Chronic cigarette smoking in healthy middle-aged individuals is associated with decreased regional brain N-acetylaspartate and glutamate levels

Abstract

Background

Cigarette smoking is associated with metabolite abnormalities in anterior brain regions, but it is unclear if these abnormalities are apparent in other regions. Additionally, relationships between regional brain metabolite levels and measures of decision-making, risk-taking, and impulsivity in smokers and non-smokers have not been investigated.

Methods

Young/middle-aged (predominately male) non-smokers (n=30) and smokers (n=35), were compared on N-acetylaspartate (NAA), choline-containing compounds (Cho), creatine-containing compounds (Cr), myo-Inositol (mI), and glutamate (Glu) levels in the anterior cingulate cortex (ACC) and right dorsolateral prefrontal cortex (DLPFC) via 4 Tesla (T) proton magnetic resonance (MR) single volume spectroscopy. Groups were also compared on NAA, Cho, Cr, and mI concentrations in the gray matter (GM) and white matter (WM) of the four cerebral lobes and subcortical nuclei/regions with 1.5T proton MR spectroscopic imaging. Associations of regional metabolite levels with neurocognitive, decision-making, risk-taking, and self-reported impulsivity measures were examined.

Results

Smokers showed lower DLPFC NAA, Cr, mI and Glu concentrations, and lower lenticular nuclei NAA level; smokers also demonstrated greater age-related decreases of DLPFC NAA, and ACC and DLPFC Glu levels. Smokers exhibited poorer decision-making and greater impulsivity. Across the sample, higher NAA and Glu in the DLPFC and NAA concentrations in multiple lobar GM and WM regions and subcortical nuclei were associated with better neurocognition and lower impulsivity.

Conclusions

This study contributes additional novel evidence that chronic smoking in young/middle-aged individuals is associated with significant age-related neurobiological abnormalities in anterior frontal regions implicated in the development and maintenance of addictive disorders.

Introduction

Chronic cigarette smoking in adults is associated with multiple neurobiological and neurocognitive abnormalities (1–3). The majority of earlier [see (1)] and recent (4–10) studies on smoking-related neurobiological abnormalities employed magnetic resonance (MR)-based morphological measures (i.e., volume and cortical thickness). Overall, the findings indicated smokers demonstrate widespread structural abnormalities that are particularly prominent in anterior frontal lobe subregions (11).

While MR morphometry provide fundamental information on the macroscopic viability of regional brain tissue, MR spectroscopy enables a more direct interrogation of the functional integrity of brain tissue. Single volume proton MR proton spectroscopy (SVS) and spectroscopic imaging (SI) methods allow the non-invasive and concurrent quantitation of several brain metabolites that collectively provide information regarding the neurophysiological viability of tissue (12, 13). Abnormalities in certain metabolite concentrations (e.g., N-acetylaspartate, NAA, choline-containing compounds, Cho) may precede macroscopic morphological and/or neurocognitive changes associated with some diseases/conditions (13). The brain metabolites most commonly quantitated via SVS and SI methods include biomarkers of neuronal integrity (i.e., NAA), cell membrane turnover/synthesis (i.e., Cho), cellular bioenergetics (i.e., creatine-containing compounds, Cr), astrogliosis and inflammation (i.e., myo-Inositol, mI), and excitatory neurotransmitter/neuromodulator pools (glutamate, Glu) [see (13, 14) for review]. Higher regional NAA and Glu levels are associated with better function in multiple neurocognitive domains (15, 16), and both of these metabolites show age-dependent decreases in concentration across adulthood (13). The few proton MR spectroscopy (MRS) studies of “healthy” chronic smokers primarily employed SVS at 3 Tesla (T), and focused on anterior frontal regions (e.g., anterior cingulate cortex, dorsolateral prefrontal cortex) and the hippocampus in young/middle-aged adults; anterior frontal subregions and the hippocampus were emphasized because neurobiological abnormalities in these regions are implicated in the development and persistence of addictive disorders (17, 18). Gallinat and colleagues reported that smokers showed lower NAA levels than non-smokers in the left hippocampus, and higher pack-years were related to higher Cho levels in the anterior cingulate gyrus (19); however, in a subsequent study, the same authors found no differences between active-smokers, former-smokers, and non-smokers on Glu levels in the anterior cingulate cortex and left hippocampus (20). O’Neill and colleagues (21) found no differences between smokers and non-smokers in thalamic Glu concentration, but within smokers, more cigarettes smoked per day and packyears of smoking were strongly related to lower thalamic Glu level. Mennecke and associates (22) reported higher left anterior cingulate Glu/glutamine (Glx) and Cho concentrations in smokers than non-smokers; after 3 days of smoking cessation, anterior cingulate Glx decreased to non-smoker levels, but no changes were observed for Cho. In the sole SI study, Durazzo and colleagues (23) observed that a small group of smokers had lower Cho concentrations in the cerebellar vermis than non-smokers. Taken together, these studies provide evidence that chronic smoking is associated with regional derangements of cortical NAA, Cho, and Glx levels. However, a limitation of the SVS method is that it does not simultaneously measure metabolites across a large number of brain regions; therefore, the regional specificity of the metabolic findings in chronic smokers (e.g., anterior vs. parietal, temporal, or subcortical) and their tissue-specificity [i.e., gray matter (GM) vs. white matter (WM)] is unclear. Additionally, previous studies did not assess associations between the regional brain metabolite levels and measures of neurocognition, decision-making, risk-taking, or impulsivity; consequently, the functional relevance of the metabolite abnormalities observed in smokers is uncertain.

This study compared healthy middle-aged smokers and non-smokers on regional brain metabolite levels. SVS at 4T measured NAA, Cho, Cr, mI, and Glu levels in the right dorsolateral prefrontal cortical region (DLPFC) and the bilateral anterior cingulate cortical region (ACC); studies at 4T facilitate more accurate quantitation of the Glu signal than lower field strengths due to greater spectral dispersion and increased signal-to-noise ratio (13). Additionally, SI at 1.5T simultaneously measured NAA, Cho, Cr, and mI (but not Glu) concentrations in the GM and WM of the bilateral frontal, parietal, and temporal lobes, occipital WM, and lenticular nucleus, thalamus, and cerebellar vermis. Associations of SVS and SI metabolite levels with performance on a comprehensive neurocognitive battery and on measures of impulsivity, decision-making, and risk-taking were examined.

Chronic smoking, independent of common smoking-related diseases (e.g., cerebrovascular disease, chronic obstructive pulmonary disorders), appears to adversely affect the integrity of brain neurobiology (1). Additionally, we observed that smoking is associated greater age-related brain volume loss observed in morphological studies with healthy individuals (9) and those with an alcohol use disorder (24, 25). Accordingly, we predicted: (i) Compared to non-smokers: (a) Smokers demonstrate lower NAA and Glu levels in the DLPFC and ACC, as well as lower NAA concentrations in the frontal, parietal, and temporal lobes, lenticular nuclei, and cerebellar vermis; (b) Smokers evidence significantly greater age-related decreases of regional NAA and Glu levels. (ii) Smokers show greater levels of risk-taking, impulsivity, and poorer decision-making. (iii) Across smokers and non-smokers, higher NAA and Glu levels are related to better neurocognition, while higher DLPFC and ACC NAA and Glu concentrations are associated with better decision-making, and with lower risk-taking and impulsivity.

Methods

Participants

Healthy, community-dwelling participants were recruited via posters, electronic billboards, and word-of-mouth. Participants were between the ages of 24 and 69 and gainfully employed at the time of study (see Table 1). Prior to engaging in procedures, participants provided written informed consent according to the Declaration of Helsinki, and the consent document and procedures were approved by the University of California San Francisco and the San Francisco VA Medical Center. For SVS, there were 35 current smokers (four females) and 30 non-smokers (four females); for SI there were 28 current smokers (two females) and 36 non-smokers (three females). SI data was gathered from 2001–2012 and SVS data was obtained from 2005–2014. Therefore, approximately 50% of participants with SI data also had SVS data; the number of smokers and non-smokers with both SVS and SI data were equivalent. SI and SVS samples were equivalent on demographic, cigarette use, and alcohol consumption variables.

Table 1.

Demographic and clinical measures

Variable	Non-smokers (n = 30)	Smokers (n = 35)
Age (years)	49.1 ± 12.0	48.6 ± 10.1
Education (years)	16.5 ± 2.1	14.9 ± 2.1*
AMNART	119 ± 9	117 ± 6
Male (%)	87	89
Caucasian (%)	63	71
Body mass index	25.5 ± 3.7	26.4 ± 3.8
Beck Depression Inventory	3 ± 3	5 ± 4
STAI-trait	31 ± 8	35 ± 9
1-yr avg drinks/month	14 ± 14	22 ± 20
Lifetime average drinks/month	19 ± 12	26 ± 20*
Biological Mother/Father positive history of problem drinking (%)	28	37
FTND	NA	5 ± 2
Cigarettes/day	NA	18 ± 6
Total lifetime years of smoking	NA	29 ± 11
Pack years	NA	27 ± 15

Note. Mean ± standard deviation;

^*p < .05;

AMNART: American National Adult Reading Test; FTND: Fagerstrom Test for Nicotine Dependence; STAI: State-Trait Anxiety Inventory, trait-score.

Primary inclusion/exclusion criteria are fully detailed elsewhere (26). In summary, participants were screened for history of neurologic (e.g., seizure disorder, neurodegenerative disorder, traumatic brain injury with loss of consciousness > 5 min), general medical (e.g., hypertension, diabetes, chronic obstructive pulmonary disease), and psychiatric (i.e., mood, thought, anxiety, substance/alcohol use disorders) conditions/disorders known or suspected to influence neurocognition or brain neurobiology. All females were pre-menopausal, by self-report. Most non-smoking participants never smoked, while a few smoked less than 40 cigarettes during their lifetime, but had no cigarette/tobacco use in the 10 years prior to study. All smoking participants were actively smoking at the time of assessment, smoked at least 10 cigarettes/day for ≥ 5 years, with no periods of smoking cessation >1 month in the 5 years prior to study. At the time of study, no smoker was engaged in any pharmacological/behavioral smoking cessation program or used any other form of tobacco or electronic cigarettes. All smokers were allowed to smoke ad libitum prior to all procedures, and take smoke breaks when requested.

Psychiatric, medical, and substance/alcohol consumption assessment

Participants were administered the screening section of the Structured Clinical Interview for DSM-IV Axis I disorders, Patient Edition, Version 2.0, as well as an in-house questionnaire designed to screen for medical, psychiatric, neurological and developmental conditions known or suspected to neurocognition or brain neurobiology. Participants were also administered semi-structured interviews for lifetime alcohol consumption (Lifetime Drinking History, LDH) and substance use (in-house questionnaire assessing substance type, and quantity and frequency of use). From the LDH, we derived average number of drinks/month (defined as containing 13.6 grams of pure ethanol) over 1 year prior to enrollment and average number of drinks/month over lifetime. Participants also completed self-report measures of depressive (Beck Depression Inventory, BDI) and anxiety (State-Trait Anxiety Inventory, form Y-2, STAI) symptomatologies, and family history of problem drinking. Smokers completed a measure of nicotine dependence level [Fagerström Test for Nicotine Dependence (FTND)], and provided information on the total number of cigarettes currently smoked per day, and the total number of years of smoking over lifetime. From this information, pack-years [i.e., (typical number of cigarettes per day/20) × total number of years of smoking] were calculated for smokers. Prior to assessment, participants’ urine was tested for common illicit substances (e.g., THC, opiates, cocaine, and amphetamines), and they were assessed for recent ethanol consumption via breathalyzer. No participant was positive for the above common illicit substances or ethanol consumption at the time of assessment. See (9) for corresponding references for the above measures.

Neurocognitive and behavioral assessment

Participants completed a comprehensive battery composed of measures commonly used in clinical and research settings in North America (27). Estimated verbal intelligence was assessed with the American National Adult Reading Test [AMNART (28)]. The battery evaluated the following domains of neurocognition reported to be adversely affected by chronic smoking (1, 26): cognitive efficiency, executive skills, general intelligence, processing speed, learning and memory (auditory-verbal and visuospatial), visuospatial skills, and working memory. See (26) for full description of the neurocognitive battery and results of comparisons of smokers and non-smokers on the comprehensive neurocognitive battery used in this study. Participants also completed task-based measures of decision-making [Iowa Gambling Task, IGT (29)], risk-taking [Balloon Analogue Risk Task, BART (30, 31)], and a self-report measure of trait impulsivity [Barrett Impulsivity Scale-11 (53)]. Scores for all the above measures were converted to z-scores based on the performance of the non-smokers in this study. A global neurocognition score was formed via the arithmetic average of z-scores for all of the individual domains from the neurocognitive battery.

MR acquisition and processing

Magnetic resonance imaging (MRI); see Supplemental Methods for details

The 4T MR data were acquired on a Bruker MedSpec system; 3D T1-weighted images were obtained with Magnetization Prepared Rapid Gradient imaging (MPR), and 3D T2-weighted images via turbo spin-echo. The 1.5T MRI data were acquired on a Siemens Vision system; MPR and T2-weighted double spin echo images were acquired. The 4T structural images were segmented into GM, WM and cerebrospinal fluid (CSF) using the Expectation Maximization Segmentation (EMS) method (32) and co-aligned with the SVS volumes of interest for determination of their tissue contribution (i.e., GM, WM, CSF) (33). The 1.5T structural images were also segmented into total brain GM, WM, and CSF via EMS. Subsequently, volumes for the four major lobes and subcortical regions were calculated and co-registered to the EMS segmentation to obtain GM, WM, and CSF fractions for the preceding regions (34). Finally the segmented 1.5T structural images were co-aligned with SI metabolite maps for anatomical localization (e.g., frontal WM) and determination of tissue contributions (i.e., GM, WM, CSF) in the corresponding SI voxels (23, 35).

Magnetic resonance spectroscopy; see Supplemental Methods for details

SVS at 4T

Volumes-of-interest (VOIs) for MRS were placed over the perigenual ACC and right DLPFC (see Supplemental Figure S1). NAA, Cr, Cho, mI, and Glu signals from both VOIs were acquired with a Stimulated Echo Acquisition Mode sequence (36) and quantitated metabolite levels were corrected for CSF contribution and scaled to water level from the corresponding VOI. For full methods details see (33).

SI at 1.5T

Spectra were acquired in three 15 mm thick parallel slices, slice gap of approximately 6 mm, with a nominal SI voxel size of 1 ml (effective size of 1.5 ml). The SI slices covered the four major cerebral lobes, subcortical nuclei, midbrain and cerebellar vermis. The reconstructed metabolite maps were co-aligned with the segmented structural images, and concentrations (CSF-corrected and scaled to water) were calculated for NAA, Cho, Cr, and mI. Metabolite concentrations for the left and right hemispheres for all regions were averaged because there were no significant hemispheric differences in either group.

Statistical analyses

Group comparisons

Generalized linear modeling (GENLIN) compared smokers and non-smokers on regional metabolite concentrations. SVS analyses focused on NAA, Cr, Cho, mI, and Glu concentrations in the ACC and DLPFC. SI analyses focused on NAA, Cr, Cho, and mI levels in the frontal, parietal, and temporal GM and WM, occipital WM, thalami, lenticular nuclei, and cerebellar vermis. Predictors in all models included smoking status (smoker and non-smoker), age, body mass index (BMI), lifetime average drinks/month, GM fraction in the VOI or SI voxel, and the smoking status × age interaction. BMI was used as a covariate because it was related to metabolite concentrations measured via SVS and SI in healthy controls (37, 38). Lifetime average drinks/month was used as a covariate since smokers drank significantly more than non-smokers. All main effects and interactions were considered statistically significant at p<.05. Significant main effects for smoking status were followed-up with t-tests (two-tailed) comparing smokers and non-smokers on mean regional metabolite levels. Despite our a priori predictions for lower regional NAA and Glu in smokers, alpha levels for all t-tests for metabolites in each SVS or SI region were adjusted for multiple comparisons via a modified Bonferroni procedure (39), which accounted for the number of metabolites (five for SVS; four for SI), number of regions (two for SVS; 10 for SI) and the moderate-to-strong intercorrelation among metabolites across regions for the entire sample. For SVS, the average intercorrelation among metabolites across the DLPFC and ACC was r=0.46. For SI, intercorrelations of metabolites among the frontal, parietal, and temporal GM and WM, lenticular nuclei and cerebellum was r=0.66. Resulting adjusted alpha levels for SVS metabolite t-tests was p=.017, and p=.020 for SI. Effect sizes for mean metabolite concentrations differences between smokers and non-smokers were calculated with Cohen’s d (40).

Decision-making, risk-taking, and impulsivity measures: GENLIN compared smokers and non-smokers on the IGT net total score, BART average adjusted pumps, and BIS total, attentional, motor, and non-planning impulsivity scores. Predictors in all models included smoking status (smoker and non-smoker), age, education, and lifetime average drinks/month. Main effects were considered significant at p<.05 and significant main effects for smoking status were follow-up with t-tests. T-tests were adjusted for multiple comparisons via the above modified Bonferroni procedure, based on the number of individual measures (six) and their average intercorrelation (r=0.44), resulting in an adjusted alpha level of p=.019. Effect sizes were calculated with Cohen’s d.

Associations of SVS and SI metabolites with neurocognitive, risk-taking, decision-making, impulsivity, and smoking severity measures: Associations of SVS and SI metabolites with neurocognitive, risk-taking, decision-making, and impulsivity measures were examined in the total sample (i.e., smokers and non-smokers) with linear regression (semi-partial correlations reported) controlling for age, education, and lifetime average drinks/month. Relationships of SVS and regional SI metabolite levels with lifetime years of smoking, and pack-years were examined in smokers with linear regression (semi-partial correlations reported) controlling for age and lifetime average drinks/month. The associations for NAA and Glu were considered significant at p<.05, given our a priori predictions; p-values for associations of Cho, Cr, and mI with the above measures were conservatively adjusted with a standard Bonferroni procedure.

Results

Participant characteristics

No significant differences were observed between smokers and non-smokers on age, AMNART, BDI, STAI-trait, BMI, and 1 year average drinks/month (all p>.10) Groups were equivalent on frequency of Caucasians and positive history of problem drinking in biological parents. Smokers had significantly fewer years of education and more lifetime average drinks/month (p<.05). See Table 1.

Group comparisons of SVS metabolite concentrations

ACC

A smoking status × age interaction was observed for Glu [χ² (1)=8.54, p=.003], where smokers showed lower Glu concentration with increasing age relative to non-smokers (see Fig. 1a); a simple slopes difference test for age indicated the effect of age on Glu level was more than 1.5 times greater in smokers than non-smokers (p=.003). Smokers showed trends for lower NAA (p=.054) and higher Cho (p=.052) than non-smokers. No significant difference in mean Glu, mI or Cr levels were observed between smokers and non-smokers (all p>.15). For all metabolites except Glu, higher lifetime drinks/month were related to lower levels (all p<.03). There was no significant difference of percent GM contributing to the ACC volume between smokers (60%) and non-smokers (58%).

Figure 1.

(A) Changes in anterior cingulate cortex glutamate levels across age for smokers and non-smokers; (B) Changes in right dorsolateral prefrontal cortex glutamate levels across age for smokers and non-smokers; ACC, anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; i.u., institutional units.

Right DLPFC

A smoking status × age interaction was yielded for NAA [χ² (1)=5.88, p=.015] and Glu [χ² (1)=8.76, p=.003], where smokers showed significantly lower NAA and Glu concentrations with increasing age relative to non-smokers (see Fig. 1b); a simple slopes test for age indicated the effect of age on NAA and Glu in smokers was twice that of non-smokers (both p<.015). Main effects for smoking status were observed for NAA [χ² (1)=13.69, p<.001], Cr [χ² (1)=6.85, p=.009], mI [χ² (1)=11.63, p=.001], and Glu [χ² (1)=6.81, p=.009]; for each metabolite, smokers demonstrated significantly lower concentrations than non-smokers and moderate-to-large effect sizes were apparent for these mean differences (see Table 2). There was no significant difference in percent GM contributing to the DLPFC volume between smokers (46%) and non-smokers (43%).

Table 2.

Group means for SVS and SI metabolite levels (institutional units) and performance on the Iowa Gambling Task and BIS

Variable		Non- smokers (n = 30)	Smokers (n = 35)	Effect size (Cohen’s d)
SVS Anterior cingulate gyrus	NAA	5.88 ± 0.79	5.49 ± 0.79	0.50
	Cho	1.29 ± 0.22	1.40 ± 0.22	0.51
	Cr	4.62 ± 0.75	4.73 ± 0.74	0.14
	mI	3.89 ± 0.86	3.59 ± 0.85	0.35
	Glu	3.95 ± 0.69	4.01 ± 0.69	0.09
SVS Dorsolateral prefrontal cortex	NAA	5.58 ± .072	4.90 ± 0.71*	0.95
	Cho	1.09 ± 0.17	1.05 ± 0.18	0.22
	Cr	4.59 ± 0.61	4.18 ± 0.60*	0.67
	mI	3.68 ± 0.66	3.11 ± 0.66*	0.87
	Glu	3.45 ± 0.53	3.09 ± 0.52*	0.67
SI Frontal gray matter	NAA	33.12 ± 3.09	31.06 ± 3.10	0.67
SI Lenticular nucleus		29.21 ± 3.52	26.64 ± 3.51#	0.73
Iowa Gambling Task		−0.07 ± 0.91	−0.64 ± 0.80&	0.67
BIS motor impulsivity		0.06 ± 1.05	0.86 ± 1.03&	0.77
BIS non-planning impulsivity		0.19 ± 1.07	0.91 ± 1.03&	0.69
BIS total score		0.20 ± 1.00	0.86 ± 0.98&	0.66

Note. Mean ± standard deviation, values obtained from estimated marginal means;

^*smokers < never-smokers, p ≤ .017;

^#smokers < non-smokers, p ≤ .020;

^&smokers < non-smokers, p ≤ .017;

BIS, Barratt Impulsivity Scale-11; Cho, choline-containing compounds; Cr, creatine-containing compounds; Glu, glutamate; NAA, N-acetylaspartate; mI, myo-inositol; SI, spectroscopic imaging; SVS, single volume spectroscopy.

Group comparisons of SI metabolites concentrations

Main effects for smoking status were observed for NAA concentration in the lenticular nuclei [χ² (1)=8.54, p=.003], with a trend for in the frontal GM [χ² (1)=4.06, p=.041], where smokers demonstrated significantly lower NAA than non-smokers in the lenticular nuclei (see Table 2). No significant smoking status × age interactions were apparent in any region, and average lifetime drinks/month were not related to metabolite levels in any region (all p>.20). No differences were observed between smokers and non-smokers on SI voxel GM fraction or voxel count in any region (data not shown).

Group comparisons of decision-making, risk-taking, impulsivity measures

Smokers demonstrated a lower IGT net total score (indicative of poorer decision-making), and higher scores on the BIS motor, non-planning, and BIS total impulsivity score than non-smokers (see Table 2). No group differences were apparent on the BART average adjusted pumps.

Associations of SVS and SI metabolite levels with neurocognitive, decision-making, risk-taking, impulsivity, and smoking severity measures

In the entire cohort (smokers + non-smokers), there were significant moderate magnitude associations of right DLPFC Glu and NAA concentrations with multiple neurocognitive domain scores as well as with BIS non-planning impulsivity score; all correlations were in the expected direction (see Table 3). Similarly, there were numerous moderate magnitude relationships, in the expected direction, between regional SI NAA levels and neurocognitive domain scores; the SI regions most consistently associated with neurocognition were the frontal GM and lenticular nuclei, which also showed the NAA reductions in smokers (see Table 3). The direction and magnitude of the above reported associations was generally consistent for both smokers and non-smokers. No significant associations were apparent for regional metabolite levels with IGT and BART measures (all p>.20). In smokers, greater lifetime years of smoking was related to lower Glu in the ACC (r= −.40, p=.012) and right DLPFC (r= −.30, p=.044). Smoking severity measures and SI metabolites levels were not significantly related.

Table 3.

Associations of regional SVS and SI-derived metabolites with neurocognitive, decision-making, risk-taking, and impulsivity measures in the total sample (smokers + non-smokers)

Measure	Metabolite	Region	r*
Cognitive efficiency	NAA	DLPFC	0.28
		Lenticular nuclei	0.36
		Temporal white matter	0.31
	Glu	DLPFC	0.40
Executive skills	NAA	Frontal gray matter	0.32
	Glu	DLPFC	0.30
General intelligence	Glu	DLPFC	0.29
Processing speed	NAA	Lenticular nuclei	0.32
	Glu	DLPFC	0.35
Visuospatial learning	NAA	Frontal gray matter	0.47
		Temporal white matter	0.31
		Occipital white matter	0.36
		Thalamus	0.31
		Lenticular nuclei	0.38
	Glu	DLPFC	0.34
Visuospatial memory	NAA	Frontal gray matter	0.35
		Occipital white matter	0.33
		Thalamus	0.31
		Lenticular nuclei	0.38
	Glu	DLPFC	0.36
Visuospatial skills	NAA	DLPFC	0.28
	Glu	DLPFC	0.33
Global neurocognition	NAA	Frontal gray matter	0.32
		Lenticular nuclei	0.41
	Glu	DLPFC	0.37
BIS non-planning impulsivity	NAA	DLPFC	−0.33
	Glu	DLPFC	−0.36

^*Semi-partial correlation coefficient (adjusted for age, education, and lifetime average drinks/month); all reported correlations p < .05;

BIS, Barratt Impulsivity Scale-11; DLPFC, right dorsolateral prefrontal cortex; Glu, glutamate; NAA, N-acetylaspartate; mI, myo-inositol.

The pattern and effect sizes for the above reported findings were essentially unchanged when females were excluded from analyses.

Discussion

The primary findings from this study were: (1) Smokers, compared to non-smokers, showed significantly lower NAA, Cr, and mI concentrations in the right DLPFC, as well as lower NAA in the total frontal GM and lenticular nuclei; smokers also demonstrated significantly greater age-related decreases of NAA in the right DLPFC, and of Glu levels in the ACC and right DLPFC relative to non-smokers. (2) Smokers showed poorer performance on a measure of decision-making (IGT) and greater self-reported impulsivity (BIS-11) than non-smokers. (3) Across the total sample (i.e., smokers + non-smokers) higher right DLPFC NAA and Glu concentrations, as well as NAA in several lobar GM and WM regions and subcortical nuclei, were associated with better performance on multiple neurocognitive domains and lower (non-planning) impulsivity.

The significantly decreased NAA level in the right DLPFC, and lenticular nuclei exhibited by smokers indicates compromised neuronal integrity in these regions (13), and smokers showed notably greater age-related decreases in DLPFC NAA concentration. Smokers also showed notable trends for lower NAA in the ACC (p = .054) and the total frontal GM (p = .041; as measured with SI), with corresponding moderate effect size (ES = 0.50–0.67). Smokers showed decreased DLPFC Glu level and markedly greater age-related reduction in Glu compared to non-smokers; ACC Glu level did not differ between the groups, but, similar to the DLPFC, smokers demonstrated greater age-related decreases in ACC Glu concentration. These findings indicate smokers showed the greatest metabolite abnormalities in frontal lobe regions, which complement quantitative MRI studies that found young/middle-aged smokers demonstrated lower volumes of the ACC, DLPFC (25, 41, 42), total frontal GM (25, 43).

Non-smokers demonstrated significantly higher DLPFC mI and Cr, in addition to higher NAA and Glu levels. Elevated cerebral GM mI and Cr have been reported in conditions with pathologically confirmed neuroinflammation [e.g., HIV-infection and Alzheimer’s Disease (14)]. The synthesis of NAA, mI, Cr, and Glu, as well as the active transport of mI, Cr, and Glu across cell membranes are energetically demanding processes (13, 44). The average intercorrelation of these metabolite levels in the DLPFC of non-smokers was moderate in magnitude (r = 0.46); therefore, the elevated DLPFC mI and Cr concentrations, in conjunction with higher DLPFC NAA and Glu levels, likely reflects the coherence of the mitochondrial function (45, 46) of the neuronal and astroglial tissue in this region in non-smokers. Additionally, regional NAA (47) and Glu (15, 48) levels show age-related declines across adulthood. The greater age-related reductions in DLPFC NAA and Glu and ACC Glu concentrations demonstrated by smokers suggest that smoking is associated with abnormally accelerated aging effects on the metabolic integrity of tissue in these regions.

Smokers demonstrated a lower IGT total score and a higher BIS-11 total score, which are indicative of poorer decision-making and greater impulsivity, respectively. These findings are consistent with previous studies reporting compromised decision-making and greater self-reported impulsivity in smokers (49, 50).

Higher Glu in the DLPFC and higher NAA levels in the DLPFC and in several lobar GM and WM regions measured by SI were associated with better performance on multiple neurocognitive domains across groups. Higher DLPFC NAA and Glu were also related to lower self-reported impulsivity. Studies with cognitively-normal adults and individuals with various biomedical and psychiatric conditions consistently reported higher regional NAA concentrations were associated with better performance on a variety of neurocognitive measures (13, 51, 52). Glu is the primary excitatory cerebral neurotransmitter and mediates approximately 70% of central nervous system synaptic transmission (53). Higher basal ganglia Glu level was correlated with better neurocognitive performance in cognitively normal adults (15, 16). The lack of associations between ACC metabolites and neurocognition may be related to the perigenual location of our ACC volume. The perigenual region of the ACC subserves affective/emotional processes, while dorsal regions are indicated to be more involved in cognitive processes, such as error monitoring and attentional regulation (54). Both the ACC and DLPFC subserve multiple cognitive processes, including decision-making, risk-taking, and impulse control (55, 56); therefore, the lower DLPFC NAA and Glu and the trends for lower NAA levels in the ACC and total frontal GM in smokers suggest a disturbance in tissue metabolic integrity in those regions, particularly with increasing age, which may, at least partially, explain the poorer performance of smokers on the IGT and greater self-reported impulsivity on the BIS-11, and the deficient performance on multiple neurocognitive domains observed in this cohort in an earlier study [i.e.,(26)]. Overall, the findings reinforce the utility of MR spectroscopy-derived brain metabolites as practical biomarkers of regional neurobiological integrity and neurocognition. See (1, 11) for a review of the potential mechanisms by which chronic smoking promotes neurobiological and neurocognitive dysfunction.

This study has limitations that may affect the generalizability of the findings. Unrecorded premorbid/comorbid group differences in lifestyle or biomedical conditions (e.g., diet/nutrition, exercise, subclinical pulmonary or cardiovascular dysfunction) and/or genetic polymorphisms [see (57)] may have influenced the results. The small number of females precluded assessment for sex effects.

This study contributes novel information to the expanding body of evidence that cigarette smoking in young/middle-aged individuals is associated with significant age-related neurobiological abnormalities, particularly in anterior frontal regions implicated in the development and maintenance of addictive disorders. Longitudinal studies on the effects of smoking cessation on the regional brain metabolites measured in this study, with a greater number of females, are clearly warranted to determine if the observed metabolite and abnormalities are persistent or normalize with smoking cessation.