Delineation of dominant and recessive forms of LZTR1‐associated Noonan syndrome

Noonan syndrome (NS) is characterised by distinctive facial features, heart defects, variable degrees of intellectual disability and other phenotypic manifestations. Although the mode of inheritance is typically dominant, recent studies indicate LZTR1 may be associated with both dominant and recessive forms. Seeking to describe the phenotypic characteristics of LZTR1‐associated NS, we searched for likely pathogenic variants using two approaches. First, scrutiny of exomes from 9624 patients recruited by the Deciphering Developmental Disorders (DDDs) study uncovered six dominantly‐acting mutations (p.R97L; p.Y136C; p.Y136H, p.N145I, p.S244C; p.G248R) of which five arose de novo, and three patients with compound‐heterozygous variants (p.R210*/p.V579M; p.R210*/p.D531N; c.1149+1G>T/p.R688C). One patient also had biallelic loss‐of‐function mutations in NEB, consistent with a composite phenotype. After removing this complex case, analysis of human phenotype ontology terms indicated significant phenotypic similarities (P = 0.0005), supporting a causal role for LZTR1. Second, targeted sequencing of eight unsolved NS‐like cases identified biallelic LZTR1 variants in three further subjects (p.W469*/p.Y749C, p.W437*/c.‐38T>A and p.A461D/p.I462T). Our study strengthens the association of LZTR1 with NS, with de novo mutations clustering around the KT1‐4 domains. Although LZTR1 variants explain ~0.1% of cases across the DDD cohort, the gene is a relatively common cause of unsolved NS cases where recessive inheritance is suspected.

Number: Biomedical Research Centre Oxford; Wellcome Trust, Grant/Award Numbers: 090532/Z/09/Z, WT098051 case, analysis of human phenotype ontology terms indicated significant phenotypic similarities (P = 0.0005), supporting a causal role for LZTR1. Second, targeted sequencing of eight unsolved NS-like cases identified biallelic LZTR1 variants in three further subjects (p.W469*/p.Y749C, p. W437*/c.-38T>A and p.A461D/p.I462T). Our study strengthens the association of LZTR1 with NS, with de novo mutations clustering around the KT1-4 domains. Although LZTR1 variants explain~0.1% of cases across the DDD cohort, the gene is a relatively common cause of unsolved NS cases where recessive inheritance is suspected. webbing of the neck, widely-spaced nipples, undescended testes and short stature. [1][2][3] Skeletal abnormalities can include pectus malformations, spinal deformities and cubitus valgus. Heart defects such as pulmonary stenosis or hypertrophic cardiomyopathy occur in a substantial fraction of patients and the association with intellectual disability has long been recognized. 2,4,5 The NS phenotype is extremely variable and not all features are observed in all patients. Heterogeneity also exists within families and many mildly affected adults remain undiagnosed until the birth of a more severely affected child. The oft quoted incidence rate of 1/1000-1/2500 live births 5 may therefore be an underestimate.
PTPN11 encodes a protein tyrosine phosphatase implicated in RAS-MAPK signalling and was the first gene associated with NS, responsible for~50% of cases. 6 Several other NS genes with roles in RAS-MAPK signalling have subsequently been identified. These include KRAS, SOS1, RAF1, NRAS, BRAF, RIT1 and SOS2. 7 Despite these gene discovery efforts, a small but significant fraction of cases remain mutation-negative for known genes.
In contrast to other studies which concentrated on components of RAS-MAPK signalling, 8 unbiased exome sequencing of two large Brazilian kindreds identified LZTR1 as the only gene harbouring rare, predicted-deleterious variants co-segregating with NS consistent with an autosomal dominant (AD) mode of inheritance. 9 Identification of three smaller families with mutations clustering around the same protein-interaction domains supported LZTR1 as a novel NS gene (MIM#616564).
Until recently, the mode of inheritance associated with NS has exclusively been AD, with mutations arising de novo or inherited from affected parents. Although sibling recurrence is occasionally seen in families where parents are unaffected, such findings are typically thought to be the result of genetic mosaicism. 10 Indeed, PTPN11 mutations undergo positive selection during spermatogenesis. 11,12 Nevertheless, speculation that an autosomal recessive (AR) form of NS exists, first proposed >25 years ago, 13

| Analysis of human phenotype ontology terms
Clinical information was collected using human phenotype ontology (HPO), a standardized vocabulary of phenotypic abnormalities (http:// human-phenotype-ontology.github.io/). The significance of phenotypic similarity between LZTR1-positive patients was estimated by comparing similarity of HPO terms between the patients of interest to that between randomly selected patients from the diverse DDD cohort, as described. 18,19 2.5 | Sanger sequencing, allele-specific PCR and relationship confirmation PCR amplification of LZTR1 was performed as described in Table S1.
An additional amplicon was included to capture intron 16 where variants can lead to retention of an alternative exon. 15 Following enzymatic purification, Sanger sequencing was performed using BigDye (version 3.1) and the ABI 3730XL (Applied Biosystems, Foster City, California).
Where parental samples were unavailable, compoundheterozygous variants were phased by allele-specific PCR whereby the 3 0 -base of the first primer was complementary to either the wildtype or mutant allele (Table S1). Sanger sequencing was then performed to determine the sequence at the second locus. A similar method was used to phase a de novo variant where the closest informative SNP was too distant to be phased by Illumina read-pairs (Table S1).
For one family where the patient underwent exome sequencing as a singleton, maternity/paternity were confirmed by genotyping nine short tandem repeat (STR) loci using the AuthentiFiler PCR Amplification Kit (ThermoFisher Scientific, Waltham, Massachusetts).

| Four likely-pathogenic de novo LZTR1 variants identified in patient-parent trios
Among 7832 parent-parent-child trios, exome sequencing uncovered five patients with de novo missense mutations in LZTR1, all called with high confidence (posterior probability for the de novo configuration, pp_dnm >0.9). Of these, four clustered around kelch domains KT1-4 (codons 79-285, UniProt Q8N653), had CADD scores of 26-34 and thus were deemed likely pathogenic. More genetic information and clinical characteristics of these patients are in Table 1, Table S2 and    Variants initially shown to be in trans by allele-specific PCR.    (Table S5). For singleton patient 269172, the closest identifiable informative SNP was >5 kb away, so phasing of the p.Y136C mutation was performed using allele-specific PCR (Table S1). In all five cases, the de novo variant had arisen on the paternal chromosome (Table S5).

| Recessive LZTR1 cases and a "blended" phenotype of NS and nemaline myopathy
Three unrelated individuals harbouring biallelic variants in LZTR1 were found. These individuals were also identified in a global analysis of the DDD cohort that sought to identify novel disease genes in addition to quantifying the overall contribution of recessive variants to developmental disorders. 19 The variants identified were all rare, with gnomAD MAFs of 1/241846 to 19/276182 and CADD scores of 23 to 40 (Table 1). Additional information for these patients is shown in  Figure 1F-G, Figure S1F-G).

| Phenotype comparisons using unbiased HPO terms
To help support that the LZTR1 variants were of clinical relevance, we tested whether the patients identified above were more similar than expected by chance. Comparison of HPO terms for the four patients in whom trio exome analysis had uncovered de novo dominant mutations in KT1-4 domains of LZTR1 showed significant phenotypic similarity (Table 1; P = 0.0478). Singleton patient 269172 was excluded from this analysis as prioritisation of this case from a group of singletons had included review of phenotype information. Comparison of the three patients with compound-heterozygous variants also showed a degree of similarity, although this did not reach a formal level of significance (P = 0.0629), as reported previously. 19 Grouping the AD and AR cases together increased the significance levels of phenotypic similarity (P = 0.0019). Finally, we removed the AR patient (284672) with biallelic mutations in NEB (as well as LZTR1) as we believed this patient to have a "blended" phenotype, with features such as myopathy and ptosis more likely due to NEB. This comparison of the remaining six cases further increased levels of similarity (P = 0.0005).  (Supplementary note 2). A reporter assay using a dual luciferase strategy was performed, as described. 23,25 The ratio of renilla to firefly luciferase was consistently reduced to 77% to 85% for the mutant 5'-UTR in comparison to the WT (Supplementary note 2). AR inheritance had been suspected in this family as a male sibling (despite normal early scans) developed polyhydramnios at 20 weeks gestation.

| Biallelic LZTR1 variants in 3/8 patients with a clinical diagnosis of NS
Severe fetal hydrops ensued and an emergency cesarean section was performed at 38 weeks due to reduced fetal movements. The baby died shortly after delivery. A postmortem showed an increased heart mass of 24.6 g (normal 16.4 ± 5.7 g) but no structural abnormality.
Histology of skeletal and cardiac muscle showed an excess of muscle spindles, noteworthy given reports of such anomalies in Costello syndrome. 26

| Clinical comparison and Face2Gene analysis
Review of available photographs indicates that a depressed nasal bridge in young children with narrow nasal root and broad nasal tip in older children are characteristic features of LZTR1-associated NS. In AD cases it appears that the face may elongate with age ( Figure 1A, C). Analysis of photos using the Face2Gene tool (www.face2gene. com) showed that for 8/10 patients where photographs were obtained, NS ranked highest for at least one age (Table 1). For 4/8 of these matches, there was a high degree of similarity to NS ( Figure S1).
Based on a patient described by Johnston et al 15  Consistent with this, we identified de novo mutations at residues p.
S244 and p.G248 ( Figure 2C). The two de novo mutations involving p.
Y136 ( Figure 2B) Figure 2D). The exception to this rule was O1504902 who harbored missense variants at adjacent codons ( Figure S2).  Using a combination of parent-child exomes and allele-specific PCR showed that 5/5 de novo mutations reported here occurred on the paternal chromosome. Although a larger case-series is required to reach significance, these results are notable given mutations in PTPN11 and other RAS-MAPK genes can influence spermatogonial selection and predominantly occur on the paternal chromosome. 31 A 2004 study phased 14 de novo mutations in PTPN11 and found all originated on the paternal haplotype. 11 The mean paternal age in that PTPN11-positive cohort was 35.6 years, 2.2 years above that for the PTPN11-negative cases and 6.1 years older than the population average. Increased paternal age at conception and a similar bias in the parental origin of de novo HRAS mutations have been documented in patients with Costello syndrome. 32,33 For the five cases with de novo LZTR1 mutations described here, paternal ages at childbirth were also elevated (mean = 35.8, Table S5) compared with the average across this DDD datafreeze (mean = 32.6).
A recent study focussing on fetal malformations detected in utero identified a case of non-immune hydrops fetalis with a homozygous variant in LZTR1. 34 Fetal hydrops was also observed in 4 non-liveborn siblings described by Johnston et al. 15 Together with siblings of O1409410/O1409412, these results indicate that LZTR1 mutations can result in a much more severe form of disease. Further studies should aim to uncover the reasons for this extreme variability and whether such lethal presentations of disease can also be associated with the AD form of LZTR1-associated NS.
Combining our results with those described in the literature, 9,15,20,27 hypertrophic cardiomyopathy was reported in 5/26 of individuals with AD-acting mutations but 19/26 of those harbouring biallelic variants (Tables S2 and S6). A systematic analysis of cardiac involvement in larger clinical cohorts of patients with LZTR1 mutations is warranted to confirm whether this bias is reproducible.
In conclusion, our study strengthens the association of LZTR1 with AD/AR forms of NS. In the dominant condition, mutations cluster around the KT1-4 domains. In the AR form, compound-heterozygosity often involves a LoF allele in trans with a presumed hypomorphic variant. Although LZTR1 mutations explain only~0.1% of cases in the DDD study, the gene is a notable cause of unsolved NS cases, especially where recessive inheritance is suspected.