Galectin-3 promotes the adipogenic differentiation of PDGFRα⁺ cells and ectopic fat formation in regenerating muscle

In recent years, the obese population has been increasing worldwide, and its prevalence has nearly doubled since the 1980s. Under obese conditions, muscle disuse and chronic inflammation caused by adipose-derived inflammatory cytokines promote muscular atrophy via the apoptosis of muscle fibers (Rubio-Ruiz et al., 2019; Sinha et al., 2017). Obesity-associated muscle atrophy causes further muscle loss, forming a vicious cycle (Stenholm et al., 2008). In addition to its effects on muscle mass, obesity increases fat content within muscle fibers (Greco et al., 2002) and also induces the accumulation of intermuscular adipose tissue (IMAT) (Goodpaster et al., 2000), which has been implicated in a decrease in muscular strength (Hilton et al., 2008; Visser et al., 2002) and muscle quality (Shirasawa et al., 2017), increased insulin resistance (Arrighi et al., 2015; Lee et al., 2012; Zoico et al., 2010) and impaired muscle regeneration (Hu et al., 2010). However, regardless of the potential pathogenic roles of IMAT, the mechanisms underlying its development under obese conditions remain largely unknown.

Recently, we identified PDGFRα⁺ mesenchymal progenitors (hereafter PDGFRα⁺ cells) as the origin of IMAT (Uezumi et al., 2014, 2010, 2011). These muscle-resident progenitors are commonly referred to as fibro-adipogenic progenitors (FAPs) because they are the major source of fibroblasts and adipocytes in the muscle (Joe et al., 2010). Upon muscle injury, these PDGFRα⁺ cells rapidly expand and promote the differentiation of muscle satellite cells, largely contributing to muscle regeneration (Joe et al., 2010; Murphy et al., 2011). These PDGFRα⁺ cells are then eventually induced to apoptosis by TNFα secreted by M1 macrophages (Lemos et al., 2015). However, it has been reported that in aged muscles or under pathological conditions, such as neuromuscular diseases, a portion of PDGFRα⁺ cells evade apoptosis by TGFβ secreted by M2 macrophages, and therefore differentiate into adipocytes (Lemos et al., 2015), although the trigger that elicits the adipogenic differentiation of surviving PDGFRα⁺ cells remains unknown. Based on this observation, treatment with nilotinib, a tyrosine kinase inhibitor that inhibits TGFβ-mediated signaling, suppresses the adipogenic differentiation of PDGFRα⁺ cells (Davies et al., 2016; Lemos et al., 2015). However, nilotinib also attenuates muscle regeneration as it inhibits the normal expansion of PDGFRα⁺ cells that are necessary for satellite cell proliferation (Fiore et al., 2016). In addition, all-trans retinoic acid (RA) has been reported as another potent agent that suppresses the adipocyte differentiation of PDGFRα⁺ cells (Mogi et al., 2016). Unfortunately, RA also exerts considerable adverse effects, such as skin disorders, liver damage, and serum lipid abnormalities that could elevate the risk of coronary heart diseases (David et al., 1988). Therefore, currently there is no effective strategy for the prevention or removal of obesity-induced IMAT, and thus a further understanding of the mechanisms underlying IMAT accumulation is urgently required. Particularly, the obesity-related factor that triggers adipocyte differentiation signals in PDGFRα⁺ cells remains elusive and clarifying this would contribute to the development of new therapeutic strategies against IMAT development under obese conditions.

Recently, the advent of single-cell RNA sequencing (scRNA-Seq) technology has enabled studies of the transcriptome within individual cells, thereby allowing for the identification and analysis of the characteristics of rare cell populations (Hwang et al., 2018). Using scRNA-Seq technology, one can even delineate the specific differentiation process of cells. Indeed, such studies on mesenchymal stem cells from adipose tissue or bone marrow have identified cells in an adipocyte progenitor-like state among cells undergoing adipocyte differentiation (Burl et al., 2018; Hepler et al., 2018; Zhong et al., 2020). Regarding PDGFRα⁺ cells in muscles, the changes during post-traumatic muscle regeneration have also been studied using this approach (De Micheli et al., 2020; Oprescu et al., 2020). These studies revealed the cellular heterogeneity of PDGFRα⁺ cells and the upregulation of genes related to extracellular matrix remodeling and immune cell infiltration during the early phase of normal muscle regeneration, but not in an obese situation. It should be noted, however, that although scRNA-Seq is a powerful tool to investigate the changes in gene expression at the cellular levels, it does not provide direct information on the extracellular/interstitial muscle environment surrounding PDGFRα⁺ cells. Importantly, it has been suggested that a muscular microenvironment plays a crucial role in regulating the adipocyte differentiation of PDGFRα⁺ cells (Uezumi et al., 2010). Consistent with this notion, it has been reported that the extracellular matrix-regulating proteins MMP14 and AnxA2 promote the adipocyte differentiation of PDGFRα⁺ cells in muscle during aging and in muscle dystrophy, respectively (Hogarth et al., 2019; Kopinke et al., 2017).

Here, we investigated the obesity-induced adipocyte differentiation process of PDGFRα⁺ cells through a combination of scRNA-Seq and proteomic analysis. Our multi-omics approach identified a small PDGFRα⁺ cell population with high adipogenic signaling activity and also found galectin-3 (Lgals3) in the interstitial environment at the early phase of muscle regeneration as a strong candidate for the obesity-related trigger of adipocyte differentiation. Our findings and datasets lay a foundation for the further study of obesity-induced IMAT formation, which could lead to the development of therapeutic strategies against IMAT development under obese conditions.

We first tested whether high-fat diet (HFD)-induced obesity itself would trigger ectopic IMAT accumulation. Long-term HFD-fed obese mice did indeed accumulate IMAT in their quadriceps muscles but not in the anterior tibialis and gastrocnemius, which did not occur in normal diet (ND)-fed mice (Fig. S1A). Although the HFD-fed obese mouse is a natural, typical model that mimics human obesity, IMAT accumulation was modest and slow in this model with only an HFD, which renders the IMAT formation process difficult to study. Though it has been recognized that intramuscular injection of glycerol induces IMAT formation (Arsic et al., 2004), we thought that this model was not appropriate for studying the IMAT formation process in obesity because it also impairs the process of muscle regeneration. Given that it has been reported that muscle regeneration under pathological conditions promotes IMAT formation (Mogi et al., 2016), we injected cardiotoxin (CTX), a cobra venom that induces myolysis of myofibers by inducing plasma membrane depolarization, into the anterior tibialis of ND- or HFD-fed mice. Two weeks after CTX treatment, IMAT with positive staining of an adipocyte membrane marker, perilipin, had accumulated in the anterior tibialis of HFD-fed mice but not in that of ND-fed mice (Fig. S1B-D). We further investigated the time point at which IMAT was formed after CTX treatment. We found that IMAT formation in HFD-fed mice predominantly occurred between days 4 and 7 after CTX treatment, and IMAT accumulation was still evident at day 14 in HFD-fed mice (Fig. 1A,B). Of note, genetically obese mice lacking leptin (ob/ob) or the leptin receptor (db/db) also exhibited CTX-induced accumulation of IMAT, indicating that this phenomenon of ectopic adipocyte formation was induced by obesity itself and is independent of diet composition (Fig. S1E,F). When the CTX-induced muscle injury was repaired, we compared the histological features of muscular regeneration between the muscles of ND- and HFD-fed mice. Although the wet weight of muscle tissues and the number of muscle fibers were recovered to a similar extent in ND- and HFD-fed mice (Fig. S1G,H), the regenerated muscle fibers were atrophied, accompanied by an increase in the fat area in HFD mice, a feature similar to sarcopenia under obese conditions (Fig. S1I,J). We next sought to confirm whether the origin of IMAT was PDGFRα⁺ cells in our model, as reported in other models (Kasai et al., 2017; Uezumi et al., 2010). To this end, we took advantage of lineage-tracing technology. Although IMAT completely loses PDGFRα expression when PDGFRα⁺ cells differentiate into adipocytes, the original PDGFRα positivity can be continuously monitored after differentiation using lineage-tracing Pdgfrα-CreERT2::Rosa26-YFP mice (Uezumi et al., 2010). As expected, CTX-induced IMAT in HFD Pdgfrα-CreERT2::Rosa26-YFP mice showed YFP expression, indicating that the cells comprising IMAT were derived from a PDGFRα⁺ cell origin (Fig. 1C-E).

Next, to gain insights into how the PDGFRα⁺ cells differentiate into adipocytes during muscle regeneration upon muscle injury under obese conditions, we conducted scRNA-Seq analysis of PDGFRα⁺ cells isolated from ND- and HFD-fed mice before and after CTX treatment (Fig. 2A). In addition to the anterior tibialis, the gastrocnemius and quadriceps muscles were also treated with CTX and collected at 4 and 7 days post-muscle injury for scRNA-Seq analysis. For each of six samples (day 0, 4 and 7 from ND- or HFD-fed mice after CTX treatment), whole hindlimb muscles (anterior tibialis, gastrocnemius and quadriceps muscles) were collected from three mice, combined, and physically and enzymatically dissociated. The PDGFRα⁺ cells were purified by cell sorting (CD31⁻/CD45.2⁻/SMC2.6⁻/PDGFRα⁺), and then single-cell cDNA libraries were constructed using the 10X Genomics Chromium platform. For each sample, at least 90.9% of analyzed cells passed the quality controls, and quality-filtered single-cell transcriptomic data were obtained for at least 8687 cells, suggesting that the resulting data were of high quality (Fig. S2A,B). Uniform manifold approximation and projection (UMAP) embedding (Butler et al., 2018) of the entire dataset resulted in the clear separation of cells collected at different time points (Fig. 2B). PDGFRα expression was confirmed in at least 92.6% of sorted cells in all samples, indicating the successful cell sorting of PDGFRα⁺ cells.

We first checked whether the marker genes of muscle regeneration are sequentially and coordinately expressed during this process (Fig. 2C). Cxcl5, which plays a role in recruiting the monocytes and macrophages necessary for muscle regeneration at the early stage (Tidball, 2017), was predominantly expressed in the cells collected at 4 days after CTX treatment. Next, Wisp1 (Ccn4), which plays a role in promoting the proliferation and differentiation of muscle satellite cells during the early to middle phase of muscle regeneration (Lukjanenko et al., 2019), was mainly expressed in the cells collected at 4 and 7 days after CTX treatment. Delta like protein-1 (Dlk1), which is known to be expressed at the middle phase of regeneration, was indeed expressed in cells collected at 7 days after CTX treatment. Finally, Dpp4 and Cxcl14, which are known to be expressed in quiescent cells, were mainly expressed in cells collected before CTX treatment (day 0). The gene expression profiles during muscle regeneration in our experiment were basically consistent with previous reports (De Micheli et al., 2020; Oprescu et al., 2020), and these gene expression changes occurred similarly in ND- and HFD-fed mice (Fig. S2C), suggesting that reasonable muscle regeneration processes were taking place in our samples.

Considering the overall similarity of ND and HFD samples, we decided to search for a rare cell population that might explain the obesity-induced ectopic adipocyte differentiation of PDGFRα⁺ cells. We previously reported that obesity induces cellular senescence (Gorgoulis et al., 2019), a state of irreversible cell cycle arrest, in a portion of hepatic stellate cells in the liver tumor microenvironment (Yoshimoto et al., 2013). We also reported that fibroblasts are induced to undergo senescence during skin wound repair, thereby contributing to skin regeneration (Demaria et al., 2014) through a so-called senescence-associated secretory phenotype (SASP), in which cells secrete various growth factors, cytokines, chemokines and matrix metalloproteinases (Coppé et al., 2010). Based on these findings, we hypothesized that cellular senescence might be differentially induced during muscle regeneration in ND- and HFD-fed mice in PDGFRα⁺ cells, which could affect the differentiation of nearby cells. Interestingly, PDGFRα⁺ cells expressing the senescence markers p16 (Cdkn2a) and p21 (Cdkn1a) appeared transiently at day 4 after CTX treatment and predominantly co-expressed the myofibroblast marker Acta2 (also known as α-SMA; Fig. 2D). We speculated that this transient emergence of senescent cells could play a role in scavenging cells produced in the tissue-regenerating stroma, as was reported in the skin (Childs et al., 2017; Demaria et al., 2014). Consistent with p16 and p21 expression as senescence markers, p16⁺/p21⁺/Acta2⁺ cells expressed relatively high levels of other senescence- and SASP-related genes, such as those involved in p53 (Trp53) and TNF signaling, compared with the other Acta2⁺ cells (Fig. 2E). By contrast, the expression levels of genes involved in PI3K-Akt signaling and cancer-related pathways were lower in these cells (Fig. 2F). These observations strongly suggest that these cells are senescent and exhibit SASP. Interestingly, among p16⁺/p21⁺/Acta2⁺ cells, a comparison of gene expression profiles between ND- and HFD-fed mice revealed that genes induced by PPARγ signaling were upregulated in cells from HFD-fed mice (Fig. S2D). This result shows that PDGFRα⁺ senescent cells from HFD-fed mice have a weak tendency for adipogenesis in addition to the SASP phenotype. The proportion of PDGFRα⁺ senescent cells, however, was not different between ND- and HFD-fed mice (χ² test P=0.11).

We next adopted an unbiased approach to identify the cell population directed towards adipogenesis using UMAP clustering (Fig. 3A,B). For each cluster, we conducted ontology enrichment analysis using a set of genes that are differentially expressed between cells derived from ND- and HFD-fed mice (Fig. S3, Table S1). Notably, we found that genes involved in the PPARγ signaling pathway were expressed at higher levels in the HFD-fed mouse-derived PDGFRα⁺ cells, and these clusters of cells were mapped to cluster 0 and 8, which were both mainly composed of cells collected 4 days after CTX treatment (Fig. 3C,D). PPARγ is a master transcription factor that regulates adipogenesis and has been shown to play a crucial role in adipocyte differentiation in PDGFRα⁺ cells in a glycerol-injected model of ND-fed mice (Dammone et al., 2018). This suggests that the IMAT formation observed in our model might originate from the cells in cluster 0 and/or 8. Single-cell trajectory analysis using Monocle 3 (Qiu et al., 2017) depicted a reasonable trajectory that starts from the early phase of regeneration and then goes through the mid-phase of regeneration, ending back in the uninjured state (Fig. 4A). However, interestingly, in cluster 0, we observed a deviation from the normal regeneration trajectory, which resulted in a cell population located at a distinct, protruded part of cluster 0. Moreover, genes targeted by adipogenic transcription factors such as CEBPα, CEBPβ, CEBPδ, PPARα, PPARδ and PPARγ showed dynamic changes along the deviated trajectory pathway toward the protrusion. Notably, the most definitive adipogenic transcription factor, PPARγ, was strongly associated with the genes that were upregulated only at the end-phase of the regeneration (Fig. 4B, Fig. S4A). Consistent with this finding, the adipocyte marker Plin1 (perilipin 1), which is also a target of PPARγ, was almost exclusively expressed in the cells comprising this protrusion (Fig. 4C). Importantly, most of these Plin1⁺ cells were PDGFRα⁺ cells from HFD-fed mice (Fig. 4D). Plin1⁺ cells exhibited a very strong signature of PPARγ signaling pathway activation (Fig. 4E) and expressed markers of adipogenesis (Adipoq, Car3, Cd36, Fabp4, Fabp5 and Pparg) (Fig. 4F). Cells expressing these adipogenic markers were responsible for PPARγ target gene expressions among cluster 0 (Fig. S4B), suggesting that IMAT could have originated from this small cell population.

Lastly, to explore which factor activates PPARγ signaling in a subset of PDGFRα⁺ cells at the early phase of muscle regeneration under obese conditions, we performed proteomic mass spectrometry of interstitial stromal tissue encompassing PDGFRα⁺ cells. The interstitial tissues were microdissected from the muscle of ND- and HFD-fed mice 4 days after CTX treatment (Fig. 5A). The most obvious difference between ND- and HFD-fed mice was that a series of histones was abundantly detected in HFD-fed mouse samples (Fig. 5B, Table S2). This could be due to the accumulation of extracellular histones that are released from dead cells, which can then function as damage-associated molecular patterns (DAMPs) (Huang et al., 2011). More interestingly, we found that galectin-3, a positive regulator of PPARγ signaling (Baek et al., 2015), was expressed at higher levels in the interstitial stromal tissue of HFD-fed mice. Therefore, we freshly isolated PDGFRα⁺ cells from ND-fed mice and treated them with recombinant galectin-3. In support of the biological significance of this observation, we confirmed that recombinant galectin-3 enhanced PPARγ expression and contributed adipogenic differentiation in primary cultured PDGFRα⁺ cells (Fig. 5C,D). By contrast, the galectin-3 inhibitor TD139 suppressed PPARγ expression in primary cultured PDGFRα⁺ cells (Fig. 5E). Next, to identify which type of cells express galectin-3, we isolated PDGFRα⁺ cells, muscle satellite cells (SMC2.6⁺), vascular endothelial cells (CD31⁺) and immune cells (CD45.2⁺) from the muscles of HFD-fed mice at day 4 post-CTX treatment. Real-time qPCR analysis of these cells showed that galectin-3 was predominantly expressed in immune cells (Fig. 5F). This finding is consistent with the previous studies reporting that macrophages are the major source of galectin-3 (Dumic et al., 2006; Ho and Springer, 1982). Importantly, intraperitoneal injection of the galectin-3 inhibitor TD139 before and after muscle injury (Fig. 5G,H) as well as galectin-3 knockout (Fig. 5I-L) significantly suppressed IMAT development under obese conditions, confirming the contribution of galectin-3 to IMAT development in vivo. Taken together, our results suggest that obesity induces excess secretion of galectin-3 from immune cells upon muscle injury/regeneration and thereby activates the PPARγ signaling pathway in PDGFRα⁺ cells, leading to a deviated differentiation process directed to ectopic intermuscular adipogenesis.

Over the past several decades, obesity has become prevalent worldwide (Batsis and Villareal, 2018). Accordingly, obesity-associated diseases, such as cardiovascular disease, type 2 diabetes, obesity-associated cancers, and sarcopenic obesity (Kalinkovich and Livshits, 2017), are increasingly recognized as serious public health problems. Ectopic intermuscular adipose tissue deposition, termed IMAT, is recognized as one of the pathological characteristics of obesity, because IMAT contributes to the impaired and unhealthy status of muscles via the accumulation of oxidated lipids and an increase in the anaerobic muscle microenvironment (Batsis and Villareal, 2018). In this study, we aimed to elucidate the mechanism underlying ectopic IMAT formation in an HFD-induced obese mouse model and searched for novel molecular targets for IMAT therapy.

Mouse models suitable for analyzing the molecular pathology of the ectopic formation of IMAT had not been established. Mogi et al. previously reported that ectopic IMAT formation occurs in both dietary and genetic obese mouse models during muscle repair processes after CTX-induced muscle injury (Mogi et al., 2016); however, they did not provide any data regarding the origin or the timing of IMAT formation. It should be noted that orthopedic injury can induce IMAT accumulation in humans (Goutallier et al., 1994) and that sarcopenia patients with such injuries tend to exhibit more severe symptoms (Chung et al., 2016; Elliott et al., 2006). These observations support the physiological relevance of our model with respect to sarcopenia. To identify the origin of the obesity-induced IMAT formation, we took advantage of a lineage-tracing system of PDGFRα-expressing mesenchymal progenitor cells (Fig. 1C-E) (Uezumi et al., 2010) because PDGFRα⁺ mesenchymal progenitor cells are known to downregulate the expression of this marker gradually during differentiation and entirely lose it when differentiation is completed. Using this system, we found that cells originally expressing PDGFRα did indeed differentiate into adipose tissue in HFD-induced obese mice but not in ND-fed mice. Moreover, we performed scRNA-Seq analysis using PDGFRα-expressing cells isolated from muscles at multiple time points during muscle regeneration after CTX-induced muscle injury. The time-course experiments confirmed that the PDGFRα⁺ cells residing in the stomal slits between muscle fibers were the origin of ectopic IMAT, and we also identified the obvious pre-adipogenic lineage cells that emerged at the early stage after muscle injury (Fig. 1B). We considered that PDGFRα⁺ cells, which are capable of differentiating into adipocytes, were plausible candidates for the origin of ectopic IMAT. Therefore, we analyzed the gene expression profile of each single PDGFRα⁺ cell isolated from skeletal muscle in HFD-induced obese mice and compared it with that of cells from ND-fed mice. Interestingly, during this scRNA-Seq, we observed a transient emergence of senescent cells highly expressing p16 and p21 in an Acta2-expressing group among PDGFRα⁺ cells only in the early stage (day 4) of the muscle regeneration process after muscle injury (Fig. 2D). Several studies have reported that transient induction of the SASP, a phenotype in which senescent cells express a series of secreted proteins, such as inflammatory cytokines, chemokines, proteases and growth factors, among others, could be observed during tissue repair. Furthermore, this is believed to be an important physiological role of the SASP, although a continuous SASP could conversely result in pathological effects, such as aging-associated chronic inflammation and cancer progression, and can even impair muscle regenerative capacity (Campisi, 2011; Sousa-Victor et al., 2014). Other reports have suggested that the accumulation of senescent adipocytes and preadipocytes is associated with exacerbation of diabetes; specifically, the upregulation of p53 in senescent adipocytes promotes the production of inflammatory cytokines from adipose tissue (Minamino et al., 2009) and the senescent preadipocytes in subcutaneous adipose tissue are associated with the hypertrophy of adipocytes (Gustafson et al., 2019), both of which lead to increased insulin resistance. Our results also revealed that adipogenic genes induced by PPARγ signaling tend to be upregulated in PDGFRα⁺ senescent cells from HFD-fed mice, suggesting that these cells are slightly shifted towards the adipogenic lineage (Fig. S2D). Furthermore, our analysis identified a list of genes, including SASP-associated genes, from PDGFRα⁺ senescent cells specifically associated with muscle repair after injury. These SASP factors can be used as possible accelerators of muscle regenerative therapy, and they might also represent useful clues to elucidate the pathogenesis of aging- and obese-associated sarcopenia.

Previous studies suggested that a transmembrane protein, delta-like protein-1 (Dlk1), also known as preadipocyte factor-1 (Pref-1) or zona glomerulosa-specific factor (ZOG), is a specific pre-adipocyte marker during adipogenic differentiation not only from preadipocytes (Merrick et al., 2019; Hudak and Sul, 2013) but also from PDGFRα⁺ mesenchymal stem cells (Oprescu et al., 2020). However, deletion of the Dlk1 gene was reported to not alter the adipose mass or muscle mass in mice, suggesting that it alone is not associated with adipogenesis (Zhang et al., 2019). In addition, we detected a Dlk1-expressing population both in ND- and HFD-fed mice at the later stage (day 7) of muscle regeneration, when IMAT had already developed. Our immunohistochemistry study showed that the development of IMAT starts before this stage. Consistent with this, our scRNA-Seq analysis identified an obesity-associated adipogenic PDGFRα⁺ cell population at the early stage (day 4) of muscle regeneration. These results suggest that Dlk1-expressing cells are unlikely to be a preadipocyte lineage for adipocyte differentiation during muscle regeneration. Moreover, although excess expression of Dpp4 has been reported to be associated with predisposition to adipogenesis in obesity (Ambrosi et al., 2017; Merrick et al., 2019), our data revealed no difference in Dpp4 expression between cells from HFD-fed and ND-fed mice (Fig. S2C). These observations suggest that PDGFRα⁺ cells expressing Dlk1 or Dpp4 might have different roles during muscle regeneration.

Interestingly, we identified galectin-3 as a possible obesity-associated inducer of adipocyte differentiation in PDGFRα⁺ cells through the proteomic analysis of interstitial stroma at an early stage of muscle regeneration (Fig. 5B). Galectin-3, an approximately 30-kDa multi-functional lectin family protein, contains a carbohydrate-recognition-binding domain that binds β-galactosides. Galectin-3 is ubiquitously localized in the nucleus, cytoplasm and extracellular membrane to cross-link non-carbohydrate and carbohydrates and is involved in cell adhesion, the cell cycle, apoptosis, inflammation, cell proliferation and differentiation by acting as monomer or oligomer depending on its density (Henderson et al., 2006). Because galectin-3 is known as an inflammatory modulator and fibrosis accelerator in the heart and liver (Sharma et al., 2004), suppression of its function is expected to be a molecular target for anti-inflammatory and anti-fibrosis therapy. Moreover recently, galectin-3 has been increasingly recognized as an adipogenesis promoter. Galectin-3 was reported to be upregulated in growing adipose tissue, to stimulate preadipocyte proliferation (Kiwaki et al., 2007) and to activate PPARγ directly to induce its target genes and other lipogenic genes. In this study, we showed that treatment with galectin-3 did indeed promote the adipogenic differentiation of primary PDGFRα⁺ mesenchymal progenitor cells derived from muscle, and galectin-3 inhibitor/deficiency suppressed IMAT formation, suggesting that the inhibition of galectin-3 could also be used to prevent ectopic IMAT formation in HFD-induced obese mice as well as humans.

The limitation of our study is that we performed the scRNA-Seq analysis focusing on a single cell type, PDGFRα⁺ mesenchymal progenitor cells isolated from regenerating muscle after CTX-induced muscle injury. Therefore, the interactions between other types of muscle comprising components such as muscle fiber, satellite cells, immune and endothelial cells were not investigated. However, owing to the abundance of initial cell numbers and focusing on this single cell type, we were able to identify a small but distinct, functional population of preadipocytes and senescent cells expressing SASP factors among the PDGFRα⁺ cells by our single-cell analysis during muscle regeneration in HFD-induced obese mice.

In conclusion, we identified a preadipocyte population among PDGFRα⁺ mesenchymal progenitor cells that could cause ectopic IMAT formation in the early stage of muscle regeneration in HFD-induced obese mice. Moreover, the increase in interstitial stromal galactin-3 produced by immune cells was found to promote the ectopic adipocyte differentiation of PDGFRα⁺ cells. In this study, we clearly showed the usefulness of the muscle regeneration model after muscle injury to elucidate the molecular mechanisms of ectopic IMAT formation, one of the important characteristics of obesity. Our data provide a deeper understanding of ectopic IMAT formation and will provide valuable new insights into the use of galectin-3 as a molecular target for the prevention and the treatment of IMAT in obesity.

Pregnant female C57BL/6 mice were purchased from CLEA Japan, and the mother mice and pups were fed a normal diet (ND; CE-2 from CLEA Japan) or a high-fat diet (HFD; D12492 from Research Diets). When mice were 20 weeks old, they were used for the muscle-regeneration experiments. PDGFRα-CreERT2::Rosa26-YFP transgenic mice were generated by Akiyoshi Uezumi (Tokyo Metropolitan Institute of Gerontology, Japan). The leptin-mutant (ob/ob) and leptin receptor-mutant (db/db) mice (C57BL/6) were purchased from Charles River Laboratories Japan, and fed a ND (CE-2 from CLEA Japan) and 20-week-old mice were used for muscle-regeneration experiments. All studies used female mice except for those using male ob/ob mice and galectin-3 knockout mice. The mice were maintained under specific pathogen-free conditions, on a 12-h light-dark cycle. All procedures using experimental animals were approved by the Institutional Animal Care and Use Committee at Osaka City University (approval No. 17026).

To create the acute muscle injury model, 100 µl of CTX (10 µM in saline, Sigma-Aldrich) was injected with a 29-gauge needle under anesthesia into the tibialis anterior muscle of mice for histological analyses or six hindlimb muscles, including bilateral quadriceps, gastrocnemius and tibialis anterior muscles, for scRNA-Seq.

Tibialis anterior muscles were isolated at days 0, 2, 4, 7 and 14 after CTX injury, embedded in tragacanth gum, and immediately frozen in liquid nitrogen-cooled isopentane (Wako Pure Chemical Industries) as described (Yoshimoto et al., 2020). Only tibialis anterior muscles from transgenic mice were fixed in 4% paraformaldehyde in PBS for 30 min, and muscles were sequentially soaked in 10% sucrose in PBS and 20% sucrose in PBS. Muscles were embedded in O.C.T. compound (Tissue-Tek Sakura) in aluminum foil and then frozen in liquid nitrogen-cooled isopentane. For immunofluorescence staining, transverse cryosections (7 μm) were stained with anti-laminin α2 (1:200, Santa Cruz Biotechnology, sc-59854), anti-perilipin (1:250, Sigma-Aldrich, P1873; or 1:100, Abcam, ab61682), anti-PDGFRα (1:80, R&D Systems, AF1062) and anti-GFP (1:1000, MBL International, #98) antibodies. After the first staining at 4°C overnight, sections were incubated with secondary antibodies conjugated with Alexa Fluor 594 or Alexa Fluor 488 (1:1000; Thermo Fisher Scientific, A11058, A11037, A11034 or A11006). Coverslips were mounted using ProLong Gold antifade reagent (Invitrogen). Sections were analyzed using a fluorescence microscope BZ-8000 (Keyence Corporation).

Cross-sections were made by cutting at the mid-belly of tibialis anterior muscles (at a position approximately 3 mm from the proximal end of the tibialis anterior muscle). After immunostaining, fluorescent images of entire cross-sections were captured with a fluorescence microscope system BZ-X710 (Keyence Corporation). Image recognition and quantification were performed using the Hybrid Cell Count Application (Keyence Corporation). The entire cross-sectional areas of tibialis anterior muscles were measured. For quantification of muscle fiber areas, laminin α2-stained sarcolemma was first recognized based on the intensity of staining by adjusting the threshold and, subsequently, the muscle fiber area was recognized using the inversion function. After recognition of the muscle fiber areas, the misrecognized small areas were excluded by adjusting the lower limit in the histogram function. For the quantification of adipocyte area, perilipin-1-stained areas were recognized based on the intensity of staining by adjusting the threshold, and the adipocyte areas inside perilipin-1-stained areas were filled manually. Finally, muscle fiber areas or adipocytes areas were calculated. A two-sided unpaired Mann–Whitney U-test was used to compare the two groups.

PDGFRα-CreERT2 mice were crossed with R26R-enhanced yellow fluorescent protein (EYFP) mice. Four milligrams of tamoxifen (Sigma-Aldrich) in corn oil was injected intraperitoneally for five consecutive days into 10-week-old PDGFRα-CreERT2::Rosa26-YFP mice to induce EYFP expression (Kasai et al., 2017).

Mononuclear cells from regenerating muscles at 4 or 7 days after CTX injection or from uninjured muscles were prepared as follows. CTX-injected or uninjured hindlimb muscles, including bilateral quadriceps, gastrocnemius and tibialis anterior muscles from three mice in each group and each time point were carefully dissected to remove attached tendons, nerves, blood vessels and fat tissue. Trimmed muscles were minced and digested with collagenase II (2 mg/ml, Worthington Biochemical Corporation) and DNase I (20 µg/ml) in Hanks Balanced Salt Solution (+) without Phenol Red (Wako Pure Chemical Industries) for 60 min at 37°C. Digested muscles were passed through an 18-gauge needle several times and further digested for 30 min at 37°C. Muscle slurries were filtered through a 100-µm cell strainer (BD Falcon) and through a 40-µm cell strainer (BD Falcon). Erythrocytes were eliminated by treating the cells with RBC lysis buffer (BioLegend). Mononuclear cells isolated from muscles were stained with anti-CD31 (1:100, BioLegend, 102513), anti-CD45 (1:100, BioLegend, 109805) and anti-SMC2.6 (1:200, a gift from So-ichiro Fukada, Osaka University, Osaka, Japan) antibodies to exclude endothelial cells, immune cells and satellite cells, respectively, as well as anti-PDGFRα (1:10, R&D Systems, FAB1062P) antibodies for collection. Anti-SMC2.6 is a monoclonal antibody reported to allow the separation of satellite and non-satellite cells (Fukada et al., 2004). Primary antibodies and secondary reagents used for cell staining are listed in Table S4. Cell sorting was performed using a Sony SH800 cell sorter (Sony). Debris and dead cells were excluded based on forward scatter, side scatter, and propidium iodide (1:100, BioLegend, 421301) gating. The analysis of stained cells was performed with a Sony SH800 cell sorter and/or Attune NxT (Thermo Fisher) analyzer, and data were processed by FlowJo Version 10 software (FlowJo).

After PDGFRα⁺ cells were sorted from murine muscles as mentioned, and single-cell suspensions were washed and resuspended in PBS at a concentration of at least 1000 cells/µl. scRNA-Seq libraries were then prepared using the Chromium Next GEM Single Cell 3′ GEM, Library & Gel Bead Kit v3.1 (10X Genomics) in accordance with the manufacturer's protocol. Following library preparation, the libraries were sequenced in multiplex (n=2 per sequencing run) on the next-generation sequencer (DNBSEQ-G400RS, MGI) to produce approximately 600 million reads per library (six libraries in total) and on average a minimum of 46,969 reads per single cell among the six samples. We were able to obtain RNA-Seq data from at least 9286 cells according to the Cell Ranger analysis (10X Genomics).

Sequencing reads were processed with the Cell Ranger version 3.1.0 (10X Genomics) using the mouse reference transcriptome mm10. From the gene expression matrix, the downstream analysis was carried out with R version 3.6.2 (2019-12-12). First, the six datasets were combined and then quality control, filtering, data clustering and visualization, and differential expression analysis were carried out using Seurat version 2.3.4 in the R package (Butler et al., 2018) with some custom modifications to the standard pipeline. Each of the six datasets was first analyzed independently before combining datasets from the same time point together for an integrated analysis. For each individual dataset, genes expressed in fewer than three cells, as well as <500 genes, were removed from the gene expression matrix. After log-normalizing the data, the expression of each gene was scaled based on the regression of the number of unique molecular identifiers in each cell. We performed principal components analysis on the gene expression matrix and used the first 55 principal components for clustering and visualization. Unsupervised shared nearest neighbor clustering was performed with a resolution of 0.5 and visualization was performed using UMAP (Becht et al., 2018). Finally, differential expression analysis was achieved using Seurat's ‘FindAllMarkers’ function based on a likelihood ratio test that assumes the data follows a negative binomial distribution and only considering genes with a >log₂(0.25) fold change and expressed in at least 10% of cells in the cluster (De Micheli et al., 2020). Ontology analyses were performed using Enrichr program (Chen et al., 2013).

We used the Monocle 3 version 0.2.2 R package (Qiu et al., 2017) to organize cells in pseudotime and infer new trajectories of PDGFRα⁺ cell subpopulations post-injury. First, we imported all cells and the coordinate data of UMAP embeddings from the Seurat dataset and across all time points and samples. Second, we constructed a k-nearest neighbor graph (k=30) on cells in the UMAP space, and then grouped the cells using Leiden community detection (resolution=1e−4). Finally, we inferred trajectories of PDGFRα⁺ cells using Monocle 3's ‘learn_graph’ function, which learns the principal graph from the reduced dimension space using reversed graph embedding.

Frozen muscle tissue of the murine anterior tibialis was sliced and stained with Sirius Red, and the interstitial areas were microdissected, as shown in Fig. 5A. Each dissected sample was dissolved in Laemmli sample buffer (Bio-Rad) and 10 µg of each boiled sample was then electrophoresed in a brand-new precast gel (4-20%; Bio-Rad). After a short period of electrophoresis, the visible whole-protein bands were dissected from the gel and used for proteomics analysis.

Proteins were digested and recovered using In-Gel Tryptic Digestion Kit (Thermo Fisher Scientific) according to the manufacturer's instruction. The recovered digests were resuspended in 0.1% formic acid and separated using Nano-LC-Ultra 2D-plus equipped with cHiPLC Nanoflex (Eksigent Technologies) in trap-and-elute mode, with trap column (200 μm×0.5 mm ChromXP C18-CL 3 μm 120 Å; Eksigent Technologies) and analytical column (75 μm×15 cm ChromXP C18-CL 3 μm 120 Å; Eksigent Technologies). The separation was carried out using a binary gradient in which 0.1% formic acid/water and 0.1% formic acid/acetonitrile were used as solvent A and B, respectively. The gradient program was as follows; 2% to 33.2% B for 125 min, 33.2% to 98% B in 2 min, 98% B for 5 min, 98% to 2% B in 0.1 min, and 2% B for 17.9 min. The flow rate was 300 nl/min. The analytical column temperature was set to 40°C. The eluates were infused on-line to a mass spectrometer (TripleTOF 5600+ System with NanoSpray III source and heated interface; SCIEX) and ionized in an electrospray ionization-positive mode. Data acquisition was carried out with an information-dependent acquisition method. The acquired datasets were analyzed using ProteinPilot software version 5.0.1 (SCIEX) for protein and peptide identification with the UniProtKB/Swiss-Prot database for Mus musculus (April 2020) appended with known common contaminants (SCIEX). The quality of the database search was confirmed by false discovery rate analysis in which the reversed amino acid sequences were used as decoy. Relative abundances of the identified proteins were estimated using Progenesis QI for Proteomics software version 4.2 (Nonlinear Dynamics). All raw data files with wiff format (SCIEX) were imported to generate aggregate, and the peptide identification results by ProteinPilot, with confidence at least 95%, were used for assignment. Label-free quantification of proteins was performed using relative quantitation using the Hi-N(3) method (Nonlinear Dynamics) (Silva et al., 2006).

First, sorted PDGFRα⁺ cells from mice were cultured on Matrigel-coated 48-well plates in growth medium consisting of DMEM supplemented with 20% fetal bovine serum, 1% penicillin-streptomycin, and 2.5 ng/ml bFGF (Thermo Fisher Scientific) and maintained at 37°C in 5% CO₂ and 3% O₂. Ten thousand PDGFRα⁺ cells were plated per well. After 3 days, for adipogenic differentiation, cells were treated with adipogenic induction medium consisting of DMEM with 10% fetal bovine serum, 0.5 mM IBMX (Sigma-Aldrich), 0.25 mM dexamethasone and 10 mg/ml insulin for 3 days at 37°C in 5% CO₂ and 20% O₂.

For galectin-3 experiments, the PDGFRα⁺ cells, as well as satellite cells, endothelial cells and immune cells from ND-fed mice, were isolated. The sorted primary PDGFRα⁺ cells were cultured and then treated daily with recombinant murine galectin-3 (200 nM, R&D Systems) or a galectin-3 inhibitor, TD139 (100 μM, Selleck Chemicals), 2 days after isolating. Six days after isolation, the total RNA from the cells was isolated using RNAiso Plus (Takara Bio), and quantitative PCR for Pparg gene expression was performed. Galectin-3 expression was also examined using isolated PDGFRα⁺ cells, as well as satellite cells, endothelial cells and immune cells. Quantitative PCR was performed as described in the ‘RNA isolation and quantitative PCR analysis’ section.

Total RNA was isolated from sorted cells or cultured cells using RNAiso Plus (Takara Bio), and the RNA was reverse-transcribed into cDNA using a PrimeScript RT Master Mix (RR036A; Takara Bio). Quantitative PCR was performed with Power SYBR Green Master Mix (4368702; Thermo Fisher Scientific) using the Applied Biosystems StepOnePlus Real-Time PCR System (Thermo Fisher Scientific) under the following cycling conditions: 50°C for 2 min and 95°C for 10 min, followed by 40 cycles of amplification (95°C for 15 s and 60°C for 1 min). Expression levels were normalized to those of glyceraldehyde 3-phosphate dehydrogenase (Gapdh). The ΔΔCt method was used to compare data. The PCR primers are shown in Table S3.

Galectin-3 knockout mice were generated using CRISPR/Cas9 genome-editing technology. A single-stranded CRISPR RNAs (crRNAs) were designed with sequences GGCACAGGAGCAAACGCCGGAGG and GAACAGGAAAGAACGACGUGAGG for targeting the galectin-3 gene. Cas9 protein (Thermo Fisher Scientific), crRNAs and trans-activating CRISPR RNA (tracrRNA) were incorporated into the cytoplasm of the zygote at the pronuclei stage by electroporation. The injected zygotes were then transferred to pseudopregnant mice. To examine genotypes of pups, PCR analysis was performed with primers (knockout primers: forward 5′-GTTGAAAAGTGGGGCAGTTC-3′ and reverse 5′-GACCAGCCTCCACAACAGAT-3′; wild-type primers: forward 5′-GCCAACACTTGAGATGCTGA-3′ and reverse 5′-GCTTCTGGGACATGAGAGGA-3′) using MightyAmp DNA Polymerase Ver. 2 (Takara Bio).

Tail tissue samples were homogenized with Micro Smash (TOMY) in RIPA buffer (Nacalai Tesque) containing protease inhibitor cocktail (Nacalai Tesque) and phosphatase inhibitor cocktail (Nacalai Tesque). Tail lysate extractions were incubated at 4°C for 20 min and then centrifuged at 4°C for 10 min at 800 g. The protein concentration of the supernatant was determined using the Bio-Rad DC protein assay kit. Protein was loaded in 10% polyacrylamide gel and transferred electrophoretically to polyvinyl difluoride membrane, and blocked in 1% skim milk in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) for 5 min at room temperature. Membranes were incubated with primary antibodies (anti-galectin-3; 1:500, Santa Cruz Biotechnology, sc-32790) at 4°C overnight in a rotor. Membranes were washed with TBS-T and incubated with horseradish peroxidase-conjugated secondary antibodies (1:2000, GE Healthcare, NA931). The images were captured using the LAS-4000 detector (Fujifilm).

Statistical significance was evaluated using GraphPad Prism 8 (GraphPad Software). Values were expressed as mean±s.d. Statistical significance was assessed by performing two-tailed Student's t-test with or without Grubb's test, Mann–Whitney U-test or Kruskal–Wallis with Dunn's multiple comparisons test. The association between ND and HFD groups with respect to the number of senescent cells was assessed by a χ² test of independence. A probability of <5% (P<0.05) was considered statistically significant.

We thank Dr Madoka Uezumi for instruction on the techniques for the collection and staining of PDGFRα⁺ cells; Dr So-ichiro Fukada for the gift of the anti-SMC2.6 antibody; the NGS Core Facility of the Genome Research Center at the Research Institute for Microbial Diseases of Osaka University for their support in conducting scRNA-Seq; Ms Kayako Tosa, Ms Hideka Miyagawa and the laboratory members in the Department of Pathophysiology of Osaka City University Graduate School of Medicine for technical support and for providing critical feedback and experimental support; and the Research Support Platform of Osaka City University Graduate School of Medicine for technical support.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199443.