Dissection of contiguous gene effects for deletions around ERF on chromosome 19

Heterozygous intragenic loss‐of‐function mutations of ERF, encoding an ETS transcription factor, were previously reported to cause a novel craniosynostosis syndrome, suggesting that ERF is haploinsufficient. We describe six families harboring heterozygous deletions including, or near to, ERF, of which four were characterized by whole‐genome sequencing and two by chromosomal microarray. Based on the severity of associated intellectual disability (ID), we identify three categories of ERF‐associated deletions. The smallest (32 kb) and only inherited deletion included two additional centromeric genes and was not associated with ID. Three larger deletions (264–314 kb) that included at least five further centromeric genes were associated with moderate ID, suggesting that deletion of one or more of these five genes causes ID. The individual with the most severe ID had a more telomerically extending deletion, including CIC, a known ID gene. Children found to harbor ERF deletions should be referred for craniofacial assessment, to exclude occult raised intracranial pressure.


| BRIEF REPORT
The gene ERF, first described in 1995, is located on chromosome 19q13.2 and encodes a member of the ETS family of transcription factors that acts as a key negative regulator of ERK1/2, effectors of the RAS-MAP kinase pathway (von Kriegsheim et al., 2009;Lavoie et al., 2020;Le Gallic et al., 2004;Polychronopoulos et al., 2006;Sgouras et al., 1995). Disease-causing heterozygous loss-of-function variants of ERF were first described in 2013, in 12 families segregating features of a newly recognized syndrome (termed ERF-related craniosynostosis or craniosynostosis type 4, OMIM# 600775), characterized by premature fusion of the cranial sutures (craniosynostosis), hypertelorism, and mild midface hypoplasia (Twigg et al., 2013). Confirmatory case reports have followed (Chaudhry et al., 2015;Korberg et al., 2020;Lee et al., 2018;Provenzano et al., 2021;Timberlake et al., 2017;Tønne et al., 2020;Yoon et al., 2020), and the clinical features of the disorder were further delineated and summarized in 16 additional families by Glass et al. (2019). In addition to craniosynostosis and facial dysmorphism, additional frequently associated features included Chiari-1 malformation, speech and language delay, poor gross and/or fine motor control, hyperactivity, and poor concentration. Importantly, craniosynostosis was often postnatal in onset, insidious, and progressive with subtle effects on head morphology, resulting in late median age at presentation of 42 months among the probands and, in some instances, permanent visual impairment occurred owing to unsuspected raised intracranial pressure (ICP) (Glass et al., 2019).
To our knowledge 26 different heterozygous variants in 39 unrelated probands/families have been described in ERF-related craniosynostosis. The pattern of ERF variants (eight frameshifts, three nonsense, three splice-site, three disrupting the initiation codon, and nine missense localized to the highly conserved DNA-binding domain) is strongly suggestive of a haploinsufficiency mechanism, and this is supported by functional studies of two of the missense variants that demonstrated loss of DNA binding (Twigg et al., 2013).
Consistent with this, ERF is depleted of loss-of-function variants in the gnomAD database, with an observed/expected ratio of 0.06 (confidence interval 0.02-0.26) and a probability of loss-of-function intolerance (pLI) score of 0.99 (Karczewski et al., 2020).
Although partial or complete heterozygous deletions of ERF would be predicted to be associated with a similar pathogenic effect, none has previously been specifically reported. Neither the analysis of ERF dosage using multiplex ligation-dependent probe amplification (MLPA) in 276 samples (Twigg et al., 2013) nor the capture-based targeted resequencing in an additional 156 samples from craniosynostosis cases without a genetic diagnosis (SRFT, unpublished data) identified any pathogenic copy number variant (CNV) affecting ERF, indicating either that such deletions are not a frequent cause of craniosynostosis, or that they could produce a more complex/severe syndrome. A few patients have been reported with large chromosome 19q13.2 deletions apparently including ERF, although the phenotype was often confounded by the inclusion of RPS19, which lies approximately 375 kb centromeric to ERF, in patients with Diamond-Blackfan anemia (Farrar et al., 2011;Kuramitsu et al., 2012;Quarello et al., 2008;Yuan et al., 2016) or ATP1A3, approximately 250 kb centromeric to ERF, in a case with a neurological disorder (Kessi et al., 2018); the names and positions of genes around ERF are given in Figure 1a and Table S1. The majority of individuals with large (≥333 kb) deletions were reported to have combinations of facial dysmorphism and/or macrocephaly, but "mild craniosynostosis" was noted in one case (Yuan et al., 2016). Here, we describe the identification of six smaller (32-314 kb) deletions at the ERF locus, four of them characterized by whole-genome sequencing (WGS) at base-pair resolution, and two by array comparative genomic hybridization (aCGH). To identify additional individuals harboring CNVs at the ERF locus, independently of the phenotype, we performed bioinformatic screening of all the 74,008 genomes of participants from families affected with rare disorders available in the 100kGP (main programme v10; RR187). This revealed two additional deletions around ERF (Subjects 2 and 3; Table 1, Figures 1a, and S1). The deletion in Subject 2 (264 kb; Figure 1a) had previously been detected by array CGH when it was reported as having arisen de novo; however, closer inspection of the paternal WGS data suggested low levels of mosaicism based on the presence of a few abnormal reads supporting the deletion ( Figure S1B). Using the same method as for Subject 1 ( Figure S1E), we estimated that approximately 5% of paternal blood cells were mosaic. The deletion in Subject 3 (31.7 kb; Figure 1a) was inherited from his father ( Figure S1C), with no indication of mosaicism ( Figure S1E). Following informed consent, we obtained DNA samples from each of the family trios and confirmed the previously deduced molecular nature of each deletion by breakpoint-PCR (Table S2) and dideoxy-sequencing ( Figure S2). No other causative pathogenic change was identified by 100kGP for any of Subjects 1-3.
In parallel, as part of a clinical genetics investigation, a further de novo deletion including ERF was identified by aCGH in Subject 4 (Table 1 and Figure 1a); following informed consent, WGS was carried out using the proband's DNA to characterize the breakpoints, demonstrating a 265 kb deletion ( Figure S1D). There was no evidence of a breakpoint-PCR product in samples from either of the parents of Subject 4, in whom the deletion was quantified as 50%, indicating a de novo origin at conception ( Figure S2). Segregation analysis of a rare SNV (chr19:g.42783791G>C, hg19) located within the deleted region established that the deletion arose on the paternal allele (data not shown).
Toward a more comprehensive analysis of genotype-phenotype correlations, additional cases harboring heterozygous deletions around ERF that had been identified by aCGH were retrieved from the DECIPHER database (Firth et al., 2009) (Subject 5,~265 kb; Subject 6,~51 kb) (Figure 1a), and the respective clinicians/scientists were contacted. However, in Subject 6, an additional confounding chromosomal abnormality was present in the proband (Table 1).
Similarly to Subject 1, this rendered it difficult to disentangle the relative contributions of the different chromosome imbalances to the phenotype. Hence, to undertake a detailed genotype-phenotype correlation of deletions surrounding ERF, we focused on Subjects 2-5 only. The major clinical features of these four subjects are summarized in Table 1; see Supplementary Case Reports for more detailed information.
Based on the relative size and extent of each deletion, and the degree of associated intellectual disability, we propose that the ERF  deletions belong to three categories. First, in the case of the smallest deletion (Subject 3, 31.7 kb), which is constitutionally inherited from the father, neither individual has ID. This deletion includes three genes (a small portion of the ZNF526 3′-untranslated region (UTR), and whole gene deletion of GSK3A and ERF), suggesting that possessing a single copy of these genes is not associated with ID.
Hence, it cannot be assumed that heterozygous deletion of ATP1A3 would cause moderate ID. Three of the five genes in the extended deletion interval (ATP1A3, GRIK5, and POU2F2) have a pLI score greater than 0.9 (Table S1), indicating evolutionary constraint against loss-of-function alleles (Karczewski et al., 2020). Both Subjects 2 and 5 had a similar degree of moderate ID but were discordant for some other clinical features (notably Jeavons syndrome-type epilepsy in Subject 2). Hence we propose that haploinsufficiency for one or a combination of genes in the ATP1A3-DEDD2 interval causes moderate ID.
In the third category, the deletion in Subject 4, who has moderate-severe ID and autistic spectrum disorder (ASD), extended more telomeric than any of the other deletions, to encompass the gene CIC. Intragenic mutations of CIC were previously described in both severe ID and ASD (Guo et al., 2019;Lu et al., 2017), which is likely to explain the more severe ID phenotype in this case.
Although our observations must be regarded as provisional given the small number of cases identified, they represent the beginnings of a map of genotype-phenotype correlations for deletions encompassing ERF. Importantly, each deletion appeared unique, with no evidence for a recurrent breakpoint mechanism. In the four cases characterized at the molecular level, most breakpoints occurred in, or in close proximity to, regions rich in repetitive elements, especially Alu elements ( Figure S3); in three of these, the sequences at the breakpoints show homology of only 2-3 nucleotides (cases 1, 2, and 4; Figure S3), indicating nonhomologous end-joining as the most likely mechanism. In Subject 3, however, nonallelic homologous recombination between two Alu elements (AluY and AluSx) evidently occurred ( Figure S3). Of note, the aCGH originally used to identify the deletion in Subject 4 suggested a smaller extent of deletion, not including CIC, in contrast to the larger 265 kb deletion determined by WGS. Moreover, the aCGH in the parents of Subject 2 had suggested that the deletion arose de novo in the child, whereas WGS demonstrated a low level of mosaicism in the father. These two examples illustrate the added value provided by WGS, both for refining molecular diagnoses and for greater precision in recurrence risks.
From a clinical point of view, deletion or functional disruption of the ERF gene itself is likely to account for the mild dysmorphic facial features (including variable hypertelorism, exorbitism, and macrostomia) in these individuals (Figure 1b). Importantly, ERF haploinsufficiency may predispose to an insidious presentation of craniosynostosis and raised intracranial pressure, without any noticeable change in skull shape (Glass et al., 2019;Twigg et al., 2013)