Sequence characterization of RET in 117 Chinese Hirschsprung disease families identifies a large burden of de novo and parental mosaic mutations

Pathogenic gene variation is a significant contributor to rare diseases, especially in children [1]. Thus, many genetic mutations of early development are inherited by children from their parents through the germline and are present in all cells of that individual, while others, mosaic or somatic mutations, may be acquired postzygotically and are present in only a subset of an individual’s cells [2]. It has long been known that cancer is a mosaic genetic disorder. However, a growing body of research suggests that analogous mosaicism may be a frequent feature in a diverse range of childhood disorders, including cerebral cortical malformations, autism spectrum disorder, epilepsies and other neuropsychiatric diseases [3‐6]. In a previous study of Hirschsprung disease (HSCR) families, we identified mosaicism in 6 of 8 (75%) isolated cases [7]. This high frequency was surprising and prompted us to further investigate the frequency and nature of RET mosaic pathogenic variants.

HSCR or congenital aganglionosis, a heterogeneous genetic disorder, is characterized by the lack of ganglion cells along varying lengths of the intestine resulting in the major cause of functional obstruction in children. According to the length of aganglionosis, the disorder is categorized into three types: short-segment (aganglionosis segment up to the upper sigmoid colon), long-segment (aganglionosis beyond the splenic flexure) and total colonic aganglionosis (TCA) [8]. The incidence of HSCR varies and is 15, 21 and 28 cases per 100,000 live births in infants with European, African and Asian ancestry, respectively. Genetic studies during the past 25 years have identified rare coding variants in 14 genes that together explain ~ 10% of HSCR cases [9‐11]. Of these, the most frequent coding mutations occur in RET, which encodes a receptor tyrosine kinase that regulates the proliferation, differentiation and migration of the enteric neural crest cells to enteric neurons. However, family studies of these pathogenic variants demonstrate incomplete penetrance and variable expressivity, the causes of which remain largely unexplained [9, 12].

Numerous studies of RET pathogenic variants in HSCR show that they occur in 8.9–16.7% of cases with a contribution from de novo variants (DNVs) which occur in the parental germline [13, 14]. However, family studies of these variants are infrequent so that the distribution of Mendelian inherited versus DNVs is unknown, making risk prediction and genetic counselling of HSCR uncertain. Here, we set out to perform a prospective study of 117 HSCR parent-affected child trios to determine the frequency of RET Mendelian inherited, parental mosaic or true DNVs. Furthermore, we explored the mutant allele distribution patterns in multiple somatic tissues and gonadal tissue, and compared the detection accuracy of three commonly used molecular methods.

Several lines of evidence suggest that the mechanism of RET involvement in HSCR is the result of partial or total loss of RET function, with mutant penetrance depending on the degree of functional loss. We presume that the threshold is > 50% loss because heterozygotes for a RET nonsense mutation do not have 100% penetrance in humans [11] but homozygotes for a Ret null mutation do have 100% penetrance in mice [19]. One missense variant (p. Asp489Asn) was confirmed to be inherited in a male patient’s unaffected mother in our study. Similarly, multiple putative RET mutations were inherited from one of the unaffected parents in a previous study. The underlying mechanism, as stated, is that although a substitution is not thought to be causative of disease in and of itself, it may influence the phenotype, especially given the multigenic nature of HSCR [20, 21]. Here, we identify 2 patients with RET nonsense mutations and 1 with frameshift mutation, all resulting in a premature stop codon that is expected to produce non-functional RET. In addition, most of the RET HSCR missense mutations involved amino-acids conserved in multiple species and were scattered in the functional domain of RET, which is consistent with the diversity of events predicted to be associated with gene inactivation [21‐24]. In brief, those lying within the extracellular domain are proposed to interfere with RET maturation and its translocation to the plasma membrane. Variants residing within the TK domain are likely to reduce the catalytic activity of the receptor, and mutations sitting in the region around Y1062 may compromise the efficiency with which RET binds to its effector molecules. Finally, we also discovered 1 synonymous and 1 splicing variant in families 14 (HSCRFM007) and 15 (HSCRFM156). At face value, these variants are likely benign; however, their absence in large databases suggests that they may have a functional effect acting through activating or abrogating cryptic splice sites or their enhancers [25].

A second intriguing part of this study is the discovery of only one full heterozygote inherited from constitutional heterozygous parent (6.7%) and three heterozygotes inherited from parental mosaics (20%). Genomic mosaicism results from postzygotic events occurring predominantly in early embryogenesis but can arise throughout life and result in genetically distinct cell lines within one individual. Human gastrulation, the process by which the three germ layers are established, is thought to occur at approximately day 16. Primordial germ cells are thought to arise from the primary ectoderm during the second week of development. Therefore, the presence of a somatic variant in blood, saliva (mesodermal tissues), urine (endodermal origin) and hair root bulbs (ectodermal tissue) indicates that the variant arose early enough to potentially also be present in germ cells and is therefore transmissible to the next generation. This high rate of mosaicism suggests that in some families with apparent DNVs, the pathogenic variant is actually mosaic in the parents, and indeed inherited, and that the risk of HSCR in subsequent children is not infinitesimal. This distinction between non-mosaic inherited DNV (heterozygous in proband and variant not detected in parent) and mosaic inherited DNV (heterozygous in proband, and variant detected mosaic in parent) is important for genetic prognosis and counseling. However, it is very difficult to distinguish true DNV from low allele fraction mosaic mutations in reality.

Here, we surveyed 14 families with both NGS and ddPCR on blood DNA. The degree of allelic ratio bias in our NGS results is larger than that in most previous studies, the source of which is still unknown. Among those four where deviation from the expected ~ 50/50 allele ratio of true heterozygosity was observed in NGS, three individuals (HSCRFM197, HSCRFM230 and HSCRFM156) were covered by less than 500X. One exception was HSCRFM198, which had a mutant ratio of 42.2% at a whole coverage of 1336X. In contrast, one sample (HSCRFM181) was covered by less than 500X but ddPCR correctly recognized the mutant status (allele ratio 49.1%), which is not surprising given the nature of the method. NGS can serve as an effective and less expensive technique for screening and quantifying variants; however, it should be noted that many factors may interfere with the results (quality) of the reads/coverage/biallelic ratio by NGS, such as DNA quality (affect baits affinity), biased PCR amplification, sequence context of the variant, pooled DNA isolated from multiple cells as template, the short-read length, sequencing errors and bioinformatic workflow which may filter out biased allele calls. In ddPCR assays, by contrast, template DNA is partitioned into tens of thousands of individual droplets so that at low DNA concentrations the vast majority of droplets contain no more than one copy of template DNA. PCR within each droplet produces a fluorescent readout to indicate the presence or absence of the target of interest, allowing for the accurate “counting” of the number of copies present in a sample [16]. The number of partitions is large enough to assay somatic mosaic events with frequencies down to less than 1%. This excellent accuracy is credited with increased signal-to-noise ratio and removal of PCR bias. As we have shown here, by examining 6 positive-control samples carrying different levels of mosaicism, both ddPCR and ADS surpass the performance of the prevailing NGS and Sanger sequencing.

Interestingly, 11 families (out of 15, 73.3%) were determined to carry non-mosaic inherited DNVs in RET, at a significantly higher rate than in any previously reported study: 42.9% in Indonesia, 43.8% in France, and 58.3% in Hong Kong, China [14, 26, 27]. Of these, 72.7% are likely pathogenic. These data raise two issues. First, the pathogenic nature of the DNV needs to be established since RET is a commonly mutable gene [28], or rather, its mutants in sperm have a survival advantage [29]. Second, why is the DNV mutation frequency so high? Although our finding may be a chance event it is unlikely because we have observed this before in our studies [7]. A possible and intriguing reason is that many RET DNVs may not be disease-causing or be penetrant on their own but can be in a specific RET genetic background that is more permissive in infants with Chinese (Asian) than European ancestry; note that the frequency is also high in the Hong Kong Chinese sample but not the Indonesian one. A candidate for this difference is the RET enhancer polymorphism rs2435357 (MCS + 9.7 or RET+ 3) at which a hypomorphic allele that significantly reduces RET transcription, has a background allele frequency of 24% (homozygotes ~ 6%) in Europe but 45% (homozygotes ~ 20%) across Asia, a ~ 4-fold difference [30, 31]. MCS + 9.7 does not act on RET transcription alone but in concert with at least two other enhancers that also contribute to this genetic background difference [31]. Thus, we hypothesize that this increased widespread susceptibility in China allows a greater number of milder RET variants to be HSCR-associated, including DNVs, accounting for the higher frequency of DNVs in Chinese HSCR patients. Regardless, both paternal age and the sequencing sensitivity of different technologies should be taken into account when making the final statement.

Every human gene is subject to random mutation multiple times within each individual. However, most variants are either benign or never reach a fraction high enough to cause disease. Thus, whether a pathogenic variant is disease penetrant or not depends on the physiological function of the encoded molecule and the fraction of cells possessing the mutation in a given tissue. Somatic mutations that lead to a gain of function or growth advantage might cause disease if they are present in even one cell, as in cancer. On the other hand, somatic mutations that lead to a loss of function might need to occur in a larger clonal fraction in order to cause a clinical phenotype. Therefore, for every deleterious somatic mutation there likely exists a threshold mosaic fraction above which the mutation causes disease but below which it does not and so remains undetected [32]. Of course, for de novo changes the penetrance is likely dependent on the number of cells affected, as well as the specific mutation, the disease involved, and the genetic background of the individual. Thus, distinguishing non-mosaic inherited DNV (germline DNV) from true postzygotic DNVs is important, as is the threshold mosaic fraction. These analyses need to be quantitative because in some cases, in clinically significant cortical malformations, the disorder can result from somatic mutations in as few as 1% of cells [33]. The threshold mosaic fraction for HSCR is important to investigate because it is likely a critical determinant of HSCR penetrance and expressivity.