Skip to main content
NCBI home page
Human Molecular Genetics logoLink to Human Molecular Genetics
. 2026 Sep 11;35(19):ddag088. doi: 10.1093/hmg/ddag088

Genetic and epigenetic contributors to cleft laterality: evidence from monozygotic mirror twins and replication cohorts

Aline L Petrin 1,✉, Waheed Awotoye 2, Christina E Spencer 3, Ligiane A Machado-Paula 4, Luke Hovey 5, Henry Keen 6, Michael Chimenti 7, Patrick Breheny 8, Fang Qian 9, Shareef Dabdoub 10, James C Thomas 11, Azeez Butali 12, Jeffrey C Murray 13, Shankar Rengasamy Venugopalan 14, Lina M Moreno-Uribe 15
PMCID: PMC13626208  PMID: 42725914

Abstract

Nonsyndromic cleft lip (nsCL) exhibits a non-random laterality pattern, with left-sided clefts occurring twice as frequently as right-sided clefts. The molecular mechanisms underlying this laterality bias remain poorly understood. We performed whole-genome sequencing and methylation profiling on a family comprising monozygotic twins with mirror-image nsCL, their affected mother, and unaffected father and brother. We conducted three independent replications via publicly available whole genome data; genome-wide methylation analysis in 38 individuals with unilateral cleft; and validation of methylation results in the top 3 candidate genes in 385 unrelated individuals with unilateral nonsyndromic cleft lip with or without cleft palate (nsCL/P) (DNA from blood or saliva). We identified a variant in FGF20 (p.Ile79Val) shared by the twins and their mother. We observed laterality and severity-associated methylation differences in three main genes. ARID5B showed higher methylation in left clefts (saliva, P = .001; blood, P = .032). ZFP57 demonstrated a strong cleft-extent effect, with cleft lip and palate (CLP) showing markedly higher methylation than cleft lip only (CL) (LCLP vs. RCL padj = 0.0004; LCLP vs. LCL padj = 0.019). HOOK2 displayed a cross-tissue cleft-extent effect in the opposite direction—CLP subtypes were hypomethylated relative to CL-only subtypes in blood (P < .0001) and saliva (P = .0008). This study provides evidence that DNA methylation patterns play a role in both the laterality and severity of cleft lip. ARID5B provides a consistent laterality signal across tissues, while ZFP57 and HOOK2 track palatal involvement independently of side. Together, these findings suggest that epigenetic variation acts downstream of genetic predisposition to shape cleft phenotypes.

Keywords: complex disease, epigenetics, DNA methylation, whole genome sequencing, epigenomics, ggenomics

Introduction

Orofacial clefts (OFCs) are among the most common congenital malformations, affecting approximately 1 in 700 live births worldwide [1]. These conditions result from disturbances during embryonic development that lead to incomplete fusion of the upper lip (cleft lip), palate (cleft palate), or both [2]. Approximately 70% of cases are classified as nonsyndromic, occurring without additional congenital anomalies [3]. OFCs impose significant functional, aesthetic, and psychosocial burdens on affected individuals, requiring multidisciplinary care from infancy through adulthood.

A striking epidemiological feature of unilateral cleft lip is its non-random laterality distribution with left-sided clefts occurring approximately twice as frequently as right-sided clefts, a pattern consistently observed across diverse populations and geographic regions [1]. This 2:1 ratio strongly suggests that cleft laterality is not a random developmental event but reflects underlying biological mechanisms that preferentially affect left-sided craniofacial development. Despite decades of research, the molecular basis of this laterality bias remains largely unexplained.

Monozygotic twins provide an invaluable model for studying the contribution of genetic and environmental factors to birth defects, as they share nearly identical genetic backgrounds [4]. Mirror-image traits, occurring in approximately 25% of monozygotic twin pairs, include hair whorls, birthmarks, and structural anomalies expressed on opposite sides of the body midline [5]. Monozygotic twins with mirror-image unilateral clefts—where one twin has a left-sided cleft and the other a right-sided cleft—represent a particularly powerful natural experiment for studying laterality determination while controlling for genetic background.

Previous whole-genome sequencing studies of mirror twins with OFCs found no discordant genetic variants that could explain the laterality differences, suggesting that epigenetic mechanisms may play a crucial role in modifying cleft laterality [6]. DNA methylation, a key epigenetic modification involving the addition of methyl groups to cytosine residues at CpG sites, has emerged as a critical factor in OFC pathogenesis [7–10]. By affecting transcription factor binding at gene regulatory regions, DNA methylation can modulate gene expression during critical developmental windows [11, 12].

Genetic studies of nonsyndromic OFCs have identified multiple susceptibility loci, implicating genes such as IRF6, PAX7, MAFB, ABCA4, and VAX1, among others [13]. While cleft lip (CL) and cleft lip and palate (CLP) share many of these risk loci, they show partially distinct genetic profiles consistent with separable developmental processes for lip fusion and palatal fusion. By contrast, far less is known about the genetic determinants of cleft laterality. The limited studies examining left- versus right-sided clefts have not identified consistently replicated variants that specifically determine cleft side [14, 15], suggesting that laterality may be influenced by developmental stochasticity, environmental asymmetry, or epigenetic variation established during embryogenesis. Furthermore, while the clinical distinction between laterality (left vs. right) and cleft extent (whether the palate is involved) is well established, these axes of phenotypic heterogeneity are rarely analyzed separately in epigenetic studies. Cleft lip results from failure of fusion of the maxillary and nasal processes at weeks 4–7 of gestation, whereas palatal clefting involves secondary palate fusion at weeks 8–12 [2]. This temporal separation suggests that laterality and cleft extent may have distinct molecular underpinnings that warrant independent investigation.

In this study, we present a comprehensive multi-omic analysis of a female monozygotic twin pair with mirror-image nonsyndromic cleft lip, born to a mother with left-sided microform cleft lip. By integrating whole-genome sequencing with genome-wide DNA methylation profiling and conducting three independent replications, we identify candidate molecular mechanisms that may contribute to both the occurrence and the laterality bias of this common birth defect.

Results

Whole-genome sequencing identifies shared protein-disrupting variants

Analysis of discordant protein-disrupting variants between the twins identified a heterozygous variant in KRT6B present only in the twin with left cleft lip (LCL). Although monozygotic twins share a common zygote, post-zygotic somatic mutations can arise independently in each twin after embryonic splitting, resulting in variants detectable in only one twin [16]. The KRT6B variant may represent such an event; however, its presence in the unaffected father and heterozygous state made it unlikely to explain the laterality difference. A de novo splice site variant in ACLY was also identified. While no evidence suggests a role for this gene in craniofacial development, the knockout mouse showed early embryonic (stage E7) lethality. Additionally, we identified a protein-disrupting variant of uncertain significance in MUC3A that was maternally-inherited (heterozygous). Based on the in-silico prediction and the likely recessive inheritance pattern, we concluded that this heterozygous maternally-transmitted variant was unlikely to be the risk variant.

Consistent with previous reports [14], we did not identify discordant pathogenic variants that could definitively explain the laterality differences. We therefore focused on maternally transmitted variants shared by both twins and their affected mother (Table 1). This approach identified 16 rare protein-disrupting variants in craniofacial candidate genes (Table 2).

Table 1.

Number of variants (CADD > 15) at each analytical step towards prioritization of pathogenic variants.

Filter Number of Variants
Protein-disrupting variants with MAF < 1% 955 (MZ-1); 939 (MZ-2)
Maternally inherited protein-disrupting variants with MAF < 1% 393 (MZ-1); 416 (MZ-2)
Shared maternally inherited protein-disrupting variants with MAF < 1% 387
Variants with CADD≥15 175
Variants with CADD≥15 + deleterious (SIFT) 90
Variants with CADD≥15 + damaging (Polyphen2) 82
Variants with CADD≥15 + deleterious (SIFT) + damaging (Polyphen2) 59
Variants with CADD≥15 in Craniofacial genes 16
Variants with CADD≥15 + deleterious (SIFT) and/or damaging (Polyphen2) in craniofacial genes 7
Open in a new tab

Note: Rows 5–7 (CADD≥15 + SIFT; CADD≥15 + PolyPhen2; CADD≥15 + both) represent parallel independent filters applied to row 4, not nested sequential subsets of each other. Row 8 applies all three conditions simultaneously to the craniofacial gene list.

Table 2.

Maternally-transmitted shared variants with evidence supporting the roles of the genes in craniofacial development.

# Chrom:Pos Ref/Alt Gene HGVS p. SIFT PolyPhen CADD Facial phenotype Craniofacial phenotype Cleft genes (DB) ACMG cat Inheritance
1 7:7452365 G/C COL28A1 NP_001032852.2:
p.Pro488Arg		 tolerated Possibly damaging 22.9 Mid forehead Nil Nil −13 (B) LR
2 9:126503192 C/T MVB12B NP_258257.1:
p.Arg297Cys		 tolerated Probably damaging 26.4 Self-reported chin dimples Nil Nil 0 (VUS) LR
3 5:113028959 C/T MCC NP_001078846.2:
p.Val952Met		 deleterious Probably damaging 23.8 Upper lip center Nil Nil 0 (VUS) VLR
4 13:37590506 C/T POSTN NP_006466.2:
p.Gly103Ser		 tolerated Probably damaging 24.9 Nil Abnormal ameloblast morphology; Abnormal tooth eruption Nil 0 (VUS) VLR
5 21:40296147 C/T DSCAM NP_001380.2:
p.Arg697Gln		 tolerated benign 20.9 Nil Abnormal cranial cavity morphology; dome cranium; small ears Nil −11 (B) VLD
6 4:2833015 C/G SH3BP2 NP_003014.3:
p.Ser505Cys		 tolerated benign 15.58 Nil Abnormal craniofacial bone morphology Nil −14 (B) VLR
7 15:89291615 A/C FANCI NP_001106849.1:
p.Leu631Phe		 tolerated benign 15.43 Nil Abnormal craniofacial bone morphology; Cleft palate Nil −21 (B) VLR
8 20:51790193 A/G SALL4 NP_065169.1:
p.Ser764Pro		 tolerated benign 21 Nil Abnormal craniofacial bone morphology; Cleft palate Nil −21 (B) VLD
9 16:2100458 C/T PKD1 NP_001009944.3:
p.Arg3169Gln		 tolerated benign 16.98 Nil Abnormal craniofacial bone morphology; short maxilla Nil −2 (LB) VLD
10 16:81044059 G/A ATMIN NP_056066.2:
p.Val521Ile		 tolerated benign 21.6 Nil Abnormal craniofacial morphology; micrognathia; Thick upper lip Nil −12 (B) VLR
11 5:84064787 T/C EDIL3 NP_005702.3:
p.Ile289Val		 tolerated benign 22.7 Nil Abnormal pinna cartilage morphology; floppy ears Nil −5 (LB) LR
12 8:17001798 T/C FGF20 NP_062825.1:
p.Ile79Val		 deleterious Possibly damaging 26.8 Nil Abnormal tooth morphology Nil −15 (B) VLD
13 8:117835473 C/T EXT1 NP_000118.2:
p.Val379Ile		 tolerated benign 16.39 Nil Cleft of Secondary palate Yes 0(VUS) VLD
14 5:37180175 G/A CPLANE1 NP_075561.3:
p.Pro1860Leu		 deleterious benign 22.6 Nil Cleft of upper lip; Cleft palate (Joubert syndrome) Yes −7 (B) VLR
15 14:39180971 C/T PNN NP_002678.3:
p.Ala421Val		 tolerated benign 17.38 Nil Cleft palate Yes −3 (LB) VLR
16 15:74722964 C/T CYP1A1 NP_001306146.1:
p.Gly45Asp		 deleterious Possibly damaging 22.8 Nil Nil Yes −14 (B) LR
Open in a new tab

Facial phenotype annotations are from the Facial Morphology GWAS [17]. Craniofacial phenotype annotations are derived from the Mouse Genome Informatics database (MGI; www.informatics.jax.org) and the Online Mendelian Inheritance in Man database (OMIM; www.omim.org). Cleft gene database (DB) annotations are from CleftGeneDB [18]. ACMG categories: B = Benign; LB = Likely Benign; VUS = Variant of Uncertain Significance. Inheritance: LR = Low Recessive; VLR = Very Low Recessive; VLD = Very Low Dominant. Nil = no annotation found in the respective database.

These variants were novel, or rare in population control databases, altered protein sequences, and ranked among the top 1% most deleterious mutations in the human genome (based on CADD scores ≥15). Several of these variants were in genes known to be associated with craniofacial development based on evidence from facial gene scan databases [17], mouse genome informatics, and cleft gene databases. Notably, variants in FANCI, SALL4, EXT1, CPLANE1, and PNN were identified in genes that, when manipulated in mice, resulted in cleft phenotypes (CleftGeneDB; https://bioinfo.uth.edu/CleftGeneDB) [18].

Table 2 shows mutations that have been considered deleterious, possibly damaging, or probably damaging by at least one prediction tool (COL28A1, CPLANE1, CYP1A1, MCC, POSTN, FGF20, and MVB12B). The mutations in COL28A1, CPLANE1, CYP1A1 and MVB12B were shared by the twins, the mother and the unaffected brother. The mutations in MCC, POSTN and FGF20 were shared by the three affected individuals (twin girls and mother) and not present in the unaffected father and brother.

Of these potentially pathogenic protein-disrupting variants in craniofacial genes, only the FGF20 variant (p.Ile79Val) was predicted by DOMINO to follow a dominant inheritance pattern; thus, becoming our main candidate gene.

Protein function analysis showed that the affected isoleucine residue resides within the fibroblast growth factor family domain, which is critical for receptor binding [19]. Analysis of human craniofacial epigenomics data revealed regulatory element signatures within FGF20, with the highest enhancer peaks observed during Carnegie stages 13–17, corresponding to critical timepoints for primary palate formation [20]. Significant FGF20 expression appeared in craniofacial tissues from CS13 through CS22 (P < .05), supporting its role in lip and palate development [21].

Our replication analysis with publicly available whole-genome sequencing data from multiethnic trios affected with nonsyndromic cleft lip and palate [22] did not show significant results; however previous genome-wide association study investigating sex as a modifier of variant risk identified a novel cleft risk locus at chromosome 8p22, which is within a topologically associating domain (TAD) containing FGF20. That study also demonstrated robust Fgf20 expression in the developing mouse maxilla between days E10.5 and E13.5 [23].

In silico analyses reveal significant FGF20 transcription in craniofacial tissues during embryonic development.

To further investigate the role of FGF20 in human craniofacial development, we examined craniofacial regulatory elements signatures within FGF20 gene and its spatiotemporal expression during embryonic development. Using data from the expanded encyclopedia of DNA elements (ENCODE) (https://www.encodeproject.org), we identified gene regulatory elements signatures within FGF20 including enhancer-like, promoter-like and CTCF- binding elements in the first exon and intron of the gene. We further analyzed the Human Craniofacial Epigenomics dataset (GSE97752) from the Cotney Lab (https://www.chopcranio.org/human-craniofacial-development) [20] and confirmed the presence of regulatory element signatures within FGF20, revealing craniofacial-specific regulatory elements, evidenced by enhancer peaks within the gene. The highest peaks were observed in developing craniofacial tissue structures between Carnegie stages (CS) 13–17, corresponding to approximately 4–6 weeks of embryonic development which are key timepoints for primary palate formation.

To assess FGF20 expression in human craniofacial tissues during embryogenesis, we analyzed bulk and single-cell RNA sequencing (scRNA-seq) data from human craniofacial tissues collected between 4–8 weeks of gestation (GSE197513, phs002008) [21]. Significant FGF20 expression (P < .05) appeared in craniofacial tissue structures at CS13 (P = .001) and remained elevated through CS14 (P = .007), CS15 (P = .00096), and CS17 (P = .01), at fold changes > 30, > 20, > 30, and > 14, respectively (Fig. 1). This spatiotemporal expression pattern correlates with critical stages of craniofacial tissue development coincident with the timing of palatogenesis, thus supporting a role for FGF20 in lip and palate development.

Figure 1.

Bar chart showing FGF20 (Ensembl identifier ENSG00000078579) expression levels across individual craniofacial tissue samples spanning 4 to 8 weeks of human embryonic development. The vertical axis is labeled `expression' and ranges from 0 to approximately 70, with horizontal gridlines at intervals of 10. The horizontal axis lists 15 individual samples, each identified by a stage-specific label. Below the axis the label reads `FGF20 / ENSG00000078579'. A color-coded legend appears at the upper right, identifying six color categories: NCC in gold, CS13 in green, CS14 in pink, CS15 in purple, CS17 in orange, and CS22 in dark teal.The 15 samples are ordered left to right as follows, with approximate expression values in parentheses:Neural Crest Cell group (gold bars): NCC_R1 (approximately 0), NCC_R2 (approximately 0), NCC_R3 (approximately 0). All three NCC samples show expression at or near the baseline with no visible bar height. Carnegie Stage 13 group (green bars): CS13-12690 (approximately 20), CS13-13365 (approximately 72). CS13-13365 is the tallest bar in the entire chart. Carnegie Stage 14 group (pink bars): CS14-12408 (approximately 1 to 2, very small), CS14-12704 (approximately 10), CS14-12913 (approximately 22).Carnegie Stage 15 group (purple bars): CS15-13000 (approximately 4), CS15-13019 (approximately 57), CS15-13128 (approximately 47). CS15-13019 is the second tallest bar in the chart. Carnegie Stage 17 group (orange bars): CS17-12320 (approximately 7), CS17-12421 (approximately 31), CS17-12588 (approximately 2).Carnegie Stage 22 group (dark teal bars): CS22_11865 (approximately 5), CS22_11878 (approximately 8). The overall pattern shows that FGF20 expression is highest at Carnegie Stages 13 and 15 and is absent in Neural Crest Cell samples. Expression is low at Carnegie Stages 14, 17, and 22, with individual sample variability within each stage group.

Open in a new tab

FGF20 expression plot from bulk and single-cell RNA seq of craniofacial tissue between 4–8 weeks post conception. *NCC: Neural crest cells; CS: Carnegie stage.

DNA methylation analysis identifies differentially methylated regions

Comparison of genome-wide methylation in blood-derived DNA between the mirror twins identified 408 CpG sites with absolute Δβ > 5% (Supplementary Table S1). Of these, 100 showed higher methylation in the twin with LCL, while 308 showed higher methylation in the twin with right cleft lip (RCL). Enrichment analysis of differentially methylated positions (DMPs) within 5 kb of gene transcription start sites revealed significant enrichment for the biological process ‘DNA methylation involved in embryo development’ (FDR p2adj = 1.32 × 10−13).

Of the 408 DMPs, 40.2% were located in gene body regions, 35.8% in intergenic regions, 11.8% in 5’UTR regions, and the remainder in TSS-proximal or 3’UTR regions. The three candidate genes selected for targeted replication occupy distinct genomic contexts: the ZFP57-associated cluster lies entirely in an intergenic region 3.3–4.2 kb upstream of the TSS; the ARID5B site (cg22368847) is located in the gene body on chromosome 10; and the HOOK2 site (cg11738485) is located in the gene body on chromosome 19.

A notable finding was a cluster of 14 differentially methylated CpG sites spanning approximately 0.9 kb (867 bp) in the intergenic region 3295–4162 bp upstream of the ZFP57 transcription start site on chromosome 6 (hg19: chr6:29648225–29 649 092). All 14 sites showed consistently higher methylation in the twin with LCL (Δβ = 7%–12%), providing convergent evidence for a coordinated regulatory change at this locus rather than an isolated stochastic difference. The site with the lowest Δβ in this cluster (cg06032337) showed borderline replication in an independent cohort, and the adjacent site cg20228636 (11 bp away) was selected as the targeted replication site for the large-scale replication study. By contrast, ARID5B and HOOK2 were each represented by a single DMP, located in gene body regions on chromosomes 10 and 19, respectively.

We conducted a replication study of the 408 CpG sites that showed a Δβ ≥ 5% in a small independent cohort. This cohort consisted of individuals with unilateral nonsyndromic CL/P, which were separated based on laterality of the cleft (right: RCL/P or left: LCL/P) and based on source of DNA (Blood or Saliva). Our analysis revealed a borderline significant CpG site in ZFP57 (saliva, P = .06) and significant CpG sites in ARID5B (blood, P = .002) and HOOK2 (blood and saliva, P = .02).

Additional replication confirms laterality and cleft extent-associated methylation patterns

To further investigate these associations (ZFP57, HOOK2 and ARID5B), we performed a larger-scale tissue-stratified analysis with an independent replication cohort (n = 385), which revealed that methylation differences among cleft subtypes operate along two distinct biological dimensions: cleft laterality (left vs. right side) and cleft extent (CLP vs. CL). Each of the three candidate genes displayed a unique combination of tissue-specific laterality and extent signatures (Tables 3 and 4).

Table 3.

DNA methylation levels by cleft laterality in the replication cohort (n = 385).

Gene Tissue n Left Mean (%) Right Mean (%) Δ Mean (%) P-value
ARID5B Blood 198 70.7 66.5 +4.2 .032
ARID5B Saliva 179 62.1 50.5 +11.6 .001
ZFP57 Blood 198 NS NS — NS
ZFP57 Saliva 187 52.3 63.5 −11.1 .016
HOOK2 Blood 196 NS NS — NS
HOOK2 Saliva 187 NS NS — NS
Open in a new tab

NS = not significant. Δ Methylation shows direction relative to left clefts.

Table 4.

Results for four cleft subtypes and comparisons by gene and tissue.

Gene Tissue n P-value Median methylation by subtype & significant post-hoc pairs
ZFP57 Blood 198 .0003 LCLP = 56.8%, RCLP = 46.1%, LCL = 31.0%, RCL = 25.4% | Sig. pairs: LCLP>RCL (p adj = 0.0004); LCLP>LCL (p adj = 0.019)
ZFP57 Saliva 187 .089 LCLP = 40.5%, RCLP = 56.2%, LCL = 32.9%, RCL = 41.4% | No significant pairs
HOOK2 Blood 196 <.0001 LCLP = 15.4%, RCLP = 18.1%, LCL = 19.5%, RCL = 20.5% | Sig. pairs: LCLP<RCL (p adj = 0.0003); RCLP<RCL (p adj = 0.002); LCLP<LCL (p adj = 0.012)
HOOK2 Saliva 187 .0008 LCLP = 14.5%, RCLP = 19.8%, LCL = 19.1%, RCL = 18.5% | Sig. pairs: LCLP<LCL (p adj = 0.001); LCL > RCL (p adj = 0.046)
ARID5B Blood 198 .189 LCLP = 72.5%, RCLP = 65.5%, LCL = 70.7%, RCL = 69.0% | No significant pairs
ARID5B Saliva 179 .006 LCLP = 60.5%, RCLP = 54.5%, LCL = 70.9%, RCL = 44.5% | Sig. pairs: LCL > RCL (p adj = 0.015); LCL > RCLP (p adj = 0.024)
Open in a new tab

LCL-left cleft lip; RCL-right cleft lip; LCLP-left cleft lip and palate; RCLP-right cleft lip and palate

ARID5B: The most striking finding emerged from analysis of ARID5B (cg22368847), which showed consistently higher methylation in left-sided clefts across both tissue types, representing a cross-tissue, directionally concordant laterality signal. In saliva, left-sided clefts demonstrated mean methylation of 62.1% compared to 50.5% in right clefts (Δ = +11.6%, P = .001). This laterality pattern was reproduced in blood-derived samples (left: 70.7%, right: 66.5%; Δ = +4.2%; P = .032), with a smaller but statistically significant effect size consistent with the greater homogeneity of blood-derived methylation (Table 3). In saliva, the four cleft subtypes (H = 12.55, P = .006) revealed that left cleft lip (LCL) showed the highest median methylation (70.9%), significantly exceeding both right cleft lip (RCL; 44.5%; padj = 0.015) and right cleft lip and palate (RCLP; 54.5%; padj = 0.024) (Table 4). The LCL vs. RCL comparison in saliva (raw P = .003) represents a laterality-within-CL subtype effect, indicating that the ARID5B laterality signal is particularly pronounced in isolated cleft lip cases.

Analysis of human embryonic craniofacial transcriptome data confirmed significant ARID5B expression during Carnegie stages 13–22, the precise window of palatal development [21].

ZFP57: Analysis of ZFP57 (cg20228636) revealed a more complex, tissue-divergent pattern. In saliva-derived DNA, right-sided clefts showed significantly higher methylation than left-sided clefts (right: 63.5%, left: 52.3%; Δ = −11.1%; U = 3488, P = .016)—the opposite direction from ARID5B. This directional contrast between genes is notable: these findings raise the possibility that CpG sites associated with the genomic imprinting machinery (ZFP57) may respond differently to asymmetric craniofacial developmental signals than those associated with chromatin remodeling factors (ARID5B).

In blood-derived DNA, ZFP57 showed no significant laterality effect (U = 5432, P = .182). Instead, blood revealed a strong and highly significant cleft-extent effect across the four subtypes. CLP subtypes showed markedly higher methylation than CL-only subtypes: median methylation in left cleft lip and palate (LCLP; 56.8%) was significantly higher than in both LCL (31.0%; padj = 0.019) and RCL (25.4%; padj = 0.0004) (Table 4). This cleft-extent effect—independent of laterality—suggests that ZFP57 methylation in blood DNA may reflect the additional epigenetic reprogramming events associated with palatal involvement in CLP, consistent with its role as a master regulator of genomic imprinting at developmental control regions [24–26].

HOOK2: Analysis of HOOK2 (cg11738485) revealed no significant laterality effect in either tissue (blood: P = .282; saliva: P = .491). Instead, HOOK2 displayed highly significant cleft-extent effects in both types of tissue, the only gene to show cross-tissue consistency for the cleft-extent dimension.

In blood (P < .0001), we observed that CLP subtypes were consistently hypomethylated relative to CL-only subtypes—a direction opposite to ZFP57. LCLP (median = 15.4%) was significantly lower than both RCL (median = 20.5%; padj = 0.0003) and LCL (19.5%; padj = 0.012). RCLP (18.1%) was also significantly lower than RCL (padj = 0.002) (Table 4).

In saliva, we confirmed that LCLP (median = 14.5%) showed significantly lower methylation than LCL (19.1%; padj = 0.001). Additionally, a saliva-specific within-CL laterality effect emerged: LCL (mean = 32.8%) showed significantly different methylation from RCL (26.5%; padj = 0.046), driven by a subset of high-methylation outliers in the LCL group (Table 4).

Discussion

By leveraging the unique power of monozygotic twins with mirror-image clefts and conducting independent replications—an unbiased genome-wide screen in 38 individuals identifying ARID5B, ZFP57, and HOOK2 as the top candidate loci, followed by a tissue-stratified validation cohort of up to 385 individuals—we have identified gene-specific, tissue-dependent epigenetic signatures that distinguish laterality effects from cleft-severity effects at single-CpG resolution.

Our whole-genome sequencing analysis identified a potentially pathogenic variant in FGF20 (p.Ile79Val) shared by both twins and their affected mother. Previous studies demonstrated co-expression of FGF20 and FGFR2 in developing mouse palatal epithelium and our analysis confirmed significant FGF20 expression in human craniofacial tissues during critical developmental windows. While this variant may contribute to cleft susceptibility, it cannot explain the laterality difference since both twins carry the same mutation.

The identification of three genes—ARID5B, ZFP57, and HOOK2—with reproducible laterality-associated methylation patterns represents a significant advance in understanding cleft etiology. Critically, tissue-stratified analyses reveal that these genes do not simply reflect a single epigenetic dimension but instead capture two distinct biological signals: cleft laterality (which side the cleft occurs on) and cleft extent (whether the palate is involved). Understanding this distinction is essential for interpreting the epigenetic architecture of CL/P and for designing future biomarker studies.

ARID5B emerged as the most robustly associated gene across all analyses, showing consistent laterality effects in both blood and saliva—the only gene to do so. The higher methylation in left-sided clefts was particularly pronounced in saliva-derived DNA and in isolated cleft lip cases. As a chromatin remodeling factor and histone demethylase co-activator that regulates transcriptional programs during development [27], ARID5B may influence the epigenetic landscape during the asymmetric cellular processes underlying facial development. Its significant expression during Carnegie stages 13–22 positions it precisely at the temporal window when laterality determination and palatal fusion occur [21]. The consistency of the ARID5B laterality signal across blood and oral epithelial tissues suggests it may reflect a systemic epigenetic asymmetry established during early embryonic patterning rather than a tissue-specific regulatory response.

ZFP57 is particularly notable as a master regulator of genomic imprinting that maintains methylation patterns at imprinting control regions during early development [24, 26]. Its tissue-divergent profile—right-over-left laterality effect in saliva, and CLP-over-CL extent effect in blood—reveals two distinct regulatory layers at this locus. The saliva-based laterality finding (right>left; Δ = −11.1%, P = .016) is the opposite in direction from ARID5B, consistent with the possibility that imprinting-associated and chromatin remodeling-associated methylation may respond differently to the same asymmetric developmental signals. The blood-based cleft-extent effect indicates that palatal involvement is associated with substantially higher ZFP57 methylation in circulating DNA. Because ZFP57 regulates imprinting at multiple genomic loci, its differential methylation may have broad downstream effects on imprinted gene expression during craniofacial morphogenesis. Protein interaction analysis confirmed that ZFP57 and TFAP2A participate in networks associated with left–right axis specification, providing a potential mechanistic link between our epigenetic findings and laterality determination [28].

HOOK2 displayed the strongest and most consistent cleft-extent epigenetic signal of the three genes, with CLP subtypes significantly hypomethylated relative to CL-only subtypes in both blood and saliva. The LCLP vs. RCL comparison in blood represents the single most statistically significant post-hoc result in the entire replication study. This CLP hypomethylation direction is strikingly opposite to ZFP57, raising the possibility that methylation at loci involved in distinct molecular pathways—imprinting maintenance and cytoskeletal gene regulation—may be differentially regulated in the context of palatal involvement. HOOK2 encodes a microtubule-binding protein critical for centrosome positioning, Golgi ribbon organization, and polarized cell migration [29]. During palate development, the mesenchymal cells of the palatal shelves must coordinate cytoskeletal remodeling for shelf elevation and horizontal reorientation—processes that depend heavily on microtubule dynamics and cell polarity [29]. Lower HOOK2 methylation in CLP patients may correspond to altered expression states in these cytoskeletal programs during palatal morphogenesis. Additionally, a saliva-specific LCL vs. RCL difference suggests that HOOK2 oral epithelial methylation captures an additional laterality-within-CL signal, albeit of borderline significance.

The tissue-specific divergence of epigenetic signals across all three genes underscores a critical methodological consideration for future epigenetic studies of CL/P. Blood and saliva-derived DNA capture fundamentally different aspects of the methylation landscape: blood-derived methylation reflects hematopoietic and systemic developmental programming, while salivary methylation may more directly capture the epigenetic state of craniofacial epithelial progenitors. Consistent with this interpretation, Substantial variation in DNA methylation profiles across purified blood cell lineages has been demonstrated, and these tissue-level differences are further amplified in comparisons between blood and mucosal tissues [30, 31]. In our data, ZFP57 exemplifies this divergence most starkly: it shows a laterality effect only in saliva and a cleft-extent effect only in blood, suggesting that distinct regulatory mechanisms operate in parallel compartments. By contrast, ARID5B’s laterality signal and HOOK2’s extent signal are more broadly shared, consistent with their roles in foundational chromatin and cytoskeletal programs that may be active across multiple tissue lineages.

Our results reveal two distinct epigenetic dimensions in the biology of unilateral CL/P that should be distinguished in future studies: cleft laterality and cleft extent. These dimensions appear to be regulated by different genes and different tissues, and they operate independently. ARID5B methylation is primarily a laterality marker; ZFP57 (blood) and HOOK2 (both tissues) are primarily cleft-extent markers; and ZFP57 (saliva) and HOOK2 (saliva, CL group) bridge both dimensions. This multi-dimensional architecture parallels the known genetic architecture of OFC, in which CLP and CL have partially distinct genetic etiologies and risk loci [2, 8]. The epigenetic findings reported here suggest that these genetically distinct developmental programs are also epigenetically distinguishable at specific CpG sites.

The stronger laterality effects observed in isolated cleft lip cases for ARID5B and HOOK2 may reflect differential sensitivity of epigenetic regulation across developmental stages. Cleft lip results from failure of fusion of the maxillary and nasal processes during weeks 4–7 of gestation, while palatal clefting involves secondary palate fusion during weeks 8–12 [2]. The presence of palatal involvement (CLP) may introduce additional epigenetic heterogeneity that dilutes the pure laterality signal, explaining why laterality-specific methylation differences are more pronounced in CL-only cases.

A notable demographic difference between the blood and saliva groups in our replication cohort (blood: 92.9% non-white; saliva: 57.5%–59.9% non-White across genes) warrants acknowledgment. This population composition difference may contribute to some of the observed tissue divergences and should be explored in future population-stratified analyses. Nonetheless, the directional consistency of findings within each tissue group, and the concordance between our discovery data in the mirror twins and the replication results for ARID5B and ZFP57 laterality effects, support the biological reproducibility of these signals.

Beyond the three candidate genes investigated in the large-scale replication cohort, several additional CpG sites showed suggestive associations in the replication 1 screen that warrant future investigation. Notably, cg27106643, located in the gene body of NFATC1 (Nuclear Factor of Activated T-cells C1; combined P = .037, blood P = .007), is embedded within a gene that encodes a transcription factor with established roles in osteoclast differentiation, bone remodeling, and craniofacial morphogenesis, processes directly relevant to palatal and lip development [32]. Similarly, cg25952596, located near the RTL1/DIO3 imprinted locus on chromosome 14q32 (combined P = .025, blood P = .005), belongs to a major genomic imprinting domain, extending the imprinting-related methylation theme identified at ZFP57 in our discovery cohort [33, 34]. Additionally, cg15256244, near SHISA5—an antagonist of WNT and FGF signaling pathways central to palatogenesis—showed a strong blood-specific signal (P = .004) consistent with the tissue-specific patterns observed for ARID5B [35]. These findings should be interpreted cautiously given the exploratory nature of the replication 1 screen, the number of sites tested, and the absence of large-scale validation. Nevertheless, they represent biologically plausible candidates for independent replication in future epigenome-wide association studies of orofacial clefts.

From a translational perspective, our findings establish the groundwork for developing epigenetic biomarkers for cleft risk assessment and phenotype prediction. ARID5B methylation, as the most consistent cross-tissue laterality marker, may have the greatest potential as an accessible biomarker from blood or saliva. The reversible nature of DNA methylation also opens possibilities for preventive interventions, as methylation patterns can be influenced by nutritional factors affecting one-carbon metabolism, including folate, methionine, and choline [12].

Our study has limitations. The discovery analysis was limited to a single twin pair, inherent to the rarity of mirror-image clefts. However, our two-stage replication strategy—an unbiased genome-wide screen followed by a targeted tissue-stratified validation in up to 385 individuals across multiple population groups—substantially strengthens our findings. Population composition differences between tissue groups limit direct cross-tissue comparisons. Future studies should incorporate matched tissue collection (blood and saliva from the same individuals), population-stratified analyses, and mechanistic studies linking the identified CpG sites to gene expression during craniofacial morphogenesis.

Materials and methods

Mirror twins and family members

The discovery cohort consisted of a family with female monozygotic twins, their unaffected brother, and their parents. One twin presented with left-sided microform cleft lip (LCL), while the other exhibited right-sided microform cleft lip (RCL). The twins’ mother had a left-sided microform cleft lip; the brother and father had no structural abnormalities (Fig. 2). Blood samples were collected from all available family members. The study was approved by the University of Iowa Institutional Review Board (protocols 199 804 081, 200 003 065, and 200 710 721), and all participants (discovery and replication cohorts) provided written informed consent.

Figure 2.

Panel A---Two-generation family pedigree. Standard genetic pedigree diagram using the conventional notation in which squares represent males, circles represent females, filled symbols indicate affected individuals, and open symbols indicate unaffected individuals. The diagram is labeled `A' in the upper left corner.Generation I (top row): an unaffected father represented by an open square on the left, and an affected mother represented by a filled circle on the right. The mother is labeled `LCL' below her symbol, indicating left cleft lip only. A horizontal mating line connects the two individuals.Generation II (bottom row): three offspring connected by a horizontal sibship line descending from the mating line above. On the far left is an unaffected son represented by an open square. In the center and right are two affected daughters, each represented by a filled circle. A downward triangle connecting the two daughters indicates they are monozygotic twins. The left twin is labeled `LCL' (left cleft lip) and the right twin is labeled `RCL' (right cleft lip) below their respective symbols, demonstrating the mirror-image laterality of their clefts. Panel B--Clinical photographs. Two sideby- side close-up facial photographs of the monozygotic twin girls, labeled `B' in the upper left corner of the left photograph. In both images, a solid black horizontal bar obscures the subjects' eyes to protect their identities.Left photograph: the twin with left cleft lip (LCL). The cleft is visible as an indentation on the left side of the upper lip and philtrum, with surrounding tissue showing characteristic scarring and displacement consistent with a microform cleft.Right photograph: the twin with right cleft lip (RCL). An analogous cleft is visible on the right side of the upper lip, positioned as the lateral mirror image of the left twin's cleft.

Open in a new tab

(A) Pedigree showing the inheritance pattern of clefting in the family. (B) Monozygotic twins with mirror cleft lip.

Whole-genome sequencing

Whole-genome sequencing was performed at an average coverage depth of 30×. Sequence data were aligned to the human genome assembly GRCh38 using the Dynamic Read Analysis for GENomics (DRAGEN) pipeline. We prioritized high-confidence variants (genotype quality ≥ 20, read depth ≥ 10) that were rare or novel (minor allele frequency < 1% in gnomAD). Protein-disrupting variants—defined as missense variants and predicted loss-of-function variants (stop-gained, frameshift, and canonical splice-site variants within ±2 bp of exon-intron junctions)—were assessed using SIFT, PolyPhen2, and CADD. Loss-of-function consequences were assigned based on Ensembl Variant Effect Predictor (VEP) annotations; we use the term ‘protein-disrupting’ throughout to encompass these variant classes. Variants were classified according to the American College of Medical Genetics and Genomics (ACMG) system, and the DOMINO algorithm was used to assess dominant inheritance patterns [36]. Our workflow is illustrated and detailed in Fig. 3.

Figure 3.

Left column (three rows, top to bottom): Box 1---`Whole Genome Sequencing': the entry point of the pipeline. A downward arrow leads to Box 2---`Variants filtered based on: Genotype quality (GQ) greater than or equal to 20 and Read depth (RD) greater than or equal to 10': quality control filter applied to all sequenced variants. A downward arrow leads to Box 3---`Novel / Rare variants (MAF less than 1%)': minor allele frequency rarity filter retaining only novel or rare variants.Center column (three rows, top to bottom): Box 4---`Prioritization of variants: CADD score greater than or equal to 15; SIFT equals deleterious (towards 0); PolyPhen2 greater than or equal to possibly damaging (towards 1)': in silico pathogenicity scoring using three prediction tools. A downward arrow leads to Box 5---`Variants segregating with phenotype (Maternally inherited variants) / 1. Discordant variants / 2. Concordant variants': phenotypic co-segregation filter, distinguishing variants present in only one twin (discordant) from those shared by both twins and their affected mother (concordant). A downward arrow leads to Box 6---`Protein-altering variants (Loss of Function (LOF), Missense)': the final output of the pipeline, representing candidate protein-disrupting variants associated with cleft risk.Right column (top two rows only): Box 7---`Variants clinical significance (ACMG criteria) / Dominant genes (DOMINO)': clinical variant classification according to American College of Medical Genetics and Genomics criteria and dominant inheritance assessment using the DOMINO algorithm. A downward arrow leads to Box 8---`Variants in genes involved in craniofacial development (Facial morphology genes, MGI, and Cleft genes DB and *Craniofacial Super-enhancer analysis)': biological relevance filter retaining variants in genes with documented roles in craniofacial development, crossreferenced against three databases and a craniofacial super-enhancer dataset. The right column has no thirdrow box.Horizontal arrows connect the columns within each row from left to right, and vertical arrows connect rows within each column from top to bottom. The final output box, `Protein-altering variants (Loss of Function (LOF), Missense)', located in the bottom center, receives input from both the left column below it (Novel / Rare variants, connected by a rightward arrow) and the center column above it, representing the convergence of all filtering and prioritization criteria.

Open in a new tab

Genomic analytical pipeline used to investigate discordant and concordant pathogenic protein-altering variants in the mirror twins contributing to the risk of cleft.

Craniofacial candidate genes were defined as those with documented roles in craniofacial development based on three curated resources: (1) the Facial Morphology Gene Scan database [17], which catalogues genes associated with normal facial variation; (2) the Mouse Genome Informatics database (MGI; www.informatics.jax.org), queried for genes whose mouse orthologs produce craniofacial or orofacial phenotypes when disrupted; and (3) the CleftGeneDB database of genes associated with cleft phenotypes in animal models (https://bioinfo.uth.edu/CleftGeneDB). The complete list of genes selected based on these criteria is provided in Supplementary Table S2.

Additional sequencing analysis

We used Sanger sequencing to confirm variant segregation in the family for top candidates from Table 2.

For replication of the FGF20 finding, we analyzed the FGF20 genomic region (full transcriptional span plus 1 kb upstream) in whole-genome sequencing data from 1611 multiethnic parent-offspring trios with nonsyndromic CL/P available through the Gabriella Miller Kids First Pediatric Research Program (phs001168, phs001420, phs001997, phs002595, phs003893) [22]. Supplementary Table S3 shows a summary of the public datasets used. Variants were extracted using bcftools and tested for preferential transmission using an allelic transmission disequilibrium test (TDT). A complementary rare variant burden test (MAF < 0.1%) was performed using the same TDT framework with gnomAD allele frequencies.

In silico functional analysis

Following the identification of the main candidate genes, we examined craniofacial regulatory elements signatures and their spatiotemporal expression during embryonic development. Using data from the expanded encyclopedia of DNA elements (ENCODE) (https://www.encodeproject.org), we searched for gene regulatory elements signatures including enhancer-like, promoter-like and CTCF- binding elements. We further analyzed the Human Craniofacial Epigenomics dataset (GSE97752) from the Cotney Lab (https://www.chopcranio.org/human-craniofacial-development) [20] to look for presence of regulatory element signatures, revealing craniofacial-specific regulatory elements active during relevant Carnegie stages (CS13-22) [21].

DNA methylation analysis in twins

Genome-wide DNA methylation profiling was performed using the Illumina EPIC BeadChip array, which evaluates over 850 000 CpG sites. Methylation levels were quantified as β-values (0–1), representing the ratio of methylated to total signal. The EpiDISH package was used for cell-type composition normalization [37]. CpG sites with coefficient of variation > 20% in gender-matched controls were excluded. Differentially methylated positions (DMPs) with absolute Δβ ≥ 5% between twins were selected for downstream analysis. The DNA methylation analysis workflow is illustrated in Fig. 4.

Figure 4.

Vertical flowchart depicting the DNA methylome analytical pipeline used to identify differentially methylated CpG sites between mirror-image cleft lip twins. The diagram contains five rounded-rectangle boxes with gray backgrounds arranged in a vertical sequence, connected by downward arrows, with two annotation labels branching to the right at specific steps.Box 1 (top): `Mirror twins (1 pair) + Unaffected Females (n=12) / Total CpGs after QC n = 735, 653'. The starting dataset consists of genome-wide methylation data from the mirror twin pair and 12 unaffected control females, with 735,653 CpG sites remaining after initial quality control. A short downward arrow leads to Box 2.Right-branching annotation 1 (horizontal arrow pointing right from the junction between Boxes 1 and 2): `Removed CpG sites with coefficient of variation (based on beta values) among unaffected greater than 20% (less likely to contribute to clefting) (n=14, 240)'. This annotation describes the filtering criterion applied at this step: 14,240 CpG sites with high methylation variability among unaffected controls are excluded as unlikely to be relevant to clefting.Box 2: `Mirror twins (1 pair) + Unaffected Female (n=12) / Filtered CpGs n = 721, 413'. After removing the 14,240 high-variability sites, 721,413 CpG sites remain for further analysis. A downward arrow leads to Box 3.Box 3: `Mirror twins CpGs absolute difference (deltabeta) greater than 5% / n=408'. A threshold of absolute deltabeta greater than 5% between the two twins is applied, yielding 408 differentially methylated CpG sites. From Box 3, a single downward arrow splits into two diverging arrows leading to two parallel boxes side by side.Box 4a (left parallel box): `100 CpGs / Left Twin greater than Right Twin'. One hundred of the 408 differentially methylated CpG sites show higher methylation in the twin with left cleft lip compared to the twin with right cleft lip.Box 4b (right parallel box): `308 CpGs / Right Twin greater than Left Twin'. Three hundred and eight of the 408 differentially methylated CpG sites show higher methylation in the twin with right cleft lip compared to the twin with left cleft lip.Right-branching annotation 2 (horizontal arrow pointing right from the junction between the two parallel boxes and Box 5): `Gene ontology analysis conducted using GREAT'. This annotation indicates that at this stage the 408 differentially methylated positions are subjected to gene ontology enrichment analysis using the GREAT tool.Box 5 (bottom): `Regions with consistent direction of methylation'. Both parallel boxes converge via arrows into this final box, which represents the pipeline output: genomic regions where the direction of methylation difference between the two twins is consistent, constituting the highest-confidence candidate differentially methylated loci carried forward for replication.

Open in a new tab

DNA methylome analytical pipeline.

Methylation replication studies

We performed a genome-wide methylation analysis (Illumina EPIC array) in 38 individuals with unilateral cleft lip and/or palate, comprising blood samples (right = 9, left = 13) and saliva samples (right = 7, left = 9). This screening identified three genes (ARID5B, ZFP57, and HOOK2) as the top three differentially methylated candidate loci.

We then followed up with a targeted methylation analysis of three priority candidate genes—ARID5B (CpG site cg22368847; n = 377), ZFP57 (CpG site cg20228636; n = 385), and HOOK2 (CpG site cg11738485; n = 383)—in individuals with unilateral nonsyndromic CL/P. Each cohort comprised approximately equal proportions of four cleft subtypes: left cleft lip and palate (LCLP), right cleft lip and palate (RCLP), left cleft lip only (LCL), and right cleft lip only (RCL) as shown in Table 5. Critically, analyses were performed separately for blood-derived (ZFP57: n = 198; HOOK2: n = 196; ARID5B: n = 198) and saliva-derived (ZFP57: n = 187; HOOK2: n = 187; ARID5B: n = 179) DNA to account for tissue-specific methylation patterns [30]. To compare methylation levels between two groups (e.g. left vs. right clefts), the normality of each group’s distribution was assessed using the Shapiro–Wilk test. The nonparametric Mann–Whitney U test (equivalent to the Wilcoxon rank-sum test) was applied for all two-group comparisons, as the majority of subgroups exhibited significant departures from normality. For comparisons across four cleft subtypes (LCLP, RCLP, LCL, RCL), the Kruskal-Wallis test was used. When the Kruskal-Wallis test indicated a significant difference, pairwise post-hoc comparisons were conducted using Mann–Whitney U tests with Bonferroni correction applied across all six pairwise comparisons (adjusted significance threshold α = 0.0083). All analyses were conducted using the SAS® System version 9.4 (SAS Institute Inc., Cary, NC, USA). Statistical significance was set at P < .05.

Table 5.

Patients distributions and characteristics per candidate gene for the targeted replication cohort.

Source n (%) Sex n (%) Race n (%) Cleft Type n (%) Laterality n (%)
Gene Saliva Blood Male Female White Non-White LCLP RCLP LCL RCL Left Right
ZFP57 187 (48.6) 198 (51.4) 203 (52.7) 182 (47.3) 91 (23.6) 294 (76.4) 99 (25.7) 96 (24.9) 96 (24.9) 94 (24.5) 195 (50.6) 190 (49.4)
HOOK2 187 (48.8) 196 (51.2) 204 (53.3) 179 (46.7) 89 (23.2) 297 (76.3) 96 (25.1) 94 (24.5) 94 (24.5) 99 (25.9) 190 (49.6) 193 (50.4)
ARID5B 179 (47.5) 198 (52.5) 201 (53.3) 176 (46.7) 90 (23.9) 287 (76.1) 94 (24.9) 92 (24.4) 97 (25.8) 94 (24.9) 191 (50.7) 186 (49.3)
Open in a new tab

LCL-left cleft lip; RCL-right cleft lip; LCLP-left cleft lip and palate; RCLP-right cleft lip and palate

To distinguish laterality effects from cleft-extent effects, two complementary analyses were performed for each gene: (1) a direct left vs. right comparison pooling all cleft subtypes (LCLP+LCL vs. RCLP+RCL) and (2) a four-subtype Kruskal-Wallis test followed by post-hoc pairwise comparisons to identify which specific subtype contrasts drove any significant omnibus result. This design allows laterality-specific signals (e.g. LCL vs. RCL; LCLP vs. RCLP) to be distinguished from cleft-extent signals (e.g. CLP subtypes vs. CL subtypes) within the same analytical framework.

Supplementary Material

Supplementary_material_ddag088
supplementary_material_ddag088.zip (115.3KB, zip)

Acknowledgements

We are grateful to all families who voluntarily participated in this study and to the funding sources.

Contributor Information

Aline L Petrin, Department of Orthodontics, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Waheed Awotoye, Department of Orthodontics, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Christina E Spencer, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Ligiane A Machado-Paula, Iowa Institute of Human Genetics, Carver College of Medicine, University of Iowa, 431 Newton Road, 117 EMRB, Iowa City, IA 52242, United States.

Luke Hovey, Department of Orthodontics, Tufts University School of Dental Medicine, 1 Kneeland St, Boston, MA 02111, United States.

Henry Keen, Iowa Institute of Human Genetics, Carver College of Medicine, University of Iowa, 431 Newton Road, 117 EMRB, Iowa City, IA 52242, United States.

Michael Chimenti, Iowa Institute of Human Genetics, Carver College of Medicine, University of Iowa, 431 Newton Road, 117 EMRB, Iowa City, IA 52242, United States.

Patrick Breheny, Department of Biostatistics, College of Public Health, University of Iowa, 145 N Riverside Dr, Iowa City, IA 52242, United States.

Fang Qian, Division of Biostatistics and Computational Biology, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Shareef Dabdoub, Division of Biostatistics and Computational Biology, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

James C Thomas, Division of Biostatistics and Computational Biology, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Azeez Butali, Department of Oral Pathology, Radiology, & Medicine, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Jeffrey C Murray, Department of Pediatrics, Carver College of Medicine, University of Iowa, 200 Hawkins Drive, Iowa City, IA 52242, United States.

Shankar Rengasamy Venugopalan, Department of Orthodontics, Tufts University School of Dental Medicine, 1 Kneeland St, Boston, MA 02111, United States.

Lina M Moreno-Uribe, Department of Orthodontics, College of Dentistry and Dental Clinics, University of Iowa, 801 Newton Rd, Iowa City, IA 52242, United States.

Author contributions

Aline Petrin (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing—original draft, Writing—review & editing), Waheed Awotoye (Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review & editing), Christina E. Spencer (Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Ligiane A. Machado-Paula (Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Luke Hovey (Methodology, Validation, Writing—original draft, Writing—review & editing), Henry Keen (Data curation, Formal analysis, Methodology, Validation, Writing—original draft, Writing—review & editing), Michael Chimenti (Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Patrick Breheny (Data curation, Methodology, Writing—original draft, Writing—review & editing), Fang Qian (Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Shareef Dabdoub (Formal analysis, Methodology, Writing—original draft, Writing—review & editing), James C. Thomas (Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Azeez Butali (Methodology, Resources, Writing—original draft, Writing—review & editing), Shankar R. Venugopalan (Conceptualization, Investigation, Methodology, Writing—original draft, Writing—review & editing), Jeff Murray (Conceptualization, Methodology, Resources, Writing—original draft, Writing—review & editing), and Lina M. Moreno (Conceptualization, Methodology, Resources, Writing—original draft, Writing—review & editing). Aline L. Petrin, Waheed Awotoye, Christina E. Spencer, Lina M. Moreno-Uribe (Contributed to conception and design, data acquisition, analysis and interpretation, and drafted and critically revised the manuscript), Christina E. Spencer, Ligiane A. Machado-Paula, Luke Hovey, Patrick Breheny, Henry Keen, Fang Qian, Michael Chimenti, Shareef Dabdoub, James C. Thomas, Azeez Butali, Jeffrey C. Murray, and Shankar Rengasamy Venugopalan (Contributed to conception, data acquisition, analysis, interpretation, and critical revision of the manuscript). All authors gave final approval and agreed to be accountable for all aspects of the work.

Conflicts of interest

All authors have no conflicts of interest to disclose.

Funding

This work was supported by NIH/NIDCR K01DE027995, University of Iowa College of Dentistry Student Research and the University of Iowa College of Dentistry Clinical/Dental Education Research Initiative Support Program (CRISP). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Ethical approval statement

The study was approved by the University of Iowa Institutional Review Board and all participants signed informed consents prior to collection of clinical data and biological samples.

References

  • 1. Mossey  PA, Little  J, Munger  RG  et al.  Cleft lip and palate. Lancet  2009;374:1773–1785. 10.1016/S0140-6736(09)60695-4. [DOI] [PubMed] [Google Scholar]
  • 2. Dixon  MJ, Marazita  ML, Beaty  TH  et al.  Cleft lip and palate: understanding genetic and environmental influences. Nat Rev Genet  2011;12:167–178. 10.1038/nrg2933. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Gowans  LJJ, Al Dhaheri  N, Li  M  et al.  Co-occurrence of orofacial clefts and clubfoot phenotypes in a sub-Saharan African cohort: whole-exome sequencing implicates multiple syndromes and genes. Mol Genet Genomic Med  2021;9:e1655. 10.1002/mgg3.1655. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Sahu  M, Prasuna  JG. Twin studies: a unique epidemiological tool. Indian J Community Med  2016;41:177–182. 10.4103/0970-0218.183593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Stanković-Babić  G, Vujanović  M, Cekić  S. Identical twins with "mirror image" anisometropia and esotropia. Srp Arh Celok Lek  2011;139:661–665. 10.2298/SARH1110661S. [DOI] [PubMed] [Google Scholar]
  • 6. Wang  L, Liu  Z, Lin  H  et al.  Epigenetic regulation of left-right asymmetry by DNA methylation. EMBO J  2017;36:2987–2997. 10.15252/embj.201796580. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Alvizi  L, Ke  X, Brito  LA  et al.  Differential methylation is associated with non-syndromic cleft lip and palate and contributes to penetrance effects. Sci Rep  2017;7:2441. 10.1038/s41598-017-02721-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Sharp  GC, Ho  K, Davies  A  et al.  Distinct DNA methylation profiles in subtypes of orofacial cleft. Clin Epigenetics  2017;9:63. 10.1186/s13148-017-0362-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Xu  Z, Lie  RT, Wilcox  AJ  et al.  A comparison of DNA methylation in newborn blood samples from infants with and without orofacial clefts. Clin Epigenetics  2019;11:40. 10.1186/s13148-019-0638-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Alvizi  L, Brito  LA, Kobayashi  GS  et al.  mir152 hypomethylation as a mechanism for non-syndromic cleft lip and palate. Epigenetics  2022;17:2278–2295. 10.1080/15592294.2022.2115606. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Boyes  J, Bird  A. DNA methylation inhibits transcription indirectly via a methyl-CpG binding protein. Cell  1991;64:1123–1134. 10.1016/0092-8674(91)90267-3. [DOI] [PubMed] [Google Scholar]
  • 12. Garland  MA, Sun  B, Zhang  S  et al.  Role of epigenetics and miRNAs in orofacial clefts. Birth Defects Res  2020;112:1635–1659. 10.1002/bdr2.1802. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Beaty  TH, Murray  JC, Marazita  ML  et al.  A genome-wide association study of cleft lip with and without cleft palate identifies risk variants near MAFB and ABCA4. Nat Genet  2010;42:525–529. 10.1038/ng.580. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Takahashi  M, Hosomichi  K, Yamaguchi  T  et al.  Whole-genome sequencing in a pair of monozygotic twins with discordant cleft lip and palate subtypes. Oral Dis  2018;24:1303–1309. 10.1111/odi.12910. [DOI] [PubMed] [Google Scholar]
  • 15. Takahashi  M, Hosomichi  K, Yamaguchi  T  et al.  Exploration of genetic factors determining cleft side in a pair of monozygotic twins with mirror-image cleft lip and palate using whole-genome sequencing and comparison of craniofacial morphology. Arch Oral Biol  2018;96:33–38. 10.1016/j.archoralbio.2018.08.009. [DOI] [PubMed] [Google Scholar]
  • 16. Jonsson  H, Magnusdottir  E, Eggertsson  HP  et al.  Differences between germline genomes of monozygotic twins. Nat Genet  2021;53:27–34. 10.1038/s41588-020-00755-1. [DOI] [PubMed] [Google Scholar]
  • 17. Shaffer  JR, Orlova  E, Lee  MK  et al.  Genome-wide association study reveals multiple loci influencing normal human facial morphology. PLoS Genet  2016;12:e1006149. 10.1371/journal.pgen.1006149. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Suzuki  A, Jun  G, Abdallah  N  et al.  Gene datasets associated with mouse cleft palate. Data Brief  2018;18:655–673. 10.1016/j.dib.2018.03.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Ohmachi  S, Mikami  T, Konishi  M  et al.  Preferential neurotrophic activity of fibroblast growth factor-20 for dopaminergic neurons through fibroblast growth factor receptor-1c. J Neurosci Res  2003;72:436–443. 10.1002/jnr.10592. [DOI] [PubMed] [Google Scholar]
  • 20. Wilderman  A, VanOudenhove  J, Kron  J  et al.  High-resolution epigenomic atlas of human embryonic craniofacial development. Cell Rep  2018;23:1581–1597. 10.1016/j.celrep.2018.03.129. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Yankee  TN, Oh  S, Winchester  EW  et al.  Integrative analysis of transcriptome dynamics during human craniofacial development identifies candidate disease genes. Nat Commun  2023;14:4623. 10.1038/s41467-023-40363-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Mukhopadhyay  N, Bishop  M, Mortillo  M  et al.  Whole genome sequencing of orofacial cleft trios from the Gabriella miller kids first Pediatric research consortium identifies a new locus on chromosome 21. Hum Genet  2020;139:215–226. 10.1007/s00439-019-02099-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Awotoye  W, Comnick  C, Pendleton  C  et al.  Genome-wide gene-by-sex interaction studies identify novel nonsyndromic orofacial clefts risk locus. J Dent Res  2022;101:465–472. 10.1177/00220345211046614. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Li  X, Ito  M, Zhou  F  et al.  A maternal-zygotic effect gene, Zfp57, maintains both maternal and paternal imprints. Dev Cell  2008;15:547–557. 10.1016/j.devcel.2008.08.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Mackay  DJG, Callaway  JLA, Marks  SM  et al.  Hypomethylation of multiple imprinted loci in individuals with transient neonatal diabetes is associated with mutations in ZFP57. Nat Genet  2008;40:949–951. 10.1038/ng.187. [DOI] [PubMed] [Google Scholar]
  • 26. Quenneville  S, Verde  G, Corsinotti  A  et al.  In embryonic stem cells, ZFP57/KAP1 recognize a methylated hexanucleotide to affect chromatin and DNA methylation of imprinting control regions. Mol Cell  2011;44:361–372. 10.1016/j.molcel.2011.08.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Baba  A, Ohtake  F, Okuno  Y  et al.  PKA-dependent regulation of the histone lysine demethylase complex PHF2–ARID5B. Nat Cell Biol  2011;13:668–675. 10.1038/ncb2228. [DOI] [PubMed] [Google Scholar]
  • 28. de  Crozé  N, Maczkowiak  F, Monsoro-Burq  AH. Reiterative AP2a activity controls sequential steps in the neural crest gene regulatory network. Proc Natl Acad Sci  2011;108:155–160. 10.1073/pnas.1010740107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Pallesi-Pocachard  E, Bazellieres  E, Viallat-Lieutaud  A  et al.  Hook2, a microtubule-binding protein, interacts with Par6α and controls centrosome orientation during polarized cell migration. Sci Rep  2016;6:33259. 10.1038/srep33259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Smith  ZD, Meissner  A. DNA methylation: roles in mammalian development. Nat Rev Genet  2013;14:204–220. 10.1038/nrg3354. [DOI] [PubMed] [Google Scholar]
  • 31. Reinius  LE, Acevedo  N, Joerink  M  et al.  Differential DNA methylation in purified human blood cells: implications for cell lineage and studies on disease susceptibility. PLoS One  2012;7:e41361. 10.1371/journal.pone.0041361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Takayanagi  H. The role of NFAT in osteoclast formation. Ann N Y Acad Sci  2007;1116:227–237. 10.1196/annals.1402.071. [DOI] [PubMed] [Google Scholar]
  • 33. Prasasya  R, Grotheer  KV, Siracusa  LD  et al.  Temple syndrome and Kagami-Ogata syndrome: clinical presentations, genotypes, models and mechanisms. Hum Mol Genet  2020;29:R107–R116. 10.1093/hmg/ddaa133. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Zhang  X, He  H, Yu  H  et al.  Maternal RNA transcription in Dlk1-Dio3 domain is critical for proper development of the mouse placental vasculature. Commun Biol  2024;7:363. 10.1038/s42003-024-06038-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. He  X. Antagonizing Wnt and FGF receptors: an enemy from within (the ER). Cell  2005;120:156–158. [DOI] [PubMed] [Google Scholar]
  • 36. Quinodoz  M, Royer-Bertrand  B, Cisarova  K  et al.  DOMINO: using machine learning to predict genes associated with dominant disorders. Am J Hum Genet  2017;101:623–629. 10.1016/j.ajhg.2017.09.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Teschendorff  AE, Breeze  CE, Zheng  SC  et al.  A comparison of reference-based algorithms for correcting cell-type heterogeneity in epigenome-wide association studies. BMC Bioinformatics  2017;18:105. 10.1186/s12859-017-1511-5. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary_material_ddag088
supplementary_material_ddag088.zip (115.3KB, zip)

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Articles from Human Molecular Genetics are provided here courtesy of Oxford University Press

RESOURCES