Type I IFN Sensing by cDCs and CD4⁺ T Cell Help Are Both Requisite for Cross-Priming of AAV Capsid-Specific CD8⁺ T Cells

Abstract

Adeno-associated virus (AAV) vectors are widely used in clinical gene therapy to correct genetic disease by in vivo gene transfer. Although the vectors are useful, in part because of their limited immunogenicity, immune responses directed at vector components have complicated applications in humans. These include, for instance, innate immune sensing of vector components by plasmacytoid dendritic cells (pDCs), which sense the vector DNA genome via Toll-like receptor 9. Adaptive immune responses employ antigen presentation by conventional dendritic cells (cDCs), which leads to cross-priming of capsid-specific CD8⁺ T cells. In this study, we sought to determine the mechanisms that promote licensing of cDCs, which is requisite for CD8⁺ T cell activation. Blockage of type 1 interferon (T1 IFN) signaling by monoclonal antibody therapy prevented cross-priming. Furthermore, experiments in cell-type-restricted knockout mice showed a specific requirement for the receptor for T1 IFN (IFNaR) in cDCs. In contrast, natural killer (NK) cells are not needed, indicating a direct rather than indirect effect of T1 IFN on cDCs. In addition, co-stimulation by CD4⁺ T cells via CD40-CD40L was required for cross-priming, and blockage of co-stimulation but not of T1 IFN additionally reduced antibody formation against capsid. These mechanistic insights inform the development of targeted immune interventions.

Graphical Abstract

Immune responses complicate the use of adeno-associated virus (AAV) vectors in human gene therapy. Shirley et al. define a mechanism by which cross-presentation of viral capsid results in activation of CD8⁺ T cells, which involves sensing of IFN I by conventional dendritic cells and co-stimulation by CD4⁺ T helper cells.

Introduction

Gene therapy using adeno-associated virus (AAV) has emerged as one of the most promising treatments for a wide range of monogenetic disorders. AAV vectors are based on a small single-stranded DNA virus comprised of a viral capsid containing a genome of <5 kb. A large diversity of capsids, both natural and engineered, with distinct tropism and the ability to transfer genes into non-dividing cells in vivo make AAV vectors ideal for many different gene therapy applications. There are now two AAV-based gene therapies that have been approved by the US Food and Drug Administration (FDA) for the treatment of Leber’s congenital amaurosis (LCA) and spinal muscular atrophy (SMA).¹ Several AAV gene therapies for the X-linked bleeding disorders hemophilia A and B by hepatic gene transfer are in phase III clinical trials.² Muscle-directed and CNS gene therapy for muscle, neuromuscular, and storage disorders are advancing, among other disease and organ targets.³

Despite these recent successes, adaptive immune responses directed against the AAV capsid remain a significant challenge and continue to limit the utility of this intervention.⁴ CD8⁺ T cell responses against capsid can cause immunotoxicities, tissue inflammation, and loss of transgene expression unless immune suppression is applied.5, 6, 7, 8, 9, 10, 11 Cells transduced with AAV vectors may become targets for CD8⁺ T cells because they cross-present capsid antigen by major histocompatibility complex (MHC) I as a result of proteasomal degradation of viral particles following endosomal escape, capsid phosphorylation, and ubiquitination.12, 13, 14, 15, 16 Pre-existing neutralizing antibodies (NAbs) stemming from natural infection exclude some patients from receiving a particular vector, and NAb formation following gene transfer complicates re-administration. Critical to achieving long-term gene therapy is a deeper understanding of the mechanisms that lead to adaptive immune responses to AAV capsid.

CD8⁺ T cells or cytotoxic T lymphocytes (CTLs) are the principal effector cells that carry out antigen-specific cell-mediated killing of infected host cells. In the naive state, CD8⁺ T cells sample peptide MHC I complexes on the surface of dendritic cells (DCs) that acquired antigen. Recognition of cognate peptide in the context of MHC I (signal 1) by CD8⁺ T cells initiates a cellular program that either results in the induction of tolerance or the activation and proliferative expansion of antigen-specific CTLs. The decision between these two fates hinges on the activation state of the antigen-bearing DC, which, if properly conditioned, can provide necessary costimulation (signal 2) and inflammatory cytokines (signal 3). DC activation is modulated by both DC-intrinsic and/or DC-extrinsic “danger signals” such as pathogen recognition through pattern recognition receptors or inflammatory cytokines, respectively. In addition, CD40 signaling in DCs that occurs through interaction with CD40L on CD4⁺ T cell helper cells represents a potent activation or “licensing” signal that can permit effective CD8⁺ T cell responses even in the absence of adequate danger signals.17, 18, 19 Thus, DCs represent the critical bridge that links innate sensing to adaptive responses, and the circumstances of their activation can tip the balance between tolerance and immunity.^20,21

Innate immune sensing of the AAV genome by the endosomal DNA receptor TLR9 (Toll-like receptor 9) has been recognized as a critical determinant for CD8⁺ T cell activation against both capsid and transgene products.22, 23, 24, 25, 26 TLR9 is highly expressed in plasmacytoid DCs (pDCs), which in response to AAV produce type 1 interferon (T1 IFN, i.e., IFNα and IFNβ). We previously established that cross-priming of capsid-specific CD8⁺ T cells requires cooperation between pDCs, which sense the AAV genome via TLR9, and conventional DCs (cDCs), which perform the actual capsid antigen presentation.²⁵ TLR9 was dispensable in cDCs and rather was specifically required in pDCs, which therefore supply the innate signal that triggers T cell activation. For example, the ability to generate a CD8⁺ T cell response to AAV capsid can be restored in TLR9-deficient mice by adoptive transfer of wild-type (WT) pDCs. Furthermore, cross-priming requires T1 IFN, albeit its exact role has not yet been defined.²⁵

In this new study, we dissect the mechanism by which cDCs are conditioned to cross-prime capsid-specific CD8⁺ T cells following treatment with AAV gene therapy vectors. We find an overall requirement for T1 IFN that augments CD8⁺ T cell priming through a mechanism that is dependent on direct signaling in cDCs, while indirect activation of natural killer (NK) cells is not involved. Furthermore, help by CD4⁺ T cells and CD40-CD40L is critical for both CD8⁺ T cell responses and antibody responses directed against the capsid. In contrast to T help, T1 IFN has little impact on antibody formation against capsid. Thus, stimulation by T1 IFN and CD4⁺ T cells are both necessary while neither alone is sufficient for cross-priming anti-capsid CD8⁺ T cells. We demonstrate that both T1 IFN and co-stimulation signaling pathways represent potential targets for therapeutic intervention to prevent the CD8⁺ T cell response.

Results

T1 IFN Signaling Is Required for Anti-capsid CD8⁺ T Cell Priming, as Judged by αIFNAR-1 Treatment

We have previously shown that T1 IFNs are important for cross-presentation of AAV capsid antigen and for activation of anti-capsid CD8⁺ T cells.²⁵ We hypothesized that blockade of the T1 IFN receptor (IFNAR) may be a potential avenue for a therapeutic intervention to prevent this response. To test this, WT C57BL/6 mice were treated intraperitoneally (i.p.) with either 1 mg αIFNAR-1 (MAR1-5A3), a monoclonal antibody (mAb) known to block IFNAR1, or 1 mg of an isotype control (MOPC-21) 1 day prior to intramuscular (i.m.) injection into the quadriceps with 1 × 10¹¹ viral genomes (vg) of AAV2-SIIN (Figure 1A). AAV2-SIIN is a modified AAV2 capsid containing the SIINFEKL peptide sequence (AAV2-SIIN), which is the immunodominant CD8⁺ T cell epitope of the model antigen ovalbumin (OVA) in C57BL/6 mice. This vector allows for anti-capsid CD8⁺ T cells to be quantified in peripheral blood over time by flow cytometry using an H-2K^b SIINFEKL tetramer.²⁵ As shown in Figures 1B and 1C, mice treated with αIFNAR had significantly impaired anti-capsid CD8⁺ T cell responses compared to isotype-treated controls at early time points. Moreover, AAV capsid-specific CD8⁺ T cell counts never resurged during more than 4 weeks later, suggesting that a single dose given immediately before administration of AAV is sufficient to prevent anti-capsid CTL formation (Figure 1C). We conclude that T1 IFN blockade prior to AAV immunization abrogates anti-capsid CD8⁺ T cell formation, which underscores the importance of this signaling pathway during the early stages of T cell priming.

Figure 1.

T1 IFN Blockade Prevents Activation of AAV Capsid-Specific CD8⁺ T Cells

WT C57BL/6 mice were i.p. injected with 1 mg of a non-depleting IFNAR1 blocking antibody MAR1-5A3 (n = 7) or an isotype control antibody MOPC-21 (n = 6) 1 day prior to i.m. injection with 1 × 10¹¹ vg of AAV2-SIIN. Peripheral blood was collected over time, and SIINFEKL-specific CD8⁺ T cells were quantified by flow cytometry. (A) Experiment schematic with timeline. (B) Representative flow plots of H-2K^b-SIIN tetramer staining of AAV capsid-specific CD8⁺ T cells 7 days after AAV2-SIIN injection. (C) Time course of the anti-capsid CD8⁺ T cell response reported as a percent tetramer⁺ of live CD3⁺CD8⁺ T cells. The dotted line at 0.08% represents the limit of detection of capsid-specific CD8⁺ T cells. Data points represent mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant.

Depletion of NK Cells Does Not Affect the Generation or Cytotoxic Capacity of AAV-Capsid-Specific CTLs

We hypothesized that AAV-elicited T1 IFNs served as a key signal for cDCs to become licensed to prime anti-capsid CD8⁺ T cells. However, it remains unclear whether AAV-elicited T1 IFN-initiated signaling in cDCs is sufficient for licensing and concomitant CD8⁺ T cell activation. Alternatively, it is plausible that T1 IFNs promote cDC licensing by indirectly eliciting inflammatory cytokine production by accessory cells, e.g., NK cells. As shown elsewhere, T1 IFNs potently induce NK cells to produce inflammatory cytokines such as interleukin (IL)-12, IL-15, and IFNγ, which could enhance cDC activation status and promote the development of effector functions in CD8⁺ T cells.^27,28 We therefore evaluated whether NK cells were required for anti-capsid CD8⁺ T cell responses to AAV. To this end, WT C57BL/6 mice were in vivo depleted of NK cells by i.p. injection with a mAb αNK1.1 (PK136), 1 day prior to AAV2-SIIN injection (Figure 2A). NK cell depletion was confirmed by flow cytometry on day 1 and day 7 after αNK1.1 injection (Figure S1). We observed no significant differences in tetramer⁺ CD8⁺ T cells in the blood at day 7 and day 10 or in the spleen or lymph node (LN) at day 11 relative to isotype controls (n = 5/group), suggesting that NK cells are dispensable for CD8⁺ T cell priming (Figure 2B). Although this result suggests that NK cells are not strictly required for anti-capsid CD8⁺ T cell responses in terms of magnitude, the contribution of accessory cytokines produced by activated NK cells may contribute to the development of CTL effector functions. To test this, we performed an in vivo killing assay based on adoptive transfer of SIIN-pulsed target cells following AAV2-SIIN immunization (Figure 2C). We found that the percentage killing of target cells was independent of NK cell depletion, implying that anti-capsid CD8⁺ T cells primed in the absence of NK cells were fully functional (Figure 2D). Taken together, both anti-capsid CD8⁺ T cell accumulation and the development of effector functions are independent of NK cells. We therefore ruled out that T1 IFN signaling facilitated cDC licensing indirectly through NK cell activation.

Figure 2.

AAV Capsid-Specific CTL Formation and Function Is Independent of NK Cells

WT C57BL/6 mice were injected i.p. with 200 μg of either NK1.1 (PK136) antibody known to deplete NK cells (n = 5) or isotype control (C1.18.4) antibody (n = 5) 1 day prior to i.m. injection with 1 × 10¹¹ vg of AAV2-SIIN. (A) Schematic with timeline for NK cell depletion and AAV capsid-specific CTL quantification. (B) Quantification of anti-capsid CD8⁺ T cell response as percent tetramer⁺ of live/CD3⁺/CD8α⁺ at day 7 and day 10 in peripheral blood (top) and at day 11 in the spleen (bottom left) and pooled inguinal and popliteal LNs (bottom right). (C) Experiment schematic for the in vivo killing assay; an additional negative control group (n = 5) received no AAV. On day 7 after AAV2-SIIN, all groups were adoptively transferred with a 1:1 mixture of 1 × 10⁶ splenocytes pulsed with SIINFEKL peptide and 1 × 10⁶ splenocytes not pulsed with peptide. Adoptively transferred cells were differentially labeled with CTV to distinguish SIIN-pulsed target cells (3.0 μM, CTV^hi) from not-pulsed control cells (0.3 μM, CTV^lo). (D) Representative histograms showing preferential killing of SIINFEKL-pulsed, CTV^hi splenocytes in mice treated with AAV2-SIIN regardless of NK cell depletion (left) and calculation of % killing (right). Data points represent mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant.

Selective Deletion of IFNAR1 in cDCs Demonstrates that T1 IFN Signaling in cDCs Is Directly Involved in Anti-capsid CD8⁺ T Cell Priming

Whether T1 IFNs impacted anti-capsid CD8⁺ T cell activation by directly signaling in cDCs was evaluated by generating and employing a mutant mouse in which IFNAR1 was selectively deleted in cDCs. The mutant mouse was generated by crossing a IFNAR1^fl/fl mouse with a CD11c-cre-EGFP mouse, which expresses cre-recombinase under the control of the CD11c promoter (Figure 3A).²⁹ Because CD11c is expressed by both cDCs and pDCs (highly expressed by cDCs while low to intermediately expressed by pDCs), we tested whether the resultant IFNAR1^fl/fl CD11c-Cre(+) mice bear the mutation primarily in cDCs. As shown in Figure 3B, Cre recombinase expression was assessed by flow cytometric analysis of the GFP reporter element located downstream of Cre recombinase. These analyses revealed a robust increase of GFP median fluorescence intensity (MFI) in Cre(+) cDC populations compared to Cre(−) controls, while no significant increase was seen in pDCs or the lineage (Lin)⁺ control population. Thus, cre expression was limited to CD11c^hi cDCs.

Figure 3.

cDC-Intrinsic T1 IFN Signaling Licenses cDCs to Prime Anti-capsid CD8⁺ T Cells but Is Not Sufficient for Cross-Presentation

(A) Transgenic mice specifically lacking IFNAR1 expression in cDCs (cDC-IFNAR^−/−) were generated by crossing mice expressing Cre recombinase under the control of a CD11c promoter (CD11c-Cre-IRES-GFP) with mice containing an IFNAR1 gene flanked by loxP sites (IFNAR^fl/fl). The specificity of Cre expression and activity was assessed in both cDCs (lineage [Lin]⁻, CD11c^hi, PDCA-1⁻) and pDCs (Lin⁻, CD11c^int, PDCA-1^hi) isolated by fluorescence-activated cell sorting (FACS) from CD11c-IFNAR^+/+ (n = 5) and CD11c-IFNAR^−/− (n = 5) mice. (B) Median fluorescence intensity (MFI) of GFP in cDCs, pDCs, and Lin⁻ (CD3, CD19, NK1.1) populations (left) and representative histograms (right). (C) Normalized copy number (2^−ΔΔCT) of IFNAR-1 exon 3 (within the LoxP-flanked region) relative to exon 7 (downstream of LoxP-flanked region) was determined by qPCR from genomic DNA isolated from either cDCs or pDCs. (D) Time course of anti-capsid CD8⁺ T cell responses in peripheral blood collected from cDC-IFNAR^+/+ (n = 6) and cDC-IFNAR^−/− (n = 7) mice treated with 1 × 10¹¹ vg of AAV2-SIIN i.m. The dotted line at 0.08% represents the limit of detection. (E) (Top) Schematic and timeline of an in vivo proliferation assay in cDC-IFNAR^+/+ and cDC-IFNAR^−/− mice (n = 6/group) treated with 1 × 10¹¹ vg of AAV2-SIIN i.m. or cDC-IFNAR^+/+ mice that received no AAV (n = 6/group). One day later, CD8⁺ T cells magnetically enriched from OT-I transgenic mice were labeled with CTV and adoptively transferred via tail vein injection at 1 × 10⁶ cells/mouse. Spleens were collected 4 days later, and proliferation was assessed based on dye dilution of CTV-labeled OT-I cells determined by flow cytometry. (Middle) Representative histograms with markers for % proliferation of CTV-labeled OT-1 CD8⁺ T cells in cDC-IFNAR^+/+ or cDC-IFNAR^−/− mice treated with AAV2 or AAV2-SIIN. (Bottom) Quantification of % proliferation (proportion of CTV-labeled cells that have divided), % divided (percent of original cells that divided at least once), and the division index (the average number of cells that a dividing cell became). Data points represent mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant.

We further verified that cre-driven recombination of IFNAR^fl/fl to IFNAR^−/− was also limited to cDCs. The relative IFNAR1 gc number in flow-sorted pDC and cDC populations was determined by qPCR. A region of exon 3 of IFNAR1, which is within the flanked LoxP region, was selected to reflect that the IFNAR null allele and copy numbers were normalized to exon 7, which is downstream of the excision site.²⁹ As expected, the relative IFNAR1 exon 3 copy number was significantly decreased in the cDCs of IFNAR^fl/fl-Cre(+) mice compared to IFNAR^fl/fl-Cre(−) mice. In contrast, no differences were observed among pDC populations from both mouse strains (Figure 3C). Taken together, the data demonstrate that cre-mediated excision of IFNAR1 is restricted to CD11c^hi cDCs in our IFNAR1^fl/fl × CD11c-Cre(+) (cDC-IFNAR^−/−) mice.

Using this well-defined model for cDC-specific IFNAR knockout (KO), we next assessed whether cDC-intrinsic IFNAR signaling is requisite for AAV capsid-specific CD8⁺ T cell responses. To this end, cDC-IFNAR^−/− or cDC-IFNAR^+/+ control mice were injected with 1 × 10¹¹ vg of AAV2-SIIN, and anti-capsid CTL formation was evaluated in peripheral blood over time. Following AAV administration, cDC-IFNAR^−/− mice had a markedly reduced proportion of anti-capsid CD8⁺ T cells on day 7 post-infection (p.i.), as compared to cDC-IFNAR^+/+ control mice. The outcomes of these experiments indicated that T1 IFNs augment AAV capsid-specific CD8⁺ T cell responses through a mechanism that involves direct sensing by cDCs (Figure 3D).

We have previously shown that cDCs are the primary antigen-presenting cells (APCs) that promote CD8⁺ T cell formation and that blockade of IFNAR1 substantially reduces cross-presentation of AAV capsid antigen.²⁵ Hence, we evaluated whether the observed reduction in anti-capsid CD8⁺ T formation was due to a reduced capacity of IFNAR^−/− cDCs to cross-present antigen using an in vivo proliferation assay. Cross-presentation of AAV capsid antigen was determined in cDC-IFNAR^−/− mice using an indirect assay based on the proliferation of fluorescently labeled, SIIN-specific CD8⁺ T cells from OT-I transgenic mice that were adoptively transferred 1 day after AAV2-SIIN immunization (Figure 3E). No proliferation above background was observed when cDC-IFNAR^+/+ mice received unmodified AAV2, confirming that OT-I proliferation is dependent on the SIINFEKL epitope (Figure 3E, middle panel). Compared to cDC-IFNAR^+/+, we observed a modest reduction in proliferation in cDC-IFNAR^−/− mice. While the reduction in % proliferation (percent of total proliferating cells) or % divided (percent of original cells that divided at least once) did not reach statistical significance, there was a significant decrease in the division index (the average number of cells a dividing cell became) in cDC-IFNAR^−/−, suggesting that cDCs in these mice had more limited capacity to support multiple rounds of division (Figure 3E, bottom panel). These findings suggest that cDC-intrinsic T1-IFN signaling is more strictly required for cDC activation to facilitate efficient priming of anti-capsid CD8⁺ T cells than for mere presentation of AAV capsid antigen.

Anti-capsid CTL Responses Require CD4⁺ T Cell Help and CD40L-CD40

Although cDCs clearly require T1 IFN signaling in order to prime AAV capsid-specific CD8⁺ T cells, it is unclear whether this inflammatory stimulus alone is sufficient. The concept of DC licensing to support proper priming of CD8⁺ T cell responses was originally described to occur through CD40-CD40L interaction with CD4⁺ T helper cells.³⁰ Of note, this pathway is differentially required for cDC licensing to prime anti-viral CTL responses depending on the degree of inflammation elicited by the pathogen. That is, T help tends to be overtly required for noninflammatory agents but can be dispensable when a pathogen elicits an adequate amount of innate activation either through damaging cells or triggering innate receptors.³¹ Thus, we tested for a dependency for T help to prime anti-capsid CD8⁺ T cells in response to AAV. First, to assess whether there was an overall requirement for CD4⁺ T help for anti-capsid CTL formation, we treated MHC II^−/− mice, which entirely lack CD4⁺ T cells, with AAV2-SIIN and found that anti-capsid CD8⁺ T cell responses were substantially diminished in these mice (Figure 4A). It is known that CD8⁺ T cell differentiation into long-lived memory cells also depends on T help.³² Consistent with this notion, we were unable to boost CTL responses upon re-challenge with an OVA-expressing vector (Figure S2), suggesting that anti-capsid CD8⁺ T cell memory formation is also significantly impaired in the absence of T help.

Figure 4.

Anti-capsid CD8⁺ T Cell Priming Depends on T Help and CD40-CD40L

(A) Schematic with timeline (left). WT C57BL/6 (n = 5) and MHC II^−/− (n = 7) mice, which are deficient for CD4⁺ T cells, were treated i.m. with 1 × 10¹¹ vg of AAV2-SIIN, and anti-capsid CD8⁺ T cells were quantified in peripheral blood over time by tetramer stain (right). (B) Schematic with timeline showing that WT C57BL/6 mice were i.p. injected with 250 μg of a non-depleting CD40L blocking antibody MR1 (n = 5) or an isotype control polyclonal Armenian hamster IgG (n = 5) 1 day prior to i.m. injection with 1 × 10¹¹ vg of AAV2-SIIN. Peripheral blood was collected over time and SIINFEKL-specific CD8⁺ T cells were quantified by flow cytometry. The dotted line at 0.08% represents the limit of detection of capsid-specific CD8⁺ T cells. Data points represent mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant.

Next, we investigated whether the required T help was mediated through CD40-CD40L, which may represent a viable target for therapeutic immunosuppression. WT C57BL/6 mice were treated with either an αCD40L blocking antibody or an isotype control 1 day prior to AAV2-SIIN (Figure 4B). Anti-capsid CD8⁺ T cells in αCD40L-treated mice were robustly reduced on day 8 compared to isotype controls, with four of five mice having undetectable responses (n = 5/group). Taken together, our experiments support that help from CD4⁺ T cells via CD40L/CD40 signaling plays a substantial role in cross-priming of CD8⁺ T cells against AAV capsid.

T Help-Dependent Anti-capsid Antibody Responses Do Not Require T1 IFN

Previously, we found that mechanisms leading to B and T cell activation against AAV capsid may be distinct.²⁴ Except for MyD88-deficient mice, antibodies against AAV2 capsid are primarily immunoglobulin (Ig)G2c subtype in C57BL/6 mice. We found no differences in anti-AAV2 IgG2c titers as measured by ELISA in αIFNAR-treated mice (Figure 5A) or in cDC-IFNAR^−/− mice (Figure 5B) compared to their respective controls. Similarly, we found no differences in anti-capsid antibody responses as measured by NAb assay in cDC-IFNAR^−/− mice (Figure 5B). These results indicated that the T1 IFN pathway is dispensable for antibody responses directed against the capsid, which is consistent with previous findings. However, there was significant reduction in IgG2c formation in mice treated with αCD40L and on average a 27-fold reduction of NAb titers (Figure 5D), indicating a requirement for T help for the B cell response. As expected, MHC II^−/− mice failed to form antibodies (Figure 5C). With further optimization, CD40/CD40L blockade could therefore be a tool to inhibit both cell-mediated and humoral responses against the AAV capsid and permit vector re-administration.

Figure 5.

AAV Capsid-Specific Antibody Responses Occur Independently of T1 IFN Signaling but Require T Help and CD40-CD40L

Anti-capsid IgG2c antibody titers were evaluated by ELISA and NAb assay in blood plasma collected from mice treated with AAV2-SIIN. (A) Time course of anti-capsid IgG2c titers in mice that received either an αIFNAR blocking antibody or an isotype control 1 day prior to AAV2-SIIN (described in Figure 1). (B) IgG2c titers (left) and day 28 NAb titers for cDC-IFNAR^+/+ or cDC-IFNAR^−/− mice (described in Figure 2). (C) IgG2c titers in MHC II^−/− versus WT mice. (D) IgG2c titers and day 21 NAb titers in mice pretreated with αCD40L 1 day prior to AAV2-SIIN (described in Figure 4). Data points represent mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant.

AAV Triggers a Heterologous Activation Phenotype in Multiple Innate Immune Subsets

After having identified multiple components of the mechanism that licenses cDCs to activate AAV capsid-specific CD8⁺ T cells, we asked whether AAV elicits a signature immunophenotype on cDCs and other innate immune cells that can be measured directly using high-throughput multi-parameter flow cytometry. In an effort to more clearly define an immunophenotype on DC subsets and myeloid populations in response to AAV, we designed an eight-color flow cytometry panel to separate pDCs, cDCs, granulocytes, and monocyte/macrophage (“mono/mac”) populations in both the spleen (Figure S3) and the inguinal LN (ingLN) (Figure S4).

WT C57BL6 mice were treated bilaterally into the quadriceps muscle with either PBS, 1 × 10¹¹ vg of AAV2-SIIN, or OVA mixed with an TLR9 agonist (a CpG-rich oligodeoxynucleotide, “CpG + OVA” group). Spleens and ingLNs were harvested 48 h p.i., a time point that was selected based on preliminary results (data not shown). Results are reported as the MFI of the indicated surface marker in each figure. We also utilized an H-2K^b-SIINKEL antibody (25-D1.16), which binds to SIIN peptide in the context of MHC I to obtain direct evidence for AAV capsid-derived antigen by MHC I. With this antibody, we found elevated surface staining on pDCs, cDCs, and mono/mac populations that was significantly different from PBS-treated controls in cDCs (Figure 6A). Furthermore, surface expression of PDCA-1 was significantly increased in splenic cDC and granulocyte populations as well as mono/mac populations in the ingLN, found in three separate experiments (Figure 6B; Figure S5; data not shown). PDCA-1 is constitutively highly expressed on pDCs but is induced in other cell types by T1 IFNs.^33,34 Thus, heightened detection of PDCA-1 48 h after treatment with AAV provides direct evidence that multiple innate immune cell types are responding to AAV-elicited T1 IFNs.

Figure 6.

Antigen Presentation of AAV Capsid by cDCs and Altered Surface Expression of PDCA-1 and CD70

WT C57BL/6 mice were treated bilaterally into the quad with either PBS (n = 5), 1 × 10¹¹ vg of AAV2-SIIN (n = 5), or 50 μg of CpG B + 10 μg of OVA (n = 3). Spleens were harvested 48 h later and the surface phenotype of innate immune cells was evaluated by flow cytometry. Innate immune subsets were identified using the gating strategy described in Figure S3, and surface expression was determined by median fluorescence intensity (MFI) in each subset for (A) H-2K^b-SIIN, (B) PDCA-1, (C) CD40, and (D) CD70. Data points represent mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. ns, not significant.

We similarly evaluated whether a set of costimulatory molecules from both the B7 family (ICOSL) and/or the TNFRS (CD40, CD70, and OX40L) would be upregulated as a consequence of AAV administration. We observed no differences in CD40 expression on any population in the spleen or ingLN despite this pathway being required for CD8⁺ T cell priming (Figure 6C; Figure S5C). Interestingly, CD70 was upregulated on mono/mac populations but not cDCs in both spleen and ingLN (Figure 6D; Figure S5D). This finding raises the possibility that non-DC subsets of innate cells may also be participating in the activation of adaptive immune responses to AAV by providing costimulatory ligands such as CD70. Finally, we did not find any changes in surface expression of OX40L but did observe a decrease in ICOSL on both splenic and ingLN APCs (data not shown).

Discussion

In addition to recent marketing approvals by the FDA, a broad range of new AAV gene therapies are under development and will soon be brought to the clinic. However, immune responses to vector and cargo continue to pose significant obstacles. Among others, these include CD8⁺ T cell responses and antibody responses against viral capsid proteins.⁴ Work by us and other laboratories has established a link between innate immune sensing of the vector’s DNA genome via TLR9 and cross-priming of capsid-specific CD8⁺ T cells, which requires cooperation between pDCs (which sense AAV via TLR9) and cDCs (which cross-present capsid antigen), as well as T1 IFN.22, 23, 24, 25, 26^,35

Licensing of DCs is required to provide all of the necessary signals to activate CD8⁺ T cells and is essential for the key handoff between the innate and adaptive immune system in the early stages of priming.³⁶ In other words, DCs need to receive an adequate amount of environmental cues in order to provide all of the activation signals needed to prime a CD8⁺ T cell response. Originally, DC licensing by CD4⁺ T cells via CD40-CD40L was described as the mechanism that CD4⁺ T cells delivered “help” to CD8⁺ T cell responses and was considered compulsory for the development of effector functions and memory differentiation.37, 38, 39 It has since been recognized that T help is not always an explicit requirement for CD8⁺ T cell priming and that direct pattern recognition receptor signaling or high levels of T1 IFN may license DCs to activate CTLs without a need for T help.^31,40, 41, 42, 43 The presence of T1 IFNs could promote DC licensing through a mechanism that involves direct signaling in DCs and/or indirectly stimulating inflammatory cytokine production by NK cells. In this study, we show that cDC licensing to prime AAV capsid-specific CD8⁺ T cells occurs independently of NK cells. However, NK cells have been observed as part of the innate immune infiltrates in tissues early after vector administration, and our data do not rule out the possibility that NK cells may still contribute to local inflammation and loss of transduced cells.²³ Instead, cDC licensing in response to AAV requires cDC-intrinsic T1 IFN signaling, and T help through CD40-CD40L co-stimulation is additionally required to cross-prime the CD8⁺ T cells. Both mechanisms are required to accomplish cross-priming, implying that neither alone is sufficient. Our data do not differentiate between a requirement for T help in cDC licensing or subsequent events that lead to the expansion of active CD8⁺ T cells. However, AAV vectors elicit only weak to modest innate responses compared to other viruses, making it unlikely that an overwhelming cytokine response occurs that activates cDCs. Others have reported that, subsequent to initial, spatially separate interactions between DCs and CD4⁺ and CD8⁺ T cells, cross-presenting cDCs interact with both T cell subsets during initiation of the anti-viral response.^44,45 Taken together, it is likely that CD4⁺ T help contributes to both activation of DCs and aiding in expansion of CD8⁺ T cells during cross-priming. Finally, we demonstrate that CD4⁺ T cell help through CD40-CD40L contributes to anti-capsid antibody responses in contrast to the T1 IFN pathway, which is dispensable for these responses (Figure 7).

Figure 7.

Linking Innate and Adaptive Immune Responses to AAV Gene Therapy

Activation of deleterious adaptive immune responses directed against the AAV capsid involves the cooperation of multiple immune cell types, surface receptors, and soluble mediators. First, AAV particles are taken up by DCs that are sampling their environment for invading pathogens. The CpG motifs present in the DNA component of AAV gene therapy vectors trigger the innate receptor, TLR9, within the endosome of pDCs. Consequently, pDCs produce T1 IFNs that signal directly in cDCs. This signaling event is required for effective priming and enhances the ability of the DC to activate T cells but only minimally effects the efficiency of cross-presentation of AAV capsid-derived antigens. T1 IFN-conditioned cDCs also interact with CD4⁺ T cells through CD40-CD40L, which may conribute to licensing the cDCs to activate AAV capsid-specific CD8⁺ T cells. Activated CD4⁺ T helper cells also promote antibody formation against AAV capsid.

A Direct Role for T1 IFN in Anti-capsid CD8⁺ T Cell Priming

T1 IFNs are highly pleiotropic cytokines that are considered key mediators of anti-viral immunity. The myriad of different immune cells that respond to T1 IFNs and the circumstantial effects they elicit make deciphering their particular role during the course of any antiviral immune response complex and context specific. Indeed, some controversy exists in the literature over how T1 IFNs shape the outcome of these responses.46, 47, 48 Here, we rule out an indirect role for T1 IFNs through the activation of NK cells and instead show that activation of AAV capsid-specific CD8⁺ T cell responses require T1 IFN signaling specifically in cDCs. In contrast to DCs, other APCs are less critical for cross-priming of CD8⁺ T cells, consistent with earlier in vitro data that pDCs are the cells that critically provide T1 IFN cytokines.^22,25 Therefore, T1 IFN production by pDCs in response to TLR9 activation promotes licensing of cDCs, which cross-present AAV capsid antigen via MHC I and ultimately prime CD8⁺ T cells. This process occurs via cDC-intrinsic T1 IFN signaling and, combined with co-stimulation by T helper cells, results in CD8⁺ T cell priming despite having little impact on AAV capsid presentation.

This discrepancy may be reconciled by a model described by Brewitz et al., ⁴⁹ who showed that early cross-presentation occurs in a T1 IFN-independent manner while subsequent expression of co-stimulation molecules in XCR1⁺ cDCs is T1 IFN-dependent. The cDC compartment can be further broken down into CD11b⁺ and XCR1⁺ cDC subsets. The latter represents the most capable subset of DCs for cross-presentation of antigen, making them also likely candidates for performing this task in the response to AAV. XCR1⁺ cDCs have also been shown to simultaneously present antigen to “pre-primed” naive CD4⁺ and CD8⁺ T cells, thereby providing a platform for the provision of T help to activating CD8⁺ T cells.⁴⁴ Importantly, expression of co-stimulation molecules on this DC subset is critical to facilitate the provision of T help to CD8⁺ T cells. In line with this notion, we show a dependency for CD4⁺ T cell help and CD40-CD40L for the activation of AAV capsid-specific CD8⁺ T cell responses. Taken together, a more general model of antiviral CD8⁺ T cell priming emerges that applies to diverse viruses such as AAV and vaccinia virus.

Targeted Immune Modulation

Current immune suppression regimens in clinical AAV gene therapy primarily employ broadly suppressive steroid drugs. We have identified multiple specific immune checkpoints that critically regulate the activation of adaptive immune responses directed against the AAV capsid. These early innate triggers are critical catalysts that govern the overall quality and magnitude of an immune response. Therefore, developing immunosuppression protocols that prevent or minimize innate activation of AAV gene therapy vectors has the potential to mitigate deleterious adaptive immune activation upstream of T and B cell activation. The provision of immunosuppressive drugs that specifically disrupt co-stimulation or IFN signaling pathways early after vector administration may circumvent activation of the adaptive immune system and the need for long-term systemic immunosuppression.

One approach is to target the T1 IFNs themselves with mAbs. This strategy has been demonstrated to be both safe and efficacious in patients with systemic lupus erythematosus (SLE) that, when treated, had substantial reductions in IFN-stimulated gene (ISG) expression.50, 51, 52 Clinical symptoms were improved in patients as well, depending on the severity of disease. Importantly, clinical efficacy was more pronounced in patients with low T1 IFN signatures following mAb therapy. Although this feature narrows the clinical reach of this approach in SLE, it is encouraging for its potential to mitigate T1 IFN responses to AAV gene therapy vectors that elicit a relatively mild IFN response.

Multiple biologics targeting the CD40-CD40L pathway are currently in clinical development for a variety of diseases, including lymphoproliferative cancers and autoimmune diseases. However, most clinically used anti-CD40 mAbs are cell depleting, which causes the recipient to become profoundly immunocompromised, a situation that is not ideal in gene therapy.⁵³ Early efforts to block CD40L with humanized mAbs (e.g., ruplizumab) were stymied by serious adverse events, including thrombosis and embolism.⁵⁴ It was later recognized that this outcome was due to the formation of immune complexes consisting of CD40L and anti-CD40L, which subsequently bound FcγRIIa on the surface of platelets and caused them to aggregate. Fc-silent domain antibodies have now been designed to circumvent FcγRIIa binding.⁵⁵ Two such antibodies, letolizumab and dapirolizumab pegol, have been shown to retain their immunomodulatory activity without inducing platelet activation and are now in clinical trials.^55,56 Furthermore, blockers of other co-stimulatory pathways such as CTLA4-Ig may be efficacious.

Expression of Co-stimulation on APCs

Herein we have shown that in vivo AAV administration induces phenotypic changes in discrete innate immune populations. These include increased PDCA-1 expression, increased pMHC complexes, and alterations in surface expression of multiple co-stimulatory molecules early after administration. PDCA-1 was upregulated on all examined populations in the splenic compartment and on monocytes/macrophages in ingLN. As PDCA-1 expression is indicative of T1 IFN signaling and/or production, this is direct evidence that T1 IFN signaling is occurring in innate cells as a result of AAV administration.⁵⁷ Surface expression of capsid-derived antigen in the context of MHC I (H-2K^b-SIIN) was primarily increased in the splenic cDC compartment. Taken together, these findings align with our model where cDCs are directly acted on by T1 IFNs and serve as the principal APCs in this system.

Interestingly, we revealed an increase in CD70 on monocytes/macrophages in both the spleen and ingLN, which suggests that this population may be an additional contributor to the establishment of the anti-capsid CD8⁺ T response in AAV gene therapy via the CD27/CD70 axis. CD27 is constitutively expressed on resting T cells, and signaling through this receptor via the provision of CD70 can induce the expression of OX40 and 4-1BB.^58,59 The complexity of the temporal regulation of these molecules needs to be further explored in future studies.⁶⁰ For example, we found that the ligand for 4-1BB was significantly upregulated in cDCs 48 h after vector administration in ingLNs, although this was not consistently the case in our experiments (data not shown), illustrating the challenging nature of such studies (which may also be influenced by serotype and strain background). Furthermore, we observed decreases in ICOSL expression on cDCs and monocytes/macrophages in the spleen. Engagement of ICOS on T cells by ICOSL skews T cells toward T helper 2 (Th2) and T follicular helper (Tfh) lineages.⁶¹ Therefore, reduced expression of this ligand may be evidence for a cytokine microenvironment favorable for Th1 skewing and effective CD8⁺ T cell priming. Although no differences were observed in CD40 surface levels, our results nonetheless indicate that this co-stimulatory molecule (which is constitutively highly expressed on cDCs) is critical for anti-capsid CD8⁺ T cell responses.

Implications for Human Gene Therapy

Cross-priming of capsid-specific CD8⁺ T cells has been observed in multiple clinical trials in humans. Both in mice and humans, CD8⁺ T cell activation against capsid is vector dose-dependent. As in mice, human pDCs produce T1 IFN in response to AAV in a TLR9-dependent manner.^11,22 Steroid drugs used clinically to counter the T cell response against AAV may in part act through suppression of pDCs in humans.⁶² Root cause analysis of loss of FIX expression in one clinical trial determined that codon optimization had inadvertently introduced more CpG motifs, suggesting that elimination of TLR9-stimulating CpG sequences from AAV vectors may improve the outcome of hepatic gene transfer for hemophilia.^26,63 These aspects support the relevance of our findings for human gene therapy. At the same time, it remains unclear why CD8⁺ T cells to AAV capsid fail to clear transduced cells in animals (despite their cytolytic activity) and why the time course of T cell activation differs. These differences may relate to levels or kinetics of antigen presentation. We have further shown that immune checkpoints can cause delayed CD8⁺ T cell responses against AAV-transduced liver.⁶⁴ In human skeletal muscle, capsid-specific CD8⁺ T cells contributed to inflammation but failed to eliminate transgene expression, suggesting additional effects of the target tissue.⁶⁵

In conclusion, pathways critical to the immune response to the AAV vector could be exploited in the development of targeted intervention to prevent cross-priming of capsid-specific CD8⁺ T cells. Some of these may also apply to CD8⁺ T cell responses against the transgene product, which were also found to be TLR9-dependent.^24,26 There are overlapping but also distinct requirements for B and T cell activation, so that for instance blockage of T1 IFN signaling prevents CD8⁺ T cell but not B cell activation against capsid. Activation of monocyte-derived DCs rather than pDCs may be more critical for the B cell response.^66,67 Our findings contribute to a more complete understanding of basic immunological mechanisms underlying deleterious immune responses to AAV gene therapy and provide insight toward the development of targeted immune suppression protocols and rational vector design in order to achieve long-lasting gene therapy in patients.

Materials and Methods

Mouse Strains

WT C57BL/6J, MHC II^−/− (B6.129S2-H2^dlAb1-Ea/J), OT-I (Tg(TcraTcrb)1100Mjb), IFNAR^fl/fl (B6(Cg)-Ifnar1^tm1.1Ees/J), and CD11c-cre-GFP (C57BL/6J-Tg(Itgax-cre,-EGFP)4097Ach/J) mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). OT-I, IFNAR^fl/fl, and CD11c-cre-GFP strains were subsequently bred in-house at the University of Florida. Generation of cDC-specific IFNAR KO mice (CD11c-IFNAR^−/−) was achieved by crossing CD11c-cre-GFP mice with IFNAR^fl/fl mice. F₁ offspring were backcrossed with F₀ IFNAR^fl/fl mice to select for mice homozygous for IFNAR^fl/fl and hemizygous for CD11c-cre-GFP. Details about mouse genotyping and validation of OT-I and CD11c-IFNAR^−/− mouse strains can be found in Supplemental Materials and Methods. Animals were housed under specific pathogen-free conditions at the University of Florida and treated under Institutional Animal Care and Use Committee-approved protocols. All mice were 6–12 weeks old at the start of each experiment, and the specific number of mice in each cohort is indicated in each figure with a minimum of four mice per group.

AAV Vectors

The OVA-derived SIINFEKL epitope was cloned into the AAV2 HI loop. A self-complementary AAV vector genome expressing GFP (scCB-GFP) was packaged into this AAV2-SIIN capsid by triple transfection in HEK293 cells.²⁵ Vector particles were purified by iodixanol gradient centrifugation, and titers were determined by dot blot hybridization and confirmed by western blot using a reference standard of known titer for comparison.⁶⁸

Animal Procedures

AAV vectors were administered i.m. into the quad diluted to 50 μL per injection, and blood samples were collected by retro-orbital bleed into heparinized capillary tubes.²⁵ All mAbs and isotype controls used for in vivo for depletion or pathway inhibition studies, including αIFNAR-1 (MAR1-5A3), αCD40L (MR-1), and αNK1.1 (PK136), were purchased from Bio X Cell (West Lebanon, NH, USA) and administered by i.p. injection at 1 mg, 250 μg, and 200 μg, respectively.

Flow Cytometry

Single-cell suspensions of PBMCs, splenocytes, or LNs were prepared as published.²⁵ Briefly, cells were mechanically disrupted, red blood cells (RBCs) were lysed (1× RBC lysis buffer, eBioscience), and cells were treated with FcR Block, TruStain FcX anti-mouse CD16/32 (93, BioLegend, San Diego, CA, USA) and stained for viability with a Fixable Live/Dead kit (Thermo Fisher Scientific, Waltham, MA). For analysis of DC populations, tissues were also pretreated with collagenase D (Roche, Basel, Switzerland) at 2 μg/mL for 20 min at 37°C to facilitate release of DCs and improve DC yield.

The antibody panel to assess AAV capsid-specific CD8⁺ T cells included CD3 (17A2), CD8α (53-6.7), and iTAg MHC tetramer (H-2K^b-SIINFEKL, MBL International, Woburn, MA, USA) bound to the fluorophore BV421. Importantly, CD8α (53-6.7) antibody was carefully titrated to avoid nonspecific binding due to interaction with the MHC I tetramer. An antibody panel to assess the DC phenotype included SIINFEKL/H-2K^b (25-D1.16), CD11c (N418), CD3 (17A2), NK1.1 (PK136), CD19 (6D5), CD11b (M1/70), CD8α (53-6.7), MHC II (I-A/I-E) (M5/114.15), PDCA-1 (927), CD40 (323), 4-1BBL (TKS-1), CD70 (FR70), and streptavadin-phycoerythrin (PE). SIINFEKL/H-2K^b complexes on cell surfaces were detected with 25D1.16 antibody bound to PE (BioLegend, San Diego, CA, USA). All antibodies were purchased from BioLegend unless otherwise indicated. Cells were stained for 20–30 min at 4°C or on ice except for the tetramer, which was stained at 25°C per the manufacturer’s recommendations. Cells were washed and fixed with intracellular (IC) fixation buffer (Invitrogen, Carlsbad, CA, USA). Data were collected no more than 48 h after staining on an LSR II Fortessa (BD Biosciences, Franklin Lakes, NJ, USA) and were analyzed with FCS Express (De Novo Software, Glendale, CA, USA).

Cell Sorting and Adoptive Transfers

For cDC and pDC sorting, splenocytes from cDC-IFNAR^+/+ or cDC-IFNAR^−/− mice were labeled with CD11c (N418), PDCA-1 (927), CD3 (17A2), CD19 (6D5), NK1.1 (PK136), and Live/Dead fixable near-IR stain (Thermo Fisher Scientific, Waltham, MA, USA), and cells were sorted on the FACSAria II (BD Biosciences, Franklin Lakes, NJ, USA). Genomic DNA was isolated using a DNeasy kit (QIAGEN, Hilden, Germany), and a SYBR Green (Bio-Rad, Hercules, CA, USA)-based qPCR assay was used to determine relative copy number of IFNAR1 EXON3 and EXON7. Primers for these reactions are as follows: EXON3 forward (5′-AAAGACGAGGCGAAGTGGTT-3′) and reverse (5′-CCTTCCTCTGCTCTGACACG-3′) and EXON7 forward (5′-AAGCAGTTCTGGAAGCCGTTC-3′) and reverse (5′-CCTCTGAGGCTTGTACATGGAG-3′).

CD8⁺ T cells were negatively selected and magnetically purified from splenocytes of OT-1 mice using a CD8α T cell isolation kit (Miltenyi Biotec, Auburn, CA, USA). For OT-I proliferation assays, sorted cells were labeled with 4 mM CellTrace Violet (CTV) (Life Technologies, Carlsbad, CA, USA) per the manufacturer’s instructions. Recipient mice received 1 × 10⁶ cells via tail vein injection. Spleens were harvested on day 5 after AAV administration, and proliferation was analyzed by flow cytometry. Dilution of CTV that occurs from proliferation reduces fluorescent intensity by ~½ increments. Cytotoxic activity of AAV capsid-specific CD8⁺ T cells was assessed using an in vivo killing assay as previously described.²⁵ In brief, total splenocytes isolated from WT C57BL/6 mice were either pulsed with 2 μg/mL SIIN peptide or with PBS for 1 h at 37°C and then differentially labeled with CTV to distinguish SIIN-pulsed target cells (3.0 μM, CTV^hi) from not-pulsed control cells (0.3 μM, CTV^lo). Cells were adoptively transferred by tail vein injection at a 1:1 ratio with a total of 2 × 10⁶ cells per mouse.

Antibody Assays

Plasma was analyzed for anti-AAV2 IgG2c by ELISA as previously described.²³ For the anti-capsid antibody ELISAs, sample wells were coated with 1 × 10⁹ vg/well intact AAV2-SIIN particles. For NAb assays, plasma samples were serially diluted and incubated for 1 h at 37°C with a fixed number of AAV2 vectors encoding luciferase (MOI of 5 × 10⁴ vg/cell) before infecting HEK293 cells. Luciferase activity was quantified 48 h later using a luciferase assay kit (Promega, Madison, WI, USA) following the manufacturer’s instructions. The final NAb titer was defined as the reciprocal of the dilution that showed >50% inhibition of transduction relative to virus alone.

Statistics

Results are reported as means ± SEM. Significant differences between groups were determined with an unpaired Student’s t test, Mann-Whitney U test, or two-way ANOVA with Bonferroni post-tests, as appropriate. A p value of <0.05 was considered significant. Analyses were performed using GraphPad Prism (San Diego, CA, USA). Experiments based on tetramer stains were done twice, while experiments on phenotyping of APCs were performed at least three times.

Author Contributions

J.L.S., A.S., and I.Z. performed experiments; J.L.S., G.D.K., B.E.H., L.M.M., M.A.W., C.T., and R.W.H. designed experiments; J.L.S., L.M.M., M.A.W., C.T., and R.W.H. analyzed and interpreted data; J.L.S., C.T., and R.W.H. wrote the manuscript; R.W.H. supervised the study.

Conflicts of Interest

R.W.H. serves on the scientific advisory boards of Applied Genetic Technologies Corporation (AGCT) and Ally Therapeutics.