Notch signaling regulates myogenic regenerative capacity of murine and human mesoangioblasts

Somatic stem cells hold attractive potential for the treatment of muscular dystrophies (MDs). Mesoangioblasts (MABs) constitute a myogenic subset of muscle pericytes and have been shown to efficiently regenerate dystrophic muscles in mice and dogs. In addition, HLA-matched MABs are currently being tested in a phase 1 clinical study on Duchenne MD patients (EudraCT #2011-000176-33). Many reports indicate that the Notch pathway regulates muscle regeneration and satellite cell commitment. However, little is known about Notch-mediated effects on other resident myogenic cells. To possibly potentiate MAB-driven regeneration in vivo, we asked whether Notch signaling played a pivotal role in regulating MAB myogenic capacity. Through different approaches of loss- and gain-of-function in murine and human MABs, we determined that the interplay between Delta-like ligand 1 (Dll1)-activated Notch1 and Mef2C supports MAB commitment in vitro and ameliorates engraftment and functional outcome after intra-arterial delivery in dystrophic mice. Furthermore, using a transgenic mouse model of conditional Dll1 deletion, we demonstrated that Dll1 ablation, either on the injected cells, or on the receiving muscle fibers, impairs MAB regenerative potential. Our data corroborate the perspective of advanced combinations of cell therapy and signaling tuning to enhance therapeutic efficaciousness of somatic stem cells.

Notch signaling consists of a conserved pathway, triggered by physical interaction between one ligand and one receptor, both transmembrane proteins exposed by contacting cells. 1 Notch signaling has been involved in different stages of muscle formation 2 and regeneration. 3,4 The canonical signaling encompasses five ligands (Dll1/3/4, Jagged1/2) and four receptors (Notch1-4); however, the axis Dll1-Notch1 appears consistently involved during myogenic fate specification, for example, neural crest-driven somite maturation. 5 Moreover, murine embryos expressing a hypomorphic allele of the Notch ligand Dll1 displayed marked impairment of skeletal muscle formation. 6 Interestingly, the Notch pathway may exert different effects according to the cell context. Culture on DLL1-coated plastic improved ex vivo proliferation and in vivo engraftment of canine satellite cells. 7 Expression of the active Notch1 intracellular domain (NICD) robustly committed murine and rat mesenchymal stem cells toward the myogenic fate both in vitro and in vivo. 8 However, Notch-mediated effects on the regenerative potential of non-satellite resident myogenic cells are still unknown.
Mesoangioblasts (MABs) are non-satellite resident myogenic stem cells, able to circulate and regenerate dystrophic skeletal muscles. 9,10 HLA-matched MABs are currently under phase 1 clinical study on Duchenne muscular dystrophy patients (EudraCT #2011-000176-33). In this view, understanding the cell-specific effects and mechanisms of myogenic cues will help improving clinical translation of MABbased therapies in vivo. Recently, it has been shown that Notch synergizes with Pdgf-bb to convert fetal myoblasts into myogenic pericytes. 11 However, knowledge about Notchtriggered effects on the regenerative potency of somatic MABs is still scant, particularly in the contexts of cell-cell (in vitro) and fiber-cell (in vivo) contact.
Therefore, we asked whether the Dll1-Notch1 axis regulates the myogenic potential of murine and human MABs and how to tune this pathway to ameliorate in vivo MAB-driven regeneration.

Results
Dll1-Notch1 regulate in vitro differentiation of murine and human MABs. For this study, we used murine 12 and human MABs, 13 both isolated from somatic muscles as alkaline phosphatase + (AP + ) cells and processed as previously reported 14 (Supplementary Figure 1). Consistently, with antecedent reports, AP + murine MABs do not spontaneously undergo myogenic differentiation in vitro, unlike human AP + MABs 9,10 (Supplementary Figure 2). We analyzed the pattern of Notch1, Dll1 and Hes1 expression and of pathway activation during spontaneous in vitro differentiation. Murine MABs showed slight upregulation of Notch1 at day 3, but progressive downregulation of Dll1 and Hes1 over time, matching Dll1 and NICD (activated receptor) protein levels (Figures 1a and b; Supplementary Figure 3a). Conversely, human MABs showed early upregulation of NOTCH1, DLL1 and HES1, paralleled by increasing protein levels of DLL1 and NICD at day 3, followed by rapid decrease (Figures 1c  and d; Supplementary Figure 3a). To investigate the role of epigenetics in the species-specific pattern of Dll1 expression, we monitored the DNA methylation propensity along the regulatory regions of the genomic Dll1/DLL1 locus. Interestingly, Dll1 regulatory regions appeared more rapidly methylated in murine than in human MABs (Figures 1e and f), being the overall methylation propensity along the murine sequences significantly higher at day 0 and 3 than in the human DLL1 locus (Supplementary Figure 3b). Thus, during spontaneous differentiation in vitro, murine MABs do not spontaneously mature into myocytes and exhibit rapid locus methylation and decreased expression of Dll1 at early stage, correlating with decreased Notch1 activation. Conversely, human MABs mature into myocytes in vitro and show Figure 1 The axis Dll1-Notch1 regulates the in vitro myogenic potential of murine and human MABs. (a-d) Expression levels of Notch1/NOTCH1, Dll1/DLL1 and Hes1/HES1 (Notch1 activation reporter), and WB analysis of Notch1/NOTCH1, its activated form (Notch1/NICD) and Dll1/DLL1 in murine and human MABs over time during spontaneous differentiation in vitro. Expression levels are reported in arbitrary units (AU) as fold change versus day 0. (e, f) Summary array of methylation propensity along upstream CpG islands (one in the murine locus, two in the human), 5′-UTR and 3′-UTR of Dll1/DLL1 genomic loci in murine and human MABs at day 0, day 3 and day 5 of spontaneous differentiation, as assayed by qPCR-based test on genomic fragments enriched in highly methylated DNA. − , unmethylated control, APC promoter; +, methylated control and propensity reference, NBR2 promoter. (g) Immunofluorescence staining on co-cultures of C2C12 myoblasts with murine MABs transduced with lentiviral vectors carrying a GFP tracer and scramble, or anti-Notch1, or anti-Dll1 interfering shRNAs. White arrows indicate chimeric GFP + /MyHC + myotubes and MAB myogenic differentiation. Scramble and scramble +γ-secretase inhibitor (gsi) conditions represent the controls of unperturbed and chemically inhibited signaling, respectively. (h) Immunofluorescence staining on differentiated human MABs after transduction with lentiviral vectors carrying a RFP tracer and scramble, or anti-NOTCH1, or anti-DLL1 interfering shRNAs. Presence of MyHC + mono/bi-nucleated myocytes indicate MAB myogenic differentiation. (i, j) Immunofluorescence staining to evaluate the dose-dependent effects of transient Dll1/DLL1 overexpression on the in vitro myogenic differentiation of murine and human MABs, both transduced with lentiviral vectors carrying conditional expression of the ligand under doxycycline (doxy) control; − , 0 μg/ml, basal control; +, 0.1 μg/ml; ++, 1 μg/ml; +++, 10 μg/ml. To confirm Notch signaling involvement in the observed effect, gsi-supplemented control conditions are also shown. To assess the myogenic contribution of murine MABs in co-culture with C2C12, GFP + murine MABs have been used in this experiment. Data in charts are depicted as mean ± standard deviation of ≥ 3 independent experiments. Scale bars indicate 100 μm Notch signaling and mesoangioblast potency M Quattrocelli et al sustained Dll1 expression and Notch1 activation at early differentiation step. We then investigated the effects of loss and gain of Notch signaling on the in vitro myogenic capacity of MABs. This was assayed in murine MABs by co-culture/fusion with C2C12 myoblasts, whereas human MABs were assayed in spontaneous differentiation toward myocytes. Specific knockdown of Notch1 and Dll1 (Supplementary Figure 4a), as well as treatment with γ-secretase inhibitor (gsi), resulted in decreased NICD levels ( Supplementary Figure 4b), and in impairment of murine MAB fusion with myoblast-derived myotubes, as compared with scramble control (Figure 1g). Analogously, gsi and NOTCH1/DLL1 knockdown led to impaired maturation of human MABs into MyHC + myocytes ( Figure 1h). We also analyzed the involvement of other receptors and ligands, for example, Notch2 receptor and Jag1 ligand, in our in vitro system. Knockdown of Notch2/ NOTCH2 or Jag1/JAG1 did not significantly alter MAB myogenic differentiation, as compared with scramble control (Supplementary Figure 5), supporting the hypothesis that the Dll1-Notch1 axis is pivotal in our experimental system. We next used a lentiviral system of doxycycline-triggered Dll1/ DLL1 overexpression to evaluate the effects of gain-offunction in the early differentiation step. After application of doxycycline till day 3 of differentiation, Dll1/DLL1 overexpression dose-dependently increased in vitro myogenic maturation of both murine and human MABs (Figures 1i and j). This effect was significantly decreased in the presence of gsi (Supplementary Figure 6), thus confirming the direct involvement of Dll1-triggered Notch signaling in the early differentiation step of murine and human MABs.
As additional proof of principle, we sought to apply our dosedependent system of transient Dll1 overexpression to a MAB model of robust myogenic differentiation. To this purpose, we used murine Sgcb-null MABs, previously reported as aberrantly and spontaneously myogenic. 15,16 Transient overexpression of Dll1 dose-dependently correlated with increased Hes1 expression at day 3 ( Figure 2a) and enhanced myotube formation at day 7 ( Figure 2b). Thus, Dll1-Notch1 signaling during the early differentiation/commitment step sustains MAB myogenic differentiation at a later stage.
We asked whether Mef2C and Maml1, reported interactors of NICD in other myogenic cell models, 17,18 were involved in Dll1dependent effects on MAB differentiation. Murine MABs showed progressive decline of gene and protein levels of Mef2C over time, whereas MEF2C was upregulated at days 3-5 in human MABs. Moreover, the levels of Maml1/MAML1 appeared stable over time in both murine and human MABs (Figures 3a and d and Supplementary Figure 7a). Notably, at day 3 of spontaneous differentiation, Mef2C and Maml1 are only partially associated to NICD and Dll1 knockdown resulted in major dissociation from the residual NICD levels. Conversely, Dll1 overexpression led to complete Mef2C recruitment, whereas only after combined overexpression of Dll1 and Mef2C did Maml1 appear completely associated to the NICD complex ( Figure 3e). Because Maml1 was reported as activator of Mef2C upon dissociation from NICD, 17 we quantified the effects of Maml1 recruitment on Mef2C activity in our cell system at day 5 of differentiation, when NICD levels are depleted in normal conditions. Using a Mef2C-responsive luciferase system, Mef2C transcriptional activity in murine MABs was decreased after Dll1 knockdown and significantly increased after combined overexpression of Dll1 and Mef2C (Figure 3f). We obtained analogous data also from human MABs (Figure 3g and h). Thus, the Dll1-Notch1 axis appears as a major regulator of the in vitro myogenic ability of both murine and human MABs, and Dll1-triggered signaling cooperates with Mef2C and Maml1 in enhancing Mef2C activity at later stages.
Priming murine and human MABs for enhanced in vivo regeneration. Subsequently, we sought to determine whether tuning the signaling without genomic integrations could ameliorate the in vivo regenerative potential of murine and human MABs. We used an adenoviral-based, nonintegrative system of temporary overexpression of Dll1 or Mef2C (Ad-Dll1 and Ad-Mef2C, respectively, Supplementary  Figure 10 and data not shown). Comparably, we tested the adenoviral priming system on human MABs and assessed their in vivo myogenic potential by bilateral intra-arterial delivery in acutely damaged muscles of Rag2null/γc-null immunodeficient mice. 19 Similarly to murine MABs, combined priming of human GFP + MABs with Ad-  14 We then tested their myogenic potential in vitro by fusion with C2C12-derived myotubes, and in vivo by intra-arterial delivery into Sgca-null mice ( Figure 5a). After tamoxifen-driven Dll1 knockout, flx GFP + MABs showed significant impairment in the contribution to GFP + myotubes in vitro ( Figure 5b) and to Sgca + /GFP + fibers in vivo (Figures 5c and d and Supplementary Figures  13a and b), as compared with vehicle-treated cells. Interestingly, Dll1 knockout in homing flx MABs resulted also in significant reduction of the functional outcome at 4 and 8 weeks post injection, as assayed by gait analysis and treadmill test (Figure 5e and Supplementary Figure 13c). To validate our transgenic model, we also assayed flx satellite cells, which, as expected, 20 showed reduced levels of Pax7 and increased spontaneous differentiation after tamoxifen addition ( Supplementary Figures 14a and b). Subsequently, we asked whether Dll1 presentation by the host fiber is also We therefore tested the engraftment levels of wildtype GFP + MABs after intra-arterial delivery into Dll1-knockout adult flx muscles (Figure 5f). In the absence of acute damage, tamoxifen-treated mice did not present significant alterations or lesions in skeletal muscles, as compared with vehicle-treated controls (Supplementary Figure 14c). In Dll1-knockout muscles, the rate of MAB engraftment was significantly reduced as compared with vehicle-treated controls, as shown by immunostaining, qPCR and WB (Figures 5g and i). Thus, Dll1 presentation by both the homing cell and the host fiber appears a positive regulator of MAB commitment and engraftment.  The directionality of Dll1 requirement for MAB homing and differentiation in vivo constitutes an interesting question, considering that Notch signals have been directly linked to the niche engraftment by satellite cells. 22 We addressed this point by means of a transgenic system of conditional Dll1 deletion. Together with the data previously discussed, the results obtained from flx MABs and flx muscles suggest a potential, yet incomplete, model of the basal role of Notch on MAB engraftment and differentiation in vivo (Figure 6b). According to this model, Notch1 activation in the homing MABs by Dll1-exposing cells or fibers may result in NICDmediated recruitment of Mef2C and Maml1. Once engrafted in the fiber, potentially also through Dll1-mediated interactions, NICD levels would decrease in the absence of receptor stimulation and Mef2C would then exert its transcriptional activity, promoting the expression of mid-late myogenic factors. Noteworthy, this model might also point to Notchbased recruitment of MABs by Dll1-exposing activated satellite cells 23 in the context of muscle injury.
Our study cannot exclude the contribution of other ligands or receptors of the Notch family, probably accounting for the remaining engraftment of Dll1-knockout MABs or in Dll1depleted muscles. However, other interactions appeared less significant than the Dll1-Notch1 axis in our experimental setup, considering that knockdown of Notch2/NOTCH2 or Jag1/JAG1 did not impair the myogenic ability of murine and human MABs in vitro. Furthermore, our in vivo data indicate that ligand-based tuning of the Notch pathway and its interactors may constitute a feasible integration-free strategy to potentiate the outcome of intra-arterial MAB therapy, possibly in combination with approaches to ameliorate microcirculation in dystrophic muscles. 24 To potentially accelerate clinical translation of such approaches, it will be fundamental to MAB-specifically assess how the Notch pathway is involved in the interaction with other resident cell types and how it intertwines with other cascades, such as Wnt and Bone morphogenetic protein pathways.
For qPCR analyses, RNA was isolated through RNA mini kit, removing gDNA traces by Turbo DNase. One microgram RNA was reverse-transcribed by means of SuperScript III kit and qPCR was performed in 384-well plates (10 μl final volume; thermal profile, 95°C 15′′ − 60°C 45′′ (×50); ViiA 7 qPCR plate reader), using Platinum Sybr Green Mix, 1 μl 1 : 5 diluted cDNA and 100 nM primers (all kits, reagents and plate reader by Life Technologies). Pgk/PGK was used as internal normalizer. The list of primers can be found in the Supplementary Materials and Methods.
DNA methylation assay. CpG islands were defined by submitting the 10 kb gDNA sequence upstream of the transcription start to CpG island searcher (http:// cpgislands.usc.edu; lower limits, CpG% = 55, ObsCpG/ExpCpG = 0.65, length = 500, distance = 100). CpG island in murine Dll1 locus, − 4257 to − 3258 bp from transcription start. CpG islands in human DLL1 locus, − 5909 to − 3112 bp and − 503 to − 0 bp from transcription start. gDNA was isolated from 10 6 cells in proliferative conditions by means of genomic DNA Mini kit (Life Technologies) and 1 μg gDNA was randomly fragmented into fragments of 200-300 bp in 50 μl TE buffer in Bioruptor sonication bath (Diagenode, Liege, Belgium) at 4°C for 15 cycles (30′′ sonication/30′′ rest) at high intensity. Sheared gDNA (200 ng) was then enriched for fragments containing ≥ 5 me-CpGs using MethylCollector Ultra kit (Active Motif, Carlsbad, CA, USA; low salt conditions) and purified via MinElute Reaction CleanUp kit (Qiagen, Venlo, The Netherlands). Purified DNA was then assayed for specific amplification of sequential fragments of regulatory regions through Sybr green-based qPCR, using APC promoter (negative control, unmethylated) and NBR2 promoter (positive control, methylated) as standards for relative quantification of the methylation propensity of the single fragments. The list of primers can be found in the Supplementary Materials and Methods.
Knockdown, overexpression and luciferase reporter vectors. For knockdown of Dll1/DLL1 and Notch1/NOTCH1, specific shRNA-mimicking oligonucleotides were PCR-amplified and cloned at the 3′ of the fluorescent tracer into pGIPZ (efficient in murine MABs; reporter, GFP) or pTRIPZ (efficient in human MABs; reporter, RFP) backbones (Open Biosystems, Huntsville, AL, USA). Vectors carrying the scramble shRNA control were purchased. Knockdown vectors were then used to produce lentiviral particles in 293T cells, using 2nd generation packaging plasmids. At 24 h after application of the viral supernatant, transduced MABs were sorted for the tracer and then plated for knockdown efficiency check and for differentiation. MABs pre-treated for 48 h with medium supplemented with 1 : 500 γ-secretase inhibitor X (Millipore) were also then incorporated in the differentiation experiment as control of chemical inhibition of the signaling. The list of oligonucleotides can be found in the Supplementary Materials and Methods.
For doxycycline-driven overexpression of Dll1/DLL1 and Mef2C/MEF2C, fulllength cDNAs were purchased as in between recombining sequences compatible for Gateway system (Genecopoeia, Rockville, MD, USA) or cloned into pENTR-11 shuttle vector (Life Technologies) from non-compatible plasmids (Open Biosystems). cDNA sequences were then recombined into pLOVE lentiviral backbone using LR Clonase II kit (Life Technologies), according to manufacturer's instructions. Lentiviral particles were produced in 293T cells, using 3rd generation packaging plasmids (5 ml viral suspension/75 cm 2 293T cells). Two milliliters of viral suspension were used to transduce 5 × 10 5 cells. Twenty-four hours after transduction, doxycycline was added for 72 h according to the different concentrations, and then removed during the second part of differentiation. In case of gsi-treated control, gsi was added in combination with the doxycycline. Doxycycline-driven overexpression of Dll1/DLL1 and Mef2C/MEF2C has been used for all in vitro experiments reported in Figures  1,2,3  For the adenoviral-based strategy, cDNAs in pENTR-11 shuttle vector were recombined, as mentioned before, into pAd-CMV-V5-DEST, whereas pAd-CMV-V5-GW-lacZ was used as Ad-mock control (both vectors by Life Technologies). Adenoviral particles were produced by transfecting PacI-linearized pAd vectors into 25 cm 2 293A cells (final volume = 2 ml). Upon 100% mortality, I supernatant was used to transduce 150 cm 2 293A cells (final volume = 10 ml). Upon 100% mortality, II supernatant was frozen in 2-ml aliquots and stored at − 80°C. One vial was used to transduce 5 × 10 5 proliferating MABs and, after 48 h, cells were carefully washed, collected and resuspended in the appropriate number/volume conditions for injection.
In vivo experiments and evaluation of the functional outcome. All animal protocols were conducted in compliance with Ethical Committee Guidelines of KU Leuven (project 095/2012) and Belgian legislation. Sgca-null (C57/Bl6 background) dystrophic mice (6-months old) were generated by the group of Prof. K.P. Campbell (University of Iowa, IA, USA). 26 Rag2-null/γc-null immunodeficient mice (2-months old) were provided by the group of C. Verfaillie (KU Leuven, Belgium). Animals were anesthetized with isofluorane. Bilateral intra-femoral artery injection was performed with 2.5 × 10 5 cells/50 μl saline solution supplemented with 1 : 10000 heparin (LEO Pharma, Ballerup, Denmark)/femoral artery, using 32-gauge needles under STEMI SV11 stereomicroscope (Zeiss, Oberkochen, Germany). Mice were kept under cyclosporine (Sandimmune Cyclosporine, Novartis, Basel, Switzerland; 10 mg/kg) regimen during the whole treatment. Functional outcome was measured through gait analysis 27 and treadmill assay. 28 Both assays were first tested and validated comparing age-matched, background-matched mice to dystrophic mice (data not shown). Gait analysis was performed inking the back paws and letting the mice freely walk along a 1-m-long paper ribbon, confined into a walled plastic path. More than three runs per time point and ≥ 25 stride length measurements were analyzed per mouse/time point. Treadmill assay was performed on 10°uphill-oriented treadmill belt, with 10 m/min starting speed and 1 m/min 2 acceleration. Mice were stopped after ≥ 5 consecutive seconds on the pulse grills.
Conditional Dll1 knockout was performed by means of seven intra-peritoneal injections (every second day) of 3 mg tamoxifen dissolved in 50 μl corn oil (both reagents by Sigma-Aldrich) into 1-month-old flx females; allele removal was checked by PCR on 1 ng gDNA 5 days after last injection (primers, Fw 5′-accttctttcgcgtatgcctcaag-3′, Rev, 5′-agagtctgtatggagggcttc-3′). Conditional Dll1 knockout in cells was performed by adding 10 μM 4-OH-tamoxifen (Sigma-Aldrich), dissolved in ethanol, to the growth medium for 5 days consecutively. Once PCRchecked for allele removal, Dll1-knockout cells were then used for in vitro differentiation or in vivo injection.
Statistical analysis. Sample size for in vitro/in vivo experiments was calculated by means of Sample Size Calculator (http://www.stat.ubc.ca/~rollin/ stats/ssize/index.html; parameters: power,.80; alpha,.05). When applicable, sample size analysis was based on average values obtained from preliminary optimization/ validation trials. To analyze data pools from methylation propensity and qPCR assays, one-way ANOVA (to test difference among 42 pools) and unpaired t-test (to compare two specific pools) were used. Significance was achieved when Po0.05 in both tests. To analyze data pools of luciferase activity, protein levels, functional outcome, fiber count, fusion index and Pax7 + nuclei count assays, Kruskal-Wallis and Mann-Whitney U test were used. Significance was accepted when Po0.05 was scored in both tests. All statistical tests were performed by means of Prism software (GraphPad, La Jolla, CA, USA).