Abstract

   Hemifacial microsomia (HFM) is a rare congenital disease characterized
   by a spectrum of craniomaxillofacial malformations, including
   unilateral hypoplasia of the mandible and surrounding structures.
   Genetic predisposition for HFM is evident but the causative genes have
   not been fully understood. Thus, in the present study, we used
   whole-exome sequencing to screen 52 patients with HFM for rare germline
   mutations. We revealed 3,341 rare germline mutations in this patient
   cohort, including those in 13 genes previously shown to be associated
   with HFM. Among these HFM-related genes, NID2 was most frequently
   mutated (in 3/52 patients). PED4DIP, which has not been previously
   associated with HFM, exhibited rare variants most frequently (in 7/52
   patients). Pathway enrichment analysis of genes that were mutated in >2
   patients predicted the “laminin interactions” pathway to be most
   significantly disrupted, predominantly by mutations in ITGB4, NID2, or
   LAMA5. In summary, this study is the first to identify rare germline
   mutations in HFM. The likely disruptions in the signaling pathways due
   to the mutations reported here may be considered potential causes of
   HFM.

   Keywords: hemifacial microsomia, whole-exome sequencing, rare germline
   mutations, pathway enrichment analysis, mandibular hypoplasia

Introduction

   Hemifacial microsomia (HFM), also known as craniofacial microsomia and
   oculo-auriculo-vertebral spectrum, is a rare congenital craniofacial
   malformation condition estimated to affect 1/3,000–1/5,000 live births
   ([31]Birgfeld and Heike, 2019). Patients exhibit a broad spectrum of
   symptoms of varying severity, ranging from microtia to complex
   developmental defects of the face (e.g., maxillary and mandibular
   hypoplasia) and surrounding soft tissue. These can be further
   complicated by ipsilateral orbital anomalies, facial paralysis,
   transverse facial cleft, and other rare phenotypes ([32]Tuin et al.,
   2015). HFM treatment is not only very painful and expensive but also
   challenging and predominantly symptomatic due to the complexity of
   induced defects.

   The pathogenic mechanisms that underlie HFM remain unknown; however,
   HFM etiology is widely recognized as being multi-factorial, including
   both environmental and genetic factors ([33]Beleza-Meireles et al.,
   2014). Despite the use of advanced gene sequencing technologies to
   identify chromosomal anomalies and candidate gene mutations in patients
   with HFM, the genetic causes of HFM remain elusive ([34]Terhal et al.,
   2006; [35]Kosaki et al., 2007; [36]Zhu et al., 2007; [37]Ala-Mello et
   al., 2008; [38]Ou et al., 2008; [39]Alasti and Van Camp, 2009;
   [40]Rooryck et al., 2009; [41]Huang et al., 2010; [42]Northup et al.,
   2010; [43]Rooryck et al., 2010; [44]Su et al., 2012;
   [45]Ballesta-Martinez et al., 2013; [46]Quintero-Rivera and
   Martinez-Agosto, 2013; [47]Torti et al., 2013; [48]Zielinski et al.,
   2014; [49]Beleza-Meireles et al., 2015; [50]Colovati et al., 2015;
   [51]Guida et al., 2015; [52]Lopez et al., 2016; [53]Zhang et al., 2016;
   [54]Berenguer et al., 2017; [55]Bragagnolo et al., 2018;
   [56]Spineli-Silva et al., 2018). Thus, in the present study, we
   analyzed rare germline mutations by using the whole-exome sequencing
   (WES) technology to detect potential genetic causes and novel
   therapeutic targets for HFM.

Materials and Methods

Patients

   Although no common HFM diagnostic criteria exist, most patients with
   HFM exhibit underdevelopment of the mandible, maxilla, ear, orbit, soft
   tissue, and/or facial nerve. In this study, our minimal diagnostic
   criterion was the presence of unilateral mandibular hypoplasia
   diagnosed by using computed tomography scans and excluded patients with
   extracranial symptoms to minimize differences caused by the
   presentation of multiple phenotypes. The study thus enrolled a cohort
   of 52 patients with a mean age of 4.99 years (range, 0.5–20 years)
   ([57]Figure 1), male:female ratio of 30:22, and a
   right:left-side-affected ratio of 26:26. All patients were further
   classified by using the OMENS+ classification system for HFM ([58]Tuin
   et al., 2015; [59]Table 1). All patients (or patient guardians)
   provided written informed consent prior to their participation in the
   study, which was approved by the Ethics Committee of Shanghai Ninth
   People’s Hospital, Shanghai Jiao Tong University School of Medicine.

FIGURE 1.

   [60]FIGURE 1
   [61]Open in a new tab

   Typical photos and computed tomography scan images of a patient with
   hemifacial microsomia. From left to right: front, left-lateral, and
   upward view.

TABLE 1.

   Clinical information for the 52 analyzed patients with hemifacial
   microsomia.
   Patient number Sample name Sex Age (years) Symptom laterality OMENS+
   classification
   1 B12815834 M 3 L O1M1E2N3S1
   2 B12816089 M 0.5 R O2M3E3N0S3
   3 WGC096878U F 1 L O1M3E3N1S2+C1
   4 WGC096879U F 7 L O0M2bE0N0S2+C1
   5 WGC096880U F 6 R O2M2aE2N2S1
   6 WGC096882U M 0.5 L O0M2aE0N0S1
   7 WGC098141U M 3 R O1M2bE3N3S1+C1
   8 ly4454 M 10 R O1M3E3N2S3
   9 ly4453 M 1 L O1M2aE2N3S1
   10 ly4693 F 2 L O1M1E0N0S1+C1
   11 ly4691 F 7 R O0M2aE3N0S2
   12 ly4692 M 11 L O1M2bE3N0S1
   13 ly4808 F 2 R O0M2bE0N0S1
   14 ly4809 M 4 R O0M2aE0N0S2+C1
   15 ly5117 F 17 L O0M2aE0N3S3
   16 ly5116 M 8 R O1M2aE0N0S1
   17 ly5115 M 18 L O1M2bE3N0S2
   18 ly5133 F 20 L O2M2aE0N0S1
   19 ly5112 M 7 L O3M2bE3N2S2
   20 ly5113 M 7 R O1M3E1N0S3
   21 ly5119 M 13 R O1M2bE3N1S2
   22 ly5172 M 7 R O1M2bE3N0S1
   23 ly5170 M 2 R O3M2bE3N2S1
   24 ly5171 F 2 R O3M2bE3N3S2
   25 ly5176 M 1 R O1M3E3N1S2+C1
   26 ly5178 F 4 L O0M1E0N0S1
   27 ly5174 M 2 L O1M1E3N0S1
   28 ly5173 M 1 L O0M3E3N3S2
   29 ly5175 F 1 R O0M2bE2N0S2+C1
   30 ly5268 F 2 L O1M2aE0N0S1
   31 ly5265 M 1 L O1M2bE0N0S1+C1
   32 ly5664 M 11 R O1M2bE0N0S2
   33 ly5758 M 6 L O3M2aE3N0S2
   34 ly5759 F 4 R O0M1E0N0S2+C1
   35 ly5830 F 3 L O3M2aE3N3S2
   36 ly5839 F 0.5 L O1M2bE1N0S2
   37 ly5838 F 4 R O2M2bE3N0S2
   38 ly5877 M 6 L O3M2bE3N2S3
   39 ly5900 M 8 R O2M2aE0N0S2
   40 ly6059 F 8 R O2M2bE3N0S2
   41 ly6058 M 7 L O2M2bE3N0S2
   42 ly6060 F 1.5 L O3M3E2N0S3+C1
   43 ly6070 M 5 R O0M2aE3N0S1
   44 ly6075 M 0.5 R O2M2bE3N0S1
   45 ly6077 M 5 L O1M2aE0N0S1
   46 ly6385 M 1.5 R O3M3E3N2S3+C1
   47 ly6278 M 1 L O1M2aE1N0S1+C1
   48 ly6676 M 0.5 R O3M3E0N0S2+C1
   49 ly6688 F 3 L O1M2aE0N0S2
   50 ly6694 F 8 R O0M2bE1N0S2+C1
   51 ly6711 F 0.5 L O1M3E3N2S2+C1
   52 ly6756 F 2.5 R O0M2bE3N0S2
   [62]Open in a new tab

   M, male; F, female; L, left side affected; R, right side affected.

WES

   Whole-exome sequencing was conducted as previously described ([63]Chen
   et al., 2018). Briefly, DNA was extracted from patient blood samples by
   using a Qiagen DNeasy Blood & Tissue Kit (QIAGEN, GmbH, Germany).
   RNaseA (QIAGEN, GmBH, Germany) was used to prevent RNA contamination.
   The purity and quality of the extracted DNA were determined by
   electrophoresis in a 1% agarose gel and by using a NanoDrop
   spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United
   States) and a Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA,
   United States). The purified genomic DNA was then sheared, and the
   samples (500 ng) were subjected to further purification, end repair,
   3’-end adenylation, indexed pair-end-adaptor ligation, ligation-product
   purification, and polymerase chain reaction (PCR) amplification. A WES
   library was constructed by using a SeqCap EZ capture kit (Roche) for
   exome capture and subjected to further PCR amplification, purification,
   validation, normalization, and pooling. An Illumina HiSeq Series
   Analyzer was used for library sequencing.

NGS (Next-Generation Sequencing) Data Processing

   Next-generation sequencing data processing was conducted as previously
   described ([64]Chen et al., 2018), with minor changes. Briefly, BWA
   software ([65]Li and Durbin, 2009) was used to map the generated
   sequencing reads to the human genome (hg19). VARSCAN2 software
   ([66]Koboldt et al., 2012) was used to detect germline mutations in the
   BAM files (minimum coverage, 20; minimum variant frequency, 0.08;
   P-value, 0.05). The fpfilter module of VARSCAN was used to identify
   false-positive variations. The resulting high-quality mutations were
   annotated by using SNPEFF ([67]Cingolani et al., 2012) and VEP
   ([68]McLaren et al., 2016) software.

Mutation Classification and Pathway Analysis

   Annotated mutations were selected for further analysis if they
   exhibited an alternative allele depth ≥ 10 and a population
   alternative-allele frequency < 0.0005 and were predicted to incur a
   ‘‘High’’ or ‘‘Moderate’’ severity impact. Small insertions and
   deletions (INDELs) and single nucleotide variants (SNVs) not predicted
   to be ‘‘benign’’ and ‘‘tolerated’’ by POLYPHEN2^[69]1 ([70]Adzhubei et
   al., 2010), SIFT^[71]2 ([72]Kumar et al., 2009), and CADD
   ([73]cadd.gs.washington.edu) ([74]Rentzsch et al., 2019) software
   tools, respectively, were used for following analysis as rare,
   potentially causative mutations.

   Pathway analysis was performed by using ConsensusPathDB software^[75]3
   ([76]Kamburov et al., 2013) by entering the official names of the
   detected SNVs and INDELs and by using all known human genes included in
   the software database as controls. The Reactome database^[77]4 was used
   as a reference database to evaluate pathway enrichment.

Results

Novel Mutations in the Known HFM Genes

   PubMed database mining identified a number of gene and chromosome
   regions that have been previously associated with HFM ([78]Table 2).
   Next, rare, potentially causative mutations that were detected in the
   52 patients in the present study were screened to identify novel
   mutations in the known HFM genes. This analysis revealed 13 novel
   mutations in HFM-associated NID2, PARD3B, CACNA1C, ERC1, IQSEC3, LZTR1,
   MED15, PLCD3, SALL4, SCR1B, SHROOM3, SLC6A13, and TCOF1, which
   collectively affected 14 (26.92%) of the analyzed patients ([79]Figure
   2A). Among these, mutations in NID2 affected the largest number of
   patients (3/52). The identified NID2 mutations, Asn202His, Asn658Lys,
   and Leu1136Arg, were each located in different protein domains
   ([80]Figure 2B). The PARD3B mutations Pro630fs and Arg773Gln were also
   detected in more than one patient ([81]Figure 2C).

TABLE 2.

   Summary of previous genetic studies of hemifacial microsomia.
   Study Mutated gene(s) symbol Abnormal chromosomal region(s)/chromosome
   abnormality Mutation type Utilized analysis method
   [82]Bragagnolo et al., 2018 HMX-1, BAPX1, EYA1, PUF60, NRBP2, SCRIB,
   GATA3, ATR-16, NF1, PRKX, RAS2 4p16.1, 4p16.3p15.33, Xp22.33–p22.31,
   22q11.21, 8q13.3, 8q24.3, 10q26.2q26.3, 10p13p14, 16p13.3,
   16p13.11p12.3, 17q11.2, Xp22.33 CNV (dup/del) Karotyping with
   G-banding, CMA
   [83]Spineli-Silva et al., 2018 YPEL1, MAPK1, ERK2, GSC2 22q11.2 CNV
   (del) MLPA, CMA
   [84]Berenguer et al., 2017 MYT1 20q13.33 SNV (c.323C>T, p.Ser108Leu)
   SNP array
   [85]Lopez et al., 2016 MYT1 20q13.33 SNV (c.314C > T, p.Ser105Leu) WES,
   transient knockdown in zebrafish
   [86]Colovati et al., 2015 ZNF74, KLHL22, MED15, SNAP29, LZTR1 22q11.21
   CNV (del) Genomic array analysis, MLPA
   [87]Zhang et al., 2016 ROBO1, GATA3, EPAS1, PARD3B, GBX2, SHROOM3,
   FRMD4A, FGF3, KLF12, EDNRB, NID2, SEMA7A, PLCD3 3p12.3, 10p14, 2p21,
   2q33.3, 2q37.2, 4q21.1, 10p13, 11q13.3, 13q22.1, 13q22.3, 14q22.1,
   15q24.1, 17q21.31 UK SNP, GWAS, GRAIL, DEPICT, qPCR, WGS
   [88]Beleza-Meireles et al., 2015 Not determined 22q11 CNV (dup/del)
   aCGH
   [89]Guida et al., 2015 ATP13A3, XXYLT1 3q29 CNV (dup) SNP array
   [90]Zielinski et al., 2014 OTX2 14q22.3 CNV (dup) WES, SNP array
   [91]Torti et al., 2013 Not determined 22q11.2 CNV (dup/del) aCGH, FISH
   [92]Ballesta-Martinez et al., 2013 OTX2 14q23.1 CNV (dup) Linkage
   analysis in families with autosomal dominant inheritance, aCGH
   [93]Quintero-Rivera and Martinez-Agosto, 2013 Not determined
   22q11.1–q11.21 CNV (tetrasomy) FISH, aCGH
   [94]Su et al., 2012 TCOF1 5q32–q33.1 UK PCR, direct sequencing
   [95]Rooryck et al., 2010 IQSEC3, SLC6A12, SLC6A13, JARID1A, CCDC77,
   B4GALNT3, NINJ2, WNK1, HSN2, RAD52, ERC1, FBXL14, WNT5B, ADIPOR2,
   CACNA2D4, LRTM2, DCP1B, CACNA1C, SPRY2 12p13.33, 47, XXX, Yp–q11.221.
   Yq11.222–q12, t(9;18) (p23;q12,2), 13q13.1 CNV
   (del/dup/trisomy/translocation) aCGH, QMF-PCR
   [96]Northup et al., 2010 Not determined inv(14) (p11.2q22.3) CNV
   (inversion) FISH
   [97]Huang et al., 2010 BIR1C, OCLN 5q13.2 CNV (del) Illumina
   HumanCNV370 Genotyping BeadChip, qPCR
   [98]Rooryck et al., 2009 WNT5B, CACNA1C del(12) (pter/p13.33) CNV (del)
   aCGH, QMF-PCR
   [99]Alasti and Van Camp, 2009 BAPX1, GSC, Hfm 4p15.33, 14q32.13, 14q32
   UK Review
   [100]Ala-Mello et al., 2008 Not determined 5p15.3–pter, 21q22.3–qter,
   21q22.11q22.12. CNV (del/dup) FISH, aCGH
   [101]Ou et al., 2008 SIX1, SIX6, OTX2 14q22.3–q23.3; 13q21.31-q21.32
   CNV (del/dup) Karotyping, FISH, aCGH
   [102]Kosaki et al., 2007 SALL1 16q12.1 SNV (c.1256T > A, p.L419X) PCR,
   direct sequencing
   [103]Zhu et al., 2007 ZIC3 Xq26.3 UK Analysis of Zic3 null mice
   [104]Terhal et al., 2006 SALL4 (exon 3) 20q13.2 UK PCR, direct
   sequencing
   [105]Open in a new tab

   aCGH, array-based comparative genomic hybridization; CMA, chromosomal
   microarray analysis; CNV, copy number variation; del, deletion; DEPICT,
   data-driven expression-prioritized integration for complex traits; dup,
   duplication; FISH, fluorescence in-situ hybridization; GRAIL, gene
   relationships across implicated loci; GWAS, genome-wide association
   study; MLPA, multiplex ligation-dependent probe amplification; PCR,
   polymerase chain reaction; QMF-PCR, quantitative multiplex
   fluorescence-PCR; qPCR, quantitative PCR; SNP, single nucleotide
   polymorphism; SNV, single nucleotide variant; UK, unknown; WES,
   whole-exome sequencing; WGS, whole-genome sequencing. XXX means XXX
   syndrome, also known as Triple X syndrome, and trisomy X, which is a
   rare and genetic disease. It is characterized by the presence of an
   extra X chromosome in each cell of a female.

FIGURE 2.

   [106]FIGURE 2
   [107]Open in a new tab

   Gene mutations identified among the analyzed hemifacial microsomia
   (HFM) patient cohort. (A) A total of 13 known HFM genes were found to
   be mutated in the studied patient cohort, affecting 14 (16.92%) of the
   52 analyzed patients. These included (B) three NID2 mutations,
   Asn202His, Asn658Lys, and Leu1136Arg, and (C) two PARD3B mutations,
   Pro630fs and Arg773Gln.

Genes With High-Frequency Rare, Potentially Causative Mutations in the HFM
Patient Cohort

   Genes with rare, potentially causative mutations that were detected in
   more than two patients were ranked by mutation frequency ([108]Figure
   3A). Of these, the most frequently mutated gene was PDE4DIP. Its
   mutations p.Ala141Thr, p.Lys154Arg, p.Cys19Gly, p.Asn1011Ser,
   p.Val120Ile, p.Gln535His, p.Pro2223Leu, and p.Gly2217Val affected the
   largest number of patients (7/52) ([109]Figure 3B).

FIGURE 3.

   [110]FIGURE 3
   [111]Open in a new tab

   Analysis of rare, potentially causative mutations identified in the
   analyzed hemifacial microsomia (HFM) patient cohort. (A) List of genes
   found to harbor variants in >2 patients, ranked by mutation frequency.
   (B) Details of mutations (p.Ala141Thr, p.Lys154Arg, p.Cys19Gly,
   p.Asn1011Ser, p.Val120Ile, p.Gln535His, p.Pro2223Leu, and
   p.Gly2217Val.) in PDE4DIP that exhibited the greatest number of
   mutations and was the most frequently mutated gene among the analyzed
   patients.

Pathway Analysis of Frequently Mutated Genes in the HFM Patient Cohort

   Next, a pathway analysis was conducted for all genes detected to harbor
   mutations among the 52 analyzed patients. This analysis showed that the
   most enriched pathways in the list of mutated genes were
   “extracellular-matrix organization,” “collagen-chain trimerization,”
   and “collagen formation” ([112]Supplementary Figure 1). Genes mutated
   in >2 patients were also subjected to the ConsensusPathDB enrichment
   analysis ([113]Figure 4), which found that the “laminin-interaction”
   pathway most frequently affected (17/52, 32.69%), predominantly by
   mutations in ITGB4, LAMA5, or NID2 ([114]Figure 5).

FIGURE 4.

   [115]FIGURE 4
   [116]Open in a new tab

   Results of the ConsensusPathDB enrichment analysis of genes found to be
   mutated in >2 patients with hemifacial microsomia (HFM).

FIGURE 5.

   [117]FIGURE 5
   [118]Open in a new tab

   Results of the pathway enrichment analysis of genes found to be mutated
   in the analyzed hemifacial microsomia (HFM) patient cohort. The
   “laminin-interaction” pathway was predicted to be most significantly
   affected by the identified mutations (most often in ITGB4, LAMA5, or
   NID2) and was likely disrupted in 17 (32.69%) of the analyzed patients.

Discussion

   To date, the molecular mechanisms that underlie HFM pathogenesis remain
   unclear; however, strong evidence suggests that these include genetic
   factors, as evidenced by the fact that several chromosomal
   abnormalities and gene mutations have been previously reported in
   patients with HFM ([119]Table 2). The largest study conducted to date
   was a genome-wide association study (GWAS) that was performed in 2016
   ([120]Zhang et al., 2016). However, that study encompassed all immature
   derivatives of the first and second pharyngeal arches. Thus, its
   results may not be closely related to the mandibular malformation.
   Furthermore, it is worth noting that very few risk variants identified
   by GWASs to date have been located in functionally significant protein
   regions (e.g., in the exons or the 5’-untranslated region)
   ([121]McClellan and King, 2010). Moreover, the GWAS method is generally
   considered insufficiently sensitive to detect rare variants for rare
   diseases ([122]McClellan and King, 2010). Given these shortcomings, NGS
   technologies such as WES have become increasingly important tools for
   studies aiming to elucidate the pathogenesis of rare diseases
   ([123]Abul-Husn et al., 2016; [124]Bilguvar et al., 2010; [125]Bowden
   et al., 2010; [126]Dewey et al., 2016; [127]Lupo et al., 2019). Here,
   we have performed the first ever WES study of a large cohort of
   patients with HFM.

   In the analyzed patient cohort comprising 52 individuals, we detected
   mutations in 13 genes that were previously associated with HFM, of
   which NID2, which encodes a member of the nidogen family of basement
   membrane proteins, was most frequently mutated. Previous studies have
   inferred a relationship between NID2 and osteogenic processes,
   consistent with the fact that the HFM phenotypic spectrum includes
   mandibular hypoplasia. For example, a previous study showed that NID2
   was markedly upregulated in demineralized osseous surfaces compared to
   its levels in mineralized osseous surfaces, suggesting that NID2 may
   act as a temporal migration guide ([128]Wischmann et al., 2018).
   Furthermore, NID2 has been implicated in osteoclastogenesis. TRAP5α,
   which is involved in osteoclast signaling and RANK signaling in
   osteoclasts, has been shown to interact with NID2 in cultured 3T3-L1
   mouse pre-adipocytes ([129]Patlaka et al., 2014). The fact that
   patients harboring NID2 mutations develop mandibular hypoplasia rather
   than mandibular agenesis is likely due to the fact that NID1 and NID2
   are found in all vertebrates and exhibit partial functional redundancy.
   Consistent with this, genetic deletion of either gene alone induces
   only mild defects in the mouse ([130]Bose et al., 2006).

   Among the genes with rare, potentially causative mutations detected in
   the present study, those in PDE4DIP were found to affect the greatest
   number of patients in the analyzed cohort. Mutations in PDE4DIP have
   been previously identified predominantly during sequencing of tumor
   samples, e.g., in prostate ([131]Gupta et al., 2017), ovarian ([132]Er
   et al., 2016), or lung cancer ([133]Li et al., 2015), as well as in
   adult pineoblastoma ([134]Snuderl et al., 2018). To date, however, to
   the best of our knowledge, there have been no published reports
   implicating PDE4DIP mutations in HFM or bone formation.

   Most previous studies investigating HFM pathogenesis have focused on a
   single gene mutation or mutations identified in a single patient;
   however, this approach is problematic given that the etiology of HFM is
   thought to be highly heterogeneous and dependent on genetic,
   epigenetic, and environmental factors. Thus, instead of focusing on
   single genes, in the present study, we conducted a pathway-based
   association analysis of generated WES data to identify common
   biological pathways that are likely to be affected in multiple
   (unrelated) patients with HFM. We suggest that this type of pathway
   analysis, which combines genomic and functional data and assesses the
   effect of multiple gene mutations, may be a better method to
   investigate the pathogenesis of rare diseases. Thus, we explored the
   distribution of signaling pathways known to be associated with genes
   that were mutated in >2 sporadic HFM cases. The “laminin-interaction”
   signaling pathway was predicted to be most frequently disrupted in the
   present cohort of patients with HFM, predominantly by mutations in
   ITGB4, NID2, or LAMA5. This is consistent with the fact that previous
   reports have demonstrated a close relationship (either positive or
   negative) between laminin interaction and bone formation during both
   osteogenesis and osteoclastogenesis ([135]Langen et al., 2017;
   [136]Susek et al., 2018).

   The limitations of the present study are that candidate genes and
   pathways were not confirmed; further functional experiments are needed
   to confirm the candidate genes and pathways. Furthermore, it should be
   noted that the diverse phenotypic spectrum of HFM is thought to be the
   result of gene-environment interactions and etiologic heterogeneity,
   which collectively cause incomplete penetrance and variable expression
   ([137]Birgfeld and Heike, 2019). Heterozygous mutations in reference
   alleles causing organ specificity in HFM may be affected by
   environmental and other factors. Many additional studies, such as RNA
   sequencing or whole-genome methylation studies, are needed to provide
   further information regarding this issue. In addition, we unfortunately
   failed to collect biological samples from the parents of the
   individuals involved. This was because the majority of patients came to
   our department with only one parent, and some of them refused to
   provide biological samples.

Conclusion

   In summary, this was the first study of rare germline mutations in a
   cohort of individuals with HFM. The likely disruptions in the signaling
   pathways due to the described mutations may be considered potential
   pathogenic causes of HFM and therefore, may serve as promising
   therapeutic targets.

Data Availability Statement

   The data presented in this study are deposited in NCBI online
   repositories. The BioProject ID is [138]PRJNA626382. The SRA accession
   numbers for these isolates are from [139]SAMN14639160 to [140]14639211.

Ethics Statement

   The studies involving human participants were reviewed and approved by
   Ethics Committee of Shanghai Ninth People’s Hospital, Shanghai Jiao
   Tong University School of Medicine. Written informed consent to
   participate in this study was provided by the participants’ legal
   guardian/next of kin. Written informed consent was obtained from the
   minor(s)’ legal guardian/next of kin for the publication of any
   potentially identifiable images or data included in this article.

Author Contributions

   XC, FL, YZ, and GC contributed to conception and design of the study.
   XC and ZM collected the clinical data and samples. FL performed the
   bioinformatics analysis. XC wrote the first draft of the manuscript.
   All authors contributed to manuscript revision, read, and approved the
   submitted version.

Conflict of Interest

   The authors declare that the research was conducted in the absence of
   any commercial or financial relationships that could be construed as a
   potential conflict of interest.

Acknowledgments