Osteogenic differentiation of skeletal muscle progenitor cells is activated by the DNA damage response

Heterotopic ossification (HO) is a pathological condition characterized by the deposition of mineralized tissue in ectopic locations such as the skeletal muscle. The precise cellular origin and molecular mechanisms underlying HO are still debated. In our study we focus on the differentiation of mesoangioblasts (MABs), a population of multipotent skeletal muscle precursors. High-content screening for small molecules that perturb MAB differentiation decisions identified Idoxuridine (IdU), an antiviral and radiotherapy adjuvant, as a molecule that promotes MAB osteogenic differentiation while inhibiting myogenesis. IdU-dependent osteogenesis does not rely on the canonical BMP-2/SMADs osteogenic pathway. At pro-osteogenic conditions IdU induces a mild DNA Damage Response (DDR) that activates ATM and p38 eventually promoting the phosphorylation of the osteogenesis master regulator RUNX2. By interfering with this pathway IdU-induced osteogenesis is severely impaired. Overall, our study suggests that induction of the DDR promotes osteogenesis in muscle resident MABs thereby offering a new mechanism that may be involved in the ectopic deposition of mineralized tissue in the muscle.

We report that 5-iodo-2′-deoxyuridine (IdU) is a potent inducer of MAB osteogenesis. The mechanism is SMAD1-5-9 independent and p38-dependent. Finally, we show that IdU incorporation into the DNA triggers the DNA damage response (DDR) and that the inhibition of the DNA damage responsive kinase ATM can hinder osteogenic differentiation. We propose that activation of the DDR in mesenchymal progenitor cells may cause heterotopic ossification in vivo.

Results
Idoxuridine induces osteogenesis and limits the myogenic differentiation. Mesoangioblasts (MABs) are multipotent myogenic progenitor cells that spontaneously undergo muscle differentiation, forming myotubes. If exposed to the Bone Morphogenic Protein 2 (BMP-2) they switch their differentiation program forming osteoblasts 18 . To identify cell perturbations that would promote MAB osteogenesis, we have developed a robust high-throughput, high-content screening by monitoring the activity of alkaline phosphatase (ALP) and myogenin expression. We performed the primary screening in the presence of 0.3 μg/ml of BMP-2. This stimulus is not sufficient to promote the formation of alkaline phosphatase positive MABs but might contribute to reveal molecules that act synergistically with BMP-2. Out of 560 compounds of the Prestwick library of FDA approved drugs, 13 small molecules (2,3% of total molecules screened) displayed some ability to induce MAB differentiation into osteoblasts when administrated in the presence of low BMP-2 concentrations. The myogenin staining also allowed to identify compound that promote (8 molecules) or inhibit (16 molecules) MAB myogenesis. The primary screening hits were retested, in the conditions of the primary screening. The compounds whose activity was confirmed in the secondary screening are listed in Supplementary Tables ST1-ST3. We identified a single drug, Idoxuridine (IdU), which significantly induces ALP activity, in a dose-dependent manner with an IC50 of approximately 5 μM (Fig. 1a,b) even in the absence of BMP-2. ALP accumulation is apparent after 48 hours of IdU exposure and reaches a maximum at 120 hours (Fig. 1c,d). IdU induces the expression of both Alpl (Alkaline phosphatase) and Bglap (Bone gamma-carboxyglutamate protein/osteocalcin) mRNAs, late markers of osteogenic differentiation (Fig. 1e,f). IdU not only promotes osteogenesis but also interferes with myogenesis, as observed by staining with anti-myogenin fluorescent antibody and by real time PCR analysis for Myod1 and Myog ( Fig. 1g-j). Notably, IdU treatment does not affect cell proliferation at the tested concentrations (Fig. S1a), suggesting that the inhibition of myogenic differentiation is cell confluence-independent. Moreover, both IdU and BMP-2 treatments cause opposite modulation of osteogenic and myogenic genes.
IdU treatment induces the expression of the osteogenic transcription factor Runx2. We next looked at the expression of Runx2 and Sp7, two master regulators of transcription of osteogenic genes. Since Runx2 expression and activity are controlled both at transcriptional and post-translational levels 19 , we assessed the modulation of both mRNA and protein levels after IdU treatment.
In IdU-treated cells the RUNX2 transcription factor is overexpressed starting at 48 hours, compared to controls. However, its expression levels, as assessed both by western blot and by immunofluorescence staining, remain lower in comparison to that of cells induced to differentiate with BMP-2 ( Fig. 2a-d). To further investigate the early phases of this process, we performed real time PCR analysis confirming a similar kinetic of accumulation of both Runx2 and Sp7 mRNA levels (Fig. 2e,f). In addition we found that IdU treatment does not induce the expression of Sox9 (Fig. S1b), the master regulator of chondrogenic differentiation 20 .
We conclude that IdU treatment promotes osteogenic differentiation by inducing the expression of the master transcription factors that regulate osteogenesis. We next asked whether the mechanisms inducing the expression of Runx2 were similar in BMP-2 and IdU-treated cells.
IdU promotes osteogenesis by a BMP-independent/p38-dependent pathway. To assess the molecular mechanism underlying IdU-induced osteogenesis we first looked at the canonical osteogenic pathway. Bone morphogenetic protein 2 (BMP-2) is one of the physiological inducers of osteogenesis in osteoblast precursors. BMP-2 induces osteogenic differentiation by the SMAD and p38-dependent pathways 19 . To evaluate if these pathways are also activated after IdU treatment we monitored the phosphorylation of both SMAD1-5-9 and p38 at early and late time points throughout the differentiation process. IdU does not induce any phosphorylation of SMAD1-5-9 at any time point indicating that IdU promotes osteogenesis by a SMAD-independent pathway ( Fig. 3a,b,c). On the other hand, an acute and transient activation of p38 is observed at 24 hours (Fig. 3b,d). Conversely, IdU treatment does not induce the activation of the ERK/MAPK-dependent signaling cascade (Fig. S2a,b). To confirm that activation of p38 is involved in IdU induction of osteogenesis we added the p38 kinase inhibitor SB 203580 to IdU-treated MABs. A significant reduction of ALP expression is observed in presence of the kinase inhibitor (Fig. 3e,f). The decrease of ALP expression after p38 inhibition is paralleled by a downregulation of Runx2 and Sp7 gene transcripts (Fig. 3g,h).
We next asked whether the mechanisms inducing the expression of p38 were similar in BMP-2 and IdU treated cells. Compounds that induce the same phenotypic transition via alternative mechanisms are more likely to show synergy 21 . We asked whether IdU synergizes with BMP-2 in the induction of MAB osteogenesis. MABs were treated in combination with BMP-2 and IdU at different concentrations. Co-treatment with both substances (g). Data are presented as percentage of MYOG positive nuclei over the total number of nuclei per field ± SEM. Statistical significance was assessed by a Student's t-test. ***p < 0.001; n = 3. (i and j) Real time PCR analysis of Myod1 and Myog genes upon IdU and BMP-2 treatment at 120 hours. Fold change calculation was performed using the 2 −ΔΔCt method relative to vehicle treated sample (horizontal line = 1). Data are presented as means ± SEM. Statistical analysis was performed by a Student's t-test. ****p < 0.0001; n = 3 (Myod1); n = 4 (Myog).
www.nature.com/scientificreports www.nature.com/scientificreports/ achieved a better than additive effect when compared to the differentiation obtained with the same concentrations of the single substances, as revealed by the quantitation of ALP expression (Fig. 3i, S2c). Two-Way ANOVA statistics confirmed the synergic effect between BMP-2 and IdU showing a strong "Interaction" effect between the two groups (Supplementary Table ST4).
We conclude that the late and transient induction of p38 is essential for IdU-mediated osteogenesis, while IdU treatment does not seem to require the activation of the SMAD signaling cascade. The synergic effect with BMP-2 prompted us to hypothesize that the upstream signals leading to p38 activation after IdU treatment are dependent on a non-canonical signaling pathway.
Idoxuridine incorporates into DNA and affects the cell proliferation rate by inducing a delay at the G2-M transition. The MAPK p38 is involved in many cellular processes such as control of staminality and differentiation 22,23 . Moreover, p38 is a key player in DNA damage stress response and cell cycle control 24 . Since IdU incorporation into DNA interferes with DNA synthesis 25,26 , we wondered whether it may cause perturbation of the cell cycle. We first asked whether IdU affects MAB differentiation by being incorporated into DNA. To this end, we performed a competition experiment exposing cells to 25 μM IdU and 25 μM thymidine, its physiological analogue. In the co-treatment condition, ALP activity is significantly reduced in comparison with samples treated with IdU only (Fig. 4a,b). Thus, by limiting incorporation of IdU into the DNA via addition of equimolar concentration of thymidine, osteoblast differentiation is severely hampered suggesting that IdU affect differentiation by incorporating into DNA.  Figure S5 (b) Densitometric quantitation of western blots as in panel (a). Vinculin was used as normalizer and data are presented as fold induction ± SEM relative to vehicle treated sample at each time point (horizontal line = 1). Statistical significance was assessed by a One-Way Anova test. *p < 0.05; **p < 0.01; ***p < 0.001; n = 3. (c) Representative anti-RUNX2 immunofluorescence of MABs treated for 48 hours with vehicle, 25 μM IdU or 1 μg/ml BMP-2. Scale bar is 100 μm. (d) Fluorescence intensity quantitation of images in panel (c). Fold intensity is relative to vehicle-treated sample (horizontal line = 1). Data are presented as mean ± SEM. Statistical analysis was performed by a Student's t-test. *p < 0.05; **p < 0.01; n = 3. (e) Real time PCR analysis of Runx2 and f) Sp7 in MABs treated with vehicle or 25 μM IdU. Fold change calculation was performed using the 2 −ΔΔCt method relative to vehicle treated sample (horizontal line = 1). Data are presented as means ± SEM. Statistical analysis was performed by a Two-Way Anova test. *p < 0.05; ***p < 0.001; n = 3. Data are presented as fold change versus untreated sample (horizontal line = 1) ± SEM. Statistical analysis was performed by a Two-Way Anova. "Asterisk" symbol (*) represents the significance versus CTRL sample. *p < 0.05; ****p < 0.0001. "Plus" symbol (+) represents the significance between Vehicle and SB pre-treated sample. ++ p < 0.01. n = 3. (i) ALP staining of MABs treated with IdU in combination with BMP-2. MABs were treated and cultivated for 120 hours in growth medium. Osteogenic www.nature.com/scientificreports www.nature.com/scientificreports/ Taken together, these results are consistent with the hypothesis that p38 activation, as a consequence of IdU incorporation into MAB DNA, negatively impact cell proliferation, while promoting osteogenic differentiation. To lend further support to this hypothesis we investigated the perturbations of MAB cell cycle induced upon IdU treatment.
When seeded at a density of 3.1 × 10 3 cells/cm 2 , IdU-treated cells arrest their growth after 48 hours, while control cells grow exponentially for 72 hours (Fig. 4c, S3a). In the exponential growth phase, the proliferation rate of IdU-treated MABs is lower than in the control (Fig. 4d).
To further analyze the cell cycle dynamics, we looked at the distribution of the cell populations in the different phases of the cell cycle. After 24 hours of IdU treatment we observe a reduction in the fraction of cells in the S phase and an increment in the fraction of cells arrested at the end of the G2 phase (Fig. 4e,f). In accordance, p21, a marker of the cell cycle arrest, accumulates at 24 hours after treatment (Fig. 4g,h).
IdU-induced osteogenesis is dependent on the DNA damage response. Given that the MAB osteogenic potential is positively modulated by IdU incorporation into the DNA, we checked whether a different non-natural nucleotide analog, 5-bromo-2′-deoxyuridine (BrdU) would display a similar activity. As shown in Fig. 5a,b, BrdU was also able to induce MAB osteogenesis albeit to a lower extent in comparison to IdU (Fig. 5a,b). Similarly to IdU, BrdU also inhibited myogenesis (Fig. 5a, S4a). In addition, treating MABs with 25 μM uridine or other DNA-interacting molecules (as 0.5 μM etoposide and cisplatin) does not cause osteogenic differentiation (Fig. S4b). This suggests that the incorporation of nucleotide analogs into the DNA, possibly by disturbing the double helix structure, could trigger a signaling cascade that modulates MAB differentiation decisions. Consistently, immunofluorescence staining of phosphorylated H2AX (γ-H2AX), 24 hours after treatment with IdU, shows an increase of γ-H2AX nuclear puncta (Fig. 5c,d), which is, however, less prominent than the one observed after treatment with the DNA damaging agent etoposide (ETO). Western blot analysis confirmed the activation of the upstream ATM kinase, which requires phosphorylation of γ-H2AX after IdU treatment. (Fig. 5e,f).
Interestingly, the ATM kinase inhibitor KU 55933, negatively modulates the induction of ALP as well as the transcription of Runx2 and Sp7. This inhibition is further reinforced by the addition of the p38 kinase inhibitor SB 203580 (Fig. 5g-j). Taken together, these results are consistent with a model whereby the osteogenic differentiation of MABs upon IdU treatment is dependent on the activation of the DDR-related ATM/p38 axis.

Discussion
Under a variety of stimuli, including inflammation and traumatic injury, skeletal muscle residing mesenchymal stem cells differentiate into osteoblast precursors 1 . This event, dubbed as heterotopic ossification (HO), causes the formation of bone foci inside the muscle parenchyma thereby affecting its function. Heterotopic ossification is a debilitating condition whose etiology is not fully understood. Although the identity of progenitor cells that promotes the formation of heterotopic bones and cartilage is still debated, cells expressing vascular endothelial markers have been proposed as candidate originators of this aberrant process 15 .
Adult mesoangioblasts (MABs) are muscle resident multipotent mesenchymal progenitors that are physically associated to blood vessels 27 . Once isolated from the muscle tissue, they have the potential to differentiate into skeletal and smooth muscle or osteoblasts 18 . With the aim of characterizing perturbations and pathways that control the differentiation choice of muscle precursor cells we developed an assay for osteoblastic differentiation suitable for high-content screening. This approach allowed us to identify Idoxuridine (IdU) as an inducer of MAB osteogenic differentiation as inferred by the expression of the early osteogenic marker alkaline phosphatase. IdU treated MABs, however, also express additional bone proteins as for instance osteocalcin, a late osteogenic marker involved in the production of the extracellular mineralized matrix 28 . Bglap, the osteocalcin encoding gene is transcribed as early as five days after treatment in our conditions. This is comparable with the kinetic reported in other in vitro differentiation models where the mRNA of Bglap is already detectable at day 4 and reaches a maximum after 21 days from induction of osteogenesis 29 .
Osteogenesis is physiologically controlled by the BMP-SMAD signaling pathway. In such path, the BMP binds to its serine/threonine-kinase receptor and triggers a phosphorylation cascade that activates the transcriptional activity of the osteogenic transcription factor RUNX2. In parallel, the same receptor induces a MAP kinase cascade through the activity of the MAP kinase-kinase 3-6 (MKK 3-6) and p38 10,11 . The activation of p38 is important for the induction of the osteogenic program, since RUNX2 transcriptional activity is promoted by phosphorylation by this MAPK on multiple residues 10,11,13,30 . BMP-2 and IdU, when co-administered, show a more than additive effect in the induction of osteogenic differentiation (Fig. 3i, S2c, Supplementary table ST4) suggesting that the two stimuli act via different mechanisms and pathways.
Consistently, we show that IdU induces the expression of both Runx2 and Sp7 (Fig. 2) independently from the activation of the BMP/SMAD signaling pathway (Fig. 3a-c), while an acute but transient activation of p38 is observed in a limited time window (Fig. 3b,d). The osteogenic phenotype and the expression of the related transcription factors after IdU treatment, are strictly dependent on the activation of p38, as shown by their sensitivities to p38 inhibitors (Fig. 3e-h).
Additional members of the MAPKs protein family, such as ERK 1-2 and JNK have been implicated in osteoblast differentiation and bone homeostasis 31 . The deletion of ERK 1-2 causes defective bone mineralization 32 . ERK activates the transcriptional activity of RUNX2 by phosphorylating S301 and S319 33 . IdU on the other hand differentiation was assessed by ALP staining. ALP positive area percentage is reported at the bottom-left side of each micrograph. Red-squared images show an additive effect between BMP-2 and IdU. The complete quantitation of ALP staining is shown in Figure S2. www.nature.com/scientificreports www.nature.com/scientificreports/ does not activate the ERK pathway (Fig. S2a,b). The activity of the JNK MAPK was also linked to the late stages of osteoblasts differentiation, according to Yamashita and Guicheux showing a decrease in ALP and osteocalcin expression during osteoblast differentiation, upon JNK inhibition 34,35 .
IdU is a nucleotide analogue that is prescribed as an antiviral agent given its ability to interfere with the DNA synthesis of Herpes simplex virus 26,36 . Moreover, the incorporation of IdU in eukaryotic cells sensitizes cancer cells to radiotherapy treatment, by inducing DNA double strand breaks (DSBs) on highly proliferative cells 25,37,38 .
Our results support a model whereby IdU induces a cell cycle delay in the G2-M transition phase (Fig. 4). Cell cycle modulation is a critical step for osteoblast differentiation. Lau and colleagues showed that low doses of X-irradiation induce G2-arrest in the pre-osteoblast cell line OCT-1 through the activation of p21 and the subsequent induction of Runx2 expression 39 . This cell cycle perturbation could be imputed to the activation of the p38 axis. Upon DNA damage ATM activates p38 through the coordinated action of the Thousand And One amino acid protein 1-3 (TAO1-3) and the MAP Kinase-Kinase 3-6 (MKK3-6). The induction of this signaling cascade results in the inhibition of the CDC25c phosphatase, resulting in the cell cycle arrest at the G2-M transition 40,41 . The observation that, in analogy with IdU, also the halogenated nucleotide BrdU promotes osteogenic differentiation (Fig. 5a,b) prompted us to investigate whether incorporation of nucleotide analogs would induce DNA damage causing a DNA Damage Response (DDR).
The activation of the ATM kinase is the first response of the DDR to DSBs and it is controlled by an autophosphorylation event 42 that triggers its kinase activity on downstream targets, as the histone isoform H2AX 43 ( Fig. 5b-f). γ-H2AX in turn plays an important role in the recruitment and assembly of the DNA repair molecular machinery 44,45 . The ATM kinase has already been implicated in osteogenic differentiation. Rasheed et al. speculated that ATM could positively modulate the BMP pathway, since ATM −/− mice display an osteoporotic phenotype due to the detrimental effect on the expression of Sp7 transcription factor 46 . By contrast, here we show that IdU treatment activates ATM and induces osteogenesis by a BMP/SMAD-independent pathway. It has been reported that the osteogenic differentiation could also be controlled through RUNX2-independent pathways. Lee and Ulsamer characterized a Dlx5-dependent/Runx2-independent mechanism directly acting on the expression of Sp7 47,48 . Dlx5 is phosphorylated by p38 on residues S34 and S217 to increase its transactivation activity. In our system IdU treatment does not promote Dlx5 expression (data not shown).
In our study, we show that IdU-induced osteogenesis requires both ATM and p38 as pharmacological inhibition of either kinase activity reduces the transcriptional activation of both Runx2 and Sp7 and, consequently, osteogenic differentiation (Figs 3e-h, 5g-j).
Overall, our data point to DDR as a potential modulator of the differentiation fate of skeletal muscle resident mesenchymal stem cells. In particular we show that in MABs the DDR can influence the differentiation "choice" between the myogenic and the osteogenic fate.
The observation that p53 −/− and c-Abl −/− mice are characterized by an osteosclerotic and an osteoporotic phenotype, respectively, points to the involvement of these two direct targets of ATM and of the DNA repair pathway [49][50][51] in osteogenic differentiation and bone development.
In direct contrast with the role of ATM in promoting osteoblast differentiation is the recent evidence of the ATM negative impact in the activation of the checkpoint kinase Chk2 52 . Chk2 deletion can partially rescue the osteogenic defects in mice lacking the parathyroid hormone-related peptide (PTHrP). Overall, these contrasting evidences highlight a Janus-like effect of ATM, stressing the role of context on its positive or negative effect on osteogenesis.
DNA damage occurs both in physiological and pathological conditions, such as aging 53,54 and acute or chronic inflammation 55,56 , conditions that also promote ectopic ossification. Myositis ossificans, as well as other myo-pathologies such as Duchenne Muscular Dystrophy (DMD), are characterized by an inflammatory response 1,57-59 . Recent evidences suggest a link between HO and the latest complications of DMD 60,61 . Here, we suggest that the inflammation-induced DDR could be one of the key factors in the development of ectopic ossification opening the possibility to develop new therapeutic strategies to counteract this pathological differentiation phenomenon.
Screening procedure. Mesoangioblasts (MABs) were plated in 96-well plates at a density of 10,000 cells per well in DM with 0.3 μg/ml bone morphogenetic protein-2 (BMP-2) (R&D Systems, #355-BM). Chemicals were added in the differentiation medium at 1 μM, 10 μM and 25 μM final concentration. Every 96-well plate in the in Figure S8. (h) Densitometric quantitation of p21 in panel (g). Values were normalized over vinculin and the fold induction was calculated over the time zero (T0 = 1) sample. Data are represented as mean ± SEM. Statistical significance was assessed by a Student's t-test. **p < 0.01. n = 3. www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ screen included 9 positive controls (3 wells with 0.3 μg/ml of BMP-2, 3 wells with 0.6 μg/ml of BMP-2, 3 wells with 1 μg/ml of BMP-2) and 10 negative controls, where BMP-2 was not added to the medium. The cells were maintained in compounds until day 5. Samples were subsequently stained for alkaline phosphatase activity and with anti-myogenin antibody as described below.
Treatments and differentiation. All treatments and differentiation were performed in GM. Idoxuridine For the anti-RUNX2 immunolabeling, permeabilization was performed by incubating cells with methanol for 10 minutes at −20 °C and washed once with dd-H 2 O and three times with PBS before blocking. Blocking was performed with 10% v/v FBS in WS (blocking solution -BS) for 1 hour at RT.
After incubation with the primary antibodies the specimens were washed 3 times with WS. Secondary antibodies were incubated for 30 minutes at RT in the dark using the following products and dilutions in BS: Goat anti-Mouse IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (Thermo Fisher Scientific, #A-11001, 1:300 v/v), Goat anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 (Thermo Fisher Scientific, #A-11008, 1:300 v/v).
After the incubation with the secondary antibodies the specimens were washed 3 times with WS. Nuclear staining was performed with Hoechst 33342 (Thermo Fisher Scientific, #H3570) diluted 1:5000 v/v in WS. The specimens were washed 3 times with PBS and maintained at 4 °C in PBS supplemented with 0.02% w/v sodium azide.
Image acquisition and analysis. All microscopy images were acquired with a Leica DMI-6000B fluorescent microscope. In the case of the screening procedure and differentiation experiments, 20X images were acquired with the Matrix Screener mode using a 3 × 3 (screening) or 5 × 5 (other experiments) matrix-layout after auto-focusing.
Efficiency of myogenic differentiation was expressed as the percentage of myogenin positive nuclei over the total number of nuclei. Efficiency of osteogenesis (Alkaline phosphatase positive cells) was estimated as the fraction of alkaline phosphatase area over the total analyzed area.
For the γ-H2AX puncta and RUNX2 intensity assays, 40X images were manually acquired for a total of 5 images per sample. Unbiased image analysis was performed using the Cell Profiler software 65 using dedicated pipelines.
Scale bar is 200 μM. (h) Bar plot of ALP positive area quantitation of images in panel (g). Data are represented as mean ± SEM of the percentage of ALP positive area of the total area of the image. Statistical analysis was performed through One-Way Anova. ** p < 0.01. n = 3. (i,j) Bar plot of RT-qPCR analysis of Runx2 and Sp7 of MABs pre-treated with vehicle or 10 μM KU 55933 ATM inhibitor alone or in combination with 10 μM SB 203580 p38 inhibitor for 3 hours in growth medium and then treated with 25 μM IdU for 48 hours in growth medium. Data are represented as fold change versus untreated sample (CTRL = 1) ± SEM. Statistical analysis was performed through Two-Way Anova. "Asterisk" symbol (*) represents the significance versus CTRL sample. **p < 0.01; ***p < 0.001; ****p < 0.0001. "Plus" symbol (+) represents the significance between DMSO and SB pre-treated sample. ++++ p < 0.0001. n = 3.
Densitometric quantitation of bands was performed with ImageJ (NIH) software. Image brightness and contrast were regulated with the Auto function in ImageJ. Cropped image are reported in Figures