Background: Skeletal muscle myofibers can be separated into functionally distinct cell types that differ in gene and protein expression. Current single cell expression data is generally based upon single nucleus RNA, rather than whole myofiber material. We examined if a whole-cell flow sorting approach could be applied to perform single cell RNA-seq (scRNA-seq) in a single muscle type. Methods: We performed deep, whole cell, scRNA-seq on intact and fragmented skeletal myofibers from the mouse fast-twitch flexor digitorum brevis muscle utilizing a flow-gated method of large cell isolation. We performed deep sequencing of 763 intact and fragmented myofibers. Results: Quality control metrics across the different gates indicated only 171 of these cells were optimal, with a median read count of 239,252 and an average of 12,098 transcripts per cell. scRNA-seq identified three clusters of myofibers (a slow/fast 2A cluster and two fast 2X clusters). Comparison to a public skeletal nuclear RNA-seq dataset demonstrated a diversity in transcript abundance by method. RISH validated multiple genes across fast and slow twitch skeletal muscle types. Conclusion: This study introduces and validates a method to isolate intact skeletal muscle myofibers to generate deep expression patterns and expands the known repertoire of fiber-type-specific genes. Keywords: Single cell RNA-sequencing, Skeletal muscle, Twitch, Fiber Background different muscles of the body reflecting different func- Skeletal muscle is a voluntary, striated muscle found tional needs [2, 4]. Our understanding of all of the genes throughout the body with contraction regulated by nerve that vary across these fiber types is limited, although impulses through the neuromuscular junction (NMJ). many well-characterized examples such as myosin heavy Skeletal muscles consist of different fiber types delin- chains, calcium ATPase pumps, and metabolic proteins eated by the isoform of the myosin heavy chain they ex- are known. Only recently has there been an effort to press, metabolic function, and other properties . catalog the entirety of fast-/slow-twitch expression dif- Mouse skeletal muscles are comprised of slow fibers ferences by single cell approaches. (type 1) and three types of fast fibers: type 2A, type 2B, The most comprehensive gene expression study was and type 2X [2–4]. These fiber types are variable across performed in mice using DNA microarrays across ten type 1 and ten type 2B fibers . Single cell RNA- sequencing (scRNA-seq) also has been performed in * Correspondence: email@example.com skeletal muscle and muscle cultures. However, until Department of Pathology, Johns Hopkins University School of Medicine, Ross Bldg. Rm 632B, 720 Rutland Avenue, Baltimore, MD 21205, USA quite recently, the large size of skeletal myofibers has Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Verma et al. Skeletal Muscle (2021) 11:13 Page 2 of 10 precluded them from these datasets, which are instead Inhibitor, Invitrogen). Single cell libraries were then pre- predominately satellite cells and other supporting cell pared using the previously described mcSCRB-seq proto- types [6–13]. A recent publication used SMART-Seq to col [21, 22]. Briefly, cells were subjected to proteinase K evaluate three fast fiber mouse fibers , and several sin- treatment followed by RNA desiccation to reduce the re- gle nucleus RNA-seq (nuc-seq) projects have also added action volume. RNA was subsequently reverse tran- to the literature [15–17]. The totality of these studies scribed using a custom template-switching primer as strongly suggests there are numerous expression differ- well as a barcoded adapter primer. The customized ences between skeletal muscle fiber types and demon- mcSCRB-seq barcode primers contain a unique 6 base strates a need for new approaches to capture this diversity. pair cell-specific barcode as well as a 10-base pair unique The Kwon lab recently established a protocol for molecular identifier (UMI). Transcribed products were scRNA-seq of large mature cardiac myocytes through pooled and concentrated, with unincorporated barcode large particle fluorescence-activated cell sorting (FACS) primers subsequently digested using Exonuclease I treat- . We ascertained if this method could be used to iso- ment. cDNA was PCR-amplified using Terra PCR Direct late the even larger skeletal muscle myofibers for Polymerase (Takara Bio). Final libraries were prepared scRNA-seq, as typically only small cells are captured in using 1ng of cDNA per library with the Nextera XT kit traditional scRNA-seq methods [6, 8]. We utilized the (Illumina) using a custom P5 primer as previously flexor digitorum brevis (FDB), a well-characterized fast- described. twitch fiber muscle of the base of the foot, made up pre- dominately of type 2A (IIa) and type 2X (IIx) fibers . scRNA-seq sequencing and analysis Our goal was to validate this whole cell capture method, Pooled libraries were sequenced on two high-output lanes compare whole cell single cell data to single nuclear of the Illumina NextSeq500 with a 16-base pair barcode data, and characterize this important model muscle. read, 8-base pair i7 index read, and a 66-base pair cDNA read design. To analyze sequencing data, reads were Methods mapped and counted using zUMIs 2.2.3 with default set- Isolation and sequencing of adult skeletal myofibers tings and barcodes provided as a list . zUMIs utilizes All animal studies were approved by the Institutional STAR (2.5.4b)  to map reads to an input reference Animal Care and Use Committee at Johns Hopkins and genome and featureCounts through Rsubread (1.28.1) to all methods were performed in accordance with the rele- tabulate counts and UMI tables [24, 25]. Reads were vant guidelines and regulations. This study used adult mapped to the mm10 version of the mouse genome. We male mice (>3 months) from the C57BL/6J and DBA2 used GRCm38 from Ensembl concatenated with ERCC backgrounds (Jackson Labs, Bar Harbor, ME). All mice spike-in references for the reference genome and gene an- were first anesthetized in an induction chamber using notations. Dimensionality reduction and cluster analysis isoflurane until breathing rate has slowed to 1 Hz and were performed with Seurat (2.3.4) . were unresponsive to rear toe pinches. This was followed by cervical dislocation prior to excision of any Seurat-based analysis muscles. To isolate skeletal myofibers, we performed Analysis was performed using the Seurat R toolkit collagenase-based digestion of the flexor digitorum bre- V3.1.1 for this dataset . Initial filtering removed vis (FDB), a short muscle of the hind feet, as per previ- lower quality cells (read count <5000 RNAs detected or ously established protocols . We performed intact >20% mitochondrial genes) before sctransform and fragmented FDB studies. The FDB was transferred normalization . We performed principal components to a dish containing DMEM with 1% penicillin/strepto- analysis (PCA) of the top 3000 variable genes based on mycin, 1% fetal bovine serum, and 2mg/mL Collagenase the Seurat sctransform algorithm and used the top 4 for Type II (Worthington). Muscle was digested for 1.5 h in downstream analysis. We generated a Seurat workflow a 37°C cell incubator with 5% CO . Subsequently, the that identifies a subset of genes with high cell-to-cell muscle was transferred to a dish containing media with- variation within the scRNA-seq data. A Uniform Mani- out collagenase and gently triturated to release single fold Approximation and Projection (UMAP) was gener- myofibers. Large undigested chunks and tendons were ated alongside a heat map representing the top genes in removed with tweezers prior to single cell isolation. A clusters as determined by each gene set used for PCA. COPAS SELECT Flow Pilot Platform (Union Biome- trica) was employed, as described below. Analysis of a public nuclear RNA-Seq dataset These sorted cells were placed individually into 96- The snRNA analysis was done in Seurat V3.1.1 taking well plates. Capture plate wells contained 5 μl of capture data available from Dos Santos et al.  https://www. solution (1:500 Phusion High-Fidelity Reaction Buffer, ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE150065. New England Biolabs; 1:250 RnaseOUT Ribonuclease Their data consisted of four sets of matrices, one of Verma et al. Skeletal Muscle (2021) 11:13 Page 3 of 10 which was a mix of tibialis, extensor digitorum longus, fluorescent dye (Opal dyes 520 and 570, AKOYA Biosci- gastrocnemius, and plantaris, which we refer to as ences) steps. Finally, the slides were counterstained with “mixed muscle” . The other three sets were separate DAPI, mounted with Prolong Gold Antifade Mounting quadriceps, tibialis, and soleus. We retained nuclei from solution (Invitrogen), and stored in a 4°C room. The all muscle samples in mixed muscle that contained 200– fluorescent images were obtained in the Johns Hopkins 2500 unique RNAs and had less than 5% mitochondrial Microscope Core Facility using a Zeiss LSM700 Laser genes. Log normalization was performed before finding scanning confocal microscope. Images were manually the top 2000 variable features and scaling through Seur- counted for co-expression, counter-expression, and non- at’s built in functions. These variable features were used expression across muscle fibers in ImageJ , and a χ to build principal components and the top 10 were used analysis, with Yates correction, was determined in Rstu- for clustering and UMAP visualization. We then subse- dio (v1.3.1093) and R (v4.0.3). lected myocyte nuclei using Ttn as a positive marker and Abca8a and Plxdc2 as negative markers of fibroadi- Human Protein Atlas pogenic progenitors. We painted the UMAP of the The HPA is a comprehensive repository of IHC-stained remaining nuclei using Myh1, Myh2, Myh4, and Myh7 tissue microarrays for numerous tissues, including skel- to identify muscle fiber types. Seurat’s Findallmarkers etal muscle [30, 31]. We cross-referenced our gene list used a Wilcoxon rank-sum test to identify differentially with the HPA to find examples of concordance and dis- expressed genes between clusters expressed in at least crepancy to our gene list for variable expression. 50% of each cluster being examined. We then used these genes and myosin heavy chains to assign identities to Gene Ontology (GO) validation slow, fast2A, fast 2B, and fast 2X clusters. GO was performed on the 557 most variable genes be- Finally, we ran differential gene expression analysis tween two fast 2X clusters (2X and 2X ) using the c1 c2 using a T-test between the Fast 2X and a combined Gene Ontology resource (http://geneontology.org/) and Slow/Fast2A group, to match the analysis of the prior selecting for the cellular component. Gene lists for terms scRNA-seq data. A logFC threshold of 0.35 filtered out “actin cytoskeleton,”“mitochondria,”“cell-cycle,” and highly abundant genes. “transcription regulator activity” were obtained from GO and used to determine the average expression of genes RISH in each category from the single cell RNA-seq and nuc- Wild-type C57Bl6 mouse skeletal muscles (extensor digi- seq datasets. The log2 normalized expression values of torum longus, gastrocnemius, soleus, diaphragm) and the datasets were normalized to each other. brain were obtained at necropsy under an approved ACUC protocol. Tissues were immediately fixed in for- Data availability malin and paraffin-embedded blocks were created, from Mouse skeletal muscle sequencing was deposited at the which 5-μm slides were made. Catalog probes for RNA Sequence Read Archive (SRA – SRP241908) and the in situ hybridization (RISH) were obtained from RNA- Gene Expression Omnibus (GSE143636). scope (ACDBio). These probes were designed to detect the following genes: Myh2 (pre-mRNA, #539031-C2), Code availability Got2 (#459111), Fhl1 (#536521), Ntrk3 (423621-C2), and All analysis scripts are available at GitHub (https:// Gabbr2 (#317971). Each probe set targeted all validated github.com/mhalushka/Skeletal_muscle_mosaicism). NCBI refseq transcript variants of the gene. One custom probe, Eno3, was designed to target all transcript vari- Results ants of Eno3 (GeneBank accession nm_007933.3). Validation of a large cell scRNA-seq method The Multiplex Fluorescent Reagent Kit v2 (ACDBio) We utilized a large particle FACS (LP-FACS) method test- was used following the manufacturer’s instructions. ing two different approaches to isolated and dissociated Briefly, FFPE tissue slides were baked for 1 h at 60°C. FDB myofibers. In the first approach, we dissected the The slides were subsequently deparaffinized with xylene, FDB from tendon to tendon prior to digestion, enabling rinsed with 100% ethanol and air-dried. After application isolation of fully intact myofibers. In the second approach, of hydrogen peroxide and washing, slides were treated we cut small portions of the FDB muscle using scissors. with the target retrieval reagent in a steamer (>99°C) for We reasoned that the latter approach would broadly 20 min. Then, the tissue was permeabilized using a pro- mimic skeletal muscle needle biopsies as might be done, tease. Hybridization of the probes to the targeted for example, from a human patient sample. We isolated mRNAs was performed by incubation in a 40°C oven for single myofibers through LP-FACS, using a flow channel 2 h. After washes, the slides were processed for the size of 500 μm. The COPAS SELECT Flow Pilot Platform standard signal amplification and application of was employed. Using time-of-flight (TOF, measuring axial Verma et al. Skeletal Muscle (2021) 11:13 Page 4 of 10 length) and optical extinction (EXT, measuring optical The two gates and pseudo-biopsy approach were used density) parameters, we found that skeletal myofibers sep- to isolate 763 cells for single cell RNA-seq using the arated into three populations—an EXT-low population, established mcSCRB-seq protocol [21, 22]. The entire EXT-high/TOF-low population, and EXT-high/TOF-high group of 763 cells/cell fragments were sequenced to a population (Fig. 1a). The EXT-high/TOF-high population median depth of 108,110 reads per cell. Preliminary ana- was comprised almost entirely of intact myofibers with lyses, however, indicated a distinct cluster of cells with a lengths > 400 μm, suggesting successful sorting of large high percentage of mitochondrial reads (Fig. 1f) or myofibers (Fig. 1b). Interestingly, the EXT-high/TOF-low otherwise low abundance reads (median 12,187 per cell). population was composed of what appeared to be rod- Notably, almost all of our pseudo-biopsy myofiber frag- shaped fragments that maintained sarcomeric proteins, al- ments and many TOF-low cells fell into this category. beit disrupted (Fig. 1c). The EXT-low population was These quality control metrics likely indicated poor qual- comprised mostly of debris and dead cells, as previously ity or sheared cells with loss of RNA. Thus, we excluded observed with cardiac myocytes (Fig. 1d). The EXT-high/ these cells, identified the EXT-high and TOF-high gate TOF-low population qualitatively resembled our pseudo- as the appropriate gate to obtain high quality myofibers, biopsy isolated myofiber fragments (Fig. 1e), which also and narrowed our analysis to the best 171 cells (>5000 shared similar TOF and EXT parameters (not shown). To genes expressed and <20% mitochondrial genes) our knowledge, this is the first FACS-based single cell remaining with a median read count of 239,572 per cell. RNA-seq study of skeletal myofibers; thus, we adopted a broad gating strategy for isolation of single cells. We Analysis of the expression patterns of single FDB sorted 700 EXT-high myofibers (comprised of both TOF- myofibers high and TOF-low populations) as well as 100 myofiber A median of 12,187 transcripts were identified in these fragments isolated through the pseudo-biopsy method. myofibers and all had the expression patterns of mature Fig. 1 FDB muscle myocyte preparation. a Flow cytometry showing three gated areas representing EXT-high/TOF-high, EXT-high/TOF-low and EXT-low populations of FDB myofibers. b Representative images of Gate 1 EXT-high/TOF-high. c Representative images of Gate 2 EXT-high/TOF- low. d Representative images of Gate 3 EXT-low. e Representative image of pseudo-biopsy isolated myocyte fragments. Gates 1 and 2 were used for library preparation. White size bar is 400 μm. f Percent of mitochondria in unused and used cells Verma et al. Skeletal Muscle (2021) 11:13 Page 5 of 10 skeletal myofibers, highly expressing a myosin heavy levels and coexpression of Myh1 and Myh2 suggesting chain isoform. higher gene plasticity and more cell hybrids (Fig. 2b) . We used the top 4 significant PCs to cluster these cell types (Fig. 2a). Three groups were observed in a UMAP Shared and variable transcripts by cell type dimensionality reduction plot. Two clusters, containing We wondered about the extent to which highly abun- 69 and 53 cells respectively (71% of all cells) had ele- dant genes were mosaic across these cell fiber states. vated expression of Myh1 and Myh8 identifying these By normalized read counts of the scRNA-seq data, we groups as fast 2X type cells. MYH8, while considered a determined the 50 most abundant transcripts by the neonatal myosin, maintains low expression in adult skel- average of each cell type in the two fast 2X clusters etal muscle [32, 33]. A third cluster containing 49 cells and the one fast 2A / slow cluster determined by was defined by high expression of Tnnt1 and Myh2. A Seurat (Suppl. Table 1). The overall most abundant deeper analysis of this group showed that 12 cells had transcripts were Ttn, Acta1,and mt-Rnr2.There was high to modestly elevated Myh7 expression (a slow- significant overlap of abundant genes, with only 9 twitch marker), indicating this cluster was a combination genes being different across the three samples. We of slow-fibers cells and fast 2A fibers (Fig. 2b, c). Of then explored differences specifically between the two note, Myh4, a myosin heavy chain associated with fiber most abundant fiber types, fast 2A and fast 2X. Of type 2B, was the dominant myosin in only a single cell 2649 evaluated genes (all expressed in ≥95% of cells that was assigned to this group (Fig. 2b, c) . As the of one cluster), 160 genes were differentially FDB is a fast-fiber muscle, the overall distribution of sig- expressed across the two groups (t-test, adj. p value < nificantly more fast (159) to slow fibers (12) is consistent 0.01). This included expected genes such as Tnni1, with expected. Tnnt1,and Myh1 and less investigated genes such as Interestingly, the expression patterns of the main fast-/ Ubash3b and Togaram2 (Fig. 2d, Suppl. Table 2). slow-fiber differentiating Myh genes was not as dichot- We validated a subset of these genes (Eno3, Fhl1, omous as noted in protein based fiber type data . Got2, Myh2) using RISH and available probes across the Here there were many more cells with intermediate extensor digitorum longus (EDL), gastrocnemius, and Fig. 2 Subtyping of skeletal myofibers. a UMAP graph of 171 skeletal muscle cells based on variable gene expression, indicating 3 clusters. b Major myosin heavy chain distributions across the 171 cells as a percentage of each heavy chain. c Assignment of each cell to a fiber type. d Heat map of major gene expression differences between fast 2A and fast 2X cells Verma et al. Skeletal Muscle (2021) 11:13 Page 6 of 10 soleus. Eno3 is a known fast fiber gene (both 2A and 2X) of tibialis, gastrocnemius, soleus, plantaris, and extensor and was identified in most cells of the fast-twitch EDL and digitorum longus (N=6 each), along with nuclei from gastrocnemius (Fig. 3a, b). Fhl1 was identified as being ele- each of quadriceps, tibialis, and soleus, identifying myo- vated in fast 2A myofibers (Fig. 2d, Suppl. Table 2). In nuclei based on Ttn expression. Whereas we focused on Fhl1-positive myofibers, Eno3 qualitative expression was sequencing depth (239,572 median reads/171 cells), Dos reduced, but not absent. In the slow-twitch soleus (Fig. Santos et al. went wide, obtaining many more skeletal 3c), levels of both genes were decreased. Of 693 myofibers myofiber nuclei (6962), but only to a median read count reviewed across all of the tissues, most (341, 49%) showed of 2785 and 1210 transcripts per nucleus in their mixed co-expression, with 186 cells being Eno3+ only, 92 being muscle sample. We processed this dataset using Seurat Fhl1+ only and 74 having no expression. A χ analysis and determined, as they reported, the presence of slow, demonstrated only a modest enrichment for co- fast 2A, fast2A/2X, fast 2B, and fast 2X nuclei clustering expression (χ =4.25, p =0.039). Fhl1 was then compared more distinctly by myosin heavy chain status on a to Myh2, a known fast 2A gene (Fig. 3d–f). The strong UMAP visualization of the data, than our whole scRNA- pre-mRNA Myh2 staining was interpreted as nuclear [15, seq data (Suppl. Fig. 1). 16]. The expression of the two genes demonstrated appro- As the whole cell versus nuclear isolation methods priate overlap in the same cells (206 co-expressed, 195 were so distinct, we evaluated how those differences non-expressed, and 30 counter-expressed, χ =320.9, p = affect the presence of abundant genes. Notably, in a 9.3e-72). Got2, also identified as elevated in fast 2A fibers, comparison of the most highly expressed genes, only 27 showed appropriate co-expression with Myh2 across all were present in the top 100 for both methods. A GO three tissues (Fig. 3g–i) and by myofiber (81 co-expressed, search of the 73 genes that were only abundant in the 69 non-expressed, 22 counter-expressed, χ =98.1, p= whole cell scRNA-seq showed these genes were enriched 4.0e−23). These patterns of fast fiber expression are con- for terms such as “myofibril” and “ATP metabolic sistent with those identified by scRNA-seq (Fig. 2d). process.” This had us wonder if we could observe differ- ences in gene classes based on the nuc-seq vs scRNA- Comparison of full cell scRNA-Seq to nuclear RNA-Seq seq methods similar to that described in other cell types A recent publication by Dos Santos et al.  described . We used normalized expression data between the nuc-seq of mouse skeletal muscles from a mixed sample studies and determined the expression differences Fig. 3 RISH staining of variably expressed genes across extensor digitorum longus (EDL), gastrocnemius (GC), and soleus (SOL). a–c Fhl1 (green) and Eno3 (red) show differing expression patterns across EDL (a), GC (b), and SOL (c). There is reduced Fhl1 in the fast-twitch GC and increased Fhl1 in the slow-twitch SOL fiber. d–f Fhl1 (green) is coexpressed with Myh2 (red), which has a perinuclear pattern. Both Fhl1 and Myh2 are reduced in GC (e) and increased in SOL (f). g–i Got2 (green) is coexpressed with Myh2 (red) showing highest expression in the EDL (g). j Neuronal tissue showing strong staining of Gabbr2 and Ntrk3. k Gabbr2 shows a variable blush across the GC, while no discernable Ntrk3 was observed in the GC. Nuclei were stained with DAPI (blue) in all images Verma et al. Skeletal Muscle (2021) 11:13 Page 7 of 10 between whole cell and nuclear data for the genes repre- Pecam1 and Smtn, as markers of endothelial cells and senting the GO terms of transcription factors, cell cycle smooth muscle cells respectively, showed comparable in- genes, mitochondria, and actin-cytoskeleton. Both tran- creases in these genes among the fast 2X cells. These c2 scription factors and cell cycle genes were more abun- data indicated that despite extremely low expression, am- dant in the nuc-seq data (3.9 and 1.9 fold respectively), bient genes, in general, have slightly elevated values in fast while actin-cytoskeleton genes were more abundant in 2X cells, perhaps consistent with more genes being de- c2 scRNA-seq data (1.9 fold). Mitochondrial genes were tected in these cells. We further note that three of the equivalent for expression across the two methods. How- most abundant genes, Ttn, mt-Rnr1,and mt-Rnr2 are con- ever, two of the most abundant scRNA-seq genes over- versely lower in fast 2X cells (Suppl Fig. 2). We take the c2 all, the mitochondrial genes mt-Rnr1 and mt-Rnr2, were summation of this data to indicate that the separation of not present in the nuc-seq data, impacting these results. fast 2X and 2X is a technical artifact related to the se- c1 c2 Also of note, Malat1, a known nuclear lncRNA, was the quencing and not a true biological distinction. most abundant transcript in the nuc-seq data, consistent with a prior report . Discussion We compared a combined slow/fast 2A group against Our study represents the first use of LP-FACS to isolate a fast 2X group to again discern differences in myofiber- single myofibers for scRNA-seq. This study was designed specific genes by sequencing method. There were 771 to prove feasibility of the method and did not attempt to differentially expressed genes (t-test, adj. p value <0.01) discern fiber type expression across a range of muscle in comparison to 152 in the whole cell dataset. Only 80 bundles or types, which will be the basis of future stud- differentially expressed genes overlapped between the ies. Skeletal myofibers are often long, stretching across two datasets and of these, three were significant in op- the length of a muscle; thus, isolation techniques (par- posite fold directions (Crim1, Myh4, Kcnc1) (Suppl. ticularly from human samples) may rely on the use of bi- Table 3). Altogether, these data indicate the ability to opsies or otherwise fragmented myofibers. To test the discern cell myofiber types, by either nuc-seq or scRNA- effect of myofiber fragmentation on scRNA-seq data seq, despite differences in the specific genes that dis- quality, we used a liberal gating strategy of our dissoci- criminate across the slow/fast 2A and fast 2X cells and ated myofibers (including both EXT-high/TOF-low and the relative expression levels of genes, by the two EXT-high/TOF-high populations) as well as directly se- methods of RNA-seq. quencing fragmented myofibers generated through a pseudo-biopsy approach. Disappointingly, we found that Is there a meaningful difference between fast 2X a large portion of our sequenced myofibers were of poor subclusters? quality, including those from our pseudo-biopsy ap- The initial Seurat analysis subsetted the fast 2X cluster proach. By contrast, the highest quality data came from into two groups. We explored if these two fast 2X clus- fully intact myofibers, in particular the EXT-high/TOF- ters (cluster 1 - 2X ; cluster 2 – 2X ) represent unique high population. Because this population is almost com- c1 c2 cell types, cell states, or some technical division. Of 5260 pletely enriched for intact myofibers, we believe that fu- genes compared, 557 genes were differentially expressed ture experiments using LP-FACS to isolate skeletal (t test; adj. p value <0.01). A GO analysis on the 557 myofibers should focus solely on the EXT-high/TOF- genes identified an enrichment of the cellular compo- high population. We are confident that this will allow nent “neuronal synapse,” suggesting variability at the for a much higher percentage of good quality scRNA- NMJ. A further review of the top significant genes seq libraries, akin to what we have observed previously showed that >20 genes appear to have neuronal origins with LP-FACS isolation of cardiac myocytes . These (Cdh4, Cdkl5, Cntn4, Dscam, Gabbr2, Kirrel3, Lingo2, results also mean that more work must be done to iden- Lrp1, L1cam, Nrcam, Ntn1, Ntrk3, Ptprt, Ptpro, Robo2, tify better isolation methods for human skeletal muscle. Sdk1, Sema5a, Sema6d, Shank2, Sox5, Tnr, and Wwox). Current methods of human skeletal muscle biopsying We attempted to exclude technical reasons for this vari- from the quadriceps only obtain muscle fragments. Al- ability before investigating a biological rationale for the though different collection reagents for these biopsies division. (high potassium and EGTA), which are used to prevent First, we noted that the vast majority of the neuronal contractions, are used, it remains to be determined if genes (20/22) were present in at least 120 of the 171 preventing contractions is sufficient to reduce RNA loss cells (Suppl Fig. 2).We surmised that some degree of as cellular integrity is always lost . If not, more cre- ambient RNA was present . We then performed ative means to obtain full length fibers or non-damaged RISH for two of these genes, Gabbr2 and Ntrk3, showing fibers must be considered, including rapid autopsy pro- robust neuronal staining (Fig. 3j) and some Gabbr2, but tocols or larger surgical resections that include skeletal no Ntrk3 in myofibers (Fig. 3k). We further noted muscles. Otherwise, human muscle data will have to be Verma et al. Skeletal Muscle (2021) 11:13 Page 8 of 10 obtained from nuclei material, which we noted had dif- PC: Principal components; PCA: Principal components analysis; UMAP: Uniform Manifold Approximation and Projection; GO: Gene Ontology; ferent, albeit complimentary, expression characteristics. LP-FACS: Large particle FACS; HPA: Human Protein Atlas The recent availability of public nuc-seq skeletal myo- fiber data allowed us to compare these two techniques. Supplementary Information As has been reported, we found significant differences in The online version contains supplementary material available at https://doi. the gene composition of these cells that was dependent org/10.1186/s13395-021-00269-2. on the sequencing approach [38, 39]. Myh gene expres- sion was more distinct in the nuc-seq data. This could Additional file 1 Supplementary Table 1. sctransformed mean expression values of the 50 most abundant genes by skeletal muscle be related to technical differences such as a sparser data fiber type. matrix of nuc-seq with more binary patterns. More Additional file 2 Supplementary Table 2. Comparison of fast2A and interestingly, it could be a biological finding if dynamic fast2X differential expression. transcriptional activity is more distinct in nuc-seq data. Additional file 3 Supplementary Table 3. Significantly variable genes The nuc-seq was enriched for transcription factors and between fast2A/slow genes and fast2X genes in both the whole cell scRNA-seq and nuc-seq datasets. Average log fold change and adjusted cell cycle RNAs, while the whole cell scRNA-seq was p. value is provided for both scRNA-seq and nuc-seq samples. biased towards other RNA types including significant ex- Additional file 4 Supplementary Figure 1. Myofiber nuclei analysis. a pression of the mitochondrial RNAs mt-Rnr1 and mt- A UMAP of all myofiber nuclei with fiber type noted. b Myosin heavy Rnr2 even after controlling for the percent of mitochon- chain expression used to assign the myofiber types. c Myosin heavy chain expression and number of nuclei by fiber type demonstrating drial RNA. Also, the genes that were variable between overlapping expression of Myh genes in some fiber types. the slow/fast 2A and fast 2X populations, across the Supplementary Figure 2. Comparison of the Fast 2X and Fast 2X c1 c2 methods, were frequently inconsistent. Nonetheless, subsets. a Twenty-two neuronal or NMJ-related genes are detected in most cells, but enriched in Fast 2X (middle left) cells. b Three neural c2 both methods successfully separated myofibers by fast/ genes (Cdh4, Kirrel3, Ntn1), an endothelial specific gene (Pecam1) and a slow type. It will take additional orthogonal approaches smooth muscle cell gene (Smtn) are all increased in Fast 2X subsets c2 such as proteomics to definitively solve this question of suggesting overall increase of ambient RNA in these cells. c Total gene counts are elevated in Fast2X subsets despite no increase in total reads c2 which genes/proteins have variable expression between or % mitochondria. d Typically abundant genes, Ttn, mt-Rnr1, and mt- myofiber types and at what expression levels. Rnr2 are all of lower expression in Fast 2X cells. c2 Characterization of the FDB identified essentially two clusters, a fast 2X cluster and a fast 2A/slow fiber clus- Acknowledgements ter. If more slow fibers were sequenced, that second The authors thank Efrain Ribeiro for his helpful comments on the project. group would have likely separated further. We were able to use RISH to validate some of the genes that had ex- Authors’ contributions R.X.V. performed all RNA-seq analysis. K.M.F. helped conceive the project and pression differences between the fast 2X and 2A cells. generated the proteomic data. S.K. and B.L.L. generated the skeletal muscle Although our analysis using Seurat subdivided the fast sequencing library. B.L.L. and C.L. obtained the mouse tissues. X.Y. performed 2X cluster, we believe the simplest explanation of this the RISH. K.F-T. performed the IHC. S.K., T.O.N, A.H.P., and M.N.M. performed the additional analysis. C.K. and D.A.K. oversaw the library preparation. A.Z.R. splitting is a technical cause related to slight differences helped develop the project. M.K.H. conceived the project, performed the in very low levels of ambient RNA. A more interesting analyses, and wrote the manuscript. All authors contributed toward revisions explanation is variable neuronal transfer of mRNAs of the manuscript. The authors read and approved the final manuscript. across the NMJ into the skeletal muscles via extracellu- Funding lar vesicles [40, 41]. This would imply a real state- M.K.H. was supported by grants 1R01HL137811, R01GM130564, and difference in these cells, but again is considered unlikely. P30CA006973 from the National Institutes of Health and 17GRNT33670405 We feel this exercise in considering technical causes of from the American Heart Association. T.O.N. was supported by grant R01GM130564. M.N.M. was supported by R01HL137811 and the University of Seurat-derived cell types is a useful reminder to groups Rochester CTSA award number UL1TR002001. A.Z.R was supported by working in the field of defining novel cell types to con- R01GM130564. C.K. and S.K. were supported by NIH R01HD086026, TEDCO sider more mundane reasons for some divisions. 2019-MSCRFD-5044, and the JHU Discovery Award. S.K. was supported by fel- lowship 20PRE35200028 from the American Heart Association. In conclusion, we introduce a method of whole skel- etal muscle cell isolation for scRNA-seq experimenta- Availability of data and materials tion. This FDB data is some of the first whole, single cell The dataset supporting the conclusions of this article is available in the skeletal myofiber data mainly identifying expression pat- Sequence Read Archive repository (SRP241908, https://www.ncbi.nlm.nih. gov/sra/?term=SRP241908) and the Gene Expression Omnibus (GSE143636, terns in fast 2A and fast 2X myofibers. Future studies https://www.ncbi.nlm.nih.gov/gds/?term=GSE143636). The analysis scripts can investigate a variety of muscle beds incorporating supporting the conclusions of this article are available at the GitHub more slow or fast 2B cells by this approach. repository, (https://github.com/mhalushka/Skeletal_muscle_mosaicism). Abbreviations Declarations scRNA-seq: Single cell RNA-seq; NMJ: Neuromuscular junction; FACS: Fluorescence-activated cell sorting; FDB: Flexor digitorum brevis; Ethics approval and consent to participate TOF: Time-of-flight; EXT: Optical extinction; UMI: Unique molecular identifier; Not applicable Verma et al. Skeletal Muscle (2021) 11:13 Page 9 of 10 Consent for publication 16. Dos Santos M, Backer S, Saintpierre B, Izac B, Andrieu M, Letourneur F, et al. Not applicable Single-nucleus RNA-seq and FISH identify coordinated transcriptional activity in mammalian myofibers. Nat Commun. 2020;11(1):5102. https://doi. org/10.1038/s41467-020-18789-8. Competing interests 17. Petrany MJ, Swoboda CO, Sun C, Chetal K, Chen X, Weirauch MT, et al. The authors have no competing interests to declare. Single-nucleus RNA-seq identifies transcriptional heterogeneity in multinucleated skeletal myofibers. Nat Commun. 2020;11(1):6374. https:// Author details 1 doi.org/10.1038/s41467-020-20063-w. Department of Pathology, Johns Hopkins University School of Medicine, 18. Kannan S, Miyamoto M, Lin BL, Zhu R, Murphy S, Kass DA, et al. Large Ross Bldg. Rm 632B, 720 Rutland Avenue, Baltimore, MD 21205, USA. particle fluorescence-activated cell sorting enables high-quality single-cell Division of Cardiology, Department of Medicine, Johns Hopkins University RNA sequencing and functional analysis of adult cardiomyocytes. Circ Res. School of Medicine, Baltimore, MD, USA. Department of Biostatistics and 2019;125(5):567–9. https://doi.org/10.1161/CIRCRESAHA.119.315493. Computational Biology, University of Rochester Medical Center, Rochester, 19. Tarpey MD, Amorese AJ, Balestrieri NP, Ryan TE, Schmidt CA, McClung JM, NY, USA. et al. Characterization and utilization of the flexor digitorum brevis for assessing skeletal muscle function. Skeletal Muscle. 2018;8(1):14. https://doi. Received: 24 December 2020 Accepted: 28 April 2021 org/10.1186/s13395-018-0160-3. 20. Shefer G, Yablonka-Reuveni Z. Isolation and culture of skeletal muscle myofibers as a means to analyze satellite cells. Methods Mol Biol. 2005;290: References 281–304. https://doi.org/10.1385/1-59259-838-2:281. 1. Okumura N, Hashida-Okumura A, Kita K, Matsubae M, Matsubara T, Takao T, 21. Soumillon M, Cacchiarelli D, Semrau S, van Oudenaarden A, Mikkelsen TS. et al. Proteomic analysis of slow- and fast-twitch skeletal muscles. Characterization of directed differentiation by high-throughput single-cell Proteomics. 2005;5(11):2896–906. https://doi.org/10.1002/pmic.200401181. RNA-Seq. BioRxiv. 2014. https://doi.org/10.1101/003236. 2. Schiaffino S, Reggiani C. Fiber types in mammalian skeletal muscles. Physiol 22. Ziegenhain C, Vieth B, Parekh S, Reinius B, Guillaumet-Adkins A, Smets M, Rev. 2011;91(4):1447–531. https://doi.org/10.1152/physrev.00031.2010. et al. Comparative Analysis of Single-Cell RNA Sequencing Methods. Mol 3. Kho AT, Kang PB, Kohane IS, Kunkel LM. Transcriptome-scale similarities Cell. 2017;65(4):631–43 e634. https://doi.org/10.1016/j.molcel.2017.01.023. between mouse and human skeletal muscles with normal and myopathic 23. Parekh S, Ziegenhain C, Vieth B, Enard W, Hellmann I. zUMIs - A fast and phenotypes. BMC Musculoskelet Disord. 2006;7(1):23. https://doi.org/10.11 flexible pipeline to process RNA sequencing data with UMIs. Gigascience. 86/1471-2474-7-23. 2018:7(6). https://doi.org/10.1093/gigascience/giy059. 4. Gonzalez-Freire M, Semba RD, Ubaida-Mohien C, Fabbri E, Scalzo P, Hojlund 24. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: K, et al. The Human Skeletal Muscle Proteome Project: a reappraisal of the ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29(1):15–21. https:// current literature. J Cachexia Sarcopenia Muscle. 2017;8(1):5–18. https://doi. doi.org/10.1093/bioinformatics/bts635. org/10.1002/jcsm.12121. 25. Liao Y, Smyth GK, Shi W. The Subread aligner: fast, accurate and scalable 5. Chemello F, Bean C, Cancellara P, Laveder P, Reggiani C, Lanfranchi G. read mapping by seed-and-vote. Nucleic Acids Res. 2013;41(10):e108. Microgenomic analysis in skeletal muscle: expression signatures of https://doi.org/10.1093/nar/gkt214. individual fast and slow myofibers. PLoS One. 2011;6(2):e16807. https://doi. 26. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell org/10.1371/journal.pone.0016807. transcriptomic data across different conditions, technologies, and species. 6. Giordani L, He GJ, Negroni E, Sakai H, Law JYC, Siu MM, et al. High- Nat Biotechnol. 2018;36(5):411–20. https://doi.org/10.1038/nbt.4096. dimensional single-cell cartography reveals novel skeletal muscle-resident 27. Hafemeister C, Satija R. Normalization and variance stabilization of single- cell populations. Mol Cell. 2019;74(3):609–21 e606. https://doi.org/10.1016/j. cell RNA-seq data using regularized negative binomial regression. Genome molcel.2019.02.026. Biol. 2019;20(1)296. https://doi.org/10.1186/s13059-019-1874-1. 7. Porpiglia E, Samusik N, Ho ATV, Cosgrove BD, Mai T, Davis KL, et al. High- 28. Dos Santos M, Backer S, Saintpierre B. F. R, Sotiropoulos A, Marie P: Single- resolution myogenic lineage mapping by single-cell mass cytometry. Nat nucleus RNA-seq and FISH reveal coordinated transcriptional activity in Cell Biol. 2017;19(5):558–67. https://doi.org/10.1038/ncb3507. mammalian myofibers. https://wwwbiorxivorg/content/101101/ 8. Dell'Orso S, Juan AH, Ko KD, Naz F, Perovanovic J, Gutierrez-Cruz G, et al. 20200416043620v1 2020. Single cell analysis of adult mouse skeletal muscle stem cells in homeostatic 29. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of and regenerative conditions. Development. 2019:146(12). https://doi.org/1 image analysis. Nat Methods. 2012;9(7):671–5. https://doi.org/10.1038/ 0.1242/dev.174177. nmeth.2089. 9. Cho DS, Doles JD. Single cell transcriptome analysis of muscle satellite cells 30. Uhlen M, Fagerberg L, Hallstrom BM, Lindskog C, Oksvold P, Mardinoglu A, reveals widespread transcriptional heterogeneity. Gene. 2017;636:54–63. et al. Proteomics. Tissue-based map of the human proteome. Science. 2015; https://doi.org/10.1016/j.gene.2017.09.014. 347(6220):1260419. https://doi.org/10.1126/science.1260419. 10. Cornelison DD, Wold BJ. Single-cell analysis of regulatory gene expression in 31. Uhlen M, Oksvold P, Fagerberg L, Lundberg E, Jonasson K, Forsberg M, et al. quiescent and activated mouse skeletal muscle satellite cells. Dev Biol. 1997; Towards a knowledge-based Human Protein Atlas. Nat Biotechnol. 2010; 191(2):270–83. https://doi.org/10.1006/dbio.1997.8721. 28(12):1248–50. https://doi.org/10.1038/nbt1210-1248. 11. Cacchiarelli D, Qiu X, Srivatsan S, Manfredi A, Ziller M, Overbey E, et al. 32. Consortium GT, Laboratory DA. Coordinating Center -Analysis Working G, Aligning single-cell developmental and reprogramming trajectories Statistical Methods groups-Analysis Working G, Enhancing Gg, Fund NIHC, identifies molecular determinants of myogenic reprogramming outcome. Nih/Nci, Nih/Nhgri, Nih/Nimh, Nih/Nida et al: Genetic effects on gene Cell Syst. 2018;7(3):258–68 e253. https://doi.org/10.1016/j.cels.2018.07.006. expression across human tissues. Nature. 2017;550(7675):204–13. https://doi. 12. Trapnell C, Cacchiarelli D, Grimsby J, Pokharel P, Li S, Morse M, et al. The org/10.1038/nature24277. dynamics and regulators of cell fate decisions are revealed by 33. Schiaffino S, Rossi AC, Smerdu V, Leinwand LA, Reggiani C. Developmental pseudotemporal ordering of single cells. Nat Biotech. 2014;32(4):381–6. myosins: expression patterns and functional significance. Skeletal muscle. https://doi.org/10.1038/nbt.2859. 2015;5(1):22. https://doi.org/10.1186/s13395-015-0046-6. 13. Rubenstein AB, Smith GR, Raue U, Begue G, Minchev K, Ruf-Zamojski F, et al. 34. Murgia M, Nagaraj N, Deshmukh AS, Zeiler M, Cancellara P, Moretti I, et al. Single Single-cell transcriptional profiles in human skeletal muscle. Sci Rep. 2020; muscle fiber proteomics reveals unexpected mitochondrial specialization. EMBO 10(1):229. https://doi.org/10.1038/s41598-019-57110-6. Rep. 2015;16(3):387–95. https://doi.org/10.15252/embr.201439757. 14. Blackburn DM, Lazure F, Corchado AH, Perkins TJ, Najafabadi HS, Soleimani VD. High-resolution genome-wide expression analysis of single myofibers 35. Bakken TE, Hodge RD, Miller JA, Yao Z, Nguyen TN, Aevermann B, et al. using SMART-Seq. J Biol Chem. 2019;294(52):20097–108. https://doi.org/10.1 Single-nucleus and single-cell transcriptomes compared in matched cortical 074/jbc.RA119.011506. cell types. PLoS One. 2018;13(12):e0209648. https://doi.org/10.1371/journal. 15. Kim M, Franke V, Brandt B, Lowenstein ED, Schowel V, Spuler S, et al. Single- pone.0209648. nucleus transcriptomics reveals functional compartmentalization in syncytial 36. Young MD, Behjati S. SoupX removes ambient RNA contamination from skeletal muscle cells. Nat Commun. 2020;11(1):6375. https://doi.org/10.1038/ droplet based single cell RNA sequencing data. Gigascience. 2020;9(12): s41467-020-20064-9. giaa151. https://doi.org/10.1093/gigascience/giaa151. Verma et al. Skeletal Muscle (2021) 11:13 Page 10 of 10 37. Pietrangelo T, Perni S, Di Tano G, Fano-Illic G, Franzini-Armstrong C. A method for the ultrastructural preservation of tiny percutaneous needle biopsy material from skeletal muscle. Int J Mol Med. 2013;32(4):965–70. https://doi.org/10.3892/ijmm.2013.1454. 38. Wu H, Kirita Y, Donnelly EL, Humphreys BD. Advantages of Single-Nucleus over Single-Cell RNA Sequencing of Adult Kidney: Rare Cell Types and Novel Cell States Revealed in Fibrosis. JASN. 2019;30(1):23–32. https://doi. org/10.1681/ASN.2018090912. 39. Selewa A, Dohn R, Eckart H, Lozano S, Xie B, Gauchat E, et al. Systematic comparison of high-throughput single-cell and single-nucleus transcriptomes during cardiomyocyte differentiation. Sci Rep. 2020;10(1): 1535. https://doi.org/10.1038/s41598-020-58327-6. 40. Ashley J, Cordy B, Lucia D, Fradkin LG, Budnik V, Thomson T. Retrovirus-like Gag protein Arc1 binds RNA and traffics across synaptic boutons. Cell. 2018; 172(1-2):262–74 e211. https://doi.org/10.1016/j.cell.2017.12.022. 41. Korkut C, Ataman B, Ramachandran P, Ashley J, Barria R, Gherbesi N, et al. Trans-synaptic transmission of vesicular Wnt signals through Evi/Wntless. Cell. 2009;139(2):393–404. https://doi.org/10.1016/j.cell.2009.07.051. Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Skeletal Muscle – Springer Journals
Published: May 17, 2021