Recurrent gain of function mutation in calcium channel CACNA1H causes early-onset hypertension with primary aldosteronism

1The steroid hormone aldosterone is normally produced in the adrenal zona glomerulosa in response to either angiotensin II, which is produced in response to volume depletion, or hyperkalemia (high plasma K+ level). Both stimuli cause membrane depolarization, activating voltage-gated Ca2+ channels; increased intracellular Ca2+ provides the signal that triggers aldosterone production. In the setting of volume depletion, aldosterone signaling in renal and intestinal epithelia produces increased salt (re)absorption, promoting restoration of intravascular volume; in hyperkalemia, aldosterone promotes increased potassium secretion, restoring electrolyte balance.

2Pathological secretion of aldosterone in the absence of normal physiological stimuli leads to primary aldosteronism (PA), producing increased salt (re)absorption and hypertension. Hypokalemia is a frequently associated finding, resulting from increased renal K+ elimination. PA is found in 10% of patients referred for evaluation of hypertension. About half of these patients have adrenal aldosterone-producing adenomas (APAs). Germline mutations in three genes have been shown to cause rare Mendelian forms of early-onset PA. Gene fusions leading to constitutive expression of aldosterone synthase (encoded by CYP11B2), a rate-limiting enzyme in aldosterone biosynthesis, cause Glucocorticoid-Remediable Aldosteronism (GRA). Mutations in and near the selectively filter of the K+ channel encoded by KCNJ5 result in channels that conduct Na+, leading to adrenal glomerulosa cell depolarization and activation of Ca2+ channels, producing a Mendelian form of aldosteronism. Gain of function mutations in the calcium channel encoded by CACNA1D cause increased Ca2+ channel activity and another form of PA. These latter patients also have seizures, neurodevelopmental and neuromuscular abnormalities owing to gain of function effects of CACNA1D in the nervous system. Families with GRA often have many affected subjects and were identified by linkage analysis in extended families. Germline mutations in KCNJ5 are typically de novo or in small nuclear families; similarly, CACNA1D mutations to date are all de novo. Germline mutations in KCNJ5 and CACNA1D were found following identification of the same or related somatic mutations as drivers of APAs.

3The causes of PA in many patients remain undetermined. Although Mendelian inheritance has been suggested by recurrence of PA in some kindreds without mutations in known genes, traditional linkage analysis has failed to identify additional causative genes, likely due to a combination of factors including locus heterogeneity, high frequency of de novo mutations, reduced penetrance and/or variable expressivity. The advent of next-generation sequencing, allowing the search for recurrent mutations or greater burden of rare variants in individual genes than expected by chance, can permit identification of such loci in the absence of classical segregation patterns. Very rare phenotypes, such as childhood PA, are promising candidates for such traits.

4Using exome sequencing, we here identify five independent occurrences of the identical mutation in CACNA1H among 40 subjects with unexplained PA in childhood. CACNA1H encodes a voltage-gated calcium channel that is expressed in adrenal glomerulosa. Electrophysiology demonstrates that this variant causes reduced inactivation and a shift of activation to more hyperpolarized potentials, effects inferred to produce increased calcium influx and PA.

5From a cohort of more than 1500 unrelated subjects referred for evaluation of genetic forms of hypertension, we identified 40 subjects diagnosed with hypertension and PA by age 10 years in whom disease-causing mutations in CYP11B2, KCNJ5, and CACNA1D were excluded. Clinical details are shown in Supplementary file 1A. All subjects had hypertension with elevated aldosterone levels despite low plasma renin activity (PRA). None of the subjects studied were the offspring of consanguineous union. DNA from peripheral blood was subjected to exome capture and sequencing; mean coverage was 73 independent reads per targeted base (Supplementary file 1B). Variants were called as described in ‘Materials and methods’.

6We performed three analyses tailored to the expectation of a rare genetic disease (‘Materials and methods’). We sought previously unreported (absent in dbSNP, NHLBI, 1000Genomes and Yale exome databases) protein-altering variants that occurred in more than one subject (Supplementary file 1C); we performed gene burden analyses to search for previously unreported or rare (minor allele frequency [MAF] < 0.01%) heterozygous variants that collectively occurred in any gene more often than expected by chance (Supplementary file 1D); we searched for rare (MAF < 0.1%) homozygous and potential compound heterozygous variants that collectively occurred in any genes more often than expected by chance (Supplementary file 1E).

7There was only one result that surpassed genome-level significance: we found five apparently unrelated subjects with the identical previously unreported heterozygous A > G variant, resulting in a p.Met1549Val substitution in CACNA1H, which encodes the pore-forming alpha subunit of a T-type, low voltage-activated calcium channel (aka CaV3.2) (Figure 1, Table 1, Supplementary file 1F). This variant is absent among more than 129,000 alleles sequenced from diverse populations in the Exome Aggregation Consortium, and Yale databases. No other CACNA1H alleles with allele frequencies <0.01% were found among our cohort. Like other Ca2+ channel alpha subunits, CACNA1H contains four homologous repeats (I–IV), each with six transmembrane segments (S1–S6). The CACNA1HM1549V variant lies in the S6 segment of repeat III. Sanger sequencing in each case confirmed the heterozygous variant (Figure 1A).

8Three index cases were of European ancestry, one Hispanic, and one African American by self-report and principal component analysis (Figure 1—figure supplement 1). Members of the extended families were recruited, and sequencing of these subjects demonstrated that CACNA1HM1549V was a de novo mutation (absent in the biological parents) in both the index case of kindred 1347, and in the affected mother of the index case in kindred 1390 (Figure 1). Analysis of highly polymorphic markers confirmed paternity and maternity in both kindreds (Supplementary file 1G). This establishes independent occurrences of CACNA1HM1549V in these two kindreds. In the remaining three kindreds, the variant was transmitted to the index case from a parent, and samples from grandparents were not available for further analysis of transmission (Figure 1). Analysis of kinship coefficients using SNP genotypes of affected subjects from Illumina Human 1M-Quad beadchips and the KING algorithm provided no evidence that these three kindreds shared recent common ancestry (Supplementary file 1H, ‘Materials and methods’). Further, haplotypes flanking the CACNA1H mutation were phased using the BEAGLE program, revealing that the maximum shared haplotype flanking the CACNA1H mutation among these three kindreds was only 53.6 kb (87.0 kb for the two European kindreds, Figure 2). From this data, the maximum likelihood estimate of the number of generations since the last shared common ancestor among these subjects is estimated to be 714 generations 95% CI 290-1268. A more conservative analysis identifying homozygous discordant SNPs (eliminating inference of haplotypes by phasing) still limited the shared haplotype to less than 127.1 kb, consistent with the results of phasing using BEAGLE. These findings indicate that the CACNA1HM1549V mutation in these three kindreds has not been inherited from a recent common ancestor, and has either arisen independently or has been inherited from an extremely remote common ancestor. The latter possibility is extremely unlikely given the absence of this mutation in more than 129,000 alleles studied to date.

9The probability of finding any recurrent protein-altering de novo mutation among 40 kindreds is ∼4.0 × 10−6 (see ‘Materials and methods’). Even with a conservative estimate of the allele frequency of CACNA1HM1549V of 0.00001 in the general population (‘conservative’ because it has never been seen among >129,000 alleles in subjects not selected for early PA), the probability of finding three additional instances of this mutation in 38 unrelated subjects is ∼8.4 × 10−12. Combined, the probability of finding these five instances of the identical variant by chance is conservatively estimated to be 3.4 × 10−17, providing extremely strong statistical support for the role of this mutation in PA.

10If the CACNA1HM1549V mutation causes early-onset PA, CACNA1H (CaV3.2) should be expressed in human adrenal glomerulosa. CACNA1H transcripts have previously been reported in human kidney, liver, heart and brain, and our prior analysis of gene expression of human adrenal cortex showed that CACNA1H was the second most highly expressed calcium channel alpha subunit, after CACNA1D. We performed immunohistochemistry with two different antibodies specific for the encoded channel protein (CaV3.2), demonstrating strong staining of human adrenal glomerulosa; this staining was abolished after preincubation with immunogenic peptide (Figure 3). These results are consistent with prior in situ hybridization and electrophysiological studies of rodent and bovine glomerulosa as well as a recent study of human adrenal gland.

11The clinical features of the index cases harboring the CACNA1HM1549V variant were uniform. All presented with hypertension by age 10 and had persistent inappropriate elevation of serum aldosterone with suppressed PRA and high aldosterone:PRA ratio, indicative of autonomous adrenal aldosterone production (Table 1). Adrenal imaging by computed tomography, magnetic resonance or ultrasound showed no evidence of mass or hyperplasia at the time of presentation. There were no other recurrent or distinctive features of the index cases, specifically no history of seizures, neurologic or neuromuscular disorders as found in patients with CACNA1D mutations. Additional details are presented in Appendix 1.

12By direct Sanger sequencing, we identified five additional CACNA1HM1549V mutation carriers among family members, including four parents and one uncle of an index case (Figure 1). Of these five, three were diagnosed with early severe hypertension while two were not, and in fact were normotensive as adults (Appendix 1 and Supplementary file 1I). For example, subject 1390-2 was diagnosed with severe hypertension and PA at age 17; her hypertension was difficult to control, leading to unilateral adrenalectomy at age 29. Her hypertension nonetheless recurred, requiring reinstitution of treatment. Interestingly, the histology of her adrenal gland showed striking microscopic hyperplasia. While the normal adrenal glomerulosa comprises only a few cell layers and is about 70 μm in depth, the glomerulosa of subject 1390-2 was ∼30 cell layers and ∼300 μm in depth (Figure 4A–C). CACNA1H was expressed in the hyperplastic glomerulosa layer (Figure 4, Appendix 1). In this kindred, the CACNA1HM1549V variant arose concordantly and segregated precisely with PA and early hypertension (Table 2). Two mutation carriers (1368-2 and 1393-3) were normotensive as adults and had not been diagnosed as hypertensive in childhood; in subject 1393-3, PRA was at the lower limit of normal with normal aldosterone level, while PRA and aldosterone levels were normal in 1368-2 (Supplementary file 1I).

13To explore the specificity of this mutation for early-onset PA, we performed targeted Sanger sequencing for the CACNA1HM1549V variant in germline DNA of 1632 additional unrelated subjects, comprising 324 subjects with PA diagnosed after age 10 years, 96 with hypertension and bilateral adrenal hyperplasia, and 1212 referred for potential genetic causes of hypertension without evidence of PA. We also sequenced tumor DNA of 90 APAs, including 40 that did not have mutations in previously implicated genes (KCNJ5, CACNA1D, ATP1A1, ATP2B3, and CTNNB1). No additional CACNA1HM1549V mutations were identified, demonstrating striking specificity for early-onset PA.

14Members of the CaV3 family are activated by small depolarizing changes in the membrane potential (activation threshold ∼ −60 mV) and display very fast voltage-dependent inactivation. Methionine at the position corresponding to CACNA1HM1549 is conserved in the S6 helix of repeat three in all identified orthologs, including invertebrates. In addition, methionine occurs at the paralogous position in other calcium channels activated by small depolarizing potential changes (Figure 1C). Prior studies of CaV3.1 (CACNA1G) have shown that methionine 1549 lies in a methionine-phenylalanine-valine (MFV) tripeptide that regulates channel inactivation. Mutation of the homologous methionine in CaV3.1 to isoleucine or alanine results in delayed channel inactivation, and related calcium channels with isoleucine at the homologous position inactivate more slowly than those with methionine (Figure 1C).

15To assess the biophysical properties of CACNA1HM1549V, we heterologously expressed either CACNA1HWT or CACNA1HM1549V in HEK293T cells and performed whole-cell patch clamp recordings (Figure 5A). Upon depolarizing voltage steps from −90 mV, CACNA1HWT showed fast activation of calcium currents followed by rapid inactivation, consistent with prior studies. In contrast, CACNA1HM1549V exhibited marginally slower activation followed by a dramatically slowed inactivation. While CACNA1HWT is virtually fully inactivated by 400 ms, CACNA1HM1549V shows strong tail currents after returning to the holding potential of −90 mV (Figure 5A), demonstrating loss of normal inactivation, an effect still evident after sustained depolarization for 5 s (Figure 5B).

16We fitted mono-exponential equations to the decay phase of the calcium current between −50 mV and +30 mV. The determined time constants represent the mean time at which the current has decreased to 1/e of its initial amplitude; the results demonstrate ∼10-fold slower inactivation of CACNA1HM1549V compared to CACNA1HWT (p < 0.001 at all voltages studied, Figure 5C). In contrast, activation and recovery from inactivation were only marginally slower in mutant channels (Figure 5D, Figure 5—figure supplement 1).

17We also observed a significant shift of activation to less depolarizing potentials (Figure 6). CACNA1HWT showed half-maximal activation (V1/2 ) at −38.9 ± 1.1 mV (N = 11); in contrast, CACNA1HM1549V showed V1/2 of −44.2 ± 1.1 mV (N = 11, p = 0.003), resulting in a lower threshold for activation and increase in size of the ‘window current’, the area under the intersection of activation and inactivation curves where a fraction of channels are constitutively open. There was no significant effect on single channel conductance (Figure 6—figure supplement 1).

18Collectively, the changes in inactivation and voltage-dependence of activation cause Ca2+ influx at membrane potentials close to the resting potential and result in channels that remain open longer, allowing increased Ca2+ entry.

19These findings define a previously unrecognized form of PA resulting from a recurrent germline gain of function mutation in the Ca2+ channel encoded by CACNA1H. The extremely strong statistical evidence implicating this mutation, its clear gain of function effect, and the known role of Ca2+ signaling on aldosterone production and cell proliferation all strongly support this conclusion. The effects of this mutation phenocopy the adrenal effects of PA-causing mutations in KCNJ5 and CACNA1D, demonstrating a shared final common pathway by which PA results from increased Ca2+ entry via voltage-gated channels. These results allow a simple genetic test for this specific cause of severe hypertension and suggest that inhibition of mutant CACNA1H activity would ameliorate hypertension in patients with this mutation. While the CACNA1HM1549V mutation explains a significant fraction of the early PA cases, the causes of the remaining cases in our cohort are still unknown.

20There is striking genotype–phenotype correlation among patients with germline and somatic mutations in KCNJ5 and CACNA1H. Several recurrent germline mutations in KCNJ5 (e.g., p.Gly151Arg and p.Thr158Ala) support robust cell proliferation leading to massive adrenal hyperplasia identifiable on CT scan, leading to adrenalectomy. In contrast, another recurrent mutation in KCNJ5 (p.Gly151Glu) shows no or minimal hyperplasia discernable by adrenal imaging. While the former mutations are also found as somatic mutations in about 40% of APAs, the latter have not been found in more than 900 APAs. This phenotypic difference is likely accounted for by different effects on Na+ conductance- the germline mutations that are not associated with APAs or hyperplasia cause markedly greater Na+ conductance, resulting in very high cell lethality, preventing sustained increases in cell mass. Similarly, adrenal glands with CACNA1HM1549V show little or no hyperplasia by CT scan and neither this mutation nor other activating mutations in CACNA1H have been seen in APAs. We have not observed high cell lethality in HEK293T cells expressing CACNA1HM1549V. Germline mutations that cause massive hyperplasia and somatic mutations that cause APA formation likely require an optimal Ca2+ signal, one that is high enough to support proliferation but not so high as to cause cell lethality.

21While germline mutations in KCNJ5 and CACNA1D were discovered following the initial identification of their somatic mutations in APAs, the discovery of the recurrent CACNA1H mutation relied entirely on brute force sequencing of patients with early severe aldosteronism and hypertension. The occurrence of de novo mutations, the reduced penetrance in adults with the absence of large multiplex families, as well as the absence of distinctive phenotypes that distinguish these patients from others with early hypertension and aldosteronism, all suggest reasons that CACNA1H mutations were not previously linked to PA.

22The reduced penetrance in adults in particular is interesting—two mutation-carrier parents were normotensive as adults, without clear evidence of PA. Incomplete penetrance among some carriers of mutations that cause aldosteronism has been previously described. The explanations for these effects remain unclear, however age-dependent activity of the renin-angiotensin system and the ability of older individuals to modulate dietary salt intake in response to physiologic demand are potential contributors. This is well described in the case of heterozygous loss of function mutation in the receptor for aldosterone (the mineralocorticoid receptor, MR). These patients have life-threatening salt-wasting and volume depletion in the first years of life due to low signaling through MR, but are asymptomatic as adults. Adult subjects show increased dietary salt intake and increase MR signaling by induction of the renin-angiotensin system, thereby markedly increasing aldosterone levels. Other possible mechanisms for incomplete penetrance include genetic modifiers either in cis or in trans, including the possibility of somatic mosaicism resulting in absence of the gain of function mutation the adrenal gland. While such mosaicism cannot be excluded, Sanger sequence traces provided no suggestion of mosaicism in circulating white blood cell or saliva DNA (Figure 1A).

23CACNA1HM1549V shows constitutive activity at membrane potentials close to the resting potential, allowing channels to be activated despite suppression of the renin-angiotensin system and absence of hyperkalemia. CACNA1HM1549V channels also show strikingly delayed inactivation, a finding similar to mechanisms in several other channelopathies. In glomerulosa, delayed inactivation is inferred to increase the period of membrane potential depolarizations. Interestingly, recent studies in mouse have implicated CACNA1H activity in regular glomerulosa membrane potential oscillations that may amplify small changes in membrane potential to produce significant Ca2+ signals. Thus this regular activation of CACNA1H, together with a shift of activation to less depolarized potentials and prolonged activity, provides a mechanism for increased Ca2+ entry, leading to aldosteronism. While a common variant in CACNA1H has been suggested to be associated with blood pressure in a small genome-wide association study of African American individuals, this result did not pass criteria for genome-wide significance, was only found after exclusion of hypertensive individuals, and was not replicated in larger studies.

24The apparent limitation of the phenotype associated with CACNA1HM1549V to PA with hypertension despite the expression of CACNA1H in other organs including heart and brain is notable, and underscores the challenges in predicting human phenotypes from knowledge of underlying mutations. No mutation carrier had a history of seizures or cardiac arrhythmia. While some prior studies have suggested a role of rare gain of function mutations in CACNA1H in epilepsy, these studies have not approached genome-wide levels of significance, do not appear to confer high risk, and have not been uniformly replicated.

25Our findings are consistent with evidence supporting a normal role for CACNA1H in the regulation of human aldosterone biosynthesis. Because CACNA1H is activated by small depolarizing changes in glomerulosa membrane potential, it is likely activated in response to small day-to-day changes in serum K+ concentration and angiotensin II levels that require fine adjustments in aldosterone production to maintain volume and electrolyte balance. In contrast, CACNA1D, which is the most highly expressed calcium channel in adrenal cortex, and which shows larger single channel conductance than CACNA1H, is only activated by large depolarizations. Activation of this channel likely contributes to the high levels of aldosterone produced in response to marked volume depletion or hyperkalemia. We suggest that CACNA1H and CACNA1D act in series in the regulation of aldosterone, with CACNA1H being activated in response to small, frequent physiologic perturbations and CACNA1D in response to more infrequent large physiologic challenges.

26These findings also raise the question whether inhibition of wild type CACNA1H would lower blood pressure or aldosterone production. In the general population, loss-of-function variants in CACNA1H are very rare (cumulative frequency of splice site, frameshift and nonsense variants in the ExAC database of 0.06%, resulting in expected compound heterozygosity or homozygosity in about 1 in 2.6 million subjects), making such studies challenging. It seems plausible that loss of CACNA1H could be compensated by activation of the renin-angiotensin system, leading to greater glomerulosa cell depolarization with consequent activation of CACNA1D, maintaining normal aldosterone production and blood pressure. Consistent with this suggestion, blood pressure was reportedly unchanged in a CACNA1H knockout mouse model, although aldosterone levels were not reported. Similarly, selective inhibitors of CACNA1H inhibit aldosterone production in vitro, but do not apparently reduce aldosterone levels or blood pressure in vivo. Whether additional non-dihydropyridine compounds will prove to be more effective in lowering aldosterone levels or blood pressure will be interesting to assess.

27PA was diagnosed based on elevated ARR (>20 ng/dl:ng/ml/hr), typically with aldosterone >15 ng/dl, or marginally elevated values in the presence of unexplained hypokalemia. Venous blood or saliva samples were obtained from subjects with unexplained early-onset PA and family members. Research protocols were approved by the local institutional review board (IRB), and informed consent was obtained from all research participants.

28DNA was prepared from venous blood or saliva samples using standard procedures. Exome capture was performed using the 2.1M NimbleGen Exome reagent (Roche NimbleGen, Madison, WI), and 75 base paired end sequencing on the Illumina (San Diego, CA) platform was performed as previously described. Coverage statistics are provided in Supplementary file 1B.

29Direct bidirectional Sanger sequencing of CACNA1HP1523-R1584 from genomic DNA of indicated subjects was performed following PCR amplification using primers

30CACNA1H_25F (5′-GACCCACCGCCTCTGTG-3′) and CACNA1H_25R (5′-AGCGCCTTACTCCTGCG-3′).

31Parent-offspring trios were genotyped as previously described, except for locus D7S820 in kindred 1390 (primers [5′-ATGTTGGTCAGGCTGACTATG-3′] and [5′-GATTCCACATTTATCCTCATTGAC-3′]). Alleles without known frequencies in the population were omitted from the analysis.

32Normal human adrenal tissue was obtained from the Yale Pathology archive, and adrenal tissue from subject 1390-2 from Pathology Services of Beaufort/Charleston (South Carolina, USA). Immunohistochemistry was performed as previously described. Primary antibodies were α-CaV3.2 (#ACC-025, Alomone, Jerusalem, Israel) or T-type Ca++ CP α1H (SC-25691, Santa Cruz Biotechnology, Santa Cruz, CA), both at dilutions of 1:100; secondary antibody was donkey α-rabbit (#035-152, 1:500, Jackson, Bar Harbor, ME). For the Alomone antibody, preincubation with the antigenic peptide (1:1, wt/wt in 10% FBS) was performed for 1 hr at RT. Both antibodies were tested on two independent glands. H&E staining was performed at Yale Research Histology using routine procedures.

33Myc-DDK-tagged CACNA1H in pCMV6-Entry was obtained from Origene (Rockville, MD) (RC212772, NM_021098.2). Site-directed mutagenesis (QuikChange, Agilent Technologies, Santa Clara, CA) was performed to introduce the p.Met1549Val mutation according to the manufacturer's instruction. Each construct was validated by sequencing of the entire coding region.

34Culturing of HEK293T cells was performed as described. Cells were transfected with 3 µg of CACNA1HWT or CACNA1HM1549V expression plasmids. For each construct, two clones were functionally tested. Whole cell patch clamp recordings were performed on a HEKA EPC10 amplifier (HEKA Elektronik, Ludwigshafen, Germany) as described previously. The extracellular solution contained: 5 mM CaCl2, 125 mM TEA-Cl, 10 mM HEPES, 15 mM Mannitol, pH 7.4. Pipette solution contained: 100 mM CsCl, 5 mM TEA-Cl, 3.6 mM PCr-Na2, 10 mM EGTA, 5 mM Mg-ATP, 0.2 mM Na-GTP, 10 mM HEPES, pH 7.4 (titration with CsOH).

35Voltage dependences of activation were determined from the peak current–voltage relation and fit by a Boltzmann function as described. The fraction of non-inactivated channels was determined by dividing the peak amplitude at −20 mV before and after 5 s long pulses to voltages between −90 and −20 mV. Time courses of activation or inactivation were analyzed by fitting a mono-exponential function. The recovery from inactivation was measured using envelope protocols consisting of an inactivation of channels during a 5 s pulse to −20 mV followed by holding the membrane potential at −90 for increasing durations. Afterwards, peak currents at −20 mV were measured and divided by the previous peak current. A plot of these ratios vs the duration of the pulse to −90 mV was fit with a mono-exponential function to obtain time constants for the recovery from inactivation.

36Non-stationary noise analysis was performed as described using a voltage protocol that activates channels at −20 mV followed by the analysis of the decay of currents and variance at −90 mV. The initial variance at the holding potential of −90 mV before activation was regarded as background variance and subtracted from the recordings. The Lorentzian noise produced by channel opening and closing depends on the unitary current amplitude (i), the number of channels (N), and the absolute open probability (P):

37Since the macroscopic current amplitude is given bythe variance-current relationship results in a quadratic distribution:

38The single channel amplitude (i) was derived from the initial slope of a plot of the variance against the mean isochronal current results. Due to a low open probability (p < 0.5) at 5 mM of external Ca2+, the recorded data points only described a small part of the usual parabola and did not allow for determination of the number of channels and open probabilities.

39Data were analyzed in FitMaster (HEKA Elektronik), SigmaPlot (Jandel Scientific, San Rafael, CA) and Python. Statistical comparisons were performed using Student's t-test or Mann–Whitney rank sum test.

40Proteins encoded by orthologs or close paralogs of CACNA1H in vertebrate and invertebrate species were identified by a BLAST search. GenBank accessions included NP_066921.2 (Homo sapiens), O88427.3 (Mus musculus), XP_414830.4 (Gallus gallus), XP_002932520.2 (Silurana tropicalis), XP_002122425.1 (Ciona intestinalis) and NP_001024496.1 (Caenorhabditis elegans). Human α1 subunit paralogs were as previously described.

41Principal component analysis was performed as previously described. For analysis of close relatedness, genomic DNA from subjects 333-1, 1347-1, 1368-1, 1390-1, and 1393-1 was genotyped on Illumina Human 1M-Quad beadchips according to the manufacturer's instructions. Data were analyzed using a combination of GenomeStudio (Illumina) and PLINK v1.07 softwares. Mean call rate was 95.7%. Kinship coefficients were calculated by using the robust algorithm in KING 1.4. For 1393-1, 1368-1 and 333-1, PLINK format was converted to BEAGLE format using Mega2. Haplotypes flanking the CACNA1HM1549V mutation were phased by observed transmission in kindred 1393 and by maximum likelihood in kindreds 1368 and 333 using BEAGLE v.3.3.2 and a reference panel (phase 1 1000Genomes project). Only SNPs called in at least two samples were used for imputation, and only called SNPs were used for determination of the shared interval. Four additional heterozygous variants in close proximity to CACNA1HM1549V were identified from the 1393-1 exome. For 1368-1 and 333-1, the inferred haplotype producing the largest shared interval was chosen for further analysis. Mutation age was determined from haplotypes including flanking 41 markers using ESTIAGE. Recombination fractions were calculated from marker distances and average recombination rate across the interval (2.9 cM/Mb, deCODE). Shared allele frequencies were from EUR population (1000 Genomes project), and mutation rate was set to 2 × 10−8.

42For statistical analysis, a de novo mutation rate of 1.4 × 10−8 was assumed. The binomial probability of observing two or more de novo mutations at a specified position in a set of 41 cases (including one affected parent) was calculated and corrected for the target size of the human exome (24.75 Mb). The likelihood of observing three additional independent mutations at the identical position in 38 patients was calculated as a binomial probability from the assumed allele frequency.

43The mutation burden per gene in the cohort of patients with PA was compared to that in a control cohort comprising 724 unaffected parents of patients with congenital heart disease sequenced to similar depth of coverage on the same exome platform using Fisher's exact test.