Infantile Hypertrophic Pyloric Stenosis (IHPS) is the leading cause of gastrointestinal obstruction in the first months of life, with an incidence of 1–3 per 1000 live births in Western countries.^1;2 It affects four to five times as many boys as girls³ and typically presents 2 – 8 weeks after birth⁴ with projectile vomiting, weight loss, and dehydration.

Although IHPS is a clinically well-defined entity, the etiology of the condition is complex and remains unclear. A genetic predisposition is well established; IHPS aggregates strongly in families and has an estimated heritability of more than 80%.² However, environmental factors, such as erythromycin exposure⁵ and feeding practice^6;7, have also been implicated in IHPS etiology. Moreover, sharp changes in incidence seen in several countries over the last decades⁸ underline the importance of modifiable environmental exposures.

Cases of IHPS sometimes occur as part of a syndrome of known genetic etiology⁹. For example, Smith-Lemli-Opitz syndrome is an autosomal recessive congenital disorder caused by mutations in the 7-dehydrocholesterol reductase (DHCR7; NCBI Entrez Gene 1717) gene.¹⁰ Affected individuals are unable to complete the final step in cholesterol biosynthesis causing a wide range of metabolic and developmental abnormalities, including IHPS in 10–15% of cases.¹⁰

Less is known about the genetic background of isolated IHPS. Several association studies have focused on genes involved in gastric contractility.^11–13 However, these studies were relatively small and produced conflicting results. In contrast, we recently conducted the first genomewide association study of IHPS and identified three susceptibility loci near the muscleblind-like splicing regulator 1 (MBNL1; NCBI Entrez Gene 4154) gene and the NK2 homeobox 5 (NKX2-5; NCBI Entrez Gene 1482) gene.¹⁴ Still these variants only explain a small fraction of the variance in disease liability.

The present study followed a 2-stage approach with the aim to identify novel genetic variants for IHPS. In the first (discovery) stage, we used a hypothesis-free approach to identify variants associated with IHPS. In the second stage, we carried forward the most promising variants for validation in three independent case-control sample sets. Finally, we also aimed to investigate the biological relevance of novel genetic loci through follow-up experiments based on prospectively collected plasma samples from cases and controls.

Table 1 shows sample characteristics of the participants contributing to the two stages of the genetic study. In the discovery stage, we analyzed the association between the disease and 9737928 imputed genetic variants in 1001 cases and 2371 controls. Genomic inflation factors were 1.05 in the complete discovery data, and 1.02 and 1.03 in the data restricted to boys and girls, respectively. Imputed SNPs at four loci showed P values <1×10⁻⁷ and were selected for further study. These included two novel loci on chromosomes 11q23.3 and 19p13.2, as well as two already confirmed loci on chromosomes 3q25.1 and 5q35.2. The third known locus on chromosome 3q25.2 was also selected for completeness. The chromosomal regions harboring the five selected loci were re-imputed with the original IMPUTE2 algorithm (i.e. without pre-phasing) for increased accuracy, and association tests were repeated for these regions. To confirm the associations at these loci, we genotyped a total of seven SNPs in validation samples from Denmark, Sweden, and the US with a total of 1663 cases and 2315 controls. One novel locus (11q23.3) was validated with genomewide significance (Table 2), the three known loci (3q25.1, 3q25.2, and 5q35.2) were confirmed, and one locus (19p13.2) could not be validated (eTable 1). Results based on hierarchical logistic regression were very similar (eTable 2). Figure 1 displays forest plots for the four genomewide significant loci.

The variant rs12721025 yielded the lowest P value (OR, 1.59; 95% CI, 1.38–1.83; P=1.9×10⁻¹⁰) at the 11q23.3 locus. This SNP is located 301 bases downstream of the Apolipoprotein A-I (APOA1; NCBI Entrez Gene 335) gene with additional apolipoprotein genes APOC3 (NCBI Entrez Gene 345), APOA4 (NCBI Entrez Gene 337), and APOA5 (NCBI Entrez Gene 116519) within 50 kb centromeric (Figure 2). A region of strong linkage disequilibrium extends several hundred kb to the telomeric side; here the variant rs77349713, which is intronic in the salt-inducible kinase 3 (SIK3; NCBI Entrez Gene 23387) gene, was also associated with genomewide significance (OR, 1.53; 95% CI, 1.33–1.75; P=1.2×10⁻⁹).

We explored the possible functional impact of the 11q23.3 associations by considering all 222 genotyped or imputed variants (SNPs and indels) at the locus with P<1×10⁻⁶. These were all correlated with rs12721025 (r² between 0.46 and 0.91) and we found no evidence for allelic heterogeneity (all of these SNPs had P>0.01 when conditioning on rs12721025). Most of the SNPs were intronic in the genes APOA1, SIK3, PAFAH1B2 (NCBI Entrez Gene 5049), SIDT2 (NCBI Entrez Gene 51092), TAGLN (NCBI Entrez Gene 6876), or PCSK7 (NCBI Entrez Gene 9159); see eTable 3. Two SNPs were in exons, but both were synonymous. Fifty out of the 222 SNPs were previously found to be associated with levels of total and/or HDL cholesterol²⁵ with P values down to 5.7×10⁻⁷ (results retrieved from http://www.wikigwa.org). For all of these SNPs, the cholesterol lowering allele consistently conferred increased risk of IHPS (eTable 4). A search of the GWAS catalog (http://www.genome.gov/gwastudies) did not reveal any associations to other phenotypes and no associations to gene expression were found in a search of the expression quantitative trait loci (eQTL) browser (http://eqtl.uchicago.edu). However, chromatin immunoprecipitation sequencing (ChIP-seq) data from the ENCODE Consortium (explored using the UCSC genome browser, NCBI build 37, http://www.genome.ucsc.edu/cgi-bin/hgGateway) showed that small islands of histone modification involving mono- and trimethylation of histone H3 on lysine 4 (H3K4me1 and H3K4me3) directly cover rs12721025.

The functional characteristics of the 11q23.3 locus suggest the hypothesis that low levels of circulating lipids in newborns are associated with increased risk of IHPS. We addressed this hypothesis by measuring plasma levels of total, LDL, and HDL cholesterol as well as triglycerides in prospectively collected umbilical cord blood from a set of 46 IHPS cases and 189 controls of Danish ancestry, most of which were also in the discovery sample (see eTable 5 for sample characteristics). eFigure 1 summarizes the distribution of the four biomarkers in cases and controls. Table 3 shows levels of total cholesterol levels in umbilical cord blood plasma for cases and controls overall, and divided into quartiles. The mean total cholesterol levels for 46 cases and 189 matching controls were 65.22 mg/dL and 75.25 mg/dL, respectively. The risk of IHPS was inversely and significantly associated with total cholesterol level with an OR of 0.77 (95% CI, 0.64–0.92; P=0.005) per 10 mg/dL. An omnibus test of differences between quartiles was also significant (P=0.02). Results for LDL and HDL cholesterol and triglycerides are shown in eTable 6. For HDL cholesterol and triglycerides, adding a quadratic term to the analyses indicated non-linear effects (P=0.05 and P=0.004, respectively). For both of these biomarkers there were significant differences between quartiles (P=0.04 and P=0.01, respectively in the omnibus test), and for both quartile 3 had the lowest ORs.

To explore the role of other lipid related variants on IHPS risk, we identified 247 SNPs representing 134 regions reported in the GWAS catalogue to be associated with lipid levels. Only one region on chromosome 19p13.2 was associated with IHPS below a Bonferroni adjusted threshold of P<3.8×10⁻⁴. For all SNPs in this region, the cholesterol lowering allele conferred increased risk of IHPS. The region was represented by rs2228671, a synonymous SNP in the LDL receptor gene (LDLR; NCBI Entrez Gene 3949), in our study, but the association was not seen in the validation cohorts (eTable 1).

A subset of 94 familial cases and 93 unaffected relatives in 37 pedigrees was excluded from the Swedish case-control analyses. These data were instead analyzed using family based association testing. This gave very limited statistical power, particularly for the rarer SNPs, and results did not reach statistical significance (eTable 7). All genomewide significant SNPs showed the same direction of effects in the two sexes and there was no interaction between sex and genotype (eTables 8 and 9).

We identified a novel genomewide significant locus for IHPS on chromosome 11q23.3 in a region harboring the apolipoprotein (APOA1/C3/A4/A5) gene cluster and we also confirmed three previously reported loci. The most significant SNP at the new locus, rs12721025, is located immediately downstream of APOA1 and is covered by several different histone modification regions.

APOA1 encodes apolipoprotein A-I which is the major protein component of high density lipoprotein (HDL) in plasma. Furthermore, rs12721025 is correlated with SNPs previously found to be associated with levels of circulating cholesterol. For these SNPs, the cholesterol lowering allele consistently conferred increased risk of IHPS. These findings suggest the hypothesis that low levels of plasma cholesterol in newborns are associated with increased risk of IHPS. We addressed this hypothesis experimentally using prospectively collected umbilical cord blood samples and found lower cholesterol levels at birth in infants who went on to develop IHPS compared with matched controls who did not develop the disease.

IHPS is a prominent clinical feature in many reports of the Smith-Lemli-Opitz syndrome (SLOS), an inborn defect of cholesterol biosynthesis in the gene DHCR7 associated with low cholesterol levels in infants at birth. In one large case series, six of 49 cases with proven DHCR7 mutations had IHPS.²⁶ Also, a large epidemiological study found that SLOS is 150 times more prevalent in IHPS cases than in the general population.³ A study of 10 patients with isolated IHPS and 8 controls found no cholesterol metabolism anomalies, but did find that plasma cholesterol levels were lower in cases compared to controls.²⁷ The study was however too small for firm statistical conclusions and appears not to have been followed up.

A number of previous findings would also be consistent with low lipid levels representing an important risk factor for IHPS. First, the protective effect of female sex could at least partly be due to higher cholesterol levels, since it is well-known that levels of LDL cholesterol and HDL cholesterol are on average higher in newborn girls compared to boys.²⁸ Second, the IHPS risk associated with bottle-feeding^6;7 could in part be caused by insufficient lipid levels since bottle-feeding is known to be associated with lower total and LDL cholesterol levels in infancy.²⁹ Finally, the drop in IHPS incidence observed in the 1990s in several countries^8;30–33 coincided temporally with increasing percentages of mothers breastfeeding their infants³⁴, suggesting that better nutritional status in infants may have prevented IHPS from developing in a fraction of potential cases.

Other previously reported lipid related variants were not significantly associated with IHPS. However, different loci regulate different aspects of lipid metabolism at different stages over a lifetime, and most lipid genetics studies have used adult participants. Further functional study is clearly needed to illuminate the biological mechanisms underlying our findings. One approach that might be revealing would focus on the essential role of cholesterol in nervous system development^35–37 given the deficiencies in enteric innervation seen in pyloric sphincter muscle tissue from IHPS patients^38–40.

The previously identified IHPS loci point to candidate genes involved in alternative splicing, cardiac muscle development, and embryonic gut development¹⁴, and further investigation is needed of the interplay between these loci and the genetic and biochemical findings reported here.

Strengths of our study include the use of samples from three different populations with a total of more than 2600 cases and 4600 controls. Furthermore, by leveraging 1000 Genomes Project data, we were able to impute and analyze almost 10 million genetic variants in the discovery stage. Also, we were able to perform a follow-up study based on prospectively collected samples that investigated the potential biological relevance of the new IHPS locus.

The study also has limitations. First, our data permit no assertions about putative causal SNPs at the associated loci, something that would require further fine-mapping studies, e.g., through targeted sequencing of affected individuals in families⁴¹. Furthermore, it is important to emphasize that our results do not establish a causal link between cholesterol levels and IHPS. If some of the risk is mediated through cholesterol further study is required to assess the relative importance of other genetic and environmental risk factors. Finally, the lipid measurement study was limited by small numbers of cases and also only covered the time right at birth, i.e. several weeks before the condition typically developed in the cases.

In conclusion, we identified a novel genetic locus that associates with IHPS at genomewide significance. Characteristics of this locus suggest the possibility of an inverse relationship between levels of circulating cholesterol in neonates and IHPS risk. Further investigation is required to illuminate the functional significance of the association identified here.