RESULTS
We found a missense homozygous ANTXR1 gene mutation in all three children affected with GAPO (NM_032208.2:c.572T>C, p.Ile191Thr). This change is novel and predicted to be a disease causing by mutation taster (0.99), PolyPhen-2 (1.0), and PROVEAN (-4.0). Another change at this position, c.572T>G, p.Ile572Ser is reported in the COSMIC database (COSV100362660).
Homozygosity mapping applied to WES data of the consanguineous family, detected one large (11Mb) homozygous region involving the TEKgene (i.e., chromosome 9) in only one of the two sisters with congenital glaucoma, among the genes known for glaucoma. However, neither CNV nor intronic splice region variant was detected in the TEK gene in either of the sisters.
No pathogenic variants or CNV were detected that co-segregated with glaucoma in the two girls with congenital glaucoma (Case 1 and 2) at any of the known loci for Mendelian forms of glaucoma (GLC1A, GLC1B, GLC1D-Q) or PCG (GLC3A-D) including the known causative genes namely,MYOC (GLC1A), CYP1B1 (GLC3A), WDR36 (GLC1G),ASB10 (GLC1F), OPTN (GLC1E), NTF4 (GLC1O),TBK1 (GLC1P), LTBP2 (GLC3C), FOXC1, PITX2 ,TEK, and PAX6. Exonic variants specific to the two sisters with congenital glaucoma and the variants they inherited from either of the parents are listed in Suppl Table 1.
Four missense variants exclusive to the two sisters with GAPO and congenital glaucoma (case 1 and case 2); PTPN4 , PCDHB4, GGT2 and TGFBI, were observed. The variants on PTPN4 andPCDHB4 were rejected as they were found to be benign. TheGGT2 variant was not considered as it was earlier reported that hGGT2 does not encode a functional enzyme and therefore is unlikely to play a role in glaucoma pathogenesis. A missense heterozygous variant inTGFBI , NM_000358.3:c.968C>T NP_000349.1:p.Ala323Val (Fig 1.a) was found, which is reported to be damaging by the variant prediction tools PROVEAN (-2.5), FATHMM-MKL (0.825), PolyPhen-2 (0.89) and disease causing by Mutation Taster (0.99). The variant is highly conserved with a Genomic evolutionary rate profiling (GERP) value of 5.65. Sanger sequencing confirmed heterozygous missense p.Ala323Val mutation in both case 1 and case 2 and was absent in case 3 and other family members. The gnomAD frequency of the variant was found to be 0.001450. The Clinvarstar rating score, extracted for the variant, was found to be 1/4 with two submissions, one for corneal dystrophy.
We then looked for the presence of TGFBI gene variants in a cohort of 40 patients with JOAG (diagnosed before 25 years of age) in whom WES was previously done. We found two TGFBI variants present in two unrelated JOAG patients. Neither of these 2 patients had any other mutations in the known glaucoma genes. In one patient, aTGFBI heterozygous missense variant was found in the same codon observed in the described cases of GAPO, but with a different amino acid residue.; NM_000358.3:c.968C>A, NP_000349.1:p.Ala323Glu. The second was a novel variant in the other JOAG patient. This was a null variant (frameshift mutation); NM_000358.3:c.1944del, NP_000349.1:p.Ser649LeufsTer22. These variants were of uncertain significance with minor pathogenic evidence and likely pathogenic respectively, as per ACMG classification. The former was damaging by Mutation taster (0.99), PolyPhen-2 (0.89), and PROVEAN (-2.5). The gnomAD frequency of NM_000358.3:c.968C>A was found to be 0.00003269, and the frequency of (NM_000358.3):c.1944del is not reported in gnomAD. A third patient with granular corneal dystrophy referred to the glaucoma clinic was also found to have JOAG. This patient had familial granular dystrophy of the cornea with 2 other first degree relatives (a brother and his daughter) also affected with granular dystrophy but without having glaucoma or raised IOP (Fig 3 ). WES analysis of this patient revealed a heterozygousTGFBI mutation NM_000358.3:c.1663C>T, NP_000349.1:p.Arg555Trp which is reported to be pathogenic inClinvar with 3 submissions associated with granular corneal dystrophies (GCD). (Table 2 )
A comparative modeling of the TGFBI variants (p.Ala323Val, p.Ala323Glu and p.Ser649LeufsTer22) was performed since the available crystal structures lack the C-terminal region of the protein. The modeling results showed α-helical region in the C-terminal, as shown inFigure 4 . The model was subjected to 200ns MD simulations for refinement. During this simulation, the C-terminal region of the protein was gradually stabilized through interactions with the FAS1-4 and FAS1-3 domains. The last frame of this simulation was therefore used as a template for the generation of the variants A323E, A323V, and 649 frameshift, which were then subjected to MD simulations to understand the effect of mutations on the local and global structure ofTGFBI and its variants. Structural deviation, compactness of the proteins, and residue fluctuations for 200ns simulations were calculated to assess the root mean square deviation (RMSD), radius of gyration (RG), and root mean square fluctuation (RMSF), respectively (Supplementary Figure 1 ). MD simulations followed by Principal Component Analysis (PCA) were performed to study the overall motions of the TGFBI and these 3 variants. Projection of the trajectory along PC1 and PC2 depicted correlated motions of various domains of the protein, as shown in Figure 5 . We found that TGFBIfluctuated between an elongated and a banana-like shape due to bending between FAS1-2 and FAS1-3 domains. The bending movement was more prominent in the variants compared to wild-type (WT) TGFBI,majorly in the FAS1-2 and FAS1-3 domains. The impact of the three variants was also evident on the Free Energy Landscape (FEL) since the free energy well was shallow and distorted compared to the deeper well of WT TGFBI (Figure 6 ).