Expanding the clinical and genetic spectrum of ALPK3 variants: Phenotypes identified in pediatric cardiomyopathy patients and adults with heterozygous variants

Citation for published version (APA): Herkert, J. C., Verhagen, J. M. A., Yotti, R., Haghighi, A., Phelan, D. G., James, P. A., Brown, N. J., Stutterd, C., Macciocca, I., Leong, KE., Bulthuis, M. L. C., van Bever, Y., van Slegtenhorst, M. A., Boven, L. G., Roberts, A. E., Agarwal, R., Seidman, J., Lakdawala, N. K., Fernández-Avilés, F., ... van de Laar, I. M. B. H. (2020). Expanding the clinical and genetic spectrum of ALPK3 variants: Phenotypes identified in pediatric cardiomyopathy patients and adults with heterozygous variants. American Heart Journal, 225, 108-119. https://doi.org/10.1016/j.ahj.2020.03.023

Expanding the clinical and genetic spectrum of ALPK3 variants: Phenotypes identified in pediatric cardiomyopathy patients and adults with heterozygous variants under 4 years of age, and myofibrillar disarray in adults.
Conclusion Biallelic damaging ALPK3 variants cause pediatric cardiomyopathy manifested by DCM transitioning to hypertrophy, often with poor contractile function. Additional extracardiac features occur in most patients, including musculoskeletal abnormalities and cleft palate. Heterozygous LoF ALPK3 variants are enriched in adults with cardiomyopathy and may contribute to their cardiomyopathy. Adults with ALPK3 LoF variants therefore warrant evaluations for cardiomyopathy. (Am Heart J 2020;225:108-119.) Pediatric-onset of cardiomyopathy, a disease of the heart muscle causing systolic and/or diastolic dysfunction, is a devastating cause of heart failure in children and the most common indication for heart transplantation in children over 12 months of age. 1 Onset occurs prenatally, at birth, or throughout childhood. Damaging variants in more than 100 genes cause either isolated or syndromic pediatric cardiomyopathy through many different pathological mechanisms. [2][3][4] ALPK3 (MIM 617608) is a recently identified pediatric cardiomyopathy gene that encodes alpha-protein kinase 3 (ALPK3), a protein with functions that remain incompletely understood. ALPK3 participates in normal intercalated disc formation and sarcomere organization in both humans and mice. [5][6][7] Alpk3-null mice develop a non-progressive cardiomyopathy characterized by predominantly myocardial hypertrophy and diminished systolic function, as typically occurs in dilated cardiomyopathy (DCM). 6 Cardiomyocytes derived from human-induced pluripotent stem cell (hiPSC-CMs) lacking ALPK3 display abnormal calcium handling. 7 We previously reported seven patients from four unrelated consanguineous families with pediatric cardiomyopathy caused by biallelic predicted proteintruncating (loss-of-function, LoF) variants in ALPK3. 5 , 7 Two additional case reports described severe congenital cardiomyopathy including features of both DCM and hypertrophic cardiomyopathy (HCM) from homozygous ALPK3 LoF variants. 8 , 9 Extracardiac manifestations have also been observed, including multiple pterygia with skeletal muscle underdevelopment, facial dysmorphisms, and skeletal features. [7][8][9] Unlike affected children, the clinical phenotypes of parents and relatives who carry only one damaging ALPK3 allele are less penetrant. Three of 21 published heterozygous carriers from two families had clinical features of HCM, described as hypertrophy of the interventricular septum, 5 , 9 whereas other heterozygous carriers had no cardiac disease. It is currently unclear if these observations indicate that damaging ALPK3 variants contribute to unexplained cardiomyopathy or modify cardiomyopathy that is caused by a pathogenic or likely pathogenic variant in an established disease gene. To address these issues, we delineated the clinical and genetic spectrum of patients with damaging biallelic ALPK3 variants and defined the prevalence of heterozygous ALPK3 variants in two cohorts with adult-onset cardiomyopathy.

Patient recruitment
Our study was carried out in collaboration with clinicians from seven different countries and institutions. Mutation analysis was performed using next-generation sequencing (NGS), either whole exome sequencing or targeted gene panels. Details on sequencing methods and data analysis are available in the Data Supplement. We reviewed clinical data of 19 patients with biallelic variants in ALPK3 (NM_020778.4), including nine previously reported patients. 5 , 7-9 HCM was defined as increased ventricular wall thickness (end diastolic wall thickness: zscore ≥2) not solely explained by abnormal loading or structural heart conditions such as valve disease, congenital heart disease, or hypertension. DCM was defined as ventricular dilation (LV end-diastolic dimension N2 SD above mean for body surface area) and systolic dysfunction (fractional shortening or LV ejection fraction N2 SD below mean for age) in the absence of abnormal loading conditions. 10

Variant interpretation
The pathogenicity of variants was assessed using Alamut Visual software (Interactive Biosoftware, Rouen, France), a gene browser that integrates missense prediction tools (Align GVGD, SIFT, MutationTaster, PolyPhen-2), allele frequencies from different population databases (gnomAD, 12 ESP, GoNL 13 ) and disease-specific databases (HGMD, ClinVar, LOVD) and mRNA splicing prediction tools (SpliceSiteFinder-like, MaxEntScan, NNSPLICE, GeneSplicer, and Branch Points). A potential splice effect was defined as a difference between reference and mutated scores greater than 10% reported by three or more of five mRNA splicing prediction tools. Deleteriousness of variants was scored using combined annotation dependent depletion (CADD). 14 A scaled CADD score of 10, 20, or 30 indicates the top 10%, 1%, and 0.1% most deleterious substitutions in the human genome, respectively. Variants were interpreted according to the 2015 ACMG guidelines. 15 Variants with a minor allele frequency (MAF) b0.1% in the Genome Aggregation Database (gnomAD) dataset (considering total population and major subpopulations) were considered rare. Nonsense and frameshift variants were considered null variants, with the exception of those residing in the last exon or the last 50 base pairs of the penultimate exon.

Histology
Paraffin-embedded or frozen cardiac tissue was available for two affected individuals (patient 1 from family 1 (F1P1) and patient 2 from family 2 (F2P2); pedigrees 1 and 2, Data Supplement). In addition, we collected muscle biopsy specimens from the lateral portion of the quadriceps femoris muscle from patient 3 from family 2 (F2P3; pedigree 2, Data Supplement) and her healthy sister (pedigree 2, Data Supplement) and from the spine, taken at scoliosis surgery, from patient 1 from family 3 (F3P1; pedigree 3, Data Supplement). Tissues from agematched donors were used as controls. All samples were histologically examined after hematoxylin and eosin staining using standard techniques. Samples from patient F3P1 were also examined by electron microscopy.

Cohort screening
ALPK3 was evaluated as part of a targeted NGS panel (gene panels B, E or F, Supplementary Table II) in 1548 patients (suspected of) having cardiomyopathy who were referred for diagnostic genetic testing at two molecular diagnostic laboratories in the Netherlands between December 2015 and July 2018. ALPK3 was also evaluated from whole exome sequence data obtained for 149 unrelated U.S. patients of European ancestry who had clinically diagnosed HCM or DCM. In each of these patients, prior NGS analyses of 83 established or putative cardiomyopathy genes (gene panel G, Supplementary Table II) had excluded a pathogenic or likely pathogenic variant. Co-segregation analysis was performed for available family members in both of these cohorts.

Haplotype analysis
To investigate whether the recurrent c.4736-1G N A, p. (Val1579Glyfs*30) variant originated from a single mutational event, haplotype analysis was performed using 13 microsatellite markers surrounding ALPK3. DNA from six probands (one homozygous carrier (F1P1) and five heterozygous carriers) was analyzed, and DNA samples of three family members of F1P1 who carry the variant were used to verify the phase and reconstruct the haplotype.

Statistical analysis
Sequence data of 64,000 unrelated Non-Finnish Europeans (NFE, assembled by gnomAD) were used as an independent control dataset. Raw data (version 2.1) were downloaded and filtered on PASS quality status. ALPK3 variants with a MAF N0.1% (2000 alleles) were excluded. The calculation of burden for LoF variants in cardiomyopathy cohorts and gnomAD subjects excluded LoF variants in the last exon of ALPK3. The prevalences of ALPK3 variants were expressed as proportions (exact 95% binomial confidence intervals [CI]). Differences in prevalence rates between cohorts were estimated as the risk difference with the exact 95% confidence interval of the risk difference and statistically compared using a onetailed binomial test. Values of P b.05 were considered significant.

Identification of ALPK3 sequence variants
All nine previously reported patients carried biallelic LoF variants in the ALPK3 gene. Of the 10 new patients described here, two carried biallelic ALPK3 LoF variants: patient F7P3 (a distant relative of F7P1 and F7P2) carried c.1018C N T, p.(Gln340*) and c.4332delC, p. (Lys1445Argfs*29) and patient F12P1 was homozygous for c.3418C N T, p.(Gln1140*). Seven patients (F7P1, F7P2, F8P1, F9P1, F10P1, F10P2 and F11P1) had compound heterozygous LoF and missense ALPK3 variants (Tables I and II and Supplementary Table I). Patient F13P1 carried a homozygous missense variant, c.5155G N C, p.(Ala1719Pro), which alters an alanine residue within the alpha-kinase domain. The Ala1719Pro substitution was absent in public exome databases and is predicted to be damaging by SIFT and PolyPhen-2.
Identified ALPK3 variants did not cluster ( Figure 1). The alpha-kinase domain has high sequence identity among ALPK family members and is required for phosphate modification of other proteins, a fundamental process involved in most signaling and regulatory processes within eukaryotic cells. 16 No additional likely pathogenic or pathogenic variants in genes associated with cardiomyopathy, including TTN, nor any pathogenic copy number variants explaining their cardiomyopathy were identified in any of the patients with biallelic ALPK3 variants (Data Supplement and Supplementary  Table II). All but two ALPK3 missense variants had high CADD scores (N 20; Table II), and most of the novel amino acids substituted residues are highly conserved across species (Data Supplement). In contrast, the p. (Val812Met) variant in F7P1 and F7P2 had a low CADD score and showed conflicting predictions of pathogenicity. However, the phenotype of both siblings carrying this variant (and a LoF variant on the other allele) showed striking similarities with other patients with biallelic ALPK3 variants. The p.(Glu199Asp) variant identified in patients F10P1 and F10P2 had a low CADD score and is also a conservative amino acid substitution that may not impact secondary protein structure given the similar properties of these residues (Grantham score: 45). While this missense variant altered a residue within a region of poor sequence alignment, thereby limiting assessment of evolutionary conservation, the variant is absent from gnomAD. Based on the shared phenotype exhibited by F10P1, F10P2, and other carriers of biallelic ALPK3 variants, we suggest that p.(Glu199Asp) is also damaging. However, functional studies should be carried out to further support our hypothesis.
Clinical features of patients with biallelic ALPK3 sequence variants Table I and Supplementary Tables I and III provide clinical summary data for the 19 patients (9 male), and additional descriptions are provided in the Data Supplement. The age at diagnosis of cardiomyopathy ranged from 20 weeks of gestation to 53 years; two patients were diagnosed after the age of 18 years. Among patients with biallelic LoF variants, median age at diagnosis was 7 days (range: 20 weeks gestation to 9 years). Median age at diagnosis in patients carrying a LoF and a missense variant was 3 months; two of them were diagnosed as adults (range birth to 53 years). Patient F13P1, harboring two missense variants, presented with cardiomyopathy at 35 weeks of gestation.

Histopathologic examination
Post-mortem microscopic examination of myocardial tissue showed (sub)endocardial fibro-elastosis in patients F1P1, F2P1, F5P2, and F5P3. At the DCM stage, no myofiber disarray was observed (patient F1P1). Histopathology of patient F2P1, who had both ventricular dilation and hypertrophy, showed focal cardiomyocyte hypertrophy without myofiber disarray. Patients F7P2 and F7P3 underwent cardiac biopsy at age 4 years and 28 years, respectively, when their DCM progressed to biventricular hypertrophy. Cardiac histopathology of patient F7P3 showed cardiomyocyte hypertrophy with myofiber disarray. A spinal muscle biopsy of patient F3P1 taken at scoliosis surgery showed variation in fiber size, fiber splitting, and numerous central cores ( Figure 5A and B). However, subsequent examination of the quadriceps muscle of the same patient did not show any ultrastructural abnormalities ( Figure 5C).

Burden of heterozygous LoF ALPK3 variants identified in patients with adult-onset cardiomyopathy
We assessed the prevalence of ALPK3 variants in two independent cardiomyopathy cohorts. The Dutch cohort comprised 1548 index patients with predominantly adultonset cardiomyopathy referred for clinical genetic testing. None had biallelic damaging ALPK3 variants, but 24 rare (MAF b0.1%) heterozygous ALPK3 variants were identified in 39 patients (2.5%; 95% CI: 1.8%-3.4%), including four LoF variants (three frameshift and one splice site variant resulting in exon 10 skipping), 18 missense variants, one stop gain variant in the last exon, and one synonymous variant with a predicted effect on splicing (Supplementary Table IV). Ten variants recurred in more than one patient. The heterozygous c.4736-1G N A, p.(Val1579Glyfs*30) variant initially observed in the unaffected parents and sister of patient F1P1 also occurred in five adult cardiomyopathy probands (P025-P029) and four of 273,120 alleles (0.0015%) in gnomAD. A shared haplotype consisting of 5 of 13 polymorphisms located within 2.33 Mb flanking ALPK3 was identified in eight individuals from five families, suggesting a common founder in the Dutch population (Supplementary Table  V).

Discussion
The clinical manifestations of biallelic and heterozygous ALPK3 variants are quite distinct. Among 19 patients with biallelic damaging ALPK3 homozygous or compound heterozygous variants, 17 patients presented with pediatric-onset cardiomyopathy. Strikingly, most cases presented initially with DCM that then transitioned to ventricular hypertrophy with reduced systolic performance-an unusual clinical sequence. ALPK3 cardiomyopathy often progressed rapidly and six patients died or underwent cardiac transplantation. Most patients had extracardiac manifestations, including craniofacial and musculoskeletal abnormalities, but these were not sufficiently consistent to delineate a recognizable syndrome.
We report, for the first time, biallelic missense variants that cause pediatric-onset cardiomyopathy. The clinical manifestations associated with these variants were similar to those associated with other damaging ALPK3 variants. These missense variants could result in a conformational change that affects protein folding or flexibility, proteinprotein or protein-DNA interaction, or the activity of the alpha-kinase domain. Unfortunately, no 3D structure of ALPK3 is available to predict the consequence of the missense variants. In addition, we demonstrated higher than expected frequencies of heterozygous ALPK3 LoF variants among adult cardiomyopathy patients in a Dutch and a U.S. cohort. Thirty-seven of these patients were clinically diagnosed with HCM, which is likely related to the hypertrophy phenotype observed in pediatric patients with biallelic ALPK3 variants. Despite this similarity, there were other notable differences between the clinical features associated with monoallelic and biallelic ALPK3 cardiomyopathy, including absence or undetected extracardiac phenotypes. Whether these differences reflect graded dose-responses to ALPK3 deficits or distinct mechanisms by which monoallelic or biallelic variants cause disease remains unknown.
Biallelic ALPK3 variants were associated with a range of morphological and functional abnormalities. Almost half of the live-born pediatric patients presented with DCM that later evolved into ventricular hypertrophy. Three individuals initially displayed a mixed cardiomyopathy with features of both DCM and ventricular hypertrophy that evolved into concentric hypertrophy of both left and right ventricles. 9 This hypertrophic phenotype differs from classic HCM caused by pathogenic variants in genes encoding sarcomere proteins. Notably, hypertrophy was atypical, often biventricular and/or concentric, or apical in distribution. LV dilatation occurred in some pediatric patients, which occurs rarely in HCM and usually decades after diagnosis, with accompanying decrease in systolic performance. 18 , 19 Like other pediatric cardiomyopathies, ALPK3 cardiomyopathy can present with features of more than one subtype. 20 , 21 However, transition from DCM to LVH has not been described before and appears to be unique to biallelic ALPK3 cardiomyopathy.
The histopathology of biallelic ALPK3 cardiomyopathy has some features observed in classic HCM, 22 including focal cardiomyocyte hypertrophy, interstitial fibrosis and, at adult age, myofiber disarray. Whether this histopathology precedes the progression to hypertrophy remains unclear. Patients with biallelic variants in ALPK3 display a variety of rhythm and conduction disturbances reminiscent of those seen in arrhythmogenic cardiomyopathy. We previously showed a reduced plakoglobin signal at intercalated disks of patients with biallelic ALPK3 variants. 5 A reduced plakoglobin signal has also been documented in ACM 23 ; this redistribution of plakoglobin from the junctional pool to the intracellular and nuclear pools likely suppresses the canonical Wnt/beta-catenin signaling, leading to enhanced fibrogenesis and myocyte apoptosis. ACM and ALPK3 cardiomyopathy may share the same pathophysiological mechanisms, thus explaining the arrhythmogenic phenotype in patients with biallelic ALPK3 variants. Alternatively, the rhythm disorders observed in ALPK3 cardiomyopathy may be secondary to progressive disease.
We observed no association between extracardiac manifestations and allelic heterogeneity: biallelic missense or LoF variants seemed to cause similar phenotypes. The majority of patients with biallelic ALPK3 variants, including those with one or two missense variants, had musculoskeletal involvement, including contractures and severe progressive scoliosis. Several patients had cleft palate, velopharyngeal insufficiency, and/or facial dysmorphisms. Jaouadi et al also described a patient with a diversified phenotype, including cleft palate, pectus excavatum, bilateral clinodactyly, and facial dysmorphic features like broad forehead, downslanting palpebral fissures, mild ptosis, and low-set posteriorly rotated ears, which fits with the extracardiac features we observed in our cohort. 9 While we cannot exclude that genome-wide inbreeding contributed to the extracardiac features seen in patients with homozygous ALPK3 variants, their occurrence in multiple unrelated patients with different allelic variants and genetic backgrounds suggests direct effects of ALPK3 variants.
The expression of ALPK3 helps to explain these extracardiac phenotypes. The prevalence of skeletal muscle phenotypes in pediatric patients likely reflects ALPK3 expression in developing skeletal and heart muscle 6 , 24 and in adult skeletal, smooth, and heart muscles (GTEx (https://commonfund.nih.gov/gtex)). In embryonic mice (E8.5), Alpk3 expression is detectable around the first branchial arch, 24 which may account for palatal abnormalities. Further support for the syndromic nature of ALPK3-related disease arises from GeneNetwork Assisted Diagnostic Optimization (GADO), a method that exploits RNA-seq data from a range of tissues and cell types and uses gene coregulation to predict gene functions. 25 For ALPK3, GADO predicts "muscle contraction" and "myogenesis" as the top phenotypes. Based on a combination of the major shared phenotypic abnormalities in our patients, GADO ranked ALPK3 in the top 1% of all coding and non-coding human genes (P = .000432) (Supplementary Table VII).

Pathogenicity of heterozygous ALPK3 variants
Among 37 relatives with a heterozygous LoF variant in ALPK3 (three ≤18 years of age), five (13.5%) were diagnosed with HCM as adults. In line with this finding, we note that gnomAD reports fewer ALPK3 L o F v a r i a n t s ( t r a n s c r i p t N M _ 0 2 0 7 7 8 . 4 (ENST00000258888)) than would be expected if these were to occur randomly, despite a pLI = 0.00. While the observed and expected differences are not statistically significant, we suggest that some individuals with ALPK3 LoF may have an undetected mild or late-onset cardiomyopathy.
To better understand the role of ALPK3 variants in adult-onset cardiomyopathy, we analyzed two independent patient cohorts. In both cohorts, the frequency of ALPK3 LoF variants was significantly higher than in the general population. Together these findings provide compelling evidence that ALPK3 plays important roles in cardiac function and pathologic remodeling. Further evidence for the role of ALPK3 in cardiac hypertrophy arises from a genome-wide association meta-analysis that identified a novel locus at chromosome 15q25.3, which encompasses ALPK3 and is strongly associated with two clinically used QRS traits (Cornell and 12-lead sum), reflecting a higher LV myocardial mass. One of the lead SNPs in this GWAS is in strong linkage disequilibrium with two nonsynonymous SNPs in ALPK3 (P = 9.94e-18). 26 Basic studies are now needed to understand the targets and pathways in which this kinase participates.

Limitations
The majority of pediatric patients with biallelic damaging ALPK3 variants were characterized in the context of clinical care, and medical examinations, imaging, and other laboratory studies were not consistently obtained across all patients. Phenotypes of heterozygous first-degree relatives from cardiac screening, interviews, and/or medical records were not systematically obtained. The mean coverage of ALPK3 sequencing data in gnomAD is much lower than in our cohorts, particularly for distal exon 1 sequences and exon 6. Therefore, the number of v a r i a n t s r e p o r t e d i n g n o m A D m a y b e a n underrepresentation.

Conclusions
Our study reinforces the role of ALPK3 in pediatric cardiomyopathy, and we describe a unique cardiac phenotype with progression of DCM to ventricular hypertrophy as a major feature of ALPK3-related disease. We further show that biallelic variants in ALPK3 can cause a syndromic form of cardiomyopathy with musculoskeletal features as well as craniofacial abnormalities. We also demonstrate an increased burden of heterozygous ALPK3 LoF variants in two adult-onset cardiomyopathy cohorts. Further study is needed to establish the pathogenicity of heterozygous ALPK3 variants.