Genetic factors influencing risk to orofacial clefts: today’s challenges and tomorrow’s opportunities

Introduction

Orofacial clefts represent a group of anatomically distinct birth defects where there is a gap or break in normal features of the mouth, most commonly the roof of the mouth (the palate) or the upper lip or both. As shown in Figure 1, the most common anatomical forms of orofacial clefts include cleft lip (CL), cleft palate (CP), and cleft lip and palate (CLP). Other features of the face and mouth can also be affected by orofacial clefts, but the overwhelming majority of all cases are infants with one of these three common forms¹. Normally, development of the face and mouth occurs very early in pregnancy and reflects a complex process of cell growth and migration followed by fusion of symmetric structures to form the palate separating the mouth and nasal cavity, with the outer structures of the face developing before the inner structures. Collectively CL, CLP, and CP are the most common craniofacial birth defects worldwide, and affected individuals face feeding difficulties early in life and typically require multiple corrective surgeries, therapeutic dental procedures, and speech therapy throughout childhood^2–4. In addition, individuals born with an orofacial cleft have increased incidence of mental health problems and higher overall mortality rates at all stages of life, even in developed countries with good medical care^5,6. Where access to medical care is severely limited, infants born with CP or CLP generally have high mortality rates due to difficulty in breastfeeding, while untreated CL and CLP cases can face social discrimination throughout their lives. Thus, these birth defects have been subject to substantial selective pressure for most of human history.

Figure 1. Examples of orofacial clefts.

A. Left unilateral cleft lip. B. Bilateral cleft lip. C. Left unilateral cleft lip plus cleft palate. D. Bilateral cleft lip plus cleft palate. E. Cleft palate.

Although clefts develop early in pregnancy, most epidemiologic data are based on cleft frequencies at birth. The worldwide average birth prevalence of all orofacial clefts is 9.92 per 10,000 (close to one per 1,000), but there is substantial variation among populations⁷. In general, East Asian and Native American populations have substantially higher birth prevalence rates than do European and South Asian populations, while African ancestry populations have lower birth prevalence rates^8–10.

Mapping genes for orofacial clefts

Different types of genetic studies create the opportunity to map genes in the absence of a well-defined model of inheritance. There are two general types of approaches: “linkage analysis” and “association analysis”. Linkage analysis always requires multiplex families and tests for co-segregation of observed genetic markers and a hypothetical gene controlling the affected vs. unaffected phenotype within a family, while association analysis simply tests for differences in frequencies of markers in samples of affected and unaffected individuals from a population, which could reflect linkage at a population level (termed “linkage disequilibrium”). Linkage analysis is a powerful approach for mapping individual genes for traits following clear Mendelian patterns within multiplex families, but it is less effective in mapping genes for complex traits (such as orofacial clefts).

Both linkage and association analysis can be used on one genetic marker or on many scattered over the entire genome. Importantly, both approaches can be adapted to accumulate evidence when multiple causal variants (in different genomic locations) exist in the data being analyzed (multiplex families or groups of cases and controls), and thus both can be used effectively to study orofacial clefts at the genome-wide level.

Numerous groups used linkage studies to map genes in multiplex cleft families²³, but limits on the number and size of these families and the underlying high level of “linkage heterogeneity” (where different families yielded evidence of co-segregation to different genes) made it extremely difficult to obtain consistent results. A consortium of multiple research groups pooling genome-wide markers (~400 microsatellite markers) identified linkage to six different chromosomal regions (chromosomes 1q32, 2p, 3q27–28, 9q21, 14q21–24, and 16q24)^24,25. Further fine-mapping studies of these regions showed consistent evidence of linkage to chromosome 1q32 near the IRF6 gene. Interestingly, mutations in IRF6 account for most of the known cases of Van der Woude syndrome, a Mendelian autosomal dominant syndrome which includes orofacial clefts as a key phenotype²³. In addition, markers in and near this gene have also consistently shown evidence of association in population-based studies. Another gene, FOXE1 on chromosome 9q21, yielded consistent evidence of linkage in multiplex families^25,26 and recent association studies^27,28. Thus, linkage studies based on multiplex families successfully identified some causal genes²⁹, but clearly multiple genes control risk to orofacial clefts and these vary from family to family.

Association studies ask a much more general question than linkage analysis, and typically this involves rather simple statistical comparisons of marker allele or genotype frequencies between groups of unrelated people. While interpreting a positive result from linkage analysis represents more compelling evidence that a putative gene controls risk, these multiplex families are rare (i.e. most cleft cases do not have any affected close relative) and individual families are not representative of the full population. Interpreting statistically significant results from an association study is more ambiguous and less definitive than results from linkage analysis, and unrecognized bias in the study design can produce spurious results (especially if the cases and controls are not drawn from the same population). As with linkage studies, one, several, or millions of markers can be compared between case and control groups, but of course as the number of tests increases some accommodation must be made for false positive results obtained by chance alone (i.e. where in truth there is no real difference between groups). Conversely, false negative results can become important too, i.e. where a key risk gene is missed because the markers used for tests of association are not informative and do a poor job of covering that gene in one population or another.

The future challenges: expanding into genomic sequencing

In many ways, the progression in genetic studies of orofacial clefts has repeatedly applied standard tools of genetics to a disorder that is unexpectedly complex and heterogeneous. Clearly there is a genetic basis to the etiology of orofacial clefts; however, it must involve multiple genes (maybe dozens) and several environmental risk factors (which may interact with different genes). Our analytical tools, both statistical and technological (for genotyping and sequencing), have greatly expanded in recent years but always seem to fall short of the mark in dealing with the high level of etiologic heterogeneity for this relatively rare disorder. Technological advances now make it feasible to sequence the entire exome (i.e. all gene regions known to code for proteins) and even the entire human genome. However, we still must carefully consider the most appropriate study design and how to analyze sequence data most effectively.

Family study designs can be very effective in identifying causal genes⁴¹, but if many rare mutations can lead to the same phenotype, it is entirely possible that each family could reflect effects of a unique gene or unique combination of genes. Bureau et al.^42,43 recently showed how whole exome sequencing (WES) data in distant affected relatives (second-degree and more distant relatives) drawn from multiplex families originally recruited for linkage studies could be used to identify rare variants that modify risk to orofacial clefts. From a long list of candidate genes, this approach identified one novel variant in CHD1 shared by three affected second cousins in a single multiplex family. CHD1 is a strong candidate gene because mutations in this gene can cause both gastric cancer and orofacial clefts, and several studies have shown evidence of association or excess of rare variants in this gene^44–47. A comprehensive whole exome analysis revealed a separate novel variant in ADAMTS9 achieving statistical significance over the entire exome in a small number of families⁴². While this gene had been identified as potentially causal in mouse models, this is the first evidence from human studies, and the risk allele is extremely rare. This same method of testing for rare shared variants can be applied to structural variants, such as deletions detected from WES data⁴⁸.

Leslie et al.⁴⁹ demonstrated the potential of sequencing studies in a large study of case–parent trios (where 13 regions around candidate genes were sequenced). This case–parent trio design allows the detection of de novo mutations, where a new mutation (not present in the parents) occurs in the child. This type of mutation simply cannot be identified in case–control studies, and such mutations are by their nature extremely rare. Also, if a de novo mutation lies outside any known coding or regulatory region, it is difficult to judge its true functional status. This group concluded de novo variants probably don’t represent a major etiologic component for orofacial clefts, although they may be relevant for other complex diseases.

This case–parent trio study of sequencing data yielded evidence consistent with previous GWAS studies, where multiple genomic regions showed strong statistical evidence of linkage and association for common polymorphic markers, but it also was able to analyze rare variants that could themselves be directly causal for orofacial clefts. Careful analysis of low-frequency and rare variants using available methods for collapsing or combining multiple variants within a gene to assess the overall impact of that gene on risk yielded mixed results: two of 13 candidate regions (NOG and NTN1) showed some statistical evidence for distinct rare variants within the gene being important in controlling risk, while four of the 13 genes yielded no such evidence (BMP4, FGFR2, MSX1, and PTCH1). Because obtaining statistical evidence for rare variants is always difficult, even in large studies, building a biological case for a putative causal mutation requires functional studies for confirmation⁴⁹.

Large-scale whole genome sequencing (WGS) studies are now underway and are being applied to a number of complex and heterogeneous diseases (including orofacial clefts). Practical challenges in using WGS data include cleaning and processing the huge volume of data. The amount of data scales up dramatically from GWAS to WES to WGS: GWAS data represent hundreds of thousands or millions of typed markers (and potentially millions more imputed markers); WES represents 1–2% of the human genome sequenced 30–50 times and aligned to yield a consensus sequence for ~50 Mb of the genome; WGS represents similarly aligned read data on all 30 billion base pairs of the genome (although even with WGS there are gaps and some regions are not well covered). The primary advantage of WGS data is that it can be used to analyze any type of sequence variant from the reference genome (single nucleotide variants or SNVs, including the common SNP markers), small deletions (termed indels), or larger structural changes (e.g. deletions and duplications, which are collectively called copy number variants or CNVs, plus more complex structural variants such as inversions and translocations where the order of genes may be changed through chromosomal breakage and exchange). The challenge comes in defining the particular assessment/comparison being made with variants identified by WGS data (which is always a reflection of the study design and hypothesis being considered), and then interpreting any “significant” differences from expected. The study design always determines the hypothesis to be tested (strict Mendelian transmission, co-segregation, differences in frequencies, etc.), but the huge number of tests must be considered, along with the potential biases inherent in the study design. Dealing with WGS data will involve tens of millions of markers or SNVs on individuals, and most of these will be quite rare. Most observed sequence variants will occur outside known coding regions, and our ability to recognize important regulatory regions remains limited.

The more important challenge facing genetic studies of orofacial clefts is to develop strategies for dealing with multiple causal genes within any data set and applying these strategies effectively to orofacial clefts as a complex and heterogeneous birth defect. The full range of study designs (selected relatives from large multiplex families, nuclear families, case–parent trios, and groups of unrelated individuals) is available, but interpreting the biological function of a novel variant detected by sequencing will not be simple. Furthermore, because any given patient could represent the effects of one or more variants in several different genes, translating new genetic findings into meaningful medical or public health applications remains a daunting task. In reality, for orofacial clefts (or any other birth defect), because the first occurrence occurs well before birth, it is difficult to define opportunities to intervene. Nonetheless, the tools to detect genes controlling risk to orofacial clefts have expanded greatly in this first part of the 21^st century, and it is our challenge to use them as effectively and efficiently as possible. It is likely that there will be classes of “causal genes” and “genetic risk factors”: a few genes with “major effects” may simultaneously control truly Mendelian malformation syndromes and show association with apparent non-syndromic forms of orofacial clefts, while possibly many “genetic risk factors” have the typical modest estimated effect sizes identified through GWAS but nonetheless achieve genome-wide significance as studies are combined via meta-analysis.

Abbreviations

CI, confidence interval; CL, cleft lip; CL/P, cleft lip with or without cleft palate; CLP, cleft lip and palate; CP, cleft palate; GWAS, genome-wide association study; GxE, gene–environment; SNP, single nucleotide polymorphism; SNV, single nucleotide variant; WES, whole exome sequencing; WGS, whole genome sequencing.