Central Norepinephrine Neurons and Behavior

Trevor W. Robbins and Barry J. Everitt


I.  Introduction

      The neurobiological data reviewed in past and present articles in the Generation of Progress series (50; see chapters by Foote and Aston-Jones and by Valentino and Aston-Jones in this volume) and summarized in Table 1 provide several clues to the functions of the LC in the behaving animal. However, extrapolations from such data to psychological processes must be made with care. Clearly, the widespread nature of the ceruleo-cortical projection indicates that activation of this noradrenergic cell group will have pervasive effects in diverse terminal regions such as neocortex, hippocampus and the amygdala. It is perhaps no surprise that the LC has been implicated in  such distinct processes as learning, memory, attention and anxiety, which clearly depend to different degrees on these discrete regions of the forebrain. Two critical questions  are: (i) what effect, if any, does such activation have on cognitive and behavioral processes ? and (ii) under what circumstances does such activation normally occur? The first question can only be answered by studying the behaving animal and inferring changes in psychological function. Thus it is predicted that manipulations of ceruleo-cortical function will affect various psychological processes including associative learning, different forms of memory and attention depending, in large part, on which terminal domains of the noradrenergic projection are especially engaged during the task under study. Presumably, the distributed nature of these effects represents an integrated, adaptive response to the environmental and behavioral setting. The electrophysiological data and also neurochemical indices, show that LC neurons are especially active during relatively high states of arousal, including exposure to stressful environments and salient, phasic stimuli (see chapter by Foote and Aston-Jones in this volume). Thus, in stressful circumstances it may be useful not only to consolidate memories more efficiently, but also to focus attention on salient features of the environment: the ceruleo-cortical noradrenergic system is clearly suited to such a function.

II.  Behavioural functions of the locus coeruleus (LC)

A.  Electrophysiological studies in the behaving animal

      Measuring LC activity concomitantly with behavior provides useful information on the functions of the LC norepinephrine (NE) neurons, although it must be borne in mind that any correlative study of this sort does not establish the overall significance of the LC in particular behavioral processes. The mere presence of neuronal activity, even if highly correlated with environmental events, do not necessarily bear implications for causal factors in behaviour ,which are best assessed by interventions which affect NE function more directly.

      The observation that LC noradrenergic neurons are active during stressful circumstances is clearly consistent with suggestions that they may be involved in the learning of aversively motivated tasks (see below). It is noteworthy, therefore, that Jacobs (50) has demonstrated marked increases in NE neuron unit activity during presentation of a CS previously paired with an aversive air puff to the face of a cat. The selectivity of this response of LC neurons was emphasized by the observation that the same CS previously paired with a rewarding stimulus did not evoke increases in NE neuronal firing. While these data strongly suggest LC involvement in learning associated with stressful stimuli, for which there is considerable support from behavioral studies (see below), they do not account for the observation that learning of some appetitively motivated tasks is  significantly retarded following lesions of the ceruleo-cortical noradrenergic projection (38).

      Indeed, recent data suggest that  LC neurons are active during appetitive learning. In rats, there was an immediate response of LC cells to any change in stimulus-reinforcer contingencies in appetitive and aversive conditioning, often prior to the behavioral expression of conditioning (75).  The changes were even more reliable than the  responses to novelty and were unrelated to movements, disappearing when the behavioral response was well established. In primates, LC neurons are active during the performance of a vigilance or  "oddball" visual discrimination task that is appetitively motivated. Monkeys were trained to release a lever after a target cue light that occurred randomly on 10% of trials and to withhold responding during non-target cues. LC neurons responded selectively to the target cues on this task,  and rapidly reversed this salience if the non-target stimuli became relevant instead (10,11,12 ). High levels of LC discharge were related to decreased foveation, restlessness and impaired task performance (10). Importantly, cortical, event-related potentials were elicited in this task selectively by the same stimuli that evoked LC responses, suggesting that the LC activation was associated with cortical processing mechanisms and in learning the significance of behaviorally important stimuli.

      At the cortical level,  P-300 like potentials  have been recorded in squirrel monkeys in other "oddball"  paradigms where the animal responds to low probability auditory  stimuli (65). The generation and modulation of the P-300 like auditory potential  was shown to be impaired by electrolytic lesions of the LC, although  more specific destruction of ascending NE  fibres failed to have the same effects.

      In general these electrophysiological studies are consistent with a role for LC NE neurons in processes related to the processing of novelty, including new contingencies that require learning.

B.  Effects of neurotoxic lesions of the dorsal bundle on behavioral function

      Another approach to investigating the functions of the LC NE system has been to define the behavioral effects of selectively removing the rostral, noradrenergic projections of the LC, and the environmental circumstances under which such effects occur. Profound levels of forebrain NE depletion can be achieved by injecting the selective, catecholamine neurotoxin 6-hydroxydopamine (6-OHDA) into the trajectory of the LC axons in the midbrain - namely, into the dorsal noradrenergic bundle (DB). The rationale underlying this neurochemically highly specific approach to studying LC function has been discussed extensively elsewhere (61,70) and will not be labored here. With optimal parameters (see Figure 1) it is feasible to deplete the telencephalic projection zones of NE to less than 10% of control values without any effect on any other known neurotransmitter system  Indeed, such potent  degrees of depletion, as assessed by tissue levels of NE, are essential if comparable reductions of extracellular NE are to be achieved (1). Smaller degrees of depletion clearly enhance possible plastic and recuperative responses of noradrenergic neurons to 6-OHDA lesions, including up-regulation of adrenoceptor populations  (2,34). Such problems may also make difficult interpretation of the effects of the alternative noradrenergic neurotoxin DSP-4, which can lead to cortical NE loss of less than 70%  (e.g. 91). Conclusions concerning the lack of involvement of central NE in particular psychological processes such as memory and learning are clearly unsafe if depletions are substantially less than 90%. On the other hand, several groups have consistently demonstrated robust and long-lasting behavioral deficits associated with depletion over 90% of telencephalic NE , which would seem hard to explain in terms of changes leading to over-compensation of the damaged system (see 70). However, there are two major problems specifically related to the effects of DB lesions: (i) assessing the possible contaminating effects of hypothalamic NE depletion caused unavoidably by diffusion of the neurotoxin to fibres projecting from the cell bodies of the medulla oblongata contributing to the so-called ventral noradrenergic bundle (VB). This can be addressed by making control lesions of this structure which, of course, also contribute to our understanding of the role of hypothalamic NE.  (ii) Isolating the critical terminal region for a particular effect; this problem will be addressed in section IIC. While there are undoubtedly problems associated with the use of neurotoxic lesions of the DB or VB to understand the behavioral functions of the central NE systems, these have to be weighed against the difficulties of interpreting effects of  systemic treatments with adrenergic receptor agents, for example  because of lack of pharmacological specificity and  peripheral factors.  However, such experiments can be helpful in  providing converging sources of evidence and will be mentioned as appropriate.

1.  Effects on unconditioned behavior. DB lesions have no obvious effect on gross behavior in the rat, such as eating, drinking or spontaneous locomotor activity. This contrasts markedly with the changes produced in these forms of behavior by central dopamine depletion, as well as the increases in body weight and feeding often observed following hypothalamic NE depletion (48,73). This lack of effect of DB lesions on gross ingestive and locomotor responses simplifies the interpretation of many of the other effects of the lesion. However, it is evident that DB lesions do affect some forms of unconditioned behavior, for example, behavioral responses to novelty, as might be expected from the single unit studies that show stimulus novelty to be an effective trigger for LC neurons. As will become obvious for many of the effects of these lesions, however the sequelae appear to depend on subtle features of the testing situation (see 23 for a review). For example, DB lesions can reduce feeding overall in the threatening situation of a highly illuminated open field, and yet attenuate the suppression of eating normally shown to a novel food. Steketee et al have replicated the attenuation of this food neophobia in novel situations following DB lesions and have found that  it can be attenuated by  icv treatment with the beta receptor agonist isoproterenol (86).

2.  Effects on conditioning and conditioned behavior      One of the most important generalizations to have emerged about the effects of  DB lesions is that they tend to impair the acquisition of new behavior to a greater extent than previously established performance. The concept that the DB is implicated in learning is by now quite venerable, but some of the original experiments were not convincing and there are many forms of learning that are not reliably affected by DB lesions.   However, there is some evidence that under well-defined conditions, DB lesions (and DSP-4 induced NE depletion, 3) reliably retard the gradual learning of an appetitive, conditional discrimination task, in which the rat is required to learn a rule (press right or left lever) to guide its response to one of two discriminative stimuli (e.g. lights flashing at one of two different frequencies). On the other hand such lesions do not impair its performance if the lesion is made following the establishment of the discrimination by pre-operative training (38). This clearly rules out many explanations of the effects of DB lesions based on simple performance factors such as changes in sensory capacity or motivation.

       The dichotomy  between acquisition and performance has also been observed for aversive conditioning: thus, conditioned suppression of food-maintained operant responding in the presence of a light acting as an aversive conditioned stimulus (CS) is attenuated in acquisition following surgery, but is unaffected if established prior to surgery (21). This dissociation argues against a simple explanation of the acquisition deficit in terms of an intervening variable such as "anxiety", and this is supported by demonstrations of a lack of effect of DB lesions on response suppression in the Geller-Seifter conflict paradigm (54).  Furthermore, the acquisition of a possible oral coping response (e.g. gnawing, eating) to an unconditioned aversive event, tail-pinch, is impaired following DB lesions, but  not if the response is established by experience prior to surgery (73). However, we should emphasize that associative processes are not always impaired by such lesions. For example,  simultaneous visual discriminations, where the rat simply has to approach the rewarded stimulus and the effects of reversing the contingencies so that the animal has to approach the formerly non-reinforced cue, are not affected by DB lesions (37). Conditioned taste aversion has also proved particularly resistant to disruption following DB lesions (35). Moreover, the effects on other forms of  aversive learning and performance are complex and may depend on such factors as the nature and intensity of the aversive reinforcer, time elapsed between the DB lesion and testing, and the precise conditioning contingencies and test situation used (see review, 80). In general, VB lesions do not affect acquisition of appetitive or aversive conditioning, although they have been found to increase resistance to extinction of the latter, both for conditioned suppression and conditioned taste aversion (21,35). It seems plausible that these effects of VB lesions on extinction represent possible interactions with neuroendocrine mechanisms in the hypothalamic-pituitary axis, and that at least some of the earlier reports of extinction deficits following DB lesions (61)  might be attributable to effects on neurons of the VB.

      Contextual aversive conditioning (that is conditioning to the background cues rather than to an explicit  conditioned stimulus (CS)) is also spared by DB lesions; in fact such conditioning may be enhanced by ceruleo-cortical NE loss, either when an explicit CS is also present (77) or when it is not (80). The difference between aversive conditioning to discrete, explicit cues and to the wider contexts in which they occur has obvious applicability to various forms of clinical anxiety.  Plasma corticosterone is also affected by such lesions, but only in a manner predicted by the conditioned behavioral response (77). However, the complementary pattern of impaired CS conditioning and enhanced contextual conditioning seen in rats with DB lesions is another piece of evidence against a simple anxiolytic view of their effects, and may instead argue for effects on attentional function (see below).

      A  possibly related effect of DB lesions on contextual conditioning is seen in spatial learning in the Morris swim maze, where the rat is required to learn to use distal cues around the room to locate a safe, but invisible platform. DB lesions again may actually enhance acquisition of the task (55), especially if the rat is swimming in cold water (78). In this situation the DB lesion appears to protect the rat against the deleterious effects on learning in water sufficiently cold to  lead to hypothermia, while having no effect by itself on body temperature in the swim-maze, or on plasma corticosterone  levels (78). In contrast, VB lesions have no effects on spatial (78) or contextual (77) conditioning, although they do affect plasma corticosteroid levels to the shock (81), thereby showing a dissociation between effects on  conditioning (by DB lesions) and endocrine status  (by VB lesions ) which probably corresponds to the relative roles of these systems in cognitive and vegetative functions, respectively.

      In attempting to resolve the, at first sight, bewildering,  pattern of effects of ceruleo-cortical NE depletion on these different forms of learning,  strong unifying themes can in fact be found. One such theme is the difficulty or sensitivity of the learning procedure, another is the task-associated level of arousal; either factor might explain the relatively greater sensitivity of aversive paradigms to learning deficits following DB lesions. Thus, the conditional appetitive discrimination task which has been repeatedly shown (3,38,70) to show deficits in DB lesioned rats typically takes many sessions to learn even for a normal rat. Another factor is  the attentional requirements of the tasks; it is possible that some of the DB effects result not from deficits in associative processes, but in the input to the associative mechanisms. Thus, the dissociation between CS and contextual learning may suggest that DB rats are utilizing more distal cues than the normal rat, which, in some situations may produce a paradoxical enhancement of acquisition.  This prediction was tested directly in the water-maze task with a discrimination between two local sets of cues (platforms with vertical or horizontal stripes). The DB lesion impaired acquisition, despite facilitating acquisition of the spatial variant of the task (78),  thus again showing a dissociation of effects on local versus distal cues, that possibly results from shifts in attention.

3.  Effects on attention  Since the original Segal and Bloom (76)  findings of enhanced S/N ratios following iontophoretically applied NE in hippocampus, there have been a rash of hypotheses concerning the role of the LC in selective attention (61,70). An earlier review (70) made it clear that direct tests of this hypothesis employing paradigms including latent inhibition, blocking and non-reversal shift have not  generally confirmed earlier positive findings (61, but see 72). However, there have been some interesting findings that probably require further investigation (e.g.57,89). Devauges and Sara (33) have suggested that activation of central NE mechanisms by the alpha-2 receptor antagonist, idazoxan, can lead to apparent improvements in shifting of attention between different types of cue, although when state-dependent  controls were included it was shown that this drug impaired non-reversal shift learning in a maze between spatial and visual cues (72).

      In assessing the role of the LC in attentional mechanisms, it is notable that there are several different forms of attention, including selective (focused) attention, divided attention and sustained attention, including vigilance. Deficits can be observed following DB lesions  in a continuous performance task which requires some of these attentional capacities, under certain conditions (17,25). As might be predicted, these deficits only occurred under very specific test conditions, namely when bursts of loud white noise were interpolated immediately prior to expected visual targets,  when the rats were treated with amphetamine, or when the stimuli were temporally unpredictable, a manipulation that increases arousal. There were no such deficits in attention when the task was merely made more difficult by, for example, reducing the brightness of the discriminanda or by increasing the frequency of their presentation. The effect of white noise is manifestly consistent with the demonstration that LC neurons fire in response to such phasic stimuli (50, see chapter by Foote and Aston-Jones in this volume) and also with demonstrations of enhanced distractibility in the maze situation (71), and possibly the enhanced reactivity of DB lesioned rats in the open field setting (15) , as described above. The effect of temporal unpredictability is important in that it suggests the LC can be activated in quite complex ways, viz. not only in response to the initial occurrence of a novel stimulus, but also to the omission of an expected event. One of the concomitant effects of white noise is to produce behavioral activation, which can be manifested as quicker reaction times and a propensity to impulsive responding (17). Such behavioral activation is also produced by the dopaminergic agonist, D-amphetamine, especially when infused into the region of the nucleus accumbens septi (22,24). Thus, it would be expected that discriminative accuracy in the visual vigilance task would also be impaired in rats with DB lesions following intra-accumbens infusions of amphetamine and this result has been found (22).  The  degree of behavioral activation produced by amphetamine was unaffected by the DB lesion (22). Thus, under conditions of equivalent behavioral activation resulting from the endogenous cue of dopamine release, DB-lesioned rats became less efficient at detecting brief visual signals. The parallel with the effects of white noise is enhanced by evidence that dopamine depletion within the ventral striatum attenuates the activating effects of this stimulus as well as those of D-amphetamine (24). The lack of effect of DB lesions in a test well-validated to measure sustained attention or behavioral vigilance has supported the general lack of effects of DB lesions on the basic version of the continuous performance test analogue (59). Perhaps the advent of relatively specific alpha-1 and alpha-2 adrenoceptor agents will enable further tests of these hypotheses if the drugs are administered centrally: thus far, however, it is evident that systemic administration of adrenoceptor agonists or antagonists do not strongly mimic effects of DB lesions (e.g. 83).

C.  Interactions of LC with specific terminal regions

      It is clear that DB lesions  can affect a variety of behavioral processes in carefully defined conditions, but it is less clear  how readily these may be attributed to the different projection fields of the LC. There are several strategies available for addressing this problem; one is to attempt to mimic the effects of  a DB lesion with a manipulation of a discrete terminal zone, including local NE depletion or the acute modulation of NE function via intracerebral administration of specific adrenoceptor agonists or antagonists. In the former case, there are problems posed by the considerable  plasticity following damage to terminal regions, which is undoubtedly greater than following damage at the level of the cell body or fibre bundle. In the latter instance, the problems of interpretation resulting from diffusion and local concentration are equally challenging, but they leave open the possibility of boosting local NE function and predicting an opposite effect to that of DB lesions. Not surprisingly, progress in this area has been limited.

      When interpreting the effects of DB lesions on aversive conditioning, there is considerable evidence of an involvement of the amygdala in learning about aversive CSs (32,44,58), whereas there is complementary evidence for a role of the hippocampus in  aversive conditioning to context (79). Local depletion of NE from the region of the amygdala impairs conditioning to aversive CSs in the same way as does DB lesions (77)  although local depletion of hippocampal NE on contextual conditioning has not been studied.  The behavioral consequences of NE depletion from both regions following DB lesions may depend, therefore, on the relative degree of engagement of hippocampal versus amygdaloid mechanisms in any particular task or situation.

      This principle may also apply to the case of choosing between familiar and novel food in a novel environment, especially as both amygdala and hippocampal lesions probably affect the responses to, perhaps different, aspects of novelty There has been relatively little investigation of the effects of terminal NE manipulations on the response to novelty. However, Borsini and Rolls (14) found that 6-OHDA lesions of, and also NE infusions into, the amygdala affected the response to novel foods, although these two manipulations did not have the expected opposite pattern of effects. In assessing a locomotor response to a novel environment, Flicker and Geyer (40) found that chronic infusions of NE into the dentate gyrus retarded the habituation of spatial exploration. These two sets of results are intriguing in that they suggest that terminal NE manipulations can affect different aspects of the response to novelty at distinct neural sites, a conclusion fully consistent with findings of the different effects of DB lesions  on response to different aspects of novelty (see above).

      With respect to the neocortex, there has similarly been little direct investigation of the role of NE in behavioral paradigms. Performance of the 5-choice attentional task described above. is known to depend on the rat neocortex, especially implicating anterior regions (63).  Other evidence has shown the involvement of NE mechanisms in processes of working memory in the primate prefrontal cortex, using the delayed response paradigm.  The decline in working memory in aged rhesus monkeys is significantly ameliorated by systemic treatment with the alpha-2 adrenergic agonist, clonidine. Arnsten and Goldman-Rakic  (7) further showed that prefrontal cortical ablation (around the sulcus principalis), or local noradrenergic denervation of this area induced by 6-OHDA, both disrupt performance on the task - confirming its dependence on this area of neocortex. Clonidine was able to reduce the NE lesion-induced deficit, but not the cortical ablation-induced deficit, emphasizing the interpretation that not only does delayed response performance depend on the prefrontal cortex, but that NE interacting with postsynaptic alpha-2 receptors in this site appears to be an important component of the processes occurring there (7). Guanfacine, a more specific alpha-2A agonist with less marked sedative and cardiovascular effects, was even more potent than clonidine in improving memory in this spatial delayed response paradigm in young (41) as well as old (9) adult  monkeys .

      While beneficial effects of clonidine have also been observed in a delayed matching to sample task in monkeys (49), other investigators have failed to observe effects of  systemic clonidine on delayed response performance in aged monkeys (31); this discrepancy might depend on subtle differences in test setting and procedure. For example, the delayed response task in these two studies was implemented in different ways; notably the latter investigation (31) employed an automated procedure whereas the studies by Arnsten and colleagues (e.g. 7,9) utilize a manual procedure in an environment susceptible to distraction. The possible importance of the attentional requirements of the task is provided by the study of Arnsten and Contant (6) showing that  alpha-2 receptor agonists  have strong protective effects against extraneous distraction in the  aged animals during performance of the delayed response task, and that these effects can be blocked by treatment with a an alpha-2 receptor antagonist acting predominantly at post-synaptic receptors. These findings are compatible with evidence reviewed above suggesting that DB lesions enhance distractibility in certain settings in the rat.

      In recent work, the disruptive effects of an alpha-1 adrenergic/imidazoline agonist (cirazoline) has been shown to impair spatial working memory performance in the delayed response task at very low doses, effects that were antagonized by the alpha-1 receptor antagonist prazosin (8). Similar findings have been reported in the rat when alpha-1 agents are infused directly into the prefrontal cortex (see 5). Arnsten has synthesized these findings  to suggest that post-synaptic alpha-1 and alpha-2 receptors may have opposing functions in the prefrontal cortex, just as they do in other brain regions (5).

      Overall, these results provide evidence that the effects of noradrenergic transmission through the ceruleo-cortical system are mediated through quite specific areas of its termination, whether neocortical, archaecortical (hippocampus) or involving subcortical structures of the limbic forebrain such as the amygdala, although the basolateral component of this nuclear complex may, in fact, be more cortical than subcortical in terms of both structure and connections. Overall, it is apparent that very different effects, e.g. of adrencoceptor agonists have been found, possibly as a function of the behavioral task under study, which in turn would relate to fundamental differences in the types of processing occurring in the terminal domains.

D.  Role of ceruleo-cortical NE in memory  and other forms of plasticity  

      Early theories of LC function emphasized its possible role in memory consolidation, which may contribute to some of the selective effects of DB lesions on acquisition (see 70).  This can be contrasted with the considerable evidence, in the rat, of spared short term memory function following DB lesions, for example, in delayed matching to position (49) and delayed alternation (66) tasks. It appears that, as well as acute effects on attentional mechanisms, the DB may also be implicated in rather longer term plastic changes in synaptic function that contribute to learning, perhaps particularly in tasks requiring lengthy training (see above).  Experiments utilizing post-trial noradrenergic manipulations on the retention of one-trial aversive conditioning tasks have examined possible direct or modulatory roles of NE in the consolidation of memory traces (44,56,58). In general, the data suggest that low doses of NE infused into the amygdala facilitate retention (56), but higher doses are either without effect (56) or are amnesic (36). Furthermore, intra-amygdala beta-blockers are also amnesic in their effects (44) Evidence suggests that central NE mechanisms, probably within the amygdala, are a final common pathway for a variety of amnestic and promnestic treatments, for example, adrenaline infusions (58) and treatment with CRF (18), respectively.

      A further locus of  plastic  change is the hippocampus, a structure typically associated with aspects of learning and memory, and there is evidence that depletion of NE can reduce LTP in the dentate gyus, a finding supported by demonstrations of an initiation of short- and long-term potentiation of the dentate gyrus response to perforant path input by the exogenous or endogenous application of NE (see review, 47). New evidence suggests that noradrenergic depletion (e.g. by DSP-4) can exacerbate working memory deficits in the rat produced by intra-hippocampal blockade of muscarinic or NMDA receptors (e.g. 64).

      In parallel with the early theorising on memory consolidation were suggestions that  manipulation of cortical NE could affect visual development (51). Destruction of central NE systems in neonatal rats does influence the degree of recovery from frontal cortex lesions when assessed behaviorally during adulthood (53). However, these effects  are  controversial and still being re-evaluated (67). A recent study has shown that the DSP-4 induced disruption of the imprinting response in the chick which has a large visual component and depends on a critical period early in development, is reversed by centrally administered NE or the beta-2 agonist salbutamol (30). Perhaps most surprisingly, destruction of the DB, apparently, via cortical alpha-1 receptors, prevents the adaptive changes in resulting from damage to the mesolimbicocortical dopamine system, leading to an abolition of both neurochemical changes (in D1 dopamine receptors) and behavioral deficits (hyperactivity, impaired delayed alternation performance) normally associated with such depletion (87). In this context, it is of considerable interest that the disruption by fornix lesions of performance of an 8 arm radial maze task, which is correlated with reductions in cholinergic  hippocampal markers, is completely ameliorated by central NE depletion induced by DSP-4 (74). Thus, again central NE depletion actually benefits behavioral recovery, and this emphasizes the importance of considering interactions and balances with  activity in other neurotransmitter systems.

III.  Theoretical syntheses

      There have been several dominant ideas in theories of LC NE function, based largely on experiments in animals, ranging from notions of reinforcement and arousal (see 70), to the mediation of anxiety and stress (46,68), and the control of  selective attention (61,70). Each theory commands its own set of supporting behavioral phenomena, but it is probably fair to say that no single construct can adequately explain the functions of the LC. While it is natural (though sometimes insightful) to ignore evidence that  does not fit into a particular theory (and we are aware that a brief chapter such as this is unlikely to be free of this tendency), it is desirable that the theory be as precise but as all-embracing as possible. It seems likely, in fact, that the LC has rather general functions which bear on aspects of attention, learning and anxiety. These functions are based on two very clear points. First, activity in  LC neurons is monotonically related to increased arousal. Second, this  activity probably improves the processing of salient events in diverse forebrain sites, whether these events are novel, salient because of conditioning, or even internalized, as representations of stimulus events receiving further processing in memory consolidation and retrieval.

      An early theoretical suggestion (4) was that the LC functioned rather like the cognitive arm of a central sympathetic ganglion,  and this notion has more recently  been taken up by others (10). "Thus activation of the peripheral sympathetic system prepares the animal physically for adaptive phasic responses to urgent stimuli, while parallel activation of LC increases attention and vigilance, preparing the animal cognitively for adaptive responses to such stimuli" (10, p516). The notion of adaptive preparation or coping with the consequences of sympathetic arousal is also to be found in the earlier review (4), and has also been stated by us (69,70) previously in terms such as the DB functions to preserve attentional selectivity especially under elevated levels of arousal (69). This is related to psychological theories such as that of Easterbrook (see 39) who argues that high levels of arousal cause attentional focusing. Another, probably related, formulation is that the LC is  implicated in "controlled" or "effortful", as distinct from "automatic", processing (25); that is, the system modulates attentional capacity.  Presumably, LC activity would be part of that mechanism that effectively focuses attention onto salient events in threatening or demanding  situations.  This theory explains why the LC is more implicated in acquisition than performance of learning tasks, why  aversive situations are more sensitive to manipulations of the LC, why attentional function is disrupted under certain conditions in DB lesioned rats, and why the LC is active during orienting responses to novel stimuli. According to this view, the LC does not mediate anxiety or stress per se, but rather a state of arousal that is correlated with anxiety  or stress, leading to attentional and cognitive change. This state  of arousal  can be self-regulated to a point, and there is evidence to implicate LC neurons in some of the phenomena of "learned helplessness". Specifically, central NE activity is affected by environmental contingencies such as the availability of effective avoidance or "coping" responses; in situations where the animal is exposed to inescapable, uncontrollable  stress, NE function is depressed (82, see chapter by Valentino and Aston-Jones in this volume for review), but can be restored by treatment with adrenoceptor agents infused in the vicinity of the LC, leading to attenuation of those responses characteristic of "behavioral despair" (82).  From a consideration of the cognitive sequelae of manipulations of central NE reviewed in this chapter, it is obvious that some of the major cognitive features not only of learned helplessness, but also depression, including problems in attention and learning, could result from LC dysfunction. 

      Several of the effects of DB lesions (e.g. reduced food neophobia and conditioned suppression) would be expected if such depletion had anxiolytic effects and the role for NE mechanisms of the  LC in mediating certain elements of the opiate withdrawal syndrome (88) is also suggestive of a role in aversive motivation. However,  the limited nature of the conditioned suppression deficits, in particular the apparent enhancement of contextual conditioning, the lack of effect on punished responding and the different effects of chlordiazepoxide and DB lesions (23,54) in the food neophobia setting, are all inconsistent with a simple form of the anxiety hypothesis. On the other hand, it is clear from  studies with humans that anxiety often leads to enhanced distractibility rather than enhanced focusing, and that there is evidence from several sources (5,10,74,78) that elevated LC activity, far from improving performance, may actually impair it. A parsimonious account would then invoke the inverted U shaped function relating arousal level to efficiency of performance (see 39,  for review ). The decrements in performance at high levels of arousal have been  often been attributed to the attention the subject begins to pay to visceral cues arising from sympathetic arousal, such as palpitation, which may become correlated  with  subjective feelings of anxiety (39).

      In contrast, low levels of coeruleo-cortical noradrenergic activity, such as those produced by low doses of clonidine, can lead to lapses in attention that are reversed by environmental stressors such as background noise (84). In certain circumstances, the task itself may theoretically induce arousal that helps to counteract the reduced arousal associated with low NE activity. One recent study has shown a correlate of this in terms of clonidine-induced reductions in basal regional cerebral blood flow being cancelled out by the requirement to perform an attentional task (29). Such a result obviously adds another dimension to the view that the coeruleocortical NE system plays a role in behavioral vigilance (10).  While many might consider that the invocation of a traditional notion such as level of arousal to summarize the range of effects associated with different degrees of noradenergic activation to be outmoded, it is quite difficult to find a more parsimonious account. What has emerged however from more contemporary analyses is that the underlying neural mechanisms are becoming clearer; for example through analyses of regional changes in blood flow, and in the involvement of different populations of adrenoceptors that regulate the effects of altered DB activity in defined ways. 

      Some of the experimental work on behavioral functions of the LC in animals appears to fit with human psychopharmacological experiments on cognitive effects of adrenoceptor agonists and antagonists. Thus, Clark et al (19) found that  the mixed alpha 1/2 agonist clonidine impaired performance in a set of dichotic listening tasks (both divided and focused attention), presumably via its presynaptic actions to reduce LC firing. They  suggested that  the drug reduced alertness or arousal,  and so increased the demands of the tasks on the volunteers. In contrast, a dopamine receptor blocker, while also impairing performance, reduced the activational state of the subjects, or their readiness for action. Clark et al (19), following a similar suggestion (69), make the useful distinction between the dual roles of the LC NE system and central dopamine pathways; the former being concerned with regulating the capacity for conscious registration of external stimuli, whereas the latter regulates the capacity to respond to them.

      Other effects of clonidine are compatible with the evidence concerning effects of DB lesions in animals. First, clonidine has been shown to impair the learning of difficult paired associates (43). Second, the drug apparently reduces the cost  (in reaction time) of shifting attention to a spatial cue (20); this is perhaps analogous to the changes in attentional distribution produced by DB lesions. On the other hand, a recent study of a similar paradigm using two rhesus monkeys found very different effects of clonidine and gunafacine, mainly to reduce the beneficial ‘alerting’ effects of a spatially uninformative cue (92). The latter authors attribute the discrepancy to several possible factors, including the method used to compute the cost of spatial shifting.

       Other studies have shown that clonidine produces significant decrements in a ‘rapid visual information processing’ test of sustained attention (28). These effects were selective in several ways: decrements were much more apparent than in a test of self-ordered spatial working memory, they contrasted with those produced by diazepam, and unusually they were more evident after the subjects had some previous experience with the task on an earlier session under placebo (27,28). Through the study of changes in regional cerebral blood flow using PET, this has recently been attributed to task-related arousal which serves to antagonize the de-arousing effects of clonidine more effectively when the task is novel (29). This study also showed a significant drug X task interaction in the right thalamus. We can expect many more sophisticated analyses of this type to pin-point central loci of systemic  adrenoceptor agents on cognition in the future. It is of interest that a study of the alpha-2 antagonist, idazoxan, actually  produced what appears to be the complementary effect to that of clonidine in psychological terms, of increased attention to the location of the previous cue (85).

      Such results are certainly have clinical potential (see also 5). One of the more remarkable examples of "cognitive enhancement" by psychotropic drugs has been the improvements in cognitive performance in Korsakoff patients by clonidine (60).  While this result is not readily predicted by the effects reviewed here on normal volunteers, it is possible that the presynaptic degeneration of NE neurons may  enhance the contribution of post-synaptic effects of the drug, in a way possibly reminiscent of the effects of clonidine in aged monkeys reviewed above. These results hold some promise for the treatment of other neurodegenerative disorders associated with ceruleo-cortical NE loss, including Parkinson's and Alzheimer's diseases. The former is associated with significant cognitive deficits, often resembling effects of frontal lobe damage, but the possible role of the LC and therapeutic possibilities of treatment with adrenoceptor agents remain largely unexplored (though see 13). While the cognitive syndrome in Alzheimer's disease is probably multifactorial, it is possible that a malfunctioning ceruleo-cortical system may make patients more susceptible to the abrupt decline in cognitive status often associated with transfer to a new environment such as a nursing institution. 

      A major problem for further work in this area is to decide on the optimal method for enhancing noradrenergic function  using drugs such as clonidine, guanfacine or idazoxan, which exert a balance of effects at both pre- and post-synaptic receptors and also have additional effects on other systems (e.g. via imidazoline or 5-HT receptors).  Experimental studies may help to resolve this potentially important clinical issue. For example,  systemic idazoxan has been shown to increase extracellular NA in the rat prefrontal cortex, a finding that might encourage the use of this drug to enhance potentially enhance impaired cortical function via boosting noradrenergic transmission in patients with dementia.  In one such study, specifically of three patients diagnosed with dementia of the fronto-temporal type, improvements were observed following systemic idazoxan in a number of tests (26), some of which are sensitive frontal lobe dysfunction, as might have been predicted from the opposite effects of clonidine in normal volunteers (27,28). However, performance on a test of spatial working memory was impaired by this treatment, showing that executive functions of the frontal lobe respond differentially to presumed noradrenergic activation . The result is however consistent with the improvement seen at certain doses of clonidine in spatial working memory tasks in both monkeys (7) and humans (27). Moreover, patients with dementia of the Alzheimer type showed only deficits with similar idazoxan treatment (45).

      In a more general sense, it is likely that  possible malfunctioning of the LC is important in those human disorders in which there is an important interface between cognition and emotion, including depression, post-traumatic stress disorder, anxiety and drug withdrawal states.  In some of these conditions, it is possible that the LC is overactive, and , as we have seen , this may also result in cognitive problems, which may exacerbate the emotional state. For example, it is plausible in panic disorder that noradrenergic overactivity  helps to concentrate attention on dominant events or pathological cognitive schemas, promoting their consolidation and thus slowing the extinction of their influence over the subject's behavior. Some striking evidence in favour of this hypothesis, which again brings out the utility of the earlier animal work  on the neuromodulation of aversive memory by  mechanisms in the amygdala, has been the demonstration that the beta-1/2 receptor antagonist propanolol selectively impairs memory for emotional material in human subjects (16,90), findings of potential clinical utility in the treatment of post-traumatic stress disorder. A range of other disorders, including cognitive aging, Korsakoff’s syndrome, depression, and certain forms of Attention Deficit Hyperactivity Disorder may suffer, alternatively, from an underactivation of coeruleo-cortical NE activity, requiring a different pharmacotherapeutic  strategy that also targets the noradrenergic system (see above and ref.5 for a fuller discussion).

       We commented in the last version of this review in 1995 (page 370) that   “many of these hypotheses will be tested in the clinical setting in the next generation of research, when its heuristic value is expected to become apparent”. We believe this statement to be ever more appropriate.

Acknowledgements

Our own research is supported by the Wellcome Trust and The Medical Research Council.  We thank our colleagues for their efforts and Dr. B.J. Cole for kindly  providing Figure 1.

published 2000