1 Introduction

Schizophrenia (SZ) is a chronic and severe mental disorder with a prevalence of 1% worldwide (Lopez & Murray, 1998). Although a number of molecular, pharmacological and clinical studies have been published, the pathogenesis and pathophysiology of SZ is poorly understood (Aleksic & Ozaki, 2017; Dawidowski et al., 2021; Owen et al., 2016). The dopaminergic hypothesis that hyperactive dopamine transmission results in the symptoms of SZ is the dominant hypothesis and current antipsychotic drug is mainly based on this hypothesis to reduce the psychiatric symptoms (Davis et al., 1991; Kapur, 2003; Meltzer & Stahl, 1976). Olanzapine (OLA) is considered to be one of the first-line atypical antipsychotic drugs for the treatment of SZ, which can be used for fist-episode SZ or long-term maintenance treatment (Correll et al., 2021). OLA binds with the following receptors: dopamine, serotonin, histamine H1, adrenergic α1, and muscarinic M (Chelkeba et al., 2017; Tollens et al., 2018). It acts an antagonist at dopamine D2 receptors in the mesolimbic pathway, blocking the potential action of dopamine on post-synaptic receptors. However, recent clinical evidence has shown that chronic OLA treatment causes severe metabolic disturbances, such as insulin resistance, obesity and dyslipidemia (Tek et al., 2016). In addition, there is a considerable individual variability with regard to clinical symptoms and clinical response to antipsychotics among patients with SZ (Messias et al., 2007). It is estimated that approximately 30% patients with SZ do not respond to olanzapine and develop treatment-resistant SZ during the course of the illness according to the epidemiological data from the research articles (Elkis & Buckley, 2016). From clinical perspective, identifying potential molecular pathogenic factors and finding predictive biomarkers for the response to OLA could improve the clinical management of SZ patients in the early stage of this disorder.

Pharmacometabolomics, as an application of metabolomics in drug efficacy research, have been used to map the global metabolite changes after antipsychotics treatment in SZ (Kaddurah-Daouk & Weinshilboum, 2014). A number of studies have shown significant changes in mitochondrial number, structure and function or the production of adenosine triphosphate (ATP) in patients with acute episode of psychosis, indicating a role for mitochondrial energetics in the early stage of SZ (Park et al., 2021). Recent studies reported that blood lipid metabolism involved in the altered carnitine levels and its derivatives are implicated in the pathogenesis and pathophysiology of SZ (Cao et al., 2019; Kriisa et al., 2017).

Carnitine (3-hydroxy-4-N-trimethylaminobutyrate, C7H15NO3) is an essential dietary amino acid obtained from animal products such as meat, fish and milk, or is biosynthesized from l-lysine and l-methionine (Flanagan et al., 2010). It has various physiological and biochemical functions in the central nervous system, such as anti-inflammatory, antioxidant, anti-apoptosis and enhancement of autophagy (Moghaddas & Dashti-Khavidaki, 2016; Traina, 2016). Studies reported that l-carnitine supplementation can promote the early recovery of cerebral infarction through its antioxidant and anti-inflammatory functions (Wang et al., 2017). In addition, carnitine plays a crucial role in the protection of DNA and protein from oxidative stress throughout its antioxidant properties (Ribas et al., 2014). Particularly, carnitines are also found to regulate and promote the synaptic neurotransmission, most notably cholinergic neurotransmission (Agarwal & Said, 2004; Calvani et al., 1992; Nałecz et al., 2004). Indeed, SZ patients do show hallmarks of abnormalities in immune response regulation, neuroendocrine and molecular pathways of neurotransmitter systems (Mould et al., 2021; Rahimian et al., 2021; Upthegrove & Khandaker, 2019; van Kesteren et al., 2017; van Mierlo et al., 2019).

Notably, carnitine and its derivatives are known for the crucial roles in the transport of long-chain fatty acids to mitochondrial matrix and oxidation of long-chain fatty acyl coenzyme A (CoAs) (Noland et al., 2009; Siliprandi et al., 1990; Young et al., 2018). The carnitine system can enhance neuron energy metabolism as a reservoir pool of acyl CoAs or as a transfer of fatty acid beta-oxidation (Jones et al., 2010). Abnormal carnitine metabolism or carnitine deficiency decreases fatty acid beta-oxidation and thus decreases the production of mitochondrial energy (ATP), which have been implicated in mental disorders (Kępka et al., 2021). Other function of carnitine is to maintain membrane integrity and stabilize the physiologic coenzyme A (CoASH)/acetyl-CoA ratio in mitochondria (Siliprandi et al., 1990). It is noteworthy that disturbances of mitochondrial morphology, function and genesis are gradually considered as components of psychiatric disorders, including SZ, bipolar disorder, depression and autism (Rajasekaran et al., 2015; Takahashi & Ogushi, 1953). Interestingly, the neuroprotective effect of a novel twin compound LR134 and LR143 containing carnitine substructure was correlated with the protection of mitochondria accompanied by the improvement of energy supply (Spagnoli et al., 1991; Wang et al., 2017).

Based on the previous studies, altered plasma or serum acylcarnitine can potentially be identified as a biomarker that indicates abnormalities in beta-oxidation (Cavedon et al., 2005; Kler et al., 1991; Rinaldo et al., 2008; Vreken et al., 1999). Particularly, preclinical studies have shown that elevated blood medium- and long-chain acylcarnitine levels have been correlated with incomplete fatty acid beta-oxidation in a model of depression (Chen et al., 2014). Therefore, the determination of carnitine concentration is crucial to characterize the metabolic status of SZ patients and may be critical for the development of potentially predictive factors for the therapeutic response to antipsychotics. Thus, in this study, we hypothesize that the analysis of carnitine and its derivatives before and after treatment with OLA for 4 weeks may identify critical biomarkers related to SZ. To test the hypothesis, we performed an untargeted metabolomics analysis to find the differential carnitines or its derivatives after treatment with OLA monotherapy for 1 month and to address the association between carnitine levels and treatment response to OLA in drug-naïve first-episode (DNFE) SZ.

2 Methods

2.1 Patients

Twenty-five female DNFE patients with SZ were enrolled, diagnosed via the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV criteria and confirmed by the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID). According to the DSM-IV, a diagnosis of SZ is made if a person has two or more core symptoms, one of which must be hallucinations, delusions, or disorganized speech for at least 1 month. The other core symptoms are gross disorganization and diminished emotional expression. The inclusion criteria include: (1) female; (2) Han Chinese; (3) age from 18 to 45 years; (4) cumulative antipsychotics treatment < 2 weeks; (5) illness duration < 5 years; (6) no substance abuse including alcohol and smoking. A detailed questionnaire including sex, age, education level, smoke status and nationality was created to collect the demographic information. All patients underwent physical examination and laboratorial tests and were inquired for medical history to exclude serious physical conditions, such as diabetes, cardiovascular disease, cerebrovascular disease, cancer and pregnancy.

The procedures of this study were approved by the Institutional Review Board. Written informed consents were obtained by all patients.

2.2 Study procedures

This was a longitudinal observational study with 1-month follow-up. DNFE patients had been treated with a flexible-dose of OLA. The dose range of OLA is from 10 to 30 mg per day according to the patient’s individual response to OLA as determined by the psychiatrist in charge. All patients were hospitalized during the study. Each patient was monitored by the nurses for the adherence to OLA treatment.

2.3 Clinical assessment

The severity of clinical symptoms was evaluated using the Positive and Negative Syndrome Scales (PANSS) by experienced psychiatrists (Kay et al., 1987). PANSS refers to two symptoms of SZ: positive symptoms, which refer to an excess or distortion of normal functioning (e.g., hallucinations and delusions), and negative symptoms, which represent a reduction or loss of normal functioning. Of the 30 items included in the PANSS, 7 constitute a positive scale, 7 is a negative scale, and the remaining 16 is a general psychopathology scale. To ensure consistency and reliability of the evaluation, three psychiatrists attended a training session for the use of PANSS scale following start of the study. The intraclass correlation coefficient (ICC) was computed to examine the test-retest reliability using SPSS 20.0. After training, a correlation coefficient > 0.8 was maintained for the PANSS total score by repeated assessments during the study. At baseline and at 4-week follow-up, the symptom was assessed by PANSS scale. The percentage PANSS total score reduction was calculated by using the formula (Tb − Ta)/Ta × 100%. Ta is the total score at baseline and Tb is the total score at 1-month follow-up. If the percentage total score reduction was more than 30%, we defined the patients as responders.

2.4 Extraction of metabolites and metabolomics data processing

At 7.00 A.M., fasting venous blood samples were collected for each patient at baseline and at 1-month follow-up. Plasma was separated and 200 μl of sample was divided into two parts for the extraction of water-soluble (75% methanol) and lipid-soluble (100% methanol) metabolites. The lipid-soluble and water-soluble supernatants were mixed (1:1), and 0.6 ml of the mixture was dried and re dissolved in 80 μl of 50% methanol and filtered through a 0.1-mm membrane. As described in our previous study, an untargeted metabolomics was used to profile the differential metabolites in patients. Principal component analyses (PCA) were performed using SIMCA-P 13 software (Umetrics). A PLS-DA (partial least squares discriminate analysis) model was used for the calculation of variable importance in projection (VIP) values. Statistical analyses of the quantitative data were performed using SPSS and SAS. Compounds with VIP > 1 and p < 0.05 were picked for further identification. Annotated compounds were identified by searching the accurate mass of the molecular ions and the fragment ions against compound databases. Carnitine was classified in HMDB or Lipidmap databases.

2.5 Statistical analysis

Since majority of carnitines were not normally distributed in patients, a log transformation was used in the analysis. All the following analyses were carried out using SPSS 20.0. Comparisons of the demographic and clinical data between pre and post treatment were conducted by analysis of variance (ANOVA). Pearson correlation was used to analyze the association between the changes in the levels of carnitine and the percentage of PANSS reduction and body weight. The changes in the metabolites, body weight and PANSS score were calculated by the values at follow-up subtract the values at baseline. Further linear regression analysis was conducted to identify the related factors to the changes in the psychotic symptoms and body weight, with the rate of reduction of PANSS total or weight gain as the dependent variables and with demographic data and the changes in carnitine levels as the independent variables. We considered p value < 0.05 as significant.

3 Results

3.1 Clinical symptoms before and after 4-week treatment

Demographic and clinical characteristics of all patients are summarized in Table 1. The average age was 27.4 ± 7.6 years, their mean education years was 9.1 ± 3.5, the mean age of onset was 26.4 ± 8.8 years and the average body mass index (BMI) at baseline was 21.0 ± 3.6 (Table 1).

Table 1 Demographic and clinical data of the patients
Full size table

After 4 weeks of OLA treatment, DNFE patients showed significant improvement in the symptoms, including positive subscore (25.0 ± 6.0 vs 16.4 ± 5.2), general psychopathology subscore (39.7 ± 7.5 vs 32.0 ± 6.0) and total score (81.8 ± 13.9 vs 63.9 ± 13.8) (all pBonferroni < 0.05). However, negative symptoms were not significantly improved after OLA treatment (p > 0.05).

3.2 Metabolomics analysis at baseline and after treatment

As described in our previous study, after screening both with p < 0.05 and VIP > 1, 175 differential compounds were identified (Liu et al., 2021b). The score plot was shown in Supplementary Fig. 1. Among these compounds, we found five types of carnitines, including linoelaidyl carnitine (HMDB06461, m/z = 424.340465), 2-Octenoylcarnitine (HMDB13324, m/z = 286.20031), Butyrylcarnitine (HMDB02013, m/z = 232.15365), 11Z-Octadecenylcarnitine (HMDB13338, m/z = 426.35648) and 9-Decenoylcarnitine (HMDB13205, m/z = 314.23135). The chemical structure of carnitine and the presented derivatives is shown in Supplementary Fig. 2. Of the five types of carnitines, linoelaidyl carnitine is a long-chain carnitines with a fatty ester lipid molecule. Thus, it is very hydrophobic, practically insoluble (in water), and relatively neutral.

After treatment, levels of 2-Octenoylcarnitine decreased and levels of linoelaidyl carnitine, 9-Decenoylcarnitine and 11Z-Octadecenylcarnitine increased significantly (p < 0.05). Butyrylcarnitine levels were not different between baseline and follow-up (p > 0.05).

3.3 Association between baseline carnitine levels and the changes in PANSS total score and body weight

Correlation analysis revealed that baseline linoelaidyl carnitine levels were correlated with the percentage of positive symptoms reduction (r = 0.39, p = 0.05) (Fig. 1). Regression analysis with age, age of onset, education years, and baseline linoelaidyl carnitine level as independent variables showed that baseline linoelaidyl carnitine level was significantly related to the reduction of positive symptoms after treatment (standardized coefficient β = 0.44, t = 2.1, p = 0.046), which accounted for the variance of 20.3%. Baseline linoelaidyl carnitine levels were not correlated with PANSS total score or subscores at baseline and at 1-month follow-up (all p > 0.05). However, no association was found between changes in carnitine level and weight gain or BMI gain (all p > 0.05).

Fig. 1
figure 1

Associations between baseline carnitine levels and symptom improvements in patients with schizophrenia (SZ). a There was a positive association between baseline linoelaidyl carnitine levels and the percentage of positive symptoms reduction in patients with SZ (r = 0.39, p = 0.05). b–e There were no significant associations between baseline levels of 2-Octenoylcarnitine, Butyrylcarnitine, 11Z-Octadecenylcarnitine and 9-Decenoylcarnitine and symptom improvements in patients with SZ

Full size image

3.4 Baseline metabolites and response to treatment

We identified 10 responders and 15 non-responders (Table 2). There is a significant difference in baseline linoelaidyl carnitine levels between non-responders (n = 15) and responders (n = 10) (4.9 ± 0.6 vs 5.8 ± 1.3, F = 6.09, p = 0.021). Further logistic regression analysis was conducted with percentage of total score reduction as dependent variable and with age, education years and age of onset as the independent variables. We found that the baseline linoelaidyl carnitine level was a predictor for 4-week treatment with olanzapine (β = 1.19, Wald X2 = 4.58, p = 0.033, odds ratio = 3.28).

Table 2 Comparisons of clinical symptoms between responders and nonresponders
Full size table

4 Discussion

This study has three main findings. (1) We found that the levels of 2-Octenoylcarnitine, 9-Decenoylcarnitine and 11Z-Octadecenylcarnitine were significantly changed after treatment with OLA for 1 month. (2) Linoelaidyl carnitine level correlated with the percentage of positive symptoms reduction after OLA monotherapy. (3) Baseline linoelaidyl carnitine was a predictive factor for the therapeutic response to OLA monotherapy.

It is the first study to simultaneously investigate the relationship of several baseline carnitine derivatives with the reduction of clinical symptoms after OLA monotherapy for 1 month in DNFE patients with SZ. Previous studies have shown that antipsychotic drugs can significantly increase the levels of acyl-carnitines (Kriisa et al., 2017; Leppik et al., 2020; Molina et al., 2021; Yi et al., 2021). In particular, in animal studies, OLA has been reported to increase the carnitines levels (Albaugh et al., 2012; Jiang et al., 2019). Our findings were also consistent with previous studies in SZ, suggesting that carnitine pathway was regulated after treatment (Kriisa et al., 2017). It should be noted that the study from Kriisa et al. used the targeted metabolomics to identify the differential carnitines in SZ, while in our study, we used the untargeted metabolomics. Therefore, only five carnitines were identified in our study. Kriisa et al. identified 39 carnitines in his study and found antipsychotic drug had an impact on 17 metabolites in DNFE patients. Interestingly, they also found that none of these carnitines were different between DNFE patients at follow-up and healthy controls, indicating that the alteration of carnitine levels returned to control levels after treatment. Taken together, our study provides further evidence that antipsychotic treatments can regulate the levels of carnitine and the derivatives.

SZ has been found to be associated with several abnormal metabolic pathways, including mitochondrial dysfunction, aberrant gut-metabolome-immune network, oxidative stress-related cell damage, changes in embryonic neurogenesis caused by molecular disturbances during development, abnormalities in serotonergic, glutamatergic and dopaminergic signaling, and abnormal sugar metabolism and insulin resistance (Bramon et al., 2008; Campeau et al., 2021; Do et al., 2009; Fan et al., 2022; Geyer & Vollenweider, 2008; Henkel et al., 2022; Lago et al., 2022; Liu et al., 2021a, 2021b; Moore et al., 1999; Yao & Keshavan, 2011; Zhang et al., 2010). Acute and chronic treatment with OLA can alter respiratory chain complexes activity in the rat brain, suggesting that metabolism energy was involved in the treatment with OLA (Agostinho et al., 2011). Carnitine and its derivatives are related to the fatty acid beta-oxidation, metabolic pathways in energy metabolism of mitochondrion, cellular homeostasis pathways, antioxidant, and anti-inflammatory response. Therefore, we speculated that the observed alterations of 2-Octenoylcarnitine, 9-Decenoylcarnitine and 11Z-Octadecenylcarnitine might imply that the changes of these related pathways exist after treatment with OLA in patients with SZ. SZ is a chronic brain disorder and it is known that the brain is composed mainly of fatty acids that need to be incorporated into structural lipids (Nałecz & Nałecz, 1996). Recent studies support that the pathological process of SZ is linked to abnormal lipid metabolism and related pathways in both the central and peripheral nervous systems (Misiak et al., 2017; Prabakaran et al., 2004; Taha et al., 2013), which is intimately associated with carnitine. Interestingly, studies have reported that carnitine deprivation can lead to a wide variety of disorders that affect brain functions (Kimura & Amemiya, 1990; Jones et al., 2010). However, it should be with more cautions in interpreting our findings, considering that any drug or any compound may lead to the modulation of the carnitine derivatives.

We also found that the baseline level of linoelaidyl carnitine was associated with better response to treatment, suggesting a close relationship between carnitine derivatives and OLA treatment in the early treatment stage of SZ. Moreover, our findings demonstrate that the higher levels in linoelaidyl carnitine at baseline, the better response to OLA monotherapy in DNFE patients. Linoelaidyl carnitine (acylcarnitine C18:2) belongs to a long-chain acylcarnitine, one of the derivatives of carnitine metabolism pathway. Many diseases caused by abnormalities in energy production and in intermediary metabolism are characterized by unusual acylcarnitines (Pierre et al., 2007). Mitochondria are the fundamental cellular compartment in lipid metabolism, particularly fatty acid oxidation contributed by acylcarnitine. Studies on SZ show that antipsychotic treatments affect the composition of brain lipids (such as polyunsaturated fatty acids and phospholipids) and lipid homeostasis (Kaddurah-Daouk et al., 2007; McEvoy et al., 2013; Polymeropoulos et al., 2009). The finding of altered acylcarnitines in this study may indirectly indicate changes of mitochondrial homeostasis, considering that the accumulation of acylcarnitine may mean mitochondrial dysfunctions. Particularly, previous studies found that measurement of the qualitative pattern of acylcarnitines can be of diagnostic and therapeutic prediction significance (Millington & Stevens, 2011). Consistent to our findings, aylcarnitines have been reported to be involved in the pathogenesis of SZ and treatment response related to antipsychotics (Kriisa et al., 2017).

However, due to the general features of SZ is not well understood, we do not give an exact explanation for this relationship. Based on previous findings, we speculated several possible reasons for the relationship. One possible explanation is that the shift in oxidative stress and inflammatory status caused by antipsychotics in DNFE may be correlated with the modifications in linoelaidyl carnitine. Linoelaidyl carnitine may reflect the severity of oxidative stress or inflammatory status (Kriisa et al., 2017). Antipsychotic drugs have been reported to directly reduce a low-grade inflammation and oxidative stress status in patients (Haring et al., 2015; Kriisa et al., 2016; Miller et al., 2011; Upthegrove et al., 2014). Specifically, the levels of oxidative stress and inflammatory markers may be altered in the different stages of the illness after treatment with antipsychotics. Previous studies provide ample evidence for the link of alteration of oxidative stress and inflammatory cytokines accompanied by the improvement of clinical symptoms after treatment with antipsychotics in SZ (Li et al., 2021). Another possible explanation is related to the pre-diabetes and glucose dysregulation in DNFE patients with SZ. A previous meta-analysis in SZ concluded that glucose abnormality exists in the early stage of this disorder and is intrinsic to SZ (Pillinger et al., 2017). Mitochondria dysfunction is implicated in the glucose dysregulation and diabetes. The role of acylcarnitines in regulating intracellular insulin signaling has been highlighted in previous studies showing that acylcarnitine induces a decrease in Akt phosphorylation in vivo (Aguer et al., 2015; Blackburn et al., 2020). Recent evidence indicates that carnitine functions in bioenergetics and β-oxidation in mitochondrion. These mechanisms play a critical role in SZ pathogenesis. Moreover, carnitine is involved in many metabolic pathways and acylcarnitine plays an essential role in β-oxidation, oxidative stress status, and neurotransmission. Thus, it is reasonable to conclude that patients with higher baseline linoelaidyl carnitine may have better mitochondria function and display some signs of recovery from the disease after treatment. Together, acylcarnitine has an impact on lipid metabolism, mitochondria functions, insulin sensitivity, inflammatory response and oxidative stress, which provide the underlying mechanism for its involvement in the improvement of positive symptoms.

Our study has several limitations. Firstly, our study includes the relatively small sample size (n = 25) for metabolite analysis using untargeted metabolomic profiling approach. Moreover, this study lacks the statistical power for association and logistic analyses. Secondly, our study volunteers are only female patients in the current study. This study design can help to minimize the sex difference in recruited patients, however, our findings restrict the generalization to male patients. Thirdly, there may be measurement error in the measurement of metabolites, since each patient only were tested once without repetition. Further study should recruit several independent cohorts, including training and test samples.

In summary, we found that several types of carnitine level were significantly altered after OLA treatment for 4 weeks. Moreover, baseline linoelaidyl carnitine level was related to the percentage of positive symptoms reduction and the response to OLA monotherapy. Our findings provide preliminary evidence for the abnormal carnitines in the treatment of SZ and maybe a potentially predictive biomarker linking antipsychotics exposure and response to antipsychotics, which is warranted to be further investigated. However, because we only recruited female patients, caution is warranted in interpreting our findings.