Phenotypic and mutational spectrum of ROR2‐related Robinow syndrome

Robinow syndrome is characterized by a triad of craniofacial dysmorphisms, disproportionate‐limb short stature, and genital hypoplasia. A significant degree of phenotypic variability seems to correlate with different genes/loci. Disturbances of the noncanonical WNT‐pathway have been identified as the main cause of the syndrome. Biallelic variants in ROR2 cause an autosomal recessive form of the syndrome with distinctive skeletal findings. Twenty‐two patients with a clinical diagnosis of autosomal recessive Robinow syndrome were screened for variants in ROR2 using multiple molecular approaches. We identified 25 putatively pathogenic ROR2 variants, 16 novel, including single nucleotide variants and exonic deletions. Detailed phenotypic analyses revealed that all subjects presented with a prominent forehead, hypertelorism, short nose, abnormality of the nasal tip, brachydactyly, mesomelic limb shortening, short stature, and genital hypoplasia in male patients. A total of 19 clinical features were present in more than 75% of the subjects, thus pointing to an overall uniformity of the phenotype. Disease‐causing variants in ROR2, contribute to a clinically recognizable autosomal recessive trait phenotype with multiple skeletal defects. A comprehensive quantitative clinical evaluation of this cohort delineated the phenotypic spectrum of ROR2‐related Robinow syndrome. The identification of exonic deletion variant alleles further supports the contention of a loss‐of‐function mechanism in the etiology of the syndrome.

All genes associated with RS, play a role in the β-cateninindependent WNT/planar cell polarity pathway. Therefore, despite the genetic heterogeneity, the genes implicated in causing RS to date converge on the WNT signaling pathway, resulting in a recognizable clinical syndrome (White et al., 2018;Zhang et al., 2022).
The receptor tyrosine kinase-like orphan receptors (RORs) are involved in the regulation of multiple biological processes during embryonic development, including development of axial and paraxial mesoderm, nervous system, and neural crest, axial and appendicular skeleton, and kidneys. The characteristic skeletal phenotype of AR-

RS includes vertebral malformations, which were observed in the
Ror2-null mouse model and are caused by the reduced size of presomitic mesoderm (Schwabe et al., 2004). Animal model studies have also identified several WNT pathway components in the mechanisms of craniofacial and limb formation (Geetha-Loganathan et al., 2009;Nohno et al., 1999;Sisson & Topczewski, 2009).

RS-associated genes not only encode components in a common
pathway, but the individual protein component directly interact with each other in signal transduction. WNT5A acts as a soluble extracellular ligand of ROR2, and together with FZD2 transmembrane receptor they trigger the DVL homologs to transduce the β-catenin independent pathway. The WNT5A-ROR2 pathway is a proposed additional branch of the noncanonical WNT-signaling network. Unlike the canonical WNT pathway, other branches of this signaling pathway are not well-defined, resulting in a paucity of information regarding constituent components (Stricker et al., 2017).
Facing the challenges in the clinical diagnosis of RS and in a first attempt to clinically differentiate the AR-RS and AD-RS forms, Mazzeu et al. (2007) investigated the frequency of clinical signs and symptoms in 88 patients with RS, considering rib fusions as indicative and potentially pathognomonic of the AR-RS form. However, despite the more severe bone involvement in AR-RS, rib fusion is not universally present in AR-RS, evident by its absence in a small proportion of molecularly-confirmed cases (Aglan et al., 2015;Mehawej et al., 2012).
The identification of the causative genes in RS further illuminated the underlying patho-mechanism of disease and enhanced the understanding of how the molecular lesions lead to the phenotypic expression. The molecular diagnosis together with quantitative deep phenotyping using Human Phenotype Ontology (HPO) terms and similarity analysis have recently become powerful tools for delineating disease contributing molecular pathways, the biology of disease, and the definition of the etiology of many syndromes, including RS . A detailed phenotypic characterization of patients with an identified disease causing variant allele allows more precise genotype-phenotype correlation, delineation of allelespecific phenotypic differences, and increases the accuracy of clinical diagnosis and management.
There are few reports of AR-RS patients with a confirmed molecular diagnosis. Thirty-two different ROR2 pathogenic variants have been identified so far in patients of different ethnicities (Table S1). Most variants were located in exons 5, 6, and 9. While truncating variant mRNAs are degraded by nonsense-mediated decay (Ben-Shachar et al., 2009), mutant protein caused by missense variants are retained in the endoplasmic reticulum and ultimately lead to the absence of the ROR2 receptor (Ali et al., 2007;Chen et al., 2005).
We report the genotype and detailed HPO-term-based quantitative phenotypic analyses of 22 patients with biallelic ROR2 variants, aiming to further characterize and expand the phenotypic and genotypic spectrum of ROR2 related AR-RS. The clinical information of three patients was partially included in previous publications: A16 (Beiraghi et al., 2011); A6 and A21 Conlon et al., 2021;Gerber et al., 2021;Schwartz et al., 2021;Shayota et al., 2020;Zhang et al., 2021).
Clinical data were collected using a standardized table including all clinical signs present in more than 25% of the patients with AR-RS, according to Mazzeu et al. (2007). Detailed family history, anthropometric data, radiographic images, and other investigations and results were obtained during the consultation or from patient's clinical records.

| Quantitative phenotypic analyses based on HPO terms
Phenotypes were annotated with HPO terms for each affected individual (N = 22). All diseases (n = 8114, including number symbol, plus sign, percent sign, and no symbol in OMIM) and genes (n = 4216, asterisk symbol in OMIM) that have been annotated with HPO terms by OMIM were downloaded from the Human Phenotype Ontology resource page (https://hpo.jax.org/app/download/annotation). Individual similarity matrices were generated with the OntologyX suite of R packages using the Lin′s semantic similarity score and the average method (Lin, 1998;Liu et al., 2019). Similarity matrices were then used to generate distance matrices of individual similarity. Hierarchical agglomerative clustering (HAC) was performed on distance matrices with the Ward's method (Ward, 1963) with the number of clusters set based on visualization of the gap statistic curve.
Individual similarity scores were visualized using the Complex-Heatmap package in R, and statistical analysis of individual groups was done using the OntologyX suite. Annotation grids were generated with the OntologyX suite of packages, and then edited to exclude ancestral terms and to order columns by phenotype frequency. A cohort-to-gene and cohort-to-disease HPO analysis was performed.
These 22 individuals, 21 unrelated research subjects with RS, were separately assessed for phenotypic similarity to all (1) genes and (2) diseases with OMIM HPO annotation. HPO-annotated phenotypes for the 22 individuals were queried against all diseaseassociated genes (n = 4216) or all diseases (n = 8114) annotated with HPO terms by OMIM for phenotypic similarity. Lin semantic similarity scores between all pairs of the 22 individuals and all genes or diseases annotated with HPO terms were calculated. The top 10 phenotypically similar gene-associated or disease HPO term sets to each disease in the group of 31 diseases described above was parsed and duplicates removed. Every combination of two that includes one member from the group of 22 individuals and one from the top phenotypically similar gene associated phenotype matches was taken, and the p value calculated via comparison of the phenotypic similarity score between that group of two and 100,000 randomly selected groups of two from all OMIM HPO annotated genes or diseases, respectively (p value cutoff < 0.001). The gap statistic was calculated for cluster number k = 1-11 (gene analysis) or 8 (disease analysis), and the resultant curve was visualized to select optimal number of clusters to use. HAC analysis and visualization of phenotypic similarity and clustering was then performed as described above for RS proband phenotypes.  Mazzeu et al. (2007) were also included.

| Analysis of variant type-associated phenotypes
Patient prevalence of each phenotype in each group were visualized by using the ComplexHeatmap package in R language.
Sanger sequencing of all ROR2 coding regions and intron-exon boundaries was performed as a first screening method for 10 patients and for confirmation of the causative variants identified through next-generation sequencing for the remaining subjects.

| RESULTS
We analyzed the genotype and phenotype of 22 patients (12 males and 10 females) from 21 unrelated families and from different ethnic backgrounds. Twelve patients have presumed consanguinity by clinical history. In four families the index cases had affected siblings ( Figure S1). The only sibling pair described in detail is the pair A17/ A18 for whom we had comprehensive clinical data on each affected family member.

| ROR2 variant screening
Using different molecular approaches, biallelic causative variants were identified in all 22 patients (Table 1, Figure 1a). All families with historical report of consanguinity presented homozygous alleles. BAF calculator provided further evidence for identity-by-descent and thus confirmed the consanguinity in the two homozygous cases where such data were available (A12 and A13) (Figure 1b). In total, 25 different putatively pathogenic variants were found in 21 patients: including 10 missense, 5 nonsense, 5 small indels, and 2 large deletions. Two patients had a splicing variant, and one patient had a variant affecting the initiation codon. Sixteen of them (64%) are novel variants not yet reported in RS.

| Immuno-localization of Ror2 mutant alleles
We generated mutant Ror2 alleles for four different missense variants identified in our cohort (R108Q, and R366W, located at the extracellular part of Ror2 receptor and P692T and R736Q located at the intracellular part of Ror2 receptor). Co-expression of Ror2 and EGFP-hRas or calnexin in Hela-cells showed that wild-type Ror2 localizes predominantly to the plasma membrane, while mutant Ror2 proteins do not migrate to the plasma membrane and are retained to the endoplasmic reticulum ( Figures S3 and S4). The results did not differ for mutations localized at the intracellular or extracellular domains (data not shown).

| Phenotype analysis
Phenotypes of the 22 patients with biallelic variants in ROR2 are summarized in Table 2 (Detailed phenotype described in Table S2) and the photographs from available patients are shown in Figure 2. All were present in more than 90% of the cohort, and 19 features, in more than 75%, pointing to an overall consistent phenotype.
Midface retrusion, wide nasal bridge, anteverted nares, downslanted mouth corners, bifid tongue, gum hyperplasia, abnormalities of the dentition, short palms, clinodactyly, hemivertebrae, and rib fusion were present in more than 75% of subjects. Therefore, these features should be considered as major defining phenotypic criteria in the clinical diagnosis of ROR2-related Robinow syndrome. Three patients did not present rib fusions, a sign formerly considered pathognomonic for AR-RS (Mazzeu et al., 2007). Intraoral manifestations were also prevalent (above 75%), including bifid tongue, gingival overgrowth, and abnormalities of the dentition. Genital hypoplasia was present in all male patients, but in less than 50% of the females.
Major congenital anomalies, such as abnormal heart and kidneys were present in less than 25% of the patients. Hypoplasia of the tongue was present in 35% of the patients and considered a novel phenotypic feature, not previously associated with AR-RS.

| Quantitative assessment of RS clinical phenotypes
To quantify and visualize genotype-phenotype correlations, semantic similarity scores were calculated using an HPO-based analysis.
Phenotypic similarity scores between each AR-RS proband and OMIM annotated gene phenotypes were calculated and visualized in a cluster heatmap.
Subjects with ROR2 variants in our cohort were clustered with DVL1, WNT5A, ROR2, DVL3, and NXN gene phenotypes (Figure 3).  Figure 3, suggesting that domain localization does not contribute to clinical phenotypic variability in this cohort.
To investigate whether there are mutation-type specific phenotypes as suggested by the initial heatmap analysis, we sorted the cohort for biallelic missense variants (N = 7) and biallelic LoF variants (N = 7) (Figures 5 and S5). Such analysis revealed that none of the patients in the missense group had camptodactyly, hypospadia, or melanocytic nevus and long palpebral fissures, whereas low-set ears, micrognathia, and retrognathia were less-frequent in this group. In contrast, patients with biallelic LoF variants do not present cryptorchidism, abnormal heart morphology, inguinal hernia, and abnormality of the kidney whereas a few patients (N = 3/7) had a broad thumb.
The overall clinical phenotype was consistent in all patients as the majority of clinical signs were present in all patients independently of the type of variant ( Figure 5). However, some signs were more prevalent (difference value > 28%) in the missense or LoF groups allowing discrimination between them ( Figure 5).
A similarity analysis between ROR2 subjects and OMIM annotated disease phenotypes showed that ROR2 subjects strongly clustered with other forms of RS caused by variants in WNT5A, DVL1, DVL3, and NXN. FZD2_OMOD2 grouped into a distinct cluster.
Diseases that have phenotypic overlap with RS are matched using a less stringent p value cutoff (p = 0.005). This aided in viewing the similar sets of diseases to ROR2 patient phenotypes, however, subclusters were more poorly resolved due to the increased number of phenotype sets to cluster.

| DISCUSSION
Here we report a cohort of 22 individuals with AR-RS caused by biallelic ROR2 variants. Twenty-five disease-causing variants in ROR2 were identified, and 16 of these were novel. Although most of the variants were missense, further description of frameshift, initiation site, splice-site variants, and large exonic deletion adds to the evidence that the syndrome is caused by biallelic loss-of-function variants and to the mutational and allelic complexity for this rare disease trait.
Six of the detected variants have been previously described (Table S1). The majority of the previous reports of AR-RS were from Turkey potentially due to the high frequency of consanguineous marriages. However, 13 different variants have been described in Turkish patients, which is inconsistent with a founder effect and more suggestive of the Clan Genomics hypothesis proposing recently arisen biallelic rare alleles are more likely to be unmasked due to identity-by-descent homozygosity (Lupski, 2021;Lupski et al., 2011).
The ROR2 gene comprises nine exons. Disease-causing variants were more frequent in exons 5, 6, and 9, usually affecting the extracellular domains, though variants affecting the tyrosine kinase domain were also identified ( Figure 1A, Table 1). None of the variants modified interdomain regions which is consistent with previous studies showing that variants affecting interdomain regions can act as gain-of-function (GoF) alleles and cause brachydactyly Type B (Schwabe et al., 2000). Considering all singlenucleotide variants in ROR2 described in patients with AR-RS (Table S1) most of them (22/36) occurred at CpG nucleotides. Cytosine residues in CpG dinucleotides might undergo modifications such as methylation, deamination, and halogenation that can contribute to the formation of mutational hotspots (Sassa et al., 2016). The preponderance of alleles involving CpG is also consistent with Clan Genomics and the derivation of the allele as a new mutation in antecedent generations of the clan that is then brought to homozygosity by IBD (Lupski, 2021;Lupski et al., 2011  Phenotypic annotation grid of phenotypes of all subjects and significantly similar known disease genes. To interpret and understand biology of phenotypes driving semantic similarity in these analyses, human phenotype ontology (HPO) terms associated with all subjects and significantly similar known disease genes were annotated and visualized in a gridded array format. Red indicates presence of a phenotype while gray represents absence or not reported. Probands and significantly similar known disease genes are labeled to the right (italicized gene symbols) and are ordered by HAC. The frequency of each phenotype in probands from this cohort is shown on top of the grid (HSANIA; MIM# 162400). Whether this specific SPTLC1 exonic deletion allele behaves as a LoF or GoF mutation remains to be explored.
According to ACMG/AMP (Richards et al., 2015), 11 variants were classified as pathogenic, six as likely pathogenic and 10 as uncertain significance. The variants classified as uncertain failed PM1 criteria for being out of mutational hotspots. In our cohort, diseasecausing variants were identified throughout the gene except for exons 2 and 4, the smaller ROR2 exons. Therefore, we did not find evidence of mutational hotspots in ROR2 and so it seems that this PM1 classification criteria is not useful for ROR2 variant classification.
Phenotypic analysis comparing missense variant alleles to LoF variants showed minor differences as depicted in Figure 5. We also showed that subjects with ROR2 variants clustered with phenotypes associated with other non-ROR2 gene forms of the syndrome confirming the identity of Robinow syndrome as a single syndrome with genetic heterogeneity and confirming that disruption of this pathway leads to a specific group of phenotypes (Figure 4).
F I G U R E 5 Phenotypic analysis of subjects with biallelic missense variants and LoF variants. Prevalence (0-1.0) of phenotypes in subjects with biallelic missense variants (A1, A2, A9, A15, A16, A19, A21), biallelic LoF variants (A3, A5, A7, A10, A14, A20), all subjects (N = 22), and subjects published in Mazzeu et al., 2007 (N = 37) is displayed by heatmap. Probands with other mutation types were not included in this analysis because of their limited numbers (N < 3). Within the heatmap, red indicates a higher prevalence while blue indicates lower prevalence; light gray indicates these specific data are not available. The phenotypes are ordered by dendrogram shown on the left based on hierarchical agglomerative clustering (HAC) analysis. A prevalence key is provided on the right This is in accordance with our recent phenotypic analysis of dominant RS showing that ROR2-RS was closely clustered with other gene forms of the syndrome .
The overall phenotype of the patients reported herein is in accordance with the previous clinical characterization of AR-RS.
Though some discrepancies were observed in relation to the report of Mazzeu et al. (2007), most clinical signs had similar frequencies in both studies (Tables 2 and S2, Figure 5). Minor discrepancies appeared more evident in clinical signs with mild clinical impact that might have been unreported or overlooked but could still be present.
Clinical signs present in all patients (prominent forehead, hypertelorism, short nose, abnormality of the nasal tip, brachydactyly, mesomelic limb shortening, short stature, and micropenis), as well as those present in more than 75% of subjects (midface retrusion, wide nasal bridge, anteverted nares, downslanted mouth corners, bifid tongue, abnormalities of the dentition, short palms, clinodactyly, hemivertebrae, and rib fusion), should be considered when evaluating variants of uncertain significance in ROR2.
All skeletal changes (short stature, brachydactyly, clinodactyly, mesomelia, rib fusion, and hemivertebrae) had frequencies above 75%. Craniofacial characteristics were also consistent between different patients, including a prominent forehead, hypertelorism, midface retrusion, wide nasal bridge, short nose, abnormality of the nasal tip, anteverted nares, and downturned corners of mouth likely providing the recognizable pattern allowing clinical diagnosis (Table 2).
As with many craniofacial disorders, facial characteristics become attenuated with age in RS patients. We have followed up five patients through adulthood. The typical facial characteristics are very prominent in early childhood, but become less pronounced in adulthood ( Figure 2b). An important consideration in the diagnosis of AR-RS is the characterization of the skeletal defects, considering their high prevalence. Therefore, thorough radiological documentation is essential for clinical diagnosis and management. As a diagnostic tool, the most important findings are mesomelia, brachydactyly, rib fusions, and hemivertebrae, as depicted in Figure 6. The variable severity of the vertebral defects is remarkable, some patients having a single hemivertebrae while others have all vertebrae involved with a major impact on prognosis ( Figure 6). The presence of rib fusions is highly suggestive of AR-RS diagnosis. Despite the previous report of two patients with AR-RS without rib fusions (Aglan et al., 2015;Mehawej et al., 2012), also absent in Patients A12, A20, and A22, other diagnoses should also be considered, including other forms of the syndrome. Scoliosis is also a common finding that In our cohort it has been described in 69% of the patients. Patients without scoliosis were usually evaluated at a very young age except for patient A5, an adult woman.
Brachydactyly was also described in all patients, although it might consist of a minimal shortening of distal phalanges or even absence of distal and medial phalanges, as shown in Figure 6.
Individuals of all ages had short stature and the final height of five adults, both females, and males, ranged between 128 and 145 cm (<3rd centile) in our cohort.
Genital anomalies do not have a major impact on female patients but are a major concern for males. At birth, the penis can be extremely small and buried, often accompanied by cryptorchidism (50%), requiring chromosomal confirmation of the genetic sex (Gerber et al., 2021). Psychological follow-up is recommended. The