A humanized knock-in Col6a1 mouse recapitulates a deep-intronic splice-activating variant

Abstract

Antisense therapeutics such as splice-modulating antisense oligonucleotides (ASOs) are promising tools to treat diseases caused by splice-altering intronic variants. However, their testing in animal models is hampered by the generally poor sequence conservation of the intervening sequences between human and other species. Here we aimed to model in the mouse a recurrent, deep-intronic, splice-activating, COL6A1 variant, associated with a severe form of Collagen VI-related muscular dystrophies (COL6-RDs), for the purpose of testing human-ready antisense therapeutics in vivo. The variant, c.930+189C>T, creates a donor splice site and inserts a 72-nt-long pseudoexon, which, when translated, acts in a dominant-negative manner, but which can be skipped with ASOs. We created a unique humanized mouse allele (designated as “h”), in which a 1.9 kb of the mouse genomic region encoding the amino-terminus (N-) of the triple helical (TH) domain of collagen a1(VI) was swapped for the human orthologous sequence. In addition, we also created an allele that carries the c.930+189C>T variant on the same humanized knock-in sequence (designated as “h+189T”). We show that in both models, the human exons are spliced seamlessly with the mouse exons to generate a chimeric mouse-human collagen a1(VI) protein. In homozygous Col6a1 ^{h+189T/h+189T} mice, the pseudoexon is expressed at levels comparable to those observed in heterozygous patients’ muscle biopsies. While Col6a1^h/h mice do not show any phenotype compared to wildtype animals, Col6a1 ^h/h+189T and Col6a1 ^{h+189T/h+189T} mice have smaller muscle masses and display grip strength deficits detectable as early as 4 weeks of age. The pathogenic h+189T humanized knock-in mouse allele thus recapitulates the pathogenic splicing defects seen in patients’ biopsies and allows testing of human-ready precision antisense therapeutics aimed at skipping the pseudoexon. Given that the COL6A1 N-TH region is a hot-spot for COL6-RD variants, the humanized knock-in mouse model can be utilized as a template to introduce other COL6A1 pathogenic variants. This unique humanized mouse model thus represents a valuable tool for the development of antisense therapeutics for COL6-RDs.

Introduction

Collagen VI-related muscular dystrophies (COL6-RDs) form a group of disorders that span a broad severity spectrum ranging from the debilitating and severe Ullrich type to the milder Bethlem form, connected by intermediate phenotypes. Characteristic manifestations include progressive muscle weakness, worsening proximal joints contractures, distal joints hyperlaxity, and progressive respiratory failure. In Ullrich type, the disease manifests at birth and without ventilatory support, can be fatal in the first two decades of life. In the milder cases, symptoms can become evident in young adulthood with slower progression but with no impact on life expectancy (reviewed in (1, 2)). Symptoms are currently managed with physical therapy and ventilatory support, but there are no specific therapies available yet.

COL6-RDs are caused by the absence or dysfunction of collagen VI, a microfibrillar extracellular matrix protein abundant in several organs including skeletal muscles where it surrounds myofibers (1). The three main genes, COL6A1, COL6A2, and COL6A3, encode the three major collagen VI alpha peptide chains (a1(VI), a2(VI), and a3(VI)) (1, 3) and are mainly expressed by the interstitial fibro-adipogenic progenitor cells (FAPs) (4–6). In FAPs, the three chains intertwine through their respective TH domains, proceeding from their C- towards their N-termini, to form monomers. The monomers form dimers and tetramers, which are secreted into the extracellular space where they polymerize and form the collagen VI meshwork (reviewed in (7, 8)).

Pathogenic COL6 variants can be inherited in an autosomal recessive manner, but, more frequently, they are acquired de novo and act dominant negatively (1, 7). The most frequent dominant-negative variants cause either glycine substitutions in the Gly-X-Y repeats or in-frame exon deletions, and they occur predominantly in the exons encoding the N-termini TH (N-TH) of any of the three a(VI) peptides. In this location, they allow the mutant collagen a(VI) chains to incorporate into monomers, and further dimers and tetramers, but they affect the function of the collagen VI tetramers in the matrix, hence their dominant-negative effect (1, 7). More recently, a surprisingly common and recurrent pathogenic variant has been identified in intron 11 of COL6A1: c.930+189C>T. This intronic variant is also located within the N-TH-encoding region and creates a splice donor (SD) site that, combined with an upstream cryptic splice acceptor (SA) site, inserts an in-frame 72-nt-long pseudoexon into the mature mRNA transcript (9). We previously showed that the pseudoexon is only inserted in approximately 50% of the mRNA transcribed from the +189T allele, yet, this expression level is sufficient to have a dominant-negative effect on collagen VI assembly and to cause the severe clinical manifestations of Ullrich type of COL6-RDs (10, 11). COL6-RDs downstream pathophysiology in muscle is complex and not fully understood (7, 8); therefore, addressing the disease upstream at the genetic source, such as with ASOs, is a promising therapeutic approach. We have shown that skipping the pseudoexon with ASOs restores nearly healthy collagen VI matrix in cultured patient fibroblasts (10), but in vivo testing has not been feasible because of the unavailability of a suitable animal model given that human intronic sequences significantly diverge from the corresponding mouse sequences.

Genomic humanization involves the introduction of human genomic sequences, coding and non-coding regulatory sequences included, into a trans-species genome (12). While the process for obtaining these humanized loci was previously tedious, with the currently available genome editing technologies it can be achieved rapidly (12, 13). Entire – or nearly entire – human genes have thus been successfully introduced, and expressed, either in a transgenic location (14), or at the endogenous locus in the trans-species genome, substituting their orthologous counterpart (15). Partial genomic humanization has also been successfully applied (16). Some of these models have proven invaluable to study basic gene regulation and function in the context of human health and disease (12). More recently, humanized animals have been proposed as models to introduce human pathogenic variants and test precision medicine therapeutics (17, 18).

Here we report the partial genomic humanization of the N-TH domain of the mouse Col6a1 gene (encompassing 1.9-kb and spanning from intron 8 to intron 14). This region includes the relevant exonic and intronic sequences for dominantly acting variants in COL6A1, including the pseudoexon-inducing c.930+189C>T variant. We created a humanized wild-type allele (harboring the reference sequence), and a humanized pathogenic allele (carrying the c.930+189T variant), and validated their use as preclinical models with molecular, biochemical, histological, and behavioral assays.

Results

Generation of two genomic humanized alleles in Col6a1

The variant to model (COL6A1 c.930+189C>T) is schematized in Figure 1A. COL6A1/Col6a1 coding sequences are highly conserved between human and mouse with 83.2% nucleotide (NM_001848.3 and NM_009933.5) and 90% amino acid (NP_001839.2 and NP_034063.1) sequence identities. In contrast, the intronic sequences are poorly conserved. For instance, the alignment of the human intron 11 sequence with the mouse sequence shows 45.1% identity and 44.3% of gaps. Importantly, the ‘aggc’ site that is converted to ‘AGgt’ due to the c.930+189C>T variant is absent in the mouse intron 11 sequence. Thus, we sought to substitute part of the mouse Col6a1 gene with the human sequence. We replaced a 1,794-nt-long segment of the mouse Col6a1 gene (encompassing introns 8 to 14) with the 1,934-nt-long orthologous human sequence (Figure 1B). We generated two humanized knock-in alleles, carrying either the reference cytosine at position c.930+189 (called “h”), or the pathogenic thymine (called “h+189T”) (Figure 1B). The humanized protein sequence spans from mouse Gly268 to Asp351 and includes the Cys344, important for collagen VI dimerization.

Figure 1. Generation and validation of a Col6a1 humanized knock-in mouse to model the c.930+189C>T pseudoexon-inducing variant.

(A) Schematic of the variant to model: the c.930+189C>T variant in intron 11 of the COL6A1 gene, which creates a donor splice site (PE-SD) and which, upon concomitant activation of an upstream cryptic splice acceptor site (PE-SA), causes the insertion of a 72-nucleotide-long pseudoexon (PE). (B) Two Col6a1 humanized knock-in alleles were generated using a CRISPR/Cas9-induced homologous recombination strategy to swap the mouse genomic sequence spanning the region between introns 8 and 14 with the orthologous human sequence. One allele contains the reference cytosine at position +189 (referred to as allele ‘h’), and the other allele contains the pathogenic thymine (referred to as ‘h+189T’). (C) Col6a1 transcripts were enriched using a probe capture method prior to long-read RNA sequencing. Sashimi plots depict the splicing of human exons with the mouse exons in quadriceps of 8-week-old males.

Expression and splicing of the Col6a1 hybrid gene

With long-read RNA sequencing, human exons 9 to 14 were found to seamlessly splice with the mouse exons and to generate full-length mouse-human chimeric Col6a1 transcripts in quadriceps tissue (Figure 1C, Supplementary Figure S1). Moreover, the c.930+189C>T variant prompted the utilization of the pathogenic splice donor site, in both heterozygous (Col6a1^h/h+189T) and homozygous (Col6a1^{h+189T/h+189T}) tissues. Consistent with the splicing outcomes we previously described in patient samples (10), at least three mRNA species produced by the variant allele were identified (as evident in Col6a1^{h+189T/h+189T} mouse tissues): 1) canonical splicing (from exon 11 to 12; isoforms 2 and 3, Supplementary Figure S1C), 2) 72-nt-long pseudoexon insertion (with activation of the cryptic SA); isoforms 1 and 4, Supplementary Figure S1C), or 3) exon 11 extension (up to the pathogenic SD), for a 187-nt-long insertion predicted to be subject to nonsense-mediated decay; isoform frequency <3% and not depicted).

We next determined the percent pseudoexon inclusion in various tissues using digital PCR assays. In 8-week-old Col6a1^h/h+189T males, the percent pseudoexon inclusion varied from 7.6% to 10.6% in quadriceps, tibialis anterior, gastrocnemius and diaphragm muscles, whereas in Col6a1^{h+189T/h+189T} tissues, the percent pseudoexon was higher (from 15.0% to 20.8%, n=4 animals per genotype and tissue; Figure 2A). In human biopsy samples, pseudoexon expression levels were comparable to those observed in the Col6a1^{h+189T/h+189T} tissues (average of 24.7%, n=3; Figure 2A). Thus, the Col6a1^{h+189T/h+189T} genotype faithfully recapitulates the human transcript isoform expression profile. The percent pseudoexon expression in quadriceps did not significantly vary between females and males or between 8-week-old and 24-week-old mice (n=4 per sex and age group, Figure 2B).

Figure 2. Pathogenic pseudoexon expression in the Col6a1 h+189T model.

(A, B) Isoform-specific digital PCR assays were used to quantify the percentage of total Col6a1 humanized transcripts that include the 72-nt-long pseudoexon, in various tissues of 8-week-old male mice compared to heterozygous patients’ biopsies (A), and in quadricep muscles of 8-week-old and 24-week-old male and female mice (B). Data points represent individual mice (n=4 per group) or patients (n=3). Data presented as mean ± SD. Quad = quadriceps; Dia = diaphragm; TA = tibialis anterior; Gast = gastrocnemius; Bx = patient biopsies; F = females; M = males. (C) Translation of the pseudoexon peptide was assessed by immunoblotting with a pseudoexon-specific antibody (Pex11-a1(VI)), in quadriceps of 8-week-old and 24-week-old Col6a1^{h+189T/h+189T} males (n=4). (D) The top of the double band (in (C)) corresponds to the size of the a1(VI) protein and was used for quantification. Data points represent pseudoexon-containing a1(VI) protein expression levels normalized to tubulin (n=4 mice per group). Kolmogorov-Smirnov test was applied and showed no significant difference between the time points.

The pseudoexon encodes a 24-amino-acid-long stretch (2.7 kDa predicted) that interrupts the Gly-X-Y repeats. We previously generated an antibody (Pex11-a1(VI)) using the 24-amino-acid-long peptide as the immunogen (10, 11). The Pex11-a1(VI) antibody showed signal in quadriceps of 8-week-old and 24-week-old Col6a1^{h+189T/h+189T} males (top band, Figure 2C), at a molecular weight comparable to the a1(VI) chain (third panel, Figure 2C). The antibody detected an additional band of lower molecular weight that is under investigation. The levels of the pseudoexon-containing a1(VI) protein and of total collagen a1(VI) do not significantly increase from 8 to 24 weeks (Figure 2D).

Mouse and muscle phenotyping

All Col6a1^h/h+189T or Col6a1^{h+189T/h+189T} mice developed normally and exhibited lifespans comparable to Col6a1^h/h and Col6a1^+/+(wild-type) control animals. There were no significant body weight differences between any of the h+189T-containing genotypes and the Col6a1^h/h or Col6a1^+/+ animals, from 4 weeks to 24 weeks of age, in males and females (n34 per group; Figure 3A, B). The muscle masses of Col6a1^{h+189T/h+189T} tibialis anterior and gastrocnemius muscles, however, were significantly decreased compared to Col6a1^+/+ and Col6a1^h/h tissues, at 8 and 24 weeks, for both males and females (n34 per group; Figure 3C, D), except for males’ tibialis anterior at 24 weeks in which the muscle masses were decreased but did not reach significance. In Col6a1^h/h+189T mice, muscles weighted less compared to Col6a1^+/+ and Col6a1^h/h mice, but this decrease did not consistently reach statistical significance.

Figure 3. Body and muscle weights of Col6a1 h+189T mice.

(A, B) Whole body weights were recorded every four weeks from 4 to 24 weeks old, in males (A), and females (B). Statistical analyses were performed with one-way ANOVA followed by Tukey’s multiple comparisons test and showed no significant difference. (C, D) Masses of tibialis anterior and gastrocnemius muscles were recorded in 8-week-old and 24-week-old males (C) and females (D) (n=4–11). Data points represent biological replicates and data are presented as mean ± SD. Statistical analyses were performed with a Kruskal-Wallis ANOVA test followed by Dunn’s multiple comparisons. *p<0.05; **p<0.005, ***p<0.0005.

We next investigated for histopathological signs of muscular dystrophy in the humanized mice skeletal muscles using hematoxylin and eosin staining. Qualitatively, we observed mild changes in the number of centrally nucleated fibers in Col6a1^h/h+189T and in Col6a1^{h+189T/h+189T} mice, and we did not observe clear signs of increased fatty infiltration or increased matrix production in these genotypes (Figure 4A). The percent of centrally nucleated fibers, an indicator for degeneration/regeneration cycles that can reflect the degree of muscle pathology, was increased in Col6a1^{h+189T/h+189T} quadricep muscles of both males and females (n=4–5 mice per group; Figure 4B), although without reaching statistical significance.

Figure 4. Muscle histology and function in Col6a1 h+189T mice.

(A) Representative images of tissue sections of tibialis anterior (top row) and quadriceps muscles (bottom row) of 24-week-old male mice stained with hematoxylin and eosin. Mild histopathological signs of muscular dystrophy were detected in Col6a1^h/h+189T and Col6a1^{h+189T/h+189T}, evidenced by the slightly increased percent of centrally nucleated fibers (arrows). (B) The percent of centrally nucleated fibers was manually counted on tissue sections stained with wheat germ agglutinin and DAPI. M = males; F = females. Data points represent biological replicates (n=4–5). Data presented as mean ± SD. Statistical analyses were performed with a Kruskal-Wallis ANOVA test followed by Dunn’s multiple comparisons. (C, D) Absolute force of the forelimb grip strength was measured using a grip strength meter and is reported in grams (g). Grip strength was measured every 4 weeks from 4 to 24 weeks of age in males (C) and females (D). Graphs depict the mean grip strength ± SD (n=5–9).

We used the grip strength test to functionally assess forelimb muscles’ strength. Strikingly, the grip strengths of both Col6a1^h/h+189T and Col6a1^{h+189T/h+189T} males and females were weaker at all ages recorded between 4 and 20 weeks compared to the control groups (Figure 4C, D). The decrease in grip strength was statistically significant between the control Col6a1^+/+ and Col6a1^h/h mouse groups compared to the disease Col6a1^h/h+189T and Col6a1^{h+189T/h+189T} mouse groups at most tested age timepoints (Supplementary Figure S2).

Discussion

In the era of personalized medicine, genomic humanized models are becoming increasingly valuable tools for the in vivo testing of nucleic acid-based therapies tailored to specific human pathogenic variants (18). Modeling a pathogenic variant within a humanized gene context provides the desired target for nucleic acid-based drugs that rely on sequence recognition, such as antisense oligonucleotides or genome editing technologies. This is particularly relevant for splice-modulating therapeutics targeted at intronic sequences. In addition, humanized models enable optimizing the delivery of these human-ready drugs to the relevant tissues and cell types in the context of the disease studied.

To facilitate the preclinical development of a splice-modulating therapy for the recurrent intronic COL6A1 variant (c.930+189C>T) as an ideal target for pseudoexon-skipping approaches, we generated a humanized mouse model that recapitulates the pseudoexon-inducing splicing event caused by the pathogenic variant. We partially humanized the Col6a1 gene as a knock-in allele (termed “h”). Like previously generated humanized mouse models, where the canonical splice sites within a humanized gene were faithfully recognized by the mouse splicing machinery (14–16), the h allele in our model expressed full-length and functional chimeric mouse/human Col6a1 transcripts. Modeling intronic and splice-altering variants, such as the c.930+189C>T, in humanized backgrounds can be more challenging than only inserting a humanized sequence, however (19, 20). In mice carrying a single allele of the humanized Col6a1 that includes the intronic variant (termed “h+189T”), the pseudoexon inclusion levels did not reach the levels detected in muscle biopsies from heterozygous patients, suggesting that the pathogenic variant was underutilized in mouse tissues despite the humanization of several flanking exons and introns to maximize the likelihood of including potential cis-acting splicing regulators. However, when quantifying the pseudoexon inclusion levels in Col6a1^{h+189T/h+189T} mice tissues, we found that the pseudoexon inclusion from both alleles combined was very close to the levels detected in biopsies from the heterozygous human patients, thus providing a relevant disease model at the transcript level. In addition, we showed that in the h+189T mouse model, the pseudoexon was translated, and that the corresponding peptide was detected in proteins of molecular weight comparable to the a1(VI) protein, thus validating that the Col6a1^{h+189T/h+189T} mouse is a relevant preclinical model to test the on-target activity of splice-modulating therapeutics.

Collagen VI deficiency or dysfunction in various mouse models results in phenotypes milder than the clinical manifestations observed in COL6-RD patients (7, 21–24). Similarly, our Col6a1^{h+189T/h+189T} mice are milder in severity when compared to the Ullrich patients heterozygous for the c.930+189C>T variant (10). Nevertheless, Col6a1^h/h+189T and Col6a1^{h+189T/h+189T} mice show muscle weakness (assessed with the forelimb grip strength testing) and histological signs of myofiber degeneration and regeneration (seen with the moderately increased percent of centrally nucleated fibers). Grip strength assessment can be used as a longitudinal functional outcome measure when testing the effect of splice-modulating antisense oligonucleotides. It is noteworthy that the heterozygous Col6a1^h/h+189T genotype results in a phenotype not that different from the homozygous Col6a1^{h+189T/h+189T} genotype, even though the abundance of pseudoexon inclusion from the single allele is less. This may suggest a cumulative pathological effect of the pseudoexon peptide over time.

While this model carries the target human sequence to optimize on-target activity of nucleic acid-targeting drugs in vivo, it is of limited use to assess sequence-dependent off-target effects, since the genomic background consists of the mouse genome. Nevertheless, sequence-dependent off-targets can be predicted and assessed in cellular models. Moreover, toxicity due to components other than the primary sequence (based on the chemistry, for example) can be tested in organisms such as non-human primates.

In the COL6A1 gene, dominant-negative variants occur within the N-TH, mostly between exons 9 and 14. Other than the intron 11 variant described here, common COL6A1 variants also located in this N-TH domain include splice-site variants that result in exon 14 skipping (25, 26), and the G284R and G293R substitutions in exons 9 and 10, respectively (27, 28). Notably, since the three COL6 genes are haplosufficient (1, 2), COL6 variants like the ones described above are also promising targets for nucleic acid therapeutics, in particular for therapeutics that mediate allele-specific silencing, such as small interfering RNAs (29, 30), RNase H-recruiting gapmers (31), and gene editing (32). Using as a template the humanized knock-in allele we created (‘h’), it is now feasible to generate humanized models for each one of these variants on this background to then test human-ready antisense therapies. We anticipate that our unique humanized knock-in model will accelerate the development of nucleic acid therapeutics for COL6-RDs.

Material and Methods

Generation of the Col6a1^h/h, Col6a1^h/h+189T and Col6a1^{h+189T/h+189T} mice

The mouse and human sequences were aligned using NCBI Blastn suite. All mouse procedures were approved by the NINDS Animal Care and Use Committee (Animal Safety Protocol #1556). Manipulations of recombinant DNA was approved by the National Institutes of Health Institutional Biosafety Committee (Registration #RD-17-VI-01). The humanized knock-in mice were created using homology recombination induced by CRISPR/Cas9 at the National Human Genome Research Institute (NHGRI) Transgenic and Gene Editing Core facility. A combination of two guide RNAs were prepared with SpCas9 into ribonucleoprotein complexes and delivered to C57BL6/N × C57BL6/J zygotes by pronuclear injections, together with a targeted construct. The guide RNAs were selected to cleave 117 bp upstream of exon 9 and 66 bp downstream of exon 14 of the mouse Col6a1 and were synthesized by Horizon (Horizon Discovery/Dharmacon, Lafayette, CO). We created two targeted constructs containing the human genomic sequence from intron 8 to intron 14 of the COL6A1 gene flanked by 800 bp of mouse homology arms. One construct contained the reference cytosine at position c.930+189 (here referred to as “h”), and the other construct carried the pathogenic thymine at the position (here referred to as “h+189T”). The constructs included 160 bp of human intron 8 sequence upstream of exon 9 (preserving two potential branch point sites) and 65 bp of sequence downstream of exon 14. Each of the two targeted constructs was 3,520 bp in length and were synthesized and cloned (Epoch LifeScience, Missouri City, TX). The plasmids were prepared and purified by 2X CsCl gradients (Lofstrand Labs, Gaithersburg, MD). The targeted vectors were isolated by restriction enzymes prior to pronuclear injection. Zygotes were implanted into CBy6F1 pseudo-pregnant females. F0 pups DNA was extracted from tail biopsies and was analyzed by PCR and Sanger sequencing. For each of the two constructs, a single pup was identified with the expected sequence, and was bred with a C57BL/6J animal. Animals were backcrossed for 5 generations with C57BL/6J animals. Breeding colonies were maintained as Col6a1^h/h x Col6a1^h/h; Col6a1^+/h+189T x Col6a1^+/h+189T; or Col6a1^h/h x Col6a1^+/h+189T to generate all genotypes used in the study (Col6a1^+/+, Col6a1^h/h, Col6a1^h/h+189T, Col6a1^{h+189T/h+189T}). Animals were housed at the Intramural NINDS Building 35 vivarium under the following conditions: 12-hour light/dark cycles, 20–25 °C room temperature, 40–65% relative humidity, access to untreated drinking water and chow ad libitum. Genotyping was performed with a PCR amplification combining a single forward primer located in the mouse intron 8 upstream of the knock-in region (5’-cagttcagccttgatgcaaa-3’) with two reverse primers each hybridizing specifically a different allele, either the mouse wild-type allele (5’-gcagaggaaatcagctcagg-3’) or the humanized knock-in allele (5’-taaagcaccttccccatcac-3’). The difference in molecular weight of these two amplicons (66 bp) was sufficient to separate and visualize with a 2.5% agarose gel electrophoresis and to identify three genotypes: unmodified wild-type, and either heterozygous or homozygous for the humanized knock-in allele. When needed, we confirmed Col6a1^h/h, Col6a1^h/h+189T and Col6a1^{h+189T/h+189T} mouse genotypes using a custom genotyping assay (ThermoFisher, Waltham, MA) in which the probes detecting the ‘T’ (h+189T) or the ‘C’ (h) alleles were labeled with different fluorescent dyes (FAM and VIC, respectively). These PCR products were detected at end point on the QuantStudio6 Real-Time PCR instrument (ThermoFisher).

Mouse phenotyping

The mice body weights and forelimb grip strength measurements were assessed on the same day once a month (every 28 days). The mice were weighed before the grip strength assessment. The mice forelimb grip strength was measured with a digital grip strength meter (Bio-GS3, Bioseb, Pinellas Park, FL) using a standard grip strength assay (33, 34). The middle three values of five measurements were averaged and reported as the absolute force generated in grams (g). One-way ANOVA test was used followed by Tukey’s HSD as a post-hoc test to determine statistical significance for body weights and grip strengths values.

Patients

Specimen collection and use was approved by the National Institutes of Health Institutional Review Board (Protocol #12-N-0095).

Tissue processing

For human muscle biopsies, RNA was isolated from fast-frozen tissue sections (prepared on the CM1950 cryostat, Leica, Wetzlar, Germany) using Trizol (ThermoFisher, Waltham, MA). For mouse specimen, muscles were either fast-frozen in dry ice-prechilled isopentane (for histology processing) or cut into ~20 mg pieces and snap-frozen in liquid nitrogen (for RNA and protein processing). To isolate RNA, tissues were homogenized in Trizol using the Red Eppendorf Lysis kits and the Bullet Blender (Next Advanced, Troy, NY). RNA was then isolated using manufacturer guidelines. To isolate proteins, tissues were homogenized and lysed in a Urea/Thiourea Buffer (7 M urea, 2 M thiourea, 4% (w/v) CHAPS in 30 mM Tris-HCl, pH 8.5) (35) containing 1X cOmplete protease inhibitor cocktail (MilliporeSigma, Burlington, MA), using the Green Eppendorft Lysis kits and the Bullet Blender (Next Advance). Reagent DX (Qiagen, Germantown, MD) was added to the lysis buffer at 0.5% v/v to prevent foaming during homogenization.

Long-range sequencing

A panel of 101 120-mer probes was designed to span the entire coding sequence of the mouse/human hybrid Col6a1 coding sequence. This set also included probes specific to each allele at the c.930+189 position (IDT xGen Pool Design, Integrated DNA Technologies, Coralville, IA). RNA samples were first cleaned up using bead purification (AMPure RNA-XP, Beckman Coulter, Indianapolis, IN). First-strand synthesis, transcripts capture, library preparation and PacBio Sequel II sequencing were performed at Sequencing Facility – Long Read Technology Center for Cancer Research of the National Cancer Institute (Frederick, MD). Reads were aligned to a custom reference genome. A publicly available workflow (v4.0) to identify transcripts in PacBio single-molecule sequencing data was used (https://github.com/PacificBiosciences/IsoSeq). The specific workflow used included two main execution steps, Clustering (https://isoseq.how/clustering/cli-workflow.html) and Classification (https://isoseq.how/classification/workflow.html). Tools used as part of this workflow included those supported in the SMRT^® Analysis software suite v13.0.0.207600 (https://www.pacb.com/support/software-downloads/). All computational analyses were performed on the NIH HPC Biowulf cluster (http://hpc.nih.gov). Sashimi plots were generated using ggsashimi (36). BAM files were loaded into VIsoQLR (37), which was used to detect and quantify splice isoform species using standard parameters, except for adjustments of the base-pair lengths made to the detected terminal exons to compensate for sequencing biases.

Digital PCRs (dPCRs)

RNA was converted to complementary DNA (cDNA) using SuperScript IV Reverse Transcriptase (ThermoFisher). Using PrimerQuest Tool (Integrated DNA Technologies) to design custom taqman-based assays, we selected two assays that specifically amplify either pseudoexon-containing transcripts only, or any humanized transcript (spanning exons 12 to 14), ordered as PrimeTime assays. We purchased a pre-designed mouse Csnk2a2 assay (Mm.PT.58.10226157) as the housekeeping control (Integrated DNA Technologies). We performed quantitative PCRs on complementary DNA (cDNA) samples by combining each isoform-specific assay with the housekeeping assay. Reaction partitioning and amplification, and fluorescence detection were done using the QIAcuity digital PCR instrument (Qiagen). For each assay, thresholds were determined manually and applied to all samples. The concentration values (copies/μL) obtained for each isoform-specific assay were normalized to the concentration obtained for the housekeeping assay. Percent pseudoexon inclusion was calculated by dividing the pseudoexon-containing transcript concentration by the concentration of all humanized transcripts. For the human biopsy samples, the same isoform-specific assays were utilized, whereas a human CSNK2A2 pre-designed assay (Hs00176505_m1, ThermoFisher) was used as the housekeeping control.

Immunoblots

Protein lysates were quantified using the 660 nm Protein Assay Kit (Pierce/ThermoFisher) and read on the Victor Nivo multimode plate reader (PerkinElmer, Waltham, MA). Protein samples were diluted in a loading dye containing 100 mM dithiothreitol (DTT). Samples were boiled at 95 °C for 5 min before loading (60 μg) on 4–12% Bis-Tris gel (ThermoFisher) and ran with 1X MES buffer (ThermoFisher) for 45 min at 125 volts. Transfer to PVDF membrane (MilliporeSigma) was done on the XCell II Blot module (ThermoFisher) in 1X Transfer buffer (ThermoFisher) supplemented with 5% methanol at 25 volts for 90 min. Membranes were blocked in Intercept Blocking Buffer (LI-COR, Lincoln, NE) for 1 hr before adding primary antibodies: anti-tubulin (T5168 (MillporeSigma), diluted 1:10,000 in blocking solution) or anti-Collagen VI (ab199720 (Abcam, Cambridge, UK), diluted 1:1,000 in blocking solution) for 90 min. The affinity-purified anti-rabbit pseudoexon antibody (Pex11-a1(VI)), gift from Raimund Wagener, was diluted 1:1,000 in blocking solution. Secondary antibodies used were 680RD goat anti-rabbit IgG (926-68071, LI-COR), diluted 1:10,000 or 800CW goat anti-mouse IgG (926-32210, LI-COR) diluted 1:15,000 in blocking solution and incubated for 1 hr. Washes were done in phosphate-buffered saline containing 0.05% tween-20 (MP Biomedicals, Santa Ana, CA). Detection was done on the Odyssee Clx Imaging System (LI-COR) using the Image Studio Software Ver 5.2 (LI-COR). Areas under the curves were measured with Fiji and used for quantification. Values for collagen a1(VI) were normalized to tubulin. Unpaired Kolmogorov-Smirnov test was applied.

Histology

Tissues were sectioned (10 μm) using the CM1950 cryostat (Leica). For hematoxilin and eosin staining, frozen slides were first thawed and incubated in filtered Harris Hematoxylin (VWR, Radnor, PA) for 8 min before being briefly dipped in water containing 0.2% ammonium hydroxide for 50 sec. The samples were then put through a series of additional dips: 95% EtOH for 5 sec, Eosin (VWR) for 1 min, 95% EtOH for 5 sec, followed by 95% EtOH for 2 min (3X), 100% EtOH for 2 min (3X), and Xylene for 2 min (3X). Specimen were covered with Permount (Fisher Scientific, Hampton, NH). For the quantification of centrally nucleated fibers, a wheat germ agglutinin and DAPI stain was used. Briefly, slides were fixed with 4% PFA (Electron Microscopy Sciences, Hatfield, PA) for 10 min and blocked with phosphate-buffered saline solution containing 5% normal goat serum and 0.3% Triton X-100 (Sigma-Aldrich, St. Louis, MO) for 30 min. Secondary antibodies used were WGA-488 (ThermoFisher) diluted 1:500 and DAPI in blocking buffer and incubated with samples for 15 min. Slides were mounted with Fluoromount-G (Southern Biotech, Birmingham, AL). The percent of centrally nucleated fibers was manually counted with the help of the Fiji software. Kruskal-Wallis test followed by Dunn’s multiple comparisons was applied.

Microscopy

Images were captured using the 20X objective on the Eclipse Ti microscope system (Nikon, Tokyo, Japan) and using the NIS-Elements AR Software (Nikon).