Do brain responses to emotional images and cigarette cues differ? An fMRI study in smokers

Participants

Smokers motivated to quit smoking and willing to participate in a clinical trial of smoking-cessation medications were recruited via local (Houston metropolitan area) radio, print and internet advertising. As part of this trial, participants were asked to complete two fMRI scanning sessions. The first session occurred before randomization to a medication group and while still smoking at their regular rate, while the second session occurred 24 h after quitting and while still receiving study medications. This article only presents data from the first session. The data from the post-quit fMRI session and its relationship to smoking cessation outcomes will be the focus of future articles, at the conclusion of the ongoing clinical trial.

Inclusion criteria for participation in the study were as follows – 18–65 years old, self-reported smoking of at least 10 cigarettes per day, baseline expired-air carbon monoxide (CO) > 10 ppm, not currently meeting diagnostic criteria for psychiatric or substance use disorders (other than smoking), being free of contraindications for the study medications (bupropion and varenicline), being fluent in English, having a working telephone, being right-handed (to reduce variability introduced by differences in brain lateralization), and self-reported European ancestry (to maintain a homogeneous sample for future genetic analyses). Of the 50 individuals who met the inclusion criteria, 40 completed the pre-quit fMRI session. Scanner-related equipment failure resulted in the loss of fMRI data for five participants, resulting in a total of 35 participants (14 female) for this analysis. Table 1 lists demographic and smoking-related characteristics for the group. All participants provided written informed consent, and the research was approved by the University of Texas MD Anderson Cancer Center Institutional Review Board. Participants were compensated $50 for completing each fMRI session.

Stimuli

We based our picture-viewing task on a paradigm that has been used to reliably measure brain responses to emotional pictures (Bradley et al., 2003; Sabatinelli et al., 2007a). Participants viewed 64 pictures from the International Affective Picture System (IAPS; Lang et al., 2005), and from smoking-related picture collections previously used in our laboratory (Carter et al., 2006) and elsewhere (Gilbert & Rabinovich, 1999). Pictures were selected from four categories of affective valence, including pleasant (eight high-arousal erotic pictures and eight low-arousal romantic pictures), unpleasant (eight high-arousal mutilation pictures and eight low-arousal sad pictures), neutral (16 neutral people) and cigarette related (16 pictures of people smoking). The high-arousal and low-arousal subcategories were selected to have the same arousal levels (based on the IAPS normative ratings) in both the pleasant and unpleasant valence categories.¹ The slides were displayed for 3 s each, separated by a 12-s intertrial interval, and delivered in a pseudorandom order with respect to category, with the restriction that no more than two pictures of the same category would be presented consecutively. Stimuli were presented with a Pentium 4 computer running e-prime software (Psychology Software Tools, Pittsburgh, PA, USA) and an MR-compatible stimulus-presentation system (IFIS-SA; Invivo, Orlando, FL, USA) such that they subtended a horizontal viewing angle of approximately 15 °.

Procedure

All laboratory sessions took place between 09:00 and 12:00 h. Participants were allowed to smoke ad libitum before the laboratory session to ensure that they would be in a non-deprived state. Upon arrival at the laboratory, the participant provided an expired-air CO sample to confirm that they had smoked, and completed smoking status and mood questionnaires. After completing the questionnaires, participants were escorted to the fMRI scanning suite, placed into the scanner, and instructed to simply look at the stimuli. The MR images were acquired using a GE Healthcare (Milwaukee, WI, USA) 3.0 T Signa HDxt (v 15.0) magnetic resonance scanner with TwinSpeed gradients in zoom mode (40 mT/m, 150 mT/m-s slew rate), and an eight-channel, high-resolution, transmit/receive brain volume coil (MRI Devices, Waukesha, WI, USA). After a short acclimation period, the BOLD signal was measured using a T2*-weighted, echo-planar imaging protocol, while the participants completed the picture-viewing task. The slice prescription consisted of 50 coronal slices (2.5 mm thick, 0.5 mm gap) that covered most of the brain, except for the most anterior and most posterior regions of the prefrontal and occipital cortices, respectively. The parameters for this pulse sequence were as follows – matrix size = 64 × 64; field-of-view = 160 × 160 mm; in-plane resolution = 2.5 × 2.5 mm; TR = 3000 ms; TE = 25.3 ms; flip angle = 90 °. The acquisition of each volume was time-locked to the slide presentation, such that, after discarding the first two volumes to allow T1 magnetization to reach steady-state, one volume was acquired while the slide was on and four volumes were acquired during the intertrial interval, for a total of 335 volumes for the entire picture-viewing task. After the picture-viewing task, a high-resolution (1 mm³ voxels), three-dimensional, T1-weighted anatomical image was obtained using a magnetization-prepared, fast spoiled gradient echo (IR-3DFSPGR) pulse sequence (TR = 6.432 ms, TE = 2.1 ms, inversion time = 400 ms, flip angle = 20 °). After obtaining the anatomical image, the participant was removed from the scanner and completed a post-experimental questionnaire before dismissal until the next experimental session.

Data reduction and analysis

The goal of this study was to determine whether BOLD responses to cigarette cues exceeded those to other categories of emotional stimuli. Primary analysis consisted of two steps. First, we identified regions of interest (ROIs) on the basis of a whole-brain analysis of the fMRI data that compared BOLD responses with cigarette and neutral cues, an approach followed in previous studies (McBride et al., 2006; Brody et al., 2007; McClernon et al., 2008, 2009; Janes et al., 2009). Second, we compared the BOLD responses across all stimulus categories within the ROIs identified in step 1 (Bradley et al., 2003). In addition, we conducted a secondary analysis comparing BOLD responses across all stimulus categories within two theoretically important ROIs that were not found to be significantly active in the whole-brain analysis from step 1. This analysis included the amygdala and the nucleus accumbens.

Participant characteristics

Table 1 shows demographic and smoking characteristics of the 35 participants included in the analyses.

Identification of ROIs from the whole-brain analysis

To identify ROIs for detailed comparisons of the BOLD response among cigarette cues and emotional stimuli, we examined the results of a whole-brain analysis of BOLD responses to cigarette vs. neutral cues. This analysis indicated that viewing cigarette related pictures, compared with neutral pictures, heightened BOLD activity in the visual association cortex in the occipital, posterior parietal and inferior temporal lobes, the dorsal striatum, the cingulate gyrus, the dorsolateral prefrontal cortex and the insula. Peak activation coordinates and statistics for the activated regions are listed in Table 2.

BOLD responses within each ROI

Visual association cortex

The whole-brain analysis revealed that responses to cigarette cues differed from neutral cues in several areas of the visual association cortex in the occipital, posterior parietal and temporal lobes (Fig. 1A). The clearest pattern of significant activation emerged bilaterally in the cuneus, inferior parietal lobule and middle temporal gyrus. Comparison of BOLD responses across all stimulus categories revealed a large increase in BOLD signal in response to erotic pictures, followed by all other emotional and cigarette-related pictures, and finally by neutral pictures (Fig. 1B), as shown by a statistically significant category × time interaction (F_20,15 = 5.32, P < 0.001, λ = 0.12), which was the result of significant differences between stimulus categories during the 3-, 6- and 9-s post-stimulus onset volumes (F_5,30 > 3.6, P < 0.05). The BOLD response peaked during the 6- and 9-s post-stimulus onset volumes, and the largest differences between picture categories were observed during this interval. Thus, subsequent analyses focused on the percent change in BOLD signal from baseline averaged across these two volumes. Figure 1C shows the BOLD responses in this time window averaged across the active clusters in the visual association cortex. The main effect of picture category was significant (F_5,30 = 12.27, P < 0.001, λ = 0.33), and post hoc comparisons showed that responses to erotic pictures were significantly larger than those to all other picture categories (P < 0.001), and that responses to neutral pictures were significantly smaller than responses to all other picture categories (P < 0.05). The significant quadratic trend (F_1,34 = 30.01, P < 0.0001) showed that in the visual cortex, BOLD activation increased as picture content became more emotionally arousing.

Dorsal striatum

The whole-brain analysis found another area of significant cigarette > neutral cue reactivity in the head of the right caudate nucleus. Across all picture categories, the pattern of results in this cluster was similar to that seen in the visual association areas (Fig. 2A and B; stimulus category F_5,30 = 4.67, P < 0.01, λ = 0.56), in which BOLD responses to erotic pictures were significantly greater than responses to all other picture categories (P < 0.05). Although the differences were not statistically significant (corrected P-values were between 0.11 and 0.39), BOLD responses to cigarette-related, romantic, sad and mutilation pictures appeared to be larger than responses to neutral pictures, similar to the result observed in the visual association areas. In fact, the significant quadratic trend (F_1,34 = 17.16, P < 0.0005) confirmed the relationship between emotional arousal and BOLD activation.

Insula

The whole-brain analysis also found a cluster of significant cigarette > neutral reactivity in the left insula (Fig. 3; main effect of picture category, F_5,30 = 2.71, P < 0.05, λ = 0.69). In this region, comparison of BOLD responses across all picture categories revealed a pattern of results that was substantially different from that seen in the other regions, as cigarette-related pictures, rather than erotic, elicited the largest BOLD response. Although Tukey HSD post hoc tests found no significant differences between individual group means (P > 0.1), an exploratory polynomial contrast of BOLD responses to cigarette-related pictures vs. the mean of all other picture categories was statistically significant (F_1,34 = 11.51, P < 0.01). This effect was primarily due to decreased BOLD responses relative to baseline in response to mutilation, neutral and pleasant pictures, and a small increase in BOLD responses to cigarette-related pictures.

BOLD responses in theory-based ROIs

Amygdala

The amygdala ROI was defined as two 512-mm³ regions centered at ±20, −4, −11 mm in Talairach–Tournoux space (Fig. 4A, inset). The mean percent signal change from 6 to 12 s after picture onset over all voxels in these cubes was extracted and analysed for differences between picture categories. For non-cigarette-related pictures, the BOLD response varied as a function of the emotional arousal level of the picture (Fig. 4; main effect of picture category F_5,30 = 3.87, P < 0.01, λ = 0.61), with the largest responses to the most highly arousing pictures (erotica and mutilation), and the smallest response to neutral pictures (quadratic trend of category –F_1,34 = 5.82, P < 0.05). The BOLD response to cigarette-related pictures did not significantly differ from the BOLD response to neutral pictures.

Reactivity to cigarette and intrinsically pleasant cues

Through a learning process, chronic drug use is believed to result in the attribution of increased incentive salience to drug cues. Thus, major theories of addiction predict that brain regions that are involved in learning, memory and reward processing should be active in response to drug cues (Koob & LeMoal, 2001; Hyman, 2005; Volkow et al., 2010). Consistent with these models, we found evidence of significant cigarette cue reactivity in the visual association cortex, dorsal striatum, anterior cingulate cortex, prefrontal cortex and insula, and a trend toward significance in the nucleus accumbens. These areas have previously been implicated in reactivity to cigarette cues (Due et al., 2002; David et al., 2005; McClernon et al., 2005, 2008, 2009; Brody et al., 2007; Franklin et al., 2007; Janes et al., 2009, 2010a,b), alcohol cues (George et al., 2001; Heinz et al., 2004, 2007) and cocaine cues (Wexler et al., 2001; Kosten et al., 2006; Volkow et al., 2006). Specifically, we found that the magnitude of the BOLD response to cigarette cues closely resembled the magnitude of the response to mildly arousing, pleasant stimuli (i.e. the romantic pictures). This finding is consistent with self-report studies in which smokers reported cigarette cues as pleasant and mildly arousing (Orain-Pelissolo et al., 2004; Carter et al., 2006), studies that showed similar startle inhibition for pleasant and cigarette-related cues in smokers (Geier et al., 2000; Cinciripini et al., 2006), and studies that compared smokers’ brain reactivity with emotional and cigarette-related cues using event-related potentials (Versace et al., 2010, 2011). These results converge in supporting the hypothesis that chronic smoking leads to overvaluation of cigarette cues by the brain’s reward circuits.

In addition to an overvaluation of drug cues, chronic drug use is believed to lead to a devaluation of natural rewards and their associated cues (Koob & LeMoal, 2001; Volkow et al., 2010). For example, in the medial prefrontal cortex, cocaine addicts, compared with non-addicts, have been shown to have larger BOLD responses to cocaine videos and smaller responses to erotic videos (Garavan et al., 2000). The lack of alternative means of activating the brain’s reward circuitry may be one reason that makes long-term drug abstinence a goal difficult to achieve (Volkow & Li, 2005). Indeed, previous research has demonstrated that detoxified alcoholics with the largest BOLD responses to pleasant pictures in the ventral striatum were more likely to be abstinent at follow-up than those with smaller responses to pleasant pictures (Heinz et al., 2007), suggesting that responding to natural rewards may have a protective effect against relapse.

Building on previous neuroimaging studies that examined responses to drug cues in isolation, this study directly compared smokers’ brain responses to cigarette cues and naturally rewarding erotic stimuli using fMRI. Contrary to the results described above for cocaine users, we found that smokers had larger BOLD responses to erotic pictures than to cigarette-related pictures in nearly all ROIs, including the medial prefrontal cortex, where the effect was previously found in cocaine users (Garavan et al., 2000).

At first glance, this result may be interpreted as being inconsistent with the hypothesis that chronic drug use results in decreased sensitivity to naturally rewarding stimuli (Volkow et al., 2010), and with the data from cocaine users and animal models that support this hypothesis (Garavan et al., 2000; Grigson & Twining, 2002). However, there are several alternative explanations for the differences in response to erotic stimuli between the studies. Although cocaine and nicotine produce similar physiological effects throughout the brain, there are subtle differences between the actions of each drug, especially in the dorsal striatum and anterior cingulate cortex (Merlo Pich et al., 1997), which is where, in comparison to drug cues, the cocaine users were less reactive (Garavan et al., 2000) and the smokers more reactive to erotic stimuli. There are also important methodological differences between the studies. For example, the smokers in our study were entering a cessation program, but the cocaine users did not report serious plans to quit. Also, the erotic stimulus was presented to the cocaine users in the form of a single motion picture, whereas we used a set of several erotic photographs randomly interspersed within a series of stimuli from several categories. These differences might explain why smokers did not show the same suppression of BOLD responses to erotic stimuli as cocaine addicts did. Further research is necessary to determine whether this discrepancy is primarily due to the differences in the neuroadaptations following cocaine and nicotine addiction, or to procedural differences between smoking and cocaine cue reactivity studies.

Cigarette-specific cue reactivity in the insula

A notable exception to our finding of larger BOLD responses to erotica than cigarette-related pictures was found in the left insula, where responses to cigarette cues were larger than responses to all other stimulus categories. Several fMRI studies have found significant responses to cigarette cues in the insula (McClernon et al., 2005; McBride et al., 2006; Brody et al., 2007; Franklin et al., 2007; Janes et al., 2009, 2010a,b), but the lack of highly-arousing control stimuli in those studies made it difficult to determine whether the insula was selectively responsive to cigarette-related cues over other emotional stimuli. Our study extended the cue reactivity literature by suggesting that BOLD responses in the insula have some degree of selectivity to cigarette cues. Interestingly, animal models also provide support for selective processing of drug cues over natural rewards in the insula – inactivation of the insula disrupted nicotine self-administration and cue-induced reinstatement of nicotine-seeking, but had no effect on food self-administration or food-induced reinstatement of food-seeking (Forget et al., 2010). Thus, diminished cue reactivity may be one of the reasons why smokers who suffer damage to the insula region find it easy to quit smoking (Naqvi et al., 2007).

Although the insula is involved in the expression of drug-taking behavior in humans (Naqvi et al., 2007) and animal models (Contreras et al., 2007), and insular cue reactivity is associated with relapse (Janes et al., 2010a), the specific processes by which the insula influences this behavior are unknown. One possibility is that insular activity in response to cigarette cues mediates an interoceptive conditioned response that is subjectively correlated with expecting the ‘taste’ of a cigarette or the sensory qualities of smoking (Naqvi & Bechara, 2010). The insula has been implicated in interoceptive conditioning in general (Craig, 2002), and in taste perception specifically (Pritchard et al., 1999). In fact, the insula has been shown to be responsive to capsaicin-induced stimulation of the upper airway (Mazzone et al., 2007), a phenomenon similar to the effects of cigarettes on the upper airway (Behm & Rose, 1994).

Another possible role of the insula in cigarette cue reactivity may be in the subjective experience of cigarette ‘urge’ or ‘craving’, and cognitive efforts to resist that craving. The insula is interconnected with the anterior cingulate gyrus and medial prefrontal cortex (Mesulam & Mufson, 1982; Mufson & Mesulam, 1982), which are involved in reward-related decision-making (Liu et al., 2011). Data in support of this hypothesis come from a study that found a significant correlation between the amount of self-reported cigarette craving and insula activity to cigarette cues (Brody et al., 2007). However, other studies did not replicate this finding (McClernon et al., 2005; Franklin et al., 2007). Thus, further research is needed to clarify the relationship between craving and cue reactivity in the insula, as well as to address the other possible roles of the insula in smoking behavior, such as interoceptive conditioned responses.

Potential limitations and future directions

Because we drew our sample from a clinical trial of smokers seeking treatment for smoking cessation, a limitation of our study is the lack of a control group of non-smokers. The lack of a control group makes it difficult to determine whether the BOLD responses that we observed were due to chronic smoking, or whether they would have been evident in never smokers or light smokers as well. However, the possibility that the effects observed in this study were not due to chronic smoking is unlikely. In fact, our results are similar to those observed in other studies of smokers that also included non-smokers, as the cigarette cue reactivity that characterized smokers was not evident in the non-smokers (Due et al., 2002; David et al., 2005; Rubinstein et al., 2011). A second limitation is that our sample included only European-American, treatment-seeking smokers who met strict inclusion criteria for a clinical trial and therefore may not be representative of the population of smokers as a whole (e.g. ethnic minorities, smokers not attempting to quit). Future research must therefore extend sampling to more representative populations of smokers.

To evaluate BOLD responses to cigarette-related stimuli relative to other pleasant pictures differing in emotional arousal, we included both romantic (low arousing) and erotic (high arousing) pictures. Previous studies suggest that BOLD responses to erotic pictures are greater in males than in females (Lang et al., 1998; Hamann et al., 2004; Sabatinelli et al., 2004), which could bias the comparison between erotic and cigarette cues when averaging across a sample of males and females. However, we found no evidence of sex differences in BOLD responses to erotic stimuli in our analyses. The lack of sex differences in BOLD activation might be due to differences between the current sample (middle-aged, treatment-seeking smokers) and the samples used in previous research (undergraduate students). The lack of sex differences may also be the result of our identification of ROIs based on reactivity to cigarette cues rather than reactivity to erotic stimuli (Lang et al., 1998; Hamann et al., 2004; Sabatinelli et al., 2004). Also, we did not ask our participants about their sexual orientation, and because all erotic stimuli depicted heterosexual couples, we may have missed potential differences in the BOLD response to the erotic stimuli across participants of differing sexual orientation (Paul et al., 2008). However, large differences in the BOLD response as a function of sexual orientation are unlikely. Previous studies have found that, although men and women produce differences in self-reported affect across images of erotic couples, opposite sex nudes, and same sex nudes, their psychophysiological responses to these three categories of erotic picture are equal (Bradley et al., 2001).

Differences in the type of reinforcement provided by erotic and cigarette cues is also an issue that should be considered when interpreting the results of this study. Cigarette cues are not intrinsically reinforcing, but are instead thought to become motivationally significant through repeated pairing with the reinforcing effects of smoking (Koob & LeMoal, 2001; Robinson & Berridge, 2003; Hyman, 2005; Volkow et al., 2010). Erotica, however, appears to be intrinsically reinforcing (e.g. Hoffmann et al., 2004). This suggests that processes other than the drug-related content stimuli may have contributed to the differences in BOLD response between these two classes of picture. One of these additional factors may have been primary vs. secondary reinforcement processes, which have been shown to produce different patterns of BOLD response (Sescousse et al., 2010). Thus, an important direction for future research is to directly compare drug- vs. non-drug-related primary reinforcers and drug- vs. non-drug-related secondary reinforcers.

Previous studies have shown that the patterns of BOLD response to emotional pictures can be altered by the presence of active tasks during picture presentation (Hariri et al., 2000; Ochsner & Gross, 2005, 2008; Pessoa et al., 2005; Blair et al., 2007). Hence, to avoid confounds due to the presence of secondary tasks, we opted for a passive viewing paradigm. The lack of overt behavioral responses during picture presentation, however, does not allow us to be certain that participants maintained attentiveness throughout the whole paradigm. Our finding of significantly greater activation to highly arousing pictures than neutral pictures in the visual association areas and the amygdala replicates several previous fMRI studies of picture viewing (Lane et al., 1997; Bradley et al., 2003; Phan et al., 2004; Sabatinelli et al., 2007b), and suggests that the participants in our study were indeed paying attention to the pictures.

In our experiment we used a fixed, 12-s interstimulus interval. The selection of this interval was based on previous fMRI studies of emotional picture viewing (Lang et al., 1998; Bradley et al., 2003; Sabatinelli et al., 2004, 2005, 2007a,b). However, the use of a fixed interstimulus interval in event-related fMRI research is problematic because: (i) the participants may be able to predict when the stimuli are presented; and (ii) the amplitude of the BOLD signal may be decreased due to variability in the duration of the hemodynamic response that is not captured by a fixed interstimulus interval (e.g. Burock et al., 1998).

Despite these concerns, the current study provides a significant extension beyond existing smoking cue reactivity research. By including intrinsically pleasant and unpleasant pictures in addition to neutral ones, we were able to conclude that smokers, although more reactive to cigarette-related pictures than to neutral ones, were not more reactive to cigarette-related pictures than to intrinsically pleasant, erotic pictures. We also found evidence suggestive of a selective bias toward cigarette-related pictures, but not toward other emotional pictures, in the insula. It is therefore possible that smoking-cessation interventions that target the insula may be more effective than existing therapies at reducing cue-induced craving and relapse (Naqvi & Bechara, 2009). Thus, future research into the neurophysiological basis of the insula’s response to cigarette cues is warranted.