Patients

We sought to identify rare inherited and de novo genetic variants causing pediatric- and adult-onset PAH, and to compare the distribution and frequency of genetic mutations in the two age groups. Patients were recruited from pulmonary hypertension centers at Columbia University and Children’s Hospital Colorado. The patients were collected over a 22-year period, from 1993 to 2015, largely recruited from a single center at Columbia University. Patients were diagnosed according to the World Health Organization (WHO) pulmonary hypertension group I classification ¹⁷ and included primarily FPAH and IPAH cases. All patients included in the analyses were unrelated. Patients with PAH-CHD, which is much more prevalent among pediatric-onset cases, were excluded to reduce heterogeneity and are the focus of a separate report. The diagnosis of PAH was confirmed by medical record review including right heart catheterization. Pediatric-onset was defined by diagnosis prior to 19 years of age, although subgroup comparison of the rate of de novo mutations considered age-of-onset within the pediatric group (ie. ≤5 vs 5–18 years). Written informed consent (and assent when appropriate) was obtained under a protocol approved by the institutional review board at Columbia University Medical Center and Children’s Hospital Colorado.

FPAH cases were screened for BMPR2 and ACVRL1 mutations by Sanger sequencing and multiplex ligation-dependent probe amplification (MLPA). Then, FPAH cases without mutations in BMPR2 and ACVRL1, and all IPAH cases, were analyzed by exome sequencing (Supplemental Figure 1).

Whole exome sequencing (WES)

WES was performed as a family-based analysis when possible and included 48 pediatric trios (proband and two biological parents), 24 pediatric-/10 adult-onset duos (proband and one biological parent), 60 pediatric-/185 adult-onset singleton probands, 10 pediatric-/13 adult-onset families with multiple affected individuals for a total of 350 probands (Table 1). DNA was extracted from peripheral blood leukocytes and WES was performed in collaboration with the Regeneron Genetics Center (RGC) or the PAH Biobank at Cincinnati Children’s Hospital Medical Center using standard procedures (Supplemental Material).

WES Data analysis

The work-flow is outlined in Supplemental Figure 1. We used a previously established bioinformatics procedure to process and analyze WES data. Specifically, we used BWA-MEM (Burrows-Wheeler Aligner) ¹⁸ to map and align paired-end reads to the human reference genome (version GRCh37/hg19), Picard MarkDuplicates to identify and flag PCR duplicated reads, GATK ^{19, 20} HaplotypeCaller (version 3) to call genetic variants and GATK Variant Quality Score Recalibration (VQSR) to estimate accuracy of variant calls. To guard against technical artifacts, we excluded variants that met any of the following conditions: minimum read depth ≤8 reads, allele balance ≤20% ²¹, genotype quality <30, mappability <1 (based on 150 bp reads), genomicSuperDups segmental duplication similarity ≥95%, or VQSR >99.7. All candidate variants were confirmed with Sanger sequencing and tested for disease segregation when family DNA samples were available.

We used ANNOVAR ²² to annotate the variants and aggregate information of allele frequency and in silico predictions of deleteriousness. We used population allele frequencies from internal databases and public data from the Exome Aggregation Consortium (ExAC) ²³ and Genome Aggregation Database (gnomAD) to define rare variants with a population allele frequency ≤0.1%. We used multiple bioinformatics prediction algorithms including PolyPhen 2, Mutation Taster, SIFT, PROVEAN, metaSVM ²⁴ and Combined Annotation Dependent Depletion (CADD) ²⁵ to predict variant deleteriousness but ultimately focused on metaSVM (damaging) to define damaging missense variants (D-Mis) for enrichment analyses. We identified de novo variants as described previously, using trios composed of proband and unaffected biological parents, and manually inspected all candidate de novo variants using the Integrative Genomics Viewer (IGV) ²⁶ to reduce the false-positive rate. Finally, we inferred copy-number variants from WES data using the CLAMMS algorithm ²⁷.

Identification of pathogenic/likely pathogenic variants in established PAH risk genes and novel candidate risk genes

Variants identified in known PAH risk genes were classified as pathogenic, likely pathogenic, or of uncertain significance according to the American College of Medical Genetics and Genomics (ACMG) guidelines ²⁸. We compared deleterious rare or de novo variants with mutations reported in the literature and in genetic databases (Online Mendelian Inheritance in Man database, Human Genome Mutation Database ²⁹ and ClinVar ³⁰). Variants in established PAH genes of uncertain significance by ACMG guidelines and variants in novel candidate risk genes were classified based on predictions of deleteriousness. We defined three levels of deleterious variants: 1) “high,” likely-gene-disrupting (LGD) variants (including premature stopgain, frameshift indels, canonical splicing variants, and deletion of exons) and missense variants predicted to be damaging by MetaSVM and CADD (phred-scale) score ≥15; 2) “medium,” missense variants predicted to be damaging by MetaSVM or CADD score ≥15; and 3) “low,” all others.

Similar BMPR2 mutation frequencies in pediatric- and adult-onset PAH patients

The combined analysis of Sanger sequencing/MLPA/WES resulted in the identification of 84 rare, predicted deleterious BMPR2 mutations in FPAH/IPAH probands (Supplemental Tables 1 and 2). Altogether, there were 58 LGD variants, 25 D-Mis variants, and 1 in-frame deletion of seven amino acids. Fourteen of the 84 mutations are novel and all are depicted in Supplemental Figure 2. Similar BMPR2 mutation frequencies were observed in pediatric- and adult-onset FPAH (14/25, 56% pediatric and 43/79, 54.4% adult) and IPAH (13/130, 10% pediatric and 14/178, 7.9% adult) patients (Table 2). However, there was a difference in the distribution of LGD and D-Mis variant types between the two age groups: LGD variants occurred more frequently in patients with older age of onset (31.3 ± 14.0 years, mean ± SD) compared to D-Mis variants (19.3 ± 12.5 years) (P =0.0006, Mann–Whitney U test). The distribution of variant locations was similar with almost all of the D-Mis variants in both age groups located within the conserved activin or protein kinase domains and most of the LGD variants located in the first half of the protein that contains the two domains (Supplemental Figure 2).

Substantial contribution of TBX4 mutations to pediatric-onset PAH

We identified 13 likely pathogenic/predicted highly deleterious TBX4 variants in 13 probands (12 pediatric- and 1 adult-onset) (Supplemental Tables 1 and 2). The variants included 9 LGD and 3 D-Mis variants, including one intragenic deletion encompassing exons 3–9. All of these variants are novel and most reside within the conserved T-box domain (Figure 1). Based on the 12 probands with at least one parental sample available, we determined the inheritance pattern of a subset of variants: 1 de novo, 4 inherited from mothers without PAH, and 4 inherited from fathers without PAH. The observed inheritance of TBX4 variants from an unaffected parent in at least 8 of 13 cases is consistent with incomplete penetrance that is also observed for BMPR2 mutation carriers³⁶, suggesting that the effect of any single risk variant on development of disease is likely dependent on genetic background, somatic mutations, environmental effects or some combination thereof. Notably, pediatric-onset IPAH cases were significantly enriched for TBX4 mutations compared to adult-onset patients (10/130, 7.7% versus 0/178, p-value <0.0001) (Figure 2A). Although a much smaller sample size, there was a similar trend for FPAH cases (Figure 2A). Additionally, TBX4 carriers had a 20-year younger age-of-onset (7.9 ± 9.0, mean ± SD; range 0.6–33 years) compared to BMPR2 carriers (28.2 ± 15.4; range 1.5–72 years) (p <0.0001) (Figure 2B).

Open in a separate window

Figure 1

Novel and previously-reported genetic mutations in PAH risk gene TBX4

Novel mutations identified by WES are indicated above the protein schematic. Previously-reported mutations not identified in this study are indicated below the schematic. D-Mis, damaging missense mutations predicted by MetaSVM (red), LGD, likely gene damaging (blue) and In-frame insertion/deletions (yellow).

Open in a separate window

Figure 2

Role of TBX4 in pediatric-onset PAH

A, Enrichment of rare, predicted deleterious variants in TBX4, but not other known risk genes, in pediatric-onset cases. P-values were calculated by binomial tests. B, Younger age-of-disease onset for TBX4 variant carriers compared to BMPR2 variant carriers (P <0.0001, Mann-Whitney U test)

Rare genetic causes of PAH

We identified a small number of patients with rare deleterious mutations in recently identified genes for PAH, PAH associated with hereditary hemorrhagic telangiectasia (PAH-HHT), and genes involved in the BMP and TGFβ signaling pathway (Supplemental Tables 1 and 2). Two pediatric-onset patients were observed to carry KCNK3 mutations: a novel paternally-inherited two amino acid insertion (c.641_642insGCAGAC:p.214insQT) (IPAH) and a previously-reported heterozygous missense variant (c.G544A:p.E182K) ¹⁵ (FPAH). A novel heterozygous frameshift mutation in CAV1 (c.471delC; p.D157fs) was found in an adult patient with a pediatric-onset daughter who died at 9 years old. Compound heterozygous LGD variants in EIF2AK4 (c.C3766T:p.R1256X and c.1150dupG:p.S383fs) were identified in a familial case with PCH. A previously-reported homozygous nonsense mutation in EIF2AK4 (c.C1387T:p.R463X) ³⁷ was present in an adult FPAH patient whose brother had PAH and died at age 33 and was not available for testing.

In our total PAH cohort, 9 patients had a diagnosis of PAH-HHT. ACVRL1 and ENG are known risk genes for PAH-HHT and targeted Sanger sequencing/WES identified 3 pediatric- and 5 adult-onset patients with mutations in ACVRL1 as well as 1 adult patient with an ENG mutation (Supplemental Tables 1 and 2). Interestingly, one patient with isolated PAH without HHT carried a paternally-inherited missense variant in ACVRL1 (c.C1199T:p.A400V) as well as a paternally-inherited missense variant in BMPR2 (c.A1509C:p.E503D).

Finally, three candidate pathogenic variants were identified in genes in the BMP and TGFβ signaling pathway. A deleterious missense variant and a nonsense variant were identified in SMAD9 and a deleterious missense variant in BMPR1B (Supplemental Tables 1 and 2). In total, the rare genetic cause mutations had similar frequencies in pediatric and adult PAH cases: 3.8% (5/130) of IPAH cases and 16% (4/25) of FPAH pediatric-onset patients and 1.7% (3/178) of IPAH and 12.7% (10/79) of FPAH adult-onset cases (Table 2).

Excess of rare deleterious variants in known PAH risk genes due to BMPR2 and TBX4

We then evaluated the burden of rare variants in all 11 established PAH-associated risk genes from WES data. We used 1319 unaffected parents from a congenital heart disease cohort in the Pediatric Cardiac Genetics Consortium (PCGC) ³¹ as population controls for allele frequency estimates. To avoid spurious association due to population stratification, we inferred ethnicity of all subjects using principal components analysis ³³, and selected 88 pediatric-onset and 160 adult-onset PAH cases of European ancestry for this analysis. We observed a significant burden of rare LGD variants in both the pediatric- (enrichment rate=22, p-value=0.002) and adult-onset cases (enrichment rate=33, p-value=6.9e-7) in this established PAH gene set (Table 3). There was also a trend toward enrichment of rare, predicted deleterious missense variants in pediatric-onset cases (odds ratio=2.6, p-value=0.06), but not in adult-onset cases. The enrichment was not observed when BMPR2 and TBX4 were removed from the analysis (Supplemental Table 3).

BMPR2 and other genes in the TGFβ pathway in PAH

The TGFβ superfamily includes the TGFβ /BMP ligands, receptors, accessory proteins, activins, and downstream signaling mediators SMADs and NOTCH3 ³⁸. In this study, we identified rare, deleterious mutations in known PAH risk genes BMPR2, BMPR1B, ACVRL1, ENG, and SMAD9 for a total yield of 48 variants in pediatric-onset (20/25, 80% FPAH and 28/130, 21.5% IPAH) and 74 in adult-onset (53/79, 67.2% FPAH and 19/178, 10.7% IPAH) patients. Additionally, we identified rare, deleterious variants in other candidate genes in this pathway: BMPR1A, ACVR2B, GDF2, SMAD6, TGFB, and TGFBR2. BMPR2 mutations accounted for ~55% of FPAH cases in both children and adults. Although the contribution of BMPR2 mutations to PAH in this study was slightly lower than in previous reports, it is likely that patients with clinically-identified BMPR2 mutations chose not to enroll in this genetic study. Exploration of the novel candidate risk genes in additional pediatric and adult PAH patients, along with functional characterization of potentially pathogenic variants, will help further our understanding of this pathway and how its dysregulation contributes to the pathophysiology of PAH.

TBX4 and pediatric PAH

TBX4 is a transcription factor in the T-box gene family expressed in the atrium of the heart, the limbs, and the mesenchyme of the lung and trachea. TBX4, jointly with TBX5, has been shown to interact with FGF10 during lung growth and branching ³⁹. TBX4 was first suggested as a candidate PAH risk gene because of its localization within chromosome 17q microdeletions of patients presenting with severe developmental delays sometimes associated with pulmonary hypertension^{40, 41}. Then, Kerstjens-Frederikse and colleagues performed sequencing of TBX2 and TBX4, both located within the deletions, in PAH patients and identified 3 rare TBX4 mutations and 3 novel 17q22q23.2 microdeletions in 20 pediatric patients as well as one rare TBX4 mutation among 49 adults; no mutations in TBX2 were identified⁸. Heterozygous loss of function mutations in TBX4 have been reported to cause small patella syndrome (SPS; MIM #147891) ⁴² and five of the TBX4 variant carriers reported by Kerstjens-Frederikse were retrospectively found to have clinical features of SPS⁸. More recent studies have identified 3/40 TBX4 mutation carriers in a children’s PAH cohort⁴³ and 3/136 adult carriers in a Spanish PAH cohort⁴⁴. Our study is the largest to date, confirming a role for rare, predicted deleterious TBX4 variants in PAH, especially pediatric-onset disease. Of the 13 PAH patients carrying LGD or D-Mis TBX4 variants, only one was diagnosed with SPS; however, no radiologic records of knees, pelvis or feet were available for retrospective assessment of the others. The clinical phenotypes of the TBX4 mutation carriers ranged from mild/moderate PAH to persistent, progressive PAH unresponsive to vasodilators and fatal or requiring lung transplantation. Notably, we observed a significant enrichment of rare, predicted deleterious TBX4 variants among pediatric- compared to adult-onset patients. Further, TBX4 variant carriers had a 20-year earlier mean age-of-onset compared to BMPR2 carriers.

Gender bias of PAH presentation is modified by age of disease onset

In studies carried out primarily in adult cohorts, PAH has been shown to occur approximately 3-fold more frequently in females than males. However, a previous study of pediatric PAH reported that females represented only ~60% of patients ⁴. In our study of both pediatric- and adult-onset patients, the female gender bias was significantly lower in pediatric compared to adult patients (Table 1). Genetically, BMPR2, ACVRL1, and ENG mutations occurred more frequently in females, while TBX4 mutations, which exhibited enrichment in the pediatric-onset patients, occurred with equal frequency among both genders. Mutations in other known PAH genes occurred less frequently and effects of age and gender could not be determined.

We thank the families for their generous contribution. Robyn Barst and Jane Morse were critical members of the team to enroll and clinically characterize patients. Patricia Lanzano provided oversight of the biorepository at Columbia University. Philip Allen provided assistance with variant curation. Children’s Hospital Colorado patient samples were enrolled and sequenced as part of the National Biological Sample and Data Repository for PAH (PAH Biobank).

Sources of Funding: Funding support was provided by NHLBI HL060056 (WKC), HL105333 (DDI, WCN, MWP, KAL), HL114753 (DDI) as well as NIH/NCATS Colorado Clinical and Translational Science Award UL1 TR001082 (DDI), The Frederick and Margaret L. Weyerhaeuser Foundation (DDI) and The Jayden de Luca Foundation (DDI).