternal plasma (viremia) or vector copy number (VCN) per diploid genome (6.6 pg of DNA); the calculated limit of detection was 1 vg per 455 diploid genomes. Equivalent loading was assessed by amplifying a 52-bp region of the macaque b-actin gene with forward 59-TCCTGTGGCATCCACGAAA-39 and reverse 59-CCA-CGTCACACTTCATGATGG-39 primers.

Anti-AAV binding and neutralizing antibodies
Binding anti-capsid antibodies were assessed by sandwich ELISA, as previously described (29), using 5 3 10 8 AAV8 or . Results were standardized to the negative baseline (readout of samples from unexposed macaques) and expressed as a ratio of the positive control (readout from a sensitized adult macaque strongly positive for anti-AAV8 or -5) (44). Neutralizing antibodies (NAbs) were assessed in vitro by incubating samples diluted 1:99 with 5 3 10 10 AAV5 or -8 particles and then with 1 3 10 5 293T cells, as described in refs. 29 and 52. At least a 50% reduction in GFP transduction, relative to the control, was indicative of NAbs. Control subjects were immune-competent, unexposed, adult male macaques capable of producing anti-AAV antibodies as a primary response, which did not produce NAb during the first month after exposure.

Cell-mediated immunity
Cell-mediated reactions to postnatal VC were analyzed by intracellular staining (ICS) adapted from a published protocol (53), as previously described (33). All antibodies and reagents were acquired from BD Biosciences unless stated otherwise. Briefly, frozen peripheral blood mononuclear cells (MNCs) were thawed, Postacquisition analyses were performed with Summit 4.2 (Beckman Coulter). Single-color controls were added via Compbeads Anti-Mouse Ig, k (BD Biosciences). Cells were gated onto CD3 and then onto CD8 or CD4 cells, which in turn were gated onto CD95 and CD28, to differentiate effector and memory cells, as previously described (33). Effector and memory cells were gated onto cytokines IL-2, IFN-g, and TNFa. Naive macaque MNCs incubated in phorbol 12-myristate 13-acetate and ionomycin (5 mg/ml) served as the positive control; the negative control consisted of MNCs incubated in peptide-free medium. Samples were read as positive if at least 0.05% of a subpopulation stained for a particular cytokine.

Integration sites analyses by linear amplification-mediated-PCR and high-throughput sequencing
Linear-amplification-mediated (LAM) PCR was performed as described in refs. 29, 33, and 54. In brief, AAV flanking sequences from both AAV-hFIX and AAV-hFX vectors were amplified from 750 ng genomic DNA from liver biopsies by using the same primers (Supplemental Table S1). Restriction digest steps were performed with the MseI and MluCI restriction enzymes (New England Biolabs). After 1 linear and 2 exponential PCRs, an additional PCR amplification with barcoded fusion primers was performed to prepare fragments for 250PE MiSeq sequencing (Illumina, San Diego, CA, USA). Raw sequences were analyzed by automated bioinformatics tools for quality filter, vector trimming, and identification of vgs [integration sites (ISs)] and vector-vector (concatemers) junctions. ISs were mapped to the macaque genome by using the University of California at Santa Cruz Blast-like alignment tool and were analyzed by automated data-mining tools to characterize the vector's integration profile (55).

Statistics
Results were analyzed using statistical software Prism v.6.04 (GraphPad Software, La Jolla, CA, USA). Data are shown as means 6 SD, and analyzed using unpaired Student's t tests. A value of P , 0.05 was considered significant. Frequency of integration sequence retrieval from tissue samples of subjects was compared by Fisher's exact test.

Immune responses to vector administration in IUGT recipients
Infants receiving either serotype demonstrated similar humoral responses to capsid and transgenic proteins. Persistent anti-capsid IgG antibodies were observed from 0.6 gestation throughout the surveillance period and remained below the positive threshold (Fig. 3A). Responses to both transgenes were subdued (Fig. 3B, C). Maternal anti-AAV5 antibody expression was more robust than anti-AAV8 (Fig.  3D) and anti-transgene responses were mild (Fig. 3E) (29).
Transgenic protein production and immune response after postnatal VC e5002, e5003, and e5002FX were negative for binding and neutralizing anti-AAV antibodies and had normal transaminases, hematologic indices, and liver histology before VC (Fig. 4).  Fig. 5A. Anti-AAV IgG antibodies remained below the positive threshold in e5002 and e5003 and were largely undetectable in e5002FX (Fig. 5B). No IgG production against hFIX or hFX was observed. Only e5002FX produced NAbs from 7 to 21 d after VC, before a resolution that preceded the increase in hFX expression (Fig. 5C, VC at d 0). NAbs were not demonstrated in e5002 and e5003. e5002 showed a .300-fold increase in hepatic VCN (from 0.01 to 2.00 copies/cell), whereas the VCN in e5003 remained at 0.10 copies/cell. Hepatic VCN in e5002FX decreased from 0.4 to 0.006 copies/cell (Fig. 5D). Activated CD4 and CD8 T-cells expressing TNF-a and IFN-g were observed against AAV5 capsid proteins (Fig. 5E). e5002FX showed increasing activated T cells from first to second postnatal VC, diminishing over the following 3 wk. Both e5002 and e5003 showed higher activated CD4 and CD8 T cells, decreasing over 4 wk. Both e5002 and e5003 perished from anesthesia-related complications following liver biopsies at 30 d after VC (Fig. 1).
(;120 d), and from 12 to 42 mo in surviving offspring. Mean VCN ranged from 0.002 6 0.001 copies/cell (cerebrum) to 6.7 6 11.1 copies/cell (skin); hepatic VCN was 0.3 6 0.1 copies/cell, and there was no demonstrable liver tropism (Fig. 6A). We observed time-dependent log-fold decreases in VCN during each 12-mo period (1-36 m; P , 0.05). VCN in skin and muscle (highest starting vector load) decreased ;2-3-log-fold by 42 mo (Fig. 6B). The decrease in VCN is shown in the context of sustained hFIX and hFX levels, as expression per hepatic VCN increased with time (Supplemental Fig. S1A). Liver-specific alanine transaminase (ALT) levels were generally normal, although it was increased at one time point in the AAV5 group (20 mo, 4-6 mo after VC; Fig. 6C). Aspartate transaminase levels were above the upper limit for the first 20 mo before decreasing; a trend toward higher levels was observed with AAV5 (Supplemental Fig. S1B). Despite this biochemical transaminitis, leukocyte counts and other hematologic indices registered only mild fluctuations within the upper normal limit, and there was no clinical evidence of hepatotoxicity (e.g., anorexia, weight loss). Liver histology 3 mo before and after postnatal challenge did not demonstrate inflammatory cell infiltrates and showed normal hepatocellular architecture (Supplemental Fig. S1C, D).
Our previous data in late-IUGT NHP demonstrated that integrated vector genome (;10% of sequences at 2-10 mo after birth), rather than concatemers, accounted for the majority of vector persistence at later time points (87.4-100% at 18-49 mo after birth) (33). In the present study, LAM-PCR analyses revealed that 43.9-73.7% of vector sequences were integrated at 10 mo, increasing to 87.5% at 42 mo (Fig. 7A). Amplicons obtained by LAM-PCR yielded 4,217,447 raw reads, enabling identification of 121 ISs corresponding to 111 unique and 14 multiple mappable ISs ( Table 2 and Fig. 7A). In e5002, the integrated vector accounted for 57.1% of retrieved sequences 3 mo after VC (19 mo of age). Integration frequencies were   Fig. 7B, C). We performed semiquantitative estimation of the sample clonal size by analyzing the 10 most frequently retrieved clones from the different samples (Fig. 7D-F). Only two of these corresponded to integration events in or near cancer-related genes RAD23B and RAB1A (refer to the retroviral tagged cancer gene databases: the Cancer Genomics Consortium, https://www.cancergenomics.org/, and the cBio Portal for Cancer Genomics, http://www.cbioportal.org/), and none was retrieved at consecutive time points in the same animal or from different animals. Considering all integration events identified, 7.2% of integration events were located within or near cancer-related genes, which did not significantly differ from the 8.9% found in a synthetic random dataset. No significant targeting of chromosomes or gene coding regions were identified (Supplemental Fig. S1E, F), and there were no integration events retrieved from the genomic regions previously reported to be involved in AAV-associated oncogenesis (48).

DISCUSSION
Recent AAV-GT hemophilia trials have demonstrated marked improvements in hemostatic stability and significantly reduced factor use in adults (3,4,57). Our data affirm the ability of early AAV-IUGT to generate curative expression of hFX and hFIX, stable for .4 yr, despite a several-fold weight increase (signifying hepatic growth), primarily represented by highly divergent and genomewide vector integration within genes and introns in equal measure without demonstrable hotspots. No conclusions can be drawn as to the impact of administration route, as the numbers are too small for comparison. We anticipate that the intravenous route is the most efficient (58), but it results in longer procedures, potentially greater fetal stress or injury, and more prolonged exposure to anesthesia than intraperitoneal injections (unpublished results). Administration and surveillance strategies used for early and late IUGT were intentionally similar, as all procedures are clinically applicable and resulted in therapeutic expression levels for hemophilia B. Although the hemophilias are not conditions that require in utero rescue from the perspective of disease onset, it is encouraging that liver-directed AAV-driven GT can be performed at 0.4 gestation with a reasonable expectation of efficacy, safety, and technical feasibility. Substituting the relevant transgene for an inborn error of metabolism or other similar condition into AAV5 or -8 can be expected to effectively rescue a disease causing early fetal damage, the most cogent argument for early IUGT.
Our data suggest that vector integration has a similarly substantial and durable influence on stable long-term expression after both early and late IUGT, considering the low percentage of remaining episomes in both 0.4 and 0.9 gestation animals by 42-47 mo (33). It is not possible to conclude at the present time whether this finding represents a translational advantage or risk in the longer term, although the biosafety data would suggest low overall integration-related risk. Our previous work showed that AAV8 is the more efficient pseudotype for liver-directed GT in 0.9 gestation fetal and adult macaques (29,33,59). However, with IUGT at 0.4 gestation, AAV8 and -5 demonstrate equivalence in their efficacies as seen by their comparable hFX expression; lower expression among AAV5-hFIX recipients was largely related to the log-fold lower dose used, given that the number of males was similar in both groups. This finding demonstrates the utility of both AAV8 and -5 for liver-directed early-IUGT strategies for a range of congenital monogenic conditions with optimized transgenes. From comparative studies of AAV8-GT in infant and juvenile rodents and NHPs, clear correlation is observed between age at treatment and hepatocyte transducibility: earlier AAV8-GT produces higher, but ultimately more unstable, transduction caused by rapid liver growth and episomal loss within 1-3 mo of vector delivery, with expression typically reaching a plateau (60,61). Hepatocyte transduction is also dose-and serotypedependent, as demonstrated by greater expression with AAV8 (90-100% transduction) than with AAV2 (5-10% transduction) (62-64) and the linear dose-response relationship observed in adult macaques given 10 10 -10 12 vg/kg of AAV8-hFIX (59).
Given these observations we had anticipated greater transgenic protein production with early-IUGT than was eventually achieved. The initial supraphysiological expression observed at 0.9 gestation was not seen at 0.4 gestation, which probably reflects superior episomal loss from rapid intrauterine growth of early-IUGT recipients, although median birth weights were comparable in both Figure 5. Postnatal VC and response. A) Three animals, e5002, e5003 (both AAV5-hFIX), and e5002FX (AAV5-hFX) received 2 3 10 11 vg/kg AAV5. Transgene levels were ,1% before VC (d 0). Brief improvements in expression were observed in e5003 after VC and were sustained over 23 d before being nullified. e5002 showed short spikes .1% at 8 and 24 d after VC, whereas e5002FX maintained ,1% hFX until an increase was seen from d 27 after VC. B) Anti-AAV IgG antibodies increased mildly only in e5002 and e5003; no such reaction was observed in e5002FX, where anti-AAV5 IgG was undetectable. No IgG production against or was observed. C) The presence of NAbs was determined by loss of green fluorescent protein transduction to #50% of baseline transduction (normalized to 100%). e5002FX demonstrated NAbs from d 7 to 21 after VC before resolution. D) Pre-and postchallenge hepatic VCN showed various responses e5002 showed an increase from 0.01 to 2.00 copies/cell, whereas VCN in e5003 remained at 0.10 vector/cell. No further follow-up was possible, given that both recipients died at the postchallenge biopsy. A substantial decline from 0.35 to 0.01 copies/cell was observed in e5002FX. E) e5002FX showed an increase in activated CD4 and CD8 T cells producing TNF-a and IFN-g from first to second postnatal VC, but this response quickly diminished by 3 wk after VC. Both e5002 and e5003 showed higher activated CD4 and CD8 T cells after VC, which similarly decreased over the following 4 wk. early and late-IUGT cohorts [275 vs. 267.5 g (range 205-290 g)] (33). In NHP pregnancies harvested at 0.5 gestation (median, 77.5 GDs; range; 71-83 d; n = 10), the median fetal weight was 46.0 g (36.2-72.0 g) and livers weighed ;1.9 g (1.6-4.1 g; unpublished results). Using this fetal weight as reference, the concentration of vector achieved with 1 3 10 13 vg was 2.2 3 10 14 vg/kg, 1-log-fold higher than the 1.6 3 10 13 vg/kg dose given to the late-IUGT recipients (29,33). At 46 g at IUGT, the injected fetus is assumed to contain ;5 3 10 10 cells (65); at the median birth weight of 275 g the total number of cells per infant is expected to reach ;3 3 10 11 cells, over a mean of 60 d. Late-IUGT animals grew 2-3-fold over the first 60 d of life (injections were performed 1 wk before birth). With 1-log-fold higher vector load at early-IUGT to 0.6-log-fold increase in cell numbers, it is anticipated that, given the same efficacy of transduction, early-IUGT recipients would have higher expression from greater cellular transduction. We instead have observed lower expression, and posit that gestation-sensitive differences in the metabolic state of hepatocytes may be important contributory factors (63). Possible saturation kinetics were seen with AAV5-hFIX and AAV5-hFX, as there was no discernible improvement in transgenic protein production with log-fold increases of vector. The reason remains unknown, but we postulate that the early fetal liver is more resistant to AAV5 transduction, perhaps because of cell surface receptors or other dynamics of vector-cell interactions. In addition, it is known that hFIX increases steadily in childhood, whereas hFX levels are significantly lower at birth, only reaching adult levels by 5 yr of age (66). Initially low transgenic protein levels reflect hepatic functional immaturity at the time of AAV-GT and eventually rise to clinically normal levels with age. The differences in individual transgenic protein production from AAV8 and -5 are not clinically relevant here, given that steady-state expression of both transgenes reached therapeutic levels.
Immune responses to the primary AAV dose remained low, and the distinct immunogenicity of AAV5 observed among immune-mature recipients (late-gestation fetuses and adults) was absent in these early recipients (33,59).
Although T-cell assays were not performed after early-IUGT, sustained expression in most recipients supports the absence of a primary cellular immune response. A transient increase in ALT in AAV5 animals was not a direct result of the postnatal challenge (occurring 4-6 mo after VC), and this biochemical observation did not coincide with loss of transgenic protein production in the surviving AAV5-treated animals. Maternal IgG production was observed during the first 12 mo with AAV5 demonstrating greater immunogenicity than AAV8, consistent with our prior observations in late-gestation IUGT (29,33). AAV5injected dams had 1.5-1.8-fold higher anti-capsid and anti-hFIX antibody levels compared to offspring (similar levels with AAV8). It is possible that antibodies detected in offspring are transplacentally trafficked maternal antibodies, which may have accounted for some of the humoral response in treated infants. It is also possible that prolonged low-level anti-AAV IgG in early recipients indicates a mild primary response in most animals. Transplacental maternal immune influence on fetal expression has been described in murine experiments (67,68). The degree to which transgenic protein production is influenced by maternal immunity is unknown; this may be a contributing factor toward low expression and warrants further investigation.
Post-VC responses in 0.4 gestation animals (compared to 0.9 gestation) demonstrated similar transgenic protein production, lower anti-AAV IgG production (remaining well below the positive limit), transient NAb expression in 1 of 3 challenged animals vs. NAb following all 3 challenges after late-IUGT, and a higher percentage of activated T cells expressing intracellular cytokines (33). Early Figure 6. Vector biodistribution and hepatotoxicity. A) VCNs showed wide variation with most subjects expressing 1 vector copy every 10-1000 cells in the absence of specific organ tropism. B) Regardless of initial VCN, all tissues showed a substantial decline in tissue vector load of 1-3-log-fold every 12 mo; there was no difference between AAV5-and AAV8-treated animals. C ) Liver-specific ALT levels increased intermittently, particularly in AAV8 recipients, but overall transaminases remained in the normal range.
NAb and activated T-cell production may have prevented transduction in e5002FX via immune-mediated clearance of transduced cells or prevention of vector entry. Recovery of expression in e5002 and e5002FX could be attributed to the eventual loss of T cells and NAb. Given this immune reaction, even the brief appearance of NAb can influence the outcomes of postnatal boosting. Although it is generally believed that early-gestation fetuses are immune naive, our data suggest that they are capable of immune responses and retain partial ability to eliminate transduced hepatocytes. Despite this, sustained transgenic protein production during this period suggests a certain degree of peripheral tolerance that facilitates postnatal boosting. A higher VC dose may overcome this barrier and remains to be tested.
Our data revealed no adverse events derived from direct AAV toxicity or insertional mutagenesis during the 4-yr surveillance in contrast to recent studies that used logfold higher doses (49,69). The sustained high transgenic protein production with rapidly decreasing VCN suggests that long-term transgene production derives from integrated vector more than from concatemers, although we did not demonstrate this with certainty; we also did not observe evidence of liver cancer.
The number of animals with long-term follow-up was small, ultimately limiting the power of important conclusions regarding durability of expression and immune response. Pregnancy losses were expected, given the high incidence among captive-bred NHPs (29,50). There was no increase in postnatal mortality compared with our late-IUGT cohort apart from procedure-related deaths and nosocomial infections. The ratio of females to males was higher in the AAV5 group, but lower in the AAV8 group, and this would have influenced transgenic protein production readouts to some extent, given the influence of gender, as we have previously described (33). We did not determine the dose-response to AAV5-FIX Figure 7. Vector ISs. A) Integration frequency increased from 40 to 70% at 10 mo to .80% at 42 mo in the different animals sampled. B) No differences in integration frequencies were noted between AAV5 and -8. C-E ) Relative sequence counts of the 10 most prominent IS retrieved were calculated in relation to all uniquely mappable IS sequences in e8001 at 10 mo (C ), e8003 at 10 mo (D, E ), and e5002 at 19 mo (after VC). The RefSeq identities of the gene located adjacent to or at the IS are listed. because of experimental constraints imposed by the realtime availability of pregnant animals. The fetuses of AAV5-seronegative mothers were the first to be treated when there was a primary need to establish procedural safety. The initial lower dose of 1 3 10 12 vg/fetus was based on previous studies that determined the effective dose of this vector in mice and adult macaques sufficient to achieve therapeutic expression (44,70). The administered dose was further constrained by the volume limitation imposed by fetal size and was deliberately kept low to avoid circulatory embarrassment. Fetuses were subsequently treated with a higher dose of AAV8-hFIX when the suboptimal therapeutic achievements of AAV5-hFIX recipients became evident. We performed ICS for cytokines in activated T cells rather than enzymelinked immunospot (ELISPOT). We chose this method because of the advantage of simultaneous assessment of multiple surface markers and effector cytokines reflecting the phenotype and function of reactive T cells. Although there are advantages to ELISPOT, we preferred the range and quantitative readouts of flow cytometry (71). Although the method may be nonstandard compared to ELISPOT, there is a growing number of studies applying ICS as the method of choice because of the greater flexibility and breadth of this assay. In NAb assays we did not perform serial dilutions, and although these semiquantitative assays suggest that high titers of NAb were not present, we cannot exclude the presence of low NAb titers in these samples. In summary, early-IUGT can make a critical contribution in the therapeutic arsenal for congenital diseases that produce pathology in early development that may otherwise result in fetal loss or severe permanent handicap. Within the parameters of this research, we have demonstrated its usefulness in achieving curative levels of transgenic protein. Compared with late-IUGT, potential benefits extend from correction of early fetal pathogenesis to a higher degree of immune tolerance to postnatal repeat treatment to boost suboptimal outcomes. It can be interpreted from this study that serotype, dose, production of NAb and a stimulated T-cell response are factors critical to a long-term approach. Further immune modulation may still be necessary, as practiced in recent AAV clinical trials (3,4,72). No adverse effects of AAV toxicity or integration were observed during 4 yr of surveillance. Long-term therapeutic expression relies significantly on an integrated vector. Individual therapeutic strategies should be optimized further in the NHP model to assess translational potential.