ABSTRACT
Controlling for confounding factors in postmortem brain studies of psychiatric disorders is crucial, particularly in gene expression analyses. Potential confounding factors include sex, age at death, medication history, agonal state, postmortem interval (PMI), tissue storage duration, tissue pH, and RNA integrity number (RIN). pH and RIN are considered particularly important in gene expression analysis because they accurately reflect mRNA quality. We previously found that pH and RIN affected the levels of genes related to the cell cycle and RNA processing, respectively, and both affected genes involved in energy production and the immune system. In this study, we investigated the influence of confounding factors on gene expression profiles in psychiatric disorders. We measured gene expression levels in the prefrontal cortex of 25 deceased patients with schizophrenia, 10 with bipolar disorder, and 21 nonpsychiatric controls via RNA sequencing, and identified genes associated with pH, RIN, PMI, sex, age at death, tissue storage duration, and antipsychotic drug dose. We identified key molecular pathways through Ingenuity Pathway Analysis. The total number of mRNA variants significantly correlated with gene expression, and each confounding factor was as follows: 139 for sex, 87 for age, 99 for PMI, 35 386 for pH, 11 373 for RIN, and 13 414 for storage duration. We identified strong associations with metabolic, immune, energy production, and DNA repair pathways. The expression of certain genes was strongly associated with different confounding factors; therefore, it is necessary to consider these factors when interpreting the gene expression profile in brain tissue.
Keywords: bipolar disorder, confounding factor, postmortem brain, RNA‐seq, schizophrenia
This study examined the impact of confounding factors on gene expression in postmortem brain tissue from individuals with psychiatric disorders. The expression was significantly associated with factors such as pH, RIN, PMI, and storage duration, highlighting the importance of accounting for these variables when interpreting transcriptomic data.
1. Introduction
The pathophysiology of most psychiatric disorders, including schizophrenia, remains largely unclear. Various psychiatric disorders, including schizophrenia and bipolar disorder, are associated with brain dysfunction; therefore, previous studies have primarily focused on functional and structural changes in the brain [1]. Postmortem brain banks are an important resource worldwide for elucidating the pathology of psychiatric disorders, such as the molecular/genetic pathogenic factors [2, 3].
Recent improvements in molecular expression analysis techniques have enabled researchers to characterize the mRNA profiles of postmortem brain tissue [4, 5]. Pathological changes in the postmortem brains of patients with psychiatric disorders are considered less pronounced than those in neurodegenerative diseases. Therefore, it is necessary to consider the impact of various variables, such as the agonal and pre‐/postmortem states. Identifying factors associated with disease pathology from gene expression data in postmortem brains presents several challenges, particularly in assessing and controlling for factors derived from nonpathological conditions [6].
We recently reported that indicators that are difficult to quantify can be validated by demonstrating a lack of significant differences in the expression of housekeeping proteins between brain tissues stored at two different brain banks [7]. pH and the RNA integrity number (RIN) are among the most representative quantifiable indicators [8] of brain tissue degradation [9] and are generally used to objectively assess brain tissue quality. Other confounding factors include prenatal status, age at death, cause of death, postmortem brain removal, storage conditions, medical and treatment histories, sex, alcohol consumption, and smoking habits [10].
We also recently investigated the effects of pH and RIN on gene expression levels using microarray gene expression data for the postmortem brain tissues of 13 patients with schizophrenia and found that pH strongly affected genes related to the cell cycle, RIN strongly affected genes related to RNA processing, and both affected genes involved in energy production and the immune system [11]. RNA sequencing (RNA‐Seq) is advantageous for detecting low‐abundance genes with higher sensitivity and quantitative accuracy compared to microarrays, provided that sufficient sequencing depth is ensured. Regarding the dynamic range, sufficient sequencing depth allows RNA‐Seq to more accurately reflect the true differences in gene expression levels, resulting in a broader detectable range of expression compared to microarray [12, 13]. RNA‐seq has recently been widely used to analyze mRNA expression in postmortem brains. Recent studies [14, 15] have highlighted the critical role of RNA splicing in disease pathogenesis, underscoring the importance of investigating RNA variants to elucidate the pathophysiology of psychiatric disorders. However, only a few studies have examined the gene groups affected by potential confounders using RNA‐seq data. Recently, Tian et al. [16] analyzed how RIN and postmortem interval (PMI) influence individual gene expression across the transcriptome using RNA‐seq data from human brain tissue. However, to our knowledge, no studies have specifically analyzed the impact of confounding factors in postmortem brains of individuals with schizophrenia or bipolar disorder. While previous differential gene expression studies [17] have accounted for potential confounding factors such as age, RIN, and medication use by including them as covariates, they have primarily focused on the impact of individual variables—such as RNA [18] quality degradation or sex differences [19, 20]—on gene expression. However, no studies have systematically examined gene groups influenced by multiple confounders or identified subsets of confounders for case–control matching in psychiatric disorders. Therefore, further research is needed to comprehensively assess the effects of potential confounders on gene expression in postmortem brain tissue. In this study, we used comprehensive RNA‐seq gene expression data for postmortem brains to identify the relationships between important gene groups and confounding factors in nonpsychiatric, schizophrenia, and bipolar disorder cases. This study aimed to clarify the importance of confounding factors for explaining the differences in postmortem brain gene expression profiles in psychiatric disorders.
2. Methods
2.1. Collection of Postmortem Brain Tissue
Postmortem brain tissue samples from patients with schizophrenia, bipolar disorder, and healthy nonpsychiatric controls were obtained from the Tohoku Postmortem Brain Bank, DNA Bank for Psychiatric Research (https://www.fmu‐bb.jp/english/index.htm), and Brain Research Institute, Niigata University. The use of human postmortem brain tissue was approved by the ethics committees of Fukushima Medical University (approval number: 1685) and was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Patients with schizophrenia or bipolar disorder met the diagnostic criteria established by the American Psychiatric Association (Diagnostic and Statistical Manual of Mental Disorders [DSM‐IV or 5]). Individuals without a history of psychiatric disorders were included as comparison participants.
Gray matter samples corresponding to Brodmann area 10 in the prefrontal cortex were collected from 25 patients with schizophrenia, 10 with bipolar disorder, and 21 nonpsychiatric controls (Table 1).
TABLE 1.
Demographic information related to RNA‐seq data.
| Control | Schizophrenia | Bipolar disorder | |
|---|---|---|---|
| Number of samples | 21 | 25 | 10 |
| Sex | 8F, 13M | 8F, 17M | 4F, 6M |
| Age at death (±SD) | 66.4 ± 17.7 | 68.4 ± 11.3 | 59 ± 24.5 |
| PMI (±SD) (h) | 11.4 ± 12.9 | 16.8 ± 11.8 | 21.3 ± 24.5 |
| Tissue pH (±SD) | 6.4 ± 0.3 | 6.34 ± 0.4 | 6.6 ± 0.35 |
| RIN (±SD) | 5.8 ± 1.2 | 6.8 ± 1.6 | 8.0 ± 1.4 |
| Years in storage (±SD) | 11.7 ± 6.9 | 13.1 ± 6.5 | 6.2 ± 6.2 |
| CPZ‐eq (±SD) (mg/day) | — | 559.1 ± 595.1 | 125 ± 156.1 |
Abbreviations: CPZ‐eq, chlorpromazine equivalent of total antipsychotic dosage; PMI, postmortem interval; RIN, RNA integrity number; SD, standard deviation.
2.2. Postmortem Brain Tissue pH Measurement
The pH of postmortem brain tissue was measured as previously described [21], with some modifications. Briefly, ~20 mg of frozen tissue was homogenized in nuclease‐free water (Ambion, Austin, TX, USA) at a ratio of 120 mg tissue/mL (5× volume). The pH of the homogenates was measured using a Twin pH‐B212 meter (Horiba, Kyoto, Japan).
2.3. RNA Isolation and Quality Measurements
Total RNA was extracted using an AllPrep DNA/RNA/Protein Mini Kit (Qiagen, Valencia, CA, USA), and genomic DNA was removed from the RNA samples by digestion with RNase‐free DNase I. RIN was assessed using a 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA), as previously described [22].
2.4. RNA Sequencing, Mapping, and Normalization
RNA isolation and mRNA sequencing were performed using a HiSeq 4000 platform (Illumina, Tokyo, Japan) with a 2 × 100 bp read length, yielding an average of 46.6 million reads per individual. The reads were cleaned using Trimmomatic (ver. 0.39) and mapped to the Ensembl human genome reference GRCh38.p13 using BWA (ver. 0.7.1). The same data were used in our previous study [23].
2.5. Demographic Factors
Sex, age, PMI, pH, RIN, and storage duration were analyzed as confounding factors. In the schizophrenia and bipolar disorder groups, we also considered the antipsychotic medication dosage in the 3 months before death (expressed in chlorpromazine equivalents).
2.6. Housekeeping Genes
We selected GAPDH, ACTB, HPRT1, RPL13A, B2M, SDHA, TBP, UBC, and HMBS as housekeeping genes, as previously used in postmortem brain tissue and central nervous system malignancy research [7, 24, 25, 26, 27]. The genes with the highest expression levels were considered representative of the group.
Correlations between the expression levels of these housekeeping genes and confounding factors were analyzed using Pearson's correlation coefficient (r).
2.7. Statistical Analysis
We analyzed the correlation between mRNA variant counts and confounding factors, namely, sex, age, PMI, pH, RIN, and storage duration, using RNA‐seq expression data from 56 postmortem brain samples from patients with schizophrenia, bipolar disorder, and nonpsychiatric controls. Pearson's correlation coefficients were calculated, and statistical significance was evaluated using Student's t‐test. Multiple comparison correction was performed using the Benjamini–Hochberg method, with a false discovery rate (FDR) threshold of q < 0.05 considered statistically significant (Supporting Information). To account for the potential influence of diagnostic groups, we performed an analysis of covariance (ANCOVA) as an additional analysis. This analysis was limited to mRNA variants that showed significant correlations with the confounding factors in the combined analysis of all 56 samples. In the ANCOVA, diagnostic group (schizophrenia, bipolar disorder, and control) was included as the independent variable, mRNA variant expression levels as the dependent variable, and each factor as a covariate. We also tested for interaction effects between diagnostic group and each confounding factor.
To identify factors that may differentially influence mRNA variant expression within each diagnostic group, Pearson's correlation coefficients were calculated separately for schizophrenia, bipolar disorder, and nonpsychiatric control groups for sex, age, PMI, pH, RIN, storage duration, and chlorpromazine equivalents (for schizophrenia and bipolar disorder only). All correlation analyses were conducted using Microsoft Excel 2019 (Microsoft Corporation, Redmond, WA, USA), and the ANCOVA was conducted using R version 4.4.3.
To characterize the biological processes associated with mRNA variants showing strong correlations with each confounding factor, we calculated the absolute values of correlation coefficients and selected the top 500 mRNA variants for pathway analysis. This approach was taken to mitigate the loss of statistical power that can result from reduced sample sizes when stratifying by diagnostic group, which may hinder the detection of significant associations. These selected variants were uploaded to Ingenuity Pathway Analysis (IPA; QIAGEN, Redwood City, CA, USA) for canonical pathway analysis.
3. Results
3.1. mRNA Variants Significantly Correlated With Confounding Factors
RNA‐seq analysis yielded expression data for 38 434 genes and detected 160 520 mRNA variants. Of these, 113 110 variants were retained for further analysis after excluding those with expression levels not significantly different from zero (Figure 1). In comparison, our previous microarray‐based expression analysis [11] with 48 702 probes identified 8286 probes with a signal intensity (SI) exceeding the threshold, which were considered significant. The dynamic range was calculated as the ratio between the maximum and minimum expression levels. The dynamic range was 3.47 (log10) for the microarray data and 6.49 (log10) for the RNA‐seq data, indicating that RNA‐seq has a higher sensitivity for detecting gene expression compared to microarrays. Figure S1 shows the comparison of expression patterns between RNA‐seq and microarray data.
FIGURE 1.
Total number of mRNA variants. A total of 160 520 mRNA variants were identified in the prefrontal cortex. After excluding those with levels not significantly different from zero, 113 110 mRNA‐encoding variants were selected for further analysis.
The total number of mRNA variants significantly correlated with sex was 139; age, 87; PMI, 99; pH, 35 386; RIN, 11 373; and storage duration, 13 414. Some mRNA variants overlapped among confounding factors (Figure 2). The results of the interaction analysis between diagnostic group (schizophrenia, bipolar disorder, and nonpsychiatric controls) and each confounding factor are presented in Table S1 and Figure S2.
FIGURE 2.
Total number of mRNA variants overlapping between tissue pH, RIN, and storage duration. The total number of mRNA variants that were significantly correlated with each confounding factor was 139 for sex, 87 for age, 99 for PMI, 35 386 for pH, 11 373 for RIN, and 13 414 for storage duration. Some mRNA variants overlapped with other confounding factors.
3.2. Pathway Analysis of Genes Correlated With Confounding Factors
In the schizophrenia group, the most enriched pathways were TGF‐β signaling for sex, kinetochore metaphase signaling for age, triacylglycerol biosynthesis for PMI, synaptogenesis signaling for pH, white adipose tissue browning for RIN, synaptogenesis signaling pathway for storage duration, and purine nucleotide degradation II (aerobic) for antipsychotic medication dosage. In the bipolar disorder group, the most enriched pathways were glycolysis I for sex, synaptogenesis signaling for both age and PMI, retinoic acid‐mediated apoptosis signaling for pH, cardiolipin biosynthesis II for RIN, oxidative phosphorylation for storage duration, and tetrapyrrole biosynthesis II for antipsychotic medication dosage. In the nonpsychiatric control group, the most enriched pathways were basal cell carcinoma signaling for gender, amyotrophic lateral sclerosis signaling for age, role of cytokines in mediating communication between immune cells for PMI, WNT/β‐catenin signaling for pH, oxidative phosphorylation for RIN, and ephrin receptor signaling for storage duration (Table 2). We compared the effects of RIN and pH on gene expression in the schizophrenia group using the data from previous microarray analyses [11] (Tables 3 and 4). Of the 320 pathways associated with RIN, 81 (25.3%) matched the microarray results. Similarly, of the 387 pathways associated with pH, 111 (28.7%) matched the microarray results.
TABLE 2.
Pathway analysis of genes correlated with each confounding factor.
| Category | Condition | Pathway | p |
|---|---|---|---|
| Age at death | Bipolar disorder | Synaptogenesis signaling pathway | 1.04 × 10−3 |
| Glutaminergic receptor signaling pathway | 1.36 × 10−3 | ||
| PI3K/AKT signaling | 1.41 × 10−3 | ||
| Control | Amyotrophic lateral sclerosis signaling | 4.42 × 10−3 | |
| Gustation pathway | 6.21 × 10−3 | ||
| DNA double‐strand break repair by nonhomologous end joining | 9.38 × 10−3 | ||
| Schizophrenia | Kinetochore metaphase signaling pathway | 3.57 × 10−3 | |
| Sertoli cell‐sertoli cell junction signaling | 5.48 × 10−3 | ||
| STAT3 pathway | 9.53 × 10−3 | ||
| CPZ‐eq | Bipolar disorder | Tetrapyrrole biosynthesis II | 2.35 × 10−3 |
| Apelin cardiac fibroblast signaling pathway | 5.29 × 10−3 | ||
| Heme biosynthesis II | 8.13 × 10−3 | ||
| Control | — | — | |
| — | — | ||
| — | — | ||
| Schizophrenia | Purine nucleotide degradation II | 2.34 × 10−3 | |
| Pathogenesis of multiple sclerosis | 6.80 × 10−3 | ||
| Unfolded protein response | 9.29 × 10−3 | ||
| PMI | Bipolar disorder | Synaptogenesis signaling pathway | 9.62 × 10−4 |
| Glutaminergic receptor signaling pathway | 1.26 × 10−3 | ||
| PI3K/AKT signaling | 1.32 × 10−3 | ||
| Control | Role of cytokines in mediating communication between immune cells | 1.06 × 10−4 | |
| Thyroid hormone metabolism II (via conjugation and/or degradation) | 2.60 × 10−4 | ||
| Nicotine degradation II | 3.53 × 10−4 | ||
| Schizophrenia | Triacylglycerol biosynthesis | 7.70 × 10−4 | |
| RHO‐A signaling | 5.50 × 10−3 | ||
| Dolichyl‐diphosphooligosaccharide biosynthesis | 8.44 × 10−3 | ||
| RIN | Bipolar disorder | Cardiolipin biosynthesis II | 2.11 × 10−4 |
| Phosphatidylglycerol biosynthesis II | 8.98 × 10−4 | ||
| Triacylglycerol biosynthesis | 1.32 × 10−3 | ||
| Control | Oxidative phosphorylation | 4.62 × 10−4 | |
| GABA receptor signaling | 1.22 × 10−3 | ||
| GABAergic receptor signaling pathway | 1.65 × 10−3 | ||
| Schizophrenia | White adipose tissue browning pathway | 2.78 × 10−3 | |
|
GPCR‐mediated integration of enteroendocrine signaling exemplified by an L cell‐adrenergic signaling |
3.49 × 10−3 | ||
| Leptin signaling in obesity | 3.69 × 10−3 | ||
| Sex | Bipolar disorder | Glycolysis I | 4.36 × 10−3 |
| Fatty acid oxidation I | 1.04 × 10−2 | ||
| UDP‐N‐acetyl‐D‐galactosamine biosynthesis II | 1.26 × 10−2 | ||
| Control | Basal cell carcinoma signaling | 9.62 × 10−3 | |
| Transcriptional regulatory network in embryonic stem cells | 1.21 × 10−2 | ||
| Fatty acid activation | 1.26 × 10−2 | ||
| Schizophrenia | TGF‐β signaling | 9.31 × 10−4 | |
| DNA double‐strand break repair by homologous recombination | 1.13 × 10−2 | ||
| Pathogen‐induced cytokine storm signaling pathway | 1.24 × 10−2 | ||
| Storage duration | Bipolar disorder | Oxidative phosphorylation | 5.05 × 10−7 |
| Sirtuin signaling pathway | 8.97 × 10−4 | ||
| Pulmonary fibrosis idiopathic signaling pathway | 2.27 × 10−3 | ||
| Control | Ephrin receptor signaling | 6.18 × 10−4 | |
| Paxillin signaling | 1.57 × 10−3 | ||
| Angiopoietin signaling | 1.92 × 10−3 | ||
| Schizophrenia | Synaptogenesis signaling pathway | 3.59 × 10−6 | |
| Endometrial cancer signaling | 6.08 × 10−5 | ||
| Estrogen receptor signaling | 2.66 × 10−4 | ||
| Tissue pH | Bipolar disorder | Retinoic acid‐mediated apoptosis signaling | 8.02 × 10−3 |
| Death receptor signaling | 8.79 × 10−3 | ||
| Guanosine nucleotide degradation III | 1.22 × 10−2 | ||
| Control | WNT/‐catenin signaling | 9.81 × 10−3 | |
| TGF‐β signaling | 9.97 × 10−3 | ||
| Mitochondrial L‐carnitine shuttle pathway | 2.69 × 10−2 | ||
| Schizophrenia | Synaptogenesis signaling pathway | 2.26 × 10−8 | |
| SNARE signaling pathway | 1.20 × 10−6 | ||
| Role of NFAT in cardiac hypertrophy | 1.75 × 10−5 |
TABLE 3.
Pathway analysis of genes correlated with tissue pH in schizophrenia.
| Current study (IPA) | p | Previous study (Reactome pathway analysis) | p |
|---|---|---|---|
| White adipose tissue browning pathway | 2.75 × 10−3 | ||
| GPCR‐mediated integration of enteroendocrine signaling exemplified by an L cell | 3.47 × 10−3 | ||
| α‐adrenergic signaling | 3.72 × 10−3 | ||
| Leptin signaling in obesity | 3.72 × 10−3 | ||
| Endocannabinoid inhibition pathway | 3.98 × 10−3 | ||
| Cellular effects of sildenafil (Viagra) | 4.37 × 10−3 | ||
| GNRH signaling | 4.57 × 10−3 | ||
| Role of BRCA1 in DNA damage response | 4.57 × 10−3 | SUMOylation of DNA damage response and repair proteins | 1.20 × 10−3 |
| Gap‐filling DNA repair synthesis and ligation in TC‐NER | 3.06 × 10−3 | ||
| TP53 regulates transcription of DNA repair genes | 7.00 × 10−3 | ||
| Dopamine receptor signaling | 4.57 × 10−3 | ||
| Corticotropin releasing hormone signaling | 4.68 × 10−3 | ||
| Glutaminergic receptor signaling pathway | 5.01 × 10−3 |
TABLE 4.
Pathway analysis of genes correlated with RIN in schizophrenia.
| Current study (IPA) | p | Previous study (Reactome pathway analysis) | p |
|---|---|---|---|
| Synaptogenesis signaling pathway | 2.26 × 10−8 | Transmission across chemical synapses | 7.83 × 10−5 |
| SNARE signaling pathway | 1.20 × 10−6 | Neurotransmitter release cycle | 1.95 × 10−3 |
| Glutamate neurotransmitter release cycle | 1.95 × 10−3 | ||
| Neuronal system | 2.95 × 10−3 | ||
| Retrograde neurotrophin signaling | 5.25 × 10−3 | ||
| Role of NFAT in cardiac hypertrophy | 1.74 × 10−5 | Calcineurin activates NFAT | 1.14 × 10−3 |
| CLEC7A (Dectin‐1) induces NFAT activation | 3.36 × 10−3 | ||
| GNRH signaling | 1.41 × 10−4 | ||
| Chemokine signaling | 1.66 × 10−4 | Interleukin‐1 signaling | 9.10 × 10−7 |
| Interleukin‐1 family signaling | 2.91 × 10−5 | ||
| Calcium signaling | 3.98 × 10−4 | ||
| Glutaminergic receptor signaling pathway | 4.68 × 10−4 | Negative regulation of NMDA receptor‐mediated neuronal transmission | 4.31 × 10−3 |
| Activation of NMDA receptors and postsynaptic events | 6.21 × 10−3 | ||
| RAC signaling | 4.79 × 10−4 | ||
| Opioid signaling pathway | 5.62 × 10−4 | ||
| Orexin signaling pathway | 7.08 × 10−4 |
3.3. Relationship Between Housekeeping Genes and Confounding Factors
We observed weak correlations between the expression of housekeeping genes and the confounding factors, including a weak correlation between sex and HPRT1 (r = −0.277, p = 0.038); correlations between CPZ‐eq and HMBS (r = 0.307, p = 0.021), tissue pH and ACTB (r = 0.269, p = 0.045), SDHA (r = 0.349, p = 0.008), UBC (r = 0.274, p = 0.04), HPRT1 (r = 0.466, p = 0.0002), RIN and HPRT1 (r = 0.270, p = 0.04), TBP (r = 0.381, p = 0.003), and between storage duration and TBP (r = −0.268, p = 0.046) (Table 5).
TABLE 5.
Pearson correlation coefficients (r: upper value) and p value (p: lower value) indicating the relationship between housekeeping genes and confounding factors.
| Sex | Age at death | CPZ‐eq | PMI | Tissue pH | RIN | Storage duration | |
|---|---|---|---|---|---|---|---|
| GAPDH | −0.14 (0.303) | 0.098 (0.474) | 0.139 (0.305) | −0.21 (0.118) | 0.338 (0.011) | 0.2 (0.140) | −0.02 (0.884) |
| ACTB | −0.12 (0.385) | 0.114 (0.403) | 0.026 (0.851) | −0.26 (0.053) | 0.269 (0.045) | 0.029 (0.833) | −0.039 (0.776) |
| HPRT1 | −0.28 (0.039) | −0.147 (0.279) | 0.168 (0.216) | 0.016 (0.906) | 0.466 (0.000) | 0.27 (0.044) | 0.127 (0.352) |
| RPL13A | 0.249 (0.064) | 0.12 (0.378) | −0.05 (0.729) | −0.24 (0.079) | −0.07 (0.609) | −0.01 (0.964) | −0.06 (0.660) |
| B2M | 0.204 (0.132) | 0.139 (0.306) | 0.05 (0.713) | −0.23 (0.087) | −0.038 (0.779) | −0.02 (0.865) | −0.17 (0.211) |
| SDHA | −0.05 (0.694) | 0.135 (0.321) | −0.04 (0.753) | −0.03 (0.831) | 0.349 (0.008) | 0.117 (0.389) | −0.144 (0.290) |
| TBP | −0.17 (0.205) | −0.077 (0.572) | −0.02 (0.865) | −0.01 (0.973) | 0.259 (0.054) | 0.381 (0.004) | −0.268 (0.046) |
| UBC | −0.02 (0.865) | 0.000 (0.998) | 0.142 (0.297) | −0.3 (0.023) | 0.274 (0.041) | −0.19 (0.170) | 0.105 (0.441) |
| HMBS | 0.12 (0.378) | −0.056 (0.681) | 0.307 (0.022) | −0.04 (0.760) | 0.052 (0.703) | 0.016 (0.910) | −0.088 (0.521) |
4. Discussion
Previous studies have used RNA‐seq to investigate RNA splicing and transcript isoform variation in psychiatric disorders, highlighting the importance of alternative splicing in schizophrenia and bipolar disorder, underscoring the relevance of transcriptomic diversity in the pathophysiology of these conditions [28, 29, 30]. In contrast, our study focuses on how confounding factors shape RNA variant expression profiles, offering a complementary perspective that enhances the interpretation of RNA‐seq data in postmortem brain studies.
To our knowledge, this study is the first to use RNA‐seq, a highly sensitive and comprehensive gene analysis method, to investigate the relationship between the gene expression profile of postmortem brain tissues from patients with schizophrenia, bipolar disorder, and nonpsychiatric controls and various confounding factors. We also identified important molecular pathways affected by each confounding factor and found that the confounding factors had little effect on the expression of representative housekeeping genes.
4.1. mRNA Variants Significantly Correlated With Confounding Factors
In the combined analysis across schizophrenia, bipolar disorder, and nonpsychiatric control groups, pH, RIN, and storage duration exhibited the highest number of mRNA variants significantly correlated with gene expression. We conducted an additional analysis to determine whether the relationships between gene expression and each factor differed across diagnostic groups. A significant interaction indicates that the correlation between gene expression and the factor varies by diagnostic group, thereby limiting the interpretability of the results obtained from the combined analysis. The mRNA variants associated with factors such as sex, age, pH, and RIN did not show significant interactions with diagnostic differences, suggesting that these associations are consistent across groups. In contrast, PMI and storage duration showed more frequent interactions with diagnostic differences. This is likely attributable to outliers in PMI within the bipolar disorder group and marked differences in storage duration among the three diagnostic groups.
A decrease in RIN may be associated with an agonal state [31] and PMI [32]. Both RIN and pH have been correlated with the degree of tissue degradation [9]. In this analysis, the number of mRNA variants affected by pH was three times greater than that affected by RIN. Previous studies found that a decrease in the pH of postmortem brain tissue was influenced by factors other than phosphate release from RNA degradation [33, 34] and was associated with the pathological progression of psychiatric disorders [35, 36, 37]. Mitochondrial dysfunction has been implicated in the pathogenesis of bipolar disorders [38, 39] owing to impaired oxidative phosphorylation, a shift to glycolytic energy production [40], lactate accumulation, and subsequent pH reduction. In schizophrenia, altered energy metabolism and lactate accumulation have been identified as potential causes of pH decrease [33]. Haloperidol and clozapine have been shown to increase lactate levels in the frontal lobe of rats [33], indicating the potential role of antipsychotic treatments. Although pH and RIN are correlated [41, 42], the factors contributing to tissue pH changes are diverse, including the pathological progression of psychiatric disorders and the effects of medication. Therefore, the higher number of mRNA variants affected by pH compared to RIN in this study can be attributed to factors that alter pH tissues, in addition to RNA degradation.
The number of mRNA variants associated with storage duration was the second highest among the confounding factors. Samples with matching RINs showed consistent gene expression levels in postmortem brain samples frozen for up to 23 years [43]. However, the same study identified genes showing a positive relationship with storage duration, though the mechanism by which storage duration affects gene expression remains unknown and requires further investigation. Among the groups, the bipolar disorder group tended to have a shorter storage duration than the schizophrenia and nonpsychiatric control groups, which may have confounded the differences in gene expression.
4.2. Insights From Pathway Analysis
In summary, the most prominently enriched pathways were related to metabolism, immune response, energy production, and DNA repair systems. High‐precision pathway analysis of genes correlated with each confounding factor identified various pathways associated with the immune system and DNA repair, consistent with a previous study [11]. Differences in confounding factors and correlated pathways were observed between the schizophrenia, bipolar disorder, and nonpsychiatric control groups. Notably, the RNA‐seq gene expression data showed similar trends to those obtained using traditional microarray methods. These findings highlight the importance of confounding factors in the interpretation of gene expression profiles between psychiatric diseases.
4.3. Relationship Between Housekeeping Genes and Confounding Factors
Few mRNA variants of the housekeeping genes were correlated with the confounding factors, except for those showing weak but significant correlations with tissue pH and RIN. This makes these genes suitable for use as standards for real‐time PCR and other expression analyses. Our study highlights the significant impact of pH and RIN on the accurate interpretation of gene expression profiles of postmortem brains. Previous studies have reported the usefulness of examining housekeeping gene expression in RNA‐seq data [44, 45]. Together with these findings, our results indicate the stable expression of housekeeping genes across different sample groups, supporting the reliability of our findings.
4.4. Limitations and Perspectives
Our study is limited in its focus on a single brain region, which necessitates further examination in other brain regions while matching the target regions for gene expression analysis. Moreover, the small sample size introduced some imbalances in variables such as age at death, PMI, and storage duration across diagnostic groups, potentially confounding the interpretation of group‐specific gene expression patterns. Further studies with larger sample sizes are required to determine whether the observed differences are associated with pathological or nonpathological factors. Finally, our research was limited to gene expression, and it remains unclear whether these changes are reflected at the protein level.
Author Contributions
Masataka Hatano, Atsuko Nagaoka, Kazusa Miyahara, Mizuki Hino, Hiroaki Tomita, and Yasuto Kunii designed the study. Mizuki Hino and Yasuto Kunii performed the experiments. Masataka Hatano, Atsuko Nagaoka, Yuto Hosogai, Risa Shishido, Hideomi Hamasaki, Mizuki Hino, Akiyoshi Kakita, Hiroaki Tomita, Yasuto Kunii, and Itaru Miura collected postmortem brain samples and clinical information. Masataka Hatano, Mizuki Hino, and Yasuto Kunii undertook the statistical analysis. Masataka Hatano wrote the first draft. All authors contributed to and have approved the final manuscript.
Ethics Statement
The use of human postmortem brain tissue was approved by the ethics committees of Fukushima Medical University (approval number: 1685).
Consent
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Appendix S1: npr270053‐sup‐0001‐AppendixS1.zip.
Figures S1–S2: npr270053‐sup‐0002‐FiguresS1‐S2.zip.
Table S1: npr270053‐sup‐0003‐TableS1.docx.
Acknowledgments
We thank H. Onuma for her contribution in coordinating donations. We appreciate all brain donors and their families for the time and effort they devoted to the consent process and interviews.
Hatano M., Nagaoka A., Miyahara K., et al., “Impact of Confounding Factors in Human Postmortem Brain Tissues on Gene Expression Profiles: A Comparison of Patients With Schizophrenia, Bipolar Disorder, and Nonpsychiatric Controls,” Neuropsychopharmacology Reports 45, no. 3 (2025): e70053, 10.1002/npr2.70053.
Funding: This study was supported in part by the Strategic Research Program for Brain Sciences from the Japan Agency for Medical Research and Development under grant number JP21wm0425019 (Y.K. and A.K.); a Grant‐in‐Aid for Scientific Research (B) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan under grant number K02376 (Y.K.); the Grant‐in‐Aid for Early‐Career Scientists from JSPS KAKENHI (grant number JP24K18734) (A.N.) and Collaborative Research Project of the Brain Research Institute, Niigata University under grant number 22002 (Y.K.).
Data Availability Statement
The data presented in this study are available on request from the corresponding author. However, our postmortem brain samples derived from the Japanese population are extremely valuable and culturally sensitive. Furthermore, we have not yet obtained explicit consent from the legal next of kin to make the full gene expression data publicly available. For these reasons, we are unable to deposit the raw sequencing data in a public repository at this time. Instead, to enhance transparency and reproducibility, we are providing a secondary dataset in the form of Supporting Information listing the correlation coefficients between gene expression and each demographic factor.
References
- 1. Yang A. C. and Tsai S. J., “New Targets for Schizophrenia Treatment Beyond the Dopamine Hypothesis,” International Journal of Molecular Sciences 18, no. 8 (2017): 1689. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Dean B., Copolov D., and Scarr E., “Understanding the Pathophysiology of Schizophrenia: Contributions From the Melbourne Psychiatric Brain Bank,” Schizophrenia Research 177, no. 1–3 (2016): 108–114. [DOI] [PubMed] [Google Scholar]
- 3. Webster M. J. and Kim S., “Collecting, Storing, and Mining Research Data in a Brain Bank,” Handbook of Clinical Neurology 150 (2018): 167–179. [DOI] [PubMed] [Google Scholar]
- 4. van Dijk E. L., Auger H., Jaszczyszyn Y., and Thermes C., “Ten Years of Next‐Generation Sequencing Technology,” Trends in Genetics 30, no. 9 (2014): 418–426. [DOI] [PubMed] [Google Scholar]
- 5. Vornholt E., Luo D., Qiu W., et al., “Postmortem Brain Tissue as an Underutilized Resource to Study the Molecular Pathology of Neuropsychiatric Disorders Across Different Ethnic Populations,” Neuroscience and Biobehavioral Reviews 102 (2019): 195–207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Tomita H., Tanaka C., and Yu Z., “Control of Confounding Factors in Brain Bank Management,” Brain Science and Mental Disorder 20 (2009): 17–24. [Google Scholar]
- 7. Nagaoka A., Hino M., Izumi R., et al., “Availability of Individual Proteins for Quantitative Analysis in Postmortem Brains Preserved in Two Different Brain Banks,” Neuropsychopharmacology Reports 44, no. 2 (2024): 399–409. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Schroeder A., Mueller O., Stocker S., et al., “The RIN: An RNA Integrity Number for Assigning Integrity Values to RNA Measurements,” BMC Molecular Biology 7 (2006): 3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Sheedy D., Harding A., Say M., Stevens J., and Kril J. J., “Histological Assessment of Cerebellar Granule Cell Layer in Postmortem Brain; a Useful Marker of Tissue Integrity?,” Cell and Tissue Banking 13, no. 4 (2012): 521–527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Lipska B. K., Deep‐Soboslay A., Weickert C. S., et al., “Critical Factors in Gene Expression in Postmortem Human Brain: Focus on Studies in Schizophrenia,” Biological Psychiatry 60, no. 6 (2006): 650–658. [DOI] [PubMed] [Google Scholar]
- 11. Miyahara K., Hino M., Yu Z., et al., “The Influence of Tissue pH and RNA Integrity Number on Gene Expression of Human Postmortem Brain,” Frontiers in Psychiatry 14 (2023): 1156524. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Wang C., Gong B., Bushel P. R., et al., “The Concordance Between RNA‐Seq and Microarray Data Depends on Chemical Treatment and Transcript Abundance,” Nature Biotechnology 32, no. 9 (2014): 926–932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Zhao S., Fung‐Leung W. P., Bittner A., Ngo K., and Liu X., “Comparison of RNA‐Seq and Microarray in Transcriptome Profiling of Activated T Cells,” PLoS One 9, no. 1 (2014): e78644. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Li Y. I., van de Geijn B., Raj A., et al., “RNA Splicing Is a Primary Link Between Genetic Variation and Disease,” Science 352, no. 6285 (2016): 600–604. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Raj T., Li Y. I., Wong G., et al., “Integrative Transcriptome Analyses of the Aging Brain Implicate Altered Splicing in Alzheimer's Disease Susceptibility,” Nature Genetics 50, no. 11 (2018): 1584–1592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Tian J., Lam T. G., Ross S. K., et al., “An Analysis of RNA Quality Metrics in Human Brain Tissue,” Journal of Neuropathology and Experimental Neurology 84, no. 3 (2025): 236–243. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Fromer M., Roussos P., Sieberts S. K., et al., “Gene Expression Elucidates Functional Impact of Polygenic Risk for Schizophrenia,” Nature Neuroscience 19, no. 11 (2016): 1442–1453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Jaffe A. E., Tao R., Norris A. L., et al., “qSVA Framework for RNA Quality Correction in Differential Expression Analysis,” Proceedings of the National Academy of Sciences of the United States of America 114, no. 27 (2017): 7130–7135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Fass S. B., Mulvey B., Chase R., et al., “Relationship Between Sex Biases in Gene Expression and Sex Biases in Autism and Alzheimer's Disease,” Biology of Sex Differences 15, no. 1 (2024): 47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Wapeesittipan P. and Joshi A., “Integrated Analysis of Robust Sex‐Biased Gene Signatures in Human Brain,” Biology of Sex Differences 14, no. 1 (2023): 36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Ono C. T., Yu Z., Kikuchi Y., et al., “Minimal Amount of Tissue‐Based pH Measurement to Improve Quality Control in Neuropsychiatric Post‐Mortem Brain Studies,” Psychiatry and Clinical Neurosciences 73, no. 9 (2019): 566–573. [DOI] [PubMed] [Google Scholar]
- 22. Ono C., Yu Z., Kasahara Y., Kikuchi Y., Ishii N., and Tomita H., “Fluorescently Activated Cell Sorting Followed by Microarray Profiling of Helper T Cell Subtypes From Human Peripheral Blood,” PLoS One 9, no. 11 (2014): e111405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Shishido R., Kunii Y., Hino M., et al., “Evidence for Increased DNA Damage Repair in the Postmortem Brain of the High Stress‐Response Group of Schizophrenia,” Frontiers in Psychiatry 14 (2023): 1183696. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Penna I., Vella S., Gigoni A., Russo C., Cancedda R., and Pagano A., “Selection of Candidate Housekeeping Genes for Normalization in Human Postmortem Brain Samples,” International Journal of Molecular Sciences 12, no. 9 (2011): 5461–5470. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Silberberg G., Baruch K., and Navon R., “Detection of Stable Reference Genes for Real‐Time PCR Analysis in Schizophrenia and Bipolar Disorder,” Analytical Biochemistry 391, no. 2 (2009): 91–97. [DOI] [PubMed] [Google Scholar]
- 26. Tan S. C., Carr C. A., Yeoh K. K., Schofield C. J., Davies K. E., and Clarke K., “Identification of Valid Housekeeping Genes for Quantitative RT‐PCR Analysis of Cardiosphere‐Derived Cells Preconditioned Under Hypoxia or With Prolyl‐4‐Hydroxylase Inhibitors,” Molecular Biology Reports 39, no. 4 (2012): 4857–4867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Valente V., Teixeira S. A., Neder L., et al., “Selection of Suitable Housekeeping Genes for Expression Analysis in Glioblastoma Using Quantitative RT‐PCR,” BMC Molecular Biology 10 (2009): 17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Takata A., Matsumoto N., and Kato T., “Genome‐Wide Identification of Splicing QTLs in the Human Brain and Their Enrichment Among Schizophrenia‐Associated Loci,” Nature Communications 8 (2017): 14519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Jaffe A. E., Straub R. E., Shin J. H., et al., “Developmental and Genetic Regulation of the Human Cortex Transcriptome Illuminate Schizophrenia Pathogenesis,” Nature Neuroscience 21, no. 8 (2018): 1117–1125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Gandal M. J., Zhang P., Hadjimichael E., et al., “Transcriptome‐Wide Isoform‐Level Dysregulation in ASD, Schizophrenia, and Bipolar Disorder,” Science 362, no. 6420 (2018): eaat8127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Tomita H., Vawter M. P., Walsh D. M., et al., “Effect of Agonal and Postmortem Factors on Gene Expression Profile: Quality Control in Microarray Analyses of Postmortem Human Brain,” Biological Psychiatry 55, no. 4 (2004): 346–352. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Barton A. J., Pearson R. C., Najlerahim A., and Harrison P. J., “Pre‐ and Postmortem Influences on Brain RNA,” Journal of Neurochemistry 61, no. 1 (1993): 1–11. [DOI] [PubMed] [Google Scholar]
- 33. Halim N. D., Lipska B. K., Hyde T. M., et al., “Increased Lactate Levels and Reduced pH in Postmortem Brains of Schizophrenics: Medication Confounds,” Journal of Neuroscience Methods 169, no. 1 (2008): 208–213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Pruett B. S. and Meador‐Woodruff J. H., “Evidence for Altered Energy Metabolism, Increased Lactate, and Decreased pH in Schizophrenia Brain: A Focused Review and Meta‐Analysis of Human Postmortem and Magnetic Resonance Spectroscopy Studies,” Schizophrenia Research 223 (2020): 29–42. [DOI] [PubMed] [Google Scholar]
- 35. Hagihara H., Catts V. S., Katayama Y., et al., “Decreased Brain pH as a Shared Endophenotype of Psychiatric Disorders,” Neuropsychopharmacology 43, no. 3 (2018): 459–468. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Park H. J., Choi I., and Leem K. H., “Decreased Brain pH and Pathophysiology in Schizophrenia,” International Journal of Molecular Sciences 22, no. 16 (2021): 8358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Hagihara H., Murano T., and Miyakawa T., “The Gene Expression Patterns as Surrogate Indices of pH in the Brain,” Frontiers in Psychiatry 14 (2023): 1151480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Cikankova T., Sigitova E., Zverova M., Fisar Z., Raboch J., and Hroudova J., “Mitochondrial Dysfunctions in Bipolar Disorder: Effect of the Disease and Pharmacotherapy,” CNS & Neurological Disorders Drug Targets 16, no. 2 (2017): 176–186. [DOI] [PubMed] [Google Scholar]
- 39. Rezin G. T., Amboni G., Zugno A. I., Quevedo J., and Streck E. L., “Mitochondrial Dysfunction and Psychiatric Disorders,” Neurochemical Research 34, no. 6 (2009): 1021–1029. [DOI] [PubMed] [Google Scholar]
- 40. Stork C. and Renshaw P. F., “Mitochondrial Dysfunction in Bipolar Disorder: Evidence From Magnetic Resonance Spectroscopy Research,” Molecular Psychiatry 10, no. 10 (2005): 900–919. [DOI] [PubMed] [Google Scholar]
- 41. Kingsbury A. E., Foster O. J., Nisbet A. P., et al., “Tissue pH as an Indicator of mRNA Preservation in Human Post‐Mortem Brain,” Brain Research: Molecular Brain Research 28, no. 2 (1995): 311–318. [DOI] [PubMed] [Google Scholar]
- 42. Stan A. D., Ghose S., Gao X. M., et al., “Human Postmortem Tissue: What Quality Markers Matter?,” Brain Research 1123, no. 1 (2006): 1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. White K., Yang P., Li L., Farshori A., Medina A. E., and Zielke H. R., “Effect of Postmortem Interval and Years in Storage on RNA Quality of Tissue at a Repository of the NIH NeuroBioBank,” Biopreservation and Biobanking 16, no. 2 (2018): 148–157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Wang Z., Lyu Z., Pan L., Zeng G., and Randhawa P., “Defining Housekeeping Genes Suitable for RNA‐Seq Analysis of the Human Allograft Kidney Biopsy Tissue,” BMC Medical Genomics 12, no. 1 (2019): 86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Chen C. M., Lu Y. L., Sio C. P., Wu G. C., Tzou W. S., and Pai T. W., “Gene Ontology Based Housekeeping Gene Selection for RNA‐Seq Normalization,” Methods 67, no. 3 (2014): 354–363. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Appendix S1: npr270053‐sup‐0001‐AppendixS1.zip.
Figures S1–S2: npr270053‐sup‐0002‐FiguresS1‐S2.zip.
Table S1: npr270053‐sup‐0003‐TableS1.docx.
Data Availability Statement
The data presented in this study are available on request from the corresponding author. However, our postmortem brain samples derived from the Japanese population are extremely valuable and culturally sensitive. Furthermore, we have not yet obtained explicit consent from the legal next of kin to make the full gene expression data publicly available. For these reasons, we are unable to deposit the raw sequencing data in a public repository at this time. Instead, to enhance transparency and reproducibility, we are providing a secondary dataset in the form of Supporting Information listing the correlation coefficients between gene expression and each demographic factor.