Analysis of Cardiac Myocyte Maturation Using CASAAV, a Platform for Rapid Dissection of Cardiac Myocyte Gene Function In Vivo

AAV-U6gRNA-cTNT-Cre plasmid (Online Figure I) was constructed by transferring a cTNT-Cre cassette from AAV-TNT-Cre plasmid¹¹ to PX552 plasmid.¹² A second U6gRNA was further added to generate AAV-U6gRNA₁-U6gRNA₂-cTNT-Cre (Addgene No. 87682). A total of 20 bp gRNA sequences (Online Table I) were cloned into these plasmids through a single ligation step.

Adeno-associated virus (AAV) was produced and purified using a standard protocol with modifications.¹³ AAV titer was quantified by quantitative polymerase chain reaction using primers indicated in Online Table II.

All animal procedures were approved by the Boston Children’s Hospital Institutional Animal Care and Use Committee. AAVs were injected into P1 Rosa^{Cas9GFP/Cas9GFP} mice¹⁴ by subcutaneous injection. CMs were dissociated by retrograde collagenase perfusion as described previously.¹⁵

The mRNA level of Jph2 was examined by quantitative reverse transcription polymerase chain reaction (qRT-PCR) using primers shown in Online Table II. The deletion of Jph2 between 2 gRNAs was detected by RT-PCR. Construction of DNA libraries for amplicon sequencing was performed by modifying a published protocol¹⁶ using primers shown in Online Table II. Paired-end, 250 bp reads were generated on an Illumina MiSeq and analyzed using CRISPResso software.¹⁷

Validation of gene knockout was performed through immunofluorescence. Antibodies and dyes used are listed in Online Table III. Hearts were dissociated into single CMs through retrograde collagenase perfusion. Then, the CMs were briefly cultured to allow attachment to laminin-coated glass coverslips. Next, the CMs were fixed by 4% paraformaldehyde, and immunostaining was performed as described.^18,19 Confocal images were taken using Olympus FV1000 inverted laser scanning confocal microscope equipped with a 60X/1.3 silicone oil objective.

Quantitative analysis by flow cytometry was performed using a Propel Laboratories Avalon cytometer with a 100 µm nozzle and standard GFP/RFP (green/red fluorescence protein) filter sets. Data were further analyzed using BioRad ProSort software.

Fluorescence-activated cell sorting was performed using a BD AriaII SORP cell sorter with a 100 µm nozzle at Dana-Farber flow cytometry core facility.

Gross heart morphology was imaged under a stereomicroscope (Zeiss SteREO Discovery V8) with an AxioCam MRc camera. Heart histology, organ weights, and expression of cardiac stress markers were assayed as described previously.¹³ Echocardiography was performed on awake animals using a VisualSonics Vevo 2100 with Vevostrain software. Echocardiography was performed blinded to AAV types and doses.

In situ T-tubule imaging and AutoTT analysis of T-tubule patterns were performed as described previously.²⁰

Intracellular Ca²⁺ was measured in isolated CMs that were loaded with Rhod-2 AM (acetoxymethyl). Cells were electrically stimulated at 1 Hz to produce steady-state conditions. All image data were acquired in the line scanning mode along the long axis of the cell using an Olympus FV1000 inverted laser scanning confocal microscope with a 60X/1.3 silicone oil objective.

Unless otherwise noted, values are displayed as mean±SD. Statistical comparisons were performed using Student t test. Kaplan–Meier survival analysis was performed using JMP software (SAS).

Previously our laboratory showed that the heart could tolerate the deletion of essential genes in a small fraction of CMs while still maintaining normal heart function.¹¹ This mosaic analysis allowed us to study CM maturation in individual cells without the interference of secondary effects caused by heart failure. To establish a versatile technical platform that allows mosaic somatic mutagenesis of any gene specifically in CMs, we combined AAV-mediated mosaic transduction¹¹ with CRISPR/Cas9-based in vivo mutagenesis.^14,21–24 We constructed AAV vectors that contain 1 or 2 U6 promoter–driven gRNA expression cassettes and a cardiac troponin-T promoter that drives CM-specific expression of Cre recombinase (Figure 1A; Online Figure I). Using these vectors, we generated AAV9,²⁵ a cardiotropic AAV serotype, and subcutaneously injected the replication-deficient virus into Rosa26^Cas9-GFP mice that harbor a Cre-inducible Cas9-P2A-GFP allele.¹⁴ cTNT-Cre activated the expression of Cas9 and GFP specifically in AAV-transduced CMs, and the U6 promoter drove expression of gRNA(s) in these cells (Figure 1A). Thus, GFP⁺ CMs express Cas9 nuclease activity that is directed specifically to target sites by gRNA(s). The nuclease creates DNA double-strand breaks at these target sites and triggers DNA damage repair by nonhomologous end joining, which is error prone and frequently results in insertions or deletions that can induce frameshift mutations silencing the targeted genes (Figure 1A).²⁶

We first designed a gRNA that targets tdTomato gene and tested whether the tdTomato protein in Rosa^{Cas9-GFP/tdTomato} mice is depleted by this CRISPR/Cas9/AAV9–based somatic mutagenesis (CASAAV) system. We injected AAV-TNT-Cre (negative control) or AAV-gRNA(tdTomato)-TNT-Cre into P1 littermate pups and isolated CMs at 1 month by retrograde collagenase perfusion. The CMs were analyzed by flow cytometry to detect GFP and tdTomato. We found that tdTomato was depleted in ≈89% GFP⁺ CMs that were transduced by AAV expressing tdTomato gRNA (Figure 1B). No depletion was detected in GFP⁺ CMs transduced with control AAV that lacked gRNA (Figure 1B; Online Figure II). Interestingly, although theoretically Cre should trigger expression of both Tomato and Cas9-P2A-GFP, we also detected tdTomato⁺ GFP⁻ cells (Online Figure II). Two effects likely resulted in these cells: (1) some CMs transiently harbor AAV, and limited Cre exposure activates tdTomato but not Cas9-P2A-GFP, as a result of differential sensitivity of these alleles to Cre activation and (2) GFP level was observed to be sensitive to CM stress, as occurs during CM dissociation and flow cytometry. The net effect is that a minority of GFP⁻ cells may have undergone Cas9 genome editing, a technical factor that needs to be considered when interpreting data (Discussion). Overall, this pilot experiment demonstrates that the CASAAV system mediates highly efficient gene silencing in CMs in vivo.

Next, we designed 2 gRNAs that target the coding sequence of Jph2 within exon 1, which encodes the first 4 plasma membrane–binding motifs of JPH2 protein (Figure 1C). These gRNAs had at least 3 mismatches to any nontargeted sequences in the mouse reference genome (Online Figure IIIA). We generated AAVs that expressed each gRNA individually (Jph2-gRNA1 or Jph2-gRNA2), or both simultaneously (Jph2-gRNA1+2), and injected 1.1×10⁹ vg/g (viral genomes per gram body weight) AAV into P1 Rosa^{Cas9-GFP/Cas9-GFP} pups. We isolated CMs from P21 hearts and performed immunofluorescence to detect JPH2 protein (Figure 1D). We found that Jph2-gRNA1, Jph2-gRNA2, and Jph2-gRNA1+2 depleted JPH2 immunofluorescence in ≈55%, ≈70%, and ≈80% of GFP⁺ CMs, respectively (Figure 1E). We also analyzed JPH2 depletion in transduced CMs by using fluorescence-activated cell sorting to purify GFP⁺ CMs, followed by Western blot (Figure 1F and 1G). Jph2-gRNA1+2 efficiently depleted JPH2 in transduced CMs. The level of BIN1, another protein implicated in T-tubule maturation,²⁷ was not affected (Figure 1G). The major effect of Jph2-gRNA1+2 was protein depletion rather than truncation (Online Figure IV), although we cannot exclude production of truncated protein in a low fraction of transduced cells. Overall, our data show that Jph2-CASAAV efficiently depleted JPH2 in CMs and that 2 gRNAs were more efficient at achieving protein depletion than each single gRNA.

To determine the mechanisms that mediate this efficient JPH2 depletion, we injected 5.5×10¹⁰ vg/g AAV at P1, which is sufficient to transduce ≈75% CMs in the heart (Figure 2). At P7, we extracted RNA from heart apexes and detected significantly decreased Jph2 mRNA after AAV-Jph2gRNA treatment by quantitative RT-PCR (Figure 3A and 3B). Analysis of Jph2 transcript structure by RT-PCR demonstrated that dual-gRNA CASAAV induced deletion of the intervening coding sequence between 2 gRNA targets in a subset of transcripts, which was not observed with either gRNA alone (Figure 3A and 3C). To more precisely define the effect of CASAAV on Jph2 transcripts, we amplified each gRNA-targeted site from cDNA and performed next-generation sequencing (Amplicon-Seq)²⁸ (Figures 2D through 2F and 3A; Online Figure IIIB and IIIC). Analysis of these sequences using CRISPResso¹⁷ showed that gRNA1 alone generated mutations in 72% of transcripts, and 71% of these mutated reads shift the translational reading frame (Figure 3D). gRNA2 alone generated mutations in 61% of transcripts, and 86% of these mutated reads shift the translational reading frame (Figure 3D). These mutations localized to the expected gRNA cleavage sites (Figure 3E). When both gRNAs were present, frameshift mutations were detected in >80% mutated reads, and these mutations occurred between the 2 gRNA-targeted sites (Figure 3D through 3F). Thus, JPH2 protein depletion induced by CASAAV is because of a combination of frame-shifting mutations and mRNA decrease, potentially resulting from nonsense-mediated decay.²⁹ The use of 2 gRNAs increased the efficiency of both of these effects. Deletion of Jph2 coding sequences that encode critical plasma membrane–binding domains between the 2 gRNA-targeted sites also provides potential mechanisms to perturb the stability and function of JPH2 protein (Figures 1C, 3C, and 3F).

To determine the effect of AAV dosage on whole organ and individual CM phenotypes, we treated mice with serial dilutions of AAV-gRNA(Jph2)1+2-TNT-Cre, referred to subsequently as AAV-gRNA(Jph2), at P1. We defined the mice receiving 1.1×10⁹, 6.1×10⁹, 1.8×10¹⁰, and 5.5×10¹⁰ vg/g AAV doses as Jph2_Low, Jph2_Mid, Jph2_High, and Jph2_Full groups, respectively. We measured the frequency of transduced CMs at P7 by detecting Cre-dependent GFP expression from the Rosa26^Cas9-GFP allele. AAV dose was positively correlated to the myocardial area occupied by GFP⁺ CMs (Figure 2A). We quantified the frequency of GFP⁺ CMs in dissociated CM preparations. Jph2_Low, Jph2_Mid, Jph2_High, and Jph2_Full groups expressed GFP in 22%, 47%, 64%, and 75% of CMs, respectively (Figure 2B).

Next, we investigated whether AAV dosage influences the efficiency of JPH2 depletion in individual transduced CMs. We performed immunofluorescence staining of JPH2 in CMs that were treated with different AAV doses at P1 and isolated at P7. The control AAV-TNT-Cre virus (lacking gRNA) was administered at the full dose (5.5×10¹⁰ vg/g). Immunostaining of JPH2 was used to measure each cell’s JPH2 expression level, and immunostaining of caveolin-3 (CAV3), a plasma membrane marker, defined the boundary of each single cell (Figure 2C and 2D). We found a ≈70% decrease of JPH2 intensity in Jph2-CASAAV–treated CMs when compared with negative controls, which further confirmed the efficient depletion of JPH2 protein. By immunostaining, there was no detectable difference in intensity between GFP⁺ CMs that were treated with different doses of AAVs (Figure 2D). To examine the impact of AAV dosage on the efficiency of protein depletion, we performed fluorescence-activated cell sorting of GFP⁺ cells in each group and analyzed JPH2 expression by Western blotting (Figure 2E). This demonstrated efficient JPH2 depletion in all groups, although the protein remained detectable in the Jph2_Low and Jph2_Mid groups but was undetectable in the Jph2_High and Jph2_Full groups. Thus, AAV dose primarily determines the frequency of transduced CMs. The dose has a small effect on protein depletion efficiency among transduced cells, but even at the lowest dose tested, the protein was efficiently depleted. These findings are important for interpreting the mosaic gene knockout data presented below.

Mice that received the full dose of AAV-gRNA(Jph2) died between P10 and P13 (Figure 4A), which is consistent with the survival curve of Myh6-Cre;Jph2-shRNA mice described previously.⁸ Mice treated with the other doses of AAV-gRNA(Jph2) survived beyond P20 (Figure 4A), when T-tubules become fully mature.² Because our goal was to examine the impact of Jph2 on T-tubule maturation, we focused the following studies on Jph2_Low, Jph2_Mid, and Jph2_High groups and studied mice at P21 to P24. Analyses of gross heart morphology showed that Jph2_High and Jph2_Mid groups had cardiac hypertrophy and dilatation, which were not evident in the Jph2_Low group (Figure 4B). This difference was quantified by measurement of the heart weight/body weight ratio, which was not elevated in the Jph2_Low group but was increased in Jph2_Mid and Jph2_High groups in a dose-dependent manner (Figure 4C). Echocardiography showed severely depressed heart systolic function in the Jph2_High group (Figure 4D). By contrast, the Jph2_Mid group had moderately depressed heart function, and the Jph2_Low group did not exhibit detectable heart dysfunction (Figure 4D). AAV-gRNA(Jph2) dosage also correlated with the extent of left ventricular dilatation (Figure 4E and 4F). Expression of cardiac stress markers Myh7, Nppa, and Nppb were consistent with these overall effects of virus dose on heart structure and functions (Figure 4G). Together, these data indicate that high and mid doses of AAV-gRNA(Jph2) caused dose-related cardiac hypertrophy and dysfunction, as previously reported in Jph2 shRNA genetic mouse models,^2,8 whereas the low dose did not impact heart function at the organ level.

We next examined the effect of the extent of Jph2 inactivation on the responses of individual CMs. Although induction of heart dysfunction correlated with increased CM size, there was no difference in CM size between GFP⁺ and GFP⁻ CMs at any viral dose (Figure 4H). Thus, JPH2 depletion does not directly affect CM size. Rather, JPH2 affects CM size indirectly through effects on overall organ function.

Collectively, these data show that a low dose of AAV-gRNA(Jph2) can be used to achieve mosaic depletion of JPH2 without affecting overall organ function. Mid and high doses increase the fraction of CMs with gene inactivation and correspondingly cause organ dysfunction and global stress responses.

Because mosaic Jph2 inactivation by low-dose AAV-gRNA(Jph2) avoided secondary effects of organ dysfunction, we were able to precisely examine the role of JPH2 in T-tubule organization without the confounding effects of cardiac dysfunction. To observe T-tubule organization in live CMs in intact heart ventricles, we labeled T-tubules with the plasma membrane dye FM 4 to 64 through retrograde heart perfusion and performed in situ confocal imaging.^2,30 We found that T-tubules were dramatically disrupted in GFP⁺ CMs in Jph2_High mice. However, this phenotype was mild in GFP⁺ CMs in Jph2_Mid mice. No overt disruption in T-tubules was observed in GFP⁺ CMs in Jph2_Low mice (Figure 5A and 5B; see additional representative confocal z stacks in Online Movies I through III). We analyzed the changes in T-tubules in details by performing unbiased quantification of T-tubule morphology using AutoTT software²⁰ (Figure 5B). This software objectively quantifies transverse T-tubule elements, longitudinal T-tubule elements, total T-tubule elements (which equal to transverse T-tubule elements+longitudinal T-tubule elements), and T-tubule regularity.²⁰ Compared with GFP⁻ internal control CMs, GFP⁺ CMs in Jph2_Low group showed no significant change in T-tubule regularity or longitudinal T-tubule elements and subtly but significantly reduced transverse T-tubule elements and total T-tubule elements (Figure 5C through 5F). The Jph2_Mid group also showed no significant change in longitudinal T-tubule elements, but a significant impairment in the other 3 parameters was greater than in the Jph2_Low group. All of these parameters were most dramatically impaired in the Jph2_High group. T-tubules in GFP⁻ CMs from Jph2_High mice exhibited T-tubule defects, which were more severe than GFP⁺ CMs in Jph2_Low and Jph2_Mid mice. These data are consistent with previous observations that T-tubules remodel and become disorganized in wild-type CMs in pathological conditions.⁹ The GFP⁺ CMs in Jph2_High mice had substantially more disrupted T-tubules than their GFP⁻ counterparts (Figure 5C through 5F), suggesting a cell-autonomous role for JPH2 to protect T-tubules from remodeling in heart failure.³¹ We also examined the effect of JPH2 depletion on sarcomere structure, as assessed by immunostaining for Z-line marker sarcomeric α-actinin (Actn2; Online Figure V). Z-line spacing and regularity, measured by AutoTT,²⁰ were not significantly affected by JPH2 depletion (Online Figure V). Together these data demonstrate that JPH2 has a minor cell-autonomous function in T-tubule organization that is exacerbated by heart failure.

To further confirm that these varied phenotypes were not caused by JPH2 depletion efficiency, we performed dual-color immunofluorescence to detect both JPH2 and CAV3 in CMs dissociated from AAV-TNT-Cre (control)–treated and AAV-gRNA(Jph2)–treated mice. This permitted us to identify true JPH2-deficient CMs and examine CAV3 pattern in these CMs (Figure 6A). Consistent with in situ T-tubule imaging results, JPH2⁻ CMs that were derived from Jph2_Low mice exhibited normal global T-tubule morphology when compared with JPH2⁺ control CMs (Figure 6A). By contrast, Jph2_Mid and Jph2_High mice exhibited disrupted CAV3 patterns in JPH2⁻ CMs (Figure 6A). AutoTT quantification of CAV3 immunostaining also showed significantly decreased total T-tubule elements and regularity in JPH2⁻ CMs isolated from Jph2_Mid and Jph2_High mice, but no statistically significant difference between Jph2_Low and control was observed (Figure 6B and 6C). It is possible that this experiment did not reproduce the mild T-tubule defects in Jph2_Low that were observed by in situ T-tubule imaging (Figure 5) because the CM dissociation procedure affects T-tubule organization in some CMs and thereby increases within-group variation (compare error bars between Figure 5 and Figure 6). However, based on both in situ T-tubule analysis and CAV3 analysis on isolated CMs, it is clear that JPH2 has a minor direct function in T-tubule maturation. Substantial T-tubule disruption was only observed when JPH2 was depleted in a large fraction of CMs, resulting in cardiac stress and associated secondary effects on CMs (Figure 6D).

We next evaluated whether the varied effect of JPH2 depletion on T-tubule organization influences Ca²⁺ handling. Freshly dissociated CMs were loaded with the Ca²⁺-sensitive dye Rhod-2 AM, and Ca²⁺ transients were measured by confocal line scan imaging. Both transduced and nontransduced Jph2_High CMs exhibited significantly reduced Ca²⁺ transient amplitude, prolonged time to peak, and slower time to 50% decay (Figure 7A and 7B), which is consistent with previous studies of JPH2’s role in Ca²⁺ handling.² However, the major effect was likely secondary to heart failure and not directly attributable to JPH2 depletion because the difference with TNT-Cre control and Jph2_High CMs was greater than the difference between GFP⁺ and GFP⁻ CMs within Jph2_High. Ca²⁺ transient amplitude was more severely depressed in GFP⁺ than GFP⁻ Jph2_High CMs, consistent with the more severe T-tubule abnormalities observed in GFP⁺ versus GFP⁻ Jph2_High CMs. Jph2_Low GFP⁺ CMs had Ca²⁺ transients with mildly but significantly reduced peak amplitude compared with GFP⁻ CMs isolated from the same heart. No significant change in time to peak or time to 50% decay was detected (Figure 7C and 7D). This is consistent with the minor effect of JPH2 depletion on T-tubule morphology in Jph2_Low. These single-cell measurements of Ca²⁺ handling paralleled our imaging data on T-tubule structure and indicate that JPH2 has a minor cell-autonomous role in establishing Ca²⁺ release unit function.

Mosaic mutagenesis by CASAAV avoids the time-consuming procedure of generating genetically modified mice. A large number of AAVs that deliver gRNAs directed at different genes can be generated side by side simultaneously. Thus, this technology allowed us to perform a genetic screen for novel T-tubule regulators. Using a candidate-based approach, we generated AAVs targeting 8 genes that could be potentially important for T-tubule maturation. These genes encode important cardiac transcriptional factors (NKX2-5 [NK2 homeobox 5], TBX5 [T-box 5], MEF2C [myocyte enhancer factor 2C], and TEAD1 [TEA domain family member 1]) and T-tubule–associated proteins (RYR2, CACNA1C [calcium voltage-gated channel subunit alpha1 C], CAV3, and NCX1 [sodium-calcium exchanger 1]). We injected each of these viruses into P1 Rosa26^Cas9-P2A-GFP mice at a low dose (similar to that used in Jph2_Low), dissociated the hearts at P21, and performed JPH2 immunostaining to image T-tubule morphology in GFP⁺ CMs (Figure 8A). GFP⁻ CMs from the same heart were used as internal negative controls. Efficient depletion of NKX2-5, TEAD1, RYR2, and CAV3 was validated by immunostaining (Online Figure VI; Figure 8C and 8D).

Through this screen, we found that RYR2 depletion caused dramatic disruption of the JPH2 staining pattern (Figure 8A and 8B). By contrast, the AAVs targeting the other 7 genes did not show this effect. Efficient depletion of RYR2 was observed in ≈65% of the AAV-infected GFP⁺ CMs (Figure 8C and 8D). Mice with mosaic RYR2 depletion in a minority of CMs (only ≈10% of total CMs were depleted of RYR2) maintained normal heart function (Figure 8E). Dramatic T-tubule disruption was observed in GFP⁺ CMs under in situ T-tubule imaging, but GFP⁻ neighboring cells retained normal T-tubule patterns (Figure 8F and 8G; see additional representative confocal z stacks in Online Movie IV). This phenotype was also confirmed in isolated CMs, as RYR2⁻ CMs exhibited disrupted CAV3 staining pattern, whereas normal CAV3 pattern was observed in RYR2⁺ CMs (Figure 8H and 8I). The effect of RYR2 depletion on T-tubule structure staining pattern was reproduced with 2 additional independent pairs of gRNAs (Online Figure VII), which ruled out off-target effects. Together, these data show a novel cell-autonomous function of RYR2 in T-tubule maturation.