Could brain-derived neurotrophic factor, irisin, and peroxisome proliferator-activated receptor-gamma be a pathway in cognitive impairment in schizophrenia?

Abstract

Objective

Previous research has indicated the brain-derived neurotrophic factor (BDNF) level is lower in schizophrenia and associated with cognitive impairment. Irisin-BDNF axis may strengthen learning and memory functions. This study examined associations between BDNF, peroxisome proliferator-activated receptor-gamma (PPAR gamma) and irisin with cognitive deficits in schizophrenia.

Methods

We enrolled 80 patients with schizophrenia and 80 healthy controls (HCs). The enzyme-linked immunosorbent assay (ELISA) method was used for biochemical analysis. The Stroop Test, Trail Making Test (TMT), and Verbal Fluency Test (VFT) were used for cognitive assessment. Statistical analyses included t-tests, correlations, and analysis of covariance (ANCOVA) controlling key confounders.

Results

In unadjusted analyses, patients had significantly lower BDNF and PPARγ levels than HCs (ps < 0.001). After controlling for covariates, the difference in BDNF was still significant (F = 11804.71, p < 0.001, partial η² = 0.989), but not for PPARγ (F = 1.32, p = 0.252, partial η² = 0.010). Irisin levels did not differ between groups. Patients performed worse on all cognitive tests. After adjustments, deficits in processing speed (TMT-A, F = 4.30, p = 0.040) and executive function (Stroop-3, F = 5.89, p = 0.016; Stroop-4, F = 6.54, p = 0.012) remained significant. Longer illness duration correlated with lower BDNF (r=-0.318, p < 0.01), while more severe negative symptoms correlated with lower irisin (r=-0.244, p < 0.05). No biomarkers-cognition correlations were found groups separately. Exploratory analyses in the combined sample suggested nominal associations, but these did not survive correction for multiple comparisons and controlling for diagnostic group status.

Conclusion

BDNF demonstrates the most robust association with schizophrenia and cognitive function. The association of PPARγ with schizophrenia is confounded by demographic and metabolic factors, and irisin showed a limited link only to negative symptoms.

Clinical trial number

Not applicable.

Introduction

Cognitive deficits are core features of schizophrenia and are related to poor functional outcomes [1–3]. These impairments are most prominent in processing speed, attention, executive function, verbal fluency, and memory [4, 5]. However, the mechanisms behind cognitive dysfunction in schizophrenia remain unknown.

Brain-derived neurotrophic factor (BDNF) plays a key role in long-term potentiation in the hippocampus, a synaptic mechanism critical for learning and memory [6]. BDNF has therefore been widely studied as a candidate biomarker of cognitive functions in schizophrenia [7–9]. A growing body of evidence indicates that peripheral BDNF levels are reduced in chronic schizophrenia and first episode psychosis [10–13]. Moreover, reduced BDNF levels have been related to greater severity of psychotic symptoms and poorer performance in several cognitive domains, suggesting that impaired BDNF signaling may contribute to both illness progression and cognitive decline [7, 13].

More recently, irisin, a myokine released during physical activity, has emerged as a neuroprotective agent [14, 15]. Research in healthy individuals and neurodegenerative conditions has shown that higher circulating irisin and BNDF levels are associated with better cognitive performance [16–18]. However, findings in schizophrenia are limited and inconsistent, with studies reporting higher [19], lower [20], or similar to healthy controls (HC) irisin levels [21], and overall, only limited evidence linking irisin to cognitive impairment in this population [22].

Peroxisome proliferator-activated receptor gamma (PPARγ) is another molecule of interest due to its central role in metabolic regulation, inflammation, and mitochondrial function. Activation of PPARγ has been shown to enhance synaptic plasticity, improve long-term potentiation, and increase BDNF expression in preclinical models [23]. This positions PPAR as an upstream regulator of neurotrophic signaling, particularly in the contex of metabolic, aging and inflammatory brain conditions. Additionally, irisin which derives from the PGC-1α–dependent FNDC5 pathway, offers neuroprotective benefits and is closely linked to PGC-1α, a crucial regulator of mitochondrial function and an important transcriptional coactivator of PPARγ. This suggests a biologically plausible pathway of irisin → PGC-1α/PPARγ → BDNF pathway, linking peripheral metabolism to central neuroplasticity [24–26]. PPARγ agonists have demonstrated cognitive benefits in some neurodegenerative disease [23, 27]. Despite one report of lower PPARγ levels in patients [28], schizophrenia research is sparse, and its relationship with cognitive performance remains undefined. Given the high prevalence of metabolic dysfunction and cognitive deficits in schizophrenia, we hypothesized that PPARγ might serve as a convergent point in this pathway, and that its dysregulation could be associated with cognitive impairment.

Although BDNF, irisin, and PPARγ each represent biologically plausible contributors to neuroplasticity, no study has comprehensively examined these three biomarkers together in schizophrenia. Moreover, the extent to which they relate to cognitive impairments across multiple domains remains uncertain. The present study intended (1) to compare serum levels of BDNF, irisin, and PPARγ between patients with schizophrenia and healthy controls; (2) to examine associations between these biomarkers and clinical characteristics; and (3) to investigate their relationships with cognitive performance, including processing speed, executive function, and verbal fluency.

Methods

Study design and participants

This cross-sectional study consisted of 80 consecutive outpatients diagnosed with schizophrenia according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria from Dışkapı Yıldırım Beyazıt Training and Research Hospital outpatient psychiatry clinic between January 2018 and December 2019. The HC group consisted of 80 consecutive outpatients who visited the Family Medicine Outpatient Clinic of the same hospital. To confirm the absence of psychiatric disorders, all potential HCs were assessed using the Structured Clinical Interview for DSM-5 Disorders, Research Version (SCID-5-RV). All participants had completed at least 8 years of primary school. Exclusion criteria for all participants were conditions that could adversely affect neuropsychological test outcome, such as significant vision or hearing impairment and intellectual disability, a history of alcohol or substance abuse, electroconvulsive therapy within the past 6 months, current pregnancy or postpartum period, and any unstable systemic, metabolic, inflammatory or neurological diseases.

Initially, 95 patients with schizophrenia were screened for eligibility. Of these, 15 were excluded from the study. Reasons for exclusion included comorbid medical conditions (n = 8), history of substance use (n = 5), and insufficient educational level (n = 2).

Ethical approval was obtained from the Dışkapı Yıldırım Beyazıt Training and Research Hospital Ethics Committee (Approval number: 2017-43/28). All participants provided written informed consent prior to assessment.

Data collection

Sociodemographic and clinical data form

A structured form containing age, gender, education level, body mass index (BMI), duration of illness, number of hospitalizations, age onset of illness, treatment and other information about disease was developed and used for data collection by researchers.

Brief Psychiatric Rating Scale (BPRS)

The BPRS, a semi-structured tool, was used to assess the severity of psychiatric symptoms in patients with schizophrenia. This scale consists of 18 questions; each rated on a scale from 0 to 6. BPRS was designed to assess psychotic and depressive symptoms, mannerisms and posture, motor retardation in schizophrenia, and other psychotic disorders [29, 30]. Higher scores indicate more severe disease. The validity and reliability of the Turkish form of the BPRS have been demonstrated [31].

The Scale for the Assessment of Negative Symptoms (SANS) and the Scale for the Assessment of Positive Symptoms (SAPS)

The SANS and the SAPS are clinician-rated instruments developed to evaluate negative and positive symptom severity in schizophrenia. The SANS focuses on affective, cognitive, and social deficits, while the SAPS assesses delusions, hallucinations, thought disorder and bizarre behavior. Each symptom rated on both scales from 0 (absent) to 5 (severe) [32, 33]. Higher scores indicate more severe disease.

Calgary Depression Scale (CDS)

The CDS was developed by Addington et al. (1990) specifically to assess depression in individuals with schizophrenia, distinguishing depressive symptoms from both negative and extrapyramidal symptoms. This tool consists of a structured interview with 9 items, each scored on a 4-point scale. Higher scores indicate more severe disease [34]. A study was carried out to study the Turkish validity and reliability of the scale [35].

All participants underwent broad neurocognitive evaluation using the Trail Making Test (TMT), Stroop Test, and Verbal Fluency Test (VFT). Test were administered in a standardized order by a trained psychologist and examiners were blinded to biomarker data.

Trail Making Test (TMT)

The TMT is employed to evaluate cognitive areas such as processing speed, sequencing, mental adaptability, and visual-motor abilities. TMT-A is typically regarded as an assessment of visual search and motor speed, while TMT- B measures a higher-level executive function, including complex attention, planning, and set shifting. TMT-B is more difficult because the stimuli are placed at greater distances and there are additional visual distractions compared to TMT-A. The primary variables of interest are the total time to completion for part A and B. In addition, error and correction scores in both TMT-A and TMT-B are recorded separately [36]. The validity and reliability in Turkish population were conducted by Türkeş et al. [37].

Stroop Test: The Stroop Test is a widely used neuropsychological assessment tool designed to evaluate cognitive interference and executive function, particularly the ability to inhibit automatic responses. The Stroop Test includes three conditions: reading color names in black, identifying the color of colored rectangles, and naming the ink color of incongruent color words. The third condition requires individuals to suppress the automatic process of reading the word to correctly identify the ink color, thereby assessing cognitive control and flexibility. Performance is evaluated based on the time taken to finish each condition and the number of errors [38]. The Turkish validity and reliability of the test were demonstrated by Karakaş and Irak (1999) [39].

Verbal Fluency Test (VFT)

The VFT is a neuropsychological tool for evaluating executive functions, semanticmemory, and language. Participants are asked to generate as many words as they can within a fixed time (typically 60 s) [40]. The Turkish validity and reliability of the Verbal Fluency Test were confirmed by Tombak and Batur (2012) [41].

Blood sampling and biochemical analysis

Venous blood samples were taken from all participants after an overnight fast of at least 12 h. Serum samples were separated by centrifugation and stored at − 80 °C until analysis. The BDNF, irisin, and PPARγ serum concentrations were assessed using commercially available enzyme-linked immunosorbent assay (ELISA) kits (BioVendor, Brno, Czech Republic).

Statistical analysis

SPSS version 23.0 (IBM Corp., Armonk, NY, USA) was used to conduct data analysis. The Shapiro-Wilk test was employed to evaluate the normality of continuous variables. For variables exhibiting normal distribution, t-tests for independent samples and Pearson correlation analyses were applied. When the normality assumption was not satisfied, the Mann-Whitney U test and Spearman correlation analysis were utilized. Logarithmic transformations were performed on variables that remained non-normally distributed to facilitate the application of parametric tests. To compare serum biomarker levels between schizophrenia and control groups while addressing potential confounders, an ANCOVA model was constructed with log-transformed BDNF, irisin PPARγ. In each model, diagnostic group was entered as the fixed factor, age, sex, education level, BMI, and smoking status were included as covariates, and a group × sex interaction term was added to assess potential effect modification by sex. Cognitive test performance between groups was compared using ANCOVA models adjusted for age, sex, education level, and BMI. Effect sizes are reported as rank-biserial correlation (r) for Mann–Whitney U tests and partial eta squared (η²p) for ANCOVA models. BMI was selected as the covariate for adiposity because it is a widely used, standardized indicator of general adiposity in psychiatric research, facilitating direct comparison with prior studies. For the correlation analyses between biomarkers and cognitive test scores, p-values were adjusted for multiple comparisons using the False Discovery Rate (FDR) method (Benjamini-Hochberg procedure) to control the expected proportion of false positives. FDR-adjusted p-values (q-values) are reported alongside nominal p-values for the combined-sample analyses. All statistical tests were two-tailed, and a p-value < 0.05 was considered statistically significant. A two-tailed p-value of less than 0.05 was considered indicative of statistical significance.

Results

Clinical and demographic characteristics

BMI, waist circumference, and gender distribution were significantly different between schizophrenia and HCs (BMI: p = 0.004; waist circumference: p < 0.001; gender: p < 0.001; smoking status: p = 0.003), with patients with schizophrenia having higher BMI and waist circumference. No significant differences were found in terms of age (p = 0.324) and education level (p = 0.968) (Table 1).

Table 1.

Demographic and clinical characteristics of participants

Variables	Schizophrenia n = 80	Control n = 80	t, χ2	p
Age, years	40.1±8.5	38.5±11.9	0.99	0.324
Gender (Female)	24 (30.0)	61 (76.3)		< 0.001
Education Level, years	11.3±3.2	11.3±3.8	0.04	0.968
Waist Circumference, cm	100.6±18.7	89.8±14.9	3.72	< 0.001
BMI	29.5±6.2	26.7±5.6	2.89	0.004
Current smoker	38 (51.4)	21 (27.6)	8.84	0.003
Duration of illness, years	16.4±7.8	-		-
Number of hospitalizations	2.5 (0–25)	-		-
Number of psychotic episodes	4.0 (1–30)	-		-
Medication/treatment
Antipsychotics
Long-acting injectable (LAI)   Paliperidone(oral/LAI)   Clozapine   Aripiprazole (oral/LAI)   Olanzapine   Quetiapine   Haloperidol (oral/depot)   Amisulpride   Trifluoperazine   Risperidone (oral/LAI)   Flupentixol (oral/depot)   Antidepressants   Mood Stabilizers   Benzodiazepines   Anticholinergic agents	43 (55.1) 27 (34.6) 24 (30.8) 24 (30.8) 23 (29.5) 18 (23.1) 14 (17.9) 5 (6.4) 3 (3.8) 2 (2.6) 1 (1.3) 14 (19) 13 (16) 5 (6.4)
BPRS-Total score	8.0 (0–38)	-		-
Calgary-Total score	1.0 (0–13)	-		-
SANS-Total score	19.0 (0–74)	-		-
SAPS-Total score	6.0 (0–47)	-		-

Data presented as mean±SD, n (%), or Median (Min-Max)

Abbreviation: BMI, body mass index; BPRS, Brief Psychiatric Rating Scale; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms

Serum biomarker levels in schizophrenia and healthy controls

In unadjusted analysis, schizophrenia group exhibited significantly lower serum levels of BDNF and PPARγ compared to the HCs (BDNF: Z = 10.85, p < 0.001 r = 0.86; PPARγ: Z = 4.18, p < 0.001, r = 0.33). No significant difference was found in irisin levels between the groups (p = 0.343) (Table 2).

Table 2.

Biomarker levels of participants

	Schizophrenia	Control	Z	p	p*	Partial η²
BDNF	2280.50 (2234.25–2303.00)	3542.50 (3515.25-3564.75)	10.85	< 0.001	< 0.001	0.989
PPAR	0.52 (0.49–0.53)	0.53 (0.53–0.54)	4.18	< 0.001	0.252	0.010
Irisin	45.83 (24.04–86.43)	31.85 (19.43–98.52)	-0.89	0.343	0.821	0.000

Data presented as median (interquartile range, 25th-75th percentile)

p-values (unadjusted) are based on Mann–Whitney U tests

p* values represent ANCOVA-adjusted p-values controlling for age, sex, education level, BMI, smoking status and sex*group. Partial η² reflects the effect size of the diagnostic group in the adjusted ANCOVA model

Abbreviations: BDNF, brain-derived neurotrophic factor; PPARγ, peroxisome proliferator-activated receptor gamma

In the fully adjusted ANCOVA model, the diagnostic group effect for BDNF remained highly significant after controlling for age, sex, education level, BMI and smoking status (F = 11804.71, p < 0.001, partial η² = 0.989). (Distributional inspection of raw BDNF values revealed marked separation between diagnostic groups with minimal overlap and low within-group variance; this data structure resulted in very small residual variance and consequently large F statistics and partial η² values in the adjusted ANCOVA models). A significant group × sex interaction was also observed (F = 5.96, p = 0.016, partial η² = 0.043), indicating that sex moderated the magnitude of the diagnostic group difference. However, the main effect of sex was not statistically significant (F = 3.73, p = 0.056), and BDNF levels were consistently lower in patients with schizophrenia across sexes. None of the covariates (age, education level, BMI, or smoking status) were independently associated with BDNF levels (all p > 0.10).

For irisin, there was no main effect of diagnostic group (F = 0.05, p = 0.821, partial η² =0,00), and the group × sex interaction was not significant (F = 1.55, p = 0.215). Sex, age, education level, BMI, and smoking status were not significantly associated with irisin levels (all p > 0.23). Similarly, for PPARγ, the adjusted ANCOVA model did not reveal a significant diagnostic group effect (F = 1.32, p = 0.252, partial η² = 0.010). There was no evidence of a group × sex interaction (F = 0.33, p = 0.564). Sex, smoking status, BMI, age, and education level were not associated with PPARγ levels (all p > 0.15).

Correlations between serum biomarkers and clinical characteristics in schizophrenia

There was a significant negative correlation between illness duration and serum BDNF levels (r=–0.318, p < 0.01). Additionally, a negative correlation was observed between SANS total scores and irisin levels (r=–0.244, p < 0.05). No significant correlations were detected between PPARγ levels and clinical variables (Table 3).

Table 3.

Correlation between serum biomarker levels and demographic and clinical characteristics of schizophrenia group

	BDNF	PPARγ	Irisin
Age, years	-0.103	-0.024	-0.011
Education Level, years	0.012	-0.132	-0.023
BMI	-0.024	-0.152	-0.178
Age onset of illness, years	0.095	-0.006	-0.006
Duration of illness, years	-0.318 *	-0.040	-0.006
Number of hospitalizations	-0.030	-0.085	0.140
BPRS-Total	0.029	0.020	-0.090
Calgary-Total	0.129	0.091	-0.109
SANS-Total	0.180	-0.102	-0.244 *
SAPS-Total	0.038	0.214	-0.027

Values are correlation coefficients (r). *p < 0.05, **p < 0.01

Abbreviations: BDNF, brain-derived neurotrophic factor; BPRS, Brief Psychiatric Rating Scale; PPARγ, peroxisome proliferator-activated receptor gamma; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms

Cognitive performance in schizophrenia and healthy controls

Patients with schizophrenia demonstrated significantly poorer performance than controls across all subtests of the TMT, Stroop, and verbal fluency tests (all p < 0.05). After adjusting for age, gender, education level, and BMI, the schizophrenia group continued to show significantly worse performance on the TMT-A (F = 4.30, p = 0.040), Stroop-3 (F = 5.89, p = 0.016), and Stroop-4 (F = 6.54, p = 0.012) tests (Table 4).

Table 4.

Cognitive performance in schizophrenia and healthy controls

	Schizophrenia	Control	t	p	Adjusted F*	p*	Partial η²
TMT-A	62.8±26.7	48.4±23.0	3.59	< 0.001	4.30	0.040	0.027
TMT-B	135.7±66.6	103.3±58.3	3.22	0.002	3.64	0.059	0.023
Stroop-1	13.0±4.6	11.6±3.8	1.95	0.052	2.38	0.125	0.015
Stroop-2	13.5±4.9	11.9±3.4	2.39	0.018	1.54	0.217	0.010
Stroop-3	17.8±6.7	14.1±3.7	4.14	< 0.001	5.89	0.016	0.037
Stroop-4	25.7±10.9	18.2±5.6	5.29	< 0.001	6.54	0.012	0.041
Stroop-5	37.8±17.5	26.8±10.6	4.68	< 0.001	3.83	0.052	0.024
Verbal Fluency-K	8.8±3.8	11.4±4.8	-3.68	< 0.001	2.22	0.137	0.014
Verbal Fluency-A	6.4±3.7	8.7±3.7	-3.71	< 0.001	0.10	0.915	0.001
Verbal Fluency-S	7.2±3.8	9.1±3.7	-3.15	0.002	0.49	0.624	0.003

*Adjusted for age, gender, education level, and BMI

Relationship between clinical characteristics and cognitive performance in schizophrenia

In the schizophrenia group, age was positively correlated with slower performance across TMT-A (r = 0.313, p < 0.01), TMT-B (r = 0.322, p < 0.01), and all Stroop tasks (r values between 0.241 and 0.360, all p < 0.01).

Higher education level was associated with better cognitive performance, showing negative correlations with TMT-A (r=–0.342, p < 0.01), TMT-B (r=–0.376, p < 0.01), and Stroop tasks (r values between − 0.331 and − 0.441, all p < 0.01), and positive correlations with verbal fluency scores (VFT-K: r = 0.404, p < 0.01; VFT-A: r = 0.444, p < 0.01; VFT-S: r = 0.356, p < 0.01). Earlier age at onset was associated with poorer performance on Stroop-4 (r = 0.343, p < 0.01) and Stroop-5 (r = 0.286, p < 0.05), but not with other cognitive tests. Higher SANS scores were negatively correlated with performance on Stroop-3 (r = 0.250, p < 0.05), Stroop-4 (r = 0.368, p < 0.01), Stroop-5 (r = 0.291, p < 0.05), and verbal fluency tasks (VFT-K: r=–0.308, p < 0.01; VFT-A: r=–0.239, p < 0.05; VFT-S: r=–0.241, p < 0.05). No significant correlations were observed between cognitive test performance and illness duration, number of hospitalizations, number of psychotic episodes, BPRS, SAPS, or Calgary scores (Table 5).

Table 5.

Correlation between cognitive test and demographic and clinical characteristics of schizophrenia group

	Age	Education Level	Age Onset	Illness Duration	Number of Hospitalizations	Number of Psychotic Episodes	BPRS	CDS	SANS	SAPS
TMT-A	0.313**	-0.342**	0.148	0.045	0.005	0.119	0.090	0.105	0.180	-0.058
TMT-B	0.322 **	-0.376 **	0.099	0.035	0.151	0.215	0.049	0.162	0.200	0.011
Stroop-1	0.341 **	-0.400 **	0.099	0.178	0.141	0.157	-0.033	0.078	0.097	-0.107
Stroop-2	0.350 **	-0.386 **	0.163	0.210	0.008	0.229	0.012	0.159	0.126	-0.075
Stroop-3	0.241 **	-0.331 **	0.109	0.079	0.204	0.227	0.053	0.123	0.250 *	-0.013
Stroop-4	0.360 **	-0.441 **	0.343 **	-0.027	0.012	0.128	0.114	0.072	0.368 **	-0.008
Stroop-5	0.284 **	-0.387 **	0.286 *	-0.091	-0.042	0.088	0.034	0.131	0.291 *	-0.084
VFT- K	-0.136	0.404 **	-0.198	0.096	-0.036	-0.033	-0.107	0.015	-0.308 **	0.005
VFT- A	-0.175 *	0.444 **	-0.069	-0.008	-0.046	0.099	-0.093	0.046	-0.239 *	0.005
VFT- S	-0.099	0.356 **	-0.220	0.211	-0.032	0.028	-0.216	-0.035	-0.241 *	-0.160

Values are correlation coefficients (r). *p < 0.05, **p < 0.01

Abbreviations: BPRS, Brief Psychiatric Rating Scale; CDS, Calgary Depression Scale; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms; TMT, Trail Making Test; VFT, Verbal Fluency Test

Correlations between serum biomarkers and cognitive performance

No significant correlations were shown between serum BDNF, PPARγ, or irisin levels and cognitive test scores when schizophrenia and control groups were analyzed separately.

In the combined sample, partial correlation analyses were conducted controlling for diagnostic group status with FDR correction. BDNF showed no significant correlation with any cognitive measure. PPARγ showed one nominally significant correlation with Stroop-2 (r = 0.215, p = 0.008) that did not survive FDR correction (q = 0.24). Irisin showed one nominally significant correlation with Stroop-1 (r = 0.184, p = 0.022) that did not survive FDR correction (q = 0.33). Accordingly, no biomarker–cognition correlations remained statistically significant after adjustment for multiple comparisons.

Discussion

This cross-sectional study examined serum BDNF, irisin, and PPARγ levels in patients with schizophrenia compared to HCs and evaluated their relationships with clinical features and cognitive performance. Our primary finding is a robust reduction in BDNF in patients with schizophrenia that persists after comprehensive adjustment for age, sex, BMI, education level and smoking status confounders, including a significant moderation by sex. Although PPARγ levels were lower in unadjusted analyses, this difference did not remain significant after covariate adjustment. No significant difference was shown in irisin levels between the groups. Patients with schizophrenia showed significant impairments across all cognitive domains, but after adjusting confounders only difference in processing speed (TMT A) and executive function (Stroop 3 and 4) remained significant. Correlation analyses revealed that lower BDNF was associated with a longer illness duration, and lower irisin was linked to more severe negative symptoms in schizophrenia. Although exploratory pooled analyses indicated potential associations between biomarkers and cognition, these associations did not remain significant after accounting for diagnostic group and adjusting for multiple comparisons.

The significantly reduced BDNF levels observed in our schizophrenia group are consistent with a substantial body of evidence identifying impaired BDNF signaling as a core neurobiological feature of the disorder. Meta-analytic findings have confirmed that peripheral BDNF levels are lower in patients with schizophrenia compared to healthy controls [42]. Although some individual studies have reported increased or unchanged BDNF levels, these discrepancies are generally attributed to differences in clinical characteristics, illness chronicity, antipsychotic exposure, and methodological variability. Overall, the literature strongly supports reduced peripheral BDNF as a robust and reproducible finding in schizophrenia [43]. Importantly, in the present study, the impact of the diagnostic group on BDNF remained highly significant even after comprehensive adjustment for demographic, metabolic, and lifestyle-related confounders, underscoring that reduced BDNF is not simply a secondary consequence of metabolic burden or smoking. A significant between group and sex observed for BDNF, indicating that sex moderated the magnitude but not the direction of the diagnostic group difference. BDNF levels were consistently lower in schizophrenia across both sexes, and the main effect of sex was not statistically significant. This pattern supports quantitative modulation (magnitude differences) rather than qualitatively distinct, sex-specific disease effects, in line with prior works where sex affected peripheral BDNF variability and its associations with cognition or metabolism, but did not alter the overall reduction in BDNF related to schizophrenia [45–46,44, 45, 46]. The relationship between BDNF and symptom dimensions has been less consistent. Several studies have found no association between BDNF and general psychopathology or PANSS scores in drug-naïve or medicated patients [47], whereas others reported that lower BDNF levels are related to more severe positive, negative, or depressive symptoms [12, 13, 48, 49]. Notably, a study in unmedicated patients with schizophrenia demonstrated that serum BDNF levels were significantly associated with psychopathology as measured by the BPRS, and these associations remained significant after correction for multiple comparisons, supporting the biological relevance of BDNF alterations independent of antipsychotic exposure [50]. In the present study, BDNF was not associated with symptom severity but showed a negative correlation with illness duration, consistent with the idea of a progressive decline in neurotrophic support over the course of schizophrenia. This pattern may reflect chronic illness burden, cumulative treatment effects, or long-standing metabolic and lifestyle factors, although these mechanisms cannot be disentangled in a cross-sectional design.

BDNF is closely linked to synaptic plasticity and cognitive processes, and accumulating data suggests lower BDNF levels are associated with poorer performance in several cognitive domains, like working memory, processing speed, attention, and executive function [7, 8, 51–53]. In our sample, we did not detect significant associations between BDNF and cognition within the schizophrenia group, which may be due to limited power, restricted variability in biomarker levels among patients, or the influence of unmeasured confounders. While pooled (patient + control) analyses initially suggested biomarker–cognition correlations, these did not persist after controlling for diagnostic group and applying FDR correction, indicating that these associations were largely driven by between-group differences rather than within-group relationships. This underscores a critical interpretive caveat: combining groups that differ fundamentally on both variables can produce misleading correlations. Our data, therefore, do not support the hypothesis that peripheral BDNF levels serve as a continuous, transdiagnostic correlate of cognitive function in this cohort. Instead, the BDNF deficit appears to be a categorical feature of the schizophrenia diagnosis itself.

In contrast to BDNF, our results contribute to the heterogeneous and uncertain literature on irisin in schizophrenia. While some studies reported no significant differences in irisin levels [21], Zhang et al. (2019) found higher irisin levels in first-episode schizophrenia (P. F. Zhang et al., 2019). In contrast, Erzin et al. (2021) observed lower irisin levels in schizophrenia despite similar physical activity levels and metabolic parameters between patients and controls [20]. Contributing to this conflicting evidence, we did not find significant difference in serum irisin levels between patients with schizophrenia and HCs.

Preclinical studies have demonstrated that irisin promotes neuronal survival, reduces oxidative and inflammatory damage, and enhances hippocampal neurogenesis and improves cognitive functions primarily through upregulating BDNF expression [54–56]. This highlights the irisin–BDNF axis as a potential pathway through which exercise may improve cognition. This relationship is supported by research in healthy population. Higher irisin levels often linked to greater physical activity have been associated with better cognitive performance and higher BDNF levels, particularly in verbal fluency, episodic memory, and global cognition evaluated by mini mental state examination (MMSE) and verbal memory tests [16]. Similarly, studies in Alzheimer’s disease (AD) have adressed that cerebrospinal fluid irisin is positively correlated with both BDNF levels and MMSE performance [17]. A recent review also suggests that irisin acts as a neuroprotective exercise-induced mediator and may slow neurodegenerative processes and also improve cognitive functions [18].

Despite this strong mechanistic foundation, studies in schizophrenia have not established a clear link between irisin and cognitive functions or symptoms. Similar to previous reports that found no correlation between irisin and symptom severity [20–22] or cognitive function especially verbal memory [22], we found no association between irisin levels and cognitive performance in our cohort, either within the patient group or across the combined sample. The only notable association was a modest negative correlation between irisin and negative symptom severity. This relationship may indicate that patients with more severe negative symptoms engage in less physical activity, have poorer metabolic health, or show greater motivational deficits, all of which could affect irisin levels. However, because physical activity and detailed metabolic status were not systematically assessed, this interpretation remains speculative. Overall, our findings suggest that the role of irisin in schizophrenia is still unclear and likely strongly influenced by lifestyle and metabolic factors that are not captured by a single serum measurement.

PPARγ is a nuclear receptor involved neuroprotection, stem cells maintenance, and inflammation [57]. Activation of PPARγ enhances mitochondrial energy production, improves long-term potentiation, increases BDNF expression, and promotes hippocampal synaptic plasticity. PPARγ agonists have also reversed age-related decreases in long term potentiation, improved spatial memory, and upregulated BDNF/TrkB signaling in models of diabetes, chronic hypoperfusion, and neurodegeneration [23]. These findings support a PPARγ–BDNF regulatory pathway that may influence cognitive function [23, 58]. A Meta-analysis indicated that pioglitazone may attenuate cognitive impairment in early or mild-to-moderate AD [24]. Similarly, rosiglitazone has been shown to ameliorate hippocampal cognitive deficits in some AD cohorts, with improvements in memory and learning reported in several trials [59, 60]. The potential therapeutic implications are significant, as PPARγ agonists, such as pioglitazone, are already in clinical use for other conditions and could be repurposed for cognitive enhancement in neuropsychiatric disorders [58]. Given their ability to support synaptic plasticity and neurite outgrowth, PPAR agonists have been widely proposed as neuroprotective agents, with numerous ligands suggested for improving cognitive performance across neuropsychiatric and neurodegenerative disorders.

Although research on PPARγ in schizophrenia is limited, emerging evidence points toward possible involvement in neurobiological and cognitive processes. In a glutamate receptor hypofunction (GluN1 knockdown) model of schizophrenia, pioglitazone treatment improved long-term memory and restored key cognitive endophenotypes [61]. One of the earliest clinical studies, by Martínez-Gras et al. (2011), showed significantly lower PPARγ levels in patients with schizophrenia compared to healthy controls, suggesting impaired PPARγ signaling [28]. Additionally, several studies indicate that PPARγ agonists can upregulate BDNF expression, raising the possibility that drugs such as rosiglitazone may benefit cognitive symptoms in schizophrenia through a PPARγ–BDNF pathway [62].

Clinical findings, however, remain inconsistent. Yi et al. (2012) investigated rosiglitazone augmentation in clozapine-treated schizophrenia patients using a comprehensive neuropsychological battery and found no significant cognitive improvements after eight weeks of treatment [63]. In contrast, studies of pioglitazone have reported reductions in depressive and negative symptoms, particularly among metabolically impaired patients [64, 65]. These mixed outcomes likely reflect heterogeneity in illness chronicity, metabolic dysregulation, antipsychotic regimens, and duration of PPARγ agonist exposure.

In our study, serum PPARγ levels were lower in patients with schizophrenia than healthy controls, but after adjusting for key confounders this difference was no longer significant, suggesting a strong influence of metabolic factors. PPARγ levels did not correlate with clinical characteristics or cognitive performance in the schizophrenia group. Any nominal pooled associations with cognitive measures did not survive diagnostic-group adjustment and FDR correction, indicating limited evidence for a robust biomarker–cognition relationship. Overall, our findings suggest that peripheral PPARγ is more reflective of metabolic status than core illness processes, and its role in schizophrenia especially regarding cognition remains inconclusive.

From a clinical perspective, these findings have important implications. The robust association between BDNF and schizophrenia, even after controlling for multiple confounders, suggests that interventions aimed at increasing BDNF levels may have therapeutic potential and warrant further investigation. Both pharmacological approaches (such as antipsychotics that upregulate BDNF) and non-pharmacological strategies (including aerobic exercise, cognitive training, and lifestyle modifications) have shown promise in preclinical and clinical studies. Future studies should adopt longitudinal designs to clarify whether reduced BDNF represents a trait vulnerability, a progressive illness marker, or a consequence of chronic disease and treatment. Interventional studies targeting BDNF-related pathways through pharmacological agents, physical activity, or cognitive remediation are also needed to determine whether enhancing neurotrophic signaling can lead to meaningful cognitive improvements in schizophrenia.

The inconsistent findings regarding irisin and PPARγ suggest that these biomarkers may be more useful as indicators of metabolic health and lifestyle factors rather than as direct targets for cognitive enhancement. Nonetheless, given the established links between metabolic dysfunction and cognitive impairment in schizophrenia, interventions that improve metabolic health through physical activity and dietary modifications may indirectly benefit cognition via the irisin-BDNF and PPARγ-BDNF pathways.

The primary strength of this study is the comprehensive evaluation of BDNF, irisin, and PPARγ within same sample, allowing direct comparison of their clinical and cognitive correlates. To our knowledge, this is the first study addressing the relationship between PPARγ levels and cognitive performance in schizophrenia, contributing novel evidence to an underexplored area. Furthermore, the use of a broad cognitive assessment battery, including processing speed, attention, executive function, and verbal fluency, enabling a more detailed characterization of cognitive deficits. The inclusion of a well-matched healthy control group, along with statistical adjustment for key demographics like age, gender, education level and BMI, further strengthens the internal validity of the findings.

However, the limitations of this study must be considered. The cross-sectional design precludes conclusions about causality or longitudinal changes in biomarker levels. It is unclear if lower BDNF is a consequence or cause of illness pathophysiology, or whether it changes over time in relation to symptom fluctuations and treatment response. Longitudinal studies tracking these biomarkers alongside cognitive performance and clinical outcomes are needed to clarify temporal relationships and identify potential targets for early intervention. Second, we relied on peripheral serum measurements, which may not accurately reflect central nervous system levels or activity of these molecules. Cerebrospinal fluid analysis or advanced neuroimaging techniques (such as PET imaging with BDNF-sensitive tracers) could provide more direct insights into brain-specific processes. The lack of objective physical activity monitoring and detailed metabolic profiling is a significant limitation, particularly for interpreting the irisin and PPARγ findings. Irisin is strongly influenced by skeletal muscle activity and exercise intensity [66], while PPARγ activity is closely linked to adiposity, insulin sensitivity, and broader metabolic regulation [67]. Without accelerometry-based physical activity assessment, dietary data, or comprehensive metabolic panels (e.g., insulin resistance indices, lipid profiles), we cannot determine whether interindividual variability in these biomarkers reflects underlying illness-related mechanisms or differences in lifestyle and metabolic health. These factors are major confounders that should be systematically controlled in future research through accelerometry, dietary assessments, and comprehensive metabolic panels. Medication effects represent an important potential source of residual confounding. Although most antipsychotic treatments have been shown to increase peripheral serum BDNF levels [68, 69], heterogeneity in medication type, dose, duration, and polypharmacy may still influence biomarker levels. Moreover, our sample consisted of chronic, medicated patients with schizophrenia, and although medication use was documented in detail and the effects of antipsychotics, antidepressants, mood stabilizers, benzodiazepines, and anticholinergic agents were not modeled individually in the primary analyses. Given the degree of polypharmacy and sample size constraints, adjusting for individual medications or medication classes risked model overfitting and reduced interpretability. Therefore, residual confounding related to pharmacological effects on serum biomarkers and cognitive performance cannot be fully ruled out. Studies in first-episode, drug-naïve patients would help disentangle illness-related from medication-related effects. Finally, we addressed multiple testing in biomarker–cognition analyses using FDR correction after controlling for diagnostic group; the absence of significant associations after these steps suggests that any apparent pooled correlations should be interpreted as between-group effects and require replication in independent samples.

Author contributions

HU and EBO wrote the main manuscript. AY assisted with the biochemical evaluation, AD with the neurocognitive battery administration and evaluation. SE was responsible for the statistical evaluation. EY, SO, and GE contributed to hypothesis testing, text suitability, and final evaluations.

Funding

This research received no specific grant and financial support from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

All data related to this study are retained by the corresponding author and can be provided upon request.

Declarations

Ethical approval

Before starting this study, permission was obtained from the ethics committee of the Dışkapı Yıldırım Beyazıt Training and Research Hospital (Approval number: 2017-43/28). All stages of the study were carried out in according to provisions of the Declaration of Helsinki as revised in Edinburgh 2000. Each participant was informed about the study beforehand and informed consent was obtained from all of them.

Consent for publication

Our article does not present any clinical or personal data that could compromise the anonymity of participants.

Competing interests

The authors declare no competing interests.