Induction of muscle stem cell quiescence by the secreted niche factor Oncostatin M

The balance between stem cell quiescence and proliferation in skeletal muscle is tightly controlled, but perturbed in a variety of disease states. Despite progress in identifying activators of stem cell proliferation, the niche factor(s) responsible for quiescence induction remain unclear. Here we report an in vivo imaging-based screen which identifies Oncostatin M (OSM), a member of the interleukin-6 family of cytokines, as a potent inducer of muscle stem cell (MuSC, satellite cell) quiescence. OSM is produced by muscle fibers, induces reversible MuSC cell cycle exit, and maintains stem cell regenerative capacity as judged by serial transplantation. Conditional OSM receptor deletion in satellite cells leads to stem cell depletion and impaired regeneration following injury. These results identify Oncostatin M as a secreted niche factor responsible for quiescence induction, and for the first time establish a direct connection between induction of quiescence, stemness, and transplantation potential in solid organ stem cells.


Introduction
Stem cells respond to tissue-specific activating signals by proliferating and giving rise to both committed progenitors as well as quiescent daughter cells. Muscle repair is mediated by resident muscle stem cells (MuSC; also known as satellite cells). In response to myofiber damage, satellite cells break from quiescence and give rise to committed progenitors known as myoblasts, which can fuse both with each other as well as with the damaged myofibers.
Importantly, a subset of activated satellite cells are then re-selected into the quiescent stem cell niche in order to maintain the stem cell pool 1 . This process is highly regulated, involving not only cell autonomous signaling pathways, but also extensive regulation by tissue resident stromal cells and invading inflammatory cell populations 2 . These regulatory pathways are critical, as defects in stem cell quiescence, activation, or self-renewal have been implicated in a variety of disease states including aging-associated sarcopenia, muscular dystrophy, and cancer cachexia 3, 4, 5, 6 .
Significant progress has been made toward elucidating the complex extracellular cues arising from the circulation and microenvironment that govern the behavior of tissue resident MuSC. Among the known soluble growth factor/receptor pathways, the HGF pathway has been demonstrated to control 'alerting' of satellite cells in response to remote injury 7 , inducing a state of metabolic activation and priming for cell cycle reentry. FGF likewise regulates satellite cell activation 8 , while a downstream negative regulator of FGF receptor tyrosine kinase signaling, Sprouty1, plays an important role in the acquisition of quiescence of activated satellite cells 9 .
Growth factor signaling eventually impinges on the p38α/β MAPK pathway, which both genetic and pharmacologic data suggests functions as an important regulator of stemness and proliferation 6,10 . Additional levels of regulation are imposed via Notch/Delta-dependent regulation of self-renewal 11,12,13 , Wnt-dependent control of myogenic fate 14 , and asymmetric division associated with differential Pax7 expression 15 .
In addition to the growth factor pathways described above, previous work has demonstrated the importance of circulating cytokines in regulating MuSC and myofiber function.
The IL-6 pathway has been a particular focus given the upregulation of this pleiotropic cytokine following exercise 16 , as well as the aberrant regulation of the downstream JAK-STAT signaling pathway during aging and in other disease states 17 . Studies with IL-6 deficient animals indicated a role for this cytokine in satellite cell and myoblast proliferation during experimentally-induced hypertrophy 18 . The IL-6 family member Leukemia Inhibitory Factor (LIF) has likewise been demonstrated to induce proliferation of human and mouse myoblasts 19 . Indeed, it has recently been reported that inhibition of STAT3-dependent signaling, which is activated by both LIF and IL-6, can rescue age-associated proliferative defects in mouse satellite cells 20 and promote satellite cell expansion 21 . Interpretation of the results of genetic and pharmacologic STAT3 inhibition is complicated, however, by the fact that this effector can be activated by numerous upstream signals, including both IL-6 and other cytokines 22 .
Despite this progress, the precise signals governing the choice between quiescence and proliferation remain poorly understood, and this constitutes a major barrier to reversing the regenerative block seen in a variety of disease states. In particular, the identities of the factors regulating quiescence of adult stem cells have remained elusive. While bioinformatic analyses have begun to address this issue by identifying genes enriched in quiescent stem cells 23, 24 , the functional significance of most of these putative regulators has not been established. To address this deficiency, we have undertaken a systematic screen for secreted and transmembrane protein regulators of stem cell proliferation and engraftment. Employing an unbiased in vitro/in vivo imaging based screening strategy, we identify Oncostatin M, a member of the IL-6 family of cytokines, as a novel regulator of stem cell quiescence, engraftment, and muscle regeneration.

A secretomic screen for muscle stem cell regulators
We hypothesized that a secreted regulator of stem cell quiescence would be capable of suppressing MuSC activation and proliferation in vitro, while conversely maintaining the ability of these cells to engraft following transplantation in vivo. To identify such regulators, we designed and carried out a protein library screen on purified MuSCs, using a combination of in vitro and in vivo imaging to distinguish between effects on stem cell proliferation and engraftment (Fig. 1a). Primary muscle stem cells isolated from luciferase transgenic mice 25,26 were cultured on uncoated plastic plates in the presence of pools of proteins derived from the GNF Secretomics library, which contains purified versions of known or predicted mammalian secreted proteins, as well as the extracellular domains of known or predicted single-pass transmembrane proteins 27 . No additional growth factors (e.g. FGF) were added. In vitro bioluminescence imaging (BLI) was used to assess stem cell proliferation ( Supplementary Fig.   1a, top), followed by cell harvesting and transplantation into the tibialis anterior (TA) muscle of hindlimb-irradiated recipients to assess engraftment, a sensitive measure of regeneration ( Supplementary Fig. 1a, bottom). Serial in vivo imaging was performed over a timecourse to quantify engraftment activity of the transplanted cells, as indicated by a progressive increase in light output, a feature of stem cells but not committed progenitors 25 . A comparison was made to both untreated cells, as well to known inducers of stem cell proliferation, FGF and the p38 MAPK inhibitor, SB202 6,10 . Protein pools demonstrating increased engraftment on serial in vivo imaging were iteratively deconvolved through rounds of subpool rescreening to identify the single protein responsible for the engraftment activity (Fig. 1b).
Pooled screening of ~600 predominantly mouse proteins yielded 5 hits in two phenotypic classes. The first class was exemplified by a pool driving both in vitro proliferation as well as in vivo engraftment (pool A; Fig. 1c, Supplementary Fig. 1b). Iterative deconvolution identified the responsible protein as FGF2, a known mitogenic factor for MuSC 28 , providing strong validation of the assay. The second phenotypic class exemplified the hypothesized behavior for an inducer of reversible stem cell quiescence, driving little in vitro stem cell proliferation but robust in vivo engraftment. Deconvolution of one such pool (pool Y; Fig. 1c, Supplementary Fig. 1b) identified the extracellular domain of the TNFα receptor ( Supplementary Fig. 1c), which acts as a receptor decoy for TNFα and which is consistent with the known pro-differentiation function of TNFα on satellite cells 29 . Pools F and H were not able to be completely deconvolved, possibly due to partial contribution of multiple protein factors. Strikingly, deconvolution of the remaining pools in this phenotypic class identified the same protein in three differentially epitope-tagged forms,  Fig. 2b), which may reflect either true heterogeneity, or limited sensitivity of the single cell detection method. To exclude effects on proliferation as a confounding variable, stem cell engraftment was next determined on a per-cell basis. Satellite cells were cultured for 2 weeks either with OSM or in standard medium containing FGF, and intramuscular transplant of 250 counted cells was then performed (Fig. 1e, left). Serial in vivo imaging revealed progressive increase of light output, indicative of engraftment, and injury of the stem cell-engrafted muscle with cardiotoxin led to a drop in BLI signal followed by a robust rebound, confirming the presence of functional satellite cells (Fig. 1e, right). BLI signal was significantly higher in muscle receiving OSM treated MuSC, consistent with increased per-cell engraftment and regeneration.

Regulation of muscle stem cell quiescence by OSM
To further investigate the anti-proliferative effect of OSM, we performed serial in vitro BLI. This demonstrated inhibition of stem cell proliferation at all timepoints (Fig. 2a, Supplementary Fig. 3a), which was attributable to decreased S-phase entry (Fig. 2b).
Proliferation of stem cells activated by culture on muscle fibers ex vivo was also strongly inhibited ( Supplementary Fig. 3b,c), and OSM treatment of primary mouse myoblasts likewise decreased levels of the M-phase marker phospho-H3S10 (Fig. 2c). Previous studies have demonstrated a critical role of the cyclin-dependent kinase inhibitor p27/Kip1 in quiescence of stem cells both in muscle as well as other lineages 30, 31 . We also observed a significant decline in p27 transcript in cultured MuSC, and moreover found that OSM treatment maintained p27 expression at levels seen in freshly isolated, non-cultured stem cells ( Supplementary Fig. 4a), suggesting that OSM might directly induce quiescence. To explore this possibility, we compared the global transcriptional profile of freshly isolated MuSCs to MuSCs cultured for 3 days with or without OSM, as well as to prior transcriptional studies of muscle stem cell quiescence 23, 24 . The published analyses were used to extract a transcriptional signature of quiescence regulated genes.
A comparison of this 'quiescence signature' with the transcriptional response to OSM revealed highly significant enrichment (Fig. 2d), which was independently confirmed by qPCR analysis ( Supplementary Fig. 4b). Over half of the signature genes were present within the OSM- Reversibility is the hallmark feature of quiescence, and we therefore tested whether OSM-induced cell cycle exit was reversible by culturing satellite cells in either activating (-OSM) or non-activating (+OSM) conditions for 3 days, after which the medium conditions were switched and EdU pulse labeling performed (Fig. 2e, left). Non-cycling OSM-treated cells robustly re-entered cell cycle following stimulation, as measured by both EdU incorporation

OSM treatment maintains stemness in vitro
Although both satellite cells and their committed progeny (myoblasts) express Pax7 and can contribute to myofiber formation 34 , only satellite cells are capable of mediating serial regeneration following injury 25 . The ability to reisolate and retransplant engrafted satellite cells is thus a strict test of stem cell identity, and we therefore tested whether OSM-treated MuSC fulfill these criteria. We employed a newly generated reporter allele in which an Internal Ribosome Entry Site and Green Fluorescent Protein (IRES-GFP) are inserted into the 3' untranslated region of the Pax7 locus (Pax7 GFP ; Supplementary Fig. 5a). This allele specifically labels CD34 + /α7 integrin + satellite cells in skeletal muscle ( Supplementary Fig. 5b). MuSC from Pax7 GFP mice were cultured in the presence or absence of OSM, and intramuscular transplant of equal numbers of treated stem cells was performed. Despite high initial engraftment of cells cultured either with or without OSM (Fig. 1d), Pax7 GFP -expressing MuSC could only be reisolated from mice engrafted with OSM-treated cells (Fig. 3a). To establish that the engrafted cells represent bona fide functional stem cells, Pax7 GFP+ cells were re-isolated from primary recipients of OSM-treated MuSC, and 50 counted cells were transferred into the hindlimbs of irradiated secondary recipients. BLI demonstrated engraftment in ~40% of recipients, and cardiotoxin injury induced initial loss followed by rebound of BLI signal, a pattern characteristic of stem cell-mediated regeneration (Fig. 3b). Previous studies have demonstrated that such increased serial transplantation ability corresponds with higher Pax7 levels 15 , and we likewise found that OSM treatment not only induced Pax7 (Fig. 3c) but also reciprocally decreased expression of the differentiation marker myogenin ( Supplementary Fig. 6a). This effect was unique amongst the related IL-6 family of cytokines ( Fig. 3d) 20,21 . Intracellular staining and flow cytometry in primary myoblasts demonstrated that OSM preferentially activated STAT3 rather than STAT5 (Fig. 3e). Analysis of the Pax7 locus revealed a previously uncharacterized binding site for STAT3, and chromatin immunoprecipitation (ChIP) from primary myoblasts confirmed direct occupancy of this site by STAT3 in response to OSM treatment (Fig. 3f). Taken together, these data reveal a MEK and STAT3-mediated response to OSM in satellite cells, leading to quiescence induction as well as maintenance of stem cell identity and function.

Functions of OSM signaling in vivo
To address the endogenous source of OSM in muscle, we assessed protein levels in muscle tissue and found they were highest in the resting state, decreased rapidly following injury, and reappeared beginning at day 4 and robustly by day 10, concomitant with myofiber regeneration (Fig. 4a). This suggests that OSM is produced in a paracrine fashion from the myofiber niche, rather than in an autocrine fashion from satellite cells. Moreover, this timecourse is inconsistent with immune cells, such as macrophages, being the primary source of OSM, as reported by others in the heart 33 . Immunofluorescence staining of OSM in uninjured muscle indeed revealed localization within myofibers of uninjured muscle, both around myonuclei and the fiber periphery (Fig. 4b), consistent with OSM production by myofibers as a secreted niche factor. To investigate the physiological significance of niche-induced OSM signaling, we derived mice deficient for the obligate OSM receptor, OSMRβ ( Supplementary Fig. 6b). Intermediate OSMRβ levels were present in heterozygous animals ( Supplementary Fig. 6b), consistent with previously reported haploinsufficiency 33 . EdU incorporation was significantly higher in OSMRdeficient satellite cells ( Supplementary Fig. 6c), demonstrating that OSM signaling in satellite cells restrains stem cell proliferation. Importantly, OSMRβ was required for the pro-quiescence effect of OSM ( Supplementary Fig. 6c Cre-loxP mutagenesis to conditionally ablate OSMRβ in Pax7 + MuSC using inducible lineagerestricted Cre expression 35 . Serial muscle injury of conditionally deleted animals led to impaired muscle regeneration, marked decrease in myofiber cross sectional area, and severe depletion of satellite cells (Fig. 4c,d,e; Supplementary Fig. 7e). The regenerative defect of satellite-cell specific OSMR deletion was indistinguishable from that of whole-body constitutive deletion, demonstrating a specific requirement for OSM signaling in satellite cells.

Discussion
Here we describe a novel in vivo imaging-based screen which identifies Oncostatin M as a key niche-derived secreted regulator of muscle stem cell quiescence. OSM is, to our knowledge, one of only two niche-localized factors known to induce quiescence in solid organ stem cells 36 , and the first shown to maintain transplantation potential. The ability of OSM to promote MuSC engraftment without inducing proliferation is distinct from previously reported factors such as regulators of p38 MAPK 6, 10 as well as related cytokines such as IL-6 and LIF ( Fig. 2h) 18,20,21,37 . The quiescence-inducing function of OSM in skeletal muscle is likewise distinct from its effect in other tissues such as the heart, where it acts to induce cardiomyocyte proliferation 33 , and more closely resembles the cytostatic effect originally described in cancer cell lines and mouse myoblasts 38, 39, 40 . This cytostatic effect was moreover shown to be associated with STAT3 activation, and inhibition of CDK2 activity. Although OSM was paradoxically reported to have a pro-proliferative effect on human myogenic progenitor cells, interpretation of this finding is complicated by the fact that human (but not mouse) OSM also signals through the proliferation-inducing LIF receptor (LIFR) pathway 41, 42 .
The temporal expression pattern of myofiber-derived OSM resembles that of Sprouty1, a satellite cell-intrinsic inhibitor of FGF signaling whose re-expression following injury is likewise critical for re-induction of stem cell quiescence 9 . Importantly, while OSM is re-expressed by fibers within 4-10 days of injury, Sprouty1 re-expression requires 3 weeks 9 . We suggest that this difference may define an early stage of quiescence, in which nascent myofibers secrete OSM in order to induce cell cycle exit and maintain stem cell identity in a subset of activated satellite cells (Fig. 4f). This model is in keeping with restricted expression of OSMR within a subset of satellite cells ( Supplementary Fig. 2b), and adds to our knowledge of heterogeneity within the stem cell pool 43,44 . In addition to its role following muscle injury, we speculate that OSM signaling may play additional roles in the steady state, for instance preventing full activation of stem cells induced by remote damage to enter the "G Alert " state 7 . Future work will address this possibility, the fate of OSMR deficient MuSC following activation, as well as whether dysregulation of OSM levels with aging or chronic injury may contribute to impaired maintenance and function of the stem cell pool 3, 4, 6, 28 . It will likewise be important to understand the mechanisms by which signaling pathways shared by multiple IL-6 family members are translated into distinct biological responses 44, 45 . Finally, we anticipate that the in vivo imagingbased screening approach reported here will prove useful in identifying additional niche components controlling the function of satellite cells and stem cells in other lineages. Subsequent experiments involving stem cell transplants utilized hindlimb-irradiated, immunocompromised NSG mice. Ubiquitously expressing transgenic CAG-luciferase-GFP mice were obtained and maintained as described previously 6,25 . Lineage reporter Pax7-IRES-GFP mice were generated using a strategy similar to one previously described 46

Stem Cell Purification and Culture
Mouse muscle stem cells were purified essentially as previously described 6 . In brief, tibialis anterior and gastrocnemius muscles were dissected and digested with Collagenase II and Dispase

Muscle stem cell transplantation and muscle injury
Stem cell harvesting, intramuscular injection, and cardiotoxin-mediated muscle injury were all performed essentially as previously described 6 , with the exception that 50 ul 10 µM cardiotoxin was used for all injuries. For counted cell transplants, stem cells were counted manually using a hemocytometer prior to transplant.

Bioluminescence imaging.
Bioluminescence imaging (BLI) was performed using either a Xenogen 100 or IVIS Spectrum CT, as previously described 6,25,48 . Briefly, 150 µL D-luciferin (reconstituted at 30 mg/mL in sterile water; Caliper LifeSciences) was administered by intraperitoneal injection into anesthetized mice. BLI were acquired using 60 second exposures for a total of 12 min following luciferin injection. Images were recorded and analyzed using Living Image software (Caliper LifeSciences). For images acquired on the Xenogen 100, a signal value of 100,000 photons per second was routinely used to define the engraftment threshold, as previously described 6,25 .
Engraftment thresholds were approximately ten-fold higher for imaging performed on the IVIS Spectrum CT, reflecting the higher sensitivity on this instrument and intrinsic detector characteristics between machines. In vitro imaging was performed similarly but with a 1:1000 final dilution of stock luciferin.

Graphical and Statistical Analysis
Graphs and statistics were generated using GraphPad Prism software. Student's two-tailed t tests were used to calculate p values.

Cell and tissue staining
Tibialis anterior (TA) muscle sections were prepared for staining essentially as previously described 6,48 . For OSMR staining, whole TA muscles were first fixed with 0.5% paraformaldehyde/PBS for 2 hr at 4C, followed by incubation in 20% sucrose/PBS overnight at

Single-cell RT-PCR
Single cell RT-PCR was performed as described 6,25 . In brief, MuSCs were obtained from digested muscle as above, and three rounds of FACS sorting was performed in order to obtain a highly purified CD34 + α7 integrin + MuSC population. In the last sorting step, the cells were

qRT-PCR
Analysis of purified cultured satellite cells was performed as described 6,25

Microarray analysis
Total RNA was prepared with Qiagen RNeasy kits. Twenty nanograms of total RNA was prepared for array analysis using Affymetrix 3' IVT Express amplification reagents. Fifteen micrograms of cRNA was hybridized to Affymetrix MOE_430_2 arrays, and arrays were washed, stained, and scanned using standard methods. CEL files were processed using MAS5.
Microarray data have been submitted to GEO (accession number GSE69976).

Bioinformatic analysis
OSM signature genes were defined as the top-quartile of the most up-regulated genes using the ratio of OSM-treatment/no-OSM treatment. Similarly a quiescence signature was defined using published data 23, 24 as the intersection of the top quartile of up-regulated genes (ratio of quiescence vs activated) in two comparisons. These two signature sets were then compared to each other and the number of overlapping genes was found to be higher than expected by chance as assessed by Fisher's Exact Test (n=754 genes, 2.01 fold enrichment, p=4e-105).

Data availability
Microarray data are available via GEO (accession number GSE69976). The data that support the findings of this study are available on request from the corresponding authors (SCS, HMB).

Author Contributions
Experiments were designed, carried out, and interpreted by S.       (f) ChIP performed using primary myoblasts cultured for 6 days either with or without OSM.
Immunoprecipitation was performed with either control IgG or STAT3-specific antibodies, and PCR was performed using primers specific for a predicted STAT3 binding site adjacent to the Pax7 locus.