Functional Heterogeneity of Central Cholinergic Systems

Peter B. Reiner and H. Christian Fibiger

Cholinergic neurons are widely distributed throughout the mammalian central nervous system, and they exist as both projection neurons and interneurons. Amongst the projection neurons, two prominent cell groups have received the lion's share of attention in recent years: basal forebrain cholinergic neurons (located in the medial septum, vertical and horizontal limbs of the diagonal band of Broca, and the nucleus basalis) and brainstem cholinergic neurons (found in the laterodorsal and pedunculopontine tegmental nuclei). Basal forebrain cholinergic neurons innervate the cerebral cortex, while brainstem cholinergic neurons primarily innervate the thalamus (see Structure and Function of Colonergic Pathways in the Cerebral Cortex, Limbic System, Basal Ganglia, and Thalamus of the Human Brain). Within the basal forebrain cholinergic system, subdivisions exist: Cholinergic neurons in the medial septum/vertical limb project primarily to the hippocampus, while the axon terminals of cholinergic neurons found in the horizontal limb and nucleus basalis are directed primarily towards the neocortex. As will be argued below, it is beginning to appear that these anatomical distinctions may have considerable functional relevance. In particular, there is growing evidence that anatomically distinct cholinergic neurons subserve distinct functions.

Based upon the continuous distribution of the "cholinergic basal nucleus complex" (56) and a continuous caudal-to-rostral gradient of development (57), earlier formulations suggested that basal forebrain cholinergic neurons might reasonably be viewed as a single functional unit. Histochemical data now argue against such a view. While most, if not all, basal forebrain cholinergic neurons express nerve growth factor (NGF) receptors (4); see also Neuronal Growth and Differentiation Factors and Synaptic Placticity), substance P receptors (23), and estrogen receptors (67); see also The Psychopharmacology of Sexual Behavior), there appears to be considerable heterogeneity with respect to colocalized neuroactive agents. In the rat, for example, only those basal forebrain cholinergic neurons localized within the medial septum and vertical limb of the diagonal band (which project to the hippocampus) express the enzyme nitric oxide synthase (25, 44) and the neuropeptide galanin (37). Thus, it is clear that not all basal forebrain cholinergic neurons exhibit the same phenotype and that cotransmitter status is correlated with terminal fields. The functional significance of this phenomenon is not yet clear, there are marked species differences in expression of several of these phenotypic markers (25, 35, 71). Nonetheless, these data demonstrate that, at least in the rat, there are chemical differences between subpopulations of basal forebrain cholinergic neurons. To date, brainstem cholinergic neurons appear relatively uniform, but this question has not been examined in sufficient detail. These observations have important implications for pharmacological manipulation of central cholinergic neurons, for therapies based upon transplantation of cholinergic neurons, and for behavioral formulations of central cholinergic function.

One way of understanding the functional role of central cholinergic neurons is to study the activity of identified cholinergic neurons in behaving animals during the execution of complex behaviors. To date, such studies have not been technically feasible. Rather, recordings have been obtained from neurons in regions that contain dense but not pure populations of cholinergic neurons, with or without the additional criterion of antidromic invasion from known postsynaptic targets. While data obtained with this approach are interesting, because it has not been possible to be certain that recordings were indeed obtained from cholinergic neurons, their precise relationship to central cholinergic function remains uncertain.

Several studies have examined state-related changes in firing rate of basal forebrain neurons in rats and cats. The results have been as diverse as the neurons from which the recordings were obtained: Some neurons increase their firing during waking, whereas others exhibit increased activity during slow-wave sleep (13, 65). In the primate, nucleus basalis neurons consistently increase their firing rate in response to reinforcing stimuli, but this cannot be construed as a "signature" of cholinergic neurons because cells in noncholinergic regions of the forebrain respond similarly (48, 53). Nonetheless, the increase in firing rate in the region of the nucleus basalis can be dissociated from both the sensory and motoric aspects of the task (50), 73). At present, it remains unclear to what extent such responses reflect a change in arousal or other complex contingencies associated with the rewarding task (49, 74).

An enduring problem, of course, is the absence of definitive phenotypic identification of these neurons as cholinergic, an issue of critical import. This obstacle has been overcome in studies of basal forebrain cholinergic neurons in brain slices, where the combination of intracellular labeling and choline acetyltransferase (ChAT) immunohistochemistry has permitted unambiguous identification of the intrinsic properties of cholinergic neurons. In nucleus basalis, identified basal forebrain cholinergic neurons have intrinsic ionic conductances which endow them with the capability of generating bursts of action potentials (32). Surprisingly, such bursting behavior is much more difficult to evoke in identified cholinergic neurons in the medial septum (Gorelova and Reiner, unpublished observations), where bursts of action potentials are commonly found "pacing" the theta rhythm (63). At present, it is difficult to place these in vitro studies into a behavioral context. However, it is of considerable interest that similar regional differences exist in both physiological and anatomical phenotypes (see above), further supporting the notion that functional subdivisions exist within the cholinergic basal forebrain.

Mesopontine cholinergic neurons have also been intensively studied in recent years. As with basal forebrain neurons, the state-related discharge of brainstem neurons in regions containing dense accumulations of cholinergic neurons is heterogeneous; at least some neurons exhibit discharge properties consistent with a role in electroencephalographic (EEG) desynchrony (see below) as well as in synchrony with ponto-geniculo-occipital (PGO) waves during rapid eye movement (REM) sleep (61). Once again, the heterogeneous nature of the nuclei in question and the absence of definitive identification of the cholinergic phenotype in such neurons has hindered the generation of strong hypotheses of brainstem cholinergic function. However, studies of identified brainstem cholinergic neurons in brain slices have produced results with clear behavioral implications. At least in the rat, it is the cholinergic neurons of the brainstem that are capable of generating bursts of action potentials (31), and this provides a plausible cellular mechanism for the generation of PGO waves (34).

During the waking state, the neocortical electroencephalogram (EEG) is dominated by a low-voltage, highfrequency pattern known as EEG desynchrony. It is well established that systemic administration of muscarinic antagonists blocks neocortical EEG desynchrony during most waking behaviors (33, 68), 70). The implication is that central cholinergic mechanisms are involved in control of the EEG, in particular in the generation of EEG desynchrony. A key question relates to the locus of the antimuscarinic effect upon the cortical EEG. The two most likely candidate cholinergic systems are those arising in the brainstem mesopontine tegmentum and the basal forebrain. The notion that brainstem cholinergic systems are involved in EEG desynchrony has a rich history. The most effective sites for electrically evoked EEG desynchrony in the cat brain (39) coincide with regions containing dense accumulations of cholinergic neurons (29, 47). Retrograde tracing studies have confirmed the early conjecture of Shute and Lewis (58) that there exists a prominent cholinergic pathway from the mesopontine tegmentum to the thalamus (28, 55). This is an important point, because it has been argued that the membrane potential of thalamic neurons is central to neuronal control of the cortical EEG, with depolarization associated with desynchrony and hyperpolarization associated with synchrony of the EEG (60). Consistent with this notion, acetylcholine depolarizes thalamic relay neurons in brain slices (36), as does stimulation of the mesopontine tegmentum in vivo (43). In unanesthetized animals, at least some mesopontine neurons fire at higher rates during EEG desynchronized states than during slow-wave sleep (20, 59). Taken together, it is plausible that activation of brainstem cholinergic neurons may be involved in EEG desynchrony.

However, not all data are congruent with this hypothesis, and in recent years an ever-increasing body of evidence has accumulated suggesting that at least some aspects of EEG desynchrony are due to release of acetylcholine from the axon terminals of basal forebrain cholinergic neurons. Acetylcholine release from the cerebral cortex is highly correlated with EEG desynchrony (8), and this finding has been supported by recent studies using in vivo microdialysis (see below). Because the predominant cholinergic innervation of the cerebral cortex derives from the basal forebrain, these observations alone suggest that basal forebrain cholinergic neurons may be central to the phenomenon. A key observation derives from studies using excitotoxic lesions. While extensive damage of the mesopontine tegmentum has only minimal effects upon the cortical EEG (72), similar lesions of the basal forebrain essentially abolish atropine-sensitive EEG desynchrony (62). When such excitotoxic lesions are confined to one side of the brain, the contralateral EEG behaves normally while the ipsilateral EEG exhibits a marked increase in slow delta waves (7), and normal EEG desynchrony can be largely restored by cholinergic agonists (69). Taken together, these data argue that atropine-sensitive EEG desynchrony is due to an increase in the release of acetylcholine in the cerebral cortex from the axon terminals of basal forebrain cholinergic neurons.

There is substantial pharmacological evidence that central cholinergic neurons are important in the acquisition and post-acquisition performance of a variety of learned behaviors (see refs. 21 and 27 for reviews). Many studies have demonstrated that antimuscarinic agents such as scopolamine and atropine have deleterious affects on such behaviors. Similarly, compounds (such as physostigmine) that enhance central cholinergic tone by inhibiting the catabolic enzyme acetylcholinesterase (AChE) can, under certain circumstances, enhance performance in learning and memory tasks. In addition, at appropriate doses a variety of muscarinic receptor agonists can enhance performance on tests of learning and memory. This body of pharmacological research has provided strong evidence that unspecified cholinergic systems in the brain play important roles in the acquisition and performance of learned tasks.

At present, there is no consensus about the psychological mechanisms underlying antimuscarinic-induced deficits. Disruptions of behavioral inhibition, working (short term) memory, retrieval from reference (long term) memory, attention, decisional processes, movement and strategy selection, and altered sensory processing are among the variables that have been proposed as mediating the disruptive effects of muscarinic receptor blockade (21). As has been indicated previously (22), hypotheses concerning the nature of the psychological mechanism that underlies antimuscarinic-induced deficits in the acquisition and performance of learned behaviors are flawed because they have occurred in the absence of neuroanatomical realities. We now know, for example, that cholinergic neurons innervate virtually the entire neuraxis and that muscarinic receptors are distributed throughout the central nervous system. This being the case, it is virtually certain that cholinergic mechanisms are involved in a disparate variety of central nervous system functions and that antimuscarinic-induced deficits are multifactorially determined. Viewed in this light, it is quite conceivable that systemically administered scopolamine disrupts the acquisition and performance of a learned behavior via simultaneous actions on, for example, attention (e.g., cerebral cortex?), working memory (e.g., hippocampus?), and sensory gating (e.g., thalamus?). In addition, given that scopolamine is a competitive antagonist, the net level of functional blockade in each structure will be determined in part by the level of ACh release in that structure, which can in turn be influenced by a variety of factors. As a result, the extent to which each of the above-mentioned mechanisms contributes to the impaired performance may be determined by variables such as level of training, motoric demands, drug dose, genetics, state of arousal, and so on (see ref. 22).

Questions regarding the behavioral and psychological correlates of cholinergic activity in discrete regions of the central nervous system are of considerable importance. Unfortunately, for the reasons outlined above, these questions cannot be addressed in experiments using systemically administered cholinergic agents. However, local intracerebral administration of such compounds are proving to be a useful alternative in this regard (5, 6). In a particularly interesting application of this strategy, Dunnett et al. (17) found that intrahippocampal injections of scopolamine produced delay-dependent impairments in the performance of a "delayed, non-matching to position" task in rats. Such delay-dependent effects are consistent with the hypothesis that the drug produces deficits in short-term memory when injected into the hippocampus, and they agree with data from many other sources that implicate the hippocampus in short-term memory functions (38, 42). In contrast to its hippocampal effects, when Dunnett et al. (17)) injected scopolamine into the prefrontal cortex it produced dose-dependent but delayindependent deficits in performance, suggesting a nonmnemonic, possibly attentional, basis to this disturbance.

Although various aspects had been considered earlier (14, 15), the formal presentation of the "cholinergic hypothesis of geriatric memory dysfunction" was proposed by Bartus et al. in 1982 (3). The two central notions of the hypothesis were that (a) forebrain cholinergic systems provide an essential substrate for a variety of cognitive processes, particularly those involved in learning and memory, and (b) the learning and memory deficits of aging are attributable, at least in part, to a decline in the functional integrity of those forebrain cholinergic systems. In its original formulation, this hypothesis was considered in the context of normal aging. However, several lines of evidence, including the loss of cortical cholinergic markers in Alzheimer's disease (AD), the discovery of a correlation between these biochemical measures and mental test scores in AD patients, and the loss/atrophy of the basal forebrain neurons themselves, led Coyle et al. (9) to extend the cholinergic hypothesis to Alzheimer's dementia—that is, to suggest that the more profound memory deficits of AD may also be attributable to extensive degeneration of the same forebrain cholinergic systems. However, many other neuroanatomical and neurochemical systems also degenerate in AD, so that it is extremely difficult to establish a causal relationship with the cholinergic decline specifically. With this issue in mind, the formulation of the cholinergic hypothesis in 1982 stimulated a large number of studies in the subsequent decade which sought to investigate whether the explicit destruction of magnocellular cholinergic neurons in the basal forebrain by axon-sparing excitotoxins would produce a profile of deficits in experimental animals similar to the changes observed in the aged animal or comparable to more profound changes in learning and memory capacity that is such a distinctive feature of human dementia. The expectation was that reproduction of comparable deficits by an explicit and selective experimental intervention would provide direct evidence in favor of the cholinergic systems having a truly causal role in their genesis. Unfortunately, the early enthusiasm that accompanied the introduction of this strategy has been tempered by subsequent findings that have identified a number of its shortcomings. These issues have been discussed extensively elsewhere and are therefore not reviewed in detail here (18, 19, 22). Suffice it to say that the most serious limitation of the lesion strategy derives from the fact that wherever they occur in the basal forebrain, cholinergic neurons are intermingled with populations of noncholinergic cells. This being the case, it is uncertain that the deficits in behavior that are produced by excitotoxin lesions of the basal forebrain are due specifically to the loss of cholinergic neurons. What is needed for lesion-based strategies to contribute definitive information concerning the functions of central cholinergic systems is a toxin that is selective for cholinergic neurons. Unfortunately, at present no such compound exists. Despite the above-mentioned limitation of excitotoxin-lesion-based strategies, it is increasingly clear that animals with extensive lesions of the cholinergic neurons that form the nucleus basalis magnocellularis (nBM) (see Structure and Function of Colonergic Pathways in the Cerebral Cortex, Limbic System, Basal Ganglia, and Thalamus of the Human Brain) can perform normally in a variety of learning and memory tests (18, 22). This is particularly evident in studies that have used quisqualic acid or a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) to lesion the basal forebrain in the region of the nBM. On the basis of such evidence, Dunnett et al. (18) have proposed that the mnemonic deficits described in the earlier studies that utilized ibotenic acid lesions of the basal forebrain are not attributable to destruction of cholinergic neurons but occur as a result of damage to corticostriatal output systems that course through the globus pallidus. If this is correct, what then are the functional consequences of damage to the telencephalic projections of the nBM? Recent data suggest that impaired attentional function may be one important consequence. One of the first studies to demonstrate this was conducted by Robbins et al. (52), who used a 5-choice serial reaction time task to show that quisqualic acid lesions in the region of the nBM produced deficits in visual attentional function in rats. Subsequently this group of investigators has demonstrated that AMPA-induced lesions of the nBM produce similar impairments in this test of visual attention and that these deficits can be ameliorated by low doses of the AChE inhibitor physostigmine (40). While reversibility by procholinergic drugs does not prove that a lesion-induced deficit is due to a damaged cholinergic system (22), this is clearly the most straightforward explanation for such a finding. In addition, and as will be seen below, data from a variety of other sources are compatible with the hypothesis that cortical projections of the cholinergic basal nuclear complex are neural substrates for some attentional functions.

It has been known for nearly 30 years that acetylcholine (ACh) release in the cortex increases markedly during EEG desynchronization (8, 30). This is consistent with data from other approaches showing that the activity of neurons in the region of the nBM is increased during EEG desynchrony (7, 13) and that lesions of the nBM produce EEG slowing (7, 51). Day et al. (10) have shown that behavioral measures of arousal such as locomotor activity also correlate positively with cortical ACh release. In the context of the attentional hypothesis of cortical cholinergic function, recent data obtained with brain microdialysis have provided significant support. Specifically, Day and Fibiger (11) have shown that d-amphetamine potently increases ACh release in the cortex of awake, behaving animals. Methylphenidate has similar actions (Day and Fibiger, unpublished observations). Inasmuch as these compounds are known to improve attention in humans (41, 64) and are the treatment of choice in attention deficit disorder, their positive actions on cortical ACh release are entirely consistent with this hypothesis. Subsequent pharmacological analysis has shown that stimulation of D1 dopamine receptors is the primary mechanism through which d-amphetamine produces these effects and raises the possibility that D1 receptor agonists may have therapeutic applications in the treatment attention deficit disorders (11, 12). The fact that scopolamine impairs performance on some attentional tasks in humans, particularly those that require active allocation of attentional capacity, is also consistent with the attentional hypothesis (16, 46). Similarly, Sahakian et al. (54) recently demonstrated that tetrahydroaminoacridine (THA), which among other actions is an AChE inhibitor, improves performance on certain tests of attentional function in patients with mild to moderate Alzheimer's disease. It is noteworthy that this occurred in the absence of significant effects on tests of mnemonic function. While these results suggest that central cholinergic mechanisms are involved in the regulation of attentional processes, they do not of course provide any information about the anatomical locus of such effects. At present, the only data pointing to a role for the nBM-cortical projection are the above-mentioned excitotoxin lesion studies in rodents, and as indicated earlier the interpretation of these results is far from straightforward. In summary, while existing evidence is compatible with a role for this cholinergic projection in attention, definitive evidence regarding the validity of this hypothesis awaits the results of future research.

As indicated above, the functions of the telencephalic projections of the cholinergic basal nuclear complex are not yet firmly established in animals. Similarly, the behavioral consequences of the degeneration of this complex in Alzheimer's disease remain unknown (see also Biological Markers in Alzheimer’s Disease, Experimental Therapeutics, and Cognitive Impairment in Geriatric Schizophrenic Patients: Clinical and Postmortem Characterization). Despite this, over the past decade many drug discovery programs have been based on the assumption that the cholinergic hypothesis of Alzheimer's dementia will eventually be validated and that pharmacological restoration of central cholinergic tone will therefore be of significant therapeutic value in the treatment of this condition. There are, however, a number of reasons to be skeptical about this line of reasoning. Perhaps the most important is that this strategy assumes that postsynaptic targets of degenerating cholinergic terminals remain relatively intact in Alzheimer's disease. While postsynaptic muscarinic receptors are generally considered to be unaffected in the hippocampus and cortex of Alzheimer's patients (1), there is abundant evidence that these structures undergo marked degeneration during the course of the disease (2, 45, 66). Indeed, it is possible that the primary pathological processes in Alzheimer's disease occur in these telencephalic structures and that the damage to the cholinergic neurons in the basal forebrain is secondary and represents retrograde degeneration (24)). Given this, the question arises as to whether pharmacological enhancement of cholinergic transmission in these cytoarchitecturally damaged target structures would be expected to reduce the cognitive deficits in Alzheimer's disease. Unless the loss of basal forebrain neurons in Alzheimer's disease is the earliest degenerative event and unless there is a significant period during which the hippocampus and cortex remain relatively intact, hopes for successful cholinomimetic replacement therapies are poorly founded. In addition, the fact that the neuropathology of Alzheimer's disease is increasingly being understood to involve many noncholinergic, chemically defined systems (26) adds to the concern that procholinergic drugs may not benefit Alzheimer's patients. In this context, it is perhaps understandable that this pharmacological strategy has not yet produced clinically significant improvements in these patients. This is not to say that the development of procholinergic drugs may not eventually find important applications in other patient populations. As discussed above, there is evidence that central cholinergic systems may be important neural substrates for attention, an important component of the larger process termed cognition. This being the case, cholinomimetic drugs may prove to be useful in the treatment of attention deficit disorder. In addition, they may also find applications in the treatment of more mildly impaired, more neurologically intact geriatric individuals such as those suffering from age-related memory loss. A better understanding of the normal functions of central cholinergic systems will be a key step in exploring these possibilities.

This work was supported by grants from the Medical Research Council of Canada, the Alzheimer's Society of British Columbia (PBR), and an unrestricted grant from Bristol-Myers Squibb (HCF).

published 2000