Cohesin Mutations Induce Chromatin Conformation Perturbation of the H19/IGF2 Imprinted Region and Gene Expression Dysregulation in Cornelia de Lange Syndrome Cell Lines

Traditionally, Cornelia de Lange Syndrome (CdLS) is considered a cohesinopathy caused by constitutive mutations in cohesin complex genes. Cohesin is a major regulator of chromatin architecture, including the formation of chromatin loops at the imprinted IGF2/H19 domain. We used 3C analysis on lymphoblastoid cells from CdLS patients carrying mutations in NIPBL and SMC1A genes to explore 3D chromatin structure of the IGF2/H19 locus and evaluate the influence of cohesin alterations in chromatin architecture. We also assessed quantitative expression of imprinted loci and WNT pathway genes, together with DMR methylation status of the imprinted genes. A general impairment of chromatin architecture and the emergence of new interactions were found. Moreover, imprinting alterations also involved the expression and methylation levels of imprinted genes, suggesting an association among cohesin genetic defects, chromatin architecture impairment, and imprinting network alteration. The WNT pathway resulted dysregulated: canonical WNT, cell cycle, and WNT signal negative regulation were the most significantly affected subpathways. Among the deregulated pathway nodes, the key node of the frizzled receptors was repressed. Our study provides new evidence that mutations in genes of the cohesin complex have effects on the chromatin architecture and epigenetic stability of genes commonly regulated by high order chromatin structure.


Introduction
Cohesin is a chromatin-associated multi-subunit protein complex that is highly conserved during evolution and involved in several aspects of chromosome biology [1], such as cell division, DNA damage repair, gene transcription and chromosome architecture [2,3].
controlled by the methylation of IC1 [27], also called the H19 differentially methylated region (DMR) or domain (DMD). IC1 acquires methylation on the paternal allele during the male gametogenesis. Depletion of either cohesin or CTCF results in reduced transcription of H19 and increased expression of IGF2, implying a role for these proteins in the expression regulation of these loci [21,28]. The putative looped structure at the IGF2/H19 domain, which brings the promoter and enhancer together in a parental allele-specific manner, is dependent on the differential methylation of IC1, and both cohesin and CTCF bind to several regions within the locus [29,30]. Recently, we provided a detailed characterization of the chromatin architecture of the 11p15.5 imprinted domain [31]; our data extended the available information regarding the structure of the IGF2/H19 domain and defined the interactome of the CDKN1C/KCNQ1OT1 domain (the second imprinted locus mapped at 11p15.5) and the long-range contacts involving the two domains. We confirmed that this domain folds in complex chromatin conformations, which facilitate the control of imprinted genes mediated by distant enhancers, and found deep alterations in the chromatin structure of the entire imprinted domain in lymphoblastoid cell lines (LCLs) from BWS and SRS patients.
Nativio and collaborators [29] reported that cohesin depletion by RNAi results in a modest reduction in looping interactions at the imprinted IGF2/H19 domain, confirming that cohesin is a stabilizing factor in chromatin looping. Although such defects may be very subtle, they hold the potential to cause changes in gene regulation. In light of this evidence, we hypothesized that genetic variations of cohesin genes may also affect the chromatin structure of loci that involve cohesin. With this aim, using chromatin conformation capture (3C) we explored whether chromatin structure and methylation of the IGF2/H19 domain may be impaired in LCLs from CdLS patients carrying different genetic alterations. We also quantified the expression profiles and methylation pattern of a panel of imprinted genes that are often regulated by a high order chromatin structure [32,33]. Expression analysis also included genes belonging to the WNT pathway, reported to be altered in CdLS [34][35][36][37].

Chromatin Conformation Capture Assay (3C)
3C was performed as previously described [31], where the 3C primer details, sequences and the extensive description of the regions analyzed were provided.
Given the importance of CTCF binding at specific sites for the development of intraand interchromosomal contacts [42][43][44], we analyzed four clusters of CTCF-binding sites in the domain: one upstream of IGF2 (CTCF Up), one downstream of H19 (CTCF Down), one in the centrally conserved domain (CCD) region, and one in IC1 (Figure 1a and Rovina et al., 2020 [31]), in addition to genes and enhancers mapped in the 11p15.5 region (Figure 1a).

DNA Methylation Analysis
Total DNA was extracted from LCLs using the QIAamp ® DNA Mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Two independent quantitative methylation experiments of IGF2, H19, GNAS, GNAS-AS1, MEST and PEG10 DMRs were performed by pyrosequencing using the Pyro Mark ID instrument (Qiagen, Hilden, Germany). Raw data were analyzed using the Q-CpG software v1.09 (Biotage Sweden AB). Details on the genomic positions, set-up and protocol were previously described [45,46].

nCounter Analysis
Total RNAs were obtained using the Qiazol reagent (Qiagen, Hilden, Germany), followed by RNA purification by the RNeasy mini kit (Qiagen, Hilden, Germany), according to the manufacturer's protocol. RNAs were eluted in 50 µL of RNase-free water. Concentration and purity were evaluated using the Nanodrop (Thermo Fisher Scientific, Wilmington, DE, USA). Biomolecules 2021, 11, x 5 of 22

DNA Methylation Analysis
Total DNA was extracted from LCLs using the QIAamp ® DNA Mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Two independent quantitative methylation experiments of IGF2, H19, GNAS, GNAS-AS1, MEST and PEG10 DMRs were performed by pyrosequencing using the Pyro Mark ID instrument (Qiagen, Hilden, The data for each cell line were derived from two independent 3C experiments. Interactions between different elements of the IC1 region are shown by red triangles; increasing color intensity corresponds to an increase in the number of interactions of the subregion. Mean association frequencies of CTRL1 and CTRL2 are indicated with black circles. Black triangles indicate the anchors used for 3C analysis. A linear representation of the domain is shown below. In accordance with the transcription of genes in the region, panel b was drawn in reverse orientation with respect to the map in panel a. Adapted from [31] following the Creative Commons Attribution 4.0 International License. Expression analysis was performed by Ncounter using the Nanostring Vantage 3D TM RNA WNT Pathways Panel (Nanostring, Seattle, WA, USA) using a panel including 180 genes associated with the WNT pathways and 12 reference genes for normalization (CC2D1B, COG7, EDC3, GPATCH3, HDAC3, MTMR14, NUBP1, PRPF38A, SAP130, SF3A3, TLK2, ZC3H14), customized with 17 imprinted and imprinted-related genes (Table S1).
The expression profiles were evaluated starting from 150 ng of total RNA for each sample, which integrity was assessed with the TapeStation 2200 (Agilent, Santa Clara, CA, USA); RNA integrity number (RIN) values > 7.0 were considered suitable for the experiments. We used Nanostring technology as it represents a medium-throughput platform to evaluate mRNA abundance profiles providing reproducible and fully automated analyses of the samples. The robustness of this technology was already validated in several papers [47,48]. The reliability of Nanostring technology is based on the ability to quantify the expression of multiple genes without amplification steps. Conversely, technical artifacts could be introduced in qPCR [49].

Statistical Analysis
3C assays: two independent 3C assays were performed for each sample. The frequencies of associations are expressed as mean ± standard deviation. The control value is derived from four independent 3C assays, two using CTRL1 and two CTRL2. The mean of each CdLS cell line is calculated from the results of two independent 3C assays. Differences in association frequencies between controls and patient's LCLs were evaluated using the two-way ANOVA test followed by the Bonferroni post-test in the GraphPad Prism program. Statistical significance is indicated as **** p ≤ 0.0001; *** p ≤ 0.001; ** p ≤ 0.01; * p ≤ 0.05.
NCounter analysis: Nanostring data were analyzed by the nSolver Advanced Analysis Software 4.0 (NanoString, Seattle, WA, USA) considering a background threshold of 20 counts and excluding from the analysis all genes with counts above the threshold. Quality assessment was performed for each sample, and two quality control parameters common to all nCounter assays were considered: the Imaging QC that measures the percentage of the requested fields of view successfully scanned in each cartridge lane and the Binding Density QC that measures the reporter probe density on the cartridge surface in each sample lane. The Benjamini-Hochberg method was applied to reduce the false discovery rate (FDR), minimizing Type I errors (false positives). Unadjusted p-value ≤ 0.05 were considered significant.

Results
A schematic overview of the experimental design is summarized in Table 1.

Chromatin Interactions at the IGF2/H19 Domain in Cells from CdLS Patients
Nativio and coworkers [29] showed a reduction in the looping interactions at the IGF2/H19 imprinted domain after cohesin depletion by RNAi, suggesting a causative relationship between cohesin anomalies and deregulation of imprinted genes. We investigated whether genetic alterations (point mutations and balanced reciprocal translocations) in SMC1A and NIPBL affect the chromatin structure of the IGF2/H19 locus. The detailed landscape of the IGF2/H19 region analyzed by 3C and the 3C coverage is depicted in Figure 1a.
To study the physical contacts in the region, we used four anchors: CTCF Up, IC1, Enh A and CTCFDw ( Figure 1). Using this approach, we recently characterized the 3D chromatin structure of the region in BWS and SRS patients and in normal cells [31]. The interactome of the control LCLs included in this study is schematically shown in Figure 1b.
We performed 3C analysis in four CdLS LCLs with genetic defects in SMC1A (CdLS1 and 2) and NIPBL (CdLS6 and 9) genes (Table 1) to identify modifications of the chromatin interactome in the IC1 domain. Two independent experiments were carried out for each cell line. The specific contacts among CTCF-binding sites, regulatory elements and genes mapped in the domain responsible for the repositioning of the regional enhancers near   indicated above. Each point in the profile is the mean ± standard deviation of two independent 3C experiments and indicates the association frequency between the anchor and the fragment on the left of the corresponding BglII restriction site. Differences (two-way ANOVA test) are indicated by asterisks. Red dots indicate the points with standard deviation > 0.1.  frequencies as described in the figure. Black triangles indicate the anchors used for 3C analysis. A linear representation of the IC1 imprinted domain is depicted below the interactome. (b) IC1 locus looping profiles for the indicated anchors in controls (dotted black) and CdLS2 (light blue) cell lines. BglII restriction sites are indicated above. Each point in the profile is the mean ± standard deviation of two independent 3C experiments and indicates the association frequency between the anchor and the fragment on the left of the corresponding BglII restriction site. Differences (two-way ANOVA test) are indicated by asterisks. Red dots indicate the points with standard deviation > 0.1. We found an aberrant chromatin structure of the IGF2/H19 region in CdLS cells compared to controls (Figures 2, 3, 4 and Figure S1). In particular, interactions of the IGF2 promoter and the CTCF Up were partially conserved, while those involving the enhancer A and the CTCF Dw were perturbed, especially in the CdLS1 and CdLS6 LCLs (Figures 2  and 3). In addition, new interactions between the CCD and the H19 region (IC1, Enh A and CTCF Dw), and between 3 IGF2 and the IC1/CTCF Up, were observed in all the CdLS LCLs (Figures 2, 3, 4 and Figure S1).
In summary, 3C results show a general impairment of the chromatin architecture in CdLS cells regardless of the specific cohesin gene involved. The emergence of many new interactions could be a further expression of the malfunction of the cohesin complex.

Expression Profile of the CdLS Cell Lines
Using a Nanostring approach, we analyzed the expression levels of a set of imprinted genes, related-imprinted genes ( Table 2) and loci of the WNT pathway (Table 3) in nine CdLS LCLs with different mutations in SMC1A (CdLS 1-5) or NIPBL (CdLS 6-9) genes, and in four control LCLs (CTRL 1-4) ( Table 1).

Imprinted and Imprinted-Related Genes Panel
Given that genetic defects in cohesin genes were found to be associated with a general impairment of the 3D chromatin structure of the IGF2/H19 imprinted domain, we investigated the expression profiles in CdLS LCLs in a panel of imprinted genes that are often regulated by a high order of chromatin structure [32,33]. We evaluated the expression levels of 13 imprinted genes and four imprinted-related genes ( Table S2a). The analysis of the imprinted genes showed that 7 out 13 were expressed in LCLs (Table 2), and no expression in LCLs was retrieved for IGF2 and H19.
Three out of seven imprinted genes were differentially expressed (DEGs) in patients' cell lines. In particular, GNAS was upregulated, whereas MEST and PEG10 were downregulated. Furthermore, GNAS-AS1 was downregulated despite being close to the threshold of significance (unadjusted p-value = 0.0058) (Table 2 and Figure 5).
These results indicate that the imprinting network was altered, suggesting an association among cohesin defects, 3D chromatin architecture impairment and imprinting network alteration.

WNT Panel
Diverse WNT pathways act as master regulators of central nervous system development [35], which is disrupted in CdLS animal models and patient-derived cells lines [35][36][37]50]. Therefore, we next analyzed a panel of 180 genes involved in WNT signaling (Table S2b), of which 163 were expressed in LCLs.  Thirty-three differentially expressed (DE) genes were observed in patients' LCLs (unadjusted p-value < 0.05) compared to controls. Among them, 16 were upregulated and 17 downregulated ( Figure 5 and Table 3). Pathway enrichment analysis by the nSolver software indicated that the most significant differences observed in CdLSs cells are in the following subpathways: canonical WNT pathway, cell cycle and WNT signal negative regulation ( Figure 6). Among the investigated genes, an altered expression was mainly observed in members of the WNT receptors of the Frizzled gene (FZD) family (FZD5, FZD8 and FZD10 downregulated and FDZ3 upregulated). igure 5. Volcano plot of DEGs CdLS compared to control cell lines. Upregulated genes are highlighted by red dots, while ownregulated genes by blue dots. In bold and underlined are reported the DE imprinted genes. FDR < 0.1 (Benjaminiochberg adjusted p value) and unadjusted p value < 0.05 are indicated by horizontal lines. The VolcaNoser tool was used r creating volcano plots [51]. Our results confirm alterations of the WNT pathways in LCLs derived from CdLS patients; in particular, we observed the most significant differences in the canonical WNT pathway, cell cycle and WNT signal negative regulation. Figure 7 shows a schematic representation of the expression alterations observed in the three main WNT signaling pathways. The nodes of the pathways that were found to be dysregulated by Pathview (nSolver Advanced Analysis Software 4.0; Figure 7 Finally, we performed three different expression analyses splitting CdLSs samples based on NIPBL and SMC1A involvement, comparing NIPBL-mutated CdLSs vs. CTRLs ( Figure S2a), SMC1A-mutated CdLSs vs. CTRLs ( Figure S2b) and NIPBLvs. SMC1Amutated CdLSs ( Figure S2c). Differently to the main cohort (CdLSs vs. CTRLs, Figures 5 and 6), in none of the three subsets, we found differentially expressed genes with Benjamini-Hochberg adjusted p-values < 0.1. Some genes reached a significant unadjusted p-value < 0.005. Interestingly, considering the NIPBL-mutated samples, 17 DEGs reached the significant threshold of 0.005, among them nine are shared with the most significant DEGs of the main cohort: SFRP1, TCF4, GNAS, MEST, MAPK10, CDKN2A, PLCB1, RUNX2 and MMP7 ( Figure S2a). Similarly, the SMC1A-mutated group showed nine out of the 25 significant DEGs in common with the main cohort: RUNX2, BMP4, MAPK10, GNAS, CDKN2A, TGFB3, TCF4, PEG10 and PLGB1 ( Figure S2b). When we consider the NIPBLvs. SMC1A-mutated CdLSs subset, the number of shared significant DEGs (p-value < 0.005) drops to 2: RUNX2 and MAPK10 ( Figure S2c). These findings are in line with the results of Boudaoud et al. that compared the gene expression profiles of LCLs from patients carrying mutations in NIPBL and SMC1A and showed reduced number of differentially expressed genes shared between the two gene groups (n = 126) compared to the totality of genes misregulated in NIPBL-mutated (n = 1431) and SMC1A-mutated (n = 1186) samples [52]. pathway, cell cycle and WNT signal negative regulation. Figure 7 shows a schematic representation of the expression alterations observed in the three main WNT signaling pathways. The nodes of the pathways that were found to be dysregulated by Pathview (nSolver Advanced Analysis Software 4.0; Figure 7

Methylation Analysis of the Imprinted DMRs
We analyzed the methylation status of the IGF2/H19, MEST, PEG10, GNAS and GNAS-AS1 DMRs by pyrosequencing to evaluate possible correlations between the observed chromatin structure and expression defects and the methylation levels of the regulatory regions of these imprinted genes.
In CdLS LCLs we observed a general instability in the methylation status of the IC1 and MEST-DMR, a trend to hypermethylation of the PEG10 and GNAS-AS1 DMRs and a trend to hypomethylation of the GNAS-DMR compared to CTRLs (Figure 8 and Table S3): IC1 methylation range was 9-45% in CdLSs and 34-41% in controls; MEST-DMR methylation range was 33-60% in CdLSs and 42-47% in controls; PEG10-DMR methylation range was 41-44% in CdLSs and 35-43% in controls; GNAS-DMR methylation range was 4-42% in CdLSs and 31-37% in controls; GNAS-AS1-DMR methylation range was 8-37% in CdLSs and 10-22% in controls. In CdLS cells, hypermethylation of PEG10, MEST and GNAS-AS1 DMRs was in line with the observed low expression, in the same way GNAS-DMR hypomethylation was in line with its overexpression; the alterations in IC1 methylation may be the result of the perturbation in chromatin architecture of the IGF2/H19 imprinted domain. Regarding the specific gene mutated in the CdLS LCLs a more pronounced variation in methylation status is apparent in SMC1A involvement, in particular for MEST, PEG10 and GNAS-AS1 DMRs.

Discussion
Cohesin has complex functions in chromosome biology, including gene expression regulation and maintenance of chromatin architecture [53]. Defects in cohesin function are associated with a group of diseases known as cohesinopathies, notably CdLS.
Starting from the evidence that cohesin depletion causes a reduction in the looping interactions in the IGF2/H19 imprinted domain [29], we studied the 3D chromatin structure of this locus in LCLs from CdLS patients with mutations in the SMC1A or NIPBL genes, and evaluated whether constitutive genetic mutations in the cohesin subunit genes can alter chromatin architecture and, consequently, gene expression.
We found a broad perturbation in the chromatin structure of the domain regardless of the CdLS causative gene, and observed a change in the interactions among CTCF-binding sites, regulatory elements and genes of the region. This scenario could be related to the existence in each cell line of a diffuse instability rather than recurrent alterations of specific interactions. It is also conceivable that compromised maintenance of chromatin architecture could lead to heterogeneous defects among cells in the same cell line. The alterations can be due to cohesin malfunction rather than lack of function. The improper sliding of the cohesin complex along the loop could, indeed, cause the observed perturbed chromatin interactions, including new associations.
In addition, we found that the 3D chromatin alterations of the locus were associated with methylation defects of IC1 in some CdLS cell lines. Methylation of this DMR controls the expression of the IGF2 and H19 imprinted genes. In our recent study we reported that, in BWS and SRS imprinting-related disorders, the causative methylation defects at IC1 are associated with alterations in the 3D chromatin architecture of the domain [31]. Differently, in CdLS it is conceivable that the observed methylation instability is related to chromatin structure alterations caused by a malfunction of the cohesin complex.
Imprinting disorders are frequently associated with growth and development abnormalities [54,55]; similarly, prenatal and postnatal growth restriction are observed in CdLS patients [56]. This feature might be also related to an imprinting network perturbation suggested by our data. In particular, we found methylation instability of the DMRs and/or defective expression of loci involved in fetal growth such as PEG10, MEST, GNAS, IGF2 and H19 [55][56][57].
DMRs methylation instability at multiple imprinted loci has already been identified in patients with multilocus imprinting disturbances (MLID), thus demonstrating a fine-tuned network involving regulatory regions of imprinting [58]. Similarly, in CdLS we surmise a broad impairment of the maintenance of the regulatory mechanisms of genomic imprinting, resulting in a general instability of the imprinting marks, which might be related, in this disease, to cohesin defects. These findings strengthen the key role of this complex in transcription regulation.
Based on the pivotal role of cohesin in the dynamics of the transcriptional regulation network, and on the hypothesis that the multiorgan defects in CdLS patients are due to global disruption of the transcriptional regulation network of developmental pathways, including the canonical WNT pathway [59][60][61], we investigated the expression profile of a panel of 180 genes of the WNT pathways. In vertebrates, the WNT signaling pathway regulates crucial aspects of the embryonic development and maintenance of adult tissue homeostasis by regulating cell proliferation, differentiation, migration, genetic stability, and apoptosis, as well as maintenance of adult stem cells in a pluripotent state [62]. These processes are obtained through two principal branches: the canonical pathway that regulates the expression of the key developmental target genes through the frizzled family receptors and, the intracellular transducer Dishevelled, and the noncanonical WNT pathway (β-catenin-independent pathway) that regulates cell polarity and dorsal mesodermal cell movements during development [37,63].
RUNX2 gene was the most significantly upregulated; it is a member of the RUNX family of transcription factors, nuclear proteins with a DNA-binding domain. RUNX2 is essential for osteoblastic differentiation and skeletal morphogenesis and acts as a scaffold for other regulatory factors involved in tissue-specific expression of the skeletal. Mutations in this gene have been associated with cleidocranial dysplasia [64,65]. The most strongly downregulated gene was MAPK10, a gene involved in neuronal proliferation, differentiation and survival. Interestingly, RUNX2 and MAPK10 are the only two significant DEGs (unadjusted p-value 0.0462 and 0.0221, respectively) when we consider the NIPBLvs. SMC1Amutated CdLSs subset.
Our results confirm alterations of the WNT pathways in LCLs derived from CdLS patients; in particular, we observed the most significant differences in the canonical WNT pathway, cell cycle and WNT signal negative regulation. In addition, we found that several genes belonging to the pathway nodes were deregulated; in particular the key node of the frizzled receptors was repressed, specifically FDZ3, FZD5, FZD8 and FZD10 genes.
Because cohesin represents the primary regulator of 3D genome organization and, consequently, of the gene expression in all cell types, we used LCLs as a model for CdLS; however, we are aware that different expression profiles may be present in cells of other embryological origin.
Our study provides new evidence that the mutations in genes of the cohesin complex may affect chromatin architecture and the epigenetic stability of genes commonly regulated by high order chromatin structures, such as the imprinted loci. Altered imprinted gene expression could, at least in part, explain the growth defects present in CdLS.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/biom11111622/s1, Figure S1. Abnormalities in chromatin architecture at the IC1 locus of the CdLS9 cell line (NIPBL mutation). Figure S2. Three different expression profiles of CdLS samples split by NIPBL and SMC1A variations evaluated by nCounter Nanostring technology. A) NIPBL-mutated CdLSs versus CTRLs. B) SMC1A-mutated CdLSs versus CTRLs. C) NIPBL-mutated versus SMC1A -mutated CdLSs. Volcano plots (Top) and panels (Bottom) of significant DEGs. Upregulated and downregulated genes shared with the main cohort (CdLSs compared to CTRLs) are highlighted by red and blue dots, respectively. Unadjusted p-value <0.05 is indicated by horizontal line. Table S1. Customized panel of imprinted and imprinted related genes analyzed by NCounter Nanostring approach. Table S2a. Panel of imprinted genes and imprinted related genes analyzed by NCounter Nanostring approach. Table S2b. Panel of genes of the WNT pathway analyzed by NCounter Nanostring approach. Table S3