Pericytes in the myovascular niche promote post-natal myofiber growth and satellite cell quiescence

Stem cells reside in specialized environments that regulate their behavior, called stem cell niches. Myogenic stem cells called satellite cells (mSCs) are involved in muscle growth and in regeneration following injury. During the late fetal development period, a basal lamina forms around myofibers (Bröhl et al., 2012; Rosen et al., 1992) and myogenic cells expressing PAX7 and PAX3 adopt the mSC position between the basal lamina and plasma membrane of myofibers (Relaix et al., 2005). mSCs have the potential to provide additional myonuclei to the myofiber or to enter a quiescent state. The mSC fate is tightly regulated by dynamic interplays between mSC intrinsic factors and extrinsic factors constituting the mSC niche (Yin et al., 2013). In undamaged adult muscle, mSCs are quiescent, binding the basal lamina through α7/β1 integrin and the parent fiber sarcolemma through M-cadherin (Yin et al., 2013), but mSCs or their progeny may also interplay by direct contact or paracrine signals with a variety of neighboring supportive cells, including endothelial cells (ECs) (Christov et al., 2007), resident connective tissue cell types, such as poorly defined ‘peri-endothelial’ cells (Abou-Khalil et al., 2009) and FCp4⁺ fibroblasts (Mathew et al., 2011; Murphy et al., 2011), and blood-borne macrophages (Arnold et al., 2007; Lescaudron et al., 1999; Lu et al., 2011; Sonnet et al., 2006).

At birth, muscle vascularization is rudimentary, with most growing myofibers being unconnected to capillaries during early post-natal development (Sallum et al., 2013), i.e. when mSCs actively proliferate to generate fusion-competent myoblasts for muscle growth (White et al., 2010). By contrast, in adult muscle, mSCs are closely associated to capillary vessels, and reside in a quiescent state in this vascular niche (Christov et al., 2007). The molecular and cellular mechanisms leading to the establishment of a functional vascular niche during development remain poorly defined. During embryonic stages, muscle patterning is governed by large vessel growth-promoting juxta-vascular connective tissue formation at the expense of muscle (Tozer et al., 2007). This progressively splits the initial bud into distinct muscle masses and explains the stereotyped organization of the muscle arterial tree with epimysial arteries feeding arcade arterioles running at the margin of each muscle fascicle (Bloch and Iberall, 1982; Gitiaux et al., 2013). Perimysial arcades produce transverse arterioles that penetrate at a right angle into the fascicle and fork four times before the terminal arteriole level, where six to eight capillaries are formed (Bloch and Iberall, 1982; Gitiaux et al., 2013).

In regenerating muscle, angiogenesis and myogenesis are spatiotemporally coordinated and appear to be regulated by common soluble factors (Christov et al., 2007) such as VEGFA (Williams and Annex, 2004), the master driver of sprouting angiogenesis (Coultas et al., 2005) that also stimulates myogenic cell growth (Chazaud et al., 2003; Christov et al., 2007). Sprouting angiogenesis takes place during late developmental stages and tissue repair (De Smet et al., 2009). In angiogenic sprouts, endothelial tip cells are the leading cells that (1) migrate in response to pro-angiogenic signals, (2) control adjacent ECs in a hierarchical manner to form stalk ECs that maintain connectivity with the parental vessel and (3) release PDGF-B to recruit PDGFR⁺ pericytes to stabilize the endothelial tube (Armulik et al., 2011; De Smet et al., 2009).

This stabilizing effect is mediated by angiopoietin (ANGPT) 1 acting on the TIE2 receptor to induce EC quiescence, survival and strong cell-cell contacts (Armulik et al., 2011). This key signaling pathway for vascular homeostasis (Shim et al., 2007; Thomas and Augustin, 2009; Wakui et al., 2006) also applies to non-vascular cells (Valable et al., 2003). In particular, ANGPT1:TIE2 signaling regulates the hematopoietic stem cell quiescence (Arai et al., 2004). In the same way, mSCs express the receptor TIE2 and respond to ANGPT1:TIE2 signaling, in terms of growth inhibition, pro-survival effects and self-renewal (Abou-Khalil et al., 2009). The exact sources of ANGPT1 that targets mSCs remain ill-defined.

Pericytes are mural cells embedded into the capillary basal lamina. There is no truly specific pericyte marker but pericytes typically express neuroglial 2 proteoglycan (NG2) (Ozerdem et al., 2001). They are detected at highly variable EC:pericyte ratios in different tissues, and are important for angiogenesis, microvasculature structural integrity and blood flow regulation (Armulik et al., 2011). Pericytes in skeletal muscle biology had been neglected until their recognition as mesenchymal stem cells with myogenic potential (Amos et al., 2008; Crisan et al., 2008; Dellavalle et al., 2007). Based on markers and morphology, pericytes are heterogeneous (Bondjers et al., 2006). In particular, muscle pericytes have been subtyped according to the expression of nestin, with type 1 nestin⁻NG2⁺ pericytes, which are profibrotic and adipogenic, and type 2 nestin⁺NG2⁺ pericytes, which have myogenic potential (Birbrair et al., 2013). The stem cell properties of pericytes may have potential therapeutic applications (Dellavalle et al., 2011; Morgan and Muntoni, 2007) and physiological relevance (Dellavalle et al., 2011). Here, we have investigated the role of pericytes in the mSC niche by analyzing their spatiotemporal and functional relationships with mSCs during muscle post-natal growth.

Temporal relationships between post-natal myogenesis and angiogenesis were analyzed on cross-sections of the gastrocnemius muscle of C57BL/6J mice, using immunostaining for PAX7 (mSCs), CD31/PECAM (ECs) and the proliferation marker KI67. The number of CD31⁺ capillaries increased from 8 to 100 per 100 myofibers from P1 to P31. At the same time, myofibers increased in size, their caliber growing from less than 10 µm to over 24 µm from P1-P9 to P15-P31 (Fig. 1A). Meanwhile PAX7⁺ mSCs dramatically decreased from 54 to 6 per 100 myofibers from P1 to P22-P31 (Fig. 1B). PAX7/KI67 double immunostaining showed a decrease in proliferating mSCs from more than 25 PAX7⁺KI67⁺ cells per 100 myofibers at P1 to none at P31, demonstrating the progressive entry of mSCs into a generalized quiescent state (Fig. 1B).

To control the specificity of the NG2 marker in post-natal mouse skeletal muscle, we examined the immunohistochemical expression of NG2, laminin 1, CD34 and CD31 in adult TA and gastrocnemius muscles. NG2⁺ cells were almost always unequivocally identified as pericytes in capillaries and smooth muscle cells in arterioles. We never detected NG2⁺ expression in PAX7⁺ mSCs or in fibroblasts expressing CD34 and devoid of laminin 1⁺ basement membrane. No NG2⁺ cells co-expressed the CD31 EC marker. During post-natal microvascular growth into the endomysium, NG2⁺ cell coverage of ECs remained stable, with a mean of 78 pericyte nuclei per 100 EC nuclei from P1 to P16 (Fig. 1C). The ratio increased at P22 to the adult value of 90±8 per 100, assessing concurrent microvascular vascular growth arrest and maturation (Fig. 1C,D). Triple immunostaining for PDGFRβ/CD146/NG2 compared with single NG2 immunostaining produced similar quantitative results at P1, P3, P5, P8, P16, P22 and P60 (data not shown).

Spatial relationships between pericytes and mSCs were examined using NG2/PAX7 double immunostaining. During post-natal development, pericytes were initially remote from mSCs (Fig. 1E) with only 24% of PAX7⁺ mSCs being at less than 5 µm from a NG2⁺ cell (Fig. 1F). Then pericytes progressively entered in the close vicinity of a growing proportion of mSCs (Fig. 1E), reaching a plateau at the end of muscle microvascular growth with 73% of mSCs being found at less than 5 µm from a pericyte at P22 (Fig. 1F). Strikingly, this novel microvascular organization set up correlates with the progressive establishment of mSCs quiescence (Fig. 1B), and is in line with the non-random association between capillaries and adult mSCs (Christov et al., 2007). Taken together with myofiber growth and angiogenesis kinetics (Fig. 1A), these results indicate that during post-natal days, pericytes associated with growing microvessels are initially remote from actively cycling mSCs, and, as microvessels extend and mature, progressively enter in the mSC vicinity while mSCs either fuse with growing myofibers or become mitotically quiescent.

Progressive reduction of the distance between pericytes and mSCs was similarly observed in humans from 2 months of age to adulthood using double immunostaining for α smooth muscle actin (αSMA), a marker of human pericytes (Ozerdem et al., 2001), and CD56/NCAM, a marker of human mSCs (Fig. 1G). Confocal microscopy allowed 3D visualization of intimate proximity of NG2⁺ pericytes and CD56⁺ mSCs, allowing paracrine signaling (Fig. 1H). Triple immunostaining for NG2, CD56 and merosin/laminin 2, a myofiber basal lamina marker, did not detect conspicuous basal lamina interruption, allowing direct cell contacts in homeostatic muscle (Fig. 1I).

Pericyte subtyping was carried out using nestin immunostaining. Both nestin⁻ and nestin⁺ cells were found among NG2⁺ cells (Fig. 1J), but most adult PAX7⁺ mSCs were associated with nestin⁺ cells (Fig. 1K). Similar proportions of NG2⁺ (72%) and nestin⁺ (76%) cells found in the vicinity of mSCs confirmed that pericytes in mSC niche are type 2 pericytes (Birbrair et al., 2013). On these grounds, co-cultures under conditions avoiding cell-cell contacts and possible confusion with pericyte myogenic differentiation were used to characterize the functional influence of human microvascular cells on human mSC behavior.

Functional interactions between either ECs or pericytes with myogenic cells were first assessed using indirect co-cultures. Human mSC-derived myogenic progenitor cells (MPCs) were conventionally extracted after pronase digestion using anti-CD56-coupled magnetic beads, and plated at normal density (3000 cells/cm²) in the lower chamber of culture wells, whereas the upper compartment, separated by a porous filter, was seeded with human umbilical vein ECs, human microvascular ECs or muscle pericytes. As previously reported (Christov et al., 2007), the two EC types induced strikingly similar increase in mSC-derived MPC growth (data not shown), and, therefore, human umbilical vein ECs were used and referred to as ECs in further experiments. Cells extracted from fresh human muscle tissue explants (Dellavalle et al., 2007), plated three or four times, and sorted as CD146⁺ and CD56⁻ using microbeads, were referred to as muscle pericytes as they expressed NG2, CD146, PDGFR and αSMA, as reported by Crisan et al. (2008), and were negative for CD45 (haematopoietic lineage cells), CD31 (ECs), CD56 (mSCs), desmin (skeletal and smooth muscle cells) and CD34 (fibroblasts and other interstitial cells).

ECs, compared with other MPCs used as controls, showed strong stimulatory effect on MPC growth in the lower chamber (+64% after 10 days of co-culture, P<0.0001) (Fig. 2A,B). By sharp contrast, pericytes significantly decreased MPC growth (−27%, P<0.0001). ECs increased the MPC fusion index in proportion to MPC density after 10 days of co-culture. By contrast, pericytes strongly stimulated myotube formation while decreasing MPC growth, suggesting a specific pro-differentiation effect (Fig. 2A,C). This was supported by RT-qPCR, showing significant upregulation of the myogenic differentiation factors MYOD1 and myogenin (Fig. 2D). Notably, pericytes directly co-cultured with ECs in the upper chamber at a 2:3 ratio, which mimics the in vivo ratio during the post-natal period, had even stronger pro-differentiation effects on MPCs, as assessed by both the increased fusion index (Fig. 2A,C) and upregulation of MYOD1 and myogenin mRNAs (Fig. 2D). Thus, indirect co-cultures suggested that ECs may induce MPC growth, whereas pericytes may exert a dual effect, inhibiting growth and inducing MPC differentiation.

To uncouple potential cell density effect and differentiation, MPCs were cultured at low density (100 cells/cm²) and exposed to EC- or pericyte-conditioned medium (CM) for 21 days. EC-CM increased MPC growth up to 4.3-fold at 4, 10 and 21 days of culture without a concurrent increase in the fusion index. By contrast, pericyte-CM increased MPC differentiation in myotubes by over fourfold at 4, 10 and 21 days of culture (Fig. 2E,F). These results were confirmed by myosin heavy chain staining (Fig. 2G).

Next, individual myofibers were cultured in human EC- or pericyte-CM, using floating conditions that preserve mSCs in their physiological environment beneath the basal lamina in contact with the myofiber (Pasut et al., 2013; Zammit et al., 2004). Mouse myofibers were used because whole myofibers cannot be isolated from human muscle, whereas human recombinant ANGPT1 has been efficiently used on isolated mouse myofibers (Abou-Khalil et al., 2009). Immunolabeling for PAX7 and MYOD1 showed that a 72 h exposure to EC-CM increased the proportion of activated PAX7⁺/MYOD1⁺ mSCs by 25% compared with controls cultured in unconditioned media (UCM). By contrast, pericyte-CM increased the proportion of mSCs at differentiation stage (PAX7⁻/MYOD1⁺) by 19%, at the expense of PAX7⁺/MYOD1⁺ cells that decreased by 25% compared with controls (Fig. 2H,I). In conclusion, ECs support proliferation of activated mSCs, and pericytes promote the differentiation program in a physiologically relevant culture system. The MPC fusion-inducing effect of pericytes was associated with maintenance of a pool of unfused mononucleated cells (Fig. 2A). To examine whether these cells could represent ‘reserve cells’ returned to a dormant state, we used pulse-chase experiments with 5-ethynyl-2′-deoxyuridine (EdU), a thymidine analog that is incorporated into the DNA of dividing cells (Fig. 3A). After a period of chase, the label dilutes in proportion to cell cycle activity: its retention at high concentration being characteristic of cells that have returned to quiescence (Chehrehasa et al., 2009). MPCs cultured in UCM for 4 days and subjected to an EdU pulse showed four levels of EdU label intensity after 2 h. After chases of 4 and 10 days, EdU dilution was observed and cells with very high labeling were not found (Fig. 3B,C). EC-CM induced more rapid EdU dilution compared with controls, with 71.8% and 84.9% of EdU^Low cells at 4 and 10 days of culture, respectively (Fig. 3C). By contrast, exposure to pericyte-CM was associated to conspicuous EdU retention, with decrease of both EdU^Low cells (7.2% at 4 days; 11.9% at 10 days) and EdU^Intermediate cells (24.1% at 4 days; 26.4% at 10 days), and marked increase of both EdU^High cells (37.8% at 4 days; 30.9% at 10 days) and EdU^{Very high} cells (36% at 4 days; 25.7% at 10 days) (Fig. 3C). These results assessed EC-induced mitogenic effect and pericyte-induced cycle arrest.

To examine whether MPC cycle arrest was related to quiescence or to differentiation, EdU detection after 10 days was coupled with various immunostainings: EdU^High and EdU^{Very high} cells expressed PAX7 (82.8%) but only a small minority expressed proliferation and differentiation markers (11.6% KI67⁺, 18.4% MYOD1⁺ and 13.2% myogenin⁺ MPCs), indicating that pericyte-induced EdU-retaining cells were mostly reserve cells (Fig. 3D). In summary, in addition to promoting myotube formation, pericytes induce MPC quiescence.

To detect secreted factors that may mediate EC and pericyte effects on mSCs, 121 cytokines and growth factors were screened by ELISA on 48 h culture supernatants. Compared with UCM, pericyte-CM showed increases in 14 molecules (ANGPT1, TGFβ1, PIGF, IGF1, HGF, MCP1, GRO, angiogenin, MIF, SGP130, TIMP1, uPAR, VEGFA and osteoprotegerin) and EC-CM showed increases in 18 molecules (ANGPT2, PDGF-BB, TGFβ1, TGFβ2, FGF6, EGF, BMP4, MCP1, GRO, angiogenin, IL8, IL1R4/ST2, I-TAC, MIF, SGP130, sRNF RII, TRAIL R3 and uPAR). Several of these factors have previously been shown to exert different effects on myogenic cells. Vascular cell factors with the highest cognate receptor expression by human MPCs were identified using mRNA profiling of 84 receptor genes. Pericyte-CM contained only two factors with receptors highly expressed by MPCs (Table 1): IGF1, a key factor in muscle differentiation and growth (Duan et al., 2010); and ANGPT1, a factor that promotes stem cell quiescence (Abou-Khalil et al., 2009). These factors remained similarly secreted when pericytes were co-cultured with ECs at a 2:3 ratio to mimic the in vivo situation (data not shown). In EC-CM, ANGPT1 and IGF1 were, respectively, undetectable and not significantly increased compared with UCM (Table 1).

Whole-genome transcript analysis of mSCs performed after GFP-based FACS sorting from muscle of PAX3^GFP/+ mice showed that mSCs differentially express receptors of the pericyte products during post-natal development (Fig. 4A-D). Consistent with progressive post-natal mSC entry in quiescence, transcription of IGFR1 and the downstream insulin signaling pathway factors involved in protein synthesis decreased progressively from post-natal P1 to P28, whereas TIE2 transcription progressively increased together with the downstream MAPK pathway molecules previously shown to mediate ANGPT1:TIE2 pro-quiescence effects on MPCs (Abou-Khalil et al., 2009).

Functional involvement of pericyte-released IGF1 and ANGPT1 was analyzed and compared with the two EC factors with the highest level of constitutive expression of cognate receptors in mSCs (Table 1), including ANGPT2, a natural ANGPT1 antagonist that competitively binds TIE2 (Maisonpierre et al., 1997), and PDGF-BB, a mitogenic factor for cells of mesenchymal origin previously shown to stimulate MPC proliferation (Christov et al., 2007). This was achieved by adding blocking agents at saturating conditions to conditioned media added to MPC cultures every 2 days, including specific blocking antibodies against IGF1, PDGF-BB and the TIE2Fc fragment, to block pericyte-released ANGPT1 or EC-released ANGPT2 (Fig. 4E-G). After 10 days, non-significant change was observed in terms of growth, EdU retention and differentiation when human MPCs were grown at normal density in the presence of UCM incubated with the blocking agents, polyclonal IgG or non-specific antibody. By contrast, addition of anti-IGF1 antibody markedly reversed the differentiating effects of pericyte-CM (Fig. 4E); anti-PDGF-BB antibody added to EC-CM significantly decreased MPC growth (Fig. 4F), as previously reported (Christov et al., 2007). TIE2Fc treatment significantly reduced pericyte-induced MPC EdU retention (Fig. 4G). As PC-CM contained tenfold more ANGPT1 than ANGPT2, and no effect was observed in UCM conditions despite similar amounts of ANGPT2, it is likely that the TIE2Fc effect in PC-CM reflected blockade of ANGPT1-induced MPC quiescence. By contrast, EC-CM contained high levels of ANGPT2. As expected from blockade of the ANGPT2 antagonistic action on MPC autocrine ANGPT1:TIE2 signaling (Abou-Khalil et al., 2009), TIE2Fc reduced EC-induced MPC EdU dilution, notably restoring the EdU^{Very high} MPC pool to control values (Fig. 4G). Consistent with this, MPC growth increased upon blockade of ANGPT1 in pericyte-CM and decreased upon blockade of ANGPT2 prominently detected in EC-CM.

The relevance of these elementary findings to post-natal muscle growth was assessed by lineage tracing and specific cell ablation using a pericyte-specific Cre/LoxP system. As NG2 is exclusively expressed by mural cells during vascular morphogenesis (Ozerdem et al., 2001), we first used a NG2-Cre mouse [B6;FVB-Tg(Cspg4-cre)1Akik/J-Jax]. Cre efficiency was assessed by crossing NG2-Cre with R26-Flox-STOP-lacZ reporter (R26R) mice. In TA muscle of adult Tg:NG2^Cre/+::R26R^lacZ mice, 14% of NG2⁺ cells strongly expressed β-gal in their cytoplasm (Fig. 5A). Only 0.7% of myofibers expressed β-gal, which was inconsistent with NG2 transcription in muscle cells during development. mSCs were examined on isolated myofibers immediately (n=233, three mice) or after 72 h of culture in high-activation medium (n=132, three mice). At t0, 972 mSCs were counted, of which 2.44% expressed β-gal. After 3 days in high-activation medium, 1584 clusters of mSC daughter cells were found, including 9.38±3.52 cells per cluster: 3% of clusters were entirely β-gal⁺; only 1.7% showed mixture of β-gal⁺ and β-gal⁻ cells; and 95.3% were totally β-gal⁻, ruling out significant reporter gene expression inherent to myoblast activation. Thus, the very small proportion of β-gal⁺ myofibers illustrated in Fig. 5A likely corresponds to the marginal contribution made by pericytes to post-natal myogenesis [estimated at 0.4-0.6% of myofibers in TA muscle by Dellavalle et al. (2011)]. The negligible proportion of non-vascular cells expressing NG2-Cre in TA muscle led us to cross the NG2-Cre mouse with the Cre-inducible Flox-STOP-diphtheria toxin (DT) receptor (iDTR) mouse, to obtain a model for selective DT-induced depletion of NG2⁺ pericytes and vascular smooth muscle cells. After 48 h of DT injection in TA muscle, adult Tg:NG2^Cre/+::R26R^iDTR mice showed a similarly low number of necrotic myofibers (<2.7%) compared with littermate controls, conspicuous pericyte loss [31 versus 57 (−46%) NG2⁺ cells per 100 myofibers] and mSC increase (4.3 versus 3 PAX7⁺ cells per 100 myofibers). DT was injected into the growing TA muscle of Tg:NG2^Cre/+::R26R^iDTR mice at P13 (Fig. 5B). Eight days later, the injected muscle appeared reduced in size compared with the contralateral TA muscle (Fig. 5C). Compared with littermate controls injected with DT, Tg:NG2^Cre/+::R26R^iDTR mice showed a conspicuous decrease of NG2⁺ pericytes (−29.5%) (Fig. 5D). The small number of myofibers with internal nuclei (due to post-injury regeneration) (0.55-2.20% in Tg:NG2^Cre/+::R26R^iDTR mice versus 0.51-1.48% in littermates) were excluded from morphometric evaluation. Pericyte depletion was associated with significant myofiber hypotrophy (16.3±0.6 versus 18.3±0.6 µm, P<0.05) (Fig. 5E), with a mild increase in total of PAX7⁺ mSCs (16.7%) and marked increase of cycling PAX7⁺KI67⁺ mSCs (+202%) (Fig. 5F). Of note, the amount of pericyte ablation in Tg:NG2^cre/+::R26R^iDTR mice was higher than predicted by lacZ reporter gene expression, and, in addition, it differed in adult and pup experiments. The non-linear relationship between X-gal staining intensity and the level of recombination has been previously documented in different muscle areas, and attributed to differences in activity of the ROSA26 promoter or to the stability of β-gal protein across different muscle areas (McCarthy et al., 2012). Minor iDTR expression at the cell surface may be sufficient to confer sensitivity to DT, whereas minute amounts of β-gal diluted into the cytosol may not be detectable microscopically. The somewhat lower extent of pericyte ablation observed after 8 days in the context of post-natal angiogenesis, compared with adult muscle examined 2 days after DT injection, was likely due to replenishment of emptied pericyte niches by novel cells expressing NG2, as previously reported (Rajantie et al., 2004). Our results therefore show that in vivo ablation of pericytes leads to mSC relief from quiescence.

We next used tissue-specific mouse mutants to characterize the role of the previously identified candidate pericyte effectors. We first crossed NG2-Cre mice with Flox-IGF1 mice to selectively suppress pericyte IGF1 production. At post-natal day P21 (Fig. 6A), mice with Igf1 gene deletion in pericytes compared with their littermates showed a decrease in total body weight (7-8%), TA muscle size, myofiber size (33.08±0.8 versus 29.08±0.7 µm, P=0.0002) (Fig. 6B,C) and number of myonuclei per myofiber (1.48±0.09 versus 2.14±0.09, P<0.0001) (Fig. 6D), supporting decreased mSC fusion as a mechanism of fiber hypotrophy.

ANGPT1 is a major pericyte factor involved in vascular development. To avoid lethality linked to lack of ANGPT1 at pre-natal stages (Jeansson et al., 2011), we looked for an inducible model to assess post-natal effects of ANGPT1 on mSCs. We took advantage of the previously characterized Tissue Nonspecific Alkaline Phosphatase (TNAP) CreERT2 mouse, in which both NG2⁺ pericytes and CD31⁺ ECs, at a ratio of about 1:1, but not mSCs, express the TNAP-Cre (Dellavalle et al., 2011) (Fig. 6E). Because ANGPT1 is secreted by pericytes but not by ECs (Table 1), we crossed TNAP CreERT2 and Flox-Angpt1 mice to generate a model to conditionally suppress pericyte-derived ANGPT1. A series of three daily subcutaneous tamoxifen injections were given at P9, P10 and P11 in Tg:TNAP^CreERT2/+::Angpt1^del/+ mice and littermate controls (Fig. 6F). Similar to wild-type mice (Fig. 1B), control mice showed a 45.9% decrease of PAX7⁺ mSCs from P18 to P28 (10.2±1.1 versus 5.5±0.1 per 100 myofibers; P=0.0003). By sharp contrast, during the same time, Tg:TNAP^CreERT2/+::Angpt1^del/+ mice showed a 25% increase in PAX7⁺ mSCs. This was associated with dramatic increase in the proportion of PAX7⁺KI67⁺ cycling mSCs, from 12.3±1 to 65.3±3.9% (Fig. 6G). The maximum mSC release from quiescence was observed at the time point (P28) associated with the highest expression of TIE2 in mSCs (Fig. 4C). This occurred in the absence of myofiber necrosis, and without concurrent changes in NG2⁺ pericyte and CD31⁺ EC numbers and myofiber size. Together with in vitro experiments, these results demonstrate that pericytes stimulate myofiber growth through IGF1 and control mSC quiescence through ANGPT1.

We show that post-natal angiogenesis is spatio-temporally coordinated with myofiber growth and with mSC entry in quiescence. During angiogenesis, pericytes control EC proliferation and survival (Franco et al., 2011), and basement membrane assembly (Stratman et al., 2009). In addition, pericytes, which are present in the niche of 70-80% of adult mSCs, significantly influenced myogenic cell behavior, through differentiation/growth-promoting effects of IGF1 and quiescence-promoting effects of ANGPT1. Notably, the high pericyte:EC ratio found in adult muscle by far exceeds classical electron microscopy-based estimations (Shepro and Morel, 1993). In contrast to pericytes, ECs were mitogenic for MPCs through ANGPT2 and PDGF-BB, and probably other growth factors (Christov et al., 2007). This indicates that different cell constituents within a stem cell niche exert distinct functions, and suggests fine-tuning of mSC behavior control by microvascular cells during post-natal growth.

The mSC microvascular niche comprises NG2⁺nestin⁺ pericytes. Pro-quiescence effects of pericyte ANGPT1 on mSCs were assessed in vitro and in vivo, using the Tg:TNAP^CreERT2/+ mouse model. Interestingly, lineage studies based on the same mouse model previously showed that TNAP⁺ pericytes have myogenic potential (Dellavalle et al., 2011), and, consistently, type 2 NG2⁺nestin⁺ pericytes detected in the mSC microvascular niche were previously characterized as mesenchymal stem cells with prominent myogenic potential (Birbrair et al., 2013). Taken together, these data strongly suggest that ‘stem cells support other stem cells’ in skeletal muscle (Leatherman, 2013). In at least three different adult stem cell niches – the Drosophila testis, the hair follicle and the bone marrow stem cell niches – two separate populations of stem cells have been identified, with one cell type producing factors that contribute to the maintenance of the other one (Leatherman, 2013; Méndez-Ferrer et al., 2010). Consistent with our study, supportive cells in the hematopoietic stem cell niche have been characterized as NG2⁺nestin⁺ smooth muscle vascular cells (Kunisaki et al., 2013).

Our results may be of practical interest as mesenchymal stem cells are increasingly used for therapeutic purposes and most functional benefits of these therapies have been linked to ill-defined trophic effects on existing tissue cells rather than to the expected differentiation of the grafted cells. We showed that pericyte-released IGF1 and ANGPT1, acting on adjacent muscle stem cells, may exert tight regulation of their post-natal fate, but the signals that may orient pericytes towards myogenic cell differentiation or a purely supportive action are unknown. To what extent pericytes may be instrumental in adaptive responses to disruption of local homeostasis, systemic signals and aging also remains to be delineated (Rojas-Ríos and González-Reyes, 2014).

Normal deltoid muscle samples from 18- to 60-year-old patients were used to obtain human mSCs or pericytes. Normal deltoid muscle biopsies from patients aged from 2 months to 21 years were used to establish post-natal evolution of the spatial relationships between pericytes and mSCs. In accordance with the Henri Mondor hospital research ethics committee, all adult patients or the parents of patients under 18 years of age gave written informed consent for participation in the study.

B6;FVB-Tg(Cspg4-cre)1Akik/J (Zhu et al., 2008), C57BL/6-Gt(ROSA)26Sortm1(HBEGF)Awai/J (Buch et al., 2005), B6.129(FVB)Igf1tm1Dlr/J (Liu et al., 1998), B6.129S4-Gt(ROSA)26Sortm1Sor/J (Soriano, 1999) (Jax Mice), Angpt1tm1.1Seq (Jeansson et al., 2011) (a gift from Susan E. Quaggin, Northwestern University, Evanston, IL, USA), Tg(Alpl-cre/ERT2)12Gcos (Dellavalle et al., 2011) (a gift from G. Cossu, UCL, London, UK and A. Dellavalle, InScientiaFides, San Marino Republic) and C57BL/6J were bred and crossed in our facilities. The Tg:NG2^Cre/+::R26R^iDTR, Tg:NG2^Cre/+::IGF1^del/+ and Tg:NG2^Cre/+::R26R^lacZ mice were generated by crossing NG2-Cre either with R26RDTR^loxP/loxP or R26RlacZ^{stoploxP/stoploxP} animals, with PCR assessment of both transgenes. The Tg:TNAP^CreERT2/+::Angpt1^del/+ was generated by crossing TNAP-CreERT2 with Angpt1^loxP/loxP animals with assessment by PCR for both transgenes. Genotyping strategies have been published for TNAP-CreERT2 (Dellavalle et al., 2011) and Angpt1^loxP/loxP (Jeansson et al., 2011), and were supplied by Jax Mice for NG2-Cre, R26RDTR^{stoploxP/stoploxP}, R26RlacZ^{stoploxP/stoploxP} and Igf1^loxP/loxP. The French Ethical Committee approved animal experiments (license number 11-00010).

For pericyte depletion, adult or P13 mice were injected in the TA muscle with DT (D0564; Sigma-Aldrich) diluted in physiological serum at 1 ng/g body weight in adult mice (10 µl/TA) and at 3 ng/g in young mice (2 µl/TA). Contralateral muscle was injected with an equal volume of serum. Mice were sacrificed 8, 24 and 48 h after injection (adult mice), and at P21 (young mice). For Angpt1 gene deletion, young mice were injected subcutaneously (50 µl) with 0.25 mg tamoxifen (T5648-1G; Sigma-Aldrich) diluted in corn oil (C8267; Sigma-Aldrich) at P9, P10 and P11, and sacrificed at P18 and P28.

All cell types were cultured at 37°C with 5% CO₂. To obtain MPCs, human mSCs were extracted from deltoid muscle samples after digestion by Pronase E (P5147; Sigma-Aldrich) at 1.5 mg/ml. After a few days of amplification, MPCs were purified using NCAM/CD56 microbeads according to the kit protocol (130-050-401; Miltenyi Biotec) and amplified in HAM F-12 medium (31765027; Life Technologies) supplemented with 20% fetal calf serum (FCS) (S1810-500; Dominique Dutscher), 1% penicillin/streptomycin (P/S; 15140122; Life Technologies), 0.2% MEM vitamins solution (11120037; Life Technologies), 2 mM non-essential amino acids (1114003; Life Technologies) and 1% Ultroser G serum substitute (15950-017; Pall Corporation).

Pericytes were extracted by explant from deltoid muscle samples (Tonlorenzi et al., 2007). After 2 weeks of culture, pericytes were purified using CD56 and CD146 microbeads (130-050-401 and 130-093-596; Miltenyi Biotec) and then amplified in M5 medium before use (Tonlorenzi et al., 2007). Purity was >98%, as assessed by both positive (CD146⁺NG2⁺αSMA⁺) and negative markers (CD45⁻CD31⁻CD34⁻desmin⁻).

Both human umbilical vein endothelial cells (C-12203; PromoCell) and human microvascular endothelial cells (C-12210; PromoCell) were cultured in endothelial cells basal medium (ECBM; C22010; PromoCell) supplemented with endothelial cell growth medium supplement mix (C-39215; PromoCell). In some experiments, ECBM supplemented with 2% FCS and 1% P/S was called unconditioned medium (UCM) and used to prepare 48 h culture-conditioned media of EC (EC-CM) or pericyte (PC-CM).

In transwell co-cultures, all cell types were grown in UCM renewed every 2 days. For 10 days, MPCs (3000 cells/cm²) in the lower well were co-cultivated with control MPCs (67,000 cells/cm²), pericytes (67,000 cells/cm²), ECs (100,000 cells/cm²) or pericytes+ECs (67,000+100,000 cells/cm²) (upper insert reference: 353095; BD Falcon), then used for RT-qPCR or stained by May Grunwald Giemsa to assess MPC growth and differentiation. In low-density cultures, MPCs (100 cells/cm²) were incubated with EdU (Click-iT EdU Alexa Fluor 488 Imaging Kit; C10337; Invitrogen) for 2 h and then cultured for 4, 10 or 21 days in the different CM renewed every 2 days (Chehrehasa et al., 2009).

Single myofibers were isolated from C57BL/6J EDL or Tg:NG2^Cre/+::R26R^lacZ TA muscles (Beauchamp et al., 2000; Rosenblatt et al., 1995). Individual intact myofibers were incubated with UCM, pericytes-CM or EC-CM mixed (vol/vol) with low-activation medium [10% horse serum; 26050088; Life Technologies), 0.5% chicken embryo extract (CE-650-J; Seralab) and DMEM (11965092; Life Technologies)]. After 72 h of incubation, myofibers were used for PAX7/MYOD1 immunostaining as previously described (Guardiola et al., 2012; Zammit et al., 2004). The myofibers isolated from Tg:NG2^Cre/+::R26R^lacZ TA muscle were stained with X-gal just after extraction or after 72 h of culture in high-activation medium (20% FCS, 1% chicken embryo extract and DMEM).

Muscle sections or single myofibers were fixed in 0.2% glutaraldehyde for 5 min at room temperature, washed and then incubated overnight at 37°C in X-gal solution [400 µg/ml of X-gal (B4252; Sigma-Aldrich), 0.02% of Tween-20 (P9416; Sigma-Aldrich), 2 mM of MgCl₂ (Y02016; Life Technologies), 4 mM of potassium ferricyanide (26810.232; Prolabo), 4 mM of potassium ferrocyanide (P9387; Sigma-Aldrich) in PBS (14190094; Life Technologies)]. After incubation, muscle sections were mounted using Eukitt (O. Kindler), single myofibers were mounted with Vectashield Mounting Medium with DAPI (H-1200; Vector) and both were observed through a Zeiss microscope.

Frozen sections (7-10 μm) and cultured cells were fixed with 4% PFA at room temperature for 5 min, permeabilized with 0.5% Triton X-100 for 5 min and blocked with 10% BSA for 30 min. Samples were incubated with primary antibody overnight at 4°C followed by secondary antibodies for 45 min at 37°C, then fluorochrome-conjugated streptavidin for 45 min at 37°C when necessary, and finally mounted with Vectashield Mounting Medium with DAPI and observed through epifluorescence or confocal Zeiss microscope.

The immunofluorescence staining was carried out with primary antibodies recognizing NG2 (AB5320; Chemicon Millipore; 1:200), αSMA (ab5694; Abcam; 1:100), PDGFRβ (#3169; Cell Signaling; 1:100), CD146 (ab75769; Abcam; 1:100), CD31 (ab7388; Abcam; 1:50 or M0823; Dako; 1:50), CD34 (14-0341-82; eBioscience; 1 µg/100 µl), CD45 (550539; BD Pharmingen; 1:50), NCAM/CD56 (123C3; Monosan; 1:50), PAX7 (DSHB; 1:50), MYOD1 (M3512; Dako; 1:50), myogenin (M3559; Dako; 1:50), MHC (MAB4470; R&D Systems; 1:20), KI67 (ab16667; Abcam; 1:500), nestin (ab81755; Abcam; 1:500), desmin (ab8592; Abcam; 1:100), laminin α1 (ab14055; Abcam; 1:500) and merosin/laminin α2 (MAB1922; Chemicon Millipore; 1:200). PAX7 immunostaining of mSCs was performed according to the M.O.M. Kit (BMK-2202; Vector). Visualization was carried out using the following secondary antibodies: Cy3-conjugated anti-mouse (715-166-150; Jackson), FITC-conjugated anti-rabbit (711-166-152; Jackson), Alexa Fluor 647 anti-chicken (A21449; Life Technologies) and biotinylated anti-rat (BA9401; Vector) coupled to Cy3-conjugated streptavidin (016-160-084; Jackson), a DAB detection Kit (SK-4100; Vector) or a Red Alkaline Phosphatase substrate kit (SK-51000; Vector).

Total RNAs were extracted from human cells by RNeasy Micro Kit (74004; Qiagen) and converted into double-strand complementary DNA using SuperScript III RT (18080-093; Life Technologies), according to the manufacturer's instructions. Real-time quantitative PCR was carried out on cDNA using 7900HT Fast Real-Time PCR System (Applied Biosystems). Each cDNA sample was amplified in triplicate by using the Platinum SYBR Green qPCR SuperMix-UDG (11733-038; Life Technologies). Analysis was carried out with the following primers: human β-actin (forward, CTGGGACGACATGGAGAAAA; reverse, AAGGAAGGCTGGAA-GAGTGC), human PAX7 (forward, CCCCCGCACGGGATT; reverse, TATCTTGTGGCGGATGTGGTT), human MYOD11 (forward, GACGG-CTCTCTCTGCTCCTTT; reverse, TCGAAACACGGGTCGTCATA) and human myogenin (forward, CACCCTGCTCAACCCCA; reverse, CACTGCCCCACTCTGGACT).

One hundred and twenty cytokines were analyzed by protein array on 48 h-conditioned media from EC, pericytes and pericytes+EC cultures (triplicate for each medium and cytokine), according to the manufacturer's instructions accompanying the RayBio Human Cytokine Antibody Array (AAH-CYT-G1000; RayBiotech). ANGPT1 and ANGPT2 were analyzed by ELISA (DANG10 and DANG20, respectively; R&D Systems). ANGPT2 was also analyzed by the RayBio protein array and ANGPT1 levels could be converted into RayBio arbitrary units using a rule of 3.

RNA profiling was carried out on human MPC using cDNA Array (R&D Systems) of 397 genes (Christov et al., 2007) and on post-natal PAX3^GFP/+ mSCs using Affymetrix Mouse Genome 430 2.0 Arrays. The data are available at GEO under accession GSE65927. To analyze the pathway enrichment, Gene Ontology (http://www.geneontology.org) and the murine KEGG pathways (www.genome.jp/kegg) were used.

The number of biological and technical replicates per point was always n=3 or more when indicated. All data are expressed as mean±s.e.m. Depending on the experiments, unpaired Student's t-test and ANOVA tests were used. P<0.05 was considered significant.

We thank Barbara Gayraud-Morel, Pierre Rocheteau and Fabrice Chrétien (Pasteur Institute, France) for useful advice.