Context-specific functions of chromatin remodellers in development and disease

Abstract

Chromatin remodellers were once thought to be highly redundant and nonspecific in their actions. However, recent human genetic studies demonstrate remarkable biological specificity and dosage sensitivity of the thirty-two adenosine triphosphate (ATP)-dependent chromatin remodellers encoded in the human genome. Mutations in remodellers produce many human developmental disorders and cancers, motivating efforts to investigate their distinct functions in biologically relevant settings. Exquisitely specific biological functions seem to be an emergent property in mammals, and in many cases are based on the combinatorial assembly of subunits and the generation of stable, composite surfaces. Critical interactions between remodelling complex subunits, the nucleosome and other transcriptional regulators are now being defined from structural and biochemical studies. In addition, in vivo analyses of remodellers at relevant genetic loci have provided minute-by-minute insights into their dynamics. These studies are proposing new models for the determinants of remodeller localization and function on chromatin.

Fig. 1 |. The family of human chromatin remodellers.

a, The 32 human chromatin remodellers cluster into families based on sequence similarity of their ATPase domains. The tree was constructed from a multiple sequence alignment of only the ATPase domains of human chromatin remodellers. b, The four canonical sub-families of chromatin remodellers are shown with their defining Snf2-family helicase-like ATPase domain highlighted along with distinguishing auxiliary domains often present in sub-family members. c, The structure of remodellers from each of the major families: the BAF or mSWI/SNF complex (Protein DataBase (PDB): PDBDEV_00000056), SMARCA5 or SNF2h (ISWI complexes) (PDB: 6ne3), CHD4 (PDB: 6ryr) and INO80 (PDB: 6hts). Each displays an example of how the ATPase domain, in red, of remodellers contacts the nucleosome, assisted by non-ATPase domains such as the SnAC (Snf2 ATP coupling), and/or by auxiliary subunits. An example of other subunits in the BAF complex is labelled. Other subunits in grey are not labelled for simplicity; see Fig. 2 and Supplementary Table 1 for a detailed list of subunits. DExDc, Asp, Glu, X, Asp motif and DEAD-like helicases superfamily; HAND, secondary structure of four α-helices, three of which are in an L-shape configuration; HSA, helicase/SANT-associated; PHD, plant homeodomain; QLQ, Gln, Leu, Gln motif; SANT, switching-defective protein 3 (Swi3), adaptor 2 (Ada2), nuclear receptor co-repressor (N-CoR), transcription factor (TF)IIIB; SLIDE, SANT-like ISWI domain.

Fig. 2 |. Chromatin remodelling complexes in human development and disease.

a–j, The human chromatin remodelling complexes are shown with their composition of subunits. Subunits, where possible, reflect the actual position and relative size in the available structures of remodeller complexes. Paralogous subunits that can be substituted for one another are displayed as A/B/C, and subunits are coloured according to their probability of intolerance to loss-of-function (pLI) scores for their encoding genes in the human genome^,. Developmental disorders associated by protein-truncating variants and predicted deleterious missense mutations found in the genes encoding remodeller subunits are labelled, compiled from large-scale sequencing studies of de novo mutations in individuals with autism spectrum disorder (ASD) and developmental delay and/or (idiopathic) intellectual disability (DDID)^,–, congenital heart disease (CHD), as well as manual curation of variants in the literature from case studies (Supplementary Table 2). *,**,***: false discovery rate (FDR) < (0.05, 0.01, 0.001) of association with ASD from the Autism Sequencing Consortium; −/−, homozygous mutation; ATR-X, X-linked alpha-thalassaemia/mental retardation; AVSD, atrioventricular septal defect; CHARGE, coloboma, heart defect, atresia choanae, growth retardation, genital abnormality, and ear abnormality; COFS, cerebro-oculo-facio-skeletal; DOORS, deafness, onychodystrophy, osteodystrophy, mental retardation, seizures; GAND, GATAD2B-associated neurodevelopmental disorder; ICF, immunodeficiency, centromeric instability facial anomalies spectrum; IHH, idiopathic hypogonadotropic hypogonadism; NEDDFL, neurodevelopmental disorder with dysmorphic facies and distal limb anomalies; NEDFASB, neurodevelopmental disorder with dysmorphic facies, sleep disturbance and brain abnormalities.

Fig. 3 |. The dosage sensitivity of human remodellers.

a, Remodeller genes are among the most sensitive to loss in human individuals. The constraint against missense variation and the intolerance to loss-of-function are plotted for all human genes. Genes encoding remodeller ATPase subunits are labelled and coloured by complex if applicable. Other complex subunits are coloured but left unlabelled for simplicity. All data is from gnomAD^,. b, Remodelling complex genes are enriched among all genes conserved in copy number across mammals and enriched in known pathogenic copy number variants (CNVs), suggesting strong selective pressure on their dosage. The tree in part b is adapted from ref. , CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

Fig. 4 |. Models of remodeller–TF interactions.

Three models of remodeller–transcription factor (TF) interactions are shown. a, A pioneer TF that can bind a motif on nucleosomal DNA recruits remodellers to a nucleosome via a biochemical interaction. b, Remodeller activity and genome-wide localization, constrained modestly by histone post-translational modifications (PTM) and/or DNA sequences that remodeller domains or complex subunits can recognize, creates accessibility for TF binding. c, Remodeller and TF activity cooperate based on their respective on- and off-rates k to nucleosomal DNA.

Fig. 5 |. Enrichment of non-synonymous cancer mutations in chromatin remodelling complexes.

Non-synonymous mutations include missense mutations, nonsense mutations, fusions, frameshifts, in-frame deletions and splice site mutations. The enrichment of mutations in a gene observed above those expected is adjusted for gene length and the calculated background mutation rate in the cancer. P values were computed as in ref. by, for a given cancer, comparing the observed number of mutations k with the cancer’s background mutation rate r, adjusted for gene length, with the assumption that k could be approximated by a Poisson distribution. The background rate was calculated from the total number of mutations per gene length for all genes in that cancer. All data was from The Cancer Genome Atlas Project (TCGA Research Network) plus refs. , accessed from the cBioPortal^, using the R package ‘cgdsr’. TP53, an important tumour suppressor, and HBB, rarely mutated in cancer, are plotted as comparisons. Adenocac., adenocarcinoma; DLBCL, diffuse large B cell lymphoma; endocer., endocervical; SCC, squamous cell carcinoma.