Attempts to transplant neural tissue into the brain go back to the last century (Thompson 1890). However, the modern period of intense investigation of such grafting has a very short history, starting with the pioneering experiments of Stenevi, Bjrklund and Svengaard (1976), Lund and Hauschka (1976), and Perlow, Freed, Hoffer, Seiger, Olson and Wyatt (1979). This history has been unusual, in that it has involved a close partnership between neuroanatomists on the one hand and neurologists and psychologists on the other, with most of the intermediate neuroscientists (physiologists, neurochemists, etc) rather lagging behind. In consequence it rapidly became clear that neural transplants are capable of restoring behavioural function, but it remained unclear (and still does) exactly how this effect is achieved. The functional success of neural transplantion soon gave rise to hopes that grafting of neural tissue might eventually provide a new therapeutic strategem for the treatment of human brain damage; and, indeed, there was a remarkably short time -- some would say, much too short -- before attempts were made to realise these hopes in the clinic. The most rapid advances in this progress from the laboratory to the clinic have been made in respect of the ascending nigrostriatal dopaminergic system, motor function and Parkinson's disease (a disorder of motor function consequent upon degeneration in the nigrostriatal pathway); and there is preliminary evidence that this condition will indeed respond to fetal dopaminergic-rich transplants into the striatum (Lindvall et al. 1992).
It is the purpose of the present paper to consider the prospects for similar progress in the case of cognitive dysfunction consequent upon damage to the brain. We shall consider two specific possibilities: (1) damage to the ascending forebrain cholinergic projection system (FCPS), like that known to occur in several forms of dementia, including senile dementia of the Alzheimer type and alcoholic dementia; and (2) damage to the hippocampal formation, like that observed after heart attack or cerebral artery occlusion, often accompanied by memory disorder. Like the nigrostriatal pathway, the FCPS is a diffuse projection system, in a manner that will be defined below; the damage that follows cerebral artery occlusion, in contrast, takes out a link in a hippocampal point-to-point neuronal circuit. Thus, consideration of both these cases will give an indication of the potential for the application of neural transplant therapy to widely differing types of brain damage, and so of the overall potential scope of this new therapy. We have recently reviewed the literature in both these fields (Sinden, Gray & Hodges, in press, for the cholinergic system; Hodges, Sinden, Meldrum & Gray, in press, for ischaemic brain damage); detailed references, if not provided below, can be found in those articles.
2. The forebrain cholinergic projection system
Acetylcholine (ACh) was the first compound in the mammalian central nervous system (CNS) to satisfy most of the pharmacological criteria for neurotransmitter status (Eccles et al. 1956). More recently, the development of immunohistochemical and in situ hybridisation techniques has enabled the unequivocal identification of CNS cholinergic projection systems, by for example the use of antibodies to the enzyme responsible for ACh synthesis, cholineacetyltransferase (ChAT). These systems have been the subject of particular interest since the observation, made at much the same time that modern research on neural transplantation began, that in Alzheimer's disease there is a loss of ChAT in the terminal areas (the neocortex, hippocampal formation and amygdala) served by the FCPS (Bowen et al. 1976; Perry et al. 1977), as well as loss of neuronal cell bodies in the nuclei of origin of the FCPS: i.e., the nucleus basalis of Meynert (or, in the rat, nucleus basalis magnocellularis), projecting to the neocortex; and the medial septal area and nucleus of the diagonal band of Broca, projecting to the hippocampal formation. Though there is, in addition, extensive damage of many other kinds in the Alzheimer brain, the loss of these cholinergic markers provides the best post mortem correlate so far available of the extent of dementia measured in vivo (Whitehouse et al. 1982). Furthermore, there is evidence of a similar loss of neurones in the FCPS in a number of other conditions involving dementia or other loss of cognitive function, including Korsakoff's psychosis, dementia pugilistica and Parkinson's disease (Arendt et al. 1983). These and a variety of other observations, made with a diversity of techniques (Morris & Kopelman, 1986), have given rise to the hypothesis that damage to the FCPS plays a key role in human cognitive dysfunction in a number of neurodegenerative diseases.
Anatomically, the disposition of cholinergic nuclei in the mammalian C.N.S. is suggestive of a coordinated, modulatory neural system. In rat brain, cholinergic neurones form continuous columns of cells extending rostrally from the corpus striatum (including the nucleus accumbens, olfactory tubercle, islands of Calleja and the caudate-putamen) through the basal forebrain (including the medial septal nucleus, the vertical and horizontal limbs of the nucleus of the diagonal band of Broca, the nucleus basalis, substantia innominata and the nucleus ansa reticularis). Caudally, these columns extend through basal diencephalic regions (including the posterior hypothalamus and motor neurones of cranial nerves 3 and 4) into the pedunculopontine and laterodorsal tegmental nuclei of the midbrain. Seen from this global point of view, the entire cholinergic system appears to be linked into a "unified complex of contiguous subsystems" (Woolf 1991). This contiguous nature of cholinergic cell bodies throughout the mammalian CNS is manifest in the confluence and overlap of each subsystem's dendritic plexus (Woolf 1991). Dendritic arbours from ChAT-positive neurones intermingle, not only within, but also between traditional architectonic boundaries.
The major projection pathways of the FCPS to the cortical mantle can be divided into three branches. A rostral branch originates in the medial septal nucleus and the vertical limb of the diagonal band of Broca (collectively abbreviated here as MSA). ChAT-positive fibres enter the cingulum bundle and travel caudally in the fornix to innervate the hippocampal formation and the entorhinal and perirhinal cortices. In the hippocampal formation, ChAT-positive afferents display a laminar terminal organisation. Fibres are particularly abundant in the hilus and the supragranular layer of the dentate gyrus; they are also densely located in the supra- and infra-pyramidal layers throughout the hippocampus proper and in the superficial stratum lacunosum moleculare of area CA1. The targets of cholinergic neurones have been studied more intensively in the hippocampal formation than anywhere else in the brain. Frotscher and Naumann (1992, p. 238) have recently summarised these targets as follows.
"Cholinergic neurons are generally regarded as a system of diffusely projecting cells. This is reflected by the variety of postsynaptic partners of cholinergic terminals. Thus, cholinergic synapses are found on cell bodies, dendritic shafts, spine heads, and spine necks of hippocampal and dentate neurons. The majority of cholinergic synapses are, however, found on dendritic shafts. Interestingly, cholinergic boutons may form symmetric and asymmetric synapses. Traditionally, symmetric contacts are regarded as inhibitory and asymmetric contacts as excitatory synapses. Finally, as revealed in double-labelling studies combining ChAT immunocytochemistry with Golgi impregnation or with double- immunolabelling for glutamate decarboxylase, cholinergic terminals establish synaptic contacts with all available target neurons in the hippocampus and fascia dentata including pyramidal neurons, granule cells, and GABAergic interneurons. By terminating on a large variety of neurons and on different postsynaptic elements, the septohippocampal cholinergic fibers may exert a diffuse, modulating effect on the target region."
A second, medial pathway arises from parts of the vertical limb of the diagonal band, the horizontal limb of the diagonal band and the magnocellular preoptic area to innervate cingulate and retrosplenial cortices as well as a range of allocortical structures, particularly the olfactory bulb, amygdala and the insular and piriform cortices. This has been the subject of less experimental attention than the other two branches of the FCPS and will not be considered further here.
The third branch of the FCPS comprises a more extensive set of projections, coursing mainly laterally to innervate the entire neocortex. This pathway originates from ChAT-positive somata in the nucleus basalis, substantia innominata and nucleus ansa lenticularis (collectively abbreviated here as NBM). The topography of the NBM projection in the rat follows, for the most part, a rostrocaudal gradient. However, a substantial degree of overlap in both corticofugal and corticopetal projections of cholinergic neurones is found in this subsystem (Saper, 1984). Species variation is also apparent in the degree of topographical specificity of the FCPS. Relative to the rat, cholinergic neurones are more densely packed in the cortically-projecting NBM cell groupings than in the rat of both old- (Mesulam et al., 1983) and new- world (Everitt et al., 1988) monkeys as well as humans (Saper and Chelimsky, 1984). The greater topographic organisation of subdivisions of NBM cholinergic projection neurones in primates (e.g., these have been differentiated into anteromedial, anterolateral, intermediodorsal, intermedioventral and posterior sectors in the human NBM; Saper and Chelimsky, 1984) appears to reflect the higher degree of structural and functional specialisation of the cortical targets of these neurones at the primate level (Ridley and Baker, 1991). As in the hippocampal formation, ChAT-positive innervation of the cortex of the rat is laminar in appearance, with dense terminal and fibre staining in layers 2-3 and 5, where there is also dense cholinergic receptor immunoreactivity, both muscarinic and nicotinic (van der Zee et al. 1992).
Both at a gross anatomical level (the widespread innervation of many brain regions by contiguous clusters of cholinergic cells) and at the level of fine structure (innervation of multiple cell types and synaptic targets), then, the FCPS appears to be organised in a diffuse manner, strikingly different in all species from the type of point-to- point topography which, as we shall see below when we consider the internal circuitry of the hippocampal formation, characterises the synaptic organisation of its target regions.
Evidence for a diffuse mode of action of the FCPS emerges from electrophysiological studies. At the population level there has long been extensive evidence that the cholinergic cells of the MSA control the hippocampal slow rhythms known as theta or rhythmic slow activity (Apostol & Creutzfeldt 1974; for review, see Gray 1982). Similar evidence has recently begun to accumulate that lateral FCPS projections are responsible for the generation of cortical rhythms associated with the level of behavioural arousal (Buzsaki et al. 1988; Riekkinen et al. 1991; Steriade et al. 1990). Since these rhythms can be recorded synchronously from spatially extended regions of the hippocampal formation or neocortex, observations of this kind imply a widely distributed modulatory or synchronising function for the FCPS rather than the transmission of discrete neuronal information. Hypotheses as to the behavioural or cognitive functions discharged by the hippocampal theta rhythm (e.g., a role in pacing the passage of neural information around the hippocampal and related circuits: Gray, 1982; O'Keefe & Nadel 1978) reflect these observations.
A similar inference to a modulatory role for cholinergic transmission is suggested by electrophysiological investigations at the single cell level. Both excitation and inhibition of neuronal firing have been reported after the application of cholinergic agonists. Krnjevik et al. (1971) were the first to demonstrate that ACh, applied by iontophoresis in vivo, increased cortical neuronal activity over long periods, along with an increase in membrane resistance, by means of a reduction in potassium conductance. Since then, a number of patch-clamp studies have identified muscarinic excitation as being due mainly to blockade of a voltage-dependent K+ current and at least three other types of K+ or calcium-activated K+ conductances (McCormick 1992). These K+ conductances normally serve to impede spike generation in response to depolarisation, and the calcium-activated K+ current potentiates the slow after-hyperpolarisation that appears after a train of action potentials. Thus the effects of these muscarinic receptor-mediated membrane changes are to enhance and prolong the neuronal response to depolarising inputs (McCormick & Prince 1986). Inhibition of neuronal activity following iontophoretic application of ACh has also been reported in all targets of the cholinergic system, often taking the form of an initial hyperpolarisation preceding a slow depolarisation. This pattern of change is due to a muscarinic activation of GABAergic interneurones (Haas 1982). The FCPS is thereby able to influence patterns of activity in cortical targets by means of both direct muscarinic excitatory activation of pyramidal neurones and via muscarinic activation of inhibitory GABAergic interneurones (Abdulla et al., in press).
Much remains to be learned about the role of these parallel actions of ACh mediating excitation and inhibition in different hippocampal and cortical target neurones in determining population-determined patterns of activity and their functional significance. At the system level, cholinergic agonists have been reported to produce a variety of electrophysiological effects: activity is suppressed in the hippocampus at Schaffer collateral-CA1 and mossy fibre-CA3 synapses (Valentino & Dingledine, 1981; Williams & Johnston, 1988); activity is blocked in intrinsic fibre laminae of pyriform cortical slices (Hasselmo & Bower, 1992); in contrast, visual input to striate cortex (Sato et al. 1987), somatic input to somatosensory cortex (Tremblay et al. 1990), and afferent input from lateral olfactory tract to piriform cortex (Hasselmo & Bower, 1992) are all facilitated; and, similarly, the response in frontal cortex to a discriminative auditory stimulus paired with reinforcing brain stimulation is facilitated (Pirch et al. 1992). A possible interpretation of this mixture of cholinergic facilitation and GABAergic- mediated suppression is that, overall, there is an enhancement of the signal-to-noise ratio, perhaps providing a neural basis for cortical/hippocampal activation and attention (Buszaki et al. 1988). A more specific role for ACh in fine tuning of functional neuronal activity has recently been suggested by Hasselmo et al. (1992), who propose a model for associative learning and memory of odour patterns in the olfactory cortex. According to this model, learning is enhanced by cholinergic facilitation of afferent synaptic input; an accompanying cholinergic inhibition of intrinsic fibre activity (Hasselmo & Bower 1992) permits the best learning and memory performance by reducing interference from previously stored patterns.
Behavioural studies of the effects of lesions to the FCPS reinforce the diffuse and integrative nature of the cholinergic system acting to coordinate function across variably distributed neural targets. Several relevant experiments have been conducted in our own laboratory. We shall describe these first, and then consider the extent to which our conclusions agree with data from other laboratories.
3. Studies of the FCPS: experimental strategy
Experimental study of the cholinergic system is hampered by the lack of a cholinergic-specific neurotoxin which could discharge the role played by, e.g., 6-hydroxydopamine for catecholaminergic systems (although recent reports suggest that immunotoxins, derived from cholinergic-specific antigens, may have the potential to overcome this problem: Nilsson et al. 1992; Dubovik et al. 1993). This problem is especially acute, since the cholinergic projection cells in the rat are scattered more (MSA) or less (NBM) densely amidst a rich variety of other neuronal systems (Alheid & Heimer 1988). In the absence of such a cholinergic-specific toxin it was necessary simultaneously to employ a number of approaches which, between them, might be capable of triangulating the cholinergic system and separating it out from other systems potentially implicated by each approach on its own. The combination of approaches we adopted consisted of two distinct methods of damaging the FCPS, either by stereotaxic injection of excitotoxins or by prolonged systemic administration of alcohol; cholinergic pharmacological challenge; and the neural transplants (cholinergic or non-cholinergic) themselves. The potential artefacts associated with each of these methods were in each case different. Thus, if all four pointed to the same conclusions, this was likely to reflect their only common feature, that is, the involvement of the cholinergic system. We shall leave discussion of the effects of cholinergic neural transplants until later. The key features of the other three methods (all fully described in the references given below) were as follows.
The injection of excitotoxins (Arendt et al. 1989; Hodges et al. 1991a; 1992; Sinden et al. 1990) targetted the nuclei of origin of the FCPS located in the MSA and NBM. This was achieved by multiple stereotaxically guided injections of small volumes, initially of ibotenate. Excitotoxic lesions can be targetted separately at NBM or MSA to reduce cholinergic markers in frontoparietal cortex and hippocampal formation respectively without overlapping cholinergic depletion (e.g. Hagan et al. 1988; Arendt et al. 1989) or made in combination to produce overlapping cholinergic depletion in the terminals of both branches of the FCPS (Arendt et al. 1989; Hodges et al. 1991a). Like other excitotoxins, ibotenate acts via glutamate receptors to destroy neuronal cell bodies, while sparing non-neuronal cells and, generally (but see Coffey et al. 1988), fibres of passage or terminals. It is not selective, however, with regard to the transmitter used by the cells it attacks. We also employed two other excitotoxins, quisqualate and `-amino-3-hydroxy-4-isoxozole propionic acid (AMPA). These different excitotoxins produce varying amounts of ChAT depletion in neocortex when injected into the region of the NBM, in the order AMPA > quisqualate > ibotenate (Dunnett et al. 1991). In addition, AMPA and quisqualate produce a relatively greater loss of ChAT-like immunoreactive structures in the lesioned NBM, accompanied by relatively less non-ChAT-positive neuronal loss (Page et al. 1991; Robbins et al. 1989b). It has therefore been proposed (Dunnett et al. 1991) that the behavioural effects of NBM lesions made with ibotenate, but not with quisqualate or AMPA, are due to disruption of a number of other neuronal systems that course through the basal forebrain, including corticostriatal, striatopallidal and amygdalofugal pathways (Alheid & Heimer 1988). Our own observations confirm these reports for the NBM. However, when the three excitotoxins are injected into the MSA, we see little if any difference in the extent or type of septal damage or in the degree of hippocampal ChAT depletion (Hodges et al. 1991a; Turner et al. 1992; Bradbury et al. submitted). A particular difficulty is posed by a group of GABAergic neurones distributed within the same region occupied by cholinergic neurones in the MSA and, like these, projecting to the hippocampus, where they synapse on GABAergic interneurones (Amaral & Kurz 1985; Freund & Antal 1988). The extent to which this GABAergic projection, or the similar projection from the NBM to the neocortex (Freund & Meskanaite 1992), is damaged by the different excitotoxins is as yet unknown; but it is virtually certain that such damage occurs with MSA injections.
Given these observations, if one wishes to damage the FCPS, the excitotoxin currently of choice would be AMPA, especially for injection into the NBM. At the time we commenced our experiments this was not known, and we used ibotenate. However, we have replicated all the key radial arm maze results described below using both quisqualate (Turner et al. 1992) and AMPA (Gray et al., in press), with qualitatively similar results to those seen after ibotenate injections. Making comparisons across many experiments using the different excitotoxins in our laboratory, there is evidence for the view (Dunnett et al. 1991) that cholinergic damage is not entirely responsible for radial maze impairments, in that ibotenic acid produced the greatest impairments with the least terminal ChAT loss and the largest amount of cell loss within the NBM and surrounding substantia innominata and globus pallidus (Hodges et al. 1991a). However, a direct comparison of ibotenic and quisqualic acid lesions, which produced equivalent reductions in cortical and hippocampal ChAT activity, also produced quantitatively similar radial maze impairments (Turner et al. 1992).
Alcohol (Arendt et al. 1988a; 1988b; 1989; Hodges et al. 1991c) was administered in the drinking water in a 20% solution by volume for periods up to 6 months. All behavioural experiments were begun at least 4 weeks after the end of the period of alcohol ingestion. Confirming Arendt et al.'s (1983) observations on the brains of Korsakoff psychotics, this alcohol regime substantially reduced the number of cells in both the MSA and NBM, as well as levels of cholinergic markers in the hippocampus and neocortex. The loss of cholinergic markers in these terminal regions was of a magnitude (c. 15- 25% in hippocampus; 30-40% in cortex) at the lower end of the range typically observed after ibotenate lesions of the nuclei of origin of the FCPS. In neither the excitotoxic nor the alcohol case (in the absence of transplant surgery, as discussed below) was any recovery observed in these biochemical parameters for up to a year (unless the period of alcohol administration was limited to 12 weeks or less). Like the excitotoxic lesions, prolonged alcohol administration also produced other, non-cholinergic damage. However, the additional damage after alcohol was radically different. Rather than being localised to the MSA and NBM regions, it occurred in other regions of the forebrain; and, whereas the excitotoxic lesions did not cause any reduction in biochemical markers of aminergic systems, there was a substantial loss of noradrenergic, serotonergic and dopaminergic markers after prolonged alcohol. Thus, the major damage common to the two methods is that to the cholinergic cells of the MSA and NBM.
Our pharmacological challenges (Hodges et al. 1991b; 1991c; 1992; Turner et al. 1992) included agonists and antagonists at both the muscarinic and nicotinic types of cholinergic receptor. These were administered acutely by intra-peritoneal injection after lesioning or alcohol treatment. It was expected that involvement of the cholinergic system in the behavioural effects of the lesions or the alcohol regime would be shown by behavioural super-sensitivity to either agonists or antagonists or both. We were particularly interested in the possibility of nicotinic effects, since one purpose of these studies was to investigate the extent to which damage to the FCPS creates a model of some of the cognitive deficits observed in Alzheimer's disease, and there is known to be a decline in the number of brain nicotinic receptors in this condition (Flynn & Mash 1986; Norberg & Winblad 1986; Perry et al. 1986; Whitehouse et al. 1986).
Our chief behavioural tool (Arendt et al. 1989; Hodges et al. 1991a; 1991b; 1991c; Turner et al. 1992) consisted of Jarrard's (1986) version of Olton's (1983) radial-arm maze. This permits repeated (e.g., daily or weekly) measurement, in the same animals, of both reference (long-term) and working (short-term) memory in both spatial mode (i.e., guided by extra-maze cues in the general laboratory environment) and cue mode (guided by floor inserts that are shuffled between trials with respect to the extra-maze environment). Normal animals achieve almost error-free performance on these tasks, even if they are tested on the same day in both the spatial and cue modes. Jarrard (1986) has shown that this method is capable of detecting double dissociations with lesions of the hippocampal formation, and so of demonstrating effects that are selective for one or other of the four components of memory tested.
4. Results of lesion experiments on the cholinergic system
We anticipated that Jarrard's task might show differences between the effects of MSA and NBM lesions on the four different memory components. In the event, however, virtually all of our experimental manipulations affected all memory components in the same manner (see, e.g., Figure 1, below), suggesting that the FCPS discharges a very general cognitive, perhaps attentional, function (Hodges et al. 1991a; 1991b; 1991c). A general function of this kind would be congruent with the conclusion, drawn above on anatomical and physiological grounds, that the FCPS acts in a diffuse or global manner, rather than by means of point-to-point or specifc input-output synaptic interactions, a distinction that will be elucidated in more detail below. The action of the FCPS could thus be to facilitate and modulate in a relatively distributed fashion the target areas that, through local circuit connections, "make their own contribution to psychological function" (Ridley et al. 1991a). In particular, such a general function could be discharged by influencing the way in which the hippocampal and cortical targets of the FCPS process more discrete types of information reaching them by way of other afferents, e.g., the FCPS may act by increasing the signal-to-noise ratio for such afferents (see above). Given the lack of difference between the results obtained in the different memory components of Jarrard's task, in what follows we shall make no distinctions between them.
A major question we posed in the experiments was this: do the cholinergic projections to the hippocampal formation (from the MSA) and neocortex (from the NBM), respectively, serve the same or different functions? If they serve different functions, we would expect to see different behavioural effects of lesions at each site; if they serve the same functions, we would expect to see similar behavioural effects of each type of lesion and, furthermore, we might expect the effects of combined lesions at both sites to be additive (at least up to a limit imposed by likely ceiling effects). Our results (Arendt et al. 1989; see Figure 1) provide clear support for the latter hypothesis. The impairments in radial- maze performance produced by excitotoxic lesions of the MSA and NBM, respectively, were in both cases substantial and enduring; more importantly, they were virtually identical, both qualitatively and quantitatively; and, when both lesions were combined, the behavioural impairment observed was well approximated by the sum of the impairments produced by each lesion on its own. These results are congruent with the inference, reached above on anatomical and physiological grounds, that the FCPS acts as a unitary system with a widely distributed function. Aspects of the findings suggested that this function might be attentional: well-trained unlesioned rats showed considerable disruption of performance when visual cues were restricted (by dimming the lights), and rats with lesions to the FCPS showed no further disruption under these conditions (Hodges et al. 1991a).
-----------------
Figure 1 about here
-----------------
If the behavioural effects we observed after the combined excitotoxic lesion of the MSA and NBM were indeed due to cholinergic damage, then we should see changes after comparable cholinergic damage produced by other means. As noted above, such damage, measured in terms of cholinergic morphological (Arendt et al. 1988b; Hodges et al. 1991c) and biochemical (Arendt et al. 1989; Hodges et al. 1991c) changes, can be produced by chronic administration of alcohol in the drinking water over a six-month period. The behavioural effects of this regime (Figure 2) were very similar to those of the combined MSA and NBM lesion.
These effects of chronic alcohol, which have been substantially replicated by another group (Casamenti et al. 1993), support Arendt's (1983) and Lishman's (1986) hypothesis that cholinergic damage underlies the cognitive deficits of Korsakoff's psychosis and alcoholic dementia, as well as those of Alzheimer's disease. Taken together with the effects of FCPS excitotoxic lesions, they also strongly suggest that, at least as tested in the radial-arm maze, the two branches of the FCPS obey a kind of principle of mass action: that is, they discharge a unitary cognitive function, in which each plays a role of approximately equal weight; while damage to the FCPS impairs performance to a degree proportional to the extent of damage, whether this is made by excitotoxic lesions or prolonged alcohol administration.
How might such a mechanism of 'mass action' work? As noted above (section 2), forebrain cholinergic projections appear to exert their effects by modulation of transmission in cortical and hippocampal targets. Clearly, the neocortex and hippocampal formation do not themselves discharge identical functions. Why, then, should cholinergic denervation of each of these regions have, apparently, similar and additive behavioural effects? One obvious possibility is that our behavioural measurements are simply too coarse to capture different functions. However, Jarrard's version of the radial maze task has demonstrated dissociations after different lesions, even when these were both within the hippocampal formation (Jarrard, 1986). An alternative explanation, which we favour, is that the psychological process which lesions to the FCPS disrupt, with the observed global impairment in radial maze performance, is one that requires constant interchange of information between (1) cortical structures subserving perception and/or memory and (2) hippocampal processing required for current task performance. A number of theories of hippocampal function have emphasised such cortico- hippocampal interaction (e.g., Hirsh 1974; Gray 1982; Squire et al. 1984; see Eichenbaum et al. in press). The kind of psychological function we have in mind is often termed 'attention', 'working memory' (Baddeley, 1986) -- although the narrow operational definition of this term applied to the radial maze by Olton et al. (1979), is ruled out by our results -- or 'vigilance' (Rusted & Warburton, 1989). Our results, therefore, suggest that the effective interchange of information between cortex and hippocampus that is necessary for this function is equally and additively compromised by cholinergic deafferentation of either region. Such an analysis does not preclude the possibility that, in other tasks dependent upon other processes, cholinergic deafferentation of these regions might produce differential effects.
-----------------
Figure 2 about here
-----------------
The critical role played by cholinergic damage in these observations received further support from the results of our pharmacological tests (Hodges et al. 1991b; 1991c; see Figure 3). These results were qualitatively similar whether the FCPS had been damaged by the combined excitotoxic lesion (MSA + NBM) or by alcohol. The experimental subjects were behaviourally super-sensitive to both cholinergic agonists and antagonists: their performance was improved by agonists and further impaired by antagonists, and both these effects were seen at doses which failed to affect performance in intact controls. These effects were seen with compounds active at both muscarinic (arecoline, scopolamine) and nicotinic (nicotine, mecamylamine) receptors.
As will be reviewed in Section 6 below, the radial maze impairments produced by either excitotoxic MSA + NBM lesions or chronic alcohol were significantly reversed by transplants of fetal basal forebrain rich in cholinergic neurones, but not by transplants of fetal hippocampus poor in such cells, placed in cortex, hippocampus or both regions. Together with evidence showing that the time-course of graft-induced recovery matched histological evidence for cholinergic innervation of their targets and that the volume of cholinergic-rich, but not - poor, grafts correlated positively with recovery, these data provide complementary evidence for the role of the cholinergic system in these cognitive deficits.
---------------------
Figure 3 about here
---------------------
The similarities, observed in these experiments, between the two branches of the FCPS are all the more remarkable given the morphological and physiological differences among the medial septum, diagonal band, nucleus basalis and substantia innominata regions of the basal forebrain. First, whereas only approximately 40-60% of the neurones of the MSA are ChAT- positive, roughly 95% of the cells projecting from the NBM are ChAT positive. Second, the various nuclei comprising the FCPS differ anatomically, neurochemically and physiologically. In addition to projecting to different telencephalic targets, for example, more caudal territories contain cholinergic cells with larger somata and more processes than those in the medial septal nucleus (Butcher and Woolf 1984; Woolf et al. 1983). Also nerve growth factor (NGF) is present in varying amounts within different FCPS subdivisions (Korsching et al. 1985). Finally, many MSA neurones display an irregular pattern of firing, whereas the electrophysiological profile of most NBM neurones is more systematic (Dutar et al. 1986). Evidently, there are some information processes, common to the cholinergic projections to neocortex and hippocampus, respectively, that override these morphological and physiological differences. We have speculated above as to what these processes might be.
Do results obtained after lesions of the FCPS in other laboratories support these conclusions? It is clear that behavioural impairments can arise in virtue of damage to the MSA or NBM that compromises systems other than the cholinergic. A number of such impairments have been seen, for example, after ibotenate lesions of the NBM, as indicated by the fact that they were not reproduced by quisqualate or AMPA lesions and/or were not reversed by cholinergic agonists or cholinergic-rich transplants. This list includes win-shift and win-stay operant memory tasks (Dunnett 1985; Etherington et al. 1987); operant conditional visual discriminations (Everitt et al. 1987); operant win-shift/lose-stay and win- stay/lose-shift conditional memory performance (Sinden et al. 1990); and navigation in the water maze (Dunnett et al. 1985; Page et al. 1991). The last of these findings is consistent with our own: the same animals which, after combined MSA and NBM lesions made by either ibotenate, quisqualate or AMPA, are permanently impaired in the radial-arm maze show in the water maze a deficit limited to acquisition of a hidden platform location (Hodges et al. 1992). Taken together with the findings reported by Dunnett et al. (1985) and Page et al. (1991), this pattern of results suggests that the cholinergic component in the water-maze navigation task is confined to learning.
As implied earlier, the prevailing view of the major behavioural role of the cholinergic projection from NBM to neocortex is one which highlights the attentional components of cognitive tasks (Dunnett et al. 1991). This has been demonstrated in some detail by a selective loss of choice accuracy in a serial reaction time task, with brief visual targets presented to rats with lesions of the NBM made with the three excitotoxins (Robbins et al. 1989a; Dunnett et al. 1991). Thus, it appears that learning in all tasks which require attention to a range of stimuli is likely to be affected by damage to the NBM-neocortex cholinergic pathway (Robbins et al. 1989a). Given sufficient training, it has been shown that animals with NBM lesions can perform a range of discrimination and memory tasks as efficiently as controls (Robbins et al. 1989b; Dunnett et al. 1991). The enduring deficits seen in Jarrard's radial maze task following cholinergic damage may thus arise from the heavy attentional demands of this task rather than impairments in any specific memory component (Hodges et al. 1991a).
Few published studies have employed the same strategy that we have pursued in attempting to establish just which of the behavioural effects of excitotoxic lesions of the MSA or NBM are due to cholinergic damage, namely, that of comparing and combining the two lesions. Hagan et al. (1988), using the water maze, found an impairment in spatial navigation after lesions to MSA but not to NBM. A possible inference from this result is that spatial navigation is not heavily dependent upon NBM cholinergic function -- an inference that we have already drawn from other evidence (see above). More strikingly, in an operant temporal discrimination task, it has been shown that rats with ibotenate MSA lesions underestimate, and those with NBM lesions overestimate, the time until they receive reward; and that rats with NBM, but not MSA, lesions are unable to perform a divided attention task (Olton et al. 1988). This is the only published evidence of which we are aware for a double dissociation between the effects of excitotoxic lesions of the NBM and MSA. Besides our own study (Arendt et al. 1989), only Hepler et al. (1985), to our knowledge, have combined the two lesions, in a study of rewarded alternation in the T-maze (a measure of working memory); the two lesions had similar effects, and their combination had effects that were equivalent to either separate lesion on its own. This pattern of results suggests a final common pathway for the two lesions, limiting their additive effects to a ceiling. Thus, although different in detail from the additive effects of NBM and MSA lesions described by Arendt et al. (1989), the pattern of results reported by Hepler et al. (1985) is consistent with the hypothesis of a unitary function for the entire FCPS.
5. Cholinergic-rich grafts
The experiments described above give reasonable support for the following conclusions: (1) combined excitotoxic lesions to the MSA and NBM give rise to cognitive deficits in the rat that involve cholinergic damage, and are measurable as global performance deficits in the radial-arm maze; (2) these behavioural deficits provide a model of the global cognitive dysfunction of Alzheimer's disease; (3) similar cognitive deficits can be produced by a prolonged regime of alcohol administration and, in spite of damage to other neurochemical systems, these too have structural and neurochemical correlates indicating damage to the forebrain cholinergic projection system; (4) these deficits provide a model of the cognitive dysfunction of alcoholic dementia or Korsakoff's psychosis. In the light of these conclusions, it becomes of great interest to determine the effects in these models of neural transplants. Such experiments can serve two goals simultaneously. First, they provide a further test of the hypothesis that the observed deficits are indeed due to cholinergic damage: if so, they should be reversible by transplants rich in cholinergic neurones but not by cholinergic-poor ones. Second, if the deficits prove to be reversible by neural transplants, this would indicate the eventual possibility of applying transplant therapy to cognitive deficits associated with human neurodegenerative disease.
There are a large number of potential biological sources of cholinergic-rich tissue for grafting studies, many of which have been successfully exploited at the anatomical, biochemical, physiological and behavioural levels in order to understand patterns of cholinergic regeneration and functional recovery. Until recently the almost universal source for studies of the recovery of cognitive function has been fetal brain, especially the basal forebrain dissected at a time, which in the rat can range from day 14-17 (E14-17) of gestational age (counting the day when vaginal plugs are found as E0), when final differentiation of cholinergic cells is complete but neurite outgrowth is still limited. The tissue is most often dissociated in trypsin prior to injection as a cell suspension into target regions (Bjrklund et al. 1983), though solid grafts are also employed. Fetal tissue grafts have the advantage of providing a source of cholinergic neurones with the potential for organisational and regulatory properties, although they will invariably contain cell types other than neurones and neurones expressing other transmitters than ACh. However, it is possible, also, to control to some extent for the non-cholinergic cell types in a basal forebrain fetal graft by employing, as an alternative source of fetal donor material, tissue from a different cholinergic-rich brain region; we have employed for this purpose tissue derived from the E13 mesencephalic fissure, containing the pedunculopontine and laterodorsal tegmental cholinergic primordia.
Alternatives to fetal-derived tissue have also been studied. The peripheral nervous system contains a rich variety of cholinergic neurones, and recent studies suggest survival and AChE-positive structures in nodosal (Itakura et al. 1990) and myenteric plexus (Lawrence et al. 1991) ganglion cells grafted to denervated cortex or hippocampus respectively. Adrenal chromaffin cells, deprived of glucocorticoid stimulation in vitro, have multipotential neuronal precursor characteristics and, after a period in culture, display cholinergic features (Boksa 1985; Ogawa et al. 1984); after transplantation into cortex of NBM-lesioned rats, enhanced AChE staining was seen around the grafted cells (Welner et al. 1990). However, in none of these cases of peripheral cell grafts was there clear evidence of graft-derived cholinergic influence on host brain. There is also increasing interest in fibroblast and tumour-derived cell lines as sources of graft material (Gage et al. 1991; and see Neuwelt et al., this volume); this is an issue to which we return below.
Our own major donor tissue has been the fetal cell suspension graft developed in the early '80s by Bjrklund et al. (1983). The development and rationalisation of this technique has encouraged uniformity in approach and permitted surprising consistency of data on the survival and integration of grafted tissue from many fetal sources. In general, grafts derived from cholinergic-rich fetal tissue taken at the optimal stage of embryonic development usually grow into a well vascularised cellular mass with boundaries clearly delimited from the host brain. Most experiments on the cholinergic system place the grafted cells ectopically, proximal to the normal targets of the FCPS and so distal to their normal location in the nuclei of origin of this system. Such procedures are adopted because fetal cells placed in the adult brain are not normally able to project their axons over large distances, such as those separating the terminal areas of the FCPS from their projection nuclei. There have been attempts to overcome this problem by the use of trophic bridges across fimbria-fornix lesion cavities provided by hippocampal grafts (Tuszinski et al. 1990), sciatic nerve fragments (Messersmith et al. 1991) or Schwann cell suspensions (Monteiro-Menei et al. 1992); these have helped to regenerate the septohippocampal cholinergic projection. However, ectopic locations have proved to be successful in dopaminergic-rich graft-aided recovery of motor function after lesions of the nigrostriatal pathway (Perlow et al. 1979; Dunnett et al. 1981). Like the FCPS, this appears to be a diffuse system (in the senses that we have attempted to define above; and see Le Moal & Simon, 1991), a similarity which encourages the belief that ectopic graft placements may similarly be effective in restoring function after cholinergic damage.
The growth of cholinergic-rich grafts in the host brain has been shown to depend upon a number of factors. Growth is better in the lesioned than the intact brain (Gage & Bjrklund 1986). Axotomy of the septohippocampal cholinergic pathway probably results in accumulation of NGF in the hippocampus, the effect of which is to enhance growth of, and fibre outgrowth from, cholinergic fetal grafts (Eriksdotter-Nilsson et al. 1989; Ernfors et al. 1989). A second factor is the degree of organotypic integration between the graft and the adjacent host tissue. In the denervated hippocampus, some studies have suggested that cholinergic cells derived from fetal septum, striatum and habenula all produce similar patterns of fibre ingrowth (Anderson et al. 1988; Gibbs et al. 1986). However, Dunnett et al. (1986) reported that septal grafts grew larger in deafferented hippocampus than cortex, while the converse was true for NBM grafts. More recently, Nilsson et al. (1988) and Clarke et al. (1990) have compared grafts into fimbria-fornix lesioned hippocampus of the same number of viable cells derived from five different cholinergic-rich sources: septum, NBM, striatum, pontomesencephalon and spinal cord. Growth was greatest for the NBM and spinal cord grafts, followed by the septal and pontomesencephalic grafts, striatal grafts coming last. Outgrowth of AChE fibres, however, followed a different pattern, being greatest for the septal grafts and least for the spinal cord ones; there was a good correlation between such fibre outgrowth and the number of ChAT-positive cells in the grafts, the septal grafts having the most of both.
Ultrastructural studies have demonstrated extensive synaptic contacts between ChAT-immunoreactive graft-derived fibres and vacated host target neurones in both hippocampus and cortex (Anderson et al. 1986; Clarke et al. 1986; Clarke & Dunnett 1986). Unlike the normal distribution of synapses, with most terminations on dendrites (Clarke 1985), graft- derived synapses were most frequently aberrantly located on perikarya. In their comparison of grafts taken from different fetal regions (see above), Clarke et al. (1990) observed ChAT- positive contacts made by all types of graft. However, most were made by septal grafts, and the grafts from striatum and spinal cord made very few; on the other hand, the contacts from septal grafts were mainly on perikarya, whereas those from the other grafts were more like normal cholinergic synapses, i.e., located on dendrites.
There are regrettably few data on changes in neurochemical and physiological function after cholinergic grafting. Muscarinic receptor binding has been investigated in fimbria-fornix lesioned hippocampus following transplants of solid fetal septal tissue into the lesion cavity; eight months after transplantation, the lesion-induced laminar- specific increases in stratum radiatum M1 binding and stratum oriens M2 binding were reversed to control values by the grafts (Dawson et al. 1989). Such receptor down-regulation suggests that the grafts in this study were secreting ACh. This suggestion has been confirmed by in vivo dialysis applied to free-moving rats. This technique has been used to show, in fornix-fimbria lesioned rats, that KCl-stimulated ACh release is increased, relative to the lesion-alone condition, by either solid septal grafts placed in the lesion cavity or by cell suspensions injected into the denervated hippocampus. Furthermore, ACh release was increased in the grafted animals in a qualitatively normal manner by either electrical stimulation of the lateral habenula (Nilsson et al. 1990) or behavioural activation via handling, immobilisation or swimming (Nilsson & Bjrklund 1992). These data clearly indicate some degree of host regulatory control over impulse- dependent ACh release from grafted cholinergic neurones. The degree of behavioural activation of the grafted neurones was, however, quite limited compared to observations in unlesioned animals; moreover, although the suspension grafts developed greater cholinergic fibre innervation of the hippocampus than the solid grafts, they showed the weaker activation of ACh release.
Evidence for functional recovery has also emerged from electrophysiological studies. In rats in which fornix-fimbria lesions had eliminated the hippocampal theta rhythm, Buszaki et al. (1987) were able to restore movement-related theta by solid E17 septal transplants into the lesion cavity; however, septal cell-suspension grafts into the hippocampus restored only limited theta, and this was seen in an inappropriate behavioural context, namely, during immobility, in spite of extensive cholinergic reinnervation of the hippocampus. These workers suggest that the solid graft may have restored theta by forming a bridge between the denervated hippocampus and host neurones that control theta; cells in the suspension grafts, while retaining some pacemaker characteristics, would lack host afferent control. Note that, as in the study by Nilsson and Bjrklund (1992), behavioural control of cholinergic hippocampal function was not a simple function of the degree of cholinergic reinnervation. There is also evidence that transplants may restore neocortical electrical rhythms disrupted by damage to the NBM. After such damage the low-voltage fast activity that normally accompanies awake immobility is suppressed by slow waves. This phenomenon was progressively reduced 2-14 weeks after neocortical grafts of E15-16 basal forebrain compared to rats with control hippocampal grafts (Vanderwolf et al. 1990). Electrophysiological evidence at a finer level of analysis is sparse. Segal et al. (1985), however, recorded intracellularly from CA1 pyramidal cells in transverse hippocampal slices taken from fornix-fimbria lesioned rats with E17-18 septal cell suspension grafts; they observed a pattern of responses to electrical stimulation of the graft indicative of muscarinic receptor activation of host pyramidal cells.
6. Results of cholinergic transplant experiments
In this section we ask, in reference to the FCPS: what is the extent to which neural transplantation is capable of restoring cognitive deficits? As we did for the data on lesion experiments in the FCPS, we shall first describe our own results and then consider the degree to which they agree with findings in other laboratories.
Our neural transplantation procedures (Arendt et al. 1989; Hodges et al. 1991a; 1991c; Sinden et al. 1990) were modelled on those employed by Bjrklund et al. (1983). Cholinergic- rich tissue was dissected from the basal forebrain of same- strain (Sprague-Dawley) fetuses at the embryonic age of 15 days (E15) and used to prepare cell suspensions after trypsinisation and gentle trituration. The cell suspensions were injected stereotaxically into two sites (dorsal and at the flexure) bilaterally in the hippocampus; and/or into two sites (frontal and temporo-parietal) bilaterally in the neocortex. In one experiment (Hodges et al. 1991c), cholinergic-rich tissue was prepared from the mesencephalic fissure (E13). Biochemical measurements at the end of the experiments showed that these transplants, as expected, increased the levels of cholinergic markers at the site of implantation. Nonetheless, the cell suspensions necessarily included many other cell types, neuronal and non-neuronal, besides the cholinergic ones of interest. For all experiments, control non-cholinergic tissue was prepared from the fetal (E17) hippocampus. At the end of all experiments the brains were subjected to thorough histological and biochemical analyses; the latter covered both cholinergic and aminergic markers.
---------------------
Figure 4 about here
---------------------
Our findings have been clearcut (Arendt et al. 1988a; 1989; Hodges et al. 1991a; 1991b; 1991c; see Figures 2 and 4): cholinergic-rich (but not non-cholinergic) cell suspensions were able to reverse the deficits in radial-arm maze performance, whether these were caused by the combined (MSA + NBM) excitotoxic lesion (Figure 1) or by prolonged alcohol ingestion (Figure 2). One of the great advantages of the radial-arm maze task is that performance can be measured repeatedly, allowing us to determine the time-course with which transplants exert their effects. As shown in Figure 4, recovery began to appear about 7-9 weeks after transplantation with basal forebrain cholinergic-rich cell suspensions and was asymptotic (and very substantial) at about 13 weeks. This time-course is consistent with Bjrklund et al.'s (1983) histological description of the growth pattern of similar cholinergic transplants into the hippocampus. Further, axonal connections between grafted cholinergic neurones and host cortex has also been demonstrated at a time point when cholinergic innervation is established. Combined retrograde tracer injection and ChAT immunohistochemistry has identified cholinergic neurones in basal forebrain grafts as innervating frontal cortical neurones over some millimeters 8-10 weeks following AMPA lesions of NBM (Calaminici et al. 1993). The transplants showed no sign of losing their effectiveness over the time (up to 1 year) for which we were able to follow the animals. Histological examination at the end of the experiment showed no sign of tumour growth, although many of the transplants had grown to a considerable size, distorting the host brain in the process. The hippocampal control transplants also grew well in the host brain, but had no detectable effect on radial maze performance. The volumes of the cholinergic- rich, but not cholinergic-poor, grafts as measured by planimetry or stereology were significantly positively correlated with behavioural recovery. Furthermore, the animals with cholinergic-rich grafts showed increases in ChAT activity to control levels in the terminal regions of the FCPS, and cholinergic neurochemical markers were the only ones to correlate with behavioural recovery. For example, in transplanted alcohol-treated rats, noradrenaline levels remained substantially reduced in animals that showed graft- induced behavioural recovery; while frontal cortex dopamine levels increased in all grafted animals, whether the grafts were behaviourally effective or not. Restoration of cholinergic function in the animals with behaviourally effective grafts was, however, incomplete, since these animals continued to respond aberrantly to cholinergic drug challenges (Figure 5): like lesioned, ungrafted animals they showed substantial behavioural deficits in response to the antagonist, scopolamine; and, unlike either intact or lesioned ungrafted animals, they responded to nicotine with an increase in errors. These interactions between drug and graft condition suggest that a complex, albeit abnormal, host- and/or auto-regulatory system may have operated to control transmitter release from the fetal grafts.
-------------------
Figure 5 about here
-------------------
As in our experiments on lesions to the FCPS (see above), we wished to compare the roles played by the two branches of this system, the projection from NBM to neocortex and that from the MSA to the hippocampus. To this end we compared the effects of placing basal forebrain cholinergic-rich transplants into either of these terminal areas or into both (Figure 4). The effects of transplants at each site were qualitatively and quantitatively similar, recovery being produced in all four measures of memory function. In ibotenate-lesioned rats the greatest recovery was seen in working memory performance, but in alcohol-treated rats improvements of similar degree were seen in both working and reference memory. The combined transplants, i.e. sited in both neocortex and hippocampus, had clearly additive effects in alcohol-treated animals (see Fig. 2). In the ibotenate- lesioned animals, recovery after either transplant on its own was so great as to preclude such dramatic evidence of additivity, although the combined graft group made significantly fewer errors than the group with grafts to cortex in three of the four measures. These findings are concordant, then, with the results of the lesion studies and provide further evidence that the FCPS plays a unitary role in cognitive function; and that this is one sufficiently general to affect all four memory components in Jarrard's (1986) version of the radial-arm maze task.
Given that cholinergic-rich transplants in both cortex and hippocampus are behaviourally effective, the possibility arises that any site in the brain would do: perhaps it is sufficient for the grafts to release a substance into the general cerebral circulation. To test this possibility, we placed basal forebrain, cholinergic-rich grafts into the sites (NBM and MSA) at which we had first injected ibotenate to damage the FCPS; however, there was no behavioural effect of these grafts. This result provides an important similarity between the FCPS and the nigrostriatal dopaminergic system; in this system, too, behaviourally effective dopaminergic-rich transplants have to be placed in terminal regions, not in the damaged nucleus of origin (Dunnett et al. 1981). A similar question concerns the source of the cholinergic-rich tissue: must it be homologous to the damaged FCPS? To address this question we injected into the neocortex and hippocampus of alcohol-treated rats cell suspensions derived from the cholinergic-rich fetal mesencephalic fissure: this source of transplant material was as effective as the basal forebrain in restoring radial-maze performance. Since, as we saw above, synaptic connectivity differs between these two sources of transplants (Clarke et al. 1990; Nilsson et al. 1988), this result must cast doubt on the role played by the detailed pattern of connectivity in mediating the behavioural effects of cholinergic-rich transplants, as must indeed the effectiveness of ectopic locations of these transplants. On the other hand, the fact that pontomesencephalic cholinergic- rich transplants are effective, whereas hippocampal cholinergic-poor transplants are not, reinforced our conclusion that the cholinergic nature of these grafts is a critical component for behavioural recovery (but see Section 11 below).
The results of these experiments demonstrate that cholinergic-rich transplants are able to produce a remarkably powerful and highly replicable recovery of cognitive function in animals in which the FCPS has been severely compromised by direct lesion or prolonged alcohol administration. There have been few studies with which we can directly compare these findings. A relatively large number of reports has been concerned with the effects of transplants in animals in which the hippocampal formation has been denervated by section of the fimbria-fornix. However, even when cholinergic fibre innervation is restored to near-normal levels by transplanted neurones, improvements in T-maze alternation, radial arm maze and water maze tasks are either very limited (sometimes dependent on cholinergic pharmacological enhancement) or non- existent; sometimes, in radial maze tasks, even cognitive impairments have been induced by grafts. (see Dunnett 1990; Sinden et al. in press for reviews). The relative failure of cholinergic-rich grafts to improve radial maze performance following fimbria-fornix lesions, compared to the improvements seen in rats with excitotoxic NBM + MSA lesions or chronic alcohol, may be due to the fact that mechanical fimbria-fornix lesions, unlike the other cases, interrupt nearly all subcortical pathways, both afferent to and efferent from the hippocampus in the rat. As a consequence, these lesions have been reported to establish, by blocking ascending inhibitory (particularly noradrenergic) inputs, an epileptic focus in the denervated rat hippocampus (Buzsaki & Gage 1988). These factors do not permit easy interpretation of the subsequent behavioural effects of intrahippocampal transplants in fimbria-fornix lesioned rats. However, recent experiments in a primate species, the marmoset, have demonstrated recovery of the capacity to learn visuospatial discrimination, one of a group of conditional learning tasks selectively impaired by fornix lesions, after fetal basal forebrain, cholinergic-rich, cell-suspension grafts, but not after hippocampal control grafts, placed into the denervated host hippocampus (Ridley et al. 1991b; 1992).
We know of no reports of cholinergic grafts into hippocampus in rats with excitotoxic lesions of the MSA alone. There are, however, some recent reports of cognitive improvements following cholinergic grafts in rats with lesions produced by intraventricular injection of ethylcholine mustard aziridinium ion (or AF64A). (This compound has been claimed to be a cholinergic-specific neurotoxin, acting by irreversible inhibition of high-affinity choline uptake [Hanin 1990], notwithstanding evidence of a range of non-specific damage at the site of its infusion [Allen et al. 1988; Jarrard et al. 1984; McGurk et al. 1987].) Grafts of fetal septum and NBM, but not striatum, improved acquisition of, and efficiency of strategy in, the radial maze, and the performance of the groups correlated with the degree of AChE-positive fibre innervation of the hippocampus (Ikegami et al. 1989; 1991). Similar effects of AF64A infusions and fetal septal grafts were demonstrated by Emerich et al. (1992); however, increasing the memory load, by imposing a 1-hr delay between the fourth and fifth choices in the maze, eliminated the improvements in the grafted group.
In contrast to studies of the behavioural effects of cholinergic grafts in the hippocampus, the majority of studies of such grafts in the neocortex have been carried out in rats with excitotoxic lesions (of the NBM). A number of studies have reported that fetal basal forebrain, but not control hippocampal grafts, implanted into multiple neocortical sites, improved the acquisition and/or retention of passive avoidance in rats with ibotenate lesions of the NBM (Dunnett et al. 1985; Fine et al. 1985; Sinden et al. 1990). Cholinergic-rich grafts have also been seen to ameliorate some, but not all, of the lateralised sensorimotor deficits tested in rats with unilateral NBM lesions made with ibotenate (Dunnett et al. 1985). Dunnett et al. (1985) also reported that, in ibotenate-lesioned rats, such grafts improved retention of a hidden platform position in the water maze, although acquisition impairments were not improved. In a T-maze delayed alternation task, Welner et al. (1988) found that neocortical grafts of E15 basal forebrain but not E15 cortical tissue gave rise to improvements in the deficit produced by quisqualate lesions of the NBM. A further positive result has been reported by Muir et al. (1992), who showed that, in a 5- choice serial reaction time task, a restricted set of deficits was induced by quisqualate lesions of the NBM and that these were reversed by fetal basal forebrain but not hippocampal grafts; one component of the deficit, however, was improved by both cholinergic and non-cholinergic grafts. As against these positive reports, one finding from our own laboratory has been negative: cholinergic-rich grafts into neocortex were unable to reverse the deficit produced in two operant variants of delayed matching and non-matching to position (win-shift/lose- stay or win-stay/lose-shift) (Sinden et al. 1990), even though a passive avoidance deficit was reversed in the same animals. Some cases of task-specificity of graft-induced recovery, such that tasks which may require greater levels of feedback control over local neural circuits do not show recovery, may be a function of the lack of opportunity for ectopically placed grafts to participate in normal neural circuit controls (Buzsaki et al. 1987; Nilsson & Bjrklund 1992). It has been argued that this principle may provide a general limit to the restoration of function following ectopic grafts of different kinds (Dunnett et al. 1987; Sinden et al. 1990).
We know of no published studies which have investigated the effects of transplants, implanted in both cortex and hippocampus, on cognitive deficits caused by combined lesions of the NBM and MSA or by prolonged administration of alcohol, other than our own, described above. Dunnett et al. (1988), however, placed cholinergic-rich grafts into either neocortex or hippocampus in aged rats which showed deficits in operant delayed matching and non-matching to position tasks. Given evidence for cholinergic impairment due to normal ageing (Bartus et al. 1982; Collerton 1986), it was expected, and observed, that the grafts would ameliorate these deficits. In line with our results in lesioned animals, the neocortical and hippocampal sites of implantation gave rise to similar improvements.
Given the number of positive reports of recovery of cognitive function after transplants of cholinergic-rich tissue in cholinergically impaired animals, it is clear that the general phenomenon is robust. The failures to produce recovery of function may be due either to a non-cholinergic origin of the cognitive deficit under study (e.g., excitotoxic damage to other neural systems), or to limitations in the transplant technique itself (e.g., the task may require greater synaptic integration than the technique is capable of achieving), or of course to other, unknown factors. One reason why our own findings, while in general in agreement with other data, have been particularly consistent is perhaps that we have taken care to establish (as summarised above) that the deficit for which we were able to show graft-induced improvement is indeed cholinergic in origin. Since the animals in which recovery of cognitive function has been induced by cholinergic-rich grafts include models of Alzheimer's disease, alcoholic dementia (our experiments) and cognitive decline during normal ageing (Dunnett et al. 1988; Gage et al. 1984), these results are clearly promising with regard to the eventual application of neural transplant therapy in equivalent human conditions. Recent experiments in a primate species, the marmoset, have demonstrated recovery of the capacity to learn complex conditional discriminations after fornix lesions, as described above, following cholinergic-rich intrahippocampal transplants (Ridley et al. 1991b; 1992). Thus, as in the comparable case of dopaminergic-rich transplants in the treatment of motor disorders, the transition from rodents to primates appears to be feasible. We emphasise, however, that much basic research needs to be done, with both rodents and primates, before the further leap to humans should be attempted. For example, our current investigations of mechanisms of action of grafts raise a serious question whether, despite the coherence of the results of the experiments on the FCPS which we have described so far, the requirement for grafts to contain neurones expressing ACh is not, after all, a necessary condition for functional recovery in rats with lesions to the FCPS. We take this issue up again in Section 11.
7. Ischaemic damage to hippocampal cell fields
In our discussion of the FCPS we have repeatedly described it as `diffuse' or 'global', justifying this term in the light of anatomical, physiological and behavioural data. We have likened the FCPS in these respects to ascending monoaminergic systems, and especially to the dopaminergic pathways for which the earliest and still most abundant evidence exists that transplants can produce behavioural recovery of function. Thus the success, documented above, that has also attended attempts to produce such recovery of function by way of cholinergic transplants prompts this question: are neural transplants capable of producing cognitive functional recovery only if the neuronal system underlying the function is of the diffuse kind? It could be, for example, that the technique can be effective if all that is required is the delivery of a substance (be it trophic, neurotransmitter, or other) into a particular region, the transplant acting rather like a biological mini-pump; but that fully regulated neuronal integration with the host brain is beyond the capacity of the transplantation method, so that damaged point-to-point projections cannot be repaired in this way. This issue was highlighted in early discussions of graft-induced functional recovery; for example, Sotelo and Alvarado-Mallart (1987) noted that most cases of successful functional recovery following neural grafting occurred in systems (such as the cholinergic one) characterised by the ratio between a small number of projection neurones to a large number of target cells, such that paracrine neurotransmitter release, without specific point-to-point synaptic connectivity, is sufficient to produce functional recovery. Point-to-point neuronal systems are ones in which each nerve cell contacts only a few target neurones. To restore function in point-to-point systems, grafted neurones would presumably need to replace missing neurones by reconstituting an equivalent synaptic circuitry. Sotelo and Alvarado-Mallart (1991) have elegantly described a transplanted point-to-point system, whereby neuroblasts from grafts of fetal cerebellar primordia are able to migrate, develop into Purkinje cells and make functional afferent and efferent connections to a Purkinje-cell-degeneration mutant strain of mouse. This archetypal case of graft-host circuit reconstruction is not, however, a universal phenomenon within the CNS. For example, many studies have failed to demonstrate any potential, beyond an early, presumably trophic effect, for grafts of fetal neocortical tissue to restore behavioural function and/or to establish appropriate graft-host interconnections following damage to neocortex (see Stein & Glasier, this volume and review by Sinden et al. 1992a). The contrast between the failures of cortical grafts in the neocortex and the success of cerebellar grafts in the Purkinje-cell-degeneration mutant mouse has prompted us recently to speculate that behavioural recovery by means of circuit reconstruction may depend on the degree to which processsing within a damaged circuit is contingent on a well-defined set of neural connections. In neocortex, where many parallel circuits may mediate recovery following brain damage, grafted neurones may not therefore be recruited for this purpose. However, in cerebellum and in other structures with more precise laminar organisation, circuit reconstruction may be possible, provided the grafted neurones are appropriate to the nature of the damage (Sinden et al. 1992a). Apart from cerebellum, evidence for tissue-type specificity in graft-induced circuit reconstruction, with parallel functional recovery, has been demonstrated in striatum (Wictorin, 1991), hippocampus (see below) and (gustatory) archicortex (Escobar et al. 1989). Such evidence for recovery of cognitive function following damage to hippocampal formation circuits is the issue we address next.
In the experiments from our laboratory described so far the hippocampus was one of the denervated regions; the damaged cells were distant from the hippocampus; and hippocampal fetal tissue provided a control for the effective, cholinergic transplant. Now we turn the tables and consider a series of experiments with the converse characteristics: the damage is inflicted on cell fields within the hippocampus (Figure 6); and the transplant that is expected to repair this damage (and its associated cognitive deficits) is taken from the fetal hippocampus.
----------------------
Figure 6 about here
----------------------
The origins of these experiments lie in two sets of previous observations. First, anatomical studies in Raisman's and Zimmer's laboratories have shown that grafted cells from particular hippocampal fields establish appropriate laminar afferent and efferent connections with host neurones (intrahippocampal reconstruction: Field et al. 1991; TQnder et al. 1990; entorhinal cortex - hippocampal formation: Zhou et al. 1989). Second, studies of experimental cerebral ischaemia (reviewed by Hodges et al., in press), simulating temporary cardiac arrest, have demonstrated cognitive impairments associated with damage that is largely restricted to one hippocampal cell field, the CA1 pyramidal cells (Davis & Volpe 1990; Volpe et al. 1984). Grafts of hippocampal tissue had also been shown to survive and become integrated into ischaemic host brain (Mudrick & Baimbridge 1991; TQnder et al. 1989). In human beings who have survived temporary cardiac arrest, cognitive deficits and hippocampal CA1 damage are also often observed (Squire 1986; Volpe & Petito 1985; Volpe et al. 1984; Zola-Morgan et al. 1986). Localised damage within this system is detectable by neuroimaging techniques (Squire et al. 1990) so that, in principle, it may be possible to identify patients suitable for transplant therapy. Given this background, we set ourselves the task of determining whether transplantation of CA1 pyramidal cells might be able to reverse the cognitive deficits caused by cerebral ischaemia. If this were possible, it might again be the basis for a future transplant therapy. A demonstration of functional recovery with grafted CA1 tissue, but not with tissue from other hippocampal fields, would also show that transplantation of appropriate neurones is applicable in the case of damage, not only to diffuse neural systems, but also to local point- to-point projections. Neurotransmission around the hippocampal circuit -- that is, from the entorhinal cortex to the granule cells of the dentate gyrus, thence to the pyramidal cells of the CA3 region, and finally on to the CA1 pyramidal cells (Figure 6) -- is topographically highly organised and mediated by the fast excitatory amino-acid, glutamate (for review, see Gray 1982). Moreover, the phenomenon of long-term potentiation (LTP), first observed in the hippocampus and thought to play an important role in learning and the formation of memory traces, is synapse- specific (Baudry & Davis 1991).
The experimental methods most commonly used at present to produce transitory global cerebral ischaemia are those of two- vessel occlusion (2VO) and four-vessel occlusion (4VO) (for review and detailed references, see Hodges et al., in press). The 2VO method involves ligation of the common carotids in paralysed and ventilated animals, with blood pressure reduced (pharmacologically or by exsanguination). The most consistent neuropathology is in dorsal hippocampus, where up to 50-60% of CA1 cells die within 3-4 days after ischaemia. The 4VO method, developed by Pulsinelli et al. (1982), is a two-stage procedure, in which the vertebral arteries are electro- coagulated and the carotids loosely snared under anaesthesia, and 24 h later the snares are tightened for a specified period. Rats lose the righting reflex within 2 min or are excluded, so that the degree of brain damage is roughly standardised. The extent of cell loss is related to the duration of tightening of the snares around the carotids. With 15-min occlusion, loss of neurones is largely confined to the CA1 area of the hippocampus, where cell loss in dorsal regions is near total (80-90%) (Le Peillet et al. 1992; Netto et al. 1993; Pulsinelli et al. 1982). After 30 min occlusion, damage within the hippocampus is more extensive, and there is also loss of cells in cortex and striatum.
8. Grafts from hippocampal subfields
The selective loss of hippocampal CA1 pyramidal cells with brief (10-20 min) episodes of 4VO provides an opportunity to investigate the effects of neural grafting to a localised area of damage within a well-defined circuit. Within the hippocampus each neuronal pathway follows an identified laminar route. For example, entorhinal cortical fibres synapse within the outer two thirds of the dentate granule dendritic field, whereas mossy fibres from the dentate gyrus project selectively to the juxtacellular area of CA3 pyramidal cells. Thus, with appropriate tracing techniques, it is possible to assess the accuracy with which grafted neurones project along normal routes and receive appropriate host afferents. Grafts of homotypic hippocampal tissue have therefore been the method of choice to examine the morphological effects of grafting in models of global ischaemic damage.
Workers in Zimmer's and Raisman's laboratories have recently shown that it is possible to microdissect and selectively implant tissue from fetal hippocampal subfields, starting from about E18, when the subfields are clearly distinguishable; and that one can trace the specificity of graft-host connections with these selective grafts, using a variety of cell stains, immunological and genetic markers and tracing techniques. TQnder et al. (1990) implanted E18-20 hippocampal CA3 suspension grafts into the host CA3 field, lesioned a week earlier with ibotenic acid so as to destroy the pyramidal cell bodies while leaving intact the host mossy fibre projection to the area. Histological examination showed that 90% of the grafts survived, contained normal CA3 neurones (pyramidal cells and cells reactive to appropriate peptidergic, cholecystokinin and somatostatin immunostaining), and were innervated by host septal AChE-positive and dentate gyrus Timm-stain-positive neurones. Antero- and retrograde tracing showed that the grafts projected to the host CA1 area, and that there were two-way commissural connections between host and graft.
The detailed synaptic pattern of connections between host mossy fibres and grafted E20 CA3 neurones was examined by Field et al. (1991), using the monoclonal antibody, Py, to label CA3 cells and Timm staining for mossy fibre projections. Host dentate granule cells projected exclusively to the normal juxtacellular region of the grafted dendritic field, and were seen at the electron-microscope level to reach 20% of normal synaptic density. CA3 grafts implanted distally in the host septum showed no mossy fibre synapses, whilst grafted CA1 cells implanted in the mossy fibre pathway similarly received an insignificant number of host projections, with only a few mossy fibre terminals. These observations show that grafts of hippocampal subfields not only become integrated in the host neural network, but do so in a highly specific manner; host innervation of the graft proceeded along normal laminar routes and faltered when presented with inappropriate pyramidal cells (CA1 instead of CA3).
These findings indicate an astonishing precision in the patterning of reinnervation following implantation of grafts from discrete hippocampal subfields into the appropriate host region. They therefore provide a powerful strategy for the study of mechanisms of graft action on cognitive function in animals with relatively discrete lesions to the CA1 field following global ischaemia. One can ask whether, in animals with selective ischaemic loss of CA1 cells, grafts of these cells are more successful in improving behavioural deficits, as well as forming more normal connections with the host brain, than grafts from other regions, including other hippocampal subfields. Since cells from the major fields (CA1, CA3 pyramidal and dentate granule cells) all release glutamate, evidence of an advantage for CA1 cells would indicate that release of an appropriate transmitter, such as we have argued above may be the sufficient requirement for functional recovery in diffuse neural systems, is not by itself a key factor in promoting functional recovery; and, since CA1 and CA3 cells are morphologically very similar, both being pyramidal neurones, such evidence would also indicate that general cell type plays a similarly limited role. Such an advantage for CA1 cells, though clearly suggested by the anatomical studies from Zimmer's and Raisman's laboratories, would be contrary to the pattern we observed for cholinergic repair, in which cholinergic cells derived from the primordia of the pedunculopontine and laterodorsal tegmental nuclei were able to produce the same behavioural effects as the basal forebrain cells homotypic to those that had been lesioned (Hodges et al. 1991c).
9. Cognitive effects of grafts from hippocampal subfields
We have recently conducted a series of experiments aimed at addressing these issues, using the 4VO method to produce transitory global cerebral ischaemia in the rat. We have been able to replicate previous observations (Pulsinelli et al. 1982) that, applied for 15-20 minutes, this technique gives rise to damage largely restricted to the CA1 pyramidal cell layer. Associated with this damage, we have shown a reliable cognitive deficit (Netto et al. 1993; Nunn et al. 1991a; 1991b), but one that is different from the deficit that we see after damage to the FCPS. The ischaemic animals behave normally in Jarrard's (1986) radial maze task, but are impaired in their ability to learn to navigate to a hidden platform in the Morris water maze; in our hands, FCPS-lesioned animals are, in contrast, severely impaired in the radial maze but have only a slight acquisition deficit in the water maze (Hodges et al. 1992; Nunn et al. 1991b). On initial acquisition in the water maze, ischaemic rats are impaired in global search strategy, but with experience the deficit becomes highly specific: they are able to navigate efficiently to the general area of the pool where the platform is hidden, but are impaired in their final localisation of its exact position. Thus, behaviourally as well as in terms of neural organisation, the ischaemic syndrome provides a clear contrast with that of cholinergic damage.
Our first study (Netto et al. 1993) compared the effects of E18-19 CA1 and dentate granule (DG) cell suspension grafts, microdissected according to the method of Field et al. (1991); E15 cholinergic-rich, basal forebrain (BF) grafts were used as non-hippocampal control tissue. Non-ischaemic and sham- transplanted ischaemic controls were also employed. Transplantation took place 3-4 weeks after 15 min 4VO. For behavioural testing we employed acquisition of a hidden platform position in the water maze (a reference memory task). Animals were tested after ischaemia and from 4 weeks to 6 months after transplantation, providing measures of retention and reversal of the initial platform position, and acquisition of a novel position, with either 4 trials/day and a 5-min inter-trial interval (ITI) or the more difficult procedure of 2 trials/day and a 10-min ITI. We also used a working memory test with 4 trials/day, a 30-s ITI, and different platform positions on each of 6 days.
----------------------
Figures 7 and 8 about here
----------------------
The major results of the experiment are shown in Figures 7 (reference memory) and 8 (working memory). When first tested, the ischaemic rats showed substantial impairment of acquisition in terms of latency to find the platform, path length and reduced % time spent in the training quadrant on a probe trial with the platform absent; impaired memory for the platform location was also evident on the probe trial. The ischaemic deficit was most apparent with the sparse training regime, and was seen in both long-term acquisition of the fixed platform location and in the working memory task, indicating a rather general impairment in spatial learning. Such deficits in spatial learning are, of course, a well-known consequence of damage (usually, however, of much greater extent) to the hippocampal formation (O'Keefe & Nadel 1978). With repeated testing in a familiar pool, the ischaemic animals showed substantial improvement in performance, indicating that they were capable of forming long-term representations; when they were transferred to a new pool, however, the acquisition deficit re-appeared. In both the reference (Figure 7) and the working (Figure 8) memory tasks the rats with CA1 grafts performed at non-ischaemic control level and learned significantly faster than ischaemic rats or the other grafted groups (BF and DG); the latter two groups were as impaired as the ischaemic animals without transplants. The efficacy of the CA1 grafts was most evident when animals were tested in a novel pool 7 months after transplantation.
The results of this experiment provided the first indication that CA1 grafts are more effective than those from another hippocampal subfield in alleviating ischaemic deficits in spatial navigation. Histological examination with Nissl and Timm staining and AChE histochemistry showed that the transplants survived in all cases examined, sited in the alveus above the region of maximal ischaemic dorsal CA1 cell loss. The hippocampal subfield grafts contained cells resembling, as appropriate, CA1 or DG cells and were innervated by AChE-positive fibres from the host medial septal area. There were, however, several problems with this study which required clarification. First, grafts were implanted in rats which had already had behavioural experience, so that their efficacy against the initial acquisition deficit had not been tested. Second, the CA1 grafts appeared to be more uniformly sized and thriving than the DG grafts, some of which were large, but others much smaller with signs of glial infiltration. Morphological studies (Bayer 1980) have shown that DG cells develop later than pyramidal cells, so that the age of the donor may have favoured survival of CA1 rather than DG grafts. If so, behavioural efficacy may have been related to the viability of the grafts rather than their connectivity and/or function. We therefore set up a second study (Hodges et al. 1993) to compare the effects of E18-19 CA1 grafts with those of DG grafts dissected at post-natal day (P) 1-2 to optimise their chances of survival. We also included E18-19 CA3 grafts to provide a more stringent test of structural specificity, by comparing the effects of two populations of pyramidal cells. These manoeuvres were not entirely successful. We found no histological evidence for survival of the grafted neurones in the DG grafted group. Moreover, survival was more marked in the CA1 graft group (80% of grafted rats) than in the CA3 group (50%). These findings confirm that donor age can critically survival of hippocampal cell-suspension grafts. They also suggest that the host environment influences graft viability and that this factor may contribute to the behavioural efficacy of homotypic intrahippocampal grafts. In the results described below, we have removed data of animals with little ischaemic cell loss and of animals with no or poor CA3 or CA1 grafts. However, we included the results of rats with failed "DG" grafts to provide a second ischaemic control group. Thus the results come from animals with comparable CA1 cell loss and, in the case of CA1 and CA3 groups, comparable sizes and sites of transplants, and the statistical analysis takes into account unequal group sizes.
The grafts were implanted in the alveus above the dorsal CA1 region 3 weeks after 15 min of 4VO. When tested in water maze acquisition (Figure 7) 11-12 weeks after grafting, the rats with CA1 grafts again showed a substantial degree of improvement compared to the marked navigational deficit shown by ischaemic and "DG" controls or the ischaemic group with CA3 grafts. The CA1 group found the platform as rapidly as non- ischaemic controls and spent a larger proportion of time in the training quadrant than the other ischaemic groups. In terms of precise recall of the platform position on the probe trial, however, as shown by time spent near the platform position, all ischaemic groups, including those with CA1 grafts, showed less accuracy than the non-ischaemic controls. All ischaemic groups were also impaired in heading angle, a measure of taking the direct path to the platform. Thus the CA1 graft improved the efficiency of spatial learning, but not the accuracy of retention or of initial orientation to the platform. The selectivity of CA1 relative to the other grafts was confirmed in a second behavioural test, matching to position in the three-door runway, shown by Sugimachi et al. (1992) to be sensitive to excitotoxic lesions of CA1: non- ischaemic controls and the group with CA1 grafts made significantly fewer errors than ischaemic animals without grafts or with CA3 grafts. As in the first experiment, the efficacy of the CA1 graft was still apparent 6 months after transplantation, when the novel pool test was conducted (Figure 7). Superiority of the CA1 group was also replicated in the water maze working memory test (Figure 8).
This second experiment, therefore, confirmed the long- lasting nature of the ischaemic impairment in spatial information processing, as well as the capacity of CA1 grafts to reverse this deficit, also in a long-lasting manner. This capacity is apparently specific to CA1 grafts, compared in the first experiment to DG cells, and to the structurally similar CA3 pyramidal cells in the second experiment. It is unlikely that CA1 but not CA3 grafts secrete growth factors, able to rescue ischaemic-lesioned host CA1 cells. We saw no evidence of such trophic effects. In Nissl and Timm stained sections, grafts of both kinds were clearly delineated from host brain. Counts of remaining host CA1 neurones did not differ between any of the ischaemic groups, whether grafted or not (Netto et al. 1993).
In a further set of experiments we have rung the changes on the area of hippocampal damage, targetting now DG rather than CA1 cells. Selective damage to this field may be achieved by neonatal X-irradiation or local injection of the neurotoxin, colchicine. Using the former method followed by transplantation of whole fetal hippocampus, Mickley et al. (1991) were able to demonstrate graft-induced alleviation of behavioural deficits, but only transiently. In our laboratory, we have employed the latter method (Xavier et al., in preparation). We initially injected colchicine at multiple sites in both dorsal and ventral hippocampus, causing extensive loss of DG cells and a profound deficit in the water maze; in these animals, fetal cell suspension DG grafts survived poorly and produced no significant recovery (Xavier et al. 1991). However, in rats with smaller lesions, made by infusing colchicine only in the dorsal hippocampus, the deficit in the water maze was smaller and comparable to that produced by ischaemia. We used these animals to compare the effects of solid grafts (which survived well) prepared from E19-20 fetal dentate gyrus or CA1 regions placed into the lesioned dorsal hippocampus. Tested 8 weeks after implantation, a time at which neural connections between graft and host should have formed, the animals with DG but not CA1 grafts showed an improvement in water maze acquisition relative to lesioned animals without grafts (Figure 9; Xavier et al., in preparation). These results are the reverse of those found in the experiments using ischaemic animals, strengthening the case that graft-induced recovery from damage at different points in the hippocampal trisynaptic circuit is field-dependent.
------------------------
Figure 9 about here
------------------------
This inference is also supported by recent electrophysiological observations made in our laboratory (Dawe et al. 1993). It had previously been reported that LTP can be elicited in hippocampal grafts placed in the lesion cavity of rats with fornix-fimbria lesions (Buzsaki et al. 1989). This is an important observation, given that LTP of hippocampal neurotransmission has been identified as an early correlate of learning, associated with activity at the NMDA class of glutamate receptors (Morris et al. 1986; and see Baudry & Davis 1991 for review). We therefore studied hippocampal slices, where graft tissue could be distinguished from host brain, taken from animals with colchicine lesions of the dorsal dentate granule cells, followed by solid grafts of E19- 20 dentate or CA1 fetal tissue, as in the behavioural experiments described above. In both cases field potentials and single unit activity were evoked in the graft by stimulation in the mid-molecular layer of the host dentate gyrus. Short- and long-term potentiation were studied by the delivery of tetanizing trains at this site, with measurements at 1 and 15 min post-tetanization. In lesioned rats with DG grafts short-term potentiation reached the levels displayed in slices taken from intact controls, and it was greater than in rats with CA1 grafts. Importantly, LTP was present in the rats with DG grafts but absent in those with CA1 grafts (Dawe et al. 1993). These findings, we believe, provide the first evidence for field-specific differences in electrophysiological responses depending on whether the grafted cells match those that have been lesioned in the host. They agree with the evidence from Xavier's study, described above, for field-specific selectivity in the effects of the same grafts on behaviour, for which indeed they may provide the neural basis.
10. How do neural transplants improve cognitive function?
It is clear from the data described above, from both our own and other laboratories, that neural transplants are able to reverse a variety of impairments in cognitive function, just as they have been shown earlier to reverse impairments in motor function (for reviews, including other data on cognitive function besides those covered here, see Dunnett & Bjrklund, in press). What is the mechanism? Or rather, mechanisms; for it is unlikely that cholinergic and hippocampal subfield transplants work in entirely the same way (for reviews of the variety of possible mechanisms, see Dunnett & Bjrklund 1987; Gage & Fisher 1991; and Gray et al. 1990).
As we have seen, the two cases differ in a number of respects. Anatomically and physiologically, the FCPS is constructed and acts in a diffuse manner; the hippocampal circuitry, in a point-to-point manner. Behaviourally, damage to different parts of the FCPS produces impairments that are equipotential, additive, and affect a very general cognitive process (perhaps attentional) that is required for adequate performance in the radial maze, but much less so for performance in the water maze; damage to the hippocampal circuitry affects a process (perhaps fine spatial localisation) that is more important for performance in the water than in the radial maze. Behaviourally effective grafts are ectopically located, in denervated terminal areas, in the case of the FCPS; but homotopically located, in the region of the lesion, in the case of hippocampal subfields. In the case of the FCPS, the effective grafts have in common that they contain cholinergic cells, and it does not matter if these are derived from tissue homotypical with the lesioned area (the basal forebrain) or elsewhere (the mesencephalic fissure); in the case of hippocampal subfield damage, the requirement for effective graft action appears to go beyond that of possessing cells specific for the lost transmitter (glutamate), although we have not measured glutamate release from the grafts in these experiments. Finally, grafts in the damaged FCPS may be behaviourally effective in spite of different morphologies and connectivities (as again instanced by basal forebrain vs. mesencephalic fetal tissue); but grafts in the hippocampus apparently must be of the same morphological type as the subfield that has been lost and capable therefore of making the same connections. These striking contrasts between the two cases, coupled with the evidence that neural transplants are able to produce substantial recovery of cognitive function in both, imply that the many types of cognitive deficits due to damage to the brain may turn out to be amenable to transplant surgery.
With regard to the FCPS the obvious implication of the data we have reviewed is that the behaviourally effective grafts work by secreting the missing transmitter, ACh. This inference is consistent with the requirement that effective graft tissue should contain cholinergic cells, with the ectopic location of the grafts, with the lack of importance of the precise type of cholinergic cell present in the grafts, and with the acute restoration of function that can be achieved by the systemic administration of cholinergic agonist drugs. Might it be the case that cholinergic grafts simply provide an elaborate drug delivery system? The difficulty with this hypothesis is that ACh is rapidly hydrolysed by AChE, so that extrasynaptic delivery of the transmitter may be rather ineffective. It is true that Nilsson et al. (1990; Nilsson & Bjrklund 1992) were able to detect increased extracellular ACh in the vicinity of cholinergic grafts; but their in vivo dialysis system included an inhibitor of AChE in the probe, so that it is difficult to interpret the significance of these observations for the mode of action of the grafts in the absence of such an inhibitor. But at least their findings strongly imply that cholinergic grafts do indeed release ACh.
If ACh secreted by cholinergic-rich grafts does not act in a quasi-pharmacological manner, might its release be in some way regulated, synaptically or otherwise? The ectopic location of the effective grafts clearly precludes that any such regulation is provided by all or even most of the afferents that are normally received by the nuclei of origin of the FCPS. However, a few of these afferents are from target areas themselves innervated by the FCPS (Saper 1984); thus one possibility is that such cholinergic neurones grafted into terminal areas establish these reciprocal connections in a quasi-normal manner. Tracer studies have also identified a variety of sparse projections to the nuclei of origin of the FCPS from noradrenergic neurones in the locus coeruleus and serotonergic neurones in the dorsal raphe (Jones & Yang 1985; Jones & Cuello 1989); since these monoaminergic pathways also innervate the cortical and hippocampal terminal areas of the FCPS, they might establish quasi-normal connections with the grafted cholinergic neurones sited in these areas (Leanza et al. 1993).
One way to investigate the importance of such regulatory processes in determining the effectiveness of fetal cholinergic-rich grafts in the induction of cognitive recovery is to examine the effects of other types of graft which are capable of delivering ACh but which are likely to be regulated differently.
Grafts derived from peripheral tissue, for example, have been shown to ameliorate cognitive function. Welner et al. (1990) reported that intra-cortical adrenal medulla chromaffin cell grafts were as effective as fetal basal forebrain grafts (Welner et al. 1988) in improving T-maze rewarded alternation performance following quisqualate lesions of the NBM. However, there was no clearcut evidence for cholinergic properties of the grafts themselves in this study; cortical AChE staining was enhanced in the grafted animals, but the authors argued that the grafts induced trophic and/or tropic effects on surviving host cholinergic terminals. Similarly, nodosal ganglion grafts (Itakura et al. 1990) into parietal cortex were found to improve NBM-lesioned rats' performance on both passive avoidance and maze tasks, again without evidence of cholinergic innervation of the host cortex from the grafts.
Similar experiments have been performed using neuroblastoma cell lines with putative cholinergic features. Kordower et al. (1987) reported that implants into hippocampus of amitotic C1300 and LAN-2 neuroblastoma cells produced a limited attenuation of T-maze alternation deficits following septal lesions, but without any evidence of recovery in cholinergic markers in the post mortem brain. In rats with ibotenate lesions of the NBM and MSA, deficits in all four memory components of Jarrard's radial maze task were shown to be improved by implants into cortex and hippocampus of chemically-differentiated NS20Y mouse neuroblastoma cells. This cell line expressed both ChAT- and GABA-like immunoreactive markers in vitro (Marsden et al., in preparation). At the end of testing, performance of the neuroblastoma graft groups was not significantly different from that of unlesioned controls (Figure 10). Moreover, the time course of recovery was similar to that produced by fetal basal forebrain grafts, suggesting a similar mechanism of graft-induced recovery (Marsden et al., in preparation).
------------------------
Figure 10 about here
------------------------
If neuroblastoma cell lines and fetal cell suspensions both induce functional recovery in FCPS-lesioned animals by the same mechanism, it is unlikely that this mechanism involves similarly regulated transmitter release. Histological observations indicate that labelled neuroblastoma cells become widely dispersed throughout the cortex and hippocampus. Supposing that these cells secrete ACh, they would not therefore be able to enter into reciprocal connection with FCPS terminal areas, nor to make contact with monoaminergic afferents to these areas, thus ruling out the two possible mechanisms for regulated transmitter release suggested above. Perhaps it is a mistake to be thinking in terms of transmitter release at all; some other feature of cholinergic-rich transplants, besides their capacity to secrete ACh, may be responsible for their effects on cognitive function. One obvious possibility is that the effective grafts may secrete growth factors which work by enhancing repair in the damaged host nervous system (see Thoenen et al., this volume).
If this `growth factor' hypothesis is correct, then recovery of function should survive removal of the graft that has produced it. The original xenogeneic (mouse) source of the NS20Y cell line provided an opportunity to test this prediction. At the end of 12 weeks of post-graft testing, when asymptotic improvements in radial maze performance were seen in the NS20Y group, the neuroblastoma cells were irradiated and subcutaneously injected once a week for four weeks into the grafted group, as well as the controls, so as to cause immune rejection of the grafts. The rats were tested weekly on the radial maze during this time and for a further five weeks. The grafted rats' performance declined progressively over the nine weeks to the level of the lesioned, unimplanted controls, although performance in the other groups was unchanged (Figure 10; Marsden et al., in preparation). Post-mortem investigation showed substantial signs of graft rejection, including an elevated peripheral lymphocyte response and, in the grafted cortical and hippocampal areas, mononuclear cell infiltration and high levels of MHC class I and II antibody staining. Thus, dysfunction in and/or destruction of the grafted cells by immune rejection removed their behavioural efficacy. Although it is possible that host tissue which, under the influence of graft-induced growth factors, has undergone repair might sustain renewed damage because of graft removal, we believe that these results more plausibly imply that it was the grafted cells themselves and not changes in the host brain that were responsible for this efficacy.
11. Neurones or glia?
A further approach we have taken, in order to unravel the multiplicity of possible neurochemical changes that may mediate the behavioural effects of cholinergic-rich grafts, has been to examine protein changes in the frontal cortex of rats with ibotenate lesions of the FCPS followed by fetal basal forebrain grafts. The grafts were removed from the brain in animals that had been behaviourally tested. Two- dimensional gel electrophoresis highlighted 33 proteins that were reliably identified in each sample and submitted to image analysis. The normalised values were related to type of transplant (cholinergic-rich vs. -poor) and individually correlated with performance in the radial maze. Of the 33 proteins, 7 showed differences which depended on type of transplant and/or were correlated with behaviour. Only one identified protein -- glial fibrillary acidic protein (GFAP), a marker of glial cells, mostly astrocytes -- was elevated in cholinergic-rich compared to -poor transplants and positively correlated with maze performance. On the other hand, some non-specific neuronal cytoskeletal proteins and neurone- specific enolase were negatively correlated with radial maze performance, but did not differ between transplant groups (Wets et al. 1991). In contrast to the inferences we have drawn earlier, these data suggest that, even when measured 12 months post-lesion and 7 months post-transplant, it is the cholinergic-rich transplant-related glia that aid cognitive recovery, whereas some neural elements may actually inhibit the recovery process.
Experimental support for this possible role of glia has recently been reported by Brckner and Arendt (1992). Rats given prolonged alcohol were implanted in cortex and hippocampus with primary cultures of P1-2 cortical astrocytes or fetal basal forebrain cell suspensions. Post-graft testing on the place version of the Jarrard radial maze test replicated the previously reported (Arendt et al. 1988a; 1989) recovery of function in the basal forebrain grafted animals, with a time-course consistent with growth of the transplant and outgrowth of cholinergic fibres. Primary astrocyte grafts also improved performance, this effect being seen soon after grafting, but reaching an asymptotic level that was marginally inferior to that of the basal forebrain group. Interestingly, both the astrocyte and, to a lesser extent, the basal forebrain grafts increased ChAT activity in the host basal forebrain, the latter effect confirming earlier observations (Arendt et al. 1989) and suggesting that grafted glia may induce recovery in damaged cholinergic neurones, possibly by providing trophic factors. In our experiments, we derived primary cultures separately enriched either in post-mitotic astrocytes or in neurones or neuronal precursors from E15 basal forebrain, and compared grafts of these cell types with E15 basal forebrain cell suspension grafts in rats with AMPA lesions of MSA and NBM. In two experiments, rats were trained four months after lesion and grafting on either a T-maze win- stay learning task or Jarrard's radial maze task. The astrocyte and fetal basal forebrain graft groups were significantly better at both tasks than the lesion-alone group. Performance of the groups with neuronal cell grafts was no better than the lesion-alone group (Figure 11). Both whole fetal basal forebrain and neuronal primary cell grafts increased ChAT activity in cortex and hippocampus; grafted astrocytes, however, had no significant effect on terminal area ChAT activity (Figure 12, Bradbury et al., submitted). Unlike whole basal forebrain and neuronal transplants, grafted astrocytes were difficult to identify in histology beyond a short post-grafting interval (Brckner & Arendt 1992; Bradbury et al., submitted); this is consistent with labelling studies which have shown that astrocytes migrate considerably after grafting (Emmett et al. 1988).
----------------------
Figures 11 and 12 about here
----------------------
The results shown in Figures 11 and 12 present us with a difficult conundrum. The results of our earlier experiments, discussed in Sections 4 and 6, were consistent at every point with the hypothesis that lesions placed in the regions of the nuclei of origin of the FCPS cause a deficit in radial maze performance in virtue of cholinergic denervation of the relevant target areas, and that fetal grafts placed in these target areas must contain cholinergic neurones if they are to reverse the radial maze deficit. The deficit seen in the radial maze in the Bradbury et al. experiment was the result of lesions similar to those made in the earlier experiments, and took a similar behavioural form. Yet the neuronal component of the same cholinergic-rich tissue which provided behaviourally effective grafts in the earlier experiments and, in this one, completely failed to improve performance. Given that the glial component of this tissue did provide a behaviourally effective graft, a natural inference might be that all our earlier results were due to glia-induced repair of host tissue. However, this interpretation of our results raises a further question: why did the hippocampal fetal tissue, used as a control in the earlier experiments, fail to provide a behaviourally-effective graft? Conceivably, glia from the fetal basal forebrain are enriched, relative to glia from the fetal hippocampus, in growth or other factors that facilitate repair of the damaged brain. But, in that case, we might have expected fetal basal forebrain, and not hippocampal, tissue to provide a behaviourally effective graft also after ischaemic or colchicine-induced damage to the hippocampal formation. Yet our results ( Section 9) were exactly the reverse. Thus, if the results of our experiments using fetal grafts were all due to the action of the glial components of these grafts, it must follow that only glia from an appropriate fetal brain region are capable of producing the factor(s) needed by a particular type of damage in the adult host brain: basal forebrain glia for cholinergic damage, CA1 cell field glia for CA1 damage, etc. While this hypothesis of region-specific glial activity cannot be ruled out, we regard it as inherently implausible. On the other hand, there is a lack of parsimony in supposing that the behavioural efficacy of undivided basal forebrain grafts observed by Arendt et al. (1989) and Hodges et al. (1991a), and that of the glial component of such grafts observed by Bradbury et al. (submitted) is due to different mechanisms in the two cases. Only further research can resolve this dilemma.
With regard to the mode of action of hippocampal subfield grafts, the situation appears clearer (though this may simply reflect the fact that they have come on the scene too recently for there yet to be many data). We have summarised above the anatomical evidence that such grafts are able to substitute for damaged host pyramidal or granule cells and to reinstate essentially normal connections (Dawe et al., 1993; Field et al. 1991; TQnder et al. 1990); and our behavioural data, indicating as they do the requirement for homotypic cell replacement if cognitive recovery is to occur, are fully consistent with this mode of action. If this inference is correct, hippocampal subfield grafts apparently induce cognitive recovery in a way that is quite different from the mechanisms -- whatever they eventually turn out to be -- that underlie the efficacy of cholinergic-rich grafts.
12. Therapeutic prospects
There is clearly much to be done before neural transplant surgery becomes a viable treatment for human cognitive deficits; and there are many pitfalls attached to transplant surgery (e.g., Sinden et al. 1992b; and see Stein & Glasier, this volume). Nonetheless, given that the effects of transplants on cognitive function appear to be as robust as their effects on motor function, and that the latter have already led to some, albeit limited, clinical success (Lindvall et al. 1992), it is reasonable to suppose that clinical applications will also be found in the cognitive field. In our view, grafting after global cerebral ischaemia has several features that make it a promising early target for such application. These features can be brought out by a comparison with neurodegenerative disorders.
First, in neurodegenerative disorders, such as Alzheimer's disease, many systems are disrupted and pathology is widespread, so that grafts based on the delivery of particular neurotransmitters or tissue from a limited brain region are unlikely to be fully effective. Pharmacological approaches using, for example, neurotrophins (Thoenen et al., this volume) or gangliosides (Garofalo et al. 1993) may be sensible alternatives to grafts for long-term enhancement of the damaged FCPS. In contrast, damage after global ischaemia is at least sometimes circumscribed (Squire 1986). Such localised damage permits the use of small grafts; furthermore, the grafts need to be targetted to the area of damage, obviating the need to extend long processes. Moreover, the relevant areas of damage, in the hippocampus, can already be detected by current neuroimaging techniques (Squire et al. 1990), so that patients likely to benefit from surgery can be identified and the growth of grafts monitored. Second, patients with neurodegenerative disease are usually elderly and fragile in health and so likely to be at risk from transplant surgery, whereas victims of heart attack are often quite young and better able to withstand surgery, and may have many healthy years ahead of them. Third, animal models do not as yet mimic the defining pathology of Alzheimer's disease (though this may change rapidly with the development of transgenic models), so we do not know whether grafted neurones would integrate in brain regions containing plaques and tangles or, indeed, succumb to the disease process. Grafts in heart attack victims, in contrast, would be placed within an otherwise healthy brain.
A key issue that will need to be resolved if these prospects are to be realised, whether for victims of heart attack or other conditions, is the development of alternatives to fetal tissue as the donor material. The use of fetal tissue poses both well-known ethical problems and also major practical problems. The experiments reviewed in this paper demonstrate the critical importance of using fetal tissue that is appropriate to the damage one is attempting to repair: cholinergic-rich tissue for damage to the FCPS, CA1 cells after ischaemia, etc. Each such tissue type has its own developmental programme, which determines the optimal time at which to dissect the tissue for transplantation and appropriate behavioural effects (e.g., Cassel et al. 1991). Even to establish the appropriate time-table for the relevant developmental programmes in the human brain, let alone to set up an ethically acceptable supply of human donor tissue, will present formidable difficulties. A further problem is that of donor-host genetic immunohistoincompatibility. Consequently, much recent effort has gone into the development of cells for grafting that are less immunologically provocative, including, as we have seen, neuroblastoma cell lines chemically treated to promote differentiation; potentially autologous cells from myenteric plexus; autologous fibroblasts, which may be transfected so as to express desired characteristics; and cells encapsulated in polymer to allow diffusion of low molecular weight substances, but otherwise minimal contact between graft and host (Sinden et al., 1992b; and see Neuwelt et al., this volume).
An interesting possibility is that application of growth factors may stimulate the de novo growth of cells of neuronal morphology in situ. Thus, two groups have recently identified neuronal cells arising from dividing precursors in dissociated cell populations from adult brain when stimulated with mitogenic growth factors (Richards et al. 1992; Reynolds & Weiss 1992). If the signals for the proliferation and differentiation of these precursor cells can be identified, the possibility would exist for replacement of damaged cells by host cells within the area of damage. An alternative approach lies in the development of neuroepithelial stem cell lines from primordial hippocampus which differentiate upon transplantation into identifiable neurones or glia, with phenotypes and connectivity appropriate to the site of implantation in the developing host brain (Renfranz et al. 1991; Snyder et al. 1992). It is not yet known whether these cells will also be sensitive to signals arising from damage in the adult brain; if so, they may be capable of adopting different phenotypes depending upon the nature of the host brain damage. These or other cells may also be transfected with a temperature-sensitive oncogene or derived from a transgenic line expressing this oncogene (Jat et al. 1991), so permitting them to proliferate readily in culture but to differentiate upon transplantation into brain. These developments hold out exciting possibilities of donor tissue which would be readily available in vitro, with low potential to evoke immunological rejection, able to take on multiple appropriate phenotypes, and into which it is possible to introduce additional desired genes.
Abdulla, F.A., Calaminici, M.R., Raevsky, V.V., Sinden, J.D., Gray, J.A. & Stephenson, J.D. (in press) An iontopheretic study of the effects of AMPA lesions of the nucleus basalis magnocellularis on cholinergic and GABAergic influences on frontal cortex neurones of rats. Experimental Brain Research.
Alheid, G.F. & Heimer, L. (1988) New perspectives in basal forebrain organization of special relevance for neuropsychiatric disorders: the striatopallidal, amygdaloid, and corticopetal components of substantia innominata. Neuroscience 27:1-39.
Allen, Y.S., Marchbanks, R.M. & Sinden, J.D. (1988) Non-specific effects of the putative cholinergic neurotoxin ethylcholine mustard aziridinium ion in the rat brain examined by autoradiography, immunocytochemistry and gel electrophoresis. Neuroscience Letters 95:69-74.
Amaral, D.G. & Kurz, J. (1985) An analysis of the origins of the cholinergic and noncholinergic septal projections to the hippocampal formation of the rat. Journal of Comparative Neurology 281:337-361.
Anderson, K.J., Gibbs, R.B. & Cotman, C.W. (1988) Transmitter phenotype is a major determinant in the specificity of synapses formed by cholinergic neurons transplanted to the hippocampus. Neuroscience 25:19-25.
Anderson, K.J., Gibbs, R.B., Salvaterra, P.M. & Cotman, C.W. (1986) Ultrastructural characterisation of identified cholinergic neurons transplanted to the hippocampal formation of the rat. Journal of Comparative Neurology 249:279-292.
Apostol, G. & Creutzfeldt, O.D. (1974) Cross-correlation between the activity of septal units and hippocampal EEG during arousal. Brain Research 67:65-75.
Arendt, T., Allen, Y., Marchbanks, R., Schugens, M.M., Sinden, J., Lantos, P.L. & Gray, J.A. (1989) Cholinergic system and memory in the rat: effects of chronic ethanol, embryonic basal forebrain transplants and excitotoxic lesions of cholinergic basal forebrain projection systems. Neuroscience 33:435-462.
Arendt, T., Allen, Y., Sinden, J., Schugens, M.M., Marchbanks, R.M., Lantos, P.L. & Gray, J.A. (1988a) Cholinergic-rich brain transplants reverse alcohol-induced memory deficits. Nature 332:448-450.
Arendt T, Bigl V, Arendt A, Tennstedt A (1983) Loss of neurons in the nucleus basalis of Meynert in Alzheimer's disease, paralysis agitans and Korsakoff's disease. Acta Neuropathologica 61: 101-108.
Arendt, T., Hennig, D., Gray, J.A. & Marchbanks, R.M. (1988b) Loss of neurons in the rat basal forebrain projection system after prolonged intake of ethanol. Brain Research Bulletin 21:563-570.
Baddeley, A.D. (1986) Working Memory. Oxford: Clarendon Press.
Bartus, R.T., Dean, R.L., Beer, B. & Lippa, A.S. (1982) The cholinergic hypothesis of geriatric memory dysfunction. Science 217:408-417.
Baudry, M. & Davis, J.L. (1991) Long-term potentiation: A debate of current issues Cambridge, Mass.: The MIT Press.
Bayer, S.A. (1980) Development in the hippocampal region in the rat -1. Neurogenesis examined with 3H-thymidine autoradiography. Journal of Comparative Neurology 190: 87-114.
Bjrklund, A., Stenevi, U., Schmidt, R.H., Dunnett, S.B. & Gage, F.H. (1983) Intracerebral grafting of neuronal cell suspensions. 1. Introduction and general methods of preparation. Acta Physiologica Scandinavica Suppl. 522:1-7.
Boksa, P. (1985) Acetylcholine synthesis by adult bovine adrenal chromaffin cell cultures. Journal of Neurochemistry 45:1254-1261.
Bowen, D.M., Smith, C.B., White, P. & Davison, A.N. (1976) Neurotransmitter-related enzymes and indices of hypoxia in senile dementia and other abiotrophies. Brain 99:459-496.
Bradbury, E., Kershaw, T., Marchbanks, R. & Sinden, J. (submitted) Astrocyte transplants alleviate lesion induced memory deficits independently of cholinergic recovery.
Brckner, M. & Arendt, T. (1992) Intracortical grafts of purified astrocytes ameliorate memory deficits in rat induced by chronic treatment with ethanol. Neuroscience Letters 141:251-254.
Butcher, L.L. & Woolf, N.J. (1984) Histochemical distribution of acetylcholinesterase in the central nervous system: clues to the localization of cholinergic neurons. In Handbook of Chemical Neuroanatomy. Vol. 3: Classical Transmitters and Transmitter Receptors in the CNS 3, Part II. Eds. A. Bjrklund & T. Hkfelt, Amsterdam, Elsevier, 1-50.
Buzsaki, G., Bickford, R.G., Ponomareff, G., Thal, L.J., Mandel, R. & Gage, F.H. (1988) Nucleus basalis and thalamic control of neocortical activity in the freely moving rat. Journal of Neuroscience 8: 4007-4026.
Buzsaki, G. & Gage, F.H. (1988) Mechanisms of action of neural grafts in the limbic system. Canadian Journal of Neurological Sciences 15:99-105.
Buzsaki, G., Gage, F.H., Czopf, J. & Bjrklund, A. (1987) Restoration of rhythmic slow activity (theta) in the subcortically denervated hippocampus by fetal CNS transplants. Brain Research 400:334-347.
Buzsaki, G., Wiesner, J., Henriksen, S.J. & Gage, F.H. (1989) Long-term potentiation of evoked and spontaneous neuronal activity in the grafted hippocampus. Experimental Brain Research 76: 410-408.
Calaminici, M., Abdulla, F.A., Sinden, J.D. & Stephenson, J.D. (1993) Direct evidence for axonal outgrowth from cholinergic grafts to cholinergically-deafferented rat cortex. Neuroreport 4:585-587.
Casamenti, F., Scali, C., Vannucchi M.G., Bartolini, L. & Pepeu, G. (1993) Long-term ethanol-consumption by rats - effect on acetylcholine-release in-vivo, choline-acetyltransferase activity, and behavior. Neuroscience 56:465-471.
Cassel, J.C., Kelche, C., Peterson, G.M., Ballough, G.P., Goepp, I. & Will, B. (1991) Graft-induced behavioral recovery from subcallosal septohippocampal damage in rats depends on maturity stage of donor tissue. Neuroscience 45: 571-586.
Clarke, D.J. (1985) Cholinergic innervation of the rat dentate gyrus: an immunocytochemical and electron microscopical study. Brain Research 360:349-354.
Clarke, D.J. & Dunnett, S.B. (1986) Ultrastructural organisation of the choline-acetyltransferase-immunoreactive fibres innervating the neocortex from embryonic ventral forebrain grafts. Journal of Comparative Neurology 250:192-205.
Clarke, D.J., Gage, F.H. & Bjrklund, A. (1986) Formation of cholinergic synapses by intrahippocampal septal grafts as revealed by cholineacetyltransferase immunocytochemistry. Brain Research 369:151-162.
Clarke, D.J., Nilsson, O.G., Brundin, P. & Bjrklund, A. (1990) Synaptic connections formed by grafts of different types of cholinergic neurons in the host hippocampus. Experimental Neurology 107:11-22.
Coffey, P.J., Perry, V.H., Allen, Y., Sinden, J. & Rawlins, J.N.P. (1988) Ibotenic acid induced demyelination in the central nervous system: a consequence of a local inflammatory response. Neuroscience Letters 84:178-184.
Collerton, D. (1986) Cholinergic function and intellectual decline in Alzheimer's disease. Neuroscience 19:1-28.
Davis, H.P. & Volpe, B.T. (1990) Memory performance after ischemic or neurotoxin damage of the hippocampus. In: The biology of memory, L.R. Squire & E. Lindenlaub ed. Stuttgart, New York: Symposia Medica Hoechst 23, FKSchattauer Verlag, 477-507.
Dawe, G.S., Gray, J.A., Sinden, J.D., Stephenson, J.D. & Segal, M. (1993) Evidence from extracellular recordings for the connectivities of fetal dentate gyrus and CA1 hippocampal subfield transplants to the colchicine-lesioned rat dentate gyrus. Brain Research 625:63-74.
Dawson, V.L., Gage, F.H., Hunt, M.A. & Wamsley, J.K. (1989) Normalization of subtype-specific muscarinic receptor binding in the denervated hippocampus by septodiagonal band grafts. Experimental Neurology 106:115-124.
Dubovik, V., Faigon, M., Feldon, J. & Michaelson, D.M. (1993) Decreased density of forebrin cholinergic neurons in experimental autoimmune dementia. Neuroscience 56:75-82.
Dunnett, S.B. (1985) Comparative effects of cholinergic drugs and lesions of the nucleus basalis and fimbria-fornix on delayed matching in rats. Psychopharmacology 87:357-363.
Dunnett, S.B. (1990) Neural transplantation in animal models of dementia. European