Reduced Myogenic and Increased Adipogenic Differentiation Capacity of Rotator Cuff Muscle Stem Cells

Background:

Fat accumulation commonly occurs in chronically torn rotator cuff muscles, and increased fat within the rotator cuff is correlated with poor clinical outcomes. The extent of lipid deposition is particularly pronounced in injured rotator cuff muscles compared with other commonly injured muscles such as the gastrocnemius. Satellite cells, which are a tissue-resident muscle stem-cell population, can differentiate into fat cells. We hypothesized that satellite cells from the rotator cuff have greater intrinsic adipogenic differentiation potential than do gastrocnemius satellite cells, and this difference is due to variations in epigenetic imprinting between the cells.

Methods:

Satellite cells from gastrocnemius and rotator cuff muscles of mice were cultured in adipogenic media, and the capacity to differentiate into mature muscle cells and adipogenic cells was assessed (n ≥ 9 plates per muscle group). We also performed DNA methylation analysis of gastrocnemius and rotator cuff satellite cells to determine whether epigenetic differences were present between the 2 groups (n = 5 mice per group).

Results:

Compared with the gastrocnemius, satellite cells from the rotator cuff had a 23% reduction in myogenic differentiation and an 87% decrease in the expression of the differentiated muscle cell marker MRF4 (myogenic regulatory factor 4). With respect to adipogenesis, rotator cuff satellite cells had a 4.3-fold increase in adipogenesis, a 12-fold increase in the adipogenic transcription factor PPARγ (peroxisome proliferator-activated receptor gamma), and a 65-fold increase in the adipogenic marker FABP4 (fatty-acid binding protein 4). Epigenetic analysis identified 355 differentially methylated regions of DNA between rotator cuff and gastrocnemius satellite cells, and pathway enrichment analysis suggested that these regions were involved with lipid metabolism and adipogenesis.

Conclusions:

Satellite cells from rotator cuff muscles have reduced myogenic and increased adipogenic differentiation potential compared with gastrocnemius muscles. There appears to be a cellular and genetic basis behind the generally poor rates of rotator cuff muscle healing.

Clinical Relevance:

The reduced myogenic and increased adipogenic capacity of rotator cuff satellite cells is consistent with the increased fat content and poor muscle healing rates often observed for chronically torn rotator cuff muscles. For patients undergoing rotator cuff repair, transplantation of autologous satellite cells from other muscles less prone to fatty infiltration may improve clinical outcomes.

Rotator cuff tears are a common upper-extremity disorder, with >250,000 surgical repairs performed annually in the U.S.¹. Achieving positive clinical outcomes following repair can be limited by fatty infiltration or myosteatosis, which is the combined atrophy, fibrosis, and fat accumulation within and around myofibers^2-4. The relative amount of fat in torn rotator cuff muscles is often greater than in other injured muscle groups^5-7, and many patients develop further myosteatosis even after undergoing a successful rotator cuff repair². Fat accumulation is also correlated with negative clinical outcomes^2,8, and identifying the cellular and molecular mechanisms that induce adipogenesis in the rotator cuff could provide new opportunities for improving muscle healing and recovery.

Satellite cells are a heterogenous stem-cell population largely responsible for postnatal skeletal muscle growth, regeneration, and repair, and they are stimulated by trauma to proliferate, differentiate, and fuse into damaged myofibers^9,10. Quiescent satellite cells are found between the muscle fibers and basal lamina and express the transcription factor Pax7 (paired box 7)^10,11. In addition to myofibers, satellite cells can also differentiate into adipocytes^12-14, and in a tenotomy and denervation muscle-injury model, greater fat accumulation and reduced healing were present in rotator cuff muscles compared with gastrocnemius muscles⁷. Because satellite cells are important in muscle regeneration and can enter the adipogenic lineage, we sought to determine differences in the myogenic and adipogenic differentiation capacity of satellite cells from gastrocnemius and rotator cuff muscles. We tested the hypothesis that, compared with gastrocnemius satellite cells, rotator cuff satellite cells have decreased myogenic and increased adipogenic differentiation capacity. To further explore mechanistic differences, as satellite cell activity can be regulated by epigenetic factors¹¹, we analyzed differential DNA methylation between gastrocnemius and rotator cuff satellite cells.

Materials and Methods

Animals

This study was approved by the University of Michigan Institutional Animal Care and Use Committee. We crossed 2 lines of genetically modified mice, obtained from The Jackson Laboratory, to generate experimental animals. First, we obtained Pax7^CreERT2 mice, which contain an IRES (internal ribosome entry site)-CreERT2 cassette between the stop codon and 3′ untranslated region of Pax7, resulting in the expression of a tamoxifen-responsive CreERT2 recombinase enzyme when Pax7 is also expressed (strain 017763)¹⁵. In the second line of mice, R26R^{flox-stop-tdTomato}, the constitutively expressed Rosa26 (R26R) locus was modified to contain a stop codon cassette flanked by loxP sites upstream of the red fluorescent tdTomato gene (strain 007909)¹⁶. In the absence of active Cre recombinase, R26R^{flox-stop-tdTomato} mice do not express tdTomato. However, upon treatment with tamoxifen, which activates the Cre enzyme, a recombination event occurs between loxP sites to remove the stop codons, resulting in the permanent expression of tdTomato, R26R^tdTomato. We crossed Pax7^CreERT2 and R26R^{flox-stop-tdTomato} mice to generate Pax7^CreERT2:R26R^{flox-stop-tdTomato} mice, which were backcrossed for several generations and maintained in the homozygous state. Quiescent satellite cells express Pax7¹⁵, and upon treatment with tamoxifen, the CreERT2 enzyme complex is activated, causing a recombination event at the R26R locus, resulting in the persistent expression of tdTomato in all cells expressing Pax7, as well as their daughter cells¹⁷, which are referred to as Pax7^CreERT2:R26R^tdTomato mice. An overview is presented in Figure 1-A. With the exception of wild-type C57Bl/6 mice that were used to determine baseline flow cytometry fluorescence, all experiments utilized 4-month-old male Pax7^CreERT2:R26R^tdTomato mice.

Fig. 1.

Overview of genetically modified mice, satellite cell labeling, and sorting. Fig. 1-A Pax7^CreERT2:R26^{flox-stop-tdTomato} mice do not normally express the red fluorescent protein tdTomato. When CreERT2 in Pax7⁺ satellite cells is activated in response to tamoxifen treatment, the stop codon cassette flanked by loxP sites is excised from the constitutively active Rosa26 (R26R) locus, causing tdTomato to be expressed in satellite cells, as well as in all subsequent daughter cells (Pax7^CreERT2:R26^tdTomato mice). Fig. 1-B Representative histology of muscles from tamoxifen-treated Pax7^CreERT2:R26^tdTomato mice, demonstrating the presence of tdTomato in satellite cells (red arrowheads). White = extracellular matrix (ECM) as marked by WGA-lectin, blue = nuclei, and red = Pax7-tdTomato. Fig. 1-C Representative flow sorting of tdTomato⁺ cells, plotted as forward scatter area (FSC-A) versus tdTomato fluorescent intensity, with the dashed area indicating the cells used for in vitro cell culture and DNA methylation sequencing studies. The inset is of an aliquot of the sorted cells demonstrating the visualization of viable tdTomato⁺ cells, visualized by overlaying the red fluorescent image onto the phase contrast image.

The labeling of satellite cells with tdTomato occurred by treating mice with an intraperitoneal injection of tamoxifen (2 mg) dissolved in corn oil (Sigma) daily for 5 days prior to muscle harvest. On the sixth day, mice were anesthetized with ketamine and xylazine. The supraspinatus and infraspinatus muscles of the rotator cuff and the gastrocnemius muscles were removed and processed for flow cytometry. The plantaris muscle was also removed as a sentinel to verify recombination and labeling of satellite cells. Following muscle removal, mice were killed by cervical dislocation and pneumothorax.

Muscle Histology

Plantaris muscles were frozen in Tissue-Tek (Sakura) with cold isopentane, and 10-µm sections were incubated with wheat germ agglutinin (WGA) lectin conjugated to Alexa Fluor 488 dye (ThermoFisher Scientific) to identify extracellular matrix. DAPI (4′,6-diamidino-2-phenylindole; Sigma)-labeled nuclei and tdTomato-identified satellite cells. Sections were visualized with a BX51 microscope (Olympus).

Satellite Cell Isolation and Flow Cytometry

Satellite cells were isolated from gastrocnemius and rotator cuff muscles, using a method modified from a previous report¹⁸. A detailed description is provided in the Appendix. Briefly, muscles were minced and digested to generate a suspension of cells enriched with satellite cells. Cells were then sorted via flow cytometry on the basis of forward scatter area (FSC-A) as a means to measure the cell size and tdTomato fluorescence. TdTomato⁺ cells were collected and used for in vitro differentiation experiments or DNA sequencing.

Cell Culture

A detailed description of cell culture is provided in the Appendix. Sorted cells were plated on culture dishes coated with growth factor-reduced Matrigel (Corning). Cells were expanded in growth media containing Dulbecco modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic (AbAm) (ThermoFisher Scientific) for 2 passages, and on reaching 70% confluence, were switched to adipogenic induction media for a period of 7 days prior to immunocytochemistry and gene-expression analysis^14,19.

Immunocytochemistry

Cells were fixed with 4% paraformaldehyde, permeabilized in 0.5% Triton X-100, and blocked in 5% goat serum. Cells were labeled with antibodies against myogenin (F5D, 1:100; Developmental Studies Hybridoma Bank), which is a transcription factor that identifies differentiated muscle cells^20,21, or with antibodies against fatty-acid binding protein 4 (FABP4, 1:500; AbCam), which is specifically expressed in adipogenic cells²². Secondary antibodies conjugated to Alexa Fluor 647 dye (ThermoFisher Scientific) detected primary antibodies. DAPI identified nuclei. Three random 10× fields per plate were quantified in an EVOS FL Imaging System (ThermoFisher Scientific). For myogenic differentiation, a cell was considered a differentiated muscle cell if it was a multinuclear myotube or contained a nucleus that was myogenin⁺, and myogenic cells were calculated as a percentage of total cells. For adipogenic quantification, the number of FABP4⁺ cells per field was calculated as a percentage of total cells.

Gene Expression

Gene-expression analysis was performed as previously described²³. RNA was isolated from cells using an miRNEasy kit (QIAGEN), reversed transcribed with iScript reagents (Bio-Rad), and amplified in a CFX96 real-time thermal cycler (Bio-Rad). Quantitative polymerase chain reaction (PCR) was performed using iTaq SYBR Green Supermix (Bio-Rad). A list of primers is shown in Appendix Table E-1. Target gene expression was normalized to the stable housekeeping gene β2-microglobulin using the 2^−ΔCt method.

DNA Methylation Analysis

DNA was isolated from freshly sorted tdTomato⁺ satellite cells, and DNA methylation analysis was performed by the University of Michigan Epigenomics Core using enhanced reduced representation bisulfite sequencing (ERRBS), as previously described^24,25. A detailed description is provided in the Appendix. Briefly, genomic DNA was digested with the methylation-insensitive restriction enzyme MspI, followed by end-repair, A-tailing, and ligation of methylated adapters. Bisulfite conversion of methylated sequences was performed prior to PCR amplification and subsequent sequencing. Gene enrichment and pathway analysis was performed with iPathwayGuide (Advaita Bioinformatics)²⁶. A full list of differentially methylated regions is provided in Appendix Table E-2 and differentially methylated cytosines, in Appendix Table E-3.

Statistical Methods

Data are presented as the mean and standard deviation. Sample sizes were selected on the basis of satellite-cell myogenic differentiation rates²¹. In immunocytochemistry and gene-expression experiments, differences between gastrocnemius and rotator cuff groups were assessed using a t test (α = 0.05), with Welch correction, in Prism 7.0 (GraphPad). MethylSig R package (The Sartor Lab, University of Michigan)²⁷ was used for differential DNA methylation analysis using a beta-binomial approach, with p values adjusted for multiple testing using a false discovery rate (FDR) adjustment to control for type-I errors. Sites were considered differentially methylated if they had a percent change in methylation of at least 20% and an FDR-adjusted p value of <0.05.

Results

To efficiently isolate satellite cells from gastrocnemius and rotator cuff muscles, and to verify that adipogenic markers were observed in myogenic lineage cells, we used a genetic approach to label the satellite cells of Pax7^CreERT2:R26R^tdTomato mice with tdTomato (Fig. 1-B). Across several runs, approximately 12% of cells isolated from both muscle groups were tdTomato⁺ (Fig. 1-C). We then evaluated the potential of satellite cells from the gastrocnemius and rotator cuff muscles to differentiate into myogenic and adipogenic cells (n ≥ 9 plates analyzed per group). There was a 23% reduction in the number of differentiated rotator cuff muscle cells per field compared with gastrocnemius cells, and a 4.3-fold increase in the number of adipogenic rotator cuff cells per field (Figs. 2-A through 2-D). Next, we measured the expression of myogenesis and adipogenesis-related genes (n ≥ 6 plates analyzed per group). While we did not see a difference between the rotator cuff and gastrocnemius groups in the early myogenic marker desmin or the muscle-fusion gene myomaker, we did see an 87% reduction in the expression of the late muscle differentiation marker MRF4 (myogenic regulatory factor 4) in rotator cuff cells (Fig. 3-A). For genes related to adipogenesis, we observed a 12-fold increase in the adipogenic transcription factor PPARγ (peroxisome proliferator-activated receptor gamma) and a 65-fold increase in FABP4 in rotator cuff cells, although no differences were observed between the rotator cuff and gastrocnemius muscle groups with respect to the adipocyte signaling molecule adiponectin or adipogenic transcription factor C/EBPα (CCAAT-enhancer-binding protein-alpha) (Fig. 3-B).

Fig. 2.

Immunocytochemistry results. Fig. 2-A The percentage of differentiated muscle cells as a fraction of total cells per field for gastrocnemius (GTN) and rotator cuff (RC) groups. Fig. 2-B Representative image of myogenin-expressing cells. Teal = myogenin, blue = nuclei, and red = Pax7-tdTomato. The teal arrowheads illustrate myogenin⁺ nuclei, and the yellow arrowheads are examples of myogenin⁻ nuclei in mononuclear cells. Fig. 2-C The percentage of adipogenic FABP4⁺ cells as a fraction of total cells per field for gastrocnemius (GTN) and rotator cuff (RC) groups. Fig. 2-D Representative image of FABP4-expressing cells. Teal = FABP4, blue = nuclei, and red = Pax7-tdTomato. Teal arrowheads are examples of FABP4⁺ cells. N ≥ 9 plates analyzed per group. Values are mean and standard deviation.

Fig. 3.

Gene expression. Quantification of the expression of myogenic genes (Fig. 3-A) and adipogenic genes (Fig. 3-B). Each target gene normalized to the stable housekeeping gene β2-microglobulin. Values are presented as the mean and standard deviation; n ≥ 6 plates analyzed per group. MRF4 = myogenic regulatory factor 4, PPARγ = peroxisome proliferator-activated receptor gamma, C/EBPα = CCAAT-enhancer-binding protein-alpha, and FABP4 = fatty-acid binding protein 4.

Because there was an increased capacity for adipogenic differentiation in rotator cuff cells, we sought to determine whether epigenetic differences existed between sorted gastrocnemius and rotator cuff satellite cells (Figs. 4-A and 4-B; n = 5 mice per group). We identified 180 hypomethylated regions and 175 hypermethylated regions in satellite cells from the rotator cuff compared with cells from the gastrocnemius (Fig. 4-C). The top 50 hypermethylated regions and hypomethylated regions are shown in Tables I and II. Finally, to identify biological processes and biochemical pathways that might be impacted by the difference in DNA methylation, we performed gene ontology analysis. For biological processes, the top 15 pathways identified were related to embryonic development and limb morphogenesis (Fig. 4-D). With regard to molecular function, the top pathways were related to transcription-factor activity and lipid metabolism, which is consistent with the in vitro findings related to adipogenesis.

Fig. 4.

DNA methylation analysis. Figs. 4-A and 4-B Overview of DNA methylation (Me) and the enhanced reduced representation bisulfite sequencing process. Fig. 4-A When cytosine is methylated in regulatory regions, gene transcription is blocked. Fig. 4-B Overview of DNA digestion, bisulfite conversion, polymerase chain reaction (PCR) amplification, and sequencing to determine methylated cytosine residues. Fig. 4-C Volcano plot demonstrating differences between the gastrocnemius (GTN) and rotator cuff (RC) groups in differentially methylated regions (DMRs). Figs. 4-D and 4-E Gene ontology analysis of relevant biological processes and molecular function predictions of DMR data. N = 5 mice per group. HMG = high mobility group.

TABLE I.

Top 50 Differentially Hypermethylated Regions of DNA in Satellite Cells from the Rotator Cuff Compared with the Gastrocnemius

ID	Symbol	Gene Name	Methylation Difference (%)	Distance (bp)	P Value
NM_001113412	Fggy	FGGY carbohydrate kinase domain containing	99.03	177,851	0.035
NR_046076	Gm4251	Predicted gene 4251	82.28	33,232	0.038
NM_025642	Mis18a	MIS18 kinetochore protein A	70.18	45,772	0.004
NM_025642	Mis18a	MIS18 kinetochore protein A	67.24	45,822	0.021
NM_029441	Cdyl2	Chromodomain protein, Y chromosome-like 2	60.33	183,549	0.007
NM_013813	Epb41l3	Erythrocyte membrane protein band 4.1 like 3	56.46	196,484	0.029
NM_198610	Igsf21	Immunoglobulin superfamily, member 21	55.18	0	0.001
NM_001286033	Stx2	Syntaxin 2	55.06	280,368	<0.001
NM_001286033	Stx2	Syntaxin 2	54.42	280,443	0.008
NM_015803	Atp8a2	ATPase, aminophospholipid transporter-like, class I, type 8A, member 2	54.25	48,132	0.040
NM_001033228	Itga1	Integrin alpha 1	53.62	245,091	0.040
NM_015803	Atp8a2	ATPase, aminophospholipid transporter-like, class I, type 8A, member 2	52.84	48,082	0.018
NM_001033228	Itga1	Integrin alpha 1	52.23	245,041	0.001
NM_029441	Cdyl2	Chromodomain protein, Y chromosome-like 2	51.93	183,449	0.005
NR_024085	BC006965	cDNA sequence BC006965	51.80	1,001,517	0.003
NM_029441	Cdyl2	Chromodomain protein, Y chromosome-like 2	50.96	183,424	0.038
NR_030709	Gm16386	Predicted gene 16386	50.32	41,229	0.009
NR_015496	1700031M16Rik	RIKEN cDNA 1700031M16 gene	48.16	9,442	0.021
NM_152895	Kdm5b	Lysine (K)-specific demethylase 5B	47.71	0	0.004
NM_152895	Kdm5b	Lysine (K)-specific demethylase 5B	46.90	0	0.012
NM_198610	Igsf21	Immunoglobulin superfamily, member 21	46.79	0	0.004
NM_009723	Atp2b2	ATPase, Ca++ transporting, plasma membrane 2	46.16	0	0.015
NR_015496	1700031M16Rik	RIKEN cDNA 1700031M16 gene	46.08	9,267	0.002
NR_015496	1700031M16Rik	RIKEN cDNA 1700031M16 gene	46.07	9,292	0.015
NM_010462	Hoxc10	Homeobox C10	45.99	717	<0.001
NM_010462	Hoxc10	Homeobox C10	45.53	692	<0.001
NR_131145	Gm29683	Predicted gene 29683	43.80	351,254	<0.001
NM_177544	Ang4	Angiogenin, ribonuclease A family, member 4	43.46	20,976	<0.001
NM_180662	Trappc9	Trafficking protein particle complex 9	43.22	0	0.029
NR_102286	4933432K03Rik	RIKEN cDNA 4933432K03 gene	43.10	37,697	0.007
NM_009834	Noct	Nocturnin	42.86	19,402	0.005
NM_172462	Zfp11	Zinc finger protein 11	42.51	187,078	0.015
NR_131145	Gm29683	Predicted gene 29683	42.34	351,279	0.011
NR_102286	4933432K03Rik	RIKEN cDNA 4933432K03 gene	42.18	37,647	0.009
NM_177544	Ang4	Angiogenin, ribonuclease A family, member 4	42.14	20,901	0.001
NR_045702	AW549542	Expressed sequence AW549542	42.10	21,482	0.049
NM_009834	Noct	Nocturnin	42.05	19,377	0.026
NM_001310738	Siglech	Sialic acid binding Ig-like lectin H	41.27	26,345	0.001
NR_035497	Mir1970	MicroRNA 1970	41.20	0	0.016
NM_180662	Trappc9	Trafficking protein particle complex 9	40.80	0	<0.001
NM_001029872	Itgad	Integrin, alpha D	40.64	22,757	0.001
NM_009834	Noct	Nocturnin	40.48	19,327	0.042
NM_001042617	Cadps	Ca2+-dependent secretion activator	40.33	0	0.003
NM_009309	T	Brachyury, T-box transcription factor T	40.30	0	0.008
NM_027188	Smyd3	SET and MYND domain containing 3	39.79	0	0.010
NM_009309	T	Brachyury, T-box transcription factor T	39.59	0	0.033
NM_015764	Greb1	Gene regulated by estrogen in breast cancer protein	39.20	0	0.005
NR_038085	Six3os1	SIX homeobox 3, opposite strand 1	38.87	66,806	0.013
NR_045342	4933411E08Rik	RIKEN cDNA 4933411E08 gene	38.21	0	0.020
NM_013723	Podxl	Podocalyxin-like	38.01	246,814	0.003

TABLE II.

Top 50 Differentially Hypomethylated Regions of DNA in Satellite Cells from the Rotator Cuff Compared with the Gastrocnemius

ID	Symbol	Gene Name	Methylation Difference (%)	Distance (bp)	P Value
NR_047528	Hotair	HOX transcript antisense RNA	−65.29	5,258	0.005
NR_047528	Hotair	HOX transcript antisense RNA	−64.92	5,283	0.001
NR_047528	Hotair	HOX transcript antisense RNA	−64.29	5,308	0.001
NM_010465	Hoxc6	Homeobox C6	−61.24	12,733	0.003
NM_010465	Hoxc6	Homeobox C6	−61.14	12,808	<0.001
NR_027899	Hoxd3os1	Homeobox D3, opposite strand 1	−61.05	13,110	<0.001
NM_010465	Hoxc6	Homeobox C6	−61.03	12,833	<0.001
NM_010465	Hoxc6	Homeobox C6	−60.90	12,858	<0.001
NR_027899	Hoxd3os1	Homeobox D3, opposite strand 1	−60.74	13,135	<0.001
NM_010465	Hoxc6	Homeobox C6	−60.72	12,883	0.002
NR_027899	Hoxd3os1	Homeobox D3, opposite strand 1	−59.88	13,060	0.005
NM_010466	Hoxc8	Homeobox C8	−58.96	0	0.031
NR_027899	Hoxd3os1	Homeobox D3, opposite strand 1	−58.94	13,185	<0.001
NR_037977	Gm53	Predicted gene 53	−57.17	0	<0.001
NM_010452	Hoxa3	Homeobox A3	−56.94	0	0.015
NM_010466	Hoxc8	Homeobox C8	−56.17	0	0.005
NR_027899	Hoxd3os1	Homeobox D3, opposite strand 1	−55.79	13,235	0.002
NM_172839	Ccnj	Cyclin J	−54.26	0	0.015
NM_010451	Hoxa2	Homeobox A2	−51.90	0	0.033
NM_010466	Hoxc8	Homeobox C8	−51.73	0	0.003
NM_010451	Hoxa2	Homeobox A2	−51.06	0	0.040
NM_008274	Hoxd12	Homeobox D12	−50.78	0	0.009
NM_010465	Hoxc6	Homeobox C6	−50.04	0	0.044
NR_131758	Hoxb5os	Homeobox B5 and homeobox B6, opposite strand	−50.01	25,927	0.023
NM_010465	Hoxc6	Homeobox C6	−48.18	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−47.36	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−47.06	0	0.003
NM_008274	Hoxd12	Homeobox D12	−47.05	0	<0.001
NR_037977	Gm53	Predicted gene 53	−47.00	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−46.63	0	0.002
NM_010465	Hoxc6	Homeobox C6	−46.52	0	0.011
NM_008274	Hoxd12	Homeobox D12	−46.48	0	0.005
NM_008274	Hoxd12	Homeobox D12	−46.38	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−46.36	0	0.008
NR_037977	Gm53	Predicted gene 53	−46.34	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−46.08	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−46.03	0	0.012
NR_037977	Gm53	Predicted gene 53	−45.94	0	<0.001
NM_008274	Hoxd12	Homeobox D12	−45.78	0	<0.001
NR_037977	Gm53	Predicted gene 53	−45.72	0	<0.001
NR_037977	Gm53	Predicted gene 53	−45.65	0	0.001
NR_037977	Gm53	Predicted gene 53	−45.64	0	<0.001
NM_026080	Mrps24	Mitochondrial ribosomal protein S24	−45.29	11,448	0.031
NM_008274	Hoxd12	Homeobox D12	−45.27	0	0.008
NR_037977	Gm53	Predicted gene 53	−44.81	0	<0.001
NM_175730	Hoxc5	Homeobox C5	−44.42	0	0.011
NM_008274	Hoxd12	Homeobox D12	−43.48	0	0.003
NR_037977	Gm53	Predicted gene 53	−43.23	0	0.004
NM_007967	Evx2	Even-skipped homeobox 2	−43.02	15,125	0.015
NM_175730	Hoxc5	Homeobox C5	−42.62	0	0.001

Discussion

Satellite cells are activated after a rotator cuff tear in mice, and their biological activity is thought to play an important role in pathological changes in rotator cuff disease²⁸. Meyer and colleagues found that cells from patients with partial-thickness tears had reduced proliferative capacity in vitro, but no difference in fusion, when compared with cells from untorn rotator cuff muscles and full-thickness tears²⁹. In an vivo study, Lundgreen et al. reported reduced satellite cell density and fewer proliferating cells in full-thickness rotator cuff tears compared with partial tears³⁰. While these studies identified differential activity of satellite cells within rotator cuff and shoulder girdle muscles in various states of injury, in the current study, we compared the activity of satellite cells from normal rotator cuff and gastrocnemius muscles, as the gastrocnemius is another commonly injured muscle that generally has better outcomes than the rotator cuff when recovering from a tendon tear^2,31,32. We observed a reduced myogenic capacity of rotator cuff satellite cells, along with a decreased expression of the differentiated myogenic transcription factor MRF4²⁰. No differences in the muscle cell-fusion gene myomaker³³ were observed, indicating a similar capacity of myogenic cells from the gastrocnemius and rotator cuff to fuse into myotubes. These results were also in agreement with a previous study of rats, which demonstrated greater fatty infiltration and reduced healing of the rotator cuff compared with the gastrocnemius⁷.

There are 3 types of differentiated adipogenic cells, including white, brown, and beige adipocytes³⁴. White adipose cells store fat and are the classical adipocyte cell type^35,36. Within muscle tissue, the primary progenitor cell for white adipocytes is thought to be fibro/adipogenic progenitor (FAP) cells, which are a distinct lineage from satellite cells^35,36, although myogenic cells can also differentiate into this lineage^12,13, with debate ongoing³⁴. Brown and beige adipocytes are related in function but arise from distinct populations of cells, with brown adipocytes having a myogenic origin and beige cells originating from a lineage that is likely similar to white adipocytes^34,37. It can be difficult to morphologically discern the 3 cell types in culture; however, a common feature among all adipogenic cells is the expression of FABP4^22,38. In our findings, all FABP4⁺ cells were also Pax7-tdTomato⁺, indicating that these adipocytes originated from a myogenic progenitor population. Further, we observed a 4-fold increase in adipogenic cells from rotator cuff muscles, and a 12-fold increase in the common adipogenic master regulator PPARγ, providing additional support for the finding of greater adipogenic differentiation capacity of rotator cuff satellite cells.

Numerous changes in DNA methylation were observed between rotator cuff and gastrocnemius satellite cells, and bioinformatics analysis identified several biochemical pathways involving adipogenesis and lipid metabolism that were predicted to be different between gastrocnemius and rotator cuff muscles. Many of the genes that were differentially methylated in rotator cuff samples were members of the HOX family of genes. The HOX genes encode transcription factors that were originally identified by their role in instructing the positional identity of progenitor cells along the anterior-to-posterior body axis³⁹. HOX genes also play important roles in myogenesis⁴⁰, and some of the differences in HOX methylation may be related to the more proximal location of rotator cuff muscles compared with the gastrocnemius in the limb, in particular with HOX9, HOX10, HOX11, and HOX13, which display a proximal-to-distal gradient of restricted expression in the developing limb³⁹. However, some of the HOX genes are also important in brown and beige adipogenesis, in particular HOXC4 and HOXC8⁴¹. In satellite cells from rotator cuff muscles, there were 2 hypomethylated regions for HOXC4, and 6 for HOXC8. HOXA3 has also been reported to be important in white adipogenesis⁴², and 8 hypomethylated regions for HOXA3 were found in rotator cuff satellite cells. As hypomethylation of a gene is associated with an increased expression of that gene, the combined results of this study indicate that satellite cells from the rotator cuff are more likely to become adipogenic cells, and this may be explained by differential methylation of adipogenic genes.

There were several limitations to this work. Humans frequently develop more profound and severe atrophy and fat accumulation than found in mouse models of rotator cuff disease^4,43,44. We only evaluated changes in adult male mice, which allows for examination of DNA methylation on both the X and Y chromosomes, although the observed results are likely applicable to both sexes. Differentiation experiments were performed in vitro, but it is possible that these findings do not entirely translate to the in vivo setting. We did not identify the specific type of adipocytes in our studies, but white and beige fat cells are known to be present in rotator cuff muscles⁴⁵, and there are brown fat depots located close to the rotator cuff^14,46. We also did not evaluate changes in chromatin packaging and histone methylation, which are also epigenetic regulatory mechanisms. ERRBS analysis also focuses gene promoter regions²⁵, but methylation of other regions of DNA could also play important roles in regulating the activity of satellite cells. Despite these limitations, the current work provides important insight into our understanding of the cellular development of rotator cuff disease.

The rotator cuff is a clinically unique muscle group with regard to pathophysiology, surgical treatment, and rehabilitation^7,47. In the current study, we sought to determine if satellite cells from gastrocnemius and rotator cuff muscles differ in their biological activity and epigenetic imprinting. Supporting our hypothesis, we found reduced myogenic and increased adipogenic differentiation of satellite cells from rotator cuff muscles, and differences in DNA methylation patterns that correspond to observed phenotypic differences between the 2 muscle groups, which helps to identify a cellular and genetic basis of the generally poor rates of rotator cuff muscle healing. As satellite cells are activated after injury to repair damaged muscle fibers⁹, and animal models have demonstrated that the repair of chronically torn rotator cuffs causes extensive injury throughout the muscle⁴⁸, it is possible that increased differentiation of myogenic cells into the adipogenic lineage contributes to the continued accumulation of fat that is observed in many patients after rotator cuff repair². Further, transplantation of satellite cells from a healthy muscle to heal diseased muscles within the same patient has shown some promise in early clinical trials⁴⁹ and has been proposed as a therapy for patients with chronic rotator cuff tears⁵⁰. Our results provide additional support for the potential use of autologous satellite cell transplantation to improve the treatment of patients with chronic rotator cuff disease.

Appendix

Details of satellite cell isolation and flow cytometry, cell culture, and DNA methylation analysis and a table listing the primer sequences used for quantitative PCR; as well as tables presenting a full list of the differentially methylated regions and the differentially methylated cytosines are available with the online version of this article as a data supplement at jbjs.org (http://links.lww.com/JBJS/F35, http://links.lww.com/JBJS/F36, http://links.lww.com/JBJS/F37).