Exome sequencing uncovers promising candidate genes for foetal structural malformations

Karyotyping, FISH, and CMA

Genomic DNA extraction from the foetal tissues/POC and peripheral blood samples of parents was performed using a QIAamp Fast DNA tissue/blood mini kit (QIAGEN, Germany). All the cases were screened for chromosomal abnormalities by karyotyping and/or FISH (n=44). For karyotyping, foetal tissues were treated with collagenase and cultured in AmnioMAX™ (ThermoFisher Scientific, USA) medium followed by incubation at 37°C for 10-12 days. Cultured cells were harvested using 40 µl colcemid and incubated at 37°C for 45 min followed by 0.25 per cent trypsin treatment. For parental karyotyping, 1 ml of blood was added to Dulbecco’s Modified Eagle Medium (DMEM) medium (4 ml DMEM mixed with 1 ml FBS) (GIBCO, ThermoFisher Scientific, USA) along with 100 µl PHA (GIBCO, ThermoFisher Scientific, USA). The cells were cultured at 37°C for 72 h in a CO₂ incubator, and colcemid was added one hour before the termination. Post-harvesting, cells (both POC and blood) were fixed in Carnoy’s fixative, and banding was performed using the standard protocol¹². About 20 cells were analysed per patient using ISCN 2020 guidelines¹³. In cases where karyotyping was not possible, FISH was performed using AneuVysion® Multicolor DNA Probe Kit (Abbott, USA). Dual colour probe – LSI 13/21 was used for the detection of chromosomes 13 and 21) and FISH was tri-colour probe – CEP 18 / X / Y was used for the detection of Chromosomes 18, X and Y¹⁴.In cases with normal Karyotype/FISH results (n=36), CMA was performed using Affymetrix CytoScan™ 750k array as per the manufacturer’s protocol. Further, CMA Data was analysed using Chromosome Analysis Suite (ChAS) version 4.3.0.71.

Whole exome sequencing and data analysis

WES was performed in POCs with normal karyotype or FISH findings (n=30), and parental WES was performed in nine cases. WES was performed on the Illumina NovaSeq platform (mean depth of 80-100X, with 97% coverage of target bases and ≥20x depth). Data processing, alignment, and variant calling were executed following the Genomic Analysis Tool Kit (GATK) best practices framework for germline variant identification. Briefly, sequencing reads were aligned to the human reference genome (GRCh38), and base quality recalibration was performed using the DRAGEN bio-IT platform. Variant calling was accomplished through the GATK pipeline, followed by annotation using ANNOVAR and OpenCRAVAT^15-17.

Clinically relevant causal variants were filtered and prioritised using Golden Helix VarSeq and Varsome workflow implementing the American College of Medical Genetics and Genomics (ACMG) guidelines for the interpretation of sequence variants. Clinically significant variants using the compilation of disease databases such as Online Mendelian Inheritance in Man (OMIM), ClinVar, and the Human Gene Mutation Database (HGMD). Further, variants were filtered based on their minor allele frequencies (MAF) in gnomAD (v3.1.2 and v4.0.0), 1000 Genome phase-3, and dbSNP (GCF_000001405.38). Rare variants with MAF≤0.01 were retained for further analysis. Additionally, in silico pathogenicity prediction tools such as PolyPhen-2, SIFT, REVEL, CADD, SpliceAI, etc.) were employed to glean further insights into the potential functional impact of the variants. Finally, the variants were classified adhering to the established guidelines set forth by the ACMG¹⁸. Pathogenic (P) and likely-pathogenic (LP) variants in genes with a known link to the specific foetal malformations observed were designated as conclusive findings and reported to the parents (Figure).

Close

Figure.

Flow diagram of the pipeline used for WES data analysis and variant prioritization. VEP, variant effect predictor (ensembl); DP, depth; GQ, genotype quality; AAF, alternate allele frequency; MAF, minor allele frequency; LoF, loos-of-function; SIFT, sorting intolerant from tolerant; CADD, combined annotation dependent depletion; REVEL ,rare exome variant ensemble learner; MoI, mode of inheritance; OMIM, online Mendelian inheritance in man; ACMG, American college of medical genetics and genomics GenCC, gene curation coalition; VUS, variant of unknown significance.

Export to PPT

Candidate gene prioritization

To address inconclusive or negative WES outcomes, a phenotype-driven variant prioritisation pipeline was employed as detailed in our previous work¹⁹. Briefly, variants that were absent or showed MAF≤0.001 in gnomAD (v3.1.2 and v4.0.0) were filtered. These variants were further filtered using ensemble variant scoring methods (CADD≥25; REVEL≥0.6; SpliceAI: ≥0.3) and variant classification was performed by following the latest ACMG variant classification guidelines. The potential candidate genes for foetal malformations were identified by using OMIM data, zebrafish ( https://zfin.org/ ), and knockout mouse model phenotype databases [Mouse Genome Informatics (MGI)], enabling cross-species comparison and validation. Furthermore, an additional phenotype-driven variant prioritisation algorithm, Exomiser (v13.2.1) was used, to prioritise variants in individual samples based on relevant Human Phenotype Ontology (HPO) terms²⁰. Lastly, the top five candidate genes in each sample were manually curated by evaluating phenotype evidence from available literature and multiple databases such as the Gene Curation Coalition (GenCC), DECIPHER, The Human - Mouse: Disease Connection (HMDC), GeneMatcher (genematcher.org), VarElect and MONDO. Only genes without GenCC evidence or genes classified as ‘supportive’, ‘moderate’ or ‘strong’ but not ‘definitive’ by GenCC for the respective foetal malformation phenotype (observed in this study) were considered as candidate genes. The exome analysis pipeline used in this study has been depicted in figure.

Aneuploidies and CNVs

Chromosomal analysis using karyotyping/FISH identified trisomies in 18.1 per cent (8 out of 44) of the cases (Supplementary Table II). This included trisomy 21 (Down syndrome) in four cases, trisomy 13 (Patau syndrome) in three cases, and one case of 47XXY mosaicism (15% mosaic). We performed CMA in cases with normal karyotype/FISH and identified CNVs in 13.6 per cent (n=6) cases, of which five showed pathogenic CNVs (Table I; Supplementary Table I, and Fig. 1). We observed a de novo CNV in the 17p11.2 region in two POCs that displayed distinct phenotypic presentations (Table I and supplementary Fig. 2). Notably, the foetus exhibiting structural heart malformations showed a 3.68 Mb copy gain in this region (Supplementary Fig. 2), contrasting with another foetus characterised by skeletal dysplasia, which showed a 3.45 Mb copy loss in the same 17p11.2 region (Supplementary Fig. 2). The 17p11.2 CNV observed in these two foetuses overlapped with the 17p11.2 microduplication syndrome region with two important dosage-sensitive genes, RAI1 and FLCN (Supplementary Fig. 2). This region is known to be associated with Smith-Magenis and Potocki-Lupski syndromes (Supplementary Fig. 2)²¹.

Variants from known genes for foetal malformations

WES analysis identified causative genetic variants in two out of 44 cases (4.5%). In foetuses AB0005 and AB0041, P/LP variants were identified in ASPM and SPECC1L genes, respectively. Reanalysis of the exome data led to the identification of VUS in two additional cases: AB0018 (CNOT1) and AB0023 (ANKRD11) (Table II). However, the two VUS variants identified in AB0018 and AB0023 need to be functionally evaluated to understand the genotype-phenotype correlation. Thus, a definitive diagnosis was achieved in two foetuses that showed P or LP variants with a matched mode of inheritance for the phenotype (Table II). Among the two cases with P or LP variants, foetus-AB0005, was aborted due to primary microcephaly and WES identified a homozygous, Loss-of-function (LoF) variant: rs199422194; ASPM(NM_018136.5):c.9697C>T; NP_060606.3:p.(Arg3233Ter) in the ASPM gene. In the second foetus (AB0041) with a definitive diagnosis, a novel, monoallelic, LP variant:SPECC1L(NM_015330.6):c.2836C>T; NP_056145.5:p.(Arg946Ter) was identified in the SPECC1L gene (Table II). In total, a conclusive genetic diagnosis was achieved in 36.3 per cent of cases (16 out of 44 cases) with various foetal malformations by using Karyotyping/FISH, CMA, and WES.

Candidate genes

We re-analysed the exomes using a phenotype-driven variant prioritization approach and a discovery analysis pipeline to identify candidate genes linked to observed FSAs. We report six potential genes: RUNX2 and PALLD (neural tube defects), KMT2D and KDM1A (structural brain abnormalities), and FBN2 and CPLANE1 (multi-system malformations). These genes were predicted to be disease-causing, with strong genotype-phenotype associations based on knockout zebrafish and mouse models (Table III).

Antenatal USG of foetus AB0002, from a non-consanguineous couple, showed a large lumbar spinal canal defect with overlying soft tissue swelling, extending almost all lumbar vertebral segments, suggesting spinal dysraphism. We identified missense, compound heterozygous variants in the RUNX2 gene, which are likely to be causative based on the phenotype-based prioritization algorithms (Exomiser combined score: 0.76, Exomiser phenotype score: 0.54, P=2.1; E-3). The RUNX2 knockout mice show abnormal vertebrae morphology (MGI phenotype; MP:0000137). In humans, RUNX2 is known to cause cleidocranial dysplasia with osseous manifestations²².

Foetus AB0004 presented Chiari II malformation (lumbosacral meningocele and obliteration of cisterna magna) at 17 wk and six days of gestation (Supplementary Table I). In this foetus, we report a homozygous, novel, splice-site variant; c.1087+1G>A:p.? in the PALLD gene (Table III). This variant was predicted to be LP and is absent in gnomAD (v4.0.0) and our Indian genome databases (IndiGenome). Our analysis predicted PALLD as a strong candidate for neural tube defects (Exomiser combined score: 0.97, Exomiser phenotype score: 0.75, P=4.4;E-4).

Foetus AB0009 presented with holoprosencephaly at 12 wk of gestation, and the mother had a history of medically terminated pregnancy due to foetal malformation. In this foetus, we identified a de novo, monoallelic, LoF (stop-gain) variant; c.2485G>T; p.E829X (LP) in the KMT2D gene. The candidacy of KMT2D is strongly supported by its role in cranial neural crest cell differentiation and forebrain development in mice. Knockout mice show abnormal dentate gyrus morphology (MP:0000812)^23,24. In humans, monoallelic germline variants in KMT2D were reported in Kabuki syndrome (OMIM: 147920). Interestingly, a recent study has reported monoallelic pathogenic variants in two unrelated individuals with holoprosencephaly²⁵.

Antenatal USG of foetus AB0012 at 20 wk and three days revealed kyphoscoliosis of the cervical spine, increased nuchal translucency, and structural heart defects. In this foetus, WES identified a de novo monoallelic missense LP variant, rs1263307630 (c.7276C>A; p.P2426T) in the FBN2 gene. In mice, Fbn2 gene disruption leads to multisystem abnormalities that include abnormal skeletal morphology (MGI phenotype- MP:0005508) and abnormal cardiovascular system morphology (MGI phenotype-MP:0000266)²⁶. In humans, monoallelic variants in FBN2 are known to cause congenital contractual arachnodactyly (CCA) (OMIM:121050)^27-29.

We identified a novel, monoallelic de novo LP variant:c.313A>T; p.K105X in the CPLANE1 gene in an aborted foetus (AB0013) presented with iniencephaly; meningocele and suspected Dandy-Walker syndrome features. CPLANE1 is a ciliopathy-associated gene that is known to cause Joubert syndrome (OMIM:614615), Oral-Facial-Digital (OFD) syndrome type VI (OMIM:277170) and Cerebellar atrophy (CA) in humans^30-33. Mice with ablated Cplane1 gene show structural brain abnormalities such as decreased brain size (MGI phenotype-MP:0000774) and abnormal cerebellum morphology (MGI phenotype-MP:0000849).

Foetus AB0021 and its twin foetus AB0020 showed different structural malformations. Unfortunately, the POC of foetus AB0020 was not available for genetic testing (Supplementary Table I). Foetus AB0021 presented with agenesis of corpus callosum up on USG at 19 wk of gestation. We identified a novel, monoallelic missense variant: c.3513C>T; p.A838V (LP), in the KDM1A gene. In humans, pathogenic variants in KDM1A were reported in individuals with developmental delay and distinctive facial features³⁴.