Modifying immune responses to adeno-associated virus vectors by capsid engineering

Abstract

De novo immune responses are considered major challenges in gene therapy. With the aim to lower innate immune responses directly in cells targeted by adeno-associated virus (AAV) vectors, we equipped the vector capsid with a peptide known to interfere with Toll-like receptor signaling. Specifically, we genetically inserted in each of the 60 AAV2 capsid subunits the myeloid differentiation primary response 88 (MyD88)-derived peptide RDVLPGT, known to block MyD88 dimerization. Inserting the peptide neither interfered with capsid assembly nor with vector production yield. The novel capsid variant, AAV2.MB453, showed superior transduction efficiency compared to AAV2 in human monocyte-derived dendritic cells and in primary human hepatocyte cultures. In line with our hypothesis, AAV2.MB453 and AAV2 differed regarding innate immune response activation in primary human cells, particularly for type I interferons. Furthermore, mice treated with AAV2.MB453 showed significantly reduced CD8⁺ T cell responses against the transgene product for different administration routes and against the capsid following intramuscular administration. Moreover, humoral responses against the capsid were mitigated as indicated by delayed IgG2a antibody formation and an increased NAb50. To conclude, insertion of the MyD88-derived peptide into the AAV2 capsid improved early steps of host-vector interaction and reduced innate and adaptive immune responses.

Graphical abstract

Büning and colleagues describe that insertion of a MyD88-derived peptide, known to block TLR downstream signaling, into the capsid of AAV vectors improved transduction efficiency, reduced innate immune responses in vitro, and attenuated adaptive immune responses in vivo.

Introduction

Vectors based on the adeno-associated virus (AAV), a member of the Dependoparvoviridae, are currently the most frequently applied in vivo gene therapy vehicles.¹ So far, six AAV vector-based gene therapy products, either administered locally (Glybera, Luxturna, Upstaza) or intravenously (Zolgensma, Roctavian, Hemgenix), have been approved in Europe for the treatment of lipoprotein lipase deficiency, RPE65 gene defect causing retinal diseases, aromatic l-amino acid decarboxylase deficiency, spinal muscular atrophy type 1, and hemophilia A and B, respectively.^2,3,4,5 Further approvals are expected given the number of late-stage clinical trials particularly for central nervous system, eye, muscle, and liver-directed gene therapies.¹

However, gene therapy using AAV vectors is challenged by adaptive immune responses, i.e., antibody and CD8⁺ T cell responses. Specifically, upon vector administration patients develop antibodies against the AAV capsid, which impair vector re-application. Moreover, cytotoxic CD8⁺ T cell immune responses directed against the transgene product or capsid epitopes causing loss of therapeutic benefit have been observed particularly when high vector doses are administered.^6,7 Occurrence of acute thrombocytopenia or hepatic or renal toxicity is also correlated with high vector doses.^6,7 Consequently, patients are closely monitored, and immunosuppressive drugs are administered.^8,9,10 However, applying broad immunosuppression is prone to side effects, including failure of inducing regulatory T cells,¹¹ indicating an urgent need for developing novel strategies. One promising approach is to decrease the immunogenicity of the vector itself.^12,13,14,15 Such modifications in principle impact on the immune response both in cells that are the intended targets of transduction and in off-target cells including antigen-presenting cells.

AAV vectors are composed of a non-enveloped icosahedral protein capsid and a single-stranded DNA genome,¹⁶ which is flanked by inverted terminal repeats serving as the origin of replication and packaging signal.¹⁶ Infection by the wildtype virus commonly takes place in childhood in conjunction with an adenoviral infection leading to seroconversion and likely to the development of a memory T cell response.^11,17

De novo induction of adaptive immune responses is dependent on the innate immune system and its pathogen-associated molecular pattern (PAMP) detection system.¹⁸ Upon AAV infection, AAV capsids are recognized by Toll-like receptor 2 (TLR2), while AAV vector genomes are sensed by TLR9.^19,20 TLR2 is located on plasma membranes and is activated by a broad range of PAMPs, among them symmetrical structures such as viral capsids,²¹ while TLR9 is located in the endosomal compartment and is known to sense unmethylated cytosine-phosphate-guanine (CpG) motifs in bacterial or viral DNA.²² Although the cytokine profile that is induced after immune sensing via TLR9 and TLR2 differs slightly, both TLRs are dependent on MyD88 as an intracellular adaptor molecule that connects TLR and interleukin (IL)-1 receptor signaling to the IL-1 receptor-associated kinase (IRAK) signaling complex.^23,24,25 Key parts of MyD88 for this function are the Toll/IL-1R homology (TIR) domain and death domain (DD) located at opposite ends of the protein.²⁵ Thus, in case a PAMP or danger-associated molecular pattern (DAMP) is sensed by its cognate TLR, TLRs form dimers and recruit MyD88 via their cytoplasmic TIR domains by binding to the TIR domain of MyD88.²⁵ This event triggers multimerization of MyD88 molecules, again involving the TIR domains.²⁵ Subsequently, kinase domain-containing IRAK molecules, including IRAK1, 2, and 4 are recruited by DD binding leading to the transmission of the TLR activation signal to downstream mediators. This finally results in the activation of transcription factors such as nuclear factor κB (NF-κB), activator protein 1, and interferon regulatory factor 7 driving the expression of proinflammatory cytokines and chemokines, which trigger the inflammatory response and, subsequently, shape the adaptive immune response.^23,25 This signaling cascade can be interrupted by blocking MyD88 dimerization.²⁶ An efficient inhibitor of MyD88 dimerization is the peptide RDVLPGT, derived from the BB-loop of the TIR domain of MyD88 itself.²⁶

Engineering the capsids has been largely exploited to re-direct tropism of AAV vectors for example by inserting receptor binding ligands at protrusions located at the 3-fold axis of symmetry.¹⁶ Here, we inserted the RDVLPGT peptide at the tip of the highest protrusion, at the insertion position 453 (I-453), aiming to interfere with MyD88 dimerization and thus TLR downstream signaling, while maintaining AAV’s natural tropism.^16,26,27 We thereby developed an AAV2 capsid variant, AAV2.MB453, that remarkably differed from AAV2 vectors with wildtype capsids. Specifically, AAV2.MB453 demonstrated superior transduction efficiency in primary human cells and triggered—compared with AAV2—a reduced expression of key mediators of the innate immune system such as IL1A, IL8, interferon (IFN)A and IFNB. In mice, the formation of IgG2a antibodies binding to AAV2 capsids was significantly delayed following intravenous as well as intramuscular administration and sera collected from AAV2.MB453 injected mice were less potent than sera from the AAV2 cohort in neutralizing AAV2 transductions. Furthermore, we observed that mice receiving vector genomes delivered by AAV2.MB453 developed significantly reduced CD8⁺ T cell responses against the transgene product as well as to the AAV2 capsid as compared with AAV2 vectors. In conclusion, engineering the AAV2 capsid to display a MyD88 blocking peptide in each of the 60 subunits resulted in an AAV2 capsid variant with improved transduction efficiency, reduced innate immune responses in vitro, and attenuated adaptive immune responses in vivo.

Results

Insertion of MyD88 blocking peptides does not compromise vector production

To maintain the natural tropism of AAV2 vectors, we decided to introduce the MyD88 dimerization blocking peptide RDVLPGT including linker alanine residues at I-453 between glycine (G) 453 and threonine (T) 454 (Figure 1A). Following capsid assembly, I-453 is located at the tip of the highest protrusion at the 3-fold axis of symmetry (Figure 1B).²⁷ The change in tertiary structure of the β-turn of the GH loop of AAV2 (also known as variable region-IV), induced by the insertion, is illustrated for a single capsid subunit in the close-up in Figure 1B and indicates that the inserted MyD88 peptide forms helical structures clearly protruding from the capsid. Albeit the modeled structure of the MyD88 peptide as part of the capsid protrusion is very similar but not identical to the structure of the same peptide in the BB-loop of the MyD88 TIR domain, docking of a MyD88 TIR domain to the peptide displayed on the AAV2.MB453 capsid was predicted based on protein structure analyses (Figure 1C).

Figure 1.

Modeling of the MyD88 peptide-presenting AAV2.MB453 capsid

The MyD88-derived peptide was inserted at I-453 flanked by five alanine residues (A). Structure prediction of the AAARDVLPGTAA peptide in the common VP3 region at I-453 (B). Three VP3 capsid subunits are depicted (white, dark gray, and light gray). The highest capsid peak formed by the β-turn of the GH loop with capsid position I-453 is illustrated in green and the inserted peptide is shown in orange. Close-up highlights the helical structures formed by the inserted MyD88 peptide (B). The interaction of the MyD88 peptide inserted at I-453 to a TIR domain of MyD88 (light brown) is shown in (C). Close-up illustrates binding of the inserted peptide (orange) to residues (light brown) of a MyD88 TIR domain (C).

The new capsid variant, termed AAV2.MB453 for MyD88 blocking peptide in capsid position 453, was produced as self-complementary (sc) vector encoding for GFP (eGFP) driven by the human cytomegalovirus (CMV) promoter. As reference, we produced in parallel AAV2 vectors with wildtype capsids delivering the identical vector genome. Since one of the downstream targets of TLR signal pathways is NF-κB, we produced the same vectors employing the spleen focus-forming virus (SFFV) promoter instead of the CMV promoter as it contains significantly less NF-κB transcription factor binding sites (Figure S1A). Following vector production and purification, genomic and capsid titers were determined by qPCR and ELISA, respectively, revealing comparable titers for AAV2.MB453 and AAV2 vector preparations (Figures S1B and S1C). Thus, AAV2.MB453 and AAV2 did not differ regarding packaging efficiency, i.e., in the empty-to-full capsid ratio (Figure S1D). In addition, western blot analysis confirmed the expected 1:1:10 ratio of the three capsid proteins VP1, VP2, and VP3 contributing to the mature capsids as well as the expected slower migration of all three engineered capsid proteins of AAV2.MB453 in SDS-PAGE compared with AAV2²⁷ (Figure S1E). Finally, we compared the infectivity of AAV2.MB453 and AAV2 vectors in HEK293 cells revealing a non-significant decrease in transduction efficiency for AAV2.MB453 (Figures S1F and S1G).

Treatment with AAV2.MB453 alters expression profile of immune response-related genes in monocyte-derived dendritic cells

Dendritic cells (DCs) are key players in orchestrating immune responses after their activation notably by sensing the presence of molecules derived from pathogens via TLRs and other innate immune receptors. To investigate whether incorporation of the MyD88 dimerization blocking peptide into the capsid of AAV2 impacts on how DCs respond to AAV vector treatment, we used human monocyte-derived DCs (moDCs). They are derived from human blood monocytes cultured in the presence of granulocyte-macrophage colony-stimulating factor (GM-CSF) and IL4, and are frequently used to study DC functions.^28,29

Incubation of moDCs with AAV2.MB453 or AAV2 vectors at a particle-per-cell ratio (GOI) of 2.5 × 10⁴ resulted in transgene expression from sc vector genomes; however, AAV2.MB453 outperformed AAV2 with regard to efficiency (Figure 2A). Specifically, AAV2 led to an average of 18% transgene expressing cells (n = 5 donors), whereas AAV2.MB453 averaged 40% transgene positive cells. As the variation between different donors was substantial, we normalized the data to AAV2 revealing a significant increase in transduction efficiency for AAV2.MB453 (Figure 2B). The median fluorescence intensity (MFI) was similar in both groups (Figure S2).

Figure 2.

Transduction of moDCs and impact on expression of immune response-related genes

Human moDCs were incubated with indicated vectors at a GOI of 2.5 × 10⁴. To determine the level of vector transgene expression (A), cells were analyzed via flow cytometry 48 h p.t. Mean with individual donors is depicted in (A) for n = 5 donors with technical triplicates. Transduction levels were normalized (AAV2 set as 1) and transduction is shown as fold change (B). Data are presented as mean with individual donors for n = 5 donors. Mann-Whitney U test was applied to calculate statistical significance. ∗p < 0.05. For multigene analysis, cells were harvested 8 h p.t., RNA isolated and transcribed into cDNA and probed by Profiler PCR Array detecting the expression level of 84 genes involved in innate and adaptive immune responses. In (C), the expression levels of key innate immune mediators compared to mock control are depicted as fold change for treatment with AAV2, AAV2.MB453, and AAV2.VSSTSPR. Figure S3 displays all genes analyzed and samples that are expressed below the threshold (Ct value 40). Data presented are from n = 1 moDC donor (donor a). The gene expression pattern in AAV2.MB453-treated cells was compared with the AAV2 sample and presented as fold change (D). Relative upregulation (in AAV2.MB453 sample compared with AAV2 sample) is highlighted in red and relative downregulation (in AAV2.MB453 sample compared with AAV2 sample) is illustrated in blue. Only genes that are upregulated or downregulated more than 2-fold are shown (D).

To study whether the presence of the MyD88 blocking peptide on AAV2.MB453 influences the cell autonomous immune response toward AAV vectors in moDCs, we determined possible changes in the mRNA expression pattern of 84 immune response-related genes (in donor a) 8 h after vector treatment (Figures 2C, S3A, and S3B). As stimulation control, moDCs were challenged with heat-killed Listeria monocytogenes (HKLM), a well-described agonist of TLR2,¹⁹ a pathogen recognition receptor expressed on human DC.^30,31 To rule out that possible effects on gene expression are solely caused by the use of an AAV2 capsid modified by peptide insertion, we included the capsid variant AAV2.VSSTSPR as a further control.²⁸ This peptide insertion variant is derived from an AAV peptide display selection in moDCs and demonstrated significantly improved transduction efficiencies in moDCs as compared with AAV2, particularly in the presence of lipopolysaccharide, a strong inducer of DC maturation.²⁸

An up- or downregulation in gene expression by 2-fold was considered significant. Compared with mock-treated moDCs, treatment with TLR2 agonists HKLM induced a strong upregulation of pro-inflammatory genes including IL1A, IL1B, IL6, IL8, and tumor necrosis factor (TNF) (Figure S3B). Treatment of moDCs with AAV2 also led to a notable upregulation of the same pro-inflammatory cytokines and chemokines, except for TNF (Figures 2C and S3B). In addition, AAV2 treatment resulted in the upregulation of type I IFNs, IFNA, and IFNB (Figures 2C and S3B). Compared with mock-treated moDCs, capsid variants AAV2.VSSTSPR as well as AAV2.MB453 also upregulated IL1B, IL6, and IL8 expression; however, in comparison with AAV2.VSSTSPR and AAV2, moDCs exposed to AAV2.MB453 responded with a less prominent upregulation of IL1B and IL8 (Figures 2C and S3B). In general, a noticeable overall reduced innate immune response was observed for AAV2.MB453 (Figures 2C and S3B). Particularly interesting was the difference regarding type I IFNs: while the capsid-engineered AAV2.VSSTSPR variant like AAV2 upregulated both IFNA and IFNB, neither IFNA nor IFNB expression was affected in AAV2.MB453-treated moDCs (Figures 2C and S3B). Gene array results (Figures 2C and S3B) were confirmed by qPCR analyses for IL8, IL1B, and IFNB, a representative subset of immune response-related genes (Figure S4).

Directly normalizing gene expression patterns induced by AAV2.MB453 treatment to the pattern triggered by AAV2 revealed reduced expression of most genes in response to AAV2.MB453, including key mediators of the innate immune system (Figure 2D). The most noticeable exception was a substantially higher expression of C3 and IL23 (Figure 2D). In contrast, in comparison with AAV2-induced gene expression levels, AAV2.MB453 triggered significantly reduced gene expression of IFNA, IFNB, IL1A, and IL8 (Figure 2D). Thus, because of the insertion of the MyD88 blocking peptide, moDCs respond to treatment with AAV vectors with a reduced innate immune response as schematically summarized in Figure S5.

Enhanced transduction of primary human hepatocyte cultures by AAV2.MB453

As a second ex vivo model, we chose primary human hepatocyte (PHH) cultures and analyzed AAV2.MB453 vector- or AAV2 vector-treated cultures either 24 h or 4 d post transduction (p.t.) (Figure S6). Interestingly, as in the case of moDCs (Figure 2A), AAV2.MB453 transduced PHH cultures from different donors with higher efficiency than AAV2 vectors (Figure S6). Higher numbers of transgene-expressing cells (Figures S6A, S6C, and S6E), as well as increased expression levels as indicated by mRNA (Figures S7A–S7D), MFI (Figures S6B, S6D, and S6F), and protein amount (Figure S7E) were observed. Improved transduction efficiency was also apparent when switching from CMV to SFFV promoter arguing for a promoter-independent effect (Figures S6C, S6D, S7B, and S7D).

To evaluate whether an improved uptake of AAV2.MB453 into PHH cultures caused the elevated transduction efficiency, we measured intracellular vector genomes at both time points post vector administration (Figures S7F and S7G). In both cases, average vector copy numbers did not differ for AAV2 compared with AAV2.MB453 arguing for a post-entry step responsible for the better performance of AAV2.MB453.

Consequently, we performed a gene expression profiling at 5 h and 24 h p.t., focusing again on immune response-related genes albeit hepatocytes are non-professional immune cells.³² Again, up- or downregulation in gene expression level by 2-fold was considered significant.

At 5 h p.t. and compared with mock, a significantly increased expression level of CD86 was observed in AAV2 vector-treated PHH cultures, while expression levels of IL13, IL17A, and IL1A were found to be reduced (Figure S8). In contrast, AAV2.MB453 vector-treated PHH cultures showed a strong upregulation of IFNA (10-fold) and a 2-fold upregulation of TLR9, while IL13, TNF, and CD40 ligand (CD40L) expression was downregulated. At our second time point, 24 h p.t., a higher number of immune response-related genes showed an altered expression level in vector-treated samples compared with mock (Figure S9). Again, CD86 expression was upregulated in response to AAV2 (27-fold). At this time point, also AAV2.MB453-treated cells showed a strong upregulation of CD86 (190-fold). Whereas AAV2 treatment led to an upregulation of TLR9 (20-fold), TLR9 was not detectable in PHH cultures exposed to AAV2.MB453. Interestingly, whereas AAV2 treatment triggered the upregulation of IFNB (5.9-fold), AAV2.MB453 did not induce any IFNB expression (Figure S9). Expression levels of IFNA were not found to be elevated in response to either AAV2 or AAV2.MB453.

Directly normalizing gene expression patterns induced by AAV2.MB453 treatment to the pattern triggered by AAV2 at 5 h p.t. revealed upregulation of expression for IFNA, IL17A, and NLRP3, while CD40L, IL4, IRAK1, and TNF were significantly downregulated (Figure S10A). At 24 h p.t., AAV2.MB453 vector-treated PHH cultures showed upregulation of CCR5, CD40L, CD86, IFNG, and IL10 (among others) and significant downregulation of CCR8, TLR9, IL4, and IFNB (Figure S10B).

AAV2.MB453 mediates a more efficient transgene expression in the liver following intravenous administration

We investigated the performance of AAV2.MB453 in vivo, focusing first on the intravenous administration route and liver as target organ. Male BALB/c mice were tail vein-injected with 1 × 10¹¹ particles of either AAV2.MB453 or AAV2 vectors encoding for eGFP controlled by the CMV promoter. At 24 h, 5 d, and 10 d post administration (p.i.), mice were harvested, and transduction efficiency was monitored (Figure 3).

Figure 3.

Vector genomes and transgene expression in liver after intravenous AAV injection

Male BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 24 h p.i., 5 d p.i., or 10 d p.i. mice were sacrificed. For each time point, two independent experiments were performed with in total seven mice per group. Liver tissue was collected, and single cells isolated. DNA was isolated from single cells and analyzed with qPCR using eGFP-specific primers. The amount of vector genomes was normalized to AAV2 per experiment to consider variations between experiments for 24 h p.i. (A), 5 d p.i. (B), and 10 d p.i. (C). In addition, RNA was isolated from single cells and transcribed into cDNA analyzed with qPCR using eGFP-specific primers. The transgene expression was normalized to AAV2 per experiment to balance variations between experiments for 24 h p.i. (D), 5 d p.i. (E), and 10 d p.i. (F). The transgene expression efficiency is presented as ratio of transgene expression (cDNA) to vector genomes for 24 h p.i. (G), 5 d p.i. (H), and 10 d p.i. (I) and normalized to AAV2. Data are presented as mean ± SEM for n = 5 to 7 mice. Mann-Whitney U test was applied to calculate statistical significance. ∗p < 0.05.

Livers from mice analyzed 24 h p.i. showed comparable amounts of AAV2.MB453 and AAV2 vector genomes indicating no major difference in infectivity (Figure 3A). Interestingly, mice cohorts differed at later time points, i.e., on day 5 and day 10 (Figures 3B and 3C), suggesting that vector genomes delivered by AAV2.MB453 are degraded faster compared with AAV2 vectors. Specifically, when calculating vector copy numbers per cell using the mouse hypoxanthine guanine phosphoribosyl transferase (HPRT) gene as reference (Figures S11A–S11C), approximately 50 vector genomes per cell were present in the mouse cohorts analyzed at 24 h p.i. for both vectors (Figure S11A). At 10 d p.i., however, the mouse cohort treated with AAV2.MB453 showed two vector genome copies per cell, while on average six copies were present in mice which had received AAV2 vectors (Figure S11C).

When exploring transgene expression at 24 h p.i. as well as at 5 d p.i., a higher transgene mRNA level for AAV2.MB453 compared with AAV2 was observed, which failed to reach statistical significance (Figures 3D, 3E, and S11D). No difference was detected at 10 d p.i. (Figures 3F and S11E). However, correlating transgene expression (mRNA/cDNA) to vector genomes (DNA) (Figures 3G–3I) revealed a statistically significant increase in liver transduction for AAV2.MB453 compared to AAV2 for the 5 d p.i. time point (Figure 3H).

Modified humoral immune response against AAV2.MB453 following intravenous administration

To determine whether display of the MyD88 dimerization blocking peptide on AAV2.MB453 impacts humoral immune responses directed against the AAV2 capsid, levels of anti-AAV2 IgG2a antibodies in the sera of mice that had received AAV2.MB453 or AAV2 vectors intravenously were determined by ELISA (identical mice as in Figure 3). We here focused on IgG2a antibodies, as induction of these antibodies occurs in a MyD88-dependent manner,^33,34 and used AAV2 vectors as bait since capsids of AAV2 and AAV2.MB453 are more than 98% identical. As expected, no IgG2a antibodies were detectable at 24 h p.i. (Figure 4A). On day 5 p.i., mice of both cohorts had produced AAV2-binding antibodies (Figure 4B) that were comparable in level (approximately 200 ng/mL). Interestingly, mice of the day 10 p.i. AAV2 cohort showed higher levels of anti-capsid antibodies compared with AAV2.MB453-treated mice (approximately 1,080 ng/mL vs. 360 ng/mL) (Figure 4C). Of note, mice within the AAV2 cohort varied, with four of seven mice developing antibody levels comparable with the AAV2.MB453 cohort, while three of seven developed a stronger response (Figure 4C).

Figure 4.

AAV2-binding IgG2a antibodies in serum after intravenous injection

Male BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes (same mice as in Figure 3). At 24 h p.i. (A), 5 d p.i. (B), or 10 d p.i. (C) mice were sacrificed, blood taken, and serum collected. Serum was analyzed regarding AAV2-binding antibodies of the IgG2a subtype by ELISA. Purified IgG2a was used to calculate concentrations. As mentioned above, for each time point, two independent experiments were performed with in total seven mice per group. In addition, female BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 8 d p.i. (D, E) or 20 d p.i. (D, F), blood was taken and serum was collected. Serum was analyzed with ELISA as described above. Data are presented as mean ± SEM for n = 6–7 mice. Mann-Whitney U test was applied to calculate statistical significance. ∗p < 0.05; ∗∗p < 0.01.

In a second experiment, we monitored the development of anti-AAV2 IgG2a antibodies in female mice over time focusing on day 8 and day 20 p.i. as time points of measurement. Lower levels of anti-AAV2 antibodies were observed during the entire course of the experiment in the AAV2.MB453-injected mice compared with the cohort that had received AAV2 (Figure 4D). These differences reached statistical significance for both time points, day 8 and day 20 p.i. (Figures 4D–4F).

A neutralization assay was performed in HEK293 cells with sera from female mice collected on day 20 using again AAV2 vectors as a target (Figure 4G). As indicated in Figure 5A, sera from the AAV2.MB453 cohort neutralized transduction of HEK293 cells with AAV2 vectors less efficiently (NAb50: $\sim$1:200) compared with sera from the AAV2 cohort. For the latter an NAb50 of $\sim$1:1,000 for neutralizing AAV2 vector-mediated transductions was calculated (Figure 5A). Since cohorts differed in the total amount of IgG2a (Figure 4F), we performed an additional neutralization assay using adjusted IgG2a binding antibody concentrations. Interestingly, this time, the serum from the AAV2.MB453 cohort did not differ noticeably in neutralizing AAV2 vector transduction compared with serum from the AAV2 cohort (Figure 5B).

Figure 5.

Neutralization assay

Serum analyzed for IgG2a that was collected on day 20 (Figure 4F) was probed for neutralization assays. In (A), different serum dilutions (1:150, 1:300, 1:600, 1:1,200, 1:2,400), and in (B) equal amounts of IgG2a antibodies (as calculated before, Figure 4F) were incubated with AAV2 at GOI of 1 × 10³ for 1 h before infection of HEK293 cells. As control, cells were incubated with the same GOI of AAV2 in the absence of serum (“no serum” set as 100). Transgene-expressing cells were analyzed via flow cytometry. Data are presented as mean ± SEM for n = 6–7 mice. Green line indicates NAb50 value.

Modified T cell-mediated immune response against AAV2.MB453 following intravenous administration

In addition to humoral responses, we assayed the development of cytotoxic CD8⁺ T cell responses directed either against the vector capsid or the transgene product following intravenous administration. First, female mice were injected with 1 × 10¹¹ vector particles of either AAV2.MB453 or AAV2 encoding for eGFP controlled by the CMV promoter. At 10 d p.i., mice were harvested and splenocytes isolated. CD8⁺ T cell responses against the transgene product and the capsid were determined by ELISpot using transgene-specific or capsid-specific peptides, respectively. Interestingly, mice of the AAV2.MB453 cohort had developed significantly fewer eGFP-specific CD8⁺ T lymphocytes per 1 × 10⁶ splenocytes compared with the AAV2 cohort (Figure 6A).

Figure 6.

CD8⁺ T cell responses to transgene product and AAV2 capsid after intravenous AAV injection

Female BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 10 d p.i., mice were sacrificed, spleens harvested, and splenocytes isolated. Splenocytes were stimulated with immunodominant epitopes for eGFP (HYLSTQSAL) and AAV2 capsid (QYGSVSTNL + PQYGYLTL), respectively, and antigen-specific, IFN-y-secreting T cells were counted. T cells reactive to eGFP (A) and to AAV2 capsid (B) are shown per 1 × 10⁶ splenocytes. N = 7 mice per group. In an additional experiment, female BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes (same mice as in Figures 4D–4F). These mice were sacrificed at 58 d p.i., spleens harvested, and splenocytes isolated. Splenocytes were stimulated with peptides PQYGYLTL and QYGSVSTNL (capsid) or eGFP peptide HYLSTQSAL, respectively, and antigen-specific, IFN-y-secreting T cells were counted. T cells reactive to eGFP (C) and to AAV2 capsid (D) are shown per 1 × 10⁶ splenocytes. Male BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes (same mice as in Figure 4C). At 10 d p.i., mice were sacrificed, spleens harvested, and splenocytes isolated. Splenocytes were stimulated with immunodominant epitopes for eGFP (HYLSTQSAL) and AAV2 capsid (PQYGYLTL), respectively, and antigen-specific, IFN-y-secreting T cells were counted. T cells reactive to eGFP (E) and to AAV2 capsid (F) are shown per 1 × 10⁶ splenocytes. As mentioned above, two independent experiments were performed with in total seven mice per group (E and F). Data are presented as mean ± SEM for n = 7 mice. Mann-Whitney U test was applied to calculate statistical significance. ∗p < 0.05; ∗∗p < 0.01.

With regard to the CD8⁺ T cell response against the AAV2 capsid at 10 d p.i., only a very mild response was detected in general (Figure 6B). When comparing mice that received AAV2.MB453 with the AAV2 cohort, a tendency toward slightly less capsid-reactive T cells was observed (Figure 6B).

To evaluate T cell responses at a later time point, we performed ELISpot assays using splenocytes from female mice harvested on day 58 p.i. (same mice as Figures 4D–4F). Analysis revealed a moderate but clearly detectable CD8⁺ T cell response to the transgene product in the AAV2 vector cohort, while transgene-directed CD8⁺ T cells were not detectable in mice injected with AAV2.MB453, thus, demonstrating a significantly lower response compared with AAV2 (Figure 6C). Levels of capsid-specific T cells were lower and about background level (Figure 6D).

Since immune responses might differ depending on gender,³⁵ we next evaluated T cell responses in male mice using the abovementioned “day 10 cohorts” (same mice as in Figures 3C, 3F, 3I, and 4C). As indicated in Figure 6E, an overall strong CD8⁺ T cell response against the eGFP transgene product was detected. Interestingly, mice of the AAV2.MB453 cohort showed almost half the number of eGFP-specific CD8⁺ T lymphocytes per 1 × 10⁶ splenocytes (Figure 6E) as compared with the AAV2 cohort. In contrast with transgene-specific CD8⁺ T cells, capsid-specific CD8⁺ T cells were found less frequently in both groups, with AAV2.MB453 treated mice tending to develop a lower level of capsid-specific CD8⁺ T cells (Figure 6F). Differences observed for transgene-specific and capsid-specific CD8⁺ T cells between AAV2 and AAV2.MB453 vector-treated cohorts did not reach statistical significance (Figures 6E and 6F).

Since CD8⁺ T lymphocyte activity is reported to lead to the destruction of vector-transduced liver cells in human clinical trials,^6,7 we analyzed serum levels of alanine transaminase and aspartate transaminase (AST) (same mice as in Figures 4A–4C, 6E, and 6F) as hallmarks of liver cell damage (Figure S12). Differences were mainly seen for AST and, interestingly, in the cohort treated with AAV2.MB453 lower AST levels were observed on average (Figure S12A).

AAV2.MB453 mediates a more efficient transgene expression after intramuscular injection

As a second major target organ of AAV gene therapy, we focused on skeletal muscle tissue, specifically the hind leg muscle, and compared two vector doses, i.e., 1 × 10¹⁰ and 1 × 10¹¹ vg/mouse in female mice.

Vector copy numbers and transduction efficiencies were determined 48 d p.i. (Figure S13). In the AAV2 vector cohort, vector copy numbers correlated with vector dose, which was not the case for AAV2.MB453-treated mice (Figure S13A). Moreover, in the high-dose cohorts, significantly fewer vector genomes were detected for AAV2.MB453 compared with AAV2 (Figure S13A). Despite this difference, transgene expression on average was comparable for AAV2 and AAV2.MB453 in the high-dose cohorts (Figure S13B), again arguing for a higher transduction efficiency (transgene expression per vector genome) of AAV2.MB453.

Modified humoral immune response against AAV2.MB453 following intramuscular injection

To characterize the humoral immune response in female mice after intramuscular injection of AAV vectors, serum had been collected at 7 d, 23 d and 48 d p.i. In the low dose cohort (1 × 10¹⁰ vg/mouse), most animals had not yet developed anti-AAV2 antibodies by day 7 (Figures 7A and 7B). However, at 23 d p.i. IgG2a levels had increased and were significantly lower in AAV2.MB453 vector-injected mice compared with AAV2 vector-treated mice (Figures 7A and 7C). Specifically, for mice of the AAV2 cohort we measured a concentration of approximately 23,000 ng/mL IgG2a, whereas a 3.9-fold lower concentration was measured in the serum of mice treated with AAV2.MB453 (approximately 5,900 ng/mL IgG2a). A decrease in antibody levels was still detectable at 48 d p.i.; however, the difference was not statistically significant (Figure S14A). For cohorts of mice that had received the one-log higher dose (1 × 10¹¹ vg/mouse) already at day 7 p.i. approximately 600 ng/mL anti-AAV2 specific IgG2a antibodies were detectable on average when AAV2 was injected (Figures 7D and 7E). In contrast, 6-fold fewer antibodies were measured in mice injected with AAV2.MB453 (approximately 100 ng/mL IgG2a). Interestingly, at later time points (day 23 and 48 p.i.), overall comparable antibody levels were observed (Figures 7D, 7F, and S14B).

Figure 7.

AAV2 capsid-binding IgG2a antibodies in serum after intramuscular injection and neutralization assay

Female BALB/c mice were injected intramuscularly with 1 × 10¹⁰ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 7 d p.i. (A, B) and 23 d p.i. (A, C) blood was taken and serum was collected. Serum was analyzed regarding AAV2 capsid-binding antibodies of the IgG2a subtype by ELISA. Purified IgG2a was used to calculate concentrations. Data are presented as mean ± SEM for n = 7 mice. Female BALB/c mice were injected intravenously with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 7 d p.i. (D, E) and 23 d p.i. (D, F) blood was taken and serum was collected. Serum was analyzed with ELISA as described above. Data are presented as mean ± SEM for n = 4 mice. Mann-Whitney U test was applied to calculate statistical significance. ∗p < 0.05. Serum analyzed for IgG2a that was collected on day 23 (C) was probed in a neutralization assay (G). Different serum dilutions (1:150, 1:300, 1:600, 1:1,200, 1:2,400) were incubated with AAV2 at GOI of 1 × 10³ for 1 h before infection of HEK293 cells. As control, cells were incubated with the same GOI of AAV2 in the absence of serum (“no serum” set as 100). Transgene-expressing cells were analyzed via flow cytometry (G). Data are presented as mean ± SEM for n = 6–7 mice. Green line indicates NAb50 value.

In addition to the total IgG2a level, we also performed a neutralization assay in HEK293 cells using different dilutions of the sera collected from the low-dose cohorts on day 23 (serum from Figure 7C). As before and as shown in Figure 7G, serum from mice injected with AAV2.MB453 was less potent in neutralizing AAV2 vectors (NAb50: $\sim$1:250) compared with serum from the AAV2 cohort (NAb50: $\sim$1:550).

Modified T cell-mediated immune response against AAV2.MB453 following intramuscular injection

For the low and high vector dose cohorts (same mice as in Figure 7), T cell responses were monitored on day 14 (using blood samples) focusing on anti-eGFP responses and on day 48 (using splenocytes at endpoint) evaluating responses against the transgene product as well as against the capsid (Figures 8A, 8B, and S15A). A mild but detectable anti-eGFP CD8⁺ T cell response was detected in the blood on day 14 p.i. for female mice injected with a low dose (i.e., 1 × 10¹⁰ vg/mouse) of AAV2 (Figure 8A). In contrast, in the AAV2.MB453 vector-treated cohort the response was significantly lower with only one in seven mice that developed a noticeable CD8⁺ T cell response, which, however, was quite strong (Figure 8A). In contrast, female mice receiving a one-log higher dose (1 × 10¹¹ vg/mouse) of AAV2 showed a strong anti-transgene T cell response in the blood on day 14 p.i. (Figure 8B). Interestingly, the clearly detectable response in mice treated with AAV2.MB453 was significantly lower (Figure 8B). Specifically, AAV2 induced 310 transgene-specific CD8⁺ T cells per 100 μL blood, whereas AAV2.MB453-treated mice only generated 50 transgene-specific CD8⁺ T cells per 100 μL blood.

Figure 8.

CD8⁺ T cell responses toward the transgene product and the capsid after intramuscular AAV injection

Female BALB/c mice were injected intramuscularly with 1 × 10¹⁰ vg/mouse or 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 14 d p.i. (A, B) blood was taken and blood cells stimulated with a peptide for the immunodominant epitope of eGFP (HYLSTQSAL) and antigen-specific, IFN-y-secreting T cells were counted for low vector dose (A) and high vector dose (B). Male BALB/c mice were injected intramuscularly with 1 × 10¹¹ vg/mouse of either AAV2 or AAV2.MB453 encoding eGFP as transgenes. At 7 d p.i. (C, E), blood was taken and blood cells stimulated with a peptide for the immunodominant epitope of eGFP (HYLSTQSAL) or with peptides for AAV2 capsid epitopes (PQYGYLTL + QYGSVSTNL) antigen-specific, IFN-y-secreting T cells were counted for eGFP (C) and capsid (E). At 14 d p.i. mice were sacrificed, spleens harvested, and splenocytes isolated. Splenocytes were stimulated with peptides for eGFP (HYLSTQSAL; D) or AAV2 capsid (PQYGYLTL + QYGSVSTNL) (F) and antigen-specific, IFN-y-secreting T cells were counted. Data are presented as mean ± SEM for n = 4–7 mice. Mann-Whitney U test was applied to calculate statistical significance. ∗p < 0.05.

As mentioned, at the endpoint (day 48 p.i.), we assayed T cell responses directed against the transgene and the capsid, respectively. ELISpot analyses of splenocytes from the low-dose cohorts revealed no significant differences in levels of transgene-specific CD8⁺ T cells between AAV2.MB453- and AAV2-treated animals (Figure S15A). Similarly, splenocytes from the high dose cohorts demonstrated generally comparable levels of transgene-specific T cells. With respect to the anti-capsid CD8⁺ T cell responses, in both low-dose and high-dose cohorts, AAV2.MB453 vectors tended to induce slightly lower levels of capsid-specific CD8⁺ T cells; however, differences observed were not considered statistically significant (Figure S15B).

To address again the issue of gender-dependent differences, we next focused on male mice and injected male BALB/c mice with 1 × 10¹¹ vg/mouse intramuscularly. Based on the result observed for female mice (Figures 8A, 8B, S15A, and S15B) we determined T cell responses directed against the transgene product and the capsid on day 7 p.i. (using blood samples) and day 14 p.i. (Figures 8C–8F).

On day 7, a mild CD8⁺ T cell response against eGFP was detected in the blood for both groups with AAV2.MB453 treated mice tending to develop lower levels of transgene-specific T cells (Figure 8C). In contrast, significantly reduced levels of anti-eGFP CD8⁺ T cells were observed on day 14 in the AAV2.MB453 cohort compared with the AAV2 cohort (Figure 8D).

In addition, a significantly reduced response toward the capsid was observed in the AAV2.MB453 cohort compared with the AAV2 cohort on day 7 (Figure 8E). Interestingly, however, the levels of anti-capsid T cells did not differ any longer when the same mice were analyzed 14 days after injection (Figure 8F).

In summary, AAV2.MB453 mitigated both humoral and cellular responses toward the AAV2 capsid and the transgene product after intravenous as well as intramuscular administration.

Discussion

Immune responses directed against the vector capsid or the transgene product are regarded as a major challenge in gene therapy. To tackle this issue in a targeted manner as an alternative approach for applying rather broad immune suppression treatments, we used a genetic capsid engineering strategy to dampen the innate immune system. Since innate immune responses shape subsequent adaptive immune responses, we hypothesized that our AAV2-derived capsid variant AAV2.MB453 would differ from AAV2 regarding innate as well as adaptive immune responses, and possibly even beyond as host responses are assumed to impact on AAV’s transduction efficiency.^36,37 Indeed, AAV2.MB453 showed not only an improved transduction efficiency on primary human cells (moDCs and PHH cultures) and following intravenous or intramuscular administration in mice (Figures 2A, 2B, 3H, S6, and S13), but also a reduced expression of key mediators of the innate immune system ex vivo, including IL1A, IL8, IFNA, and IFNB (Figures 2C, 2D and S3–S5). In addition, a significant delay in the formation of AAV2-specific antibodies was observed (Figures 4 and 7) and sera from mice injected with AAV2.MB453 also demonstrated a lower neutralizing potency (Figures 5 and 7G). Furthermore, CD8⁺ T cell responses against the transgene product eGFP and against the AAV2 capsid (early after intramuscular administration) were significantly reduced in mice injected with AAV2.MB453 vectors compared with AAV2 vector-treated mice (Figures 6 and 8). Thus, inserting a short peptide known to possess innate immune response blocking activity into the AAV2 capsid yielded an AAV2 capsid variant that showed improved transduction efficiency, i.e., that required fewer vector genomes to express the transgene product, and that reduced innate and attenuated adaptive immune responses.

The RDVLPGT peptide was first described by Loiarro and colleagues²⁶ as blocker of dimerization of MyD88. Thereby, MyD88, a key mediator of the PAMP and DAMP sensing part of the innate immune system, cannot exert its downstream function. Consequently, the signaling cascade that connects the recognition of a pathogen or a danger signal with the effector arm consisting of cytokines, chemokines, and type I IFNs is blocked.

The peptide was inserted into the highest of the three peaks at the 3-fold axis of symmetry of the capsid (i.e., at I-453) enabling display of 60 peptides per capsid without interfering with capsid assembly or vector genome packaging (Figures S1B–S1E). Structure and binding prediction studies argued that the inserted MyD88 peptide on AAV2.MB453 can interact with the TIR domain of MyD88, a key requirement for our strategy to work (Figure 1C).

Interestingly, when AAV2.MB453 vectors were applied, both in primary cells (PHH cultures and moDCs) and in vivo (after intravenous as well as intramuscular injection) an improved transduction efficiency compared with AAV2 vectors was observed (Figures 2A, 2B, 3H, S6, and S13). This enhanced transduction efficiency was not caused by improved cell entry, as intracellular vector copy numbers were equal or even lower for AAV2.MB453 compared with AAV2 (Figures 3B, 3C, S7F, S7G, and S13A). Improved intracellular processing appears as an alternative explanation and might be caused by the distinct cellular response triggered by AAV2.MB453 as opposed to AAV2 (Figures 2C, 2D, S3, S4, and S8–S10). This hypothesis is supported by the increased transgene expression in liver reported for AAV vectors with a low capability to stimulate TLR9 compared with a potent TLR9-activating AAV vector that expressed poorly.³⁷ In line, we observed for PHH cultures, a significant downregulation of proinflammatory cytokines TNF (5 h p.i.) and IFNB (24 h p.i.) in AAV2.MB453-treated compared with AAV2-treated PHH cultures (Figures S8–S10). Interestingly, TLR9 expression was below the detection limit in AAV2.MB453 vector-treated cells at 24 h p.t., while upregulated in cells treated with AAV2 vectors (Figures S9 and S10B).

In AAV2 vector-treated moDCs (donor a) elevated expression levels of several genes involved in innate immune responses such as IL1A, IL1B, IL6, IL8, IFNA, and IFNB were observed (Figures 2 and S3–S5). Upregulation of type I IFNs is in line with Zhu and colleagues,²⁰ who reported on activation of human plasmacytoid DCs (pDCs) purified from peripheral blood mononuclear cells (PBMCs) upon encounter with AAV2 vectors. Upregulation of IFNB was also observed for moDC in response to AAV2 (Figure 2; ³⁸), or AAV2.VSSTSPR (Figure 2C), or the two I-587 variants AAV2.GL and AAV2.NN.³⁸ In case of AAV2.MB453, however, levels of IFNB and IFNA in moDCs remained unchanged or were not detectable, respectively (Figures 2C, 2D, and S3–S5). Likewise, AAV2 vectors triggered a significant upregulation of IL1A and IL1B, whereas cells responded to AAV2.MB453 vectors by a less prominent upregulation of IL1B and no change in IL1A (Figures 2C, 2D, and S3–S5). In summary, results for type I IFNs and IL1A/IL1B clearly indicate an attenuated activation of the innate immune system in moDCs upon encounter with AAV2.MB453 compared with AAV2 (Figures 2C, 2D, and S3–S5). Reduced expression levels of type I IFNs in DCs in response to AAV2.MB453 might as well contribute to the modified adaptive immune response seen in our in vivo studies, since it is known that several types of DCs (pDCs and conventional DCs) cooperate to establish an anti-AAV capsid-directed CD8⁺ T cell response.^39,40 Another interesting difference concerns the cytosolic dsRNA sensor RIG-I that, like MDA5, was reported to be involved in AAV innate immune recognition.⁴¹ We observed upregulation of DDX58, the gene encoding RIG-I, following treatment with AAV2 and AAV2.VSSTSPR, but not in case of AAV2.MB453 (Figure S3B).

The involvement of TLR9 in the development of antibodies directed against AAV is unclear; conflicting results exist in the literature.^20,33,34,42 However, there is clear evidence for the importance of MyD88 in B cells for antibody generation.^33,34 Therefore, targeting MyD88 might be more promising than focusing on the inhibition of TLR9. Indeed, we observed lower levels of capsid-directed antibodies in mice receiving AAV2.MB453 either intravenously or intramuscularly, compared with mice treated with AAV2 (Figures 4 and 7). The production of antibodies was vector dose dependent, and AAV2.MB453 was more efficient in delaying antibody production after an intramuscular administration at a lower vector dose (1 × 10¹⁰ vg/mouse) (Figure 7). Consequently, we also observed a reduced potential to neutralize AAV2 in vitro using sera collected from mice that were injected with AAV2.MB453 (Figures 5A and 7G).

Since other antibody subtypes might also play a role such as IgG1 antibodies, which develop independent of MyD88 signaling,^33,34 we followed this hypothesis but could not detect antibodies of the IgG1 subtype in mouse sera (at 23 d p.i.; data not shown).

In line with a reduced humoral response to AAV2.MB453, CD8⁺ T cell responses against the transgene product (Figures 6A, 6C, 6E, and 8B) and against the capsid at early time points after intramuscular administration (Figure 8E) were lower when AAV2.MB453 was injected. We observed notable differences in T cell responses toward AAV2.MB453 between male and female mice. After intravenous injection, T cell responses against the eGFP transgene were significantly reduced in female mice (Figure 6A) injected with AAV2.MB453 compared with the AAV2 cohort (Figure 6A), but not when male mice were used (Figure 6E). For AAV2-treated mice, the level of eGFP-specific T cells was similar in female (684 spot forming cells [SFC]) (Figure 6A) and male mice (715 SFC) (Figure 6E). However, mice injected with AAV2.MB453 showed a higher eGFP-specific T cell response in male mice (472 SFC) (Figure 6E) compared with female mice (290 SFC) (Figure 6A). In contrast, there was no significant difference in terms of anti-capsid CD8⁺ T cells detected between AAV2 and AAV2.MB453 after intravenous injection (Figures 6B, 6D, and 6F).

After intramuscular injection of vectors, we detected significantly reduced levels of eGFP-specific T cells in AAV2.MB453 cohorts compared with AAV2-treated mice for both female (Figures 8A and 8B) and male mice (Figure 8D). Remarkably, when vectors were administered intramuscularly in female mice, we also observed a significant reduction in capsid-reactive T cells in mice injected with AAV2.MB453 compared with AAV2 cohorts (Figure 8E).

The here observed differences in gender are in line with previously published reports that described gender differences in multiple species for both the innate as well as the adaptive arm of the immune system and suggested the involvement of both genes and hormones.⁴³

In mouse studies, responses against the AAV2 capsid are generally weak.⁴⁴ Thus, to further determine the impact of MyD88 blockage on anti-capsid CD8⁺ T cell immune responses, we either need to change the animal species or might need to combine our peptide insertion with a more immunogenic AAV capsid. One possibility for the latter is the use of the ovalbumin-derived SIINFEKL epitope, which was used, for example, in the context of AAV2 and AAV8 capsids to study mechanisms of anti-AAV capsid immune response induction.^37,45,46 Using such an AAV2.SIINFEKL capsid, the Herzog group demonstrated that the RDVLPGT peptide—provided as membrane-penetrating peptide—can be used to reduce capsid-directed T cell responses in vivo.³⁹ In the same study, the humoral arm was also investigated. Authors reported on unaltered anti-AAV2 binding antibody levels upon administration of the membrane-penetrating version of the RDVLPGT peptide at 2 months after intravenous injection,³⁹ a finding in line with our results for late time points (Figures 7A and 7D). However, at earlier time points p.i. (Figures 7A and 7D), the AAV2 and AAV2.MB453 cohorts differed with a lower level of anti-AAV2 binding antibodies in serum of mice receiving AAV2.MB453.

Chan and colleagues¹³ recently presented an alternative strategy to reduce immune responses against AAV vectors by directly interfering with TLR9-mediated recognition of vector genomes. The use of a TLR9-inhibitory sequence that had been incorporated into the DNA sequence enabled vectors to partially evade deleterious anti-capsid cytotoxic T cell responses.¹³ More precisely, when injected intravitreally in mice AAV vectors with built-in TLR9-inhibiting sequences reduced the infiltration of CD8⁺ T cells and, simultaneously, bolstered transgene expression.¹³ The same strategy revealed some promising efficacy also in pig and non-human primate models.¹³ At the same time, however, results suggest that additional pathways of immune recognition that are independent of TLR9 might exist for mediating inflammation.

Another promising approach to circumvent TLR9-mediated recognition of AAV vectors is the reduction or elimination of CpG motifs from the vector genome. Promising results in this regard were published by Faust and colleagues,¹⁴ who reported on reduced anti-AAV CD8⁺ T cell responses and preserved transgene expression in muscle. More recently, Bertolini and colleagues¹² confirmed the potency of this strategy in reducing levels of capsid-specific CD8⁺ T cells and demonstrated in addition, a reduced level of transgene-specific CD8⁺ T cells. Martino and colleagues¹⁵ followed a different strategy to reduce anti-capsid T cell responses. By using a capsid-modified AAV vector for which surface-exposed tyrosine (Y) residues were replaced by phenylalanines (F), they could reduce proteasomal degradation of vector capsids and consequently, the amount of capsid antigen to be presented at major histocompatibility complex I. As a result, they observed reduced hepatocyte destruction by capsid-specific T cells and stabilized transgene expression. Ongoing clinical trials for inherited retinal diseases using AAV2(YF) mutants reinforce the great potential of this strategy (NCT02161380, NCT02416622, and NCT02599922).

Altogether, increasing our efforts to unravel cell autonomous immune responses contributing to AAV-directed adaptive immune responses is urgently needed. To reduce side effects, strategies for general immune suppression should be avoided. Instead, approaches that solely address cells that are transduced by AAV vectors or into which AAV vectors are internalized should be preferred. First steps in these directions have been performed with the seminal studies described above.^12,13,14,15 Our strategy is following the same line, albeit covering a broad spectrum of possible innate immune responses as we are targeting an upstream key molecule in most TLR signaling pathways. Our approach can easily be combined with one or more of the other above-mentioned approaches such as adding the TLR9 inhibitory sequence to the vector genomes delivered by AAV2.MB453 to name one example.¹³ Such combinations are expected to further enhance the inhibition of innate immunity obtained by our capsid engineering strategy and will be the focus of follow-up studies. Moreover, our approach can be transferred to AAV vectors based on other naturally occurring serotypes or to capsid-engineered variants.

Material and methods

Modeling of AAV capsids

Three VP3 proteins of AAV2 (PDB: 6IH9 ⁴⁷) or of capsid variant AAV2.MB453 were modeled with the Rosetta software suite,⁴⁸ using RosettaCM for loop construction⁴⁹ and docked to the TIR domain of MyD88 (PDB: 7BER ⁵⁰) with the Rosetta protein docking protocol⁵¹ through Rosetta Scripts.⁵² Figures were generated with ChimeraX.⁵³ A detailed protocol of the modeling and docking procedure can be found in the supporting information.

Cloning of capsid variant

The capsid variant AAV2.MB453 is based on AAV serotype 2 (AAV2). The MyD88-derived peptide, flanked by in total five alanine residues (AAARDVLPGTAA), was ordered as gene synthesis (GeneArt Gene Synthesis, Invitrogen) and cloned into the pRC helper plasmid⁵⁴ to produce pRC.MB453.

AAV vector production

Vectors used in this study encoded for eGFP under the transcriptional control of the CMV promoter or, if indicated, of the SFFV promoter in a sc AAV vector genome configuration. AAV vectors were produced and purified as described.^55,56 For AAV2, the helper plasmid pRC⁵⁴ and for AAV2.MB453 the helper plasmid pRC.MB453 was used in combination with pscGFP⁵⁷ (CMV or SFFV promoter) and pXX6.⁵⁵

After purification by discontinuous iodixanol gradient ultracentrifugation, a further purification and concentration step was used using filter column ultracentrifugation (Amicon Ultra Centrifugal Filter Units, Merck Millipore) to concentrate and to buffer to PBS/M/K (1× PBS/1 mM MgCl₂/2.5 mM KCl).

AAV vector characterization

Genomic particle titers were determined using absolute qPCR quantification with transgene-specific primers (eGFP_fw: CACAACGTCTATATCATGGC; eGFP_rev: TGTGATCGCGCTTCTC) and the corresponding vector plasmid standard curve. Capsid titers were quantified using the AAV2 Titration ELISA Kit (Progen). The transducing titer was determined after transducing HEK293 with serial dilutions of vector stocks and flow cytometry analysis as described.²⁷ For western blot analysis of vector capsid composition, capsid protein-specific antibody B1 (Progen) and HRP-coupled secondary antibody (10004302, Cayman chemical) were used as described previously.²⁷

Transduction experiments and expression analysis of immune-related genes

To generate moDCs, CD14⁺ cells were isolated from human PBMCs (buffy coat) using a CD14⁺ selection system (CD14 MicroBeads, Miltenyi Biotech). Buffy coats were provided by the Institute for Transfusion Medicine, Leipzig University Medical Center, and their use for the project was approved by the local ethics committee in Leipzig (Germany, 486-17-ek-005). Cells were cultured with 20 ng/mL IL4 (Immunotools) and 100 ng/mL GM-CSF (Immunotools) to differentiate into CD14^– moDCs. In experiments, moDCs were incubated with 2.5 × 10⁴ vg/cell and transduction efficiency was analyzed 48 p.t. via flow cytometry. For RNA isolation and gene expression analysis (Profiler PCR Array and qPCR), moDCs were collected 8 h p.t.

PHH were isolated and prepared as described elsewhere.⁵⁸ All patients provided written informed consent for research use of their specimen approved by the Ethics Committee of Hannover Medical School (no. 252–2008). PHH cultures were incubated with 2.5 × 10⁴ vg/cell and transduction efficiency was analyzed 24 h p.t. or 4 d p.t. via flow cytometry. For RNA isolation and gene expression analysis, PHH were collected 5 h p.t. or 24 h p.t. RNA isolation was performed with RNAzol (Sigma-Aldrich).

For cDNA synthesis, up to 1 μg RNA was used (QuantiTect Reverse Transcription Kit; Qiagen). For qPCR analysis of vector genomes and transcripts (PHH and moDC samples) the following primer pairs were used: hPLAT_fw: ACCTAGACTGGATTCGTG; hPLAT_rev: AGAGGCTAGTGTGCAT; hGAPDH_fw: GGTATCGTGGAAGGACT; hGAPDH_rev: GGGTGTCGCTGTTGAA; eGFP primer (see above); hIL8_fw: AAGAACTTAGATGTCAGTGC; hIL8_rev: ACTTCTCCACAACCCT; hIL1B_fw: GTACGATCACTGAACTGC; hIL1B_rev: GAGTGGGCTTATCATCTTT; hIFNB_fw: AATCTCCTCAGGGATGTCAAAGT; hIFNB_rev: TCCTCCTTCTGGAACTGCTGCA. For PHH samples, transgene mRNA expression was normalized to hGAPDH and vector genomes were normalized to hPLAT (Figure S7). For moDC samples, mRNA expression of IL8, IL1B, and IFNB was normalized to hGAPDH (Figure S4).

Expression levels of immune-related genes were measured with a qPCR-based array (RT² Profiler PCR Arrays, Innate & adaptive immune responses, Qiagen) according to manufacturer’s instructions (Figures 2C, 2D, S3, S8, and S9).

Figures 2D, S10A, and S10B show the fold change values in the AAV2.MB453 group were calculated relative to the AAV2 group: Formula fold change: 2ˆ(-delta Ct (AAV2.MB453 group)/2ˆ(-delta Ct (AAV2 group). Delta Ct is calculated as follows: Delta Ct: (Ct(gene of interest)—average Ct(housekeeping genes)). Figure 2C shows the fold change values were calculated relative to the mock control: Formula fold change: 2ˆ(-delta Ct (group 1, 2 or 3)/2ˆ(-delta Ct (mock control group).

Fold change values greater than 1 indicate an up-regulation. Fold change values less than 1 indicate a down-regulation, they are depicted as the negative inverse (−1/fold change).

Mice

Experimental procedures and animal housing were approved by the Lower Saxony State Office for Consumer Protection and Food Safety (LAVES) or by the local institutional ethic committee (Cenomexa, project #19709) and followed the European directive 2010/63/EU. BALB/c wildtype mice were purchased from Charles River and housed at the Central Animal Facility at Hannover Medical School or were purchased from Janvier Labs and housed in the dedicated animal facility of UMRs1234 Labs (University of Rouen). AAV vectors were injected into 7- to 8-week old mice via intravenous (tail vein) or intramuscular (hind leg muscles) route of administration at either 1 × 10¹⁰ vg/mouse (intramuscularly only) or 1 × 10¹¹ vg/mouse (both intravenously and intramuscularly) in a total volume of 100 μL PBS as indicated. For intramuscular injections, 25 μL was injected into four different sites. Blood samples were taken via retrobulbar venous plexus puncture at indicated time points. At the endpoint, blood was drawn by heart puncture and serum, splenocytes, liver, and muscle tissue were collected and analyzed.

DNA and RNA were isolated from murine liver cells and muscle tissue. The following primer pairs were used for qPCR detection of vector genomes and eGFP transcripts, respectively: mHPRT_fw: CACGTTTGTGTCATTAGTGAA; mHPRT_rev: AAGATAAGCGACAATCTACC; mGAPDH_fw: TACCCCCAATGTGTCCGTC; mGAPDH_rev: AAGAGTGGGAGTTGCTGTTGAAG; eGFP primer (see above). For murine liver samples, the level of vector genomes was normalized to HPRT and transgene mRNA expression was normalized to GAPDH (see Figure S11). In a subsequent step (Figures 3A–3F), vector genomes as well as transgene expression were normalized to the mean value of AAV2 (AAV2 set as 1).

Anti-AAV2 binding antibodies and neutralization assay

The level of anti-AAV2 IgG2a binding antibodies was determined by ELISA. In detail, 96-well plates were coated with 2.5 × 10⁹ capsids (as determined via AAV2 Titration ELISA). For a standard curve, wells were coated with purified IgG2a (PP102, Sigma-Aldrich). Serum was diluted in 1:3 dilution steps with a starting dilution of 1:50. Anti-AAV2 antibodies in the serum bound to coated AAV2 capsids were captured with an anti-mouse, HRP-coupled IgG2a antibody (1:25,000, ab97245, Abcam). Antibody complexes were visualized by addition of TMB (Sigma-Aldrich) and subsequently sulfuric acid. The absorbance was measured at 450 nm. To analyze the capacity of serum antibodies to neutralize AAV2 infection, either serum dilutions (1:150, 1:300, 1:600, 1:1,200, or 1:2,400) or equal amounts of IgG2a antibodies (as calculated before) were mixed with AAV2 and incubated for 1 h at room temperature before being added to HEK293 cells at GOI (vector particles per cell) 1 × 10³. As a control, cells were incubated with the same GOI of AAV2 in the absence of serum leading to 70%–80% of eGFP expressing cells. This condition was set as 100. NAb50 value is indicated as dilution or concentration needed to reduce transduction levels by 50%. Levels of transgene-expressing cells were determined via flow cytometry 48 h after transduction.

ELISpot

For isolation of splenocytes, excised spleens were mashed through a pre-wetted 70 μM strainer. Red blood cell lysis was performed using 1× ACK buffer (80.24 g/L ammoniumchloride, 10.01 g/L potassium bicarbonate, 1 mM EDTA, pH 7.4). Isolated splenocytes were counted and their viability checked. For ELISpot analysis, plates were coated with IFN-y capture antibody (AN-18, BioLegend) overnight. Splenocytes were stimulated by specific peptides (final 10 μg/mL). For the eGFP transgene, the peptide HYLSTQSAL (all produced by GeneCust) was applied and the peptides QYGSVSTNL and PQYGYLTL were used as immunodominant capsid epitopes.^59,60 Splenocytes were seeded on top at a density of 2–3 × 10⁵ cells/well. As positive control, T cell activator Concanavalin A (InvivoGen) was combined at 10 μg/mL with 5 × 10⁴ cells/well. As negative control, cells were left unstimulated. In this control, 10 to 20 SFC per 1 × 10⁶ splenocytes were observed and defined as background level. In case, ELISpot was performed with blood cells (100 μL blood), red blood cell lysis was conducted as well. To the peptide-cell mixture 1 × 10⁵ irradiated splenocytes were added to increase the number of antigen-presenting cells. IFN-y released from activated T cells was captured by an IFN-y detection antibody (R4-6A2, eBioscience/Invitrogen) and visualized by a secondary streptavidin-coupled ALP-conjugated antibody (Mabtech) and BCIP/NBT substrate (Mabtech). The number of developed spots indicates the number of specific T cells and was counted (ImmunoSpot ELISpot reader, CTL).

Statistical analysis

Statistical analysis was performed using unpaired t tests, Mann-Whitney U tests, or one-way ANOVA with Tukey’s Multiple Comparison Test and data are presented as mean ± SEM. Significance values are defined ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001. All statistical analyses were performed in GraphPad Prism 5 (GraphPad Software).

Data and code availability

This study includes no data deposited in external repositories.