Intramuscular adipose tissue: from progenitor to pathology

Abstract

The accumulation of intramuscular adipose tissue (IMAT) is a nearly ubiquitous feature of skeletal muscle pathology, strongly correlating with impaired contractility and metabolic dysfunction across a wide spectrum of clinical conditions, from aging and obesity to genetic myopathies and orthopaedic injuries. For decades, a critical question has persisted: is IMAT a passive biomarker of disease progression or an active pathogenic agent? This review synthesizes emerging evidence to address this question by exploring several key areas. We first evaluate the mechanisms by which IMAT impairs muscle function, examining evidence for its dual role as both a physical disruptor and a local source of unbalanced paracrine signals. By integrating findings from human studies with insights from diverse animal models, we also highlight significant translational challenges, particularly the resistance of common rodent models to developing human-like IMAT pathology. Furthermore, we review the cellular origin of IMAT to resident fibro/adipogenic progenitors (FAPs), a highly plastic cell population that supports regeneration in healthy muscle but can differentiate into adipocytes under pathological conditions. We then dissect the complex signaling network that governs this fate switch—specifically the balance between pro-adipogenic ‘triggers’ and inhibitory ‘brakes’ that becomes dysregulated in disease. The evidence increasingly points to IMAT as an active contributor to muscle decline. Therefore, future progress requires a multi-pronged approach: the continued elucidation of the specific molecular ‘brakes’ and ‘triggers’ that govern FAP fate, the development of more translationally relevant preclinical models, and the standardization of IMAT quantification methods to improve diagnostic accuracy and clinical trial endpoints.

Graphical Abstract

Intramuscular adipose tissue (IMAT) expands across a range of conditions with varied etiologies – from chronic aging to acute injury. The signals that control IMAT expansion are still being uncovered, but several adipogenic brakes and triggers have been identified. The differentiation of fibro/adipogenic progenitors (FAPs) into adipocytes is thought to be driven by both loss of adipogenic brakes and gain of adipogenic triggers. Once present, IMAT disrupts both physical architecture and paracrine signaling in muscle, which is thought to contribute to impaired regeneration, insulin resistance and functional deficits across populations. This figure was created using a licensed version of BioRender.com.

Introduction

Adipose tissue is distributed throughout the body in anatomically discrete regions, or depots. Some are large and continuous, like subcutaneous fat, and some are small and isolated, like the pericardial fat that surrounds the heart. In the musculoskeletal system, adipose is a distributed network of small depots, owing to the functional connectivity of the 800+ muscles, bones and joints that each have associated, discrete adipose tissues. Adipose tissue is frequently found surrounding tissues and organs, but only in bone and skeletal muscle is it found within the anatomical boundary of healthy tissue. While these adipose depots have been observed for centuries in livestock, their physiological function remains largely unknown. In muscle, intramuscular adipose tissue (IMAT) has garnered much more interest for its role in pathology than physiology.

The expansion of IMAT, which is the buildup of adipocytes between individual myofibers within skeletal muscle, has been recognized as a feature of muscle disease since the 1940s, when it was discovered that the large, but weak, calf muscles of boys with Duchenne Muscular Dystrophy (DMD) contained excessive amounts of adipose tissue(1). As it was associated with necrosis of muscle fibers, investigators coined this phenomenon “fatty infiltration”(1, 2) or “fatty replacement,”(3, 4) where necrotic contractile fibers were increasingly filled in or replaced by an adipocyte-laden connective tissue rather than a regenerated fiber. Shortly after its identification on muscle biopsy, “fatty replacement” in DMD was more extensively characterized by the non-invasive imaging modalities of CT and MRI(3, 5). These studies were the first to note that the expansion of IMAT was heterogeneous, with some muscles more affected than others, and correlated with functional outcomes such as strength and mobility(3).

In the early 1990s, expansion of IMAT was recognized as a feature of aging, obesity and metabolic disease and extensively measured in these populations over the next decades (reviewed in detail in(6)). Elegant and comprehensive studies again tied IMAT expansion to reduced strength and physical function(7–12), and many found this relationship to be independent from reductions in lean mass(7, 9, 13). Furthermore, IMAT expansion was also strongly correlated with insulin resistance in many studies (reviewed in detail in(14)). As the endocrine function of adipose tissue and its role in the pathogenesis of type 2 diabetes was emerging, these findings led to the hypothesis that, rather than simply replacing contractile material, IMAT secreted inflammatory adipokines which acted on neighboring muscle fibers to actively impair metabolic and contractile function.

In parallel, IMAT was extensively characterized in the muscles of the rotator cuff as a pathological feature of chronic rotator cuff tears. Here again, “fatty infiltration” was strongly negatively correlated with muscle strength and functional recovery(15–18). In fact, this association was so strong that a grading system to assess IMAT was developed for CT scans(19), and later MRI(20), to help inform surgical decision making. This was followed by work in DMD using IMAT measures to predict functional decline and loss of mobility in patients(21). Now, as the tools to assess IMAT have become more accessible and widespread, IMAT expansion is understood to be a nearly ubiquitous feature of muscle injury and disease, where it is nearly universally associated with an adverse clinical outcome.

Despite this extensive characterization across diverse clinical populations, some major outstanding questions remain. First, why does IMAT form? Is it the pressure of an adipogenic driver(s), shared across these varied conditions, or does it result from the loss of an inhibitory brake on the fibro/adipogenic progenitor (FAP) which gives rise to IMAT? Second, how does IMAT affect muscle function? Is there dysregulated cytokine cross-talk or physical disruption or both? And finally, can IMAT be therapeutically targeted to reduce or restructure its negative influence? These three questions are the crux of this review. We will describe existing findings in humans and animal models and discuss leveraging new tools to test mechanistic hypotheses.

Definitions and measurement tools

IMAT by many names

Lack of consistent definitions for adipose tissue in muscle has challenged the field. In this review, we are using the term intramuscular adipose tissue (IMAT) to refer to adipocytes within the muscle fascia. As noted above, these are also variably referred to by other names such as fatty replacement, fatty degeneration, muscle fatty infiltration (MFI), intramuscular fat (IMF), intermuscular adipose tissue (also frequently called IMAT) and myosteatosis. We have chosen intramuscular adipose tissue for three reasons: 1) a closer alignment with the latin intramuscular meaning “within muscle,” 2) distinction from the adipose tissue outside the muscle fascia which is also frequently termed intermuscular adipose tissue as it resides between muscles and 3) distinction from intramyocellular lipid which is frequently referred to as a muscular fat but is not an adipose tissue. Most studies agree that IMAT is the salient measure – that the adipocytes within the muscle fascia are the ones likely tied to metabolic or contractile dysfunction. However, it is common to see IMAT defined as “any lipid in the muscle compartment” because it can be difficult to distinguish between anatomical lipid locations (i.e. the lipid droplets of the adipocytes that make up IMAT vs the intramyocellular lipid droplets within the myofibers) using non-invasive measurement tools in people, especially in advanced pathology. While impactful data can be (and has been) gained without anatomical divisions, it is clear that the three lipid deposits in muscle – intermuscular adipose tissue, IMAT and intramyocellular lipid - are functionally and anatomically distinct. Therefore, developing ways to measure them separately is likely to provide additional insight into pathology.

Measurement of IMAT in people

The only cellular level measurements of IMAT in human muscle are by biopsy. Adipocytes appear as vacant space on most stained sections of biopsy (Fig 1A.a–f) and can be confirmed as adipocytes with a lipid soluble dye such as Oil Red O or BODIPY or by immunostaining for the lipid coat protein PERILIPIN. Measurements of adipocyte area or the adipose area fraction of the BODIPY can be made from these images. However, a major limitation of such dyes is that lipid droplets are labile and easily displaced during tissue processing (Fig 2). In fact, in our hands, only 20–35% of BODIPY⁺ lipid droplets perfectly overlap with PERILIPIN⁺ adipocytes, with the remainder displaced to other parts of the section (Fig 2C–D). This resulted in a stark difference in the area occupied by BODIPY (~5%) vs PERILIPIN (~10–15%). This variability, which also depends on fixation and handling, can lead to inaccurate quantification of intermuscular adipose tissue (IMAT) and the misidentification of intramyocellular lipids and so care should be taken when analyzing/interpreting these images. Biopsy is also a powerful tool to study tissue morphology at the cellular level, differentiating intra- and peri-fascicular IMAT(22), measuring proximity patterns between features such as adipocytes and necrotic fibers(23) and other relevant cells such as macrophages. Emerging tools such as spatial transcriptomics and proteomics will make biopsy even more informative, especially for investigating cellular cross-talk. However, muscle biopsy will always have two major drawbacks: 1) it is an exceedingly small sample of a likely heterogeneous pathology and 2) the associated pain makes it challenging to acquire. IMAT is most commonly quantified by CT or MRI. Both of these methodologies are non-invasive, include anatomical visualization, have good discernment between muscle and fat and are typically accessible. Additionally, they can image any muscle of the body, including those inaccessible to biopsy, through its full extent or cross-section. For both modalities, quantification of IMAT typically relies on setting a threshold of signal intensity that differentiates muscle and fat and then subdividing the composition of a region of interest. For some studies, the region of interest is a specific muscle and the “fat” signal reflects IMAT and intramyocellular lipid(24), for other studies the region of interest is everything deep to the subcutaneous fat and the “fat” signal reflects intermuscular adipose as well. As the raw pixel intensities are continuous, rather than binary, where the division is set can influence the result, and is not strictly consistent between studies (reviewed in detail in(25)). The continuous nature of the image is partially due to the resolution limits of the scanners, which capture on the order of 200 fibers and adipocytes per voxel. Thus, the cellular composition of a voxel will decide how it is classified. Other non-invasive methodologies are sporadically used, notably magnetic resonance spectroscopy (MRS) which can distinguish IMAT from intramyocellular lipid, but lacks anatomical information (26) and ultrasound which offers a cost-effective and portable way to estimate IMAT, but at low resolution(27, 28).

Figure 1. IMAT comparison across species and conditions.

A). Histological images of muscle cross-sections from six human conditions commonly featuring IMAT (top) and the matched mouse model (bottom). Image detail: a-c) representative human biopsies reproduced with permission from neuromuscular.wustl.edu, d) gastrocnemius biopsy from an individual with diabetic peripheral neuropathy (DPN) from (151), e) supraspinatus biopsy from and individual with chronic rotator cuff tear (RCT) from (152), f) vastus lateralis biopsy from an individual with obesity from (153), g) tibialis anterior muscle from a wildtype mouse, h) gastrocnemius muscle from the D2.mdx model of DMD, reproduced from (154) with permission under the Creative Commons Attribution 4.0 international license, i) quadriceps muscle from the dysferlin-null (Dysf-KO) model of LGMD2B, reproduced from (155) with permission under the Creative Commons Attribution 4.0 international license, j) gastrocnemius muscle from a streptozotocin (STZ) treated model of DPN from (156), k) supraspinatus muscle from the rotator cuff (RC) tenotomy model of RCT from (157), l) quadriceps muscle from high fat diet (HFD) fed model of obesity from (158). B). Top: Schematic depicting common mouse models used to study IMAT development. Bottom: Representative immunofluorescent images of mouse tibialis anterior muscle cross-sections for each model, ranging from low adipogenic (mdx, denervation), to moderate (cardiotoxin (CTX)) and high adipogenic (glycerol (GLY)) injury. Nuclei are stained with DAPI (magenta), adipocytes with PERILIPIN (yellow), and muscle fibers with PHALLOIDIN (gray). C) Quantification of IMAT percentage by area for each condition. Statistical analysis: N=4–17 per group (mixed sex), Welch’s ANOVA with Dunnett’s T3 multiple comparisons test. **** p<0.0001, ** p<0.005 compared to healthy 2 month old mice (2m). All scale bars: 100 μM. Figure created in part using a licensed version of BioRender.com.

Figure 2.

A) Oil Red O staining of unfixed, snap frozen 10–12 week old 129s female mouse tibialis anterior muscle sections showing displacement of lipid droplets (denoted by white arrowhead) from adipocytes (denoted by black asterisk). B) Schematic depicting protocol used for BODIPY (1:10⁶; MedChem Express, HY-W090090), PERILIPIN (1:1000, Cell Signaling #9349S) and DAPI (Invitrogen #D1306) immunofluorescent staining. Tibialis anterior muscle was harvested from 10–12 week old 129S mice 21 days post GLY injury. The tissue was either directly fixed in 4% paraformaldehyde (PFA) (male, n=3) or snap-frozen (female, n=3) before being embedded in OCT for sectioning. Snap-frozen sections were post-fixed in 4% PFA for 10 minutes prior to staining, whereas fixed tissue was directly stained. After staining, both groups underwent two mounting techniques. “Good mounting” refers to careful placement of the coverslip without lateral pressure or movement, whereas “bad mounting” involves lateral shifting or pressing of the coverslip after initial placement onto the slide. C) Representative immunofluorescent images of stained sections under different fixation and mounting conditions as outlined in B. Lipid droplets are stained with BODIPY (magenta), adipocytes with PERILIPIN (gray), and nuclei with DAPI (cyan). Merged images are shown on top with respective PERILPIN and BODIPY channels split and shown below. D) Quantification of BODIPY ⁺/PERILIPIN ⁺ double positive cells expressed as a percentage of total PERILIPIN⁺ cells (left). Quantification of area occupied by PERILIPIN (center) or BODIPY (right) expressed as a percentage of total area. Statistical analysis: N=3 per group, 2-way ANOVA with Bonferroni’s multiple comparisons test. *p<0.05. All scale bars: 100 μM. Figure created in part using a licensed version of BioRender.com.

Measurement of IMAT in animal models

Animal models offer the opportunity to comprehensively study IMAT mechanistically at the cellular level throughout the muscle volume, avoiding the sampling and resolution problems of human measurements. While IMAT can be measured non-invasively in small animal scanners, most studies opt for histological assessment or decellularization based quantification(29). Histological assessment has all the benefits described for human biopsy, but is typically only done on a single cross-section therefore could be subject to sampling bias. This bias can be, at least partially, addressed by analyzing serial sections at different depths throughout the muscle or by splitting the muscle for parallel histological and molecular (RNA/protein) analyses(30). Decellularization based quantification allows 3D mapping of IMAT, but since the remainder of the muscle is cleared, it becomes useless for any further analysis. Larger animal model assessment more frequently involves paired MRI/CT and histopathology(31, 32).

Experimental models to study IMAT

The presence of IMAT in rodent muscle is extremely limited in homeostasis. More importantly, mice and rats are resistant to pathological IMAT expansion in most models of human disease, including aging, diabetes and DMD(33–36) (Fig 1A), and as a consequence, few studies in rodents have exclusively focused on IMAT. The exceptions are studies using intramuscular injection of glycerol (GLY), which induces regeneration with substantial IMAT formation (30, 37–41) (Fig 1B). Below we review the strengths and limitations of these models.

Injury models

The most extensive study of IMAT in animals has been in models of rotator cuff injury, with more than 50 pre-clinical studies characterizing IMAT expansion across mice, rats, rabbits and sheep(42–44). These models typically involve transection of one or more tendons of the rotator cuff muscles with or without transection of the suprascapular nerve. IMAT expands more dramatically when nerve transection is included in the model(43, 45), which aligns with the effect of denervation on IMAT expansion discussed below. However, it should be noted that nerve involvement in human chronic rotator cuff tears is not highly prevalent(46). Thus, it is possible that mechanisms uncovered from the tenotomy+denervation model would not be broadly translational(45). In rodents, tenotomy alone causes IMAT expansion to only ~5% of the muscle volume(45, 47) (Fig 1A.k). Whereas in rabbits and sheep, IMAT may occupy as much as 50% of the muscle following chronic tenotomy which more closely aligns with chronic rotator cuff tears in people(48, 49). Denervation alone can also induce IMAT expansion, which has been well characterized in the rotator cuff with suprascapular nerve transection (45, 49) and in the hindlimb by transection of the sciatic nerve(36, 50). Similar to tenotomy, while mice and rats form limited amounts of IMAT (36, 50) (Fig 1B–C), rabbits display large conversion of muscle into adipose tissue(51). Thus, the choice of animal model significantly impacts the study of IMAT. Rodent models, while offering powerful genetic tools, may underestimate the extent of fatty infiltration that occurs in larger mammals, including humans.

Aging and metabolic dysregulation

Similar to people, mice and rats exhibit expansion of IMAT with advanced age(34, 52). Age-related IMAT expansion in rodent models is worsened when combined with high-fat diet (HFD) induced metabolic dysregulation – a model for sarcopenic obesity(53, 54). HFD alone can also induce IMAT expansion(55, 56)(Fig 1A.l), but it only reaches ~5% of the muscle volume even after months of HFD(53, 55). These observations are likely true for larger animals as well, but investigations in these animals have focused almost exclusively on the determinants of meat quality.

Genetic models

A major benefit of studying disease in mice is the capacity for genetic manipulation. Many of the primary human dystrophies and myopathies that feature IMAT accumulation have an established mouse model, but few studies have investigated IMAT expansion in these models with the notable exception of DMD and limb-girdle muscular dystrophy 2b. Several mouse models for DMD exist. The most common (mdx) carries a spontaneous point mutation in the dystrophin gene, the same gene that is disrupted in DMD patients(57). While mdx mice exhibit muscle fiber necrosis and elevated creatine kinase levels, they generally have a milder disease course compared to human DMD(33). Another significant limitation of the mdx model, is the minimal development of IMAT, which only reaches ~2% of the muscle volume(58, 59) (Fig 1A.h, 1B). This lack of progressive IMAT replacement in the mdx mouse is an issue with the newer DMD models and has been recognized as a major hurdle in the field as it precludes the use of this model to study the long-term impact IMAT might have on disease progression. Attempts have been made to force IMAT formation in mdx mice such as by combining mdx mice with a knockout of apolipoprotein E (ApoE) mutant allele coupled with HFD, with the objective of “humanizing” the mice with hyperlipidemia(58, 60). Hyperlipemic mdx ApoE^−/− mice displayed significantly more IMAT formation than naïve mdx mice, occupying up to ~20% of the muscle area. A similar strategy successfully increased IMAT formation in a mouse model of limb-girdle muscular dystrophy type 2B, dysferlin-null (Dysf^−/−) mice(61), where IMAT reached ~40% of the muscle area. Other transgenic models of myopathy similarly develop little IMAT naturally, despite mimicking many other features of human muscle pathology(62, 63). On the other hand, high-throughput screening of mouse knockout models has discovered that knockout of Asparagine Synthetase Domain Containing 1 (Asnsd1), a gene without a known role in human muscle pathology, induces widespread IMAT expansion, replacing more than 50% of the gastrocnemius muscle(64). While the lack of IMAT expansion in mice has historically hindered our understanding of its role in human disease, these less conventional approaches to “force the effect” may turn out to be surprisingly informative. These less translationally-relevant conditions that are required to get substantial IMAT expansion in mice may offer insight into why mice are resistant to IMAT and offer clues as to how we can increase resistance in people.

Intramuscular GLY injection

As noted above, the gold-standard model for mechanistic studies of IMAT development is intramuscular GLY injection. This is because GLY 1) induces much more IMAT expansion than typical models of injury/disease (Fig 1B–C), 2) shortens the time course of expansion from months to a few weeks and 3) can be easily overlayed on any model, from wildtype to complex transgenics. The primary limitation of the GLY model is that it is an acute injury model without a translational equivalent. Consequently, the systemic effects of IMAT and the broader disease context might differ significantly in chronic human diseases where ongoing metabolic dysregulation, systemic inflammation, or multi-organ involvement play a larger role. However, the GLY model is particularly relevant for studying human conditions characterized by significant fatty infiltration following acute muscle injury and for mechanistic investigations looking to decouple the local effects of IMAT from systemic disease. The GLY model also remains valuable for studying the basic cellular and molecular mechanisms that allow FAPs (the IMAT progenitor) to adopt an adipogenic fate.

Mechanistic insights into the pathological function of IMAT

While the accumulation of IMAT strongly correlates with muscle dysfunction across human disease, its exact mechanistic impact remains unresolved. It has been long debated whether this association is correlative or causative – whether IMAT contributes to muscle dysfunction or is just an “innocent bystander.” Emerging evidence from our laboratories and others is beginning to establish its role as an active agent in muscle pathology(65–67).

IMAT as a physical disruptor of muscle architecture

In humans, quantities of IMAT are highly negatively correlated with muscle strength and physical function. These studies span aging(7), obesity(68), muscular dystrophy(69), chronic rotator cuff tears(15), diabetes(8), stroke(10), rheumatoid arthritis(70), chronic obstructive pulmonary disease(71), cancer(72), chronic kidney disease (73) and healthy young people with muscle unloading(13) in a non-exhaustive list. As these conditions frequently also feature muscle atrophy and degeneration, it has been proposed that replacement of contractile material by IMAT simply modifies the composition of muscle, reducing the amount of material within the cross-section capable of contraction. However, the relationship between IMAT and reduced muscle strength is frequently independent of loss of lean mass(7, 9, 13), suggesting a direct effect of IMAT on muscle contraction. In 2020, the Meyer lab, we demonstrated that this same strong negative correlation exists in mice following intramuscular GLY injection, and it cannot be explained by reduction in lean mass(65). Furthermore, we showed that a genetic block on adipogenesis, through adiponectin-driven diphtheria toxin expression, completely eliminated the contractile force deficit following GLY injection. The Kopinke lab took a parallel approach to preventing IMAT development by inducible genetic deletion of Pparγ, the master regulator of adipogenesis, specifically in FAPs(74). Following a GLY injury in this mouse model, termed mFATBLOCK, we found that preventing IMAT formation led to a dramatic enhancement in the functional recovery and greater force production of the muscle. Surprisingly, this improved recovery appeared to be independent of post-regenerative, muscle stem cell-driven myogenesis or major shifts in intercellular signaling. Instead, our data strongly argue that IMAT acts as a physical barrier with a two-phase impact: initially, it hinders the formation of nascent myofibers early in regeneration, and later, it restricts their subsequent post-regenerative hypertrophic growth. This structural mechanism also provides a compelling explanation for the findings by the Meyer lab, where IMAT-laden whole muscle is weak, but isolated single fibers are not(65). Intrafasicular IMAT is typically found as chains of adipocytes running along the length of myofibers, and this would be expected to disrupt lateral transmission of force between myofibers which is important for efficient contraction (88). Physical disruption of neuromuscular connectivity (95) or blood flow could also contribute to both reduced contractility and regenerative capacity. Since the association between IMAT and muscle contractile/regenerative deficits exists in the mouse GLY model, these hypotheses are ripe for mechanistic exploration.

IMAT as a potential paracrine signaling center

Beyond its physical effects, IMAT has long been hypothesized to act as a local, pathological signaling hub. This idea is strongly supported by human studies where the link between IMAT quantities and muscle and whole-body insulin resistance in people is equal or greater in strength than the link between IMAT and muscle weakness (reviewed in detail in(14)). Excess IMAT correlates with peripheral insulin sensitivity, glucose tolerance, fasting glucose, fasting insulin and HOMA-IR (selected(75–78)), not only across populations with metabolic dysfunction but also in healthy older individuals(78), individuals post-stroke(10), individuals with spinal cord injury(79), and those with HIV(80). The close proximity of IMAT adipocytes to myofibers and the known role of adipokines in development of muscle insulin resistance have led to dysregulated IMAT-myofiber paracrine signaling as the most prominent mechanistic hypothesis to explain this association(81). In vitro data strongly support this hypothesis. To date, only two studies have directly tested whether IMAT secreted factors impair insulin signaling(66, 82). Laurens et al. found that exposure of cultured primary myotubes to media conditioned by adipogenically differentiated progenitors isolated from skeletal muscle of obese participants impaired insulin signaling(82). Similarly, Sachs et al. found that exposure of primary myotubes to IMAT explant conditioned media decreased insulin sensitivity to a similar degree as visceral adipose conditioned media(66). Since prior work has implicated pro-inflammatory adipokines secreted by visceral adipose tissue in the development of muscle insulin resistance(83–85), these studies support a similar, but local, role for IMAT. Indeed, the composition of IMAT explant conditioned media is more similar to visceral than subcutaneous fat in people and includes several pro-inflammatory adipokines including TNFα, IL-6 and MCP-1(67). Beyond adipokines, IMAT could modify the local myofiber environment by secretion of free fatty acids and other bioactive lipids, which can also contribute to the development of insulin resistance in muscle(86). A recent study found that rates of IMAT lipolysis were similar to visceral adipose and much higher than subcutaneous(66). While intriguing, more experiments are needed to identify a definitive cause-and-effect relationship for any of these potential signals.

To date, mechanistic studies of the relationship between IMAT and myofiber insulin resistance have been limited to culture models and it is unknown whether development of IMAT in absence of metabolic disease (e.g. via GLY injury) or the genetic inhibition of IMAT in metabolic disease (e.g. via mFATBLOCK) modulate muscle insulin signaling in vivo. If so, these models open more avenues for mechanistic exploration. Future models enabling targeted manipulation of IMAT (e.g. to modify paracrine signaling or modulate adipocyte size) will shed light on whether IMAT impacts muscle predominantly by physical disruption, biochemical signaling or both. Our recent work suggests that the role of IMAT may be context dependent as preventing IMAT formation after GLY injury rescued muscle function without causing detectable changes in the secretome of key cytokines and myokines or affecting myogenesis via secreted factors in vitro(87). Thus, it is possible that a paracrine signaling role of IMAT is most prominent in chronic diseases, particularly metabolic diseases, where IMAT accumulates progressively under the biological stressors that drive pathological adipose signaling generally, while physical disruption plays a larger role in conditions of musculoskeletal injury.

The IMAT threshold: How much IMAT is needed to see an impact?

Multiple studies, including from both of our laboratories, have reported a strong negative impact of IMAT on muscle function, which raises the question: how much IMAT is needed to see a pathological impact? Our work suggests the existence of an “IMAT threshold” needed for the muscle to be negatively affected. As established in the GLY injury model, functional deficits only become apparent once IMAT accumulation crosses a critical point, which we determined to be ≥12% of the muscle area(74). This finding has profound implications because this threshold is far more readily reached in human pathologies than in most preclinical mouse models. First, this discrepancy has caused the field to underestimate IMAT’s role as an active contributor to pathology, as its detrimental impact is not observed in models that remain below this pathogenic threshold. Second, it means that therapies are often tested in an environment that fails to replicate the detrimental impact of IMAT, potentially generating misleading results. The existence of a threshold also raises the question of whether different pathological effects require different amounts/distributions of IMAT. It is plausible that the impairment of muscle contraction may be sensitive to the total volume and distribution of IMAT, which disrupts lateral force transmission across the entire muscle. In contrast, the impairment of regeneration might be more dependent on the local density of adipocytes during the critical early phase of repair, where they physically compete for space with nascent myofibers. The paracrine effect of IMAT may be heterogeneous in muscle, where a relatively small but dense cluster of adipocytes could be sufficient to disrupt local signaling. This highlights the need for future studies to define these context-specific thresholds to better inform therapeutic strategies.

The cellular origin of IMAT

A brief history of FAPs

The search for the IMAT progenitor cell has been ongoing since the late 1990s. Efforts originally focused on a heterogeneous group called “side population” cells(88) and the adipogenic conversion of satellite cells(89) and then expanded to identify adipogenic potential of pericytes(90), PW1-expressing cells (PICs)(91) and FAPs(92, 93). The primary cellular source of IMAT has now been established to be FAPs expressing Platelet-Derived Growth Factor Receptor Alpha (PDGFRα), the “gold-standard” marker in the field. For example, our studies, which used genetic lineage tracing and functional adipogenic disruption utilizing a PDGFRa-CreERT allele to target FAPs, indicate that these cells are the primary origin of IMAT. Importantly, some other populations, including “side population” cells, PICs and fibroblasts share FAP surface marker expression and are thought to include the population now defined as FAPs(94–96).

In healthy muscle and during the acute phase following muscle injury, FAPs play an indispensable supportive role via inter-cellular signaling (reviewed in(97)). Conversely, under conditions of chronic injury, such as those observed in muscular dystrophies, during aging, or in response to specific types of acute insults like GLY-induced muscle damage, FAPs can adopt diverse pathological fates. In these scenarios, FAPs can differentiate into adipocytes, contributing to the formation of IMAT, or into myofibroblasts, leading to fibrosis(39, 92, 93, 98). This dynamic nature of FAPs—capable of switching between beneficial and detrimental roles—allows them to sense changes in the muscle microenvironment and actively shape tissue outcomes. Consequently, therapeutic interventions absolutely require a detailed and refined understanding of FAP regulation in order to preserve their beneficial functions while mitigating their pathological contributions(99).

FAP heterogeneity

FAPs exhibit considerable heterogeneity in both healthy and diseased muscle, and in both mice and humans. Single-cell and single-nuclei transcriptomic (sc and snRNAseq) studies have identified FAP subpopulations with distinct gene expression signatures indicative of progenitor states, adipogenic commitment, or fibrogenic commitment(100–106). The existence of these distinct FAP subpopulations (pro-regenerative vs adipogenic or fibrogenic prone) suggests that FAP heterogeneity is a key determinant of muscle injury outcome. One may speculate whether adipogenic conditions selectively promote the expansion and differentiation of adipogenic FAP subpopulations or promote adipogenesis of other subpopulations. Dysregulated FAP signaling could also end up in a positive feedback loop where aberrant signaling also suppresses/overrides the protective, pro-myogenic functions of FAP subpopulations which is expected to worsen fibro/adipogenic pathology and subsequently further impair regeneration. Future research employing single-cell/nuclei technologies combined with maintaining spatial information will be essential to further characterize the diverse FAP subpopulations present in both the healthy and diseased muscle. Identifying specific markers and understanding the unique signaling pathways that define and regulate pro-regenerative versus pro-degenerative FAP phenotypes are key research priorities. This knowledge could pave the way for therapies that selectively “reprogram”, rather than ablate, FAPs - boosting beneficial activities while suppressing detrimental ones such as their differentiation into IMAT.

Regulation of FAP adipogenesis

The adipogenic fate of FAPs is controlled by a network of extracellular ligand-receptor pathways that either promote adipogenic differentiation (acting as “triggers”) or inhibit it (acting as “brakes”). In healthy muscle, multiple redundant brakes restrain FAP adipogenesis. Conversely, aging and disease shift this balance: adipogenic triggers are upregulated or acquired, while protective brakes fail or are lost. Many excellent reviews have recently been published discussing the plethora of pathways that control the fate and function of FAPs(107–109), so here we will focus specifically on the key upstream signaling pathways at the receptor-ligand level that might represent a “brake” or “trigger” for FAP adipogenesis.

Before dissecting the complex network of upstream signals, it is important to mention that many intracellular signaling pathways (such as cAMP/PKA, PI3K-AKT, and JAK/STAT3) have been shown to modulate FAP adipogenesis and all roads to adipogenesis appear to go through PPARγ (Peroxisome Proliferator-Activated Receptor gamma). PPARγ is the central downstream effector that executes the adipogenic program within all adipose depots including in FAPs as we have recently demonstrated(87). Because these downstream pathways have been well established across adipogenic progenitors(110) and can be activated by various upstream signals in a tissue-specific manner, this review focuses on the upstream signals that ultimately determine whether the expression and activity of PPARγ is permitted or repressed.

FGF signaling: An age-associated trigger

Fibroblast Growth Factor (FGF) signaling, particularly through FGF2, has emerged as a potent pro-adipogenic “trigger” that becomes pathologically activated in the context of aging(111). In young muscle, FGF2 is transiently expressed and supports muscle stem cells(112), but with age muscle fibers begin secreting excess FGF2, creating an adipogenic-prone niche(111). AAV-mediated FGF2 overexpression in mouse muscle was sufficient to induce IMAT formation(111). Mechanistically, FGF2 downregulated SPARC, an extracellular matrix protein that normally inhibits adipogenesis, through a MAPK–miR-29a pathway. The reduction of SPARC “released a brake,” permitting the induction of PPARγ and allowing FAPs to differentiate into adipocytes. Interestingly, elevated FGF signaling also disrupts the muscle stem cell niche(112), suggesting a dual pathogenic role where a single signal—chronically elevated FGF2 from the aged muscle fiber itself—could drive both IMAT formation and a decline in regenerative capacity. While the evidence that FGF2 is sufficient to drive IMAT formation is strong, FAP- specific loss of function experiments are still missing. Future experiments need to be explored to determine whether this adipogenic trigger is specific to age-related IMAT.

Annexin A2: A disease-specific trigger

Certain disease-specific signals can also function as adipogenic triggers. One striking example occurs in dysferlinopathy (Limb-Girdle MD type 2B): chronic muscle membrane tears lead to extracellular release of Annexin A2, a protein that normally aids membrane repair(113). Extracellular Annexin A2 then directly stimulates neighboring FAPs to undergo adipogenesis. In dysferlin-deficient mice, deleting Annexin A2 markedly reduces IMAT, while administration of Annexin A2 protein greatly accelerates FAP adipogenesis(114). Although FAP-specific experiments are still lacking, Annexin A2 appears to represent a “damage-trigger” that is necessary and sufficient in driving IMAT formation, illustrating how unique disease contexts can create novel adipogenic ligands.

Hedgehog signaling: The master endogenous brake?

We and others have demonstrated that, in healthy muscle, the Hedgehog (Hh) signaling pathway functions as a powerful intrinsic brake on FAP adipogenesis(39, 41, 115). With minimal activity at rest, Hh signaling is potently activated upon injury to maintain FAPs in a non-adipogenic state. Schwann cells and other niche cells produce the ligand Desert Hedgehog (DHH) during regeneration(41), which FAPs sense via their primary cilia, small cellular antennas(39). Genetic or pharmacological inhibition of Hh activity in vitro and in vivo removes the brake and allows FAP to form fat. Conversely, genetic or pharmacological activation of Hh signaling potently inhibits PPARγ-driven adipogenesis of FAPs both after GLY injury and in mdx mice(39, 41, 115). Mechanistically, Hh signaling in FAPs induces the secreted inhibitor TIMP3, which blocks MMP14 and downstream adipogenic programs(39). Alternatively, or in addition, Hh signaling might repress adipogenesis via GDF10(116), which we have identified as a potent Hh-controlled FAP-derived pro-myogenic factor(41). Thus, Hh signaling is both necessary and sufficient in limiting IMAT formation by directly repressing FAP adipogenesis. An important question remaining is whether Hh is the endogenous brake that is lost in all conditions experiencing IMAT formation. Future experiments are also needed to define if and why Hh signaling is blunted or ineffective with age and in chronic diseases.

Retinoic acid signaling: A vitamin-derived brake

Retinoic acid (RA), the biologically active metabolite of vitamin A, is a necessary and sufficient pathway which suppresses the formation of IMAT when active. RA functions as a ligand for a family of nuclear receptors, the retinoic acid receptors (RARs) and retinoid X receptors (RXRs). Upon binding RA, this heterodimeric complex directly modulates the transcription of target genes, including PPARγ. Activating RA signaling via exogenous RA (117) or pharmacological RAR agonist treatment (118, 119) represses FAP-derived adipogenesis in vitro and vivo. In contrast, FAP-specific expression of a dominant negative RARα allows FAP adipogenesis and increased IMAT formation(117). While these results demonstrate that RA is sufficient to block FAP adipogenesis, future experiments should involve FAP-specific RAR knockout experiments especially as the dnRARα allele can cause non-canonical effects(120). In addition, it is important to establish whether RA signaling is diminished with age and/or disease thereby providing a physiological argument for RA to serve as a major adipogenic brake.

Wnt signaling: A brake on many roads

The Wnt signaling pathway is a potent and well-characterized inhibitor of adipogenesis in mesenchymal progenitors, including FAPs(109, 121). It functions as a critical brake through multiple, convergent mechanisms, including the canonical β-catenin pathway and non-canonical branches that are integrated with other key fate-determining pathways like Hippo(122). A comprehensive study by Reggio et al. identified the canonical Wnt5a–GSK3–β-catenin axis as a crucial pathway restraining FAP adipogenesis(123). Pharmacological inhibition of GSK-3 (which mimics Wnt activation by preventing β-catenin degradation) stabilized β-catenin, repressed PPARγ, and prevented FAP adipogenesis in vitro and vivo. Similar adipogenic repressive properties were found using a different pharmacological means to activate β-catenin-dependent transcription(124). In addition to the canonical axis, ligands like WNT7A can inhibit FAP adipogenesis through a β-catenin-independent, Rho-YAP-dependent mechanism(125). To date, no study has employed conditional, cell-type–specific ablation of individual canonical Wnt ligands (e.g., Wnt7a or Wnt5a) or of β-catenin within FAPs. Only circumstantial evidence exists(126). Thus, it remains to be determined whether Wnt, either through canonical or non-canonical means, is required to restrain the adipogenic conversion of FAPs and prevent IMAT deposition in vivo.

Notch signaling: How muscle blocks IMAT

The Notch pathway is a highly conserved system that mediates contact-dependent, or juxtacrine, communication. In muscle, it functions as a critical checkpoint that links the fate of FAPs to the regenerative state of the surrounding myogenic lineage, thereby acting as an adipogenic brake. Pharmacological blockade of Notch signaling using the γ-secretase inhibitors DAPT promotes FAP adipogenesis in vitro and in vivo following muscle injury(127). Thus, active Notch signaling is necessary to restrain FAP adipogenesis. Activating Notch signaling is also sufficient to block FAP adipogenesis in vitro, as demonstrated by co-culturing FAPs with myotubes or plating them on surfaces coated with a Notch ligand(127). A fascinating aspect of Notch signaling is its failure in chronic disease. FAPs from mdx mice become pathologically insensitive to the anti-adipogenic effects of Notch activation(127). Combined with the lack of FAP-specific gain-and loss-of-function models to confirm the pharmacological results, the importance of Notch in IMAT formation remains to be determined.

Inflammatory cytokines: A complex cocktail of potent anti-adipogenic cues

The immune system is an active regulator of FAP fate during regeneration. Cytokines released by infiltrating immune cells can function as powerful adipogenic brakes or triggers. The first of these to be identified was interleukin-4 (IL-4), a canonical type-2 cytokine, which was found to function as an important anti-adipogenic signal during acute muscle repair (128). This study found that following an acute injury, infiltrating eosinophil-secreted IL-4 is sensed by FAP and promotes FAP proliferation while simultaneously blocking PPARγ-dependent adipogenic differentiation. It also found that exogenous IL-4 was sufficient to block FAP adipogenesis, while FAPs lacking the receptor displayed enhanced adipogenesis post GLY injury. However, this model has recently been challenged, ranging from IL-4 promoting adipogenesis (129) to IL-4 signaling being completely dispensable for normal muscle regeneration and IMAT formation (130) and, thus, this mechanism awaits further investigation. The Cytokines IL-1α and IL-1β, produced by infiltrating immune cells (mainly macrophages) during muscle regeneration, act as potent inhibitors of mouse and human FAP adipogenic differentiation in vitro(129, 131). While in vivo evidence is currently lacking, this suggests a role for macrophages in controlling IMAT formation. The cytokine IL-15 is a potent mitogen released from myofibers that stimulates FAP proliferation via the activation of the JAK/STAT pathway. IL-15 also efficiently restricts FAP adipogenesis in vitro and in vivo. Interestingly, IL-15 also induces weak Hh activation suggesting that this cytokine might be upstream of DHH. Thus, a key question remains as to whether IL-15 inhibits FAP adipogenesis via preventing exit from the cell cycle, a prerequisite for efficient adipogenesis(132–134), or by inducing the Hh pathway(41).

Trigger vs brake: Where to go from here?

The accumulation of IMAT is likely a “dual-hit” phenomenon resulting from a progressive shift in the muscle’s signaling landscape. In this model, pathological fat formation occurs when the homeostatic “brakes” on FAP adipogenesis fail while pro-adipogenic “triggers” become active. The challenge before us is to now answer the following questions: (a) Are all “brakes” equally potent? Can a “trigger” override a “brake,” or is the reverse true? (b) Which specific “brake” offers the most promising therapeutic target? (c) Furthermore, is there a universal “master” control, or do different diseases require personalized approaches? (d) How can we therapeutically manipulate a “brake” or “trigger” without causing detrimental effects on other essential biological processes? Answering these questions is critical, as the most effective future therapies for IMAT will likely be combinatorial and context-specific. Such strategies will need to restore the function of key brakes while simultaneously blocking pathological triggers, with the ultimate goal of silencing the master adipogenic regulator, PPARγ.

Experimental Considerations

It is clear that more mechanistic insight into the development of IMAT and its influence on muscle is needed. As more experiments are designed and conducted, some lessons already learned may be valuable to highlight here.

Consideration of sex

Studies across experimental models consistently find higher IMAT accumulation in female mice compared with males. This includes the GLY injury model(135, 136), genetic models(137), and aging (138) and aligns with data in rats(139). However, this sex-specificity does not universally align with data collected in humans. Most studies find that IMAT is higher in the calf and thigh of healthy men compared with women(75, 140, 141), but higher in the rotator cuff of women compared with men(142–144). Among individuals with obesity and type 2 diabetes, thigh IMAT is typically similar between men and women(12, 145, 146), and its association with insulin resistance is not sex-specific(12).

Another thing to consider in both animal models and human studies is whether the variable of interest is a total volume of IMAT or a percentage of IMAT per muscle area or volume. Males tend to have larger muscles and could thus have a larger volume of IMAT, but a similar percentage when compared with females. Since measures in the human thigh and calf are typically total volume, while those in the rotator cuff are typically a percentage scale, this could explain some of the difference between these studies. Another thing to consider is hormonal state, as evidence from animal models suggests that sex hormones contribute to injury-associated IMAT in rats(139). Many human studies include women over 50, but not many differentiate by menopausal status.

Consideration of strain

The extent of GLY-induced IMAT is also influenced by the genetic background of the mouse strain, with the most commonly used background strain (C57BL/6J mice) being the most resistant to IMAT formation and 129/Sv and CD1 developing considerably more(29, 135, 147). Hybrid strains can make it easier to incorporate IMAT susceptibility in transgenic models on a BL/6J background(135). An additional consideration is that the phenotype of IMAT may differ between strains of mice(147). Our data and that of others suggests that C57BL/6J mice develop IMAT that resembles white adipose tissue and does not express UCP-1(147, 148). As most data suggest that adult human IMAT is a white phenotype(82, 149), further characterization of IMAT phenotype across strains will be valuable in confirming translational relevance.

Consideration of muscle of interest

IMAT expansion in rodents appears to be exacerbated in certain muscle groups with the diaphragm and rotator cuff muscles more susceptible than limb muscles(34, 47, 55). This is a particularly striking observation as other features of pathology like myofiber atrophy and fibrosis are similar between muscles and only IMAT differs(47, 51). Heterogeneity of IMAT accumulation is typical of human aging and disease(8, 146, 150), but is not consistent between conditions. This suggests local factors within the muscle microenvironment may modulate the systemic effects of aging and obesity and argues for assessing IMAT expansion outside of the gold-standard hindlimb muscles.

Considerations for FAP culture models

For cell culture experiments testing antagonists or inhibitors, the pathway of interest must be endogenously active - especially in a monoculture. For example, there is no evidence that any of the Hh ligand are present in the serum or produced by FAPs, thereby acting in an autocrine fashion. Thus, one cannot inhibit a silent pathway. Meaningful in vitro assessment of an antagonist requires that its target signaling cascade is either basally active or deliberately stimulated through experimental means to produce valid and interpretable results.

Conclusions and perspectives

IMAT is a nearly ubiquitous feature of human muscle pathology with a strong association to tissue dysfunction that has been documented and tracked for decades. With recent advances in sequencing, imaging and cellular and model organism engineering, we are beginning to understand the origin and nature of this association (summarized in Fig 3). The identification of FAPs as the cellular source of IMAT, has significantly advanced our understanding of their regulatory mechanisms and enabled manipulation of IMAT in vivo (e.g. mFATBLOCK). Notable advances over the past two decades include 1) characterization of sub-populations of FAPs with enhanced adipogenic potential and differential sensitivity to disease states, 2) identification of key adipogenic triggers and brakes originating from the muscle, the immune system and other FAPs and 3) demonstration that preventing IMAT formation improves muscle regeneration and function, the first confirmation of causation in vivo. This review was not intended to be comprehensive of all recent advances as there is now sufficient data to fill entire reviews on a single signaling pathway, and we encourage the reader to consult these detailed works(107–109). Instead, we highlighted the parallels in human and animal work, the challenges in translating work between species and the tools that are poised to enhance this effort. Some areas for integration include 1) consistent definitions and communication across disciplines, 2) matched cellular-level characterization of FAPs and IMAT in human and mouse muscle and 3) development of models/systems that better mimic human pathology. Advanced techniques such as single cell and spatial transcriptomics are beginning to be applied to human biopsies and, in parallel, genetic tools to manipulate FAPs/IMAT in mice are being developed. The key will be to integrate these strategies for a truly translational understanding of IMAT.

Figure 3. FAP-IMAT signaling schematic.

FAPs (pink), dispersed in healthy muscle (left circle), regulate homeostasis and response to injury by paracrine signaling to satellite cells, vasculature associated cells, Schwann cells and immune cells. With aging, genetic disease, chronic injury or metabolic dysfunction, this signaling becomes dysregulated. Adipogenic triggers become active and adipogenic brakes are repressed, initiating expression of Pparγ in FAPs and promoting IMAT formation (center circle). As dysregulated signaling progresses to pathology, IMAT (yellow) replaces healthy muscle tissue (right circle), acting as a physical disruptor of muscle architecture and source of detrimental paracrine signals. Concurrently, undifferentiated FAPs and myofibroblast-converted FAPs become more numerous also modifying the muscle’s physical and biochemical pathology. Figure created in part using a licensed version of BioRender.com.