Directed evolution of a family of AAV capsid variants enabling potent muscle-directed gene delivery across species

Abstract

Replacing or editing disease-causing mutations holds great promise for treating many human diseases. Yet, delivering therapeutic genetic modifiers to specific cells in vivo has been challenging, particularly in large, anatomically distributed tissues such as skeletal muscle. Here, we establish an in vivo strategy to evolve and stringently select capsid variants of adeno-associated viruses (AAVs) that enable potent delivery to desired tissues. Using this method, we identify a class of RGD motif-containing capsids that transduces muscle with superior efficiency and selectivity after intravenous injection in mice and non-human primates. We demonstrate substantially enhanced potency and therapeutic efficacy of these engineered vectors compared to naturally occurring AAV capsids in two mouse models of genetic muscle disease. The top capsid variants from our selection approach show conserved potency for delivery across a variety of inbred mouse strains, and in cynomolgus macaques and human primary myotubes, with transduction dependent on target cell expressed integrin heterodimers.

Keywords: AAV capsid engineering; Duchenne muscular dystrophy; MyoAAV; X-linked myotubular myopathy; directed evolution; integrin heterodimers; muscle gene therapy; non-human primates.

Figure 1. DELIVER identifies a class of muscle-tropic AAV capsid variants containing an RGD motif.

A) Schematic of virus library production and capsid variant selection using DELIVER. B) Comparison of rAAV titers produced using ITR-containing constructs that express the AAV9 capsid coding sequence under the control of CMV, CK8, or MHCK7 promoters. Data are presented as mean ± SD (n=4). P-value calculated by one-way analysis of variance (ANOVA) with Tukey-Kramer multiple comparisons test (MCT). C, D) In vivo expression of the AAV9 capsid library mRNA expressed under the control of CMV, CK8, or MHCK7 promoters in skeletal muscle (C) and heart (D) of 8 weeks old C57BL/6J mice systemically injected with 1E+12 vg of the capsid library. Data are presented as mean ± SD (n=3). P-value calculated by one-way ANOVA with Tukey-Kramer MCT. *: P<0.05, **: P<0.01. E) Graphs showing enrichment of capsid variants expressed under MHCK7 promoter over virus library at the DNA and mRNA level in different mouse skeletal muscles. F) Sequence of the 7-mer insertion in the top highly expressed capsid variants in muscles of 8 weeks old C57BL/6J mice injected with 1E+12 vg virus library after the second round of transcript-based selection. Variants with the same color in each group are encoded by synonymous DNA codons. Variant rank is based on the sum of mRNA expression for each variant in quadriceps, tibialis anterior, gastrocnemius, triceps, abdominal, diaphragm, and heart. See also Figure S1.

Figure 2. MyoAAV 1A transduces mouse skeletal muscles with high efficiency after systemic injection.

A, B) Whole mount fluorescent (A) and cross section (B) images of skeletal muscles, heart and liver from 8 weeks old C57BL/6J mice systemically injected with 1E+12 vg of AAV9- or MyoAAV 1A-CMV-EGFP. Green: EGFP, Red: laminin for muscles, Lectin for Liver, Blue: Hoechst. Scale bar in cross sections: 100 μm. C) Quantification of fold difference in EGFP mRNA expression in various tissues of male and female 8 weeks old C57BL/6J mice injected with 1E+12 vg of AAV9- or MyoAAV 1A-CMV-EGFP. Dashed red line indicates relative expression from AAV9-CMV-EGFP. Data are presented as mean ± SD (n=3–4); *: P<0.05, **: P<0.01 (Student t-test between AAV9 and MyoAAV 1A injected mice for each group). D) Quantification of in vitro transduction in mouse (left) and human (right) primary myotubes transduced with vehicle, AAV9- or MyoAAV 1A-CK8-Nluc. Data are presented as mean ± SD (n=5); **: P<0.01 (Student t-test). Donor 1: 29 year old male, donor 2: 19 year old female, donor 3: 20 year old male, donor 4: 34 year old female. E) Whole body in vivo bioluminescence images of 8 weeks old BALB/cJ mice systemically injected with 4E+11 vg of AAV8-, AAV9-, or MyoAAV 1A-CMV-Fluc, taken over 120 days. F) Quantification of total luminescence from forelimbs and hindlimbs of animals injected with AAV8-, AAV9-, or MyoAAV 1A-CMV-Fluc, assessed over 120 days. P-value calculated between AAV8, AAV9, and MyoAAV 1A groups by two-way ANOVA with Tukey-Kramer MCT; Data are presented as mean ± SD (n=5). **: P<0.01 for both MyoAAV 1A vs AAV8 and MyoAAV 1A vs AAV9. Difference between AAV8 and AAV9 groups is not statistically significant at any of the time points. See also Figures S2 and S3.

Figure 3. Systemic administration of MyoAAV 1A-Dmd CRISPR and MyoAAV 1A-human MTM1 results in therapeutic benefit in mouse models of DMD and XLMTM, respectively.

A) Representative immunofluorescence images for dystrophin (red) in muscles from 8 weeks old mdx mice injected with AAV9- or MyoAAV 1A-Dmd CRISPR (4.5E+12 vg of AAV-SaCas9 and 9E+12 vg of AAV-gRNA). Scale bar: 400 μm. B) Western blots detecting dystrophin and GAPDH in muscles of 8 weeks old mdx mice injected with AAV9- or MyoAAV 1A-Dmd CRISPR, with relative signal intensity determined by densitometry at the bottom. A.U.: arbitrary unit, normalized to GAPDH. C) Taqman-based quantification of exon 23-deleted mRNA in different muscles of 8 weeks old mdx mice injected with AAV9- or MyoAAV 1A- Dmd CRISPR. Data are presented as mean ± SD (n=9–10); **: P<0.01 (Student t-test). D, E) Tibialis anterior muscle specific force (D) and decrease in isometric force after 5 eccentric contractions (E) for wild-type C57BL/6J mice injected with vehicle (n = 11), and mdx mice injected with vehicle (n = 15), AAV9-Dmd CRISPR (n=15), or MyoAAV 1A-Dmd CRISPR (n=17). **: P < 0.01 (one-way ANOVA with Tukey-Kramer MCT). F) Schematic of the experiment to investigate the efficacy of 2E+12 vg/kg of AAV9- or MyoAAV 1A-MHCK7-human MTM1 (hMTM1) systemically delivered to 4 weeks old Mtm1 knockout (KO) mice. G) Total body weight of Mtm1 KO mice injected with vehicle or 2E+12 vg/kg of AAV9- or MyoAAV 1A-MHCK7-hMTM1, and wild type littermate controls injected with vehicle. Data are presented as mean ± SD (n=6 for KO AAV9, n=6 for KO MyoAAV 1A, n=3 for wild type vehicle, n=4 for KO vehicle). P-value calculated between MyoAAV 1A and AAV9 groups; **: P<0.01 (Multiple t-tests with Holm-Sidak MCT). H) Pictures of Mtm1 KO mice injected with 2E+12 vg/kg of either MyoAAV 1A-hMTM1 or AAV9-hMTM1 16 weeks after injection of the virus. I) Mean hourly passive activity assessed by in-cage running wheel rotation from wild type mice injected with vehicle, or Mtm1 KO mice injected with vehicle, AAV9-hMTM1, or MyoAAV 1A-hMTM1, both at 2E+12 vg/kg. Data are presented as mean ± SD (n=6 for KO AAV9, n=6 for KO MyoAAV 1A, n=3 for wild type vehicle, n=3 for KO vehicle) for weekly measurements averaged across three-week time periods. P-value calculated between MyoAAV 1A and AAV9 groups; **: P<0.01 (Multiple t-tests with Holm-Sidak MCT). J) Survival curve for Mtm1 KO animals injected with vehicle, 2E+12 vg/kg AAV9-hMTM1, or MyoAAV 1A-hMTM1, as well as wild type littermates injected with vehicle. (n=6 for KO AAV9, n=6 for KO MyoAAV 1A, n=3 for wild type vehicle, n=4 for KO vehicle). Data points for the Mtm1 KO mice injected with vehicle are from a previous experiment. P-value calculated between MyoAAV 1A and AAV9 groups; **: P<0.01 (Mantel-Cox test). K) Quantification of fold difference in hMTM1 mRNA expression in gastrocnemius, quadriceps, heart, and liver of Mtm1 KO mice injected with 2E+12 vg/kg AAV9- or MyoAAV 1A-hMTM1 at 4 weeks of age and analyzed 4 weeks after injection. Dashed red line indicates relative expression from AAV9-hMTM1. Data are presented as mean ± SD (n=4); *: P<0.05, **: P<0.01 (Student t-test between AAV9 and MyoAAV 1A injected groups for each tissue). L) Quantification of vector genome per diploid genome in various tissues of Mtm1 KO mice injected with 2E+12 vg/kg AAV9- or MyoAAV 1A-hMTM1 at 4 weeks of age and analyzed 4 weeks after injection. Data are presented as mean ± SD (n=4). *: P<0.05, **: P<0.01 (Student t-test). M) Western blots detecting hMTM1 and GAPDH in muscles of Mtm1 KO mice injected with vehicle, or 2E+12 vg/kg AAV9- or MyoAAV 1A-hMTM1 at 4 weeks of age and analyzed 4 weeks after injection, with relative signal intensity determined by densitometry at the bottom. A.U.: arbitrary unit, normalized to GAPDH. N) Extensor digitorum longus (EDL) muscle specific force for wild-type C57BL/6J mice injected with vehicle (n = 4), and Mtm1 KO mice injected with vehicle (n = 4), 2E+12 vg/kg AAV9-hMTM1 (n=4), or MyoAAV 1A-hMTM1 (n=4) at 4 weeks of age and analyzed 4 weeks after injection. **: P < 0.01 (one-way ANOVA with Tukey-Kramer MCT). See also Figure S4.

Figure 4. MyoAAV 1A transduction is dependent on integrin heterodimers

A, B) Quantification of in vitro transduction (A) and viruses bound to cell surface (B) in HEK293 cells transfected with plasmids encoding for RGD-binding integrin heterodimers or with pUC19, and transduced with MyoAAV 1A-CMV-Nluc. Data are presented as mean ± SD (n=3); *: P<0.01 (one-way ANOVA with Dunnett’s MCT with the pUC19 transfected cells set as the control). C, D) In vitro transduction efficiency in human primary myotubes treated with different concentrations of CWHM-12 (C) or GLPG-0187 (D) pan-integrin ɑV antagonists and transduced with AAV9- or MyoAAV 1A-CK8-Nluc. Data are presented as mean ± SD (n=5). P-value calculated compared to the 0 nM small molecule condition in each group. *: P<0.01 (one-way ANOVA with Dunnett’s MCT). E, F) In vitro transduction efficiency in human primary myotubes transduced with MyoAAV 1A-CK8-Nluc (E) or AAV9-CK8-Nluc (F) incubated with different concentrations of ɑVb1, ɑVb3, ɑVb6, ɑVb8, or MBP recombinant proteins. Data are presented as mean ± SD (n=5). *: P<0.01, **: P<0.001 (one-way ANOVA with Dunnett’s MCT with the 0 nM recombinant protein set as the control for each group). G, H) In vitro transduction efficiency in human primary myotubes treated with different concentrations of anti-ɑVb6 (G) or isotype control (H) antibody and transduced with AAV9- or MyoAAV 1A-CK8-Nluc. Data are presented as mean ± SD (n=5). P-value calculated compared to the 0 ng/ul antibody condition in each group; *: P<0.01, **: P<0.001 (one-way ANOVA with Dunnett’s MCT). See also Figures S5 and S6.

Figure 5. Further evolution of MyoAAV 1A using DELIVER generates more enhanced muscle-tropic capsid variants.

A) Structure of the AAV9 VR-VIII surface loop and the predicted structure of the MyoAAV 1A VR-VIII surface loop with the amino acids annotated. B) Schematic of virus library design and sequence of the top hits identified from mouse muscles after injection of the second-round virus library at two different doses. C) Different tissues of 8 weeks old C57BL/6J mice systemically injected with 2E+11 vg (~8E+12 vg/kg) of MyoAAV 1A-CMV-EGFP (left) or MyoAAV 2A-CMV-EGFP (right) illuminated by blue light. D) Whole mount fluorescent images of gastrocnemius, triceps, TA, and quadriceps of 8 weeks old C57BL/6J mice systemically injected with 2E+11 vg of AAV9-, MyoAAV 1A-, or MyoAAV 2A-CMV-EGFP. E) Quantification of fold difference in EGFP mRNA expression in various tissues of 8 weeks old C57BL/6J mice systemically injected with 2E+11 vg of MyoAAV 1A-, or MyoAAV 2A-CMV-EGFP compared to mice injected with the same dose of AAV9-CMV-EGFP. Dashed red line indicates relative expression from AAV9-CMV-EGFP. Data are presented as mean ± SD (n=11 for MyoAAV 1A and MyoAAV 2A, n=8 for AAV9). P-value calculated compared to the AAV9 group. *: P<0.05, **: P<0.01 (one-way ANOVA with Dunnett’s MCT). F) Quantification of vector genome per diploid genome in various tissues of 8 weeks old C57BL/6J mice injected with 2E+11 vg of AAV9-, MyoAAV 1A-, or MyoAAV 2A-CMV-EGFP. Data are presented as mean ± SD (n=11 for MyoAAV 1A and MyoAAV 2A, n=8 for AAV9). P-value calculated between MyoAAV 2A and AAV9 groups; *: P<0.05, **: P<0.01 (Student t-test). G) Quantification of in vitro transduction in human primary myotubes transduced with AAV9-, MyoAAV 1A-, or MyoAAV 2A-CK8-Nluc. Data are presented as mean ± SD (n=5). *: P<0.01 (one-way ANOVA with Tukey-Kramer MCT). H) In vitro transduction efficiency in human primary myotubes treated with different concentrations of GLPG-0187 integrin ɑV antagonists and transduced with AAV9- or MyoAAV 2A-CK8-Nluc. Data are presented as mean ± SD (n=5). P-value calculated compared to the 0 nM small molecule condition in each group. *: P<0.05, **: P<0.01 (one-way ANOVA with Dunnett’s MCT). I, J) In vitro transduction efficiency in human primary myotubes transduced with MyoAAV 2A-CK8-Nluc (I) or AAV9-CK8-Nluc (J) incubated with different concentrations of ɑVβ1, ɑVβ3, ɑVβ6, ɑVβ8, or MBP recombinant proteins. Data are presented as mean ± SD (n=5). **: P<0.01, ***: P<0.001 (one-way ANOVA with Dunnett’s MCT with the 0 nM recombinant protein condition in each group as the control). K) In vitro transduction efficiency in human primary myotubes treated with different concentrations of anti-ɑVb6 antibody (left) or 10 ng/ul isotype control antibody (right) and transduced with AAV9- or first or second-generation RGD-containing capsid variants encoding for Nluc under the control of CK8 promoter. Data are presented as mean ± SD (n=5). P-value for the anti-ɑVb6 antibody data calculated between the first-generation and second-generation groups. *: P<0.05, **: P<0.01 (Two-way ANOVA with Sidak’s MCT). P-value for the isotype control data calculated between the isotype control and 10 ng/ul anti-ɑVb6 antibody condition for each group; **: P<0.01 (Student t-test). L) Whole body in vivo bioluminescence images of 8 weeks old BALB/cJ mice systemically injected with 2E+11 vg (~8E+12 vg/kg) of AAVrh74-, AAV9-, or MyoAAV 2A-CMV-Fluc, taken over 21 days. Color scale: 6E+6 – 1E+9. M) Quantification of total luminescence from hindlimbs of animals injected with AAVrh74-, AAV9-, or MyoAAV 2A-CMV-Fluc, assessed over 21 days. P-value calculated between AAVrh74, AAV9, and MyoAAV 2A groups by two-way ANOVA with Tukey-Kramer MCT; **: P<0.01 for both MyoAAV 2A vs AAVrh74 and MyoAAV 2A vs AAV9. Difference between AAVrh74 and AAV9 groups is not statistically significant at any of the time points.

Figure 6. Systemic injection of MyoAAV 2A-CK8-microdystrophin at the low dose of 2E+13 vg/kg results in widespread microdystrophin expression and effective restoration of muscle function in adult DBA/2J-mdx mice.

A) Representative immunofluorescence images for microdystrophin-FLAG (red) in muscles of 8 weeks old DBA/2J-mdx mice injected with 2E+13 vg/kg AAV9- or MyoAAV 2A-CK8-microdystrophin. Scale bar: 400 μm. B) Western blots detecting microdystrophin-FLAG and GAPDH in muscles of 8 weeks old DBA/2J-mdx mice injected with 2E+13 vg/kg AAV9- or MyoAAV 2A-CK8-microdystrophin, with relative signal intensity determined by densitometry at the bottom. A.U.: arbitrary unit, normalized to GAPDH. C) Quantification of fold difference in microdystrophin mRNA expression in various muscles and liver of 8 weeks old DBA/2J-mdx mice systemically injected with 2E+13 vg/kg of MyoAAV 2A-CK8-microdystrophin compared to mice injected with the same dose of AAV9-CK8-microdystrophin. Dashed red line indicates relative expression from AAV9-CK8-microdystrophin. Data are presented as mean ± SD (n=9–10); *: P<0.05, **: P<0.01 (Student t-test). D) Quantification of vector genome per diploid genome in various muscles and liver of DBA/2J-mdx mice injected with 2E+13 vg/kg of AAV9- or MyoAAV 2A-CK8-microdystrophin. Data are presented as mean ± SD (n=9–10). **: P<0.01 (Student t-test). E, F) Muscle specific force (E) and decrease in isometric force after 5 eccentric contractions (F) for DBA/2J mice injected with vehicle (n=10), and DBA/2J-mdx mice injected with vehicle (n = 10), 2E+13 vg/kg AAV9-CK8-microdystrophin (n=10), or 2E+13 vg/kg MyoAAV 2A-CK8-microdystrophin (n=10). *: P < 0.05, **: P < 0.01, ***: P < 0.001 (one-way ANOVA with Tukey-Kramer MCT).

Figure 7. MyoAAV class of capsid variants evolved in NHPs transduce different muscles of Cynomolgus Macaques with high efficiency.

A, B) Schematic of virus library design and sequence of the top hits identified from NHP muscles after two rounds of in vivo selections in cynomolgus macaques starting from capsid variants containing a random 7-mer insert (A), or from one round of in vivo selection in cynomolgus macaques using the top 120,000 variants identified from the first round of RGD-fixed selection in mice (B). C) Comparison of in vitro transduction between the 11 muscle-tropic capsid variants selected in mice in human primary myotubes from four different donors. Data are presented as individual data points with mean; **: P<0.01 (one-way ANOVA with Dunnett MCT with MyoAAV 1A as the control). D) Schematic of the barcoded human Frataxin transgene and the pool of capsid variants used for characterization of the top muscle-tropic variants in NHPs and mice. E) Fold change mRNA expression over AAVrh74 in different skeletal muscles, heart, and liver of 3 Cynomolgus Macaques 1 month after systemic administration of 3E+13 vg/kg of the pooled virus mix. mRNA expression quantified by deep sequencing of the barcodes associated with each capsid variant. Data are presented as mean ± SD (n=3). *: P<0.05, **: P<0.01, ***: P<0.001, ****: P<0.0001 (one-way ANOVA with Dunnett’s MCT with AAVrh74 as the control). F) Comparison of AAVrh74, MyoAAV 3A, MyoAAV 4A, MyoAAV 4C, and MyoAAV 4E, and AAV9 titers with 6 different transgenes. Data are presented as mean ± SD (n=6). *: P<0.05; ****: P<0.0001 (one-way ANOVA with Dunnett’s MCT with AAVrh74 as the control). G) In vitro transduction efficiency in human primary myotubes treated with different concentrations of GLPG-0187 integrin ɑV antagonist and transduced with AAV9-, MyoAAV 3A-, MyoAAV 4A-, MyoAAV 4C-, or MyoAAV 4E-CK8-Nluc. Data are presented as mean ± SD (n=5). P-value calculated with a one-way ANOVA with Dunnett’s MCT with the 0 nM condition for each group set as the control. δ: P<0.0001 for MyoAAV 3A, MyoAAV 4A, MyoAAV 4C, and MyoAAV 4E; λ: P<0.0001 for MyoAAV 4A and MyoAAV 4E. H, I) In vitro transduction efficiency in human primary myotubes transduced with AAV9-CK8-Nluc (H) or MyoAAV 3A-, MyoAAV 4A-, MyoAAV 4C-, or MyoAAV 4E-CK8-Nluc (I) incubated with different concentrations of ɑVβ1, ɑVβ3, ɑVβ6, ɑVβ8, or MBP recombinant proteins. Data are presented as mean ± SD (n=5). **: P<0.01, ***: P<0.001, ****: P<0.0001 (one-way ANOVA with Dunnett’s MCT with the 0 nM recombinant protein condition in each group as the control). J) In vitro transduction efficiency in human primary myotubes treated with different concentrations of anti-ɑVβ6 antibody or mouse isotype control and transduced with AAV9-, MyoAAV 3A-, MyoAAV 4A-, MyoAAV 4C-, or MyoAAV 4E-CK8-Nluc. Data are presented as mean ± SD (n=5). δ: P<0.0001 for MyoAAV 3A, MyoAAV 4A, MyoAAV 4C, and MyoAAV 4E; θ: P<0.0001 for MyoAAV 3A, MyoAAV 4A, and MyoAAV 4E (one-way ANOVA with Dunnett’s MCT with the 0 nM condition for each group as the control). K) Comparison of in vitro transduction between HEK293FT and HEK293FT AAVR KO cells transduced with AAV2-, AAV4-, AAV9-, MyoAAV 3A-, MyoAAV 4A-, MyoAAV 4C-, or MyoAAV 4E-CMV-Nluc. Data are presented as mean ± SD (n=4). ****: P<0.0001 (Student t-test on log transformed data). See also Figure S7.