A multi-functional AAV-CRISPR-Cas9 and its host response

Abstract

CRISPR-Cas9 delivery by AAV holds promise for gene therapy but faces critical barriers due to its potential immunogenicity and limited payload capacity. Here, we demonstrate genome engineering in postnatal mice using AAV-split-Cas9, a multi-functional platform customizable for genome-editing, transcriptional regulation, and other previously impracticable AAV-CRISPR-Cas9 applications. We identify crucial parameters that impact efficacy and clinical translation of our platform, including viral biodistribution, editing efficiencies in various organs, antigenicity, immunological reactions, and physiological outcomes. These results reveal that AAV-CRISPR-Cas9 evokes host responses with distinct cellular and molecular signatures, but unlike alternative delivery methods, does not induce extensive cellular damage in vivo. Our study provides a foundation for developing effective genome therapeutics.

Introduction

The CRISPR-Cas9 system enables programmable genetic and epigenetic manipulations^1–3 that present exciting opportunities for personalized therapeutics. As CRISPR-Cas9 approaches the clinic, efficacy and safety for the patient are paramount. With this concern in mind, many have endeavoured to deploy CRISPR-Cas9 with adeno-associated viruses (AAVs), delivery vectors that are prevalent, serologically compatible with a large fraction of the human population^4,5, and generally considered non-pathogenic. In addition, AAVs allow semi-selective tissue-tropism via local or systemic delivery⁶. We and others recently highlighted the preclinical promise of AAV-CRISPR-Cas9 for modification of inherited genetic defects in mice^7–12. Further application of AAV-CRISPR-Cas9 for modulating postnatal chromatin status or gene expression would vest profound biological control, particularly in treating diseases resulting from epigenetic alterations irresolvable by genome-editing. However, this ability has yet to be realized, in part because the large Cas9 transgenes leave little space for additional function-conferring elements within current designs^7–14 (AAV payload limit ≤ 4.7 kb). This obstacle is exacerbated with the most widely used Streptococcous pyogenes Cas9 (SpCas9, 4.2 kb), which makes packaging of even the minimum functional cassette extremely challenging^10,13,14. Hence, in this study, we sought to first establish a flexible AAV-CRISPR-Cas9 platform that enables the wide spectrum of unrealized applications in vivo. Second, we track how the host responds to our system. This is important because the exogenous nature of AAV-CRISPR-Cas9 might incite detrimental host reactions against the encoded transgenes and/or viral capsid (reviewed in ¹⁵). Understanding the host responses towards AAV-CRISPR-Cas9 would identify confounding factors that impact experimental rigor, highlight relevant considerations for clinical translation, and provide a roadmap for engineering efficient genome manipulation systems.

Results

Full activity and facile AAV packaging of split-Cas9

Of the various CRISPR-Cas9^1–3,7,16 and recently characterized CRISPR-Cpf1¹⁷ orthologs, we chose to examine SpCas9 because it possesses multiple attractive features (see Supplementary Note). Most importantly, SpCas9 has the least restrictive protospacer adjacent motif (PAM) requirement among Cas9 orthologs, and thus provides the highest density of possible target sites per given genome (Supplementary Fig. 1a). Conversely, more restrictive PAM requirements [e.g. Streptococcus thermophilus (St1)¹⁶, Neisseria meningitides (Nm)¹⁶ and Staphylococcus aureus (Sa)⁷ Cas9s; Acidaminococcus sp. (As) and Lachnospiraceae bacterium (Lb) Cpf1s¹⁷] render many target sites inaccessible. SpCas9 also adopts a bi-lobed protein architecture^18,19 [similar to Sa²⁰ and Actinomyces naeslundii (Ana)¹⁹ Cas9s, but distinct from Francisella novicida (Fn)²¹ Cas9]. We hypothesized that splitting SpCas9 at its disordered linker (V713-D718) would maintain protein folding for each lobe, allowing seamless reconstitution of full-length Cas9 (Cas9^FL) in vivo by split-intein protein trans-splicing²² (Supplementary Fig. 1b). This structure-guided design might be necessary because prior reports in cell cultures^20,23–27, each with distinct design principles, have demonstrated that splitting Cas9 resolves its unwieldy size but often interferes with Cas9 function. Here, we fused the Cas9 N-terminal lobe with the Rhodothermus marinus N-split-intein (Cas9^N) (2.5 kb) and the C-terminal lobe with C-split-intein (Cas9^C) (2.2 kb), which shortens the coding sequences below that of all known Cas9 orthologs, and liberates > 2 kb in each AAV vector for additional elements. Split-Cas9 was fully active in transfected cells, targeting the endogenous genes tested at efficiencies equivalent to Cas9^FL (Supplementary Fig. 1c and 1d). Full activity from structure-guided split-intein reconstitution²⁶ contrasts with sub-optimal activity from non-covalent heterodimerization^20,23–25, suggesting that scarless protein ligation preserves Cas9 structure and function. Next, we packaged Cas9^C-P2A-turboGFP and Cas9^N-U6-gRNAs into AAV serotype DJ (AAV-Cas9-gRNAs) (Supplementary Fig. 2a) and applied the viruses to cultured cells. AAV-Cas9-gRNAs modified target genes in differentiated myotubes, tail-tip fibroblasts, and spermatogonial cells (Supplementary Fig. 2b–f), demonstrating robustness in three distinct cell types representing proliferative and terminally differentiated cell states.

Delivery efficiency dictates genome-editing rate

To evaluate functionality in vivo, we next pseudotyped AAV-Cas9-gRNAs targeting Mstn to serotype 9 (AAV9-Cas9-gRNAs^M3+M4) and intraperitoneally injected the viruses into neonatal mice (5E11 or 4E12 vector genomes, vg) (Fig. 1a). Deep-sequencing of whole tissues from injected mice revealed a range of editing frequencies (up to 10.9%), similar to those observed in cell culture (Supplementary Fig. 1 and 2) and in single myofibers following an alternative delivery method of intramuscular DNA electroporation (Supplementary Fig. 3a). Interestingly, editing frequencies exhibited inter-tissue bias for both on-target (Mstn) and off-target (chr16:+3906202) sites (Supplementary Fig. 4a–c), which correlated strongly with the concentrations of AAV9-Cas9-gRNAs per cell (vector genomes per mouse diploid genome, vg/dg) (Fig. 1b, Supplementary Fig. 4d–f). This dose-dependency indicates that delivery efficiency dictates editing rate, and despite sufficiency in infecting most cells (Supplementary Fig. 4g and 4h), higher AAV9-Cas9-gRNAs copies per cell continues to increase editing frequencies. Next, to demarcate viral biodistribution, we tracked AAV9-Cas9-gRNAs activity at single-cell resolution, using the Ai9 mouse line²⁸ that accurately couples genomic excision of a 3×Stop cassette with tdTomato fluorescence activation (Fig. 1a, Supplementary Fig. 3b–d). Systemic delivery of AAV9-Cas9-gRNAs^TdL+TdR (5E11 or 4E12 vg), targeting sequences flanking the 3×Stop cassette, generated excision-dependent tdTomato+ cells in all examined organs (Fig. 1c, Supplementary Fig. 5 and 6). These data demonstrate that the AAV-split-Cas9 platform is functional for genome-editing of multiple organs.

Figure 1.

Postnatal genome-editing with AAV9-Cas9-gRNAs and transcriptional activation with AAV9-Cas9-VPR-gRNAs. (a) AAV9-Cas9-gRNAs targeting the endogenous Mstn gene or the 3×Stop cassette in neonatal mice. (b) Mutation frequency correlates with AAV9 transduction efficiency (Pearson’s R = 0.73, Spearman’s ρ = 0.74, P < 0.05) (n = 4 mice, 4E12 vg of AAV9-Cas9-gRNAs^M3+M4). Horizontal dashed line, sequencing error rate; vertical dashed line, qPCR false positive rate. Error bars denote s.e.m. for sequencing and qPCR replicates. (c) Fluorescent images of AAV9-Cas9-gRNAs^TdL+TdR-edited tdTomato+ cells in the indicated organs (2 upper rows, n = 3 total mice at 4E12 vg; 2 lower rows, n = 4 total mice at 4E12 vg). Gray, tdTomato. Scale bar, 5 mm. (d) AAV-Cas9-VPR-gRNAs-mediated gene activation of the indicated genes in myotubes (black) and spermatogonial cells (red). Closed dots denote single-gRNA and open circles denote dual-gRNAs (AAV-Cas9^N-gRNA:AAV-Cas9^C-VPR, 1:1). Error bars denote s.e.m. (e) AAV9-Cas9-VPR-gRNAs-mediated gene activation in adult mice (FDR = 0.05). Volcano plot shows total mRNA-sequencing of the same muscle samples used for qRT-PCR in Supplementary Figure 7e (n = 3 mice per condition).

Split-Cas9 enables AAV delivery of Cas9-fusion proteins

We next capitalized on the additional viral capacity of AAV-split-Cas9 to incorporate transcription-activator fusion domains (Cas9^C fused to the 1.6 kb tripartite VPR²⁹) for targeted upregulation of gene expression (AAV-Cas9-VPR). We further made use of the fact that nuclease-active Cas9 programmed with truncated gRNAs can bind genomic loci without inducing DNA breaks, which allows use of a single Cas9-activator fusion protein for either gene-editing or gene-activation, depending on the gRNA spacer length^30,31. However, AAV-Cas9-VPR programmed with full-length gRNAs targeting Mstn showed reduced endonucleolytic activity compared to AAV-Cas9 (Supplementary Fig. 7a). In contrast, the same nuclease-active AAV-Cas9-VPR, programmed with truncated gRNAs (14–15 nt spacers), upregulated gene expression of the target Pd-l1 (Cd274), Fst, and Cd47 genes (up to 23-, 9- and 2- fold, respectively) (Fig. 1d, Supplementary Fig. 7b and 7c). Gene activation by AAV-Cas9-VPR-gRNAs inversely correlated with the basal expression levels of the target genes, potentially circumscribed by a bottleneck in the endogenous transcriptional machinery, a trend also observed with transfected nuclease-inactive ‘dead’ Cas9^FL (dCas9)-activators plasmids^29,32. Importantly, AAV-Cas9^FLs are unable to accommodate the VPR domain fusion. Hence, split-Cas9 enables AAV delivery of engineered Cas9-fusion proteins.

Postnatal gene activation with AAV9-Cas9-VPR-gRNAs

Next, we tested Cas9-mediated transcriptional regulation in vivo. Our initial comparison of samples treated with AAV9-Cas9-VPR-gRNAs and AAV9-turboRFP (4E12 vg and 1E11 vg, respectively) against those with AAV9-turboRFP only (1E11 vg) could not be accurately analyzed amidst global transcriptome perturbations (Supplementary Fig. 7d), a host reaction which we describe in detail below. Hence, we focused on mice intramuscularly injected with equivalent dosages of AAV9-Cas9-VPR-gRNAs and AAV9-turboRFP, with only the gRNA spacer sequences varied to target different sets of genes. Compared to control mice injected with AAV9-Cas9-VPR-gRNAs targeting only Mstn and Fst, mice injected with AAV9-Cas9-VPR-gRNAs targeting Mstn, Fst, Pd-l1 and Cd47 showed modest activation of the Pd-l1 and Cd47 genes (2.1-fold and 2.3-fold, respectively, as assessed by mRNA-sequencing two weeks post-treatment) (Fig. 1e, Supplementary Fig. 7e). This result demonstrates, for the first time, postnatal transcriptional regulation with CRISPR-Cas9.

Cas9 evokes cellular immune responses while AAV9 does not

The transcriptome perturbations observed alongside on-target gene activation provide a segue into our second aim, as these differentially expressed genes (> 1000 loci) were significantly enriched for a spectrum of immunological processes (Supplementary Fig. 8, Supplementary Table 1). This result places immunogenicity of AAV-CRISPR-Cas9 as a key property that destabilizes the host system, and could conceivably impact its use in vivo. To address this imminent concern, we proceeded to dissect the immunological response to AAV9-split-Cas9, comparing it in parallel with the alternative delivery method of intramuscular DNA electroporation (encoding Cas9^FL). Regardless of delivery method, expression of Cas9 in the adult tibialis anterior muscle (Fig. 2a, Supplementary Fig. 9a) induced enlargement of the draining lymph nodes with increased cell counts (Supplementary Fig. 9b). In Cas9-expressing muscles, the frequencies of CD45+ hematopoietic cells were also significantly elevated (Fig. 2b, Supplementary Fig. 9b) and enriched around transgene-expressing myofibers (Supplementary Fig. 9c). Administering the same vectors without the Cas9 coding sequence did not elicit significant cellular infiltration/expansion, indicating a Cas9-driven immune response.

Figure 2.

AAV9 and Cas9 evoke host immune responses. (a) Intramuscular Cas9-expression via AAV9-split-Cas9 injection or plasmid-Cas9^FL electroporation. (b) Heat maps depict fold-difference of each immune cell-type fraction compared to that of vehicle-injected muscles (right column) (n = 4 mice per condition). (c) Lymphocyte TCR-β CDR3 repertoires after Cas9-exposure (n = 4 mice per condition; 6 pair-wise comparisons) (Welch’s t-test, Bonferroni corrected). (d) Clonotypic abundance of Vβ16 CDR3 CASSLDRGQDTQYF (Welch’s t-test). Numbers in parentheses denote clonotypic rank within each TCR-β CDR3 repertoire after Cas9 re-stimulation. (e) Epitope mapping by M13 phage display (all Ig subclasses). (f) Cas9 epitopes from Cas9-exposed animals (top, n = 4 DNA-electroporated; bottom, n = 4 AAV9-delivered). P-values from Wald test, Benjamini-Hochberg adjusted for FDR = 0.1. (g) Capsid epitopes from AAV9-exposed animals (n = 8). Counts denote number of animals with capsid-specific antibodies covering each amino acid position (x-axis), and red bars denote positions of immunodominant epitopes. AAV9 capsid expresses as three isoforms (VP1/2/3).

Within the expanded CD45+ population, myeloid cells (CD11b+Gr1-monocytes/macrophages/dendritic cell subsets) and T-cells (CD3+CD4+ and CD3+CD8+) were particularly enriched (Fig. 2b, Supplementary Fig. 9d–f). T-cells orchestrate antigen-specific immune responses, with each clonal lineage expressing a unique T-cell receptor β-chain (TCR-β) CDR3 motif that mediates most of the antigen-contact. The unique “barcodes” of T-cell clonotypes allow us to identify Cas9-responsive T-cells through deep-sequencing of the TCR-β repertoires. We observed that Cas9-exposure skewed expansion of T-cell clonotypic subsets (Fig. 2c), implying antigen-specific T-cell activation and proliferation. Four TCR-β clonotypes were common to all Cas9-exposed animals (n = 4) and undetected in all unexposed animals (n = 8). Because bona fide T-cells would proliferate with antigen recall, we sought to confirm antigen-specific T-cell expansion by challenging extracted lymphocytes with purified Cas9 protein. Of the four initial clonotypes, we identified one (Vβ16, CDR3: CASSLDRGQDTQYF) as a true Cas9-responsive T-cell clonotype, which proliferated according to Cas9 protein re-stimulation (Fig. 2d). Hence, Cas9 activates a cellular immune response, with at least one antigen-specific T-cell clonotype common among injected mice (suggestive of a public response), and with the other infiltrating T-cells largely dissimilar between individuals (private response).

Cas9 evokes humoral immune responses

Cas9-specific antibodies were also elicited post-exposure, as confirmed by ELISA (Supplementary Fig. 10a). Individual mice exhibited variable titers of Cas9-specific antibodies, suggesting a largely private humoral response. To map Cas9 epitopes, serum from each animal was co-incubated with M13 phage display libraries tiling the Cas9 transgene, and antibody targets were determined by Ig:phage pull-down (Fig. 2e, Supplementary Fig. 10b). Epitope-mapping shows that individual Cas9-exposed animals exhibited an antibody repertoire targeting unique residues of Cas9; but three linear epitopes were observed more than once (Fig. 2f, Supplementary Fig. 10c, Supplementary Table 2). 1352-ITGLYETRI-1360 consists of residues recognizing gRNA stem loop 2¹⁸; 122-IVDEVAYHEKYP-133 resides in the REC1 domain also contributing to Cas9:gRNA interactions, but does not include residues mediating the contact¹⁸. 1126-WDPKKYGGFD-1135 resides in the PAM-binding loop and contains conserved residues, but maintains Cas9 endonucleolytic function when selectively mutated (1125-DWD→AAA¹⁹, or D1135E³³ for increasing Cas9 specificity). Combining these residue changes retains Cas9 activity (Supplementary Fig. 10d). While speculative, future large-scale functional variant profiling could reveal more residues amenable to epitope-recoding.

AAV9 evokes humoral immune responses

In contrast to Cas9, AAV9 elicited capsid-specific antibodies (Supplementary Fig. 10e) against epitopes that were shared among injected animals at surprisingly high degrees (Fig. 2g, Supplementary Table 2), reminiscent of a public response to viruses recently observed also in humans³⁴. Epitope-mapping provides intriguing support that AAV9 antigenicity derives from biophysical and functional aspects, instead of purely sequence-level motifs. The metastable VP1 unique and VP1/2 common regions are antigenic, suggesting their externalization from the viral interior for antigen capture. Immunodominant epitopes in VP3 are overrepresented on the capsid surface (Supplementary Fig. 10f). Notably, while many of these residues can be separately double-alanine mutated without disrupting viral assembly³⁵, they are predominantly implicated³⁵ in maintaining viral blood persistency (Supplementary Fig. 10g) and liver-selective tropism (Supplementary Fig. 10h). Together, the immunodominant VP3 epitopes (372-FMIPQYGYLTLNDGSQAVG-390, 436-MNPLIDQYLY-445, and 494-TQNNNSEFAWPG-505) cover 12 out of the 18 residues associated with AAV9 hepatotropism. Hence, AAV9 elicits humoral immunity that overlaps among animals, across substantial regions of the capsid protein that modulate viral biodistribution.

Transcriptome signatures following AAV-CRISPR-Cas9 exposure

Having established that AAV-CRISPR-Cas9 activates host immunity, we next investigated the functional impact. We returned to the total mRNA-sequencing data, and by using the DCQ algorithm³⁶ against reference gene expression profiles derived by the Immunological Genome Project (ImmGen), deconvoluted the admixture transcriptomes into > 200 immune cell types of the hematopoietic lineage tree (Fig. 3a, Supplementary Fig. 11, Supplementary Table 3). This deconvolution identified cell-type compositions concordant with bulk fractions quantified via FACS (Fig. 3b), but affords higher inferential power for generating testable hypotheses. Firstly, gene signatures for differentiated B-cells, monocytes and dendritic cells were overrepresented in the draining lymph nodes (Supplementary Fig. 11a), consistent with the observed humoral response and expected physiological niche. Secondly, in muscles treated with AAV9-Cas9-VPR-gRNAs, gene signatures for subsets of dendritic cells and T-cells were dominant (Fig. 3a).

Figure 3.

Immunological phenotypes towards AAV-CRISPR-Cas9. (a) Deconvoluted hematopoietic lineage tree from muscle samples. Node sizes scale with fold-differences in gene signatures induced by AAV9-Cas9-VPR-gRNAs^{set 2} treatment (n = 3 mice per condition). Detailed annotations are available in Supplementary Figure 11. (b) Cellular compositions determined by deconvolution reflect fractions quantified by FACS. Two groups of AAV9-Cas9-VPR-gRNAs-injected mice were compared to AAV9-turboRFP-injected negative control mice (n = 3 mice per condition). Error bars denote within-group s.e.m. (c) Tissue sections immunostained for the indicated proteins in muscles injected with AAV9-Cas9-VPR-gRNAs (4E12 vg) (n = 3 mice per condition). All injections included 1E11 vg of AAV9-turboRFP to demarcate transduction. Additional data are presented in Supplementary Figure 12. (d) Quantification of the fraction of centrally nucleated myofibers (n = 3 mice per condition) (one-way ANOVA, followed by Dunnett’s test against mice injected with 1E11 vg of AAV9-turboRFP).

The presence of CD8+ T-cells in Cas9-expressing tissues is particularly intriguing, because mature antigen-specific CD8+ T-cells can induce tissue damage by cytolysis; however, from our dataset, we detected gene signatures that assigned the CD8+ lineage predominantly in the naïve or activated yet immature cell fates (Fig. 3a). A closer look at genes encoding key cytolytic T-cell differentiation signals (e.g. IL-12, Ifn-α/β, IL-2, Tbx21, Eomes) and cytolytic effector proteins (e.g. Prf1, Gzmb, Fasl) revealed that these genes were not altered at statistically significant levels two weeks post-treatment (Supplementary Table 1). This led us to examine functional readouts more sensitively by intra-tissue immunofluorescence and histology.

AAV-CRISPR-Cas9 does not evoke extensive cellular damage

Interleukin-2 (IL-2) is pivotal for cytolytic T-cell differentiation³⁷. Downstream, perforin (Prf1) is the essential pore-forming protein released by cytolytic T lymphocytes and NK cells to destroy target cells. In line with mRNA-sequencing, immunofluorescence indicated basal IL-2 and perforin protein levels within AAV9-Cas9-VPR-gRNAs-treated muscles, except for sparse interstitial areas that correspond to the highest density of cellular infiltrates (Fig. 3c, Supplementary Fig. 12). The unremarkable perforin levels suggest minimal downstream cytolysis. To quantitatively assess cellular damage, we quantified the fraction of centrally nucleated myofibers within muscle histological sections, a physiological measure indicative of myofiber degeneration and repair. Through this assay, we did not observe significant muscle cellular damage and repair responses at two weeks after administration of AAV9-Cas9-VPR-gRNAs (Fig. 3d). As positive controls for our assays, we examined muscles electroporated with DNA encoding Cas9^FL. In contrast, IL-2 and perforin levels were both strongly increased in muscles two weeks after electroporation of Cas9-encoding DNA, and significant myofiber degeneration-repair was observed (Supplementary Fig. 13 and 14a). These observations following DNA electroporation could not simply be attributed to physical disruptions, because transgene-induced intramuscular CD45+ cellular infiltration, IgG and IgM antibody levels, IL-2 and perforin levels, and myofiber degeneration-repair (Supplementary Fig. 13 and 14a–d) were significantly elevated over vehicle-electroporated control mice, and could be partially reduced by immunosuppression (FK506). Hence, our assays, which are adequately sensitive for detecting the extensive myofiber cytolysis evoked by electroporation of Cas9-encoding DNA, did not indicate significant myofiber damage and replacement after AAV9-Cas9-VPR-gRNAs treatment.

Discussion

Research use and eventual clinical implementation of CRISPR-Cas9 require comprehensive evaluation of its in vivo profile, including the potential to evoke deleterious host responses. Delivery methods such as DNA electroporation and adenoviruses³⁸ cause severe inflammation and immunological reactions. Immunosuppression could arguably mitigate some of these detrimental consequences, but nonetheless is still associated with well-appreciated side-effects (Supplementary Fig. 14e). More immunologically inert delivery vectors are hence particularly attractive. Precedents of immune evasion/tolerance induction by viruses have been documented in natural and experimental settings^15,39,40, which have fuelled ongoing efforts towards understanding and reducing the risk profiles of therapeutic vectors. Here, we contribute by directly interrogating the in vivo profile of AAV-CRISPR-Cas9, specifically that of our platform that allows flexible incorporation of function-conferring domain fusions that AAV-Cas9^FLs cannot accommodate in their current forms^7–14 (Supplementary Note). We established on multiple fronts that AAV-CRISPR-Cas9 activates the immune system, but had difficulty detecting muscle cellular damage and repair responses with adequately sensitive assays in the time frame of our experiments. We cannot rule out that immunological cytotoxicity could be tissue context-specific, but this is unlikely to affect our conclusion because prior reports suggest that the muscle is more susceptible to cytolytic responses than organs such as the liver¹⁵. Another caveat is our use of wild type mice for platform characterization, animals with immunological features that may differ in important ways from those of human subjects, especially in the inflammatory and pro-cytotoxic disease contexts where AAV-CRISPR-Cas9 therapeutics might most readily be employed⁴¹. Interestingly, prior reports have shown that delivery of other transgenes with AAV also activates CD8+ T-cells lacking full functionality, with these transgene-specific CD8+ T-cells exhibiting impaired cytolytic activity⁴² or cell viability⁴³. It is here that further investigation towards bridging these two implications of our system – immune activation and undetected cellular damage – could reveal elements that may illuminate ways to further minimize its immunogenicity. In addition, whether the host immune response to AAV-CRISPR-Cas9 may impact other aspects of its therapeutic implementation, such as the effects towards on- and off-target gene modification and activity, remains an important topic for future investigation. In subsequent studies, the Cas9-reponsive T-cell clonotype identified here could serve as a distinctive biomarker for Cas9-specific immunity. On a parallel front, additional targeted engineering of AAV-CRISPR-Cas9 could be highly fruitful in defusing unintended side-effects that remain prominent, bringing us towards a genome manipulation tool that is efficacious on-target and inert otherwise. Cellular and molecular dissection of the host responses evoked by AAV-CRISPR-Cas9, as reported here, should aid in such efforts.

Methods

Methods and any associated references are available in the online version of the paper.

Online Methods

Constructs and sequences

U6-driven gRNA plasmids were constructed as described¹. AAV plasmid backbone was derived from pZac2.1-CASI-EGFP-RGB, a gift from Luk Vandenberghe. Minicircles parental plasmids were cloned in ZYCY10P3S2T, and minicircles were generated as described⁴⁴. AAV plasmids were cloned in Stbl3 (Life Technologies C7373-03). All other plasmids were cloned in DH5α (NEB C2987H). Protein transgenes were expressed from ubiquitous hybrid promoters: SMVP promoter (generated by fusing SV40enhancer-CMV-promoter-chimeric intron), CASI promoter⁴⁵, or CAG promoter⁴⁶. SMVP plasmid was derived from pMAXGFP (Lonza). pCAG-GFP was a gift from Connie Cepko (Addgene plasmid # 11150). pAAV-CMV-HI-EGFP-Cre-WPRE-SV40pA was obtained from the University of Pennsylvania Vector Core. Sequences of gRNA spacers, genotyping primers, qPCR probes and primers are stated in Supplementary Table 4.

AAV packaging and purification

AAVs were packaged via the triple-transfection method^47,48. HEK293 cells (Cell Biolabs AAV-100 or Agilent 240073) were plated in growth media consisting of DMEM+glutaMAX+pyruvate+10%FBS (Life Technologies), supplemented with 1× MEM non-essential amino acids (Gibco). Confluency at transfection was between 70–90%. Media was replaced with fresh pre-warmed growth media before transfection. For each 15-cm dish, 20 μg of pHelper (Cell Biolabs), 10 μg of pRepCap [encoding capsid proteins for AAV-DJ (Cell Biolabs) or AAV9 (UPenn Vector Core)], and 10 μg of pAAV were mixed in 500 μl of DMEM, and 200 μg of PEI “MAX” (Polysciences) (40 kDa, 1 mg/ml in H₂O, pH 7.1) added for PEI:DNA mass ratio of 5:1. The mixture was incubated for 15 min., and transferred drop-wise to the cell media. For large-scale AAV production, HYPERFlask ‘M’ (Corning) were used, and the transfection mixture consisted of 200 μg of pHelper, 100 μg of pRepCap, 100 μg of pAAV, and 2 mg of PEIMAX. The day after transfection, media was changed to DMEM+glutamax+pyruvate+2%FBS. Cells were harvested 48–72 hrs after transfection by scrapping or dissociation with 1×PBS (pH7.2) + 5 mM EDTA, and pelleted at 1500 g for 12 min. Cell pellets were resuspended in 1–5 ml of lysis buffer (Tris HCl pH 7.5 + 2 mM MgCl + 150 mM NaCl), and freeze-thawed 3× between dry-ice-ethanol bath and 37 °C water bath. Cell debris was clarified via 4000 g for 5 min., and the supernatant collected. Downstream processing differed depending on applications.

For preparation of AAV-containing lysates, the collected supernatant was treated with 50 U/ml of Benzonase (Sigma-Aldrich) and 1 U/ml of Riboshredder (Epicentre) for 30 min. at 37 °C to remove unpackaged nucleic acids, filtered through a 0.45 μm PVDF filter (Millipore), and used directly on cells or stored in −80 °C.

For purification of AAV via chloroform-ammonium sulfate precipitation, 1/10^th volume of chloroform and NaCl (1 M final concentration) was added to the lysate and shaken vigorously. After centrifugation, the supernatant was incubated with PEG-8000 (10% final w/v) on ice for ≥ 1 hr or overnight. PEG-precipitated virions were centrifuged (4000 g, 30 min., 4 °C), and resuspended in 50 mM HEPES buffer (pH8). 50 U/ml of Benzonase (Sigma-Aldrich) and 1 U/ml of Riboshredder (Epicentre) were added and incubated for 30 min. at 37 °C. An equal volume of chloroform was then added, and vigorously vortexed. After centrifugation, the aqueous phase was collected and residual chloroform evaporated for 30 min. Ammonium sulfate precipitation of AAVs were performed with a 0.5 M – 2 M cut-off. AAVs were then resuspended and dialyzed in 1× PBS + 35 mM NaCl, quantified for viral titers, and stored in −80 °C.

All experiments with purified AAVs utilized the iodixanol density gradient ultracentrifugation purification method^47,48, unless otherwise stated. The collected AAV supernatant was first treated with 50 U/ml Benzonase and 1 U/ml Riboshredder for 30 min. at 37 °C. After incubation, the lysate was concentrated to <3 ml by ultrafiltration with Amicon Ultra-15 (50 kDa MWCO) (Millipore), and loaded on top of a discontinuous density gradient consisting of 2 ml each of 15%, 25%, 40%, 60% Optiprep (Sigma-Aldrich) in an 11.2 ml Optiseal polypropylene tube (Beckman-Coulter). The tubes were ultracentrifuged at 58000 rpm, at 18 °C, for 1.5 hr, on an NVT65 rotor. The 40% fraction was extracted, and dialyzed with 1×PBS (pH 7.2) supplemented with 35 mM NaCl, using Amicon Ultra-15 (50 kDa or 100 kDa MWCO) (Millipore). The purified AAVs were quantified for viral titers, and stored in −80 °C.

AAV2/9-CMV-HI-EGFP-Cre-WPRE-SV40 (Lot V4565MI-R), AAV2/9-CB7-CI-EGFP-WPRE-rBG [Lot CS0516(293)], AAV2/9-CB7-CI-mCherry-WPRE-rBG (Lot V4571MI-R), and AAV2/9-CMV-turboRFP-WPRE-rBG (Lot V4528MI-R-DL) were obtained from the University of Pennsylvania Vector Core.

AAV titers (vector genomes) were quantified via hydrolysis-probe qPCR⁴⁹ against standard curves generated from linearized parental AAV plasmids.

Cell culture transfection and transduction

All cells were incubated at 37 °C and 5% CO₂.

C2C12 cells were obtained from the American Tissue Collection Center (ATCC, Manassas, VA), and grown in growth media (DMEM+glutaMAX+10% FBS). Cells were split with TypLE Express (Invitrogen) every 2–3 days and before reaching 80% confluency, to prevent terminal differentiation. Passage number was kept below 15. For transfection of C2C12 myoblasts, 10⁵ cells were plated per well in a 24-well plate, in 500 μl of growth media. The following day, fresh media was replaced, and 800ng of total plasmid DNA was transfected with 2.4 μl of Lipofectamine 2000 (Life Technologies). 1:1 mass ratio of vectors encoding Cas9:gRNA(s) was used. Media was replaced with differentiation media (DMEM+glutaMAX+2% donor horse serum) on the 1^st and 3^rd days post-lipofection.

For differentiation of C2C12 into myotubes, 2×10⁴ cells were plated per well in a 96-well plate, in 100 μl of growth media. At confluency, 1–2 day(s) after plating, media was replaced with fresh differentiation media (DMEM+glutaMAX+2% donor horse serum), and further incubated for 4 days. Fresh differentiation media was replaced before transduction with AAVs. Culture media was replaced with fresh differentiation media 1d after transduction, and cells incubated for stated durations.

The 3×Stop-tdTomato reporter cell line was derived from tail-tip fibroblasts of Ai9²⁸ mouse (JAX 007905), and immortalized with lentiviruses encoding the large SV40 T-antigen (GenTarget Inc, LVP016-Puro). Cells were cultured in DMEM+pyruvate+glutaMAX+10%FBS. Lipofectamine 2000 (Life Technologies) was used for transfection of plasmids, and images were taken 5 days after transfection. For transduction with AAVs, cells were plated at 2×10⁴ per well in a 96-well plate, in 100 μl of growth media. AAV-containing lysates or purified AAVs were applied at confluency of 70–90%. Culture media was replaced with fresh growth media the next day, and cells incubated for stated durations.

The GC-1 spg mouse spermatogonial cell line (CRL-2053) was obtained from ATCC. Cells were cultured and transduced similarly to the 3×Stop-tdTomato cell line, with a Cas9^N:Cas9^C of 1:1.

Cell lines were not tested for mycoplasma contamination after commercial purchase or primary isolation.

Animals

All animal procedures were approved by the Harvard University Institutional Animal Care and Use Committee.

Ai9²⁸ mice (JAX No. 007905) were used for tdTomato activation, and for systemic AAV9-Cas9-gRNAs and AAV9-GFP-Cre experiments, randomized at approximately 50:50 male:female ratios, with sex of each animal depicted in the figures. C57BL/6 male mice were used for in vivo electroporation and intramuscular AAV injections.

Standard sample sizes were chosen for all experiments. Exact sample sizes for experiments are stated in figure legends and text. Statistical methods were not used for pre-determination of power. All animals were randomly allocated to treatment and control groups and handled equally.

In vivo electroporation

Animals were anesthetized using isoflurane and injected with 50 μl of 2mg/ml hyaluronidase (Sigma-Aldrich, H4272) in the tibialis anterior muscle. After 1 hr, plasmids in vehicle (10 mM Tris-HCl pH 8.0) were injected into the muscle, followed by electroporation⁵⁰ (10 pulses of 20 ms at 100 V/cm with 100 ms intervals) using an ECM 830 Electro Square Porator (BTX Harvard Apparatus) and a two-needle array.

For Ai9 reporter activation, 30 μg of pSMVP-SpCas9^FL, 60 μg total of pCRII-U6-gRNA, and 15 μg of pCAG-GFP were used for CRISPR-Cas9 injections; 30 μg of pSMVP-Cre and 15 μg of pCAG-GFP were used for Cre injections; and 30 μg of pSMVP-SpCas9^FL and 15 μg of pCAG-GFP were used for no-gRNA control injections. FK506 (Sigma-Aldrich, F4679) was administered daily at 5mg/kg (body weight), commencing 1 day before electroporation.

For C57BL/6 gene-targeting, 30 μg of vectors encoding SpCas9^FL (pSMVP-SpCas9 or minicircle-SMVP-SpCas9), 30 μg total of pCRII-U6-gRNA, and 15 μg of pCAG-GFP were used. With co-translating P2A-turboGFP, pCAG-GFP was omitted, and 30 μg pSMVP-SpCas9-P2A-turboGFP and 30 μg total of pCRII-U6-gRNA were used.

For immunogenicity comparison to AAV-split-Cas9, both tibialis anterior muscles of 11-week old male C57BL/6 mice were each electroporated with 30 μg of pSMVP-Cas9^FL. Control mice were electroporated with 30 μg of plasmid vector control (consisting of the same plasmid with Cas9 coding sequence removed) per muscle. Vehicle electroporations were similarly performed. Animals were euthanized via CO₂ asphyxia and cervical dislocation 2 weeks after treatment.

Single myofiber isolation

Mice were euthanized via CO₂ asphyxia and cervical dislocation 10 days after in vivo electroporation. Dissected muscles were digested with 0.2% collagenase type II in DMEM for 50 min. at 37 °C. Muscles were triturated with fire-polished Pasteur pipettes and single GFP+ myofibers isolated under a SMZ1500 (Nikon) fluorescence dissection stereomicroscope.

Genotyping and analysis

C2C12 cells were harvested 4 days post-lipofection, with 100 μl of QuickExtract DNA Extraction Solution (Epicentre) per well of a 24-well plate; and C2C12 myotubes were harvested 7 days post-AAV transduction, with 20 μl of DNA QuickExtract per well of a 96-well plate. Cell lysates were heated at 65 °C for 10 min., 95 °C for 8 min., and stored at −20 °C. Each locus was amplified from 0.5 μl of cell culture lysate per 25 μl PCR reaction, for 20–25 cycles.

Single myofibers were each placed in 10 μl of QuickExtract DNA Extraction Solution, and heated at 65 °C for 15 min., 95 °C for 10 min. For genotyping, the entire lysate was used in a 100 μl PCR reaction, and thermocycled for 30–35 cycles due to low genomic copy numbers.

Bulk tissues were each placed in 100 μl of QuickExtract DNA Extraction Solution, and heated at 65 °C for 15 min., 95 °C for 10 min. 0.5 μl of lysate was used per 25 μl PCR reaction, and thermocycled for 25 cycles.

For barcoding for deep sequencing, 1 μl of each unpurified PCR reaction was added to 20 μl of barcoding PCR reaction, and thermocycled [95 °C for 3 min., and 10 cycles of (95 °C for 10 s, 72 °C for 65 s)]. Amplicons were pooled, the whole sequencing library purified with self-made SPRI beads (9% PEG final concentration), and sequenced on a Miseq (Illumina) for 2×251 cycles. FASTQ were analyzed with BLAT (with parameters -t=dna -q=dna -tileSize=11 -stepSize=5 -oneOff=1 -repMatch=10000000 -minMatch=4 -minIdentity=90 -maxGap=3 -noHead) and post-alignment analyses performed with MATLAB (MathWorks). Alignments due to primer dimers were excluded by filtering off sequence alignments that did not extend >2 bp into the loci from the locus-specific primers. To minimize the impact of sequencing errors, conservative variant calling was performed by ignoring base substitutions, and calling only variants that overlap with a ±30 bp window from the designated Cas9-gRNA cut sites. Negative controls were equally analyzed for baseline sequencing error rates, to which statistical tests were performed against.

Off-target sites for Mstn gRNAs were predicted using the online CRISPR Design Tool⁵¹ (http://crispr.mit.edu/). Off-target sites were ranked by number of mismatches to the on-target sequence, and deep sequencing performed on top hits. Sequencing reads were analyzed equally between experimental samples (AAV9-Cas9-gRNAs^M3+M4) and control samples (AAV9-Cas9-gRNAs^TdL+TdR) using BLAT. Variant calls were performed for insertions and deletions that lie within a ±15 bp window from potential off-target cut-sites.

Quantitative reverse-transcription PCR (qRT-PCR) for gene expression

Cells were processed with Taqman Cells-to-Ct kits (Thermo Fisher Scientific #4399002) as per manufacturer’s instructions, with the modification that each qRT-PCR reaction was scaled down to 25 μl. Taqman hydrolysis probes (Thermo Fisher Scientific) used: Pd-l1 (Mm00452054_m1), Fst (Mm00514982_m1), Cd47 (Mm00495011_m1), and house-keeping gene Abl1 (Mm00802029_m1). Gene expression from targeted genes was normalized to that of Abl1 (ΔCt). For cell culture experiments, fold-changes were calculated against AAV-Cas9^C-VPR-only controls (no-gRNA) (2^−ΔΔCt). Basal gene expression percentiles for C2C12 myotubes and GC-1 spermatogonial cells (type B spermatogonia) were retrieved from the Gene Expression Omnibus (GEO) repository (GDS2412 and GDS2390 respectively).

Total RNA from skeletal muscle tissues was extracted via TRIzol. Reverse transcription was conducted with High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems #4368814), and 5 μl of each reaction was used for qRT-PCR in 1× FastStart Essential DNA Probes Master (Roche #06402682001). Gene expression from targeted genes were normalized to that of Abl1 (ΔCt) and fold-changes were calculated against AAV9-turboRFP-only controls (2^−ΔΔCt).

AAV administration in mice

All AAV experiments in vivo were conducted in a randomized and double-blind fashion. The allocation code was unblinded only after analyses were completed. Injections utilized AAV9-Cas9^N-gRNAs:AAV9-Cas9^C-P2A-turboGFP and AAV9-Cas9^N-gRNAs:AAV9-Cas9^C-VPR ratios of 1:1.

3-day old neonates were each intraperitoneally injected with 4E12, 5E11, or 2.5E11 vector genomes (vg) of total AAV9. Vector volumes were kept at 100 μl. Animals were euthanized via CO₂ asphyxia and cervical dislocation 3 weeks following injections. For AAV9-GFP and AAV9-mCherry co-transduction experiments, animals were euthanized 9 days after injection. For qPCR and deep sequencing of whole tissues, samples were taken from the heart body wall, liver, gastrocnemius muscle, olfactory bulb, ovary, testis, and diaphragm.

AAV9-Cas9-VPR-gRNAs were intramuscularly injected at Cas9^N-gRNAs:Cas9^C-VPR ratio of 1:1, at a total of 4E12 vg. To demarcate transduced tissues, 1E11 of AAV9-turboRFP was co-administered in the same mix. Controls mice were injected with 1E11 of AAV9-turboRFP only, with the final injection mix at the same volume. Animals were euthanized via CO₂ asphyxia and cervical dislocation 2 weeks after AAV9-Cas9-VPR-gRNAs injection.

For determining AAV- and Cas9- specific immune responses, both tibialis anterior muscles of 11-week old male C57BL/6 mice were each injected with AAV9-Cas9^N and AAV9-Cas9^C (2E12 vg each). For control mice, 4E12 of AAV9 vector control (consisting of the same AAV genome with the Cas9 coding sequence removed) was injected per muscle. Vehicle injections were similarly performed. Animals were euthanized via CO₂ asphyxia and cervical dislocation 2 weeks after treatment.

Quantitative PCR (qPCR) for AAV genomic copies in tissues

Each qPCR reaction consists of 1× FastStart Essential DNA Probes Master (Roche #06402682001), 100nM of each hydrolysis probe (against the AAV ITR and the mouse Acvr2b locus), 340nM of AAV ITR reverse primer, 100nM each for all other forward and reverse primers, and 2.5 μl of input tissue lysate. A mastermix was first constituted before splitting 22.5 μl into each well, after which tissue lysates were added. Thermocycling conditions were: [95°C 15 min.; 40 cycles of (95°C 1 min., 60°C 1 min.)]. FAM and HEX fluorescence were taken every cycle. AAV genomic copies per mouse diploid genome were calculated against standard curves. For each tissue sample, two repeated samplings were performed for qPCR and deep-sequencing, all on separate days, and the means plotted with s.e.m. qPCR false positive rate were calculated similarly from two vehicle-injected negative control mice, to which statistical tests were performed against.

Immune cells profiling by FACS

Muscles were harvested two weeks after injections, chopped, and digested with 0.2% collagenase and 0.05% dispase in DMEM for 15 min. Mononuclear cells were separated from the tissue debris using 70 μm nylon sieves, and resuspended in Fc blocking solution (1:50) (BD Pharmingen) in staining media (1 mM EDTA and 10% FBS in PBS) for 30 min. on ice. For experiments using AAV9-GFP as transduction reporter, immune cells were stained with anti-CD45 APC/Cy7 (1:125) (BioLegend 103115), anti-CD11b APC (1:125) (BioLegend 101211), anti-Ly6G/Gr1 Pacific Blue (1:200) (BioLegend 127611), anti-CD3 PE-Cy7 (1:50) (BioLegend 100220), anti-CD8a PE (1:50) (BD Pharmingen 553033), anti-CD4 PerCP (1:100) (BD Pharmingen 561090), anti-CD19 PE-CF594 (1:200) (BD clone 1D3 562291) and 7AAD (1:20) (eBiosciences 420403) for 20 min. on ice. For experiments using AAV9-turboRFP as transduction reporter, all antibodies were the same except that anti-CD8a-FITC (BD clone 53-6.7 553031) was substituted. Flow cytometry was performed on a BD LSR II, provided by the Harvard Stem Cell Institute Flow Cytometry Core. Flow cytometry data was collected using FACSDiva software (Becton Dickinson) and analyzed using FlowJo (Tree Star) and MATLAB (MathWorks).

Cas9 re-stimulation and TCR-β repertoire sequencing

Lymphocytes were isolated from both inguinal lymph nodes and one popliteal lymph node per bilaterally injected mouse. Lymph nodes were cut, and incubated for 30 min. in RPMI + 1 mg/ml collagenase at RT. Lymphocytes were released by meshing through 70 μm nylon sieves, washed twice with 1× PBS+5 mM EDTA, and resuspended in 500 μl of growth media (RPMI 1640+10% FBS+1×Pen/Strep/AmphoB+50 μM 2-βME). Cell counting was performed on a Countess device (Life Technologies). For Cas9 restimulation experiments, 2.5 μg of recombinant Cas9 (NEB) was incubated with > 10⁶ cells in 500 μl of growth media for 3 days. Experiments were conducted in parallel with wells containing cells without re-stimulation (i.e. without Cas9 protein). Cellular RNA was extracted using QIAshredder and RNeasy micro (Qiagen). Reverse transcription was performed with SMARTscribe (Clontech), using SMARTNNN template-switching adapter as described⁵². KAPA HiFi polymerase was used for PCR. Individual RNA molecules were counted based on Unique Molecular Identifiers using MIGEC⁵³, aligned with MIXCR⁵⁴, and post-analysis performed with MATLAB (MathWorks). Morisita-Horn indices per exposure condition were calculated by pairwise comparisons among 4 mice (2 animals from electroporation dataset and 2 animals from AAV dataset).

ELISA

A modified protocol adapted from Wang et al., 2015³⁸ was used. Briefly, recombinant SpCas9 protein (PNA Bio, cat # CP01) was diluted in 1× coating buffer (KPL) and 0.5 μg was used to coat each well of 96-well Nunc MaxiSorp Plates (Thermo Scientific) overnight at 4°C. Plates were washed 3 × 5 min. with 1× wash buffer (KPL) and blocked with 1% BSA Blocking Solution (KPL) for 1 hr at RT. Standard curves were generated using a mouse monoclonal antibody against SpCas9 (Epigentek, clone 7A9, cat # A-9000-100), with final concentrations of 0, 1, 2.5, 5, 10, 25, 50, 100 ng ml⁻¹ and in triplicates. Serum samples were added at 1:40 dilution and plates were incubated for 5 hrs at 4°C with shaking. Wells were washed 3 × 5 min., and 100 μl of HRP-labelled goat anti-mouse IgG1 (Santa Cruz; diluted 1:4000 in 1% BSA Blocking Solution) was added to each. After incubating for 1 hr at RT., wells were washed 4 × 5 min., and 100 μl of ABTS ELISA HRP Substrate (KPL) was added to each. Optical density (OD) at 410 nm was measured using a plate reader (BioTek).

Fluorescent immunoassay

To determine antibody specificity and class-switching, serum levels of AAV9-specific IgM, IgG, and IgG2a from AAV9-treated mice were compared to that from vehicle-injected control mice. AAV9 viruses (1E9 vg) were coated on each well of a 96-well PVDF MaxiSorp plate for 1 hr in 1× TBST, followed by 1 hr of blocking in 1× TBST + 3% BSA. After 3 washes with 1× TBST, 1:100 diluted mouse serum was applied at 25 μl per well, for 1 hr. After 3 washes with 1× TBST, 1:200 diluted anti-mouse secondary antibodies were added [goat anti-mouse IgG-CF633 (Biotium 20120), goat anti-mouse IgG2a-CF594 (Biotium 20259) and goat anti-mouse IgM-Dy550 (Pierce PISA510151)], and incubated for 1 hr. Wells were then washed 5× with 1× TBST, and fluorescence readings in 100 μl of 1× TBS were taken via a plate reader. All steps were conducted at RT. To account for autofluorescence from sera, fluorescence readings were normalized against that from wells treated similarly except for the exclusion of secondary antibodies.

Epitope mapping by M13 phage display

M13KE genome was amplified by PCR, with one end terminating with the pIII peptidase cleavage signal, and the other end terminating with a 4×Gly linker followed by the mature pIII. Cas9 and AAV9 VP1 capsid coding sequence PCR products were each randomly fragmented with NEBNext dsDNA Fragmentase until about 50–300 bp. Purified fragments were treated with NEBNext End-Repair Module. After DNA purification, fragments were blunted ligated into the M13KE PCR product overnight at 16 °C. The entire ligation reaction was purified, and transformed into ER2738 (Lucigen), at 200 ng per 25 μl of bacteria, with electroporation conditions of 10 μF, 600 Ω, and 1.8kV. After 30 min. recovery in SOC media, the culture was amplified by combining with 20 ml of early-log ER2738 culture. After 4 hrs, the culture supernatant was collected, and incubated to a final concentration of 3.33% PEG-8000 and 417 mM of NaCl, overnight at 4 °C. M13 phage was pelleted, and resuspended in 2 ml of TBS. Phage titers were determined by LB/IPTG/X-gal blue-white plague counting, averaging > 1E11 pfu / μl. For Ig:phage pull-down, 20 μl of each phage library was incubated with 5 μl of mouse serum or titrated amount of purified antibody controls [7A9 (Novus Bio), Guide-IT (Clontech), bG15 (Santa Cruz), bS18 (Santa Cruz), bD20 (Santa Cruz), non-binding mouse IgG isotype control (Santa Cruz)], and made up to 50 μl with TBST, for 1 hr at RT. For each reaction, 25 μl of Protein A/G magnetic beads (Millipore PureProteome) was first washed twice with TBST, resuspended to 10 μl, and added to the reaction for additional 30 min. incubation. The beads were then washed 5× with TBST, and captured Ig:phage eluted with 100 μl of 200 mM glycine-HCl, pH 2.2, 1 mg/ml BSA for 8 min. The eluant was neutralized with 15 μl of 1 M Tris-HCl, pH 8.5. 5 μl of captured phage display eluant was used per PCR reaction, with 20 cycles of spacer amplification and 10 cycles of barcoding, and sequenced on a Miseq (Illumina). Each serum sample was processed for technical replication on separate days. Differential binding of phage was determined using DESeq2⁵⁵ in R⁵⁶, with all Cas9-unexposed (n = 16 animals) or AAV-unexposed (n = 16 animals) samples as appropriate controls for non-specific binding. Alignments and further analyses were performed with MATLAB (MathWorks). Visualization of epitopes on Cas9 (PBD ID: 4CMP, chain A) and AAV9 VP3 (PBD ID: 3UX1) structures was conducted with Pymol. Phenotypic data of double-alanine AAV9 mutants were obtained from ref. ³⁵, with mutant viral blood persistency calculated as the difference in blood viral levels 72 hr. and 10 min. post-injection (both normalized to that of wildtype AAV9, with 0 denoting wildtype phenotype, and negative values denoting loss of blood persistency). Mutant tropism represented by ‘Phenotypic Difference’ values as described³⁵. Solvent-accessibility surface area (sasa) ratios for AAV9 capsid (PDB ID: 3UX1⁵⁷) were first calculated as described⁵⁸, and the final sasa ratio per residue calculated as the mean from a ± 5 bp sliding window centered on the residue.

Total mRNA-sequencing

1 μg of TRIzol-extracted RNA from muscle tissues or draining lymph node tissues were enriched for polyA-tailed mRNA and processed with NEBNext Ultra Directional RNA Library Prep Kit (New England Biolabs), followed by sequencing on a NextSeq sequencer (Illumina), giving ~30 million reads per sample. Reads were aligned to the mm10 reference genome and FPKM quantified with the Cufflinks workflow⁵⁹, with differential expression tested with Cuffdiff⁶⁰. GO-terms network was visualized with ClueGO⁶¹ Cytoscape⁶² plug-in. Immune cell reference transcriptomes were downloaded from the ImmGen database. DCQ deconvolution was conducted as described³⁶ in R⁵⁶.

Histology and immunofluorescence staining

Mouse organs and tissue samples were dissected, fixed in 4% paraformaldehyde in 1× DPBS for 1.5 hr, followed by 3× 5min. washes with 1× DPBS. Samples were immersed in 30% sucrose until submersion, embedded in O.C.T. compound (Tissue-Tek), frozen in liquid-nitrogen-cold isopentane, and cryosectioned on a Microm HM550 (Thermo Scientific). Skeletal muscles were sectioned to a thickness of 12 μm, while the liver and heart were sectioned at 20 μm.

For immunofluorescence, TA muscle sections were blocked in 1× PBST + 3% BSA for 1hr at RT, immunostained with primary antibodies at RT for 1 hr, followed by 3× washes with PBS/T. Slides were then incubated with secondary antibodies at RT for 1 hr., followed by 3× washes with PBS/T.

Anti-mouse laminin (Abcam ab14055) was used at 1:400, followed by 1:250 of secondary antibody goat anti-chicken IgY Alexa Fluor 555 (Life Technologies A-21437). Anti-CD45 APC (BioLegend 103112) was used at 1:100. Goat anti-mouse IgG-CF633 (Biotium 20120) and goat anti-mouse IgM-Dy550 (Pierce PISA510151) were used at 1:200 to detect endogenous mouse antibodies.

Anti-mouse IL-2 and perforin antibodies were used at 1:100 (Santa Cruz sc-7896 and sc-9105 respectively), followed by 1:200 of secondary anti-rabbit CF633 (Biotium). Immunostaining was conducted on three separate days, each with a mix of electroporated and AAV-injected muscle sections.

Immunostained slides were mounted with mounting media containing DAPI (Vector Laboratories, H1500).

Western blot

Muscles were harvested 2 weeks after AAV injections or plasmid electroporation. ~10 mm³ tissue clippings were flash-frozen in liquid nitrogen, followed by lysis in 300–500 μl of T-PER Tissue Protein Extraction Solution (Thermo Scientific) supplemented with 1× Complete Protease Inhibitor (Roche), and homogenized in gentleMACS M tubes (Miltenyi Biotec). 10–15 μl of each tissue lysate was ran on 8 % Bolt Bis-Tris Plus gels (Life Technologies) in 1× Bolt MOPS SDS running buffer at 165 V for 50 min. Protein transfer was performed with iBlot (Life Technologies) onto PVDF membranes, using program 3 for 13 min. Western blots were conducted with 1:200 of anti-Cas9 Guide-IT polyclonal antibody (Clontech 632607), 1:400 of anti-Gapdh polyclonal antibody (Santa Cruz sc-25778), and 1:2500 of anti-rabbit IgG-HRP secondary antibody (Santa Cruz sc-2004), using an iBind device (Life Technologies). Stained membranes were developed with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific) and imaged on Chemidoc MP (Bio-Rad). Band intensities were quantified with ImageJ (NIH).

Immunosuppression

FK506 was dissolved in 100% DMSO, and the stock solution further diluted 1:100 in vehicle for final concentrations of 1% DMSO, 10% Cremophor (Sigma-Aldrich, C5135), and 1× PBS. Mice were injected daily with 5 mg/kg (body weight) of FK506, with the first injection commencing 1 day before in vivo electroporation.

For Cas9-gRNAs-injected muscles, a mix of 30 μg of minicircle-SMVP-Cas9, 15 μg of pCAG-GFP, and 30 μg total of pCRII-U6-gRNA plasmids was injected. 30 μg of minicircle-SMVP-Cas9 was injected for Cas9-only injections, 30 μg of pCRII-U6-gRNA plasmids for gRNAs-only injections, 15 μg of pCAG-GFP for GFP-only injections. 4 mice were injected per condition.

For body weight measurements, 2-month-old mice were injected with 5mg/kg/day of FK506 or vehicle for 2 weeks. 3 male C57BL/6 mice were injected per condition.

Imaging and analyses

Confocal images were taken using a Zeiss LSM780 inverted microscope. For live cell-imaging, each image consists of 3× z-stacks (7 μm intervals) and 2×2 tiles. For muscle sections, 3× z-stacks (7 μm intervals) were used. For liver and heart sections, 4× z-stacks (10 μm intervals) were used. Tiling was used to image entire samples, followed by stitching. Stacked fluorescence images were projected by maximum intensity with Zen 2011 (Carl Zeiss).

Epifluorescence images were taken with an Axio Observer D1 (Carl Zeiss) or Axio Observer Z1 (Carl Zeiss).

For co-transduction analysis after AAV9-GFP and AAV9-mCherry administration, pixels that contain both GFP and mCherry fluorescence intensities above the background thresholds were identified, and the lower intensity values from either channel were used to populate a merged image. All other pixels in the merged image were set to null.

Centrally nucleated myofibers were quantified from whole section images with CellProfiler⁶³ and random inserts were counted manually.

Whole organ images were taken on a SMZ1500 (Nikon) fluorescence dissection stereomicroscope equipped with a SPOT RT3 camera (Diagnostic Instruments), for an imaging area of 16 mm by 12 mm, with 3 s exposure for the liver and 4 s exposure for the heart, muscle, brain and gonads. All images were acquired with a gain setting of 8 using the SPOT imaging software (Diagnostic Instruments, Sterling Heights, MI). Images for each organ were inverted and thresholded equally across animals.

Images were analyzed with ImageJ (NIH), CellProfiler⁶³ and MATLAB (MathWorks).

Availability of software and scripts

Software is cited by primary literature and available on online repositories. All custom scripts used in the study will be readily provided to any requestor.