Functional Imaging of Craving

Daniel W. Hommer, M.D.

To visualize brain activity associated with mental states, such as craving for alcohol and other drugs (AODs), researchers have begun to use functional imaging techniques. Three commonly used techniques are single photon emission computed tomography (SPECT), positron emission tomography (PET), and functional magnetic resonance imaging (fMRI). Studies using these three approaches have been reviewed in order to evaluate the validity of a proposed model of the brain regions involved in alcoholism and the craving for alcohol. This model suggests a central role for a connected group of brain regions that include the basal ganglia, thalamus, and orbital cortex. A study using SPECT technology in alcoholics, however, found altered brain activity in only some of those regions during craving. Additional studies in alcoholics, as well as cocaine users, identified several other brain regions whose activities appeared to change in response to craving. These studies have led to the development of a revised model of brain regions involved in craving for AODs. Numerous questions remain, however, that must be answered before the brain areas involved in craving can be identified conclusively. KEY WORDS: AOD (alcohol and other drug) craving; single photon emission computed tomography; positron emission tomography; magnetic resonance imaging; brain imaging; blood flow measurement; cocaine; brain; basal ganglia; limbic system; brain function; neurotransmission; dopamine; glucose metabolism; literature review

rimates, including humans, are neurochemical processes in specific brain tions already exist between the inside visual animals—that is, they like regions (i.e., brain states). Although and the outside of a resting or inactive Pto see things—giving rise to the one cannot see mental states, one can neuron, the process of moving addi- adage “seeing is believing.” One cannot generate images that help visualize the tional ions to generate a nerve signal see thoughts, feelings, or mental states, intensity and location of physiological requires energy. In the brain, this energy however, whether one’s own or those processes associated with brain states. is supplied by the brain’s metabolism of of other people. One can only perceive At the most basic level, the physio- the sugar glucose, which is delivered these states either directly, in one’s own logical processes underlying brain states to the brain through the bloodstream. consciousness, or indirectly, through involve changes in and the maintenance Accordingly, when a certain brain the reports and behavior of other people. of differences in the concentrations of “Craving” is a term derived from pop- molecules that carry electrical charges ular psychology that is used to describe (i.e., ions) across the membrane form- DANIEL W. H OMMER, M.D., is chief of one of these mental states—namely, the ing the boundary of each nerve cell, or the Section of Brain Electrophysiology intense desire for a certain object or neuron. Changes in the concentrations and Imaging, Laboratory of Clinical experience (e.g., alcohol or other drugs of various ions inside and outside of the Studies, Division of Intramural Clinical [AODs]). cell serve as signals that can be trans- and Biological Research, at the National Cognitive neuroscience postulates mitted from one neuron to another. Institute on Alcohol Abuse and Alcoholism, that mental states are the product of Because differences in ion concentra- Bethesda, Maryland.

Vol. 23, No. 3, 1999 187 region is actively involved in the gener- to be detected or quantified accurately people engaged in an activity unrelated ation of mental states, the energy by simple visual inspection. Therefore, to drinking. At least in theory, the only requirements of that region increase a new image representing the differ- difference between the mental states and, consequently, so does the blood ences between the two original images evoked by the two videos would be flow there. By measuring blood flow or (i.e., a difference image) is created using that one video would induce a desire to glucose metabolism (a process known computerized analysis. This can be done drink, and the other video would not. as functional brain imaging), one can because each image is made up of small Thus, the difference image produced in determine which brain areas are most units, or pixels, each of which repre- this experiment would represent the active during a particular brain state. sents a small brain area. For each pixel blood flow or glucose metabolism asso- This article reviews the small num- the blood-flow or glucose-metabolism ciated with alcohol craving. However, ber of studies that have applied func- researchers cannot be absolutely certain tional brain imaging to investigating that any differences between the brain brain states associated with craving for states induced by the two videos would AODs, particularly alcohol and cocaine. By measuring result only from the presence or absence It includes studies on cocaine craving of craving. Other influences not related because (1) to date, only two studies blood flow or glucose to craving, such as differences between have attempted to image craving for the two videos with regard to their alcohol, and (2) the brain states associ- metabolism, one can visual properties (e.g., color and move- ated with craving may be similar for all determine which brain ment), also may affect the difference AODs. The article first describes the image. Consequently, it is important three most commonly used functional areas are most active that investigators try to minimize dif- imaging techniques. It then introduces during a particular ferences between the stimuli used to a preliminary model of the functional evoke the desired brain states other anatomy of craving to provide a frame- brain state. than the one difference that they are work for understanding the studies on attempting to visualize. craving conducted among alcoholics and cocaine abusers. After comparing value in the first image is “subtracted” Single Photon Emission the results of studies that examine the from the value in the corresponding Computed Tomography preliminary model, the article discusses pixel in the second image. For example, future directions in the functional assume that the blood flow in a given SPECT is the oldest functional imaging imaging of craving. pixel is low in the first brain state (e.g., technique and generally provides images no craving) and is assigned a value of 2 with the lowest resolution, although on a scale from 0 to 10. In the second newer SPECT equipment provides Methods of Functional brain state (e.g., craving), the blood flow images with significantly greater detail. Brain Imaging in the same pixel is high (i.e., is assigned With this approach, the subject is injected a value of 9). Thus, in the difference with a weakly radioactive tracer—usu- Currently three major techniques are image, this pixel will have a blood flow ally 99mtechnetium-hexamethylpropy- used to visualize the brain activity asso- value of 7, representing a relatively large lene-amineoxide (99mTc-HMPAO)— ciated with various mental states: single difference between the two brain states. to measure blood flow. This technique photon emission computed tomography This subtraction process is performed allows the investigator to visualize the (SPECT), positron emission tomogra- for all pixels in the images, generating blood flow distribution within the phy (PET), and functional magnetic an image that represents the extent of brain during the first 1 or 2 minutes resonance imaging (fMRI). Each tech- changes in blood flow between the two after 99mTc-HMPAO injection. The nique involves measuring local changes brain states being assessed. Statistical major advantage of this approach is in cerebral1 blood flow or metabolism. analysis of several difference images that it is more widely available and To that end, researchers usually obtain then allows the identification of the much less expensive than other imag- at least two separate images representing brain regions whose activities vary most ing techniques, such as PET. two different states of the brain regions strongly between the two brain states. SPECT also has substantial disadvan- of interest. In theory, one could just visu- The key to this approach is to select tages, however. The biggest disadvan- ally compare the two images to identify tasks that will induce brain states that tage is that because 99mTc-HMPAO brain regions in which brain activity differ only with regard to the mental decays slowly (i.e., has a long half-life), differs depending on mental state. In state being investigated. In a study of the two images needed to produce a many cases, however, the differences alcohol craving, for example, investiga- difference image must be recorded sev- between the two images are too subtle tors might have alcoholics watch two eral days apart, leading to two practical different videotapes: One video might problems. First, each time a subject is 1For definitions of this and other technical terms show people drinking alcoholic beverages, placed in the machine (i.e., the scanner) used in this article, see the glossary, p. 196. whereas the other video might show that measures the radioactivity and

188 Alcohol Research & Health Functional Imaging of Craving

generates the image, his or her posi- eration of more accurate images com- already delivered its oxygen to the tis- tioning will be slightly different. This pared with 15O PET. A disadvantage sues (i.e., deoxygenated hemoglobin). inconsistency complicates the genera- of the prolonged measurement period This fMRI approach is based on the tion of an accurate difference image, needed to generate one image, how- fact that oxygenated and deoxygenated because the two original images must ever, is that FDG PET allows the gen- hemoglobin differ slightly in their mag- be aligned so that their size, shape, and eration of only one pair of images, and netic properties, resulting in the emis- orientation match perfectly. Second, thus of one difference image, per day. sion of a signal that can be detected by mental states and the resulting brain A major advantage of PET over special scanners. These scanners are so states are likely to vary more when they SPECT is that it generates images with sensitive that they can detect signals are recorded a few days apart than when a greater resolution: The decay of each from a volume (i.e., voxel) of brain they are recorded a few minutes or hours radioactive isotope used in PET (i.e., tissue as small as a few millimeters on apart. As a result of the long separation 15O and 18F) eventually generates two each side. The signal, which is propor- between scans required by SPECT, dif- photons that can be detected by a spe- tional to the ratio of the concentration ference images can reflect changes in cial camera. Conversely, the decay of of oxygenated to deoxygenated hemo- brain states merely associated with the each tracer molecule used in SPECT globin inside the voxel, is called the passage of time rather than specifically generates only one photon, thereby blood oxygen level-dependent (BOLD) with the mental state under investigation. allowing less precise localization of the signal. tracer molecules. When a particular brain region is Positron Emission Tomography PET also is associated with two major being used, the small arteries that sup- disadvantages, however. First, because ply the region with oxygen and other For PET, subjects are injected with the 15O and 18F tracers have such short nutrients expand. As a result, the molecules that naturally occur in the half-lives, they must be generated amount of oxygenated hemoglobin tissues, such as water (H2O) and glucose, immediately before use—a labor-inten- flowing into that region increases. In and which have been tagged with a sive and expensive process. Thus, a fact, the amount of oxygenated hemo- radioactive variant (i.e., isotope) of one machine to generate these tracers (i.e., globin flowing into the tissue far of their atoms. These radioactive variants a cyclotron) must be located on-site at exceeds the amount of deoxygenated are taken up into the tissues just like every PET facility. Second, because hemoglobin produced by any increased the normal molecules, and their distri- researchers can produce only a few images metabolism in the active brain tissue. bution can be measured using a scanner. per day for each subject, statistical anal- Thus, the ratio of oxygenated to deoxy- The most commonly employed PET yses combining data from several (i.e., genated hemoglobin—and, therefore, technique measures blood flow using usually at least six) subjects are necessary the BOLD signal—increases in the 15 radioactively labeled [ O] H2O. Each to generate reliable results. Images pro- activated tissue, a change that can be PET image represents the blood flow duced from a group of subjects, how- measured by fMRI. during the first 1 or 2 minutes following ever, have considerably lower resolution fMRI offers many advantages over 15O administration. This approach than images produced from a single earlier functional imaging techniques. offers several advantages over SPECT. subject, because human brain anatomy First, fMRI does not expose subjects For example, because 15O has a half-life varies slightly from person to person. to radioactive substances, allowing of only a few minutes, six to eight images researchers to study each subject on can be produced during one imaging Functional Magnetic Resonance multiple occasions. Second, investiga- session. This feature reduces imaging Imaging tors can collect numerous images dur- problems that result from the subject’s ing a single imaging session, thereby movement, incorrect alignment of the fMRI is the newest functional imaging allowing statistical analysis of data images, and day-to-day variations in technique. Like normal MRI, this obtained from a single subject. This mental state. approach does not use radioactive tracers, ability to analyze data from individual Another tracer used in PET imaging but exposes the subject to radio waves subjects means that the investigator can is radioactively labeled glucose (i.e., in the presence of a strong magnetic avoid the reduction in accuracy that [18F]-fluorodeoxyglucose [FDG]). This field. This treatment causes the nuclei occurs when brain images from many compound, which is taken up by cells of certain atoms in the brain to emit subjects are combined. Third, fMRI that are in the process of generating signals that can be detected by a scanner. measurements represent brain activity energy from glucose breakdown (i.e., The most commonly used form of fMRI during only a few seconds, enabling metabolically active cells), is then trapped analyzes the distribution of hemoglobin, researchers to detect rapid changes in within the cells so that its location can the molecule in red blood cells that brain activity (i.e., enhancing temporal be imaged. In contrast to 15O PET’s helps transport oxygen to the tissues. resolution). (PET and SPECT cannot short measurement period, FDG PET Specifically, this approach determines measure brain activity during such a measures neuronal glucose use during the ratio of hemoglobin that is still short interval.) Finally, the spatial reso- a period of 20 to 30 minutes. This carrying oxygen (i.e., oxygenated lution of fMRI exceeds that of the extended measurement allows the gen- hemoglobin) to hemoglobin that has other modalities.

Vol. 23, No. 3, 1999 189 A Preliminary Model proposed to be affected connects the behavioral context (i.e., ongoing behav- of the Brain Regions orbital cortex, ventral striatum (includ- ior, motivation, and memories of simi- Involved in Craving ing the nucleus accumbens and portions lar circumstances) and provides the of the medial caudate nucleus), and basis for selecting which small cortical Modell and colleagues (1990) developed thalamus. (For a list of the functions of regions, or circuits, need to be activated. the first model of neural systems that these and other brain regions mentioned In the human brain, multiple variants may play a role in alcoholism and par- in this article, see table 1 below. For the of these circuits affect different regions ticularly in craving. The model is based locations of those regions, see the figure of the cortex and function in parallel on the observation that the emotions on p. 191.) Failure of this circuit may (Alexander et al. 1986). The particular associated with craving (as determined cause poor control of urges to drink. circuit proposed to be relevant to alco- by patients’ self-reports) are similar to Striatal-thalamocortical loops are found hol craving involves the orbital cortex. the compulsions described by patients in all terrestrial vertebrates (Reiner et al. The orbital cortex is part of the lim- with obsessive-compulsive disorder 1998). In higher mammals, these cir- bic system, the brain system that is (OCD). Because OCD appears to be cuits augment activity in small regions thought to deal with emotions. One of associated with dysfunction in a brain of the cortex that are involved in cogni- the orbital cortex’s functions is to evaluate area called the orbital cortex, Modell tion, emotion, memory, or some spe- the motivational significance of stimuli and colleagues (1990) argued that a cific actions, simultaneously inhibiting that predict reward (Tremblay and similar dysfunction could underlie craving other nearby cortical regions. Striatal- Schultz 1999). In addition, the orbital for alcohol. Specifically, the investiga- thalamocortical loops receive nerve sig- cortex inhibits actions that are inappro- tors postulated a dysfunction in a type nals (i.e., input) from a set of cortical priate for the person’s current situation of brain circuit called a striatal-thalam- regions with related functions. This (i.e., context-inappropriate actions). ocortical loop. The particular circuit input serves to monitor the current Damage to the orbital cortex can lead

Table 1 Various Brain Regions Involved in Craving and Their Main Functions

Brain Region Function Subcortex

Striatal thalamocortical loop Connects functionally related areas of the cortex with the striatum, which projects to the thalamus via the globus pallidus; in the thalamus, information is combined with information from the related cortical areas and then sent back to one region of the frontal cortex Functions to select one output (e.g., emotion, thought, or motor action) from among several competing potential responses based on behavioral context

Nucleus accumbens Connected to limbic systems and orbital prefrontal cortex; related to emotion and motivation (ventral striatum) (1)

Caudate nucleus Connected to medial and lateral prefrontal cortex; related to cognition and motivation (dorsal striatum) (2)

Thalamus (3) Relays information between cortical regions and other brain areas, including the striatum

Amygdala (4) Learning of stimuli predicting reward and punishment; fear response

Cortex Prefrontal cortex Working memory; planning and motor preparation; inhibition of context-inappropriate behavior

Orbital (5) Olfaction, behavioral inhibition, and evaluation of stimuli’s motivational significance

Dorsolateral (6) Spatial working memory and cognition

Anterior cingulate (7) Attention and motivation

Insular Cortex (8) Visceral function, olfaction, taste, and emotion

NOTE: The numbers in parentheses refer to the locations of the brain regions in the accompanying four-part figure in this article (see page 191).

190 Alcohol Research & Health Functional Imaging of Craving

to antisocial behavior. Thus, it is rea- weight [g/kg]). The alcohol dose was brain regions that they expected to be sonable to expect that a circuit serving too small to produce a direct pharma- activated during craving. Those ROIs, this region may be involved in craving. cological effect, but the researchers which included the right and left caudate hypothesized that the dose would nucleus, the thalamus, and the orbital induce a desire to drink more alcohol. frontal cortex, were chosen based on Craving Studies The subjects, who knew that they were the neural systems model of alcoholism in Alcoholics receiving either water or alcohol but discussed in the previous sections of did not know which one they were this article. Of these regions, only the Although several imaging studies have receiving on a given day, always right caudate nucleus showed a signifi- examined the effects of craving among received the water on the first day of cant increase in blood flow (and thus cocaine abusers, only one such study scans and alcohol on the second day of brain activity) during the craving con- has successfully induced and assessed scans. Self-ratings obtained following dition. In addition, the extent of the craving among alcoholics (Modell and the SPECT scans confirmed that the subjects’ self-reported desire for alcohol Mountz 1995). In that study, the inves- low alcohol dose did, in fact, induce a predicted the magnitude of the increase tigators used SPECT technology to desire to drink in most subjects. in caudate blood flow following alcohol examine changes in cerebral blood flow To facilitate the analysis, the investi- administration. in nine alcoholics2 who had ingested gators conducted a region-of-interest Modell and Mountz (1995) inter- either water or a low dose of alcohol (ROI) analysis in which they measured preted these results as confirming (0.03 grams per kilogram of body changes in blood flow only in a few Modell and colleagues’ (1990) neural systems model of alcoholism. The lack of activation in the other components of the proposed circuit (i.e., the thala- mus and the orbital cortex), however, weakens that interpretation. Further- more, the results are suggestive at best, because the study included no nonalco- holic control subjects. Accordingly, it is uncertain whether the increased caudate activation resulted from alcohol admin- istration itself or from the craving elicited by alcohol. The use of control subjects would have enabled researchers to distinguish between the two inter- pretations; the control subjects presum- ably would have exhibited similar blood flow changes if these changes had resulted from alcohol’s pharmacological effects but would not have experienced changes resulting from craving. Despite these limitations, however, the correlation between craving score and magnitude of blood flow change suggests that the activation of the right caudate nucleus may be related to craving. Only one other published imaging Locations of anatomical brain regions associated with craving as shown on a series of four MRI cross-sections taken from the top of the head (i.e., coronal slices). The study has analyzed brain regions poten- images were generated using a TÐ1-weighted magnetic resonance scan of a normal tially associated with craving in alcoholics. human brain. This type of scan shows the brain as it usually appears in fresh or In that study, Hommer and colleagues fixed post-mortem sections, with white matter appearing white, gray matter appear- (1997) attempted to induce craving for ing gray, and the fluid surrounding the brain (i.e., the cerebrospinal fluid) appearing alcohol by injecting 18 hospitalized dark. alcoholics3 and 12 control subjects with

2All study participants were treated for alcoholism NOTE: Numbers indicate the following brain regions: 1 = nucleus accumbens; 2 = caudate nucleus; 3 = thala- mus; 4 = amygdalas; 5 = orbital prefrontal cortex; 6 = dorsolateral prefrontal cortex; 7 = anterior cingulate cortex; on an outpatient basis and had been abstinent for 8 = insular cortex. For a description of the function of these brain regions, see table 1. at least 1 week. MRI = magnetic resonance imaging. 3All of these subjects had been abstinent for at least 3 weeks.

Vol. 23, No. 3, 1999 191 m-chlorophenylpiperazine (mCPP), • Two groups of neurons located Craving Studies an agent that has been used to induce bilaterally in the thalamus (i.e., the in Cocaine Abusers craving among alcoholics. Widely used anterior and medial dorsal nuclei of in psychiatric research, mCPP has been the thalamus) In contrast to the few imaging studies a probe for the activity of neuronal sys- conducted on alcohol craving, six studies tems that use the neurotransmitter sero- • The dorsolateral prefrontal cortex have investigated craving for cocaine. tonin to relay signals from one neuron Three of those studies used drug-related to the next. For this signal transmission, • The left insular cortex cues to elicit craving for cocaine, and serotonin released by the signal-emitting three studies used drug administration neuron must interact with docking • A gyrus on the temporal cortex (one study used cocaine itself and two molecules (i.e., receptors) on the surface located at the side of the brain (i.e., studies used the chemical methylpheni- of the signal-receiving neuron. Several the left middle temporal gyrus) date [Ritalin®]). Grant and colleagues different serotonin receptors exist. The (1996) used FDG PET to compare agent mCPP primarily mimics serotonin’s • A gyrus on the cingulate cortex (i.e., brain glucose metabolism in 13 non- effects at one particular serotonin recep- the posterior cingulate gyrus). treatment-seeking cocaine abusers and tor (i.e., the 5HT2C receptor) while 5 control subjects. The two scanning blocking serotonin’s effects at other Among the alcoholic subjects, mCPP sessions, separated by 1 week, took receptors. Alcohol administration also significantly increased brain glucose place while the subject viewed video- results in serotonin release and activa- metabolism in larger areas of the cere- tapes of objects associated with cocaine tion of the 5HT2C receptor. bellum and posterior cingulate gyrus use or objects used in crafts. An ROI To identify brain regions involved than it did in the nonalcoholics. Con- analysis demonstrated that among the in craving, Hommer and colleagues versely, the alcoholics showed a smaller control subjects, no significant differences (1997) determined changes in glucose area of mCPP-induced activation in the in brain glucose metabolism existed in metabolism following injection of mCPP thalamus compared with the healthy response to the two sets of cues. Among or an inactive saline solution (i.e., a volunteers, almost no activation in the the cocaine abusers, however, brain placebo) using FDG PET. In all cases orbital cortices, and no activation in metabolism increased significantly in the placebo was always administered the head of the caudate nucleus and response to the cocaine-related cues in first, followed by mCPP 60 minutes the dorsolateral prefrontal cortex. several regions of the cortex, including later. Although other studies have found These results indicate that among the medial orbital, dorsolateral prefrontal, that mCPP can elicit both a desire to nonalcoholics, treatment with a substance and medial temporal (including the drink and subjective effects similar to that induces or mimics serotonin release amygdala) cortices as well as other cor- alcohol intoxication among alcoholics, (i.e., a serotonergic challenge) activates tical regions.5 No significant changes in none of the alcoholics in this study striatal-thalamic circuits involving the brain metabolism occurred in the basal (who were in the PET scanner when orbital and dorsolateral prefrontal cor- ganglia or thalamus. In addition, the being assessed) expressed any increase in tices and that these responses are reduced, subjects’ self-ratings of craving showed a desire to drink or experienced any alcohol- or blunted, among alcoholics. Unfort- significant positive correlation with meta- like effects. In fact, neither alcoholics unately, the relevance of these findings bolism in the dorsolateral prefrontal nor control subjects reported any signif- for identifying brain regions involved cortex and the medial temporal cortex. icant subjective effects following mCPP in craving remains unclear. The brain Recently Childress and colleagues administration other than a mild regions activated by mCPP in nonalco- (1999) reported the results of an 15O increase in anxiety. Both groups, how- holics include the parts of the striatum, PET study of cocaine craving that ever, exhibited increases in brain glucose thalamus, and cortex that have been included 14 treatment-seeking cocaine metabolism. Among the control sub- implicated in mediating craving for abusers and 6 non-cocaine-abusing jects, the increases occurred in the fol- alcohol by Modell and colleagues (1990). control subjects. The investigators per- lowing brain regions: Conversely, alcoholics showed a blunted formed six sequential scans on each response to mCPP in these regions. subject, both during rest periods and • Two ridges (i.e., gyri) on the surface The findings could indicate that the while the subjects were watching cocaine- of the orbital cortex (i.e., the right difficulties alcoholics experience with related and non-cocaine-related video- medial and posterior orbital gyrus) craving and with controlling their alco- tapes. Using ROI analysis the study hol intake are associated with lower- demonstrated that in the cocaine abusers, • The cerebellar4 hemispheres on both than-normal activity in this circuit. In but not in the control subjects, cocaine sides of the brain (i.e., bilaterally) the absence of studies demonstrating a link between subjective craving and the 4The cerebellum is a brain region located at the • The left nucleus accumbens function of striato-thalamic and orbito- lower back of the brain that is primarily involved in frontal circuits, however, this model for motor functions. • The head of the caudate nucleus the functional neuroanatomy of craving 5These additional cortical regions include the retro- bilaterally remains speculative. splenial, temporo-parietal, and peri-striate cortices.

192 Alcohol Research & Health Functional Imaging of Craving

craving elicited by the videotapes resulted by a rapid return to pre-cocaine levels correlated with craving for cocaine. How- in increased blood flow in the amygdala correlated with euphoria. Conversely, the ever, the relevance of this correlation is and anterior cingulate cortex and decreas- activity of the regions that showed a unclear, because both the release of dop- ed blood flow in the caudate nucleus sustained increase in activity (i.e., the amine and the concentration of dopamine and in the globus pallidus, which con- nucleus accumbens/subcallosal gyrus receptors are low in the thalamus. nects the striatum and the thalamus. region and the basal forebrain) correlated The videotapes that Childress and positively with craving. Finally, the time Comparison of the Study Results colleagues (1999) used to elicit cocaine course of the BOLD signal in the amyg- craving also were used in an fMRI study dala negatively correlated with craving. The six imaging studies conducted among of six cocaine abusers (whose treatment Two studies used methylphenidate cocaine abusers and the one successful status was not specified) and six control to elicit craving in cocaine abusers who study conducted in alcoholics identified subjects (Maas et al. 1998). The study had been hospitalized for treatment for numerous brain regions that appear to was designed to determine whether the at least 3 weeks. In one of those studies, be involved in craving (for a summary, changes observed in the two previously changes in brain glucose metabolism see table 2, p. 194; also see the figure on described studies (Childress et al. 1999; were measured in 20 cocaine abusers page 191 for the locations of those brain Grant et al. 1996) also could be demon- before and after intravenous adminis- regions). The specific pattern of brain strated using fMRI. To this end the tration of methylphenidate (Volkow et regions that is activated appears to depend investigators used a ROI analysis of 10 al. 1999). ROI analysis demonstrated to some extent on the technique used brain regions that had been identified in that methylphenidate increased brain to elicit craving. One group of studies the other two studies as being responsive glucose metabolism in the anterior cin- induced craving by administering small to cocaine craving. Among those regions, gulate gyrus, right thalamus, and cere- amounts of the abused drug itself or however, only the anterior cingulate gyrus bellum. Furthermore, methlyphenidate- of an agent with similar effects. Those and the left dorsolateral prefrontal cortices induced increases in brain glucose studies reported increased activity in were activated in the cocaine abusers, and metabolism in the right orbital frontal the thalamus during craving. In con- only the results for the anterior cingulate cortex and right caudate nucleus corre- trast, studies that used visual cues (i.e., gyrus reached statistical significance. lated positively with craving for cocaine. videotapes) to induce craving reported Another fMRI study of cocaine crav- Because the study included no control increased brain activity primarily in the ing used cocaine administration to induce subjects, however, it is unclear whether dorsolateral prefrontal cortex. both euphoria and craving in 10 non- this pattern of activation is unique to The prefrontal cortex generally is treatment-seeking cocaine abusers (Breiter cocaine abusers or would also occur activated during tasks involving both et al. 1997). The study included no among non-drug users. working memory (i.e., the short-term control subjects. Cocaine administration The other study involving meth- storage of information) and episodic induced BOLD-signal increases in a ylphenidate examined changes in the memory (i.e., the recall of sequences large number of limbic and other brain release of the neurotransmitter dopamine of past events). Accordingly, activation regions, whereas a saline solution caused using PET technology and an agent of the dorsolateral prefrontal cortex only small changes in the BOLD sig- called 11C-raclopride, which can inter- during cue-induced craving may occur nals measured in the lateral prefrontal act with a specific receptor for dopamine because craving based on visual cues and temporo-occipital cortices. The (i.e., the D2 receptor) (Volkow et al. requires the activation of certain memory largest increases following cocaine admin- 1997). Dopamine release may be asso- circuits, whereas such memories may istration occurred in the region of the ciated with the pleasurable or rewarding not be required for the occurrence of nucleus accumbens, caudate nucleus, effects of AODs and with the motivation drug-induced craving. It is unclear why thalamus, and anterior cingulate and to use AODs to experience those reward- cue-induced craving fails to activate the insular cortices, as well as in a few other ing effects. Methylphenidate causes thalamus, because both drug-induced brain regions.6 In the amygdala the the release of dopamine in the brain, and cue-induced craving are associated BOLD signal decreased in response to which then displaces 11C-raclopride with activity changes in the anterior cocaine administration. bound to D2 receptors. Thus, changes cingulate gyrus, caudate nucleus, amyg- The researchers took advantage of in 11C-raclopride binding provide a dala, and orbital cortices. All these the fact that fMRI allows the collection measure of how much dopamine is regions, along with the thalamus, are of many images during a brief period released in response to methylphenidate components of various basal ganglia- of time and examined the relationship administration. The study found that thalamocortical circuits; because of between subjective response and the time compared with control subjects, cocaine their anatomical connections, one course of brain activation. They found abusers exhibited decreased dopamine would expect these regions to be func- that the activity of the brain regions that release in the striatum, specifically the tionally linked (Alexander et al. 1986). showed an early peak response followed caudate nucleus, but increased release Although the studies discussed in in the thalamus. Furthermore, the mag- this article indicate that the brain regions 6These additional areas include the subcallosal nitude of thalamic dopamine release mentioned earlier are activated during gyrus, putamen, and basal forebrain region. among the cocaine abusers was positively craving, one cannot exclude the possibility

Vol. 23, No. 3, 1999 193 that other regions are activated as well. nucleus accumbens and insular cortex, all participate in various basal ganglia- All but two of the studies used ROI as well as the other activated regions, are thalamocortical circuits, which influ- methods of image analysis and therefore components of striatal-thalamocortical ence a person’s sequential choice of examined only limited portions of the circuits. Although the precise function behaviors (e.g., actions, thoughts, and brain. In fact, most of the studies limited of those circuits in humans is unknown, emotions) based on the behavioral their analysis to the caudate nucleus; they are thought to be involved in context (see table 1). Three of those thalamus; amygdala; and orbital, dor- emotion, motivation, and social behav- regions (i.e., the caudate nucleus, thala- solateral frontal, and anterior cingulate ior (Alexander et al. 1986). In addition, mus, and orbital frontal cortex) were cortices. The only two studies to analyze the insular cortex is involved in visceral included in the initial craving model data from the entire brain (Breiter et al. functions related to feeding and emotion. developed by Modell and colleagues 1997; Hommer et al. 1997) detected (1990). In addition, the dorsolateral activation in the nucleus accumbens and prefrontal cortex may be involved in insular cortex in addition to the regions A Revised Model of the craving, as discussed earlier in this listed above. (The results of the study by Functional Neuroanatomy article. The anterior cingulate cortex, Hommer and colleagues are not sum- of Craving which was activated in most studies of marized in table 2, because the subjects craving, is involved in the regulation of in that study did not report craving The brain regions identified as being attention and emotion (Posner and during the scanning procedure.) The involved in craving for alcohol or cocaine Rothbart 1998) and may be particu-

Table 2 Changes in Activity in Several Brain Regions Under Conditions of Craving for Alcohol or Cocaine as Determined in Studies Using Various Funtional Imaging Techniques

Imaging Dorsolateral Anterior Technique Nucleus Caudate Orbital Frontal Cingulate Insular Used Accumbens Nucleus Thalamus Amygdala Cortex Cortex Cortex Cortex

Alcohol SPECT Right only (Modell and Mountz 1995) + correlation Cocaine

FDG PET (Grant et al. 1996) + correlation + correlation 15O PET (Childress et al. 1999) fMRI Left only (Maas et al. 1998) + correlation + correlation fMRI (Breiter et al. 1997) + correlation Ð correlation FDG PET Right only Right only Right only Right only (Volkow et al. 1999) + correlation + correlation 11C-raclopride PET (Volkow et al. 1999) + correlation

NOTE: Arrows represent increases and decreases in activity; “+ correlation” and “– correlation” refer to a positive or negative correlation, respectively, between the degree of brain activity and the extent of craving; the designations “right” and “left” refer to structures located in the right and left hemispheres of the brain, respectively. SPECT = single photon emission computed tomography; FDG = fluorodeoxyglucose; PET = positron emission tomography; fMRI = functional magnetic resonance imaging; 15O = radioactively labeled oxygen; 11C = radioactively labeled carbon. Raclopride is an agent that interacts with a receptor for the neurotransmitter dopamine.

194 Alcohol Research & Health Functional Imaging of Craving

larly important in learning associations rons of the midbrain. Thus, it appears (Schultz 1998). This observation leads to between certain stimuli and rewards that dopamine-releasing neurons pro- two questions. First, if a burst of dop- (Bussey et al. 1997). Finally, the amyg- vide a signal indicating the presence amine neuron activity signals the organ- dala is commonly associated with fear of an important new association to be ism to learn an association, does a decrease but also relays information about stim- learned between a stimulus and a sub- in dopamine neuron activity indicate uli associated with reward to the orbital sequent reward. With continued pair- that a connection must be “unlearned”? cortex (Hatfield et al. 1996). ings of the stimulus and the reward, Second, if a burst of dopamine neuron The brain regions associated with however, the dopamine-releasing neu- activity leads to a pleasant or arousing craving also are generally associated rons stop responding, indicating that state, does a decrease in dopamine neuron with motivation (Koob et al. 1998), a those neurons do not react to stimuli activity lead to a dysphoric state, per- state that can be described as inducing with a well-established, consistent rela- haps akin to frustration? Although the those behaviors that are related to satis- tionship to a reward. In other words, answers to those questions are unknown, fying an organism’s needs. Further- once an association is learned, no further the questions themselves suggest that the more, with the exception of the thalamus, dopamine activity is required to main- model of craving underlying the design of all these brain regions receive informa- tain that connection. In fact, if dopamine the studies reviewed here is too simplistic. tion from dopamine-releasing neurons activity indicates that the organism To date, all functional imaging studies that are located in the midbrain but must learn a new association between have treated craving as a state charac- extend their axons to these regions and a stimulus, a behavior, and a reward, terized by only one pattern of behaviors release dopamine there. Dopamine is then responses of dopamine-releasing or emotions (i.e., a unitary state). The often considered a reward neurotrans- neurons in familiar, unchanging situa- studies by Schultz (1998) and the result- mitter whose release provides the drive tions might even be counterproductive. ing questions, however, indicate that behind behavior undertaken in antici- Both alcohol and cocaine increase the desire to drink elicited by a cue or pation of a pleasurable outcome (i.e., dopamine concentrations throughout low alcohol dose in a situation in which appetitively motivated behavior). Con- the various basal ganglia-thalamocorti- the subject knows that more alcohol is versely, reductions in dopamine activity cal circuits that have been implicated in available likely differs from the desire to may be involved in motivating behav- craving in the studies reviewed in this drink elicited in a situation in which ior undertaken in order to avoid the article. Consequently, both drugs should further alcohol consumption is impos- loss of pleasure (i.e., aversively moti- put a person’s brain into a state similar sible. Although subjects in one of the vated behavior) (Schultz 1998). This to the state that occurs when a person studies reviewed here were allowed to bivalent nature of dopamine’s motiva- makes the initial connection between a use cocaine after the cue exposure and tional function may mirror the mixture stimulus or behavior and a desired reward. scanning were completed (Grant et al. of motivations that appear to comprise Such a state may be both pleasant and 1996), no study has systematically craving. On the one hand, craving can arousing and therefore may encourage examined the effect of the perceived result from an alcoholic’s desire for the further drug consumption (i.e., is rein- availability of the drug on brain states pleasurable effects of alcohol (i.e., an forcing). In addition, the elevated dop- associated with craving. appetitive drive). On the other hand, amine concentrations should prime the craving also can result from an alcoholic’s person to associate the stimuli preceding desire to avoid the unpleasant experi- alcohol or cocaine use (e.g., certain objects, The Future of Functional ences of withdrawal or another negative places, people, thoughts, or feelings) with Imaging of Craving emotional state (i.e., an aversive drive). the pleasure experienced when the drug Thus, craving may involve either aver- is taken. Thus, when those stimuli are Although the imaging studies discussed sive or appetitive motivation and, at encountered again, they should initially in this article have begun to shed light times, perhaps both simultaneously. elicit a burst of activity in the dopamine- on the brain regions involved in craving, Studies of the activities of individual releasing neurons, thereby leading to future studies should address several dopamine neurons in the monkey brain anticipation of the reward and AOD- additional issues. First, researchers in response to certain stimuli (Schultz seeking behavior. This scenario corre- should design studies that can distinguish 1998) have helped generate a framework sponds to appetitively motivated craving. the effects of appetitive and aversive for understanding the neurophysiology What happens, however, when an motivation. For example, experiments of craving and may provide suggestions alcoholic encounters stimuli associated could compare the effects of craving on for the design of future imaging stud- with alcohol use in a situation in which brain activity under conditions in ies. In those studies, animals repeatedly he or she cannot drink? The experi- which the subject knows that AODs were exposed to arbitrary stimuli (i.e., ments in monkeys mentioned earlier will or will not be available. colored lights) that predicted the also demonstrated that when a stimulus Second, imaging studies in healthy opportunity to gain a reward. When which previously had been paired with people without AOD addiction should the stimulus initially was paired with a reward stopped predicting that reward, attempt to identify patterns of brain a reward, it rapidly elicited a burst of appearance of that stimulus soon began activity generally associated with appet- activity in the dopamine-releasing neu- to elicit a decrease in dopamine activity itive and aversive motivation (i.e., not

Vol. 23, No. 3, 1999 195 GLOSSARY Axon: Part of the nerve cell (i.e., neuron) that extends to other Neurotransmitter: A chemical messenger used by nerve cells to neurons and transmits information to those cells. relay signals from one cell to the next; common neurotrans- Basal ganglia: Groups of nerve cells located above the thala- mitters include dopamine and serotonin. mus in the white matter of the brain. These nerve cells, Photon: The smallest quantity of electromagnetic energy. This which include the caudate nucleus, putamen, and pallidum, energy may occur in the form of X-rays, gamma rays, or a help control motor and cognitive activity. quantum of light. Cerebral: Pertaining to the cerebrum, the largest and upper- Resolution: The ability of an imaging technique to distinguish most section of the brain, which is divided into the right between two adjacent structures; greater resolution indicates and left hemispheres. better image quality. Half-life: The time required for one-half of a given amount of Tracer: A radioactive molecule used in imaging techniques that a radioactive compound to decay. enables the investigator to visualize a biological process. associated with AODs). The results of References ethanol reward and dependence. Alcoholism: such studies could be compared with Clinical and Experimental Research 22:3–9, 1998. ALEXANDER, G.; DELONG, M.; AND STRICK, P. MAAS, L.C.; LUKAS, S.E.; KAUFMAN, M.J.; WEISS, the patterns of brain activation found Parallel organization of functionally segregated cir- R.D.; DANIELS, S.L.; ROGERS, V.W.; KUKES, T.J.; during craving for AODs, thereby cuits linking basal ganglia and cortex. Annual Review AND RENSHAW, P.F. Functional magnetic reso- enabling researchers to understand of Neuroscience 9:357–381, 1986. craving in the context of how the brain nance imaging of human brain activation during BREITER, H.C.; GOLLUB, R.L.; WEISSKOFF, R.M., cue-induced cocaine craving. American Journal of deals with motivation in general. KENNEDY, D.N.; MAKRIS, N.; BERKE, J.D.; GOOD- Psychiatry 155:124–126, 1998. Third, although this review has MAN, J.M.; KANTOR, H.L.; GASTFRIEND, D.R.; MODELL, J.G., AND MOUNTZ, J.M. Focal cerebral RIORDEN, J.P.; MATHEW, R.T.; ROSEN, B.R.; AND focused on dopamine, researchers know blood flow change during craving for alcohol mea- that many other neurotransmitters are HYMAN, S.E. Acute effects of cocaine on human brain activity and emotion. Neuron 19:591–611, 1997. sured by SPECT. Journal of Neuropsychiatry and involved in motivational processes. Thus, Clinical Neurosciences 7:15–22, 1995. the neurotransmitters gamma-aminobu- BUSSEY, T.J.; MUIR, J.L.; EVERITT, B.J.; AND ROBBINS, MODELL, J.G.; MOUNTZ, J.M.; AND BERESFORD, tyric acid (GABA), glutamate, and opiate T.W. Triple dissociation of anterior cingulate, poste- rior cingulate, and medial frontal cortices on visual T.P. Basal ganglia/limbic striatal and thalamocorti- peptides—and, perhaps, serotonin—all discrimination tasks using a touchscreen testing cal involvement in craving and loss of control in help mediate the reinforcing properties procedure for the rat. Behavioral Neuroscience alcoholism. Journal of Neuropsychiatry and Clinical of alcohol (Koob et al. 1998). Once the 111:920–936, 1997. Neurosciences 2:123–144, 1990. nature of the brain states associated CHILDRESS, A.R.; MOZLEY, P.D.; MCELGIN, W.; POSNER, M.I., AND ROTHBART, M.K. Attention, with craving has been characterized in FITZGERALD, J.; REIVICH, M.; AND O’BRIEN, C.P. self-regulation and consciousness. Philosophical more detail, it may be possible to Limbic activation during cue-induced cocaine crav- Transactions of the Royal Society, London: Biological explore systematically how drugs that ing. American Journal of Psychiatry 156:11–18, 1999. Science 353:1915–1927, 1998. selectively affect specific neurotransmit- GRANT, S.; LONDON, E.D.; NEWLIN, D.B.; REINER, A.; MEDINA, L.; AND VEENMAN, C.L. ters can modify those brain states. VILLEMAGNE, V.L.; LIU, X.; CONTOREGGI, C.; Structural and functional evolution of the basal In summary, the concept of craving, PHILLIPS, R.L.; KIMES, A.S.; AND MARGOLIN, A. ganglia in vertebrates. Brain Research and Brain which has its origin in popular psychol- Activation of memory circuits during cue-elicited Research Reviews 28:235–285, 1998. cocaine craving. Proceedings of the National Academy ogy, may not correspond to unitary of Sciences of the USA 93:12040–12045, 1996. SCHULTZ, W. Predictive reward signal of dopamine brain states observable with functional neurons. Journal of Neurophysiology 80:1–27, 1998. imaging techniques. By carefully design- HATFIELD, T.; HAN, J.S.; CONLEY, M.; GALLAGHER, M.; AND HOLLAND, P. Neurotoxic lesions of baso- TREMBLAY, L., AND SCHULTZ, W. Relative reward ing and analyzing functional imaging lateral, but not central, amygdala interfere with preference in primate orbitofrontal cortex. Nature studies, however, researchers may be Pavlovian second-order conditioning and reinforcer 398:704–708, 1999. able to determine whether it would be devaluation effects. Journal of Neuroscience 16: VOLKOW, N.D.; WANG, G.J.; FOWLER, J.S.; LOGAN, 5256–5265, 1996. more useful for researchers to divide J.; GATLEY, S.J.; HITZEMANN, R.; CHEN, A.D.; the process of craving into two or more HOMMER, D.; ANDREASON, P.; RIO, D.; WILLIAMS, DEWEY, S.L.; AND PAPPAS, N. Decreased striatal components. If functional imaging is W.; RUTTIMANN, U.; MOMENAN, R.; ZAMETKIN, dopaminergic responsiveness in detoxified cocaine- to be more than colored pictures that A.; RAWLINGS, R.; AND LINNOILA, M. Effects of m- dependent subjects. Nature 386:830–833, 1997. chlorophenylpiperazine on regional brain glucose humans use to convince themselves VOLKOW, N.D.; WANG, G.J.; FOWLER, J.S.; utilization: A positron-emission tomographic com- HITZEMANN, R.; ANGRIST, B.; GATLEY, S.J.; that psychological constructs have a parison of alcoholic and control subjects. Journal of physical reality in the brain, researchers Neuroscience 17:2796–2806, 1997. LOGAN, J.; DING, Y.S.; AND PAPPAS, N. Association must be able to use this approach to of methylphenidate-induced craving with changes test hypotheses and change their mod- KOOB, G.F.; ROBERTS, A.J.; SCHULTEIS, G.; in right striato-orbitofrontal metabolism in cocaine PARSONS, L.H.; HEYSER, C.J.; HYYTIA, P.; MERLO- abusers: Implications in addiction. American els of how the mind works. PICH, E.; AND WEISS, F. Neurocircuitry targets in Journal of Psychiatry 156:19–26, 1999.

196 Alcohol Research & Health