Fibro-Adipogenic Remodeling of the Diaphragm in Obesity-Associated Respiratory Dysfunction

Obesity-associated respiratory complications range from simple dyspnea on exertion to life-threatening obesity hypoventilation syndrome (OHS) (1). OHS—defined by PaCO₂ >45 mmHg in individuals with BMI ≥30 kg/m² and no alternative cause of hypercapnia—impacts 2 million Americans (2) and exacerbates risks of heart failure and early mortality (3).

OHS pathophysiology is incompletely understood; however, restricted lung expansion in the setting of excess thoracic and visceral adipose tissue is assumed to be a major driver (4). Clinical studies demonstrate respiratory muscle weakness parallels this physical restriction (5,6), while autopsy samples from OHS patients contain prominent intradiaphragmatic adipocyte inclusions (7). These findings suggest that anatomic remodeling of the diaphragm itself may contribute to obesity-induced respiratory impairment.

Preclinical studies of respiratory function in obesity have largely employed genetically obese Zucker diabetic fatty (ZDF) rats. In this model, ex vivo diaphragmatic testing has yielded equivocal results, with contractile force reduced in older animals (8) but unchanged (9,10) or increased (11) in younger animals. ob/ob mice demonstrate hypoventilation reversible by leptin infusion, indicating that this adipokine may regulate central respiratory drive (12). Nonetheless, ventilatory function in diet-induced obese (DIO) models is unaffected by this maneuver (12), suggesting that exploration of other mechanisms, including intrinsic diaphragm compromise, is warranted to better understand the pathogenesis of obesity-induced respiratory dysfunction.

Accumulation of intramuscular adipose tissue (IMAT) is a complication of immobility (13) and muscular dystrophy (14). IMAT increases with normal aging (15) and quantitatively correlates with reduced muscle strength in the elderly (16). Recent studies demonstrate that IMAT expansion is also associated with weakness in individuals with obesity and type 2 diabetes (17–19).

Intramuscular fibrosis accompanies impaired regeneration, increased tissue stiffness, and reduced contractile force in skeletal muscle disorders (20). Intramuscular extracellular matrix (ECM) deposition is associated with insulin resistance in obese mice (21–23), while upregulation of skeletal muscle collagen gene expression occurs in humans after lipid infusion (24) and experimental overfeeding (25) and with chronic obesity (26). Despite links between overnutrition and ECM remodeling, direct effects of these fibrotic changes on muscle contraction remain largely undefined. Thrombospondin 1 (THBS1), a circulating ECM protein, activates transforming growth factor β (TGF-β) (27), promotes mesenchymal cell proliferation (28), and underlies fibrosis in limb muscles of obese mice (21).

Fibro-adipogenic progenitors (FAPs) are mesenchymal cells residing within skeletal muscle that give rise to adipocytes and fibroblasts in mice and humans (29–31). Largely quiescent at baseline, FAPs proliferate in response to muscle injury, facilitating muscle regeneration (32,33). In the mdx mouse model of muscular dystrophy, disordered FAP dynamics contribute to pathological adiposity and fibrosis (34,35). While FAPs may remodel skeletal muscle in obesity, their response to metabolic challenge remains unexamined. In this study, we tested whether fibro-adipogenic diaphragm remodeling occurs in obesity-associated respiratory impairment and whether FAPs contribute to the process.

Mice were obtained from The Jackson Laboratory (Bar Harbor, ME). The Jackson Laboratory maintains Sca-1 GFP (cat. no. 12643), PDGFRα-Cre (013148), and ob/ob (000632) mice on a C57BL/6J background and mT/mG (007576) mice on a 129X1/SvJ background. Animals were maintained in pathogen-free housing with a 12-h light-dark cycle and ad libitum food and water. For DIO studies, mice received a normal chow diet (CD) (5L0D; LabDiet, St Louis, MO) until 8 weeks of age. Control mice continued CD, while experimental mice switched to a high-fat diet (HFD) containing 45% calories from lipid (D12451; Research Diets, New Brunswick, NJ). Mice consumed the CD or HFD for 1, 3, or 6 months before analyses. For comparisons of ob/ob and wild-type (WT) C57BL/6J mice, all animals consumed CD and were analyzed at 16 weeks of age. The University of Michigan Institutional Animal Care and Use Committee approved all studies.

As previously described (36), diaphragms were localized by ultrasound (US) using a transversely oriented MS250 transducer (frequency 24 MHz) (Vevo 2100; Visual Sonics, Toronto, ON). Diaphragm motion, observed in M-mode, was recorded for three or more respiratory cycles. Excursion amplitude, inspiratory duration, inspiratory velocity, expiratory velocity, and peak-to-peak time were measured on still images. Inspiratory duty cycle was calculated as the quotient of inspiratory duration/peak-to-peak time.

Twitch properties and tetanic force (37) were measured on 2- to 4-mm-wide lateral costal diaphragm strips. In a Krebs-Ringer bath containing 0.03 mmol/L tubocurarine chloride, held at 25°C and bubbled with 95% O₂ and 5% CO₂ (maintaining pH 7.4), an attached rib was sutured to a servomotor (model 305B; Aurora Scientific, Aurora, ON) and the free central tendon edge to a force transducer (model BG-50; Kulite Semiconductor Products, Leonia, NJ). The bath was electrically stimulated through a field generated between two platinum electrodes by a biphasic current stimulator (model 701A; Aurora Scientific). LabVIEW 2014 software (National Instruments, Austin, TX) controlled electrical pulse properties and servomotor activity while recording transducer data. Strips were adjusted to optimal length (L_o) at which a stimulus pulse elicited maximum isometric force (P_o). Resting tension was measured at L_o. Contraction and half-relaxation times were measured during individual isometric twitches. Muscle cross-sectional area (CSA) was calculated using L_o and muscle mass. Specific force was calculated as the quotient of P_o/CSA.

Diaphragmatic stromal cells were isolated through a protocol adapted from one previously described (30). Costal diaphragms were excised, minced, and digested in collagenase type II (Worthington Biochemical, Lakewood, NJ) diluted to 0.067% by weight in serum-free DMEM (Thermo Fisher Scientific, Waltham, MA). After 1 h at 37°C, collagenase was inactivated with DMEM containing 10% FBS. Resultant suspensions were sequentially passed through 100-μm and 40-μm cell strainers.

Flow cytometry used established marker profiles (29). Briefly, cells isolated from Sca-1 GFP mice were fixed with 4% paraformaldehyde (PFA), permeabilized/blocked in 0.3% Triton X-100 with 5% normal goat serum, and then incubated with fluorophore-conjugated CD31, CD45, integrin α7, and platelet-derived growth factor receptor α (PDGFRα) antibodies (Supplementary Table 1). After DAPI staining, cells were analyzed on a MoFlo Astrios EQ running Summit software (version 6.3; Beckman-Coulter, Brea, CA). Gates were established using fluorescence minus one for labeled antibodies and WT littermate for (Sca-1) GFP. Cell cycle analysis was performed based on nuclear DAPI staining. For gene expression and ex vivo analyses of stromal isolates from WT mice, cells were purified by two-step magnetic-assisted cell sorting employing MiniMACS columns (Miltenyi, Bergisch Gladbach, Germany) following the manufacturer’s instructions. Cells were first subjected to negative selection using Lin (CD31 and CD45) antibodies (130-09-858). Resulting Lin⁻ cells were sequentially incubated with FITC-conjugated Sca-1 antibody (120-003-751) and anti-FITC microbeads (120-003-754). Lin⁻/Sca-1⁺ (FAP), and Lin⁻/Sca-1⁻ (non-FAP)-enriched populations were subsequently analyzed.

Whole tissues—costal diaphragm and brown adipose tissue (BAT)—were cleaned of adherent tissues, snap frozen, and digested in Trizol (Thermo Fisher) with mechanical homogenization, while stromal cells subjected to magnetic-assisted cell sorting (MACS) isolation were pelleted by centrifugation. Both underwent RNEasy-based RNA isolation following the manufacturer’s instructions (Qiagen, Germantown, MD). cDNA synthesis with SuperScript II (Invitrogen, Carlsbad, CA) preceded quantitative PCR with SYBR Green (Thermo Fisher Scientific) or Taqman (Applied Biosystems, Foster City, CA) on a StepOnePlus machine (Applied Biosystems). Primer/probe sequences are indicated in Supplementary Table 2.

After isolation, stromal cells were cultured for 1 day in 24-well plates containing standard medium: DMEM with 10% FBS and antibiotic/antimycotic (penicillin, streptomycin, and amphotericin B; Sigma-Aldrich, St Louis, MO). Cells were then seeded at 30% confluence in four-well chamber slides (Lab Tek; Thermo Fisher Scientific) with standard medium. Those allocated for Ki67 and extracellular collagen staining were evaluated after 3 days. Those allocated for adipogenesis were cultured for 5 days (reaching full confluence), then switched to standard medium plus adipogenic cocktail (1 μg/mL insulin, 250 nmol/L dexamethasone, 10 μmol/L troglitazone, and 10 nmol/L triiodothyronine) for 7 days, and then switched to standard medium plus insulin (1 μg/mL) for 3 days. After each experiment, cells were fixed, blocked, and immunostained as previously described (38). For Ki67 staining, wash buffers and antibody diluents contained 0.2% Tween-20. This was excluded for extracellular collagen staining. Cells subjected to adipogenesis analysis were fixed and then incubated in BODIPY-543 (Thermo Fisher Scientific). Cells subjected to THBS1 treatment were transferred to slide wells containing standard medium for 12 h and then switched to DMEM with 1% FBS. Subsequently, individual groups were incubated in this medium with or without THBS1 (cat. no. 605225; Calbiochem, San Diego, CA) at 0, 1, or 5 μg/mL, followed by Ki67 staining. Counterstain with Alexa Fluor 488–conjugated phalloidin (Thermo Fisher Scientific) was used in specific experiments.

As previously described (39), whole diaphragms were decellularized in 1% SDS for 5 days, fixed in 4% PFA, and then incubated in Oil red O (ORO) dissolved to 0.3% by weight in 60% isopropanol. After overnight incubation in 1% SDS, they were affixed to wax plates or dissection petri dishes for analysis.

Six-μm-thick formalin-fixed, paraffin-embedded sections of costal hemidiaphragm, BAT, and inguinal and gonadal white adipose tissue (WAT) were prepared as previously described (21) and stained with Sirius red, hematoxylin-eosin (H-E), and Masson trichrome by standard procedures. Immunohistochemistry on formalin-fixed, paraffin-embedded sections was performed as described in (21), with primary and secondary antibodies indicated in (Supplementary Table 1). Full-thickness diaphragm chunks (4 mm²) fixed in 1% PFA for 5 min at room temperature were used for immunofluorescent detection of type I and VI collagen by whole-mount confocal microscopy. All processing was performed in detergent-free solutions. Samples were imaged using an Olympus DP72 camera mounted on an Olympus SZ61 microscope (Tokyo, Japan) or a Nikon A1 confocal microscope running NIS-Elements software (Olympus). Three nonconsecutive sections per animal were used for quantitative morphometry, with analysis performed using NIH ImageJ.

Plasma (1 μL) was diluted in SDS lysis buffer, denatured, and run in 5% SDS-PAGE. THBS1 was detected by Western blot using rabbit anti-THBS1 antibody (Santa Cruz Biotechnology, Santa Cruz, CA) as previously described (21).

Data analysis employed Student t test for two-group comparisons, one-way ANOVA (with Tukey post hoc test) comparisons of two or more groups, two-way ANOVA (with Sidak post hoc test) for two variables, and linear regression for correlation analysis. P value <0.05 indicated statistical significance. Data are shown as mean ± SD.

To generate a model of obesity-induced respiratory dysfunction, we placed male C57BL/6J mice on a 6-month DIO time course, observing marked weight gain (Fig. 1A) and glucose intolerance (Supplementary Fig. 1A) in HFD-fed animals versus CD-fed counterparts. We analyzed diaphragm motion with noninvasive US (Fig. 1B). This technique is validated in rodents, and US-based measurement of diaphragm excursion amplitude correlates with ex vivo isometric force values in WT and mdx mice (36). In our animals, amplitude was unchanged between groups at 3 months but significantly lower in HFD-fed animals after 6 months (Fig. 1C). Similar HFD-induced declines occurred in inspiratory and expiratory velocity (Fig. 1D and E). Additionally, 6-month HFD mice displayed increased respiratory rates, indicated by reduced peak-to-peak time (Fig. 1F). Duty cycle, denoting percentage of each breath spent in inspiration (inspiratory duration/peak-to-peak time), also increased significantly with 6-month HFD feeding (Fig. 1G). These data indicate that long-term DIO impairs diaphragm motion, inducing restrained, low-amplitude contractions occurring at shorter intervals.

We analyzed diaphragm motion in ob/ob mice, which, at 16 weeks of age, had weights comparable with those of 6-month DIO animals (Supplementary Fig. 2A). When compared with age-matched WT controls, ob/ob mice also showed reduced amplitude (Supplementary Fig. 2B), inspiratory velocity (Supplementary Fig. 2C), expiratory velocity (Supplementary Fig. 2D), and peak-to-peak time (Supplementary Fig. 2E); however, duty cycle was unchanged (Supplementary Fig. 2F). As such, ob/ob animals displayed similar, but not identical, diaphragm motion defects versus DIO mice.

We next tested whether observed ultrasonographic changes resulted from functional defects intrinsic to the diaphragm. We specifically evaluated its costal region (40) (Supplementary Fig. 1B), previously shown to be responsible for contractile force generation (41), by performing ex vivo isometric testing (Supplementary Fig. 1C). Indeed, specific force, defined as maximum tetanic force per unit CSA, was >15% lower in 6-month HFD versus CD mice (Fig. 1H). Specific force measurements from individual animals correlated with excursion amplitudes (Supplementary Fig. 1D) and exhibited negative linear relationships with body weight (Supplementary Fig. 1E) and inguinal adipose mass (Supplementary Fig. 1F). Resting tension, time to peak tetanus, and half-relaxation time were unchanged with HFD feeding (Supplementary Fig. 1G–I). Notably, the reduced force generation induced by HFD feeding was not associated with significant changes in costal diaphragm thickness, average myofiber diameter, or percentage of muscle CSA occupied by myofibers (Supplementary Fig. 1J–M). In contrast with DIO mice, specific force measurements from ob/ob diaphragms were nearly identical to those of controls (Supplementary Fig. 2G). This result parallels prior reports of normal diaphragm contractile force in young ZDF rats (9,10) and is consistent with findings that ob/ob hypoventilation can be quickly reversed by leptin infusion (12).

Since long-term DIO yielded intrinsic diaphragmatic contractile compromise without obvious differences in myofiber morphology, we assessed whether other anatomic changes accompanied physiological impairment. First, we analyzed adiposity using decellularized diaphragms of 6-month CD and HFD mice. In CD-fed animals, ORO staining revealed well-demarcated lipid droplets closely juxtaposed with apparent branches of the phrenic vasculature (40) (Fig. 2A and B). HFD feeding increased ORO-positive droplet number and size, especially in the lateral costal diaphragm (Fig. 2A and B). H-E staining of costal diaphragm sections also showed unilocular adipocytes (Fig. 2C), with their identity confirmed by perilipin immunohistochemistry (Fig. 2D). These localized near blood vessels in CD-fed mice but extended farther into the tissue with HFD feeding (Fig. 2C). CSA of individual adipocytes increased after 1 month of HFD feeding (Fig. 2E) and was threefold greater versus that in controls at 3 and 6 months (Fig. 2E). Size distribution profiling demonstrated adipocytes in CD diaphragms to be small: CSA ≤250 μm² for the majority (Fig. 2F). On the contrary, in HFD diaphragms many adipocytes displayed CSA ≥1,000 μm² (Fig. 2F). Adipocyte number was statistically unchanged between groups at 1 and 3 months but was significantly elevated with HFD feeding at 6 months (Fig. 2G). Percentage of diaphragm CSA occupied by adipocytes also increased significantly with HFD feeding only at the 6-month time point (Fig. 2H). Notably, intradiaphragmatic adipocyte number negatively correlated with specific force in individually assessed diaphragm strips (Fig. 2I).

Transcripts encoding mature adipocyte markers fatty acid binding protein (FABP4) and leptin, as well as a WAT marker, TCF21 (42), increased with HFD feeding. On the contrary, those encoding lipoprotein lipase (LPL), adiponectin, and PPARγ were unchanged. As seen in other adipose depots (43), diaphragmatic adipsin declined with DIO (Fig. 2J). WAT exhibits obesity-induced inflammation and macrophage infiltration (44). While DIO diaphragms showed no change in Il1b, Tnfa, and Il6 (Supplementary Fig. 3A), Emr1, encoding macrophage marker F4/80, increased marginally with HFD feeding (Supplementary Fig. 3A). Nonetheless, we observed few F4/80⁺ cells and no discernable crown-like structures in diaphragms of CD- or HFD-fed mice (Supplementary Fig. 3B). Interestingly, DIO diaphragms showed strong Ucp1 induction, albeit at a low level versus BAT (Fig. 1K). Consistent with this, immunohistochemistry revealed isolated UCP1⁺ unilocular adipocytes among many UCP1⁻ cells (Fig. 1L). Together, these findings suggest that intradiaphragmatic IMAT exhibits WAT-like characteristics and ability to express UCP1 after metabolic challenge (45).

Notably, diaphragms of ob/ob mice, which demonstrated no ex vivo contractile defects, had larger adipocytes (Supplementary Fig. 2H and I) but equivalent adipocyte number relative to those of WT controls (Supplementary Fig. 2J).

Given association of intramuscular ECM deposition with obesity (21–26) and contractile dysfunction (20), we postulated that DIO diaphragms might undergo fibrotic remodeling. Indeed, percentage of costal diaphragm CSA occupied by polymerized collagen progressively increased during HFD feeding (Fig. 3A), becoming statistically elevated versus controls by 6 months (Fig. 3B). With DIO, collagen deposition was especially apparent near adipocytes (Fig. 3C). Moreover, within strips subjected to isometric testing, polymerized collagen area negatively correlated with specific force measurements (Fig. 3D).

Gene expression analysis of 6-month costal diaphragms revealed HFD-induced increases in transcripts encoding type I and VI collagen (Fig. 3E). On the contrary, levels of those encoding type III and IV collagen, as well as collagen cross-linking enzyme lysyl oxidase (LOX), were unchanged between groups (Fig. 3E and F). Among matrix metalloproteinases (MMPs), Mmp2 declined, Mmp9 did not change, and Mmp14 increased with DIO (Fig. 3G). Expression of profibrotic mediator Tgfb1 was unchanged, while that of Tgfb3 increased in HFD-fed mice (Fig. 3H). Additionally, levels of Pdgfa and Fgf2 were unchanged, while those of Ctgf and Thbs1 increased with HFD feeding (Fig. 3I). These profiles suggest active and complex gene regulation underlying diaphragm ECM remodeling in long-term DIO.

Unlike DIO animals, ob/ob mice showed no increase in intradiaphragmatic collagen deposition versus WT controls (Supplementary Fig. 4).

We surmised that FAPs—skeletal muscle progenitors capable of differentiation into adipocytes and fibroblasts (29,30)—might underlie the fibro-adipogenic diaphragmatic remodeling observed in DIO mice. FAPs express PDGFRα and can be defined by a surface marker profile positive for Sca-1 and negative for CD31 (endothelial marker), CD45 (hematopoietic marker), and integrin α7 (satellite cell marker) (29,30). We intercrossed PDGFRα-Cre mice, which express Cre recombinase under the PDGFRα promoter, with mT/mG reporter mice to trace distribution of GFP⁺ (FAP-derived) cells within the costal diaphragm (Fig. 4A). Significantly, GFP⁺ cells constituted all adipocytes in 6-month CD and HFD PDGFRα Cre-mT/mG diaphragm samples (Fig. 4B and Supplementary Fig. 5A). These adipocytes were larger and more abundant with HFD feeding (Fig. 4B). We then assessed distribution of extracellular type I collagen (COL1) and type VI collagen (COL6) within costal diaphragms of the same animals. We noted expected collagen staining in epimysium and central tendon, while observing no signal in samples stained with control antibody (Supplementary Fig. 5B–D). CD costal diaphragms contained thin COL1 fibrils between myofiber bundles (Fig. 4C). COL1 bands appeared larger in HFD-fed samples, while irregular densities also surrounded clusters of GPF⁺ stromal cells and adipocytes (Fig. 4C). Bands of COL6 appeared between myofibers in both groups; however, HFD diaphragms showed more prominent signal surrounding adipocyte inclusions (Fig. 4C).

Since PDGFRα⁺ cells gave rise to all adipocytes and many collagen-depositing cells, we specifically analyzed response of intradiaphragmatic FAPs to HFD feeding at the 6-month time point. First, we noted increased costal diaphragm PDGFRα expression in HFD-fed mice (Fig. 4D). Next, we isolated stromal cells from costal diaphragms of Sca-1 GFP mice. These animals harbor an EGFP construct driven by Sca-1 promoter, enabling analysis of Sca-1⁺/CD31⁻/CD45⁻/integrin α7⁻ FAPs by flow cytometry (Supplementary Fig. 6A and B). Consistent with prior reports, we found that these cells were broadly PDGFRα⁺ (Supplementary Fig. 6A). In DAPI-stained stromal cell isolates, we found that both FAP number and proliferation dramatically increased with long-term obesity (Fig. 4E).

To explore the link between DIO-induced diaphragm remodeling and FAP dynamics, we analyzed gene expression profile and ex vivo fibro-adipogenic potential of these cells. Specifically, we used immunomagnetic beads to isolate stromal populations enriched for FAPs (Lin⁻/Sca-1⁺) and non-FAP stromal cells (Lin⁻/Sca-1⁺) from costal diaphragms of WT 6-month CD and HFD mice. Lin^−/Sca1⁺ FAPs preferentially expressed Pdgfra, while Lin⁻/Sca1⁻ non-FAPs preferentially expressed satellite cell marker Pax7 (Fig. 5A). Critically, HFD feeding induced elevated expression of Col1, Col3, and Col6 in FAPs but not non-FAPs, implicating the former as the likely source of increased tissue-level collagens (Fig. 5A). Interestingly, Tgfb3, also globally elevated in the DIO costal diaphragm (Fig. 3H), increased with HFD feeding in both populations (Fig. 5A), suggesting that FAP-mediated fibrosis may be regulated by multiple stromal cell types.

Consistent with in vivo data, we observed greater ex vivo proliferation (Ki67 staining) in FAPs isolated from HFD-fed mice (Fig. 5B). COL1 deposition also increased markedly in these cells (Fig. 5C). Analysis of unsorted stromal cells from PDGFRa Cre-mT/mG mice confirmed these findings, with increased GFP⁺ cell proliferation and collagen deposition seen in DIO costal diaphragm isolates (Supplementary Fig. 7).

In FAPs, HFD feeding induced expression of Tcf21, Il6, and Ucp1 but not classic early adipogenesis–related genes Pparg2, Cebpa, and Pgc1a (Fig. 5D). In agreement with this, FAPs from CD- and HFD-fed mice showed equivalent ex vivo adipogenesis (Fig. 5E). Along with our in vivo findings, these data indicate that DIO induces FAP proliferation and collagen deposition but does not impact adipogenic potential. Increased diaphragmatic adiposity after long-term HFD feeding therefore likely results from increased FAP number.

We next sought trophic factors that might underlie FAP proliferation in the obese diaphragm. We focused on THBS1, a circulating ECM protein, produced by visceral adipose tissue (21) that acts as a potent mesenchymal cell mitogen (28). Its genetic ablation protects DIO mice from skeletal muscle fibrosis and insulin resistance (21). We first analyzed serum THBS1 level, finding it to be markedly increased in 6-month HFD mice (Fig. 5F). We next treated isolated FAPs with increasing THBS1 concentrations over a physiological range seen in obese humans (46). FAPs from CD-fed mice exhibited dose-dependent proliferation, whereas those from HFD-fed mice showed enhanced proliferation even at the 1 μg/mL concentration. These data highlight THBS1 as a candidate regulator of FAP proliferation and fibro-adipogenic remodeling of the DIO diaphragm.

In this study, we demonstrated that the mouse diaphragm undergoes progressive fibro-adipogenic remodeling with long-term HFD-feeding, ultimately exhibiting impaired motion in vivo and contraction ex vivo. The temporal correlation between fibro-adipogenic changes and diaphragm dysfunction, coupled with identification of proliferating FAPs as a likely remodeling driver, provides a framework for better understanding respiratory impairment in obesity.

Our data demonstrate that DIO of considerable duration is needed to cause respiratory compromise in mice. Noninvasive US found no significant diaphragm motion changes after 3 months of HFD feeding but revealed shallow, short-interval breaths with increased duty cycle by 6 months. This approximates breathing patterns of obese humans (1,5). As previously shown in mdx mice (a muscular dystrophy model with primary diaphragm dysfunction) (36), low excursion amplitudes seen on US correlated with reduced specific force measured ex vivo. In turn, specific force correlated negatively with body weight and adiposity measures. This parallels the increasing risk of obesity-associated respiratory dysfunction clinically observed at higher BMI (2). Elevated intradiaphragmatic adipocyte number and polymerized collagen content emerged only at the 6-month time point, when diaphragm dysfunction manifested, and correlated quantitatively with specific force decrements.

Equally obese but relatively young (16-week-old) leptin-deficient ob/ob mice showed similar diaphragm motion defects, with lower amplitude, inspiratory velocity, and expiratory velocity; however, they had nearly identical duty cycle and ex vivo specific force versus age-matched controls. Ob/ob diaphragms also did not demonstrate increased adipocyte number or polymerized collagen as seen with long-term DIO. These findings suggest that hypoventilation in ob/ob mice occurs independent of primary diaphragm compromise. Indeed, leptin is a key regulator of central respiratory control (47), and its administration reverses ventilatory defects observed in ob/ob but not DIO mice (12). We therefore surmise that chronicity of HFD feeding, with potential contribution from dietary content, underlies diaphragm anatomic remodeling and contractile dysfunction not seen in the younger ob/ob mice. The premise that long-term obesity is needed to induce intrinsic diaphragm defects may explain discrepancies in previous analyses of ZDF rats. In that model, while diaphragm isolates from young animals (3–4 months old) demonstrate unchanged or even increased (9–11) isometric force, those from older animals (8 months old) show force reduction (8).

IMAT accumulation and tissue fibrosis are both associated with skeletal muscle contractile dysfunction (16–20) and occur in obesity (17–19,21–26). In our DIO model, FAPs emerged as likely drivers of diaphragm fibro-adipogenic remodeling, giving rise to all adipocytes and many collagen-producing fibroblast-like cells. HFD feeding caused FAPs to proliferate and assume a fibrogenic phenotype. Since FAPs from HFD-fed mice had similar adipogenic potential versus controls, we reason that adipocyte accumulation in long-term DIO results from slow but progressive adipocyte differentiation from a larger FAP pool. FAPs from HFD-fed mice were nonetheless qualitatively different, exhibiting elevated expression of specific adipocyte-related genes. Tcf21, a WAT marker (42), recently shown to be enriched in visceral adipose stem cells (38), was induced in FAPs with HFD feeding, as was Il6, a TCF21-regulated gene (38). HFD feeding also upregulated FAP-specific expression of Ucp1, corresponding with presence of scattered UCP1⁺ unilocular adipocytes in DIO diaphragms. Interestingly, Ucp1 upregulation in FAPs and the presence of adipocytes positive for UCP-1 were demonstrated in mouse tibialis anterior muscles subjected to toxin-induced injury (48). Specific contributions of increased intramuscular adiposity and fibrosis to contractile dysfunction remain elusive but represent tractable areas of inquiry with improved understanding of FAP dynamics in obesity.

Our identification of THBS1—a WAT-derived, circulating ECM protein increased in obesity and type 2 diabetes (46)—as a mediator of FAP proliferation provides one link between overnutrition and diaphragm fibro-adipogenic remodeling. Recent work has identified THBS1 as a stromal cell mitogen, acting in concert with known FAP growth factors TGF-β and PDGF-AA (28). Since THBS1 also regulates skeletal muscle fibrosis (21), further evaluating its impact on FAPs will enable novel mechanistic exploration of diaphragm structure/function compromise in obesity.

Funding. Research reported here was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) of the National Institutes of Health (NIH) under award P30AR069620, National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant R01-DK-095137 (to T.-H.C.), and the University of Michigan Biomedical Research Council Fund (to T.-H.C.). E.D.B. is supported by the Marilyn Fishman Grant for Endocrinology Research from the Endocrine Fellows Foundation and NIDDK T32-DK-007696. D.E.M. is partially supported by NIAMS R01AR068428. The University of Michigan Animal Phenotyping Core, in which ultrasound experiments were performed, also receives support for training from NIH Office of the Director K26OD016502 to D.E.M.

The content of this publication is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. E.D.B., K.C.-B., C.S.D., and T.A. performed the experiments. E.D.B., D.E.M., S.V.B., and T.-H.C. evaluated data. E.D.B. and T.-H.C. conceived of the research idea and designed experiments. E.D.B. and T.-H.C. wrote the manuscript. E.D.B., K.C.-B., C.S.D., T.A., F.H., D.E.M., S.V.B., and T.-H.C. reviewed the manuscript. T.A., F.H., D.E.M., and S.V.B. provided technical consultation. T.-H.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.