The association of tobacco smoking and metabolite levels in the anterior cingulate cortex of first-episode psychosis patients: A case-control and 6-month follow-up 1 H-MRS study

Tobacco smoking is highly prevalent among patients with psychosis and associated with worse clinical outcomes. Neurometabolites, such as glutamate and choline, are both implicated in psychosis and tobacco smoking. However, the specific associations between smoking and neurometabolites have yet to be investigated in patients with psychosis. The current study examines associations of chronic smoking and neurometabolite levels in the anterior cingulate cortex (ACC) in first-episode psychosis (FEP) patients and controls. Proton magnetic resonance spectroscopy ( 1 H MRS) data of 59 FEP patients and 35 controls were analysed. Associations between smoking status (i.e., smoker yes/no) or cigarettes per day and Glx (glutamate + glutamine, as proxy for glutamate) and total choline (tCh) levels were assessed at baseline in both groups separately. For patients, six months follow-up data were acquired for multi-cross-sectional analysis using linear mixed models. No significant differences in ACC Glx levels were found between smoking (n = 28) and non-smoking (n = 31) FEP patients. Smoking patients showed lower tCh levels compared to non-smoking patients at baseline, although not surving multiple comparisons correction, and in multi-cross-sectional analysis (p FDR = 0.08 and p FDR = 0.044, respectively). Negative associations were observed between cigarettes smoked per day, and ACC Glx (p FDR = 0.02) and tCh levels (p FDR = 0.02) in controls. Differences between patients and controls regarding Glx might be explained by pre-existing disease-related glutamate deficits or alterations at nicotine acetylcholine receptor level, resulting in differences in tobacco-related associations with neurometabolites. Additionally, observed alterations in tCh levels, suggesting reduced cellular proliferation processes, might result from exposure to the neurotoxic effects of smoking.


Introduction
With a prevalence of almost 70 %, tobacco smoking among patients with schizophrenia is 2-3 times higher compared to the general population (de Leon and Diaz, 2005;Vermeulen et al., 2019).Smoking in schizophrenia is associated with worse clinical outcomes, including readmission (Kagabo et al., 2019).Shared genetic variance between smoking behaviour and schizophrenia has been identified at the Cholinergic Receptor Nicotinic Alpha (CRNA) gene cluster, which encodes the nicotine acetylcholine receptor (nAChR) in the brain (Hartz et al., 2018;Ohi et al., 2019).Nicotine, the primary addictive substance in tobacco, exerts its effects on the brain through the activation of nAChRs, and chronic tobacco exposure causes upregulation of these receptors (Gentry and Lukas, 2002).Positron emission tomography (PET) studies and post-mortem ligand binding assays investigating nAChR deficits in the brain of smoking patients with schizophrenia show a reduction in α7 and α4β2 subtypes of the nAChR, likely representing down-regulation, and impairment of the upregulation seen in healthy controls of α4β2 receptors in patients (Adams and Stevens, 2007;Breese et al., 2000;Durany et al., 2000;Wong et al., 2018).However, the exact aetiology underlying the smoking-schizophrenia association remains unclear.
Few neuroimaging studies have explored the association of tobacco and the brain in psychosis.These studies suggest that smoking in combination with psychosis is associated with more pronounced structural and functional alterations compared to non-smoking patients (for a brief review, see (Smucny and Tregellas, 2017)).The anterior cingulate cortex (ACC) is involved in smoking and schizophrenia.The ACC integrates information between limbic regions and cortical networks, thereby regulating emotions and reward-based decision-making (Rolls, 2019).These processes underlie the motivational value of drugs (Ashare et al., 2014;Daly et al., 2016) and are thus essential in tobacco smoking addiction.In schizophrenia, alterations in the ACC such as reduced synaptic density and grey matter are consistently reported (Fornito et al., 2009;Roberts et al., 2015).The ACC is part of the salience network, which is thought to be disrupted in schizophrenia (White et al., 2010).Moreover, nAChR expression is altered in both schizophrenia and smoking addiction, and these receptors are densely expressed in the ACC (Picard et al., 2013).These shared ACC abnormalities may contribute to the development or persistence of smoking in patients with schizophrenia.
Considering altered ACC neurometabolite levels in schizophrenia and tobacco smoking, it is important to investigate their interrelationship and how this interaction may impact the development or manifestation of schizophrenia.Furthermore, research examining the impact of smoking duration on the brain of patients through longitudinal analyses is lacking.Longitudinal studies could help clarify inconsistent results, and ascertain the impact of cumulative neurotoxic effects due to prolonged smoking exposure on patients with psychosis.To our knowledge, this is the first study to investigate the association of chronic tobacco smoking and ACC Glx (as a proxy for glutamate) and tCh levels of first-episode psychosis (FEP) patients, using a longitudinal design.In addition, this association is evaluated in a control sample.Furthermore, this study evaluates the association of alcohol and cannabis use and neurometabolites considering their strong correlation with tobacco use (Agrawal et al., 2012;Falk et al., 2006).
We hypothesized that smoking controls show higher Glx and lower ACC tCh levels than non-smoking controls.More importantly, considering the pre-existing aberrant glutamatergic and cholinergic systems in schizophrenia, we expect these differences to be more pronounced between smoking and non-smoking patients than in controls, and these effects to progressively increase with time due to both the progression of the disease and the prolonged exposure to the neurotoxic effects of smoking.

Subjects
Data from two studies were collected (van der Pluijm et al., 2024;van Hooijdonk et al., 2024).FEP patients with a primary diagnosis in the schizophrenia spectrum were recruited from the specialized Early Psychosis Clinic at the Amsterdam Medical Centre (AMC) and two Dutch mental healthcare institutions (GGZ Rivierduinen and GGZ inGeest).Controls were recruited through online advertisement.Groups were matched for age, sex, and smoking status (i.e., smoker yes/no).All subjects were between 18 and 35 years old and had a good knowledge of the Dutch language.Exclusion criteria included antipsychotic medication use ≥1 year, use of hard-or psychotropic drugs such as amphetamine or attention deficit hyperactivity disorder medication, benzodiazepine use the night before or on day of the magnetic resonance imaging (MRI) scan, (history of) substance dependence (excluding tobacco, alcohol, and cannabis), inability to provide informed consent, a neurological disorder or brain damage, and MRI contraindications (including pregnancy).In addition, controls were not allowed to have (a history of) a psychiatric disorder, or first-degree relative with a psychotic disorder.Studies were approved by the Medical Ethics Committees of the AMC and Leiden-Den Haag-Delft.Participants provided informed written consent before participation, were screened for drug use through a urine test before the MRI scan, and received monetary compensation.

Study design and measures
Subjects had a baseline measurement, including a clinical interview coveringdemographics and medication use, and an MRI scan.In patients, symptom severity was measured using the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987).Controls underwent the Mini-International Neuropsychiatric Interview to ensure absence of psychiatric disorders (Sheehan et al., 1998).Alcohol and drug use (including tobacco) were assessed using the Composite International Diagnostic Interview (CIDI) (Andrews and Peters, 1998).Only patients were invited for a follow-up (FU) visit 6 months after baseline, which consisted of a clinical interview and an MRI scan.Participants were included as smokers when they smoked daily ≥1 month in the year before study participation, or non-smokers when they reported current complete abstinence from smoking.Further, participants were defined as cannabis users if they smoked cannabis in the past year and alcohol users if they consumed ≥12 alcoholic beverages in the past year.Information on cigarettes smoked per day at peak intensity, weeks cannabis use, and the average alcoholic beverages per week in the past year were also collected.

1 H-MRS data processing
Spectra were fitted and quantified using Linear Combination Model (LCModel) version 6.3-1L (Provencher, 2001), with the LCModel standard basis set for PRESS at 35 ms.The basis set was composed of alanine, aspartate, creatine, GABA, glucose, glutamine, glutamate, GPC, PCh, lactate, myo-inositol (mI), NAA, N-acetyl-aspartyl-glutamate (NAAG), scyllo-Inositol, taurine, and GSH.Spectra were fitted between 0.2 and 4.2 ppm.The node spacing for the spline function (DKNTMN) was set to 0.5 ppm to stiffen the baseline (Bhogal et al., 2017).Metabolites were corrected in reference to the unsuppressed water peak and quantified in institutional units [I.U.].In addition to Glx and tCh, we explored the association of smoking and other neurometabolites, including total creatine (tCr, creatine and phosphocreatine), mI, GSH, and total NAA (tNAA, NAA, and NAAG).We report on these neurometabolites in our supplement.
Acquired spectra were visually inspected to check for low data quality.Spectra with Full Width at Half Maximum (FWHM) ≥ 0.1 ppm, Cramer-Rao lower bound (CRLB) ≥ 15 % for Glx or tCh, or signal-tonoise ratio (SNR < 22; based on total mean SNR + 2*SD (Egerton et al., 2012)) were considered of poor quality and excluded from analysis (Li et al., 2020).

Covariates
A priori, age (Durazzo et al., 2015) and sex (Hellem et al., 2015) were selected as confounders for all analyses, as they potentially influence neurometabolite levels.Alcoholic drinks per week (Falk et al., 2006) and weeks of cannabis use (Agrawal et al., 2012) in the past year were added as covariate in all analyses, due to their strong correlation with tobacco smoking, and cannabis its association with glutamate changes in the cingulate cortex (Watts et al., 2020;Zuo et al., 2022;Zuo and Lukas, 2023).

Statistical analysis
Statistical analyses were performed in R version 4.3.2using RStudio (version 2023.12.1, RStudio Inc.).If metabolite data were not normally distributed, a square-root transformation was applied for statistical analysis.Differences were considered statistically significant at p < 0.05.
We first compared Glx and tCh levels across all four groups at baseline using an analysis of covariance (ANCOVA) with post hoc Tukey-Kramer test.Due to significant tCh differences between patients and controls (Table S4), further analyses were performed within patients and controls separately.This approach removed inherent group differences as confounding factors, ensuring that findings are more accurate, and directly applicable and clinically relevant to each group.Thus, ANCOVAs with age, sex, and alcohol and cannabis use as covariates were used to evaluate the association between smoking status and metabolite levels in both groups.Linear regression analyses were conducted to investigate associations between cigarettes smoked per day and metabolite levels.
For patients, linear mixed-effects models were performed as multicross-sectional analysis to assess associations between smoking status and metabolite levels across baseline and FU.We chose this approach to enable analyzing data from multiple time points concurrently, whilst accounting for correlations between repeated measurements.Smoking status and covariates (age, sex, weeks of cannabis use, and alcoholic drinks per week) were added as fixed effects.Intercepts for subjects and random slopes for time were included as random effects.Models were fitted using maximum likelihood estimation (Harrison et al., 2018), and continuous variables were centred to improve model performance and interpretability (Bolger and Laurenceau, 2013).A step-by-step procedure was employed to incorporate each variable, whilst comparing the model fit using the Akaike information criterion.Patients' data were included if data were available for at least one timepoint on the outcome variable of interest and smoking characteristic, as mixed modelling enables calculating valid estimates of missing data under the assumption of missing at random even if data for one timepoint were missing.Additional mixed model analyses were conducted with cigarettes smoked per day as independent variable to investigate a dose-response relationship.In all mixed model analyses, p-values were calculated using the Satterthwaite method.To address multiple testing, false discovery rate (FDR) correction (Benjamini-Hochberg) was applied to the two main metabolites of interest (Glx and tCh) in all analyses.
Exploratory analyses were performed at baseline to evaluate associations between metabolite levels and cannabis/alcohol status using separate ANCOVAs.Linear regression analyses were run with weeks of cannabis use or average alcoholic beverages per week in the past year as independent variables.Lastly, most patients stayed in the psychiatric clinic at the time of the scan, which potentially constrained their smoking behaviour.As tobacco smoking has short-lasting and reversible effects on nAChRs in otherwise healthy subjects (Goriounova and Mansvelder, 2012), we performed sensitivity analyses excluding patients with altered smoking behaviour during their stay.

Demographics and baseline characteristics
Seventy-three patients and 38 controls were scanned.Fifty-nine patients and 35 controls were included in baseline analysis (see Fig. S1 for reasons for exclusion).Thirty-nine patients underwent the FU MRI scan, 35 of whom were included in analysis.Except for one patient who smoked daily for two weeks in the year before participation, all nonsmokers reported complete abstinence from smoking for ≥1 year.Demographical and clinical characteristics of the final sample after quality control are listed in  4 Chi-squared test. 5Level of education was scored on an 8-point scale ranging from 0 (no education) to 7 (Master's degree or equivalent). 6At FU, the study sample consisted of 35 patients, of which 17 (49 %) were smokers.
described in Table S2.
At baseline, no significant differences were found between smokers and non-smokers within both the patient and control group concerning age, sex, current alcohol use, alcoholic beverages per week, antipsychotic medication dosage (patients only), and the negative and general subdomains of the PANSS (patients only) (Table 1).Smoking patients more often used cannabis (p < 0.001) and for more weeks in the past year at baseline (p < 0.001).In addition, smoking patients had a lower level of education compared to non-smoking patients (p = 0.03).At FU, smoking patients scored higher on the negative subscale of the PANSS (p = 0.029) and more often used cannabis than non-smoking patients (p = 0.01).All patients used antipsychotic medication, and 3 non-smoking patients used concurrent antidepressant medication.Twenty-eight (47 %) patients and 12 (34 %) controls at baseline, and 17 (49 %) patients at FU reported daily smoking within the last 12 months.Smoking patients and controls smoked on average 13.2 ± 6.04 and 13.5 ± 9.05 cigarettes per day at baseline, respectively.

Spectral quality
GSH levels were not analysed, as approximately half of the spectra were of poor quality (Table S3).There were no significant differences in SNR, FWHM, or CRLB for any of the metabolites (i.e.tCr, Glx, mI, tCh, and tNAA), or voxel fractions (white matter, grey matter, or cerebrospinal fluid) between smoking and non-smoking subjects within patients or controls at any time point (Table S3). 1 H-MRS voxel placement was visually inspected and consistent across subjects (Fig. 1A).For statistical analysis, myo-inositol was transformed with a square-root transformation. tCh

Baseline analysis in patients
Smoking patients exhibited lower ACC tCh levels compared to nonsmokers, although this difference did not survive correction for multiple comparisons (F(1,53) = 4.41, p FDR = 0.082).No significant group differences were observed between smoking and non-smoking patients or associations between cigarettes per day and metabolite levels were observed for Glx (Fig. 2A, B, and Table S5).See Table S5 for results on exploratory metabolites.

Baseline analysis in controls
No significant differences between smoking and non-smoking controls were found for Glx or tCh (Fig. 2A and Table S6).Negative associations were found between cigarettes per day and Glx (t(27) = − 2.40, p FDR = 0.02) and tCh (t(27) = − 2.67, p FDR = 0.02; Fig. 2B and Table S6).See Table S6 for results on exploratory metabolites.

Multi-cross-sectional analysis in patients
To assess associations across time points between smoking status and metabolite levels in patients, we conducted linear mixed-effects models with 59 baseline and 35 FU measurements.A main effect of smoking status on tCh levels was observed across baseline and FU with smoking patients displaying lower tCh levels (p FDR = 0.044) (Table 2).No significant effects were observed for Glx.See Tables S7 and S8 for results on exploratory metabolites.
No significant associations were found between (weeks of) cannabis use and any of the metabolites for both patients and controls (Tables S11  and S12).

Sensitivity analyses
Sensitivity analyses were performed using a subsample of patients (n = 20) that smoked as usual during their stay in the clinic and excluding the non-smoking patient who reported smoking for two weeks in the past.No significant differences were demonstrated between smoking and non-smoking patients for Glx or tCh (Tables S13-S15).

Discussion
To our knowledge, this is the first study examining associations between tobacco smoking and neurometabolite levels in controls and FEP patients over 6 months.We observed no significant differences in ACC Glx levels between non-smoking and smoking patients.Across baseline and FU, smoking patients displayed lower tCh levels than non-smokers.Further, we demonstrated negative associations between cigarettes smoked per day, and ACC Glx and tCh levels in controls.Group analysis showed increases in ACC tCh, including smoking patients versus smoking controls, and non-smoking patients versus smoking and nonsmoking controls.
Our finding of a negative association between cigarettes per day and ACC Glx levels in controls contradicts earlier findings of higher Glx levels of smoking subjects (Mennecke et al., 2014;Schulte et al., 2017).Noteworthy, the voxel placement of Mennecke et al. (2014) is more posterior than ours and voxel placement can influence neurometabolite concentrations (Nakahara et al., 2022;Simmonite et al., 2023), whilst Schulte et al. (2017) employed a similar voxel placement to our study.However, lower glutamate levels have been reported in the dorsolateral prefrontal cortex of daily smokers (Durazzo et al., 2015;Faulkner et al., 2021).Tobacco activates nAChRs, resulting in glutamate release in the brain (D' Souza and Markou, 2013).Presynaptic inhibitory mGluR2/3 receptors may partially compensate for chronic nAChR activation by glutamate in smokers (Markou, 2007), potentially reducing the excitatory effects of nAChR activation, and decreasing glutamate transmission in smokers.Supporting this hypothesis, increased mGlu2/3 receptor activity has been observed in tobacco-dependent rats (Markou, 2007).Unfortunately, MRS does not provide insights into glutamate function at receptor level, as it measures the combined intra-and extracellular levels of neurometabolites in the voxel.
We expected more pronounced differences in patients, yet no significant Glx differences were demonstrated.This might be attributed to disease-specific disrupted glutamate signalling (Duarte and Xin, 2019;Uno and Coyle, 2019;Wijtenburg et al., 2015) and nAChR changes.Abnormalities in glutamate signalling in schizophrenia are hypothesized to result from divergent nAChR signalling (Wing et al., 2012).Patients with schizophrenia show genetic variance at the CRNA gene cluster, which encodes for nAChR (Ohi et al., 2019), and reduced nAChR expression (Parikh et al., 2016).These receptors regulate the release of many neurotransmitters including glutamate (Subramaniyan and Dani, 2015;Wonnacott, 1997).Chronic tobacco exposure results in desensitization (Pidoplichko et al., 1997), inactivation, and upregulation of these receptors (Gentry and Lukas, 2002).Notably, the upregulation of high-affinity nAChRs was observed in the cortex, thalamus, hippocampus, and caudate nucleus of healthy smokers but not in smokers with schizophrenia (Breese et al., 2000;D'Souza et al., 2012).These diseasespecific differences at nAChR level between healthy and patient smokers may underlie the differences observed between patients and controls in the association between tobacco and glutamate.The divergent neurobiological profile of patients with schizophrenia could influence how these patients respond to chronic tobacco exposure.As the nAChR system might represent a common substrate for schizophrenia and tobacco dependence, future research could investigate this hypothesis by focussing on nAChR (dys)functioning, e.g. using [ 18 F]ASEM PET imaging (Coughlin et al., 2019;Wong et al., 2018).Shifting our focus away from Glx and the nAChR system, we observed, in line with earlier research (Yamasue et al., 2002), higher ACC tCh levels in patients compared to controls.Further, smoking patients exhibited lower tCh levels compared to non-smokers, and across baseline and FU smoking patients displayed lower tCh levels.Moreover, controls show significant negative associations between cigarettes per day and tCh levels, consistent with earlier research demonstrating significantly decreased ACC tCh levels in smokers compared to nonsmokers (Mennecke et al., 2014;O'Neill et al., 2023).The tCh signal mostly arises from GPC and PCh, related to membrane phospholipids and proliferation processes, with a small possible contribution from free choline (Rae, 2014).Only free choline is a acetylcholine precursor, making abnormalities in tCh levels a different mechanism from abnormalities in nAChR function.These multiple factors influencing tCh levels complicates interpreting tCh changes.Decreased tCh levels suggest reduced cell density, and membrane turnover and synthesis (Rae, 2014), indicating a decrease in cellular proliferation processes.Speculatively, these decreases might be due to prolonged exposure to the neurotoxic effects of smoking, linked to impaired neuroinflammation (Brody et al., 2017) and increased oxidative stress (Liu et al., 2020), leading to damage to proteins and membranes (Liguori et al., 2018).
Drinking patients showed significantly higher ACC Glx levels compared to non-drinking patients, which was not observed in controls.Noteworthy, only four controls reported not consuming alcohol.Previous research demonstrated that persistent and excessive alcohol consumption reduces glutamate levels in the dorsal ACC of otherwise healthy individuals (Prisciandaro et al., 2016).Even though subjects in the current study only engaged in moderate alcohol consumption (Table 1), alcohol consumption may have a more pronounced impact on the already disrupted glutamate system in patients compared to controls.We observed no significant differences in alcohol status or drinks per week between smoking and non-smoking patients, and drinks per week were included as covariates in all analyses.Consequently, we do not expect that alcohol consumption significantly influenced differences between smokers and non-smokers.
To address potential effects due to changes in smoking behaviour, sensitivity analyses were performed in a subsample of patients that continued to smoke as usual during their stay in a psychiatric clinic.These analyses showed similar results, validating our initial results.See supplementary discussion S1 for the discussion of findings of the exploratory metabolites.
The current study was limited by the lack of detailed information on smoking behaviour and cannabis use.Specific details such as the timing of the last cigarette or serum cotinine levels before the scan, or age of smoking initiation were not documented.Further, cannabis use was only measured by weeks of use in the past year.Although cannabis use and tobacco smoking are highly correlated (Agrawal et al., 2012), as supported by the current findings (Table 1), no associations between cannabis use and metabolite levels demonstrated in both controls and patients.Therefore, we assume that the observed differences between smoking and non-smoking controls are mainly driven by smoking.More detailed information on smoking history and dependence, for example using the CIDI extension substance abuse module (CIDI-SAM (Cottler et al., 1989)), could help uncover nuanced differences, or allow investigation of effects such as withdrawal.Furthermore, the relatively small control sample size, with 12 smokers and 23 non-smokers, limited the statistical power of the study.Moreover, results in this study are crosssectional, thus it is not clear whether the effects are a consequence of smoking or a reason that people smoke.Furthermore, analyzing metabolite concentrations separately in patients and controls ignores potential interactions between the groups, and limiting exploration of nuanced differences or similarities.Lastly, all patients in this study were taking antipsychotic medication.Reductions in glutamate levels following antipsychotic treatment after 1-80 months have been reported (Egerton et al., 2017).However, there were no significant differences in antipsychotic medication dosage among our patient samples and thus any potential effect will likely be minimal.
Concluding, we observed an association between chronic tobacco smoking and ACC Glx and tCh levels in controls, but not in FEP patients.Possibly, pre-existing disease-related glutamate changes or nAChR deficits in FEP patients preclude a measurable impact of smoking on glutamate.Further, observed decreased tCh levels in smokers, indicating a decrease in cellular proliferation processes, could be a consequence of the prolonged exposure to the neurotoxic effects of smoking.

Funding
This work was supported in part by a Veni grant (91618075) from the Netherlands Organisation for Health Research and Development (ZonMw) (EvdG) and Stichting J.M.C. Kapteinfonds (JPS).The funders were not involved in the study design, data collection or analysis, decision to publish or preparation of the manuscript.

Declaration of competing interest
None.
Fig. 1. a) Placement of MRS voxel in all subjects.Voxel was placed in the anterior cingulate cortex (ACC).The colour bar represents the fraction of overlap between the voxels, with red indicating greater overlap between participants.b) Representative MRS PRESS spectrum of a patient fitted by LCModel.SNR = 31 and linewidth (FWHM) = 0.034 ppm.Abbreviations: FEP, first-episode psychosis.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2 .
Fig. 2. a) Boxplots with overlaid data points showing the levels at baseline of the main metabolites of interest in the ACC of first-episode psychosis patients and controls, with boxplots presenting the group mean ± one standard deviation.b) Scatterplots showing the association between levels of the main metabolites of interest at baseline in the ACC and the number of cigarettes per day.The bold lines indicate the predicted relationships.The shaded areas indicate the 95 % confidence intervals.Abbreviations: Glx, glutamate + glutamine, I.U., institutional units; tCh, total choline. 1 Not surviving correction for multiple comparisons.*p < 0.05.

Table 1 .
Characteristics of the full sample are

Table 1
Demographical and clinical characteristics of FEP patients and controls at baseline and FU 1 .

Table 2
Linear mixed model results of all main metabolites for smoking status analyses in patients. 1Smoking status and covariates (age, sex, and, weeks of cannabis use, alcoholic drinks per week in the past year) were entered as fixed effects in the model.Intercepts for subjects and random slopes for time were entered as random effects.Significant p-values are in bold.*p < 0.05, **p < 0.01.
1 2 Not surviving correction for multiple comparisons.