The Dual-State Theory of Prefrontal Cortex ... - Semantic Scholar

0 downloads 0 Views 2MB Size Report
Seamans JK, Durstewitz D, Christie BR, Stevens CF, Sejnowski TJ (2001):. Dopamine ... 9151. 63. Trantham-Davidson H, Neely LC, Lavin A, Seamans JK (2004): Mecha- ... Granon S, Passetti F, Thomas KL, Dalley JW, Everitt BJ, Robbins TW.
REVIEW

The Dual-State Theory of Prefrontal Cortex Dopamine Function with Relevance to Catechol-OMethyltransferase Genotypes and Schizophrenia Daniel Durstewitz and Jeremy K. Seamans There is now general consensus that at least some of the cognitive deficits in schizophrenia are related to dysfunctions in the prefrontal cortex (PFC) dopamine (DA) system. At the cellular and synaptic level, the effects of DA in PFC via D1- and D2-class receptors are highly complex, often apparently opposing, and hence difficult to understand with regard to their functional implications. Biophysically realistic computational models have provided valuable insights into how the effects of DA on PFC neurons and synaptic currents as measured in vitro link up to the neural network and cognitive levels. They suggest the existence of two discrete dynamical regimes, a D1-dominated state characterized by a high energy barrier among different network patterns that favors robust online maintenance of information and a D2-dominated state characterized by a low energy barrier that is beneficial for flexible and fast switching among representational states. These predictions are consistent with a variety of electrophysiological, neuroimaging, and behavioral results in humans and nonhuman species. Moreover, these biophysically based models predict that imbalanced D1:D2 receptor activation causing extremely low or extremely high energy barriers among activity states could lead to the emergence of cognitive, positive, and negative symptoms observed in schizophrenia. Thus, combined experimental and computational approaches hold the promise of allowing a detailed mechanistic understanding of how DA alters information processing in normal and pathological conditions, thereby potentially providing new routes for the development of pharmacological treatments for schizophrenia. Key Words: Attractor dynamics, computational model, dopamine, GABA currents, NMDA currents, prefrontal cortex, schizophrenia

P

erhaps the most debilitating problems for medicated patients with schizophrenia are their enduring cognitive deficits. Cognitive deficits are present from onset of the illness, are often independent of psychotic symptoms, are stable over time, and predict functional outcome (1–3). The cognitive deficits associated with schizophrenia may even be considered endophenotypic characteristics that mark the presence of a genetic predisposition (4). By definition, an endophenotypic marker found in schizophrenic patients should also be found in nonaffected family members at a higher rate than in the general population. Accordingly, the most robust findings among nonaffected relatives are impairments in the domains of verbal and working memory, cognitive flexibility, and executive function/inhibition (2,5–7). Therefore, deficits in these domains may be considered core features of the disease, perhaps even providing the grounds on which other classes of symptoms develop (8,9). There is now general consensus that the neural substrate for at least some of the cognitive deficits in schizophrenia, such as working memory, involve the prefrontal cortex (PFC) and the mesocortical dopamine (DA) system (10,11). Working memory refers to the ability to maintain trial-unique, goal-relevant information, to process it to derive predictions, and to integrate it with current sensory input to guide behavioral decisions. It is important to note that working memory is not the same as short-term memory (STM) and can be dissociated from it both on anatomical

From the Centre for Theoretical and Computational Neuroscience (DD), University of Plymouth, Plymouth, United Kingdom; and Department of Psychiatry & The Brain Research Centre (JKS), University of British Columbia, Vancouver, British Columbia, Canada. Address reprint requests to Daniel Durstewitz, Ph.D., Centre for Theoretical and Computational Neuroscience, Faculty of Science, University of Plymouth, Portland Square Building, A 220, Plymouth, PL4 8AA, United Kingdom; E-mail: [email protected]. Received February 21, 2008; revised April 18, 2008; accepted May 20, 2008.

0006-3223/08/$34.00 doi:10.1016/j.biopsych.2008.05.015

and behavioral grounds (12–19). Rather, working memory involves the active handling and monitoring of goal-related information to derive predictions and guide decisions. The cellular basis of working memory has been studied in animals mainly using variations of the delayed response paradigm, so far most extensively in the primate dorsolateral PFC. Neurons within this region increase their activity during the cue, delay, and response phases of delayed response tasks in an often stimulus-selective manner (20 –23). The selectively enhanced firing rates during the delay periods of these tasks in the absence of external cues have been interpreted as a neural basis for the active maintenance of memory items. Although such stimulus- or response-specific elevations in mean firing rates during the delay are found in many brain areas (24 –26), delay activity in the primate PFC is closely correlated with behavioral performance (22,27,28) and persists in the presence of distracters (29). Computational models have successfully captured the essence of this type of activity (reviewed in 30).

The Dual-State Theory of PFC DA Function Dopamine modulates both working memory performance and task-dependent neuronal firing rates within the PFC in a complex manner (31–37). Dopamine exerts its impact on working memory and neural activity through a multitude of effects on presynaptic release, N-methyl-D-aspartate (NMDA) and gammaaminobutyric acid type A (GABAA) currents, the persistent sodium (Na⫹) current, various calcium (Ca2⫹) currents, the slowly inactivating potassium (K⫹) current, and the H current (Figure 1; 38 – 48), mediated by D1-class (D1 and D5 [D1]) and D2-class (D2, D3, D4 [D2]) receptors. To understand the functional implications of this intricate pattern of DA-induced cellular and synaptic changes for neural processing and PFC-dependent cognition, biophysically realistic computational models have proven useful (Figure 2; 49 –53). Such models consist of sets of differential equations for each neuron that describe the evolution of the somatic and dendritic membrane potentials according to various voltage-gated, Ca2⫹-gated, and synaptically gated ionic currents (e.g., 54). These, in turn, are often modeled by HodgkinBIOL PSYCHIATRY 2008;64:739 –749 © 2008 Society of Biological Psychiatry

740 BIOL PSYCHIATRY 2008;64:739 –749

D. Durstewitz and J.K. Seamans

Figure 1. Dopaminergic modulation of PFC cellular and synaptic properties. (A) DA has a plethora of effects via D1- and D2-class receptors on voltage-gated and synaptic ionic conductances of PFC pyramidal cells and interneurons. Reprinted with kind permission from Springer Science ⫹ Business Media: Psychopharmacology (Berl), The ability of the mesocortical dopamine system to operate in distinct temporal modes, 2007;191:609 – 625, Lapish et al., Figure 1 (98). (B) D1 and D2 receptor stimulation act antagonistically on GABAA synaptic currents. Application of a D1 agonist subsequently to a D2 agonist reverses the D2 effects, and vice versa. Reprinted with permission from Seamans et al. (44). DA, dopamine; GABAA, gamma-aminobutyric acid type A; PFC, prefrontal cortex.

Huxley-like gating kinetics as first formulated by Hodgkin and Huxley (55) in their Nobel prize-winning work on action potential generation. Hence, this approach translates neuronal struc-

tures into equivalent electrical circuits that mimic current flow across active and passive membrane channels and between neuronal compartments as illustrated in Figure 2.

Figure 2. Biophysical modeling in a nutshell: in this computational approach to single neuron and network function, the morphologies of real neurons are first translated into a structure of connected compartments (20 in this example), each of them in turn being represented by an equivalent electrical circuit that captures all the passive and active (ligand-, voltage-, or ion-gated) currents flowing across the cell membrane. Each of these membrane currents is generated by a static (passive) or adjustable (active) conductance (the zig-zag lines) in series with an ionic battery driving that current. All currents are in parallel to the membrane capacitance (Cm), and patches of membrane are connected through the intracellular (cytoplasmatic) resistance (Ri). The operation of these circuits is described by a set of nonlinear differential equations for the membrane voltages in all somatodendritic compartments and the gating variables mimicking the voltage-dependent transitions between different states of the underlying ion channel. The voltage is regulated by the sum of all passive, active, and synaptic currents, which in turn are given by Ohm’s law and the product of different gating (activation and inactivation) variables. The whole system of differential equations describing a network of such neurons is implemented on a computer and then integrated numerically yielding solutions of all system variables as functions of time (see Figures 3 and 4).

www.sobp.org/journal

D. Durstewitz and J.K. Seamans The appeal of these models is their close relation to biophysical quantities measured electrophysiologically. They allow effects of DA measured in vitro to be implemented rather directly with no or only few additional assumptions. For instance, the ⬃40% change of NMDA conductances revealed in vitro translates into a ⬃40% change of the parameter regulating the maximum NMDA conductance in the model (50,53). Unlike the in vivo situation, however, these models make it possible to dissect the network behavior in detail, manipulating or monitoring every single parameter/variable of the system, and applying experimental protocols that are not feasible in vivo. They therefore allow one to relate the modulation of single ion channels to the impact of these modulations on large-scale network dynamics. This makes this approach a very powerful tool for understanding the biophysics and neuromechanics of cognitive phenomena. As illustrated in Figure 3, these systems can reproduce the low and high activity states classically associated with spontaneous (baseline) activity and stimulus-specific delay activity in working memory tasks (22,56). This is achieved by embedding cell assemblies, i.e., groups of functionally related neurons that share a strong excitatory connectivity among each other, into the network (Figure 3A). Once the firing activity within a cell assembly is driven across a certain threshold by an external stimulus, the assembly can maintain activity autonomously due to this strong recurrent excitation that is mainly supported by slowly decaying NMDA currents (Figure 3B; an idea first made explicit by Wang [57] and experimentally supported, for instance, by Seamans et al. [58]). Formally, the spontaneous rest state and the stimulus-specific high activity states correspond to different attractor states of the system as illustrated in Figure 3C and explained in the corresponding legend. Starting from such a configuration that mimics certain functional characteristics of a working memory network, it can now be investigated in detail how the D1 or D2 receptor-mediated changes reported in vitro affect the network dynamics. The results of such simulations revealed that the combined effects of D1-induced conductance modulations (Figure 1) led to a change in network dynamics that made it more difficult to switch between various high activity (active memory) states, i.e., to an increase in the energy barrier between different discrete states of network activity (Figures 3C and 4B; 49 –53,59). These effects are partly rooted in the differential contribution of various D1modulated currents to different activity regimes: while the D1induced increase in NMDA and other voltage-dependent currents (38,39,43) fosters the currently active memory state by boosting recurrent excitation within cell assemblies, the concomitant increase in GABAA currents (44) leads to fiercer competition among different active ensembles of neurons, thereby limiting the set of items encoded in working memory. At the same time, this D1-mediated enhancement of GABAA currents, as well as the reduction in glutamate release probability (41,43), make it harder to evoke activity in cell assemblies in the first place (50). At the in vivo electrophysiological level, these dynamical changes would predict an increased signal-to-noise (S/N) ratio (60) in the sense of an increased differentiation of firing rates associated with currently active memory states and those associated with nonactivated states or spontaneous activity (as illustrated in Figures 3B and 3C). A related and complementary idea, namely an increased S/N ratio via a DA-induced change in the gain of the single neuron input/output function, had been proposed in more abstract terms within connectionist-like models more than 15 years ago by Servan-Schreiber et al. (61) and has received experimental support recently (59).

BIOL PSYCHIATRY 2008;64:739 –749 741 In vitro D2 agonists tend to act opposite from D1 receptors on NMDA and GABAA currents, as well as on pyramidal cell excitability in PFC neurons (Figure 1B; 38,39,43,62,63). Such opposing effects of D1 versus D2 stimulation have also been observed for various molecular markers of intracellular cascades, like cyclic adenosine monophosphate (cAMP) production and phosphorylation of dopamine and cyclic AMP-regulated phosphoprotein with molecular weight 32 kDa (DARPP-32) (64,65). As a result, simulated D2 activation reduces the barrier among activity states in the model networks (Figures 4A and 4B), i.e., the valleys of the energy landscape become so flat and nearby that noise may easily push the system from one state into the other. Hence, this would cause spontaneous pop-out of activity states caused by noisy fluctuations, highly unstable representations, and fast and spontaneous transitions between many different activity states as illustrated in Figure 4C (44,50,66,67). Thus, D1 versus D2 receptor activation has opposing dynamical and functional implications for neural network behavior. These two different D1- and D2-induced regimes of network activity will be termed D1-state and D2-state, respectively, in the following. For clarity, it should be added that we are purely referring to electrophysiological effects of D1/D2 stimulation induced on the postsynaptic site here, since D2 autoreceptors that regulate DA release are much less abundant or even missing from dopaminergic axons projecting into the prefrontal cortex (68). Each of the two model states has its associated computational advantages and disadvantages. The increased differentiation among attractor states in the D1-state has been shown to strongly boost the robustness of items in working memory, protecting them from distracting stimuli and noise as it becomes harder to switch the system among different activity states (49 –53,59; see also 69). On the other hand, one major disadvantage of this state could be that the flexible manipulation of information is reduced due to these hampered transitions among states. A consequence of this may be a decrease in divergent modes of thinking. In the extreme, active representations may become so robust that they are immune to new information and updating, leading not only to perseverative thoughts and actions but also to a decreased alignment of the PFC with the external world. The D2-state, due to the decreased energy barrier among activity states, would allow easier access to PFC networks (see also 69) and faster switching among different PFC activity patterns such that the network may quickly cycle through multiple representations that could become active nearly simultaneously (Figure 4C). Although the implications of these dynamical changes for specific cognitive tasks have not yet been explicitly tested in biophysical simulations, it is conceivable that they enhance set shifting and flexible problem solving as they facilitate moving among representations and increase the variety of recalled memory patterns. It may also speed up decisionmaking processes at the potential cost of accuracy. A disadvantage of the D2-state is that in its extreme, representations become so fleeting and there is no filtering of information flow (i.e., blocking of distracting inputs) into the PFC that working memory processing cannot be maintained, in turn, causing thinking to become disorganized and tangential. In summary, we propose that PFC network dynamics exist on a continuum where the general activity regime at any time is determined by prevailing D1 versus D2 receptor activation, i.e., the D1/D2 receptor activation ratio (Figure 5A). When the PFC approaches the extremes of the continuum, pathology could emerge. These pathological states may be at the core of schizowww.sobp.org/journal

742 BIOL PSYCHIATRY 2008;64:739 –749

D. Durstewitz and J.K. Seamans

Figure 3. Stimulus-selective persistent activity, presumably underlying working memory (see text), and D1 modulation in a PFC network model. (A) Structure of a network model underlying simulations as those shown in (B) and (C) and in Figure 4. Pyramidal, cells consist of a somatic and a dendritic compartment and are recurrently coupled via both AMPA and NMDA excitatory synapses. They also excite a population of interneurons that feeds back inhibition mediated by GABAA synaptic conductances into all pyramidal cells. Stimulus-specific persistent (delay) activity could be produced by cell assemblies (highlighted by the yellow square) consisting of subpopulations of pyramidal neurons with stronger than average mutual synaptic connections. Once a threshold of activation is crossed, these strong recurrent excitatory inputs could keep high firing levels within a stimulated cell assembly going and thus maintain stimulus-specific enhanced firing rates as observed experimentally. (B) An external stimulus switches a cell assembly from a low spontaneous state (labeled 1) into a stimulus-specific high persistent activity state (labeled 2) that is maintained even after withdrawal of the initiating stimulus (until terminated by some other event), thus encoding an online memory of the stimulus in agreement with experimental observations (22,55). D1 stimulation differentially suppresses low (spontaneous) and enhances high (stimulus-selective memory) activity in these simulations (right-hand side). (C) The dynamical basis of these phenomena revealed by a numerically derived state space representation of the model dynamics: the graph shows a two-dimensional plane spanned by the average firing rate of the pyramidal cells (x axis) and the average firing rate of the interneurons (y axis) within a cell assembly. Arrows indicate the flow, i.e., give the direction of change of average firing rates as a function of the current state of the network (the length of all arrows was normalized to 1). Following these arrows, one sees that firing rates converge to either one of two points (labeled 1 and 2), corresponding to the low spontaneous and high persistent firing rates illustrated in (B). These points of convergence are called attractor states of the system dynamics, and they are more formally given by the intersection of two lines (called nullclines), one giving the steady-state firing rates of the pyramidal cells as a function of a fixed average rate of the interneurons (dark blue curve) and the other, vice versa, showing the steady-state firing rate of the interneurons as a function of a fixed pyramidal cell rate (red curve). Hence, where these lines intersect both pyramidal neurons and interneurons are in their steady states yielding a fixed point. The two regions of convergence for the low and high firing rate attractors are separated by the green dashed line and are called their basins of attraction. This separating line between the basins of attraction can be seen as a threshold: to elicit memory activity, a stimulus has to drive the network from its low spontaneous state across this border between the basins such that it converges toward one of the high activity states. Within this representation, D1 stimulation leads to a stretching of the pyramidal cell nullcline (light blue curve) along the x and y axes, which underlies the increase in energy barriers among activity states illustrated more explicitly in Figure 4B. AMPA, alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; GABAA, gamma-aminobutyric acid type A; NMDA, N-methyl-D-aspartate; PFC, prefrontal cortex. Reprinted from Neural Net 15, Durstewitz D and Seamans JK, The computational role of dopamine D1 receptors in working memory, copyright 2002, with permission from Elsevier (53).

phrenia from which other symptom classes emerge as discussed below. Empirical Support for the Dual-State Theory: Animal Studies As mentioned above, one indication of the D1-induced changes in network dynamics is an increased differentiation of firing rates associated with target- or memory-related activity as www.sobp.org/journal

compared with nontarget, background, and spontaneous activity. This D1-mediated differentiation reflects the underlying dynamical changes that cause the increased robustness of working memory representations. Indeed, these effects in the model replicate both early electrophysiological observations suggesting a (relative to baseline) stronger amplification of delay- and response-related single unit activity by DA during working

D. Durstewitz and J.K. Seamans

BIOL PSYCHIATRY 2008;64:739 –749 743

Figure 4. D1-state and D2-state dynamics in the biophysical model network. (A) State space representation of the network dynamics (see Figure 3C for explanation) in D1-dominated, balanced, and D2-dominated regimes (only the nullclines are shown). While D1 stimulation leads to a stretching of the pyramidal cell nullcline along the x and y axes, D2 stimulation leads to a contraction along both dimensions. (B) Representation of the information in (A) (with corresponding line colors) in terms of an energy landscape (note that this graph is just a schema). Minima of the energy correspond to the fixed-point attractors in (A), and the state of the system may be envisioned as a ball rolling down into the nearest minimum. The local slopes in this graph depend on the sign and magnitude of the derivatives of the underlying system as indicated by the flow field in Figure 3C. The graph makes clear that it becomes much harder to switch between different attractor states in the D1-dominated regime as the troughs move apart and the valleys become much steeper. Conversely, in the D2-dominated regime, the valleys become so flat and nearby that noise may easily push the system from one state into the other. Also note that the ease of “attractor hopping” could in principle be regulated purely by the steepness of the valley slopes, without any change in the position of the minima, hence without any change in average firing rates. For simplicity, the energy landscape is shown just as a 2-D graph. For a 2-D state space as in (A) and Figure 3C, the full energy landscape would be a surface in a 3-D space (the axes of the state space plus the energy axis) that may be obtained, for instance, by integrating along the flow field. (C) Network simulation illustrating the fact that the system spontaneously switches or cycles among different attractors (neural representations) in the D2-dominated regime, while robustly maintaining a once elicited attractor in the D1-dominated regime. Modified with kind permission of Springer Science⫹Business Media, Figure 15.3 (66). 2-D, two-dimensional; 3-D, three-dimensional.

memory (32,70), as well as very recent findings suggesting that D1 agonists diminish nontarget-related activity to a much larger degree than target-related activity (37,71). In both cases, the outcome is an increase in S/N ratio, as predicted by the computational models (cf. Figure 3). Furthermore, stimulation of the origin of the DA pathway in the ventral tegmental area increased current pulse-evoked high-rate firing while decreasing spontaneous low firing of PFC neurons recorded intracellularly in vivo (72), again consistent with an increase in S/N ratio. Therefore, in accordance with simulations, DA in vivo appears to enhance the S/N ratio in PFC either through an increase in signal, a decrease in noise, or both, or—in terms of the dynamical models— through an increase in the energy barrier between different activity states. These changes in network dynamics underlie the increased stability of working memory representations in the network simulations, and in fact numerous studies in different species have convincingly demonstrated that local application of D1 agents or DA blockade in the PFC is highly detrimental to working memory performance (31–36). While D1 receptors are critical for the ability to preserve and use mental representations over delay periods (10,71), evidence for a functional role for PFC D2 receptors has been less clear. However, Wang et al. (73) showed in primate PFC that saccade (response)-related firing at the end of working memory trials was

selectively responsive to D2 agents. In other words, D2 receptor stimulation had an influence on neural activity at times when the current working memory contents had to be released and the system had to be prepared for new input but not during the maintenance phase of the task, consistent with a preferential contribution of D2 receptors to the flexible integration of new information proposed by our dual-state model. An increasing number of behavioral studies also support a specific role for D2 receptors in modulating response flexibility (for review, see 74). In monkeys, the D2 receptor antagonist raclopride impaired the performance on a reversal task probing response flexibility (75). In the rat, the role of PFC D2 receptors in set shifting was demonstrated clearly by Floresco et al. (76). Using a four-arm plus-maze, rats were trained to respond using an egocentric strategy of always going to the arm that was in the same spatial location relative to their frame of reference (i.e., always go right or left). The rule then changed and to perform optimally rats now had to visit arms according to a visual cue, regardless of its egocentric location (or vice versa). This was termed a set shift. Floresco et al. (76) found that a D2 antagonist but not a D1 antagonist disrupted performance and caused perseverative errors on the task, consistent with a shift to a D1-state dynamic. In contrast to this D2 modulation of set shifting, a D1 antagonist but not a D2 antagonist attenuated performance on an www.sobp.org/journal

744 BIOL PSYCHIATRY 2008;64:739 –749

D. Durstewitz and J.K. Seamans

Figure 5. Hypothesized [DA]-dependency of PFC dynamical regimes in normal subjects and schizophrenic patients. (A) Normal relationship between DA concentration, the D1/D2 activation ratio, and the resulting dynamical states and associated attractor landscapes. Arrows indicate where the COMT genotype val/val and met/met carriers would be located on this curve at baseline. The curve was generated by assuming that the activation of D1 receptors and D2 receptors in their high and low affinity states, respectively, can be described by three Boltzmann functions of dopamine concentration (A ⫻ [1 ⫹ exp(⫺B ([DA] ⫺ [DAhalf]))] ⫺1). Based on empirical data (63,91), the three functions for D2high, D1, and D2low receptors would have different half-activation values ([DAhalf]: D2high ⬍ D1 ⬍ D2low). The curve shows the ratio D1/(D2high ⫹ D2low) based on these three functions, which is assumed to determine the dynamical regime in which the PFC resides. (B) Deformation of this curve and associated changes in attractor landscapes under the assumption that some proportion of D2 receptors move from their low into their high affinity states, giving rise to positive symptoms. (C) Deformation of this curve and associated changes in attractor landscapes under the assumption that there is an increase in D1 receptors, giving rise to negative symptoms. DA, dopamine; met, methionine; PFC, prefrontal cortex; val, valine.

eight-arm radial maze working memory task (34). In this case, the rats did not exhibit perseverative errors (as would be expected for an extreme D1-state) but responded randomly, consistent with a shift into a D2-state dynamic. D1 agonists have also been observed to cause disruption in working memory tasks, but unlike D1 antagonists, they can cause perseverative errors (35,77). These data agree with the idea that excessive D1 stimulation increases the energy barrier among PFC activity states so much (cf. Figures 3C and 4B) that the currently active memory contents that guide the animal’s responses become immune to new inputs. Thus, the previously successful representation would continue to guide responding in the face of failure (i.e., perseveration). On the other hand, a D1 agonist or a D2 or D4 antagonist can improve performance on working memory and attentional tasks when delivered to a DA system that is working suboptimally (77– 80). In summary, there is converging evidence from in vivo electrophysiological and behavioral pharmacological studies of a double dissociation between D1 and D2 receptor effects in accordance with our model predictions: while a D1-state favors stable working memory performance and increases cortical S/N, a D2-state is associated with higher response flexibility. Empirical Support for the Dual-State Theory: Human Studies Most human studies working with pharmacological manipulation of DA receptors have focused on the effects of D2 receptors or sometimes combined D1/D2 receptor agonists, due to a lack of specific D1 receptor agents suitable for use in humans. Studies that employed both specific D2 agonists and combined D1/D2 agonists generally support a specific role for D1 but not D2 receptors in enhancing working memory (81). The D2 antagonists, in contrast, have been shown to deteriorate www.sobp.org/journal

performance on tests of attentional set shifting and response flexibility (82). Moreover, in the same group of healthy volunteers, Mehta et al. (83) showed that the D2 antagonist sulpiride not only impaired set shifting but also actually protected against the deleterious effects of intervening task-irrelevant distractors on a spatial working memory task, as would be expected by a shift of PFC networks into a D1-dominated regime caused by D2 receptor blockade. Conversely, activation of D2 receptors with the D2 agonist bromocriptine improved performance on the Wisconsin Card Sorting Test (WCST) in patients with traumatic brain injuries (84), and D2 receptor density in the medial PFC correlates with performance on the WCST, with higher densities being associated with more errors and a lack of cognitive efficiency (85). Behavioral studies in humans have furthermore supported the partly antagonistic nature of working memory and cognitive flexibility requirements (86). However, there are also empirical results that fit less well with the dual-state theory. For instance, in contrast to the predictions, D2 antagonists were shown to impair and D2 agonists to improve spatial working memory in some studies (87– 89), while having just the opposite effect on measures of behavioral flexibility in this work. In interpreting these data it is important to note, however, that pharmacological agents in human experiments are applied systemically, thus affecting all brain structures including the basal ganglia and D2 autoreceptors (see Supplement 1 for additional discussion).

Differential Regulation of D1-State and D2-State In Vivo Since the normal brain obviously does not use specific D1 or D2 agonists and antagonists to establish a certain PFC dynamical regime, the question arises how this differential regulation could

D. Durstewitz and J.K. Seamans be achieved in vivo. We proposed that the two receptor classes may be differentially regulated by the prevailing DA concentration. In vitro, D1- and D2-class receptors appear to be differentially regulated within the concentration range observed in vivo (Supplement 1). The D1 receptors enhanced NMDA and GABAA currents at DA concentrations that were lower (⬍500 nmol/L) than those (⬎1 ␮mol/L) at which D2 receptors act to reduce these currents (39,63,90). On the other hand, at very low DA concentrations, D2 receptor-dependent modulations have also been observed. For instance, West and Grace (91) in striatum and Lavin et al. (72) in PFC showed that a D2 antagonist increased evoked firing of neurons recorded intracellularly in vivo. This implies that there was a D2 receptor-mediated effect that occurred at the low nmol/L levels of DA that are present in these brain areas of anesthetized animals. These electrophysiological findings therefore suggest that the dependence of DA-modulated dynamical regimes on DA concentration takes the form of an inverted U-shape (Figure 5A): at very low or very high DA concentrations, a D2-state would prevail, while at medium DA concentrations the network dynamics would be dominated by D1 receptor-mediated effects. This idea of an inverted U-shape relation for DA was used previously by Meyer-Lindenberg and Weinberger (92) to explain behavioral and functional magnetic resonance imaging (fMRI) effects in humans with different catechol-O-methyltransferase (COMT) genotypes and pharmacological manipulations (see next section). Originally, it was proposed by Williams and Goldman-Rakic and colleagues (37,93), although with regard to D1 receptor effects alone (see Supplement 1 for a discussion of how this integrates with the present proposal). While electrophysiological studies support an inverted Ushape relation, the underlying mechanisms producing this relation are less clear. One potential mechanism could involve the preferential stimulation of D2 receptors in their high (D2high) or low (D2low) affinity states at the two ends of the inverted U-curve as illustrated in Figure 5A (94 –97; the support for this assumption is detailed in Supplement 1). Another factor contributing to preferential stimulation of D1 versus D2 receptors could be anatomical (90,98): D2 receptors appear to be located within or in close proximity to the synapse, while D1 receptors appear to be exclusively extrasynaptic (99,100). This arrangement may favor a preferential activation of D2 receptors at very high initial DA concentrations near the release sites, which then degrade and dilute before reaching the more distant D1 receptor sites. Thus, D1 and D2 receptors would be activated at all DA levels, but there would be greater relative influence of one receptor over the other at different DA concentrations. We stress, however, that while the dual-state theory needs some physiological mechanism to differentially regulate the activation of D1-class versus D2class receptors in line with cognitive requirements, the precise nature of this relationship (inverted-U, linear, etc.) or the precise mechanisms of regulation (DA concentration, spatial factors like extrasynaptic location, etc.) do not affect the basic idea of two different dynamical regimes (D1-state and D2-state) and their relation to computational and cognitive properties. COMT: Endogenous PFC DA Regulation and Its Effect on Cognition From the discussion above, inferring which of the two theorized dynamical regimes (D1-state vs. D2-state) PFC networks might occupy depends on knowledge of the regulation of endogenous DA levels. In PFC, this process is strongly regulated by the enzyme COMT, and as a result of the unique biochemical-

BIOL PSYCHIATRY 2008;64:739 –749 745 anatomical profile of the PFC and its DA innervation, this regulation by COMT is quite specific to the PFC, and in addition it is relatively specific for DA in the PFC (101–103; see Supplement 1 for further details). A functional polymorphism for COMT involves a methionine (met) to valine (val) substitution at codon 158 (104,105). The met allele has a quarter the enzyme activity as the val allele (104), and therefore met/met individuals should have higher DA levels specifically in PFC. Hence, within the context of our theory, we would argue that at baseline, val/val carriers would be normally in a low [DA]/D2high dominated regime (D2-state), while met/met carriers would be closer to the moderate [DA]/D1 receptor-dominated regime (D1-state), moving into the high [DA]/D2low receptor-dominated regime when challenged (Figure 5A). Accordingly, human COMT val/val genotype individuals often exhibit poorer performance on tests of working memory that involve the PFC compared with met/met individuals (106). Yet, in accordance with our dual-state theory, on tests of cognitive flexibility the pattern reverses. Methionine carriers are less and val carriers are more flexible during reversals on a competing programs task (107). Methionine carriers were also more inflexible in processing emotional stimuli (108). Perhaps even more striking is that these patterns of behavior can be changed by manipulating DA levels pharmacologically: val/ val subjects showed improved performance on tests of executive function when given a COMT inhibitor (tolcapone) or amphetamine, consistent with moving them from the D2high-state into the D1-state, whereas the met/met individuals got less efficient, consistent with moving them from the D1-state into the D2lowstate (3-back task in 109,110). On the WCST, however, a task that involves switching among rule sets as one of its components, the val allele is associated with a reduction in performance compared with the met allele (106,111). At first glance, this appears to be in disaccord with the present theory, but it is important to note that the WCST, in fact, involves a variety of cognitive processes, and poor performance on the task has also been related to an inability to maintain rule sets across delays (i.e., working memory) and an inability to generate a plan (112–114). Given the multidimensional nature of the task, it is therefore conceivable that poor performance on the WCST could emerge as a result of a dysregulation of either the D1-state or the D2-state. Besides alterations in task performance, COMT-dependent differences in PFC activation as assessed by fMRI are even more robust. On the n-back working memory task in humans, met/met individuals at baseline (D1-state) or val/val individuals on amphetamine or tolcapone (moving them from a D2high-state into a D1-state, see above) exhibit a relative decrease in the area of PFC that is recruited to solve the task (11,109,110,115), consistent with the narrowing down on a few spatially localized and robust representations in the D1-state (Figure 4). In contrast, met/met individuals receiving amphetamine or tolcapone (moving them from a D1-state to a D2low-state) or val/val individuals at baseline (innately closer to a D2high-state) exhibit an increased area of PFC activity during the working memory task, making the PFC less efficient (11,109,110), consistent with the spatial broadening of activity associated with the frequent hopping between many different widely distributed representations in the D2-state (Figure 4). As noted above, met individuals with intrinsically high DA levels exhibited an activity pattern that was like that of val carriers at baseline when given amphetamine or tolcapone to increase PFC DA levels even further. This counter-intuitive result is consistent with the model formulation above and suggests that more DA over an already high basal level actually looks the same www.sobp.org/journal

746 BIOL PSYCHIATRY 2008;64:739 –749 as a low DA state and means that the [DA]-PFC activity relationship may indeed be an inverted U as very low and very high DA levels produce similar cortical D2-dominated dynamics and degraded working memory performance (Figure 5A; 92). In addition to this evidence from behavioral and imaging studies, electroencephalogram (EEG) recordings in humans with identified COMT genotypes also suggest a modulation of the signal/noise ratio in line with the dual-state theory (11,115–119). Noise in these studies was defined as the variability in the phase relationship of evoked electroencephalic activity (event-related potentials [ERP]) with respect to a stimulus. Winterer et al. (116) suggested that some noise may actually be beneficial in promoting flexibility by overcoming “local energy minima,” as it indeed would in the D2-state in the biophysical simulations (cf. Figure 4; 120). These authors showed that on attentional tasks frontal noise in the EEG was lower in met/met (putative D1-state) individuals (121), while val/val (putative D2high-state) individuals had greater frontal noise (118). Prefrontal noise in these individuals was also negatively correlated with performance on the n-back working memory task (122). Along with the differences in noise, there is also a difference in the “signal” between groups, as homozygous met/met individuals had a peak blood oxygenation level-dependent (BOLD) response that was stronger than the one found in val/val carriers (115). Thus, consistent with the dualstate theory, the studies on individuals with different COMT genotypes support the idea that the cortical S/N ratio and working memory performance are increased in subjects closer to a D1 regime (met/met), while tasks that require a high degree of flexibility benefit from a relatively more D2-controlled dynamics (val/val).

Implications of the Dual-State Model for Schizophrenia This section tries to work out some of the potential implications of the proposed D1- and D2-dominated states for understanding symptoms in schizophrenia and is more speculative in nature than the previous sections. The idea that the PFC and mesocortical DA system are critically involved in the cognitive symptoms of schizophrenia owes much to the pioneering work of Weinberger (123) and Goldman-Rakic (124). Our models may help to expand on these ideas by placing their theories into a biophysical computational framework. As described above, by regulating the balance of D1- versus D2-mediated cellular and synaptic effects, DA concentrations would adapt the cortical dynamics to different cognitive requirements, favoring either robust working memory or cognitive flexibility (Figures 4 and 5A). In schizophrenia, however, this adaptive process may be deeply disturbed. Remarkably, virtually every manipulation that has been linked to psychosis, including administration of amphetamine or phencyclidine or neonatal hippocampal lesions, all shift D2 receptors from D2low to D2high (97). Within the context of our model, such a shift would essentially mask the D1-state and make D2 effects predominant at all DA levels, resulting in a strongly D2-controlled dynamical regime. As depicted in Figure 5B, such a regime would be characterized by extremely shallow basins of attraction such that even moderate DA levels would result in an inability to hold and manipulate information, i.e., poor working memory, and therefore potentially disorganized and tangential thinking, thought derailment, drifting, and loose or overinclusive associations. In this sense, within our framework, the D2mediated flattening of attractor landscapes and the relative absence of a D1-dominated state would create a situation that www.sobp.org/journal

D. Durstewitz and J.K. Seamans disrupts cognition and in turn favors the emergence of positive symptoms of schizophrenia (see also 67). Accordingly, agents that block D2 receptors (as antipsychotic medications do) would tend to limit the impact of D2 receptor activation at moderate DA levels and help to restore the normal operating regime of the D1 receptor. However, it has been argued that a hypo-DA state may also exist in the PFC of patients with schizophrenia that produces an upregulation of D1 receptors (123,125). This would cause the D1-state to occupy a much larger portion of the curve such that elevations in DA levels that would normally produce a balanced D1/D2 state would now produce a very strong D1-state (Figure 5C). Recall that the very high energy barrier among activity states associated with an extreme D1-state would not only interfere with set shifting but also with normal working memory and goal-directed behavior, as it would lead to perseverative thoughts and an effective decoupling of the PFC from other cortical and environmental inputs. One may speculate that this decoupling may prevent emotional and motivational processes from driving the goal-directing machinery of the PFC and thereby lead to a lack of interest in the external world and social surroundings, with the PFC becoming overly preoccupied by internal processing of recurring thoughts. A decreased production of emotion (flattened affect) and decreased goal-directed behavior (avolition), as characteristic of the negative symptoms of schizophrenia, may therefore be the result. Indeed, increased D1 receptor binding in schizophrenia is correlated with poorer working memory performance (125), and poorer working memory is, in turn, highly associated with global negative symptomatology (126). Thus, within our framework, the extreme D1mediated deepening and widening of basins of attraction may create a situation that favors the emergence of negative symptoms. In this sense, changes that arise as a result of genetic variations in the neural machinery that normally mediates working memory, attention, and executive function form the core symptoms of schizophrenia from which other symptom classes could emerge (8,9,11,92).

Conclusions In vitro electrophysiological measurements have revealed a perplexing variety of D1 and D2 receptor-mediated effects on synaptic and voltage-gated ion channels (reviewed in 90). In an attempt to unravel the functional and neurodynamical implications of these complex and sometimes apparently opposing effects, we and others have constructed a series of biophysically realistic neurocomputational models that could reproduce basic characteristics of PFC network activity and task performance (49 –53,59,66,127). Simulation results led to the proposal of two distinct regimes, termed the D1-state and the D2-state, which are associated with different dynamical and computational properties. A D1-state is characterized by a high energy barrier among different network states with the consequence that active working memory representations become highly robust to distraction and noise but the associated disadvantage of less flexible switching among activity states. In contrast, a D2-state, leading to a reduced energy barrier and shallow basins of attraction (Figure 4B), would favor cognitive flexibility and fast switching among representations. As reviewed here, the existence of these two regimes with their associated electrophysiological and cognitive properties is now supported by a wealth of experimental findings in rodents, nonhuman primates, and humans. Hence, there

D. Durstewitz and J.K. Seamans is a remarkable convergence of computational results derived from biophysical models based on in vitro electrophysiological data, in vivo single-unit recordings, and behavioral paradigms in animals, with imaging, EEG, and behavioral data in humans. Computational tools as outlined here may not only offer explanations for the modulation of PFC-dependent cognitive abilities on a biophysical and neurodynamical basis, they may also help to better understand in mechanistic terms the derailment of cognitive and emotional functions in nervous system diseases like schizophrenia and may even offer a new route for developing efficient pharmacological treatments.

Authors are in alphabetical order. This work was funded by a Grant from the Canadian Institutes of Health Research (MOP-84319). The authors report no biomedical financial interests or potential conflicts of interest. Supplementary material cited in this article is available online. 1. Green MF, Kern RS, Heaton RK (2004): Longitudinal studies of cognition and functional outcome in schizophrenia: Implications for MATRICS. Schizophr Res 72:41–51. 2. Heydebrand G (2006): Cognitive deficits in the families of patients with schizophrenia. Curr Opin Psychiatry 19:277–281. 3. Lewis R (2004): Should cognitive deficit be a diagnostic criterion for schizophrenia? J Psychiatry Neurosci 29:102–113. 4. Gottesman II, Gould TD (2003): The endophenotype concept in psychiatry: Etymology and strategic intentions. Am J Psychiatry 160:636 – 645. 5. Snitz BE, Macdonald AW 3rd, Carter CS (2006): Cognitive deficits in unaffected first-degree relatives of schizophrenia patients: A metaanalytic review of putative endophenotypes. Schizophr Bull 32:179 – 194. 6. Kéri S, Janka Z (2004): Critical evaluation of cognitive dysfunctions as endophenotypes of schizophrenia. Acta Psychiatr Scand 110:83–91. 7. Sitskoorn MM, Aleman A, Ebisch SJ, Appels MC, Kahn RS (2004): Cognitive deficits in relatives of patients with schizophrenia: A meta-analysis. Schizophr Res 71:285–295. 8. Beck AT, Rector NA (2005): Cognitive approaches to schizophrenia: Theory and therapy. Annu Rev Clin Psychol 1:577– 606. 9. Barch DM (2005): The cognitive neuroscience of schizophrenia. Annu Rev Clin Psychol 1:321–353. 10. Goldman-Rakic PS, Muly EC 3rd, Williams GV (2000): D(1) receptors in prefrontal cells and circuits. Brain Res Brain Res Rev 31:295–301. 11. Winterer G, Weinberger DR (2004): Genes, dopamine and cortical signal-to-noise ratio in schizophrenia. Trends Neurosci 27:683– 690. 12. Hebb DO (1939): Intelligence in man after large removals of cerebral tissue: Report of four left frontal lobe cases. J Gen Psychol 21:73– 87. 13. Hebb D (1977): Wilder Penfield: His legacy to neurology. The frontal lobe. Can Med Assoc J 116:1373–1374. 14. Stuss DT, Benson DF (1986): The Frontal Lobes. New York: Raven Press. 15. Passingham RE (1975): Delayed matching after selective prefrontal lesions in monkeys (Macac mulatta). Brain Res 92:89 –102. 16. Petrides M (1989): Frontal lobes and memory. In: Boller F, Graffman J, editors. Handbook of Neuropsychology, vol. 3. Amsterdam: Elsevier, 75–90. 17. Petrides M (1995): Functional organization of the human frontal cortex for mnemonic processing: Evidence from neuroimaging studies. Ann N Y Acad Sci 769:85–96. 18. Petrides M (2000): Impairments in working memory after frontal cortical excisions. Adv Neurol 84:111–118. 19. Pribram KH, Tubbs WE (1967): Short-term memory, parsing, and the primate frontal cortex. Science 156:1765–1767. 20. Kubota K, Niki H (1971): Prefrontal cortical unit activity and delayed alternation performance in monkeys. J Neurophysiol 34:337–347. 21. Fuster JM, Alexander GE (1971): Neuron activity related to short-term memory. Science 173:652– 654.

BIOL PSYCHIATRY 2008;64:739 –749 747 22. Funahashi S, Bruce CJ, Goldman-Rakic PS (1989): Mnemonic coding of visual space in the monkey’s dorsolateral prefrontal cortex. J Neurophysiol 61:331–349. 23. Rao SC, Rainer G, Miller EK (1997): Integration of what and where in the primate prefrontal cortex. Science 276:821– 824. 24. Miller EK, Li L, Desimone R (1993): Activity of neurons in anterior inferior temporal cortex during a short-term memory task. J Neurosci 13: 1460 –1478. 25. Aksay E, Gamkrelidze G, Seung HS, Baker R, Tank DW (2001): In vivo intracellular recording and perturbation of persistent activity in a neural integrator. Nat Neurosci 4:184 –193. 26. Supèr H, Spekreijse H, Lamme VA (2001): A neural correlate of working memory in the monkey primary visual cortex. Science 293:120 –124. 27. Quintana J, Yajeya J, Fuster JM (1988): Prefrontal representation of stimulus attributes during delay tasks. I. Unit activity in cross-temporal integration of sensory and sensory-motor information. Brain Res 474: 211–221. 28. Sobotka S, Diltz MD, Ringo JL (2005): Can delay-period activity explain working memory? J Neurophysiol 93:128 –136. 29. Miller EK, Erickson CA, Desimone R (1996): Neural mechanisms of visual working memory in prefrontal cortex of the macaque. J Neurosci 16: 5154 –5167. 30. Durstewitz D, Seamans JK, Sejnowski TJ (2000): Neurocomputational models of working memory. Nat Neurosci 3(suppl):1184 –1191. 31. Brozoski TJ, Brown RM, Rosvold HE, Goldman PS (1979): Cognitive deficit caused by regional depletion of dopamine in prefrontal cortex of rhesus monkey. Science 205:929 –932. 32. Sawaguchi T, Matsumara M, Kubota K (1990): Catecholaminergic effects on neuronal activity related to a delayed response task in monkey prefrontal cortex. J Neurophysiol 63:1385–1400. 33. Sawaguchi T, Goldman-Rakic PS (1994): The role of D1-dopamine receptor in working memory: Local injections of dopamine antagonists into the prefrontal cortex of rhesus monkeys performing an oculomotor delayed-response task. J Neurophysiol 71:515–528. 34. Seamans JK, Floresco SB, Phillips AG (1998): D1 receptor modulation of hippocampal-prefrontal cortical circuits integrating spatial memory with executive functions in the rat. J Neurosci 18:1613–1621. 35. Zahrt J, Taylor JR, Mathew RG, Arnsten AFT (1997): Supranormal stimulation of D1 dopamine receptors in the rodent prefrontal cortex impairs spatial working memory performance. J Neurosci 17:8528 – 8535. 36. Aujla H, Beninger RJ (2001): Hippocampal-prefrontocortical circuits: PKA inhibition in the prefrontal cortex impairs delayed nonmatching in the radial maze in rats. Behav Neurosci 115:1204 –1211. 37. Arnsten AF (2007): Catecholamine and second messenger influences on prefrontal cortical networks of “representational knowledge”: A rational bridge between genetics and the symptoms of mental illness. Cereb Cortex 17(suppl 1):i6 –i15. 38. Yang CR, Seamans JK (1996): Dopamine D1 receptor actions in layer v-vi rat prefrontal cortex neurons in vitro: Modulation of dendriticsomatic signal integration. J Neurosci 16:1922–1935. 39. Zheng P, Zhang XX, Bunney BS, Shi WX (1999): Opposite modulation by cortical N-Methyl-D-aspartate receptor-mediated responses by low and high concentrations of dopamine. Neuroscience 91:527–535. 40. Nicola SM, Surmeier J, Malenka RC (2000): Dopaminergic modulation of neuronal excitability in the striatum and nucleus accumbens. Annu Rev Neurosci 23:185–215. 41. Gao WJ, Krimer LS, Goldman-Rakic PS (2001): Presynaptic regulation of recurrent excitation by D1 receptors in prefrontal circuits. Proc Natl Acad Sci U S A 98:295–300. 42. Gao WJ, Goldman-Rakic PS (2003): Selective modulation of excitatory and inhibitory microcircuits by dopamine. Proc Natl Acad Sci U S A 100:2836 –2841. 43. Seamans JK, Durstewitz D, Christie BR, Stevens CF, Sejnowski TJ (2001): Dopamine D1/D5 receptor modulation of excitatory synaptic inputs to layer V prefrontal cortex neurons. Proc Natl Acad Sci U S A 98:301–306. 44. Seamans JK, Gorelova N, Durstewitz D, Yang CR (2001): Bidirectional dopamine modulation of GABAergic inhibition in prefrontal cortical pyramidal neurons. J Neurosci 21:3628 –3638. 45. Gorelova N, Seamans JK, Yang CR (2002): Mechanisms of dopamine activation of fast-spiking interneurons that exert inhibition in rat prefrontal cortex. J Neurophysiol 88:3150 –3166.

www.sobp.org/journal

748 BIOL PSYCHIATRY 2008;64:739 –749 46. Gorelova NA, Yang CR (2000): Dopamine D1/D5 receptor activation modulates a persistent sodium current in rat prefrontal cortical neurons in vitro. J Neurophysiol 84:75– 87. 47. Young CE, Yang CR (2004): Dopamine D1/D5 receptor modulates state-dependent switching of soma-dendritic Ca2⫹ potentials via differential protein kinase A and C activation in rat prefrontal cortical neurons. J Neurosci 24:8 –23. 48. Rosenkranz JA, Johnston D (2006): Dopaminergic regulation of neuronal excitability through modulation of Ih in layer V entorhinal cortex. J Neurosci 26:3229 –3244. 49. Durstewitz D, Kelc M, Güntürkün O (1999): A neurocomputational theory of the dopaminergic modulation of working memory functions. J Neurosci 19:2807–2822. 50. Durstewitz D, Seamans JK, Sejnowski TJ (2000): Dopamine-mediated stabilization of delay-period activity in a network model of prefrontal cortex. J Neurophysiol 83:1733–1750. 51. Compte A, Brunel N, Goldman-Rakic PS, Wang XJ (2000): Synaptic mechanisms and network dynamics underlying spatial working memory in a cortical network model. Cereb Cortex 10:910 –923. 52. Brunel N, Wang XJ (2001): Effects of neuromodulation in a cortical network model of object working memory dominated by recurrent inhibition. J Comput Neurosci 11:63– 85. 53. Durstewitz D, Seamans JK (2002): The computational role of dopamine D1 receptors in working memory. Neural Net 15:561–572. 54. Koch C, Segev I, editors (1997): Methods in Neuronal Modeling, 2nd ed. Cambridge, Massachusetts: MIT Press. 55. Hodgkin AL, Huxley AF (1952): A quantitative description of membrane current and its application to conduction and excitation in nerve. J Physiol 117:500 –544. 56. Fuster JM (1973): Unit activity in prefrontal cortex during delayedresponse performance: Neuronal correlates of transient memory. J Neurophysiol 36:61–78. 57. Wang XJ (1999): Synaptic basis of cortical persistent activity: The importance of NMDA receptors to working memory. J Neurosci 19: 9587– 9603. 58. Seamans JK, Nogueira L, Lavin A (2003): Synaptic basis of persistent activity in prefrontal cortex in vivo and in organotypic cultures. Cereb Cortex 13:1242–1250. 59. Thurley K, Senn W, Luscher HR (2008): Dopamine increases the gain of the input-output response of rat prefrontal pyramidal neurons [published online ahead of print April 9]. J Neurophysiol. 60. Rolls ET, Thorpe SJ, Boytim M, Szabo I, Perrett DI (1984): Responses of striatal neurons in the behaving monkey. 3. Effects of iontophoretically applied dopamine on normal responsiveness. Neuroscience 12:1201– 1212. 61. Servan-Schreiber D, Printz H, Cohen JD (1990): A network model of catecholamine effects: Gain, signal-to-noise ratio, and behavior. Science 249:892– 895. 62. Gulledge AT, Jaffe DB (1998): Dopamine decreases the excitability of layer V pyramidal cells in the rat prefrontal cortex. J Neurosci 18:9139 – 9151. 63. Trantham-Davidson H, Neely LC, Lavin A, Seamans JK (2004): Mechanisms underlying differential D1 versus D2 dopamine receptor regulation of inhibition in prefrontal cortex. J Neurosci 24:10652–10659. 64. Nishi A, Snyder GL, Greengard P (1997): Bidirectional regulation of DARPP-32 phosphorylation by dopamine. J Neurosci 17:8147– 8155. 65. Greengard P (2001): The neurobiology of dopamine signaling. Biosci Rep 21:247–269. 66. Durstewitz D (2007): Dopaminergic modulation of prefrontal cortex network dynamics. In: Tseng K-Y, Atzori M, editors. Monoaminergic Modulation of Cortical Excitability. New York: Springer. 67. Loh M, Rolls ET, Deco G (2007): A dynamical systems hypothesis of schizophrenia. PLoS Comput Biol: e228. 68. Cubeddu LX, Hoffmann IS, Talmaciu RK (1990): Is the release of dopamine from medial prefrontal cortex modulated by presynaptic receptors? Comparison with nigrostriatal and mesolimbic terminals. Ann N Y Acad Sci 604:452– 461. 69. O’Reilly RC, Noelle DC, Braver TS, Cohen JD (2002): Prefrontal cortex and dynamic categorization tasks: representational organization and neuromodulatory control. Cereb Cortex 12:246 –257. 70. Sawaguchi T, Matsumara M, Kubota K (1988): Dopamine enhances the neuronal activity of spatial short-term memory task in the primate prefrontal cortex. Neurosci Res 5:465– 473.

www.sobp.org/journal

D. Durstewitz and J.K. Seamans 71. Vijayraghavan S, Wang M, Birnbaum SG, Williams GV, Arnsten AF (2007): Inverted-U dopamine D1 receptor actions on prefrontal neurons engaged in working memory. Nat Neurosci 10:376 –384. 72. Lavin A, Nogueira L, Lapish CC, Wightman RM, Phillips PE, Seamans JK (2005): Mesocortical dopamine neurons operate in distinct temporal domains using multimodal signaling. J Neurosci 25:5013–5023. 73. Wang M, Vijayraghavan S, Goldman-Rakic PS (2004): Selective D2 receptor actions on the functional circuitry of working memory. Science 303:853– 856. 74. Floresco SB, Magyar O (2006): Mesocortical dopamine modulation of executive functions: Beyond working memory. Psychopharmacology (Berl) 188:567–585. 75. Lee B, Groman S, London ED, Jentsch JD (2007): Dopamine D2/D3 receptors play a specific role in the reversal of a learned visual discrimination in monkeys. Neuropsychopharmacology 32:2125–2134. 76. Floresco SB, Magyar O, Ghods-Sharifi S, Vexelman C, Tse MT (2006): Multiple dopamine receptor subtypes in the medial prefrontal cortex of the rat regulate set-shifting. Neuropsychopharmacology 31: 297–309. 77. Floresco SB, Phillips AG (2001): Delay-dependent modulation of memory retrieval by infusion of a dopamine D1 agonist into the rat medial prefrontal cortex. Behav Neurosci 115:934 –939. 78. Granon S, Passetti F, Thomas KL, Dalley JW, Everitt BJ, Robbins TW (2000): Enhanced and impaired attentional performance after infusion of D1 dopaminergic receptor agents into rat prefrontal cortex. J Neurosci 20:1208 –1215. 79. Druzin MY, Kurzina NP, Malinina EP, Kozlov AP (2000): The effects of local application of D2 selective dopaminergic drugs into the medial prefrontal cortex of rats in a delayed spatial choice task. Behav Brain Res 109:99 –111. 80. Zhang K, Grady CJ, Tsapakis EM, Andersen SL, Tarazi FI, Baldessarini RJ (2004): Regulation of working memory by dopamine D4 receptor in rats. Neuropsychopharmacology 29:1648 –1655. 81. Müller U, von Cramon DY, Pollmann S (1998): D1- versus D2-receptor modulation of visuospatial working memory in humans. J Neurosci 18:2720 –2728. 82. Tost H, Meyer-Lindenberg A, Klein S, Schmitt A, Hohn F, Tenckhoff A, et al. (2006): D2 antidopaminergic modulation of frontal lobe function in healthy human subjects. Biol Psychiatry 60:1196 –1205. 83. Mehta MA, Manes FF, Magnolfi G, Sahakian BJ, Robbins TW (2004): Impaired set-shifting and dissociable effects on tests of spatial working memory following the dopamine D2 receptor antagonist sulpiride in human volunteers. Psychopharmacology (Berl) 176:331–342. 84. McDowell S, Whyte J, D’Esposito M (1998): Differential effect of a dopaminergic agonist on prefrontal function in traumatic brain injury patients. Brain 121(Pt 6):1155–1164. 85. Lumme V, Aalto S, Ilonen T, Nagren K, Hietala J (2007): Dopamine D2/D3 receptor binding in the anterior cingulate cortex and executive functioning. Psychiatry Res 156:69 –74. 86. Dreisbach G, Müller J, Goschke T, Strobel A, Schulze K, Lesch KP, Brocke B (2005): Dopamine and cognitive control: The influence of spontaneous eyeblink rate and dopamine gene polymorphisms on perseveration and distractibility. Behav Neurosci 119:483– 490. 87. Luciana M, Collins PF (1997): Dopminergic modulation of working memory for spatial but not object cues in normal volunteers. J Cogn Neurosci 9:330 –347. 88. Mehta MA, Manes FF, Magnolfi G, Sahakian BJ, Robbins TW (2004): Impaired set-shifting and dissociable effects on tests of spatial working memory following the dopamine D2 receptor antagonist sulpiride in human volunteers. Psychopharmacology (Berl) 176:331–342. 89. Mehta MA, Swainson R, Ogilvie AD, Sahakian J, Robbins TW (2001): Improved short-term spatial memory but impaired reversal learning following the dopamine D(2) agonist bromocriptine in human volunteers. Psychopharmacology (Berl) 159:10 –20. 90. Seamans JK, Yang CR (2004): The principal features and mechanisms of dopamine modulation in the prefrontal cortex. Prog Neurobiol 74:1–58. 91. West AR, Grace AA (2002): Opposite influences of endogenous dopamine D1 and D2 receptor activation on activity states and electrophysiological properties of striatal neurons: Studies combining in vivo intracellular recordings and reverse microdialysis. J Neurosci 22:294 –304.

D. Durstewitz and J.K. Seamans 92. Meyer-Lindenberg A, Weinberger DR (2006): Intermediate phenotypes and genetic mechanisms of psychiatric disorders. Nat Rev Neurosci 7:818 – 827. 93. Williams GV, Goldman-Rakic PS (1995): Modulation of memory fields by dopamine D1 receptors in prefrontal cortex. Nature 376:572–575. 94. Sibley DR, De Lean A, Creese I (1982): Anterior pituitary dopamine receptors. Demonstration of interconvertible high and low affinity states of the D-2 dopamine receptor. J Biol Chem 257:6351– 6361. 95. Richfield EK, Penney JB, Young AB (1989): Anatomical and affinity state comparisons between dopamine D1 and D2 receptors in the rat central nervous system. Neuroscience 30:767–777. 96. Seeman P, Weinshenker D, Quirion R, Srivastava LK, Bhardwaj SK, Grandy DK, et al. (2005): Dopamine supersensitivity correlates with D2High states, implying many paths to psychosis. Proc Natl Acad Sci U S A 102:3513–3518. 97. Seeman P, Schwarz J, Chen JF, Szechtman H, Perreault M, McKnight GS, et al. (2006): Psychosis pathways converge via D2high dopamine receptors. Synapse 60:319 –346. 98. Lapish CC, Kroener S, Durstewitz D, Lavin A, Seamans JK (2007): The ability of the mesocortical dopamine system to operate in distinct temporal modes. Psychopharmacology (Berl) 191:609 – 625. 99. Smiley JF, Levey AI, Ciliax BJ, Goldman-Rakic PS (1994): D1 dopamine receptor immunoreactivity in human and monkey cerebral cortex: Predominant and extrasynaptic localization in dendritic spines. Proc Natl Acad Sci U S A 91:5720 –5724. 100. Sesack SR, Aoki C, Pickel VM (1994): Ultrastructural localization of D2 receptor-like immunoreactivity in midbrain dopamine neurons and their striatal targets. J Neurosci 14:88 –106. 101. Gogos JA, Morgan M, Luine V, Santha M, Ogawa S, Pfaff D, Karayiorgou M (1998): Catechol-O-methyltransferase-deficient mice exhibit sexually dimorphic changes in catecholamine levels and behavior. Proc Natl Acad Sci U S A 95:9991–9996. 102. Matsumoto M, Weickert CS, Akil M, Lipska BK, Hyde TM, Herman MM, et al. (2003): Catechol O-methyltransferase mRNA expression in human and rat brain: Evidence for a role in cortical neuronal function. Neuroscience 116:127–137. 103. Tunbridge EM, Bannerman DM, Sharp T, Harrison PJ (2004): Catecholo-methyltransferase inhibition improves set-shifting performance and elevates stimulated dopamine release in the rat prefrontal cortex. J Neurosci 24:5331–5335. 104. Lachman HM, Papolos DF, Saito T, Yu YM, Szumlanski CL, Weinshilboum RM (1996): Human catechol-O-methyltransferase pharmacogenetics: Description of a functional polymorphism and its potential application to neuropsychiatric disorders. Pharmacogenetics 6:243– 250. 105. Weinberger DR (2001): Effect of COMT val108/158 met genotype on frontal lobe function and risk for schizophrenia. Proc Natl Acad Sci U S A 98:6917– 6922. 106. Egan MF, Goldberg TE, Kolachana BS, Callicott JH, Mazzanti CM, Straub RE, et al. (2001): Effect of COMT val108/158 met genotype on frontal lobe function and risk for schizophrenia. Proc Natl Acad Sci U S A 98: 6917– 6922. 107. Nolan KA, Bilder RM, Lachman HM, Volavka J (2004): Catechol O-methyltransferase val158met polymorphism in schizophrenia: Differential effects of val and met alleles on cognitive stability and flexibility. Am J Psychiatry 161:359 –361. 108. Drabant EM, Hariri AR, Meyer-Lindenberg A, Munoz KE, Mattay VS, Kolachana BS, et al. (2006): Catechol O-methyltransferase val158met genotype and neural mechanisms related to affective arousal and regulation. Arch Gen Psychiatry 63:1396 –1406.

BIOL PSYCHIATRY 2008;64:739 –749 749 109. Mattay VS, Goldberg TE, Fera F, Hariri AR, Tessitore A, Egan MF, et al. (2003): Catechol O-methyltransferase val158-met genotype and individual variation in the brain response to amphetamine. Proc Natl Acad Sci U S A 100:6186 – 6191. 110. Apud JA, Mattay V, Chen J, Kolachana BS, Callicott JH, Rasetti R, et al. (2007): Tolcapone improves cognition and cortical information processing in normal human subjects. Neuropsychopharmacology 32:1011–1020. 111. Barnett JH, Jones PB, Robbins TW, Müller U (2007): Effects of the catechol-O-methyltransferase val158met polymorphism on executive function: A meta-analysis of the Wisconsin Card Sort Test in schizophrenia and healthy controls. Mol Psychiatry 12:502–509. 112. Koren D, Seidman LJ, Harrison RH, Lyons MJ, Kremen WS, Caplan B, et al. (1998): Factor structure of the Wisconsin Card Sorting Test: Dimensions of deficit in schizophrenia. Neuropsychology 12:289 –302. 113. Lanser MG, Berger HJ, Ellenbroek BA, Cools AR, Zitman FG (2002): Perseveration in schizophrenia: Failure to generate a plan and relationship with the psychomotor poverty subsyndrome. Psychiatry Res 112: 13–26. 114. Stratta P, Daneluzzo E, Bustini M, Prosperini P, Rossi A (2000): Processing of context information in schizophrenia: Relation to clinical symptoms and WCST performance. Schizophr Res 44:57– 67. 115. Winterer G, Musso F, Vucurevic G, Stoeter P, Konrad A, Seker B, et al. (2006): COMT genotype predicts BOLD signal and noise characteristics in prefrontal circuits. Neuroimage 32:1722–1732. 116. Winterer G, Ziller M, Dorn H, Frick K, Mulert C, Dahhan N, et al. (1999): Cortical activation, signal-to-noise ratio and stochastic resonance during information processing in man. Clin Neurophysiol 110:1193–1203. 117. Winterer G, Ziller M, Dorn H, Frick K, Mulert C, Wuebben Y, et al. (2000): Schizophrenia: Reduced signal-to-noise ratio and impaired phase-locking during information processing. Clin Neurophysiol 111:837– 849. 118. Winterer G, Egan MF, Kolachana BS, Goldberg TE, Coppola R, Weinberger DR (2006): Prefrontal electrophysiologic “noise” and catecholO-methyltransferase genotype in schizophrenia. Biol Psychiatry 60:578 –584. 119. Winterer G, Musso F, Beckmann C, Mattay V, Egan MF, Jones DW, et al. (2006): Instability of prefrontal signal processing in schizophrenia. Am J Psychiatry 163:1960 –1968. 120. Durstewitz D (2006): A few important points about dopamine’s role in neural network dynamics. Pharmacopsychiatry 39(suppl 1):S72–S75. 121. Gallinat J, Bajbouj M, Sander T, Schlattmann P, Xu K, Ferro EF, et al. (2003): Association of the G1947A COMT (val(108/158)met) gene polymorphism with prefrontal P300 during information processing. Biol Psychiatry 54:40 – 48. 122. Winterer G, Coppola R, Goldberg TE, Egan MF, Jones DW, Sanchez CE, Weinberger DR (2004): Prefrontal broadband noise, working memory, and genetic risk for schizophrenia. Am J Psychiatry 161:490 –500. 123. Weinberger DR (1987): Implications of normal brain development for the pathogenesis of schizophrenia. Arch Gen Psychiatry 44:660 – 669. 124. Goldman-Rakic PS (1999): The physiological approach: Functional architecture of working memory and disordered cognition in schizophrenia. Biol Psychiatry 46:650 – 661. 125. Abi-Dargham A, Moore H (2003): Prefrontal DA transmission at D1 receptors and the pathology of schizophrenia. Neuroscientist 9:404 – 416. 126. Bowie CR, Harvey PD (2005): Cognition in schizophrenia: Impairments, determinants, and functional importance. Psychiatr Clin North Am 28: 613– 633, 626. 127. Durstewitz D, Gabriel T (2007): Dynamical basis of irregular spiking in NMDA-driven prefrontal cortex neurons. Cereb Cortex 17:894 –908.

www.sobp.org/journal