Our ability to understand how genomic DNA is spatially organized into eukaryotic cells has dramatically increased with the development of robust proteomics and deep sequencing techniques. These technologies in conjunction with techniques like chromatin conformation capture (3C), ChIP sequencing, high resolution imaging and genome modeling have started to address the important question of how higher order structure of DNA regulates cell-specific expression of genes. However, NuMat attachment sites of chromatin have seldom been applied in this context. In the present study, we have sequenced the MARs of D. melanogaster at developmental stages to understand the role of these sequences in genome packaging and regulation.
Before investigating the dynamics of MARs during development and differentiation, we re-visited the NuMat isolation protocols. Initial studies have suggested that the nuclear protein matrix is mostly made up of lamins. Subsequently, the protein composition has been found to be more complex with lamins being the most prominent and reproducibly associated protein (19, 20). It is observed that protein composition depends on subtle variations in the isolation protocol. The protein composition is notably enriched when NuMat is prepared in the presence of oxidative agents or divalent cations (21, 22). Similar effect is seen if the NuMat is prepared with inclusion of a heat stabilization step where the isolated nuclei is briefly exposed to physiologic temperature of 37oC prior to salt extraction (15, 23). EM images show that stabilization is necessary for detection of fibro-granular network inside the nuclei. The stabilization step is particularly questioned, as it has been shown to render a set of nuclear proteins insoluble by creating disulfide cross-links (21, 24). It is important to mention here that nuclear proteins are not cross-linked by disulfide bonds in vivo. Such cross-links appear spontaneously during isolation of nuclei and NuMat, and thus if alkylating agents are used to inhibit the formation of disulfide bonds, the NuMat prepared subsequently is barely recognizable (21, 25). These observations lead to a critical question that whether the isolation process stabilizes a pre-existing structure or causes changes in the nuclear organization to create such structure. We addressed this in our present study by comparing the NuMat proteome and MAR sequences isolated from unstabilized and stabilized nuclei. Our results demonstrate that in agreement with the published literature, intricate and dense fibro-granular meshwork can be visualized after stabilization. Protein content of NuMat isolated with or without stabilization remain qualitatively similar, although there is a significant quantitative increase after stabilization. Therefore, upon stabilization, more amounts of same proteins are retained in the NuMat, increasing the amounts of nuclear proteins retained in NuMat from 4–10%. MARs too show a similar trend as the unstabilized MARs are further enriched upon stabilization. These observations strengthen the possibility that internal NuMat exists in live cells and interventions like exposure to 37oC stabilizes it and facilitate its isolation.
Further, it has been shown that upon RNase A treatment, the meshwork of fine filaments populating the NuMat, collapses and aggregates onto the nucleolus and peripheral lamina. In order to understand the molecular details of this morphological effect and to determine the structure-function relationship of NuMat RNA, we have compared the NuMat proteome of RNase A treated versus untreated nuclei. Imaging analysis shows that large gaps appear in the meshwork on RNase A digestion. Many of the structural proteins like zipper (non-muscle myosin heavy chain), beta-spectrin, beta-tubulin, etc, are lost from the NuMat. Along with them, several transcription factors and RNA binding proteins are also lost upon removal of RNA from the structure. This confirms that the dynamic association of various proteins in the NuMat is managed by NuMat RNA, which plays an integral part in holding the functional matrix together.
Sequencing of MARs isolated from unstabilized, stabilized and RNase A treated NuMat provides further insight into the molecular dynamics of nuclear architecture. Apart from a set of sequences, that we call as Core-MARs here, most of the MARs are dynamic. The MARs occurring in promoter region and 5’-UTR of genes are the most dynamic of them all. They get associated with the NuMat during the stabilization of nuclei at 37°C, in an RNA dependent manner, since the same promoter and 5’-UTR MARs that are gained by stabilization, are lost after RNase A treatment of NuMat. The association of the MARs during stabilization possibly happens due to initiation of transcriptional activity and pausing of Pol II. Stalled Pol II has been shown to accumulate just downstream of the promoters in several previous studies and such promoters have already been found enriched in NuMat (10, 26, 27). Pol II pauses proximal to the promoter, where it waits for proper cellular cues to proceed for elongation. This mechanism provides a number of advantages in terms of rapid and synchronous activation of genes. Interestingly, some promoters show a strong tendency for Pol II pausing and this property is tightly linked to DNA sequence of the promoter (26, 28). We also observe that RNase A digestion not only detaches the promoter and 5’-UTR MARs from protein matrix, it also disrupts the architecture significantly. Our study sheds light on the critical debate about the mechanism by which stabilization makes the NuMat filaments more robust to EM observation. Our results suggest that the transcriptional activity and stalling of Pol II makes the NuMat fibrils more stable, and that this stability is dependent on, and is mediated by the transcribed RNA.
We next explored the dynamics of MARs as development proceeds in D. melanogaster embryos. Earlier studies from our group has shown that 65% of NuMat proteome in fruitflies is dynamic during development (29). Here we find that just like the protein composition of NuMat, MARs can also be either stable or dynamic. Stable MARs do not change between cell types as the embryo develops, which we have classified as Core-MARs. Dynamic MARs are the ones which change during development and are cell type dependent. We observe that 1277 MARs are stably associated with NuMat in 0–16 hour old embryos of which 943 MARs are maintained at all the three stages of development examined. They are distributed throughout the genome, with enrichment at pericentromeric region and on X chromosome. Published studies from our lab has shown that in pericentromeric heterochromatin, MARs correlate well with TADs and possibly play an instrumental role in expression of genes embedded in heterochromatin (30, 31). Also in Drosophila, centromeres of chromosomes are clustered into discrete nuclear structure known as chromocenter. Chromocenter ensures encapsulation of all chromosomes into a single nucleus and its disruption leads to micronuclei formation ultimately resulting in cell death (32). It is possible that Core-MARs play a basic role in maintenance of the nuclear architecture by helping in bundling and anchoring of chromocenter. More than half of the Core-MARs are in intergenic regions of the genome suggesting that these regions help in stably anchoring the genome to the underlying NuMat. Increased density of Core-MARs on X chromosome might be linked to the higher activity of genes for dosage compensation. The next abundant category is intronic MARs, which account for almost 1/4th of Core-MARs. About 22% of the Core-MARs lie in promoter region, 5’-UTR, 3’-UTR and exons of genes. Here they show an interesting enrichment at pseudogenes and non-protein coding genes. Most of these pseudogenes are from rRNA cluster on X chromosome and suppressor of stellate cluster on Y chromosome. Pseudogenes are fossil copies of parent genes, and are traditionally thought to be nonfunctional elements. Recently, however, some of them have been shown to be transcribed and are implicated in regulation of protein-coding genes. A comparative genomic study demonstrates that pseudogenes are highly linage specific and reflect genome history (33). NuMat association of such lineage specific elements as constitutive Core-MARs, proposes an interesting lineage specific feature of MAR elements. However, the significance of pseudogenes as Core-MARs, remains to be explored further.
Among the dynamic MARs, the most dramatic changes are observed in the MARs associated with the promoters and 5’-UTR of genes. The genes with MARs in their promoters and 5’-UTR regions are amongst the most highly expressed genes and are pre-disposed to Pol II pausing at promoter. Accordingly, only three MARs show up in the 5’-UTR regions in early embryos where transcription is negligible. It is known for a long time that the positioning of a gene at the NuMat increases transcription. For example, in “halo” preparations where loops of DNA stripped of histones remain attached to the NuMat, active genes localize to the bases of loops while inactive genes are peripheral (34). Our data too indicates that attachment to NuMat at promoter, UTRs and exonic region is, more often than not, involved in enhanced gene expression and NuMat provides the necessary platform for active transcription as well as anchors genes poised for transcription.
A curious observation from our study was that the total number of MARs increased from 0–2 hour embryos to 6–8 hour embryos but decreased in 14–16 hour embryos. This is intriguing because, as the number of cell types increase with development, number of MARs are expected to increase at later stages. One possible reason might be that for many of the dynamic MARs, originating from the increased number of cell types at later stages, may have increased the signal to noise ratio. In such a case, the below thresh-hold reads coming from a small number of cells, may have been removed from the dataset during analysis. The other possibility can be that MARs represent more of active regulatory elements like enhancers, repressors and insulators, that are in action at middle stages of development when the cell lineages are being decided. The regulatory activity reduces at later stages when cell fates are already assigned with concomitant decrease in number of MARs.
Epigenetic signature can be used to predict regulatory elements de novo. Cis-regulatory elements such as promoters, enhancers and insulators can be defined not only by the DNA sequence motifs but also by the patterns of bound proteins and histone modifications (35). Going by the logic that function of MARs as cis-regulatory elements must be relative to the trans-acting factor that binds to it, we looked into the types of protein factors that have overlapping binding sites with MARs. Our analysis indicates that a significant proportion of MARs overlap with regulatory elements. Several MARs overlap with transcription factor binding sites, underlining the involvement of these elements in facilitating transcription as promoters and enhancers. Many promoter and 5’-UTR MARs overlap with binding sites of architecturally important and insulator binding proteins like CP190, CTCF, BEAF 32, GAF, Su(Hw). This is not surprising as these factors have been mapped by ChIP-Seq to frequently associate with TSS (36). The HP1 family of proteins are found to bind mostly in intergenic region. However, a significant number of HP1 family of proteins binds with promoter MARs. This is also not surprising as a recent paper examined the genome-wide binding profile of HP1 family of proteins to find that in euchromatin, these proteins target genes that are highly expressed and exhibit Pol II pausing at TSS (37). The clustered binding of many of these proteins at paused Pol II promoter MARs consolidates the idea that such promoters can also act as insulators (38). A good number of intergenic and intronic MARs do not overlap with the binding of any of the protein factors examined. This maybe because ChIP data for known NuMat proteins (like NuMA, HnRNP-U/SAF-A, Lamin etc) is not available in Drosophila.
Epigenetic profile of MARs shows an enrichment of H3K4me3 along with H3K9ac and H3K27ac, a pattern that signifies active promoters, over promoter and 5’-UTR MARs. Intronic and intergenic MARs mostly have repressive chromatin mark of H3K9me3. However, only ~ 30% of the MARs in intronic/intergenic MARs show such an epigenetic profile. A majority of MARs appear to be devoid of histones and can be categorized as nucleosome free regions. Summing up the protein binding and epigenetic profiles, we find that around ~ 35% of the intronic/intergenic MARs neither carry epigenetic profile conducive of enhancers, insulators nor do they have repressive chromatin marks. These MARs appear to be devoid of nucleosomes and probably associate with NuMat, purely based on their DNA sequence feature. This property of MARs needs to be analyzed further in detail.