Our experiments comparing MyoD binding between normal human myotubes and rhabdomyosarcoma cells have identified (i) regions of decreased MyoD binding that include critical myogenic cofactors, such as MEF2C, (ii) local impairment of MyoD binding at peaks associated with specific cofactors (MEF2C, RUNX, NFIC, and JDP2), and (iii) local DNA hypermethylation at the promoter of the transcription factor JDP2, a factor capable of inducing myogenic differentiation in RMS, in numerous primary tumors, as well as the RD cell line. Together, these findings indicate both local and regional suppression of cofactors that cooperate with MyoD to regulate muscle gene expression and result in a decreased expression of a subset of MyoD-regulated genes.
Widespread genome-wide binding of MyoD in RD cells agrees with our prior findings that MyoD activity is compromised in RMS but that DNA binding is not globally affected (8
). Given the global similarity in MyoD binding between myotubes and RD cells, the observed differences in subsets of MyoD-bound sites, such as decreased MyoD binding adjacent to myogenic genes that fail to activate in RMS, begin to provide mechanistic insight into how RMSs fail to differentiate. Rather than a wholesale failure of MyoD activity, MyoD function fails at a discrete subset of its targets.
We identified two causes for decreased MyoD binding at subsets of sites in RD cells. Differences in binding over large regions spanning 100 or more kilobases correlated with differential nuclease access and indicated that large regional differences in chromatin compaction might restrict MyoD access to some critical genes, e.g., the MEF2C gene ( and ). In addition to regionally decreased MyoD peaks, many peaks were locally decreased (i.e., in a region where the surrounding peaks were not different between RD and primary muscle cells). The de novo motif analysis of the DNA sequences under these peaks identified potential cofactors that are expressed at lower levels in RD cells than primary muscle cells and can induce the expression of muscle genes when expressed in RD cells. Our demonstration that MEF2C expression also increases MyoD binding at some peaks with an adjacent MEF2 binding motif suggests that MEF2C and, possibly, the other cofactors increase the accessibility of these sites to MyoD and/or stabilize the binding of MyoD at some of these sites. We should note, however, that site-specific ChIPs for MyoD with increased RUNX1 expression did not detect an increase in MyoD binding (K. L. MacQuarrie, unpublished data), but it is possible either that the appropriate sites were not interrogated or that some myogenic cofactors identified by our analysis increase MyoD binding, while others do not.
In normal myogenesis, MyoD causes histone acetylation on a genome-wide scale (1
). We have previously demonstrated the presence of histone acetylation at MyoD targets in RD cells (9
), demonstrating that MyoD can still function in that capacity in RMS cells. While a low level of histone acetylation would be expected across the regions that we identify to be binding less MyoD in RD cells than in primary cells, it is not clear what occurs in the situation of local changes, such as individual low peaks. It is possible that local acetylation, such as on the scale of a single nucleosome, might be impaired, but it is also possible that acetylation occurs at some or all such sites. Such possibilities will need to be tested directly to determine how specifically targeted MyoD-mediated histone acetylation is at such sites.
We have previously proposed that rhabdomyosarcoma cells are poised on the verge of differentiation and require only a final push for the process of myogenesis to proceed to completion, after the activation of nested feed-forward circuits that center on MyoD and utilize cooperative factors such as RUNX1. The data presented in this paper not only offer further support for that model but also suggest that myogenesis may be controlled by a dosage of myogenic factors, both the myogenic regulatory factors and cooperating factors. RMS differentiation can be induced by JDP2, NFIC, RUNX1, ZNF238, and MEF2C, and other groups have demonstrated similar effects with other components of the normal myogenic gene network (9
), all of which interact with and/or potentiate MyoD-mediated activity in some manner. The ability of a variety of MyoD-cooperating factors to complete the final steps of differentiation in RMS cells and complete the process of terminal differentiation suggests that myogenic cells have evolved to respond to a cumulative dosage of promyogenic factors. A similar model has previously been proposed for B-lymphocyte development, with the relative dosage of three of the E proteins being critical for development of the cells in vivo
Using a genome-wide DNA methylation assay developed in our lab, we have recently shown hypermethylation of CpG islands in rhabdomyosarcoma cells relative to normal skeletal muscle (31
). While this assay robustly identifies differential DNA methylation in large CpG islands, it does not detect more subtle differences in DNA methylation outside these regions (32
). Methylation of CpG shores, areas adjacent to but outside CpG islands, has been implicated in both developmental processes and tumor biology and linked to reduced gene expression (33
). The identification of hypermethylation in the CpG shore in the promoter region of JDP2
in primary tumors as well as the RD cell line demonstrates that the differences seen in myogenic pathways in cell culture models can be relevant to primary tumor biology and suggests that the silencing of promyogenic genes apart from the myogenic regulatory factors may be a mechanism that these tumors use to escape terminal differentiation. It is of interest to note that the AP-1 motif was identified as being enriched with myoblast-specific peaks in the primary cell comparison (), while it was identified as being enriched with myotube-specific peaks in the later comparison (); but JDP2 can homodimerize as well as form heterodimers with a variety of partners, including ATF-2, c-Jun, and C/EBPγ (35
), and the primary cell comparison may represent an AP-1 complex with functional differences from the activity that occurs with increased JDP2 expression in RMS cells. JDP2 activity has been implicated in the control of cellular senescence (38
), and HES1, a bHLH protein involved in the control of the quiescence-senescence decision in cells, contributes to the failure of terminal differentiation in RMS cells (39
). The relation of both factors to each other and MyoD activity during normal myogenesis or in RMSs is unclear at this time.
The hypermethylation of the JDP2 promoter will need to be further investigated to determine if it is capable of tumor induction or is related to continued tumor growth after formation. Regardless of whether JDP2 hypermethylation is a causative factor in the genesis of rhabdomyosarcoma, it suggests that the development of prodifferentiation-based therapies that will impair or halt tumor growth by screening for their ability to affect specific cellular targets inducing terminal differentiation may be a possibility.