Causal immune-inflammation mapping of the GERD-Barrett-adenocarcinoma cascade identifies FGF19-HLA-DR⁺ T-cell axis driving esophagogastric junction metaplasia

Abstract

Background

Gastroesophageal reflux disease (GERD), Barrett’s esophagus (BE), and esophageal adenocarcinoma (EAC) form a recognized pathological continuum, but the causal immune and inflammatory processes driving progression remain incompletely defined.

Methods

Summary-level genome-wide association study (GWAS) data were analyzed for 91 circulating inflammatory proteins (n = 14,824), 731 immune cell phenotypes (n = 3,757), and three esophageal diseases (GERD, BE, and EAC). Bidirectional and two-step Mendelian randomization (MR) were used to infer causal effects and mediation, with Cochran’s Q, MR-Egger, and MR-PRESSO applied to assess heterogeneity and pleiotropy. Causal interaction networks were reconstructed to map immune- and inflammation-dominant regulatory patterns across disease stages, and MR-prioritized signals were experimentally validated in esophagogastric junction (EGJ) organoids and mouse models.

Results

MR supported a causal GERD-BE-EAC sequence, with BE mediating 31.95% of the total GERD-to-EAC effect. In total, 139 immune cell traits and 29 inflammatory proteins showed causal links to disease risk. Mediation analyses highlighted M-CSF1 and HLA-DR+CD4 + T cells as central hubs. Guided by the MR-prioritized FGF19-HLA-DR + T-cell axis, experimental studies demonstrated that FGF19 promotes EGJ glandular conversion and increases infiltration of HLA-DR+ CD4+/CD8 + T cells, validated in EGJ organoids and mouse models.

Conclusions

This integrated genetic and experimental framework delineates a bidirectional immune-inflammation regulatory network underlying progression from GERD to BE and EAC. FGF19 emerges as a candidate cytokine driving EGJ glandular remodeling through HLA-DR+ CD4+/CD8 + T-cell associated immune activation, providing candidate molecular targets for early prevention and intervention.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-08552-w.

Highlights

Genetic causal inference supports a GERD-BE-EAC progression, with Barrett’s esophagus mediating 31.95% of the total GERD to EAC effect.

Large-scale MR profiling identifies immune-inflammation drivers of esophageal disease risk, implicating 139 immune cell traits and 29 circulating inflammatory proteins.

Mediation and network analyses prioritize key immune-inflammation axes, highlighting CD11c+ monocytes, CXCL10, FGF19, and HLA-DR+CD4 + T cells, with M-CSF1 and HLA-DR+CD4 + T cells as central hubs.

In vivo validation confirms the MR-prioritized FGF19-HLA-DR + T-cell axis, showing FGF19 promotes EGJ glandular conversion and increases HLA-DR+ CD4+/CD8 + T-cell infiltration in organoid and mouse models.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-08552-w.

Introduction

Esophageal adenocarcinoma (EAC) has shown a marked increase in incidence over the past few decades, particularly in Western populations, and remains associated with poor prognosis despite therapeutic advances [1, 2]. Mounting evidence indicates that EAC arises through a pathological continuum beginning with gastroesophageal reflux disease (GERD), followed by Barrett’s esophagus (BE), and culminating in malignant transformation [3, 4]. While this GERD-BE-EAC cascade is well established clinically, the underlying biological mechanisms driving the stepwise progression remain incompletely understood.

Chronic inflammation has long been recognized as a central process linking tissue injury to malignant transformation in the gastrointestinal tract [5]. Reflux-induced mucosal damage initiates local immune activation and cytokine secretion, which may shape the microenvironment toward persistent inflammation and genomic instability [6]. However, whether systemic inflammatory mediators causally contribute to the transition from reflux to metaplasia and cancer, or merely reflect downstream effects, remains uncertain [7]. Likewise, immune cells represent both effector and regulatory components in this process, yet the directionality and hierarchy of immune-inflammation interactions across disease stages are poorly defined [8].

Mendelian randomization (MR) offers a powerful framework to infer causality between exposures and outcomes using genetic variants as instrumental variables, thereby minimizing confounding and reverse causation inherent in observational studies [9]. Recent large-scale genome-wide association studies (GWAS) have generated comprehensive datasets for circulating inflammatory proteins and immune cell traits, providing unprecedented opportunities to explore immune and inflammatory mechanisms at a causal level [10, 11]. Integrating these datasets with esophageal disease GWAS enables a systematic dissection of the immune-inflammation axis throughout disease evolution.

In this study, we applied bidirectional and two-step MR analyses to examine causal relationships among GERD, BE, and EAC, as well as between immune cell phenotypes, inflammatory proteins, and disease risk. We further quantified mediation effects to identify key intermediates and constructed immune and inflammation dominant causal networks to visualize regulatory hierarchies. By mapping the causal architecture connecting immune dysregulation, inflammatory signaling, and esophageal disease progression, this work aims to clarify the mechanistic links bridging reflux, metaplasia, and carcinoma, and to identify potential immuno-inflammatory targets for early intervention and cancer prevention (Fig. 1).

Fig. 1.

Work flow for causal immune-inflammation mapping of the GERD-Barrett esophagus-adenocarcinoma cancer cascade identifies FGF19-HLA-DR⁺ T-cell axis driving esophagogastric junction metaplasia

Materials and methods

Data sources

The dataset of 91 circulating inflammatory proteins was obtained from the study by Zhao et al., which performed GWAS analyses in 11 cohorts comprising 14,824 individuals of European ancestry using the Olink Target Inflammation panel [12]. Summary statistics for these proteins (GCST90274758-GCST90274848) were retrieved from the GWAS Catalog (https://www.ebi.ac.uk/gwas/home). In addition, GWAS data for immune traits (GCST90001391-GCST90002121) were obtained from the same catalog, including comprehensive information on 731 immune cell phenotypes collected from 3,757 European individuals [13].

Summary-level GWAS data for GERD were derived from Verma et al., including 175,991 cases and 236,737 controls [14]. BE data were obtained from Schröder et al. [15] (3,867 cases and 12,534 controls), and EAC data were obtained from Gharahkhani et al. (4,112 cases and 17,159 controls) [15, 16]. Collectively, 12,911,041 SNPs were included. All participants were of European ancestry to minimize potential bias from population stratification and geographic or ethnic heterogeneity. The complete list of GWAS accession IDs is provided in Table S1.

Selection of instrumental variables (IVs)

For esophageal diseases, genome-wide significance was defined as P < 5 × 10⁻⁸. To ensure adequate instrument numbers, a relaxed threshold of P < 1 × 10⁻⁵ was applied when fewer than one SNP was available for esophageal diseases, circulating inflammatory proteins, or immune cell traits. To avoid linkage disequilibrium (LD) bias, clumping was performed with an LD threshold of r² < 0.001 and a window size of 10,000 kb, ensuring independence of selected SNPs. Weak instrument bias was evaluated using the F-statistic, calculated as:

where N is the sample size, K is the number of IVs, and R² is the proportion of variance in the exposure explained by IVs. All retained instrumental variables across the MR analyses showed F-statistics > 10, supporting the robustness of the causal estimates against weak instrument bias (Tables S2).

MR analysis

We conducted MR analyses to systematically investigate causal relationships and regulatory networks between esophageal diseases (GERD, BE, and EAC), immune cell phenotypes, and circulating inflammatory proteins. The overall study design was based on a bidirectional two-sample MR framework combined with mediation MR to improve causal inference and mechanistic interpretability. This design was specifically chosen to (1) evaluate directional relationships among GERD, BE, and EAC, (2) distinguish exposure-to-disease effects from reverse causation, and (3) identify intermediate immune-inflammatory mediators within the disease continuum. Such a framework is particularly suitable for complex, multi-stage diseases where temporal progression and biological pathways need to be disentangled.

Bidirectional two-sample MR was used to rigorously evaluate causal directions among esophageal diseases and to establish the esophageal disease causal chain. Subsequently, to delineate the dynamic involvement of immune cells and inflammatory proteins during the disease process, muti-bidirectional MR analyses were performed to assess: the causal effect of immune cells on esophageal diseases; the reverse effect of esophageal diseases on immune cells; the causal effect of inflammatory proteins on esophageal diseases and the reverse effect of esophageal diseases on inflammatory proteins.

Inverse-variance weighted (IVW) MR was applied as the primary method to estimate causal effects, with weighted median (WME), weighted mode (WM), simple mode (SM), and MR-Egger regression used as complementary approaches. Heterogeneity was assessed using Cochran’s Q test; if significant heterogeneity was present (P < 0.05), IVW with a random-effects model was applied, otherwise a fixed-effects model was used. Horizontal pleiotropy was tested with MR-Egger regression (P > 0.05 indicating no pleiotropy). MR-PRESSO was employed to detect and correct for outlier SNPs, reducing pleiotropy-related bias. When outlier SNPs were detected, the corresponding variants were removed according to the MR-PRESSO outlier test, and causal estimates were re-evaluated using the corrected results. Mediation MR analysis was conducted using a two-step procedure. In the first step, the causal effect of the exposure on candidate mediators was estimated (β₁). In the second step, the effect of the mediator was estimated (β₂). The mediated effect (β₁ × β₂) and the proportion of mediation were evaluated using the delta method.

All analyses were conducted in R (version 4.4.2) using the “TwoSampleMR” package for two-sample MR and the “MR-PRESSO” package for outlier detection. All statistical tests were two-sided, with P < 0.05 considered statistically significant. Results are reported as odds ratios (ORs) with 95% confidence intervals (CIs).

EGJ-derived organoid culture and treatment

Briefly, EGJ tissues were enzymatically dissociated and embedded in growth factor-reduced Matrigel, followed by culture in advanced DMEM/F12-based organoid medium supplemented with essential growth factors. For stimulation experiments, organoids were treated with recombinant mouse FGF19 (50ng/ml) for defined time periods. Control organoids received vehicle treatment only. To assess glandular differentiation, organoid morphology was evaluated by phase-contrast microscopy, and histological analyses were performed on paraffin-embedded organoid sections stained with hematoxylin and eosin (H&E). Expression of glandular and squamous epithelial markers (KRT7 and p63) was assessed by immunofluorescence staining.

Mouse model of chronic bile reflux-associated EGJ remodeling

Genetically engineered Lgr5-CreERT2; IL1B^tg; p16^flox/flox; Kras^LSL-G12D mice (C57BL/6 background) were obtained from GemPharmatech (Nanjing, China) and maintained under specific pathogen-free conditions. This model has been previously validated to recapitulate Barrett’s-like dysplastic progression at the EGJ under chronic bile reflux conditions [17]. Mice were genotyped by PCR before enrollment and randomly assigned to experimental groups.

At 3 months of age, recombination was induced by intraperitoneal tamoxifen (75 mg/kg/day; Sigma-Aldrich) for five consecutive days to activate Kras^G12D and delete p16. To ensure appropriate experimental controls, littermate mice receiving the same tamoxifen regimen but without deoxycholic acid (DCA) exposure were included as genetic and induction controls. In addition, wild-type C57BL/6 mice subjected to identical housing and treatment schedules served as baseline physiological controls.

After recombination, mice in the reflux model group received 0.2% DCA (Sigma-Aldrich) in drinking water for six months to model chronic bile reflux, while control groups received regular drinking water under otherwise identical conditions. For FGF19 intervention, mice were treated with recombinant FGF19 (200 ng, subcutaneous injection, three times per week) during the DCA exposure period according to the experimental design, whereas vehicle-treated mice were used as treatment controls to exclude injection-related effects. Animals were monitored weekly for weight, behavior, and general health. At 9 months of age, mice were euthanized and tissues containing the EGJ were collected for immunohistochemistry. Histological inflammation and metaplasia were independently scored by two blinded gastrointestinal pathologists using established criteria to ensure objective comparison across control and experimental groups.

Flow cytometry analysis

Single-cell suspensions from organoid co-cultures or mouse EGJ tissues were prepared using mechanical and enzymatic dissociation. Cells were stained with fluorochrome-conjugated antibodies against CD45, CD4, CD8, and HLA-DR. Dead cells were excluded using viability dyes. Flow cytometry data were analyzed using FlowJo, and gating strategies were standardized across experiments.

Statistical analysis

Quantitative experimental data are presented as mean ± standard error of the mean (SEM). Comparisons between two groups were performed using student’s t-test, while multiple-group comparisons were conducted using one-way ANOVA followed by appropriate post hoc tests, as appropriate for normally distributed data. For MR statistics, effect estimates were primarily derived using the IVW method as the main estimator, with complementary sensitivity analyses including MR-Egger, WME, WM, and MR-PRESSO to improve robustness and assess potential pleiotropy and outlier effects. These statistical approaches were selected to strengthen causal inference under instrumental variable assumptions and to reduce bias arising from invalid instruments. To ensure consistency and transparency of statistical reporting, effect sizes were uniformly expressed as ORs with corresponding 95% CIs for binary outcomes and as β coefficients with 95% CIs for continuous traits, together with their associated P values. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant.

Results

Causal risk chain of esophageal diseases

As shown in Fig. 2A, the IVW analysis demonstrated a significant positive association between GERD and the risk of BE (OR = 1.163, 95% CI: 1.053–1.284, P = 0.003). Moreover, both GERD (OR = 2.071, 95% CI: 1.538–2.789, P < 0.001) and BE (OR = 1.216, 95% CI: 1.066–1.387, P = 0.003) were associated with an increased risk of EAC. In contrast, reverse MR analyses yielded no significant associations (Fig. 2B), supporting the directionality of the forward MR results, namely that GERD is a risk factor for BE, and both GERD and BE further contribute to an elevated incidence of EAC (Tables S3).

Fig. 2.

Exploring the causal links between gastroesophageal reflux, Barrett’s esophagus, and esophageal adenocarcinoma using bidirectional MR analysis. (A) Forward MR analysis results of the esophageal disease axis. (B) Reverse MR analysis results of the esophageal disease axis

The MR-Egger intercept suggested no evidence of horizontal pleiotropy. MR-PRESSO analysis indicated no potential outlier SNPs. Cochran’s Q test revealed heterogeneity in the MR analyses of EAC on BE and EAC on GERD, for which the IVW random-effects model was applied. No heterogeneity was detected in the other analyses (Table 1).

Table 1.

Assessment of heterogeneity and pleiotropy in the causal relationship analysis among gastroesophageal reflux disease, Barrett’s esophagus, and esophageal adenocarcinoma

	Exposure	Outcome	Heterogeneity		Pleiotropy		Presso
			Q	Pval	Intercept	Pval	RSSobs	Pval
forward MR analysis	Gastroesophageal reflux disease	Barrett’s esophagus	79.323	0.259	-0.015	0.366	81.596	0.339
	Barretts esophagus	Esophageal adenocarcinoma	29.974	0.052	0.010	0.772	33.158	0.065
	Gastroesophageal reflux disease	Esophageal adenocarcinoma	74.815	0.080	-0.006	0.758	77.279	0.102
reverse MR analysis	Barrett’s esophagus	Gastroesophageal reflux disease	32.847	0.107	0.006	0.215	36.056	0.109
	Esophageal adenocarcinoma	Barretts esophagus	5.859	0.016	-	-	-	-
	Esophageal adenocarcinoma	Gastroesophageal reflux disease	11.400	0.010	0.066	0.092	18.978	0.064

Causal relationships between immune cells and esophageal diseases

After establishing the causal chain among esophageal diseases, we next investigated the bidirectional causal associations between immune cell phenotypes and different stages of disease progression (Fig. 3A-C).

Fig. 3.

Causality of immune cells in the esophageal disease axis. Comprehensive overview of bidirectional MR analysis among gastroesophageal reflux disease, barrett esophagus, and esophageal adenocarcinoma in relation to immune cell regulation (A-C), highlighting causally significant associations (D-F)

For GERD (Fig. 3D), a total of 68 immune cell traits demonstrated significant causal effects. Among these, the majority (n = 49) exhibited positive associations with GERD risk (OR range: 1.008–1.046), whereas a smaller subset (n = 8) was inversely associated with risk (OR range: 0.967–0.991). Conversely, GERD exerted causal effects on 18 immune cell traits, of which 8 showed positive associations (OR range: 1.245–1.415) and 10 exhibited negative associations (OR range: 0.690–0.795).

For BE (Fig. 3E), 31 immune cell traits were significantly associated with disease risk, including 11 with risk-enhancing effects (OR range: 1.038–1.145) and 20 with risk-reducing effects (OR range: 0.867–0.973). In the reverse direction, BE exerted broad immunological impacts, influencing 50 immune cell traits, of which 36 were positively associated (OR range: 1.083–1.154) and 14 were negatively associated (OR range: 0.863–0.923).

For EAC (Fig. 3F), 40 immune cell traits were identified as having significant causal relationships. Of these, 25 traits increased EAC risk (OR range: 1.054–1.217), while 15 traits conferred protective effects (OR range: 0.884–0.998). Conversely, EAC influenced 37 immune cell traits, including 7 with positive associations (β range: 1.193–1.443) and 30 with negative associations (β range: 0.751–0.855).

Collectively, these findings suggest a stage-specific and bidirectional interplay between immune cell phenotypes and esophageal disease progression, characterized by predominantly risk-enhancing effects in GERD and BE, and a more complex pattern of both deleterious and protective influences in EAC. Comprehensive results, including specific immune cell traits and effect estimates, are provided in Tables S4-S6.

Causal relationships between inflammatory proteins and esophageal diseases

Using bidirectional MR analyses of 91 circulating inflammatory proteins, we identified several proteins significantly associated with the risk of GERD, BE, and EAC (Fig. 4A-C).

Fig. 4.

Causality of inflammatory proteins in the esophageal disease axis. Comprehensive overview of bidirectional MR analysis among gastroesophageal reflux disease, barrett esophagus, and esophageal adenocarcinoma in relation to inflammatory proteins regulation (A-C), highlighting causally significant associations (D-F)

For GERD (Fig. 4D), elevated levels of C-X-C motif chemokine 10, fibroblast growth factor 21, C-X-C motif chemokine 11, and C-C motif chemokine 20 were positively associated with GERD risk (OR range: 1.042–1.065). In contrast, higher levels of SIR2-like protein 2, macrophage colony-stimulating factor 1 (M-CSF1), and C-C motif chemokine 23 were inversely associated with GERD risk (OR range: 0.954–0.956). Furthermore, GERD exerted causal effects on circulating proteins, leading to an increase in TGF-α (OR = 1.115, P = 0.011) and a decrease in IL-13 (OR = 0.887, P = 0.042).

For BE (Fig. 4E), higher levels of leukemia inhibitory factor receptor were associated with increased risk (OR = 1.225, P = 0.008). Conversely, elevated levels of adenosine deaminase, fibroblast growth factor 19, and latency-associated peptide transforming growth factor beta 1 were significantly associated with reduced BE risk (OR range: 0.865–0.866). BE also influenced circulating protein levels, resulting in reductions in vascular endothelial growth factor A, M-CSF1, transforming growth factor alpha, tumor necrosis factor receptor superfamily member 9, monocyte chemoattractant protein-1, interleukin-8, tumor necrosis factor ligand superfamily member 14, oncostatin-M, and sulfotransferase 1A1 (OR range: 0.933–0.964).

For EAC (Fig. 4F), elevated levels of M-CSF1, tumor necrosis factor, and interleukin-10 receptor subunit beta were associated with increased disease risk (OR range: 1.135–1.254), whereas higher levels of signaling lymphocytic activation molecule were inversely associated with risk (OR = 0.813, P = 0.012). Additionally, EAC exerted causal effects on circulating proteins, leading to increased stem cell factor (OR = 1.114, P = 0.012) and reduced interleukin-17 A (OR = 0.904, P = 0.037).

A complete summary of the MR analyses between esophageal diseases and inflammatory proteins is provided in Tables S7-S9.

Mediation effects

Mediation analyses provided further insight into the mechanistic pathways linking esophageal diseases, immune traits, and inflammatory proteins (Fig. 5; Tables S10). Notably, the pathway from GERD to EAC via BE demonstrated a robust mediation effect, with BE accounting for nearly one-third of the total effect (mediation proportion: 31.95%), underscoring its role as a pivotal intermediary in the GERD-EAC causal cascade (Table 2).

Fig. 5.

Causal mediation effects of immune traits and inflammatory proteins in the GERD-Barrett’s esophagus-esophageal adenocarcinoma sequence

Table 2.

Mediation analysis of inflammatory and immune factors across the GERD-Barrett esophagus-esophageal adenocarcinoma sequence

Exposure	Mediator	Outcome	Direct effect a			Direct effect b			Total effect			Mediation effect
			β	SE	Pval	β	SE	Pval	β	SE	Pval	Effect Size	Proportion
GERD	BE	EAC	1.016	0.122	< 0.001	0.237	0.069	0.001	0.754	0.162	< 0.001	0.241 (0.116∼0.366)	31.95%
CD11c+ monocyte %monocyte	C-X-C motif chemokine 10 levels	GERD	0.041	0.017	0.017	0.063	0.024	0.008	0.017	0.008	0.046	0.003 (0.001∼0.004)	15.80%
CD4 + CD8+ T cell %leukocyte	C-X-C motif chemokine 10 levels	GERD	-0.051	0.02	0.009	0.063	0.024	0.008	-0.022	0.01	0.031	-0.003 (-0.005∼-0.001)	14.59%
Lymphocyte Absolute Count	SIR2-like protein 2 levels	GERD	0.04	0.019	0.033	-0.047	0.02	0.021	-0.016	0.007	0.03	-0.002 (-0.003∼-0.001)	11.90%
BAFF-R on IgD+ CD24 + B cell	SIR2-like protein 2 levels	GERD	-0.023	0.01	0.028	-0.047	0.02	0.021	0.011	0.005	0.025	0.001 (9.44E-05∼0.002)	9.37%
CD4 + CD8+ T cell %leukocyte	C-X-C motif chemokine 11 levels	GERD	-0.047	0.022	0.038	0.044	0.021	0.043	-0.022	0.01	0.031	-0.002 (-0.003∼-0.001)	9.14%
CD38 on IgD+ CD38dim B cell	SIR2-like protein 2 levels	GERD	-0.039	0.015	0.009	-0.047	0.02	0.021	0.022	0.008	0.004	0.002 (0.001∼0.003)	8.08%
IgD+ CD38dim B cell %B cell	C-C motif chemokine 20 levels	GERD	0.057	0.024	0.017	0.042	0.02	0.038	0.03	0.012	0.011	0.002 (0.001∼0.004)	7.79%
BAFF-R on CD24 + CD27+ B cell	SIR2-like protein 2 levels	GERD	-0.021	0.01	0.041	-0.047	0.02	0.021	0.013	0.005	0.017	0.001 (2.510E-05∼0.002)	7.60%
Fibroblast growth factor 19 levels	HLA DR+ CD4 + T cell %lymphocyte	BE	-0.139	0.058	0.017	0.091	0.042	0.031	-0.144	0.049	0.004	-0.013 (-0.022∼-0.004)	8.73%
Fibroblast growth factor 19 levels	HLA DR+ CD4 + T cell %T cell	BE	-0.128	0.057	0.024	0.085	0.037	0.021	-0.144	0.049	0.004	-0.011 (-0.019∼-0.003)	7.56%
Signaling lymphocytic activation molecule levels	Naive CD8 + T cell Absolute Count	EAC	-0.118	0.057	0.036	0.123	0.051	0.017	-0.207	0.082	0.012	-0.015 (-0.024∼-0.005)	7.02%
Interleukin-10 receptor subunit beta levels	SSC-A on plasmacytoid Dendritic Cell	EAC	-0.109	0.055	0.048	-0.081	0.034	0.018	0.257	0.1	0.01	0.009 (0.002∼0.015)	3.44%

Beyond this disease-disease mediation, both immune cells and inflammatory proteins emerged as critical mediators of esophageal disease risk. In immune-inflammation-disease pathways, several immune cell traits influenced GERD risk through specific inflammatory proteins (e.g., CD11c⁺ monocytes and CD4⁺CD8⁺ T cells mediated via CXCL10), while in inflammation-immune-disease pathways, inflammatory proteins such as fibroblast growth factor 19 and signaling lymphocytic activation molecule exerted their effects on BE or EAC through defined immune cell subsets.

Collectively, these findings indicate that esophageal disease progression is may driven by intertwined regulatory mechanisms in which immune and inflammatory processes reciprocally act as mediators, forming a complex network that amplifies disease susceptibility and progression.

Interaction networks

To further delineate the regulatory architecture underlying esophageal disease progression, we integrated the results of bidirectional and mediation MR analyses to construct two complementary causal networks: an immune-dominant network and an inflammation-dominant network. Both models contributed substantively to disease susceptibility and progression, underscoring the dual and intertwined roles of immune regulation and inflammatory signaling.

The networks revealed numerous significant causal pathways, highlighting the centrality of specific immune subsets and inflammatory mediators in bridging distinct disease stages. Within the immune-dominant network (Fig. 6A), GERD-driven immune traits such as CD45RA on terminally differentiated CD8⁺ T cells not only modulated BE risk via CXCL10 but also exhibited direct associations with EAC, exemplifying their upstream regulatory role. Comparable patterns were observed for CD4⁺CD8⁺ T cells, central memory CD4⁺ T cells, and HLA-DR⁺CD4⁺ T cells, which collectively influenced disease risk through shaping the downstream inflammatory milieu.

Fig. 6.

Dominant MR network esophageal diseases progression. (A) Network for the progression of immune cell-dominated esophageal disease axis progression. (B) Network for the progression of inflammatory protein-dominated esophageal disease axis progression. Note: Factors enclosed in blue circles represent immune cells; those in orange triangles denote circulating inflammatory proteins; and those in gray rectangles indicate esophageal diseases. Factors labeled in red font act as potential mediators. Dashed lines represent causal relationships among non-disease traits (excluding GERD, BE, and EAC). Red solid lines indicate positive causal associations with the diseases, whereas blue solid lines denote negative causal associations

Conversely, the inflammation-dominant network (Fig. 6B) illustrated protein-to-cell cascades in which inflammatory proteins served as upstream regulators of immune phenotypes during BE and EAC progression. For example, BE exerted its effects on EAC susceptibility through MCP-1-mediated modulation of plasmacytoid dendritic cell activity. Notably, M-CSF1 emerged as a hub node, displaying causal links with all three esophageal diseases, thereby positioning itself as a potential central mediator of the immune-inflammation interplay.

Taken together, these networks underscore that esophageal disease evolution is driven by a core set of immune cells and inflammatory proteins functioning as regulatory hubs. Through complex and reciprocal interactions, they form a causal framework that integrates and transmits signals across disease stages, ultimately amplifying disease risk and progression.

FGF19 promotes glandular metaplasia at the gastroesophageal junction through HLA-DR⁺ CD4⁺/CD8⁺ T-cell activation

Building on the MR and mediation analyses identifying FGF19 as a potential circulating factor for BE and EAC, we next sought to determine whether FGF19 exerts a functional role in glandular epithelial differentiation at the gastroesophageal junction (GEJ) and to elucidate the underlying immune mechanisms.To further investigate the causal relationship between FGF19 and glandular differentiation, mouse GEJ-derived organoids were treated with recombinant FGF19 (Fig. 7A). FGF19 stimulation significantly promoted glandular morphogenesis, characterized by enhanced luminal formation and upregulation of KRT7 (Fig. 7B). Notably, FGF19-treated EGJ organoids exhibited a transition from squamous epithelial cell aggregates to glandular epithelial structures (Fig. 7C). Furthermore, FGFR4, the primary receptor of FGF19, was knocked down and this markedly attenuated FGF19-induced glandular epithelial conversion at the EGJ (Supplemental Fig. 1A and B).

Fig. 7.

FGF19-HLA DR + T cells axis promotes glandular metaplasia at the gastroesophageal junction. (A) Workflow for establishing mouse EGJ-derived organoids and FGF19 treatment (50 ng/mL). (B) Immunofluorescence of organoids showing DAPI (blue), p63 (red), and KRT7 (green). Scale bars, 50 μm. (C) Bright-field images and H&E staining of control and FGF19-treated organoids. Scale bars, 100 μm. (D) Schematic of the in vivo bile acid exposure model with vehicle or FGF19 treatment and subsequent SCJ analysis. (E) Representative EGJ gross images and IHC staining for p63 and KRT7 in control vs. FGF19-treated mice. Scale bars, 50 μm. (F) Quantification of p63 and KRT7 H-scores. (G) Representative flow cytometry plots of HLA-DR expression on CD4⁺ and CD8⁺ T cells from EGJ tissues. (H) Quantification of HLA-DR⁺ CD4⁺ and HLA-DR⁺ CD8⁺ T cells. Data are mean ± SEM

Consistent with the in vitro findings, in vivo mouse models of gastroesophageal junction injury revealed that FGF19 administration significantly enhanced glandular epithelial conversion at the EGJ (Fig. 7D and E). Immunohistochemical analyses demonstrated that FGF19 treatment resulted in increased KRT7 expression and a characteristic loss of p63 expression in the squamous epithelium at the EGJ (Fig. 7F). Notably, regions with high FGF19 expression exhibited increased infiltration of HLA-DR⁺ CD4⁺ and HLA-DR⁺ CD8⁺ T cells, suggesting a close spatial association between FGF19 signaling and antigen-presenting T-cell activation at the metaplastic niche (Fig. 7G and H).

Discussion

This study integrates large-scale Mendelian randomization analyses with experimental validation to delineate the immune-inflammatory mechanisms underlying the GERD-BE-EAC sequence. By leveraging genetic instruments for circulating inflammatory proteins and immune cell phenotypes, we first inferred a directionally consistent causal relationship at the genetic level in which GERD increases the risk of BE and subsequently EAC, with BE mediating nearly one-third of the total effect. Importantly, we further translated a genetically prioritized mediator into a tractable biological mechanism by experimentally validating the FGF19-HLA-DR⁺ T-cell axis may as a driver of glandular epithelial conversion at the EGJ.

A major contribution of this work lies in distinguishing immune and inflammatory factors that are genetically associated and potentially involved in disease progression from those that may reflect secondary responses. Notably, chemokine-driven inflammatory amplification has long been implicated in reflux-associated esophagitis and EAC pathobiology; prior mechanistic and translational work has highlighted the importance of chemokine networks in shaping leukocyte trafficking and tumor-promoting inflammation in EAC [18, 19]. Through bidirectional and mediation MR analyses, we identified multiple immune-inflammation regulatory pathways operating across disease stages. However, our MR results clarify their directionality and hierarchical relationships. In particular, inflammatory mediators such as CXCL10 and CXCL11 were genetically linked to an increased risk of GERD, consistent with their roles in chemotactic amplification of reflux-associated inflammation, whereas proteins such as M-CSF1 exhibited protective or stage-dependent effects, highlighting the complexity of immune regulation across disease evolution.

At the immune cell level, our analyses revealed stage-specific patterns of immune involvement. GERD was primarily associated with immune activation signatures, including CD11c⁺ monocytes and cytotoxic T-cell subsets, suggesting an inflammation-driven immune response to epithelial injury. BE, in contrast, showed enrichment of regulatory and memory-associated immune phenotypes, indicative of immune reprogramming under chronic inflammatory stimulation. EAC was characterized by immune profiles compatible with immune exhaustion and tumor-associated immune remodeling. Importantly, across these stages, HLA-DR-expressing CD4⁺ T cells consistently emerged as a functionally informative immune state, implicating antigen-presentation-associated T-cell activation in metaplastic and neoplastic contexts. This interpretation is supported by recent clinical immunology work validating HLA-DR as a robust, quantifiable surface signature of systemic T-cell activation across inflammatory and immune-dysregulation states, reinforcing the biological plausibility of using HLA-DR⁺ T-cell states as mechanistically informative readouts [20]. While the MR analyses provided a population-level causal map, translating these findings into biologically meaningful mechanisms required experimental validation [21]. Among the prioritized mediators, FGF19 emerged as a key cytokine linking immune regulation to BE risk through an HLA-DR⁺ CD4⁺ T-cell-dependent pathway.

FGF19 is a canonical endocrine output downstream of bile acid-activated FXR signaling, functioning as a negative feedback regulator of bile acid synthesis and a broader metabolic/immunomodulatory cue [22]. Given that BE is widely considered a reflux-associated squamocolumnar metaplasia, in which acid and bile reflux constitute key injurious exposures, the FXR-FGF19 axis provides a compelling conceptual bridge between refluxate composition and mucosal remodeling [23, 24]. Notably, FXR agonism is being explored clinically in BE prevention/interception settings, underscoring community interest in bile acid-FXR pathways as actionable levers in early disease [25, 26]. A key interpretive nuance is that FGF19/FGFR4 signaling is a well-described double-edged sword in oncology [27, 28]. Aberrant activation of the FGF19-FGFR4 axis has been implicated in tumor growth and progression across multiple solid tumors, and therapeutic strategies targeting this pathway are under active investigation [29]. Importantly, this existing translational drug-development landscape de-risks the targetability question, selective FGFR4 inhibitors have been evaluated clinically using FGF19 as a biomarker for pathway activation in advanced cancers, establishing both feasibility and precedent for biomarker-guided intervention along the FGF19–FGFR4 axis [30]. Therefore, our finding that FGF19 promotes EGJ glandular conversion should be interpreted as a stage-dependent remodeling program rather than uniformly “protective” or “oncogenic” In early reflux-metaplasia settings, FGF19-driven epithelial plasticity and immune activation may facilitate glandular re-epithelialization/adaptation; however, whether persistent or dysregulated activation of the same axis could later contribute to pro-neoplastic signaling requires explicit temporal dissection. This stage-dependence offers a coherent explanation for why endocrine growth factor pathways can exhibit apparently divergent associations across GERD, BE, and EAC risk strata in genetic studies, and motivates future work stratified by disease stage and exposure duration.

Our results also refine the immunological interpretation of HLA-DR⁺ T cells in Barrett-related remodeling [31]. HLA-DR expression on T cells is widely recognized as a marker of recent activation/proliferation and antigen-experienced effector differentiation. Accordingly, the upregulation of HLA-DR⁺ CD4⁺/CD8⁺ T cells in response to FGF19 in organoids and mouse models supports a model in which epithelial fate conversion at the EGJ is accompanied by an antigen-presentation-linked T-cell activation state. Rather than treating immune traits as purely quantitative abundance measures, our mediation framework and experimental data together suggest that HLA-DR⁺ T cells may be mechanistically closer to metaplastic niche formation than many bulk immune readouts [32]. From a translational perspective, this may support a testable biomarker strategy: cross-sectional validation of circulating FGF19 and HLA-DR⁺ T-cell frequencies across normal/GERD/BE/dysplasia/EAC and prospective assessment of whether baseline or longitudinal changes predict progression within BE surveillance would clarify whether these markers add incremental value beyond established clinical predictors and histopathology.

Several limitations should be noted. First, the immune cell traits used in the MR analyses were genetically inferred traits derived from GWAS data rather than direct measurements in esophageal tissues, and thus may not fully reflect dynamic, tissue-specific immune states during disease progression. Second, although organoid and mouse models provided functional support for the FGF19-HLA-DR⁺ T-cell axis, they cannot completely recapitulate the complexity of human GERD-BE-EAC progression, including the full tissue microenvironment, chronic disease exposure, and patient heterogeneity. Additionally, the current analyses were restricted to populations of European ancestry, and validation in diverse populations will be necessary to assess generalizability. Accordingly, these findings should be interpreted with caution and further validated in human samples.

Conclusion

In conclusion, this work provides genetically informed evidence for an immune-inflammation framework underlying the GERD-BE-EAC continuum and adds mechanistic plausibility by identifying and experimentally supporting a potential role of FGF19 in EGJ glandular remodeling in association with HLA-DR⁺ CD4⁺/CD8⁺ T-cell activation. This integrated MR-to-experiment strategy provides a roadmap for translating systemic mediator prioritization into tissue-level mechanisms and actionable interception targets at the reflux-metaplasia interface.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

YL and JZ designed the research strategy. SL and ZG worked with the in vitro experiments and immunofluorescence analysis. JD and YL performed the organoid culture. YT, GM and YH performed the statistical analysis and wrote the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by funds from the National Science Foundation of China (82403748), the Natural Science Foundation of Sichuan Province (2024NSFSC1293 and 2024NSFSC1873) and Basic Research Projects of Qinghai Provincial Department of Science and Technology (2023-ZJ-791).

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethical approval

All animal experimental protocols were approved by the Ethics Committee of West China Hospital, Sichuan University (Approval number: No.20220107005). All animal experiments were in accordance with the guide for the care and use of laboratory animals established by the United States National Institutes of Health.

Consent for publication

This manuscript does not include details, images, or videos relating to an individual person; therefore, consent for publication is not needed beyond the informed consent provided by all study participants as described above.

Competing interests

The authors declare that they have no competing interests.