Metabolic changes in patients with bipolar disorder in spring

ABSTRACT Bipolar disorder (BD) is a common mental condition with a seasonal pattern (SP) of onset. In the spring, there is a higher incidence rate of mania or mixed onset and suicide. However, the underlying mechanism of this SP remains unclear. In this study, targeted metabolomics was used to understand metabolic changes in patients with BD before and after the spring equinox. Nine patients with BD and matched healthy controls were tested for serum metabolomics at the spring equinox and 15 days before and after the spring equinox. The results showed that 27 metabolite levels changed significantly, three of which interacted between three time points and groups involving triglyceride (TG, 20:4_34:2), TG (20:4_34:3) and TG (16:0_36:6). The identified metabolic pathways mainly involved arginine biosynthesis, D-glutamine and D-glutamate metabolism, and nitrogen metabolism. Changes in solar radiation and lunar cycle during spring may be the external causes of metabolic changes. These findings help to further explore seasonal metabolic changes in patients with BD and provide insights into the mechanisms of patients’ emotional changes in spring.


Introduction
Bipolar disorders (BD) are severe and chronic psychological disorders that consist of manic/hypomanic episodes and depressive episode (McIntyre et al. 2020). According to the Global Burden of Diseases, Injuries, and Risk Factors Study 2019, there were 39.5 million (95% confidence interval: 33.0-46.8) estimated cases of BD in 2019 (Collaborators 2022). There is still limited understanding of the etiology of BD, which involves genetic and environmental components (Aguglia et al. 2017;Lichtenstein et al. 2009). Increasing evidence suggests that abnormalities in chronobiological rhythms are critical pathophysiological elements in BD (Brochard et al. 2018;Harvey 2008;Maruani et al. 2018). Seasonal variability may play a pivotal role in mood-related behavior among patients with BD (Aguglia et al. 2018;McClung 2011). The first evidence of seasonal variation in BD is monthly fluctuations in manic and depressive episodes, which show a consistent spring peak (Frangos et al. 1980;Lee et al. 2007;Parker and Walter 1982). The steepest increase in pure manic admissions is related to the greatest lengthening of the photoperiod in spring (Cassidy and Carroll 2002). Cognitive functions, psychotic symptoms, aggression, and suicidality all show significant spring-related variation in patients with BD (Bauer et al. 2019;Christodoulou et al. 2012;Geoffroy et al. 2014). These findings suggest that BD patients may have trouble adjusting to environmental perturbations during spring (Postolache et al. 2010).
Despite numerous findings supporting the impact of spring on BD, the pathophysiology of spring's effect on patients with BD remains unclear. Metabolomics estimates the downstream effects of proteomic variation in individuals by identifying and quantifying small molecules (Patti et al. 2012). Also, metabolomics has been used to evaluate chronobiological changes in plasma metabolites that may show seasonal variation (Ang et al. 2012;Ma et al. 2020). In particular, seasonal variation in blood lipids is thought to affect the clinical diagnoses and management decisions (Cambras et al. 2017;Ma et al. 2020). However, relevant biomarkers or biochemical signatures of spring in BD remain unknown. To address this research gap, we aimed to explore the metabolic effects of spring on patients suffering from BD and the underlying physiological mechanism. In addition, time-series analyses can identify synchronies between mood cycles and the 14.8-day spring-neap cycle in individuals with rapid cycling BD (Wehr 2018). The moon's semidiurnal gravimetric tides may drive patients' BD cycles. Thus, we considered the influence of the lunar tidal cycle and selected the observation time as the spring equinox and 15 days before and after, to explore spring-related pathophysiological mechanisms. The spring equinox is the center of spring and the time point at which the solar radiation rate changes most rapidly in spring. A 15-day period is sufficient to allow the body to establish a stable photoperiodic memory (Prendergast et al. 2000). Furthermore, these three time points contain a complete lunar cycle: from the second quarter to the second quarter (http://www.hko.gov.hk/sc/gts/astron omy/Moon_Phase.htm). We hope to probe the underlying mechanism of metabolic changes in patients with BD in spring, thereby providing suggestions for managing patients' emotional changes in spring.

Participants
During 2021, we recruited 9 patients with BD from the Beijing Anding Hospital, Capital Medical University as well as 9 healthy controls (HCs). All the participants were Han Chinese and long-term residents of Beijing aged between 19 and 49. To avoid the impact of rapid emotional changes on metabolism, these patients with BD were in a stable period. Serum metabolomics analysis was performed for the patients with BD and healthy controls at three time points: 15 days before the spring equinox, the spring equinox, and 15 days after the spring equinox. Due to operational reasons, patients were screened during clinical visits 2 days before and after the time points. This study was approved by the Institutional Ethics Committee of Beijing Anding Hospital, Capital Medical University, and all participants provided written informed consent.
The inclusion criteria for patients with BD were as follows: (1) being diagnosed by attending psychiatrists (two or more) from the Psychiatry Department and meeting the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) criteria for BD: (2) Hamilton Depression Rating Scale (HDRS) score equal to or less than 7; (3) Young Mania Rating Scale (YMRS) score equal to or less than 6; and (4) at clinically stable phase more than 8 weeks with the mood-stabilizing treatments, which were stably used for patients more than 4 weeks. The exclusion criteria were as follows: (1) experiencing suicidal or homicidal behavior; (2) having neurodegenerative diseases, serious endocrine diseases, or metabolic disorders; (3) receiving hormonal medication or electroconvulsive therapy in the past 6 months; (4) the occurrence of any major life events; or (5) being pregnant or breastfeeding. HCs were matched with patients with BD according to age, gender, height and weight. HCs were recruited through solicitation and screened using M.I.N.I. for previous psychiatric disorders. None of the HCs had a history of acute or chronic illness or any drug allergies.

Serum sample collection
Prior to blood sample collection, participants were instructed to fast for 12 h. Consumption of tea, coffee, and alcohol was limited/restricted. Blood sample collection was concentrated from 8:00 am to 10:00 am. BD Vacutainer SST II Advance Tubes were used to collect peripheral blood samples from all participants. The blood samples were processed within 4 h of collection and stored at −80°C. Serum samples were shipped on dry ice to Beijing Protein Innovation Co., Ltd. in Beijing for metabolite analyses.

Biospecimen handling, data acquisition, and data preprocessing
A metabolomics kit (MxP® Quant 500 kit: BIOCRATES Life Sciences AG, Innsbruck, Austria) was used to analyze the samples collected at the three time points. This non-targeted kit can detect the concentrations of 26 classes of up to 630 metabolites. The samples were placed into a 96-well sample preparation device to facilitate the subsequent measurement. In the 96-well plates, the concentration of different standard solutions was used as the calibration standard. A stable isotope labeled internal standard was used as an internal standard for quantifying the metabolite concentration of the sample and was pre-added to the 96-well plate. After adding the samples and standards to the sample preparation device, phenyl isothiocyanate solution was added. Phenyl isothiocyanate solution can be used to derive other analytes such as amino acids. Organic solvent was used to extract the target analyte after derivatization. Subsequently, we diluted the samples to facilitate analysis of the metabolite concentration. Flow injection analysis-tandem mass spectrometry (FIA-MS/MS) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) were the main methods used for detection. Different detection methods are suitable for different metabolites. For example, in the present study, lipids were measured by FIA-MS/MS, while amino acids were measured by LC-MS/MS. The MxP Quant 500 Column System was used for LC-MS/MS analysis. Solvent a (water containing 0.2% formic acid) and solvent digitalis a (methyl green containing 0.2% formic acid) made up the mobile phase. For FIA-MS/MS, the flow injection solvent was used and the flow rate was 30 μL/min mobile-phase operation flow injection analysis plate. The Sciex Analyst® mass spectrometry software package was used to quantify the data. The results were then loaded into Biocrates MetIDQTM software for analysis. For all analytes, the accuracy of detection was within the normal range of the method, as measured by internal calibrators. There was no imputation and no normalization between the plates.

Statistical analysis
Statistical analyses were performed using the MetaboAnalyst 5.0 platform (Pang et al. 2021) (http:// www.metaboanalyst.ca/) and IBM SPSS Statistics 26.0 and RStudio 4.0.5. We used the mice package in RStudio to perform multiple imputations for missing data and some data were natural log transformed to meet normality assumptions. After metadata checking, data filtering and normalization, a two-way repeated analysis of variance (ANOVA2) with the model Time series + one experimental factor was used to analyze differences in the data using MetaboAnalyst 5.0. Metabolites with p < 0.05 and a false discovery rate (FDR) in multiple testing correction (p < 0.1) were included for further investigation. For data with differences, we used the Repeated measure of a general linear model module and obtained the differences between time points by multiple comparisons (EMMEANS) in SPSS 26.0. Finally, we performed pathway analysis of timevarying metabolites using the MetaboAnalyst 5.0 platform. During parameter selection, Enrichment method was Hypergeometric Test, Topology analysis was Relative-Betweeness Centrality and A pathway library was Homo sapiens (KEGG).

Demographic characteristics
The study involved nine patients with BD and nine healthy controls. Each group consisted of three males and six females. The patients with BD and HC groups did not differ significantly in age (t = 0.128, p = 0. 900) or BMI (t = 0.689, p = 0.501). Of the nine patients, four were diagnosed with type I bipolar disorder (BP-I) and five were diagnosed with type II bipolar disorder (BP-II).

Metabolites in serum
The peaks analyzed by LC-MS/MS and FIA-MS/MS were matched with compounds in the mzCloud, mzVault and MassList databases. A total of 630 serum metabolites were identified in patients with BD and HCs. For quality control, metabolites with a coefficient of variation (CV) >0.2 were removed. After these samples were removed, 355 metabolites were retained for statistical analysis.
In the "time series + one empirical factor" model analysis, 34 metabolites showed significant changes between patients with BD and HCs (before adjusting the p value). Furthermore, 92 metabolites significantly changed with time and 67 metabolites showed a significant interaction between three time points and groups before adjusting the p values (p < 0.05, details shown in Tables S1 and S2 in supplementary materials). After adjusting the p values (p adjust ), only 27 metabolites showed significant changes with time and three metabolites showed a significant interaction between three time points and groups (p adjust <0.1; Table 2 and Table S3).

Changes in metabolites over time
The 27 significantly changed metabolites were divided into eight categories: amino acids and amino acid-related, carboxylic acids, ceramides, cholesteryl esters, diglycerides, indoles and derivatives, lysophosphatidylcholines, phosphatidylcholines and triglycerides. At the three time points, the concentrations tended to show a "checkmark" shape: the concentration was lower at the second time point and higher at the first and third time points (Figure 1). There was a higher concentration of Gln and Cit in HCs than in patients with BD, but no significant difference was observed in HArg and Sarcosine. Lac also showed no significant difference between patients and HCs. Both ceramides were higher in patients than in HCs. Lysophosphatidylcholines were generally higher in patients, except that lysoPC a C18:0 was higher in HCs at the first time point. For phosphatidylcholines, the concentration of PC aa c34:2 was higher in HCs at the first and second time points, but higher in patients suffering from BD at the third time point. Patients had a higher concentration of PC ae C40:2 at all three time points. The concentrations of three kinds of cholesterol showed different situations: CE (16:0) was higher in patients, CE (18:2) was higher in HCs, and CE (17:1) was significantly higher in patients than HCs only at the third time point. At all time points, the concentrations of all triglycerides were higher in patients with BD than HCs. Overall, HCs mostly showed significant changes after the spring equinox, while the metabolite concentrations in patients with BD changed significantly around the spring equinox. In terms of types of metabolites, diglycerides and triglycerides showed significant changes at the first and second time points in patients with BD. Metabolites in HCs changed significantly between time point 2 and time point 3. Other metabolites in all participants were prone to significant changes between the second and third time points (shown in Table S4).

Pathway analysis
A pathway analysis was performed in MetaboAnalyst 5.0 for metabolites that changed significantly over time (shown in Figure 2). The included metabolites were seven non-lipid differential metabolites. Three significant pathways were identified: arginine biosynthesis, D-glutamine and D-glutamate metabolism, and nitrogen metabolism (shown in Figure 3, p < 0.05).

Discussion
The current study explored the possibility of springrelated variation in serum metabolites in patients with BD compared to HCs by utilizing LC-MS/MS and FIA-MS/MS methods. The principal observation was that 27 metabolites showed significant changes in the spring. Additionally, the concentration of differential metabolites generally reached a low point at the spring equinox. Together, these findings suggest that arginine biosynthesis, D-glutamine and D-glutamate metabolism and nitrogen metabolism may be involved in spring-related variations in BD.
Metabolomics has been used to identify and probe the molecular mechanisms of BD (Guest et al. 2016). Recently, a metabolomics systematic review identified 80 metabolites and nine metabolic pathways in plasma might act as biomarkers to distinguish patients with BD from controls (Wei et al. 2021). The previous study has revealed nine metabolic pathways in which glutamate metabolism and arginine metabolism also showed significant differences in our study. Glutamine and lactic acid have shown significant regulatory trends in many previous studies comparing patients with BD and HCs (MacDonald et al. 2019). In addition, differences in lipid metabolism between controls and patients with BD at different attack stages have been reported by several studies (Huang et al. 2018;Wysokiński et al. 2015). Metabolic changes caused by spring may have something in common with metabolic differences caused by BD, suggesting that the metabolic changes we observed may be part of the molecular mechanism of BD recurrence in spring.
Importantly, the disturbed seasonal rhythms of BD patients may be affected by climate changes and environmental factors even if they are symptom-free  (Rosenthal et al. 2020). Several studies have underlined solar insolation, lunar tidal cycle, ambient temperature, and other factors impacting on BD (Aguglia et al. 2019;Yao et al. 2023). During the spring and fall equinoxes in the Northern Hemisphere, solar insolation changes are maximal; in contrast, at the solstices, the rate of change is greatly reduced (Rosenthal et al. 2021). The seasonality of BD symptoms may be affected by seasonal changes in solar insolation, particularly at locations away from the equator (Goodwin and Jamison 2008). A study of the solar radiation at 51 locations in the southern and northern hemispheres found that the solar radiation change rate may be a seasonal trigger for BD symptoms (Rosenthal et al. 2020). The continuous increase in glutamine observed in the present study is consistent with an increase in light intensity. The concentration changes of diglycerides and triglycerides in HCs showed a symmetrical checkmark shape, which is consistent with the solar radiation change rate. However, the change in concentration rate of metabolites in patients with BD between time points 2 and 3 was smaller than that in HCs. Thus, it is critical to further explore the regulation of blood lipid concentrations in patients suffering from BD. Furthermore, the lunar tidal cycle cannot be ignored. The patients' BD cycles may be affected by the moon's semidiurnal gravimetric tides. In the present study, we considered the influence of the lunar tidal cycle and made observations at three time points. The decrease in metabolite concentration at time point 2 may also be related to changes in the lunar tidal cycle. Further research is warranted to observe the effects of light intensity, lunar tidal cycles, and other environmental factors on metabolism in BD.
Notably, our findings indicated a decrease in cholesterol and triglycerides in patients with BD at the vernal equinox, which is consistent with prior research (Huang et al. 2018;Wysokiński et al. 2015). Lipid-related studies have shown that elevated cholesterol and triglycerides are significantly associated with emotional symptoms. Additionally, we observed that the serum glutamine level of patients with BD in spring was lower than that of HCs, and the increase rate was smaller with time, which could be due to glutamate metabolism in patients with BD. Since the recruited patients with BD were stable due to medication, we did not observe mood fluctuations. Nevertheless, changes in lipid and glutamate-related metabolism might shed light on the mechanisms of the peak of mania in spring. These results suggest that biological rhythms in spring display a range of abnormalities in BD, making biological rhythms a potential candidate to probe the mechanism of seasonal affective behaviors in patients with BD. Further investigation of the biological mechanisms of seasonal changes in BD could thus provide a therapeutic direction for psychiatrists and clinical treatment. Future studies should also explore biological changes in summer, autumn and winter to observe the underlying mechanisms of seasonal changes in patients with BD.
This study is subject to several limitations. First, the sample size is relatively small. A larger sample would provide more reliable findings. Secondly, due to the need to rule out the impact of emotion on metabolite changes, we enrolled patients with BD in the stable state without mood swings. Future studies should further probe metabolite changes in patients during the attack period in spring. Thirdly, we estimated the alterations of metabolites in the spring rather than other time periods. The patterns of metabolites in summer, autumn and winter should be further explored in patients with BD in the future.
In conclusion, we observed longitudinal changes in metabolites in patients with BD and healthy subjects during the spring. The results indicate that the spring period affects lipid metabolism, arginine biosynthesis, D-glutamine and D-glutamate metabolism, and nitrogen metabolism in patients with BD. These changes may be affected by solar radiation or lunar tidal cycle changes, thereby affecting glutamate and lipid metabolism in the body and ultimately resulting in spring mood changes. This relationship needs to be confirmed by further animal models and cell experiments. Our findings may provide new directions for exploring the impact of time on patients with BD and understanding why time affects mood, potentially improving patients' ability to cope with the spring season.

Disclosure statement
No potential conflict of interest was reported by the author(s).