Neurometabolic profile of the amygdala in smokers assessed with 1H-magnetic resonance spectroscopy

Tobacco smoking is one of the main causes of premature death worldwide and quitting success remains low, highlighting the need to understand the neurobiological mechanisms underlying relapse. Preclinical models have shown that the amygdala and glutamate play an important role in nicotine addiction. The aims of this study were to compare glutamate and other metabolites in the amygdala between smokers and controls, and between different smoking states. Furthermore, associations between amygdalar metabolite levels and smoking characteristics were explored. A novel non-water-suppressed proton magnetic resonance spectroscopy protocol was applied to quantify neurometabolites in 28 male smokers (≥15 cigarettes/day) and 21 non-smoking controls, matched in age, education, verbal IQ, and weekly alcohol consumption. Controls were measured once (baseline) and smokers were measured in a baseline state (1-3 hours abstinence), during withdrawal (24 hours abstinence) and in a satiation state (directly after smoking). Baseline spectroscopy data were compared between groups by independent t-tests or Mann-Whitney-U tests. Smoking state differences were investigated by repeated-measures analyses of variance (ANOVAs). Associations between spectroscopy data and smoking characteristics were explored using Spearman correlations. Good spectral quality, high anatomical specificity (98% mean grey matter) and reliable quantification of most metabolites of interest were achieved in the amygdala. Metabolite levels did not differ between groups, but smokers showed significantly higher glutamine levels at baseline than satiation. Glx levels were negatively associated with pack-years and smoking duration. In summary, this study provides first insights into the neurometabolic profile of the amygdala in smokers with high anatomical specificity. By applying proton magnetic resonance spectroscopy, neurometabolites in smokers during different smoking states and non-smoking controls were quantified reliably. A significant shift in glutamine levels between smoking states was detected, with lower concentrations in satiation than baseline. The negative association between Glx levels and smoking quantity and duration may imply altered glutamate homeostasis with more severe nicotine addiction.

Tobacco smoking is one of the main causes of premature death worldwide and quitting success remains low, highlighting the need to understand the neurobiological mechanisms underlying relapse.Preclinical models have shown that the amygdala and glutamate play an important role in nicotine addiction.The aims of this study were to compare glutamate and other metabolites in the amygdala between smokers and controls, and between different smoking states.Furthermore, associations between amygdalar metabolite levels and smoking characteristics were explored.
A novel non-water-suppressed proton magnetic resonance spectroscopy protocol was applied to quantify neurometabolites in 28 male smokers (≥15 cigarettes/day) and 21 non-smoking controls, matched in age, education, verbal IQ, and weekly alcohol consumption.Controls were measured once (baseline) and smokers were measured in a baseline state (1-3 h abstinence), during withdrawal (24 h abstinence) and in a satiation state (directly after smoking).Baseline spectroscopy data were compared between groups by independent t-tests or Mann-Whitney-U tests.Smoking state differences were investigated by repeated-measures analyses of variance (ANOVAs).Associations between spectroscopy data and smoking characteristics were explored using Spearman correlations.
Good spectral quality, high anatomical specificity (98% mean gray matter) and reliable quantification of most metabolites of interest were achieved in the amygdala.Metabolite levels did not differ between groups, but smokers showed significantly higher glutamine levels at baseline than satiation.Glx levels were negatively associated with pack-years and smoking duration.
In summary, this study provides first insights into the neurometabolic profile of the amygdala in smokers with high anatomical specificity.By applying proton magnetic resonance spectroscopy, neurometabolites in smokers during different smoking states and non-smoking controls were quantified reliably.A significant shift in glutamine levels between smoking states was detected, with lower concentrations in satiation than baseline.The negative association between Glx levels and smoking quantity and duration may imply altered glutamate homeostasis with more severe nicotine addiction.

Introduction
Tobacco use disorder (TUD) is still one of the leading causes of morbidity and premature death worldwide (Drope et al., 2018), and despite many smokers wanting to quit and the availability of pharmacological and cognitive behavioral therapy options, relapse rates remain high (Fiore et al., 2008;Hughes et al., 2004).To improve cessation therapy, understanding the neurobiological mechanisms underlying smoking relapse is crucial.Prolonged vulnerability to relapse is a defining characteristic of substance use disorders (SUDs), and the amygdala has been implicated to play a crucial role in it.The amygdala is part of the limbic system and is vital for regulating emotional and behavioral responses.It is not only involved in fear conditioning (LeDoux, 2000) but is also central for processes underlying SUDs including associative learning between neutral and innate emotional stimuli, salience attribution, and reward learning (LeDoux, 2000;Pryce, 2018;See et al., 2003).Due to its interconnection with regions involved in learning and memory like the hippocampus, and in motivation like the prefrontal cortex and the striatum, this network allows the detection of relevant stimuli in the environment and, in response to learned associations, the induction of appropriate behavior (Everitt et al., 2003).In SUDs, dysfunction of the amygdala causes the attribution of increased incentive salience to the substance and associated stimuli, promoting drug-seeking (Jentsch and Taylor, 1999).Thus, environmental stimuli and drug-associated memories as well as stress (Sharp, 2017) can induce craving and facilitate relapse.This involves the amygdala, as lesions or pharmacological inactivation of amygdalar nuclei attenuated cue- (See et al., 2003), context- (Fuchs et al., 2005), and stress-induced reinstatement of drug-seeking (Shaham et al., 2000).The extended amygdala is implicated in forming the negative emotional state during withdrawal, characterized by a reduced sensitivity of the brain reward system and the activation of the brain stress system.The desire to alleviate the negative symptoms promotes substance use and, hence, relapse (Koob and Volkow, 2010).Taken together, these findings imply that the amygdala plays an important role in SUDs, particularly in the states of withdrawal and relapse.
Recent preclinical findings highlight the importance of disrupted glutamate homeostasis in SUDs (Gipson et al., 2013;Kalivas, 2009).While acute nicotine administration is associated with increased glutamate transmission in brain reward circuitries, nicotine withdrawal has been associated with decreased transmission (Kenny et al., 2003).Furthermore, reinstatement of nicotine-seeking after withdrawal was accompanied by increases in glutamate release from the prefrontal cortex into the NAc core along with signs of synaptic potentiation (Gipson et al., 2013).The systemic administration of antagonists of postsynaptic glutamate receptors as well as activating presynaptic, inhibitory metabotropic glutamate receptors inhibited acute effects of nicotine, attenuated self-administration and cue-induced reinstatement of nicotine-seeking during withdrawal (Li et al., 2014).Moreover, the glutamatergic projection from the amygdala to the NAc seems to play an important role in SUDs, with pharmacological disconnection greatly reducing cue-controlled drug-seeking (Di Ciano and Everitt, 2004).Regarding nicotine, self-administration increased the expression of N-methyl-D-aspartate (NMDA) receptor subunits and the receptor-mediated excitatory postsynaptic currents in the amygdala, while amygdalar administration of an NMDA receptor antagonist decreased self-administration of nicotine (Kenny et al., 2008).These findings imply that glutamatergic transmission in the amygdala plays an important role in SUDs and in relapse.
In humans, it is unclear to what extent these glutamatergic adaptations occur in individuals suffering from SUDs and specifically TUD.However, neuroimaging studies have shown that the amygdala and glutamate do appear to play a key role in human drug addiction.For instance, a meta-analysis investigating positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) studies found consistent drug/nicotine cue-related activity in the amygdala and a positive association between this neural activity and craving for different addictive drugs (Chase et al., 2011;Kühn and Gallinat, 2011).Interestingly, craving was associated with increased glutamate levels in the NAc of cocaine users (Engeli et al., 2020).Furthermore, higher γ-aminobutyric acid (GABA) receptor availability was found in the amygdala of people with smoking history (Stokes et al., 2013), and current smokers had a lower mGluR5 (glutamate receptor) density throughout the brain compared to non-smokers, including the amygdala (Akkus et al., 2013;Hulka et al., 2014).With these findings suggesting an important role for the amygdala and glutamate signaling in TUD, investigating the translational relevance of findings from preclinical models may yield a more comprehensive insight into mechanisms underlying relapse vulnerability.This can be achieved by proton magnetic resonance spectroscopy ( 1 H-MRS), a non-invasive method to quantify neurometabolites.However, to our knowledge, no 1 H-MRS study has investigated the amygdala in smokers to date.This is probably due to technical challenges for 1 H-MRS of small subcortical brain areas like the amygdala (Hock et al., 2013), with the small voxel diameter negatively affecting the signal-to-noise ratio (SNR).Furthermore, surrounding circulating fluids and natural tissue interfaces induce motion artefacts and susceptibility changes, which distort line shape and broaden the full width at half maximum of peaks (FWHM), a measure of spectral quality.Long measurement times are required to improve spectral quality by extensive signal averaging; this increases the likelihood of subject motion and scanner-induced frequency drifts (Hock et al., 2013).By refining the metabolite cycling (MC) technique, this problem could be solved for the spinal cord (Hock et al., 2013) and the protocol was adapted and successfully implemented for the NAc (Engeli et al., 2020;Steinegger et al., 2021).In this study, the 1 H-MRS protocol was further adapted for the amygdala and optimized for measuring glutamate concentrations.With this protocol, glutamate and other neurometabolites were measured in the amygdala of non-smoking controls and smokers in three different states: baseline (1-3 h abstinence), withdrawal (24 h abstinence), and satiation (directly after smoking).Furthermore, the relationships between amygdalar glutamate levels and smoking and clinical parameters were investigated.Additionally, other metabolites of interest were quantified and compared between groups, smoking states, and in correlation with clinical parameters.

Participants
Sixty-six male participants (34 smokers, 32 controls) were recruited from the general population by means of public billboards, university websites and word-of-mouth communication.The inclusion criteria comprised 1) age of 18-50 years, 2) right-handedness, 3) fluency in German, 4) no current or previous Axis I DSM-IV psychiatric disorder (except TUD for smokers), 5) no family history of severe psychiatric disorders, 6) no neurological disorder or head injury, 7) no use of illegal drugs or prescription drugs affecting the central nervous system, 8) no contraindication for magnetic resonance imaging (MRI), and 9) no concurrent participation in another study.Smokers were required to: meet the DSM-V criteria for TUD, smoke at least 15 cigarettes a day, and be willing to abstain from smoking for at least 24 h prior to the second study day.The Ethics Committee of the Canton of Zurich approved the study and the Declaration of Helsinki was followed.All participants were informed about the study protocol, provided written informed consent and received financial compensation.Six smokers and eleven controls were excluded due to non-compliance with study protocol, cancelling their participation, or insufficient data quality on the first study day (details in Supplementary).Additionally, six smokers were excluded from smoking state analyses, due to at least one measurement on the second day failing data quality criteria.The final sample comprised 28 smokers and 21 controls for the investigation of C.A. Steinegger et al. glutamate, glutamine and their sum Glx (22 smokers for smoking state analyses).To investigate the other metabolites of interest, good spectral quality in the respective spectral sections was required (details in Supplementary).Due to insufficient data quality, one smoker and one control were excluded from these analyses (group: n = 27 smokers, 20 controls; state: n = 21 smokers).

Experimental procedure
The procedure was comparable to a previous 1 H-MRS study investigating glutamate levels of smokers in the NAc in a separate cohort (Steinegger et al., 2021).Controls participated in one study day and smokers in two (at least one week apart).Smokers were measured in three different smoking states, defined as baseline, withdrawal and satiation.All participants were asked to abstain for at least 48 h from alcohol and one hour from caffeine, prior to the start of each study day.Study days began at 8 a.m. at the MRI center of the Psychiatric Hospital of the University of Zurich.Urine was sampled to control for recent illegal drug use (amphetamine, benzodiazepines, cocaine, morphine/heroin, tetrahydrocannabinol).To verify smoking abstinence, expired carbon monoxide (CO) was quantified with UBLOW-CO monitors (NEOMED, Uechtelhausen, Germany), using cutoff values recommended by the manufacturer.On the first study day, smokers were requested to abstain from smoking 30 min prior to start.State-relevant questionnaires were administered, and clinical interviews conducted.As part of a multimodal study (also see Steinegger et al., 2021), participants underwent 1 H-MRS (baseline), followed by additional imaging sequences, after which they provided demographic data and completed clinical questionnaires.On the second study day, smoking abstinence was verified with CO values ≤8 ppm, state-relevant questionnaires were administered, followed by a 1 H-MRS measurement (withdrawal).Participants were then asked to smoke ad-libitum in the ensuing short break.Subsequently, they underwent another 1 H-MRS measurement (satiation), followed by the administration of state-relevant questionnaires.The 1 H-MRS measurements lasted approximately one hour, including anatomical sequences, voxel placement time, and the 1 H-MRS sequences (25 min long).

Acquisition and processing of magnetic resonance imaging data
1 H-MRS and MRI were performed on a 3T MRI scanner (Achieva, dStream architecture, DDAS, Software Release 5.4, Philips Healthcare, Best, The Netherlands) equipped with a 32-channel SENSE head coil.A three-dimensional T1-weighted scan (FOV=240 × 240 mm, 160 slices, flip angle=8 • , voxel size=1 × 1 × 1mm 3 ) was performed to place the 1 H-MRS voxel of interest.To rule out structural abnormalities, a neurologist examined the anatomical images.Scans were segmented to obtain the relative volume of gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) in the voxel using SPM8 (Ashburner et al., 2012).
Non-water-suppressed 1 H-MRS using the MC technique was performed on the left amygdala by implementing an adapted protocol originally developed for the spinal cord (Hock et al., 2013) and previously evaluated for the NAc and amygdala (Engeli et al., 2020;Hock et al., 2014;Steinegger et al., 2021).Voxel size was predefined as 13 × 15 × 15mm 3 (APxRLxFH).Inner volume saturation (Edden et al., 2006;Henning et al., 2008;Hock et al., 2013;Schulte et al., 2007) was applied to reduce the chemical shift displacement error and anomalous J-modulation, resulting in an effective voxel size of 9.0 × 12.2 × 10.0mm 3 .The transmitter frequency of the radiofrequency pulses used for localization was set to 3.03 ppm.512 averages were obtained from the total of four MC sequence block runs (TE=32 ms, TR=2500 ms, 2000 Hz band width, 2048 samples, flip angle=90 • , pulse duration=25 ms, spectral averages=128 each).Inversion pulses were applied with the transmit frequency varying between 170 and 190 in order to avoid partial inversion of the water peak in the case of strong eddy current effects or bad shimming (Hock et al., 2013).A spectrum of each individual MC sequence block was visually inspected upon completion to estimate data quality.Between MC sequence blocks, coronal T2-weighted images were acquired to verify voxel position.If substantial voxel displacement was detected, a new T1-weighted scan was run, the voxel repositioned, and the calibration phases as well as the last MC sequence block repeated.
After raw data exportation, 1 H-MRS data were frequency aligned, phase corrected and corrected for eddy currents and channel combined (Hock et al., 2013).LCModel (Provencher, 1993) was used to fit the peak areas of the different metabolites.Glutamate, glutamine and their sum, Glx, were the main metabolites of interest.Other metabolites of interest comprised creatine, N-acetylaspartate plus N-acetylaspartyl-glutamate (NAA+NAAG), glycerophosphocholine plus phosphocholine (GPC+PCh), myoinositol plus glycine (mI+Gly) and GABA.Spectra with SNR<7 and a water peak with FWHM>9 Hz were excluded from analyses.Molar concentrations were calculated with tissue water as a concentration reference (Gasparovic et al., 2006).

Statistical analysis
Data were analyzed using SPSS (version 25, IBM Corp., Armonk, N. Y., USA) and visualized with R Core Team (2018), using the R package ggplot2 (Wickham, 2016) for figures.Group comparisons regarding metabolic, demographic, clinical and drug use data were carried out with independent t-tests.Mann-Whitney-U tests were used for nonparametric data.Repeated-measures analyses of variance (ANOVAs) were applied to investigate metabolic differences between smoking states and state-specific clinical data.Pairwise comparisons were Bonferroni-corrected. Bayes factors (BF) were calculated for group and smoking state comparisons using the R package BayesFactor (version 0.9.12-4.2)(Moreyand Rouder, 2018) and interpreted using cutoff values according to Jeffreys (1961).Potential associations between metabolite levels and clinical parameters were explored, applying Spearman's rank-order correlations.Pearson's correlations are reported for the significant associations.Glutamine and GABA were excluded from correlation analysis due to their larger data variance (see Cramér-Rao lower bounds (CRLB), Supplementary Table S1).Due to the exploratory nature of this investigation, no correction for multiple comparisons was applied.

Participant characteristics
Groups were matched in age (ranging from 19-45 years), years of education, verbal IQ and weekly alcohol consumption (Table 1).Smokers scored higher on the Barratt Impulsiveness Scale, but did not differ significantly from controls regarding other clinical symptoms (Table 1).
Smoking characteristics are reported with mean and standard deviation in Table 1.Smokers smoked between 15 and 30 cigarettes a day and displayed a moderate level of nicotine dependence.Questionnaire scores on craving (QSU), withdrawal (CWS, MNWS) and anxiety symptoms (STAI-S) in the different states are shown in Table 2. Comparison between smoking states showed that there were main effects of state (Table 2), with smokers displaying stronger craving and withdrawal symptoms during withdrawal than in baseline (Bonferroni corrected posthoc comparisons, QSU: p<0.001,CWS: p<0.001,MNWS: p = 0.022) and in satiation (QSU: p<0.001,CWS: p = 0.005).Withdrawal symptoms were also stronger in satiation than baseline (CWS: p = 0.009, MNWS: p = 0.038).Furthermore, there was a main effect of state on anxiety symptoms, which did not survive correction for multiple comparison (baseline<withdrawal: p>0.05).

Tissue composition and metabolite concentrations in the amygdala: group effects
Withstanding the small voxel size and location (Fig. 1A), good data quality was achieved (Fig. 1B), as shown with the measures SNR, FWHM (Table 3), and CRLB (Supplementary Table S1), while voxel segmentation revealed high anatomical specificity for the amygdala (details in Supplementary).
For the included spectra, FWHM and SNR did not differ between groups (see BF). Voxel segmentation mean tissue composition values are shown in Table 3.No significant difference between groups was found for GM, WM or CSF volume, indicating reliable voxel placement across participants.
Mean molar concentrations for both groups are depicted in Table 4 for all metabolites of interest.Glutamate, glutamine and Glx concentrations did not differ between groups (Fig. 2).Furthermore, no group difference was detected for creatine, NAA+NAAG, mI+Gly, GPC+PCh and GABA (Table 4).

Tissue composition and metabolite concentrations: smoking state effects
Within smokers, GM volume did not differ conclusively between states (Table 5).A main effect of state was found for CSF, with larger mean volume during withdrawal than in baseline (p = 0.025), but not supported by BF.CSF voxel fraction did not exceed 2% in smokers for any state.However, the number of measurements with CSF values larger than 0% were more for withdrawal (n = 7) than for baseline (n = 2).Similar data quality was achieved across smoking states, with no significant difference in FWHM and SNR between smoking states (Table 5).
Regarding glutamine, a main effect of state was found with baseline levels higher than satiation levels (p = 0.016, Fig. 2C and Table 6).There was also a main effect of state for Glx (Fig. 2B), with baseline concentrations higher than those in satiation, which did not survive Bonferroni correction (p = 0.142), nor was it supported by BF (Table 6).By contrast, no difference between smoking states was found for glutamate, nor for any other metabolite of interest.

Associations between smoking characteristics and metabolite concentrations
In the baseline state, lower Glx levels were associated with more pack-years (rs( 26 In withdrawal and satiation, no significant association between smoking parameters and glutamate, Glx, or any other metabolite of interest was found.

Discussion
The neurometabolic profile of the amygdala was quantified successfully and with high anatomical specificity (98% gray matter) with 1 H-MRS in twenty-eight smokers during the states of baseline, withdrawal and satiation, and twenty-one controls.At baseline, neurometabolites did not differ between smokers and controls.State analyses in smokers revealed no difference for glutamate, but significantly higher glutamine and Glx levels at baseline compared to satiation.More packyears and longer smoking duration were associated with lower Glx levels at baseline.Moreover, more pack-years were also associated with less creatine at baseline.
Compared to controls, smokers did not show significantly altered glutamate and glutamine levels in the amygdala.In preclinical models, acute nicotine causes an increase in glutamate transmission in the mesolimbic system (Reid et al., 2000) and in amygdalar explants (Barazangi and Role, 2001), which both subsided within an hour.Hence, short-lived glutamatergic alterations in response to nicotine may no longer have been present in the baseline state measured in the smokers in this study, obscuring potential group differences.Time-sensitive, glutamatergic adaptations have also been found in humans, with       decreased glutamate receptor availability reported in the amygdala of smokers, particularly after recent smoking (Akkus et al., 2013;Hulka et al., 2014), highlighting the importance of timing.Regarding other brain areas, baseline glutamate levels in smokers were also not altered compared to controls in the NAc (Steinegger et al., 2021) or the hippocampus (Gallinat and Schubert, 2007), two brain areas connected with the amygdala and shown to play an important role in SUDs (Koob and Volkow, 2010).By contrast, reduced levels were found in the anterior cingulate cortex and dorsolateral prefrontal cortex in smokers compared to controls (Durazzo et al., 2016), which may suggest local changes in glutamate levels in TUD.It is also possible that the concentration differences were too small to be detected by 1 H-MRS, and changes occurring in specific amygdalar nuclei or cellular compartments were obscured by measuring the amygdala as a whole.
Regarding smoking-state comparison of the main metabolites of interest, baseline glutamine and Glx levels were significantly higher than satiation levels.In contrast, glutamate levels did not differ between smoking states.It is noteworthy that the lack of significant change in glutamate levels between smoking states stands in contrast to preclinical studies investigating different states (Gipson et al., 2013;Kenny et al., 2003), which may be attributable to species differences or study-specific temporal, brain-region specific or methodological differences.It is possible that our three states were too similar and that 24 h was not long enough to result in substantial shifts in glutamate.Furthermore, although our participants went immediately into the scanner after the smoking break, running the anatomy sequence and placing the voxel were time-consuming, such that immediate, large effects of nicotine on glutamate signaling may have been missed in the satiation measurement.Furthermore, withdrawal in human TUD and in preclinical models is not the same: while animals are usually kept in a controlled environment without access to the drug, humans with TUD are exposed to environmental cues while choosing to abstain from the drug.This may result in crucial differences in the neurobiological mechanisms underlying the respective withdrawal states.Regarding our finding that glutamine and Glx levels were reduced during satiation compared to baseline, multiple interpretations are possible.The reduced glutamine and Glx levels could reflect altered glutamate-glutamine cycling in smokers.The short-term reduction in glutamine levels in satiation may be due to an increased mitochondrial metabolism, using glutamine to generate energy or synthesize proteins (McKenna, 2007).An additional aspect to consider for the interpretation of the findings is that mean glutamine CRLBs were around 20-25%, implying greater measurement uncertainty compared to other metabolites investigated, thus possibly affecting the results.
Regarding associations between metabolite levels and clinical parameters, baseline Glx was associated negatively with pack-years and smoking duration.This may imply that with heavier smoking, glutamate and/or glutamine levels are lower after the acute effects of nicotine subside.Interestingly, this is in line with our prior study assessing the NAc, where we found a negative association between glutamate and craving as wells as Glx and severity of smoking (Steinegger et al., 2021), which could imply altered glutamate homeostasis in the NAc and amygdala after prolonged nicotine use and more severe nicotine addiction.Basal glutamate levels are reduced in the NAc in preclinical studies modeling cocaine addiction, due to decreased cystine-glutamate exchange in the extrasynaptic space, impairing glutamate homeostasis (Kalivas, 2009).Similar findings were found during nicotine withdrawal in a preclinical study, in which nicotine withdrawn rats showed decreased expression of the catalytic subunit of the cystine-glutamate exchanger in the NAc (Knackstedt et al., 2009), providing a possible mechanism of change.Additionally, in the present study, lower creatine was associated with more pack-years, possibly implying that cell energetics are negatively affected with prolonged smoking.Creatine levels have been found to decrease with age in the brain (King et al., 2008), but did not correlate with age in this study (not shown).Regarding the correlations of Glx and creatine to pack years and the smoking state changes in Glx levels, we believe that two separate mechanisms may be involved.Animal models suggest that amygdalar glutamate (as a neurotransmitter) is sensitive to smoking state (Barazangi and Role, 2001) and our results in this study imply a smoking state sensitivity for glutamine, which may explain the short-term effects of Glx with smoking state.As creatine plays a different role in the brain, this metabolite may not be as sensitive as Glx to smoking state, explaining the lack of change between smoking states.However, both Glx and creatine may show long term changes in response to smoking, as suggested by the correlations with pack years.Yet the correlation between creatine and pack-years was not significant anymore after controlling for age.
The following limitations need to be considered for this study.Only male participants were included in the study, limiting generalizability.Prior studies applying structural and functional MRI analysis revealed sex-related effects of smoking on corticostriatal circuits.Male smokers exhibited lower left caudate volume and altered resting state functional connectivity (rsFC) between subcortical and cortical structures (Wen et al., 2022) as well as lower rsFC regional homogeneity in the right ventral striatum, left cerebellum crus1, and left fusiform gyrus (Lin et al., 2021), whereas female smokers showed a trend for smaller right amygdalar volume.Future studies should investigate sex-specific effects of smoking.The amygdala was measured as a whole and 1 H-MRS cannot reliably distinguish neurometabolite pools in different nuclei within the voxel or in separate cellular compartments (Mullins, 2018).Furthermore, our 1 H-MRS protocol was primarily optimized for the detection of glutamate and glutamine, but with 3T the two are still difficult to differentiate due to spectral overlay, potentially obscuring individual changes.Mean CRLBs of glutamine and GABA were above 20% and 30%, respectively, implying larger uncertainties regarding LCModel fit in comparison to the other metabolites.Finally, due to the long measurement times, it is also possible that short-lived changes in neurometabolite signaling were diluted, particularly during satiation.

Conclusions
In summary, the neurochemical profile of the amygdala, including glutamate, glutamine and Glx levels, was measured reliably with high anatomical specificity by applying 1 H-MRS in three different smoking states.Shifts in glutamine levels, with lower concentrations in satiation than baseline, were revealed.Lower baseline Glx levels were associated with more pack-years and longer smoking duration, possibly implying altered glutamate homeostasis in smokers with more severe addiction.Future 1 H-MRS investigations at higher field strengths or in different states, e.g., during acute nicotine administration or after prolonged smoking abstinence, may provide further insight on the temporal and anatomical specificity of neurometabolic amygdalar changes.Finally, this 1 H-MRS protocol can be used to investigate the amygdalar neurochemical profile in other psychiatric disorders and different SUDs, potentially shedding light on shared and drug-specific adaptations in human drug addiction.

Funding and disclaimer
This work was supported by the Forschungskredit of the University of Zurich (FK-15-041), the EMDO Stiftung and the Fonds für wissenschaftliche Zwecke im Interesse der Heilung von psychischen Krankheiten of the Psychiatric Hospital of the University of Zurich (8702) all awarded to L.M. Hulka.

Declaration of competing interest
None.

Table 1
Demographic and clinical data, alcohol use pattern and smoking characteristics (means and standard deviation).

Table 3
Tissue composition, quality control measures and 1 H-MRS metabolite concentrations of the amygdala (means and standard errors) for group analyses.

Table 4 1
H-MRS metabolite concentrations of the amygdala (means and standard errors) for group analyses in [mmol/L] quantified using tissue water as a reference.

Table 5
Tissue composition, quality control measures and 1 H-MRS metabolite concentrations of the amygdala (means and standard errors) for state analyses.

Table 6 1
H-MRS metabolite concentrations of the amygdala (means and standard errors) for state analyses in [mmol/L] quantified using tissue water as a reference.