Germline Mutations in MSR1, ASCC1, and CTHRC1 in Patients With Barrett Esophagus and Esophageal Adenocarcinoma FREE

The incidence of esophageal adenocarcinoma (EAC) in the United States and Europe has increased 350% since 1970, with uncertain etiology.¹ Although early-stage EAC is curable, most cases are detected at an advanced stage with poor survival. Esophageal adenocarcinoma is believed to be preceded by Barrett esophagus (BE), a premalignant metaplasia caused by chronic gastroesophageal reflux disease (GERD).^2- 6 GERD-related inflammation and the transforming growth factor β (TGFB) pathway have been implicated in sporadic BE and EAC, just as the role of inflammation has become prominent in a range of human cancers.⁷ Although acknowledged, the role of inflammation in BE and EAC has not been thoroughly studied.⁸ Barrett esophagus is common in the general population, estimated to occur in 1% to 10%⁶; it develops in 12% to 15% of patients with GERD.⁵ The risk of EAC in patients with BE is approximately 0.4% per year.⁹

Although most BE and EAC are believed to be sporadic, genetic (heritable) etiologies have been supported by observation of familial clustering of cases noted over several decades, although few large families co-segregating BE/EAC have been reported.^10- 11 In 1 referral series, clinical epidemiologic analyses suggest 7% of individuals with BE, EAC, or both have at least 1 affected blood relative.¹² Although shared environmental factors may contribute to such familial aggregation, an autosomal dominant mode of inheritance with incomplete penetrance is consistent with most published studies, and rare reported cases are consistent with autosomal recessive inheritance.¹⁰

The discovery of germline mutations in a gene or genes that predispose to BE/EAC may have ramifications regarding cancer risk assessment, genetic counseling, premorbid diagnosis, and targeted surveillance and management, and also add to the fundamental understanding of the pathophysiology of sporadic BE and EAC. We therefore sought to identify a gene or genes associated with BE/EAC predisposition.

Our study (2005-2010), approved by respective institutions' review board for research participants, involved prospective recruitment of all 298 consenting adults with histologically proven BE, EAC, or both, as well as families with 2 or more cases with BE, EAC, or both from 16 academic and community hospitals and clinics nationally (two-thirds originated from Cleveland Clinic, Cleveland, Ohio, and Johns Hopkins Medical Institutions, Baltimore, Maryland; <1% of research participants declined participation). All BE cases were long segment (eMethods). For discordant sibling pair studies, the nonaffected sibling had endoscopy documented unaffected status. Only white participants of northern or western European descent were selected and sex-matched in cases and controls.

Model-Free Linkage Analysis. Twenty-one concordant-affected sibling pairs (42 individuals with BE/EAC) and 11 discordant sibling pairs (11 with BE/EAC and 11 without BE/EAC) (2005-2006) were genotyped using Affymetrix GeneChip Human Mapping 100K SNP set (Affymetrix, Santa Clara, California) (Figure 1). Significant linkage to chromosomal regions found by 1 model-free linkage analysis method was self-replicated by a second model-free linkage analysis approach, which is used for small sample-sized data sets.^13- 15 Genomic regions were considered potentially interesting when −log₁₀P value(pP) ≥ 2.2 by SIBPAL analysis¹³ had logarithm of odds  3.2 by LODPAL analysis.¹³ These regions from this pilot linkage-association analysis were considered “potentially interesting” and served as regions to be validated (eMethods and Figure 1).

Independent Validation and Fine Mapping Significant Regions. It is standard in this field to single out significant genomic regions from pilot analysis to follow up with increased sample sizes from independent cases (validation), more genetic markers (fine mapping),^16- 19 or both in a second validation stage (Figure 1). We followed this strategy of independent validation and fine mapping of the “potentially interesting” regions identified by the pilot linkage-association analysis. We also paid particular attention to 2 additional regions (1q23 and 8p22), because these regions were found previously to be frequently somatically lost (by array comparative genomic hybridization) in EAC or gastroesophageal junction cancers.²⁰

Population substructure of cases and controls was determined by PLINK and EIGENSTRAT (eMethods).^21- 22 Analysis using EIGENSTRAT software^21- 22 and principal component analysis identified the top eigenvalues from the 376 available eigenvalues. Regression analyses were used to allow for potential population substructure by 2 separate analyses (PLINK-derived and EIGENSTRAT-derived analyses). After population substructure was assessed to be similar (>85%),^21- 22 single single-nucleotide polymorphism (SNP) association analyses of the above targeted genomic regions were performed with an independent validation series totaling 176 patients with BE/EAC and 200 ancestry-matched population controls (2007-2010) whose SNP data were derived from the denser Illumina Human610-Quad BeadChips (Illumina Inc, Hayward, California). Although we only were validating specific regions, we reasoned that it would be more cost-efficient to genotype all markers in a commercially available Chip instead of creating a new automation process for a reduced marker set. If the underlying genetic effect had been negligible, we would not have expected to see any savings on the average sample size, but fortunately the underlying genetic effect was large enough to warrant savings on sample size.

Statistical simulations have indicated that haplotype analysis with multiple SNPs may be more powerful than single SNP analysis,^23- 25 because multiple alleles at different loci on the same chromosome that are in linkage disequilibrium (LD) are likely to interact with each other to result in a phenotype. Thus, haplotype analysis was performed using PLINK²⁶ to predict the most likely haplotypes and those that were significantly associated with the BE/EAC phenotype. To account for type I error, an empirical P value corrected for testing multiple markers was obtained by permuting (10 000 permutations) the affectation status across the individual genotypes, as described in PLINK.²⁶

To narrow in on one or a subset of genes within and in proximity to the significant SNPs/haplotypes germane to BE/EAC (by tissue-specific expression in oncologic pathways and for functional-genomic validation), we integrated our significant regions with publicly available somatic gene expression data derived from 19 patients with BE/EAC (GDS3472 or GSE13083),²⁷ followed up by unsupervised hierarchical clustering²⁸ of genes within 250 kb flanking the significant SNPs and haplotypes across BE/EAC and unaffected individuals (eFigure 1, Table 1, and Table 2).

A final list of biologically plausible candidate genes (“priority” candidate genes) was then scanned for germline mutations in BE/EAC cases and compared with ancestry-matched population controls (Figure 1). Genes with mutations in cases but not in controls were screened in an independent validation series of 58 cases prospectively accrued from outpatient endoscopy units (2010) (Figure 1 and eTable 1).

Proteins were extracted from immortalized lymphoblastoid cells obtained from patients with BE/EAC and normal controls. After processing, protein lysates were loaded onto sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels. Antibodies specific to CCND1 (Cell Signaling Technology Inc, Danvers, Massachusetts), MSR1 (Abcam, Cambridge, Massachusetts), and α-tubulin (Sigma-Aldrich, St Louis, Missouri) were used for Western blotting.

Wild-type MSR1 or pCMV-FLAG empty vector were transiently transfected into MSR1-null HEK293 cells using Lipofectamine LTX and Plus Reagent (Invitrogen, Carlsbad, California). Cells were harvested after 24 hours and lysates (30 μg of protein) were analyzed by Western blotting using antibodies against FLAG (Sigma-Aldrich, 1:1000), CCND1 (Thermo-Fisher Scientific Inc, Waltham, Massachusetts, 1:200), and α-tubulin (Sigma-Aldrich, 1:5000).

CCND1 immunohistochemical analysis was performed using an avidin-biotin complex immunoperoxidase technique.

A pilot combined linkage-association analysis based on modification of established criteria²⁹ revealed 5 candidate regions (1q24.1-25.3, 1q41, 8q21.11-22, 10q21-22, and 11q21) (Table 1 and Figure 1).

Subsequently, we performed a validation study in an independent series of 176 cases and 200 controls (Figure 1), using a denser SNP-marker set but focusing only on the germline regions of interest and the 2 somatically lost hot spot regions (1q23 and 8p22). We were able to validate 4 (1q24.1-25.3, 1q41, 8q21.11-22, and 10q21-22) of these 5 pilot-derived germline candidate regions, while excluding one (11q21). Three additional loci at locations remote from the pilot linkage peaks (1q21.2, 8p22, and 11q25) were also found (Table 1). The most significant SNPs from this validative association analysis were located within or in the vicinity of the most promising pilot-derived linkage peaks (Table 1).

Haplotype and LD analysis conducted on the 176 cases and 200 controls confirmed our findings—any single SNP that was significant in the above single SNP analysis always revealed a haplotype block containing significant SNPs within the haplotypes (at least in LD) (P < .005) (Table 2). In the haplotype analysis, we considered regions of highest priority as those haplotypes exhibiting significance across multiple SNPs. The combined P values from the significant single SNPs and the significant haplotypes facilitated prioritization of regions of interest for further follow-up. There were 4 significant regions that overlapped from the linkage, single SNP association, and haplotype-LD analyses (1q24.1-25.3 [encompassing 1q24.2, 1q24.3, and 1q25.2-25.3 fine-mapped regions], 1q41, 8q21.11-22 [encompassing 8q21.11-22 and 8q22.1-24.22 fine-mapped regions], and 10q21-22). Additionally, there were 3 significant regions that overlapped in the single SNP association and haplotype analyses (1q21.2, 8p22, and 10q22.1). Thus, we selected these regions, shown in Table 2, as “regions of interest” (Figure 1, eTable 2, and eTable 3). Each of these regions contained SNPs that were statistically significant at P < .005 and also had multiple haplotype windows showing significance (P < .01).

Integration of our significant SNP and haplotypes with publicly available so matic BE/EAC transcriptome data (Figure 1) yielded 38 genes located within 250 kb flanking each significant SNP, within significant haplotypes, or both that accurately clusteredBE/EAC cases from controls (eFigure 2). An additional filtering step based on known organ-specific functions resulted in a final short list of 12 priority candidate genes (LHX4, DIRC3, MARK1, KIF26B, MSR1, TMEM67, WDSOF1, CTHRC1, KCNQ3, PRKG1, ASCC1, and OPCML), which were also functionally plausible, within our regions of interest (Table 1 and Table 2).

Mutational analyses of these 12 priority candidate genes in BE/EAC cases and controls revealed germline mutations in 3 genes (MSR1 [macrophage scavenger receptor 1] [MIM153622], ASCC1 [activating signal co-integrator 1 complex subunit 1] [NC_000010.10], or CTHRC1 [collagen triple-helix repeat-containing 1] [MIM610635]) in 13 of 116 patients (11.2%) with BE/EAC (Table 3 and eTable 1). No sequence variants were found in the remaining 9 genes that were not also present to the same degree in controls.

Among the 116 patients with BE/EAC, 8 patients (proportion, 0.069; 95% confidence interval [CI], 0.030-0.130; P < .001) had a germline truncating mutation in MSR1 c.877C>T, resulting in p.R293X (Figure 2 and Table 3), and 2 additional patients (proportion, 0.017; 95% CI, 0.021-0.061; P = .19) with BE/EAC carried germline MSR1 p.L254V (c.760C>G) in exon 5 (Table 3 and eFigure 3A). These mutations were not found in 139 ancestry-matched population controls. Additionally, we identified 2 germline missense mutations—c.869A>G in exon 8 of ASCC1, resulting in p.N290S in 2 patients (proportion, 0.021; 95% CI, 0.003-0.074; P = .18); and c.131A>C in exon 1 of CTHRC1, resulting in p.Q44P in 1 patient (proportion, 0.011; 95% CI, 0.0003-0.061; P = .42), neither of which were found among 125 controls (Table 3, eFigure 3B, and eFigure 3C).

To confirm the mutations found in the 3 candidate genes (Table 3), mutational analyses were then performed in an independent series of 58 cases obtained from outpatient endoscopy units. These samples confirmed the presence of germline MSR1 c.877C>T, p.R293X mutation in 2 of 58 cases (proportion, 0.034; 95% CI, 0.004-0.120; P = .09) and CTHRC1 c.131A>C, p.Q44P mutation in 1 of 58 cases (1.7%). After pooling the original 116 cases with the validation series of 58, a total of 10 cases with BE/EAC carried p.R293X (proportion, 0.054; 95% CI, 0.026-0.098; P = .006) (Table 3).

Western blotting of germline protein lysates from 5 MSR1 mutation-positive patients with BE/EAC and 7 controls revealed variable decreases in MSR1 protein levels in 3 cases (Figure 3). All 5 MSR1-mutation positive patients had increased CCND1 levels compared with controls (Figure 3). Barrett esophagus tissues from patients who were mutation-positive showed increased nuclear expression of CCND1 by immunohistochemistry compared with control esophageal specimens (Figure 4). We then proceeded with the converse experiment by overexpressing wild-type MSR1 in HEK293 cells, resulting in decreased CCND1 protein (Figure 3).

Barrett esophagus is prevalent in the general population and has the potential to progress to EAC. Because late-stage EAC carries a poor outcome, it is desirable to identify predisposition or risk alleles that will eventually allow premorbid risk assessment and affect subsequent management. Herein, we have identified germline mutations in 3 candidate genes in approximately 11% of our series of patients with BE/EAC, with the most commonly affected being MSR1 (approximately 7%), followed by ASCC1 and CTHRC1. Findings of germline MSR1 and CTHRC1 mutations were replicated in an independent validation series.

MSR1 on 8p22 encodes the class A macrophage scavenger receptor, which are macrophage-specific trimeric integral membrane glycoproteins implicated in many macrophage-associated, hormonal, and pathological processes, including inflammation, innate and adaptive immunity, oxidative stress, and apoptosis.^30- 32 The MSR1 c.877C>T sequence variant, resulting in p.R293X, located within a highly conserved collagen-like domain of the MSR1 protein,³³ would be expected to disrupt function. The MSR1 p.R293X was previously shown to associate with prostate cancer in specific ancestries, although this association is controversial.^30,33- 36 Given our observations and taking the p.R293X-prostate cancer association at face value, one possible explanation is that this particular mutation is associated with both BE/EAC and prostate cancer predisposition, with the latter at lower penetrance. In many inherited neoplasia syndromes, single gene mutations predispose to cancers in more than 1 organ.³⁷ We found both p.R293X and p.L254V (also within the conserved coiled-coil domain and near the glycosylation site at the 249th amino acid) in the germline of our BE/EAC cases, but not in the ancestry-matched population controls, which strongly suggest that these mutations contribute to BE/EAC risk, or at least are necessary for BE/EAC predisposition. Whether they are also sufficient is currently unknown.

Accumulating evidence links MSR1 to inflammatory events.³⁸ Barrett esophagus/EAC may also be associated with inflammatory events,⁷ thus supporting our observations that MSR1 is a plausible candidate susceptibility gene for BE/EAC. The molecular mechanisms underlying the pathogenesis of inflammation-associated cancer are complex and involve both the innate and adaptive immune systems.^39- 42

More and more examples linking inflammatory and carcinogenic pathways, such as the cell cycle, are surfacing. For example, signal transducer and activator of transcription 3 and nuclear factor kappa-B link phosphatase and tensin homologue, deleted on chromosome 10 to inflammatory pathways.⁴³ Beyond genetic evidence, we have additionally shown up-regulation of key cell cycle molecule CCND1 by both Western blotting of germline proteins and immunohistochemistry of MSR1 mutation-related BE tissue, in which CCND1 is overexpressed in the nucleus. MSR1 p.R293X results in a truncated protein (affecting cytoplasmic topology, the transmembrane and parts of the collagen-like motifs), which still expresses, but variably. We observed germline MSR1 mutation, with variably decreased MSR1 protein levels, was associated with overexpression of nuclear CCND1 in BE tissues in MSR1 -mutation carriers (but not in control normal epithelium). This observation suggests a linkage of inflammation to the cell cycle and a potential etiology for BE, via loss of control of the G1-S transition consistent with checkpoint-mediated cell cycle delays.^44- 45 CCND1 elevation in both sporadic and heritable BE/EAC as an important final common pathway has precedent,^46- 49 linking the WNT and adenomatous polyposis coli protein (APC) cascades, and lending credence to our observations. It remains to be determined whether increased expression of CCND1 in the setting of germline MSR1 mutation can by itself, or in combination with other oncogenic events, lead to neoplastic transformation in BE.

We also found that ASCC1 germline p.N290S, affecting a region conserved across species, occurred in 2.1% of BE/EAC cases, but not in ancestral-matched population controls. The candidacy of ASCC1 as a risk allele for BE/EAC is supported by somatic expression array data comparing normal esophageal epithelium, premalignant BE, and EAC samples.⁵⁰ ASCC1 enhances nuclear factor kappa-B and activator protein 1 transactivation by directly binding to JUN kinase.⁵¹ Although little else is known about ASCC1/TRIP4 function, its putative cross-talk with JUN and nuclear factor kappa-B again links inflammatory to tumor suppressive/oncogenic pathways. Its role in potentially co-regulating the androgen receptor⁵² may also begin to explain the known epidemiologic increased BE risk in men. In addition, although we found only 1 germline missense mutation in CTHRC1, this gene is intriguing in its cross-talk with 2 established pathways in sporadic BE/EAC pathogenesis, TGFB, and WNT⁵³ via APC.

Although this sequence alteration is a rare variant, and this study may not be powered to differentiate between 1 of 89 and 0 of 125, a change from glutamine to proline results in an alteration from a single-branch polar (negative) amino acid to a small hydrophobic cyclic amino acid. This change is predicted to disrupt the protein's ability to form secondary structure and the helix-loop-helix structure necessary for collagen deposition, fibrosis, and involvement in the WNT and APC pathways. CTHRC1 is expressed in tissue repair processes and may be important in the host's response to GERD. Given its role in collagen matrix deposition and its expression in myofibroblasts,⁵³ an alteration of CTHRC1 might predispose to decreased lower esophageal sphincter tone and consequent tendency toward GERD and BE.

Because we sought to identify risk alleles that contribute to a reasonable subset of heritable BE/EAC that also yield clinically useful attributable risks (or protection), we consciously used a relatively small series. We did this because the larger the series, the more likely the genetic effect sizes shrink, such that case-control series in the thousands result in odds ratios in the 1.1 to 1.3 range. We added rigor by using a second-stage independent validation series of cases and controls that were sufficiently powered based on our pilot data. In addition, rather than performing brute-force sequencing of at least 38 genes in just our 9 top priority regions (which would have increased the likelihood of finding “noise”), we used a systems-biology approach of statistically prioritizing regions of interest by integrating multiple platforms to finally come up with 12 top priority genes, which eventually yielded germline mutations in MSR1, ASCC1, and CTHRC1 in patients with BE/EAC but not in population controls. These 3 genes together accounted for 11% of our cases, reflecting what is normally considered a moderate- to high-penetrance genetic load for a disease. The functional interrogation and immunohistochemistry results support the pathogenicity of these germline MSR1 mutations. Nonetheless, future independent studies are needed to replicate our data in other patient populations to confirm the conclusions.

In summary, germline mutations in MSR1, ASCC1, and CTHRC1 in patients with BE/EAC appear physiologically relevant to BE, encoding proteins involved in apoptosis, innate immunity, polarity, and mobility that affect inflammatory and TGFB/WNT signaling pathways. Larger cohort studies may be necessary to determine the usefulness of these genes and their variants in risk assessment and premorbid diagnosis.