Development of a platform of 3D adipogenesis to model, at higher scale, the impact of LY2090314 compound on fibro/adipogenic progenitor adipogenic drift

LY2090314 suppresses 2D and 3D adipogenesis of fibro/adipogenic progenitors by stabilizing a transcriptional competent  -catenin complex

Recently, endogenous WNT ligands are emerging as a critical regulator of FAP behavior, suggesting that the manipulation of WNT pathway components, via small molecules, may represent a feasible strategy for limiting the adipogenic drift of these cells (Reggio et al., 2020b). In the last years, our group uncovered GSK3 kinase as a transducer that controls adipogenesis, while identifying GSK3 inhibitors as valuable candidates to impair FAP adipogenic fate (Reggio et al., 2020b). Specifically, we demonstrated that a short-term exposure to LY2090314, a highly potent GSK3 inhibitor, is capable to limit fatty degeneration in glycerol-degenerating muscles as well as inhibit adipogenesis of FAPs extracted from dystrophic mdx muscles (Reggio et al., 2020b).
In the present work, we provide pharmacological evidences that an active -catenin transcriptional complex intimately represses the adipogenic program of adipose progenitor cells, including human FAPs. Using LC-MS/MS profiling, we demonstrated that LY2090314 exerts WNT mimicking properties to inhibit human FAP adipogenesis through, at least, two independent mechanisms, both mediated by -catenin. A primary mechanism is ascribed to the role of LY2090314 in inhibiting GSK kinase, while a second indirect mechanism further strengths GSK blockage via WNT5A up regulation.
To demonstrate the clinical value of LY2090314, we established a novel method to study fat infiltrates in a three-dimensional (3D) scale. Biomimetic matrices are paving the way for the generation of affordable humanized models for medical and research applications, allowing to scale up cell-based evidence in more complex and clinically-valuable tissuelike systems (Celikkin et al., 2021;Fernández-Garibay et al., 2022;Fuoco et al., 2012). Using the photopolymerizable Poly-Ethylene-Glycol-Fibrinogen (PF) hybrid biomimetic matrix with

Disease Models & Mechanisms • DMM • Accepted manuscript
FAPs, we generated a miniaturized fatty infiltrate 3D model. Employing this novel platform, we showed that a two-digit nanomolar dose of LY2090314 is effective to repress 3D adipogenesis, thus providing the proof of principle for the use this compound to limit FAPderived fat infiltrates in diseased muscles. In addition, our miniaturized 3D fatty depot represents a scalable system to collect cellular and molecular details regarding the impact of drugs while predicting in a 3D scenario clinical outcome, unwanted side effects as well as complementary molecular targets for combinatorial anti-adipogenic therapies.

A transcriptional competent -catenin complex is the most downstream effector repressing adipogenesis
3T3-L1 cells are immortalized murine preadipocyte cells that are widely used to shed light in the molecular routes that drive the transition of preadipocytes into lipid-laden cells (Rosen and Spiegelman, 2000). Adipogenesis of this cell line is promoted by prodifferentiation hormones supplemented in the adipocyte differentiation medium (ADM): insulin, synthetic glucocorticoids, such as dexamethasone, and the phosphodiesterase inhibitor 1-methyl-3-isobutyl xanthine (IBMX) (Rosen and Spiegelman, 2000).
After three days of ADM culture, 3T3-L1 started to express canonical adipogenic markers as a sign of the activated adipogenic differentiation. Specifically, the expression of two members of the Cebp family (Cebpa and Cebpb) as well as the master adipogenic gene Pparg were significantly induced by ADM exposure (Fig.S1A). Incubation with ADM increased PPARγ protein level that exclusively localizes in the nucleus of cells undergoing adipogenesis (Fig.S1B). Time course experiment confirmed the differentiation kinetics of these cells (Fig.S1C). Specifically, PPARγ expression (used as an early adipogenic marker) was detected at the 3 rd day after ADM incubation while Perilipin (that marks terminally differentiated cells) started to appear at the 5 th day and progressively increased until complete maturation of 3T3-L1 cells (Fig.S1C). The phenotype conversion into adipocytes can be revealed using a highly specific dye, such as Oil Red O (ORO), highlighting mature lipid-laden cells (Fig.S1D, E). Therefore, the presented results demonstrated that 3T3-L1 cells do not spontaneously differentiate into adipocytes.
Conversely, ADM incubation triggers a well-defined kinetics leading to the conversion of these progenitors into mature lipid-laden cells.

Disease Models & Mechanisms • DMM • Accepted manuscript
Embryonic pathways such as NOTCH, SHH and WNT/-catenin control several aspects of adipose biology, including embryogenesis, postnatal development, induction of differentiation and homeostasis (Rosen and MacDougald, 2006). Even though this notion is well-consolidated in literature, our knowledge mostly comes from experiments that consider one of these pathways at a time. Thus, such lack of information is possibly hindering the relevance of one of these pathways compared to the others. For these reasons, we sought to identify which signaling pathway, among NOTCH, SHH and WNT/catenin, has a prominent role in repressing the adipogenic programs of 3T3-L1. We selected SAHM1, GANT61 and iCRT-3 to block with high selectivity the transcriptional program of NOTCH, SHH and WNT/-catenin, respectively ( Figure 1A). Briefly, SAHM1 disrupts the binding between NCID and RBPJ blocking the activation of NOTCH target genes. Similarly, iCRT-3 displaces the interaction between -catenin and its transcriptional partner TCF/LEFs. By contrast GANT61 blocks Gli1 and Gli2 activity by aborting the SHH mediated signaling ( Figure 1A). Such an approach offers the advantage to target as downstream as possible such signaling routes with a reduced interference of upstream and collateral pathways.
We tested the inhibitory effect of these small molecules in unstimulated 3T3-L1 with the aim of identifying whether a single inhibition pathway may stimulate adipogenesis in the absence of the ADM (Fig.1B). Each compound has been tested at two different doses (10 and 20 M) in agreement with literature reports demonstrating the safety and noncytotoxicity of these compounds (Ashley et al., 2015;Mazumdar et al., 2011;Sharma et al., 2017). We found that only the inhibition of the transcriptional activity of -catenin by iCRT-3 is sufficient to stimulate adipogenesis in cultured 3T3-L1, in the absence of adipogenic stimuli ( Figure 1B), as revealed by the fractions of ORO-positive cells ( Figure   1C). This suggests that the transcriptional activity of -catenin is solely responsible for repressing adipogenesis of 3T3-L1 preadipocytes. Combinatorial treatments of these compounds, tested at 20 M, clearly demonstrated that only in the presence of iCRT-3 adipogenesis takes place ( Figure 1D and 1E), suggesting that -catenin is the most downstream effector repressing adipogenesis. Hence, the control of NOTCH and SHH on this phenotype is either dependent on -catenin or dependent on an intermediate factor that in turn impairs -catenin transcriptional activity ( Figure 1F).

LY2090314 raises β-catenin concentrations and abrogates adipogenesis of 3T3-L1 and hASCs
The data collected pointed out -catenin as a molecular determinant controlling the adipogenic differentiation of 3T3-L1 cells. Thus, we sought to investigate -catenin dynamics during 3T3-L1 adipogenesis. To this end, we monitored the levels of -catenin protein, during the first three days of 3T3-L1 cell differentiation. Notably, western blot analysis revealed a progressive temporal decrease of -catenin expression during 3T3-L1 adipogenesis (Fig.S2A). Immunofluorescence analysis displayed a significant reduction in the -catenin signal throughout the cellular compartment of these cells undergoing adipogenesis, especially in the nucleus ( Fig.S2B and S2C). To prove these observations, from a biochemical point of view, we processed whole cell lysates to enrich cytoplasm and nuclear protein fractions. Three days of incubation with the ADM were sufficient to cause an almost total depletion of -catenin from the nuclear compartment confirming the reliability of our microscopy data (Fig.S2D).
To clarify the molecular event(s) controlling the intracellular levels of -catenin, we performed a pharmacological-based experiment to monitor -catenin levels upon early ADM stimulation in the presence of cycloheximide, MG132 and Concanamycin A, compounds that block translation, proteosome and autophagy activities, respectively. Intriguingly, only MG132 was effective in rising -catenin concentrations suggesting that catenin is subjected to a proteasome-dependent degradation (already at 24 hrs) after ADM exposure ( Fig.S2E and S2F). MG132 exposure promoted the accumulation of a small band with higher molecular weight corresponding to the mono-ubiquinated form of catenin ( Fig.S2E), an event that is consistently lost when the translation is inhibited by cycloheximide. Overall, our data demonstrated that the activation of the adipogenic program requires a decrease of the cellular concentrations of -catenin, especially its nuclear fraction, and that such reduction is mediated by proteasome degradative processes.
To assess whether interfering with -catenin levels, namely a decrease, may have an impact on the adipogenic fate of 3T3-L1, we focused our attention on the GSK3 kinase that is known to prime -catenin for its degradation, via phosphorylation. To test this hypothesis, we exploited LY2090314 to stimulate -catenin accumulation within cells ( Fig.2A). LY2090314 is a highly selective GSK3 inhibitor capable to exert its action in a nanomolar concentration range, thus having a great translation potential (Atkinson et al.,

Disease Models & Mechanisms • DMM • Accepted manuscript
2015; Kunnimalaiyaan et al., 2018). Short-term incubation (3 days) with 20 nM of LY2090314 is sufficient to prevent -catenin degradation while rising -catenin amount within 3T3-L1 cells undergoing adipogenesis (Fig.2B). Moreover, we found that such stabilization parallels with the failure of PPAR expression, making these cells incapable to engage the adipogenic program. Consistently, 3T3-L1 fails to differentiate into mature adipocytes upon LY2090314 exposure as revealed by ORO staining (Fig.2C) and by the expression of Perilipin and PPAR at the end point of the differentiation program (Fig.2D).
Remarkably, short-term exposure to LY2090314 was sufficient to sustain the total as well as the non-phospho active form (i.e., nuclear) of -catenin in these cells, thus promoting catenin accumulation within the nucleus (Fig.2E).
To corroborate such data in a primary adipose progenitor cell model, we collected primary human adipose stromal cells (hASCs), a primary cell suspension that is enriched of adipose progenitors capable of differentiating into primary adipocytes. Primary hASCs were cultured and differentiated using standard protocols giving rise to highly differentiated ORO-positive adipocytes (Fig.S2G). Like 3T3-L1 cells, -catenin demonstrated to be subjected to the same lowering kinetics also in differentiating hASCs (Fig.S2H), suggesting that -catenin degradation hallmarks a genuine adipogenic program.
Incubation with 20 nM of LY2090314 is sufficient to prevent -catenin degradation while significantly rising the intracellular concentrations of this protein effector (Fig.S2I). Such event parallels the failure of hASCs to acquire a terminally differentiated phenotype, as revealed by the absence of ORO positive adipocytes (Fig.S2J). Contextually, PPAR and Perilipin expression (Fig.S2K) were halted in hASCs exposed to anti-adipogenic doses of LY2090314.
To mechanistically prove that LY2090314 is acting by stabilizing a nuclear transcriptionally active complex containing -catenin, we attempted to rescue adipogenesis in LY2090314treated 3T3-L1 by co-exposing cells to increasing doses of iCRT-3. Incubation with iCRT-3 significantly re-sensitizes these cells to the ADM, leading to the formation of mature adipocytes. Therefore, GSK blockage by LY2090314 raises -catenin intracellular levels and inhibits adipogenesis by stabilizing active -catenin transcriptional complex (Fig.2F).
All together, these data indicate that adipogenesis of adipose progenitors can be modulated by altering -catenin concentrations and that LY2090314 is highly effective in abrogating the differentiation fate of cell models endowed with adipogenic potential (Fig.2G).

LY2090314 impairs adipogenesis of primary fibro/adipogenic progenitors of murine and human origin.
Alteration of fibro/adipogenic progenitor (FAP) homeostasis is commonly found in muscle degenerative disorders. In muscular dystrophies, the myofibers breakdown coupled to an inefficient muscle regeneration leads to the progressive replacement of the contractile tissue with fat, that largely infiltrates muscle fibers. The cell origin of this non-muscle tissue has been ascribed to the altered behavior of FAPs (Giuliani et al., 2021;Theret et al., 2021). In a diseased or massively injured environment, as in the case of glycerol damage, Consistently, we previously demonstrated that once isolated from mdx dystrophic mice FAPs are prone to engage adipogenesis and that short-term exposure to LY2090314 desensitizes this detrimental fate (Reggio et al., 2020b). To confirm these findings in a more robust setup, we isolated wild-type murine FAPs (mFAPs) from 3 different C57/BL6J mice and subjected these mFAP preparations to adipogenic differentiation ( Consistently, iCRT-3 increased the fraction of mFAPs-derived adipocytes in a dosedependent manner (Fig.S3D), by inhibiting active -catenin transcriptional complexes.
Therefore, mFAPs-derived adipocytes were not present in cell cultures exposed to LY2090314 compound (Fig.S3E). Collectively these data indicate LY2090314 as a candidate agent to counteract fat infiltrates in muscular dystrophies, by halting FAP adipogenic propensity.
To formally prove the biological relevance of our model and clinical impact of this pharmacological agent, we decided to explore the impact of LY2090314 compound on human derived FAPs. To collect multidimensional information from human-derived samples, we decided to couple standard differentiation assays with state-of-the-mass spectrometry-based proteomics allowing the dissection of molecular details from human FAP cultures, at high-resolution and unbiased level.
Next for each donor, we exposed the resulting primary cultures to ADM either supplemented with the vehicle or with 20 nM LY2090314, for three days. Upon three days of adipogenic priming, hFAPs were harvested for proteomics (Fig.3A). In-solution (LC)-MS/MS quantitation approach allowed us to identify ~3,300 unique proteins (Table S1) Next, we interrogated Gene Ontology (GO) libraries to unbiasedly identify biological terms that are positively (Table S2) or negatively (Table S3) enriched by LY2090314 compound.
According to their adjusted p-value, WNT-related terms were found to be over-represented in cultures exposed to LY2090314 (Fig.3D, full enrichment analysis is provided as Table   S2), implying that the anti-adipogenic properties of this compound may be ascribed to its ability of acting as a WNT phenocopying agent. Consistently, volcano plot representation shows that -catenin (CTNNB1) protein is significantly and massively induced by LY2090314 while the adipogenic marker FABP4 is among the most depleted proteins ( Fig.3E). Consistently, short-term incubation with 20 nM LY2090314 sustained -catenin expression and completely abrogated the ability of hFAPs to differentiate into adipocytes in the presence of ADM (Fig.3F,G). Therefore, LY2090314 represses adipogenesis by acting as a WNT surrogate in human FAPs.
Surprisingly, we found that LY2090314 also increased the protein abundance of WNT5A ( Fig.3E), that we previously demonstrated to be capable, via a -catenin-dependent mechanism, of restricting adipogenesis of murine dystrophic FAPs (Reggio et al., 2020b).
To rationalize the content of information of our proteomic survey, we mapped the 71 significantly modulated proteins (input nodes) onto a literature-derived network of signaling and physical interactions. For the purpose we used SIGNOR database through a recently developed app in the cytoscape environment. To increase the connectivity between our protein entities, we used the possibility to include "bridge" proteins via a native SIGNOR algorithm (Lo Surdo et al., 2022).

Disease Models & Mechanisms • DMM • Accepted manuscript
The resulting network encompasses 51 nodes including receptors, signal transducers, transcription factors and phenotypes while relationships between nodes are either positive (activations) or negative (inhibitions) (Table S4). Here, we were capable to identify a network module centered on -catenin (Fig.S3F). Moreover, to increase the coverage of information, we also included the expression values for those protein entities that were dosed/quantified throughout this study (Fig.3H).
The subnetwork accurately describes incoming and outcoming stimuli of -catenin while highlighting its negative relation over the adipogenic transcription factors, especially on PPARG and CEBPA. Such inhibitory role is potentiated by the upregulation of WNT5A that contributed to stabilize -catenin by GSK3 blockage (Fig.3H).
Altogether, these data demonstrated that the anti-adipogenic role of LY2090314 is ultimately mediated by -catenin through, at least, two independent mechanisms: a primary mechanism that is ascribed to the role of LY2090314 of inhibiting GSK kinase, and a second indirect mechanism that further strength GSK blockage role via WNT5A.

Development of a 3D adipogenic model to test LY2090314 efficacy in mitigating fatty degenerative diseases.
In our lab, we developed and optimized reliable methods for the generation of 3D organoids using muscle progenitor cells (Fuoco et al., 2015). The system is based on photopolymerizable mixture of Poly-Ethylene-Glycol (PEG) and Fibrinogen (PEG-Fibrinogen: PF) that was extensively demonstrated as (i) biocompatible and FDAapproved, (ii) suitable for 3D culture of myogenic progenitor cells (Testa et al., 2020) and (iii) adapted for the generation of mature and functional human and murine myosubstitutes that efficiently graft in host recipient muscles (Costantini et al., 2021). Starting from these notions, we tested the possibility for generating 3D models of adipogenesis by culturing adipose progenitors in a PF-based 3D microenvironment. To this end, hFAPs were resuspended in 8 mg/ml PF matrix scaffold (Fig.4A). The resulting mixture was then loaded in a polytetrafluoroethylene (PTFE) caster and polymerized by exposing the mold chamber to non-toxic and low-penetrating UV light (365nm) (Fig.4A,B). The polymerized PF 3D-construct was stable, anchored to the mold fork, with a volume approximately of 1 cm 3 and easily manipulable in classical culture vessels (Fig.4B). Suprascapular transplantation of hFAP organoids in immunodeficient mice resulted in vascularized fibro/adipose depots (Fig.4C), resembling those that can be found in some human fatty degenerative conditions. As a consequence of a spontaneous fibro/adipose differentiation Disease Models & Mechanisms • DMM • Accepted manuscript drift, cryosections of these organoids showed the presence of hFAPs that express myofibroblast and adipocyte markers such as Collagens and Perilipin, respectively ( Fig.4C).
Moreover, we generated surrogate models of fat pad by using hASCs. Indeed, in this 3D culture state hASCs maintained the ability to respond to the ADM and differentiated in mature adipocytes with huge Perilipin-positive lipid droplets (Fig.4D), suggesting the suitability of this systems to generate 3D models of fat infiltrates.

LY2090314 inhibits murine and human 3D FAP adipogenesis
To this end, we tested LY2090314 at the concentration of 20 nM for its ability to inhibit 3D m/hFAP adipogenesis. In the absence of the GSK3 inhibitor, mFAPs differentiated into adipocytes as revealed by evident ORO-positive area that are visible throughout the 3D construct (Fig.5A). By contrast the incubation with 20nM LY2090314 prevented the formation of such visible ORO-positive area (Fig.5A). The longitudinal microscopy-based monitoring of these constructs clarified that LY2090314 prevented the accumulation of lipid droplets in these 3D models of fatty infiltrates (Fig.5B), as previously observed for those cells cultured in standard 2D culture vessels. Whole mount confocal microscopy demonstrated that highly mature mFAP-derived adipocytes were only observed in the absence of LY2090314 (Fig.5C). Consistently, hFAP 3D constructs failed to form highly differentiated 3D fat depots when exposed to 20 nM of LY2090314 (Fig.5D, upper and lower panel). The overall data presented in this study demonstrated that a two-digit nanomolar dose of LY2090314 compound is still effective to fully inhibit adipogenesis in a complex 3D environment, thus offering the rational proof for engaging valuable studies for testing clinical applications of LY2090314.

Discussion
Over the ruinous nature of the dystrophic damage, human muscular dystrophies are further worsened by intramuscular infiltrations of ectopic tissues, including fat, scar and occasionally mineralized tissue. As an aggravating factor, adipose infiltrates hinder the supply of nutrients to muscle fibers, limit muscle regeneration and exacerbate muscle deterioration. Non-physiological adipocytes also infiltrate the skeletal muscles of sarcopenic individuals or in obese and type II diabetes patients, contributing to the morbidity.

Disease Models & Mechanisms • DMM • Accepted manuscript
The origin of this deleterious fat is now ascribed to the altered differentiation behavior of fibro/adipogenic progenitors (FAPs) (Uezumi et al., 2011), a cell population that reside in the interstitial and perivascular space of the skeletal musculature (Joe et al., 2010;Santini et al., 2020). In the last ten years, several groups specialized their research in the growing field of FAP biology, elucidating roles and molecular mechanisms that helped to rationalize Therefore, these and other studies imply that, despite certain challenges, ad hoc manipulation of WNT pathway components is eligible for designing novel therapies to selectively target FAP adipogenicity.
To increase the clinical value of these observations, in the present work, we identified the -catenin transcriptional complex as an entity that intimacy represses adipogenesis of adipose precursor cells. Contextually, we tested the potential use of GSK3 inhibitor LY2090314 as a WNT phenocopying agent that stimulates stabilization and accumulation of -catenin. LY2090314 is a highly potent GSK3 inhibitor exerting its anti-adipogenic effect in the nanomolar range, significantly reducing promiscuous actions as well as unwanted off-target effects.
We demonstrated that the antiadipogenic effect of LY2090314 is conserved in a panel of murine and human adipose progenitor cells, including human FAPs. Therefore, this suggest that -catenin-mediated adipogenic repression is a highly conserved mechanism that can be manipulated to control the differentiation of adipose precursors cells.

Disease Models & Mechanisms • DMM • Accepted manuscript
Moreover, using LC-MS/MS profiling, we uncovered that LY2090314 exerts WNT mimicking properties to inhibit human FAP adipogenesis through, at least, two independent mechanisms, both mediated by -catenin: a primary mechanism ascribed to the role of LY2090314 of inhibiting GSK kinase, and a second indirect mechanism that further strength GSK blockage via up regulation of WNT5A.
Translating promising data in successful pre-clinical studies and/or clinical trials remains one of the hardest challenges for research that is focused on identifying novel pharmacological compounds for dystrophies treatment. Indeed, while promising, most of these data are limited to basic in vitro systems (i.e., 2D cultures) and fail to be verified in humanized animal models. Such drawbacks are often caused by limited testing steps in most complex culture scenarios, that would be needed to adjust doses, evaluate cellular responses and to model the clinical outcomes at cell resolution level. To overcome such limitation, we presented the possibility to generate 3D models of fat infiltrate by combining the photopolymerizable PEG-Fibrinogen biomimetic matrix with murine and human FAPs.
In this 3D state, FAPs are viable and graft in host recipient mice. Most importantly, in this 3D system FAPs maintained their ability to respond to external stimuli and differentiate into adipocytes. Moreover, one of the great advantages of this ectopic surrogate is the possibility to plan a long-term culturing/live monitoring of these cells, allowing longitudinal analyses and tests. By exploiting this system, we confirmed that a single two-digit nanomolar dose of LY2090314 is still sufficient to abrogate 3D FAP adipogenesis, concretizing the possibility to use LY2090314 inhibitor to limit fatty degeneration in higher systems. The integration of our current molecular knowledge on FAPs with the development of 3D humanized disease models will help in concretizing clinical therapeutic interventions for treating secondary complication of human dystrophy.

Human samples
The studies involving human participants were reviewed and approved by IFO (Istituti Fisioterapici Ospitalieri, Rome) ethics committee. The patients/participants provided their written informed consent to participate in this study, in line with Helsinki Declaration.
Donors were all healthy: XX53, XY65 and XX57. Biopsies have been taken during orthopedical intervention. FAPs have been isolated from muscle biopsies. ASC have been isolated from subcutaneous fat tissue.

Isolation of primary murine FAPs
FAPs were isolated from the hind limbs of male wild type C57BL/6J mice. Briefly, hind limbs were surgically removed and then minced in HBSS (GIBCO) supplemented with 100 U/ml P/S (Roche) and 0.2% BSA (AppliChem). For each mouse, the homogeneous muscle tissue preparation was enzymatically digested in 2 μg/μl collagenase A, 2.4 U/ml dispase II, and 10 μg/ml DNase I (Roche) in Dulbecco's phosphate-buffered saline (BioWest) w/calcium and magnesium. Enzymatic digestion was performed for 1 h at 37°C with gentle shaking. The homogenate underwent consecutive filtration through 100, 70, and 40 μm cell strainers (Corning). Before each filtration step, cells were centrifuged at 700g for 10 min at 4°C and then resuspended in fresh HBSS. Red blood cells were lysed in RBC lysis buffer (Santa Cruz). Freshly isolated muscle mononuclear cells were then resuspended in Magnetic beads buffer (0.5% BSA and 2 mM EDTA in 1× PBS) and filtered through a 30-μm Pre-Separation Filter (Miltenyi) to remove large particles. The whole cell suspension underwent subsequential incubations with the microbeadconjugated antibodies used for magnetic sorting. The sorting procedures and the labelling procedures with the microbead-conjugated antibodies were performed according to the manufacturer's instructions. FAPs were selected as Lin-/α7-int-/Sca-1+ cells.
Freshly sorted murine FAPs were resuspended in FAPs-GM consisting of high-glucose (25 mM) DMEM GlutaMAX supplemented with 20% FBS, 10 mM Hepes, 1 mM sodium pyruvate, and 100 U/ml P/S. After four days the FAP-GM was fully refreshed and cells cultured for two additional days before the induction of the adipogenic differentiation.

Isolation of primary human FAPs
Muscle biopsies were obtained as res nullius from surgeries on healthy donors, thanks to a collaboration with Istituti Fisioterapici Ospitalieri (IFO) hospital. Muscle homogenates were subjected to in the presence of dispase, collagenase and DNase I adopting the same protocol used to release murine FAPs. The resulting suspension was filtered and plated for 10 days in Cytogrow (Resnova) to allow cell amplification. For enriching human FAPs, amplified cells were stained with microbead-conjugated antibodies and selected as CD56-/CD15+ cells. Freshly sorted human FAPs were cultured and differentiated into adipocytes by adopting the protocols tested for the murine counterpart. Notably, to enhance adipogenesis of these cells, 1 M rosiglitazone was added to standard ADM.

Generation of a 3D model of fatty infiltration
All the experiments were performed by using PolyEthyleneGlycol-Fibrinogen (PF) biomimetic matrix. 3D fat infiltrate models were fabricated by a casting method. Briefly, PF 8mg/ml, supplemented with 0.1% of Irgacure ™ 2959 photoinitiator solution (Ciba Specialty Chemicals) was loaded with hFAPs or mFAPs. The fork was placed in the mold and the cell-hydrogel mixture poured in a polytetrafluoroethylene (PTFE) caster. The mixture was polymerized by exposing the mold chamber to non-toxic and low-penetrating UV light, for 5 minutes. The polymerized constructs were gently removed and transferred in normal dishes containing FAP-GM and then used according to the experimental needs.

Transplantation of 3D constructs loaded with hFAPs
Two-month-old male NOD/SCID mice (n=5) were anesthetized with a 1:1 mixture of ketamine (5 mg/mL) and xylazine (1 mg/mL) at a dose of 10 mL/kg i.m. A limited skin incision on the medial side of the back has been practiced, dorsal muscle was separated from the skin, the hFAP derived construct was carefully positioned and skin closure was performed by non-absorbable 6-0 silk sutures (Clinsilk). Mice were sacrificed at 28 days after implantation for morphological analysis.

Isolation of primary hASCs
Fat biopsies were obtained as res nullius from surgeries on healthy donors, thanks to a collaboration with Istituti Fisioterapici Ospitalieri (IFO) hospital. To release hASCs, fat biopsies were digested with an enzymatic mix containing Collagenase type II for 1hr at

Disease Models & Mechanisms • DMM • Accepted manuscript
37°C. hASCs were cultured and differentiated using the same protocols adopted for human FAPs.

Drug compounds
All the compounds used in this study were purchased from Selleckchem and reconstituted according to manufacturer's instructions.

Glycerol injury and muscle sections preparation and labelling
For glycerol muscle injury 50 μl of 50% v/v of hypertonic solution of glycerol was administered intramuscularly into the Tibialis Anterior of C57BL/6J mice. The contralateral limb was equally injected with saline solution, as internal control. After 14 days, mice were sacrificed and the hind limb muscles were surgically removed, embedded in optimal cutting temperature and snap-frozen in liquid nitrogen. Muscle sections with 10 M thickness were obtained using a Leica cryostat. Muscle sections were immunolabelled for PDGFR and Perilipin expression.

Real-time PCR
Total RNA was extracted using TRIzol. Before resuspension, total RNA was precipitated overnight in the presence of 10 μg of glycogen. RNA concentration was assessed using NanoDrop Lite Spectrophotometer. Total RNA (1,000 ng) was reverse transcribed into cDNA with PrimeScript RT Reagent Kit. qPCR reactions were carried out with SYBR Premix Ex Taq (Tli RNaseH Plus) and performed in technical duplicates for each biological repeat. Each reaction mixture (final volume of 20 μl) contained 50 ng of cDNA. Actb was used as reference genes.

Immunoblot
Cells were washed in PBS 1X and lysed in ice cold lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 1% Nonidet P-40, 1 mM EDTA, 1% Triton X-100) supplemented with 1 mM ortovanadate, 1 mM NaF, protease inhibitor mixture 1X (Sigma-Aldrich, Catalog P8340), inhibitor phosphatase mixture II 1X (Sigma-Aldrich, catalog 5726) and inhibitor phosphatase mixture III 1X (Sigma-Aldrich, catalog P0044) prior to use. Cell lysates were incubated in ice for 30 minutes and then separated at 15,500×g in a refrigerated centrifuge. Protein concentration was estimated using Bradford reagent (BioRad, catalog 500-0006). Total protein extracts were resolved by 10%, 15% SDS-PAGE or 4-20% Bio-Rad CRITERION gradient gel according to the needs. Proteins were transferred to Trans-Blot Turbo mini or midi nitrocellulose membranes (Bio-Rad, catalog 1704156-1704157) using a Trans-Blot® Turbo™ transfer System (Bio-Rad) and the non-specific binding were saturated for 1 hour at room temperature (RT) in blocking solution (5% milk, 0.1% Tween-20 in 1X Tris Buffered Saline). Saturated membranes were incubated with specific primary antibodies diluted in blocking solution according to the manufacturer's instruction. The binding of primary antibodies was revealed using host-specific secondary antibodies. Chemiluminescent detection was performed using Clarity™ Western ECL Blotting Substrates (Bio-Rad, catalog 1705061) and the Fujifilm Las-3000 imaging system.
Band densities were quantified using ImageJ. The brightness and contrast of each blot was adjusted using the "Auto Contrast" function in Adobe Photoshop. The full list of the antibody used is reported as TableS5.

Oil Red O
Oil Red O (Sigma-Aldrich) stock solution was prepared according to the manufacturer's instructions. Fixed cells were washed with 1× PBS and incubated for 10 min with filtered ORO working solution (3:2 ratio, ORO:ultrapure water). Stained cells were washed twice for 10 min with 1× PBS and counterstained using Hoechst 33342. ORO-stained cells were acquired via fluorescence or brightfield microscopy.

Image acquisition and analysis
Immunolabeled cells and section were acquired using the DMI6000B fluorescent microscope (Leica). Images were scored manually using Fiji by two independent collaborators in a blind analysis. Results are expressed as a ratio of the total objects counted.

Proteome sample preparation
Cells were harvested as indicated in the text and directly lysed in ice cold RIPA buffer.
Proteome preparation was done using the in StageTip (iST) method (Kulak et al., 2014;Reggio et al., 2021). Samples were separated by HPLC in a single run (without prefractionations) and analysed by LC-MS/MS.

LC-MS/MS measurements
Instruments for LC-MS/MS analysis consisted of a NanoLC 1200 coupled via a nanoelectrospray ionization source to the quadrupole-based Q Exactive HF benchtop mass spectrometer. Peptide separation was carried out according to their hydrophobicity on a home-made column, 75 um ID, 8 Um tip, 400 mm bed packed with Reprosil-PUR, C18-AQ, 1.9 um particle size, 120 Angstrom pore size (New Objective, Inc., cat. PF7508-250H363), using a binary buffer system consisting of solution A: 0.1% formic acid and B: 80% acetonitrile, 0.1% formic acid. Total flow rate: 300nl/min. LC linear gradient: after sample loading, run start at 5% buffer B for 5 min, followed by a series of linear gradients, from 5% to 30% B in 90 min, then a 10 min step to reach 50% and a 5 min step to reach 95%. This last step was maintained for 10 min.
MS spectra were acquired using 3E6 as an AGC target, a maximal injection time of 20 ms and a 120,000 resolution at 200 m/z. The mass spectrometer operated in a data dependent Top20 mode with subsequent acquisition of higher-energy collisional dissociation (HCD) fragmentation MS/MS spectra of the top 20 most intense peaks. Resolution for MS/MS spectra was set to 15,000 at 200 m/z, AGC target to 1E5, max injection time to 20 ms and the isolation window to 1.6 Th.
The intensity threshold was set at 2.0E4 and Dynamic exclusion at 30 second.

Proteome data processing
All acquired raw files were processed using MaxQuant (1.6.2.10) and the implemented Andromeda search engine. For protein assignment, spectra were correlated with the Human (v. 2021) including a list of common contaminants. Searches were performed with tryptic specifications and default settings for mass tolerances for MS and MS/MS spectra.

Disease Models & Mechanisms • DMM • Accepted manuscript
Missing values have been replaced by random numbers that are drawn from a normal distribution. Two -sample T-test analysis was performed using FDR = 0.05. Proteins with difference Log 2 Difference≥ ± 1 and q value <0.01 were considered significantly enriched.
Categorical annotation was added in Perseus in the form of gene ontology (GO) biological process (GOBP), molecular function (GOMF), and cellular component (GOCC), and KEGG pathways.

Protein Network generation
This strategy has been previously developed and applied to query complex proteome datasets (Reggio et al., 2020a;Sacco et al., 2019) Casual relationships between significant protein entities (coming from the LY2090314-vsvehicle comparison) were retrieved from SIGNOR database (Lo Surdo et al., 2022)using a dedicated app in the cytoscape platform (Shannon et al., 2003). Nodes were color-coded according to their difference value (LY2090314-vs-vehicle comparison) while the dimension of nodes is proportional to their q-value.

Data availability
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD040479. repeats. All data are represented as mean ± SEM and the statistical significance is defined as **p < 0.01; ***p < 0.001. represented as mean ± SEM and the statistical significance is defined as ***p < 0.001.     Table S1. Full table reporting proteins profiled via high-resolution mass spectrometry. Table S2. List of GO terms that are positively enriched in the LY2090314-vs-vehicle comparison. Table S3. List of GO terms that are negatively enriched in the LY2090314-vs-vehicle comparison. Table S5. Reagents used in this study.

Disease Models & Mechanisms • DMM • Accepted manuscript
Click here to download Table S1 Click here to download Table S2 Click here to download Table S3   Table S4. Casual interactions between protein entities that are part of LY2090314specific network.
Click here to download Table S4 Click here to download Table S5 Disease Models & Mechanisms: doi:10.1242/dmm.049915: Supplementary information