Recombinant adeno-associated virus (rAAV) vectors are one of the most promising in vivo gene delivery tools. Several features make rAAV vectors an ideal platform for gene transfer. However, the high homology with the parental wild-type virus, which often infects humans, poses limitations in terms of immune responses associated with this vector platform. Both humoral and cell-mediated immunity to wild-type AAV have been documented in healthy donors, and, at least in the case of anti-AAV antibodies, have been shown to have a potentially high impact on the outcome of gene transfer. While several factors can contribute to the overall immunogenicity of rAAV vectors, vector design and the total vector dose appear to be responsible of immune-mediated toxicities. While preclinical models have been less than ideal in predicting the outcome of gene transfer in humans, the current preclinical body of evidence clearly demonstrates that rAAV vectors can trigger both innate and adaptive immune responses. Data gathered from clinical trials offers key learnings on the immunogenicity of AAV vectors, highlighting challenges as well as the potential strategies that could help unlock the full therapeutic potential of in vivo gene transfer.

Low‐affinity MHC class I‐associated cryptic epitopes derived from self proteins overexpressed in a wide variety of human tumors or derived from antigens of viruses exhibiting a high mutation rate, could be interesting candidates for tumor and virus immunotherapy, respectively. However, identification of low‐affinity MHC‐associated epitopes comes up against their poor immunogenicity. Here we describe an approach that enhances immunogenicity of nonimmunogenic low‐affinity HLA‐A2.1‐binding peptides. It consists of modifying their sequence by introducing a tyrosine in the first position (P1Y). P1Y substitution enhances affinity of HLA‐A2.1‐associated peptides without altering their antigenic specificity. In fact, P1Y variants of ten nonimmunogenic low‐affinity peptides exhibited a 2.3‐ to 55‐fold higher binding affinity and/or stabilized the HLA‐A2.1 for at least 2 h more than the corresponding native peptides. More importantly, P1Y variants efficiently triggered in vivo native peptide‐specific CTL which also recognized the corresponding naturally processed epitope. The possibility for generating CTL against any low‐affinity HLA‐A2.1‐associated peptide provides us with the necessary tool for the identification of cryptic tumor and virus epitopes which could be used for peptide‐based immunotherapy.

Wiskott-Aldrich syndrome (WAS) is an immune deficiency with thrombopenia resulting from mutations in the WASP gene. This gene normally encodes the Wiskott-Aldrich syndrome protein (WASP), a major cytoskeletal regulator expressed in hematopoietic cells. Gene therapy is a promising option for the treatment of WAS, requiring that clinically applicable WASP gene transfer vectors demonstrate efficacy in preclinical studies. Here, we describe a selfinactivating HIV-1-derived lentiviral vector encoding human WASP and show that it effectively transduced bone marrow progenitor cells of WASP knockout (WKO) mice. Transplantation of these transduced cells into lethally irradiated WKO recipients led to stable expression of WASP and correction of immune, inflammatory and cytoskeletal defects. Splenic T-cell proliferation was restored, podosomes were reinstated on bone-marrow-derived dendritic cells and colon inflammation was reduced. This shows for the first time (a) that cytoskeletal defects can be corrected in WKO mice, (b) that human WASP is biologically active in mice and (c) that a lentiviral vector is effective to express human WASP in vivo over several months. These data support further development of such lentiviral vectors for the gene therapy of WAS. Gene Therapy (2005) 12, 597-606.

Most of the human tumor-associated antigens (TAAs) characterized thus far are derived from nonmutated "self"-proteins. Numerous strategies have been developed to break tolerance to TAAs, combining various forms of antigens with different vectors and adjuvants. However, no study has yet determined how to select epitopes within a given TAA to induce the highest antitumor effector response. We addressed this question by evaluating in HLA-A*0201-transgenic HHD mice the antitumor vaccination efficacy of high-and low-affinity epitopes from the naturally expressed murine telomerase reverse transcriptase (mTERT). Immunity against low-affinity epitopes was induced with heteroclitical variants. We show here that the CTL repertoire against high-affinity epitopes is partially tolerized, while that against low-affinity epitopes is composed of frequent CTLs with high avidity. The highaffinity p797 and p545 mTERT epitopes are not able to protect mice from a lethal challenge with the mTERT-expressing EL4-HHD tumor. In contrast, mice developing CTL responses against the p572 and p988 low-affinity epitopes exhibit potent antitumor immunity and no sign of autoimmune reactivity against TERT-expressing normal tissues. Our results strongly argue for new TAA epitope selection and modification strategies in antitumor immunotherapy applications in humans.

Tolerance to tumor-nonmutated self proteins represents a major obstacle for successful cancer immunotherapy. Since this tolerance primarily concerns dominant epitopes, we hypothesized that targeting cryptic epitopes that have a low affinity for HLA could be an efficient strategy to breach the tolerance to tumor Ags. Using the P1Y heteroclitic peptide approach, we identified low affinity cryptic HLA-A*0201-restricted epitopes derived from two widely expressed tumor Ags, HER-2/neu and hTERT. The P1Y variants of four HER-2/neu (neu391, neu402, neu466, neu650)- and two hTERT (hTERT572 and hTERT988)-derived low affinity peptides exhibited strong affinity for HLA-A*0201 and stimulated specific CTL from healthy donor PBMCs. These CTL specifically recognized HER-2/neu- and hTERT-expressing tumor cells of various histological origins. In vivo studies showed that HLA-A*0201 transgenic HHD mice vaccinated with the P1Y variant peptides generated CTL that specifically lysed Ag-expressing tumor cells, thus recognizing the cognate endogenous Ags. These results suggest that heteroclitic variants of low affinity, cryptic epitopes of widely expressed tumor Ags may serve as valid tools for tumor immunotherapy.

Thymus-derived regulatory T cells (Tregsautoimmunity ͉ immune tolerance ͉ nervous system

CD4+ T cell antitumor responses have mostly been studied in transplanted tumors expressing secreted model antigens (Ags), while most mutated proteins in human cancers are not secreted. The fate of Ag-specific CD4+ T cells recognizing a cytoplasmic Ag in mice bearing autochthonous tumors is still unclear. Here we show, using a genetically engineered lung adenocarcinoma mouse model, that naive tumor-specific CD4+ T cells are activated and proliferate in the tumor-draining lymph node (TdLN) but do not differentiate into effectors or accumulate in tumors. Instead, these CD4+ T cells are driven toward anergy or peripherally-induced Treg (pTreg) differentiation, from the early stage of tumor development. This bias toward immune suppression is restricted to the TdLN, and is maintained by Tregs enriched in the tumor Ag-specific cell population. Thus, tumors may enforce a dominant inhibition of the anti-tumor CD4 response in the TdLN by recapitulating peripheral self-tolerance mechanisms.

Gene transfer efficacy is limited by unwanted immunization against transgene products. In some models, immunization may be avoided by regulating transgene expression with mir142.3p target sequences. Yet, it is unclear if such a strategy controls T-cell responses following recombinant adeno-associated viral vector (rAAV)-mediated gene transfer, particularly in muscle. In mice, intramuscular rAAV1 gene delivery of a tagged human sarcoglycan muscle protein is robustly immunogenic and leads to muscle destruction. In this model, the simple insertion of mir142.3p-target sequences in the transgene expression cassette modifies the outcome of gene transfer, providing high and persistent levels of muscle transduction in C57Bl/6 mice. Such regulated vector fails to prime specific CD4 and CD8 T cells; although, transgene tolerance seems to result from ignorance and could be broken by a robust antigenic challenge. While effective in normal mice, the mir142.3p-regulated transgene remains immunogenic in sarcoglycan-deficient dystrophic mice. In these mice, transgene expression is only prolonged but does not persist as effector CD4 and CD8 T-cell responses develop. Thus, using a mir142.3p-regulated transgene can improve rAAV muscle gene transfer results, but the level of efficacy depends on the context of application. In normal muscle, this strategy is sufficient to prevent immunization and functions even more effectively than tissue-specific promoters. In dystrophic models, additional strategies are required to fully control T-cell responses.