4q-D4Z4 chromatin architecture regulates the transcription of muscle atrophic genes in facioscapulohumeral muscular dystrophy

Abstract

Despite increasing insights in genome structure organization, the role of DNA repetitive elements, accounting for more than two thirds of the human genome, remains elusive. Facioscapulohumeral muscular dystrophy (FSHD) is associated with deletion of D4Z4 repeat array below 11 units at 4q35.2. It is known that the deletion alters chromatin structure in cis, leading to gene up-regulation. Here we show a genome-wide role of 4q-D4Z4 array in modulating gene expression via 3D nuclear contacts. We have developed an integrated strategy of 4q-D4Z4-specific 4C-seq and chromatin segmentation analyses, showing that 4q-D4Z4 3D interactome and chromatin states of interacting genes are impaired in FSHD1 condition; in particular, genes that have lost the 4q-D4Z4 interaction and with a more active chromatin state are enriched for muscle atrophy transcriptional signature. Expression level of these genes is restored by the interaction with an ectopic 4q-D4Z4 array, suggesting that the repeat directly modulates the transcription of contacted targets. Of note, the up-regulation of atrophic genes is a common feature of several FSHD1 and FSHD2 patients, indicating that we have identified a core set of deregulated genes involved in FSHD pathophysiology.

Figure 1.

4q-D4Z4–specific 4C-seq highlights FSHD1 impaired interactome. (A, left) Circos plot (Krzywinski et al. 2009) depicting cis and trans 4q-D4Z4 interactions in CN (CN-3, CN-4) myoblasts (MBs) called by 4C-ker. Common interactions with FSHD1 (FSHD1-3, FSHD1-4) MBs are in gray, whereas interactions specifically lost in FSHD1 are highlighted in light blue. (Right) Zoomed-in Circos plot representation of common (gray) and FSHD1 lost (light blue) cis interactions on Chr 4. Genes are indicated for a region extending up to 4 Mb from the VP. Black triangles in Circos plots depict the VP localization. (B) Representative nuclei of 3D multicolor DNA FISH using probes mapping to 4q35.1 region (4q; green), a 4q-D4Z4–positive interacting region (8q24.3, C+; red), and a 4q-D4Z4 not interacting region (3q11.2, C−; magenta) in CN (CN-1, CN-2, CN-3, CN-4) and FSHD1 (FSHD1-1, FSHD1-2, FSHD1-3, FSHD1-4) MBs. Nuclei are counterstained with DAPI (blue). All images at 63× magnification. Scale bar, 5 µm. (C) Cumulative frequency distributions of distances (below 1.5 µm) between 4q and C+ and between 4q and C− in CN (dark and light gray; left) and FSHD1 (dark and light blue; right) MBs. n = 1296 (CN 4q/C+), 1708 (CN 4q/C−), 884 (FSHD1 4q/C+), and 1128 (FSHD1 4q/C−). P-values were calculated by unpaired one-tailed t-test with a confidence interval of 99%. Asterisks represent statistical P-values: for 4q/C+ versus 4q/C− in CN and FSHD1, P < 0.0001. (D) Percentage of nuclei positive for the interactions (under the cutoff of 1.5 µm). n = 427 (CN 4q/C−), 324 (CN 4q/C+), 282 (FSHD1 4q/C−), and 221 (FSHD1 4q/C+). P-values were calculated by Fisher's exact one-sided test with a confidence interval of 99%. Asterisks represent statistical P-values: for 4q/C− versus 4q/C+ in CN, P < 0.0001; for 4q/C− versus 4q/C+ in FSHD1, P = 0.0046.

Figure 2.

Chromatin segmentation analysis revealed chromatin state switches consistent with transcriptional changes in FSHD1 muscle cells. (A) ChromHMM 15-state model obtained with ChIP-seq data sets for H3K36me3, H3K4me1, H3K27ac, H3K4me3, and H3K27me3. Heatmaps display histone marks emission probabilities and transition probabilities between chromatin states. (B) Schematic representation of the strategy used to assign genes as activated or repressed in FSHD1. (C) Expression levels from RNA-seq data sets for FSHD1 activated and repressed genes in MBs (left) and MTs (right) in CN (CN-3, CN-4, and Yao's data sets C20, C21, C22) and FSHD1 (FSHD1-3, FSHD1-4, and Yao's data sets F4, F6) (Yao et al. 2014). Box-and-whisker plots show the median of matched expression values of each gene for CN and FSHD1, and whiskers extend to the 5 to 95 percentiles. P-values were calculated by paired two-tailed Wilcoxon matched-pair signed-rank test with a confidence interval of 99%. Asterisks represent statistical P-values: for CN versus FSHD1 activated in MBs, P < 0.0001; for CN versus FSHD1 repressed in MBs, P < 0.0001; for CN versus FSHD1 repressed in MTs, P < 0.0001.

Figure 3.

Genes that have lost the interaction with 4q-D4Z4 have a more active chromatin state and are enriched for muscle atrophy signature in FSHD muscle cells. (A) Flowchart of filtering steps to identify FSHD1-altered genes. Genes within lost 4q-D4Z4 interactions were filtered as activated (red) or repressed (blue) in FSHD1. (B) Gene Ontology analysis (Biological Processes) of FSHD1 lost-activated and lost-repressed genes. Bars correspond to −log₁₀ of the P-value. (C) Gene set enrichment analysis (GSEA) results of the 319 FSHD1 lost-activated genes performed on expression data from unloading-induced muscle atrophy subjects (Reich et al. 2010). Genes up-regulated in atrophic condition are depicted in red, whereas genes not enriched are depicted in blue. (NES) normalized enrichment score. (D) Expression levels from RNA-seq data sets for atrophic genes (Reich et al. 2010) in CN (CN-3, CN-4, and Yao's data sets C20, C21, C22) and FSHD1 (FSHD1-3, FSHD1-4, and Yao's data sets F4, F6) (Yao et al. 2014) MBs and MTs. Box-and-whisker plots show the median of matched expression values of each gene for CN and FSHD1, and whiskers extend to the 5 to 95 percentiles. P-values were calculated by paired two-tailed t-test with a confidence interval of 99%. Asterisks represent statistical P-values: for CN versus FSHD1 in MBs, P = 0.0099. (E) Expression levels from RNA-seq data sets for atrophic genes (Reich et al. 2010) in CN (CN-3, CN-4, and Yao's data sets C20, C21, C22) and FSHD2 (Yao's data sets F12, F14, F20) (Yao et al. 2014) MBs and MTs. Box-and-whisker plots show the median of matched expression values of each gene for CN and FSHD2, and whiskers extend to the 5 to 95 percentiles. P-values were calculated by paired two-tailed t-test with a confidence interval of 99%. Asterisks represent statistical P-values: for CN versus FSHD2 in MBs, P = 0.0251; for CN versus FSHD2 in MTs, P = 0.0041.

Figure 4.

FBXO32 gene has a deregulated chromatin structure, and it is overexpressed in FSHD1 and FSHD2 muscle cells. (A) Representative nuclei of 3D multicolor DNA FISH using probes mapping to 4q35.1 region (4q; green) and FBXO32 (red) in CN (CN-1, CN-3, CN-4), FSHD1 (FSHD1-1, FSHD1-3, FSHD1-4), and FSHD2 (FSHD2-1, FSHD2-2) MBs. Nuclei are counterstained with DAPI (blue). All images are at 63× magnification. Scale bar, 5 µm. (B) Cumulative frequency distribution of distances (below 1.5 µm) between 4q and FBXO32 in CN (gray), FSHD1 (blue), and FSHD2 (dark blue) MBs. n = 3652 (CN), 2464 (FSHD1), and 1020 (FSHD2). P-values were calculated by unpaired one-tailed t-test with a confidence interval of 99%. Asterisks represent statistical P-values: for CN versus FSHD1, P = 0.0473; for CN versus FSHD2, P = 0.0036. (C) Percentage of nuclei positive for the interactions (under the cutoff of 1.5 µm). n = 913 (CN), 616 (FSHD1), and 255 (FSHD2). (D) 4C normalized coverage tracks at the FBXO32 locus for FBXO32-4C VP in CN (CN-3, CN-4; gray) and FSHD1 (FSHD1-3, FSHD1-4; blue). (E, top) Schematic representation of the FBXO32 locus and HindIII sites. (Middle) Chart showing the frequencies of 3C interaction between FBXO32 promoter and the indicated HindIII restriction sites (sites 4–32), using the same bait of the 4C VP (light gray vertical bar) in CN (gray) and FSHD (blue). n = 3 (CN) and 3 (FSHD1). SEM is indicated. P-values were calculated by two-way ANOVA followed by Bonferroni posttest correction. Asterisks represent statistical P-values: for P19, P25, and P30 CN versus FSHD1, P < 0.001; for P32 CN versus FSHD1, P < 0.01. (Bottom) 4C normalized coverage tracks as well as ChromHMM chromatin states tracks at the FBXO32 locus for FBXO32-4C VP in CN (gray) and FSHD1 (blue). The arrow represents the promoter region; enhancers are highlighted in yellow. (F) Bar plot showing enrichment of RNA Pol II at FBXO32 promoter (left) and an intragenic region (right) assessed by ChIP-qPCR experiment in CN (gray) and FSHD1 (blue) MBs. Results are presented as a percentage of input. n = 2 CN (CN-3, CN-4) and 2 FSHD1 (FSHD1-3, FSHD1-4). SEM is indicated. P-values were calculated by unpaired one-tailed t-test with a confidence interval of 99%. Dots represent the values of each replicate; asterisks represent statistical P-values: for FBXO32 intragenic region CN versus FSHD1, P = 0.0050. (G) Expression levels of FBXO32 gene during CN (gray), FSHD1 (blue), and FSHD2 (dark blue) differentiation. (MT2) Myotubes day 2, (MT4) myotubes day 4, (MT6) myotubes day 6. Data were normalized on GAPDH expression and on MBs. n = 4 CN (CN-1, CN-2, CN-3, CN-4), 4 FSHD1 (FSHD1-1, FSHD1-2, FSHD1-3, FSHD1-4), and 2 FSHD2 (FSHD2-1, FSHD2-2). SEM is indicated. P-values were calculated by two-way ANOVA followed by Bonferroni posttest correction. Dots represent the values of each replicate; asterisks represent statistical P-values: for MT4, CN versus FSHD1, P < 0.0290; CN versus FSHD2, P < 0.0001.

Figure 5.

Ectopic 4q-D4Z4 array restores the expression of FSHD1-lost interacting genes. (A, top) Representation of the BAC containing a 4q upstream region and wild-type D4Z4 (BAC 4q-D4Z4n) (see Methods). (Bottom) Representative nucleus of 3D multicolor DNA FISH using probes for the transfected BAC backbone (red) and 4q35.1 region (4q; green) in MBs transfected with BAC 4q-D4Z4n. Nuclei are counterstained with DAPI (blue). All images are at 63× magnification. Scale bar, 5 µm. (n) Number of nuclei analyzed. (B) Representative nucleus of 3D multicolor DNA FISH using probes for the transfected BAC backbone (red) and 4q35.1 region (4q; green) in MBs transfected with Ctrl BAC (RP11-2A16, representative of an unrelated and not interacting genomic region, Chr 17q21.33). Nuclei are counterstained with DAPI (blue). All images are at 63× magnification. Scale bar, 5 µm. (n) Number of nuclei analyzed. (C) Representative nucleus of 3D multicolor DNA FISH using probes for the transfected BAC backbone (red) and FBXO32 region (FBXO32; light blue) in MBs transfected with BAC 4q-D4Z4n. Nuclei are counterstained with DAPI (blue). All images are at 63× magnification. Scale bar, 5 µm. (n) Number of nuclei analyzed. (D) Bar plots showing expression levels of FBXO32, TRIB3, and ZNF555 (FSHD1 lost-activated genes) in CN (gray) and FSHD1 (blue) MBs transfected with Ctrl BAC and BAC 4q-D4Z4n. Data were normalized on GAPDH expression. n = at least three (with the exception of TRIB3 and ZNF555 CN Ctrl BAC, n = 2). SEM is indicated. P-values were calculated by paired one-tailed t-test with a confidence interval of 99%. Dots represent the values of each replicate; asterisks represent statistical P-values: for FBXO32 Ctrl BAC versus BAC 4q-D4Z4n in CN, P = 0.0182; for FBXO32 Ctrl BAC versus BAC 4q-D4Z4n in FSHD1, P = 0.0073; for TRIB3 Ctrl BAC versus BAC 4q-D4Z4n in FSHD1, P = 0.0281. (E) Bar plot showing expression levels of LZTS3 (FSHD1 lost-repressed gene) in CN (gray) and FSHD1 (blue) MBs transfected with Ctrl BAC and BAC 4q-D4Z4n. Data were normalized on GAPDH expression. n = 3 (with the exception of CN Ctrl BAC, n = 2). SEM is indicated. P-value was calculated by paired one-tailed t-test with a confidence interval of 99%. Dots represent the values of each replicate; asterisks represent statistical P-values: for Ctrl BAC versus BAC 4q-D4Z4n in FSHD1, P = 0.0296. (F) Bar plots showing expression levels of FOXO3 and MYOG (not interacting genes) in CN (gray) and FSHD1 (blue) MBs transfected with Ctrl BAC and BAC 4q-D4Z4n. Data were normalized on GAPDH expression. n = at least three. SEM is indicated. Dots represent the values of each replicate.

Figure 6.

Model of 4q-D4Z4–mediated regulation of atrophic genes transcription. (A) 4q-D4Z4 array is interacting with a subset of atrophic genes, organizing their chromatin structure and keeping on hold their transcription in healthy donor muscle cells. (B) In FSHD1 patients’ muscle cells, the deleted and hypomethylated 4q-D4Z4 array causes an impairment of D4Z4 interactome, leading to a chromatin switch toward an active state (mainly enhancer and promoter regions), which in turn results in the transcriptional up-regulation of the atrophic genes.