Antigen-selective modulation of AAV immunogenicity with tolerogenic rapamycin nanoparticles enables successful vector re-administration

Gene therapy mediated by recombinant adeno-associated virus (AAV) vectors is a promising treatment for systemic monogenic diseases. However, vector immunogenicity represents a major limitation to gene transfer with AAV vectors, particularly for vector re-administration. Here, we demonstrate that synthetic vaccine particles encapsulating rapamycin (SVP[Rapa]), co-administered with AAV vectors, prevents the induction of anti-capsid humoral and cell-mediated responses. This allows successful vector re-administration in mice and nonhuman primates. SVP[Rapa] dosed with AAV vectors reduces B and T cell activation in an antigen-selective manner, inhibits CD8+ T cell infiltration in the liver, and efficiently blocks memory T cell responses. SVP[Rapa] immunomodulatory effects can be transferred from treated to naive mice by adoptive transfer of splenocytes, and is inhibited by depletion of CD25+ T cells, suggesting a role for regulatory T cells. Co-administration of SVP[Rapa] with AAV vector represents a powerful strategy to modulate vector immunogenicity and enable effective vector re-administration.

G ene therapy mediated by recombinant adeno-associated virus (AAV) vectors is one of the most promising approaches for the treatment of a variety of inherited and acquired diseases 1 . Human clinical gene therapy trials with AAV have demonstrated durable expression at therapeutic levels when targeting tissues like the liver 2-7 , motor neurons 8 , and the retina 9 . Despite the exciting results to date, one of the limitations of the AAV vector gene transfer platform is the durability of the effect. For many metabolic and degenerative diseases, treatment is critically needed early in life 10,11 , prior to the onset of irreversible tissue damage. However, because of their non-integrative nature, systemic gene therapy with AAV vectors in pediatric patients is expected to be limited by tissue proliferation associated with organ growth, which results in significant vector dilution over time [12][13][14] . Thus, maintaining the possibility to re-administer AAV is an important goal to achieve sustained therapeutic efficacy over time in pediatric patients. In addition, vector readministration in both pediatric and adult patients would be desirable to enable vector titration, to increase the proportion of patients that achieve therapeutic levels of the transgene expression, while avoiding supra-physiological transgene expression 7 and potential toxicities associated with large vector doses 15 .
However, vector immunogenicity represents a major limitation to re-administration of AAV vectors 16 . Persistent high-titer neutralizing antibodies (NAbs) are triggered following vector administration 5 , which abolishes any benefit of repeated AAVbased treatments. In addition, experience in human trials has shown that induction of capsid-specific CD8 + T cell responses can lead to clearance of AAV vector-transduced cells 3,5,6,17 . Thus, safe and effective strategies aimed at reducing AAV vector immunogenicity that allow for stable transgene expression and vector re-dosing are urgently needed.
Here, we demonstrate that co-administration of SVP[Rapa] with AAV vectors induce safe and effective control of capsid immunogenicity in an antigen-selective manner. Importantly, this approach allows for productive repeated dosing of the same AAV serotype in mice and in nonhuman primates. Successful vector readministration enabled by SVP[Rapa] allows for dose titration in the liver via targeting of additional populations of hepatocytes at each vector administration. In addition to inhibiting AAVspecific B cell activation, germinal center formation, and antibody production, SVP[Rapa] treatments also reduce antigen-specific T cell recall responses and prevent the appearance of CD8 T cell infiltrates in the liver. Our results suggest that SVP[Rapa] induces a population of regulatory cells that mitigate immune responses specific to the AAV serotype co-administered at the time of SVP [Rapa] treatment and are capable of transferring tolerance to naive recipients. Thus, SVP[Rapa]-mediated immunomodulation represents an attractive strategy to reduce AAV vector immunogenicity.

SVP[Rapa] treatment allows for AAV vector re-administration.
To evaluate the ability of SVP[Rapa] to enable productive redosing of AAV vectors, male C57BL/6 mice were treated first with an AAV8 vector expressing luciferase (AAV8-luc) at day 0, followed by a second administration of an AAV8 vector encoding for human coagulation factor IX (AAV8-hF.IX) on day 21. In this setting, expression of the hF.IX transgene following the second injection of AAV is expected to be inhibited by the immune response induced by the first injection of AAV. Three experimental conditions were tested, (i) administration of both vectors with SVP[Rapa], (ii) administration of both vectors with SVP[empty] control, (iii) administration of the AAV8-hF.IX vector at day 21 only (no additional treatment control). Both AAV8-luc and AAV8-hF.IX vectors were infused at a dose of 4 × 10 12 vg kg −1 (Fig. 1a). SVP [Rapa] inhibited the formation of anti-AAV8 IgG antibody responses after both the first and second injection of AAV vector (Fig. 1b). Conversely, animals treated with empty nanoparticles (SVP[empty]), and control naive animals infused with only 4 × 10 12 vg kg −1 of an AAV8-hF.IX vector alone at day 21, developed high-titer anti-AAV IgG antibodies. Similarly, neutralizing antibody (Nab) titers, measured with an in vitro cell-based assay 24 , were significantly higher in SVP [empty] control vs. SVP[Rapa]-treated animals (Fig. 1c). Accordingly, successful vector re-administration, measured by plasma levels of hF.IX transgene product, was achieved only in animals receiving SVP [Rapa] and in the control group of naive animals dosed at day 21 with AAV8-hF.IX vector only (Fig. 1d). Vector genome copy number (VGCN) in liver was consistent with plasma hF.IX levels, showing lack of transduction in livers of SVP[empty]-treated animals (Fig. 1e). VGCN was lower in SVP [Rapa] vs. naive animals treated at day 21 ( Fig. 1e), possibly reflecting the presence of low titer anti-AAV NAbs which were not detected in the assays used. Importantly, co-administration of SVP[Rapa] at the time of each antigen exposure appeared to be critical for the inhibition of the anti-AAV humoral immune response, as re-administration of AAV8-hF.IX vector alone, with no SVP[Rapa], resulted in the induction of anti-capsid IgG response ( Supplementary Fig. 1a, b). Accordingly, readministration was not successful in animals not treated with SVP[Rapa] ( Supplementary Fig. 2), and co-administration of SVP [Rapa] with AAV8-Luc was more effective in blocking anti-AAV antibodies than administration of SVP[Rapa] one day prior to vector administration ( Supplementary Fig. 3).
We next evaluated the ability of SVP[Rapa] to enable a boost in transgene expression following a repeat dose of AAV vector administered three months after the initial treatment. C57BL/6 mice were dosed with an AAV8 vector encoding secreted human embryonic alkaline phosphatase (AAV8-SEAP) together with SVP[Rapa] or SVP[empty], then rested for 93 days prior to the second treatment (Fig. 1f). Mice treated with SVP[Rapa] showed no significant levels of anti-AAV8 IgG after two vector administrations (Fig. 1g). Importantly, SEAP transgene expression increased approximately twofold after the second injection (Fig. 1h). In contrast, control animals treated with AAV vector only or vector mixed with SVP[empty] showed a boost in antibody titers against AAV8 and no significant increase in SEAP transgene expression after the second injection (Fig. 1g, h). These results indicate that SVP[Rapa] nanoparticles, when coadministered at the same time of AAV vectors, efficiently block anti-capsid antibody responses, allowing for vector re-dosing and modulation of therapeutic efficacy.

SVP[Rapa]
permits AAV vector re-dosing in nonhuman primates. We next assessed the activity of SVP[Rapa] administered in combination with AAV vectors in a pilot study with nonhuman primates (NHP, Macaca fascicularis), using a similar strategy as described in Fig. 1a. Three seronegative animals were treated with AAV8 vectors with either rapamycin nanoparticles (n = 2, SVP[Rapa]#1 and SVP[Rapa]#2) or SVP[empty] control (Fig. 2a). Two vectors encoding for secreted transgenes were administered to the animals, one at day 0 encoding for alpha-acid glucosidase (AAV8-Gaa) and one at day 30 encoding for hF.IX (AAV8-hF. IX.). Both vectors were given at a dose of 2 × 10 12 vg kg −1 . Animals that received the vector and SVP[Rapa] showed efficient inhibition of anti-AAV8 antibody responses, both IgG and IgM, which lasted for the duration of the follow up (Fig. 2b, c). Accordingly, SVP[Rapa], but not control SVP[Empty] nanoparticles inhibited the formation of anti-AAV8 NAbs (Fig. 2d). This was confirmed with an in vivo neutralization assay 25,26 in which naive mice were injected with AAV8-hF.IX together with serum from the experimental monkeys collected on day 30 prior to vector re-administration. Serum from the animal treated with SVP[empty], but not that from monkeys treated with SVP [Rapa], inhibited expression of hF.IX ( Supplementary Fig. 4a).      (Fig. 2a) were biopsied, sectioned, and stained for the Gaa and the hF.IX transgene products. Hepatocytes positive for Gaa, the product of the first vector infused, AAV8-Gaa, were detectable in all 3 animals at similar levels ( Fig. 3a, b and Supplementary Fig. 5a, b). Conversely, staining for hF.IX, derived from the AAV8-hF.IX vector administered at day 30, was detectable only in the SVP [Rapa] #1 and #2 animals but not in the control animal receiving SVP[empty] (Fig. 3a, b and Supplementary Fig. 5a, b). Quantification of hepatocytes from SVP[Rapa]-treated animals showed the presence of hepatocytes that were single positive for Gaa, single positive for hF.IX, and double positive for Gaa and hF.IX (Fig. 3a, b). Similar experiments were conducted in mice with two vectors encoding for the non-secreted transgenes green florescent protein (AAV8-GFP) and human uridine diphosphate (UDP) Glucuronosyltransferase 1A1 (AAV8-hUGT1A1) 13 , administered at day 0 and 21, respectively ( Supplementary Fig. 6a). Liver sections showed that the second transgene (hUGT1A1) was readily detectable in hepatocytes in animals receiving SVP[Rapa] at each vector administration compared to control animals which showed no double staining ( Supplementary Fig. 6b, c).
These results indicate that repeated AAV vector administration enabled by SVP[Rapa] treatment can enhance liver gene transfer in part by increasing the number of hepatocytes that are transduced.

Inhibition of humoral and cellular responses with SVP[Rapa].
Liver enzyme elevation and loss of transgene expression after AAV vector administration in humans correlates with the appearance of capsid-specific CD8 + T cells in the peripheral blood 17 . We evaluated the liver of mice treated with AAV vectors and SVP [Rapa] or SVP[empty] (Fig. 1) for the presence of cells expressing CD8 mRNA by quantitative RT-PCR. CD8 mRNA was significantly elevated in control mice treated with SVP[empty]. In contrast, the level of CD8 mRNA detectable in animals treated with AAV vector and SVP[Rapa] was not significantly different from that of naive mice (Fig. 4a). To further characterize the effect of SVP[Rapa] on antigen-specific B cells and T cells, C57BL/6 mice were immunized with an AAV8 vector encoding the VP1 structural protein of the AAV8 capsid (AAV8-VP1), an approach that has been shown to elicit robust humoral and cellmediated immunity to the capsid 27 Fig. 4i) showed an increase in Tregs and Tfr, suggesting a role for these regulatory T cell subsets in SVP[Rapa]mediated immunomodulation.
Next, we compared the ability of B cell depleting anti-CD20 antibodies 28,29 with SVP[Rapa] to modulate anti-AAV humoral responses ( Supplementary Fig. 9a). Anti-CD20 antibody treatment efficiently depleted the B cells ( Supplementary Fig. 9b), however, it failed to inhibit the anti-AAV8 IgG response following AAV vector administration ( Supplementary Fig. 9c). Conversely, co-administration of SVP[Rapa] with the AAV8 vector blocked anti-vector antibody formation, independent of anti-CD20 treatment ( Supplementary Fig. 9c), supporting the notion that efficient targeting of T helper cells through SVP [Rapa] treatment is more effective than directly targeting B cells through anti-CD20 depletion in preventing anti-AAV antibody responses.

SVP[Rapa]
control memory T cell responses to AAV. As many humans are naturally exposed to wild-type AAV, and thus carry a pool of capsid-specific primed T cells 30 (Fig. 5a, b). Next, we assessed the ability of SVP[Rapa] to control primed T cell responses following adoptive transfer of CD4 + T cells from AAV-treated mice into naive mice (Fig. 5c). Donor animals were immunized with AAV-SEAP at a dose of 4 × 10 11 vg kg −1 . All donor mice developed antibodies to AAV8, indicating immune exposure (Fig. 5d). Sixty-two days later, animals were killed and 1 × 10 7 CD4 + T cells isolated from spleens by negative selection were transferred to naive recipient mice which were then dosed with AAV8 vectors combined with SVP[Rapa] or SVP[empty] at day 1 and 21 post adoptive transfer. No anti-AAV antibodies were observed in SVP[Rapa]-treated recipient mice after the first challenge of with 4 × 10 11 vg kg −1 AAV8-RFP, and only low titer anti-AAV8 antibodies were observed in some animals at late time points after the second injection (Fig. 5e, diamond symbols). Conversely, all animals that received memory CD4 + T cells but then treated with SVP[empty] developed high-titer AAV8 IgG antibodies (Fig. 5e, circles). These results suggest that SVP[Rapa] is capable of controlling re-activation of memory CD4 + T cells upon re-exposure to the AAV capsid antigen.
Antigen-selectivity and effect on Tregs of SVP [Rapa]. Immunosuppression can be associated with adverse events, including infections and malignancies. To test whether the control of immune responses to AAV vectors mediated by SVP[Rapa] was associated with general immunosuppression, we injected mice with an AAV8-Luc vector with SVP[Rapa] and then challenged them three weeks later, with the same capsid (an AAV8-hF.IX vector, Fig. 6b) or with a different capsid (an AAV5-hF.IX vector, Fig. 6c). An additional group of animals was challenged with human F.IX protein formulated in complete Freund's adjuvant (Fig. 6d)      showed good control of humoral responses against the AAV8 capsid (Fig. 6b), animals developed a robust IgG antibody response against the AAV5 capsid (Fig. 6c) and hF.IX protein (Fig. 6d). These results indicate that the immunomodulatory effect of SVP[Rapa] is limited to the antigen co-administered with the nanoparticles. As antigen-specificity could reflect immune tolerance mechanism(s), and because upregulation of regulatory T cells was observed in lymphoid organs of SVP[Rapa] treated animals (Fig. 4f, g), we hypothesized that Tregs had a role in the immunomodulatory effect of SVP[Rapa]. One of the hallmarks of antigen-specific tolerance is the ability to transfer tolerance from tolerized mice to naive mice. Accordingly, splenocytes from donor mice treated with AAV8-luc vector alone or together with SVP[Rapa] or SVP[empty] were collected 14 days post treatment and adoptively transferred into recipient mice, which were then challenged one day later with AAV8-hF.IX vector (Fig. 6e). Cells from donor mice treated with AAV8-luc vector co-administered SVP[Rapa] were able to suppress anti-AAV8 IgG response in recipient mice after challenge (Fig. 6f). Conversely, robust anticapsid IgG formation was observed in mice that received cells from donor mice primed with AAV8-luc and SVP[empty] control (Fig. 6f). To further confirm the role of Tregs cells in SVP [Rapa] control of vector immunogenicity, we performed a depletion experiment using an anti-CD25 antibody (clone PC61) to deplete Tregs. Animals were primed at day 0 with an AAV8-Luc vector (Fig. 6g). Before AAV8-hF.IX vector re-administration at day 21, animals received two injections of the antibody PC61 at day 19 and 20 (200 μg per dose). PC61 administration resulted in profound depletion of CD4 + CD25 + T cells, as measured in spleen and lymph nodes after killing (day 32, Supplementary  Fig. 10). Treg depletion led to a partial restoration of anti-AAV8 IgG antibody responses in SVP[Rapa] treated animals compared to animals treated with SVP[Rapa] with no Treg depletion (Fig. 6f). Conversely, PC61 administration together with SVP [empty] control had no effect on anti-AAV antibody levels.
Together these findings suggest that Tregs have a role as mediators of SVP[Rapa] immunomodulatory activity, although other factors may contribute to the observed complete control of capsid immunogenicity.

Discussion
Mitigating the immunogenicity to AAV is particularly challenging because of its size, the repetitive display of antigenic epitopes on the capsid 32,33 , and the high degree of antibody suppression required to prevent vector neutralization 3,34 . Our results indicate that SVP[Rapa] has the potential to be a safe and effective approach to control capsid immunogenicity associated with systemic delivery of AAV vectors. SVP[Rapa] prevented the formation of NAbs and enabled successful vector re-administration in mice (Fig. 1) and nonhuman primates (Fig. 2) in an antigenselective manner (Fig. 6b). Furthermore, we demonstrated that repeat dosing of AAV vectors with SVP[Rapa] can be used to increase expression of a transgene for example in the liver via targeting of additional hepatocyte populations at each vector infusion ( Fig. 3 and Supplementary Fig. 5). These findings are of high relevance to gene therapies for diseases with early lethality 10 , in which early intervention may require subsequent treatments to maintain efficacy over time 12 . They also address the need for strategies to carefully dose-titrate AAV vectors to avoid overexpression 7 , or to target systemic diseases 35 in which multiple vector administrations are likely needed to achieve full therapeutic efficacy 36 . Despite numerous attempts, the ability to re-administer AAV vector still represents an unmet goal for the field of in vivo gene transfer. Some degree of success has been achieved in the context of intramuscular gene transfer 37 or gene transfer to immunoprivileged body sites 38,39 , in which vector neutralization by circulating antibodies is not a concern 40 . The goal of addressing AAV capsid immunogenicity is not limited to prevention of antibody responses to allow for vector administration. Results in human trials indicated that induction of CD8 + T cell responses directed to the AAV capsid correlates with liver inflammation and can limit the duration of transgene expression [2][3][4][5][6][7]17 . While administration of corticosteroids has been effective in blocking these responses in some cases 2,5,6 , in other instances expression was lost despite immunosuppression 41,42 , indicating that more effective strategies need to be developed. Here, we observed a marked suppression of de novo capsid T cell responses with SVP[Rapa], even in the context of the expression of the capsid VP1 protein 27 (Fig. 4b). Although liver enzyme elevation is not typically observed in animal models of AAVmediated gene therapy 17,43 , we did observe evidence of CD8 T cell infiltrates in the liver after vector administration, which were inhibited by SVP[Rapa] treatment (Fig. 4a). Additionally, we observed inhibition of memory T cell responses, in a setting closer to that of gene transfer in humans pre-exposed to the virus 30,31 (Fig. 5). Thus, despite the limitations of animal models in predicting anti-capsid CD8 + T cell responses 17,43 , our results suggest that administration of SVP[Rapa] may be beneficial in terms of modulation of both humoral and cell-mediated responses to the AAV capsid.
Immunosuppression, including the use of free rapamycin in combination with other immunosuppressive agents 44 , has been tested in gene therapy to prevent anti-vector antibody formation. This approach has proved poorly efficacious in preventing anti-AAV induction, even when potent drug combinations were used 45 , in some cases resulting in detrimental interference with the induction regulatory T cells, which are essential for transgene maintenance in liver gene transfer 44,46 . Recently, a clinical case study was published on the effect of long-term depletion of B cells using rituximab combined with rapamycin and steroids administration around the time of AAV vector administration, which resulted in the abrogation of the anti-capsid antibody response 28 . However, data available indicate that long-term immunosuppressive treatment is likely needed in order to be efficacious. Our results indicate that transient depletion of B cells with an anti-CD20 antibody had no substantial effect on the anti-capsid antibody response in mice and provided no additional benefit to SVP[Rapa] treatment ( Supplementary Fig. 8).
Plasmapheresis has also been proposed to re-administer AAV vectors 47 ; however, this maneuver has several limitations, as human studies indicate that repeated sessions will be needed to eliminate anti-AAV antibodies present in the extravascular space, and that this approach is unlikely to completely eradicate high-titer antibodies 48 . Similarly, capsid switching has been proposed 49 , however this is also a very complex procedure from a drug development and manufacturing perspective, and it is further complicated by high degree of cross-reactivity of anti-AAV antibodies 50 .
In the setting of AAV liver gene transfer, administration of SVP[Rapa] concomitant to each vector exposure was sufficient to Fig. 6 Antigen-specificity of SVP[Rapa] treatment and role of regulatory T cells. a Protocol outline. Male C57BL/6 mice (n = 5) were first injected i.v. with 4 × 10 12 vg kg −1 of AAV8-luciferase vector together with SVP[Rapa] or with SVP[empty] control. Animals were then challenged i.v. on day 21 with 4 × 10 12 vg kg −1 of AAV8-hF.IX vector or AAV5-hF.IX vector or injected s.c. with hF.IX in complete Freund's adjuvant (cFA). b Anti-AAV8 IgG in animals primed with AAV8-luc and challenged with AAV8-hF.IX. c Anti-AAV5 IgG in animals primed with AAV8-luc and challenged with AAV5-hF.IX, and d anti-hF.IX IgG antibodies determined by ELISA in animals primed with AAV8-luc and challenged with hF.IX protein in cFA. e Protocol outline. Donor mice (n = 5) were left untreated or treated i.v. with 4 × 10 11 vg kg −1 of AAV8-luciferase vector together with SVP[Rapa] or SVP[empty] control. 14 days later, spleens were harvested and 3 × 10 7 splenocytes were adoptively transferred into recipient mice (n = 5). All animals were then treated one day later with 4 × 10 11 vg kg −1 of AAV8-hF.IX vector. f Anti-AAV8 IgG antibodies measured 14 days post AAV8-hF.IX vector injection and measured by ELISA. g Protocol outline. Male C57BL/6 mice (n = 5 per group) were treated i.v. with 4 × 10 12 vg kg −1 of AAV8-luciferase vector together with SVP[Rapa] or SVP[empty] control on day 0, followed by Tregs depletion (anti-CD25, clone PC61, 200 µg per mouse) on days 19 and 20. All animals were then challenged with 4 × 10 12 vg kg −1 of AAV8-hF.IX vector alone on day 21. h Analysis anti-AAV8 IgG antibodies measured by ELISA. SVP[Rapa] treatment consisted of 200 µg of rapamycin. Data are shown as mean ± s.d. Statistical analyses were performed by two-way ANOVA with Tukey post hoc test in b-d, by one-way ANOVA with Tukey post hoc test in f, and by non-parametric unpaired, two-tail t-test in h (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001) completely inhibit anti-capsid immune responses. Coadministration of SVP[Rapa] with each dose of AAV vector appeared to be essential to completely block anti-AAV antibody responses, as repeated vector administration with no SVP[Rapa] resulted in antibody formation even when SVP[Rapa] was coadministered with the first vector dose (Supplementary Fig. 1). Consistently, data generated indicate that the effect of SVP [Rapa] is selective for the co-administered antigen, as we showed that there was no effect on the antibody response to a different AAV serotype or to an unrelated protein antigen administered at a later time point (Fig. 6b). Regulatory T cells seem to be involved as suppressors of capsid immunogenicity mediated by SVP[Rapa], as previously shown for other antigens 18,19 and as demonstrated by the observed increase in Treg frequency in lymphoid organs of treated mice (Fig. 4g), and by adoptive T cell transfer (Fig. 6f) and Treg depletion experiments (Fig. 6h). These observations are consistent with the published beneficial effect of rapamycin on the homeostasis of Tregs 51 , and further supports the safety of the approach, as SVP[Rapa]-induced immune modulation appears to be highly antigen-selective, and because Tregs require antigen persistence for homeostasis 52 . The inability of anti-CD25 treatment to completely restore the anti-AAV antibody response may reflect incomplete deletion of CD25 + T cells in lymphoid organs or involvement of CD25-negative regulatory cells 53 . Alternatively, it is plausible that other mechanisms are mediating the effect of SVP[Rapa] on capsid immunogenicity, including the immunosuppressive effect of rapamycin. Whether the effect of SVP[Rapa] observed here, in the context of liver gene transfer with AAV8 vector doses that are clinically relevant 5,6 , will be identical for other capsids or at higher vector doses remains to be established. Future studies in larger cohorts of nonhuman primates will be required for the clinical translation of the results shown here.
We have previously demonstrated the functional benefit of using SVP[Rapa] to mitigate the formation of anti-drug antibodies (ADAs) against a variety of biologic drugs, including coagulation factor VIII in a model of hemophilia A 23 , anti-TNF monoclonal antibody in a model of spontaneous arthritis 18 , acid alphaglucosidase in Pompe disease mice 20 , immunotoxins in a model of mesothelioma 21 , and pegylated uricase in uricase-deficient mice and in nonhuman primates 18 . Additionally, the safety and ability of SVP[Rapa] to inhibit the formation of ADAs against pegisticase, a highly immunogenic enzyme therapy, in a dose-dependent manner was demonstrated in a phase 1 clinical trial (Clinicaltrials. gov NCT02648269) 22 . SVP[Rapa] is currently being evaluated in combination with pegsiticase in an ongoing multi-dose phase 2 clinical trial (Clinicaltrials.gov NCT02959918) 22 .
One open question about the work presented here is whether SVP[Rapa] administration will impair the ability of the immune system to fight infections. Because of the need for SVP[Rapa] to be co-administered with the antigen, in order to guarantee uptake from the same subset of antigen presenting cells 18,21 , it is less likely that SVP[Rapa] will lead to global immunosuppression, although future studies are needed to carefully address this point. The lack of immune cell depletion observed in animals treated with AAV vectors and SVP[Rapa], both rodents and nonhuman primates, provides additional data supporting the safety of the approach. Initial clinical data in small numbers of patients treated with SVP[Rapa] are consistent with these results, as they do not show any increased risk of infection (Clinicaltrials.gov NCT02648269; NCT02959918) 22 . Although AAV is considered to be a non-pathogenic virus and is not associated with any disease, it is unknown if there would be any consequences of inducing immune tolerance to AAV. Further evaluation of safety will have to be addressed in preclinical studies and clinical trials. Finally, another question about the technology is whether it could be used to eradicate established antibody responses to AAV. SVP [Rapa] has been shown to reduce antibody titers against protein antigens in primed animals 21,23 . However, since very low titers of neutralizing antibodies can inhibit AAV transduction, it is unclear if a reduction in anti-AAV titer would be sufficient to enable productive transgene expression, particularly in the presence of long-lived plasma cells. We are investigating whether SVP[Rapa] combined with drugs targeting B cells or plasma cells 54 may enable vector administration in the setting of preexisting neutralizing antibodies.
In conclusion, the work presented here addresses AAV vector immunogenicity and specifically the longstanding challenge of vector re-administration. SVP[Rapa] co-administration with AAV vectors is a powerful, yet highly specific technology that enables the prevention of anti-AAV humoral immune responses and T cell reactivity to the capsid. This is a promising approach for safe and effective repeated vector dosing to dose titrate the therapeutic effect and to re-treat recipients of gene therapy in case efficacy is lost over time, thus opening new therapeutic avenues for AAV vector-mediated gene transfer for diseases requiring systemic transduction 35 or treatment in childhood due to early lethality 10 .
Methods SVP[Rapa] nanoparticles production. Rapamycin-loaded poly(lactic acid) (PLA) nanoparticles (SVP[Rapa]) were prepared using oil-in-water single-emulsion evaporation method 18,19 . Briefly, rapamycin, PLA, and PLA-PEG block copolymer were dissolved in dichloromethane solution to form the oil-phase. The oil-phase was then added to an aqueous solution of polyvinylalcohol in phosphate buffer followed by sonication (Branson Digital Sonifier 250A, Branson Ultrasonics, Danbury, CT). The emulsion thus formed was added to a beaker containing phosphate buffer solution and stirred at room temperature for 2 h to allow the dichloromethane to evaporate. The resulting nanoparticles containing rapamycin were washed twice by centrifugation at 76,600×g at +4°C and the pellet was resuspended in phosphate buffer solution. The bare nanoparticles without rapamycin (SVP[empty]) were prepared in identical conditions. Rapamycin content was measured by first transferring the nanoparticles into an organic solvent, followed by analysis via reverse-phase HPLC using a water/acetonitrile gradient containing 0.1% TFA. Typically, SVP[Rapa] contained~10% rapamycin load; in a given preparation of SVP[Rapa], the concentration of rapamycin was adjusted to 2 mg ml. Nanoparticle size was ∼200 nm as determined by dynamic light scattering. Each SVP[Rapa] treatment consisted of 50 or 200 μg of rapamycin as indicated in figure legend.
AAV vector production. AAV vectors were produced using adenovirus-free transfection method and purified by CsCl gradient centrifugation 55,56 . Genomecontaining AAV vectors and empty AAV capsid particles were titrated using a quantitative real-time polymerase chain reaction (qPCR) 55,56 . The transgene expression cassettes for hF.IX, Gaa, and UGT1A1 were under the control of the apolipoprotein E (Apo E) enhancer/alpha1-antitrypsin (hAAT) liver-specific promoter. The expression cassettes of luciferase (luc), AAV8 capsid protein VP1, RFP, GFP, and SEAP were under the control of a CMV promoter. All vectors were free of endotoxin.
Animal studies. Animal studies were performed in accordance to the American and European legislations and approved by the intuitional ethical committee of the Centre d'Exploration et de Recherche Fonctionnelle Expérimentale (Evry, France, protocol number APAFIS 3055-20151019213299180) and by the animal care and use committee of Avastus Preclinical Services (Cambridge, Massachusetts). C57BL/ 6 mice were obtained from Jackson Laboratories. All mice were 4-to 8-weeks-old male at the onset of the experiments. Experimental procedures using nonhaplotype matched male cynomolgus monkeys (Macaca fascicularis) were approved by the institutional animal care and use committee (protocol number APAF1S 2651-2015110216583210 15) and were conducted at Nantes veterinary school (Oniris, Nantes, France). Animals were randomized to treatment groups at the beginning of each study and the investigator was blinded to the group allocation. For mouse studies, the minimal sample was determined 57 to allow to detect significant differences among treatment groups. In all mouse experiments, with the exception of Figure 5b (n = 4), a minimum of five animals per group was used. In an additional set of experiments, male C57BL/6 mice were left untreated or treated with AAV8-Luc vector (4 × 10 12 vg kg −1 ) together with SVP[Rapa] or SVP [empty]. 14 days later, spleens were collected and 3 × 10 7 splenocytes were adoptively transferred into recipient C57BL/6 mice. All animals were then treated with an AAV8-hF.IX vector (4 × 10 12 vg kg −1 ).
Antigen-specificity studies. Male C57BL/6 mice were treated i.v. with an AAV8luc vector (4 × 10 12  Anesthesia performed before injection was induced first via intramuscular injection of demetomidine (30 mg kg −1 ) and ketamine (7 mg kg −1 ) then maintained with isoflurane (Vetflurane®) mixed with oxygen. AAV8 vectors and nanoparticles were infused in a volume of 10 or 5 ml, respectively, via the saphenous vein.
In all animal experiments, peripheral blood was collected and sera were isolated or immediately transferred to tubes containing citrates or EDTA to isolate plasma, at baseline and indicated time points. Spleen and inguinal lymph nodes were collected at necropsy in fresh RPMI medium and diverse organs were collected and stored at −80°C for further analysis as indicated in figure legend. Cells counts in mouse peripheral blood were performed using MS9 automated cell counter with veterinary parameters and reagents (MS9; Schloessing Melet, Cergy Pontoise, France). For monkey samples, blood cell counts and biochemical analysis were performed at Nantes veterinary school (Oniris, Nantes, France).
The samples derived from experiments performed at Selecta Bioscience were analyzed with a different AAV8 antibody assay 60 and results are expressed as EC50 titers or top OD value. Briefly, 96-well plates were coated overnight with AAV8 vector at 2.2 × 10 9 vg ml −1 , washed and blocked with 1% casein the following day, then diluted serum samples (1:40) added to the plate and incubated overnight. Plates were then washed; a rabbit anti-mouse IgG specific-HRP (1:1500) was added to the wells and incubated for two hours. The presence of bound IgG antibodies was detected by adding TMB substrate and measuring at an absorbance of 450 nm with a reference wavelength of 570 nm.
Selected serum samples were also analyzed for anti-AAV neutralizing antibody titer using an in vitro cell-based test 24 . Briefly, in this assay, serial dilutions of heatinactivated test samples were mixed with a vector expressing luciferase and incubated for one hour. After incubation, samples were added to cells and residual luciferase expression was measured after 24 h. The neutralizing titer was determined as the highest sample dilution at which at least 50% inhibition of luciferase expression was measured compared to a non-inhibition control. In this assay, a NAb titer of 1:10 represents the titer of a sample in which after a 10-fold dilution a residual luciferase signal lower than 50% of the non-inhibition control is observed.
Factor IX plasma levels determination. Plasma levels of human F.IX transgene were determined with an ELISA assay 29,59 . The detection of hF.IX antigen levels in mouse plasma was performed using the monoclonal antibodies against hF.IX (GAFIX-AP, Affinity Biologicals, Ancaster, Canada) for coating and anti-hFIX-HRP for detection (GAFIX-APHRP, Affinity Biologicals, Ancaster, Canada). In NHP samples, anti-hFIX antibody (MA1-43012, Thermo Fisher Scientific Waltham, MA), and anti-hF.IX-HRP antibody (CL20040APHP, Tebu-bio, Le Perrayen-Yvelines, France) were used for coating and detection, respectively. The presence of hF.IX antigen levels was determined by adding SIGMAFAST TM OPD substrate (Sigma-Aldrich) following manufacturer's instructions and measuring OD at an absorbance of 492 nm.
AAV vector genome copy number determination. Tissue DNA was extracted from whole organ using the Magna Pure 96 DNA and viral NA small volume kit (Roche Diagnostics, Indianapolis IN) according to the manufacturer's instructions. Vector genome copy number (VGCN) was performed using qPCR with primers specific for hAAT promoter for mice and for the hF.IX transgene cDNA for NHPs. Titin and albumin were used as normalizing genes in murine or NHP samples, respectively. VGCN was quantified by TaqMan real-time PCR with the ABI PRISM 7900 HT sequence detector (Thermo Fisher Scientific, Waltham, MA). The sequence of primers and probes is listed in Supplementary Table 4. Immunohistochemistry. For Gaa, hF.IX and UGT1A1 staining, multiple wedge biopsies (left, right, quadrate and caudate lobes) from livers were fixed in 4% paraformaldehyde (PFA) for 10 min at room temperature (RT). Tissues were then incubated in 15% sucrose overnight then in 30% sucrose. The tissues were next placed in OCT, frozen in cold isopentane and subjected to cryosection (8 µm). Tissue sections were blocked with 5% BSA in PBS plus 4% goat serum and 0.1% Triton for 2 h at RT. Tissue Samples were incubated overnight at +4°C with a monoclonal antibody against Gaa (clone EPR4716(2), dilution 1:100, Abcam, Cambridge, UK) and hF.IX (clone MA1-43012, dilution 1:100, Thermo Fisher Scientific Waltham, MA) or with monoclonal antibody against GFP (clone 1020, dilution 1:300, AVES Labs, Tigard, OR) and serum from rats immunized with UGT1A1 protein (Dilution 1:100). Following washes in PBS with 0.05% Tween-20, samples were incubated with secondary antibodies conjugated to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 594 (dilution 1:400, Thermo Fisher Scientific Waltham, MA). Following incubation and washes, sections were counterstained with Dapi. Images were captured using Confocal Microscope Leica SP8 (Leica MicroSystems, Wetzlar, Germany). Analysis of images was performed using the Cellprofiler software; staining signal was isolated from background signal (intensity 1 to 19 random units). A minimum of four fields per animal were acquired and analyzed to generate the data on distribution of signal intensity. The signal intensity was then divided by the surface in mm 2 to obtain number of positive hepatocytes per mm 2 .
Anti-AAV T and B cell ELISpot assays. Spleens were isolated, single-cell suspensions were then prepared and 2 × 10 5 cells were stimulated for 24 h with an AAV8 peptide pool 6 (1 μg ml −1 of each peptide, Mimotopes, Victoria, Australia) in 100 μl of complete medium (RPMI supplemented with 10% FBS, Pen/Strep and L-Glutamine) (Thermo Fischer Scientific, Waltham, MA). Medium only was used as negative control and ConA (Concanavalin A Lectin Type IV, dilution 1:1000, Sigma-Aldrich, St. Louis, MO) was used as positive control. Additionally, cells were also stimulated in vitro for 7 days with AAV8 peptides.
For quantification of AAV8 antibodies secreting cells (ASC), 96-multiscreen plates (Merck-Millipore, Seattle, WA) were coated with empty AAV8 particles (1 × 10 10 particles per well). After overnight incubation at +4°C, plates were washed three times and blocked with complete medium (RPMI supplemented with 10% FBS, Pen/Strep and L-Glutamine) (Thermo Fischer Scientific) for 2 h at RT. 2 × 10 5 splenocytes per well and 5 × 10 5 splenocytes per well were incubated for 24 h at 37°C and 5% CO 2 . Plates were then washed and ASC were detected by incubation with biotin conjugated anti-IgG (for murine samples, Southern Biotech, Brimingham, AL; for NHP samples, Mabtech, Nacka Strand, Sweden) or biotin conjugated anti-mouse IgM (southern Biotech, Brimingham, AL) for 2 h at RT. After three washes, streptavidin-ALP (Mabtech, Nacka Strand, Sweden) was added for 1 h at RT, followed by the addition of the substrate for spot development (Mabtech). All antigens and controls were tested in triplicate. The plates were read on AID ELISpot reader system using the ELISpot 6.0 iSpot software (Autoimmun Diagnostika, Strassberg, Germany).
RNA extraction and RT-PCR. Total RNA were isolated from livers using the Magna Pure 96 cellular RNA large volume kit (Roche Diagnostics, Indianapolis IN) according to the manufacturer's instructions. RNA samples were then pretreated with DNAse (Thermo Fischer Scientific, Waltham, MA) and reverse transcribed using RevertAid H minus first strand cDNA synthesis kit (Thermo Fischer Scientific). RT-qPCR was performed using SybrGreen (Thermo Fisher Scientific) with primers specific for CD8 (Supplementary Table 4). Gene expression was calculated using the ΔΔC t method relative to GAPDH housekeeping gene (Supplementary  Table 4) and by normalizing to the mean expression of untreated mice.
Statistical analysis. Statistical analysis was performed using GraphPad Prism (GraphPad Software Inc, La Jolla, CA) version 6.0. Analysis of statistical significance between two groups was assessed using the two-tailed unpaired t-test. For multiple group comparisons, one-way or two-way ANOVA with Tukey multiple comparison test was used.

Data availability
Data supporting the findings of this manuscript are available from the corresponding author upon reasonable request.