ERF‐related craniosynostosis: The phenotypic and developmental profile of a new craniosynostosis syndrome

Mutations in the ERF gene, coding for ETS2 repressor factor, a member of the ETS family of transcription factors cause a recently recognized syndromic form of craniosynostosis (CRS4) with facial dysmorphism, Chiari‐1 malformation, speech and language delay, and learning difficulties and/or behavioral problems. The overall prevalence of ERF mutations in patients with syndromic craniosynostosis is around 2%, and 0.7% in clinically nonsyndromic craniosynostosis. Here, we present findings from 16 unrelated probands with ERF‐related craniosynostosis, with additional data from 20 family members sharing the mutations. Most of the probands exhibited multisutural (including pan‐) synostosis but a pattern involving the sagittal and lambdoid sutures (Mercedes‐Benz pattern) predominated. Importantly the craniosynostosis was often postnatal in onset, insidious and progressive with subtle effects on head morphology resulting in a median age at presentation of 42 months among the probands and, in some instances, permanent visual impairment due to unsuspected raised intracranial pressure (ICP). Facial dysmorphism (exhibited by all of the probands and many of the affected relatives) took the form of orbital hypertelorism, mild exorbitism and malar hypoplasia resembling Crouzon syndrome but, importantly, a Class I occlusal relationship. Speech delay, poor gross and/or fine motor control, hyperactivity and poor concentration were common. Cranial vault surgery for raised ICP and/or Chiari‐1 malformation was expected when multisutural synostosis was observed. Variable expressivity and nonpenetrance among genetically affected relatives was encountered. These observations form the most complete phenotypic and developmental profile of this recently identified craniosynostosis syndrome yet described and have important implications for surgical intervention and follow‐up.

mild exorbitism and malar hypoplasia resembling Crouzon syndrome but, importantly, a Class I occlusal relationship. Speech delay, poor gross and/or fine motor control, hyperactivity and poor concentration were common. Cranial vault surgery for raised ICP and/or Chiari-1 malformation was expected when multisutural synostosis was observed. Variable expressivity and nonpenetrance among genetically affected relatives was encountered. These observations form the most complete phenotypic and developmental profile of this recently identified craniosynostosis syndrome yet described and have important implications for surgical intervention and follow-up.
ERF-related craniosynostosis was first described in 2013 in 12 unrelated families accounting for 7.1% of a cohort of 127 patients with undiagnosed clinically syndromic craniosynostosis, and 2.9% of a total cohort of 412 undiagnosed patients with syndromic or nonsyndromic craniosynostosis (Twigg et al., 2013). More recently, the overall prevalence in all syndromic craniosynostosis has been estimated at 2% and in clinically nonsyndromic craniosynostosis at 0.7% (Wilkie et al., 2017). It appeared to be associated particularly with sagittal and lambdoid synostosis, but also multisutural craniosynostosis and pansynostosis. Chiari-1 malformations appeared to be more common, and there was a relatively high risk of pathologically raised intracranial pressure (ICP), behavioral problems, and speech and language delay. A presumptive diagnosis of Crouzon syndrome had been made for many of these patients. Examples of variable expression and nonpenetrance were also reported (Twigg et al., 2013).
Since the initial report, two patients with ERF mutations have been described in a cohort of 40 patients with sagittal or multisutural synostosis (Chaudhry et al., 2015) and three patients with ERF mutations have been described in a cohort of 309 individuals with craniosynostosis who did not have a prior molecular diagnosis (Lee et al., 2018). A recent exome sequencing study of 291 parent-offspring trios with nonsyndromic midline craniosynostosis reported a novel frameshift ERF mutation in a father and his two offspring each of whom had nonsyndromic metopic synostosis (Timberlake et al., 2017). Elsewhere, a specific heterozygous ERF missense p.(Y89C) substitution has been found to cause Chitayat syndrome in four unrelated probands and one parent with hyperphalangism, characteristic facies, hallux valgus, and bronchomalacia (Balasubramanian et al., 2017). None was noted to have craniosynostosis although only one had been assessed by cranial computed tomography (CT), at 5.5 years of age.
Here, we report our experience of 16 unrelated probands and 20 additional family members with heterozygous ERF mutations confirming that they contribute significantly to the craniosynostosis caseload, and highlight particular issues of importance in the clinical management of patients and their wider families.

| Editorial policies and ethical considerations
Patients known to the U.K. supra-regional craniofacial units at Great Ormond Street Hospital (London), the John Radcliffe Hospital (Oxford), and Birmingham Children's Hospital and who had been diagnosed since the initial description of ERF-related craniosynostosis (Twigg et al., 2013) (Twigg et al., 2013).
Common to all three services, genetic investigation for patients with multisuture or suspected syndromic craniosynostosis and without a known familial etiology includes screening for mutations in FGFR1 (Exon 7), FGFR2 (Exons 8 and 10), FGFR3 (Exons 7 and 10) and TWIST1 (Exon 1) sequencing and multiplex ligation-dependent probe amplification as a minimum. Those with normal results have further testing of FGFR2 (Exons 3, 5, 11, 14-17), EFNB1, ERF, TCF12, IL11RA and in some instances array-CGH chromosome testing (although exact protocols vary slightly between centers and clinicians).
Following a diagnosis of ERF-related craniosynostosis, a family history was obtained for all probands. Parents were offered genetic counseling, testing for the mutation and, where indicated, mutation screening was offered for other "at-risk" family members, in line with standard clinical genetics practice.
All probands and related children identified with a familial ERF mutation were evaluated through the craniofacial service. In the majority this included CT head scanning with three-dimension reconstruction to visualize the cranial vault and, in some cases magnetic resonance imaging of the brain depending on departmental protocol.
Those with confirmed craniosynostosis were evaluated by a multidisciplinary team drawn from plastic and maxillofacial surgery, neurosurgery, otolaryngology, dental surgery, developmental pediatrics, audiology, ophthalmology, speech and language therapy, psychology, and clinical genetics.
After detailed review of the medical history, including the identification of any potentially confounding variables, a clinical evaluation for craniofacial dysmorphology was completed. Other noncraniofacial phenotypic features were noted. Radiological review for Chiari-1 malformation was undertaken in each case. Ophthalmological assessment included visual acuity, fundoscopy and, at one center, visual evoked potentials. Visual impairment was defined as worse than 0.3 LogMAR with refractive correction and both eyes open. Audiological assessment included a hearing test and otoscopy. Language assessments were selected from a battery of standardized tests used routinely in the United Kingdom and based on the child's age (Wiig, Secord, & Semel, 2006a, 2006bZimmerman, Pond, & Steiner, 2009). Speech was assessed using a nonstandardized assessment (Grunwell, 1995).
Speech, expressive and receptive language were rated as being normal or with mild, moderate or severe delay determined by the test scores obtained. Similarly, age-specific gross and fine motors skills were evaluated by developmental pediatricians with the severity of delay summarized as before. Developmental pediatricians and/or child psychologists evaluated learning and behavior. While the reports on these evaluative domains were complex and bespoke, for the purpose of this study the findings were stratified by severity of learning and behavioral difficulty. The systematic assessment concluded with a multidisciplinary debriefing and data were recorded prospectively on a standardized proforma. Developmental assessments were carried out on a regular basis for all probands as part of their clinical evaluation and follow-up. Family members were assessed on an ad hoc basis and on the basis of their self-reported clinical history.

| RESULTS
A total of 16 apparently unrelated probands with suspected pathogenic ERF mutations were identified. Of the genetically-related family members identified by family tree who consented for genetic testing, an additional 20 individuals with ERF mutations linked to those of the probands was found. The ERF mutations and associated phenotypes are summarized in Table 1. The case history for each individual is summarized in Supporting Information.

| Genotype
Thirteen different heterozygous mutations (eight of which are reported for the first time) were identified in the 16 families comprising one mutation within the translation initiation codon, three nonsense mutations, three frame-shifting mutations predicted to result in premature protein truncation, and six mutations predicted to result in missense substitutions (Figure 1).
Only one patient (P1) was confirmed by parental testing to have a de novo mutation. In two patients (P35, P36) the mutations are suspected to be de novo on clinical grounds but parental samples are awaited. In three families (K6, 10, 12) the parents were either unavailable or had declined testing. Of these, one parent was suspected to be affected in two families (one father because of exorbitism and one mother because of her facial appearance and history of mild learning difficulties) but have been classified as unknown for the purpose of this study. In one additional family (K2), although the parents were not available for assessment or testing, the available parental history and the identification of affected maternal half-siblings infers maternal inheritance. In nine families one parent was found to carry the ERF mutation (five fathers; four mothers) but the grandparents and other relatives on that side had not been tested. In one further family where nine individuals have been found to carry the ERF mutation to date, it was traced back to the proband's maternal grandmother. Overall we observed 15 maternal transmissions (including three inferred) and seven paternal.
(K401Efs*10)) were each shared by two families in our cohort. From our results we are not able to distinguish whether these mutations are recurrent or originate from a founder relative. Although the respective probands are not known to be related through available family histories, in each instance they originate from the same broad geographical area. Importantly, only three of our 16 probands (P20, P25, and P35) had a sufficiently abnormal head shape and/or facial appearance to raise the suspicion of a craniofacial syndrome in the neonatal period.

| Craniosynostosis
A further patient (P23) presented during infancy (Table 2). For the majority, the onset of craniosynostosis was insidious and progressive. This is illustrated by Patient 14 in whom early testing and monitoring was undertaken owing to his family history and in whom an evolving pattern of craniosynostosis was observed. In the neonatal period his head shape was normal and his fontanelle and sutures were patent    Abbreviations: DP = distal phalanx; F = female; K = kindred; M = male; mo = months; N/K = not known; OHT = orbital hypertelorism; P = patient. Note. Unconfirmed clinical diagnoses are in italics. Where the parental origin has been stated it has been confirmed by direct mutation testing. Where parental samples were not available the origin has been recorded as "unknown". For the parents of a proband who carried the same ERF mutation, the origin in them was generally "unknown" as the grand-parental samples were not available. In K2 the parental origin could be inferred. In two patients the mutation was suspected to be de novo on clinical grounds but parental samples were not available.
Mutation nomenclature is based on NM 006494 (cDNA) and NP 006485 (protein).
The bold text represents the Probands.
clinically. CT scanning at 8 months of age showed bisquamosal synostosis and the inferior coronal sutures were felt to be indistinct radio-    Table 1.

| Developmental assessment
The developmental assessment profiles, as summarized in Table 3, show that most of the 16 probands demonstrated ophthalmological, audiological, neurological, speech and language, or behavioral anomalies. Ten of the 14 (71%) probands old enough to assess exhibited speech and/or receptive and expressive language delay, which generally responded well to therapy. In addition, seven of the family members reported speech delay and/or required speech and language therapy in childhood.
Ten of the 16 (63%) probands exhibited poor gross motor and/or fine motor skills with deficits in gross motor control in five subjects, fine motor skills in two subjects and components of both in three subjects. Gross motor delay was a feature noted in the history of only one family member. Poor concentration and/or hyperactivity was observed in four of the 13 (31%) probands over 3 years of age and noted in the history of six of the 19 (32%) family members over 3 years of age. Six of the 12 (50%) probands and seven of the 19 (37%) family members older than 4 years needed support within their mainstream school or nursery. Importantly, of the nine probands with evidence of raised ICP, neurocognitive disturbance was identified in six and audiovisual disturbance in four. However, of the seven probands without evidence of raised ICP, neurocognitive disturbance was identified in four and audio-visual disturbance in three, suggesting that raised ICP was not the causative factor in these features.
Recurrent otitis media was identified in five (31%) probands and was a reported feature in the history of three family members. Associated hearing loss was variable.

| DISCUSSION
We describe 36 previously unreported individuals from 16 kindreds in whom we have found 13 different heterozygous ERF mutations. Only one mutation in our cohort was confirmed to have arisen de novo, with a further two (P35, P36) suspected. The ERF mutation has been confirmed or can be inferred to have arisen from one of the parents in 10 of the probands. Two additional probands have one parent who is suspected to be mildly affected clinically.
Four of the ERF mutations found in our cohort have been reported previously (Twigg et al., 2013). One of those (p.(G299Rfs*9)) was confirmed in our patient to have arisen de novo and is therefore recurrent. For three others (p.R183*, p.K401Efs*10, and p.Q424*), we are unable to exclude the possibility of a founder effect since we have not been able to demonstrate a de novo origin and neither could Twigg et al. (2013) in their earlier cohort (Twigg et al., 2013). Mutations affecting the initiator codon have been reported twice previously (Chaudhry et al., 2015;Twigg et al., 2013) but the underlying nucleotide change in our patient was novel.
In keeping with the earlier findings, the predicted missense mutations in our cohort all occurred in highly conserved residues of the DNA-binding ETS domain of the ERF protein between amino-acids 29 and 106. The six mutations predicted to result in protein truncation were all located further towards the C-terminus and to cause loss of the repressor domain, or ERK interaction and repressor domains, if they did not result in nonsense mediated mRNA decay. Overall the pattern of heterozygous mutations observed is consistent with a predominant haploinsufficiency mechanism of pathogenesis, as previously proposed (Twigg et al., 2013).
The most consistent clinical features of the probands include multisutural synostosis with the Crouzonoid triad of OHT, exorbitism and malar hypoplasia, as well as Chiari-1 malformation, speech and language delay, poor fine and/or gross motor skills, and learning difficulties and/or hyperactivity, in keeping with previous findings (Twigg et al., 2013). While pansynostosis or sagittal and bilambdoid synostosis were the most frequent patterns of suture involvement accounting  for 8 of 24 (33%) and 6 of 24 (25%), respectively, the sutural involvement in our cohort is more diverse than indicated from the initial report. Excluding the seven cases of pansynostosis, the sagittal suture was involved in 11 of 16 patients (69%) while both lambdoid sutures were involved in 7 of 16 (44%) and one lambdoid suture was involved in an additional 4 of 16 cases (25%). At least one coronal suture was involved in a third of cases (unilateral in three and bilateral in two).
Given these findings we recommend a low threshold for testing for ERF mutations in patients with pansynostosis or multisuture synostosis of any pattern but particularly with sagittal and lambdoid involvement.  A notable feature in our cohort has been the relatively subtle change in head shape in many of the patients. We speculate that delayed evolution of the craniosynostosis in patients with ERF mutations may result in preservation of a normal head shape because it develops after the period of very rapid skull growth between the third trimester of pregnancy and the end of the first year of life. Moreover, while facial dysmorphism appears to be a common feature of ERFrelated craniosynostosis, we observed that it is usually symmetrically so. We speculate that the reason for this lies with the predominance of symmetrical synostotic patterns and this may contribute to delayed recognition of the condition.
The associated OHT and exorbitism is similar to that seen in Crouzon syndrome which was the commonest misdiagnosis in our series. It is interesting to speculate that the overlapping facial phenotypes result from a shared downstream constitutive activation of the RAS/MAPK pathway (Twigg & Wilkie, 2015). However, we have observed a number of distinctive differences between the two conditions, aside from the relative delay in presentation discussed above.
Crucially, in the case of ERF-related craniosynostosis mid-facial hypoplasia was typically mild, and in no case was sufficiently severe to merit surgical intervention for airway management, ocular protection or appearance, even in adulthood. With the exception of one patient, all exhibited a Class I occlusal relationship. Moreover, the notably consistent pattern of developmental anomalies including speech and language delay, poor motor skills, and learning difficulties and/or behavioral problems typified by hyperactivity or poor concentration are not typical features of Crouzon syndrome. Encouragingly the speech and language and motor delays improved with supportive interventions. In addition, all the adult ERF mutation carriers were living independently as far as we could establish.
An important observation was that both neurocognitive and audio-visual abnormalities were equally likely among the probands with raised ICP as those without. This would suggest that ICP alone is not solely responsible for these deficits but, rather, they are intrinsic features of the phenotype.
We note that the frequency of neurodevelopmental issues recorded in the adult ERF mutation carrying family members was much lower than expected given the results from the pediatric cohort.
This may reflect a recall bias or alternatively, may suggest that the neurodevelopmental problems exhibit variable penetrance. Interestingly, four children from two kindreds within our cohort have been fostered or taken into social services care for neglect and in both families one biological parent carries the ERF mutation. We speculate that unrecognized learning and behavioral issues in unascertained adult ERF mutation carriers may have contributed to educational underachievement and/or social issues that may predispose to this occurrence.
Only two individuals in the entire cohort had sensory processing problems or features suggestive of autistic spectrum disorder, one of whom had a coincidental common recurrent 16p13.11 duplication which is a recognized neurosusceptibility variant enriched in patients with autism.
Overall, the observations in our cohort suggest that children with ERF mutations are likely to benefit from closer general pediatric surveillance and early interventions for their development and behavioral issues.
One patient had a radio-ulnar synostosis, cervical vertebral  (Balasubramanian et al., 2017). Chitayat syndrome is also associated with facial dysmorphism (of a nature strikingly similar to that observed in our cohort), speech and language and motor delay (which is, again, similar in pattern to that observed in our cohort), and significant respiratory compromise from early childhood (which we did not observe in our cohort).
Although none of the reported patients with Chitayat syndrome was considered to have craniosynostosis, only one had been assessed by cranial CT scan at 5.5 years of age. As our cohort demonstrates, the absence of a clearly abnormal skull shape in patients with ERF mutations does not exclude the possibility of craniosynostosis.
Somatic loss-of-function mutations in ERF have been reported in tumors including prostate, stomach and colorectal adenocarcinomas and Ewing's sarcoma (Bose et al., 2017;Huang et al., 2017) at frequencies of 3-5%. In several instances the somatic ERF mutations found in tumor tissue have been identical to the constitutional ERF mutations found in patients with ERF-related craniosynostosis. There are other precedents for genes where identical mutations have been observed somatically in tumors and constitutionally in a variety of craniosynostosis and other dysmorphic syndromes, including genes encoding other components of the RAS-MAPK pathway. As a general principle oncogenesis is a multistep process with progression dependent on the sequential accumulation of mutations within the tissue cells, such that the presence of a single constitutional mutation is not necessarily associated with a substantially increased cancer risk. We did not seek detailed cancer family histories in our cohort and have not undertaken extended testing to identify ERF carriers in the wider family of our cohorts and so we cannot address whether there is an increased cancer risk in these families.
In conclusion, ERF-related craniosynostosis is a newly recognized disorder characterized by multisutural synostosis (with a predilection for pansynostosis or sagittal and bilambdoid involvement), facial dysmorphism with a mild Crouzonoid phenotype, Chiari-1 malformation, delays in language development which generally resolve, behavioral abnormalities in the attention deficit and hyperactivity spectrum and mild learning disabilities which can usually be managed with support in mainstream education. The craniosynostosis may develop after birth in the first few years, evolve insidiously, and be associated with a relatively normal head shape. Cascade screening to identify children at risk in early childhood and close follow-up of those identified as mutation carriers is strongly recommended to minimize the risk of serious visual sequelae of raised ICP and for early pediatric intervention for expressive and/or receptive language delay and behavioral issues. We advocate a low threshold for testing for ERF mutations in  Abbreviations: + = mild; ++ = moderate; +++ = severe; A&T = adenotonsillectomy; Co = conductive hearing loss; Ex = expressive language delay; ICP = intracranial pressure; OME = recurrent otitis media with effusions; OSA = obstructive sleep apnea; P = papilledema; Re = receptive language delay; SN = sensorineural hearing loss; Sp = speech delay; U/K = unknown.
The bold text represents the Probands.
patients with multisutural or pansynostosis, or patients presenting with a Crouzonoid appearance and negative FGFR genetic screen.