TBX6 missense variants expand the mutational spectrum in a non-Mendelian inheritance disease

Abstract

Congenital scoliosis (CS) is a birth defect with variable clinical and anatomical manifestations due to a spinal malformation. The genetic etiology underlying about 10% of CS cases in the Chinese population is compound inheritance by which the gene dosage is reduced below that of haploinsufficiency. In this genetic model, the trait manifests as a result of the combined effect of a rare variant and common pathogenic variant allele at a locus. From exome sequencing data of 523 patients in Asia and two patients in Texas, USA, we identified six TBX6 gene-disruptive variants from 11 unrelated CS patients via exome sequencing and in vitro functional testing. The in trans mild hypomorphic allele was identified in 10 out of the 11 subjects; as anticipated these 10 shared a similar spinal deformity of hemivertebrae. The remaining case has a homozygous variant in TBX6 (c.418C>T) and presents with a more severe spinal deformity phenotype. We found decreased transcriptional activity and abnormal cellular localization as molecular mechanisms for TBX6 missense loss-of-function alleles. Expanding the mutational spectrum of TBX6 pathogenic alleles enabled an increased molecular diagnostic detection rate, provided further evidence for the gene dosage-dependent genetic model underlying CS, and refined clinical classification.

INTRDUCTION

Congenital scoliosis (CS) is a curvature of the spine secondary to vertebral malformations arising during embryogenesis. It is estimated that the incidence of CS is 0.5-1/1000 livebirths (Giampietro et al., 2009), presenting either as an isolated malformation or as part of a syndromic condition (Al Kaissi, Ben Chehida, Ganger, Klaushofer, & Grill, 2015; Callier et al., 2008; Cornier et al., 2008). CS can range in severity from mild (i.e. single level vertebral malformation) to severe (i.e. multiple levels of vertebral malformations) spinal deformities (Campbell et al., 2003; Flynn et al., 2013), reflecting the biological complexity of somitogenesis and spine development. Consequently, defining the underlying molecular pathogenesis of the many forms of CS has been challenging.

Mutations in TBX6 encoding the T-box 6 transcription factor, a key regulator of the segmental patterning of the paraxial mesoderm, have been implicated in CS (Chapman & Papaioannou, 1998; P. H. White, Farkas, McFadden, & Chapman, 2003). We previously reported that a rare heterozygous loss-of-function (LoF) TBX6 allele (i.e. a deletion CNV [Copy Number Variation] or a stop gain/frameshift variant allele) in trans with a hypomorphic common variant TBX6 haplotype (T-C-A composed of three SNPs, rs2289292-rs3809624-rs3809627 (Liu et al., 2018; Wu et al., 2015)) could explain about 10% of CS cases in the Chinese Han population (Wu et al., 2015). We use the term ‘TBX6 compound inheritance model’ to refer to this pattern of TBX6 biallelic variation. The proposed TBX6 compound inheritance model was further confirmed in independent worldwide patient populations (Liu, Wu, Yang, & Takeda; Takeda et al., 2017). We provided additional evidence for the gene dosage hypothesis and the requirement for biallelic variation at Tbx6/TBX6 by engineering a mouse model to explore compound inheritance and the gene dosage model (N. Yang et al., 2018). CS patients affected with compound inheritance at the TBX6 locus present a distinctive pattern of clinical features: simple hemi-/wedge-shaped vertebrae in lower half of the spine, and was thus defined as TBX6-Associated Congenital Scoliosis (TACS) (Liu et al.; Wu et al., 2015). TACS is readily amenable to surgical correction by wedged resection.¹¹ Consistent with our observations in humans, a high penetrance of the CS phenotype was observed only in mice with the allelic combination (biallelic: LoF + hypomorph) modeling compound inheritance in the Tbx6 gene/locus edited mice. Moreover, these animals exhibit a strikingly similar phenotype to that observed in humans with TBX6 variants including the type and anatomical location of malformation, i.e. defects of vertebral column formation involving the lower part of the spine (N. Yang et al., 2018).

The TBX6/Tbx6 compound inheritance model was further supported by similar analyses from independent Japanese and European CS cohorts (Lefebvre et al., 2017; Takeda et al., 2017). These studies revealed that different TBX6 variant types are associated with vertebral malformations; including deletion CNV, frameshift and nonsense LoF alleles. In an analysis of 94 Japanese CS cases, the in trans risk haplotype was identified in all nine patients having TBX6 null mutations (Lefebvre et al., 2017; Takeda et al., 2017). In one French cohort, CS was associated with compound heterozygosity for two TBX6 rare missense variant alleles, as well as with proximal 16p11.2 microdeletion in trans with a TBX6 rare variant (Lefebvre et al., 2017). In addition, a Macedonian family has been described with CS characterized by multi-vertebral malformations and a TBX6 variant (c.1311A>T) that disrupted the natural stop codon, resulting in the addition of an 81 amino acid tail to the C-terminus (p.∗437Cext∗81) (Sparrow et al., 2013). As in other –COOH terminal extensions, the functional consequences i.e. whether the mutation conferred LoF versus GoF (Gain-of-Function), remains unclear. The nature of the in trans allele at the TBX6 locus was unknown (Coban-Akdemir et al., 2018; Inoue et al., 2007).

Interestingly, the vertebral malformations reported in association with TBX6 variant alleles vary in expression from the core TACS defining simple phenotype observed with the compound inheritance model. To what extent this reflects allelic heterogeneity or the selected combination of alleles at the TBX6 locus remains to be explored. Congenital scoliosis patients with TBX6 gene variants can present with a range of spinal deformity, i.e., from simple hemivertebrae/butterfly vertebrae involving the lower part of the spine, simple rib anomaly, and fewer vertebrae and intraspinal defects) (Liu et al.) to a more severe clinical phenotype, defined as spondylocostal dysostosis (SCD; MIM# 122600) that involves extensive segmentation defects with multiple abnormalities of the ribs (Kusumi & Dunwoodie, 2018; Sparrow et al., 2013; Sparrow, Sillence, Wouters, Turnpenny, & Dunwoodie, 2010). We reasoned that the heterogeneity of the TACS clinical phenotype could reflect the complexity and unclear mechanism of individual alleles, LoF or GoF, and the extent of gene dosage perturbations that might contribute to the compound inheritance gene dosage model. To address the allelic heterogeneity hypothesis, decipher the underlying genetic and functional mechanisms of TBX6 associated congenital scoliosis, and further establish the precise cellular phenotypes associated with TBX6 variant alleles based on the compound inheritance model for TBX6 null variants, we further explored the mutational spectrum and functional alterations of TBX6 missense variant alleles that were originally categorized initially as variants of unknown clinical significance (VUS). We therefore investigated missense variation in a large Chinese Han CS cohort (523 Chinese CS patients, 724 subjects totally), as well as one Hispanic patient with CS from Baylor Genetics clinical diagnostic laboratories (BG) and another Egyptian patient from Texas Scottish Rite Hospital for Children in Dallas (TSRH) with a clinical diagnosis of SCD.

MATERIALS AND METHODS

Editorial Policies and Ethical Considerations

Informed consent was obtained from each participant or the corresponding guardian for the DISCO study (Deciphering Disorders Involving Scoliosis and Comorbidities study); a waiver of consent was approved for the de-identified US subjects. The study was approved by the institutional review boards of Peking Union Medical College Hospital (PUMCH, Approval No. JS-908), Baylor College of Medicine (Protocol NO. IRB-42680), Baylor Genetics clinical diagnostic laboratories and the University of Texas Southwestern Medical Center/Texas Scottish Rite Hospital for Children (Protocol No. 122011-054).

Participants

We enrolled 523 patients clinically diagnosed with CS in PUMCH between October 2010 and November 2017 as part of the DISCO study. The CS diagnosis was confirmed with radiological imaging and patients with additional syndromic features were excluded. Detailed inclusive and exclusive criteria were previously described (Liu et al.; Wu et al., 2015). De-identified subjects carrying TBX6 rare missense variants from Baylor Genetics clinical diagnostic laboratories (BG) and Texas Scottish Rite Hospital for Children in Dallas (TSRH) were identified and then enrolled in this study. Informed consent was obtained from each participant or the corresponding guardian for the DISCO study; a waiver of consent was approved for the de-identified US subjects and the TBX6 variant participating subjects provided informed consent for participating including clinical information. Patient US-P2 was enrolled under protocol number 122011-054 as approved by the University of Texas Southwestern Medical Center Institutional Review Board. She was initially diagnosed with CS of unknown pathogenesis and was later noted to have mild headaches, decreased sensation and some muscle weakness. Genetic testing for Charcot-Marie-Tooth disorders was negative. She also later developed deep vein thrombosis. She was given a final diagnosis of SCD, unknown axonal neuropathy, and May-Thurner syndrome.

Exome sequencing and data analysis

Exome sequencing (ES) was performed on DNA extracted from peripheral blood from 417 sporadic singleton cases and 105 cases with family members, a total of 523 cases, 745 subjects. In brief, Illumina paired-end libraries were prepared from DNA samples and subjected to exome capture using four different capture kits in a chronological order (xGEN targeted capture kit (IDT), seqcap pure capture kit (NimbleGen), VCRome SeqCap EZ Chice HGSC 96 Reactions (Roche), and All Exon V6+UTR r2 core design (91 Mb, Agilent)), followed by sequencing on an Illumina HiSeq 4000 platform. Variants were called and filtered using a PUMCH developed pipeline as described (Wang et al., 2018). Rare variants (MAF < 0.001) were selected for analysis based on 1000 Genomes (October 2013), the Exome aggregation Consortium (ExAC; http://exac.broadinstitute.org), genome Aggregation Database (gnomAD, http://gnomad.broadinstitute.org/) and in-house databases of 2103 ES data from phenotypically well-characterized subjects. Besides, the accession Number of the TBX6 reference sequence is NM_004608.3. To evaluate the pathogenicity of the TBX6 missense variants, in silico prediction tools were adopted: Sorting Intolerant From Tolerant (SIFT; v1.03; https://sift.bii.a-star.edu.sg/), Combined Annotation-Dependent Depletion (CADD; v1.4; https://cadd.gs.washington.edu/) score, Genomic Evolutionary Rate Profiling (GERP++; May 2011; http://mendel.stanford.edu/SidowLab/downloads/gerp/index.html) score. The variant with a score below 0.05 is predicted as deleterious in SIFT.

Sanger sequencing

The DNA samples were extracted according to standard protocols (Supplemental Methods). All variants presented in this study were validated by Sanger sequencing. Variant-encoding amplicons were amplified using EX Taq™ kits (TaKaRa) from genomic DNA samples. Sequencing was performed on a 3730XL DNA Analyzer (Applied Biosystems).

Haplotyping

The genomic region of the TBX6 gene including rs2289292, rs3809624, and rs3809627 was PCR-amplified using primers in Supplemental Methods. ClonExpress One Step Cloning Kit (Vazyme, China) and the pGEM-T vector (Promega, USA) were applied. The PCR amplicons were cloned and insert DNA fragments were transformed to E. coli competent cells. The bacterial colonies were collected and verified by Sanger sequencing.

Plasmid construction

A full-length cDNA clone of human TBX6 (GenBank: NM 004608.3) was PCR-amplified using KOD Fx and primers 5’-CCAGTGCGGCCGCGATGTACCATCCACGAGAATTGTACCC −3’ (forward) and primers 5’-CCAGTAGATCTTCAGTACATGGGTTTGGAGCCC −3’ (reverse). The PCR amplicons were cloned into the NotI and BglII sites of the pFLAG-CMV-4 expression vector (Sigma–Aldrich, St. Louis, MO). Inverse PCR-based site-directed mutagenesis was performed using KOD -Plus- Neo Kits (TOYOBO). Mutated plasmid was used to transform E. coli clones and validated by Sanger sequencing. The details are provided in Supplementary Data.

Dual-luciferase reporter assay

A HeLa cell line was a gift from the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (Beijing, PR China). Cells were grown in medium-glucose DMEM (Life Technologies) containing 10% FBS (Life Technologies) and 1% penicillin-streptomycin solution (Life Technologies). Transfections were performed using Lipofectamine 3000 Transfection Reagent (Invitrogen) in 24-well trays (Costar). Cells were transfected with plasmids encoding wild-type or mutant TBX6 (100ng), luciferase reporter vector (firefly-pGL3-P2Ewt, 400ng), co-expression vector encoding Notch intracellular domain (NICD, 46ng) and pRLTK control vector (10ng). The transfection efficiency was examined by transfecting cells with eGFP-CMV-TBX6 plasmid. Cells were then harvested in 100 ml of Passive Lysis Buffer (Promega) 24 hours after transfection. Firefly and Renilla luciferase activities were assayed using the Dual-Luciferase reporter system (Promega) and measured on a LUX multimoder reader (Thermo-Fisher). Firefly luciferase counts were normalized against Renilla luciferase counts to account for differences in transfection efficiency. The experiments were performed three times for each assay and replicated three times.

SDS-PAGE and immunoblotting

The constructs were transfected into HeLa cells using Lipofectamine 3000 Transfection Reagent (Invitrogen) in 6-cell trays (Costar) and harvested 48 hours later. The transfected efficiency was examined by transfecting cells with eGFP-pFLAG-CMV-TBX6 plasmid. SDS-PAGE and immunoblotting analysis was performed on whole-cell extracts by standard methods, using rabbit-derived monoclonal DDDDK antibody (ab1162, Abcam) as the primary antibody and secondary horseradish peroxidase (HRP)-conjugated antibodies (ab7090, Abcam) were used to detect protein expression levels. These experiments were performed in triplicate.

Immunofluorescence

A full-length cDNA clone of human TBX6 (GenBank: NM 004608.3) was obtained and cloned into the NotI and BglII sites of the pFLAG-CMV and Myc-CMV expression vector (Sigma–Aldrich, St. Louis, MO). The HeLa cells were implanted on glass slides in 12-cell trays and transfected in Lipofectamine 3000 Transfection Reagent (Invitrogen) system for 24 hrs. Each unit was co-transfected with an equal amount of wild-type Myc-CMV-TBX6 vector and mutant pFLAG-CMV-TBX6 vector. The glass slides were rinsed three times in PBS buffer (Life Technologies) and fixed in methanol and acetone solution (1:1). The fixed cells were blocked with 1% BSA buffer (Invitrogen) and then incubated with PBS containing 0.1% Triton™ X-100 (Sigma-Aldrich) and primary antibodies for 1 hour at room temperature. The primary antibodies include rabbit monoclonal DDDDK antibody (ab1162, Abcam) and mouse monoclonal anti-Myc tag antibody (ab32, Abcam) targeting pFLAG tag and Myc tag, respectively. Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Cyanine 3 and Goat anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor Plus 488 were adopted as secondary antibodies (A10520, A32723, Invitrogen). Dyed slides were mounted in VECTASHIELD mounting medium with DAPI (Vector Laboratories, Peter Borough, United Kingdom) and observed using confocal microscopy. All the above experiments were replicated three times.

Statistical Analysis

The unpaired t-test was used for statistical analysis of luciferase reporter assays and immunoblotting. Statistical analyses were performed with SPSS version 15.0 (SPSS, USA). P<0.05 was considered as statistically significant.

RESULTS

Identification and in silico analysis of TBX6 variants of unknown significance (VUS)

Exome sequencing (ES) was performed on 523 Chinese CS cases from the DISCO (Deciphering Disorders Involving Scoliosis and COmorbidities, http://discostudy.org/) study. The compound inheritance model of TBX6 LoF variants combined with a T-C-A risk haplotype accounts for 43/523 patients (8.2%) (Liu et al., 2018; Wu et al., 2015). In the present study, we explored the role of different TBX6 variant types with potential functional effects distinct from predicted likely complete LoF of TBX6 protein function (eg. stop gain, frameshift) in the pathogenesis of CS. We identified ten rare TBX6 heterozygous nonsynonymous variants and one in-frame insertion variant in 17 patients (five unrelated patients carrying the identical missense allele, c.434C>T, p.Pro145Leu, and two patients carrying the same missense allele, c.356G>A, p.Arg119His). These variants were classified clinically as having unknown functional consequences and therefore as VUS in accordance with the American College of Medical Genetics and Genomics (ACMG) guidelines for classification of variant pathogenicity (Richards et al., 2015). The variants were confirmed by Sanger sequencing (Figure S1).

Among the 11 rare variants studied, 6 alleles were absent from 1000 Genomes, ExAC (Exome Aggregation Consortium), ESP6500, ARIC (Atherosclerosis Risk in Communities Study Description) and our in-house PUMC database consisting of 2103 exomes from individuals from China from DISCO study. Of the 6 novel missense variants, 3 variants (c.424G>T, p.Asp142Tyr; c.440A>T, p.Asp147Val; and c.929C>T, p.Pro310Leu) were predicted to be damaging/likely damaging (CADD = 21, 16.01, 20.6 respectively; SIFT prediction as deleterious for all; GERP = 6.04, 6.04, 5.36 respectively), and 2 variants (c.1184G>A, p.Gly395Asp; c.1300C>T, p.Pro345Ser) were predicted to be deleterious by SIFT, while CADD and conservation scores were more consistent with tolerated variants (CADD = 8.965, 11.06 respectively, GERP = −0.425, 4.69 respectively). The 6th variant was a novel in-frame insertion (c.473_475insGGG, p.Trp158_Glu159insGly). The remaining five variants had an observed frequency < 0.001 in reference databases. Four of the five missense variants (c.356G>A, p.Ard119His; c.434C>T, p.Pro145Leu; c.499C>T, p.Arg167Cys; c.700G>C, p.Gly234Arg) were predicted to be damaging/likely damaging (CADD = 34, 16.65, 20.9 and 19.27 respectively, SIFT predicts all as deleterious, GERP = 6.04, 5.008, 4.83, 5.95 respectively) while one (1112C>T, p.Pro371Leu) was predicted to be tolerated (CADD = 11.36, SIFT prediction as deleterious, GERP = 3.18) (Table 1 and Supp. Table S1).

Table 1.

Genotype and phenotype of patients diagnosed as TBX6- associated congenital scoliosis.

Cohort	Patient Number	Ethnicity	TBX6 variants		gnomAD_ALL	gnomAD_EAS	Gerp	SIF T	CADD	In vitro functional test		Phenotype
			Allele 1	Allele 2						TA assay	Mis- localization	Vertebral Malformation	Other Congenital Malformation
DISCO	XH026	Chinese Han	c.356G>A(p.Arg119His)	T-C-A	0.0000497	0.0000323	6.04	D	34	Decreased	+	T12 SH-L	12th right rib absent
	XH261	Chinese Han	c.356G>A(p.Arg119His)	T-C-A	0.0000497	0.0000323	6.04	D	34	Decreased	+	L1/L2 SH-R, L2 SW-D	None
	XH421	Chinese Han	c.424G>T(p.Asp142Tyr)	T-C-A	0	0	6.04	D	21	Decreased	−	T10 SH-L	10th right rib absent
	XH071	Chinese Han	c.434C>T(p.Pro145Leu)	T-C-A	0.0002	0.0013	5.008	D	16.65	Decreased	−	T5-T12 SD, FF (T5 SD, T6 NSH-L, T7 -T11 SD-B, T12 SD)	Thoracic deformity; extensive rib fusion and deformity; ventriculomegaly; cerebellar vermis anomaly
	XH246	Chinese Han	c.434C>T(p.Pro145Leu)	T-C-A	0.0002	0.0013	5.008	D	16.65	Decreased	−	T3-T SD, FF (T3 & T6 & T8 NSB, T6-T7 NSH-L)	Thoracic deformity; 4th & 6th right rib absent
	XH329	Chinese Han	c.434C>T(p.Pro145Leu)	T-C-A	0.0002	0.0013	5.008	D	16.65	Decreased	−	T5-T7 SD, FF	Thoracic deformity
	XH171	Chinese Han	c.434C>T(p.Pro145Leu)	T-C-A	0.0002	0.0013	5.008	D	16.65	Decreased	−	T6-T12 SD, FF (T7 & T9 & T10 hemivertebrae)	Thoracic deformity; extensive rib fusion and deformity; diastematomyelia
	XH070	Chinese Han	c.434C>T(p.Pro145Leu)	T-C-A	0.0002	0.0013	5.008	D	16.65	Decreased	−	C6-T4 SD, FF; T8 NSH-L; T11 NSH-R	Thoracic deformity; bilateral 13th ribs; atrial septal bulge; congenital hernia
	XH635	Chinese Han	c.473_475dupGGG(p.Trp158_Glu159insGly)	T-C-A	0	0		.	.	Decreased	+	L2-L3, SH-R	None
BG	BG-10	Hispanic	c.1133G>A(Arg378His)	T-C-A	0.000048	0	2.23	T	9.618	Decreased	−	T8-T10 SD, FF	NA
TSRH	TRSH-P1	Egyptian	c.418C>T(p.Leu140Phe)	c.418C>T(p.Leu140Phe)	0	0	6.04	D	20.5	Decreased	+	Multiple thoracic SD, FF	Thoracic deformity

Abbreviation: DISCO, Deciphering Disorders Involving Scoliosis and COmorbidities; BG: Baylor Genetics Laboratories; TRSH, Texas Scottish Rite Hospital; CS, congenital scoliosis; TACS, TBX6-associated CS; SD, segmentation defects; FF: formation failure of vertebrae; SH: segmented hemivertebra; SW: segmented wedge-shaped vertebra; NSH: unsegmented hemivertebra; NSB: unsegmented butterlfy-vertebra; NSW: unsegmented wedge-shaped vertebra; L: vertebra on the left; R: vertebra on the right; D: dorsal; C, cervical vertebra; T, thoracic vertebra; L, lumbar vertebra; TA: transcriptional activity; NS: non-significant; NA: unavailable.

To further explore the functional effects of these variants, and experimental functional evidence for potential pathogenicity of TBX6 VUS in different populations, we investigated two de-identified subjects from a US cohort who were originally ascertained from the Baylor Genetics (BG) clinical diagnostic laboratories and the Texas Scottish Rite Hospital for Children in Dallas (TSRH) with indications of vertebral malformations. The former patient, who is of Hispanic decent, presented with CS; while the other one, who is of Egyptian decent, presented with SCD clinically. The heterozygous variant (c.1133G>A, p.Arg378His) identified in the Hispanic patient was defined as rare by a minor allele frequency < 0.001 in reference databases, and was predicted to be tolerated (CADD =9.618, SIFT prediction as tolerated, GERP = 2.23). The novel variant (c.418C>T, p.Leu140Phe) identified in the Egyptian patient was predicted to be damaging/likely damaging (CADD = 20.5, SIFT prediction as deleterious, GERP = 6.04) and was interpreted to be in the homozygous state and not an ‘apparent homozygous allele/haplotype’ versus a hemizygous state due to 16p11.2 deletion (Liu et al., 2018) (Table 1 and Supp. Table S1).

Identification of the hypomorphic risk haplotype

Given the compound inheritance pattern of biallelic variation previously associated with TACS (Liu et al.; Takeda et al., 2017; Wu et al., 2015), we examined the haplotype of the in trans allele in the 17 cases carrying TBX6 variants of unknown functional consequences. A long-range PCR followed by sequencing the sub-cloned amplicons was performed. Consistent with our reported compound inheritance model, we identified the hypomorphic haplotype ‘T-C-A’ (the co-occurrence of three common SNPs, rs2289292-rs3809624-rs3809627) in trans with the identified TBX6 VUS (c.356G>A, c.424G>T, c.434C>T, c.473_475dupGGG, c.929C>T, c.1112C>T and c.1184G>A) in 12 (5 unrelated patients carrying the identical missense allele, c.434C>T, and 2 patients carrying the same missense allele, c.356G>A)/17 cases in the DISCO cohort. The remaining 5 cases (c.440A>T, c.700G>C, c.1300C>T, and 2 unrelated patients carrying the identical missense allele, c.499C>T) were instead found to have the haplotype ‘C-T-C’ in trans; a similar common haplotype in the Chinese population found at a frequency of 43.1% versus the frequency of 44.4% for the ‘T-C-A haplotype (Table 1 & Supp. Table S1). The heterozygous variant (c.1133G>A) from the case of Hispanic descent was also identified in trans with the hypomorphic haplotype ‘T-C-A’, and the Egyptian case carried a homozygous haplotype ‘C-T-A’ (Table 1).

Functional analysis of TBX6 variants

We investigated whether these TBX6 variants (all the 17 Chinese patients and 2 patients from the US cohort) could affect TBX6 transcriptional activity (TA) in vitro. We established a wild type pFLAG-CMV-TBX6 vector using the cDNA copy of human TBX6. The variant pFLAG-CMV-TBX6 vectors were generated via targeted mutagenesis. Considering that TBX6 directly binds to MESP2 upstream to mediate the somitogenesis process (Yasuhiko et al., 2006), we adopted a previously reported transfection system to assess transcriptional activity in each of the eleven mutant plasmids. MESP2 ORF sequence was subcloned into the pGL3-P2Ewt (Promega) vector as a luciferase reporter construct. A co-expression vector encoding the Notch intracellular domain (NICD) together with control expression vector pRLTK (Promega) were co-transfected in human HeLa cell lines (Sparrow et al., 2013; Yasuhiko et al., 2006), followed by dual-luciferase activity assessment. TBX6 variants from all 17 Chinese patients and 2 US patients along with a positive control nonsense variant (c.699G>A) were tested by in vitro functional analysis, among them, six variants (c.356G>A, recurrent in 2 unrelated patients; c.418C>T; c.424G>T; c.434C>T recurrent in 5 unrelated patients; c.473_475dupGGG; and c.1133G>A) from 11 patients were identified to have significantly decreased relative transcriptional activity compared to the wild type construct (P<0.001, Figure 2). We next examined TBX6 protein expression level in the same cell lines. HeLa cells were transfected with wild-type or mutated vectors and were harvested after 48 hours culture. SDS-PAGE and immunoblotting revealed no significant difference between mutant and wild type proteins (Figure S2), thus documenting that the variants do not appear to perturb either protein translation or protein stability. Of note, five of the six TA-disrupting variants were located at the T-box domain (89-272 amino acids, Figure 1A and 1B) that contains a binding motif for TBX6 degradation by downstream RIPPLY2 (Zhao et al., 2018). Notably, molecular modeling showed two of the functionally deleterious variants (p.R119H and p.W158_E159insG) were located close to the DNA-binding region (Figure 1C). Thus, these variants might have pathogenic consequences by disrupting downstream transcriptional processes mediated by RIPPLY2 or by otherwise abrogating DNA binding affinity.

Figure 2: Relatively transcription activity of mutated TBX6.

A previously reported nonsense variant c.699G>A (p.Trp233Ter) was used as a positive control, whose transcriptional activities (TA) were shown to be significantly decreased compared to the wild-type control. *** P < 0.001; ** P < 0.01. The decreased transcriptional activity was observed in c.356G>A, c.424G>T, c.434C>T, c.473_475dupGGG TBX6 plasmid transfection from DISCO cohort, as well as the c.1133G>A from US-P1 (black bars). These variants were all found in trans with the T-C-A haplotype in patients with hemi-/wedge-shaped vertebrae. The remaining variants in our cohort were identified statistically nonsignificant less transcriptional activity (gray bars). The variant (c.418C>T) from US-P2 also showed statistically significant decreased transcriptional activity (black bar), which was found in homozygous state in a patient clinically diagnosed as SCD (multiple levels of vertebral malformations).

Figure 1: The schematic view of six TBX6 gene-disruptive variants.

(A) Schematic view of exons 1-9 and locations of variants (orange stars). (B) Overview of TBX6 protein with the DNA binding region (T-box domain, orange bar). The TBX6 variants are shown above the protein with black balls and counting. Four missense variants (p.R119H, p.L140F, p.D142Y, p.P145L) and the in-frame insertion variant (p.W158_E159insG) are enriched in the T-box region domain, one variant (p.R378) was not located in the T-box region domain. (C) Space distribution of the 5 gene-disruptive variants in the T-box domain in TBX6 protein. The model of T-box domain from TBX6 (89-272 amino acids) in complex with DNA was built by Swiss-Model (https://www.swissmodel.expasy.org/) using the crystal structure of TBX3 (PDB ID code: 1H6F) as model and presented by PyMOL. Two of the identified variants (p.R119H and p.W158_E159insG, red) locate close to the DNA-binding region.

We next used immunofluorescence to profile the intracellular distribution of tagged TBX6 protein. Myc-tagged wild type TBX6 vector was constructed as a control reporter to trace the distribution of TBX6 protein encoded by the in trans allele. Co-transfection of a control vector and a pFLAG-tagged mutant TBX6 vector was performed in a HeLa cell line to mimic the heterozygous state. Under these conditions, abnormal distribution of mutant TBX6 protein was observed in 3 variants (c.356G>A, c.418C>T and c.473_475dupGGG), in which wild type Myc-TBX6 (green) accumulated in the nucleus as normal while mutant pFLAG-TBX6 (red) was distributed in plasma as extra-nuclear particles (Figure 3), suggesting the variants may cause cellular mis-localization of the encoded protein and consequent reduction of transcriptional activity. The 8 cases with 7 variants (c.440A>T; c.499C>T, recurrent in 2 unrelated patients; c.700G>C; c.929C>T; c.1112C>T; c.1184G>A; and C.1300C>T) that did not disrupt transcriptional activity or show evidence for mis-localization were excluded from further study (Supp. Table S1).

Figure 3: Intracellular distributions of mutant TBX6 protein.

The cell nucleus could be visualized by DAPI (blue). The wild-type TBX6 protein (Myc-tagged, green) located inside the cell nucleus, while the mutated TBX6 (FLAG-tagged, red) (c.356G>A, c.418C>T and c.473_475dupGGG) showed obviously mis-localized at the extra-nuclear area (white arrows).

Improved molecular diagnostic rate in patients with TBX6 VUS

By genetic and in vitro functional assays, from 10 patients we identified 4 nonsynonymous variants (c.356G>A, recurrent in 2 unrelated patients; c.424G>T; c.434C>T recurrent in 5 unrelated patients; and c.1133G>A) and 1 in-frame indel variant (c.473_475dupGGG) all with deleterious functions consistent with LoF alleles and in trans with a T-C-A haplotype at the TBX6 locus. In addition, one patient had a disruptive homozygous missense variant (c.418C>T). Each of these six variants therefore meet the criteria now defined as “likely pathogenic variants” according to ACMG guidelines (Richards et al., 2015). Based on the compound inheritance model, these 11 patients, 9 from the Chinese DISCO study and 2 from US ascertained subjects of Hispanic and Egyptian descent, could potentially be molecularly diagnosed as having TBX6-associated congenital scoliosis with one LoF allele in trans with the hypomorphic haplotype or biallelic pathogenic variation (Table 1). Therefore, the molecular diagnosis of compound inheritance in the 9 Chinese Han patients contributed an additional 1.7% TACS cases (9/523), raising the overall diagnostic rate of TACS to 11% in this cohort (Liu et al.).

Functional alleles contributing to the compound inheritance model and TACS

Among the 11 molecularly diagnosed patients, 10 patients (XH026, XH070, XH071, XH171, XH246, XH261, XH329, XH421 XH635, and US-P1, Table 1) were identified to have a significant disrupted TBX6 allele through functional experiments and a hypomorphic risk haplotype in trans, providing further evidence for the compound inheritance model (Liu et al.; Takeda et al., 2017; Wu et al., 2015). Detailed phenotypic assessment demonstrated that all of the 10 cases presented with a spinal malformation phenotype similar to those observed in patients with the compound inheritance model at the TBX6 locus (Wu et al., 2015), i.e. a constellation of simple hemivertebrae/butterfly vertebrae involving the lower part of the spine, simple rib anomaly, and fewer vertebrae and intraspinal defects (Figure 4). These patients therefore received a molecular diagnosis defined as TACS. Notably, the patient carrying a bi-allelic TBX6 (i.e. homozygous) functionally disrupted pathogenic variant (c.418C>T) allele, with both decreased transcriptional activity and abnormal protein localization, presents a distinct and clinically more severe phenotype compared to the TACS group. The patient had more extensive segmentation defects with multiple abnormalities of the ribs, multiple levels of vertebral malformations.

Figure 4: Spinal radiographs of TACS patients.

Ten patients (XH026, XH070, XH071, XH171, XH246, XH261, XH329, XH421, XH635, US-P1) present as TACS (simple hemivertebrae/butterfly vertebrae involving the lower part of the spine, simple rib anomaly, and fewer vertebrae and intraspinal defects), the similar phenotype as the 16p11.2 del patient (XH522) we previously reported (Liu et al.). The vertebral malformations were indicated in the white arrows. The remaining one patient (US-P2) present a much more severe phenotype compared with TACS (a more severe one, extensive segmentation defects with multiple abnormalities of the ribs manifesting as CS clinically, SCD-like). The extensive vertebral malformations were indicated by the white bracket.

DISCUSSION

Here we present genetic, clinical and functional data interrogating the role of missense or in-frame indel variant alleles of TBX6 in the pathogenesis of CS. We also established a workflow to study the functional effects of rare variant alleles contributing to CS and integrating in vitro assays to further explore the VUS and potentially implicate them as likely-pathogenic variants (Figure S4). By providing experimental evidence for the compound inheritance model and the molecular pathogenicity of these variants, we now introduce a more refined dosage-dependent model to further describe the phenotypic spectrum of TACS (Figure 5). We speculate that the subtle differences in gene expression/dosage sensitivity of this locus might reflect developmental gene expression gradients required during somitogenesis for proper organismal development.

Figure 5: Gene dosage model of TBX6-Associated Congenital Scoliosis (TACS).

In our putative model, the phenotypes in TACS patients vary along with the dosage alterations of TBX6 expression. Slight loss of TBX6 expression caused by heterozygous hypomorphic haplotype or bi-allelic hypomorphic haplotype could be regarded as tolerable mutational dosage (green part) and would not lead to any spinal phenotype. One heterozygous severe hypomorphic (gene-disruptive) or null allele would still be tolerable (yellow part). Both of severe hypomorphic or null allele in trans with a mild hypomorphic haplotype could reduce the TBX6 expression level to threshold of a high CS penetrance (upper half part of the red part), also known as compound inheritance model, in which TACS is the most common phenotype. Bi-allelic series consist of two severe hypomorphic alleles or a null allele in trans with a severe hypomorphic allele, which would lead to an extreme loss of expression of TBX6 (lower half of the red part) contributing to a much more severe phenotype. Bi-allelic null variants would be embryonic lethal.

We sought to further explore the TBX6 compound inheritance model and investigate the locus-derived gene dosage hypothesis by investigating TBX6 variant alleles for which the functional consequences were uncertain. These nonsynonymous, mostly missense, variant alleles were identified by ES in subjects with CS. They were found as heterozygous TBX6 alleles with the exception of one homozygous allele. Ten unrelated patients (nine from a Chinese Han population and one Hispanic) with rare heterozygous TBX6 nonsynonymous variants or an in-frame indel variant alleles were shown to have the well-established hypomorphic haplotype in trans, consistent with the compound inheritance model previously confirmed in independent worldwide cohorts (Lefebvre et al., 2017; Liu et al.; Takeda et al., 2017; Wu et al., 2015), and were clinically defined as TACS (simple hemivertebrae/butterfly vertebrae involving the lower part of the spine, simple rib anomaly, and fewer vertebrae and intraspinal defects) (Liu et al.). One subject of Egyptian descent was found to have a rare, homozygous likely pathogenic allele in TBX6, without the hypomorphic haplotype. While this observation was consistent with a CS caused by TBX6 pathogenic variation, the phenotype is much more severe than TACS arising from compound inheritance of TBX6 alleles. In this scenario we posit that two copies of the TBX6 LoF allele versus 1 in the compound inheritance model significantly impairs TBX6 activity and results in a clinically more severe scoliosis and vertebral malformation defects, i.e. extensive segmentation defects with multiple abnormalities of the ribs manifesting as a spondylocostal dysostosis, or SCD-like.

A parallel independent study was conducted exploring missense alleles identified in patients of Japanese origin. Three heterozygous missense variants (c.400G>A, p.Glu134Lys, c.1129G>A, p.Gly377Arg, and c.1189G>A, p.Gly397Arg) in trans with a hypomorphic haplotype were identified in 3 Japanese CS patients. In addition, bi-allelic compound heterozygous missense variants, c.356G>A, p.Arg119His, and c.449G>A, p.Arg150His were found in a Japanese SCD patient. One variant (c.400G>A) showed a decreased transcriptional activity compared to wild type. The remaining four variants all showed a loss of function by mis-localization of mutant TBX6 proteins. That study suggests that bi-allelic loss of function mutations of TBX6 would cause a spectrum of phenotype including CS and SCD according to the severity of the loss of TBX6 function during somitogenesis (Nao Otomo, et al. manuscript in submission).

Through identification of different combinations of rare, functional variants and hypomorphic alleles, we have further refined a locus-dependent dosage model to explain TBX6-mediated molecular mechanisms causing CS (Figure 5) (Liu et al.; Wu et al., 2015; N. Yang et al., 2018). In this model, a dosage of 50% is sufficient for the normal development of the embryo, i.e., a wild type allele in trans with a null/severe hypomorphic allele/mild hypomorphic haplotype or a bi-allelic mild hypomorphic haplotype would not lead to a CS phenotype in human or mice. The pathogenic threshold is a dosage loss of more than 50%, i.e., a null/severe hypomorphic variant in trans with a mild hypomorphic haplotype, which would cause an insufficient dosage of TBX6 expression leading to the TACS phenotype that includes simple defects in vertebral formation (hemivertebrae) and segmentation. In contrast, in the setting of bi-allelic variation involving a rare deleterious homozygous variant (s), severe hypomorphic variant, the greater relative loss of TBX6 function, leads to a more severe phenotype (SCD-like, multiple vertebral segmentation defects). Complete depletion of TBX6, i.e., bi-allelic null variants (i.e. amorphic alleles), is proposed to cause embryonic lethality as in mouse knockout studies(Chapman & Papaioannou, 1998; P. H. White et al., 2003) and CRISPR/Cas9 engineered mouse models (N. Yang et al., 2018).

Of note, similar dosage models have been proposed for CLCC1 in autosomal recessive retinitis pigmentosa (arRP) and PMP22 mutations that cause hereditary neuropathy with liability to pressure palsies (HNPP)-like neuropathy. For CLCC1, decreased endoplasmic reticulum membrane levels of CLCC1 might be sufficient for most tissues but insufficient in the developmentally higher-demanding tissues like the retina (Li et al., 2018). For myelin proper development and function is dependent upon PMP22 gene dosage alterations of which could lead to different neuropathy phenotypes (Shy et al., 2006). Gene dosage and expression models may be important to other birth defects and also to the penetrance of adult onset common complex traits such as age-related macular degeneration (AMD) versus childhood onset arRP or autosomal recessive stargardt macular dystrophy (Allikmets, Shroyer, et al., 1997; Allikmets, Singh, et al., 1997; Lewis et al., 1999; Shroyer et al., 1999; Shroyer, Lewis, Yatsenko, & Lupski, 2001; Yatsenko, Shroyer, Lewis, & Lupski, 2001).

As noted for many genotype-phenotype relationships, the precise location of variants within a protein can impact both the mechanism of pathogenesis and the resulting clinical phenotype. For example, missense variants in NAA10 encoding N-terminal acetyltransferase (NatA) complex lead to a spectrum of congenital anomalies, while truncating variants located in the last exon of the gene may escape from nonsense-mediated decay (NMD) resulting in a distinct phenotype (Poli et al., 2018; J. White et al., 2015). In our study, we found that five out of six TBX6 functional deleterious variants were enriched in the T-box domain. As a member of a phylogenetically conserved gene family, TBX6 depends on the crucial T-box domain to perform DNA-binding. TBX6 plays a key role during somitogenesis by regulating differentiation,(Chapman & Papaioannou, 1998) the specification of paraxial mesoderm from the axial stem cells (Martin & Kimelman, 2012; Takemoto et al., 2011), as well as defining the boundaries of each successively formed vertebrae through the interaction with its downstream target genes, such as MESP2 and RIPPLY2 (Takahashi et al., 2010; Zhao, Ajima, Ninomiya, & Saga, 2015). Variants in the T-box domain thus hypothetically may lead to alteration of transcriptional signaling, impacting downstream target genes (Takahashi et al., 2010; Zhao et al., 2015). Notably, MESP2 and RIPPLY2 have also been reported as causative genes for congenital vertebral malformations (Karaca et al., 2015; Makino et al., 2013). Additionally, a motif in the T-box domain was recently reported to be essential for Tbx6 protein degradation in vivo, suggesting an influence on protein stability from this region (Zhao et al., 2018). Intriguingly, the immunofluorescent protein localization assay provides insights potentially implicating another distinct pathogenic mechanism: a cytosolic accumulation of mutant TBX6 protein which blocks its access to nuclear DNA (Figure 3). However, we are unable to identify the signaling peptide critical for protein localization in vivo, which also requires further studies.

Considering the variant location, TBX6 gene function and related biological processes (Kusumi & Dunwoodie, 2018), we adopted in vitro assays including transcriptional activity assessment and immunofluorescence, respectively, to evaluate the disruptive effects of these potentially pathogenic variants. Haplotype analysis was also performed to identify the common hypomorphic haplotype in trans. By combining these functional analyses, experimental evidence was obtained for six TBX6 variants, previously interpreted as VUS, as likely-pathogenic due to deleterious effects on protein function. We note that each of these variants was in trans with a risk allele. These functional studies thus yielded an additional molecular diagnostic rate of compound inheritance of 1.7% and expand the mutational spectrum of TBX6-associated congenital scoliosis due to the compound inheritance and gene dosage model from 9.3% to 11% in the DISCO cohort (Liu et al.). Although genetic testing provides an increasing diagnostic yield with the recent findings of novel variants or genotype-phenotype association, most of the identified variants during clinical genetic practice remain as VUS (Y. Yang et al., 2013; Y. Yang et al., 2014). In this study, we provide a potential approach to the functional assessment of TBX6 VUS in the clinical-genetics diagnostic workflow (Figure S4).

There are several limitations to this study. Four VUS identified in the DISCO cohort have not been fully functionally characterized due to nonsignificant disrupted transcriptional activity or protein mis-localization. We did consider the possibility that rare variants in other genes associated with vertebral malformations may contribute to CS in this cohort. However, no candidate variants were identified in known CS candidate genes, including TBXT (MIM# 601397) (Ghebranious et al., 2008), WNT3A (MIM# 606359) (Ghebranious et al., 2007), DLL3 (MIM# 602768) (Bulman et al., 2000), MESP2 (MIM# 605195) (Oginuma, Niwa, Chapman, & Saga, 2008), HES7 (MIM# 608059) (Sparrow, Guillén-Navarro, Fatkin, & Dunwoodie, 2008). Three patients with TBX6 VUS alleles in trans with the T-C-A hypomorphic haplotype, also exhibit a typical TACS type I clinically. However, no gene-disruptive effects were observed in both transcriptional activity and protein mis-localization for these variants (Supp. Table S1). These observations indicate the possibility of limitations of the current functional assays and bioinformatic predictions and suggest the compound inheritance genetic model may be a powerful model to explain gene dosage dependent phenotypes beyond that which we can be computationally predicted or for which one is unable to experimentally illustrate functional effects. Considering the potential for multi-genic modifiers indicated by recent advances (Gonzaga-Jauregui et al., 2015), further study based on genome sequencing and enhanced variant detection (e.g. indels and exonic CNV rare variant alleles; noncoding common variant alleles) and more comprehensive genotype-phenotype correlation is necessary. For example, gene dosage modeling in mice may prove informative in future CS studies (N. Yang et al., 2018).

CONCLUSION

Our work suggests that nonsynonymous variants can contribute to the compound inheritance model at the TBX6 locus by contributing heterozygous LoF alleles. Consistent with the gene dosage model for compound inheritance, homozygosity for rare nonsynonymous variant in TBX6 (c.418C>T) results in a more a severe phenotype as the LoF, as evidenced by decreased transcriptional activity and mis-localization mechanism further perturbs dosage/expression than that with compound inheritance.