Pooled serum proteomic profiling reveals shared and distinct molecular features in nonerosive reflux disease and reflux esophagitis: A retrospective observational study

Abstract

Objective

Gastroesophageal reflux disease comprises heterogeneous phenotypes, and symptom severity does not always correlate with endoscopic findings. This retrospective observational study aimed to characterize serum proteomic alterations in gastroesophageal reflux disease and to explore whether nonerosive reflux disease and reflux esophagitis exhibit distinct circulating protein signatures.

Methods

Seventy individuals were included, comprising 26 controls, 14 patients with nonerosive reflux disease, and 30 with reflux esophagitis. For exploratory group-level proteomic discovery, pooled serum samples from each group were analyzed using liquid chromatography-tandem mass spectrometry. Clinical characteristics, upper gastrointestinal endoscopy, high-resolution manometry, and 24-h pH-impedance monitoring were assessed as appropriate. Differentially expressed proteins were identified using a fold change threshold of 1.3 together with nominal statistical significance, followed by functional annotation and enrichment analyses.

Results

Age and sex were comparable among the groups. Nonerosive reflux disease and reflux esophagitis showed no significant differences in gastroesophageal reflux disease questionnaire score, disease duration, body mass index, or most reflux-monitoring parameters. However, patients with reflux esophagitis had more gas reflux episodes, lower distal contractile integral values, and a higher frequency of esophageal motility disorders, suggesting impaired motor function. Proteomic profiling identified 872 proteins, including 666 quantified proteins. Differential expression analysis identified 48 proteins in nonerosive reflux disease versus controls, 43 in reflux esophagitis versus controls, and 7 in nonerosive reflux disease versus reflux esophagitis. The exploratory nonerosive reflux disease-associated protein profile was mainly characterized by cytoskeletal remodeling, adhesion-related regulation, redox adaptation, and selective immune rearrangement, whereas the exploratory reflux esophagitis-associated profile showed features related to actin remodeling, injury-defense responses, and alterations in lipid homeostasis.

Conclusions

Nonerosive reflux disease and reflux esophagitis exhibited distinct but partially overlapping serum proteomic profiles in this exploratory pooled serum analysis. These preliminary findings suggest phenotype-related biological differences beyond reflux burden alone; however, they should be interpreted as hypothesis-generating and require validation in independent, individual-level cohorts.

Introduction

Gastroesophageal reflux disease (GERD) is a common and clinically heterogeneous disorder in which reflux symptoms, endoscopic findings, and physiologic reflux burden do not always show a simple linear relationship. ¹ In routine practice, patients with reflux symptoms may present with erosive reflux esophagitis (RE), whereas others have no visible mucosal breaks on endoscopy and are classified as having nonerosive reflux disease (NERD) in the presence of reflux-related symptoms and supportive clinical and physiologic evidence. ² Current diagnostic frameworks emphasize that GERD should be defined by converging clinical, endoscopic, and physiologic evidence, reflecting the biological complexity underlying symptom generation and mucosal injury. ³

NERD has long been recognized as more than a milder form of erosive disease. A substantial body of evidence suggests that symptom generation in NERD may be related to abnormalities in mucosal integrity, altered epithelial permeability, and enhanced sensitivity to refluxate, even in the absence of visible endoscopic injury. ⁴ By contrast, RE is more readily associated with overt epithelial damage, inflammatory injury, and tissue repair. These observations raise an important question: do NERD and RE represent distinct molecular phenotypes, or do they share a broader biological background with only limited divergence? Previous studies have suggested that nonerosive and erosive reflux disease differ in esophageal mucosal protein expression, indicating that molecular heterogeneity exists beyond conventional endoscopic classification. ⁵

However, most previous mechanistic studies have focused on local mucosal tissue rather than the circulating proteome. Blood-based biomarkers may provide a noninvasive means of capturing circulating signals related to tissue remodeling, inflammation, cellular stress, or metabolic adaptation. ⁶ In GERD, this approach may be particularly relevant for understanding why nonerosive and erosive phenotypes differ beyond symptom burden and endoscopic appearance. Existing noninvasive and tissue-based biomarkers remain informative but incomplete: salivary pepsin demonstrates only moderate diagnostic accuracy, mucosal impedance requires further standardization and validation, and histology is inherently invasive and limited by sampling variability and diagnostic specificity.^7,8 These shortcomings underscore the need for validated circulating biomarkers that can complement conventional reflux phenotyping. However, before such biomarkers can be clinically evaluated, exploratory discovery studies are needed to identify candidate circulating protein signals and generate testable hypotheses for individual-level validation.

In the present study, we performed an exploratory pooled serum proteomic analysis in patients with NERD, patients with RE, and controls without reflux symptoms or erosive esophagitis. By comparing proteomic profiles across these groups, we sought to determine whether NERD and RE exhibit common circulating molecular alterations relative to a clinically normal background and whether the two reflux phenotypes can also be distinguished by phenotype-specific proteomic signatures. Improved characterization of these shared and distinct features may enhance our understanding of the biological heterogeneity of GERD and identify candidate targets for future biomarker research.

Participants and methods

Study design and participants

This retrospective observational study enrolled patients with GERD who attended the Department of Gastroenterology, Beijing Tongren Hospital, between January 2017 and August 2019 and had stored serum samples available for analysis. Eligible participants were identified from hospital medical records and the institutional serum archive. Consecutive eligible patients who met the inclusion criteria were included. In total, 44 patients with GERD were enrolled. All had a gastroesophageal reflux disease questionnaire (GERD-Q) score ≥8 and completed upper gastrointestinal endoscopy, high-resolution esophageal manometry (HRM), and 24-h multichannel intraluminal impedance-pH monitoring. ⁹ Among them, 30 patients with endoscopic evidence of erosive esophagitis were assigned to the RE group, whereas 14 patients with reflux-related symptoms but no endoscopic erosive lesions were classified as the NERD group. In addition, 26 individuals without reflux-related symptoms, with a GERD-Q score <8 and no endoscopic evidence of erosive esophagitis, were included as the control group (N group).

The study protocol was approved by the Ethics Committee of Beijing Tongren Hospital, Capital Medical University, Beijing, China (Approval Number: TRECKY2019-114; Approval Date: 30 October 2019). Written informed consent was obtained from all participants. All patient data were deidentified before analysis. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki of 1975, as revised in 2024. The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline. ¹⁰

Inclusion and exclusion criteria

Eligible participants were adults aged 18 years or older who had archived serum specimens available and had undergone upper gastrointestinal endoscopy, HRM, and 24-h impedance-pH monitoring. Patients in the GERD cohort were required to have a GERD-Q score ≥8. Group assignment was based primarily on symptoms and endoscopic findings: patients with erosive esophagitis were categorized as having RE, whereas patients with reflux-related symptoms without endoscopic erosions were categorized as having NERD. Individuals in the N group had no reflux-related symptoms, a GERD-Q score <8, and no endoscopic evidence of erosive esophagitis. Exclusion criteria included prior upper gastrointestinal surgery, upper gastrointestinal malignancy, severe systemic illness, major esophageal motility disorders, active gastrointestinal bleeding, pregnancy, incomplete clinical data, or the lack of an adequate serum sample for analysis.

Clinical assessment

Baseline demographic and clinical information was retrieved retrospectively from the medical records, including age, sex, reflux-related symptoms, GERD-Q score, endoscopic findings, HRM results, and 24-h impedance-pH monitoring parameters.

Endoscopy

Upper gastrointestinal endoscopy was performed in all participants as part of the routine clinical evaluation using an Olympus GIF-H290 endoscope (Olympus Medical Systems, Tokyo, Japan). Patients with visible mucosal breaks were diagnosed with RE and graded according to the Los Angeles classification. ¹¹ Patients with reflux-related symptoms, a GERD-Q score ≥8, and no endoscopic erosive lesions were classified as having NERD. Control participants had no reflux-related symptoms and no endoscopic evidence of erosive esophagitis.

High-resolution esophageal manometry

HRM was performed in all patients with GERD using the Solar GI HRM system (MMS, Enschede, The Netherlands). The examination was used to assess esophageal motor function and to exclude major esophageal motility disorders. ¹² Relevant findings were reviewed retrospectively and incorporated into the overall clinical evaluation.

24-h pH-impedance monitoring

All patients with GERD underwent 24-h impedance-pH monitoring using the Solar GI system (MMS, Enschede, The Netherlands). Standard reflux parameters, including acid exposure time and total reflux episodes, were obtained from the original reports. ¹³ These findings were interpreted together with symptom assessment and endoscopic results as part of the clinical characterization of reflux disease.

Serum collection

Fasting peripheral venous blood samples were collected in the morning. After centrifugation, serum was separated, aliquoted, and stored at −80°C until proteomic analysis. Only participants with stored serum specimens suitable for analysis were included in the study.

Serum sample preparation and liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based proteomic analysis

Peripheral venous blood was collected and processed using standard procedures. After centrifugation, serum aliquots were stored at −80°C and submitted for proteomic analysis within 6–12 months of collection. Proteomic sample preparation, LC-MS/MS acquisition, database searching, and downstream bioinformatics analyses were performed by Jingjie PTM BioLab Co., Ltd (Hangzhou, China). After centrifugation at 12,000 ×g for 10 min at 4°C, supernatants were transferred to fresh tubes. The 12 most abundant serum proteins were depleted using Pierce™ Top 12 Abundant Protein Depletion Spin Columns (Thermo Fisher Scientific), and protein concentration was measured using a bicinchoninic acid (BCA) assay.

For exploratory proteomic discovery, serum samples were pooled before analysis. Five biological pools were generated for each group. In the N group, four pools contained serum from five individuals and one from six individuals. In the NERD group, four pools contained serum from three individuals and one from two individuals. In the RE group, each of the five pools contained serum from six individuals. Within each pool, equal volumes of serum from each participant were combined before protein depletion and subsequent proteomic processing.

Proteins were reduced, alkylated, digested with trypsin, desalted, labeled with tandem mass tag (TMT) reagents, fractionated by high-pH reverse-phase high-performance liquid chromatography (HPLC), and analyzed using an EASY-nLC 1000 UPLC system coupled to a Q Exactive™ Plus mass spectrometer (Thermo Fisher Scientific). Because pooled serum samples were used, the proteomic findings were interpreted as reflecting group-level molecular characteristics rather than direct individual-level biomarker evidence.

Database search and bioinformatics analysis

Data were processed using MaxQuant (version 1.5.2.8) and searched against the human UniProt database concatenated with a reverse decoy database. Trypsin/P was specified as the digestion enzyme with up to four missed cleavages allowed. Carbamidomethylation of cysteine was set as a fixed modification, whereas methionine oxidation and protein N-terminal acetylation were specified as variable modifications. The false discovery rate was controlled at <1% at the identification level.

Functional annotation was performed using the Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), protein domain, and subcellular localization databases. Enrichment analyses were conducted using two-tailed Fisher's exact tests, with corrected p values <0.05 considered statistically significant. Protein–protein interaction analysis was performed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING; version 10.1), retaining only interactions with confidence scores ≥0.7.

TMT reporter ion intensities were used for relative protein quantification according to the standard processing workflow of the proteomics facility. Quantitative protein abundance values were log2-transformed before statistical comparison, and proteins without valid quantitative values were excluded from differential expression analysis.

Differential expression analysis and reproducibility assessment

For each protein, quantitative abundance values were obtained for each pooled sample. Group-level mean abundance values were then calculated across the five pools within each group, and fold changes between groups were calculated as the ratios of group means. Relative quantitative values were log2-transformed, and between-group differences were assessed using a two-sample, two-tailed t-test. Proteins with p <0.05 and a fold change >1.3 or <1/1.3 were considered differentially expressed. No protein-level q values or multiple-testing-adjusted p values were available for differential protein screening. Therefore, the differentially expressed proteins identified in this study should be interpreted as exploratory candidate signals rather than as a definitively statistically validated biomarker panel. The nominal p value threshold was used only for exploratory candidate screening and not for confirmatory statistical inference.

Because proteomic measurements were performed on pooled serum samples rather than individual-level serum samples, formal individual-level sensitivity analyses, including leave-one-out analyses or subgroup analyses according to clinical covariates, could not be performed. Instead, reproducibility was assessed at the pooled-sample level using principal component analysis (PCA), relative standard deviation (RSD), and Pearson correlation analysis.

Statistical analysis

Statistical analyses were performed using Statistical Package for the Social Sciences (SPSS) version 26.0 (IBM Corp.; Armonk, NY, USA). Continuous variables were expressed as the mean ± SD or median (interquartile range), and categorical variables were expressed as frequencies and percentages. Comparisons among the three groups were performed using one-way analysis of variance or the Kruskal–Wallis test for continuous variables and the chi-square test or Fisher's exact test for categorical variables, as appropriate. For two-group comparisons, the independent-samples t-test or Mann–Whitney U test was used for continuous variables, as appropriate. All tests were two-sided, and p <0.05 was considered statistically significant.

Use of artificial intelligence (AI) tools

No artificial intelligence tools were used for data collection, clinical assessment, proteomic data generation, statistical analysis, or scientific interpretation of the study results. During manuscript revision, an AI-assisted language-editing tool was used only to improve readability and clarity. The authors reviewed and edited all AI-assisted text and take full responsibility for the final content of the manuscript.

Results

Study design and pooled serum proteomic workflow

The overall study design and pooled serum proteomic workflow are shown in Figure 1. A total of 70 individuals were included in the clinical cohort, comprising 26 controls, 14 patients with NERD, and 30 patients with RE. All participants underwent clinical evaluation, and patients with reflux phenotypes were further assessed by upper endoscopy, HRM, and 24-h pH-impedance monitoring. For exploratory proteomic discovery, serum samples from each group were pooled and subjected to LC-MS/MS, followed by differential expression analysis, functional annotation, and enrichment analysis.

Figure 1.

Study design and serum proteomic workflow. A total of 70 individuals were included in the study. All participants underwent clinical evaluation and upper endoscopy. The study population comprised controls (N; n = 26), patients with nonerosive reflux disease (NERD; n = 14), and patients with reflux esophagitis (RE; n = 30). Pooled serum samples from each group were subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS), followed by differential expression analysis, functional annotation, and enrichment analysis.

Baseline demographic characteristics

Baseline demographic characteristics of the study population are summarized in Table 1. The three groups were comparable in age and sex distribution. The mean age was 60.35 ± 9.44 years in the N group, 64.07 ± 9.35 years in the NERD group, and 59.77 ± 11.50 years in the RE group, with no significant between-group difference (p = 0.424). Likewise, the proportion of male participants did not differ significantly among the groups (26.9%, 21.4%, and 36.7%, respectively; p = 0.537). These findings indicate that the study groups were broadly comparable with respect to baseline demographic characteristics.

Table 1.

Baseline demographic characteristics of the study population.

Variable	N group (n = 26)	NERD group (n = 14)	RE group (n = 30)	p value
Age, years	60.35 ± 9.44	64.07 ± 9.35	59.77 ± 11.50	0.424
Male sex, n (%)	7 (26.9)	3 (21.4)	11 (36.7)	0.537

Baseline demographic characteristics of the study population. Data are presented as mean ± SD or number (%), as appropriate. p values were calculated using one-way analysis of variance for continuous variables and the chi-square test or Fisher's exact test for categorical variables, as appropriate.

N: controls; NERD: nonerosive reflux disease; RE: reflux esophagitis.

Clinical and esophageal functional characteristics of patients with NERD and RE

Clinical and esophageal functional characteristics of patients with NERD and RE are shown in Table 2. GERD-Q score, disease duration, and body mass index were similar between the two groups (all p > 0.05).

Table 2.

Clinical and esophageal functional characteristics of patients with NERD and RE.

Variable	NERD group (n = 14)	RE group (n = 30)	p value
Clinical characteristics
GERD-Q score	11.5 (8.0 to 15.5)	9.0 (8.0 to 11.25)	0.084
Disease duration, years	6.5 (1.75 to 10.0)	5.0 (3.0 to 10.0)	0.909
Body mass index, kg/m2	23.56 ± 3.33	24.31 ± 4.17	0.557
24-h pH-impedance monitoring
DeMeester score	6.21 (1.90 to 20.96)	7.74 (2.06 to 14.08)	0.930
Acid exposure time (AET), %	1.5 (0.4 to 11.18)	2.5 (0.6 to 5.0)	0.715
Upright AET, %	3.0 (0.43 to 6.25)	3.6 (0.85 to 5.1)	0.836
Supine AET, %	0.45 (0 to 1.73)	0.3 (0 to 1.9)	0.594
Total reflux episodes	9.4 (4.75 to 39.05)	29.3 (8.4 to 40.05)	0.194
Longest reflux episode, min	3.85 (2.6 to 16.03)	4.2 (2.25 to 8.1)	0.724
Acid reflux episodes	2.85 (0 to 12.55)	5.5 (0.55 to 7.6)	1.000
Weakly acidic reflux episodes	26.3 (9.25 to 44.8)	28.5 (9.1 to 49.8)	0.826
Gas reflux episodes	50.6 (40.1 to 61.0)	75.7 (58.45 to 111.0)	0.012
Positive symptom association, n (%)	4 (28.6)	7 (23.3)	0.709
High-resolution manometry
No esophageal motility disorder, n (%)	11 (78.6)	14 (46.7)	0.047
Lower esophageal sphincter pressure (LESP), mmHg	7.31 ± 5.47	8.91 ± 5.77	0.388
Integrated relaxation pressure (IRP-4 s), mmHg	1.15 (−0.65 to 3.33)	1.75 (−0.13 to 3.93)	0.520
Distal contractile integral (DCI), mmHg·s·cm	1185.5 (473.0 to 1879.5)	342.5 (74.25 to 871.25)	0.010
Peristaltic break, cm	1.45 (0 to 7.03)	4.9 (1.68 to 9.75)	0.062

Clinical and esophageal functional characteristics of patients with NERD and RE. Data are presented as mean ± SD, median (interquartile range), or number (%), as appropriate. p values were calculated using the independent-samples t-test or Mann–Whitney U test for continuous variables and the chi-square test or Fisher's exact test for categorical variables, as appropriate.

GERD-Q: gastroesophageal reflux disease questionnaire; IRP: integrated relaxation pressure; NERD: nonerosive reflux disease; RE: reflux esophagitis.

Most parameters derived from 24-h pH-impedance monitoring were also comparable between NERD and RE, including DeMeester score, acid exposure time (AET), upright AET, supine AET, total reflux episodes, longest reflux episode, acid reflux episodes, weakly acidic reflux episodes, and positive symptom association (all p > 0.05). However, patients with RE had significantly more gas reflux episodes than those with NERD (75.7 (58.45 to 111.0) vs. 50.6 (40.1 to 61.0), p = 0.012).

HRM findings further demonstrated functional differences between the two reflux phenotypes. The proportion of patients without esophageal motility disorders was significantly lower in the RE group than in the NERD group (46.7% vs. 78.6%, p = 0.047). In addition, distal contractile integral (DCI) was significantly lower in RE than in NERD (342.5 (74.25 to 871.25) vs. 1185.5 (473.0 to 1879.5) mmHg·s·cm, p = 0.010). Lower esophageal sphincter pressure, integrated relaxation pressure, and peristaltic break did not differ significantly between the two groups.

Overall, these findings indicate that NERD and RE were similar in symptom burden and most conventional reflux metrics. In contrast, RE was characterized by more frequent gas reflux and more impaired esophageal motor function.

Global characteristics of the serum proteomic dataset

The global characteristics and quality assessment of the serum proteomic dataset are presented in Figure 2. In total, 872 proteins were identified, of which 666 contained quantitative information. Overall, the proteomic workflow showed acceptable performance. RSD distributions were broadly comparable among the groups, indicating reasonable analytical reproducibility. Pearson correlation analysis showed moderate within-group consistency, whereas PCA suggested only partial separation among the three groups, with evident overlap between NERD and RE and broader dispersion in the N group. These findings suggest that serum proteomic differences existed at the group level but were not sharply dichotomous, particularly between the two reflux phenotypes. Accordingly, the proteomic differences observed in this study should be interpreted as group-level exploratory patterns rather than robust diagnostic separation among individual patients.

Figure 2.

Global characteristics and quality assessment of the serum proteomic dataset. (a) Pearson correlation heatmap of protein quantification between paired pooled serum samples, illustrating the pairwise quantitative similarity among samples from the N, NERD, and RE groups. (b) Two-dimensional principal component analysis (PCA) scatter plot of protein quantification across repeated pooled samples, showing the overall distribution and partial separation of the three groups. Each point represents one pooled sample, and ellipses indicate the sample distribution within each group. (c) Boxplot of relative standard deviation (RSD) values for protein quantification across repeated pooled samples in each group, used to assess quantitative reproducibility. (d) Basic statistical summary of the mass spectrometry results, including total spectra, matched spectra, peptides, unique peptides, identified proteins, and quantified proteins. Together, these analyses provide an overview of the quality, reproducibility, and depth of the serum proteomic dataset.

N: controls; NERD: nonerosive reflux disease; RE: reflux esophagitis.

To balance biological interpretability and robustness in this exploratory pooled design, differentially expressed proteins were primarily defined using a fold change threshold of 1.3 together with a nominal p <0.05. Under this cutoff, 48 differentially expressed proteins were identified in the NERD versus N comparison, including 35 upregulated and 13 downregulated proteins. In the RE versus N comparison, 43 proteins were differentially expressed, including 32 upregulated and 11 downregulated proteins. By contrast, only seven proteins differed between NERD and RE, including three upregulated and four downregulated proteins. These findings indicate that both reflux phenotypes showed detectable circulating proteomic alterations relative to the N group, whereas direct divergence between NERD and RE was limited.

The distribution of differentially expressed proteins across the three pairwise comparisons is shown in Figure 3. Both NERD and RE displayed broader proteomic deviation from the N group than from each other. This overall pattern supports substantial molecular overlap between the two reflux phenotypes at the circulating level, despite detectable phenotype-related differences.

Figure 3.

Differentially expressed proteins in pairwise comparisons among reflux phenotypes. (a to c) Volcano plots of differentially expressed proteins in the (a) NERD versus N comparison, (b) RE versus N comparison, (c) and NERD versus RE comparisons. The x-axis shows the log2-transformed fold change in protein abundance between the comparison groups, and the y-axis shows the −log10-transformed p value. Proteins meeting the predefined criteria for significant upregulation or downregulation are distinguished from proteins without significant differential expression by different plotting patterns, as indicated in figure key. (d) Summary of all differentially expressed proteins across the three pairwise comparisons. Differentially expressed proteins were defined using a fold change threshold of >1.3 or <1/1.3 together with a nominal p <0.05 based on a two-sample, two-tailed t-test after log2 transformation of relative quantitative values.

N: controls; NERD: nonerosive reflux disease; RE: reflux esophagitis.

Functional enrichment patterns of serum proteomic alterations

Functional enrichment patterns are summarized in Figure 4, demonstrating both shared and phenotype-specific biological features.

Figure 4.

Functional enrichment patterns of differentially expressed proteins across reflux phenotypes. (a and b) Directed acyclic graph (DAG) visualization of Gene Ontology (GO) biological process enrichment for differentially expressed proteins in the (a) NERD versus N comparison and (b) RE versus N comparison. Each node represents a GO category containing differentially expressed proteins. Node pattern indicates the level of enrichment significance, as specified in the figure key. Directed edges indicate hierarchical relationships between parent and child GO categories. Node size represents the degree of enrichment, expressed as fold enrichment. (c) Bubble plot showing the KEGG pathway enrichment distribution of differentially expressed proteins in the NERD versus RE comparison. (d) Bubble plot showing the protein domain enrichment distribution of differentially expressed proteins in the NERD versus N comparison. Because differentially expressed proteins were identified using fold change thresholds and nominal p values without protein-level multiple-testing correction, these enrichment results should be interpreted as exploratory functional patterns.

KEGG: Kyoto Encyclopedia of Genes and Genomes; N: controls; NERD: nonerosive reflux disease; RE: reflux esophagitis.

Proteomic characteristics of NERD relative to N comparison. Compared with the N group, NERD exhibited a proteomic pattern mainly characterized by structural regulation and adaptive stress-associated changes. Functional classification showed that the altered proteins were mainly involved in cellular processes, biological regulation, and responses to stimuli. The predominant molecular functions were binding, catalytic activity, and molecular function regulation. Subcellular localization analysis indicated that these proteins were mainly distributed in the extracellular space, cytoplasm, and nucleus.

Enrichment analysis further clarified the biological profile of NERD. Protein domain enrichment highlighted LIM-type zinc finger domains, represented by proteins such as ZYX, PDLIM1, and CSRP1, suggesting involvement in cytoskeletal anchoring and adhesion-associated structural regulation. In parallel, enrichment of thioredoxin-like fold domains, represented by TXN, SH3BGRL, and SH3BGRL3, indicated participation in redox-responsive and stress-adaptive processes. In addition, enrichment of helicase-related domains, including proteins such as DICER1 and DHX29, suggested that post-transcriptional or RNA-processing-related regulation may also contribute to the NERD-associated serum profile.

On the downregulated side, NERD showed enrichment of immunoglobulin-related domains, including proteins such as ICOSLG, IGHM, and IGHV3-30-5, indicating that the circulating profile in NERD was not simply characterized by a uniform elevation of immune-related proteins. Rather, the overall pattern suggested selective rearrangement of immune-associated signals together with cytoskeletal remodeling and stress adaptation.

Proteomic characteristics of RE relative to N comparison. The RE versus N comparison yielded a more structured and functionally concentrated proteomic profile than that observed in NERD. Functional classification again indicated enrichment of extracellular and cytoplasmic proteins, but pathway-level signals were more clearly organized. Among the upregulated proteins, GO enrichment indicated significant overrepresentation of the cytoskeleton, actin cytoskeleton organization, actin filament organization, focal adhesion, adherens junction, cell-substrate junction, and related molecular functions such as actin binding and cytoskeletal protein binding. These findings consistently pointed to extensive structural remodeling and altered cell-matrix or cell-junction interactions.

In addition to this structural axis, upregulated proteins in RE were enriched in platelet activation and blood coagulation-related processes. This pattern was supported by proteins such as FGA, GP6, and SERPINA1/SERPINA5, suggesting that the circulating profile of RE carried signatures related to injury-associated defense or hemostatic responses.

A second major feature of RE was the coordinated reduction of lipid-associated proteins. Downregulated enrichment terms were strongly concentrated in protein-lipid complex, lipoprotein particle, plasma lipoprotein particle, lipid transport, lipoprotein metabolic process, and cholesterol metabolism. This pattern was supported by reduced expression of APOC2, APOC3, APOC4, and LPA. Domain enrichment further aligned with this interpretation, highlighting LIM-type zinc finger, calponin homology, thioredoxin-like fold, and FERM/acyl-CoA-binding-related domains, thereby linking the RE profile to both actin-associated remodeling and altered lipid-handling processes.

Overall, the RE-associated serum proteome was characterized by coordinated structural remodeling, injury-defense responses, and alterations in lipid homeostasis.

Limited serum-level divergence between NERD and RE

In contrast to the comparisons with the N group, the comparison between NERD and RE identified only seven differentially expressed proteins. Functional enrichment was sparse and mainly limited to nucleus-related categories, DNA binding, and cellular macromolecule biosynthetic processes. KEGG signals were weak and largely driven by isolated proteins, without stable pathway-level clustering, and no significantly enriched protein domains were identified.

Taken together, these results indicate that the circulating proteomic distinction between NERD and RE was modest and not dominated by broad coordinated pathway-level remodeling. Instead, the residual differences between the two reflux phenotypes appeared to be restricted to a small number of regulatory proteins. Thus, the serum proteomic data support substantial overlap between NERD and RE, with only limited subtype-specific divergence detectable at the circulating level.

Integrated interpretation of clinical and proteomic findings

An integrated view of the key clinical differences between NERD and RE is presented in Figure 5. Notably, the molecular differences observed between these two phenotypes were not accompanied by broad differences in conventional reflux burden. Instead, the main clinical distinctions between the two groups were more frequent gas reflux, a lower proportion of patients without esophageal motility disorders, and lower DCI in RE. Together with the enrichment results, these findings suggest that the proteomic divergence between NERD and RE is unlikely to be explained solely by acid exposure and may instead reflect phenotype-specific differences in esophageal motor function, clearance capacity, and mucosal response patterns. To provide an integrated overview of these findings, Figure 6 summarizes the shared and phenotype-specific proteomic alterations together with the key clinical distinctions between NERD and RE.

Figure 5.

Key clinical and esophageal functional distinctions between NERD and RE. (a) Gas reflux episodes were more frequent in patients with RE than in those with NERD. (B) Distal contractile integral (DCI) was lower in the RE group than in the NERD group, indicating reduced esophageal contractile vigor. (c) The proportion of patients without esophageal motility disorders was lower in the RE group than in the NERD group, indicating more frequent motility abnormalities in RE. Continuous variables are presented as the median with interquartile range, and categorical variables are presented as percentages. p values were calculated using the Mann–Whitney U test for continuous variables and the chi-square test or Fisher's exact test for categorical variables, as appropriate.

NERD: nonerosive reflux disease; RE: reflux esophagitis.

Figure 6.

Integrated overview of shared and distinct clinical-proteomic patterns in NERD and RE. (a) Venn diagram showing the overlap of differentially expressed proteins identified in the NERD versus N and RE versus N comparisons. A total of 48 proteins were altered in the NERD versus N comparison, 43 proteins were altered in the RE versus N comparison, and 17 proteins were shared between the two comparisons. (b) Schematic summary of the NERD-associated proteomic signature. Representative proteins, including ZYX, PDLIM1, TXN, SH3BGRL3, FKBP1A, ICOSLG, IGHM, and IGHV3-30-5, were associated with cytoskeletal remodeling, adhesion regulation, redox adaptation, and selective immune rearrangement. (c) Schematic summary of the RE-associated proteomic signature. Representative proteins, including SERPINA1, GP6, FGA, APOC2, APOC3, and LPA, were associated with actin remodeling, injury-defense responses, and alterations in lipid homeostasis. (d) Integrated clinical-proteomic interpretation of the two reflux phenotypes. NERD was associated with higher DCI, fewer gas reflux episodes, and fewer motility disorders, whereas RE was associated with lower DCI, more gas reflux episodes, and more motility disorders. Together, these findings suggest distinct host-response patterns despite broadly similar reflux burden. This figure is intended as an exploratory conceptual summary of the observed clinical and proteomic patterns and does not imply causal relationships or validated biological mechanisms.

DCI: distal contractile integral; N: controls; NERD: nonerosive reflux disease; RE: reflux esophagitis.

Discussion

In this exploratory serum proteomic study, we found that NERD and RE exhibited distinct but partially overlapping circulating protein profiles despite broadly similar symptom burden and conventional reflux-monitoring results. Although RE is an endoscopically visible phenotype of GERD, increasing evidence suggests that NERD and RE may represent partially distinct disease phenotypes rather than merely sequential stages of the same process. ¹⁴ In our cohort, the distinction between NERD and RE could not be fully explained by greater acid exposure alone but appeared to involve differences in esophageal motor function, clearance-related physiology, and phenotype-enriched biological responses.

Consistent with this interpretation, the two reflux phenotypes showed no significant differences in GERD-Q score, disease duration, body mass index, DeMeester score, AET, acid reflux episodes, or symptom association. By contrast, patients with RE had more gas reflux episodes, a lower proportion without esophageal motility disorders, and a lower DCI. Although these findings should be interpreted cautiously, given the modest sample size and single-center design, they raise the possibility that the erosive phenotype may be associated, at least in part, with impaired motor function and reduced clearance capacity rather than with a simple increase in acid burden alone.

The proteomic findings further support this view. PCA showed only partial separation among the groups, with evident overlap between NERD and RE, and differential expression analysis identified substantially more altered proteins in comparisons with controls than in the direct comparison between the two reflux phenotypes. Thus, NERD and RE shared a broad background of circulating proteomic alterations relative to controls and showed only limited serum-level divergence from each other. This pattern is compatible with partial biological overlap within GERD, accompanied by detectable but not sharply dichotomous phenotype-related differences.

Compared with controls, the enrichment pattern in NERD may reflect a response oriented more toward barrier adaptation than overt tissue-destructive injury. Current models of GERD pathogenesis increasingly emphasize impaired epithelial barrier function, altered intercellular junctions, and cytoskeletal reorganization as early events in reflux-induced mucosal damage rather than a purely chemical burn effect. Tight junctions, adherens junctions, desmosomes, and their linkage to the cortical actin cytoskeleton are central to epithelial integrity, and abnormalities in these structures have been documented even in endoscopy-negative reflux disease. ¹⁵ Within this framework, the predominance of pathways related to cytoskeletal remodeling and adhesion-associated regulation in NERD may indicate ongoing maintenance or restoration of epithelial architecture under reflux stress.^16,17 The concurrent enrichment of redox-related functions is also biologically plausible, given the recognized role of oxidative stress and antioxidant responses before gross erosive injury becomes evident. ¹⁸ The observed enrichment pattern in NERD may be compatible with a relatively adaptive or regulatory response, although this interpretation remains speculative and requires experimental validation.

By contrast, the enrichment profile in RE was more compatible with structural injury and injury-response remodeling. Previous proteomic studies comparing nonerosive and erosive reflux phenotypes similarly suggested that these forms differ at the molecular level, with differences involving stress-response and epithelial repair-related proteins rather than simply reflecting a linear increase in severity. ⁵ Recent tissue proteomic studies have shown that different forms of esophageal inflammation may display distinct molecular signatures and functional networks despite partially overlapping histologic appearances. In GERD-related tissue, such alterations have been linked to inflammatory cell recruitment, oxidative stress-associated processes, extracellular matrix remodeling, and peroxisomal dysfunction, suggesting that reflux-related injury is shaped by broader host-response programs rather than acid exposure alone. ¹⁹ In this context, the actin-associated remodeling observed in RE is consistent with repeated junctional disruption, epithelial restitution, ¹⁷ and inflammatory remodeling. In addition, enrichment of injury-defense processes is consistent with evidence implicating nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB)-associated inflammatory signaling, cytokine activation, and oxidative injury in reflux esophagitis.^20,21 The signal for lipid homeostasis-related alterations in RE is also notable, as recent proteomic and experimental data have linked reflux esophagitis to ACSL4 activation, lipid peroxidation, and ferroptosis-related mucosal damage. ²² Together, these findings suggest that the serum enrichment profile of RE was compatible with, but does not prove, a less compensated epithelial response involving actin-dependent remodeling, defense/injury signaling, and lipid-related stress. Given the exploratory nature of this serum-based analysis and the lack of tissue-level validation in our cohort, however, these interpretations should be regarded as hypothesis-generating.

Importantly, this distinction argues against a purely linear severity model in which RE simply reflects greater exposure to the same injurious stimulus present in NERD. Instead, the relative similarity of conventional reflux metrics, together with the differences in gas reflux, motor function, and enrichment profiles, suggests that RE and NERD may differ in how the esophagus responds to reflux-related stress. In this setting, the observed association between motor dysfunction and RE should not be interpreted as evidence of causality, and further mechanistic studies are needed. In contrast, NERD may be characterized by symptom generation and regulatory biological responses in the absence of overt erosive damage.

Our findings are consistent with the concept that NERD and RE are not merely interchangeable stages of the same disorder, ²³ but are instead related yet biologically heterogeneous reflux phenotypes. This interpretation is also consistent with broader evidence that esophageal dysfunction may be shaped by systemic determinants beyond local mucosal injury alone. Accumulating evidence suggests that esophageal dysfunction in metabolic disease is multifactorial, involving hyperglycemia, autonomic neuropathy, altered hormonal homeostasis, and impaired sensorimotor regulation. In this context, adipokine-related signaling may represent another systemic mechanism linking metabolic status to epithelial vulnerability and reflux phenotype expression. Leptin, a key adipocyte-derived hormone involved in energy homeostasis and inflammatory regulation, has been implicated in obesity-associated metabolic dysfunction, oxidative stress adaptation, and altered tissue responses through pathways including Janus kinase/signal transducer and activator of transcription (JAK/STAT), phosphoinositide 3-kinase/protein kinase B (PI3 K/Akt), suppressor of cytokine signaling 3 (SOCS3), and related signaling networks. ²⁴ Although leptin signaling was not directly assessed in the present study, these observations provide a plausible biological framework in which systemic metabolic and hormonal signals may modulate mucosal susceptibility and injury-response patterns in reflux disease. Hyperglycemia has been linked to reduced lower esophageal sphincter pressure, impaired esophageal clearance, and abnormalities of distal peristaltic function. At the same time, autonomic dysfunction may further influence reflux susceptibility, gastric emptying, and symptom generation. ²⁵ In parallel, Mendelian randomization evidence has implicated circulating metabolites and metabolic pathways, including lipid-related metabolites, bile acid biosynthesis, tryptophan metabolism, and carnitine synthesis, in GERD susceptibility, further supporting a broader metabolic contribution to reflux disease pathophysiology. ²⁶ Although these data do not specifically distinguish NERD from RE, they reinforce the concept that reflux phenotypes may reflect not only reflux burden but also differences in host metabolic background, mucosal vulnerability, and injury response. A similar concept is supported by salivary proteomic data in GERD, showing that patients without erosive tooth wear had relatively higher levels of acid-resistant and antimicrobial proteins than those with erosive damage, suggesting that local protective capacity may modify how reflux-related injury is expressed. ²⁷

Another implication of our findings is that serum proteomic profiling may provide complementary information beyond conventional physiologic testing. Endoscopy, HRM, and reflux monitoring remain essential for phenotypic classification, but they do not fully capture the molecular processes underlying tissue injury or symptom generation. The identification of distinct circulating protein patterns therefore raises the possibility that blood-based molecular profiling could complement current phenotypic stratification in reflux disease, particularly when symptom presentation, endoscopic findings, and physiologic metrics are not fully concordant. ⁸ In this respect, recent morphoproteomic studies of esophagitis have also highlighted the potential value of molecular signatures in refining difficult or borderline classifications when morphology alone is insufficient. ¹⁹

Several limitations should be acknowledged. First, the most important limitation of this study is the use of pooled serum for proteomic discovery. Although pooling may reduce analytical costs and provide an initial overview of group-level proteomic patterns, it prevents assessment of interindividual variability, limits estimation of confidence intervals at the individual level, and precludes direct evaluation of the diagnostic performance of individual candidate biomarkers. Second, this was an exploratory, single-center retrospective study with a relatively small sample size, particularly for patients with NERD, which limits statistical power and generalizability. Third, pool composition was not fully balanced across the groups, particularly in the NERD group, which may have introduced additional variability into the exploratory analysis. Fourth, differential protein screening was based on fold change thresholds and nominal p values rather than protein-level multiple-testing-adjusted significance. Although protein identification was controlled using the standard database-search workflow, the absence of protein-level q values for differential expression analysis limits the statistical strength of the candidate protein list. Consequently, without multiple-testing correction, these candidate protein lists may include false-positive findings and should be interpreted as preliminary exploratory signals. Therefore, the identified differentially expressed proteins should be regarded as preliminary candidates for future validation rather than definitive biomarkers. Fifth, no independent cohort or orthogonal validation method, such as enzyme-linked immunosorbent assay (ELISA), targeted mass spectrometry, or immunoblotting, was available to confirm selected candidate proteins. Further individual-level studies with independent validation are needed to confirm these preliminary observations. Finally, because this study was cross-sectional and serum-based, causal relationships between circulating proteomic alterations and reflux phenotype cannot be established, and tissue-level validation is needed to clarify the biological relevance of these findings.

In summary, NERD and RE showed distinct but overlapping serum proteomic characteristics despite no substantial differences in most conventional reflux burden metrics. The clinical differences were related to more frequent gas reflux and greater esophageal motor impairment in RE. These findings suggest that the molecular separation between NERD and RE may reflect differences in clearance function and mucosal response patterns rather than acid exposure alone, further supporting the view that reflux phenotypes are biologically heterogeneous. To facilitate interpretation of the overall findings, we developed a conceptual summary integrating the principal clinical and proteomic observations identified in this study (Figure 6). Although exploratory in nature, this framework highlights the substantial overlap between NERD and RE and illustrates the distinct clinical and molecular patterns observed in each phenotype.

Conclusion

In conclusion, this exploratory pooled serum proteomic study provides preliminary evidence that NERD and RE may exhibit related but heterogeneous circulating proteomic patterns. Despite broadly similar conventional reflux burden, RE was associated with more frequent gas reflux, greater motor impairment, and a more injury-oriented proteomic profile, whereas NERD showed features more consistent with structural regulation and adaptive stress-response patterns. Because this study was based on pooled serum samples, nominal p value-based differential screening, and lacked an independent validation cohort, these findings should be regarded as hypothesis-generating and require confirmation in future individual-level studies with independent validation.