An ongoing controversy in the gene therapy field is whether an effective CD8+
T cell response to vector-derived capsid sequences arises following transduction by AAV vectors. Studies of AAV-mediated gene transfer in rodents, canines, and non-human primates demonstrate long-term transgene expression and no evidence of a CD8+
T cell response to capsid (17
). These results are in contrast to findings from a clinical trial in which hepatic infusion of AAV resulted in a transient, asymptomatic rise in liver enzymes coincident with loss of transgene expression (1
). Using MHC pentamers, our group previously documented the expansion and contraction of capsid-specific CD8+
T cells in a human subject following infusion with AAV vector (2
). The temporal kinetics of expansion and contraction of this antigen-specific population closely matches the asymptomatic rise and fall in liver enzymes. These observations suggested that host CTL responses directed against AAV capsid resulted in destruction of transduced hepatocytes, thereby accounting for the transient rise in hepatic enzymes and loss of transgene expression. However, there is some skepticism regarding this hypothesis, as the vector does not carry gene sequences for capsids. Consequently, it is unclear whether pre-formed capsids could gain entry into MHC class I antigen presentation pathways in sufficient quantity to drive effective capsid-directed CTL responses (23
). The studies presented herein were thus undertaken in an effort to determine the molecular and cellular basis for what appeared to be destruction of transduced hepatocytes in humans infused with AAV vector.
Using our TCR multimer reagent, we have demonstrated that low levels of AAV capsid–derived antigen are indeed presented on MHC I by hepatocytes. Endogenous expression levels of MHC I on hepatocytes in vivo are not as trivial as once thought, as class I is detectable on normal hepatocytes in mice and humans and therefore is less likely to be rate limiting for antigen presentation (24
). However, the molecular fate of AAV capsid after cellular entry, endosomal trafficking, and delivery of its genetic payload is not completely understood. In particular, it remains unclear how AAV engages class I antigen presentation pathways in vivo, as capsid is not being expressed but rather is an exogenous antigen (27
). It has been demonstrated that ubiquitination of AAV capsid proteins occurs after endocytosis (28
) and is facilitated by EGFR-PTK signaling (30
). Proteasomal inhibitors can augment transduction, suggesting that AAV capsids are degraded by proteasomes and that this process limits transduction efficiency (28
). If a pre-formed capsid engages antigen processing machinery while still within endosomes, then cross-presentation pathways are likely involved (34
). In this regard, both liver sinusoidal endothelial cells (36
) and hepatocytes (37
) have been shown to be capable of presenting exogenous antigen, although the molecular details have yet to be completely elucidated. On the other hand, cross-presentation may not be required, as AAV must escape from the endosome to release its DNA. Intact pre-formed proteins in the cytosol can access MHC I pathways, as cytosolic microinjection of ovalbumin protein results in subsequent antigen presentation on MHC I (38
). Thus, free capsid in the cytosol may be similarly inducted into MHC I pathways without de novo protein synthesis. These possibilities are not mutually exclusive, and it is not clear whether species differences in hepatocyte antigen presentation exist as well.
We have also shown that both AAV vectors containing DNA and empty capsid particles were able to sensitize hepatocytes for cytolysis, but only the former is capable of mediating gene transfer. This underscores the importance of minimizing contamination by empty capsids during the manufacturing process (40
), to increase transduction efficiency while reducing capsid antigen burden. The low numbers of pMHC complexes we observed by confocal microscopy after AAV transduction were immunologically relevant, as they rendered the transduced hepatocytes effective targets for cytolysis. Indeed, T cells have been shown to have exquisite sensitivity for antigen. Using cell lines that stably express HIV-1 to recover endogenously processed viral peptides, it has been demonstrated that target cells can be sensitized by as few as 10–100 peptides for lysis by HIV-specific CTLs (41
). In an alternative approach, CTLs have been shown to have detectable increases in intracellular calcium after stimulation by 1 pMHC complex, inducible cytotoxic activity by 1–3 complexes, and maximal calcium flux with mature immunologic synapse formation by 8–10 complexes (42
). Thus it is not surprising that capsid-derived pMHC complexes were undetectable by flow cytometry, at the threshold for detection by confocal microscopy, and yet immunologically sufficient to trigger CTL-mediated destruction, even at 1/30th the dose of AAV vector used for visualization by confocal microscopy.
One unresolved puzzle is why the loss of transgene expression observed in a human clinical trial (1
) was not predicted by rodent, canine, and nonhuman primate animal models of AAV-mediated gene transfer in which transgene expression is long lived (17
). Perhaps the most obvious influence is that many humans have had antecedent infections with wild-type AAV2 and harbor memory CD8+
T cells not present in animal models. Also, clearly this is not the first example in which animal models have failed to recapitulate human immune responses. In a phase I clinical trial the superagonist anti-CD28 monoclonal antibody TGN1412 induced a massive cytokine storm in human recipients, resulting in multiorgan failure in 6 healthy volunteers at 1/500th the dose that appeared safe in mice and nonhuman primates (44
). Subsequent investigation demonstrated that superagonist activation of CD28 leads to a sustained calcium influx in human T cells, but not cynomolgus or rhesus monkey T cells (45
). Human T cells also proliferate more vigorously after TCR activation than do T cells from chimpanzees, and this difference has been attributed to loss of inhibitory Siglec expression during human evolution (46
). Differences between human and murine T cells also exist, as type I IFNs can drive T cell IFN-γ production from humans but not mice due to divergent regulation of STAT1 signaling pathways (47
). Together, these observations indicate that human T cells are more sensitive to stimulation than their rodent or nonhuman primate counterparts and may account for the differences observed between human and animal studies with AAV-mediated gene therapy.
Finally, we have demonstrated a potential therapeutic application for soluble TCR multimers in specifically blocking T cell responses. By competing with CTLs for capsid-derived pMHC complexes, TCR multimers can interfere with immunologic synapse formation between the T cell and the antigen-presenting cell. TCR multimers have previously been shown to inhibit CTL expansion (13
) and secretion of MIP-1β (12
) following antigenic stimulation. We now show, for what we believe is the first time, that soluble TCR multimers are also capable of protecting cells from CTL-mediated cytolysis. This approach provides specificity to the inhibition of T cell responses, in contrast to the broad immunosuppression induced by corticosteroids or other nontargeted agents. We have demonstrated the utility of TCR multimers in preventing destruction of AAV-transduced hepatocytes in culture. This approach could potentially be generalized for other CTL-mediated diseases, such as type I diabetes mellitus or autoimmune hepatitis. However, the TCR would need to be tailored to an individual’s HLA allele and to the immunodominant epitope of interest. Further studies are necessary to delineate the optimal dosing, pharmacokinetics, and potential immunogenicity of soluble TCR therapeutics.
In conclusion, our studies provide mechanistic evidence to account for the observed loss of F9
expression and self-limited rise in aminotransferases following hepatic gene transfer by AAV in humans. These data provide the rationale for an ongoing clinical study in which AAV-mediated gene transfer to the liver is accompanied by a short course of immunomodulation to block the CD8+
T cell response to capsid (49
). These results have implications for future clinical trials of AAV-mediated hepatic gene transfer and provide insights into how successful long-lived gene transfer might be achieved.