Chromatin and transcription analysis of the 4q35 region reveals distinct functional domains
To investigate the mechanism(s) by which the D4Z4 macrosatellite array affects the transcriptional regulation of the 4q35 locus (Figure 1A), we measured RNA levels of genes located at different distances from the D4Z4 array. We analyzed human primary myoblasts (HPMs) and trophoblast derived cells (HTCs) obtained from FSHD subjects heterozygous for a D4Z4 reduced allele (DRA) and matched controls (Figure 1B) [46]. On the centromere-proximal side of 4q35, we analyzed RNAs from these protein-coding genes (Figure 1A): ANT1 (Adenine Nucleotide Translcator1; known also as Solute carrier family 25 member 4 SLC25A4), FAT1 (FAT atypical cadherin 1), and FRG1 (FSHD region gene 1), which are are positioned at 4.9 Mb, 3.5 Mb and 127 kb from the D4Z4 array, respectively. We also analyzed telomere-proximal RNAs, including FRG2 (FSHD region gene 2; 37 kb from the array), and DBE-T (D4Z4 Binding Element-Transcript), a lncRNA transcribed from the 5’ end of the D4Z4 array. We also analyzed two additional transcripts derived from D4Z4: DUX4 exon1 (Ex1)-containing transcripts, hereafter named D4Z4-T (D4Z4-Transcript), which arise from each D4Z4 repeat, and DUX4FL (DUX4 Full Length) pLAM-containing transcripts, the FSHD disease-associated mRNAs derived from the most telomere-proximal D4Z4 repeat (Fig. 1A and Supplemental_Fig_S1).
Our qPCR analyses (Fig 1C-D) showed that 4q35 genes are differentially expressed depending on their chromosomal position, confirming previous evidence [27,28,41,47]. Specifically, in control cells the centromere-proximal ANT1, FAT1 and FRG1 genes displayed high levels of expression, whereas FRG2, DBE-T and D4Z4-T were barely detectable. Consistent with the expected derepression of the locus associated with reduced D4Z4 copy number, FRG2, DBE-T and D4Z4-T transcripts were significantly upregulated in FSHD1 cells (Figure 1C and D). In contrast, ANT1 and FRG1 transcripts were found at comparable levels both in control and FSHD cells. Therefore, the loss of D4Z4 satellites in the FSHD patients’ cells correlated with an altered regulation of the telomere-proximal transcripts.
We could not reliably quantify the DUX4FL transcript via qPCR with commonly used primers [48,49] due to the very low amount of the detected amplicon (Ct values over 35) and because of the presence of multiple peaks in the melting curve analysis of the PCR products (Supplemental_Fig_S1). These observations are consistent with previous detection of DUX4FL transcripts only in a small percentage of FSHD-derived myoblasts (1 per 1000 cells) [44] and in cell lines only by using multi-step nested PCR [44,49–51].Since our aim was to conduct unbiased analyses of the physiological levels of 4q35 transcripts to compare their expression and regulation, we avoided pre-amplification steps.
The distinct regulation of genes located at different distances from the D4Z4 array prompted us to investigate the chromatin features of the 4q35 region. Chromatin immuno-precipitation (ChIP) experiments were performed in primary HPMs and HTCs using antibodies raised against histones tail modifications associated with transcriptional regulation: AcH3, AcH4 and H3K4me3, H3K9me3 and H3K27me3 (Figures 1E-F, values in Supplemental_Table1). We detected three classes of modification patterns. First, in both primary cell types, active chromatin marks were found at the ANT1 and FAT1 promoters within the centromere-proximal region of 4q35 region, with enrichment of AcH3, AcH4, and H3K4me3, and low levels of H3K9me3 and H3K27me3. Second, at the FRG1 and FRG2 promoters, both repressing and activating marks were detected. This “bivalent” pattern of histone modification is characteristic of “poised” promoters that are inactive but able to respond to external stimuli [52–54]. Specifically, we detected the enriched levels of AcH4 and H3K9me3 at the FRG1 promoter in both cell types, at comparable levels in control and DRA-bearing cells. At the FRG2 promoter, we observed comparable levels of AcH4 in control and FSHD-derived cells, whereas H3K9me3 enrichment was significatively more elevated in FSHD-derived myoblasts compared to control cells. A poised chromatin modification pattern at the FRG2 promoter is a conserved feature in multiple cell types, as confirmed by Chromatin State Segmentation by HMM from studies by ENCODE/Broad [55] (Supplemental_Fig_S2). The third modification pattern observed occurred at the D4Z4 repeats themselves and was dominated by strong enrichment of H3K9me3 associated with low levels of the other chromatin marks analyzed (Figure 1E-F). A similar signature was also found at the gene desert region LILA5 (Supplemental_Fig_S3), as expected in heterochromatic regions. Notably, all three classes of modification patterns were largely similar comparing samples from healthy individuals and from FSHD patients with D4Z4 deleted alleles (DRAs). However, H3K9me3 levels were significantly increased at the FRG2 promoter and at D4Z4 in primary cells carrying a DRA [13,38]. This might seem to conflict with our observation that both FRG2 and D4Z4-T were derepressed in cells carrying a DRA (Fig 1B and C). However, these data are consistent with previous results cells from patients with ICF (Immunodeficiency, Centromeric region instability, Facial anomalies syndrome), in which D4Z4 transcription is detected in spite of retention of H3K9me3 [56]. In sum, 4q35 contains three subdomains with euchromatic, poised, and heterochromatic features, arrayed in a centromere-to-telomere order.
Figure 1. 4q35 genes expression and epigenetic profile: A) Schematic representation of the chromosome 4q35 showing physical distances between ANT1, FAT1, FRG1 and FRG2 genes and the D4Z4 macrosatellite within the AF146191-U85056 contig, based on GenBank entry U85056.1. The positions of oligonucleotides used in ChIP experiments (red) and qPCR (blue) are shown. B) Table showing the sizes of the two 4q35 and 10q26 alleles in the selected human primary cells used in this paper, together with the 4q-ter (4qA) haplotype. Control human trophoblast cells (HTCs) and human primary myoblasts (HPMs) cells carry normal-sized 4q alleles (>10 D4Z4 repeat units), whereas FSHD-derived HTCs and HPMs bear a reduced D4Z4 allele (DRA), i.e. < 8 D4Z4 repeat units (U=Units); C-D) RT q-PCR quantification of ANT1, FAT1, FRG1, FRG2, DBE-T, D4Z4-T and DUX4FL mRNAs in (C) human primary myoblasts (HPMs) and (D) human trophoblast cells (HTCs). Data were normalized using RPLP0 as a reference mRNA. E-F) Chromatin immunoprecipitation (ChIP) analysis performed in (E) HPMs and (F) HTCs. IPs were performed using the indicated antibodies recognizing H3K4me3, H3K9me3, H3K27me3 and pan-acetylated Histone 3 and 4 (AcH3 and AcH4), or a non-specific control (IgG), followed by qPCR amplification using primers described in Fig.1A. Data are displayed as the percent enrichment for each antibody over total input chromatin. Experiments were done in triplicate and analyzed using two-way Anova statistical tests. Asterisks indicate the statistical significance of data obtained in DRA cells compared to control cells for each antibody, as follows: * 0.05<p-value<0.01; ** 0.01<p-value<0.001; *** 0.001<p-value<0.0001; **** p-value<0.0001.
The FRG2 promoter is activated by inhibition of histone acetylation or PARP1, in a manner regulated by D4Z4 repeat length
The detection of FRG2 and D4Z4 transcription despite the enrichment of heterochromatin-associated histone marks at their promoters suggested complex modes of regulation. To investigate the role of chromatin modification across 4q35, we pharmacologically inhibited different classes of chromatin-modifying enzymes both in HTCs and HPMs (Fig. 2) and measured the RNA levels by RT-qPCR. We treated cells with trichostatin A (TSA), an inhibitor of class I, II, IV, histone deacetylases (HDACs) [57], nicotinamide (NAM) [58], an inhibitor of class III HDACs (sirtuins), or PJ34, an inhibitor of Poly (ADP-ribose) polymerase-1 (PARP-1) [59–61]. These treatments were performed either in presence or absence of 5-Aza-dC (Aza), an inhibitor of DNA methylation [57]. After these treatments, minor changes in expression of ANT1, FAT1 and FRG1 (Figure 2A-C and G-I) were observed. In contrast, strong transcriptional induction of FRG2 was observed upon TSA and PJ34 treatments (Figure 2D and J, note the y-axis scale). The effects of both these compounds were enhanced by 5-aza-dC, indicating that DNA methylation contributes to FRG2 silencing. Like the centromere-proximal genes, transcription of DBE and D4Z4-T transcripts was not induced by the selected compounds (Figure 2E-F and K-L). We conclude that the FRG2 promoter is particularly sensitive to local chromatin modifications.
To determine how TSA-induced transcriptional changes correlated with altered histone modifications, we performed ChIP experiments (Supplemental_Fig_S4, data in Supplemental Table 3). Both in control and DRA-bearing cells, TSA led to a general increase of ‘open chromatin’ marks (acetylated H3/H4 and H3K4me3) at 4q35 genes. In particular, and consistent with its transcriptional upregulation, we observed increased H3 and H4 acetylation at the poised FRG2 promoter in HPMs and HTCs bearing one DRA allele (Supplemental_Fig_S4B-C and E). Together, our data indicate that histone acetylation and D4Z4 repeat length both contribute to the robust inducibility of the FRG2 promoter.
Figure 2. 4q35 gene expression is affected by epigenetic drugs depending on 4q allele size. Expression data of 4q35 genes in human primary myoblasts (HPMs) (A-F) and in human trophoblasts cells (HTCs) (G-L) carrying a normal sized allele (CTRL (>10U)) or D4Z4 reduced alleles (DRA:7U and DRA:4U). Cells were treated or not treated with the indicated compounds: 5-Aza-2'-deoxycytidine (5-Aza-dC), Trichostatin A (TSA), nicotinamide (NAM), PARP inhibitor (PJ34). ANT1 (A), FAT1 (B), FRG1 (C), FRG2 (D), DBE-T (E) and D4Z4-T (F) RNAs were measured by RT q-PCR and normalized over the RPLP0 reference gene. Experiments were done in triplicate and the results were analyzed using two-way Anova tests to perform multiple comparisons. Hashtags (#) indicate the statistical significance of data from treated samples compared to untreated samples (NT) in each group. Asterisks (*)indicate statistical significance of data from treated cells carrying DRA compared to the same treatment in control cells. P-value ranges are as follows: *, # 0.05<p-value<0.01; **, ## 0.01<p-value<0.001; ***, ### 0.001<p-value<0.0001; ****, #### p-value<0.0001.
Transcription of FRG2 and D4Z4 macrosatellite sequences is induced by genotoxic agents
To test whether 4q35 transcription was affected by a wider array of environmental perturbations, we analyzed the effects of genotoxic agents. We treated primary cells with Cisplatin (CIS) [62], Etoposide (ETO) [63] and Doxorubicin (DOXO) [64,65] (Figures 3 and Supplemental_FigS5). The centromere-proximal genes (ANT1, FAT1 and FRG1) were mildly repressed upon genotoxic injury. In contrast, expression of FRG2, DUX4-T and DBE-T increased significantly in the presence of all these compounds both in control and FSHD-derived cells. Additionally, genotoxic agents increased the amounts of 4q35 telomeric transcripts FRG2, DBE-T and D4Z4-T significantly more in HPMs and HTCs bearing a DRA in comparison with cells bearing normal sized D4Z4 alleles. We conclude that RNA levels from telomere-proximal 4q35 genes are induced by genotoxic agents, and the magnitude of this effect is increased by the presence of shortened D4Z4 arrays.
To investigate more deeply the chromatin changes at 4q35 in response to DNA damage, we evaluated the amounts of histone isoforms that serve as DNA damage indicators: phosphorylated H2AX (gH2AX), which appears at DNA double-strand breaks, and macroH2A1.1, which is recruited to sites of DNA damage-induced PARP1 activation [65–67] (Figure 3G-H, values in Supplemental_Table2). At FRG2, we detected low levels of both these histones in the absence of DOXO treatment. At D4Z4, macroH2A.1 and 𝛄H2AX are present in basal levels untreated control cells, and these levels increased in cells carrying a 4U DRA or when DNA damage was induced by DOXO (Figure 3N). These observations suggest distinct chromatin architectures at 4q35 alleles that contain a DRA, in which the macrosatellite deletion renders the locus more accessible to DNA damaging agents and/or to DNA damage response factors.
We also evaluated histone modifications at the 4q35 genes in response to exogenous stresses (Figure 3I-N and Supplemental_Fig_S6, values in Supplemental_Table 2 and 4). Our analysis revealed that the DOXO-mediated increase of FRG2, DBE-T and D4Z4-T transcripts was not associated with local enrichment of histone acetylation or other transcription-associated histone modifications (Figure 3I-K and Supplemental Fig. S6A-B). Instead, a significant increase of H3K9me3 at D4Z4 loci (p-values <0.001 and <0.01) was observed. Furthermore, H3K9me3 levels became significantly greater in cells bearing a DRA than in control cells, both at the FRG2 promoter (p-value <0.001) and the D4Z4 array (p-value <0.05). Treatment with PARP inhibitor PJ34 also induced a robust increase in H3me3K9 (p-value <0.001) at FRG2 promoter and D4Z4 in DRA cells (Figure 3L-N and Supplemental Fig S6C-D). Therefore, the increased RNA levels observed at the 4q35 telomere-proximal genes upon genotoxic stress or PARP inhibition are paradoxically accompanied by increased H3K9 methylation.
Figure 3 .4q35 genes show different responsiveness to DNA damage depending on D4Z4 size and subtelomeric localization.
Control HPMs and HPMs bearing 7U and 4U D4Z4 arrays were untreated or treated with genotoxic drugs: Doxorubicin (DOXO), Etoposide (ETO) and Cisplatin (CIS), at the reported concentrations. Expression data of ANT1 (A), FAT1 (B), FRG1 (C), FRG2 (D), DBE-T (E) and D4Z4-T (F) was evaluated 24h after treatments and normalized over RPLP0 reference gene levels. Error bars represent standard deviation values for three independent replicates. Hashtags refer to statistical significance of treated samples in respect to not treated samples. Asterisks refer to statistical significance of treated cells carrying DRA in respect to the same treatment in control cells (NT) in each group. G-L) Chromatin immunoprecipitation assays (ChIP) conducted in control and DRA HPMs that were untreated or treated with Doxorubicin (G-I) or PJ34 (J-L). Antibodies directed to H3K4me3, H3K9me3, H3K27me3 and pan-acetylated Histone 3 and 4 (AcH3 and AcH4) were used, followed by qPCR amplification using primers described in Fig.1A. Anova statistical test with multiple comparison was performed (*0.05<p-value<0.01; ** 0.01<p-value<0.001; *** 0.001<p-value<0.0001; **** p-value<0.0001). Different symbols: * (asterisk) + sign (plus sign) and # (hashtag) refer to different antibodies used in ChIP experiments (*=AcH4; +=AcH34me3; #=H3K9me3 to show the statistical significance of data obtained in treated cells in respect to the same in not treated cells). M-N) Chromatin Immunoprecipitation (ChIP) in HPM cells that were untreated or treated with Doxorubicin (M) or PJ34 (N) . Antibodies directed to γH2Ax and macroH2A1.1 (mH2A1.1) were used followed by qPCR amplification of 4q35 genes as indicated. Anova statistical test with multiple comparison was performed (*0.05<p value<0.01; ** 0.01<p value<0.001; *** 0.001<p value<0.0001; **** P value<0.0001. * (asterisks) refer to each different antibody used in ChIP experiment to show the statistical significance of data obtained in treated cells in respect to the same in not treated cells.
Transcripts from telomere-proximal 4q35 genes are post-transcriptionally stabilized upon DNA damage
We observed that steady-state FRG2 transcript levels were induced by TSA, in cells with DRA alleles this was accompanied by increased histone H3/H4 acetylation (Figure 2). In contrast, the increased transcript levels of FRG2, DBE-T and D4Z4-T upon genotoxic injury were not correlated with histone modifications typical for transcriptional activation increased H3K9me3 levels (Figure 3).
These observations were inconsistent with typical gene activation scenarios, but we reasoned that they could be consistent with post-transcriptional stabilization of telomeric 4q35 transcripts in presence of genotoxic damage. To test this, we treated control or FSHD HTCs with Actinomycin D (ActD), at concentration sufficient to inhibit transcription by both RNA polymerase I and II [68] and then evaluated the stability of 4q35 transcripts over time. Experiments were performed with ActD alone or in presence of DOXO (Figure 4). Notably, the expected increase in FRG2, DBE-T and DUX4-T transcript levels in the presence of DOXO was also observed when transcription was inhibited by ActD treatment (Fig. 4D-F), supporting our idea about post-transcriptional stabilization. Also, quantification of the data detected longer half-lives for these three telomere-proximal RNA species in cells with DRA alleles (Fig. 4A). We conclude that the major regulatory event for the 4q35 telomere-proximal transcripts upon genotoxic stress is post-transcriptional stabilization.
Figure 4. 4q35 telomeric transcripts are stabilized upon DNA damage and transcriptional inhibition dependently on D4Z4 size reduction.
A) Table reports the half-life of 4q35 transcripts measured after Actinomycin D (ActD) treatment in Control (CTRL) and DRA-containing (4U) HPMs. B-F) Cells were treated with ActD for the indicated times (30’, 1h, 2h, 4h and 6 h) in presence or absence of DOXO, and the levels of 4q35 gene transcripts were evaluated by qPCR. The half-lifes of each RNA was calculated as the time needed to reduce the transcript level to half (50%) of its initial abundance at time 0. Data shown are means ± s.e.m. of 3 replicates.
Transcripts from 4q35 telomere-proximal genes are chromatin-associated
The observation that FRG2, DBE-T and DUX4-T transcript levels are affected by the same stimuli raises the question whether these RNAs have additional commonalities. Since repetitive element RNAs often function as components of chromatin fibers [69] , we performed RNA fractionation experiments in primary control or FSHD-derived myoblasts (Fig. 5A-B). In both cell samples, FRG2 and D4Z4-T RNAs were enriched in the chromatin-associated fraction and behaved similarly to the previously characterized chromatin-associated transcripts lncDBE-T and TERRA [39,70]. As controls for the fractionation, we confirmed that the lncRNA NEAT1, was prevalently found in the nuclear fraction, and the protein-coding mRNA GAPDH was preferentially enriched in the cytoplasm.
The chromatin association of the FRG2, DBE-T and DUX4 transcripts was confirmed by Chromatin-RNA Immuno-Precipitation (ChRIP) [71] conducted in in primary myoblasts from control and FSHD subjects using H3K4me3, H3K9me3 and H3K27me3-specific antibodies (Fig. 5C). FRG2, DBE-T and DUX4 transcripts were selectively and significatively enriched in H3K9me3 and H3K27me3-marked chromatin. As a control for the selectivity of our analysis, we confirmed that lncRNA Neat1, known to be associated with actively transcribed genes, was enriched in H3K4me3-marked but not H3K9me3 or H3K27me3-marked chromatin [72].
Together, our findings reveal that the telomere-proximal 4q35 genes share important regulatory features: their transcript levels are induced by genotoxic stress via post-transcriptional stabilization, and these RNAs are all chromatin-associated (Figure 6). Furthermore, the regulation of these telomeric transcripts is affected by the size of the D4Z4 sub-telomeric array (Figures 2 and 3). Therefore, the regulatory potentialc of this locus is expected to be variable in the human population.
Figure 5. 4q35 genes regulation upon different stimuli reflects architectural and epigentic patterns. A-B) RNA fractionation experiments were conducted in CTRL (A) and DRA (B) human primary myoblasts (HPMs). Transcripts from the indicated 4q35 genes were measured by qPCR analysis of cytoplasmic, nuclear and chromatin-associated RNA fractions, and the percentage detected in each fraction over total RNA was graphed. GAPDH, TERRA and NEAT1 transcripts were also assessed as positive controls that are most enriched in cytoplasmic, chromatinic and nuclear fractions, respectively. C) ChRIP experiment performed in HPM cells. Antibodies directed to H3K4me3, H3K9me3, H3K27me3 were used to precipitate RNA from control and DRA cells. Data shown are means ± s.e.m. of 3 replicates. * (asterisks) refer to each different antibody used in ChRIP experiment to show the statistical significance of data obtained for each antibody over control IgG.
Figure 6. 4q35 genes regulation upon different stimuli refects architectural and epigenetic patterns. Top diagram: A topological domain (TAD, indicated by the magenta triangle) at 4q35 includes the FRG1 and FRG2 genes [73]( 190–191 Mb of Chr 4). Additional cis-interactions between D4Z4 and nearby genes have also been reported [27,74] (curved lines). Our present study indicates functional subdomains within the 4q35 subtelomere, arrayed in a gradient along the chromosome. The different chromatin configurations at each subdomain correlate with the different response of these regions to external stimuli. The centromere-proximal genes ANT1 and FRG1 display active histone marks and are constitutively expressed at high levels (Chromatin domain 1). In contrast, the FRG2 promoter displays a poised promoter (Chromatin domain 2). Finally, the telomeric genes at the D4Z4 repeats (Chromatin domain 3) display repressive chromatin marks and are transcriptionally repressed in normal individuals in the absence of genotoxic stress. Middle diagram: Drug-induced epigenetic derepression (i.e. TSA treatment) results in enrichment of active histone marks at chromatin domain 1 promoters and a switch toward active chromatin at the FRG2 promoter leading to increased RNA levels. Bottom diagram: DNA damage (i.e. DOXO treatment), globally reduces the transcriptional activity across 4q35 and mediates a switch towards increased repressive chromatin markings at D4Z4 and the FRG2 promoter. Additionally, transcripts from Chromatin domains 2 and 3 are stabilized through a posttranscriptional event. This model applies to control and to cells carrying a reduced D4Z4