Section 1: demographic and clinical characteristics of adult and paediatric cardiomyopathy in 100KGP

Characteristics of the cohort are summarised in Fig. 3 and Table 1. We identified 1918 individuals of all ages with cardiomyopathies from 90,109 participants (2%) as of October 2022. The 1563 probands comprise 1290 adults (83%) and 273 children (17%). One hundred forty-four (9%) had an age at consent ≤ 16 years and were designated as children meeting stringent criteria. Overall, 1368 (88%) probands were recruited under a CM specific disease category and were designated as Primary CM, and 195 (12%) probands were recruited under another category but had an HPO term containing ‘cardiomyopathy’ and were designated as Complex CM.

Fig. 3

Workflow depicting the number of participants in 100,000 Genome Project (100KGP) queried and filtered for cardiomyopathy (main-programme_v16_2022-10–13). Participants were defined as paediatric if they met any of the following criteria: age at consent ≤ 16 years old; age at diagnosis ≤ 16 years old; age at onset between 1 and 16 years; age at first cardiology appointment (using hospital episode statistics (HES) outpatient codes 320 or 170) ≤ 16 years old. Primary CM: participants recruited under a specific cardiomyopathy disease category. Complex CM: participants recruited under a non-cardiomyopathy disease category with a human phenotype ontology (HPO) term containing ‘cardiomyopathy’. Diagnosis returned represents the number of probands designated as solved in 100KGP. CM, cardiomyopathy

Table 1 Demographic characteristics of 1563 cardiomyopathy probands in the 100,000 Genome Project (100KGP)Comparing adult and paediatric cardiomyopathy—demographics

Cardiomyopathies in children are much rarer than in adults, and previous work has suggested they may have a distinct genetic architecture. A proportion present almost uniquely in childhood particularly during infancy, and a proportion represent early presentations of CMs that can have a broad age of onset (seen in particular in adolescents). Given this distinction, we sought to separately characterise and compare the adult and paediatric groups. Male probands predominate across all ages: 58% of children with CM and 63% of adults. This contrasts with the broader 100KGP pilot study on all rare disease which found a higher proportion of males to females in paediatric but not adult rare disease probands [20]. Ancestry (self-reported) is not stated for a large proportion 284/1563 (18%). Excluding those where it is not stated, a greater proportion of adults (85%) than children (77%) are recorded as White, p = 0.0028. Asian probands make up 14% of the paediatric CM cohort and only 7% of the adult, p = 0.0012; this is in keeping with findings from the overall rare disease pilot [20]. Ancestry by diagnosis also differed. Notably, Asian ancestry comprises 14% of the overall paediatric cohort but 23.8% of children with ‘Complex CM’. Family history was reported in more adults than children, 44% vs 35% (see Table 1). This difference is more marked when stringent criteria were used to define children (see Additional file 1: Table S6). Those without a family history were more likely to be recruited as a trio with mother and father compared to those with a family history, p < 0.0001.

Comparing adult and paediatric cardiomyopathy—family structure, recruitment and natural history

As expected, many more children were enrolled as trios, 54%, compared to 5% of adults (see Fig. 4). Most adults were recruited as singletons (69%). When paediatric probands are defined using stringent criteria, the percentage enrolled as trios is even higher (75%). Interestingly trios did not contribute more to the solved cases than unsolved cases, p = 0.83, and this stands when using stringent criteria to define children (see Additional file 1: Table S8).

Fig. 4

Percentages of different family structures at enrolment for paediatric and adult cardiomyopathy probands and ‘solved’ probands in 100,000 Genome Project (100KGP). Stacked bars are labelled with numbers of probands. Paediatric—paediatric probands with cardiomyopathy (CM); Paediatric solved—paediatric probands with CM and a molecular diagnosis; Adult—adult probands with CM; Adult solved—adult probands with CM and a molecular diagnosis. Singleton refers to a proband for whom no other family member was recruited, duo with other biological relative refers to a proband–non-parent pair, duo with mother or father refers to a proband-parent pair, families with three or more participants refers to a proband recruited with 3 or more biological relatives (but not both mother and father), and trio with mother and father refers to a proband and both parents ± other individuals from the family. Participants are labelled as ‘solved’ in 100KGP if they had a genetic diagnosis that explained their presenting disease

Recruitment

A much larger proportion of children are grouped under Complex CM compared to adults, 34% vs 8% respectively. The proportion of Complex CM is even higher amongst children meeting stringent age criteria, 65/144 (45%) (see Additional file 1: Table S6). Those with Complex CM were recruited under a broad list comprising 52 different disease categories (see Additional file 1: Table S9). The most frequently used at any age was ‘Mitochondrial’.

Co-existing congenital heart disease (CHD)

Co-existing CHD was documented in 32/273 children (11.7%). Twenty-one (7.7%) had an atrial septal and/or ventricular septal defect. Eleven (4%) had another congenital heart lesion (including but not limited to aortic valve stenosis, tetralogy of Fallot, interrupted aortic arch and transposition of the great arteries). Over half of the children (17/32) with co-existing CHD were recruited under a specific CM disease category (designated as Primary CM). In comparison, only 12/1290 (0.9%) of adults had a documented CHD, the majority with atrial septal and/or ventricular septal defects.

Natural history: 100KGP cardiomyopathy cohort have severe disease

Amongst children, 38/273 (14%) have died. This included 15 children designated as Primary CM (15/180, 8%) and 23 with Complex CM (23/93, 25%). Nearly half of these children were recruited with an undiagnosed metabolic condition or a mitochondrial disorder, 17/38 (45%). The majority of children who died did so before the age of 5 years, 27/38 (71%). The median follow-up time from diagnosis was 88 months (maximum 787 months).

For adults, 126/1290 (9.8%) have died. This includes 102 adults designated as Primary CM (102/1188, 8.6%) and 24 as Complex CM (24/102, 23.5%). Over half of the adult deaths occurred in individuals recruited under HCM, 67/126 (53%). The majority of adults who died were over the age of 50, 103/126 (81.7%). The median follow-up time from diagnosis was 115 months (maximum 785 months).

There was no significant difference overall between adults and children recruited to 100KGP in survival probability from the time of disease onset (see Fig. 5a). However, for children, the curve illustrates a steep drop off in survival probability at diagnosis before levelling off. This is also evident when comparing adults and children from birth (Fig. 5b). Survival probability is impacted by CM type. Those presenting with ‘Complex CM’ had the lowest survival probability (see Fig. 5c–f). These survival probabilities reflect only the patients recruited to the 100KGP and do not account for those who may have died before recruitment or those who were not recruited for any other reason.

Fig. 5

Kaplan–Meier survival analysis for adult and paediatric probands with cardiomyopathy (CM) in 100,000 Genomes Project. a and b Comparing adults and children from diagnosis and birth respectively. c and d Comparing survival probability for adults with different CM diagnoses from diagnosis (c) and birth (d). e and f Comparing survival probability for children with different CM diagnoses from diagnosis (e) and birth (f). p values derived from a log rank test. DCM—dilated cardiomyopathy (probands recruited under DCM and any Mixed_CM or Mixed_other probands where at least 1 recruitment category was DCM); HCM—hypertrophic cardiomyopathy (probands recruited under HCM and any previous Mixed_CM or Mixed_other where at least 1 recruitment category was HCM); Other includes probands recruited under LVNC—left ventricular non-compaction cardiomyopathy or ARVC—arrhythmogenic right ventricular cardiomyopathy. Complex CM: probands recruited under a non-cardiomyopathy disease category, but with a human phenotype ontology (HPO) term containing ‘cardiomyopathy’

Section 2: genetic architecture of adult and paediatric cardiomyopathy in 100KGPDiagnostic yield following GEL’s initial analysis

After GEL’s initial analysis, the rate of genetic diagnosis was significantly lower in adults with CM than in children (11% vs 19%, p = 0.0007). The solved rate for children remained at ~ 20% when using stringent criteria to define age (see Additional file 1: Table S8). An additional 10 children and 27 adults were reported as ‘partially solved’. Full variant details can be found in Additional file 1: Tables S10 and S11.

Diagnostic yield varies by age and cardiomyopathy subtype

The diagnostic yield was highest for those children designated as Complex CM, 23%. For children recruited specifically under HCM, it was 18% and under DCM, 17%. This is lower than other studies of paediatric CM [10, 13] reflecting what should be a discovery cohort depleted for known causes of CM.

Adults with DCM had a higher diagnostic yield (18%) compared to those with HCM (9%) reflecting the lower diagnostic yield of NHS testing for DCM vs HCM at the time (see Table 2).

Table 2 Number of paediatric and adult cardiomyopathy probands by disease recruitment category in the 100,000 Genome Project (100KGP) and the number and percentage of solved probandsMost diagnoses are attributed to known cardiomyopathy genes

Despite eligibility criteria requiring participants to have undergone standard genetic testing, most positive findings involve known genes that are expected to be analysed in routine diagnostic panel sequencing. For example, variants in MYBPC3 and MYH7 are still the most frequent cause of HCM in both adults and children (see Fig. 6).

Fig. 6

Genes implicated in the solved paediatric (n = 52) and adult (n = 146) cardiomyopathy probands in 100,000 Genome Project split by recruitment disease categories: Primary—DCM, HCM, Other (which includes ARVC and LVNC) and Complex CM (participants recruited under a non-cardiomyopathy disease category but who have a human phenotype ontology (HPO) term containing ‘cardiomyopathy’). Labels on bars indicate number of probands. Genes are colour coded: black genes are on the current R135 CM panel ‘green’ list; grey genes are on the R135 CM panel ‘green’ list but are not robustly associated with the type of cardiomyopathy reported in the solved case, e.g. PKP2 and HCM; dark purple genes are not on the current R135 CM panel ‘green’ list but CM is a recognised feature of the associated disease (NDUFA4 and FKRP are on the ‘amber’ list); light purple genes in bold are not on the current R135 CM panel ‘green’ list and CM is not a recognised feature of the associated disease. Full variant details can be found in Tables S10 and S11. *One adult proband recruited under HCM has both a pathogenic variant in PKP2 and a likely pathogenic variant in LZTR1 (see Table S11 for details). R135 CM panel—NHS Genomic Medicine Service paediatric or syndromic cardiomyopathy panel (R135 v3.44); ‘green’ list—genes with the highest level of evidence; ‘amber’ list—genes with moderate evidence; ARVC, arrhythmogenic right ventricular cardiomyopathy; CM, cardiomyopathy; DCM, dilated cardiomyopathy; HCM, hypertrophic cardiomyopathy; LVNC, left ventricular non-compaction cardiomyopathy

Is solved really solved? 11% of CM diagnoses in 100KGP involve genes not on an existing UK CM panel

Twenty-two CM cases where the recruiting GMC concluded the case to be solved involve genes not on the current R135 CM panel ‘green’ list (see Fig. 6 genes in bold). Most of these genes (12/22) are associated with syndromic conditions where CM can be a feature but is unlikely to be found in isolation, e.g. NEB related nemaline myopathy. In keeping with this, it is mostly those with complex CM, i.e. those who were recruited under a different disease category who have findings in these genes. Variants in these genes would have been tiered by GEL for analysis because the patient’s additional phenotype terms triggered other disease panels to be applied (see dark purple genes in Table 3). Two of these twelve genes, NDUFA4 and FKRP, are already on the R135 CM panel ‘amber’ list. However, going forward it would be reasonable to include the other 10 genes on a syndromic paediatric CM panel and they may already be on more comprehensive gene lists used by laboratories outside of the UK.

Table 3 Disease and cardiomyopathy association for twenty-two genes not on the R135 CM panel ‘green’ list where variants were identified and considered diagnostic for cardiomyopathy probands in 100,000 Genomes Project (100KGP)

In contrast, for 10/22 genes responsible for solved cases and not on the R135 CM panel, CM is only very rarely, or not known to be associated, e.g. ANKRD11 related KBG syndrome or LDLR related familial hypercholesterolaemia (see Table 3). It is not possible from the information available in 100KGP to be sure why the recruiting centres concluded these CM cases were solved. These could be partial diagnoses where the CM phenotype remains unexplained, or the CM could be a secondary finding. In some instances, the reported CM could be an expansion of the known phenotype. Further input from the recruiting clinical centre will be needed to resolve these cases.

There are also cases documented as solved where the gene is on the CM panel, but not robustly associated with the type of CM reported, e.g. PKP2 seen in HCM (see grey genes in Fig. 6). Again, further input regarding the phenotype of the patient is needed to unravel these cases.

How does the diagnostic yield compare with other rare disease in 100KGP?

As of October 2022, 17% of all rare disease participants in GEL were reported as ‘solved’—i.e. they have a genetic diagnosis that explains their presenting disease. In comparison, 12.5% of the CM cohort were reported as solved.

Section 3: re-analysis of the paediatric cardiomyopathy probands—can we improve diagnostic yield?

We re-examined 210 unsolved PCM probands using the methods outlined in Fig. 7. In addition, we analysed any paediatric proband with a solved or partially solved label where their CM did not appear to be fully explained.

Fig. 7

Overview of analyses and sources of diagnoses for 272 paediatric cardiomyopathy probands with available genome sequencing data in 100,000 Genomes Project. Initial GEL analysis pipeline identified (partial) diagnoses for 62 probands; 11 probands had variants identified by other researchers and reported via the Genomics England diagnostic discovery pathway. In our re-analysis, 60 probands had 65 variants of interest identified by 5 methods: identifying rare protein altering variants in nuclear-encoded genes in a candidate list; identifying structural variants in a candidate gene list; reviewing all protein coding de novo variants where trio data was available; reviewing variants flagged as pathogenic or likely pathogenic by ClinVar; reviewing top 5 Exomiser hits. GEL, Genomics England; P/LP, pathogenic/likely pathogenic; PAV, protein altering variant; SV, structural variant; VUS, variant of uncertain significance

We identified a probable diagnosis (at least one pathogenic or likely pathogenic (P/LP) variant) in 31 unsolved PCM probands and a possible diagnosis (at least one Hot VUS) in a further 18 individuals. A classification of ‘Hot VUS’ was assigned when there was evidence towards pathogenicity (but insufficient to reach LP) and there is a reasonable likelihood that additional evidence such as RNA splicing studies, functional assays or co-segregation with disease in multiple affected family members would clarify its significance. An additional LP variant has been identified in one previously solved proband (see Table 4). No additional potentially causative variants were found for those participants labelled as ‘partially solved’ by GEL. Further variant details are given in Additional file 1: Table S12.

Table 4 Probable and possible diagnoses identified from the re-analysis of paediatric cardiomyopathy probands in 100,000 Genomes Project. Probable diagnosis—at least one pathogenic or likely pathogenic (P/LP) variant identified; possible diagnosis—at least one Hot VUS identified. A classification of ‘Hot VUS’ was assigned when there was evidence towards pathogenicity but it was insufficient to reach LP and there is hope that additional evidence such as RNA splicing studies, functional assays or co-segregation with disease in multiple affected family members would clarify its significance. One previously solved proband has a LP and VUS in ALPK3 identified. The majority of variants were identified using a candidate cardiomyopathy panel (CM panel). Variants were also sourced using ClinVar, Exomiser top 5 hits and reviewing structural variants (SV) and de novo variants. Some variants were also identified by the Genomics England (GEL) diagnostic discovery pathway (diag) or identified by GEL on a second analysis. HGVSc, Human Genome Variation Society coding DNA sequence; HGVSp, Human Genome Variation Society protein sequence; CM, cardiomyopathy; comp het, compound heterozygous; het, heterozygous; hom, homozygous; hemi, hemizygous. aPMID 34732400 [40]; bdoi-https://doi.org/10.1111/bjd.21325 [41]; cPMID 32396390 [42]; dPMID 34212438 [43]. Further variant details are available in Table S12

Potentially causative variants were predominantly identified in unsolved cases with Complex CM (19/49, 39%) or HCM (17/49, 35%), with a smaller proportion (10/49, 20%) identified in those with DCM. Family history was reported in 45% (22/49).

Twenty percent (10/49) were compound heterozygous or homozygous variants and 12% (6/49) were de novo dominant variants. For most, inheritance is unknown due to recruitment as singletons or duos. Structural variants and intronic and splice region variants predicted to impact splicing were found in 12/49 (24%) individuals.

Which advances have improved our yield?

Variants in several known genes were identified, including 8 variants in MYH7 and MYBPC3. For MYH7, variant level evidence has grown over time allowing some previously classified variants of uncertain significance to be upgraded [44]. For MYBPC3, several intronic variants known to impact splicing have been implicated in HCM [42]. Most of these are deep intronic variants that would not have been returned for review initially but are now reported as P/LP in ClinVar. Using ClinVar helped to identify 11/49 (22%) potential diagnoses.

Using our broader candidate gene list, we identified 14/49 (29%) potential diagnoses in genes not on the original CM panels used by GEL (see Fig. 2). Six probands have variants in FHOD3 and FLNC, genes more recently associated with CM. FLNC was updated to diagnostic grade (‘green’) for DCM on PanelApp in September 2019 and curated as definitively associated with DCM by ClinGen [45] in 2020 [46]; FHOD3 was updated for HCM in PanelApp in December 2019. In FHOD3, we identified two variants affecting the same essential splice donor site of exon 12 reported previously in 3 unrelated families [16, 47].

Findings in rarer causes of CM include a heterozygous frameshift variant in NAA15, a gene associated with intellectual disability and congenital heart disease, but also found to cause HCM in 2 unrelated children [48]; a homozygous variant in DOLK, which is associated with congenital disorder of glycosylation type Im which includes a DCM phenotype; a homozygous variant in TREX1 which is associated with Aicardi-Goutieres syndrome in which mouse models develop CM; and a homozygous variant in KLHL24, a gene known to be a rare recessive cause of HCM.

Using a gene agnostic de novo analysis, we identified four probands with LP de novo single nucleotide variants (SNVs). The genes were RYR2, LZTR1, JAK1 and MAP3K7. In this PCM cohort, there are two probands with de novo missense variants in MAP3K7. Pathogenic variants in MAP3K7 have been associated with cardiospondylocarpofacial syndrome (CSCF); phenotypically this can overlap with Noonan syndrome (NS). A study looking specifically at a cohort of patients with CSCF and MAP3K7 variants observed 4/12 patients with CM (one HCM and three DCM) [49]. This gene is not routinely assessed in patients with either syndromic or isolated CM. LZTR1 is associated with Noonan syndrome which includes HCM as part of the disease. There is limited evidence to support a relationship between RYR2 and HCM, therefore further phenotype information will be required before a diagnosis can be established. JAK1 is not known to be associated with CM.  

Updated analysis of SVs > 50 bp, filtered for rarity and our broad gene list, revealed deletions and duplications in 4 probands affecting FLNC, TTN and RYR2. A further two duplications affecting the ATAD3 cluster were flagged through the diagnostic discovery pathway. Finally, other SNVs were prioritised using Exomiser [26] (see Table 4 and Additional file 1: Table S12 for full details).

Overall, by using our broader approach we identified variants in 34 unique genes. Seven of these 34 genes are not on the R135 CM panel but CM is a recognised feature of the associated disease, e.g. MTO1 related infantile hypertrophic cardiomyopathy and lactic acidosis (see Table 4). For a further three genes there is no association with CM reported (JAK1, SCN8A and FGFR3). These genes were identified from the de novo and ClinVar analysis and their known disease associations appear in keeping at least in part with the proband’s phenotype terms. However their contribution to CM is unknown therefore they may represent only partial diagnoses.

How does this analysis compare with other diseases that have been re-analysed in 100kgp

This re-analysis of the PCM cohort has identified a probable or possible diagnosis in 49 previously unsolved participants and has the potential to increase diagnostic yield by up to 18%. These results have been submitted to the diagnostic discovery pathway in GEL for further assessment before being sent to diagnostic laboratories for final variant classification.

Our outcomes are in keeping with those of other diseases that have been re-analysed. Hyder et al. demonstrated a similar diagnostic uplift from 14 to 29.8% in craniosynostosis patients and found a much higher success rate for syndromic presentations [50]. Similarly, Best et al. made a research diagnosis in an additional 19.3% of ciliopathy patients [51].