A new syndrome of moyamoya disease, kidney dysplasia, aminotransferase elevation, and skin disease associated with de novo variants in RNF213

Abstract Ring‐finger protein 213 (RNF213) encodes a protein of unknown function believed to play a role in cellular metabolism and angiogenesis. Gene variants are associated with susceptibility to moyamoya disease. Here, we describe two children with moyamoya disease who also demonstrated kidney disease, elevated aminotransferases, and recurrent skin lesions found by exome sequencing to have de novo missense variants in RNF213. These cases highlight the ability of RNF213 to cause Mendelian moyamoya disease in addition to acting as a genetic susceptibility locus. The cases also suggest a new, multi‐organ RNF213‐spectrum disease characterized by liver, skin, and kidney pathology in addition to severe moyamoya disease caused by heterozygous, de novo C‐terminal RNF213 missense variants.


| Editorial policies and ethical considerations
Consent for participation was obtained from each family, and appropriate consent forms were signed. For the patient evaluated at The Children's Hospital of Philadelphia, the study was approved by the IRB (Protocol number 16-013278).

| Genetic testing methodology
Clinical trio exome sequencing was performed on Patient 1 at Invitae, and at Gene Dx for Patient 2. Exome at Invitae was performed by isolating genomic DNA, enriching targeted regions using a hybridizationbased protocol, and sequencing using an Illumina platform. Targeted regions (95% of mappable exome ±10 base pairs of flanking region) were covered at >Â20 depth. Reads were aligned to the reference sequence (GRCh37). All reported pathogenic and likely pathogenic variants were independently confirmed either by sanger sequencing, Pacific Biosciences SMRT sequencing, or MLPA-seq. Regarding Patient 2, after negative clinical exome testing, Patient 2 was recruited into the Center for Applied Genomics (CAG) at CHOP for exome reanalysis. Variant call format (VCF) were obtained from Gene Dx and reanalyzed using an in-house variant annotation, filtration and prioritization platform developed within the CAG and validated for clinical use. Variants were initially filtered at 0.5% gnomAD MAF and annotated with a combination of multiple tools and databases, including Variant Effect Predictor, HGMD, ClinVar, dbSNP, OMIM, HPO, PolyPhen-2, and SIFT, and a custom-built splice-site annotator. Variants are assigned a priority score of likelihood as the causal variant for the patient's disease, ranked using a weighted combination of (a) overlap with HPO terms, (b) patient and family genotypes, (c) predicted functional impact, (d) inheritance modeling, (e) presence in mutation databases such as HGMD and ClinVar; and other factors.
The identified RNF213 variant was validated by sanger-sequencing using the following primers: F: CCTGTAAACCTAGCCCCTCAT; R: TCCCCAAGATCATGTACTAGCT. Gene Matcher was used to match these two cases (Sobreira, Schiettecatte, Valle, & Hamosh, 2015  Elevated aminotransferases were detected within the first weeks of life (400 s U/L). These continued to rise, ultimately peaking at 7-months of age (5000 s U/L), and then decreased without intervention. Liver echotexture was mildly coarse and bright on ultrasound.
Liver biopsy at 7-months of age showed preserved liver architecture with scattered hemosiderin-laden histiocytes, mild portal tract expansion, and fine portal to portal tract bands of fibrosis without evidence of autoimmune hepatitis.
Patient was noted to have hematuria and nephrotic range proteinuria with normal creatinine during her initial workup. She developed hypertension, requiring multiple antihypertensive medications.
Echocardiogram showed mild to moderate left ventricular hypertrophy with normal function, consistent with longstanding hypertension.
Renal ultrasound was normal; CT angiogram showed bilateral renal Father has no history of stroke, skin, liver, or kidney disease. Other identified de novo and rare variants were felt to be unrelated to phenotype (Table S1). Parents have since had a second child who had a normal brain MRI at 6-weeks of age. No genetic testing has been done.

| Patient 2
Patient 2 was the product of a full-term, uncomplicated pregnancy to a G2P1!2 mother. Birth weight was 3.997 kg (75-90%); length was 53 cm (75%). Poor weight gain was noted shortly after birth with associated poor feeding, and he was diagnosed with gastroesophageal reflux disease. At 4 months of age, patient was hospitalized with persistent failure to thrive and was found to have elevated aminotransferases (AST 310 U/L; ALT 317 U/L) with normal liver ultrasound, hypertension with evidence of left ventricular hypertrophy, stage III chronic kidney disease, and normal-sized but diffusely echogenic kidneys bilaterally with no evidence of vesicoureteral reflux. Metabolic workup including plasma amino acids, urine organic acids, very-long chain fatty acids, pipecolic acid, carbohydrate deficient transferrin, and N-glycan testing was nondiagnostic. Protease inhibitor (PI) typing was notable for IZ, consistent with a diagnosis of alpha-1 antitrypsin deficiency (AATD). A chromosomal microarray, congenital anomalies of the kidney and urinary tract gene panel, and a cystic kidney disease gene panel were nondiagnostic. Patient was discharged home on amlodipine monotherapy.
Patient was re-admitted at 6-months of life with vomiting, hypertension, and new onset rash confirmed by biopsy to be erythema multiforme. Due to concern for amlodipine allergy, patient was switched to propranolol therapy with good response. He continued to have intermittent recurrence of his rash: large, flat, red lesions with central clearing that were nonpruritic and typically appeared on the arms and legs (Figure 1f). His weight gain improved, but he continued to have short stature with normal endocrine laboratories. At 11-months of age he was admitted due to an episode of hypoglycemia, unresponsiveness, and facial droop of unclear etiology. Trio whole exome sequencing was sent and was notable for biallelic Other identified de novo and rare variants were felt to be unrelated to phenotype (Table S2).

| DISCUSSION
MMD is a progressive cerebral vasculopathy and stroke syndrome, typically of nongenetic basis, but at times demonstrating autosomal dominant inheritance with variable penetrance. RNF213 is a wellestablished susceptibility gene for MMD, with the East-Asian founder variant p.(Arg4810Lys) conferring over a 100-fold increase in risk of disease (Kamada et al., 2011;Liu et al., 2011;Miyatake et al., 2012).
RNF213 encodes a protein of unknown function, but with suspected roles in protein turnover (Morito et al., 2014). We have identified two patients with heterozygous de novo variants in RNF213 with severe, early-onset moyamoya disease and stroke, but also with novel extra-cerebral phenotypes, including chronic kidney disease, liver disease with elevated aminotransferases, and skin disease.
The specific association between RNF213 gene variants and pediatric MMD and stroke has been described previously, both in children heterozygous or homozygous for the founder East-Asian p.(Arg4810Lys) variant as well as in children with rare, de novo variants (Guey et al., 2017;Ito et al., 2015;Miyatake et al., 2012;Zhang et al., 2019). The observation that individuals with biallelic RNF213 variants have more severe and earlier-onset disease suggests that gene-dosage plays a critical role in dictating disease pathology, and would suggest that disease-associated heterozygous variants in pediatric stroke have a severe and deleterious effect on RNF213 function that cannot be overcome by the wild-type allele.
Though the exact function of RNF213 and the mechanism by which gene variants cause disease is unknown, in vitro and in vivo models support a role for RNF213 in angiogenesis. Specifically, abnormal RNF213 function in vitro causes abnormal cellular proliferation and dysregulation of inflammatory and extracellular matrix gene expression (Ohkubo et al., 2015), morpholino-mediated and TALENmediated suppression of rnf213 expression in zebrafish causes abnormal vascular sprouting (Lin et al., 2020;Liu et al., 2011;Wen et al., 2016), and murine models of rnf213 deficiency demonstrate abnormal angiogenesis and response to vascular injury (Ito et al., 2015;Sonobe et al., 2014). The exact mechanism by which variants in RNF213 disrupt angiogenesis and vascular injury response is unknown, but RNF213 is known to play a role in noncanonical Wnt and calcium signaling and NF-kB pathway activation, which are pathways known to regulate growth and inflammatory signaling (Amal et al., 2019;Scholz et al., 2016;Takeda et al., 2020).
A role for RNF213 outside the central nervous system has been suspected, given the near-ubiquitous expression of the gene. Additional roles uncovered for RNF213 include fast muscle formation and neuromuscular signaling (Kotani et al., 2015), cellular response to hypoxia (Banh et al., 2016), beta-cell function (Kobayashi et al., 2013), and energy and lipid metabolism (Banh et al., 2016;Piccolis et al., 2019;Sugihara et al., 2019). RNF213 promotes lipid droplet stability by interfering with adipose triglyceride lipase activity, and its deficiency protects against lipotoxicity and palmitate-induced cell death, though interestingly the common East-Asian variants have no effect on lipid droplets in vitro (Piccolis et al., 2019;Sugihara et al., 2019). Extra-cerebral manifestations associated with RNF213 variants have also been described clinically, including an infant who presented with elevated aminotransferases found on biopsy to have lysosomal neutral fat accumulation and cytoplasmic cholesterol crystals in addition to MMD (Harel et al., 2015), a child who presented with a dysplastic right kidney in addition to MMD (Dibi, Maana, Jabourik, & Bentahila, 2017), and a young adult with polycystic kidneys as well as MMD (Pracyk & Massey, 1989). Interestingly, in all cases where genetic data was available, mutations clustered in the RNF213 C-terminal region, which is characteristic of isolated MMD as well as with our syndromic patients (Harel et al., 2015).
Given the precedence in the literature of other patients with MMD and liver and kidney pathology (Dibi et al., 2017;Harel et al., 2015;Pracyk & Massey, 1989), the patient reported by Harel and colleagues (Harel et al., 2015) with MMD and liver disease with a known de novo RNF213 variant, as well as the near-identical phenotypes of our reported patients, both with de novo RNF213 variants, we are highly suspicious of a distinct hepatorenal phenotype associated with RNF213 gene variants. The etiology of the syndromic presentations for these patients is unclear. Patient 2 harbors biallelic variants in SERPINA1, resulting in AATD, though these variants are unlikely to explain his phenotype, as PI typing IZ usually manifests in milder liver disease in childhood (Stoller, Hupertz, & Aboussouan, 2006). There is a known association between AATD and cytoplasmic staining anti-neutrophil cytoplasmic antibodies (c-ANCA) vasculitis, specifically granulomatosis with polyangiitis, with increased risk in those with PI Z and S (Mahr et al., 2010); however, there have been no reports or mechanistic literature to support AATD manifesting in MMD, and patient's skin biopsy results were felt to be inconsistent with those typically seen in AATD. Therefore, we concluded the Z allele may be a disease modifier and not the main cause of phenotype.
Regarding kidney phenotype, patients had disparate findings of structurally normal kidneys with biopsy evidence of glomerulosclerosis, interstitial fibrosis and tubular atrophy (Patient 1) and cystic kidney dysplasia (Patient 2). We hypothesize that both of these patients experi- It is unknown at this time why these patients demonstrate such severe cerebral and extra-cerebral manifestations. All variants map to the RNF213 C-terminal region, which has been established as a hotspot for severe disease (Cecchi et al., 2014;Guey et al., 2017;Zhang et al., 2019) (Figure 2). RNF213 variants may cause disease via a threshold effect, and the variants identified in our patients may more severely impact RNF213 function. This model is certainly supported by the finding that individuals harboring two copies of the East-Asian RNF213 risk allele have earlier-onset and higher penetrant disease (Miyatake et al., 2012;Zhang et al., 2019). Patient 1 was found to harbor two changes in the RNF213 gene, and had more severe disease with early death. Importantly, we are unable to determine whether these changes are in cis or trans. It is also possible that only certain variants affect extra-cerebral RNF213 function, as the East-Asian p.
(Arg4810Lys) variant is an established risk factor for MMD, but does not affect lipid droplets in vitro (Sugihara et al., 2019).
The skin pathology seen in our two patients has not been reported previously. Patient 1 was steroid dependent, suggesting an inflammatory component to the skin disease, whereas Patient 2 has not required treatment. Skin biopsy on Patient 1 was nonspecific but skin biopsy performed on Patient 2 was suggestive of erythema multiforme, which is an immune-mediated condition usually triggered by viral infection or by an environmental or toxic trigger. It is tempting to speculate that the RNF213 variants seen in these patients disrupt NF-kB signaling and the immune response more severely than previously described variants. This phenomenon may also be relevant F I G U R E 2 Schematic of the RNF213 protein showing the 2 AAA+ ATPAse domains as well as the RING domain. Amino acid numbers listed below the figure. Identified patient variants noted by yellow arrow [Color figure can be viewed at wileyonlinelibrary.com]