|Home | About | Journals | Submit | Contact Us | Français|
The role of nutrients and metabolism in cellular differentiation is poorly understood. Using RNAi screening, metabolic profiling and small-molecule probes, we discovered three metabolic enzymes whose knockdown induces differentiation of mouse C2C12 myoblasts even in the presence mitogens: phosphoglycerate kinase (Pgk1), hexose-6-phosphate dehydrogenase (H6pd) and ATP citrate lyase (Acl). These enzymes and the pathways they regulate provide novel targets for the control of myogenic differentiation in myoblasts and rhabdomyosarcoma cells.
Numerous studies have pointed to the central role of transcription factors in cell differentiation (1). We sought to determine if carbon metabolism is connected to the cellular differentiation network (2) and, if so, whether it could provide a novel method for controlling myogenic differentiation. Here, we provide evidence that perturbations at specific points in the metabolic network can induce muscle myoblasts to differentiate into myosin heavy chain-expressing myotubes and we elucidate the associated metabolic pathways. We show that this finding can be applied to small molecule-based differentiation of human rhabdomyo-sarcoma cells.
We developed a method for RNAi-based screening using mouse C2C12 myoblasts. These cells are proliferative in the presence of serum (10% fetal bovine serum (FBS)) and can be induced to differentiate into multi-nucleated myotubes by serum withdrawal within 5 days (3). To determine whether metabolic enzymes play a role in inducing muscle differentiation, we performed a phenotypic, high-content screen by knocking down 50 enzymes of carbon catabolism using short hairpin RNAs (Supplementary Table 1) in the presence of 10% FBS (Supplementary Methods). Differentiation induced by the hairpins was quantitated by image analysis of myosin heavy chain expression using immunofluorescence. Integrated myosin heavy chain intensity per well was calculated relative to that in control wells. Negative controls included cells infected with non-targeting hairpins (e.g. shLacz, shRFP and shGFP), and positive control cells were treated with 2% horse serum (HS), which induces differentiation (3). Three enzymes were identified that induced the differentiation of C2C12 myoblasts following knockdown: phosphoglycerate kinase (Pgk1), hexose-6-phosphate dehydrogenase (H6pd) and ATP citrate lyase (Acl) (Figure 1a–b). Successful knockdown by at least two distinct shRNAs targeting each of these metabolic genes and expression of various differentiation markers (myogenin and myosin heavy chain) were confirmed by Western blot analysis and quantitative RT-PCR, respectively (Supplementary Figure 1a–g). We further confirmed the on-target effects of Pgk1 and H6pd knockdown by establishing cell lines stably expressing the human orthologs of Pgk1 and H6pd. Knockdown of the endogenous mouse genes in these cell lines failed to induce differentiation (Supplementary Figure 1h).
Because some metabolic enzymes have been shown to possess metabolism-independent regulatory functions (4) we next determined whether myogenic differentiation was accompanied by changes in levels of the metabolites regulated by, Pgk1, H6pd and Acl. We analyzed global changes in levels of intracellular metabolites that accompany myogenic differentiation using LC-MS/MS-based metabolic profiling (Supplementary Methods). Myogenic differentiation was induced by culturing C2C12 cells in reduced serum, and relative levels of intracellular metabolites were measured on days 0, 3 and 6 after initiating differentiation (complete list of measurements is provided in Supplementary Table 2). Upon differentiation, two salient changes were observed: 1) a decrease in glycolytic intermediates 3-phosphoglycerate and phosphoenol pyruvate and 2) a decrease in the bile acids taurodeoxycholic acid (TUDCA) and glycochenodeoxycholic acid (Supplementary Figure 1i). Pgk1 catalyzes the formation of 3-phosphoglycerate, which is then metabolized to phosphoenol pyruvate during glycolysis. Both of these metabolites were found to decrease significantly during myogenic differentiation. Bile acids including TUDCA are endogenous inhibitors of endoplasmic reticulum (ER)-based stress (5). H6pd is an ER-localized isoform of glucose 6-phosphate dehydrogenase, and inhibition of H6pd causes ER stress in vivo (6).
Previous studies suggest that ER signaling can activate the calcium/calcineurin/NFAT pathway; both ER stress and NFAT signaling via the phosphatase calcineurin are known mediators of skeletal muscle differentiation and function (7–9). We found that H6pd induces differentiation of C2C12 cells in a calcineurin-dependent fashion (Figure 1c) as judged by the inhibition of differentiation upon treatment with cyclosporin A (CsA). This effect was specific for shH6pd and not seen with shPgk1 or shAcl (Figure 1c). Thus, we infer that H6pd knockdown promotes myogenic differentiation through an ER-signaling mechanism that is dependent on calcineurin.
We were unable to identify metabolites from the metabolic profile that reveal a direct role for Acl, the third enzyme identified in the RNAi screen. Therefore, we separately examined whether Acl impacts differentiation via its downstream effects on metabolism. Acl is a primary source of nuclear acetyl coA for histone acetylation (10) and cytoplasmic acetyl CoA for cholesterol and fatty acid synthesis (11). We explored the effect of Acl knockdown in C2C12 cells on both of these pathways.
We determined that C2C12 differentiation by Acl knockdown is dependent on histone deacetylases (HDACs), as judged by the inhibition of shAcl-mediated differentiation by the HDAC inhibitor trichostatin A (TSA) (Figure 1d). We hypothesized that depletion of Acl would reduce nuclear acetyl coA, which is a necessary cofactor for histone acetyl transferases. We therefore measured levels of histone acetylation in shAcl-treated cells (10) and found that knockdown of Acl reduces levels of acetylated histones (Figure 1e). This suggests a mechanistic connection between acetyl CoA metabolism, chromatin and myogenic differentiation.
Our primary screen indicated that lipid metabolism per se was not an effector of differentiation since knockdown of enzymes in fatty acid s-oxidation did not cause myoblast differentiation. We therefore examined the role of cholesterol metabolism by treating cells with statins, small-molecule inhibitors of cholesterol biosynthesis, and found that they stimulate differentiation. C2C12 myoblasts were treated with three different statins: pravastatin, atorvastatin and fluvastatin. Fluvastatin and atorvastatin are potent inducers of C2C12 differentiation (at 1 μM), while pravastatin is less effective but also capable of inducing myogenic differentiation (Supplementary Figure 2a, b). Furthermore, cholesterol abrogates the differentiation-promoting effect of statins, suggesting that cholesterol mediates C2C12 differentiation (Supplementary Figure 2c). We have observed that both cholesterol and TSA inhibit differentiation induced by serum withdrawal (Supplementary Figure 2d), implicating both the nuclear and cytoplasmic acetyl coA in myoblast differentiation.
Treatment of the human RD embryonal rhabdomyosarcoma cell line with fluvastatin induced rhabdomyosarcoma cell differentiation, reduced cancer cell proliferation and inhibited anchorage-independent growth (Figure 2a–c, Supplementary Figure 2e, Supplementary Methods). Rhabdomyosarcoma is the most common pediatric soft-tissue sarcoma and is presumed to derive from cells of the skeletal muscle lineage (12). Interference with any of three key nodes of the metabolic network described herein may offer new approaches to cancer differentiation therapy for rhabdomyosarcoma.
In this study, we identify three novel metabolic enzymes whose knockdown with RNAi causes myoblast differentiation: Pgk1, H6pd and Acl. During myogenic differentiation, Pgk1 mRNA levels decrease by about 50% within 4 days after proliferating cells are induced to differentiate (13). We provide evidence that metabolic intermediates regulated by Pgk1, 3-phosphoglycerate and phosphoenol pyruvate decrease during myogenic differentiation.
Unlike its cytosolic isoform glucose-6-phosphate dehydrogenase, H6pd localizes to the ER and is used to regenerate reduced nicotinamide adenine dinucleotide phosphate (NADPH) within the ER when glucose-6-phosphate and NADP are available (14, 15). H6pd-null mice develop skeletal myopathy characterized by switching of type-II fibers to type-I (slow) fibers. The absence of H6pd impairs protein folding and activates the unfolded protein response pathway (6). Here we show that H6pd knockdown induces myoblast differentiation, and that myogenic differentiation is accompanied by a decrease in endogenous inhibitors of ER stress: the bile acids, TUDCA and glycochenodeoxy-cholic acid. Further, we provide a mechanistic connection between H6pd, ER stress and calcium/calcineurin signaling by showing that CsA overcomes shH6pd-mediated differentiation.
Acl catalyzes the synthesis of acetyl-CoA, which is used in nuclear histone acetylation and cytoplasmic cholesterol biosynthesis. We provide evidence that both pathways are relevant to shAcl-mediated differentiation. Inhibition of Acl reduces global chromatin acetylation (10). Similarly, we found a decrease in chromatin acetylation following Acl knockdown. However, the differentiation response elicited by shAcl appears to be cell-context specific. In pre-adipocytes, shAcl blocks differentiation (10) whereas and in myoblasts, it induces differentiation (as shown in this report). In cancer models, inhibition of Acl by either RNAi or the small molecule SB-204990 limits proliferation and survival of tumor cells displaying aerobic glycolysis in vitro (16), and induces tumor differentiation and reduction in tumor growth in vivo (16). We show here that inhibition of cholesterol metabolism through treatment with fluvastatin induces rhabdomyosarcoma cell differentiation, blocks proliferation and prevents anchorage-independent growth.
Thus, we discovered that myogenic differentiation and cellular metabolism are coordinated via at least three cellular pathways: calcium/calcineurin signaling, chromatin acetylation and the cholesterol biosynthetic pathway. These data suggest that metabolic enzymes may be important for the control of cancer cell growth and differentiation, and that metabolic perturbations may result in important changes in cell states. Alterations in the metabolic network identified herein may serve as novel therapeutic approaches to muscle sarcoma differentiation therapy in the future.
We thank the RNAi Consortium for shRNAs and members of the RNAi platform of the Broad Institute for their scientific advising: Alan Derr, Jen Grenier, Serena Silver, Glenn Cowley and Ozan Alkan. We also thank Steve Carr, Ru Wei, Elaine Yang and members of the Broad Institute metabolic profiling platform for scientific advice and analysis of metabolite extracts, and Jeanne Nisbet of Children’s Hospital Boston for thoughtful comments. This work was supported in part by DoD Breast Cancer Innovator Award #BC074986 (to D.E.I.) and NIGMS 38627 (to S.L.S.). S.L.S. is an investigator with the Howard Hughes Medical Institute.
Author ContributionsA.L.B. designed and performed RNAi screen and all validation experiments qRT-PCR, generation of stable cell lines and studies on cyclosporin, cholesterol and trichostatin. A.R. designed and performed metabolite experiments and histone acetylation study. S.L.S. participated in the design of the overall study and S.H., D.E.I. and S.L.S. contributed to the preparation of the manuscript. D.E.I. and A.L.B. conceived of the idea for this study. All authors interpreted data, commented on results, and participated in writing the manuscript.
Competing Financial Interests
The authors declare no competing financial interests.