Evaluating the Causal Relationship of Levo-Carnitine and Risk of Schizophrenia: A Bidirectional Two-Sample Mendelian Randomization Study

1. Introduction

Schizophrenia (SZ), a highly incapacitating mental disorder, is estimated to afflict around 1% of the worldwide populace[1], being widely acknowledged as the most severe among all mental illnesses[2]. This complex and multifaceted disorder is typified by significant impairments in language acquisition, working memory, executive function, and processing speed[3]. SZ is a mental disorder with a significant hereditary component, characterized by a wide range of genetic variations including both common single nucleotide polymorphisms (SNPs) and rare mutations[4,5,6]. Furthermore, advancements in molecular biology research have led to the identification of numerous risk genes associated with SZ, as well as genetic evidence supporting familial transmission. These findings further substantiate the role of genetics in the development of SZ and provide compelling evidence for comprehending the susceptibility to SZ from a genetic variation perspective[7,8]. In summary, SZ is a complex disorder influenced by multiple genes, as supported by genetic variants[9]. Therefore, exploring the origins of psychiatric disorders or identifying potential therapeutic targets using the framework of genetic variants has substantial practical implications. Remarkably, despite extensive research conducted over several decades, the underlying pathological mechanisms of SZ still lack comprehensive understanding.

The endogenous pool of carnitine consists of Levo-carnitine (LC) and acylcarnitine, wherein LC serves as the active form for carnitine while acylcarnitine is a derivative states derived from LC[10,11]. LC plays a crucial role in the β-oxidation of fatty acids within the human body[12,13], and its synthesis occurs via biosynthesis, utilizing the amino acids lysine and methionine[14,15]. Among the derivative states in the LC pool, acetyl-LC (ALC), propionyl-LC (PLC), and isovaleryl-LC (ILC) are three types that exhibit significant biological activity and have garnered attention. LC is ubiquitously present in the majority of cells within the human body[16], and assumes a vital role in upholding the integrity of cell membranes and exhibits specific functionalities within mitochondria[17]. Prior investigations have substantiated that an insufficiency of LC results in the enlargement of astrocytes and the expansion of mitochondria in nerve cells[18]. The presence of mitochondrial dysfunction in schizophrenia is substantiated by a confluence of evidence derived from genetic and peripheral investigations[19]. This implies that LC may play a role in the pathological mechanisms underlying SZ, exerting an impact on the structural and energetic metabolism aspects of nerve cells. Furthermore, the study conducted by Kriisa et al. posited that impairment in LC function could potentially give rise to psychiatric disorders, including SZ[20]. In a prospective cohort study, the efficacy of olanzapine, the primary antipsychotic medication prescribed for individuals with SZ, was observed to significantly diminish the concentrations of LC metabolites in patients with SZ. Additionally, the reduction in LC metabolite levels exhibited a significant correlation with cognitive enhancement subsequent to treatment[21]. Acyl-LC is an endogenous compound that is prominently present in various bodily tissues, including muscles, the brain, and sperm. The significance and neuroprotective properties of these carnitines have been extensively validated through contemporary scientific investigations. In light of observational research, the collective evidence substantiates the potential involvement of LC and its derivatives in the etiology, progression, and therapeutic intervention of SZ. [22]Nevertheless, the existence of a definitive causal association between the two remains uncertain.

In recent years, the employment of Mendelian randomization (MR) has facilitated the ability to deduce causal connections between modifiable environmental exposures and outcomes, thus garnering growing attention[23]. MR leverages genetic variation found in SNPs as instrumental variables (IVs) within observational contexts. By virtue of genetic variations being randomly allocated during gamete formation and being unaffected by environmental and lifestyle factors, the estimates derived from MR exhibit reduced vulnerability to confounding bias. Moreover, it is imperative that the individual’s lineage genotype is established prior to examining the outcome of interest, and the measurement of genetic variants must be conducted with utmost accuracy. This meticulous approach in conducting MR analysis reduces the susceptibility to biases arising from reverse causation and measurement errors. Consequently, we have chosen to investigate the causal association between different subtypes of LC, as determined by genetic factors, and the susceptibility to SZ using a two-sample bidirectional MR analysis.

2. Materials and Methods

2.1. Study Design

In order to assess the potential causal relationship between different derivatives of LC and SZ, a Two-sample MR analysis was conducted. This analysis was guided by three fundamental principles[24]: (1) The use of SNPs identified through genome-wide association studies (GWAS) as IVs for the exposure variable; (2) Ensuring that the IVs are not associated with confounding factors; (3) Confirming that the IVs solely influence the risk of the outcome through their association with the exposure, without exerting a direct impact on the outcome itself. The forward MR analysis primarily aimed to examine the correlation between LC and its derivatives as the exposure, and SZ as the primary outcome. Conversely, in reverse analysis, SZ was considered as the exposure variable, while LC and its derivatives were regarded as the outcome variables.

Figure 1. Principles of bidirectional mendelian randomization study and our study design. MR, Mendelian randomization; GWAS, genome-wide association studies. p, r², and kb represent the parameters in the “TwoSample” R package.

Figure 1. Principles of bidirectional mendelian randomization study and our study design. MR, Mendelian randomization; GWAS, genome-wide association studies. p, r², and kb represent the parameters in the “TwoSample” R package.

2.2. Data Sources

For the purpose of forward MR analysis, the keywords "carnitine" and "schizophrenia" were employed to retrieve data from the OpenGWAS database (https://gwas.mrcieu.ac.uk/). The research data consisted of the European population with the most substantial sample size or number of cases, where LC and its derivatives data were considered as the exposure, and SZ related data served as the outcome. Finally, a total of[25] datasets pertaining to LC and its derivative states, along with one data set specifically addressing SZ. The detailed information for each data set can be found in Table s1 [26,27].

2.3. Identified Ivs

Firstly, IVs were selected from the exposure data using a quality control procedure. The “TwoSample” R package in the R software was utilized to extract IVs associated with exposure by setting parameters (P=5e-8, clump=True, r2=0.001, kb=10000)[28]. Due to the inherent uncertainty regarding the causality of SZ, potential confounding variables related to exposure-associated IVs were not adequately controlled for, and we employed a filtration process to eliminate and exclude confounding variables specifically linked to SZ utilizing "phenoscanner" R package. Hence, the efficacy of IVs was assessed by examining the F statistics, which quantifies the level of association between the SNPs and the exposure. Subsequently, relevant information pertaining to IVs closely aligned with exposure was extracted from the outcome data set. After harmonized the extracted data, a screening process was implemented to exclude IVs that were closely associated with outcomes (P<5e-8) from the data. In order to clarify the potential impact of outliers in IVs on the MR analysis results, the MR-PRSSO package[29] was utilized to verify the presence of outliers in the harmonized data and subsequently eliminate them.

2.4. MR analysis

After completing the aforementioned procedures, we utilized five distinct methodologies (MR-Egger[30], weighted median[31], IVW[32], simple mode[33], and weighted mode[33]) to evaluate the overall effect outcomes during the MR analyses. The IVW approach was prioritized for result interpretation. It is crucial to ensure that the outcomes derived from these five MR analyses demonstrate consistent positive or negative effects, which can be determined by examining the b value. A P value below 0.05 signifies statistical significance within the framework of MR analysis.

2.5. Sensitivity Analysis

Sensitivity analysis incorporates the examination of heterogeneity and pleiotropy, as well as the assessment of the impact of an individual SNP on the overall outcomes of MR analysis. Heterogeneity was evaluated using ME-Egger methods, specifically by implementing the Q-test[25]. In instances where heterogeneity was detected, a random effects model was utilized to reevaluate the magnitude of the effect. Moreover, the representation of heterogeneity was achieved by employing Scatter plot and Funnel plot techniques. Additionally, the presence of horizontal pleiotropy in the IVs was assessed using the Egger intercept method. Furthermore, the influence of excluding a single SNP on the results of the MR analysis was examined through the application of the leave-one-out method.

2.6. Evaluation of sample overlap bias

In order to further substantiate the influence of sample overlap on the outcomes of MR analysis, we employed the mrmsample overlap R software package to assess the bias and Type I error associated with varying levels of sample overlap rates. If the findings of this examination demonstrate that the bias and Type I error persist at a relatively consistent level despite an increase in the repetition rate of samples, it indicates the relative robustness of the MR analysis results.

3.1. Forward analysis

3.1.1. Forward MR analysis

Based on the findings from the forward MR analysis, it is apparent that LC and its derivatives demonstrate inconsistent causal effects on SZ, as indicated in Table 1. Among the results from three primary derivatives of LC (ALC, PLC, ILC), only ILC exhibits a negative causal relationship with SZ (Table 1, OR=0.435, 95% CI: 0.247-0.765, P_IVW=0.004). Conversely, the remaining substances do not exhibit a significant causal relationship with SZ(Table 1, P_IVW>0.05). Detailed information regarding the IVs employed in the analysis program can be found in the Supplementary Materials 1.

3.1.2. Forward sensitivity analysis

The sensitivity analysis results provided confirmation that there was no statistically significant presence of pleiotropy or heterogeneity in the IVs associated with each groups during the forward MR analysis process (Table s2, P>0.05). Furthermore, the scatter plot distributions indicated potential linear associations within all three analysis groups (Figure 2A–C). Moreover, the absence of conspicuous abnormal SNP distributions was also verified through examination of the funnel plot (Figure 3A–C). The varying impact of individual IVs on SZ is evident across different IVs ( Supplementary Materials 2). However, employing the leave-one-method for evaluation substantiates that the influence of a single IVs on the main findings of MR analysis remains insignificant (Supplementary Materials 3). These findings collectively affirm the robustness and dependability of the analytical outcomes in the present study.

3.1.3. Sample overlap bias in forward analysis

The results indicated a slight increase in both bias and type Ⅰ error with an increasing rate of sample duplication. However, the overall level remained relatively stable (Supplementary Materials 4). These findings suggest that there is no significant duplication of samples among the population included in this study, thereby reinforcing the reliability and validity of our MR analysis results.

3.2. Reverse analysis

The findings from the reverse MR analysis suggest that there is no statistically significant causal association between schizophrenia and L-carnitine and its derivatives (Table 2, P>0.05). Furthermore, the sensitivity analysis results support the absence of significant pleiotropy and heterogeneity among the instrumental variables in each groups (Table s3, P>0.05). The diverse effects of individual IVs on SZ are apparent when considering different IVs (see Supplementary Materials 2). Nevertheless, employing the leave-one-out method for evaluation confirms that the impact of a single IV on the primary outcomes of MR analysis remains statistically insignificant (see Supplementary Materials 3). Additionally, the examination of sample overlap’s influence on the results of the MR analysis yielded comparable findings to the forward analysis (see Supplementary Materials 4).

Table 3. Reverse MR results of causality of SZ on LC and its derivatives. N._SNP, number of single nucleotide polymorphisms; OR, odds ratio; P_IVW, P for inverse-variance weighted method; OR, odds ratio; CI, confidence interval.

Outcome	N.SNP	PIVW	OR	95%CI
Carnitine	10	0.74	1.01	0.99-1.01
Palmitoylcarnitine	2	0.79	0.99	0.94-1.05
Acetylcarnitine	8	0.32	0.99	0.97-1.01
Isovalerylcarnitine	4	0.55	0.99	0.96-1.02
Hexanoylcarnitine	3	0.94	1.02	0.97-1.03
Butyrylcarnitine	3	0.80	0.99	0.95-1.04
Propionylcarnitine	9	0.37	0.99	0.98-1.01
3-dehydrocarnitine	5	0.79	1.01	0.98-1.02
Isobutyrylcarnitine	4	0.78	0.99	0.96-1.03
Octanoylcarnitine	4	0.69	1.01	0.97-1.04
Decanoylcarnitine	2	0.79	1.01	0.96-1.05
Stearoylcarnitine	3	0.84	1.01	0.97-1.04
Laurylcarnitine	1	0.88	1.01	0.93-1.09
Oleoylcarnitine	6	0.88	1.01	0.98-1.02
X-13431--nonanoylcarnitine	3	0.89	1.01	0.96-1.04
2-methylbutyroylcarnitine	6	0.88	0.99	0.97-1.02
Hydroxyisovaleroyl carnitine	7	0.98	1.00	0.97-1.02
Glutaroyl carnitine	3	0.64	1.00	0.98-1.02
2-tetradecenoyl carnitine	4	0.83	1.00	0.97-1.04
Succinylcarnitine	1	0.82	0.99	0.97-1.03
Cis-4-decenoyl carnitine	4	0.62	1.01	0.98-1.04

Table 3. Reverse MR results of causality of SZ on LC and its derivatives. N._SNP, number of single nucleotide polymorphisms; OR, odds ratio; P_IVW, P for inverse-variance weighted method; OR, odds ratio; CI, confidence interval.

Outcome	N.SNP	PIVW	OR	95%CI
Carnitine	10	0.74	1.01	0.99-1.01
Palmitoylcarnitine	2	0.79	0.99	0.94-1.05
Acetylcarnitine	8	0.32	0.99	0.97-1.01
Isovalerylcarnitine	4	0.55	0.99	0.96-1.02
Hexanoylcarnitine	3	0.94	1.02	0.97-1.03
Butyrylcarnitine	3	0.80	0.99	0.95-1.04
Propionylcarnitine	9	0.37	0.99	0.98-1.01
3-dehydrocarnitine	5	0.79	1.01	0.98-1.02
Isobutyrylcarnitine	4	0.78	0.99	0.96-1.03
Octanoylcarnitine	4	0.69	1.01	0.97-1.04
Decanoylcarnitine	2	0.79	1.01	0.96-1.05
Stearoylcarnitine	3	0.84	1.01	0.97-1.04
Laurylcarnitine	1	0.88	1.01	0.93-1.09
Oleoylcarnitine	6	0.88	1.01	0.98-1.02
X-13431--nonanoylcarnitine	3	0.89	1.01	0.96-1.04
2-methylbutyroylcarnitine	6	0.88	0.99	0.97-1.02
Hydroxyisovaleroyl carnitine	7	0.98	1.00	0.97-1.02
Glutaroyl carnitine	3	0.64	1.00	0.98-1.02
2-tetradecenoyl carnitine	4	0.83	1.00	0.97-1.04
Succinylcarnitine	1	0.82	0.99	0.97-1.03
Cis-4-decenoyl carnitine	4	0.62	1.01	0.98-1.04

4. Discussion

To determine if there is a bidirectional causal relationship between L-carnitine’s subtypes and its related metabolites and schizophrenia, two-sample MR analysis was performed. Our findings provide further depth to prior observations of metabolic disturbances in schizophrenia, as L-carnitine is an essential component of beta-oxidation of fatty acids, essential for the production of cellular energy. The results indicate a strong negative relationship between ILC and schizophrenia, suggesting that higher levels of ILC may protect against schizophrenia onset or severity. We also found no causal link between any L-carnitine and schizophrenia in the reverse two-sample MR analysis, demonstrating the stability of our experiments.

Previous research has predominantly concentrated on the broad concept of carnitine and its relevance to neurological disorders. It has been previously ascertained that individuals with schizophrenia have a predisposition to metabolic irregularities and bioenergetic dysfunction[34]. Notably, Cao B. et al[35,36,37] have implicated ALC in the cellular bioenergetic abnormalities observed in schizophrenia, suggesting that acyl-carnitines may serve as prospective subjects for future investigations into their involvement in the pathoetiology of schizophrenia.

According to the study conducted by M.A. VIRMANI and R. BiSELLIT et al[38], it was observed that L-carnitine and ALC possess the ability to impede neurotoxicity resulting from various forms of mitochondrial dysfunction, such as uncoupling or inhibition of oxidative phosphorylation. In a separate investigation by M. Pennisi, G. Lanza et al[39], the involvement of ALC in dementia was explored, revealing its potential in decelerating cognitive decline. The underlying mechanisms behind this phenomenon may encompass the restoration of cellular membrane and synaptic functionality, promotion of mitochondrial energy metabolism, safeguarding against toxins, and exertion of neurotrophic effects. According to preclinical and laboratory findings, ALC has been investigated in human clinical trials as an adjunctive therapy for patients with dementia or MCI[40,41], geriatric depression[42,43], and various other prevalent medical conditions. These trials provide support for the notion that certain forms of L-carnitine supplementation may have a mitigating effect. Consequently, these pieces of evidence imply that L-carnitine and its derivatives possess extensive therapeutic possibilities in the management of schizophrenia.

Previous observational studies have demonstrated that the activity of nerve growth factor (NGF) significantly impacts the maturation of neurons and sustains the differentiation status of diverse neurons in the peripheral and central nervous systems. In their experiment, J. W. Pettegrew et al. discovered that aged rats treated with ALC exhibited amelioration of the age-related decline in NGF-binding capacity observed in the hippocampus and basal forebrain regions. This finding implies that specific sublines of L-carnitine may enhance the responsiveness of neurons to neurotrophic factors within the central nervous system in older rats. Furthermore, ALCAR exerts an impact on the hypothalamic-pituitary-adrenal (HPA) axis within the brain, thereby potentially mitigating pathological brain deterioration under stressful circumstances[44]. Certain variants of L-carnitine additionally modulate the activity of nerve growth factor (NGF) and various hormones, while also playing a role in regulating synaptic morphology and transmission of diverse neurotransmitters, including acetylcholine[45,46]. These findings suggest that specific sublines of L-carnitine possess the ability to affect multiple targets within the central nervous system. Additionally, scientific evidence substantiates the notion that L-carnitine can influence concentrations of neurotransmitters. Based on the findings reported by various sources[47], the utilization of ALC in a child injury model has been shown to enhance mitochondrial function, mitigate brain swelling, and prevent tissue loss[48,49,50]. Furthermore, sustained administration of ALC has demonstrated the potential to enhance the cerebral energy state in healthy mice[51]. Additionally, primary cilia play a crucial role in minimizing oxidative stress and safeguarding neuronal cells41. Building upon the aforementioned studies, our investigation delved into the causal association between select sublines of L-carnitine and schizophrenia. Our findings revealed a consistent and noteworthy inverse correlation between ILC and schizophrenia, thereby bolstering our hypothesis that certain sublines of L-carnitine confer a safeguarding influence against schizophrenia.

The study’s merits encompass the utilization of two-sample MR analysis, which, in contrast to observational studies, can effectively mitigate the influence of confounding variables and reverse causality. Leveraging publicly accessible GWAS summary statistics data, we derived advantages from a substantial sample size, thereby augmenting the accuracy of our estimations and the statistical robustness of our discoveries. This methodological framework guarantees a heightened level of result reliability.

However, it is imperative to acknowledge the constraints inherent in our study. Firstly, our findings predominantly apply to populations of European ancestry. Although this approach may alleviate biases stemming from population stratification, the applicability of our results to other ethnic groups is yet to be ascertained.

Secondly, it is important to acknowledge that our analysis, like any other MR study, is susceptible to potential unobserved pleiotropy, which may introduce biases into our conclusions. Therefore, it is crucial to conduct additional comprehensive research, incorporating various methodological approaches and diverse populations, to validate and enhance the accuracy of our findings.

5. Conclusions

Our study aims to evaluate the causal association between different subtypes of L-carnitine and related metabolites, as determined by genetic factors, and the risk of developing schizophrenia. This analysis utilizes a two-sample Mendelian randomization approach, which reveals a statistically significant inverse correlation between ILC levels and the occurrence of schizophrenia. These results suggest that higher levels of ILC may be associated with a decreased risk of developing schizophrenia. Consequently, our findings suggest that ILC may have potential implications in the prevention and treatment of schizophrenia, thereby providing a novel avenue for future research exploring the relationship between metabolism and this psychiatric disorder.