Clinical, in vitro, and in vivo evidence of WAPL as a novel cohesinopathy gene and phenotypic driver of 10q22.3q23.2 genomic disorder

Abstract

Cohesin is a fundamental genome-organizing complex that orchestrates three-dimensional chromosome folding and gene expression via DNA loop extrusion. Alterations to genes encoding cohesin subunits and cohesin loaders cause Mendelian disorders, including Cornelia de Lange syndrome (CdLS). By contrast, disruption of factors that remove cohesin from DNA, including WAPL and its binding partners PDS5A and PDS5B, have not yet been associated with human disease. Here, we explored the relevance of these cohesin release factors in Mendelian disease by establishing a rare disease cohort of deeply phenotyped individuals with heterozygous, predicted damaging variants in WAPL (n=27), PDS5A (n=8), and PDS5B (n=8), by modeling WAPL deficiency in human cell lines and mice, and by aggregating rare disease association statistics from consortia studies. We identified a WAPL-related disorder characterized by developmental delay, intellectual disability, and risk of other developmental anomalies including clubfoot. Similarities between individuals with damaging WAPL variants and those with large, recurrent 10q22.3q23.2 (10q) deletions (which encompass WAPL) nominate WAPL as a driver gene within this genomic disorder region. While carriers of PDS5A or PDS5B variants exhibited features of developmental disorders, neither cohort-based statistics nor case phenotyping associated these genes with specific phenotypes. We used CRISPR engineering to generate truncating variants in WAPL, as well the 7.8 Mb 10q deletion or duplication in human iPSCs and induced neurons. Transcriptomic analyses identified differentially expressed genes in both models, with highly significant overlap between WAPL haploinsufficiency and 10q deletion signatures. Mice with 50% residual Wapl expression exhibited mild deficits of growth and learning/memory, whereas those with 25% residual Wapl expression displayed birth defects and postnatal lethality, revealing a dosage liability threshold below the level of heterozygosity. In summary, we delineated a novel genetic condition caused by cohesin release factor deficiency, nominated WAPL as a driver gene within a genomic disorder region, and further illuminated dosage sensitivity of human cohesin.

Keywords: PDS5A; PDS5B; WAPL; cohesin release factor; transcription.

Figure 1.. Disease-causing variants in WAPL, a cohesin release factor, were predicted then discovered.

a. Simplified depiction of cohesin balance. Cohesin-mediated genome organization is initiated by loading of the cohesin ring (dark grey) by NIPBL and MAU2 onto chromatin (dark blue), followed by loop extrusion. WAPL resets this process by removing cohesin from chromatin. PDS5 facilitates both cohesin release by WAPL and cohesin stability by sororin, among other functions (Zhang et al., 2021). b-c. Probability of haploinsufficiency (pHaplo, b) of 18,641 autosomal genes, via (Collins et al., 2022), and loss of function observed/expected upper bound fraction (LOEUF, c) of 18,567 genes, via gnomAD (Karczewski et al., 2020). Red lines, average of known autosomal dominant CdLS genes NIPBL, RAD21, and SMC3. Blue lines, average of cohesin release factor genes WAPL, PDS5A, and PDS5B. See also Table S1. d. 27 WAPL variants identified in subjects including predicted damaging missense (ms, top) and loss-of-function (pLoF, bottom), plotted in protein space in genomic orientation. Domains of the canonical protein (1190 amino acids) via UniProt (https://www.uniprot.org) and Ensembl (https://ensembl.org). gnomAD pLoF variants passing quality filters are denoted by black (frameshift, stop gain) and grey (splice) circles. The diamond denotes the point after which escape from nonsense-mediated decay would be predicted (codon 1151). The triangle is the start position for minor 3′ isoforms in (h). The cut site for CRISPR/Cas9 experiments is in gold text. The p.Val863AlafsTer2 variant was seen in two individuals. Missense and pLoF variants both cluster in the C-terminal half of WAPL (p=0.0020 and p=0.0024, respectively by two-tailed binomial test). e. Location of WAPL within the 7.8 Mb 10q22.3q23.2 reciprocal genomic disorder region (black bar; via ClinGen Pathogenic CNV track of the UCSC Genome Browser (https://genome.ucsc.edu/)). f. AlphaFold per-residue structural confidence (pLDDT, or predicted local distance difference test), via (Kwon et al., 2024). Y-axis range is 0–100, with scores <50 (orange) being a reasonably strong predictor of disorder (https://alphafold.ebi.ac.uk/). g. Deep protein language model-derived predicted pathogenicity of every possible missense change in WAPL, via (Brandes et al., 2023). Y-axis, amino acid. X-axis, residue. Higher LLR (log-likelihood ratio) indicates more likely deleteriousness h. Regional missense constraint in amino acid space, via gnomAD v2.1.1 (https://gnomad.broadinstitute.org/). Scores indicate the fraction of expected amino acid variation in a population cohort. i. WAPL transcript expression by tissue, via gnomAD. C-terminal isoforms exist but are more lowly expressed. j-k. 3D structures of WAPL via AlphaFold (j) (https://www.alphafold.ebi.ac.uk/entry/Q7Z5K2) that demonstrates low-confidence structure for its N-terminal half, and via crystallography of its C-terminal half (k) (Ouyang et al., 2013) (pdb ID 4k6j, amino acids 631–1190). Case missense variants (red) show possible clustering. Plots created via (https://g2p.broadinstitute.org; (Kwon et al., 2024)).

Figure 2.. Cohesin release factor variants are associated with neurodevelopmental and other features.

a. Phenotypes (rows) in subjects (columns) with heterozygous predicted damaging variation in WAPL, PDS5A, or PDS5B. Most individuals with WAPL SNVs have mild-moderate developmental problems, and facial dysmorphism and neurological issues are each present in about half of subjects. Many individuals have behavioral challenges, growth delays, and musculoskeletal defects (e.g., clubfoot). WAPL subjects 3–4 were removed from this figure for lack of phenotype detail. PDS5A and PDS5B cases are enriched for developmental and variable neurological issues, but they do not coalesce into defined syndromes. Phenotypes are described in Tables 1 and S4. b. A comprehensive literature review identified 27 individuals with recurrent 10q dels. Sibs were each retained rather than collapsed to one per family. The similar frequency of some features between 10q del patients and WAPL point mutation patients (e.g. developmental, dysmorphism, musculoskeletal) suggests that WAPL may be a driver of these phenotypes. c. Methylation signature of a pLoF WAPL variant p.(Arg674LeufsTer3) compared to known CdLS genes. Subjects with WAPL SNVs do not have a DNA methylation pattern that clusters with the overall methylation signature of CdLS.

Figure 3.. CRISPR-engineered human iPSC and iN models of WAPL^+/− and heterozygous 10q del or dup reveal disease-relevant transcriptional disturbance.

a. The CRISPR-Cas9 system was employed to introduce heterozygous fs indels in WAPL or to generate the 7.8 Mb 10q del or dup. Edited iPSCs were differentiated into iNs. iPSCs and iNs were RNA-sequenced. b. Genotypes and number of differentially expressed genes (DEGs) at FDR <0.1. ↑ upregulated. ↓ downregulated. See panels (i-j) for description of “dose responsive.” c-h. Volcano plots showing DEGs. Purple, FDR<0.1. Green, FDR≥0.1 and p<0.05. Gray, p≥0.05. The x-axis range is limited to log₂ of −2 to 2 for clarity. i-j. Volcano plots of genes, genome-wide, that are reciprocally dosage sensitive in 10q iPSCs (i) and iNs (j). Data are derived from an analysis in which del, dup, and wt samples were assigned to a (−1,1,0) vector. With this definition, 10q region genes correlate positively with 10q copy number and have positive fold change (FC). k-l. Expression changes, plotted as log2 FC of genes within and flanking the 10q del and dup in iPSCs (k) and iNs (l).

Figure 4.. Signatures of altered gene expression in models of WAPL^+/− and heterozygous 10q del or dup.

a-d. DEG comparisons between cell types within a given genotype. WAPL^+/− DEGs (a) significantly overlap between cell types (iPSCs and iNs); however, the fold change (FC) of the intersection of these genes is not correlated. 10q del (b), 10q dup (c), and 10q dose responsive (d) DEGs, with genes within the 10q GD region removed, each significantly overlap between iPSCs and iNs; the FC of the intersection of these genes is correlated for 10q del and 10q dose responsive DEGs. Overlap of DEGs at the p<0.05 threshold are shown as Venn diagrams, and the correlation of the FC of the intersection of these genes as dot plots. P values of gene overlap are from Fisher’s exact test. Linear regression lines are shown, and correlation statistics are listed below each dot plot. The axes are limited to −2 to 2 for clarity (see Table S9 for complete DEG list). e-f. DEG comparisons between genotypes within a given cell type. WAPL^+/− and 10q del DEGs significantly overlap and the FCs of the intersection of these genes are correlated in iPSCs (d) and iNs (e). g-l. The most significant biological processes enriched among DEGs (at a threshold of FDR<0.1) from each model, via gene ontology (GO) analysis. Dotted lines are FDR <0.1

Figure 5.. Analyses of neonatal behavior in Wapl^+/+ and Wapl^+/− pups.

a. Righting Reflex. Wapl^+/− pups performed better than their wt cohorts (p_genotype = 0.03). b. Geotaxis. Wapl^+/− pups outperformed their wt cohorts (p_genotype = 0.0006). c. Four Limb Hang. Younger Wapl^+/− pups (≤8 days) performed better than their wt cohorts (p_genotype = 0.02) but the effect of genotype diminishes with age. d. Gait. Genotype has no discernible effect. e. Growth curves. Wapl^+/− pups are significantly lighter than wt (p_genotype = 0.0001). Data in (a-c) are presented as median, the 25–75% quartiles, and max/min values. Data in (d-e) are presented as mean ± SEM. Wapl^+/+ female = solid magenta; Wapl^+/− female = hatched magenta; Wapl^+/+ male = solid blue; Wapl^+/− male = hatched blue. For (d) and (e), circles/solid lines = wt; triangles/dashed lines = het. N = 32: 8 Wapl^+/+ females; 12 Wapl^+/− females; 6 Wapl^+/+ males; 6 Wapl^+/− males. Statistical significance was evaluated by 3-way ANOVA with genotype, sex, and age as independent variables.

Figure 6.. Analysis of adult behavior in Wapl^+/+ and Wapl^+/− 5- to 7-month mice.

a. Open Field Test, to assess mobility and exploratory behaviors. Total distance (a-i), speed (a-ii), time mobile (a-iii), time in the center region (a-iv), time immobile in the center region (a-v), and time immobile in the edge region (a-v) during the 30-minute test. Statistics are p_genotype from 2-way ANOVA. b. Y-Maze. Wapl^+/− female and male mice alternated 11% and 10% more than wt cohorts. Statistic is p_genotype from 2-way ANOVA. Additional metrics of this assay are shown in Fig. S14a. c. Morris Water Maze, to assess the ability of mice to use and remember visual clues to navigate to a hidden platform. Latency (time to platform) is measured on 6 consecutive days. Wapl^+/− mice performed significantly worse. Statistic is p_genotype from 3-way ANOVA. See Supplemental Fig. 14b for full summary of ANOVA analyses. Performance is especially poor in female Wapl^+/− (see Fig. S14c–d). d. Fear Conditioning. On day 1, we trained mice to associate environmental cues – context and an auditory signal – with a mild electric foot shock. After 24 hours, mice were re-exposed to the original context and then subsequently to the auditory cue in a novel context and their resultant freezing behaviors were used to evaluate associative learning. No significant differences in freezing were measured. Statistics are p_genotype from 2-way ANOVA. e. Rotarod. Het female and male mice performed 15% and 19% better that wt cohorts. Statistic is p_genotype from 2-way ANOVA. Data in (a-e) are presented as median, the 25–75% quartiles, and max/min values. In (a-c), N = 64: 16 Wapl^+/+ females; 16 Wapl^+/− females; 16 Wapl^+/+ males; 16 Wapl^+/− males. In (d), N = 32: 8 Wapl^+/+ females; 8 Wapl^+/− females; 8 Wapl^+/+ males; 8 Wapl^+/− males. In (e), N = 53: 12 Wapl^+/+ females; 13 Wapl^+/− females; 15 Wapl^+/+ males; 13 Wapl^+/− males.