New diagnoses of EDS and HTAD

We identified 10 diagnostic P or LP variants in genes that are known causes of EDS and HTAD (table 1, online supplemental table 6). Two novel heterozygous pathogenic COL12A1 variants were considered diagnostic. Splice site variant 4 was identified in patient 402 (bilateral congenital hip dislocation): the variant was found in one other individual in gnomAD and had high in silico prediction of pathogenicity (ADA score 0.999). COL12A1 variant 10 resulted in a helical glycine substitution in patient 755 with multiple features suggestive for myopathic EDS (mEDS), including neonatal hypotonia and kyphoscoliosis.

Variant 19 resulted in loss of function in COL5A1 in patient 1528, who had previously declined clinical diagnostic testing (ClinVar ID 280931). Patient 34 with hyperextensible skin, distal joint hypermobility and a carotid artery dissection had an overlapping HDCT/cEDS phenotype and carried the synonymous variant 2 in COL5A1. We had previously classified this as a variant of uncertain significance (VUS).8 14 The variant impacts the last nucleotide of exon 51, with high in silico pathogenicity, and we now consider this likely pathogenic (ClinVar ID 212971). This patient also carried a pathogenic variant in ITGB3 (variant 3) (autosomal recessive Lanzmann thrombasthenia MIM 173470), a gene that has been found to be abnormally expressed in skin fibroblasts from patients with hEDS,15 and a novel variant in candidate gene PGTER4 (see below).

HDCT patient 814 carried novel LP TGFBR2 variant 14 in the Ser/Thr kinase domain, without known vascular involvement. A recent report of this variant and accompanying functional data support LP classification.16 HDCT patient 564, with pectus carinatum and aortic root dilatation, carried a TGFB2 pathogenic variant 9 (CADD=34). A different variant at the same nucleotide was reported as LP in association with syndromic aortic aneurysm (ClinVar ID 440982). Two patients (patient 33 and patient 479) had complex HDCT phenotypes and LP variants in TGFB3 (variant 1) and SMAD2 (variant 8). hEDS patient 1484 had LP variant 18 in COMP (multiple epiphysial dysplasia type 1, MIM 600310). HDCT patient 1420 had LP variant 17 in ALPL causative for hypophosphatasia (MIM 171760).

VUS in EDS, LDS, HTAD and other syndromic genes with potential for pathogenicity reclassification

Thirty variants met the ACGS criteria where further segregation/functional work may enable reclassification as pathogenic or LP (online supplemental table 7).17 Two patients with a clear cEDS phenotype harboured variants in COL5A1 exon/intron 64, which encodes two transcripts in the C-propeptide domain, with alternate splicing in different tissue.18 Patient 583 with COL5A1 LoF variant 29 had cEDS major features: skin hyperextensibility, widened atrophic scars, generalised and small joint hypermobility with additional features of hEDS. cEDS patient 806 has a novel variant 35 at position +6 of intron 64. While a single multi-exon deletion including exon 64 (exons 63i-65i) has been reported as pathogenic, other exon 64 variants remain VUS (https://databases.lovd.nl/shared/genes/COL5A1).

cEDS patient 595 with missense TGFB3 variant 31 (CADD=25) had Mitral Valve Prolapse (MVP) and a high arched palate. hEDS patient 107, with a second-degree relative with an aneurysm, carried an ULK4 splice variant 23. Loss of Function (LoF) variants in ULK4 have been reported to increase the risk of aortic thoracic dissection in a single small study.19 In syndromes with EDS-like features, patient 1530 (female) had splice variant 45, a VUS* in the UPF3B gene, Lujan syndrome (MIM 309520, intellectual development disorder X linked, associated with Marfanoid habitus).20 hEDS patient 107 carried variant 22, a VUS* in KCNH1 (MIM 135500, Zimmerman-Laband syndrome), which may have cartilage abnormalities and gingival hyperplasia as associated features. hEDS patient 967 carried variant 36, a VUS* in FLCN1 (MIM 607273, Birt-Hogg-Dube syndrome), associated with recurrent pneumothoraces and an increased risk of renal carcinoma.

We identified variants in genes associated with a skeletal dysplasia phenotype. cEDS patient 1451 had COL9A3 variant 40, a glycine substitution in the triple helical domain (MIM 120270, AD multiple epiphysial dysplasia type 3 with and without proximal myopathy) and also carried two VUS in COL5A1 (online supplemental table 9). cEDS patient 1002 carried a novel cysteine substitution (variant 37) in MAP3K7 (cardiospondylocarpofacial syndrome, MIM 157800) within the protein kinase domain.

We interrogated our data for Mendelian causes of symptomatology associated with EDS. Erythromelalgia is a SCN9A channelopathy associated with abnormal pain sensation and small fibre neuropathy (MIM 133020). We identified a novel SCN9A variant 27, at a transmembrane domain mutation hotspot, in patient 482 with a vEDS-like phenotype with thin skin and tissue fragility.

We identified patients with two or more rare/novel variants, for example, HDCT patient 72, with terminal digital and nail anomalies and a family history of HTAD had missense variant in WNT10A (variant 21, CADD=30, odontoonychodermal dysplasia/tooth agenesis MIM 606268)) and a VUS in ROBO4 (aortic valve disease 3 MIM 618496) (online supplemental table 10). Multiple patients in the cohort had complex symptoms, signs and/or family histories, suggesting possible enrichment for patients with more than one rare Mendelian disorder.

Variants of uncertain significance in genes associated with risk of ICA

We identified multiple variants in genes previously reported as associated with risk of intracranial aneurysm (ICA) (online supplemental tables 7; 8). hEDS patient 65 with a femoral artery aneurysm and family history of ICA carried ROBO4 VUS and a second VUS in the fibrinogen-like domain of ANGPTL6. Rare variants in this domain have been reported as associated with familial ICA risk.21 Variant 42 (VUS*) in PCNT was found in hEDS patient 1495 who was not known to have a personal or family history of ICA; this variant has been previously reported in familial ICA.22

Autosomal recessive disorders

A further eight heterozygous LP/P variants were identified in autosomal recessive EDS genes and other autosomal recessive genes overlapping with EDS symptomatology, ZNF469, LAMA2, ITGB3, ELP1, ADAM22, C1QC and PRSS56 (table 1, online supplemental tables 6; 7; 9–11). Seven heterozygous VUS* were identified in LAMA2, TNFSF11, TONSL, RYR3, SLC2A10 and CANT1. Multiple VUS in ZNF469, PRDM5, DSE, CHST14, ELP1, AEBP1, CCN6, RYR3, DYSF and LAMA2 (data not shown). HDCT patient 620 with an occipital horn syndrome phenotype, and consanguineous parents, was homozygous for a VUS in SDSL (NM_138432.3 c.626C>T, p.Ala209Val) (MIM 618752, severe congenital neutropenia type 8). Phenotypic review did not show haematological abnormalities: these variants were therefore considered unlikely to be causative.

VUS in EDS, HTAD, myopathy and inborn errors of metabolism genes

Additonal VUS were identified in genes associated with EDS, HTAD, myopathy and inborn errors of metabolism (online supplemental tables 7; 9–11). A VUS in BGN was identified in hEDS patient 1393 (female) with increased arm span to height ratio and talipes, and aortic root dilatation; loss of function mutations in this gene have been reported to result in Meester-Loeys.23 A number of patients carried ultrarare variants in genes associated with non-syndromic HTAD (ROBO4, PRKG1, SMAD6, ULK4, MAT2A, SMAD2, MFAP5). HDCT patient 453 with carotid dissection had a 64 bp insertion predicted to result in out of frame/loss of function transcript in PRKG1 (pLi=1). hEDS patient 1629 without known cardiovascular involvement had a novel SMAD6 VUS in the MH1 domain. hEDS patient 1443 had a family history of abdominal aortic aneurysm in maternal relatives and ICA in a paternal relative carried novel VUS in SMAD6. Patient 526 had MVP and a family history of multiple individuals with cardiac valvular disease, with novel VUS in IFIH (CADD=31), in the helicase domain (MIM 606951, Singleton-Merten syndrome, acroosteolysis and aortic valve calcification).24 HDCT patient 79 carried EMILIN1 VUS at amino acid residue 28, close to residue 22, thought to affect N terminal signal peptide cleavage.25 HDCT patient 422, with camptodactyly and Asperger’s syndrome, carried a novel VUS, resulting in an in-frame deletion mutation in MED12.

We found a single VUS* variant 43, and multiple VUSs in EDS and Bethlem myopathy genes (online supplemental table 9), HTAD (online supplemental table 10), myopathy, inborn errors of metabolism and dysautonomia genes (online supplemental table 11), many of which are similarly classified in ClinVar. These patients did not have specific clinical features (eg, contractures for Bethlem myopathy, cauliflower ears for Beals syndrome or aggressive periodontal disease for pEDS) which might contribute to ACMG criteria PP4.

EDS gene candidates based on linkage and skin fibroblast gene expression studies

We reviewed our data for germline variants in loci previously reported in a linkage study of a large family with hEDS, which identified LZTS1 as a candidate gene (online supplemental tables 12–16).26 A single patient with hEDS in our cohort (patient 703) had a LZST1 missense variant, with limited in silico evidence of pathogenicity (CADD=23). We also identified multiple rare variants (CADD >15) in genes within the reported region of linkage (online supplemental table 12). These included SORBS3 (vinculin binding domain) reported to regulate extracellular matrix (ECM) stiffness in vitro,27 ADAM7, ADAM27 (variants in protease domains), multiple variants in the CCAR1 gene (a regulator of cell division) and DOCK5 (mouse model has reduced skeletal muscle, zebrafish has abnormal fast muscle.28 In addition, we identified multiple rare variants in genes previously reported in a linkage study of Pelvic Organ Prolapse,29 for example, LAMC1, ROBO2 (online supplemental table 13, online supplemental methods).

Gene expression data from skin fibroblasts for patients with hEDS, cEDS and vEDS have been published, suggesting candidacy for several dysregulated genes.15 30 31 We identified multiple rare germline variants with CADD >15, in several of these genes (online supplemental methods and online supplemental tables 14-16). These included integrin signalling, innate immune system function, TRAIL and TRAIL receptor genes, reported to affect integrin signalling in the ECM, controlling vascular remodelling.32 We identified multiple rare heterozygous variants in HSPG2 (Perlecan) (online supplemental table 15). Homozygous variants in HSPG2 cause AR Schwartz-Jampel syndrome (MIM 142461) via disordered cartilage maintenance, osteonecrosis and endomysial dysfunction via a channelopathy mechanism. A knock-in HSPG2 mouse model demonstrated disordered acetylcholinesterase endplate morphology with abnormal patch clamp and a fatigability phenotype.33 Two POSTN variants were found in FAS1 domains (online supplemental table 16): periostin is reported as contributing to tissue repair after injury via upregulating collagen (I) and multiple other ECM component proteins.34

Rare variants in loci associated with GJH/self-reported Beighton score, rotator cuff injury and knee pain GWASs

We identified multiple rare variants with CADD >15 in genes associated (p<5×10^-8) with self-measured Beighton score >5 in a published GWAS6: These included the PIEZO Type Mechanosensitive Ion Channel Component 1 (PIEZO1) and NEDD4 E3 ubiquitin protein ligase (NEDD4) (online supplemental table 17). PIEZO1 is a mechanotranducer protein, important in the cellular responses to shear stress, maintenance of the vascular endothelium and mechanosensation in chondrocytes and epithelium .35 NEDD4 is a mediator of abnormal fibroblast proliferation in keloid scarring.36

HTAD candidate genes

Multiple patients in this cohort had a personal or family history of HTAD, carotid, intracranial and other aneurysmal disease . Careful review of all novel variants with CADD >15 in non-annotated genes revealed a small number of variants with high CADD scores (>20) in candidate genes with published data supporting a role in vascular disease and remodelling (online supplemental table 18). HDCT patient 1625 with a dilated aortic root and megacolon had a novel missense variant 63, in transforming growth factor beta 1–induced transcript 1 gene (TGFB1/1). This gene is regulated by TGF beta signalling; mice lacking its homologue, hic5, show deficient smooth muscle cell response to vascular injury (MIM 602353).37 This variant at TGFB1/1 Arg 67, neighbours phosphoserine 68, hence may disturb signal transduction. kEDS patient 1396 carried variant 59, a nonsense mutation INO80D (MIM 610169). Homozygous missense variants in INO80D were reported in a single family with aortic hypoplasia, aggressive atherosclerotic disease and periodontal disease,38 pLi=1. Patient 34, with HDCT and carotid artery dissection, harboured variant 50 in prostaglandin E receptor 4 (PTGER4) (MIM 601586). Dysregulated expression of PTGER4 has been reported in abnormal wound healing, regulation of vascular tone and blood pressure, in abdominal and thoracic aortic aneurysm and the regulation of cerebral blood flow.39

Reviewing murine and functional studies reported for Marfan syndrome, we identified germline variants in TMBIM1 (MIM 610364), SCUBE3, IRF7, IGFBP2 and TMEM176B and MMP2.40 hEDS patient 1491 with kyphosis and a high arched palate carried FBN3 variant 61 in the TGFbeta binding domain, disruption of the equivalent domain in FBN1 cause Marfan syndrome. hEDS patient 1695 had a loss of function variant 64 in NOTCH4, (LOEUF=0.32), with livedo reticularis and a maternal aunt with pulmonary artery atresia. This gene is known to affect vascular morphogenesis in mice, but has not been associated with disease in humans.41 HDCT patient 446 with carotid dissection carried four variants, including novel variant 54 in NFAT5 (MIM 604708). Osmoregulatory stimulus has previously been found to upregulate NFAT5 expression, resulting in abdominal aortic aneurysm and dysregulated immune function.42 Two other NFAT5 variants were also identified, in hEDS patients 1595 and 922 without aneurysms (online supplemental table 19). We identified an hEDS patient 566 with Marfanoid habitus, arterial rupture and collagen fibril irregularity, who carried a novel loss of function variant in the SYAP1 gene (variant 56); a knockout mouse model for this gene has a highly distinctive motor deficit phenotype43 (the pLi score is 0.94).

Matrisome genes

We searched for rare variants with CADD >15 in genes known to interact with fibrillar collagen biosynthesis and signalling, chondroitin synthesis and modification (https://reactome.org/PathwayBrowser) (online supplemental table 19). Collagenases I/II/III (MMP1, 8, 13 and 4) are known regulators of the fibrillar collagens in the ECM. Variant 60 substituted a histidine residue of Zinc binding site in MMP8, which was previously reported in GWAS as associated with premature rupture of the membranes (MIM 120355). The patient had hEDS with a family history of recurrent miscarriage. Heterozygous missense variant 51 in MMP25 (608482) (online supplemental table 18) was identified in a patient with hEDS: this gene is functional in the innate immune system and abnormal expression has been associated with tendinopathy in a mouse model.44 45 We also noted multiple heterozygous VUS in autosomal recessive skeletal dysplasia genes, CANT1, TONSL, OSTM1 (data not shown).

Biallelic pathogenic variants in ADAMTS2 cause dermatosparaxis type EDS. We identified a patient with HDCT (patient 446) with heterozygous Variant 52 in ADAMTS5 and variant 53 in ADAMTS16. Both variants were in the spacer domains, known to regulate aggrecanase activity. Heterozygous missense variants were also identified in ADAMTS20, ADAMTS22, ADAMTS23, ADAMTS28. Pathogenic variants in C1R/C1S cause pEDS, by gain of function on as-yet unidentified targets,46 we found multiple rare variants in other (non-annotated) serine proteases (online supplemental table 19).

Integrins, ephrin, ciliopathy, TSPANs, DOCK, circadian rhythm pathways

Within the entire cohort, we noted clusters of variants in genes not currently associated with EDS and in novel genes and pathways with biologically plausible links to EDS, including integrins (ITGA3, ITGB4, ITGA8, ITGAV and ITGB1BP1) (online supplemental table 19). Integrin-collagen interactions are integral to wound healing, inflammation, innate immunity and via TGFBeta signalling and other pathways.47 We identified multiple rare variants in ephrins and their receptors (data shown for EPHA8, EFNA1), known to regulate vascular endothelial and corneal proliferation, tissue fibrosis, wound healing and catecholamine synthesis.48 Ciliopathies are generally associated with complex phenotypes; however, variants in IFT88 and NFATC3 were recently reported with bicuspid aortic valve.49 We identified two novel variants in these genes. Wound healing is known to be under circadian rhythm control through local and central mechanisms.50 We identified a small number of variants in PER1 (MIM 602260), PER2 (MIM 603426) and ZFHX3 (MIM 104155). It is possible that abnormal wound healing seen in patients with EDS is due to the disruption of these control mechanisms. We identified multiple variants in DOCK5 (MIM 616904), in the linked region for hEDS. While it has not yet been annotated as causative of disease in humans, a mouse model has a reduced skeletal muscle phenotype and a zebrafish model has abnormal fast muscle.28 We also identified multiple variants in various TSPANS. TSPAN2 regulates TGFB1/SMAD expression in vascular endothelium (MIM 613133).

Genetic burden analysis

In view of the large number of rare variants identified in multiple pathways, a formal burden analysis was carried out to seek statistically significant associations. Burden analysis was carried out using the TASER software13 (table 2). While LOC283685 was close to meeting the criteria for significance (p=2.34e-6, adjusted p=7.41e-6), we identified that the coding sequence of the final exon of GOLGA6L2 transcript ENST00000312015 (Glu308-Ter415), annotated separately in USC GRCh38, probably overlaps the C-terminal sequence of LOC283685 (Glu61-Ter168). The overall burden of rare variants in GOLGA6L2 including this terminal region did not meet significance (p=2.67e-3, adjusted p=4.36e-3). The lack of statistically significant results of this analysis is likely related to the small sample size. A number of the top scoring loci, however, had biological plausibility. The LRTTM4-HSPG (heparan sulfate proteoglycome) complex has been proposed a tetrapartite model for synaptic plasticity involving interactions with the ECM and HSPG has been noted in the vEDS transcriptome. GOLGA6L2 is of unknown function; golgins are a large group of vesicle tethering proteins with tissue-specific effects, other golgins are known to result in reduced bone mineral density and neuromuscular phenotypes (GOLGA2 MIM 602580). ANKFY1 is involved in transport to the Golgi apparatus. ADCY1 (MIM 103072) causes autosomal recessive deafness with abnormalities of circadian rhythm.50