CRISPR-Cas9 generated Pompe knock-in murine model exhibits early-onset cardiomyopathy and impaired skeletal muscle function

The goal of this study is to generate and characterize a knock-in model of Pompe disease (PD) – a rare, progressive, fatal disorder primarily affecting the cardiac and musculoskeletal systems. While a murine model of PD exists, it bears a Cre/loxP induced exonic insertion of a neomycin cassette and does not completely recapitulate severe human PD - displaying nonfatal hypertrophic cardiomyopathy only late in its natural history. We therefore designed a CRISPR-Cas9 knock-in system targeting the Gaa gene to introduce the known pathogenic CRIM negative Gaa mutation c.1826insA (p.Y609*). Following optimization of our knock-in strategy in cultured murine myoblasts, we successfully generated a Gaac1826insA mouse model using a dual sgRNA with ssODN donor template approach. Whole genome sequencing and analysis of the Gaac1826insA murine model establishes that our system is highly specific for the Gaac1826 target locus and does not induce any off-target mutations or genomic rearrangements. Next, we examined GAA mRNA transcript, protein expression and enzymatic activity levels in our PD knock-in mice. Gaac1826insA mice display significantly reduced levels of GAA expression and enzymatic activity relative to wild-type mice. We performed echocardiography on Gaac1826insA mice to assess cardiac structure and function. Gaac1826insA mice exhibit early-onset, progressive cardiac hypertrophy as measured by significant increases in left ventricular wall thickness and mass index by 3 months of age. We also conducted functional tests – grip strength, inverted screen, gait analysis – on Gaac1826insA mice every 3 months to assess overall motor performance. Gaac1826insA mice display impaired motor strength and coordination relative to wild-type mice. Altogether, our results demonstrate that the Gaac1826insA murine model recapitulates human infantile-onset Pompe disease and is better suited for evaluation of therapeutic strategies such as genome correction.


Introduction
Generation of transgenic murine knock-in models of human disease once relied solely upon targeted insertion of the desired sequence via Cre-Lox recombination and embryonic stem cell implantation. Though this strategy has resulted in many successful murine model systems, it is labor-intensive, time-consuming, and expensive. The advent of genome editing via engineered nucleases, especially clustered regularly interspaced short palindromic repeats (CRISPR)based systems has allowed for a potentially accurate, efficient, and relatively inexpensive alternative to the traditional method of transgenic knock-in model generation 1,2 .
As an intriguing example of disorders that may uniquely benefit from genome editing, inherited metabolic disorders (IMDs) are a diverse group of genetic diseases affecting the proper breakdown or synthesis of essential compounds such as carbohydrates, amino acids, or organic acids. Many of these disorders are caused by single gene defects that alter the expression and/or activity of critical metabolic enzymes. Given the monogenic nature of IMD pathogenesis, this class of genetic disorders is quickly becoming an area of high interest for CRISPR-mediated genome editing therapeutics 3,4 . Pompe disease can be treated with intravenous enzyme replacement therapy (ERT) using recombinant human acid α-glucosidase (rhGAA) enzyme, which significantly reduces cardiac hypertrophy and increases overall and ventilator-free survival 6 . Unable to endogenously synthesize GAA, Pompe patients are infused indefinitely and may produce an anti-rhGAA antibody response that may limit or neutralize treatment efficacy. Regardless of immune response, glycogen storage, autophagic buildup, and fibrosis within skeletal myocytes are observed even in early-treated IOPD patients 7,8 . Consequently, a phenotype of sensorineural hearing loss, central nervous system white matter abnormalities, slowly progressive muscle weakness and delayed mortality is now observed in rhGAA-treated survivors with IOPD 9,10 .
The limitations of current Pompe disease treatment underscore the necessity of new therapeutic development. CRISPR-based therapeutic strategies may address the impermanence of ERT, effecting permanent, highly-specific somatic correction of genomic GAA mutations within myocytes and subsequent intramuscular, endogenous synthesis of enzyme. Reduced GAA enzyme within the bloodstream may also mitigate the immunogenicity of intravenous ERT.
First, though, an animal model with molecular, biochemical, physiological and functional analogy to human Pompe disease must be developed. Currently, there is a widely-utilized knockout murine model of Pompe disease (B6;129-Gaa tm1Rabn/J ) that demonstrates survival into adulthood with muscle glycogen storage and progressive muscle weakness 11 . This model bears a neomycin resistance cassette (~800bp) in its Gaa gene, not an analog of a human GAA mutation, complicating preliminary efforts at in vivo genome correction. We report the successful generation of a Gaa c.1826dupA (p.Y609*) 12 murine knock-in model of Pompe disease utilizing a novel, dual-single guide RNA approach flanking the intended Gaa insertion site, and early characterization that demonstrates GAA enzymatic deficiency, hypertrophic cardiomyopathy, muscular glycogen storage and pathology recapitulating human Pompe disease.

Dual overlapping gRNA approach achieves highest HDR levels in vitro
Three guide RNAs (Gaa c.1826 gRNA-1,2,3) were selected based on best predicted on-target and off-target scoring and proximity of expected cut site to Gaa c.1826 target locus (Figure 1a). To determine in vitro on-target editing activity and HDR efficiency, Gaa c.1826 gRNA expression vectors and respective ssODN donor templates were nucleofected into C2C12 mouse myoblasts. On-target editing activity and HDR efficiency was highly dependent on target sequence ( Table 1). For single gRNA approach, Gaa c.1826 gRNA-3 demonstrated the highest ontarget editing activity (47.90.1%) with Gaa c.1826 gRNA-2 achieving the best HDR efficiency (7.71.4%). Given that multiple gRNAs with overlapping sequences are known to enhance CRISPR/Cas9-mediated knock-in efficiency 24 , we evaluated whether a dual overlapping gRNA approach could improve our on-target editing activity and HDR efficiency. We chose to test Gaa c.1826 gRNAs-2 and -3 at an equimolar ratio as they achieved the highest levels of on-target editing activity and HDR efficiency. Furthermore, Gaa c.1826 gRNAs-2 and -3 are a senseantisense pair with fully overlapping target sequences thereby reducing the likelihood of added off-target activity. We found that the dual overlapping gRNA approach achieved high on-target editing activity (47.90.1%) with the highest overall HDR efficiency (12.62.9%). Further testing will need to be performed to confirm that this dual overlapping gRNA approach can be broadly applied to increase HDR efficiency in other target loci and cell lines.

Generation & characterization of Gaa c.1826dupA knock-in C2C12 cell line
We used the dual overlapping Gaa c.1826 gRNA strategy (Figure 2a-b) followed by puromycinresistant selection to isolate Gaa c.1826dupA knock-in C2C12 clonal cells. Sequencing results confirmed presence of desired Gaa c.1826dupA knock-in mutation as well as silent PAM and seed region mutations to prevent gRNA editing of the donor template (Figure 1b). Gaa c.1826dupA knockin cells exhibited enhanced PAS staininga marker of glycogen accumulationrelative to Gaa wt cells (Figure 1c). Gaa c.1826dupA knock-in cells display a 96% reduction in Gaa transcript levels and GAA enzymatic activity was completely abolished (Figure 1d). Together, these results demonstrate that the dual overlapping gRNA approach can increase overall HDR efficiency in vitro thus improving the probability of isolating clonal cells with a desired knock-in mutation. Moreover, our Gaa c.1826dupA knock-in cell line exhibits molecular and biochemical analogy to Pompe disease thereby validating its use as an in vitro model.

Generation & characterization of Gaa c.1826dupA transgenic mice
We next applied the dual overlapping Gaa c.1826 gRNA strategy in vivo (Figure 2a and 25% HDR efficiency (founder mice positive for Gaa c.1826dupA mutation) ( Table 2). Following founder mice genotyping, we selected a founder with the lowest levels of mosaicism -as determined by TIDER 16 analysisfor mating and segregation of Gaa c.1826dupA mutation. We successfully generated a homozygous Gaa c.1826dupA knock-in mouse after 2 generations of breeding (Figure 2c). Sequencing results confirmed presence of desired Gaa c.1826dupA knock-in mutation as well as silent PAM and seed region mutations in G0 founder (Gaa Mosiac ), G1 heterzygous (Gaa wt/c.1826dupA ) and G2 homozygous knock-in (Gaa c.1826dupA ) mice (Figure 2d).
To screen for off-target integration of the donor template, we searched for called single nucleotide variants (SNVs) that had the unique donor template motif. We then repeated this step for the reverse complement and found that the only positive result was the intended mutation at the Gaa c.1826 target locus. Next, we screened the top 5 genomic regions predicted by GT-Scan 14 to be potential off-target sites (Figure 2f). There were no detected SNVs within 500bp of these sites.
Altogether, these results suggest that the dual overlapping gRNA approach is an efficient strategy to generate transgenic Gaa c.1826dupA knock-in mice and did not result in any detectable off-target activity in the genomes of our founder mouse and its progeny. Gaa wt/c.1826dupA heterozygous (HET) mice exhibited a 75% and 46% reduction in Gaa transcript levels relative to WT, respectively. To compare these results with the commercially available Gaa tm1Rabn knockout (KO Rabn ) mice, we found an 81% reduction in KO Rabn Gaa transcript levels relative to WT. Notably, there was no significant difference in Gaa transcript levels between KI and KO Rabn mice.

Gaa transcript levels, GAA enzyme activity and glycogen load in Gaa
Next, we measured GAA enzyme activity levels in Gaa c.1826dupA transgenic and WT liver, diaphragm and gastrocnemius muscle. For all tissues, transgenic mice demonstrated a gene dose-dependent decrease in GAA enzyme activity levels relative to WT mice (Figure 3b). Liver GAA activity was reduced by 100% and 35% reduction in KI and HET mice relative to WT, respectively. Diaphragm GAA activity was reduced by 95% and 54% in KI and HET mice relative to WT, respectively. Gastrocnemius muscle GAA activity was reduced by 98% and 57% in KI and HET mice relative to WT, respectively. We also assessed GAA activity in KO Rabn mice and found a 99%, 88% and 97% reduction in liver, diaphragm and gastrocnemius GAA activity relative to WT, respectively. Notably, there was no significant difference in GAA enzymatic activity between KI and KO Rabn mice for all tissue types.
We then measured glycogen load in Gaa c.1826dupA transgenic and WT diaphragm and gastrocnemius muscle. For both tissues, transgenic mice demonstrated a gene dose-dependent increase in glycogen load relative to WT mice (Figure 3c). Diaphragm glycogen levels were increased 108-fold and 2-fold in KI and HET mice relative to WT, respectively. Gastrocnemius muscle glycogen levels were increased 28-fold and 2-fold in KI and HET mice relative to WT, respectively. We also measured glycogen load in KO Rabn mice and found an 82-fold and 89-fold increase in diaphragm and gastrocnemius glycogen levels relative to WT, respectively.
Interestingly when compared to KI, KO Rabn glycogen load was 1.3-fold lower and 3.2-fold higher in diaphragm and gastrocnemius, respectively.
Taken together, these results demonstrate that our Gaa c.1826dupA knock-in mouse model exhibits molecular and biochemical analogy to the established preclinical model of Pompe disease (Gaa tm1Rabn ) and is an appropriate in vivo model for genome-based therapeutic evaluation.

Cardiac anatomy and function in Gaa c.1826dupA transgenic mice
To assess overall cardiac anatomy and function in Gaa c.1826dupA transgenic mice, we performed echocardiography on 3-month old WT, HET and KI mice (Figure 4a). Relative to WT and HET mice, KI mice display significant increases in intraventricular septal diameter (IVSd), left ventricular posterior wall diameter (LVPWd) and left ventricle mass indexanatomical hallmarks of hypertrophic cardiomyopathy (Figure 4b). KI mice also display alterations in left ventricular internal diameter end systole (decreased; relative to HET) and fractional shortening (increased; relative to WT)early indicators of abnormal cardiac function (Supp. Figure 1b).
These results demonstrate that Gaa c.1826dupA knock-in mice exhibit early-onset hypertrophic cardiomyopathya primary clinical feature of infantile-onset Pompe disease. Moreover, this work displays the utility of murine echocardiography as a robust diagnostic tool to assess cardiomyopathy in preclinical disease models.

Forelimb grip strength performance in Gaa c.1826dupA transgenic mice
To assess forelimb muscle strength in Gaa c.1826dupA transgenic mice, we measured peak tension force exerted by 3-month old WT, HET and KI mice using a murine-specific grip strength meter. In males, KI mice exhibited significant decreases in peak tension force when compared to WT and HET mice as well as a significant reduction in body mass relative to WT mice (Figure 5a). Female KI mice also exhibited significant decreases in peak tension force when compared to WT and HET mice and a significant reduction in body mass relative to WT mice (Figure 5b).
These results show that Gaa c.1826dupA knock-in mice exhibit early-onset musculoskeletal impairmenta key feature of infantile-onset Pompe disease. Gaa c.1826dupA knock-in mice also display a reduction in overall body mass which may be an early indicator of muscular dystrophy.

Gaa c.1826dupA transgenic mouse histology
To examine cardiac and skeletal muscle tissue structure and glycogen load, we performed periodic acid-Schiff (PAS) staining of heart, diaphragm and gastrocnemius muscle from 3-month old WT and KI mice. In contrast to WT mice, KI mice displayed aggregation of PAS staining in all three tissue types (Figure 6). Furthermore, when compared to WT fibers, KI cardiac and skeletal muscle fibers appear irregular in shape and display abnormal extracellular spacing between cells.
Altogether, these results show that Gaa c.1826dupA knock-in mice display early signs of muscle tissue pathology. The observed PAS-positive aggregates and irregular myocyte structural features are key markers of disease pathogenesis and progression in Pompe tissue.

Discussion
Currently, preclinical development of novel therapeutic options for Pompe disease rely primarily on the Gaa tm1Rabn/J knockout mouse model, developed in the late 1990s. While the Gaa tm1Rabn/J mouse is an appropriate model for evaluating new enzyme replacement and gene therapy strategies, a Pompe disease knock-in model bearing a Gaa mutation homologous to a known human pathogenic variant is much preferred for development of genome correction-based therapeutics.
This study demonstrates the successful generation of a new knock-in model of Pompe disease using CRISPR-Cas9 genome editing. Our data show the importance of optimizing HDRmediated knock-in efficiency via in vitro gRNA and donor template testing prior to in vivo application. We found that in silico combined rank scoring of gRNAs does not always correlate with actual experimental results. In fact, our Gaa c.1826 gRNA with the highest predicted rank score (gRNA-1) demonstrated the lowest in vitro on-target efficacy. Our data also suggest the enhancement of both gene editing and HDR events by utilizing a dual/multiple gRNA approach versus a single gRNA approach 24 , especially if there is overlap of candidate gRNA target sequences. We found that using a dual gRNA approach, with complete overlap in gRNA target sequences, resulted in the highest level of knock-in efficiency (12.6%) when compared to each gRNA alone (0 to 7.7%). and higher than the additive knock-in efficiencies of the gRNAs (11.2%).
Our preliminary in vitro results provide empirical evidence that the dual gRNA approach could increase the probability of generating cellular and murine knock-in disease models.
Consequently, we successfully isolated and characterized a clonal murine C2C12 myoblast line bearing a known pathogenic Pompe disease mutation -Gaa c.1826dupAusing the dual gRNA strategy. We found that Gaa c.1826dupA knock-in cells display molecular and biochemical analogy to human Pompe disease with significantly reduced Gaa transcript levels, undetectable GAA enzymatic activity and increased glycogen load. Further testing will need to be performed to confirm that this dual overlapping gRNA approach can be broadly applied to increase knock-in efficiency in other target gene loci and cell lines. Long-term phenotyping of Gaa c.1826dupA knock-in mice will provide further evidence of analogy to human Pompe disease. We are currently performing longitudinal physiological and histological assessment of Gaa c.1826dupA knock-in mice to determine natural history and disease progression.
Given the importance of determining cross-reactive immunogenic material (CRIM) status as it relates to severity of disease progression and immune response to GAA enzyme replacement therapy 26 , we will also aim to determine immune response of Gaa c.1826dupA knock-in mice to recombinant GAA protein. Altogether, this study provides evidence that Gaa c.1826dupA knock-in C2C12 myoblast cells and mice recapitulate, with the exception of infantile mortality, infantileonset Pompe disease. Our results validate their use as models of Pompe disease for preclinical evaluation of genome correction-based and other therapeutic strategies.

Gaa c.1826 target locus guide RNA and donor ssODN design
In silico design of CRISPR-Cas9 guide RNAs (gRNAs) specific for the Gaa c.1826 target locus was performed using Genetic Perturbation Platform (GPP) sgRNA Designer 13 (Broad Institute).
Candidate gRNAs were selected using the following criteria: 1) top combined rank score (based upon on-target efficacy and off-target specificity scores) and 2) proximity of predicted Cas9 nuclease cut site to the Gaa c.1826 target locus. Further potential gRNA off-target analysis was performed using Genome Target Scan (GT-Scan) 14 . Three gRNAs were used in this study (Table 1) Following puromycin-resistant selection, single cell clones were selected by standard serial dilution methods in 96-well plates at the presence of 2.5μg/mL puromycin dihydrochloride.
Single cells clones were identified and maintained until sequencing results confirm clonal cell genotype.

Experimental Animals.
The use and care of animals used in this study adhere to the guidelines of the NIH Guide for the Care and Use of Laboratory Animals, which are utilized by the CHOC Children's Institutional Animal Care and Use Committee. All study procedures were reviewed and approved under CHOC Children's IACUC protocol #160902.
Whole genome sequencing and analyses were performed on G0 founder (Gaa Mosaic ) and G1 heterozygous (Gaa wt/c.1826dupA ) tail samples. In brief, 1 μg fragmented genomic DNA was ligated with adaptors using TruSeq DNA libraries and whole genome sequencing was performed on an Illumina HiSeq X Ten Sequencer at >40x read depth (Fulgent Genetics). WGS on-target and on-and off-target analysis was analyzed on OnRamp BioInformatics platform. Data were aligned to the Mouse genome (mm10) using BWA 18 . PCR artifacts were identified with the memtest utility from Sentieon 19 , and filtered out using samtools 20

Murine echocardiography
Prior to echocardiography, a depilatory cream was applied to the anterior chest wall to remove the hair. 3-month old mice were anesthetized with 5% isoflurane for 15 seconds and then maintained at 0.5% throughout the echocardiography examination. Small needle electrodes for simultaneous electrocardiogram were inserted into one upper and one lower limb.

Forelimb grip strength assay
One hour prior to grip strength measurement, 3-month old mice were transferred to behavioral room to acclimate subjects to test conditions. Following acclimatization, each mouse was weighed and placed on a forelimb pull bar. Peak tension force exerted by each animal was recorded by a mouse grip strength meter (Columbus Instruments). Each mouse performed 3 pulls per day over 3 consecutive days for a total of 9 pulls per test session. Peak tension force (N) was calculated as the average of each subject's 9 pulls over the test session.

Tissue Harvesting, Processing, and Histological Staining
Tail biopsies for genotyping were collected on postnatal day 7 (founder mice) or postnatal day 21 (G1 & G2 mice). Genomic DNA was extracted using Agencourt® DNAdvance TM genomic DNA isolation kit (Beckman Coulter) with proteinase K and DTT. 3-month old mice were euthanized using CO2 asphyxiation and transcardially perfused with phosphate-buffered saline (PBS) followed by 4% PFA for histological staining or PBS alone for biochemical analyses.
Heart, diaphragm and gastrocnemius muscle tissue were harvested in this study. Tissue samples for biochemical studies were rapidly frozen and stored at -80°C; tissues for histological staining were processed and embedded in paraffin blocks for sectioning at 4μm thickness and Periodic acid-Schiff (PAS) staining was performed.

Quantitative real-time PCR
Total RNA was extracted from C2C12 myoblasts or postnatal day 21 tail tip samples using Direct-zol RNA miniprep kit (Zymo Research) and reverse-transcribed using High Capacity cDNA Reverse Transcription kit (Applied Biosystems) following manufacturer's instructions.

GAA enzymatic activity assay
For biochemical analysis, frozen C2C12 myoblast pellets or mouse tissues were homogenized in CelLytic M cell lysis reagent (MilliporeSigma). α-glucosaidase enzyme activity was assessed as previously described with minor modifications 22  Fluorescence measurements were obtained using an FLx800 spectrofluorophotometer (BioTek) at excitation and emission wavelengths of 360 nm and 460 nm, respectively. In order to eliminate the background activity caused by maltase-glucoamylase in liver tissues, a final concentration of 3 µM acarbose (Cayman Chemical Company) was added in α-glucosaidase enzyme activity reaction for liver samples. One activity unit was defined as 1 nmol converted substrate per hour. Protein concentration was estimated using Pierce BCA assay kit and bovine serum albumin was used as a standard. Specific activity was defined as units of activity per mg of protein.

Glycogen assay
Tissue glycogen levels were measured using a glycogen assay kit (Sigma-Aldrich) following manufacturer's instructions. In brief, 10 µL tissue homogenate was incubated with hydrolysis enzyme reaction mixture in a total volume of 50 µL at room temperature for 30 min before adding 50 µL development enzyme reaction mixture for 30 min incubation at room temperature.
Absorbance at 570 nm was measured using a spectrophotometer (Multiskan FC Microplate Photometer, Thermo Fisher). A standard curve was generated using standard glycogen solution in the reaction. A reaction without hydrolysis enzyme treatment was used for background correction (endogenous glucose) for each sample.

Periodic acid-Schiff staining
C2C12 cell lines were seeded on Matrigel®-coated 18mm glass coverslips at low density (2.5x10 3 cells) in culture media at 37°C with 5%CO2. 24h post-plating, culture media was replaced with serum-free culture media and maintained for an additional 72 hours with daily replacement of serum-free culture media. 96h post-plating, cells were fixed with 4% paraformaldehyde (4% PFA, Electron Microscopy Sciences) for 30 min at room temperature.

Statistical analysis
All graphs and statistical comparisons were generated using GraphPad Prism 8. Statistical analyses were performed using the two-tailed unpaired t-test or one-way ANOVA followed by Tukey's HSD test. All data are presented as mean ± SD.