Shifts in Serum Bile Acid Profiles Associated With Barrett's Esophagus and Stages of Progression to Esophageal Adenocarcinoma

Abstract

INTRODUCTION:

Reflux bile acids are believed to promote esophageal adenocarcinoma (EAC), but the role of systemic bile acids is unknown. This study aimed to assess associations between systemic bile acids and stages of Barrett's esophagus (BE) progression.

METHODS:

Subjects with and without BE were enrolled in this multicenter cross-sectional study. Targeted serum bile acid profiling was performed, and a subset of subjects completed a validated food frequency questionnaire. RNA sequencing was performed on BE or gastric cardia tissue to assess bile acid associations with gene expression.

RESULTS:

A total of 141 subjects were enrolled with serum bile acids profiled (49 non-BE; 92 BE: 44 no dysplasia, 25 indefinite/low grade dysplasia, 23 high-grade dysplasia/EAC). Lower Healthy Eating Index score, older age, higher body mass index, and no proton pump inhibitor use were associated with increased levels of multiple bile acids. Global bile acid pools were distinct between non-BE and stages of BE neoplasia (P = 0.004). Increasing cholic acid was associated with high-grade dysplasia/EAC compared with non-BE, even after adjusting for EAC risk factors (adjusted odds ratio 2.03, 95% confidence interval 1.11–3.71) as was the combination of unconjugated primary bile acids (adjusted odds ratio 1.81, 95% confidence interval 1.04–3.13). High cholic acid levels were associated with tissue gene expression changes including increased DNA replication and reduced lymphocyte differentiation genes.

DISCUSSION:

Alterations in serum bile acids are independently associated with advanced neoplasia in BE and may contribute to neoplastic progression. Future studies should explore associated gut microbiome changes, proneoplastic effects of bile acids, and whether these bile acids, particularly cholic acid, represent potential biomarkers or viable therapeutic targets for advanced neoplasia in BE.

INTRODUCTION

The incidence of esophageal adenocarcinoma (EAC) has been increasing since the late 1960s in Western countries, contributing to significant morbidity and mortality (1). Gastroesophageal reflux disease (GERD) and obesity are 2 of the most well-studied modifiable risk factors for EAC. However, the incidence of GERD began to rise in the 1970s, and the obesity pandemic only began in the 1980s, suggesting that other risk factors are at play (2,3).

Notably, increased dietary fat and declining fiber consumption in the United States have coincided with the rise in EAC incidence (4), and high dietary fat and fiber intake are associated with increased and decreased risks of EAC, respectively (5,6). In experiments with the L2-IL1B mouse model of Barrett's esophagus (BE), mice fed a high-fat diet developed esophageal neoplasia and tumors more rapidly compared with mice fed a control diet, with these effects mediated in part by the gut microbiome (7). Bile acid composition is regulated by gut microbiome composition, and bile acids play a key role in lipid digestion. Primary bile acids are made by hepatocytes, conjugated with the amino acids taurine or glycine (which makes them water-soluble and allows them to emulsify fats), and secreted into the intestine. These bile acids are then deconjugated by bacteria through the enzyme bile salt hydrolase (BSH). In the distal small intestine and colon, bacteria then further convert primary to secondary bile acids. Most of these bile acids are ultimately reabsorbed into the enterohepatic circulation. In humans, consumption of a high-fat diet results in increased relative abundance in gut bacteria—namely Bacteroides, Clostridium, Bifidobacterium, and Lactobacillus—that express BSH, the enzyme essential for bile acid deconjugation, resulting in increased levels of unconjugated and secondary bile acids in stool (8).

Bile acids have increasingly been recognized as endogenous etiologic agents in gastrointestinal (GI) cancer, especially well-studied in colorectal cancer (9), possibly through interactions with bile receptors such as G protein-coupled bile acid receptor 1 (GPBAR1, also known as TGR5), vitamin D receptor, and farnesoid X receptor (10,11). Bile acids also have regulatory effects on T cell and macrophage differentiation in GI tissue (9). Studies in BE and EAC have focused on the potential role of bile acids in gastroesophageal reflux; in vitro studies have shown that bile acids increase production of reactive oxygen species and DNA damage in Barrett's and squamous esophageal epithelial cells (12). Furthermore, in the aforementioned transgenic L2-IL1B mouse model, administration of deoxycholic acid accelerated the development of Barrett's-like metaplasia and dysplasia (13). However, in a clinical trial in humans, ursodeoxycholic acid markedly impacted gastroesophageal refluxate bile acid composition but had no effect on tissue markers of oxidative DNA damage, cell proliferation, or apoptosis (14), suggesting that esophageal exposure to bile acids in refluxate may not be a major promoter of esophageal neoplasia.

Circulating bile acids, derived from the gut and influenced by dietary intake, may exert biological effects, and BE tissue is likely exposed to systemic and refluxate bile acids. Circulating bile acid composition was shown to predict future risk of colon cancer in a nested case-control study (15). However, little is known regarding the potential contribution of circulating bile acids to BE progression. Thus, in this study, we sought to evaluate the association between circulating bile acids in the progression of BE and explore the role of diet as a potential mediator.

METHODS

Study design

Patients with and without BE who underwent upper endoscopy for clinical indications at Columbia University Irving Medical Center, Mayo Clinic—Rochester, and University of Pennsylvania were prospectively enrolled from February 2018 to September 2021 in this multicenter study. Patients were 18 years or older, and for patients with BE, had histologically confirmed BE and took proton pump inhibitors (PPIs) at least daily for 3 months before enrollment. Patients with a history of gastric or esophageal squamous cell cancer or gastric or esophageal surgery, who used antibiotics or systemic immunosuppressants within 3 months before the date of endoscopy, or had known untreated esophageal stricture or uninvestigated dysphagia were excluded. Patients with BE length <2 cm were also excluded for study purposes to ensure that enough Barrett's mucosa would be present to allow for necessary additional biopsies for tissue analyses.

Demographics and clinical data

Data were collected on demographic and clinical information including medical history, history of endoscopies and associated histology, family history of BE or esophageal cancer, medication use, and smoking history. In addition, anthropometric measures were recorded to calculate body mass index (BMI) and waist-to-hip ratio.

Dietary questionnaire

Dietary information was collected using the Diet History Questionnaire II (DHQ II) to assess for dietary intake over the preceding 12 months (16). Diet*Calc (program version 1.5.0, 2012), a software provided by the National Cancer Institute to analyze standard versions of the DHQ II, was used to generate nutrient and food group intake estimates. Outliers were excluded if the reported total energy intake (kcal) was less than or greater than 3 SDs from the mean. Percent energy from fat was obtained directly from the Diet*Calc output, whereas energy adjusted fiber intake was calculated as total fiber intake (grams) per 1,000 kcal total energy intake. The Healthy Eating Index (HEI), a validated measure of diet quality (17,18) was calculated using a version that aligns with the 2015 Dietary Guidelines for Americans (19). The HEI-2015 scores 13 components for a total of 100 points, including 9 adequacy categories: total fruits, whole fruits, total vegetables, greens and beans, whole grains, total protein foods, seafood and plant proteins, and fatty acids and 4 moderation categories: refined grains, sodium, added sugars, and saturated fats. For each of these measures (i.e., percent energy from fat, energy adjusted fiber intake, and HEI score), outliers that were less than or greater than 3 SD from the sample mean were excluded.

Upper endoscopy

Various biospecimens were collected, while subjects were in the fasting state, before and during endoscopy, including saliva and oral rinse as well as blood. During the endoscopy, gastric aspirates as well as additional brushings and biopsies were collected for study purposes. BE length, hiatal hernia size, and presence of focal lesions within the BE segment were noted. Biopsies were taken per standard of care. Subjects were classified by the highest degree of neoplasia and categorized as one of the following: no BE, nondysplastic BE (ND), indefinite for dysplasia (IND), low-grade dysplasia (LGD), high-grade dysplasia (HGD), or adenocarcinoma. All subjects with EAC had intramucosal (T1a) adenocarcinoma. For the study endoscopy, biopsies were interpreted by pathologists at the corresponding institution where the endoscopy was performed. Slides were also reviewed by a central pathologist (A.D.P.). If there was disagreement between the local and central pathologists' readings, the final diagnosis was determined by consensus review with a second central pathologist (S.M.L.). If the diagnosis was indefinite for dysplasia, slides were re-reviewed to determine a consensus diagnosis. If a subject had a history of a higher degree of neoplasia detected on a prior endoscopy (compared with the study endoscopy), the subject was categorized by the historical findings if these biopsies had been reviewed by an expert GI pathologist at one of the study sites. Additional biopsies were taken for gene expression analyses from the mid-BE segment (from subjects with BE) or the gastric cardia within 1 cm of the squamocolumnar junction (from non-BE subjects), avoiding any focal lesions. The biopsies were placed in Qiagen Allprotect and stored at −80° C.

Serum bile acid levels

Bile acids were measured from human serum samples using ultra performance liquid chromatography-tandem mass spectrometry (LC-MS) as reported previously (20). The following bile acids were measured: cholic acid (CA), taurocholic acid (TCA), glycocholic acid (GCA), chenodeoxycholic acid (CDCA), taurochenodeoxycholic acid (TCDCA), glycochenodeoxycholic acid (GCDCA), deoxycholic acid (DCA), taurodeoxycholic acid (TDCA), glycodeoxycholic acid (GDCA), hyodeoxycholic acid (HDCA), lithocholic acid (LCA), taurolithocholic acid (TLCA), ursodeoxycholic acid (UDCA), tauroursodeoxycholic acid (TUDCA), and glycoursodeoxycholic acid (GUDCA).

Bile acids were extracted by spiking human serum samples with deuterated internal standards and mixing with chilled acetonitrile for protein precipitation. The extracted bile acids were resuspended in methanol for LC-MS analysis. LC-MS analysis was performed using Waters Xevo TQS mass spectrometer and integrated with a Waters Acquity UPLC system (Milford, MA). Ten microliters of the sample were injected onto a Phenomenex Kinetex C18 column (50 × 2.1 mm, 1.7 μm, 100 Å) maintained at 40° C and at a flow rate of 0.250 mL/min. The initial flow conditions were 40% solvent A (water containing 5 mM ammonium formate) and 60% solvent B (methanol containing 5 mM ammonium formate). Solvent B was raised to 80% linearly over 8 minutes, increased to 97% in 2 minutes and returned to initial flow conditions by 11.30 minutes with a total run time of 14 minutes. Quantitative measurements were performed in selective ion monitoring mode and negative electrospray ionization. The lower limit of quantitation for the bile acids was 1 nM. Intra-assay precision for the measured bile acids ranged from 2.9% to 5.8% with an intra-assay accuracy from 98.6% to 105.7%. The assay showed an interassay precision for all bile acids ranging from 1.49% to 5.07%. Absolute bile acid levels were recorded.

Tissue gene expression

Esophageal tissue gene expression analyses were performed by bulk RNA sequencing. Esophageal biopsies were homogenized in Qiazol on Tissuelyser II (Qiagen, Hilden, Germany) at frequency 25/sec for 2 minutes, twice. Total RNA was then purified using the miRNeasy micro kit (Qiagen) following the kit protocol. RNA was eluted with 30 μL RNase-free water. Quantitation was performed using Nanodrop and Bioanalyzer. Poly-A pull-down was performed to enrich mRNAs from total RNA samples, then library construction was performed using Illumina TruSeq chemistry. Libraries were then sequenced using Illumina NovaSeq 6000. Samples were multiplexed in each lane, yielding the targeted number of paired-end 100 bp reads for each sample. RTA (Illumina, San Diego, CA) was used for base calling and bcl2fastq2 (version 2.19) for converting BCL to fastq format, coupled with adaptor trimming. A pseudoalignment was performed to a kallisto index created from transcriptomes (human: GRCh38; mouse: GRCm38) using kallisto (0.44.0).

Given our findings that CA was independently associated with HGD/EAC, differential gene differential expression analyses were performed to identify genes associated with high CA by comparing samples from subjects in the fourth vs first quartiles. Analyses were restricted to those with RNA Integrity Number values >5. The statistical algorithm based on the negative binomial generalized linear model, DESeq2 (implemented in R programming language) (21), was used for the 2 groups comparison. The model was adjusted for underlying histology as well as risk factors for EAC, including age, BMI, sex, history of GERD, smoking history (ever/never), and family history of BE/EAC. Genes were ranked based on the Wald statistic obtained from the DESeq2 analysis, and gene set enrichment analysis (22) was performed using the R package “fgsea” (23) in Bioconductor. The reference database including 186 KEGG canonical pathways was used to identify significantly altered pathways (adjusted P value ≤0.01).

Statistical analysis

Fisher exact tests were used to analyze categorical variables and rank-sum tests were used to analyze continuous variables as appropriate. Spearman correlation coefficients were calculated to assess within-individual associations between bile acids. The distribution of bile acids was skewed, so a logarithmic transformation was applied to normalize the distribution and used for all regression analyses. Multivariable linear regression analyses were performed to assess for factors associated with both dietary intake and bile acid levels. MANOVA was performed to assess global bile acid differences across groups, simultaneously analyzing levels of all 15 bile acids. Principal components analyses were performed to identify drivers of global bile acid composition. Kruskal-Wallis was used to assess for differences in individual bile acid levels across all groups; for bile acids with P < 0.05, pairwise comparisons and logistic regression analyses were performed to compare subjects with and without BE as well as subjects with advanced neoplasia (HGD/EAC) with subjects without BE or with BE without dysplasia. Multivariable logistic regression analyses were then performed for these individual bile acids, adjusting for EAC risk factors: age, sex, BMI, a modified metabolic syndrome proxy (score ranging from 0 to 3, with 1 point each for hypertension, hyperlipidemia, and diabetes; obesity was not included because BMI was adjusted for separately), smoking history (ever vs never), and family history of BE or esophageal cancer. Additional models also adjusted for HEI score, fat, and fiber intake. BE length was also included in models restricted to subjects with BE. Race and ethnicity were excluded from these models because the overall study population was almost exclusively composed of non-Hispanic White people. PPI use was not included as a covariate due to collinearity because all patients with BE were on PPIs. Statistical significance was defined as P < 0.05. All analyses were performed in Stata version 17.0. The data generated in this study are available on request from the corresponding author.

Ethics statement

This study was approved by the respective Institutional Review Boards of each participating institution. All patients provided written informed consent.

RESULTS

A total of 166 subjects were enrolled in the study, 157 were eligible for inclusion in the study analyses (9 were found to have BE length <2 cm), and 141 had blood collected and analyzed for serum bile acids (see Supplementary Figure 1, Supplementary Digital Content 1, http://links.lww.com/CTG/B185). Of these, 49 were non-BE controls and 92 had BE (44 ND, 9 IND, 17 LGD, 16 HGD, 6 EAC). The mean age was 59.9 (SD 14.6), 63.1% were male, 96.2% were White, and 98.1% were non-Hispanic. Compared with controls, those with BE were older, and a higher proportion were male, on statins, ever-smokers, and had a family history of BE or EAC (Table 1).

Table 1.

Characteristics of study participants

	Non-BE (n = 57)	BE (n = 100)	P valuea
Demographics
Age, yr, mean ± SD	49.7 ± 15.6	65.8 ± 10.1	<0.001
BMI, mean ± SD	29.7 ± 6.9	30.2 ± 5.4	0.29
Male sex, n (%)	22 (38.6)	77 (77.0)	<0.001
Non-Hispanic ethnicity, n (%)	57 (100)	97 (97.0)	0.70
White race, n (%)	53 (93.0)	98 (98.0)	0.11
Modified metabolic syndrome proxy,b median (IQR)	0 (0–1)	1 (0–1)	0.24
PPI use, n (%)	30 (52.6)	100 (100)	<0.001
Statin use, n (%)	12 (21.1)	50 (50.0)	<0.001
Ever smoker, n (%)	19 (33.3)	59 (59.0)	0.003
Family history of BE or esophageal cancer, n (%)	5 (8.8)	24 (24.2)	0.02
Dietc
Energy-adjusted fiber intake (g/1,000 kcal), median (IQR)	9.1 (7.3–12.3)	8.3 (6.2–10.1)	0.08
% Energy from fat, median (IQR)	36.5 (33.4–40.0)	36.5 (32.0–39.9)	0.67
HEI score, median (IQR)	68.1 (57.7–71.8)	64.6 (57.3–69.5)	0.40

BE, Barrett's esophagus; BMI, body mass index; PPI, proton pump inhibitor; HEI, Healthy Eating Index; IQR, interquartile range.

a

Wilcoxon or Fisher exact P values reported for difference between controls and BE.

b

Modified metabolic syndrome proxy variable was created, where 1 point was given for hypertension, hyperlipidemia, and diabetes each, allowing for a score of 0–3 points.

c

Subset (n = 103) of participants completed the Diet History Questionnaire II food frequency questionnaire.

Dietary intake

One hundred three subjects (73%) completed the food frequency questionnaire, with the following median values: energy-adjusted fiber, 8.6 g/1,000 kcal (interquartile range [IQR] 6.5–10.3); energy-adjusted fat, 36.5% of kcal (IQR 32.0–40.0); HEI score, 66.1 (IQR 67.7–69.9) (Table 1). Participants who completed the DHQ II were older than those who did not (mean age 61.6 years [SD 14.6] vs 56.8 years [SD 14.2]; P = 0.02); there were no other differences between these 2 groups. In multivariable polytomous logistic regression analyses, there were no significant associations comparing the fourth vs first quartiles of energy-adjusted fiber intake, fat intake, or HEI score with any EAC risk factors (data not shown).

Dietary intake patterns were then analyzed comparing controls and subjects with BE. In multivariable logistic regression analyses, increasing fiber intake was associated with a nonsignificant reduced odds of BE (adjusted odds ratio [aOR] 0.82, 95% confidence interval [CI] 0.66–1.02, P = 0.08). There were no significant associations between energy-adjusted fat intake or HEI score with BE (see Supplementary Table 1, Supplementary Digital Content 4, http://links.lww.com/CTG/B188).

Serum bile acids

Among the entire study group, the bile acids with greatest absolute abundance were GCDCA (median 266.5 nM), DCA (median 243.5 nM), GDCA (median 110.6 nM), and GCA (median 76.5 nM) (see Supplementary Table 2, Supplementary Digital Content 5, http://links.lww.com/CTG/B189). There were strong within-individual correlations between the taurine-conjugated and glycine-conjugated forms of bile acids (e.g., TCA:GCA, Spearman ρ = 0.85, P < 0.0001) as well as between the unconjugated primary bile acids CA and CDCA (Spearman ρ = 0.90, P < 0.0001; see Supplementary Figure 2, Supplementary Digital Content 2, http://links.lww.com/CTG/B186). Multivariable linear regression analyses were performed to identify clinical and dietary factors associated with serum bile acid levels. Older age and increased BMI were associated with increased total bile acids as well as increased levels of numerous individual bile acids, whereas PPI use and increasing HEI score were associated with lower levels of various bile acids (see Supplementary Table 3, Supplementary Digital Content 6, http://links.lww.com/CTG/B190). The significant associations with PPI use persisted in analyses restricted to control subjects. Other established risk factors for EAC such as male sex, smoking status, and family history were not significantly associated with serum bile acid levels. Only minimal associations were seen between dietary fat or fiber intake and levels of individual bile acids.

Analyses were then performed to assess associations between serum bile acid levels and stages of BE progression. Global bile acid pools were distinct comparing non-BE subjects and subjects with BE (MANOVA P = 0.002) and between non-BE and stages of BE neoplasia (MANOVA P = 0.004). There were clear shifts in individual bile acids with progression to HGD/EAC (Figure 1). In principal components analyses, the top 4 principal coordinates (PCs) collectively explained 86% of the variance in bile acids. Associations between bile acids and HGD/EAC were driven by forms of UDCA (main drivers of PC3, which explained 16% of variation) and the unconjugated primary bile acids CA and CDCA (main drivers of PC4, which explained 13% of variation) (see Supplementary Table 4, Supplementary Digital Content 7, http://links.lww.com/CTG/B191). In multivariable logistic regression analyses with adjustment for EAC risk factors, PC3 was inversely and PC4 was positively associated with HGD/EAC compared with ND and non-BE, respectively (P = 0.02 for both) (see Supplementary Figure 3, Supplementary Digital Content 3, http://links.lww.com/CTG/B187, Supplementary Table 5, Supplementary Digital Content 8, http://links.lww.com/CTG/B192).

[Figure omitted. See PDF]

When examining individual bile acids, CA was the only bile acid to demonstrate significant differences across non-BE and BE stages (Kruskal-Wallis P = 0.01), so we focused additional pairwise comparisons on CA (see Supplementary Table 2, Supplementary Digital Content 5, http://links.lww.com/CTG/B189). Increased CA was associated with HGD/EAC compared with non-BE in univariate analyses (Figure 2) and remained significant after adjusting for EAC risk factors (per log-fold change, aOR 2.03, 95% CI 1.11–3.71). Similar trends were seen after adjustment for HEI, fat, and fiber intake. As CA and CDCA were highly correlated, differences in the sum of the 2 unconjugated primary bile acids across BE stages were assessed in post hoc analyses and were also found to be independently associated with HGD/EAC (per log-fold change, aOR 1.81, 95% CI 1.04–3.13). There were nonsignificant trends toward decreased levels of unconjugated and conjugated forms of UDCA associated with HGD/EAC (see Supplementary Table 2, Supplementary Digital Content 5, http://links.lww.com/CTG/B189).

[Figure omitted. See PDF]

Tissue gene expression alterations associated with CA

Transcriptomic data were available for analysis in 102 subjects (34 controls, 32 NDBE, 21 IND/LGD, 15 HGD/EAC). We chose only to assess gene expression differences for those bile acids with significant differences in levels across groups. In light of the association between CA and HGD/EAC, tissue gene expression analyses were performed by RNA-sequencing comparing the highest vs lowest quartile of CA level, adjusted for underlying histology as well as EAC risk factors. There were 23 significantly differentially expressed genes comparing high vs low CA levels (Figure 3a, see Supplementary Table 6, Supplementary Digital Content 9, http://links.lww.com/CTG/B193). High CA levels were associated with increased expression of CDC25B, a promoter of DNA replication that is upregulated in EAC (24), as well as THBS2, which encodes thrombospondin 2, mediates immune infiltration in tumors, and is associated with poor prognosis in gastric and pancreatic adenocarcinoma (25–27). Notably, there were reductions in several genes responsible for lymphocyte recruitment, differentiation, and activation, including CDH26, CCL5, PTPRCAP, CD7, CD8A, and CD48. Several pathways were significantly altered in subjects with high CA levels, including increased cell cycle and DNA replication and decreased T-cell receptor signaling and natural killer cell mediated cytotoxicity (Figure 3b). In total, these findings suggest that CA may play a biological role in promoting esophageal neoplasia, in part by regulating the immune microenvironment.

[Figure omitted. See PDF]

DISCUSSION

This cross-sectional analysis of patients with and without BE found that the bile acid profile was markedly different in the progression of BE, independent of established EAC risk factors including diet. Notably, those with advanced neoplasia demonstrated an increase in the primary unconjugated bile acids, particularly CA, and trends toward a decrease in forms of UDCA. Tissue gene expression analyses demonstrated that high serum CA was associated with alterations potentially consistent with cancer promoting effects through immune microenvironment regulation and other mechanisms.

In this study, unconjugated primary bile acids and CA in particular were associated with HGD/EAC. CA has been shown to have proinflammatory and carcinogenic effects, with increases in CA observed in active inflammatory bowel disease in humans (28,29) and increased colonic neoplasia seen in male Fischer rats administered CA (30). The mechanism by which CA may promote neoplasia may be related to its hydrophobicity because hydrophobic bile acids generate reactive oxygen species and induce multiple cell stresses including DNA and mitochondrial damage (31). In vitro experiments have also shown that esophageal squamous carcinoma cells treated with CA had increased expression of matrix metalloproteinase-9 (32) and cells treated with CDCA showed increased production of prostaglandin E2 and vascular endothelial growth factor (33), suggesting these bile acids may mediate tumor-induced angiogenesis. Thus, circulating unconjugated primary bile acids may have direct effects on Barrett's epithelium that promote neoplastic progression. As such, they may represent a potential biomarker of advanced neoplasia in BE.

High levels of CA were associated with numerous tissue gene expression alterations. There was upregulation of CDC25B, which encodes M-phase inducer phosphatase 2, promotes cell cycle progression, and was previously shown to be upregulated in EAC (24). THBS2 encodes thrombospondin 2, which is associated with poor prognosis in gastric and pancreatic adenocarcinoma, potentially through mediation of immune infiltration in tumors (25–27). In the L2-IL1B mouse model, development of EAC is associated with a shift in the immune microenvironment towards increased myeloid lineage cells (myeloid-derived suppressor cells and neutrophils) and reduced lymphocyte lineage cells (such as natural killer cells) (7,13). Similar findings were seen in this study because high CA was associated with altered expression of numerous genes that would suggest a reduction or inhibition of lymphocyte lineage cells. These findings together suggest a potential relationship between the circulating bile acid pool and the immune microenvironment, potentially through immune evasion, to regulate cancer development in BE.

There were nonsignificant trends toward decreased levels of UDCA and its conjugated forms associated with advanced neoplasia, consistent with prior research suggesting UDCA may have anti-inflammatory properties. UDCA has been shown to inhibit proliferation of cancer cells in animal and patient models of colon (34–37) and esophageal cancer (14,38). The mechanisms by which UDCA may exert its effects continue to be explored, with studies suggesting it may inhibit DNA damage and oxidative stress (39) and stimulate intestinal epithelial cell migration after cell injury (40). In mice, UDCA also enhances antitumor immunity through transforming growth factor β degradation (41). In a prior clinical trial in patients with BE, UDCA administration for 6 months did not cause significant changes in tissue markers of oxidative DNA damage, cell proliferation, or apoptosis (14). However, it is unknown whether the treatment changed serum UDCA levels; circulating rather than refluxate UDCA may have antineoplastic properties in BE, and interventions aimed at increasing serum UDCA should continue to be explored.

Individual serum bile acids in this study were most associated with age, BMI, PPI use, and diet (specifically, HEI score). Although fiber was not significantly associated with most bile acids, there was a trend toward fiber intake being inversely associated with BE, suggestive that fiber-rich foods may have beneficial effects via mechanisms other than shifting the bile acid profile, including decreasing proinflammatory cytokines and altering the glycemic response (42). Many of these relationships between patient characteristics and the circulating bile acid pool are likely mediated in part by gut microbiome composition. High levels of unconjugated primary bile acids in advanced neoplasia suggest that these subjects may have high numbers of bacteria that express BSH, the enzyme responsible for bile acid deconjugation. Microbiome-based therapies that reduce the expression or activity of BSH, or alternatively deplete the unconjugated primary bile acid pool via conversion to secondary bile acids, represent potential approaches to shape a favorable bile acid pool in patients with BE.

This study has numerous strengths. This is a large population of subjects with untreated BE, with and without dysplasia, as well as non-BE controls. Extensive demographic and clinical data were collected, which allowed for analyses to show that CA is independently associated with Barrett's-associated advanced neoplasia, even after adjusting for EAC risk factors. Long-term dietary intake patterns were assessed using a validated food frequency questionnaire, allowing for incorporation of this potentially key confounder in analyses. Tissue gene expression analyses offered insights into the potential biological effects of circulating bile acids in this patient population. Finally, rigorous histological categorization was performed using a combination of local expert and central pathologist reads. This study was not without limitations. This was a cross-sectional analysis; thus, conclusions cannot be drawn regarding causality. It is possible that short-term dietary intake patterns may have had an important influence on circulating bile acid composition; however, the measures used to assess dietary intake in this study have demonstrated validity and reliability (16) and have been used in multiple studies evaluating the effects of long-term dietary habits on disease and mortality (43,44). Only a subset of subjects completed the dietary questionnaire, and 12-month dietary assessment may not be reflective of intake over the life course. While studies have suggested that PPI users have altered gut microbiomes (45,46) which may contribute to changes in the bile acid composition, we were unable to assess extensively for potential effects of PPI use on serum bile acid levels due to collinearity with BE status because all patients with BE were on a PPI. Stool samples were not collected for gut microbiome and bile acid analyses, which could have led to further insights into the relationship between diet, gut bacteria and bile acid metabolism, and progression to EAC. However, because primary bile acid metabolism and reabsorption largely occur in the small intestine, stool analyses may not have been as meaningful.

In conclusion, serum bile acid composition and CA in particular was independently associated with advanced neoplasia in patients with BE. In addition, high levels of CA were associated with a variety of alterations to tissue gene expression, with a pattern suggestive of a possible relationship between circulating bile acids and the immune microenvironment. While previous studies have mainly examined the role of refluxate bile acids, this study suggests that systemic bile acids may have important biological effects contributing to esophageal carcinogenesis. Future studies are needed to explore associated gut microbiome changes such as bacterial deconjugation through BSH, to further evaluate CA and CDCA as potential risk factors and UDCA derivatives as potential protective factors in the progression of BE, and to determine whether these bile acids represent viable therapeutic targets for EAC prevention.

CONFLICTS OF INTEREST

Guarantor of the article: Julian A. Abrams, MD, MS.

Specific author contributions: A.K.: literature review, data analysis, main author of manuscript, approved final draft submission. P.G., Y.L., H.L., and Z.J.: data analysis, approved final draft submission. P.G.I., K.K.W., G.W.F., and C.J.L.: data acquisition, critical review of manuscript, approved final draft submission. G.G.G., A.D.P., and S.M.L.: data acquisition, approved final draft submission. J.G.: study design, critical review of manuscript, approved final draft submission. A.K.R. and T.C.W.: critical review of manuscript, approved final draft submission. H.H.W.: data analysis, critical review of manuscript, approved final draft submission. M.Q.: critical review of manuscript, approved final draft submission. J.A.A.: study design, data acquisition, manuscript writing and review, principal investigator, approved final draft submission.

Financial support: This research was supported in part by the National Institutes of Health (U54CA163004, R01CA255298, R01CA272898). This research was also funded in part through the NIH/NCI Cancer Center Support Grant P30CA013696 and used the Genomics and High Throughput Screening Shared Resource and by the Columbia University Digestive and Liver Disease Research Center grant P30DK132710. This research was further supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant Number UL1TR001873, and used the Biomarkers Core Laboratory. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Potential competing interests: None to report.

Study Highlights

WHAT IS KNOWN

• ✓ While reflux bile acids are thought to promote esophageal adenocarcinoma, the role of systemic bile acids is not known.

WHAT IS NEW HERE

• ✓ Serum bile acid composition was found to be markedly different in patients with Barrett's esophagus and associated advanced neoplasia, independent of dietary intake.

• ✓ High levels of cholic acid were associated with a variety of alterations in tissue gene expression, with a pattern suggestive of a possible relationship between circulating bile acids and the immune microenvironment.

• ✓ These findings suggest that systemic bile acids may have important biological effects contributing to esophageal carcinogenesis.

Author Notes

Correspondence: Julian A. Abrams, MD, MS. E-mail: [email protected].

SUPPLEMENTARY MATERIAL accompanies this paper at http://links.lww.com/CTG/B185, http://links.lww.com/CTG/B186, http://links.lww.com/CTG/B187, http://links.lww.com/CTG/B188, http://links.lww.com/CTG/B189, http://links.lww.com/CTG/B190, http://links.lww.com/CTG/B191, http://links.lww.com/CTG/B192, http://links.lww.com/CTG/B193

REFERENCES

1.Uhlenhopp, DJ, Then, EO, Sunkara, T, et al. Epidemiology of esophageal cancer: Update in global trends, etiology and risk factors. Clin J Gastroenterol 2020;13(6):1010–21.

2.Friedenberg, FK, Hanlon, A, Vanar, V, et al. Trends in gastroesophageal reflux disease as measured by the National Ambulatory Medical Care Survey. Dig Dis Sci 2010;55(7):1911–7.

3.Kong, CY, Nattinger, KJ, Hayeck, TJ, et al. The impact of obesity on the rise in esophageal adenocarcinoma incidence: Estimates from a disease simulation model. Cancer Epidemiol Biomarkers Prev 2011;20(11):2450–6.

4.He, D, Huang, X, Wang, ZP, et al. Dietary fat intake and risk of esophageal carcinoma: A meta-analysis of observational studies. Oncotarget 2017;8(58):99049–56.

5.Coleman, HG, Murray, LJ, Hicks, B, et al. Dietary fiber and the risk of precancerous lesions and cancer of the esophagus: A systematic review and meta-analysis. Nutr Rev 2013;71(7):474–82.

6.Ibiebele, TI, Hughes, MC, Whiteman, DC, et al; Australian Cancer Study. Dietary patterns and risk of oesophageal cancers: A population-based case-control study. Br J Nutr 2012;107(8):1207–16.

7.Münch, NS, Fang, HY, Ingermann, J, et al. High-fat diet accelerates carcinogenesis in a mouse model of Barrett's esophagus via interleukin 8 and alterations to the gut microbiome. Gastroenterology 2019;157(2):492–506.e2.

8.Wan, Y, Yuan, J, Li, J, et al. Unconjugated and secondary bile acid profiles in response to higher-fat, lower-carbohydrate diet and associated with related gut microbiota: A 6-month randomized controlled-feeding trial. Clin Nutr 2020;39(2):395–404.

9.Cai, J, Sun, L, Gonzalez, FJ. Gut microbiota-derived bile acids in intestinal immunity, inflammation, and tumorigenesis. Cell Host Microbe 2022;30(3):289–300.

10.Ajouz, H, Mukherji, D, Shamseddine, A. Secondary bile acids: An underrecognized cause of colon cancer. World J Surg Oncol 2014;12(1):164.

11.Nguyen, TT, Ung, TT, Kim, NH, et al. Role of bile acids in colon carcinogenesis. World J Clin Cases 2018;6(13):577–88.

12.Bernstein, H, Bernstein, C, Payne, CM, et al. Bile acids as endogenous etiologic agents in gastrointestinal cancer. World J Gastroenterol 2009;15(27):3329–40.

13.Quante, M, Bhagat, G, Abrams, J, et al. Bile acid and inflammation activate gastric cardia stem cells in a mouse model of Barrett's-like metaplasia. Cancer Cell 2012;21(1):36–51.

14.Banerjee, B, Shaheen, NJ, Martinez, JA, et al. Clinical study of ursodeoxycholic acid in Barrett's esophagus patients. Cancer Prev Res (Phila) 2016;9(7):528–33.

15.Kühn, T, Stepien, M, López-Nogueroles, M, et al. Prediagnostic plasma bile acid levels and colon cancer risk: A prospective study. J Natl Cancer Inst 2020;112(5):516–24.

16.Subar, AF, Thompson, FE, Kipnis, V, et al. Comparative validation of the Block, Willett, and National Cancer Institute food frequency questionnaires: The Eating at America's Table Study. Am J Epidemiol 2001;154(12):1089–99.

17.Kennedy, ET, Ohls, J, Carlson, S, et al. The Healthy Eating Index: Design and applications. J Am Diet Assoc 1995;95(10):1103–8.

18.Reedy, J, Lerman, JL, Krebs-Smith, SM, et al. Evaluation of the Healthy Eating Index-2015. J Acad Nutr Diet 2018;118(9):1622–33.

19.Krebs-Smith, SM, Pannucci, TE, Subar, AF, et al. Update of the Healthy Eating Index: HEI-2015. J Acad Nutr Diet 2018;118(9):1591–602.

20.Haeusler, RA, Camastra, S, Nannipieri, M, et al. Increased bile acid synthesis and impaired bile acid transport in human obesity. J Clin Endocrinol Metab 2016;101(5):1935–44.

21.Love, MI, Huber, W, Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15(12):550.

22.Subramanian, A, Tamayo, P, Mootha, VK, et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA 2005;102(43):15545–50.

23.Korotkevich, G, Sukhov, V, Sergushichev, A. Fast gene set enrichment analysis. bioRxiv 2019. doi:10.1101/060012

24.Hourihan, RN, O'Sullivan, GC, Morgan, JG. Transcriptional gene expression profiles of oesophageal adenocarcinoma and normal oesophageal tissues. Anticancer Res 2003;23(1A):161–5.

25.Yang, H, Sun, B, Fan, L, et al. Multi-scale integrative analyses identify THBS2+ cancer-associated fibroblasts as a key orchestrator promoting aggressiveness in early-stage lung adenocarcinoma. Theranostics 2022;12(7):3104–30.

26.Zhang, S, Yang, H, Xiang, X, et al. THBS2 is closely related to the poor prognosis and immune cell infiltration of gastric cancer. Front Genet 2022;13:803460.

27.Liao, X, Wang, W, Yu, B, et al. Thrombospondin-2 acts as a bridge between tumor extracellular matrix and immune infiltration in pancreatic and stomach adenocarcinomas: An integrative pan-cancer analysis. Cancer Cell Int 2022;22(1):213.

28.Franzosa, EA, Sirota-Madi, A, Avila-Pacheco, J, et al. Gut microbiome structure and metabolic activity in inflammatory bowel disease. Nat Microbiol 2019;4(2):293–305.

29.Lloyd-Price, J, Arze, C, Ananthakrishnan, AN, et al. Multi-omics of the gut microbial ecosystem in inflammatory bowel diseases. Nature 2019;569(7758):655–62.

30.Deschner, EE, Cohen, BI, Raicht, RF. Acute and chronic effect of dietary cholic acid on colonic epithelial cell proliferation. Digestion 1981;21(6):290–6.

31.Chen, X, Overcash, R, Green, T, et al. The tumor suppressor activity of the transmembrane protein with epidermal growth factor and two follistatin motifs 2 (TMEFF2) correlates with its ability to modulate sarcosine levels. J Biol Chem 2011;286(18):16091–100.

32.Li, S, Ung, TT, Nguyen, TT, et al. Cholic acid stimulates MMP-9 in human colon cancer cells via activation of MAPK, AP-1, and NF-κB activity. Int J Mol Sci 2020;21(10):E3420.

33.Soma, T, Kaganoi, J, Kawabe, A, et al. Chenodeoxycholic acid stimulates the progression of human esophageal cancer cells: A possible mechanism of angiogenesis in patients with esophageal cancer. Int J Cancer 2006;119(4):771–82.

34.Kohno, H, Suzuki, R, Yasui, Y, et al. Ursodeoxycholic acid versus sulfasalazine in colitis-related colon carcinogenesis in mice. Clin Cancer Res 2007;13(8):2519–25.

35.Serfaty, L, De Leusse, A, Rosmorduc, O, et al. Ursodeoxycholic acid therapy and the risk of colorectal adenoma in patients with primary biliary cirrhosis: An observational study. Hepatology 2003;38(1):203–9.

36.Huang, WK, Hsu, HC, Liu, JR, et al. The association of ursodeoxycholic acid use with colorectal cancer risk: A nationwide cohort study. Medicine (Baltimore) 2016;95(11):e2980.

37.Alberts, DS, Martínez, ME, Hess, LM, et al. Phase III trial of ursodeoxycholic acid to prevent colorectal adenoma recurrence. J Natl Cancer Inst 2005;97(11):846–53.

38.Peng, S, Huo, X, Rezaei, D, et al. In Barrett's esophagus patients and Barrett's cell lines, ursodeoxycholic acid increases antioxidant expression and prevents DNA damage by bile acids. Am J Physiol Gastrointest Liver Physiol 2014;307(2):G129–139.

39.Goldman, A, Condon, A, Adler, E, et al. Protective effects of glycoursodeoxycholic acid in Barrett's esophagus cells. Dis Esophagus 2010;23(2):83–93.

40.Golden, JM, Escobar, OH, Nguyen, MVL, et al. Ursodeoxycholic acid protects against intestinal barrier breakdown by promoting enterocyte migration via EGFR- and COX-2-dependent mechanisms. Am J Physiol Gastrointest Liver Physiol 2018;315(2):G259–71.

41.Shen, Y, Lu, C, Song, Z, et al. Ursodeoxycholic acid reduces antitumor immunosuppression by inducing CHIP-mediated TGF-β degradation. Nat Commun 2022;13(1):3419.

42.Sun, L, Zhang, Z, Xu, J, et al. Dietary fiber intake reduces risk for Barrett's esophagus and esophageal cancer. Crit Rev Food Sci Nutr 2017;57(13):2749–57.

43.Panizza, CE, Shvetsov, YB, Harmon, BE, et al. Testing the predictive validity of the Healthy Eating Index-2015 in the multiethnic cohort: Is the score associated with a reduced risk of all-cause and cause-specific mortality? Nutrients 2018;10(4):452.

44.Chiuve, SE, Fung, TT, Rimm, EB, et al. Alternative dietary indices both strongly predict risk of chronic Disease123. J Nutr 2012;142(6):1009–18.

45.Yang, YCSH, Chang, HW, Lin, IH, et al. Long-term proton pump inhibitor administration caused physiological and microbiota changes in rats. Sci Rep 2020;10:866.

46.Imhann, F, Bonder, MJ, Vich Vila, A, et al. Proton pump inhibitors affect the gut microbiome. Gut 2016;65(5):740–8.