To diagnose SD conditions, radiographic imaging in early childhood is key; thus, our cohort of only paediatric patients consisted of suitably appropriate clinical-radiographic data for further genetic investigation. None of the patients had molecular confirmation of their condition before WES testing. Upon WES testing, a three-stage approach to variant analysis (as outlined in the methods) was undertaken. This allowed molecular scientists to focus their search, identifying pathogenic variants efficiently and saved time spent analysing less likely variants.
Significant variants leading to a likely or confirmed diagnosis were identified in 53.3% (n = 8/15) of patients; 46.7% (n= 7/15) having a definite molecular diagnosis and 6.7% (n= 1/15) having a likely molecular diagnosis. In comparison, Bae et al. reported a likely or confirmed molecular diagnosis in 52.5% (n=84/160) of similar patients (excluding those patients who had molecular confirmation prior to WES testing).14 This means that our yield was comparable to Bae et al., and was marginally higher. If we include patient 15 as a confirmed diagnosis, who after negative WES testing was concluded to have an acquired SD aetiology, then our yield reached 60% (n= 9/15).
The possible reasons for a slightly higher yield as compared to Bae et al. includes increased gene discovery linked to SD conditions since Bae et al. was published (2016), improved bioinformatic pipelines, technological advances in next-generation sequencing machinery (e.g. improved coverage and read depth) and triple radiological review within our analysis (all three reviewers being expert contributors to the International Skeletal Dysplasia Society).14 We note that radiological input was key with patients 2 and 4. For patient 2, the radiology review led to a change in the pre-test working diagnosis from spondyloepiphyseal dysplasia congenita (SEDC) to campomelic dysplasia (CD), enabling a better genotype-phenotype match, thus securing the diagnosis. Likewise, the radiological diagnosis with patient 4 identified the correct spectrum of disorders so that when a COL11A2 (collagen type XI alpha-2 chain) variant was found, the appropriate final diagnosis was more easily identified.
Of patients with a known diagnosis pre-WES (n=10), WES led to a confirmed molecular diagnosis in 7/10 cases (rising to 8/10 if we include patient 8 as another confirmed diagnosis, though of an acquired cause). This highlights the significant return when there is a known diagnosis pre-test.
Of patients with unknown clinical diagnoses (n= 5/15), WES led to molecular confirmation (likely or highly likely) in 20% (n =1/5) of cases which is the same detection rate as Bae et al.14 This was expected considering low pre-test diagnostic SD hypotheses.
WES yields for non-SD conditions (25-51%, see table 1) are generally lower than for SD disorders (this study; up to 60% and Bae et al. 52.5%).14 This is likely due to the rich combination of radiology and clinical features, and synergism between multiple experts, enhancing phenotypic-driven bioinformatics analysis.
We further discuss selected patients to extrapolate key learning points and have grouped the patients into themes.
Theme 1: Known clinical diagnosis with clear single-gene cause, but testing unavailable.
Patient 1 had a clinical diagnosis of spondyloepiphyseal dysplasia tarda (SEDT; OMIM #313400). SEDT is caused by heterozygous TRAPPC2 (tracking protein particle complex subunit 2) variants. Several males across multiple generations had a similar diagnosis but no known genetic testing. At the time of diagnosis in patient 1 (2014), TRAPPC2 testing was not readily available in the UK (except on a research basis).16 Thus, WES was performed, with primary analysis directed towards TRAPPC2. A pathogenic TRAPPC2 variant was identified (see Table 2). With many laboratories moving to WES and WGS, the availability of many single-gene tests is decreasing; thus, WES approaches with targeted analysis are becoming ubiquitous. It is therefore crucial that clinicians make a pre-test diagnosis to enable targeted analysis and prevent the generation of unwanted and unrelated variants.
Theme 2: Known clinical diagnosis (heterogeneous condition), multiple potential genetic causes.
Patient 4 had macrocephaly, a flat facial profile, nasal bridge depression, small nose, micrognathia, a small bell-shaped thorax and short long-bones with widened metaphyses. The original diagnosis was ‘mild’ fibrochondrogenesis type 2 (FBCG2; OMIM #614524), which is caused by COL11A2 variants and can be lethal. The differential included fibrochondrogenesis type 1 (OMIM #228520) caused by COL11A1 as well as Stickler syndrome (OMIM #108300) hence multiple genes could be potentially causative; thus, WES was an appropriate strategy. The radiological review noted vertebral coronal and sagittal clefts and with increasing age (see Figure 1); enlarged epiphyses and clinically hearing loss became manifest. This led to an alteration of the diagnosis to otospondylomegaepiphyseal dysplasia (OSMED; OMIM #215150), an allelic disorder to FBCG2. WES confirmed compound heterozygous COL11A2 variants which cause a spectrum of disorders from mild deafness to OSMED to potentially lethal FBCG2. This case illustrated radiological (e.g. mega-epiphyses) and clinical clues (hearing loss) that may only become apparent with time. It also highlights the need for clinicians to be careful of using ‘old’ diagnostics labels as milder forms of previously ‘lethal’ conditions or what we considered ‘extremely’ severe phenotypes are emerging. In such instances, umbrella terms like COL11A2-spectrum disorder as a pre-molecular diagnosis may be more apt rather than FBCG2.
Patient 5, a six-year-old male of normal stature who presented with leg pain, genu valgum, pes planus, muscle weakness and decreased hip mobility, thus had possible multiple epiphyseal dysplasia (MED; OMIM #132400) with non-specific myopathy. Creatinine kinase was normal, and electromyography suggested myopathic disorder, muscle biopsy showed mild type 2 fibre atrophy and array Comparative Genomic Hybridisation (aCGH) was normal. MED is characterised by multiple long and or short bone epiphyseal abnormalities and has autosomal dominant (ADMED; OMIM #132400) and autosomal recessive (ARMED; OMIM #226900) forms.17,18 ADMED can be due to; COMP (cartilage oligomeric matrix protein), COL9A1-COL9A3 (collagen type IX alpha 1-3 chain) or MATN3 (matrilin 3) and presents in childhood with joint pain, exercise-induced fatigue, restricted mobility, short stature, and early-onset osteoarthritis.18 Radiology showed delayed epiphyseal (long-bone) ossification and small and irregular epiphyses (especially in the knees and hips)(see Figure 2). Due to MED heterogeneity, WES was an appropriate strategy for molecular diagnosis and identified a pathogenic MATN3 variant confirming MED Type 5 (OMIM #607078).
Patient 7 presented aged 3 years with an undefined SD, multiple falls, complex congenital heart disease (CCHD), relative macrocephaly and mild developmental delay. Skeletal survey at 10 months highlighted irregular metaphyses. He had leg-length discrepancy, right leg bowing, mild joint hypermobility and hypotonia. Further radiology revealed worsening metaphyseal dysplasia (expansion, cup-shaped irregularity) with radiolucent non-ossified cartilage affecting major joints. At 6.5 years, hand radiology showed severe widespread bilateral multiple enchondromas (see Figure 3d), that were also present in the lower limbs (see Figure 3a). The diagnosis was felt to be an enchondromatosis-like condition and upon radiological review; metaphyseal chondromatosis with D-2-hydroxyglutaric aciduria (MC-D2HGA; OMIM #614875) was suspected. Though MC-D2HGA was highly suspected, other genetic disorders with very similar phenotypes remained thus WES was an appropriate strategy.
Enchondromatosis (EC) is a rare heterogeneous condition with multiple enchondromas (benign hyaline cartilage forming tumours in the metaphysis).19 The commonest two subtypes are Ollier disease (OMIM #166000) and then Maffucci syndrome (OMIM #614569).19 MC-D2HGA is another rarer subtype with only 11 reported cases (four of which are due to somatic mosaicism of IDH1 (isocitrate dehydrogenase 1), the remainder lacking molecular confirmation).20
WES testing identified a de novo heterozygous IDH1 c.395G>A p.(Arg132His) germline variant (approximately indicative allele frequency of 50/50, wild type: a variant on Sanger). Urinary organic acid testing showed marked isolated increase in 2-hydroxyglutarate, confirming MC-D2HGA (this result came after WES was initiated but prior to the WEST results). MC-D2HGA is associated with macrocephaly, developmental delay, hypotonia, significant metaphyseal dysplasia, enchondromatosis, dysmorphia and CCHD. This is the first reported MC-D2HGA case caused by germline IDH1 changes (with read depth over the IDH1 variant position of >x250 depth) though the author is aware of one unreported MC-D2HGA case caused by germline changes (Zankl A. [Presentation] 6th Nordic Workshop on Skeletal Dysplasia, Karolinska Institute, Sweden. 5th March 2020.) Had a WES approach not been used, it would have been difficult to reach this diagnosis.
Patient 9 presented with osteopetrosis, a heterogeneous condition, classically divided into; fatal infantile malignant form (OMIM #259700/611490/259720), Albers-Schonberg disease (OMIM #166600/259700) and a milder adult form. She presented with facial nerve palsy, and subsequent radiography noted Rickettsial-like bony ends (see Figure 5). Foraminal impingement led to hearing loss and visual problems and a working diagnosis of infantile-onset osteopetrosis was established, suitable for WES. The proband had compound heterozygous variants in CLCN7 (chloride voltage-gated channel 7); a pathogenic variant inherited from her unaffected father and a VUS (variant of unknown significance) from her unaffected mother, though her maternal grandmother was affected (late-onset osteopetrosis) and had the same VUS. It later transpired that the mother actually had mild radiographic and clinical features demonstrating segregation and striking intrafamilial variable expression of osteopetrosis type 4 (OMIM #611490). If PP4 (protein phosphate-4) or PP1 (protein phosphate-1) panels were applied, the VUS would be upgraded to likely pathogenic. The paternal variant alone did not manifestly cause disease in the father but was felt to negatively modify the severity of the condition in the presence of the maternal variant.
Theme 3: Known diagnosis, no known gene
Patient 11 presented with severe short stature, delayed motor milestones, a high-pitched voice, dysmorphisms, severe scoliosis, discoloured dentition, and no fractures. The working diagnosis after the radiological review was odontochondrodysplasia (OMIM #604505). At the time, this condition had no known gene cause. Thus, an agnostic approach via WES was appropriate for new gene discovery. No pathogenic variants found. Some years later, a causative gene was found, TRIP11 (thyroid receptor-interacting protein 11) though recent re-analysis of the case still did not yield a molecular cause.
Theme 4: Unknown SD diagnosis
Patient 13 presented with SD (short stature, macrocephaly, and scoliosis) with non-skeletal syndromal features (congenital heart disease and hearing loss). CHD7 (chromodomain helicase DNA binding protein 7) gene testing for CHARGE syndrome (OMIM #214800) was negative. WES identified two likely pathogenic variants in EXT2 (exostosin glycosyltransferase 2) found in trans with clinical findings overlapping for AREXT2 syndrome (autosomal recessive exostosis-2 gene syndrome). AREXT2 is an ultra-rare recessive disorder with an unclear and expanding phenotype as only four families have been reported.21 First described in 2015, it was termed Seizures, Scoliosis and Macrocephaly syndrome (SSM; OMIM #616682). Later authors suggested the ‘AREXT2’ label to recognise the lack of uniformity of scoliosis or seizures. It is unclear whether patient 13 has AREXT2 thus, functional work is required. In this unknown syndromal case, WES has ruled out many potential diagnoses and the WES-based agnostic approach allowed for further investigation of an emerging ultra-rare condition.
Theme 5: WES to exclude a genetic diagnosis and support an acquired cause
Chondrodysplasia punctata (CDP) is a rare SD characterised by punctiform calcification of cartilage and is acquired or genetic in origin.22 Genetic forms (OMIM #302950/118650) are heterogeneous. Acquired forms can be due to maternal malabsorption of vitamin K, maternal warfarin or some anticonvulsant.23 Patient 8, a 2-year-old female, presented with faltering growth and antenatal exposure to lamotrigine and topiramate. Antenatal scans demonstrated short long-bones. Post-natal radiology showed sacral stippling with delayed tarsal ossification, consistent with CDP (see Figure 4b).24 Since several genes cause CDP, WES is an appropriate strategy to reasonably exclude underlying genetic causes. Thus, negative WES increases antenatal anticonvulsant exposure as the likely cause. Although there is no reported association of lamotrigine and topiramate, causing stippling, other Hydantoin anticonvulsants (e.g. phenytoin) have been reported to do so.23 This case highlights the need for early radiography, as stippling is often only seen in the first year of life and rarely after age three.25
Patient 15 was born with normal birth and growth parameters until age 11 months when an unexplained prolonged fever led to a diagnosis of haemophagocytic lymphohistiocytosis (OMIM #603553). Biallelic PRF1 (Perforin-1) pathogenic variants (compound heterozygote) were confirmed. At 18 months, matched unrelated cord blood HSCT (homologous stem cell transplantation) was performed. Subsequent progressive growth failure developed associated with radiographic spondyloepimetaphyseal chondrodysplasia, functional asplenia and sensorineural hearing loss. Development of skeletal changes throughout childhood including fixed flexion hip deformity, marked pes planus, marked genu valgum and pectus carinatum. Extensive endocrine investigations returned no cause. WES (and subsequent WGS) was negative for a skeletal cause. An international group of SD experts noted the similarity of this case and several others concluded a potentially new disorder is manifesting with growth failure and a chondrodysplasia phenocopy post early HSCT for non-oncological disorders.26 This case highlights the utility of WES to reduce the likelihood of a genetic cause significantly and increase the confidence of an acquired cause.26
Further Discussion
As many previous studies have discussed, a broad approach to genetic testing through WES or WGS allows for the identification of conditions that may not have been suspected clinically and thus the expansion of previously known phenotypes.
In our cohort of 15 patients, the use of WES has led to three novel findings. Firstly, Patient 7 is the first reported case of MC-D2HGA due to a germline variant. Secondly, patient 13, is potentially the fifth reported family with AREXT2 syndrome providing further expansion of the phenotype. Lastly, patient 15, is one of seven patients who have collectively provided evidence for a new disease, ‘chondrodysplasia phenocopy post early HSCT for non-oncological disorders’ and has been submitted for publication.26
Although WES remains the current ‘go-to’ diagnostic test in many rare disease scenarios, we increasingly see a shift to WGS, especially for the acutely ill child. Even then, a large body of undiagnosed patients remain. When WES returns negative or inconclusive results, for many, the diagnostic odyssey is abandoned or halted. Yet it is clear that re-analysis of WES data can often result in diagnosis in 10-15% of these cases.2
Additional testing such as long read sequencing (LRS), copy number variants (CNV) in non-coding regions, non-coding variants (NCV), repeat expansion (RE), methylation testing (MT) and other structural changes will need to be explored to increase diagnostic yield. Burdick et al. explored the proportion of diagnoses from additional testing in 54 patients with clinical diagnoses, enrolled in their Undiagnosed Disease Network.2
Of the 54 participants, the molecular diagnosis was obtained in 36 (67%) through WES, and in 15 (28%) through additional testing. Of these, 7/15 (47%) had an NCV, 6/15 (40%) a CNV, and 2/15 (13%) had a RE or a DNA methylation disorder. A yield figure could not be given since there were many other patients within the programme, who were at different stages of assessment. Nevertheless, the report highlights the benefit of testing beyond WES and the approaches that can be considered.
This combined with the advance of radiomics (the systematic use of artificial intelligence to provide diagnostic processing and analysis of ever sophisticated imaging data) will usher in an increased yield. The careful combination of the physician (endocrinologist, orthopaedic surgeon), radiologist and geneticist working together will be crucial.
The difficulty with large scale sequencings, such as WES or WGS is the generation of large numbers of variants of unknown significance (VUS), especially as genetic testing, is mainstreamed. Likewise, bioinformatic pipelines can present de-novo variants with strong computational evidence suggestive of particular conditions, that are not easy to dismiss. The present challenge is, therefore, that of ‘variant interpretation’. Further familial segregation is not always possible.
Traditionally, one way to assess a VUS is to perform functional studies (e.g. fresh blood samples to undertake RNA studies to see if the DNA change has a functional impact on RNA production and therefore on the given protein). This method is challenging as it requires fresh blood samples (as RNA degrades quickly), a difficulty, especially in the paediatric population and blood samples are not always the appropriate sample to test a particular variant as the gene of interest may not particularly be expressed.
Additionally, RNA studies are not always possible in the NHS diagnostic lab, and clinicians often had to partner with university academics or other institutions for functional work (often by providing fresh blood samples, or skin samples or saliva). The problems with this are manifold; time-intensive and dependent on knowing suitable partners. Furthermore, obtaining the required sample can be difficult; we have already discussed the difficulty with blood samples. Skin samples require a skin biopsy which is a relatively invasive procedure. Saliva samples are often difficult to work with, of poor quality/limited in what can be assessed. Lastly, with more and more functional study requests to academics, the boundaries of responsibility for what are essentially research analyses are blurred, thus increasing reluctance from academia to perform such work (culpability issue). Suitable new ‘quick and easy’ tests need developing for variant analysis, and hair pluck analysis may be one such avenue in the SD domain, due to beneficial expression profiles of SD genes.