Plasma miRNA Profiles in Chronically Treated Bipolar Disorder Patients: A Case–Control Study

Abstract

Objective

This study aimed to determine the molecular repercussions of chronic treatment and potential biomarker candidates by comparing the expression profiles of 34 selected miRNAs in peripheral plasma of patients with bipolar disorder receiving pharmacotherapy for at least one year with healthy controls.

Methods

The study included 40 patients with bipolar disorder and 40 age- and sex-matched healthy controls. The miRNA fraction was obtained from plasma samples isolated from peripheral blood. 34 target miRNAs were quantified on the Biomark Real-Time PCR Dynamic Array^TM IFC platform. Control and bipolar groups were compared based on ΔCt values.

Results

After applying multiple comparison corrections, we found that the following miRNAs significantly decreased in the bipolar group hsa-miR-222-3p, hsa-miR-574-3p, hsa-miR-145-5p, and hsa-miR-195-5p. Conversely, hsa-miR-25-3p exhibited an increase. The most notable increases were seen in hsa-miR-92a-3p, with a fold change of 1.25 (p < 0.001; q = 0.009), and hsa-miR-486-5p, with a fold change of 1.67 (p = 0.002; q = 0.033). Additionally, other miRNAs showed raw pvalues less than 0.05, but they lost statistical significance after false discovery rate correction.

Conclusion

Peripheral plasma miRNA profiles in chronic bipolar disorder revealed elevated miR-92a-3p and miR-486-5p and decreased miR-222-3p, miR-574-3p, miR-145-5p and miR-195-5p. These miRNAs may be suitable for evaluation as minimally invasive biomarker candidates in bipolar disorder, and their potential clinical use in diagnosis, prognosis, and treatment monitoring should be investigated with further studies.

INTRODUCTION

Bipolar disorder (BD) is a chronic and severe neuropsychiatric disease characterized by episodes of mania, hypomania and depression as well as irregularities in various cognitive, psychomotor, and vegetative functions such as sleep and appetite [1]. With a lifetime prevalence of 2.4% worldwide, GAD significantly reduces individual functioning and quality of life and constitutes a heavy social and economic burden for both patients and their immediate families [2]. Although the current diagnosis of BD is primarily based on clinical interviews and Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5)/International Classification of Diseases 11th Revision symptomatic criteria, the heterogeneous nature of the disease and the exacerbation-remission cycle pose severe limitations in developing early intervention and personalized treatment strategies due to the lack of objective biomarkers [3].

MicroRNAs (miRNAs) are approximately 22 nucleotides in length and are classified as non-coding RNA molecules. They play a crucial role in the regulation of the translation and stability of target mRNAs at the post-transcriptional level [4]. Their critical role in various biological processes such as immune response, neuropla-sticity, cell death, and neurodevelopmental signaling pathways has made miRNAs determinant modulators of neuropsychiatric disease pathogenesis and treatment response [5]. Peripheral circulating miRNAs are found in exosomal vesicles or RNA-binding protein complexes and exhibit high stability. They are promising biomarker candidates obtained by minimally invasive blood, plasma, or serum sampling [6].

There is rapidly increasing evidence that circulating miRNA profiles in psychiatric disorders reflect intracerebral pathological processes [7]. Expression dysregulation of many selected miRNAs has been reported in peripheral blood samples obtained from patients with BD [8,9]. The multidimensional pathogenesis of BD, shaped by the interaction of genetic, environmental, and neurobiological factors, increases the need to identify broader miRNA panels covering immune-neurological, oxidative stress, and neurodevelopmental axes [10,11]. However, existing studies in the literature need to be re-validated in large, well-defined, and longitudinally treated BB cohorts, and the sensitivity, specificity, and correlations of specific miRNA signatures with treatment response need to be systematically evaluated.

In this study, we aim to systematically examine changes in miRNA signatures after long-term treatment by comparing the expression profiles of 34 selected miRNAs in peripheral blood samples isolated from patients with BD who have been on treatment for at least one year with healthy controls. The findings will provide a basis to reveal the molecular reflections of chronic pharmacotherapy and the effects of treatment efficacy on potential biomarker profiles.

METHOD

Participants

A total of eighty individuals, comprising 40 patients and 40 controls, took part in our research. Patients diagnosed with BD were evaluated via a Structured Clinical Interview (SCID) based on DSM-5 diagnostic criteria by a psychiatrist before they participated in the study. These patients had experienced no manic or depressive episodes and had been on the same pharmacological regimen (including maintenance therapy) for at least 12 consecutive months prior to study entry. At the time of enrollment, the patients displayed no emotional symptoms and were clinically stable. The control group was formed from hospital staff who were matched for sex, age, and body mass index and who had no history of psychiatric hospitalization. Before being included in the study, the control participants also completed a SCID by DSM-5 diagnostic criteria. The criteria for the patient group required participants to be aged between 18 and 65, free from significant medical conditions, including endocrine disorders, to have a regular menstrual cycle, not to be undergoing hormone replacement therapy, and to have no comorbid psychiatric disorders. The control group was required to have no history or current diagnosis of psychiatric conditions or treatments. All participants provided written informed consent before their inclusion in the study. The ethical approval for this study was granted by Kahramanmaraş Sütçü İmam University, Faculty of Medi-cine (Approval No.: 2024/20-10).

miRNA Selection

The selection criteria for understanding the pathogenic processes associated with BD have been divided into functional categories, each reflecting a fundamental mechanical axis in the biology of the disease.

miRNAs related to neurodevelopment and synaptic plasticity

Since neurodevelopmental disorders are thought to be present in BD, miRNAs that play a role in neuronal differentiation and synaptic plasticity processes were selected. For example, miR-34a is overexpressed in brain development and targets risk genes in BD [12]. miR-92a-3p has also been shown to alter synaptic protein expression [13]. The necessity of miR-26a for long-term potentiation and dendritic spines [14]. The let-7 family of miRNAs has been reported to be important modulators of neuronal survival [15].

Inflammatory response and immune regulators

Microglia-derived exosomes release miR-146a-5p and suppress neurogenesis in stressful situations [16]. Pro-inflammatory miR-155-5p is known for its effect of increasing inflammation in the nervous system [17].

miRNAs as circulating biomarkers

miRNAs released from brain cells enter the circulation and remain stable, making them potential indicators of genetic and environmental risks. The literature reports decreased or increased levels in the blood of patients with BD. For example: miR-221, miR-223-3p, miR-23a-3p, miR-484, miR-574-3p, hsa-miR-210-3p, miR-145-5p, miR-15a-5p, miR-15b-5p [7].

Sample Collection and Plasma Preparation

Peripheral blood samples collected in tubes with EDTA were centrifuged at 4,000 rpm for 15 minutes to separate the plasma fraction. The plasma was centrifuged once more at 13,000 rpm to remove cellular debris and stored at −80°C.

RNA Isolation and cDNA Synthesis

The plasma miRNA fraction was isolated with Roche’s High Pure miRNA Isolation Kit (Roche Diagnostics) according to the manufacturer’s protocol. The isolate was stored at −80°C. Total RNA was reverse transcribed into cDNA with the miScript II RT Kit (Qiagen). The resulting cDNA was subjected to 12 cycles of preamplification using the miScript PreAmp Kit (Qiagen). Preamplified cDNA was evaluated with exonuclease and primer-dimer, and primer residues were removed.

miRNA Expression Analysis

34 miRNA levels of preamplified cDNAs (Table 1) were measured using a High Capacity Biomark Real-Time PCR system (Fluidigm) and Dynamic Array^TM IFC (Fluidigm). Primer and sample loadings were performed as recommended in the kits; thermal profile and melting-curve analysis were performed automatically through the instrument’s software.

Table 1.

Examined miRNA panel

miRNA
hsa-let-7a-5p	hsa-miR-375
hsa-miR-122-5p	hsa-miR-34a-5p
let-7b-5p	hsa-miR-423-5p
hsa-miR-26a-5p	hsa-miR-499a-5p
hsa-miR-143-3p	hsa-miR-574-3p
hsa-miR-146a-5p	hsa-miR-92a-3p
hsa-miR-150-5p	hsa-miR-324-3p
hsa-miR-155-5p	hsa-miR-210-3p
let-7b-3p	hsa-miR-145-5p
hsa-miR-373-3p	hsa-miR-15a-5p
hsa-miR-340-3p	hsa-miR-193a-5p
hsa-miR-221-3p	hsa-miR-486-5p
hsa-miR-222-3p	hsa-miR-196a-5p
hsa-miR-223-3p	hsa-miR-195-5p
hsa-miR-23a-3p	hsa-miR-15b-5p
hsa-miR-25-3p	hsa-miR-133b
hsa-miR-484	Cel-miR-39

miRNA, microRNA.

Statistical Analysis

All statistical analyses were performed using GraphPad Prism (version 10.3.1). Mean ΔCt values and their standard deviations were calculated for each miRNA, and unpaired ttests were conducted to compare control and vitiligo groups for all miRNA and oxidative/nitrosative stress parameters. A multiple unpaired ttest approach was applied for the 34 miRNAs. Raw pvalues were adjusted for false discovery rate (FDR) using the Benjamini, Krieger, and Yekutieli method to generate q‑values. For the seven miRNAs with raw p < 0.05, receiver operating characteristic (ROC) curve analyses were performed to evaluate their discriminative performance; area under the curve (AUC) values with sensitivity and specificity were deter-mined.

Demographic variables were compared between groups as potential confounders. Age was analyzed using an unpaired two‑tailed ttest. Categorical variables—sex, education level (collapsed into “Lower” [primary + secondary] vs. “Higher” [high school + college]), and marital status—were analyzed using Fisher’s exact test.

To further assess independence and potential demographic influences on the two FDR‑significant miRNAs (miR‑92a‑3p and miR‑486‑5p), we conducted subgroup analyses by treatment regimen (mood stabilizer, antipsychotic, combination) using one‑way ANOVA, examined correlations with participant age via Pearson correlation, and compared levels between males and females with unpaired two‑tailed ttests. A two‑sided p < 0.05 was considered statistically significant.

RESULTS

Participant Demographics

The mean ages of healthy controls (n = 40) and patients with BD (n = 40) were 33.1 ± 7.35 years and 34.3 ± 8.6 years, respectively, and the sex distribution was 21 females/19 males and 18 females/22 males. In the bipolar group, the mean duration of illness was 11.1 ± 6.02 years, the mean number of manic episodes was 2.98 ± 1.99, and the mean number of depressive episodes was 1.75 ± 1.33. Treatment regimens included mood stabilizers only (n = 13), antipsychotics (n = 10), or a combination of both (n = 17). All data of the participants are summarized in Table 2. Subgroup analysis of miR‑92a‑3p and miR‑486‑5p expression across these three treatment arms demonstrated no significant differences (miR‑92a‑3p, p = 0.52; miR‑486‑5p, p = 0.07) (Table 3).

Table 2.

Demographic and clinical characteristics

	Control (n = 40)	Bipolar (n = 40)	p value
Age, yr	33.1 ± 7.35	34.3 ± 8.6	0.65
Sex (female/male)	21/19	18/22	0.45
Education status			0.10
Lower education (primary + secondary school)	7	17
Higher education (high school + college/university)	33	33
Marital status (married/single)	24/16	18/22	0.26
Duration of disease, yr	-	11.1 ± 6.02	-
Number of episodes			-
Manic episode	-	2.98 ± 1.99
Depressive episode	-	1.75 ± 1.33
Treatment regimen			-
Mood stabilizer	-	13
Antipsychotics	-	10
Mood stabilizer + antipsychotics	-	17

Values are presented as number only or mean ± standard deviation.

Group differences in demographic characteristics were assessed as follows. Age was compared using an unpaired two‑tailed ttest. Categorical variables (sex, education level, and marital status) were analyzed with Fisher’s exact test after collapsing education into “Lower” (primary + secondary) versus “Higher” (high school + college). A two‑sided p < 0.05 was considered statistically significant.

-, not available.

Table 3.

Comparison of miR‑92a‑3p and miR‑486‑5p expression across treatment subgroups

miRNA	Mood stabilizer (n = 13)	Antipsychotic (n = 10)	Combination (n = 17)	p value
hsa-miR-92a-3p	0.74 ± 1.68	1.28 ± 0.99	0.67 ± 1.31	0.52
hsa-miR-486-5p	−2.19 ± 0.82	−2.47 ± 0.48	−2.81 ± 0.74	0.07

Mean ΔCt ± standard deviation values for miR‑92a‑3p and miR‑486‑5p are shown for patients receiving mood stabilizer monotherapy (n = 13), antipsychotic monotherapy (n = 10), and combination therapy (n = 17). pvalues are from one‑way ANOVA testing for differences among the three treatment groups. No significant differences were observed, indicating that miRNA levels did not vary by medication regimen.

miRNA, microRNA.

miRNA Expression Analyses

Mean ΔCt values of 34 selected miRNAs isolated from peripheral plasma were measured by the Biomark Real-Time PCR system. Group means, bipolar minus control differences, standard error of the difference, raw pvalues from the ttest, and q-values after the Benjamini-Krieger-Yekutieli FDR correction were analyzed. According to the raw pvalues, significant differences were detected in seven miRNAs (hsa-miR-222-3p, hsa-miR-25-3p, has-miR-574-3p, hsa-miR-92a-3p, hsa-miR-145-5p, hsa-miR-486-5p and hsa-miR-195-5p). However, after multiple comparison correction, only hsa-miR-92a-3p (80% reduction; fold change = 0.20; p < 0.001; q = 0.009) and hsa-miR-486-5p (40% reduction; fold change = 0.60; p = 0.002; q = 0.033) remained below the q < 0.05 threshold and maintained statistical significance. The other five miRNAs, although significant with individual pvalues, ceased to be significant after FDR correction to control the risk of Type I error (Table 4).

Table 4.

Mean ΔCt values of control and bipolar groups, difference between groups (bipolar – control), SE of the difference, raw pvalue and q-value after Benjamini, Krieger, and Yekutieli FDR correction

miRNA	Mean of control	Mean of bipolar	Difference	SE of difference	p value	q value (FDR)
hsa-let-7a-5p	0.5730	0.4373	0.1357	0.1244	0.278	0.671
hsa-miR-122-5p	1.984	1.594	0.3905	0.3119	0.214	0.649
let-7b-5p	−0.8955	−0.8502	−0.04533	0.09402	0.631	0.751
hsa-miR-26a-5p	3.734	3.666	0.06884	0.2505	0.784	0.844
hsa-miR-143-3p	5.109	4.990	0.1190	0.1942	0.542	0.722
hsa-miR-146a-5p	3.956	3.666	0.2894	0.2785	0.302	0.671
hsa-miR-150-5p	0.4330	0.6856	−0.2526	0.2338	0.283	0.671
hsa-miR-155-5p	5.109	4.990	0.1190	0.1942	0.542	0.722
let-7b-3p	5.109	4.990	0.1190	0.1942	0.542	0.722
hsa-miR-373-3p	2.804	2.599	0.2055	0.2170	0.346	0.679
hsa-miR-340-3p	5.109	4.990	0.1190	0.1942	0.542	0.722
hsa-miR-221-3p	3.784	3.619	0.1655	0.3411	0.629	0.751
hsa-miR-222-3p	4.907	4.508	0.3992	0.1927	0.042	0.198
hsa-miR-223-3p	0.2016	0.2090	−0.007351	0.3423	0.983	0.964
hsa-miR-23a-3p	1.692	1.701	−0.009851	0.3250	0.976	0.964
hsa-miR-25-3p	0.4273	0.7498	−0.3225	0.1325	0.017	0.144
hsa-miR-484	4.097	3.828	0.2692	0.1727	0.123	0.456
hsa-miR-375	4.500	4.432	0.06789	0.1820	0.710	0.816
hsa-miR-34a-5p	1.574	1.327	0.2472	0.1865	0.189	0.629
hsa-miR-423-5p	−0.1127	−0.1594	0.04670	0.1709	0.785	0.844
hsa-miR-499a-5p	5.109	4.990	0.1190	0.1942	0.542	0.722
hsa-miR-574-3p	5.109	4.733	0.3756	0.1750	0.035	0.198
hsa-miR-92a-3p	−0.1113	0.8348	−0.9460	0.2474	< 0.001	0.009
hsa-miR-324-3p	4.373	4.210	0.1632	0.1669	0.331	0.679
hsa-miR-210-3p	5.109	4.990	0.1190	0.1942	0.542	0.722
hsa-miR-145-5p	5.109	4.650	0.4590	0.1793	0.012	0.138
hsa-miR-15a-5p	5.109	4.990	0.1190	0.1942	0.542	0.722
hsa-miR-193a-5p	3.424	3.271	0.1530	0.2635	0.563	0.722
hsa-miR-486-5p	−3.036	−2.510	−0.5253	0.1639	0.002	0.033
hsa-miR-196a-5p	3.974	3.484	0.4905	0.3070	0.114	0.456
hsa-miR-195-5p	0.9173	0.6948	0.2225	0.1068	0.040	0.198
hsa-miR-15b-5p	3.604	3.581	0.02301	0.2857	0.936	0.964
hsa-miR-133b	0.5516	0.3106	0.2410	0.2156	0.267	0.671
Cel-miR-39	−2.184	−2.645	0.4611	0.5158	0.374	0.693

According to raw pvalues, significant differences were observed in seven miRNAs (hsa-miR-222-3p, hsa-miR-25-3p, hsa-miR-574-3p, hsa-miR-92a-3p, hsa-miR-145-5p, hsa-miR-486-5p and hsa-miR-195-5p); however, after multiple comparison correction, only hsa-miR-92a-3p and hsa-miR-486-5p miRNAs remained below the q < 0.05 threshold value, but only hsa-miR-92a-3p and hsa-miR-486-5p miRNAs continued to show reliable differences. Thus, the prominent changes of the five miRNAs based on individual pvalues lost significance to control the risk of Type I error with FDR correction.

miRNA, microRNA; SE, standard error; FDR, false discovery rate.

Fold change values are shown graphically in Figure 1; the red dashed line is the 1.00 reference line, with values below representing down-regulation and above representing up-regulation. These results reveal a reliable decrease in the levels of hsa-miR-92a-3p and hsa-miR-486-5p in the peripheral plasma of patients with chronic BD. These two miRNAs have been prioritized for further studies as candidates for minimally invasive biomarkers in BD. These findings must be confirmed in future larger cohort studies and functional analyses.

Fig. 1.

Fold change values of 34 miRNAs were measured in patients with bipolar disorder and healthy controls (bipolar/control).

Fold change values show the relative miRNA expression differences between the bipolar patient and control groups. The red dashed line (horizontal) indicates equal expression with a value of 1.00; values below the line indicate down-regulation, and values above the line indicate up-regulation. In particular, significant upregulation of hsa-miR-92a-3p and hsa-miR-486-5p, and significant downregulation of hsa-miR-222-3p, hsa-miR-574-3p, hsa-miR-145-5p, and hsa-miR-195-5p are highlighted.

miRNA, microRNA.

Diagnostic Performance of Candidate miRNAs

To assess the individual discriminative ability of the seven miRNAs with raw p < 0.05, ROC curve analyses were performed (Fig. 2). The AUC, optimal cut‑off value, sensitivity, and specificity for each miRNA are summarized below:

Fig. 2.

Receiver operating characteristic (ROC) curves for the seven miRNAs.

Panels A−G depict the ROC analyses for (A) hsa‑miR‑222‑3p, (B) hsa‑miR‑25‑3p, (C) hsa‑miR‑574‑3p, (D) hsa‑miR‑92a‑3p, (E) hsa‑miR‑145‑5p, (F) hsa‑miR‑486‑5p, and (G) hsa‑miR‑195‑5p. The area under the curve (AUC), and sensitivity/specificity for each miRNA are indicated on the corresponding plot, demonstrating their individual discriminative performance in distinguishing bipolar disorder patients from controls.

miRNA, microRNA.

hsa‑miR‑222‑3p: AUC = 0.638; cut‑off = 30; sensitivity 45%; specificity 85%. hsa‑miR‑25‑3p: AUC = 0.641; cut‑off = 38; sensitivity 65%; specificity 73%. hsa‑miR‑574‑3p: AUC = 0.646; cut‑off = 26; sensitivity 43%; specificity 83%. hsa‑miR‑92a‑3p: AUC = 0.739; cut‑off = 48; sensitivity 65%; specificity 83%. hsa‑miR‑145‑5p: AUC = 0.675; cut‑off = 41; sensitivity 73%; specificity 68%. hsa‑miR‑486‑5p: AUC = 0.728; cut‑off = 43; sensitivity 65%; specificity 78%. hsa‑miR‑195‑5p: AUC = 0.595; cut‑off = 45; sensitivity 45%; specificity 100%. Among these, miR‑92a‑3p and miR‑486‑5p demonstrated the highest diagnostic accuracy (AUC > 0.72), indicating their potential as individual biomarkers to distinguish bipolar patients from controls.

Association with Age

Correlation analyses between participant age and the two FDR‑significant miRNAs showed no significant associations, suggesting that age is not a confounding factor for miR‑92a‑3p or miR‑486‑5p expression (Table 5).

Table 5.

Correlation between age and miRNA expression levels

miRNA	Pearson r	95% confidence intervals	R2	p value (two‑tailed)
hsa-miR-92a-3p	0.1870	−0.1322 to 0.4711	0.03497	0.25
hsa-miR-486-5p	0.06085	−0.2555 to 0.3654	0.003703	0.71

Pearson correlation coefficients (r), 95% confidence intervals, coefficient of determination (R²), and two‑tailed pvalues are shown for the associations between participant age and ΔCt levels of miR‑92a‑3p and miR‑486‑5p. Neither correlation reached statistical significance, indicating no age‑dependent variation in these miRNA expression levels.

miRNA, microRNA.

Association with Sex

Unpaired two‑tailed ttests comparing miRNA expression between male (n = 22) and female (n = 18) participants revealed no significant sex‑based differences for miR‑92a‑3p or miR‑486‑5p (Table 6). These findings indicate that sex is unlikely to confound the diagnostic performance of these miRNA biomarkers.

Table 6.

Correlation between age and miRNA expression levels

miRNA	Male (n = 22) ΔCt ± SD	Female (n = 18) ΔCt ± SD	p value
hsa-miR-92a-3p	0.71 ± 1.45	1.01 ± 1.27	0.49
hsa-miR-486-5p	−2.46 ± 0.80	−2.60 ± 0.71	0.58

Mean ΔCt ± standard deviation (SD) values for miR‑92a‑3p and miR‑486‑5p are shown separately for male and female participants. pvalues were calculated by unpaired two‑tailed ttest to assess sex‐based differences in miRNA expression; no significant differences were observed, indicating that sex is unlikely to confound these miRNA biomarkers.

miRNA, microRNA.

Inter‑marker Correlation

Pearson correlation analysis between miR‑92a‑3p and miR‑486‑5p across all participants revealed no significant association (r = 0.06, p = 0.71), indicating minimal overlap in their expression patterns.

DISCUSSION

This study systematically examined the differences in miRNA profiles detected in the peripheral circulation of individuals with BD. Research suggests that peripheral circulating miRNAs serve as minimally invasive biomarkers for psychiatric disorders, including BD. Through a systematic review of circulating miRNA profiles in BD, candidate miRNAs, particularly miR-222-3p, miR-25-3p, miR-574-3p, miR-92a-3p, miR-145-5p, miR-486-5p and miR-195-5p, have emerged as molecules of interest for further investigation.

Among the listed miRNAs, miR-145-5p is a miRNA for which direct, case-control quantitation was achieved in human peripheral blood samples comparing BD patients with healthy controls. In a case-control study, Tekin et al. [18], found that miR-145-5p was down-regulated in the whole blood of patients with bipolar I disorder compared to healthy controls In contrast, another study by Lee et al. [19] focusing on the serum of patients with bipolar II disorder (BD-II) found no significant difference in miR-145-5p levels between BD-II patients and healthy con-trols. These discordant findings may reflect several factors, such as differences in BD subtype (I vs. II), type of sample matrix (whole blood vs. serum), or cohort cha-racteristics. Nevertheless, these data highlight the potential role of miR-145-5p as a marker and its complexity and underscore the need for further stratified studies to resolve these discrepancies. In our study, a significant decrease in miR-145-5p levels was observed in bipolar patients compared to healthy controls. This suggests that miR-145-5p is a biomarker that may give clues about the presence and severity of the disease. Mechanistically, low miR-145-5p may lead to emotion dysregulation by affecting synaptic plasticity and neuroinflammation [20]. From a psychiatric perspective, this decrease may be a dynamic indicator that can be used to monitor episode risk and treatment response.

For miR-222-3p, miR-25-3p, miR-574-3p, miR-92a-3p, miR-486-5p, and miR-195-5p, there are no clear and detailed findings on their measurement in peripheral samples from BD patients versus healthy controls in known studies. Although miRNome- or panel-wide analyses have been reported [21-24], some of which refer to large-scale profiling of circulating miRNAs, there is limited evidence for these specific miRNAs in BD versus control comparisons in the material. These studies have performed extensive profiling in whole blood and plasma but are limited, especially in case-control comparisons.

In our study, miR-222-3p levels were lower in bipolar patients compared to controls. This reflects findings in anxiety/depression models where reduced miR-223-3p is associated with neuroinflammation, and its restoration suppresses NLRP3 inflammasome activation and inflammatory cytokines [25]. These data suggest that the downregulation of specific miRNAs, such as miR-222-3p in BD, may contribute to increased neuroinflammation and that normalizing their levels may provide therapeutic benefits.

In our cohort, miR-25-3p showed a trend toward upregulation in bipolar patients relative to controls, consistent with previous findings in unmedicated manic-psychotic bipolar subjects [21], miR-25-3p modulates DNA repair, cell cycle, proliferation, and apoptosis. It has been linked to diseases such as diabetes, nephropathy, and heart failure through its effects on the renal epithelium, brain endothelium, and vascular smooth muscle [26]. Together, these data suggest that elevated miR-25-3p may reflect and potentially contribute to altered neurodevelopmental and metabolic pathways underlying the psychotic features of BD.

In our research, miR-574-3p levels were significantly lower in bipolar patients compared to controls. Although miR-574-3p is being studied for the first time in BD, it is known to act as a tumor suppressor—downregulated in early gastric and bladder cancers and chronic myeloid leukemia, where its overexpression inhibits cell proliferation, migration, and invasion—and is implicated in cardiovascular disease, suggesting important functions in cellular growth control and vascular health [27].

Our results of increased miR-92a-3p in BD are in line with upregulation research in schizophrenia, Alzheimer’s disease, and mild cognitive impairment [28,29]. miR-92a-3p is known to regulate synaptic plasticity and neurotransmission genes, and its dysregulation has been linked to mood and neuropsychiatric conditions such as post-stroke depression and autism spectrum disorders [30]. These findings suggest that miR-92a-3p overexpression may contribute to synaptic dysfunction underlying bipolar pathology.

Although miR-486-5p has not been previously studied in BD, its upregulation in our cohort parallels findings in ischemic heart disease, acute lung injury, and sepsis [31]. As a muscle-enriched plasma miRNA, miR-486-5p modulates PI3K/Akt, TGF-β, and IGF-1 pathways, affecting angiogenesis and inflammation [32,33]. Therefore, its elevated levels in bipolar patients may indicate vascular or inflammatory dysregulation in the disorder.

Moreover, our ROC curve analyses (Fig. 2) demonstrated that the two miRNAs which remained significant after FDR correction—miR‑92a‑3p and miR‑486‑5p—also exhibited the highest diagnostic accuracy (AUC = 0.739 and 0.728, respectively). This concordance between q‑value significance and superior AUC performance underscores their predictive power for BD. In contrast, despite showing raw p < 0.05, the other miRNAs yielded lower AUCs (< 0.68) and thus appear less robust as standalone biomarkers.

In our study, miR-195-5p was found to be down-regulated in bipolar patients, which may have important neuroprotective effects. Preclinical studies show that miR-195-5p protects against dendritic degeneration and neuronal death by suppressing APP, DR-6, and BACE1 in the hippocampus and cortex [34]. It is also at the center of the miRNA-transcription factor network associated with schi-zophrenia and regulates BDNF levels to modulate GABAergic markers by targeting various linked genes [29]. Therefore, reduced miR-195-5p in BD may contribute to reduced neurotrophic support and synaptic stability in affective pathology.

In conclusion, this study systematically revealed profile differences of seven candidate miRNAs in the peripheral circulation of individuals with BD; among them, the reliable increases of miR-92a-3p and miR-486-5p, and the significant decreases of miR-222-3p, miR-574-3p, miR-145-5p, and miR-195-5p highlighted these molecules as potential biomarkers involved in the pathophysiology of the disease. The findings suggest that miRNA dysregulation may affect biological axes such as synaptic function, neuroinflammation, cellular growth, and vascular pro-cesses. Future studies with larger and longitudinal cohorts examining the relationship between functional validation and treatment response will be conducted to clarify the clinical value of these miRNAs in diagnosis, prognosis, and targeted treatment strategies.

Limitations

Our study has several limitations. First, the fold changes observed, particularly for miR-92a-3p and miR-486-5p, are modest. This may limit direct biological interpretation, but suggests that they may have potential clinical value. Second, the study’s cross-sectional design allowed miRNA levels to be measured at only a single time point, which may limit our ability to assess potential changes that could occur during participants’ mood episodes. Third, although all patients had been on a stable pharmacological regimen for at least 12 months and no subgroup differences were observed, this may not entirely rule out long-term drug effects on miRNA expression, which may remain a potential confounder. Finally, while our ROC and correlation analyses indicate that miR-92a-3p and miR-486-5p capture distinct diagnostic information, the absence of a multivariate logistic regression model prevents us from definitively demonstrating each marker’s unique predictive contribution. Although our study was cross-sectional and single-center, our findings suggest that miR-92a-3p and miR-486-5p may be potential biomarkers for clinical diagnosis and targeted approaches in BD.