This article reports an interesting set of experiments showing that ravens repeatedly choose among a set of different objects, an object that will prove several minutes or hours later to be useful to obtain food. The authors interpret this preference as evidence that ravens would be able to plan for the future. However, since the preferred object was repeatedly and selectively paired with food reward during initial training, one cannot definitively rule out the involvement of non-planning processes. For instance, ravens could prefer the would-be useful object, not because they anticipate its future utility, as hypothesized by the authors, but merely because they attach more affective and/or motivational value to the object due to its past selective association with food reward. To rule out this associative mechanism, it is important to add a control condition where, all else being equal, choice of the previously food-paired object is not followed by an opportunity to use it to obtain food (for a similar criticism and a specific example of a control condition, see: Redshaw et al. https://www.ncbi.nlm.nih.gov/pubmed/28927634).
This is a must-read and highly accessible Review Article on preclinical choice models of drug addiction. The authors first remind us that: 1) drug addiction can be best described as a behavioral disorder that arises from a harmful misallocation of behavior toward drug use at the expense of other competing nondrug activities; and 2) one of the major, perhaps primary, goals for addiction treatment is not only to stop or reduce drug use but also to promote a reallocation of behavior toward nondrug activities, the latter being essential for maintaining drug abstinence. Then, the authors explain that by providing alternative, competing nondrug activities during drug access, preclinical choice models are uniquely suited to capture and recapitulate the behavioral core of addiction. Finally and most importantly, they also provide novel, compelling evidence for the translational utility of these models that may surpass other more classical, non-choice models of addiction. Notably, they have begun to establish the existence of an unprecedented strong concordance between the clinical effectiveness of a drug treatment and its effectiveness in preclinical choice models of addiction. I have no doubt that this Review will finish to convince preclinical researchers who are still hesitant that choice models of addiction are the best choice for future research in the field!
There are many different types of data in this ambitious Research Article. I will only comment on the behavioral effects of medial prefrontal (mPFC) activation. Though those manipulations induced large-scale reorganization of both cortical and subcortical brain activities, as measured by fMRI, they had only marginal behavioral repercussions.
Notably, contrary to what the authors seem to imply, there is little or no evidence for anhedonia during mPFC activation (see Figure 4 of the paper). At best, there is a small reduction in preference for a low concentration of sucrose (i.e., from 90 to about 85%) but rats still continued to largely prefer sucrose over water. The behavioral effects of mPFC activation are simply magnified by setting the origin of the Y-axis at 70% - a value well above indifference. The behavioral significance of this tiny decrease in preference, if any, is largely unclear. In addition, there is no direct evidence that it results from a prefrontal top-down control over dopamine-dependent reward signaling in the striatum, as suggested by the authors. Finally, there is no evidence in Figure 5 that rats seek to spend more time in a place associated with stimulation of midbrain DA neurons. Rats are initially globally indifferent. The small, albeit significant, effect of mPFC activation likely reflects a direct aversive effect that might also be observed alone with no concurrent stimulation of midbrain DA neurons – an important control experiment that the authors apparently failed to conduct. It would have been more relevant to test whether and to what extent mPFC activation directly modulates the self-stimulation behavior of midbrain DA neurons reported in Figure 2.
Overall, the behavioral effects reported in this Research Article are too marginal and too disparate to offer a clear picture of the role of mPFC activation in regulating dopamine-dependent reward-seeking behavior.
This review provides a nuanced and intelligent summary of the research on drug habits from the UK “Cambridge School” of Neuropsychology. It reminds us several obvious, albeit too often overlooked, facts about drug habits. First, drug habits, like other habits, develop in stationary and thus highly predictable environments; much less so in changing environments. Second, drug habits per se are adaptive behaviors and their formation is generally not pathological. Third, in theory, drug habits may become pathological only when they escape to the control of the drug user (i.e., self-control), be it a human or a non-human animal. Loss of self-control entails that the drug user attempts to stop or reduce drug use, generally to avoid or minimize the negative consequences, but with no or little success. Finally and most importantly, this review helps us to better realize that there is still no direct evidence that experimental animals have lost control over their drug seeking and/or taking habits. There is only evidence in some individual animals that drug habits can partly escape to the control of the experimenter. For instance, an individual rat that continues to seek cocaine despite punishment can be considered to have escaped to the control of the experimenter (perhaps because the punishment is too weak) but not necessarily to its own self-control. In my view, this important distinction between these different sources of control (self versus others) should be better appreciated by researchers who attempt to model addiction in animals.
This comment is a follow-up of my previous comment about the difficulty in interpreting this study that contradicts most previous similar studies. After a careful analysis of this paper and after collecting all elements of methods “ectopically scattered” through the text, I think I finally arrived at a satisfactory explanation for why most rats preferred cocaine over sweet water in the present study. Briefly, everything was made to make access to sweet water reinforcement less direct and more difficult than access to cocaine reinforcement, thereby biasing choice towards cocaine!
More specifically, rats had to go through an unusually long chain of behavioral events before getting access to sweet water. A similar chain was not required for cocaine delivery. First, once rats turned the wheel on the operant panel, they had to cross the cage to reach a magazine on the opposite panel inside which there was a retractable drinking spout that delivered sweet water. This arrangement introduces a spatial and thus a time gap between responding and sweet water reinforcement. Both gaps are known to reduce conditioning. Second, once rats have reached the magazine, they did not have directly access to the drinking spout that delivered sweet water. They had first to insert their head into the magazine to make the retractable drinking spout appears. This behavior amounts to a second operant response which thus defines with the first response (i.e., wheel turning) an operant chain. In addition, once rats inserted and maintained their head in the magazine, the drinking spout was not continuously available but came “back and forth in the magazine during 50s.” This is a rather unusual method of fluid delivery (note: the frequency and duration of these back-and-forth movements are not indicated in the Methods).
Thus, to repeat, everything was made in the present paper to make access to sweet water reinforcement more difficult and less direct than access to cocaine reinforcement, thereby biasing choice towards cocaine. This unusual approach may be appropriate for addressing some scientific questions but it is misguided and inappropriate for studying the vulnerability to cocaine addiction which was the main goal of the present paper. If one wants to pursue such a goal, one better tries to make access to cocaine reinforcement equal to or more difficult than access to the nondrug option and not the other way around! Indeed, if one sufficiently weakens the nondrug option, then one will eventually reach a point where most individual rats, even the non-addicted ones, will prefer the drug! To take an extreme example, if one provides rats with ready access to cocaine but ask them to play piano or climb Mt Everest to get access to sweet water, they will surely choose cocaine over sweet water. This is not surprising, this is just trivial! In contrast, if rats take cocaine despite and at the expense of an equally or a more accessible potent nondrug option, then one has got something much less trivial and probably more relevant for studying the vulnerability to cocaine addiction.
This study is interesting and also quite embarrassing. It is interesting because of the important questions that it asks. It is embarrassing because it shows that when given a choice, most rats prefer cocaine over sweet water – a finding that is strictly the opposite of what we and others have found over the past few years. Of course, contradiction and refutation are the “game of science”. We should not be embarrassed by them and instead welcome them.
My embarrassment comes from the fact that these opposite outcomes were obtained by a former master student of mine – Nathalie Vanhille who is the first author of this study – using a choice protocol initially developed in our lab. When Nathalie was working in our lab using this protocol, she observed that most rats preferred sweet water over cocaine – the opposite of what she now reports in this study despite the use of an identical choice protocol.
But were the choice protocols really identical? Of course not! Like always, the devil lurks into the details and details can sometimes matter a lot! Apparently, this study differs from our previous choice studies in the way rats were given access to sweet water. In our study, access to sweet water was pretty straightforward. Rats had to press a lever to fill a nearby receptacle with sweet water. Then they could obtain additional volumes of sweet water during 20s by continually licking the receptacle. In this study, however, access to sweet water was really contrived for reasons that remain unclear until one reaches the middle of the Discussion. In fact, despite my best efforts and those of other members of the team, we were unable to get a clear final picture of how rats get access to sweet water in this study.
So here is my challenge for the interested readers and researchers: I would really appreciate if someone could help me figure out how exactly rats get access to sweet water in this study.
This is an interesting series of experiments on choice between nicotine and sucrose in rats. In experiments 1, 3, 4 & 5, hungry rats were first trained to respond for sucrose or nicotine on alternate days before being provided with a choice between the two options. Overall, virtually all rats responded more for sucrose than for nicotine under a variety of choice conditions. Thus, all else being (approximately) equal, sucrose surpasses nicotine reward in rats!
In experiment 2, rats were first trained to respond for nicotine before being provided with a choice between nicotine and sucrose. In this condition, about 50% of the rats responded more for nicotine than for sucrose, suggesting that “nicotine self-administration does not only occur in the absence of alternative reinforcement options”, at least in some rats.
Though the results of experiment 2 are promising, they are difficult to interpret univocally. Experiment 2 lacks an important control group that controls for the difficulty in learning to respond for sucrose during choice testing. Briefly, rats were asked to respond on a novel lever under a random ratio 4 schedule of sucrose reinforcement, WITHOUT ANY PRIOR PROGRESSIVE TRAINING on this lever and after a long period of operant extinction. It is likely that many rats will fail to learn to respond for sucrose under these specific conditions, even in the absence of the opportunity to self-administer nicotine! Future research should resolve this important issue.
In this new Shattuck Lecture, Eric Kandel and Denise Kandel (K2) announce that they will let science speaks for itself about the role of “nicotine as a gateway drug” to other drug addictions and will not distort “what science does and does not tell us” about this controversial matter, as many others would generally do.
However, this paper is more about cross-sensitization between nicotine and other drugs in mice than about “nicotine as a gateway drug” to other drug addictions in human populations. Basically, K2 present evidence showing that forced pre-exposure to nicotine can enhance the proportion of individual mice that become sensitive to the stimulant and rewarding effects of a low dose of cocaine. In fact, after nicotine pre-exposure, a large majority of mice become sensitized to cocaine (i.e., up to 78-98% of mice). This proportion contrasts with the small minority of cigarette smokers who go on to develop cocaine addiction in human populations (i.e., 20%). This large discrepancy complicates the extrapolation between mice and humans but is ignored by K2. Would it imply that nicotine-exposed mice are more vulnerable to cocaine addiction than cigarette smokers? Or, more likely, would it indicate that nicotine-induced sensitization to cocaine per se should not be confused with cocaine addiction?
Research on cross-sensitization between different drugs, including nicotine and cocaine, in rats and mice were very popular among addiction researchers in the 80s and 90s. However, because drug sensitization is clearly not drug addiction, these studies have since been replaced by other, more valid animal models of drug addiction. This research is entirely ignored by K2. Thus, it remains to be demonstrated whether and to what extent initial nicotine self-administration in mice or rats can increase the proportion of individuals that develop more relevant behavioral features of cocaine addiction. Thus, we are back to the initial “scientific question of the role of nicotine as a gateway drug” to other drug addictions.
It is perhaps the privilege of a Nobel laureate (Eric Kandel is one of the recipients of the 2000 Nobel Prize in Physiology or Medicine) to write on any subject matter by ignoring the relevant work of most other prize-free scientists. Ironically, however, a laureate is awarded this prize precisely because many other scientists have recognized the importance and relevance of his/her own previous work.
Reading this Shattuck lecture, one is reminded of the novel “Pierre Menard, author of The Quixote” by Jorge Luis Borges. This novel illustrates how knowing the author of a text can dramatically impact our reading and interpretation of it. So I propose the following test for the interested reader. Please try to read this article as if it had not been co-authored by Eric Kandel. I bet that you will be surprised by how little this article contributes to the scientific question that it seeks to address.
Thanks for your comments on my previous comments which, contrary to what you claim, were not based on “incorrect assumptions”. I leave the readers to judge for themselves. Here I just want to persist and sign for the sake of clarification. There is no solid evidence for compulsive cocaine use in your study. Mice gave up easily on cocaine and this apparently, I now learn from your comments, regardless of strain differences. I agree that your study was not designed to directly compare the motivation for food with that for cocaine. But still it shows that mice are able to produce hundreds of responses to obtain certain rewards of importance to them. The fact that mice are able to but do not expand a comparable level of effort to obtain cocaine shows that they are not that motivated for cocaine. More specifically, it shows that even after several weeks of cocaine intake, the motivation for cocaine fails to acquire a degree of intensity that approaches that for food in hungry mice. This is perhaps one of the reasons why mice were trained on a FR1 for cocaine in your study and not on a FR10 like for food. Finally, your operational measure of “perseverative responding” is difficult to interpret because the introduction of the drug-off periods also considerably influences cocaine self-administration during the drug-on periods. Judging from Figure 1, cocaine self-administration rapidly becomes very irregular, with short and long inter-injection intervals. As a result, some of the inter-injection intervals during the drug-on periods are much longer than the drug-off periods themselves, making behavior during the latter periods quite difficult to interpret!
This study is very interesting but it also has several potential problems that may limit its relevance to understand the neural basis of compulsive cocaine use. First, there is no solid evidence for compulsive cocaine use in this study. For instance, mice were clearly able to inhibit cocaine seeking at the onset of the signaled drug-off periods. What they seemed unable to do was to wait until the end of the drug-off periods before resuming responding. This behavior may merely reflect an anticipation of cocaine availability at the end of the drug-off periods and not a compulsion (see: http://www.ncbi.nlm.nih.gov/pubmed/22985696). In addition, PR responding for cocaine was very low in this study and, in fact, 10-30 times lower that PR responding for food (i.e., about 50 versus 1000 responses per PR sessions for cocaine and food, respectively). In other words, mice gave up quickly on cocaine, without spending much effort. Calling this behavior “compulsive” or even “compulsive-like” is excessive. Finally, the facilitatory effects of CNO-induced inhibition of D2-MSNs on PR responding for cocaine should be interpreted with caution. These effects were seen in a small group of mice (n = 6) that had an atypical low baseline (i.e., saline) level of PR responding compared to the other groups from this study (i.e., 18.7±6.1 versus 35±9.5, 41.5±18.2 or 59.2±24.1). In fact, PR responding following CNO-induced inhibition of D2-MSNs did not differ from baseline responding in these other groups.
The authors propose to reclassify addictive drugs, including cocaine, according to their common property of directly evoking a high frequency of de novo dopamine transients that would surpass that normally evoked by “natural” rewards. This property is particularly obvious in awake and behaving rats. This property would be necessary and sufficient to explain why “a higher reward magnitude [would be] conferred to abused drugs compared to natural rewards” and why “drug-associated cues [would acquire] the powerful ability to reinstate drug seeking and taking”.
Imagine that a rat has come to expect a limited daily access time to a desired good and that it has no way to save it for future consumption. Intuitively, you might think that the shorter the expected access time to the good, the faster the rat will consume it. In other words, the rate of consumption should be inversely related to the daily access time to the good (at least within a certain range of time values).
In this elegant study, the authors report the opposite outcome in rats self-administering cocaine. Rats did not self-administer cocaine at a lower rate but instead at a higher rate when they expected a long (6h) versus short (1h) daily access time to the drug. Importantly, when rats had no way to predict whether access time to the drug will be short or long, they self-administered cocaine as if they were expecting a short access time!
How to make sense of these rather unexpected findings? To address this question, we will probably have to address first the following other questions: Is the positive relationship between rate of consumption and access time reported in this study TYPICALLY observed with other goods or is it specific to cocaine self-administration? What does this relationship tell us about how and what do rats exactly learn in this sort of differential access time setting? Does it imply that rats have a longer future time horizon than previously thought (i.e., hours instead of minutes)? I hope that future research will shed some light on these different questions.
This clever study shows that responding for cocaine-associated cues can be entirely and persistently abolished by canceling out cocaine-induced long-term plasticity at hippocampal and cortical synaptic inputs to D1R-expressing neurons in the nucleus accumbens. Importantly, the latter feat was achieved by acutely co-opting endogenous mechanisms of synaptic plasticity using astute optogenetic-based LTD protocols.
The efficacy of these LTD protocols was specific to responding for cocaine-associated cues and had no impact on responding for sucrose or sucrose-associated cues - which were both notably more intense and more robust than responding for cocaine-associated cues.
One may regret that the authors did not test whether and to what extent the same LTD protocol that acutely abolished responding for cocaine-associated cues would also affect the maintenance of responding for cocaine itself and/or the ability of other well-known factors to trigger “relapse.” These additional experiments would have been helpful to better define what has exactly been erased by the LTD protocols. It would also be important to know whether the same efficacy could be achieved after more prolonged exposure to cocaine self-administration.
Finally, this study paradoxically shows that endogenous mechanisms of synaptic plasticity are apparently largely intact in cocaine-exposed animals and can be co-opted to reverse, persistently and rather easily, cocaine-induced synaptic plasticity. One may then wonder whether and to what extent similar mechanisms could be self-recruited by people who successfully quit cocaine.
This study provides important novel insights into the neurobiological mechanisms that drive escalation of cocaine intake in rats. It shows that escalation of i.v. cocaine self-administration is selectively associated with a loss of cue-induced phasic dopamine in the nucleus accumbens. Reversing this neurochemical deficit with l-Dopa (30 mg/kg) was sufficient to reverse escalated levels of cocaine intake, suggesting that cue-induced phasic dopamine plays a causal role in regulating cocaine intake.
This conclusion is rather unexpected because cocaine cues are generally invoked as key triggers of drug seeking but not as key regulators of drug taking. Perhaps one way to test this hypothesis would be to measure the effects of cue omission (or reduced cue intensity) on maintenance of cocaine self-administration. This intervention should prevent (or reduce) cue-induced phasic dopamine and thus induce an increase in cocaine intake, at least initially.
Finally, to better understand the action of l-Dopa on escalation of cocaine intake, it will be important to check whether and to what extent it also affects tonic dopamine levels during a session of cocaine self-administration.
This study is outstanding. Together with other studies, it confirms that sugar is a double reward. Sugar activates brain dopamine (DA) neurons twice: once during sweet tasting (via an initial polysynaptic brainstem pathway) and once after blood glucose absorption (via an activation of MCH neurons). The latter delayed activation of DA neurons plays a critical role in learned food preferences and may also contribute to maintain the ability of sweet taste to activate DA neurons.
Rather surprisingly, this study also shows that concurrent optogenetic activation of hypothalamic MCH neurons during access to water sweetened with sucralose (a non-caloric sweetener) increases consumption of sweet water and also DA levels in the striatum. The time course of this effect is somewhat paradoxical, however, if one supposes that it mimics activation of MCH neurons by post-absorptive glucose. In fact, this effect is more consistent with a direct modulation of sweet palatability by MCH neurons. It would be interesting to know how glucose activation of MCH neurons during access to sweet water would affect consumption.
This remarkable study suggests, perhaps for the first time, that the orbitofrontal cortex (OFC) imposes “its” values to other value-coding brain regions (e.g., dorsal striatum) to guide choice and preference. Under normal circumstances, the values of hierarchically lower brain regions are in line with those of the OFC and no conflict arises. However, when these values diverge, the OFC would take the lead and impose “its” values to guide behavior, even if they are less accurate or up-to-date than those of hierarchically lower brain regions. A key challenge for future research will be to better understand how and under what circumstances the OFC acquires a set of values that diverges from reality and from that acquired by other brain regions.
Excellent review. The authors present the VTA as a unique brain hub that integrates a variety of humoral and neuronal signals to regulate the "drive" to eat palatable foods for pleasure, comfort and relief. They appreciate the difficulty of targeting this intricate brain region to selectively reduce the excessive pursuit of food reward without altering other life pursuits.