Heterozygous Variants in FREM2 Are Associated with Mesiodens, Supernumerary Teeth, Oral Exostoses, and Odontomas

Background: Supernumerary teeth refer to extra teeth that exceed the usual number of dentitions. A mesiodens is a particular form of supernumerary tooth, which is located in the premaxilla region. The objective of the study was to investigate the genetic etiology of extra tooth phenotypes, including mesiodens and isolated supernumerary teeth. Methods: Oral and radiographic examinations and whole-exome sequencing were performed on every patient in our cohort of 122 patients, including 27 patients with isolated supernumerary teeth and 94 patients with mesiodens. A patient who had multiple supernumerary teeth also had odontomas. Results: We identified a novel (c.8498A>G; p.Asn2833Ser) and six recurrent (c.1603C>T; p.Arg535Cys, c.5852G>A; p.Arg1951His, c.6949A>T; p.Thr2317Ser; c.1549G>A; p.Val517Met, c.1921A>G; p.Thr641Ala, and c.850G>C; p.Val284Leu) heterozygous missense variants in FREM2 in eight patients with extra tooth phenotypes. Conclusions: Biallelic variants in FREM2 are implicated in autosomal recessive Fraser syndrome with or without dental anomalies. Here, we report for the first time that heterozygous carriers of FREM2 variants have phenotypes including oral exostoses, mesiodens, and isolated supernumerary teeth.


Introduction
Supernumerary teeth refer to extra teeth that exceed the usual number in dentitions, which either erupt or remain unerupted [1][2][3]. Supernumerary teeth are rare in primary dentition [4]. The prevalence of supernumerary teeth has been reported to be 0.04-3.8%, depending on the studied population and the method of selecting the samples [2,3,[5][6][7].

Patients
Oral and radiographic examinations were performed on our cohort of 122 patients affected by various kinds of isolated extra tooth phenotypes, which included 94 patients with isolated mesiodens and 27 patients (10 females; 17 males) with isolated supernumerary teeth. Regarding 94 patients with mesiodens, 64 (68.1%) were males and 30 (31.9%) were females. Seventy-eight patients (82.9%) had single mesiodens, while 16 (17.1%) of them had double mesiodentes. The orientation of the mesiodens was noted in 64 mesiodentes of 54 patients; 43 (67.1%) of them had a normal orientation, 20 (31.3%) were inverted ones, and 1 (1.6%) had a transverse orientation. The eruption status was noted in 60 mesiodentes of 49 patients: 32 (53.3%) were erupted and 28 (46.7%) were unerupted. Regarding the patients with isolated supernumerary teeth that were not mesiodens, the information of the supernumerary teeth was not available. A patient with supernumerary teeth also was found to have odontomas. Inclusion criteria of the study were patients with isolated extra tooth phenotypes, including supernumerary teeth or mesiodens. Exclusion criteria of the study were patients with the usual number of or fewer teeth or syndromic supernumerary teeth.

Whole-Exome Sequencing, Mutation Analysis, and Bioinformatic Analyses
The genomic DNA of all patients was isolated from saliva using the Oragene-DNA (OG-500) Kit (DNA Genotek, CANADA). One hundred and twenty-two patients with isolated mesiodens or isolated supernumerary teeth underwent the whole-exome sequencing (WES) service from Macrogen Inc., Seoul, Korea. The WES service employed SureSelect V6 kit (PR7000-0152; Agilent Technologies, Santa Clara, CA, USA), including untranslated regions, resulting in a 80× read depth on average. We used GATK3.8 best practices to identify variants, while BWA-mem was used to map the raw sequencing reads (FASTQ) with the GRCh38+decoy human genome reference sequence. In order to forecast deleterious mutational effects, the reported variants from GATK3.8 were loaded into the Variant Effect Predictor (VEP) with the plugin Database of Nonsynonymous Functional Prediction (db-NSFP), which is a mutation prediction database containing >30 variants' impact prediction algorithms. The predicted variants of every patient were reported in a variant calling file (VCF file). We collected these files and stored them in our in-house database, which allowed us to query pathogenic variants according to different segregation modes. In addition, variant allele frequencies were determined by comparing them against well-known public variant databases, including gnomAD (https://gnomad.broadinstitute.org (accessed on 8 January 2023), 1000 Genomes (https://www.genome.gov/27528684/1000-genomes-project (accessed on 8 January 2023), GenomeAsia (https://browser.genomeasia100k.org), and the Thai Reference Exome database (T-Rex; https://genomicsthailand.com/Search/T-Rex (accessed on 8 January 2023). The following factors were taken into account when we were prioritizing the variants: (1) the rarity of the allele; (2) the CADD score > 15; (3) localization to or near an essential functional region of the protein. Then, if familial samples were available, the validated variants of interest were examined for segregation using PCR-based amplification and Sanger sequencing. The effects of mutations were predicted by MutationTaster (https://www.mutationtaster.org (accessed on 8 January 2023)). PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/ (accessed on 8 January 2023)), and SIFT (https://sift.bii.a-star.edu.sg (accessed on 8 January 2023)).

Results
We identified a novel (c.8498A>G; p.Asn2833Ser) and six heterozygous missense variants (c.1603C>T; p.Arg535Cys (rs201457616), c.5852G>A; p.Arg1951His (rs201806885), c.6949A>T; p.Thr2317Ser, c.1549G>A; p.Val517Met (rs566143955), c.1921A>G; p.Thr641Ala (rs116802472), and c.850G>C; p.Val284Leu (rs770004356) in FREM2 in five patients with mesiodens, two patients with supernumerary teeth, and a patient with supernumerary teeth and odontomas (         The heterozygous missense variant c.1603C>T; p.Arg535Cys was identified in patient 1, who had a mesiodens and torus palatinus. According to gnomAD, this variant is rare, with a global allele frequency of 0.0001712. It has not been reported in the South Asian population. Its allele frequency in the East Asian population is 0.002164. According to the T-Rex database, this allele was found in 4 out of 2184 alleles (allele frequency = 0.0018315 or 0.18%). The allele frequency of this variant in the study group (0.409836%) is higher than those in the normal Thai population (T-Rex database) and in gnomAD ( Table 2). This variant is predicted to be disease causing, possibly damaging, and damaging by MutationTaster (prob = 0.99999999580175), PolyPhen-2 (0.994), and SIFT (0.001), respectively. The CADD and DANN scores of this variant are 25.4 and 0.9992, respectively (Table 1).  The heterozygous missense variant c.1603C>T; p.Arg535Cys was identified in patient 1, who had a mesiodens and torus palatinus. According to gnomAD, this variant is rare, with a global allele frequency of 0.0001712. It has not been reported in the South Asian population. Its allele frequency in the East Asian population is 0.002164. According to the T-Rex database, this allele was found in 4 out of 2184 alleles (allele frequency = 0.0018315 or 0.18%). The allele frequency of this variant in the study group (0.409836%) is higher The heterozygous missense variant c.5852G>A; p.Arg1951His was identified in patient 2, who had erupted double mesiodens. This variant is extremely rare, with a global allele frequency of 0.000003982 according to gnomAD. It has not been reported in the South Asian population. Its allele frequency in the East Asian population is 0.00005448. This variant has not been reported in the normal Thai population (T-Rex database). The allele frequency of this variant in the study group (0.409836%) is higher than those in the T-Rex and gnomAD databases ( Table 2). This variant is predicted to be disease causing (prob = 0.99712280816916) and possibly damaging (0.873) by MutationTaster and PolyPhen-2, respectively. The CADD and DANN scores of this variant are 19.18 and 0.9985, respectively (Table 1). Patient 3 had double mesiodens; one had erupted and the other had not. He carried a heterozygous missense variant, c.6949A>T; p.Thr2317Ser, in the FREM2 gene. This variant cannot be found in e, LOVD, and HGMD databases (Table 1), but according to the T-Rex database, it is reported in one of two thousand, one hundred, and eighty-four alleles. The allele frequency of this variant in the study group (0.409836%) is higher than those in the normal Thai population (T-Rex database) and gnomAD ( Table 2). This variant is predicted to be disease causing (prob: 0.999998598060573) by MutationTaster. The CADD and DANN scores of this variant are 22.4 and 0.9985, respectively (Table 1). Patient 4, who had an inverted and unerupted double mesiodens, carried a heterozygous missense variant, c.1549G>A; p.Val517Met, in the FREM2 gene. This variant is very rare, with a global allele frequency of 0.00002012 according to gnomAD. This variant has not been reported in the South Asian population. The allele frequency of this variant in the East Asian population is 0.0002730. This allele is not reported in T-Rex database. The allele frequency of this variant in the study group (0.409836%) is higher than those in T-Rex and gnomAD databases ( Table 2). This variant is predicted to be   The heterozygous missense variant c.1921A>G; p.Thr641Ala was identified in patient 5, who had two mandibular supernumerary teeth and taurodontism, and patient 6, who had a mesiodens, agenesis of the maxillary permanent third molars, and taurodontism of the maxillary left permanent second molar. According to gnomAD, this variant is very rare, with a global allele frequency of 0.00002829. The allele frequency of this variant in the East Asian population is 0.0003510. It has not been reported in the South Asian population. According to the T-Rex database, this allele was found in 5 out of 2184 alleles (allele frequency = 0.00228938 or 0.22%). The allele frequency of this variant in the study group (0.409836%) is higher than those in T-Rex and gnomAD databases ( Table 2). This variant is predicted to be disease causing (prob: 0.999961580268265), possibly damaging (0.474), and damaging (0.008) by MutationTaster, PolyPhen-2, and SIFT, respectively. The CADD and DANN scores of this variant are 22.1 and 0.9889, respectively. In addition to the FREM2 variant, patient 5 also had a heterozygous missense variant (c.136A>T; p.Met46Leu; chr1 g.68659881T>A; rs368633951) in WLS, and this has been reported [23]. This allele is rare, with a global allele frequency of 0.0000398. Patient 7 who had two supernumerary mandibular premolars, unseparated roots of the permanent molars, torus palatinus, and torus mandibularis and carried a heterozygous missense variant, c.850G>C; p.Val284Leu, in the FREM2 gene. This variant is rare with a global allele frequency of 0.0002231 according to gnomAD. The allele frequency of this variant in the East Asian population is 0.003158. This variant has not been reported in the South Asian population. According to the T-Rex database, this allele is found in 4 out of  (Table 2). This variant is predicted to be disease causing (prob: 0.999963783102466) by MutationTaster. The CADD and DANN scores of this variant are 23.8 and 0.9953, respectively. Patient 8, who had multiple supernumerary teeth and multiple odontomas, carried a heterozygous missense variant, c.8498A>G; p.Asn2833Ser, in the FREM2 gene ( Figure 6). This variant was not found in the T-Rex, gnomAD, LOVD, and HGMD databases; therefore, it is considered to be novel. This mutation is predicted to be disease causing (0.999943878525573) and pathogenic (0.9953) by MutationTaster. The CADD and DANN scores of this variant are 17.14 and 0.9802, respectively (Tables 1 and 2). p.Met46Leu; chr1 g.68659881T>A; rs368633951) in WLS, and this has been reported [23]. This allele is rare, with a global allele frequency of 0.0000398. Patient 7 who had two supernumerary mandibular premolars, unseparated roots of the permanent molars, torus palatinus, and torus mandibularis and carried a heterozygous missense variant, c.850G>C; p.Val284Leu, in the FREM2 gene. This variant is rare with a global allele frequency of 0.0002231 according to gnomAD. The allele frequency of this variant in the East Asian population is 0.003158. This variant has not been reported in the South Asian population. According to the T-Rex database, this allele is found in 4 out of 2184 alleles (allele frequency = 0.0018315 or 0.18%). The allele frequency of this variant in the study group (0.409836%) is higher than those in the T-Rex and gnomAD databases ( Table 2). This variant is predicted to be disease causing (prob: 0.999963783102466) by Mu-tationTaster. The CADD and DANN scores of this variant are 23.8 and 0.9953, respectively. Patient 8, who had multiple supernumerary teeth and multiple odontomas, carried a heterozygous missense variant, c.8498A>G; p.Asn2833Ser, in the FREM2 gene ( Figure 6). This variant was not found in the T-Rex, gnomAD, LOVD, and HGMD databases; therefore, it is considered to be novel. This mutation is predicted to be disease causing (0.999943878525573) and pathogenic (0.9953) by MutationTaster. The CADD and DANN scores of this variant are 17.14 and 0.9802, respectively (Tables 1 and 2).

FREM2 Variants and Their Pathogenicities
None of the seven variants have been reported before as being pathogenic. The variant c.8498A>G p.Asn2833Ser is not reported in gnomAD, LOVD, and NCBI and is considered to be novel. The finding of the c.1921A>G; p.Thr641Ala variant in two unrelated patients, one (patient 6) with mesiodens, and the other one (patient 5) with two supernumerary mandibular premolars and taurodontism supports its pathogenicity. The allele frequencies of the variants found in our patients are higher than those in the normal Thai (T-Rex), South Asian, East Asian, and global populations, according to gnomAD (Table 2).
For rare variants as the causes of disease, statistical analyses are not appropriate as cases due to them being rare, and thus, not sufficient. It is only possible for more common conditions. As it is, the phenotypic presentation of supernumerary tooth phenotypes is likely to be under-recognized in the general population, and especially, an isolated feature is not likely to be considered as a pathology warranting exclusion or mention in public genomic databases such as gnomAD (considered 'healthy' individuals) from which variant frequencies that underpin such causal relationships are made.
With respect to 'control' populations, it is necessary to note that the prevalence of supernumerary teeth has been reported to be 0.04-3.8% [2,3,[5][6][7], but most supernumerary teeth including mesiodens have not erupted into the oral cavities. Notably, more than 50% of mesiodens are unerupted [21][22][23][24][25]. Therefore, unless all those healthy people included in gnomAD have had radiography-based oral examinations, it is ideally unsuitable to include them as a control cohort, as hundred or more of them might have supernumerary tooth phenotypes.
Rare genetic variants have been claimed as predisposing factors or causal factors for genetic diseases. We are aware that not all rare or novel variants are associated with genetic diseases. However, we are convinced that the rare FREM2 variants found in our patients are associated with the supernumerary tooth phenotypes because it is well known that biallelic variants in FREM2 genes are associated with Fraser syndrome with dental anomalies [27,28].

FREM2 Variants, Probability of Being the Loss-of-Function Intolerant (pLI), and the Clinical Significance
In the present cohort of 122 patients with supernumerary tooth phenotypes, the frequency of FREM2 variants that are missense variants is 0.05737705 (7/122). An estimate of the pLI can be used to determine whether or not a single disrupting variant is of likely clinical significance [30]. pLI is derived from comparing the number of protein-truncating variants in a sample with the number of expected mutations under neutrality, given an estimated mutation rate for the gene of interest. In the present study's cohort of individuals with mesiodens and supernumerary teeth, the frequency of missense variants was 0.05737705. Assuming that the number of segregating missense variants observed in a gene is Poisson-distributed, it will correspond to the mean of the number of segregating protein-truncating variants expected in the sample under neutrality times the depletion in the number due to selection. These estimates were categorized as being either neutral (mean = 1), recessive (mean = 0.463), or haploinsufficient (mean = 0.089) (28). The frequency of FREM2 variants in the present cohort (0.05737705) is very similar to the estimate of missense variants leading to haploinsufficiency, and this agrees with the findings of heterozygous variants only. It is noteworthy that, such as FREM2, the WNT10A and WNT10B genes have pLI scores of 0, and heterozygous variants in WNT10A and WNT10B are well known to be implicated in dental anomalies [31]. It has been argued that pLI scores are unsuitable for evaluating autosomal recessively acting genes due to the lack of consequences in heterozygotes [32]. Indeed, pLI scores may not be suitable to evaluate variants that cause dental phenotypes, as these are generally mild or may even go unnoticed, and pLI is calculated based on individuals mentioned in gnomAD. Therefore, if the phenotype is not life-threatening or does not result in a severe disability, such as having a supernumerary tooth, the pLI score is not likely to be meaningful.

The Absence of Rare Variants in Other Known Dental Anomaly-Related Genes
In addition to FREM2 variants, a rare variant (c.130A>T; p.Met44Leu) in the WLS gene was found in patient 5 and has been reported [23]. WES also identified rare variants: the c.33dup; p.Lys12Ter variant in EVC2 in patient 1 and the c.461C>T; p.Pro154Leu variant in MSX1 in patient 4, which might contribute to the phenotypes found in both patients (Table 3). Otherwise, exome sequencing in our patients did not reveal rare variants in other known, dental anomaly-related genes such as WNT10A, WNT10B, PAX9, AXIN2, LRP4, LRP5, LRP6, GREM2, LAMB3, TSPEAR, TFAP2B, PITX2, BMP4, EDA, EDAR, EDARADD, EVC, EVC2, CREBBP, COL1A2, ANTXR1, FGF10, SMOC2, KDF1, KREMEN1, and DKK1 [21][22][23][24][25]. The present study demonstrates, for the first time, that heterozygous carriers of FREM2 variants may have isolated supernumerary teeth, mesiodens, and super-numerary teeth with odontomas. Notably, the dental anomalies reported in biallelic patients with FREM2-associated Fraser syndrome were tooth agenesis and short roots [27,28]. Since our patients were heterozygous, it is hypothesized that the FREM2 variants in them might have caused gain-of-function and resulted in supernumerary teeth, mesiodens, and odontomas. Therefore, the nature of the variants might influence the phenotype. Table 3. Genetic variants with allele frequencies less than 0.0001 (MAF < 0.0001) according to gnomAD in tooth-related genes in patients with FREM2 variants with mesiodens or isolated supernumerary teeth. radiographic evaluations of parents of children with FREM2-related Fraser syndrome for dental anomalies, which would further prove the pathogenicity.

FREM2, Tooth Development, and Supernumerary Tooth Formation
During tooth development, Frem2 is expressed in the dental epithelium at the cap stage [27]. Of note, nephronectin plays critical roles in Sox2 expression, and a mutation in SOX2 was implicated in SOX2 anophthalmia syndrome with supernumerary teeth [33]. Therefore, it is hypothesized that heterozygous missense variants in FREM2 result in alterations of the FRAS1-FREM2-FREM1 complex, the disruption of the basement membrane assembly of nephronectin [26], the abnormal expression of SOX2 [33], the overactivation of WNT/β-catenin signaling [34], and the subsequent formation of supernumerary teeth, mesiodens, and odontomas [29,35,36] (Figure 7). Most cases of mesiodens are sporadic. If cases are familial, the mode of inheritance is autosomal dominance with incomplete penetrance [24,37]. This finding implies that many genes are involved in its etiology. Recently, pathogenic variants in LRP5, LRP6, WLS, DKK1, and LRP4 have been implicated in mesiodens or odontomas [21][22][23][24][25]. Our study suggests that FREM2 variants are likely the "predisposing factors" to the mesiodens phenotype.

FREM2 Variants and Oral Exostoses
Patients 1 and 6 also had torus palatinus or torus mandibularis, respectivel exostoses, including torus palatinus, torus mandibularis, and buccal exostoses, hav reported to be associated with aberrant WNT/β-catenin signaling [21,23,24]. The o ostoses found in our patients might have been coincidental or caused by the un genetic pathways involving FREM2.

FREM2 Variant and Odontomas
We report eight patients, five with mesiodens, two with supernumerary teet one with supernumerary teeth and odontomas. Odontomas are benign mixed o genic tumors involving all odontogenic tissues (enamel, dentin, and cementum) [2 though supernumerary teeth and odontomas are classified as distinct entities, both pathologic processes, topographic distribution, and pathologic manifestations [

FREM2 Variants and Oral Exostoses
Patients 1 and 6 also had torus palatinus or torus mandibularis, respectively. Oral exostoses, including torus palatinus, torus mandibularis, and buccal exostoses, have been reported to be associated with aberrant WNT/β-catenin signaling [21,23,24]. The oral exostoses found in our patients might have been coincidental or caused by the unknown genetic pathways involving FREM2.

FREM2 Variant and Odontomas
We report eight patients, five with mesiodens, two with supernumerary teeth, and one with supernumerary teeth and odontomas. Odontomas are benign mixed odontogenic tumors involving all odontogenic tissues (enamel, dentin, and cementum) [2]. Even though supernumerary teeth and odontomas are classified as distinct entities, both share pathologic processes, topographic distribution, and pathologic manifestations [2]. Recently, odontomas have been reported to be associated with a variant in LRP6 [21]. Notably, odontomas and supernumerary teeth can be found in patients with APC-associated familial adenomatous polyposis syndrome. However, patient 8, who had supernumerary teeth and odontomas, did not have a rare variant in the APC gene. Therefore, the phenotype of patient 8 was not related to familial adenomatous polyposis syndrome [11].

Clinical Implication
Our study shows that heterozygous variants in FREM2 are associated with extra tooth phenotypes. If a patient has an FREM2 pathogenic variant, it is recommended to perform thorough clinical and radiographic examinations before orthodontic treatment. Supernumerary teeth, mesiodens, and oral exostoses may be presented later in life. If we suppose that a patient's family member has an FREM2 variant or a history of having supernumerary teeth, mesiodens, or odontomas, in that case, it is recommended to perform thorough clinical and radiographic examinations and genetic testing of the patient prior to orthodontic treatment to rule out the presence of the FREM2 variant or other supernumerary tooth-associated genetic variants.
The presence of supernumerary teeth or mesiodens may cause a range of dental complications due to dental crowding, eruption failure, midline diastema, rotation or displacement of the adjacent teeth, root resorption, root dilaceration, dentigerous cyst formation, and the root maldevelopment of permanent teeth [38]. It is important to note that extraction is not always the treatment of choice for supernumerary teeth. Several supernumerary teeth are incidental and harmless. Many of them are best left in place and kept under observation, mainly when the normal eruption of the neighboring teeth has occurred and no associated pathology is seen [3].
Our study also shows that patients with heterozygous variants in FREM2 may have oral exostoses. Torus mandibularis and torus palatinus in patients may affect the design of the removable orthodontic appliances. It is important to note that these oral exostoses will appear after the patient has passed through the teenage years. Therefore, the phenotypes associated with FREM2 variants may affect the orthodontic treatments.

Future Studies
Since FREM2 protein is a vital component of the FRAS1-FREM2-FREM1 complex and subsequent abnormal epithelial-mesenchymal interactions, the disruption of epithelialmesenchymal interactions as a result of FREM2 mutations may have deleterious effects on tooth development including tooth mineralization [39]. It is suggested that structural composition of the supernumerary teeth should be studied in the future.
We report heterozygous carriers of FREM2 in patients with extra tooth phenotypes. Biallelic variants of FREM2 are implicated in Fraser syndrome, as previously mentioned [27,28]. The study of dental conditions of the heterozygous parents of the patients affected with Fraser syndrome has not been conducted. It is suggested that dental evaluations be performed in the parents of patients with Fraser syndrome.

Conclusions
We report for the first time that heterozygous carriers of FREM2 may have phenotypes including oral exostoses, mesiodens, and isolated supernumerary teeth.

Study Limitations
Our cohort consists of 122 patients with supernumerary tooth phenotypes. However, we only have DNA samples and dental information of the affected patients who came for oral and radiographic examinations. We are aware that it would have been ideal if we had each patient's family members to study the co-segregation between genotype and phenotype. It would have strengthened the association between the heterozygous variants in FREM2 and the phenotypes.
Regarding the use of T-Rex and gnomAD databases as "normal" populations, it is ideally unsuitable to include them as control cohorts, as hundred or more patients might have supernumerary tooth phenotypes because generally having supernumerary teeth is not painful, and most patients who have supernumerary teeth do not even know they have them. In addition, more than 50% of mesiodens have not erupted into the oral cavities.

Data Availability Statement:
The data underlying this article are available.