ReviewBehavioral functions of the mesolimbic dopaminergic system: An affective neuroethological perspective
Introduction
The ML-DA system (Fig. 1) has received considerable attention due to its involvement in a range of psychological processes and neuropsychiatric diseases. In fact, after the development of a DA theory of schizophrenia (Carlsson, 1974, Carlsson, 1978, Snyder, 1972, Meltzer and Stahl, 1976), additional ML-DA hypotheses have been proposed to explain addiction (Wise and Bozarth, 1981, Wise and Bozarth, 1987, Koob, 1992), attention deficit hyperactivity disorder (ADHD) (Oades, 1987, Levy, 1991, Russell, 2000), and depression (Willner, 1983a, Willner, 1983b, Dailly et al., 2004) as well as global behavioral activation (Gray, 1995) ranging from response persistence to behavioral compulsions (Salamone and Correa, 2002, Everitt and Robbins, 2005).
Localized electrical brain stimulation studies (Olds and Milner, 1954, Heath, 1964, Olds, 1977, Wauquier and Rolls, 1976) have implicated the ML-DA in positive rewarding states (Wise, 1978, Wise, 1981, Wise and Rompre, 1989) as well as in appetitive motivated behaviors (Panksepp, 1971, Panksepp, 1981a, Panksepp, 1982, Panksepp, 1986, Panksepp, 1998, Blackburn et al., 1987, Blackburn et al., 1989, Berridge and Robinson, 1998, Ikemoto and Panksepp, 1999). Since DA is also released in response to aversive stimuli and stress (Abercrombie et al., 1989, Puglisi-Allegra et al., 1991, Rouge-Pont et al., 1993, Pruessner et al., 2004), it appears to promote generalized behavioral arousal under both positive as well as negative emotional conditions, perhaps in terms best conceptualized as the seeking of safety (Ikemoto and Panksepp, 1999). Moreover, the ML-DA system has recently been recognized for its role in the determination of personality traits, including “novelty” or “sensation” seeking (Bardo et al., 1996, Zuckerman, 1990), “extraversion” (Depue and Collins, 1999), and “impulsivity” (Cardinal et al., 2004).
Current interpretations of ML-DA functions diverge with respect to emphasis on unconditioned or behavioral priming effects (motivational theories) versus conditioned effects (learning theories). The “psychomotor activation” hypothesis (Wise and Bozarth, 1987), the “behavioral activation system” hypothesis (Gray, 1995), the “behavioral facilitation” hypothesis (Depue and Collins, 1999), the “SEEKING system hypothesis” (Panksepp, 1981a, Panksepp, 1981b, Panksepp, 1998, Ikemoto and Panksepp, 1999), the “wanting” hypothesis (Berridge and Robinson, 1998), and the “effort-regulation” hypothesis (Salamone and Correa, 2002, Salamone et al., 2003) all acknowledge a motivational interpretation of ML-DA functioning. They share a common perspective based on the classic distinction between appetitive and consummatory phases of motivated behaviors (Sherrington, 1906, Craig, 1918), and with relatively minor differences, consider the DA system as a fundamental drive for the expression of appetitive-approach behaviors.
The “reinforcement” (Fibiger, 1978, White and Milner, 1992, Everitt and Robbins, 2005) and the “reward” hypotheses (Wise, 1978, Wise and Rompre, 1989, Schultz, 1997, Schultz, 1998, Schultz, 2001, Spanagel and Weiss, 1999, Di Chiara, 2002, Wise, 2004), on the other hand, have largely focused on DA as a learning mediator. While motivational theories are interested in the proactive actions of DA transmission on future behaviors, learning theories tend to consider retroactive effects on strengthened associations among past events. Although modern incentive motivation concepts view rewards as promoters of motivational arousal and increased behavioral readiness (Bolles, 1972, Bindra, 1974, Toates, 1986, Berridge and Robinson, 1998), learning theories consider that the “most important role of DA in incentive motivation is historical; it is the stamping-in of stimulus–reward association that has established incentive motivational value for previously neutral stimuli” (Wise, 2004).
Multiple attempts to integrate motivational and learning perspectives of ML-DA transmission have been pursued (e.g., Berridge, 2004, Toates, 2004, Koob, 2004), but a coherent evolutionary-ethological view of how brain DA promotes certain types of unconditional psychobehavioral tendencies is typically missing in most formulations. Therefore, a comprehensive hypothesis integrating new findings with earlier literature on rewarding electric brain stimulation has yet to emerge. In our opinion, such needed integration may be achieved by postulating a role of ML-DA in modifying primary-process emotional behaviors1 and internal affective states (Panksepp, 1998, Panksepp, 2005).2 In fact, emotions and affects have repercussions both on the way animals act in the world and learn through experience. As extensively described in previous works (Panksepp, 1981a, Panksepp, 1981b, Panksepp, 1998, Ikemoto and Panksepp, 1999), ML-DA promotes the emergence of the SEEKING emotional disposition,3 which we envision as an affective urge that characterizes all motivated behaviors. This view has been around as long as the more recent incentive salience and reinforcement-type theories but has been typically ignored by those committed to behaviorist learning paradigms.
In mammals, most DA-containing neurons are clustered within three major mesencephalic groups: A8 cells in the retrorubral field, A9 cells in the substantia nigra (SN) and A10 cells in the ventral tegmental area (VTA) (Dahlstrom and Fuxe, 1964, Ungerstedt, 1971, Lindvall and Bjorklund, 1974, Fallon and Moore, 1978, German et al., 1983, Arsenault et al., 1988, German and Manaye, 1993). Similar organizations of DA cell bodies have been demonstrated in reptiles (Smeets et al., 1987, Smeets, 1988, Gonzalez et al., 1994) and birds4 (Smits et al., 1990, Durstewitz et al., 1999). In addition, less dense aggregations of DA neurons inhabit the supramammillary region of the hypothalamus, the dorsal raphe and the periaqueductal gray (Swanson, 1982, Gaspar et al., 1983). Morphological characteristics, anatomical locations, ascending projections and their associations with arousal functions have led many to assign DA neurons to the classical “reticular formation” (Moruzzi and Magoun, 1949, Schiebel and Scheibel, 1958, Leontovich and Zhukova, 1963). Placed within the context of the reticular activating system (Parvizi and Damasio, 2001), DA neurons are sensitive to various global states of organisms, and their ascending projections modulate brain arousal in accordance with those states (Geisler and Zahm, 2005).
The mesencephalic DA cell groups (A8, A9 and A10) lack clear anatomical boundaries, develop in parallel from common embryonic tissues (Olson and Seiger, 1972, Fallon and Moore, 1978, Hu et al., 2004), and partly overlap in their projection fields (Nauta et al., 1978). Their axons project largely to structures located in the anterior part of the forebrain and modulate the activity of cognitive–executive re-entrant circuits between the cortical mantle and the BG (Alexander et al., 1986, Kalivas et al., 1999) (Fig. 2). Such circuits are involved in the organization of practically all motivated behaviors, both highly flexible and more automatic. It is thought that BG–thalamocortical circuits produce adaptive behavioral flexibility, while their dysregulation underlies a whole plethora of neuropsychiatric diseases, from depression to obsessive-compulsive disorders, from addiction to Parkinson's, etc. (Swerdlow and Koob, 1987, Robbins, 1990, Deutch, 1993, Kropotov and Etlinger, 1999, Jentsch et al., 2000, Graybiel and Rauch, 2000, Joel, 2001, Groenewegen, 2003). Resembling a spiraling, functional organization (Zahm and Brog, 1992), a special type of “state” process, information flow appears to exist between different loops of such circuitries with feed-forward processing from limbic regions (especially medial frontal areas) to executive and motor circuits (Heimer and Van Hoesen, 2006). DA neurons thereby act as an intermediary of limbic-emotional and motivational action outflow (Haber et al., 2000, Joel and Weiner, 2000, Mogenson et al., 1980b).
Although DA cell groups form an anatomical continuum, the ML-DA system has been differentiated from the nigrostriatal (NS) DA system on the basis of anatomical and functional criteria (Bernheimer et al., 1973, Ungerstedt et al., 1974). The ML-DA system (Fig. 1), situated more medially in the brain, is more ancient in brain evolution than the more laterally situated NS-DA circuitry, and it has been more clearly implicated in the regulation of intentional, motivated movements, in flexible-emotive behaviors and in the process of “reward” than the laterally situated NS-DA fields (Papp and Bal, 1987, Wise and Bozarth, 1987, Blackburn et al., 1989, Berridge and Robinson, 1998, Ikemoto and Panksepp, 1999). The NS-DA system, in contrast, controls procedural aspects of movements and motivated behaviors as it reaches more dorsal areas of BG, where behavioral and cognitive habits are learned, stored and expressed (Hornykiewicz, 1979, Carli et al., 1985, Carli et al., 1989, Graybiel, 1997, Jog et al., 1999, Haber, 2003).
DA receptor activated molecular pathways have been partially unraveled (Greengard et al., 1999, Greengard, 2001a), but the precise mechanisms by which DA influences behavioral and psychological phenomena remain unclear. As a modulator of neural activity, DA interacts with fast synaptic transmission (Greengard, 2001b) and thereby influences the way specific external information is processed within the brain (Mesulam, 1998). One hypothesis posits that DA regulatory function increases the signal-to-noise ratio and enhances the efficacy of neural networks in elaborating biologically significant signals (Rolls et al., 1984, DeFrance et al., 1985, Kiyatkin and Rebec, 1996, Nicola et al., 2000). Based on in vivo and in vitro single-cell studies, the signal-to-noise ratio hypothesis explains how behavioral and motivational arousal processes may be linked to specific cognitive or perceptual representations. However, for understanding how behavioral and psychological arousal is processed in the nervous system, large-scale energetic states of the brain, instead of electrical activity of single neurons, need to be considered (Steriade, 1996, Steriade, 2000, Ciompi and Panksepp, 2004, Llinas et al., 2005, Freeman, 2005). DA modulates global-field dynamics, desynchronizes cortical-derived oscillatory rhythms and promotes high-frequency waves along the gamma band within BG–thalamocortical circuits (Brown and Marsden, 1998, Brown, 2003, Magill et al., 2004, Lee et al., 2004). In our view, these rhythms may be accompanied by the release of neurodynamic instinctual sequences, which are essential infrastructures for intentional behaviors.5 Neurodynamic sequences are repetitive sequential activity patterns reverberating across specific areas and circuits of the brain. Recently, they have been called “avalanches” (Beggs and Plenz, 2003, Beggs and Plenz, 2004), and their influence on brain activity may be described with the concept of “dynamic attractors” (Freeman, 2000, Freeman, 2001, Freeman, 2003).
The sequential patterns favored by DA in ventral BG–thalamocortical circuits may relate to an instinctual drive to seek life-supportive aspects of the environment and to actively escape those aspects that could be destructive. These neurodynamic sequences are evolutionarily intrinsic, but epigenetically refined, procedural patterns associated with the expressions of exploring and approach behaviors (i.e., locomotion, sniffing, head movements, saccades). The reverberation of such sequential patterns within brain circuits changes the individual's attitude towards the environment, promoting the SEEKING disposition to dominate the motivational landscape of the organism (Panksepp, 1998). This establishes a variety of expectancy states that energize and coordinate the anticipation of life-supporting events with characteristic reward seeking behavioral tendencies (Panksepp, 1981a, Panksepp, 1981b, Panksepp, 1986). In this way, primary-process “intentions in action” get transformed into learning and thought-related “intentions to act” (Panksepp, 2003).
Our interpretation of the behavioral functions of the ML-DA system is based on a theoretical perspective we have called the affective neuroethological view. Such a perspective has characteristic features that diverge from current dominant theoretical models and that focus on a series of currently neglected elements.
(1) Energy. Modern brain research often fails to account for the energetic and dynamic aspects of neural, behavioral and mental activities. We should ask why animals perceive the world as they do and are spontaneously active in globally energetic ways. How can cognitive computations arise in the brain without the support of global dynamic states that channel an organism's needs via large-scale brain network functions? Where do such global states arise, and how do they interact with informational processes?
New neurodynamic approaches that grant organisms intrinsic behavioral urges are needed to make sense of why organisms do what they do (Panksepp, 1998, Kandel, 1999, Freeman, 2000, Freeman, 2003, Solms and Turnbull, 2002, Ciompi and Panksepp, 2004). It is time to introduce such concepts into the discussion of brain DA functions since mesencephalic DA and ascending reticular activating system (ARAS) are fundamental energetic sources for many types of neural activity6 (Moruzzi and Magoun, 1949, Lindsley et al., 1949, Lindsley et al., 1950, Jones, 2003). In particular, behavioral activating properties of DA may depend on its capacity to influence global field dynamics in the forebrain, as reflected in DA facilitation of the emergence of fast-wave oscillatory rhythms in BG and cortical areas (Brown and Marsden, 1998, Levy et al., 2000, Tseng et al., 2001, Brown, 2003, Magill et al., 2004, Sharott et al., 2005).
(2) Internal procedural sequences. Behavior is not limited to learning and associative processes; neuro-behavioral instinctual processes, shaped by evolution, are essential for almost all aspects of goal-directed learning. Neurocognitive behaviorism denies (or at least ignores) an organism's intrinsic behavioral identity and thus neglects certain inborn adaptive capacities as fundamental determinants of learning (Lorenz, 1965). In addition to neural plasticity and top–down hierarchical brain processes, we must harness ethological traditions in order to better understand intrinsic capacities of organisms and thereby emphasize the importance to evolutionary constraints on learning (Tinbergen, 1951, Lorenz, 1965, Burkhardt, 2005). In vertebrates, such constraints emerge substantially from the influences that subcortical brain structures exert over neocortical functions (MacLean, 1990, Panksepp, 1998).
In particular, basal forebrain and BG are involved in the expression of sequential, species-specific movements, such as instinctive and unlearned sequential grooming movements in rodents (Cromwell and Berridge, 1996), which are the Fixed Action Patterns (FAPs) of ethologists7 (Lorenz, 1950, Tinbergen, 1951, MacLean, 1990). Moreover, the BG influence learning, especially when different sequences of actions are linked into a single functional unit (Knowlton et al., 1996, Graybiel, 1998, Jog et al., 1999, Packard and Knowlton, 2002, Bayley et al., 2005). Basal forebrain areas, including BG, extended amygdala, septum and nucleus of Meynert (Heimer and Van Hoesen, 2006), represent the deep, subcortical parts of the cerebral hemispheres (Swanson, 2000), and they are essential foundations for higher information processing regions of neocortex to operate. Housing abundant GABA inhibitory neurons, they form reciprocal networks and send inhibitory outputs to thalamic, hypothalamic and midbrain nuclei (Kitai, 1981, Berardelli et al., 1998, Kropotov and Etlinger, 1999). Situated between the cortex, the diencephalon and the brainstem, the basal forebrain is viewed as largely inhibitory with tonical suppression of behavioral actions (Swanson, 2000). Nevertheless, when something perturbs its intrinsic equilibrium, particular sequences of activity are released. Therefore, basal forebrain nuclei have been considered “doors that, when unlocked, may release into action large functions outside them” (Llinas, 2002).
(3) Emotions. Dorsal BG areas control habitual behaviors, whereas other basal forebrain nuclei (ventral BG, extended amygdala, and septum) are involved in emotional behaviors (Koob, 1999, Swanson, 2000, Alheid, 2003, Heimer and Van Hoesen, 2006). Emotions comprise sequences of FAPs that characterize their expressive and communicative aspects (Darwin, 1872, MacLean, 1990, Llinas, 2002), but one main characteristic of emotion is to regulate the organism's behavioral repertoire in flexible ways. Behavioral plasticity arises when each emotional operating system orchestrates a wide range of potential responses in accordance with environmental conditions (Panksepp, 1998). When an emotion is activated, the organism's attention is focused largely on a particular set of stimuli, memories and responses. For example, an animal does not eat while experiencing intense fear; food is transiently excluded from its interests. Diffusion of basal forebrain/BG characteristic patterns communicates an emotional disposition within the brain. Such patterns represent the basic action tendencies characteristic of various primary-process emotions, whose neural representations influence the activity of many different brain regions and help match perceptual and cognitive representations into a global action tendency. In such a way, basal forebrain changes intentional states and orients behavior in specific directions.
From this perspective, it is inadequate to try to explain motivations, intentions and emotions simply from top–down cognitive or representational perspectives. Intentions-in-action, as intrinsic impulses to act, may best be viewed as neural dynamic sequences, which, once activated, constitute internal procedural drives8 (Llinas, 2002). In our model, such neurodynamic sequences emerge from within basal forebrain and BG areas (Knowlton et al., 1996, Graybiel, 1998), and associated medial diencephalic and mesencephalic circuits, with parallel roles in learning and expression of motor habits and emotions (MacLean, 1990, Graybiel, 1997, Jog et al., 1999).
(4) Affective feelings. Neuronal activity is not limited to the production of computational representations of the world; it also helps organize a large variety of states, among which the emotions and associated affects have been ignored for perhaps too long (Panksepp, 1998, Panksepp, 2005). Removing affectivity from neuroscience may lead to a profound misunderstanding of intrinsic brain organization and functioning and hinder scientific understanding of how brains truly operate. A recently re-introduced James-Lange type view of emotions considers affective feeling to be produced by “somatic marker” representations of body changes (Damasio, 1996, Damasio et al., 2000). However, the nature of feelings should also incorporate the intrinsic intentionality of many instinctual behaviors; emotions are not only a consequence of “what happened” (Damasio, 1999), but also “what is happening”, “what is going to happen” and “what may happen”. Such processes are not uniquely human characteristics; an affective core underlying subjectivity appears to have emerged early in vertebrate brain evolution (Panksepp, 1981a, Panksepp, 1981b, Panksepp, 1998, Panksepp, 2005), derived from brain systems that regulate the inner states of the organisms (MacLean, 1990, Damasio, 1999, Craig, 2003, Thompson and Swanson, 2003, Schulkin et al., 2003, Berntson et al., 2003, Porges, 2003, Sewards and Sewards, 2003, Alheid, 2003, Denton, 2006). The core affective substrate of every emotional feeling seems to be generated and, in part, informs hierarchically related neural networks that include, most prominently, the periaqueductal gray, the hypothalamus and the extended amygdala (Panksepp, 1998). Indeed, accumulating evidence for some kind of primary-process psychological experiences arising from such primitive subcortical circuits is becoming substantial (Panksepp, 2005, Merker, 2007). In our view, the core affective states are communicated to higher brain levels through the emergence of specific neurodynamic sequences, so that the cognitive–evaluative aspects of emotion can be elaborated in a coordinated fashion by various forebrain areas, especially orbitofrontal and medial frontal regions.
Section snippets
Electrical self-stimulation of the brain (ESSB)
The discovery of ESSB by Olds and Milner (1954) represented a major breakthrough in understanding the neurobiological bases of reward. Electrical stimulation of various brain sites in association with specific behaviors increased the probability that animals would repeat those behaviors. These studies led to the recognition of reward areas in the brain (Olds et al., 1971, Wise, 1996, Wise, 2005, Chau et al., 2004) with the medial forebrain bundle (MFB) being a primary neural pathway
Theoretical interpretations
Complex relationships among neural, behavioral and psychological levels guarantee the presence of substantial gaps in our understanding that remain to be filled. The adoption of novel integrative hypotheses may be essential for promoting empirical predictions that can help fill the remaining gaps.
New inroads of the affective neuroethological perspective
In the previous section, we described how the behavioral functions of ML-DA emerge from its ability to activate the SEEKING emotional disposition. It is now important to provide new hypotheses describing how this disposition is processed in the brain. Obviously, this proposal needs an elucidation of the role of DA in modulating neural activity across brain circuitries. Indeed, correlative neurophysiological observations obtained from recording DA neurons (which tell us much about what DA cells
Current theories
Drug abuse has been defined as a chronically relapsing disorder, in which the addict experiences uncontrollable compulsion to take drugs, while the repertoire of behaviors not related to drug seeking, taking and recovery declines dramatically (White, 2002). The development of addiction is attributed to the action of drugs in the brain (Leshner, 1997). Chronic drug use causes permanent neural changes at many levels of analysis, from molecular and cellular levels to neural circuits (Hyman and
Conclusion
The analysis of ML-DA functions has become an enormous field of inquiry, and new findings and theoretical interpretations are emerging at a steady pace. As this paper was completed, a whole issue of the journal “Psychopharmacology” (2007, vol. 191, issue 3) appeared that was dedicated to the topic. There is no need to modify our position with respect to the cornucopia of these additional perspectives, which are mostly elaborations of previous positions. We would simply highlight that the view
Acknowledgments
Studies reported in this paper were supported by a grant to R.H. and J.P. (NIH/NIDA 1R21DA016435-01A1) and by the help of Hope for Depression Research Foundation to J.P. We would also like to thank the Department of Biological Sciences and the J.P. Scott Center for Neuroscience (Bowling Green State University) for their support.
References (666)
- et al.
The functional anatomy of basal ganglia disorders
Trends Neurosci.
(1989) - et al.
Instrumental learning, but not performance, requires dopamine D1-receptor activation in the amygdala
Neuroscience
(2005) - et al.
Neurobiology of executive functions: catecholamine influences on prefrontal cortical functions
Biol. Psychiatry
(2005) - et al.
GABA in the ventral tegmental area: differential regional effects on locomotion, aggression and food intake after microinjection of GABA agonists and antagonists
Life Sci.
(1979) - et al.
Brain substrates for increased drug seeking during protracted withdrawal
Neuropharmacology
(2004) - et al.
Dopamine modulates acute responses to cocaine, nicotine and ethanol in Drosophila
Curr. Biol.
(2000) - et al.
Psychobiology of novelty seeking and drug seeking behavior
Behav. Brain Res.
(1996) - et al.
Differential responsiveness of dopamine transmission to food-stimuli in nucleus accumbens shell/core compartments
Neuroscience
(1999) The fruit fly: a model organism to study the genetics of alcohol abuse and addiction?
Cell
(1998)- et al.
Addiction, dopamine, and the molecular mechanisms of memory
Neuron
(2000)