Hmga2 protein loss alters nuclear envelope and 3D chromatin structure

The absence of Hmga2 alters H3K27me3 deposition upon the induction of the transition from undifferentiated pluripotent stem cells into EpiLCs

Considering the general thought that Hmga2 is a chromatin protein, we explored the effects of Hmga2 suppression (Additional file 1: Fig. S1A) on histone modification dynamics taking place upon the transition of PSCs into EpiLCs. As shown in Fig. 1A (and Additional file 1: Fig. S1B), the trimethylation of K27 of histone H3 (H3K27me3) is affected in Hmga2 KO cells [16] induced to develop into EpiLCs.

Fig. 1figure 1

Perturbation of H3K27me3 levels in Hmga2 wt and KO cells upon the induction of EpiLCs. A Western blotting images showing the levels of histone modifications (H3K27me3, H3K4me3, H3K9me2, H3K9me3, and H3K27ac) in wt and KO PSCs at day 0 (undifferentiated cells) and at three different times after the induction of the transition into EpiLCs (day 1, day 2, and day 3). The levels of the histone H3 were used as control. Two wt and two KO clones were used. Quantification and statistical analysis are reported in Additional file 1: Fig. S1B. B Western blotting and relative quantification graphs showing the levels of H3K27me3, H3K27me2, and H3K27ac during EpiLC establishment (time course experiments: days 1, 2, and 3) of wt and KO cells. C Western blotting images and relative quantification graph showing the levels of H3K27me3 in wt and KO PSCs during SFEB differentiation. D Western blotting images and relative quantification graph showing the levels of the PRC2 proteins Suz12 and Ezh2 during EpiLC establishment (time course experiments: days 1, 2, and 3) of wt and KO cells. The data in the graphs represent the mean ± SD (n=3 biological replicates) of H3K27me3, H3K27me2, and H3K27ac against total H3 (panels B and C) or of Suz12 and Ezh2 signal intensity against Gapdh expressed as arbitrary units (a.u.). Student’s t-test, two tailed (ns: not significant, *p<0.05, **p<0.01)

Indeed, while in basal conditions (undifferentiated cells) the levels of H3K27me3 were comparable in Hmga2 wt and KO cells, they were maintained higher in Hmga2 KO cells throughout the three days of EpiLC induction. This difference between wt and KO clones is clearly appreciable in time course experiments, where the expected decrease of H3K27me3 is less pronounced in Hmga2 KO cells than in the wt cells, while the amount of H3K27me2 and H3K27ac showed similar dynamics in both cells (Fig. 1B).

To understand whether the observed effect on the variations of the epigenetic state due to Hmga2 absence is a general mechanism to allow pluripotent stem cells to leave the naïve state, we induced the differentiation of wt and KO Hmga2 PSC clones through the formation serum-free embryoid bodies (SFEBs). This 3D-model mimics different aspects of cell differentiation during early mammalian embryogenesis and it is particularly useful for the interpretation of KO phenotypes, allowing to by-pass the limitations related to 2D-systems [17, 18]. In this experimental setting, the different behavior of H3K27me3 between wt and KO cells was clearly confirmed, with a sustained presence of this histone mark at 2, 3, and 4 days after the induction of differentiation (Fig. 1C). Considering that H3K27me3 modification is catalyzed by polycomb repressive complex 2 (PRC2), we analyzed the expression of two subunits of the core PRC2, Suz12, and Ezh2, and we observed a slight but significant change of Suz12 levels that remained higher in KO cells throughout the induction of the transition (Fig. 1D). A similar behavior, although less pronounced, was also observed for Ezh2 (Fig. 1D).

Gene expression changes upon the induction of EpiLCs in the absence of Hmga2

The above-described results revealed that the absence of Hmga2 influences the abundance of the repressive histone modification H3K27me3 upon the exit of PSCs from the undifferentiated naïve state. To address whether this phenomenon is accompanied by a Hmga2-dependent change in the gene expression profile, we performed an RNA sequencing (RNA-seq) experiment to compare Hmga2 wt and KO cells at 24h after the induction of EpiLCs. The expression profiles of wt cells and two independent clones of Hmga2 KO cells were similar, with some exceptions (Fig. 2A). Indeed, we found 8341 genes whose expression was significantly modified 24 h after the exit from the naïve state of wt cells. Most of these genes were similarly modified in Hmga2 KO cells (Additional file 2: Table S1). There were 2311 genes whose expression was significantly modified only in wt cells and in none of the two Hmga2 KO clones, while the expression of 343 genes was significantly modified only in both Hmga2 KO clones and not in the wt cells. Then, we compared the list of genes, whose expression changes upon EpiLC induction appeared to be Hmga2-dependent, with the list of PRC2 putative target genes, available in GEO GSE74330 [21]. This comparison showed no significant enrichments.

Fig. 2figure 2

Hmga2 association with chromatin. A Comparison of the expression profiles of wt and Hmga2 KO cells induced to differentiate into EpiLCs. The profiles of the two cell types are similar (see also Additional file 2: Table S1) with some differences indicated in the Venn diagram. The heatmap shows the 2634 genes with a common behavior in the three cell clones, according to the Venn diagram. B DNA sequence motifs in the regions of the peaks of Chip-seq experiment with chromatin from wt EpiSC, immunoprecipitated with the anti-Hmga2 antibody and analyzed by MEME-chip software with default options. The three consensus sequences emerging from the analysis are shown. C Venn diagram of the alignment between Hmga2 peaks and those of chromatin markers H3K4me1 (GSM1382217 from ref. [19]) and H3K27ac (GSM1382219 from ref. [19]). D ChIP-seq data of Hmga2 in wt EpiSCs were aligned with available data of DamID in mouse ESCs (ref. [20], GSE17051). The results show a significantly overlap with the regions associated with Lamin b1 (about 51% Spearman correlation). The image shows a 196 Mb region of chromosome 1

Hmga2 association with chromatin in EpiSCs

There are numerous results indicating that Hmga2 interacts with chromatin at specific sites, thus contributing to gene regulation [11,12,13,14]. A genome-wide analysis of Hmga2 interaction with chromatin was already done in undifferentiated mouse ESCs overexpressing the recombinant protein [22] that demonstrated a diffuse interaction of Hmga2 with chromatin. We sought to explore endogenous Hmga2 association with chromatin and this was possible in EpiSCs (see the “Methods” section) where the levels of Hmga2 are enough high.

Thus, we performed a ChIP-seq experiment using a primary antibody against Hmga2 and this experiment allowed us to detect a diffuse interaction of endogenous Hmga2 with chromatin, confirming the previously published results obtained in undifferentiated ESCs [22]. In addition, we observed 2673 sites where Hmga2 accumulated as discrete peaks in our non-replicated experiment (2-fold, p< 0.001, see Additional file 3: Table S2). The analysis of the sequences of these Hmga2 peaks, by MEME-ChIP software, allowed us to identify three significant motifs (Fig. 2B).

These peaks were preferentially located at the inter-nucleosomal spacers, as demonstrated by the alignment of Hmga2 ChIP-seq with the data deriving from the MNase-seq experiments (ref. [23]) indicating that the number of the overlapping peaks is negligible.

A small number of Hmga2 peaks overlapped or were in close spatial relationship with putative enhancer elements marked by H3K4me1, H3K27ac, or both (Fig. 2C), while no significant overlapping between Hmga2 peaks and H3K27me3 peaks was observed. In all the cases analyzed, the histone marks of the putative enhancers were not significantly modified upon the silencing of Hmga2 in EpiSCs (Additional file 4: Fig. S2). Furthermore, no significant associations were found between the genes whose transcription start site (TTS) is located within 100 kb from these putative enhancers and the genes whose expression was different in wt vs Hmga2 KO cells.

The above-reported results do not support a direct relationship between Hmga2 association with chromatin, the observed changes in histone mark deposition and/or with gene expression modification in Hmga2 KO cells. However, it should be considered that the ChIP-seq experiment was done with chromatin from EpiSCs; thus, a different interaction pattern of Hmga2 during the early steps of this transition cannot be excluded.

Hmga2 loss alters the structure of the nuclear lamina in differentiating cells

The analysis of Hmga2 ChIP-seq data showed a significant association (~51% Spearman correlation) of this protein with non-transcribed regions and with Lamin-associated domains (LADs) [24] (Fig. 2D), confirming the results already observed in ESCs overexpressing Hmga2 [22]. Lamins guarantee the proper structure of the nuclear periphery and are essential to maintain an adequate 3D genome organization [25]. We asked whether the cells lacking Hmga2 suffer from alterations of the NL. Therefore, we performed Lamin b1 (Lmnb1) immunofluorescence on Hmga2 wild-type and knock-out cells during the transition into EpiLCs. At day 1, Hmga2 KO cells showed a remarkable distortion of the NL compared to the wt cells, with many nuclear protrusions and a clear nuclear blebbing (Fig. 3A, B).

Fig. 3figure 3

Hmga2 suppression causes a remarkable distortion of the nuclear lamina. A Lmnb1 immunofluorescence (red) on Hmga2 wt and KO cells at day 1 of EpiLC induction, showing the distortion of the NL in Hmga2 KO cells. Maximum projection of z-slices (ROI 1024×1024) is shown; quantification graphs obtained by counting >200 nuclei/condition. B Representative images (single Z-plane, ROI 1024×1024) of the nuclear abnormalities (nuclear protrusions and nuclear blebbing) observed in Hmga2 KO cells at day 1 of EpiLC induction and absent in wt cells; quantification graphs obtained by counting >200 nuclei/condition. C Three-dimensional analysis of z-slices maximum projections showing the distortion of Lmnb1 (red) as well as the reduced porosity (grey dots inside the NL) detected in Hmga2 KO cells compared to the wild-type ones. Scale bar= 10 μm. D Lmnb1 immunofluorescence (red) on Hmga2 KD ESCs induced into EpiLCs (day 3) showing an evident distortion of the NL, with some enlarged and wrinkly nuclei. Non-silencing siRNA was used as control. Quantification graphs obtained by counting >400 nuclei/condition. E Lmnb1 immunofluorescence on Hmga2 KD cells showing the nuclear abnormalities (nuclear blebbing, enlarged nuclei) typical of cells devoid of Hmga2. DAPI (blue) was used to counterstain the nuclei. A single plane of the z-stack projection is shown. Scale bars of A, B, D, and E = 50 μm. Error bars represent standard deviation. Statistical significance on three biological replicate experiments was determined using Student’s t-test (ns: not significant, *p<0.05, **p<0.01)

Tridimensional analysis of the z-slices maximum projections showed that Hmga2 KO cells suffered from nuclear weakness, and they were also characterized by a reduced porosity compared to the wt cells that, instead, had an intact NL, which appeared as circular, thick, and definite (Fig. 3C).

The observed altered immunostaining of Lmnb1 was only dependent on the structure of the NL, because Lmnb1 expression was unchanged in Hmga2 KO cells compared to wt cells (Additional file 5: Fig. S3A).

Noteworthy, no differences were observed between Hmga2 wild-type and KO cells in the undifferentiated state (Additional file 5: Fig. S3B). This evidence indicated that the distortion of the NL in Hmga2 KO cells occurred only when the cells were induced into EpiLCs, thus coinciding with the block of transition caused by Hmga2 loss [16].

To confirm the data obtained in Hmga2 KO cells, we examined the effects of Hmga2 silencing in wt mouse ESCs, upon the induction of EpiLCs. Hmga2 downregulation (see Additional file 5: Fig. S3C and D) strongly altered the structure of the NL, causing nuclear abnormalities similar to those observed in Hmga2 KO cells (Fig. 3D, E).

The nuclear lamina interposes between condensed chromatin and the inner nuclear membrane [26]. To examine the status of the inner nuclear membrane, we carried out the immunostaining with the anti-Emerin antibody in Hmga2 wt and KO cells as well as in Hmga2-silenced EpiLCs, as this protein of the inner nuclear membrane is molecularly connected to the nuclear lamina. As reported in Fig. 4, the structural abnormalities detected using Lmnb1 antibody were also observed with Emerin antibody: the immunostaining revealed that the inner nuclear membrane of the Hmga2 KO and Hmga2-silenced cells suffered from the distortion of the nuclear lamina. As for Lmnb1, no nuclear envelope defects were observed in Hmga2 KO cells compared to the wild-type cells in the undifferentiated conditions, confirming that the distortion occurred during the exit from the undifferentiated state (Additional file 6: Fig. S4).

Fig. 4figure 4

Hmga2 silencing induced a phenotype identical to that observed in Hmga2 KO cells. A Lmnb1 (red) and Emerin (green) immunofluorescence on Hmga2 wt and KO cells (day 1 of EpiLC induction) showing the nuclear abnormalities typical of Hmga2 KO cells (single z-plane, ROI 1024×1024, scale bar= 50μm). Quantification graphs obtained by counting >150 nuclei/condition. B Lmnb1 (red) and Emerin (green) immunofluorescence at day 3 after the induction of EpiLCs upon Hmga2 silencing. Nuclear blebbing and enlarged nuclei were surrounded by both Lmnb1 and Emerin. DAPI (blue) was used to counterstain the nuclei. A single plane of the z-stack projection is shown. Scale bar= 50 μm. Quantification graphs obtained by counting >300 nuclei/condition. Error bars represent standard deviation. Statistical significance on three biological replicate experiments was determined using Student’s t-test (***p<0.001)

Altogether, these data demonstrate that the absence of Hmga2 inside the nucleus prevents maintaining the proper structure of the nuclear lamina.

Hmga2 loss affects the intranuclear distribution of H3K9me3

Chromatin associated with nuclear lamina is marked by heterochromatic histone modifications; thus, we investigated whether changes regarding the distribution and the abundance of these histone modifications occurred in Hmga2 KO cells. First, we conducted immunofluorescence experiments to investigate H3K9me2/3 and H3K27me3 localization in Hmga2 KO cells at day 1 of transition into EpiLCs. These experiments allowed us to detect a widespread mis-localization of H3K9me3 in differentiating Hmga2 KO cells (Fig. 5A).

Fig. 5figure 5

Mislocalization of H3K9me3 histone mark in Hmga2 KO and KD cells upon EpiLC induction. A Immunoflurescence experiments on Hmga2 wt and KO cells at day 1 after the induction of EpiLCs of Lmnb1 (red) and H3K9me3 (green). B Lmnb1 (red) and H3K9me3 (green) immunofluorescence on Hmga2 KD cells (day 3 after the induction). Non-silencing siRNA was used as control. In all immunofluorescence experiments, DAPI (blue) was used to counterstain the nuclei. All pictures are shown as single z-plane, ROI 1024×1024, scale bar= 50μm. Quantification graphs show the percentage of Hmga2 wt and KO cells with H3K9me3 mislocalization (n≥200 nuclei per condition) and the percentage of Hmga2 KD cells showing H3K9me3 mis-localization (n≥ 300 nuclei per condition). The count for Hmga2 wt and KO cells was performed at day 1 of differentiation, while the count for of Hmga2 KD cells was performed at day 3 of EpiLC transition, on three biological replicates for each condition. Error bars represent standard deviation. Statistical significance was determined using Student’s t-test (ns: not significant, ***p<0.001)

In detail, in the absence of Hmga2, H3K9me3 signal was diffuse inside the nucleus and not organized in intense dots at chromocenters (pericentric constitutive heterochromatin represented by DAPI-dense foci), as observed in the wild-type cells (Fig. 5A). H3K9me2 signal was mostly located close to nuclear envelope so its distribution in KO cells confirmed the distortion of nuclear lamina (Additional file 7: Fig. S5A). These results suggested that Hmga2 loss affected H3K9me3 distribution, with effects on peripheral and internal heterochromatin. The distribution of H3K27me3 histone modification was not affected: Hmga2 KO cells were characterized by small intense dots marked by H3K27me3 whose distribution was similar to that observed in wt cells (Additional file 7: Fig. S5B). A similar behavior was also observed in case of H3K4me3 and H3K27ac, which were similar in wt and KO cells (Additional file 7: Fig. S5C-D). In addition, we explored whether the increased levels of H3K27me3 and the slight accumulation of PRC2 subunits, Suz12 and Ezh2 (see Fig. 1), could have any role in the lamina-associated phenotype observed in Hmga2 KO cells. To this aim, we inhibited PRC2 by treating the cells with GSK126 for 72 h and inducing the transition towards EpiLCs for 24h. Although the treatment led to a clear decrease of H3K27me3, we observed the same alterations of the nuclear lamina as in mock treated cells (Additional file 8: Fig. S6).

Hmga2 silencing in ESCs induced toward EpiLCs also confirmed the altered localization of H3K9me3 (Fig. 5B). It is worth noting that H3K9me3 and H3K27me3 accumulate within or close to nuclear blebs observed in Hmga2 KO cells thus confirming the heterochromatin enrichment in nuclear abnormalities (HENA) observed in other experimental settings [27].

Inter-TAD interactions and their association with nuclear lamina are altered in Hmga2 KO cells

The crosstalk between LADs with whole genome organization [25] and with the dynamics of TAD cliques during differentiation [20, 28] was clearly demonstrated in several conditions. On the other hand, in some cases, LAD borders are marked by H3K27me3 [24, 29], and the suppression of Ezh2, the catalytic subunits responsible for H3K27me3 deposition, prevents NL interaction with chromatin [27]. Thus, we hypothesized that defects in nuclear shape and heterochromatin mis-localization accompanying Hmga2 suppression could be related to defects in 3D genome assembly.

To explore the effects of Hmga2 on genome 3D structure, we performed the Hi-C mapping in two independent clones of wt cells and in two clones of Hmga2 KO cells, at day 3 after the induction of the transition into EpiLCs. After the filtering and mapping of Hi-C data of the four clones, we obtained 38,483,563; 55,451,307; 45,019,690; and 53,506,473 valid interactions, respectively (Additional file 9: Table S3).

The contact maps of all the chromosome and of an example (chromosome 11) are shown in Fig. 6A. Visual inspection of these maps indicate that the contact pattern is very similar in wt cells and in KO cells. To quantify this observation, we performed PCA analysis and identified A/B compartments genome-wide in wt and KO (Additional file 10: Fig. S7A) [30]. Comparison between compartments in different conditions reveals a general high correlation (genome-wide average Pearson r=0.96, Additional file 10: Fig. S7B), with few, short regions, typically located at the compartment periphery, exhibiting changes in compartment identity (Additional file 10: Fig. S7C, Additional file 11: Table S4). Next, we quantified the similarity of TAD boundaries [31, 32] in the different systems, called by their Insulation Score [33]. Such an analysis confirmed that TAD boundaries are similar in the four samples, with an overlap, in the case of chromosome 11, that was always close to 90% at 25 kb (see Fig. 6B and C and the “Methods” section), i.e., comparable to replicate similarity. Genome-wide analysis made with data at 25 kb indicated a global overlap close to 90%. Thus, the results indicated that the suppression of Hmga2 does not alter TAD structure. The correlation of boundaries with epigenetic marks and with the binding of Hmga2 to chromatin confirmed that H3K4me3 is tightly associated with the boundaries [31], while no significant associations were observed in the case of H3K27me3 as previously reported. Interestingly, though, Hmga2 abundance on chromatin appears to be negatively associated with the boundaries (see Fig. 6D), highlighting a novel potential involvement of Hmga2 in TAD architecture.

Fig. 6figure 6

3D genome structure in wt and Hmga2 KO cells. A Left: Genome-wide Hi-C data in wt and Hmga2 KO conditions at 5Mb resolution, right: Hi-C data of chromosome 11 in both conditions at 250Kb resolution. B Hi-C maps in the region chr11:105000000-116000000, 25-Kb resolution. Insulation score and boundaries are reported below. C Overlap between boundaries (detected at 50kb resolution) in wt and Hmga2 KO for chromosome 11. D Enrichment in WT of H3K4me3, H3K27me3, and Hmga2 (normalized) at TAD boundaries, shaded area represents the standard error of each point

Next, we explored the effects of Hmga2 KO on inter-TAD interactions. To this aim, we considered the average interaction between two given TADs and calculated the log ratio (or log fold-change, indicated as Log FC) between their interaction in wt cells vs Hmga2 KO cells. Figure 7A summarizes the computational approach to address this point.

Fig. 7figure 7

Hmga2 KO alters inter-TAD interactions and their association with nuclear lamina. A The computational approach to analyze the effects of Hmga2 KO on inter-TAD interactions was to calculate the Log fold change between the interactions between TADs in wt and Hmga2 KO. B Fold changes between all TAD pairs were calculated for each chromosome; the graph reports the distribution of inter-TAD interactions in chromosome 1, showing that there are numerous interactions that are decreased or increased in Hmga2 KO cells compared to wt cells. C Genome-wide enrichment of DamID signal [33] in TAD pairs exhibiting the most prominent interaction change

The analysis of the log fold changes allowed us to identify the TAD pairs with most prominent differences between wt and KO cells, defined as those where LogFC >1 or <−1 (Fig. 7B, chromosome 1). In a significant number of cases, the interaction present in wt cells is decreased or increased by a factor of 2, highlighting that Hmga2, although not affecting TAD boundaries, has a role in the formation of chromatin 3D structure. Considering the phenotype of Hmga2 KO EpiLCs, and the demonstrated role of Lamins in the organization of global 3D genome structure [34], we asked whether there is any relationship between the changes we observed in the interactions among TADs and the chromatin regions preferentially associated with the nuclear lamina (DamID data taken from ref. [34]). To this aim, we selected all 787 pairs of TADs whose interaction is significantly decreased in Hmga2 KO cells (Log FC >1) and all 645 pairs that interact more in Hmga2 KO cells than in wt cells (LogFC<-1). The results show that, in most cases (77%) of TAD pairs with decreased interaction in KO cells, only one of the TADs is located in a region associated with Lamin, while only 23% of the pairs are either both associated with Lamin or not associated with Lamin (Fig. 7C). On the contrary, the TAD pairs showing an increased interaction in Hmga2 KO cells compared to wt cells are almost uniformly distributed (54% with only one of the two TADs associated with Lamin and 46% both associated with Lamin or not associated with Lamin). To check the robustness of these results, we repeated the above-described analysis using a random control system where we permuted (bootstrap) the boundary positions along the genome (Additional file 12: Fig. S8). Although TAD pairs were selected with the same criterion, no specific DamID enrichment pattern is observed in TAD pairs with increased or decreased interaction in KO as compared to the wt condition.

Finally, we checked that both types of TAD pairs exhibit analogous size distributions (Kolmogoroff-Smirnov test p-val > 0.05), so to exclude size-related effects. Taken together, our results highlight that the known role of Hmga2 in defining chromatin interaction with the lamina affects the establishment of TAD borders and inter-TAD interactions.

留言 (0)

沒有登入
gif