What this is
- This research examines gut barrier integrity in acutely ill, antipsychotic-free schizophrenia (Sz) patients.
- It focuses on two markers: () and (I-FABP).
- The study assesses their associations with immune activation and , considering confounding factors like smoking.
Essence
- Schizophrenia patients show distinct gut barrier alterations, with elevated linked to smoking and reduced I-FABP indicating epithelial injury. These findings suggest separate mechanisms of gut dysfunction in Sz.
Key takeaways
- levels were higher in schizophrenia patients (21.96 µg/mL) vs. controls (18.10 µg/mL), but this difference was not significant after adjusting for smoking. This indicates that elevation may be influenced by smoking rather than being a specific indicator of schizophrenia.
- I-FABP was lower in schizophrenia patients (218.2 pg/mL) compared to controls (315.0 pg/mL), suggesting potential gut epithelial injury independent of smoking. This reduction may reflect chronic alterations in gut integrity.
- The lack of correlation between and I-FABP highlights distinct pathophysiological processes in gut dysfunction among schizophrenia patients, suggesting that multiple markers are necessary to capture the complexity of gut health.
Caveats
- The cross-sectional design limits causal inferences about gut barrier changes and immune activation. Longitudinal studies are needed to clarify the temporal relationship between these factors.
- Differences in smoking status between schizophrenia patients and controls may confound the interpretation of findings, as smoking could influence levels.
- The study lacked direct permeability testing and comprehensive dietary or microbiome data, which are important for understanding gut health in schizophrenia.
Definitions
- lipopolysaccharide-binding protein (LBP): A protein that indicates endotoxin exposure and systemic immune activation.
- intestinal fatty acid-binding protein (I-FABP): A marker of gut epithelial damage and permeability changes.
- metabolic syndrome (MetS): A cluster of conditions increasing the risk of heart disease, stroke, and diabetes.
Simplified
Background
Schizophrenia (Sz), once viewed solely as a brain disorder, is now recognized as a systemic illness with immune and metabolic components [1 –3]. Meta-analyses have reported increased neutrophil and monocyte counts and slightly elevated C-reactive protein (CRP) in Sz [4, 5]. In our previous study of 253 acutely ill, unmedicated patients and 294 controls – excluding individuals with infections, trauma, or recent surgery – we found large and medium effect size elevations in neutrophils and monocytes, respectively, even after adjusting for stress and smoking [6]. This "sterile" myeloid activation raised the question of its underlying drivers.
One possible upstream mechanism is subclinical gut barrier dysfunction, which may allow translocation of microbial products such as lipopolysaccharide (LPS) into the circulation, triggering systemic inflammation. Supporting this, Gokulakrishnan et al. [7] found elevated gut permeability markers, including lipopolysaccharide-binding protein (LBP) and zonulin, especially in antipsychotic-naïve patients, suggesting intestinal barrier changes are intrinsic rather than drug-induced. Similarly, Weber et al. [8] reported elevated soluble CD14 (sCD14), a marker of monocyte activation, before Sz onset in U.S. military personnel, consistent with low-grade bacterial translocation and innate immune priming. LPS, a potent activator of neutrophils and monocytes [9 –12], can stimulate cytokines such as interleukin (IL)−6, which drives hepatic CRP synthesis [13].
Although human data remain scarce, animal models suggest that gut-derived inflammation can impair neurodevelopment and behavior relevant to Sz phenotypes, including social cognition [14, 15]. These effects are mediated via the gut–microbiota–brain axis, encompassing immune signaling, the vagus nerve, neuroendocrine pathways, and microbial metabolites that influence brain function and behavior [15, 16]. Consistent with this, Sz patients exhibit gut microbiota dysbiosis alongside immune-inflammatory abnormalities [17, 18].
Increased intestinal permeability may also contribute to development of metabolic syndrome (MetS)1 by enabling endotoxin-driven low-grade inflammation [19 –21], which impairs insulin signaling and promotes insulin resistance [22, 23]. While second-generation antipsychotics, particularly clozapine and olanzapine, contribute to weight gain and metabolic dysfunction [24, 25], such changes also occur in antipsychotic-naïve first-episode Sz (FESz) patients, and first-degree relatives show increased MetS risk [26 –31], suggesting intrinsic immunometabolic vulnerability. These findings suggest that immunometabolic dysfunction may precede medication effects and could reflect intrinsic vulnerability. A bidirectional relationship between MetS and gut permeability has been proposed, in which systemic inflammation and oxidative stress disrupt the mucosa, creating a self-reinforcing cycle [32 –35]. This link is particularly relevant in Sz, where both systemic inflammation and metabolic risk are elevated, even in early stages of the illness.
Studies on gut permeability markers in Sz have yielded mixed results. Some reported elevated LBP, intestinal fatty acid-binding protein (I-FABP), or zonulin [7, 36, 37], while others find no LBP differences [8, 38, 39] (Supplementary Table 1). LBP correlated with CRP [8, 39], IL-6 and tumour necrosis factor (TNF)-α [37], sCD14, [8, 39] and with body mass index (BMI) in some cohorts [37, 39]. However, leaky gut-MetS links have not been systematically studied. Variation across studies may reflect differences in medication status, disease stage, biomarker selection, control of lifestyle factors, and geographic microbiome differences [7, 8, 37, 39], highlighting the need for standardized protocols and antipsychotic-free samples.
To address these gaps, we examined gut permeability markers and their immune–metabolic relationships in unmedicated acutely ill Sz patients, including both FESz and relapsed Sz (RSz) cases, to determine whether barrier dysfunction is an early, trait-like feature or develops later in the illness course. We focused on LBP, a liver-derived marker of endotoxin exposure and systemic immune activation [40], and I-FABP, an enterocyte-derived marker of gut epithelial injury [41], as they are frequently used in Sz research and distinguish endotoxin-related inflammation from enterocyte stress [7, 8, 36 –39].
We aimed to (1) compare LBP and I-FABP levels between Sz patients and controls, and between FESz and RSz patients; (2) examine associations with innate immune markers; (3) assess links to MetS features; and (4) evaluate relationships with clinical symptom severity. This multi-dimensional approach sought to clarify the role of gut barrier integrity in Sz immunometabolic and clinical profiles, independent of medication and illness chronicity.
Methods
Patients and controls
Exclusion criteria were psychosis secondary to medical conditions, infections (respiratory, gastrointestinal, urinary tract), substance use disorders, diabetes, immune diseases, cancer, or cardiovascular disease. Screening followed German AWMF schizophrenia guidelines and included physical examination, differential blood counts, CRP, kidney function, lipids, fasting glucose, urine drug screen, MRI, and EEG. Pathological results were defined using institutional reference ranges. Symptoms were rated with the Positive and Negative Syndrome Scale (PANSS).
Two patient subgroups were analyzed (Suppl. Table 2): (1) FESz (n = 61) naïve to antipsychotics at baseline (T0), and (2) RSz (n = 35) patients with longer illness duration(median 6.0 years) but antipsychotic-free at least 6 weeks before sample collection. Healthy controls (n = 96), matched for age, sex, and smoking status (Table 1), were recruited from the community and underwent identical screening. Controls with psychiatric illness, substance use, infections, diabetes, immune disorders, cancer, or cardiovascular disease were excluded.
| Variables | median (Q1,Q3,n)Sz | median (Q1,Q3,n)Controls | Test | Test value | p-value(FDR) | (Cliff's delta)Effect size |
|---|---|---|---|---|---|---|
| Demographic data & severity of clinical symptoms | ||||||
| Age (years) | 33.0 (26.3;44.5;96) | 34.5 (27.0;45.8;96) | U-Test | W = 4471.5 | 0.724 (0.965) | −0.030 |
| Sex (female/male) | m: 56/f: 40 | m: 56/f: 40 | χ²-Test | χ² = 0.0 | 1.000 (1.000) | 0 |
| BMI (kg/m)2 | 23.56 (20.69;27.13;96) | 23.69 (21.78;27.57;96) | U-Test | W = 4426.5 | 0.638 (0.965) | −0.039 |
| Tobacco smoking (yes/no) | yes: 54/no: 42 | yes: 12/no: 84 | χ²-Test | χ² = 38.811 | < 0.001 (< 0.001) | 0.461 |
| Duration of illness (years) | 0.0 (0.0;3.3;94) | - | - | - | - | - |
| PANSS total corr. (score) | 35.00 (28.00;48.75;96) | - | - | - | - | - |
| PANSS-P corr. (score) | 12.00 (9.00;16.00;96) | - | - | - | - | - |
| PANSS-N corr. (score) | 7.00 (4.00;13.00;96) | - | - | - | - | - |
| PANSS-G corr. (score) | 15.00 (12.00;20.00;96) | - | - | - | - | - |
| Leaky gut markers | ||||||
| Plasma LBP (µg/mL) | 21.96 (16.66;29.60;95) | 18.10 (13.89;23.48;95) | U-Test | W = 5457.0 | 0.013 (0.021) | 0.209 |
| Serum I-FABP(pg/mL) | 218.2 (116.4;369.9;93) | 315.0 (174.5;533.5;94) | U-Test | W = 3518.5 | 0.021 (0.021) | −0.195 |
| Immune-related parameters | ||||||
| Neutrophils (× 10⁹/L) | 4.81 (3.53;6.48;96) | 3.01 (2.36;3.86;94) | U-Test | W = 7258.0 | < 0.001 (< 0.001) | 0.609 |
| Monocytes (× 10⁹/L) | 0.56 (0.43;0.73;96) | 0.42 (0.32;0.60;94) | U-Test | W = 6158.0 | < 0.001 (< 0.001) | 0.365 |
| CRP (mg/L) | 1.35 (0.60;3.40;95) | 0.90 (0.60;1.70;95) | U-Test | W = 5371.0 | (0.0540.023) | 0.19 |
| IL-18 (pg/mL) | 214.6 (172.3;300.3;94) | 208.3 (159.4;274.2;96) | U-Test | W = 5072.0 | 0.140 (0.245) | 0.124 |
| IL-6 (pg/mL) | 35.03 (22.68;52.99;84) | 30.25 (18.00;67.19;90) | U-Test | W = 4033.5 | 0.446 (0.446) | 0.067 |
| TNF-α (pg/mL) | 17.05 (11.04;26.97;87) | 20.50 (10.84;35.62;91) | U-Test | W = 3581.5 | 0.273 (0.319) | −0.095 |
| MCP-1 (pg/mL) | 277.4 (216.7;372.6;94) | 308.5 (221.2;407.0;95) | U-Test | W = 4026.0 | 0.244 (0.319) | −0.098 |
| Metabolic syndrome-related parameters | ||||||
| MetS (yes/no) | yes: 10/no: 86 | yes: 12/no: 84 | χ²-Test | χ² = 0.051 | 0.821 (0.902) | 0.033 |
| Waist circumference (cm) | 88.00 (80.50;97.25;94) | 90.00 (80.25;97.00;96) | U-Test | W = 4465.0 | 0.902 (0.902) | −0.010 |
| Systolic blood pressure (mmHg) | 125.5 (115.0;140.0;96) | 120.0 (110.0;127.5;96) | U-Test | W = 5799.0 | 0.002 (0.015) | 0.258 |
| Diastolic blood pressure (mmHg) | 80.0 (70.0;90.0;96) | 80.0 (70.0;80.0;96) | U-Test | W = 4491.0 | 0.754 (0.902) | −0.025 |
| Triglycerides (mmol/L) | 0.89 (0.64;1.21;96) | 1.015 (0.695;1.427;96) | U-Test | W = 3974.5 | 0.100 (0.225) | −0.137 |
| HDL cholesterol (mmol/L) | 1.43 (1.20;1.78;96) | 1.54 (1.25;1.79;96) | U-Test | W = 4073.5 | 0.165 (0.298) | −0.116 |
| Glucose (mmol/L) | 4.91 (4.49;5.53;94) | 5.01 (4.71;5.30;94) | U-Test | W = 4212.0 | 0.582 (0.873) | −0.047 |
| sRAGE (pg/mL) | 1349 (785;1946;95) | 1538 (1042;2547;96) | U-Test | W = 3667.0 | (0.059)0.019 | −0.196 |
| VEGF (pg/mL) | 232.5 (155.2;325.7;92) | 175.7 (139.9;262.1;96) | U-Test | W = 5287.0 | (0.059)0.02 | 0.197 |
Blood samples
Overnight-fasted samples were collected within 24 h of admission (8:00 a.m.). Differential white blood cell (WBC) counts were obtained from EDTA tubes within one hour after blood take collection. Serum tubes were clotted for two hours, then centrifuged at 1000 g for 10 min; plasma EDTA tubes were centrifuged immediately. Supernatants were aliquoted and stored at − 80 °C until analysis.
Assays
Serum I-FABP and plasma LBP were measured using commercial ELISAs (Hycult Biotech, Uden, Netherlands). Neutrophil and monocyte counts were obtained with an XN-3000 counter (Sysmex). CRP was measured on a Cobas 8000 c701 analyzer (Roche Diagnostics). IL-18, IL-6, TNF-α, MCP-1, soluble receptor for advanced glycation end products (sRAGE), and vascular endothelial growth factor (VEGF) were quantified with the LegendPlex Human Neuroinflammation Panel (BioLegend, San Diego, CA, USA).
Statistics
Analyses were performed in R 4.3.1, with p < 0.05 (two-tailed) considered significant. False discovery rate (FDR) correction was applied within each variable domain using the method of Benjamini-Hochberg (Table 1). Data distributions were assessed by Shapiro–Wilk tests; non-parametric tests were used as most variables were non-normal.
Group comparisons
Demographic differences in sex and smoking status were assessed using χ²-square tests. Group comparisons for continuous variables (e.g., LBP and I-FABP) were conducted using non-parametric Mann–Whitney U-tests, and included comparisons between Sz patients and controls, FESz and RSz patients, and MetS status (MetS was defined as having three or more of the following risk factors for heart disease, stroke, or type 2 diabetes: abdominal obesity, high blood pressure, high triglycerides, low HDL cholesterol, and high blood sugar). Additional comparisons of LBP and I-FABP levels by sex and smoking status were performed using Mann–Whitney U-tests. Cliff's delta (δ) was used to assess effect sizes (ǀδǀ ≥ 0.147 "small," ǀδǀ ≥ 0.330 "medium," ǀδǀ ≥ 0.474 "large").
Correlation analyses
We computed Spearman's rank correlation coefficients between LBP/I-FABP and (a) age, BMI; (b) duration of illness, PANSS scores; (c) innate immune markers; (d) MetS-related parameters (waist circumference, blood pressure, triglycerides, HDL cholesterol, glucose, sRAGE, VEGF).Analyses were run in the full sample and separately in Sz and controls. 2
Exploratory prediction models
To complement the correlation analyses, we conducted random forest regression with backward variable selection in the full sample, the Sz and control groups to identify the most robust predictors of LBP and I-FABP levels. Case weights adjusted for unequal observation counts (e.g., repeated measures or varying subject contributions). Model selection was based on highest pseudo-R2 (interpreted as: < 0.20 negligible/weak, 0.20–0.40 moderate, > 0.40 strong predictive power).
Smoking adjustment
Group differences in LBP and I-FABP were retested via Aligned Rank Transform (ART) ANOVA with smoking as covariate. A subgroup analysis of non-smokers used Mann–Whitney U-tests.
Results
Sensitivity and validity of I-FABP and LBP blood measures
Of 192 samples, 190 were within the detection range for LBP (2 below limit); 187 were within range for I-FABP (5 below limit).
Group comparisons of demographic data and severity of clinical symptoms
Patients and controls were matched for age, sex, and BMI (Table 1, Supplementary Table 2). Smoking was more frequent in patients (p < 0.001; Table 1). As expected, FESz patients had shorter illness durations than RSz patients (p < 0.001), while PANSS total scores were similar, with a nominally higher PANSS-G score in FESz (p = 0.031; not significant after FDR correction) (Supplementary Table 2).
Group comparisons in LBP and I-FABP levels
Median LBP was higher in Sz (21.96 µg/mL) vs. controls (18.10 µg/mL; p = 0.013, FDR-adjusted = 0.021, δ = 0.209), while I-FABP was lower (218.2 vs. 315.0 pg/mL; p = 0.021, FDR-adjusted = 0.021, δ = − 0.195; Table 1). No differences were found between FESz and RSz (Supplementary Table 2).
MetS status did not significantly affect LBP or I-FABP in patients (Supplementary Table 3). In controls, LBP was higher with MetS (25.78 vs. 17.24 µg/mL; FDR-adjusted < 0.001, δ = 0.610; Supplementary Table 4).
Group comparisons in immune- and metabolic syndrome-related parameters
Compared to controls, Sz patients had elevated neutrophils (FDR < 0.001, δ = 0.609) and monocytes (FDR-adjusted < 0.001, δ = 0.365), with CRP showing a trend level difference (FDR-adjusted = 0.054, δ = 0.190). No group differences were seen for IL-18, IL-6, TNF-α, or MCP-1, nor between FESz and RSz (Supplementary Table 2).
MetS prevalence was similar between groups (Table 1). Waist circumference, diastolic blood pressure, triglycerides, HDL, and glucose did not differ, but systolic blood pressure was higher in patients (FDR-adjusted = 0.015, δ = 0.258). VEGF and sRAGE showed trend-level differences (FDR-adjusted = 0.059 each). No metabolic differences were found between FESz and RSz (Supplementary Table 2).
Correlations and prediction models for LBP and I-FABP
To examine links between LBP/I-FABP and innate immune, metabolic, demographic, and clinical parameters, we used Spearman's correlations (including Mann–Whitney U-tests for sex and smoking) and random forest regression in the full sample, and separately in Sz and control groups (Supplementary Tables 5 and 6).
LBP findings
LBP was significantly associated with most immune markers. CRP and neutrophil count were top predictors in all immune-based models, explaining a moderate proportion of variance (pseudo-R2: 0.354 full sample; 0.273 Sz; 0.449 controls), consistent with strong correlations (all FDR-adjusted < 0.001). Monocytes and IL-18 correlated in the full sample and/or controls, but contributed less to prediction. IL-6, TNF-α, and MCP-1 showed no significant correlations and low importance.
MetS-associations were weaker. Waist circumference and systolic blood pressure were the most relevant predictors, with waist circumference consistently correlated with LBP (FDR-adjusted < 0.001 full sample; p = 0.017 Sz/controls) and high importance in models. Predictive power was negligible to low (pseudo-R2: 0.048–0.104).
LBP correlated with BMI (r = 0.304, p < 0.001) and smoking (p = 0.005) in the full sample; BMI was a top-ranked predictor but with low utility (pseudo-R2: 0.097 full; 0.139 controls; –0.011 Sz). Age and sex were not associated. PANSS scores and illness duration were unrelated to LBP and explained no variance in the Sz model (pseudo-R2 = –0.279).
I-FABP findings
Associations with innate immunity were weaker than for LBP. IL-18 was the only immune marker significantly correlated with I-FABP—in the full sample (r = 0.199, FDR-adjusted = 0.046) and more strongly in Sz (r = 0.336, FDR-adjusted = 0.008). In models, IL-18 predicted I-FABP in Sz (pseudo-R2 = 0.167) but did not account significantly for variance in the full sample (0.024) or controls (–0.023). Monocytes and IL-6 showed moderate to high importance in Sz, but there were no significant correlations.
Metabolic predictors did not show significant correlations with I-FABP in any group and prediction models had negligible explanatory power (pseudo-R2 range: −0.117 to 0.015).
No associations with BMI, age, sex, or smoking. Models performed poorly (pseudo-R2: 0.015 full; –0.061 Sz; –0.042 controls). Clinical parameters were also unrelated to I-FABP, with negative pseudo-R2 in Sz (–0.132).
Correlation between LBP and I-FABP
Across all groups, no significant correlations were found between LBP and I-FABP levels. In the full sample, the correlation was near zero (r = –0.017, FDR-adjusted p = 0.819), with similarly negligible and non-significant correlations observed in controls, all Sz patients, FESz, and RSz subgroups (all FDR-adjusted p-values > 0.8; Supplementary Table 7).
Testing robustness of LBP and I-FABP findings
Group differences adjusted for smoking and non-smoker subgroup analyses
After adjusting for smoking via ART ANOVA, LBP group differences were no longer significant (FDR-adjusted = 0.199), while I-FABP remained significant (FDR-adjusted = 0.033).
Similarly, focusing on the subgroup of non-smoking participants, LBP differences were non-significant (FDR-adjusted = 0.211), but I-FABP remained lower in patients (FDR-adjusted = 0.003, δ = − 0.350; Supplementary Table 8).
Correlation and prediction models for LBP in non-smokers
In non-smokers (42 Sz, 84 controls), CRP remained the strongest LBP predictor (pseudo-R2: 0.347 Sz; 0.385 controls; Supplementary Table 9). Neutrophils remained significant but less predictive; monocyte associations weakened.
Regarding metabolic predictors, waist circumference correlated with LBP in Sz (r = 0.448, p = 0.027) and was consistently a high-importance predictor. Systolic BP was a low-importance, but significant, predictor in Sz (p = 0.034). Cholesterol correlated with LBP only in controls (r = 0.333, p = 0.016).
Discussion
This is the first study to examine LBP and I-FABP as gut barrier markers in a well-characterised cohort of acutely ill, antipsychotic-free Sz patients, and to examine their relationships with innate immune activation, metabolic features, and clinical characteristics. Our findings revealed a robust reduction in I-FABP levels in patients compared to controls, independently of smoking. However, the initially observed elevation in LBP was no longer present after accounting for smoking. These results point to a selective alteration in gut barrier function in Sz, specifically affecting enterocyte integrity as reflected by the I-FABP marker, rather than endotoxin exposure as measured by LBP.
Importantly, although LBP showed moderate correlations with immune markers such as CRP and neutrophils, these relationships also appear to be driven or amplified by smoking, which was prevalent among patients. The small effect sizes for LBP and I-FABP contrast with the larger elevations in innate immune cells, suggesting that gut barrier changes, while detectable, may not be the primary driver of immune activation during acute psychosis. However, we cannot rule out the possibility that earlier, transient disruptions in gut integrity may have contributed to the immune profile seen at presentation. Longitudinal studies, ideally including ultra-high-risk and prodromal populations, are needed to determine whether gut barrier dysfunction precedes or follows immune activation in Sz.
LBP: Links to endotoxin burden and innate immune activation
Our findings align with prior reports of elevated LBP in Sz [7, 36, 37] (Supplementary Table 1), and its consistent correlations with CRP and neutrophils support its role as a gut-derived inflammation marker [40]. In contrast, Severance et al. [39] and Scheurink et al. [38] have observed no increase in LBP (Supplementary Table 1). These discrepancies likely reflect differences in study populations and unaddressed confounders – for example, variability in illness stage (first-episode vs. chronic patients), prior medication exposure, and lifestyle factors. Importantly, smoking emerged as a critical factor, as the initially higher LBP in our antipsychotic-free patient sample was entirely attributable to smoking. Once smoking status was accounted for, the case–control difference in LBP disappeared. Given that many previous studies did not control for smoking, any observed LBP elevations may have been spurious, driven by higher smoking rates in patient groups rather than by schizophrenia itself. This finding offers a unifying explanation for prior inconsistencies and underscores the importance of controlling for smoking when evaluating LBP as an indicator of gut barrier dysfunction in schizophrenia.
Smoking may increase LBP through multiple mechanisms, including pulmonary exposure to bacterial endotoxins [42, 43] and increased disruption of gut barrier integrity [44]. Given the high prevalence of smoking in Sz populations, these pathways could confound attempts to interpret LBP as a gut-specific marker of barrier dysfunction.
This suggests that LBP reflects low-grade endotoxin burden or immune system activation in a subset of patients. However, these associations must be interpreted with caution, as they may reflect shared variance with smoking or other lifestyle-related factors. Taken together, these findings highlight the importance of controlling for smoking when using LBP as a biomarker of gut permeability, and suggest that LBP is not a reliable indicator of gut-derived immune activation in Sz without proper stratification. These findings may help to clarify the inconsistencies observed in earlier research.
I-FABP: Links to enterocyte integrity and MetS
I-FABP, reflecting enterocyte injury [41, 45], was reduced in Sz and unaffected by smoking, suggesting a more stable alteration. This pattern may indicate chronic epithelial changes, e.g., mucosal atrophy from long-standing low-grade inflammation, possibly linked to shared genetic risk with inflammatory bowel disease (IBD)3 and/or dietary patterns in Sz such as a preference for sweet and high-fat foods [46 –49]. Unlike our results, Jensen et al. [36] found elevated I-FABP in medicated chronic Sz, and González-Blanco et al. [37] found no differences. These discrepancies could reflect differences in disease phase, antipsychotic exposure, and/or methodological designs (Supplementary Table 1).
Divergent gut marker profiles and implications for pathophysiology
The opposite patterns of elevated LBP and reduced I-FABP indicate that gut barrier dysfunction in Sz may reflect distinct processes such as paracellular leakiness and chronic enterocyte dysfunction, respectively [45, 50]. Their lack of correlation supports this distinction. Only LBP was consistently linked to inflammation, emphasizing the importance of using multiple markers rather than single marker approaches in capturing gut barrier complexity.
The moderate association of LBP with waist circumference and systolic BP fit with evidence that visceral adiposity and metabolic stress impair gut integrity [51 –53]. Emerging evidence supports a bidirectional relationship between MetS and gut barrier integrity [51, 53 –55].
Metabolic disturbances like hyperglycemia, which may play a role in Sz [28, 29], can compromise gut integrity. Elevated glucose levels disrupt tight junctions through GLUT2-mediated metabolic reprogramming in intestinal epithelial cells, increasing paracellular permeability [56, 57]. Moreover, systemic inflammation driven by MetS-associated cytokines such as TNF-α and IL-6 can downregulate junctional proteins and impair epithelial structure [58 –60]. Dyslipidemia and oxidative stress may further damage enterocytes, potentially contributing to reduced I-FABP levels [52 –54].
In the reverse direction, increased gut permeability facilitates translocation of bacterial products like LPS, which activate innate immune responses and promote systemic inflammation. This, in turn, can exacerbate metabolic dysregulation and perpetuate intestinal barrier dysfunction, creating a self-reinforcing inflammatory-metabolic loop [51, 55, 61].
Strengths and limitations
Key strengths include a systematically recruited, antipsychotic-free cohort, matched controls, and a dual-marker strategy distinguishing endotoxin-related inflammation from enterocyte stress (Supplementary Table 1). By employing a dual-marker approach, we were able to differentiate between distinct dimensions of gut dysfunction. Moreover, combining traditional correlation analyses with machine learning-based random forest models enhanced the robustness and interpretability of our findings.
Limitations of the study include the cross-sectional design, which precluded causal inference. We cannot determine whether gut barrier changes preceded or followed immune alterations. Furthermore, we lacked direct permeability testing (e.g., lactulose/mannitol or zonulin), dietary and microbiome data. Additionally, smoking differences between Sz and controls limit the interpretation of LBP findings.
Conclusions and future directions
We identified two dissociable gut-barrier alterations in Sz: smoking-sensitive LBP elevations linked to systemic inflammation and smoking-independent I-FABP reductions. I-FABP may be a comparatively stable marker of gut epithelial status.
Future studies should determine whether these changes are state-dependent or trait-like by tracking individuals from pre-psychotic through acute and remitted phases and, critically, by assessing unaffected first-degree relatives to evaluate LBP and I-FABP as potential endophenotypes and indicators of familial vulnerability. Longitudinal designs that integrate gut-barrier markers with microbiome and dietary assessments, functional permeability testing, and targeted interventions (e.g., probiotics, anti-inflammatory agents, dietary modification) are needed to clarify mechanisms and pave the way toward more personalized treatments for Sz.
Supplementary Information
Supplementary Material 1. Supplementary Material 2.
Gut-Brain Axis