Exome sequencing findings in children with annular pancreas

Abstract Background Annular pancreas (AP) is a congenital defect of unknown cause in which the pancreas encircles the duodenum. Theories include abnormal migration and rotation of the ventral bud, persistence of ectopic pancreatic tissue, and inappropriate fusion of the ventral and dorsal buds before rotation. The few reported familial cases suggest a genetic contribution. Methods We conducted exome sequencing in 115 affected infants from the California birth defects registry. Results Seven cases had a single heterozygous missense variant in IQGAP1, five of them with CADD scores >20; seven other infants had a single heterozygous missense variant in NRCAM, five of them with CADD scores >20. We also looked at genes previously associated with AP and found two rare heterozygous missense variants, one each in PDX1 and FOXF1. Conclusion IQGAP1 and NRCAM are crucial in cell polarization and migration. Mutations result in decreased motility which could possibly cause the ventral bud to not migrate normally. To our knowledge, this is the first study reporting a possible association for IQGAP1 and NRCAM with AP. Our findings of rare genetic variants involved in cell migration in 15% of our population raise the possibility that AP may be related to abnormal cell migration.

remain asymptomatic, but older studies have estimated it to be approximately 1 in 10,000 from autopsies and surgical cases (Ravitch & Woods, 1950).

| Embryologic development
The pancreas develops from two endodermal buds, the ventral and the dorsal.They develop by the fourth week post-conception.Around the 7th week, the ventral bud rotates clockwise, fuses with the dorsal bud and forms the adult pancreas (Patra et al., 2011).

| Current theories
It has been generally accepted that the origin of AP lies in the ventral pancreatic bud, but the exact mechanism has yet to be determined.Three major theories have been proposed: (1) Lecco's theory in which the ventral bud adheres to the duodenal wall before rotation (Gross & Tague, 1944;Lecco, 1910;Lehman, 1942;Suda, 1990), (2) Baldwin's theory in which there is persistence and enlargement of the ventral bud (Baldwin, 1910;Moore, 2003;Nobukawa et al., 2000), and (3) Tieken's theory in which hypertrophy of the ventral and dorsal buds form an annular structure (Hill, 1993;Tieken, 1901).So far, none of the proposed theories have provided an adequate explanation of the variable ductal anatomy and no consensus has been reached (Etienne et al., 2012;Fu et al., 2005;Godil & McCracken, 1997;Nobukawa et al., 2000).There may be more than one mechanism for the development of AP as the existing embryological hypotheses cannot adequately explain it.

| Study design
The data (including information on birth defects and medical and demographic data) were obtained from the California Birth Defects Monitoring Program (CBDMP), a population-based surveillance system for collecting information on children born between 1984 and 2014 with congenital malformations (Croen et al., 1991).The study excluded all aneuploid and known syndromic cases of AP.Demographic and diagnostic information from medical records for liveborn and stillborn fetuses (defined as ≥20 weeks gestation) were collected by trained data collectors.Overall, ascertainment for major malformations has been estimated as 97% complete (Schulman & Hahn, 1993).The DNA for the study was obtained from newborn screening filter paper, so only live born infants were included.
Data were de-identified and personally identifiable information was not provided to the investigators to ensure confidentiality and patient privacy.Demographic characteristics of all births in the same region and time from which the AP infants were identified were obtained from birth certificates.These data included maternal raceethnicity, maternal age, gestational age at delivery, infant's sex, and birthweight.Differences in demographic factors were assessed by chi-squared and Fisher's exact tests and associations were estimated by odds ratios (ORs).The protocol for this study, including the state-approved guidelines for the use of newborn bloodspots, was approved by the California Department of Public Health Institutional Review Board (IRB) 14-08-1693 and by the NIH Office of Human Subjects Research Protections.The guidelines permit research use of archived newborn bloodspots without further individual consent.

| Specimen processing and sequencing
Residual filter paper after newborn screening was obtained and four 3 mm dried blood spot (DBS) punches were washed at 4C for 18-24 h in PBS with 0.1% BSA.DNA was extracted from each dried DBS punch individually using GenSolve reagents (GenTegra) and purified using QIAmp DNA Blood Mini reagents (Qiagen).After extraction and purification, DNA from the four punches was combined and dried to a final volume of 100 μL prior to quantification by PicoGreen (Molecular Probes).Samples were then further dried to 2 μL.
Exome sequencing of the DNA samples from 115 patients was then performed at the University of Minnesota Genomics Center (UMGC).Targeted capture libraries were generated using the Agilent SureSelect Human All Exon v7 kit and were sequenced on an Illumina NovaSeq SP producing 151 bp paired end reads.

| Filtering and prioritization
Our genotyping pipeline identified 479,669 unique variants.Of those 372,783 (77.7%) were deemed high quality (not in a GATK tranche, depth > 10, genotype quality [GQ] >50).Of those, 75,949 affected an amino acid (missense or nonsense variants) and were deemed putatively functional.We then focused on genes in which loss-offunction variants were found in more than one individual or in which missense or/and loss-of-function variants were found in more than five individuals.Using the gene damage index (GDI; Itan et al., 2015) we prioritized novel and known gene candidates, focusing on genes with mutational burdens in the lowest 75th percentile.Of the 75,949 variants, 10,462 met our primary criteria (maximum allele frequency [AF] ≤0.005 and in a gene with a GDI <75).In addition, all putatively functional (missense or loss-of-function) rare variants (AF ≤0.005) in genes previously associated with AP were manually reviewed using the Integrative Genomics Viewer (IGV) v2.4.13 (Robinson et al., 2011), regardless of the quality.Filtering criteria are summarized in Table S1 which specifically describes each of the filtering thresholds applied to nominate the variants and genes of interest.

| RESULTS
The prevalence of AP in this study population was 1.7 per 100,000.Demographic characteristics of the cases were compared to the same underlying population of births from which the AP cases were ascertained based on a total of 6,783,570 births in the time period and region (Table 1).Among the 115 cases that were sequenced, 64 (55.6%) were females and 70 (61%) were full term.AP-affected infants were more frequently African American than the source livebirth population (13% vs. 6.5%;OR = 2.20, 95% CI 1.27-3.81,p = 0.011).As expected with many major birth defects, many children were delivered prematurely.Therefore, the percentage of preterm births was higher than in the general population and there was a significantly higher percentage of low birth weight infants among cases (34.8% vs. 5.2%).Accompanying birth defects of AP-affected infants are shown in Table 2.
Of the total of 115 cases sequenced, 91 passed sample quality filters (FREEMIX <0.05).Of the 91, 78 cases harbored at least one variant passing our variant thresholds.Based on our criteria for prioritizing variants (≥2 loss-of-function variants or ≥6 loss-of-function or missense variants) two genes emerged as candidates for AP, IQGAP1 (IQ motif-containing GTPase activating protein 1), and NRCAM (neuronal cell adhesion molecule).Seven cases had a single heterozygous missense variant in IQGAP1 (none were recurrent).All seven variants were present in the gnomAD database and their allele frequencies are noted in Table 3.None of these variants are present in ClinVar.In addition, five of seven variants are predicted by CADD score to be within the top 1% of deleterious variants in the genome.Seven other infants had a single heterozygous missense variant in NRCAM (none were recurrent), five of them with CADD scores >20 (Table 3) Only one of the seven variants was present in gnomAD, with allele frequency 0.0003247.Neither of these genes was associated with any previously described syndromes.We also looked at known databases (including OMIM, GeneCards and Uniprot) but they did not provide any additional information related to AP.According to the criteria for variant classification from the ACMG, all variants in both genes are classified as VUS except one that is classified as likely benign and one as likely pathogenic.
The classification of each specific variant is described on Table 3.The likely pathogenic variant (c.3469C > T, p. Pro1157Ser) has a relatively low frequency in gnomAD (0.0002).It notably falls in IQGAP1, a gene expressed during human pancreatic development.In addition, this variant disrupts the RAS-GTP activation domain, which supports its potential damaging effect.
We also looked for variants in genes previously reported to be associated with AP.We checked RFX6, FOXF1, PDX1, IHH, and SHH, and we identified (≥1 loss-of-function or missense variants, AF ≤0.005) two rare heterozygous missense variants in the PDX1 gene (p.Ala104Asp) and the FOXF1 gene (p.Met257Ile; Table 4).Next, we checked for X-linked recessive genes and we identified three variants in IQSEC2 (p.His1246_His1247del, p.Pro1109Leu, and p.Gly115Cys; Table 5).No compound heterozygotes or autosomal homozygotes passed our IGV review.Additional variants that passed our criteria of the autosomal dominant model are noted on Table S2.

| DISCUSSION
We identified two genes, IQGAP1 and NRCAM, that have a high number of rare functional variants in the 91 AP cases that were analyzed.Notably, both genes are involved in cell migration.Single heterozygous missense variants in IQGAP1 were identified in seven cases.IQGAP1 is a scaffold protein participating in multiple cellular functions including actin organization and regulation of cell motility (Bensenor et al., 2007; Figure 1).Cell migration is a multistep process that requires changes to the cytoskeleton (Ridley, 2001).IQGAP1 regulates cytoskeletal function by interacting with Cdc42 and Rac1 (Hart et al., 1996;Joyal et al., 1997), CLIP-170 (Fukata et al., 2002), and actin (Erickson et al., 1997;Mateer et al., 2002).Cell migration is initiated in response to extracellular signals such as cytokines and growth factors (Choi et al., 2013) that activate Rac1 and Cdc42 (members of the Rho family GTPases); activated Rac1 and Cdc42 recruit IQGAP1, which stabilizes them in the active GTP-bound state (Brandt et al., 2007;Le Clainche et al., 2007;Swart-Mataraza et al., 2002).In addition, in directionally migrating cells, IQGAP1 accumulates on the leading edge where it facilitates actin polymerization and directly crosslinks actin filaments (Briggs & Sacks, 2003;Hart et al., 1996;Mataraza et al., 2003).Then, it recruits adenomatous polyposis coli (APC) to actin filaments and captures the plus-ends of the microtubules through CLIP-170 (a microtubule-binding protein; Fukata et al., 2002;Noritake et al., 2005).Following that, APC stabilizes the microtubules, which is essential for the actin meshwork at the leading edge to be stable (Noritake et al., 2005).IQGAP1 has also been characterized as an oncogene (Dong et al., 2016;Xia et al., 2014) implicated in several types of cancer including pancreatic (Casteel et al., 2012;Hu et al., 2019;Liu et al., 2010;Wang et al., 2013Wang et al., , 2014;;Xia et al., 2014;Zhao et al., 2014).In addition, IQGAP1 is known to mediate the Wnt/β-catenin signaling pathway to induce cancer metastasis (Peng et al., 2021).Based on the cancer data, we hypothesize that genetic variants in IQGAP1 may result in decreased cell migration.Thus, the pancreatic buds may not migrate normally.
Our hypothesis is supported by a report that a knock down of IQGAP1 by both transient and stable expression of small interfering RNA (siRNA) resulted in significantly slowed cell migration (Mataraza et al., 2003).That study showed that changing the intracellular concentration of IQGAP1 changed cell migration; overexpression of IQGAP1 resulted in increased cell motility, while decreased levels decreased cell motility (Mataraza et al., 2003).In addition, in a different study, a dominant negative IQGAP1 decreased active levels of Cdc42 and Rac1 and so cell motility was decreased significantly (Swart- Mataraza et al., 2002).
The second gene we identified, NRCAM, which was found in seven cases, encodes a neuronal cell adhesion molecule of the superfamily of immunoglobulins (Grumet, 1991(Grumet, , 1997;;Grumet et al., 1991;Wang et al., 1998).First described in neuron-neuron adhesion, NRCAM, was then reported to potentially play a role in cell-cell communication via signaling from its intracellular domain to the actin cytoskeleton during directional cell migration.Like IQGAP1, it is overexpressed in various types of cancer, including pancreatic carcinoma (Dhodapkar et al., 2001;Wang et al., 1998) again suggesting that under-expression due to genetic variants could interfere with normal migration.
In addition, we provided confirmatory evidence regarding two genes previously reported to be associated with AP: PDX1 and FOXF1.PDX1 is a key regulatory element of pancreas development (Pethe et al., 2021).The association between AP and the PDX1 gene was previously suggested in a case study of a patient with AP and other related abnormalities including duodenal atresia, hypoplastic gallbladder, permanent neonatal diabetes mellitus, and exocrine pancreatic insufficiency (Kulkarni et al., 2017).There are reports that mutations in PDX1 resulted in pancreatic agenesis (Schwitzgebel et al., 2003;Stoffers et al., 1997).In a case report a male patient had AP, while his sibling had a FOXF1 variant and appeared to have, among other defects, dorsal pancreatic agenesis (Reiter et al., 2016).FOXF1 has been found to be involved in the development of the foregut (Madison et al., 2009;Mahlapuu et al., 2001).
4.1 | How do our findings fit to the existing hypotheses?
Going back to the three theories regarding the pathogenesis of AP, it is clear that our findings do not Note: AD allelic depth for the reference allele (listed first) and the alternate allele (listed second); since three of these infants are male, we expect the depths of the reference allele to be zero for these hemizygous calls; the small number of reference allele depths between 2 and 4 seen here are likely due to sequencing error or other technical artifact.
Abbreviations: AF, allele frequency, CADD, combined annotation dependent depletion score, GDI, gene damage index, GQ, genotype quality.specifically fit any of them.Both genes we identified are involved in actin cytoskeleton which mediates various important cellular processes such as cell migration.This might be a fourth causal pathway that may be relevant in a subset of cases not explained by the previous three theories.Strengths of our analysis include a well-characterized, population-based study sample in which a very high percentage of major malformations were identified.Additionally, our study is the largest exome sequencing study of AP cases to date.Limitations of the study include that only limited amounts of DNA could be recovered from the newborn blood spots.This led to lower coverage for some individuals and also means that all of the DNA available was used for WES, leaving none that could be used for confirmatory Sanger sequencing for variants of interest.That said, we have done a careful review of the raw data using IGV and have included screenshots of them in Supplemental Material.Many technical artifacts leading to spurious genotype calls, such as mismapping, can be observed when visualized in this way, and we did not find any such patterns for the variants reported in our tables.In addition, whether the parents were affected is not known.Lastly, functional studies were not conducted because none of the variants were recurrent.Even though our study has limitations, it presents new data underlying the genetic etiology of non-syndromic AP to stimulate future research.

| CONCLUSIONS
To our knowledge this is the largest exome sequencing study in children with AP.Fifteen percent of our cases had variants in two genes involved in cell migration.We hypothesize that these variants in IQGAP1 and NRCAM may help explain the failure of the ventral pancreatic bud to migrate.Our results require confirmation, but they may provide new directions for research.

AUTHOR CONTRIBUTIONS
All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.J.L.M, and G.M.S. made the concept.N.P. and J.L. performed and analyzed the sequencing experiments.W.Y. performed the statistical 2 analysis.G.P interpreted the sequencing results and drafted the manuscript.G.P, N.P, J.L, W.Y, S.R, G.M.S, and J.L.M. edited the manuscript.

aAD
>10, GQ >50, AF ≤0.005 in public databases, GDI <75th percentile.b Maximum AF observed in public databases for any subpopulation.F I G U R E 1 Role of IQGAP1 in cell polarization and directional cell migration.APC adenomatous polyposis coli, GDP guanosine diphosphate, GEF guanine nucleotide exchange factors, GTP guanosine triphosphate.Cell migration is initiated by extracellular signals (GEFs), such as chemokines and growth factors, that activate Rac1 and Cdc42 (GTP-bound) at the leading edges.Activated Rac1 and Cdc42 induce polymerization of actin filaments and recruit IQGAP1 that crosslinks the actin filaments.Following that, IQGAP1 recruits APC to the actin filaments and IQGAP1, through CLIP-170, captures the plus-ends of microtubules.Then, APC stabilizes the microtubules, which are essential for stable actin meshwork at the leading edges.

Characteristic AP cases (n = 115; %) CBDMP live births (n = 6,783,570; %) Overall p value a
Note: Gestational weeks were not included in the analysis.aWaldchi-squared test excluding missing.bFisher'sexact test comparing African Americans versus White.cThestate only provided data in categories to preserve confidentiality.
Variants in IQGAP1 and NRCAM meeting our primary filtering criteria.aTranscriptsusedfor annotation are provided in TableS1with Integrative Genomics Viewer screenshots.Abbreviations: ACMG, American College of Medical Genetics; AD, allelic depth for the [Allele1, Allele2]; AF, allele frequency; CADD, combined annotation dependent depletion score; GDI, gene damage index; GQ, genotype quality; LB, likely benign; LP, likely pathogenic; VUS, variant of unknown significance.Variants in genes previously reported to be associated with annular pancreas that met filtering criteria a , ordered by gene damage index.The following genes previously reported to be associated with AP were evaluated: RFX6, FOXF1, PDX1, IHH, SHH.
T A B L E 3Note: a AF < 0.005, GDI < 75th percentile.bMaximumAF observed in public databases for any subpopulation.T A B L E 4Note: Variants in X-linked genes (recessive inheritance model) meeting filtering criteria.a T A B L E 5