dMec is a novel dMi-2-containing complex
Mi-2 has been shown to reside within NuRD complexes and, although interactions between Mi-2 and other proteins have been reported, there has been no indication so far that Mi-2 also exists in other stable and abundant remodelling complexes (Bowen et al, 2004). Here, we provide evidence for a novel dMi-2-containing complex, dMec, which we have purified to apparent homogeneity from nuclear extracts of Drosophila melanogaster. dMec consists of two subunits: dMEP-1 and dMi-2. Unlike NuRD, dMec does not contain a histone deacetylase activity.
Previous studies had suggested that Caenorhabditis elegans MEP-1 binds both to an Mi-2 homologue (LET-418) and to an HDAC1 homologue (HDA-1) (Unhavaithaya et al, 2002). It was therefore hypothesised that MEP-1 associates with the NuRD complex in C. elegans. In contrast, our immunoprecipitations, depletions and chromatographic fractionations strongly suggest that dMec and dNuRD or dRPD3 do not physically associate in solution. Given that the dNuRD subunit dp66 and dMEP-1 co-occupy a number of promoters in the AS-C locus, it remains possible that dMec and dNuRD or dRPD3 interact when bound to chromatin. In addition, we cannot rule out that dMec and dNuRD or dRPD3 associate in contexts or cell types not investigated in this study. It is clear, however, that dMEP-1 is not a stable subunit of dNuRD. Rather, it forms a distinct complex with dMi-2 in nuclear extracts derived from Drosophila cell lines.
The tight physical association of dMi-2 and dMEP-1 is mirrored by their virtually identical temporal and spatial coexpression profiles during embryogenesis. The subunits of the dMec complex are detectable in early stages of embryogenesis before the onset of zygotic gene expression. The robust expression of dMi-2 and dMEP-1 in ovaries suggests that dMec is maternally contributed. dMi-2 and dMEP-1 levels increase during the first 9 h of embryogenesis and sharply decline thereafter. This suggests an important role of dMec during embryogenesis. Although the levels of dMec subunits in larvae are much lower than in embryos, we have detected robust interactions between dMi-2 and dMEP-1 in extracts derived from larval tissues including larval brains and wing imaginal discs (data not shown).
dMec represses proneural genes
We were surprised to find that the bulk of dMi-2 resides in dMec and only a minor fraction is associated with dNuRD. This raises the possibility that phenotypes that have been observed in dMi-2 mutant flies are not a consequence of a defective dNuRD complex, but a result of a lack of dMec activity. Genetic analysis has demonstrated that dMi-2 and TTK69 cooperate to repress differentiation into neuronal cells during development of the Drosophila nervous system. TTK69 achieves this in part by repressing the proneural genes of the AS-C (Murawsky et al, 2001; Badenhorst et al, 2002). Yamasaki and Nishida have recently shown that dMi-2 and TTK69 colocalise to the AS-C locus on polytene chromosomes and that an ectopically expressed HA-tagged dMi-2 protein associates with the ac gene in wing imaginal discs (Yamasaki and Nishida, 2006). These experiments did not distinguish between dNuRD and dMec. Therefore, we asked whether endogenous dMec and/or dNuRD are binding to and are involved in repression of proneural genes. Our ChIP experiments show that both complexes localise to an overlapping set of genes within this locus. However, whereas dNuRD binds to promoters of proneural genes as well as the promoter of the unrelated pcl gene, dMec appears to bind preferentially to the proneural genes. We have recently identified dMi-2 and dMEP-1 in a screen for corepressors involved in SUMOylation-mediated transcriptional repression (Stielow et al, 2008). Both proteins can bind to the SUMO-group in vitro (Stielow et al, 2008). Given that TTK69 is SUMOylated in vivo and that SUMOylation is important for TTK69 function, it is conceivable that dMec is recruited to the AS-C locus by interacting with DNA-bound, SUMOylated TTK69 (Bajpe et al, 2008).
We note that the analysis of ChIP by quantitative PCR has also revealed an association of dMi-2, dMEP-1 and dp66 with non-promoter regions. It is conceivable that this is a reflection of a general, non-targeted association of these factors with chromatin. It is interesting to note that such a ‘constitutive’ association with chromatin has been demonstrated for mammalian Mi-2 (Li et al, 2002).
dNuRD is considered to contribute to transcriptional repression through its associated histone deacetylase activity. dMec lacks a histone deacetylase subunit. Nevertheless, depletion of dMEP-1 results in derepression of proneural genes of the AS-C locus. This suggests that dMec actively contributes to the repression of proneural genes in an HDAC-independent manner. Several findings suggest that dNuRD, although present at the promoters, does not have a significant function in maintaining proneural gene repression. First, depletion of dMi-2, which is expected to affect both dMec and dNuRD, does not result in stronger derepression than dMEP-1 depletion. Second, codepletion of dMEP-1 and dMi-2 does not produce stronger derepression effects than depletion of dMEP-1 alone. Third, reducing the levels of the dNuRD subunit dp66 does not alter proneural gene transcription. Fourth, TSA, which is expected to inhibit the dNuRD-associated HDAC activity, does not derepress transcription of proneural genes. Taken together, our findings suggest that dMec represses transcription in an HDAC-independent manner, although it remains possible that other (TSA-insensitive) HDAC-containing complexes are involved. Consistent with this view, SUMOylation-mediated transcriptional repression depends on dMi-2 and dMEP-1 but does not require histone deacetylases and is insensitive to TSA (Stielow et al, 2008). It is likely that the ATP-dependent nucleosome remodelling activity of dMec is involved in transcriptional repression. In agreement with this hypothesis, dMec displays nucleosome-stimulated ATPase activity in vitro.
The identification of dMec raises the question of whether mammalian cells also contain Mi-2 complexes in addition to NuRD. On the basis of amino acid sequence comparison, it may be inferred that there is no obvious homologue of dMEP-1 encoded in mammalian genomes. Moreover, the region of dMEP-1 responsible for dMi-2 binding does not contain easily recognizable structure or sequence motifs. However, the region of dMi-2 that is bound by dMEP-1 is highly conserved in mammalian Mi-2 homologues (Klinker, Kunert and Brehm, unpublished data). This conservation of the dMEP-1-binding region suggests that mammalian cells might contain dMi-2-binding proteins with functional similarity to dMEP-1.
Several chromatin remodelling complexes can share a common ATPase subunit (Eberharter and Becker, 2004). ISWI ATPases are present in NoRC, ACF, CHRAC and NURF. Brahma is a component of both BAP and PBAP. The functional characteristics of each complex are determined by which subunits the ATPase is associated with. Our discovery of dMec extends this theme to the CHD family of ATP-dependent nucleosome remodellers and uncovers an unexpected complexity in the composition and function of Mi-2 complexes.
In the future, it will be important to determine the molecular mechanisms of transcriptional repression by dMec and dNuRD, to identify their respective sets of target genes, to elucidate their modes of recruitment and to analyse the relative contributions of dMec and dNuRD to neural development in vivo.