11

GETTING THE CUE

AS I WALK down the street in downtown Bethesda on a warm Thursday evening, an inexplicable, intense urge to smoke hits me out of the blue. This is quite strange. I have not smoked more than the occasional cigarette in the close to twenty years since I met my wife. Yet in an instant, it is all upon me. I have the intense feeling that something is missing. My mouth is wetting in anticipation of the taste of a smoke, and I’m almost feeling a miniature preview of that taste, out of thin air. Similarly, my head is going just a tiny bit dizzy, the way it would do a lot more were I really to inhale some smoke. Yes, this is strange indeed. Why would this long forgotten set of feelings be upon me now, after a busy day at work, heading for the restaurant I favor for an occasional happy hour with colleagues from the lab? It is only a moment later that the answer becomes clear to me. As I turn the corner, standing on the sidewalk are two men, engaged in a conversation, each drawing heavily on a cigarette. I realize now that a faint smell of smoke must have hit me before I saw them. That, in itself, is not at all strange. The strange thing is that my first percept was not “there is a smell of cigarette smoke here.” The first percept was “I need a smoke.”

Here is a less benign version of the same thing. In 1977, when much of addiction psychiatry was still busy debating what kind of unresolved intrapsychic conflicts caused and maintained addiction and what kind of psychoanalytic approach might be appropriate to treat it, Chuck O’Brien brought heroin addicts into the lab at the University of Pennsylvania. Over several weeks he and his colleagues used an experimental procedure to let subjects learn to associate two simple stimuli—a 700 Hz tone and an odor—with the experience of heroin withdrawal that the researchers induced using a medication.¹ At the end of the experimental series, it was sufficient to present the tone and the odor alone, without any drug injection, for subjects to experience all the signs of withdrawal. These signs included the well-known objective symptoms, such as an increased rate of breathing, decreased skin temperature, and changes in pupil diameter. But subjects also reported experiencing the subjective feelings of withdrawal, which include an urge to take heroin. Clearly stimuli that were initially completely unrelated to any drug experience had by now somehow taken on a role as triggers for an entire coordinated program of responses, one that included unpleasant bodily as well as psychological elements. These responses were of the same nature that out on the street would provide a powerful incentive to resume drug seeking and taking.

I recall Chuck telling us a rather sad story related to this shortly after I first met him, one that brings these experimental findings out of the lab and into the real world of addiction better than most things I have since read in research papers. Chuck told us of giving a lecture for a group of physicians. As he had explained the objectives, the experimental design, and finally the results of the study described above, he said, a person in the audience, seated in the front row, seemed more and more distressed. As the lecture unfolded, the listener became increasingly tense and pale and started fidgeting with his hands. As often is the case after a good talk, people came up afterward and tried to catch the speaker for some additional questions. When his turn finally came, the distressed listener started asking questions that made it clear that for him, this was personal. In the end he explained that he was a successful anesthesiologist. In those days, that was a high-risk profession for developing opioid addiction because of easy access to morphine and less strict routines for controlling its use in hospitals than we have today.² After several years of morphine addiction, he had successfully kicked the habit, which is an impressive achievement. He had then worked for several years without relapsing. But Chuck’s vivid description of the experimentally induced withdrawal had been enough of a cue for him to precipitate a powerful urge for morphine.

Patients and clinicians alike have long known that individuals with addictive disorders report craving and relapse when exposed to stimuli associated with their drug use, be that the street corner, the bar, or the drinking buddies. Understanding how these stimuli start playing an important role for triggering craving and relapse is probably one of the most important questions facing addiction research. Our ability to develop science-based therapies in part relies on our ability to understand mechanisms of cue-induced relapse and find ways of preventing this phenomenon from occurring. Abraham Wikler of the University of Kentucky is often credited with introducing the concept of conditioned drug cues as an important relapse factor, for instance, in a much cited paper from 1973.³. This initial proposition was based on clinical experience—some might say anecdotes—and theorizing. The Science paper by O’Brien and colleagues was the first scientific demonstration that the phenomenon postulated by Wikler actually happens in people. It was followed by work in the alcohol field carried out by Peter Monti at Brown University, who established that letting an alcohol-dependent patient handle and smell—but not consume—his or her preferred alcoholic beverage gave rise in a vast majority of patients to intense alcohol craving. Yet even after these demonstrations, it took a while before the importance of these findings was fully appreciated.

Unsurprisingly, the extinction–reinstatement model introduced in the previous chapter has also provided solid support for a critical role of drug-associated cues for relapse. Rats or other animals that have learned to self-administer drug in the presence of previously neutral stimuli and then undergone extinction will quickly resume lever pressing when they are reexposed to the drug-associated cues, much the same way as they do following a priming dose of the drug. This phenomenon is reliably reproduced for all major classes of addictive drugs, including alcohol. To date, many laboratories, including my own, rely on this model for our attempts to understand relapse induced by drug memories and the brain machinery that produces it. The model is also widely used in efforts to develop medications that could help prevent relapse triggered by drug or alcohol cues. In these experiments, we use cues that are very similar to those used in the original human experiments, such as a tone or an odor. The use of this animal model has allowed studies of brain circuitry that contributes to cue-induced relapse. A caveat is that most of the mapping studies have examined relapse to cocaine seeking. We do not know at this stage how similar the circuitry is for different drugs. Nevertheless, a part of the amygdala,⁴ a structure in the temporal lobe that is critical for emotional aspects of learning, has through this work reliably been identified as a brain center that is critical for relapse triggered by drug-associated cues.

But we can’t ask experimental animals whether they really crave drug when they run back to a lever and resume pressing it. At the same time, in attempting to identify the brain networks that produce subjective states such as craving in humans, research encounters two major obstacles. One is practical, the other conceptual. On a practical level, subjective states may be possible to assess in human subjects with addictive disorders through the use of carefully designed self-report measures, but not until recently did these studies allow an analysis of what brain circuitry generates the subjective states. Perhaps more important, many hard-nosed neuroscientists are inherently skeptical whether people can reliably report what is going on in their brains, such as craving, or causing their behavior, such as relapse. This is not to question the honesty of people who provide the self-reports; it is merely to consider the limits of our ability to look into our own brains.⁵ Introspection may have fed an entire business of psychoanalysis but has simply not stood up to the test of science to provide us with an understanding of how the brain works to produce the mind. In the end, this is perhaps not entirely shocking. After all, you would not rely on introspection to find out if you have a brain tumor either, would you?

Advances in human functional brain imaging have to some extent helped break through these barriers. Using one of these imaging techniques, positron emission tomography, a groundbreaking paper in 1996 opened up a line of research that has since produced a convergence of human and animal studies.⁶ That paper and many others that followed examined the relationship between cravings reported in response to drug cues and patterns of brain activity. In that study experienced cocaine users were shown videos of drug cues, such as white powder, a mirror, a razorblade, and a rolled-up dollar bill. Their responses were compared with those of research volunteers who had never used cocaine. As expected, showing the drug paraphernalia resulted in reports of powerful craving in the cocaine users but not in the participants who did not have experience with cocaine. The cocaine-user group also responded to the cocaine-related cues with increased brain activity, measured as consumption of sugar by nerve cells, which was visible in the PET images. Among the brain regions activated were parts of the frontal lobe and the amygdala,⁷ which I have already mentioned is critical for emotional aspects of learning and memory. The greater the reports of craving, the greater the activity in these brain areas, the researchers found. In contrast, the volunteers who had never used cocaine and did not report cocaine cue-induced craving showed no signs of cue-induced brain activity in this area. Many similar studies have since followed, with increasingly sophisticated methodologies. Studies using both PET and fMRI have looked at the effects of drug-associated cues such as pictures of glasses or bottles with alcoholic beverages. Depending on the drug, experimental conditions, and study design, somewhat different patterns of network activations have been seen, but overall it seems clear that the subjective state of craving is associated with distinct changes in brain activity patterns. The intensity of these changes frequently correlates with the degree of craving.

The patterns that have emerged from the fMRI studies are slightly different from those initially suggested by the PET studies. This may be related to several things. As mentioned, the early PET studies employed sugar utilization to measure how nerve cell activity changed in response to the stimuli. Also, because of the nature of the method, the PET studies sum up data over a period of time that lasts many minutes. In contrast, fMRI studies rely for activity measures on how much of the oxygen nerve cells extract from blood and presumably use up. Because of that, these studies are able to measure second-to-second variation. One interesting study that used fMRI found that alcohol-associated cues, compared with cues that were neutral, resulted in activation of the nucleus accumbens, which is a key node of brain reward circuitry, as well as areas in the frontal lobes of the brain.⁸ An important finding was that the magnitude of activation in these regions correlated with the subjective alcohol craving reported by alcohol-dependent patients. In contrast, no such correlation was found in social drinkers. This shows that the relationship between stimuli that elicit memories of alcohol use takes on a different role as alcoholism develops.

Other studies have shown that presentation of alcohol-associated pictures to alcoholics results in an enhanced activation of what is called the ventral visual stream. This is one of the two main brain networks that process visual information and is thought to deal with recognizing objects and representing their form. It has strong connections to the medial temporal lobe, which stores long-term memories, the limbic system, which controls emotions, and the dorsal visual stream, the other main vision network, which deals with object locations and motion. The increased activation of the ventral visual stream by alcohol-associated stimuli in alcoholics probably reflects that these visual cues have taken on an increased salience compared with other stimuli a person might attend to. And one classic regulator of assigning salience to things we see is our old friend the amygdala, which in part does so based on prior experience of those stimuli.

Overall the imaging studies have provided support for an involvement of a network of brain structures that include the amygdala, the insula, the nucleus accumbens, and parts of the frontal lobes in cue-induced cravings. That is in good agreement with available animal studies.⁹ These insights are not only of academic interest. They are also potentially very useful in developing new addiction medications. If a medication can prevent the brain networks described here to become activated after exposure to drug cues, then chances are it will also be able to prevent cue-induced relapse. This approach to early drug development has the attraction that we can go by an objective measure rather than relying on subjective self-reports of patients.

And seeing is believing, isn’t it? Not necessarily. The fact that a brain area lights up in conjunction with a particular behavior or experience does not necessarily mean that this brain area causes that behavior or experience. I cannot repeat enough times: correlation does not equal causation. This is the main weakness that plagues the exploding field of functional brain imaging, and the field of addiction research is no exception. But when brain imaging results converge with the results of animal studies in which we can isolate the causal role of a manipulation, then we can most likely gain important insights about mechanisms that cause rather than just correlate with addictive behaviors. That does seem to be the case with many of the studies on cue-induced craving and the brain networks that produce it.

In the future it may actually become possible to silence brain structures involved in representing drug memories and translating them into cravings and drug seeking. One approach that is of theoretical importance but is not likely to ever become a staple of clinical addiction medicine is deep brain stimulation (DBS). Rapidly adopted for the treatment of symptoms produced by Parkinson’s disease, and more recently for obsessive-compulsive disorder and severe depression, this approach has been reported by a German group to dramatically reduce relapse when targeting one of the structures activated by drug-associated cues, the nucleus accumbens. This was observed in a small number of cases and could clearly be the result of a powerful placebo effect. But controlled studies are now under way and will allow data to be collected that can establish whether DBS actually causes an improvement in addictive disorders. A second approach that is being tried is much less invasive but also much blunter. Neurologists working on movement disorders have for some time known that sending a strong magnetic field into the brain, by placing a coil on the outside of the skull, can temporarily silence or stimulate underlying brain tissue. This method, called transcranial magnetic stimulation (TMS), can as effectively as innocently paralyze the thumb for a few minutes, demonstrating which part of the cerebral cortex produces thumb movements. The limitation until recently has been that the magnetic fields used have not been able to reach deeply enough into the brain to affect structures that are important for drug memories. Recent technological developments have, however, led to the construction of coils that can do just that. Studies on addictive disorders are currently ongoing using this approach, both to understand what brain structures do what and to prevent relapse.

Based on the developing understanding of how drug cues cause relapse, attempts have also been made to develop behavioral treatments. The principle is quite appealing. If the response to the drug cues could be extinguished in the lab, then perhaps the risk of craving and relapse upon encountering these cues in the real world could be decreased? It hasn’t quite worked out that way. Yes, based on this expectation, Peter Monti and his colleagues developed “cue exposure therapy” (CET) for the treatment of alcoholism. Although the number of studies is still small, evidence from these controlled and well-designed trials does in fact support the notion that CET can reduce the frequency and severity of relapse.¹⁰ I will revisit this approach in the chapter that discusses behavioral treatments for alcohol addiction. At this point, however, we should note two things. First, CET involves a considerably more sophisticated approach to relapse prevention than simply extinguishing cue-reactivity. Rather than relying on extinction alone, it involves teaching patients the coping skills needed to manage cravings that arise in response to alcohol cues without relapsing. Second, despite this thoughtful and rational approach, the effect size is modest.

One might wonder what is going on here. Presumably part of the answer is that cue-induced cravings are after all only one class of relapse triggers. Relapse in response to stress, which I will discuss next, is another major category and will not necessarily be affected by treatments that target cue-induced craving. Furthermore, some degree of drug sampling may occur in a mostly random fashion, or perhaps out of habit. This will effectively result in the delivery of priming drug doses, which we have already seen are powerful triggers for craving and relapse themselves.

But according to a hypothesis put forward by Yavin Shaham of NIDA, another important part of the answer may be related to the nature of the drug cues used in these procedures. There are different kinds of cues in the world around us. Some are discrete. In the artificial environment of the lab, the 700 Hz sound from the tone generator, played for a few seconds, is one of them. In the real world, the honking of a horn from a car might fall in the same category. Once one has learned to associate it with the image of a rapidly approaching car, a brief presentation will certainly be enough to make one step back to the curb. This kind of cue is simple in terms of what sensory modality it is picked up by, and it has a distinct onset and an equally distinct offset. But most cues that become associated with drug use in the real world, or for that matter with many other important processes, are far more complex. They provide an entire context in which the actions of the individual occur and with which they become associated—one that is woven from different strands of visual, auditory, tactile, and olfactory fabric. Accordingly, these cues are referred to as contextual.

The brain circuitry involved in relapse triggered by these two types of cues seems to be overlapping, but there are also important differences. As one might expect, the more complex contextual cues rely on a more widespread network of brain structures. In a parallel to what has been found in studies of learned fear, relapse triggered by discrete cues is for the most part related to the function of the amygdala and the nucleus accumbens. In contrast, relapse triggered by the more complex contextual cues in addition relies on another structure classically important for contextual learning and memory: the hippocampus.

It may well be that the more widespread brain networks that drive context-induced relapse inherently make this phenomenon harder to extinguish. But that aside, effectively extinguishing the right contextual cues in the laboratory is going to be difficult simply because the context is very different. The patient is in a hospital, not on the street. Around the patient are people whose looks, sounds, and smells are entirely different from those of people at a bar. I could go on and on, but you get the picture. It is clear that simulating and then extinguishing every possible context that may have become associated with using drugs or alcohol is a much more challenging task than simply extinguishing the response to a cue such as the look or smell of the alcoholic beverage itself. For all we know, the former task may pose a challenge that may simply turn out to be insurmountable. Whether that is going to be the case will obviously be a matter of research, not of opinion.

A fascinating recent development offers some hope that another psychological approach may circumvent the challenges outlined above. Rather than attempting to induce new learning that would extinguish drug responses to drug cues in every conceivable situation, this approach attempts to erase them. This research is based on major discoveries in the field of fear learning, which for many years has provided some of the best models to study how memories are made and unmade. In an extreme oversimplification, memories are first acquired, in an unstable form, into working memory that uses parts of the frontal lobes. They are then consolidated with the help of parts of the amygdala and the hippocampus. Once consolidation has occurred, they are available to be retrieved in response to the right cue or context. But about a decade ago, Karim Nader, now at McGill University, and his mentor at the time, Joseph LeDoux at New York University,¹¹ provided provocative data in support of an additional process, called reconsolidation. The idea is that when a memory trace is retrieved, it transiently becomes unstable, just as it was originally, and will now require reconsolidation to be stored once more. Manipulations during a critical time window following retrieval, while the memory trace is still unstable, will prevent its reconsolidation, essentially making the memory go away.¹² Recently this procedure was applied to relapse behavior in rats and drug craving in heroin addicts, in both cases triggered by drug-associated cues. Both in rats and in people, the appropriate manipulation during the time window of reconsolidation seemed to eliminate the behavioral responses to the respective cues, much the way it has been able to erase learned fear responses.¹³ If replicated, this approach clearly holds the promise to add an important tool to the treatment toolkit and help prevent relapse triggered by drug memories.

Those psychologically inclined may find it appealing to work on designing therapies that will extinguish responses to multiple, varying contexts, hoping that the extinction process will somehow generalize, or erase drug memories altogether. These are reasonable, promising approaches to addressing the challenge of cue-induced relapse. I certainly hope for them to be successful. For the more neuropharmacologically inclined scientists, it might instead be appealing to think that there is a final common pathway through which different types of cues promote relapse. If that final common pathway uses specific transmitters and receptors, then blocking the function of these might hold the same kind of promise. What the outcome will be is a purely empirical question. I don’t care much who “wins,” as long as we get something that works. As will we will see in a coming chapter, there is evidence that a pharmacological approach might hold some promise as well.