Functional genomics J Med Genet: first published as 10.1136/jmedgenet-2020-107095 on 11 September 2020. Downloaded from ORIGINAL RESEARCH Low tolerance for transcriptional variation at cohesin genes is accompanied by functional links to disease- relevant pathways William Schierding ,1 Julia A Horsfield,2,3 Justin M O’Sullivan 1,3,4 ► Additional material is ABSTRACT of functional cohesin without eliminating it alto- published online only. To view Background: The cohesin complex plays an essential gether. Complete loss of cohesin is not tolerated please visit the journal online in healthy individuals.2 Thus, cohesin is haplo- (http:// dx. doi. org/ 10. 1136/ role in genome organisation and cell division. A full jmedgenet- 2020- 107095). complement of the cohesin complex and its regulators is insufficient such that normal tissue development important for normal development, since heterozygous and homeostasis requires that the concentrations 1Liggins Institute, The University mutations in genes encoding these components can of cohesin and its regulatory factors remain tightly of Auckland, Auckland, New be sufficient to produce a disease phenotype. The regulated. Zealand 2Department of Pathology, implication that genes encoding the cohesin subunits The human mitotic cohesin ring contains four Dunedin School of Medicine, or cohesin regulators must be tightly controlled and integral subunits: two structural maintenance University of Otago, Dunedin, resistant to variability in expression has not yet been proteins (SMC1A, SMC3), one stromalin HEAT- New Zealand 3 formally tested. repeat domain subunit (STAG1 or STAG2) and one Maurice Wilkins Centre 6 for Molecular Biodiscovery, Methods: Here, we identify spatial- regulatory kleisin subunit (RAD21). Genes encoding cohesin The University of Auckland, connections with potential to regulate expression of subunits are mutated in a wide range of cancers. Auckland, New Zealand cohesin loci (Mitotic: SMC1A, SMC3, STAG1, STAG2, Cohesin mutations are especially prevalent in acute 4 MRC Lifecourse Epidemiology RAD21/RAD21- AS; Meiotic: SMC1B, STAG3, REC8, myeloid leukaemia.7–9 Notably, STAG2 is 1 of just Unit, University of Southampton, RAD21L1), cohesin- ring support genes (NIPBL, MAU2, 12 genes mutated in four or more cancers,5 partic- Southampton, UK WAPL, PDS5A, PDS5B) and CTCF, including linking their ularly in bladder cancer, Ewing sarcoma, glioblas- 10 Correspondence to expression to that of other genes. We searched the toma and myeloid malignancies. Dr William Schierding, The genome- wide association studies (GWAS) catalogue for In meiotic cohesin, SMC1A is replaced by University of Auckland Liggins SNPs mapped or attributed to cohesin genes by GWAS SMC1B; STAG1/2 by STAG3 and RAD21 by Institute, Auckland 1142, New (GWAS- attributed) and the GTEx catalogue for SNPs REC8 or RAD21L1.1 Mutations in meiotic cohesin Zealand; mapped to cohesin genes by cis- regulatory variants subunits are associated with infertility in men,11 w. schierding@ auckland. ac. nz and Dr Julia A Horsfield, in one or more of 44 tissues across the human body chromosome segregation errors and primary 12 Department of Pathology, (expression quantitative trail locus- attributed). ovarian insufficiency in women. Dunedin School of Medicine, Results: Connections that centre on the cohesin ring Cohesin is loaded onto DNA by the SCC2/SCC4 University of Otago, Dunedin subunits provide evidence of coordinated regulation complex (encoded by the NIPBL and MAU2 genes, 9016, New Zealand; 1 http://jmg.bmj.com/ julia. horsfield@ otago. ac. nz that has little tolerance for perturbation. We used the respectively). Mutations in the cohesin loading CoDeS3D SNP- gene attribution methodology to identify factor NIPBL are associated with >65% cases of Received 14 April 2020 transcriptional changes across a set of genes coregulated Cornelia de Lange syndrome (CdLS). Remarkably, Revised 8 June 2020 with the cohesin loci that include biological pathways features associated with CdLS are observed with Accepted 20 June 2020 such as extracellular matrix production and proteasome- less than 30% depletion in NIPBL protein levels.13 mediated protein degradation. Remarkably, many of The release of cohesin from DNA is achieved by the genes that are coregulated with cohesin loci are WAPL, which opens up the interface connecting the themselves intolerant to loss- of- function. SMC3 and RAD21 subunits. The PDS5A/PDS5B on October 6, 2021 by guest. Protected copyright. Conclusions: The results highlight the importance cohesin associated subunits affect this process by of robust regulation of cohesin genes and implicate contacting cohesin to either maintain (with STAG1 novel pathways that may be important in the human and STAG2) or remove (with WAPL) the ring from cohesinopathy disorders. DNA.1 Spatial organisation and compaction of chro- mosomes in the nucleus involves non- random folding of DNA on different scales. The genome © Author(s) (or their INTRODUCTION is segregated into active A compartments and employer(s)) 2020. Re- use The cohesin complex has multiple essential roles inactive B compartments,14 inside which further permitted under CC BY- NC. No during cell division in mitosis and meiosis, genome organisation occurs into topologically associating commercial re- use. See rights and permissions. Published organisation, DNA damage repair and gene expres- domains (TADs) interspersed with genomic regions 1 14 by BMJ. sion. Mutations in genes that encode members of with fewer interactions. Cohesin participates in the cohesin complex, or its regulators, cause devel- genome organisation by mediating ‘loop extrusion’ To cite: Schierding W, opmental diseases known as the ‘cohesinopathies’ of DNA to form loops that anchor TAD boundaries. Horsfield JA, O’Sullivan JM. 2 J Med Genet Epub ahead of when present in the germline or contribute to the At TAD boundaries, cohesin colocalises with the 3–5 print: [please include Day development of cancer in somatic cells. CCCTC binding factor (CTCF) to form chromatin Month Year]. doi:10.1136/ Remarkably, cohesin mutations are almost always loops between convergent CTCF binding sites. jmedgenet-2020-107095 heterozygous and result in depletion of the amount Fine- scale genomic interactions include chromatin Schierding W, et al. J Med Genet 2020;0:1–9. doi:10.1136/jmedgenet-2020-107095 1 Functional genomics J Med Genet: first published as 10.1136/jmedgenet-2020-107095 on 11 September 2020. Downloaded from loops that mediate promoter- enhancer contacts. Notably, the and regulatory hubs). We also identify all variants within each time- specific and tissue- specific formation of the fine- scale loops gene locus that had a previously determined cis- eQTL (GTEx also requires cohesin.15–17 catalogue) to the cohesin gene (eQTL-attributed variant list). As The spatial organisation of the genome is particularly dynamic with the GWAS variants, we tested these variants for the pres- and susceptible to disruption during development. For example, ence of spatial-regulatory relationships involving genes outside changes to TAD boundaries are associated with developmental of the locus. Only a few of these eQTL-attributed variants are disorders.18 Furthermore, disruption of TAD boundaries by currently implicated in disease pathways, but their regulatory cohesin knockdown can lead to ectopic enhancer- promoter relationships with cohesin suggest that they may be significant interactions that result in changes in gene expression.19 Rewiring for cohesinopathy disorders. of the patterns of course and fine- scale chromatin interactions also contributes to cancer development,20–22 including the gener- ation of oncogenic chromosomal translocations.23 24 RESULTS Disease- associated genome- wide association study (GWAS) 140 genetic variants with regulatory potential are associated variants in non- coding DNA likely act through spatially organ- with cohesin loci ised hubs of regulatory control elements, each component of Mitotic cohesin genes (SMC1A, SMC3, STAG1, STAG2 and which contributes a small amount to the observed phenotype(s), RAD21), meiotic cohesin genes (SMC1B, STAG3, REC8 and as predicted by the omnigenic hypothesis.25 26 Non- coding muta- RAD21L1), cohesin support genes (WAPL, NIPBL, PDS5A, tions at cohesin and cohesin- associated factors were found by PDS5B and MAU2) and CTCF were investigated to determine GWAS (GWAS- attributed variants) to track with multiple pheno- if they contain non- coding genetic variants (SNPs) that make types.27 However, the impact of genetic variants located within contact in 3D with genes and therefore could directly affect gene cohesin and its associated genes has not yet been investigated expression (GWAS- attributed and eQTL-attributed; table 1, with respect to phenotype development. online supplementary table S1). We hypothesised that cohesin- associated pathologies can be A total of 106 GWAS-attributed genetic variants associated affected by subtle, combinatorial changes in the regulation of with disease were identified (methods) that mapped to a cohesin cohesin genes caused by common genetic variants within control gene (49 GWAS; 50 SNPs) or cohesin- associated (49 GWAS; 56 elements. Here, we link the 3D structure of the genome with SNPs) gene (online supplementary table S1). Twelve SNPs (blue, expression quantitative trail locus (eQTL) data to determine if online supplementary table S1) are not listed in the current GWAS variants attributed to cohesin genes affect their transcrip- GTEx SNP dictionary