Disconnecting hippocampal projections to the anterior thalamus produces de cits on tests of spatial memory in rats

Transcription

1 European Journal of Neuroscience, Vol. 12, pp. 1714±1726, 2000 Ó European Neuroscience Association Disconnecting hippocampal projections to the anterior thalamus produces de cits on tests of spatial memory in rats E. C. Warburton,* A. L. Baird, A. Morgan, J. L. Muir and J. P. Aggleton School of Psychology, University of Cardiff, Wales CF10 3YG Keywords: anterior thalamus;hippocampus;recognition, spatial memory Abstract A disconnection procedure was used to test whether projections from the hippocampus to the anterior thalamic nuclei (AT), via the mbria-fornix (FX), form functional components of a spatial memory system. The behavioural effects of combined unilateral lesions in the AT and FX were compared when they were either in contralateral hemispheres (AT-FX Contra) or the same hemisphere (AT-FX Ipsi). Other groups received bilateral FX lesions and Sham surgeries. Expt 1 demonstrated that none of these lesions affected performance of an object recognition task, while performance of an object location task, which tests the subjects' preference for an object that has changed location, was impaired in the AT-FX Contra and FX groups. In a T-maze alternation task, however, the FX group was severely impaired while both the AT-FX Ipsi and AT-FX Contra lesion groups showed only a mild impairment. In order to test whether spared crossed projections might support spatial performance in the AT-FX Contra group we then examined the effects of a combined AT-FX Contra lesion coupled with transection of the hippocampal commissure. This combination of lesions produced a severe disruption in spatial memory performance in the water maze, radial arm maze and T-maze, which was signi cantly greater than that produced by ipsilateral and contralateral AT-FX lesions alone. These results support the notion that disconnection of the AT from their hippocampal inputs produces impairments on a range of spatial memory tasks, but indicate that there are an array of different routes that can subserve this function. Introduction The mbria-fornix (FX) is one of the major routes by which afferent and efferent information reaches the hippocampus. It is therefore not surprising that in rats cutting the FX has been shown to disrupt tests of spatial memory that are also impaired by hippocampectomy (Gray & McNaughton, 1983; Barnes, 1988; Aggleton et al., 1991). However, the identi cation of the critical pathways responsible for the similar spatial de cits following hippocampal and FX lesions has not proved straightforward. While most attention has focused on the cholinergic inputs from the basal forebrain and medial septum to the hippocampus, a quite different set of fornical connections that may be of critical importance are those from the hippocampus to the anterior thalamic nuclei (AT; Aggleton & Brown, 1999). These include the indirect connections via the mammillary bodies. These diencephalic projections principally arise from the subicular complex (Swanson & Cowan, 1977; Aggleton et al., 1986) which with the hippocampal elds CA1±4 and the dentate gyrus, comprise the `hippocampus'. While there are reciprocal connections from the AT to the hippocampus (Shibata, 1993) these use a different pathway, i.e. the cingulum bundle (Mufson & Pandya, 1984). The present study therefore sought to test Correspondence: Dr E. C. Warburton, at present address below.* e.c.warburton@bristol.ac.uk *Present address: Department of Anatomy, University of Bristol, School of Medical Sciences, University Walk, Bristol BS8 1TD, UK. Received 9 August 1999, revised 24 January 2000, accepted 25 January 2000 the functional contribution of the hippocampal±anterior thalamic connections served by the FX. Like the hippocampus, the AT appear to play an important role in spatial memory. As evidence of this, lesions of the AT impair a similar array of spatial tasks as those disrupted after hippocampectomy. These include place nonmatching in the T-maze and the radial arm maze, navigation in the Morris water maze, and delayed nonmatching to position in an automated test chamber (Aggleton & Sahgal, 1993; Byatt & Dalrymple-Alford, 1996; Warburton et al., 1997; Warburton & Aggleton, 1999). In spite of the similarities between these lesion effects, there is no direct evidence concerning the functional nature of the hippocampal projections to the AT via the FX. We tested this directly in two sets of disconnection studies. The rst study compared the effects of unilateral lesions placed in the FX and the contralateral AT, with unilateral lesions placed in the FX and the ipsilateral AT. The rationale behind such experiments is that unilateral lesions in different sites in opposite hemispheres will only have a severely disruptive effect if the two regions form part of a functional system, whereas unilateral lesions made in the same hemisphere will have relatively little effect on behaviour. Thus, if the hippocampal projections to the AT via the FX are critical for spatial memory we would predict that the former (crossed) lesions will be more disruptive than the latter (ipsilateral) lesions. The second study used the same approach but the unilateral fornix lesion was extended caudally to include the hippocampal commissure. This was in response to evidence that this pathway can support cross hemispheric function and so attenuate the effects of hippocampal system disconnections (Olton et al., 1982).

2 Disconnecting hippocampal-thalamic connections 1715 Materials and methods Expt 1. Crossed lesions of the FX and AT In this experiment the postoperative performance of four groups of rats was compared. The lesion groups were as follows: surgical controls (Sham); bilateral FX lesions; unilateral FX and unilateral anterior thalamic lesions in the ipsilateral hemisphere (AT-FX Ipsi); and unilateral fornix and unilateral anterior thalamic lesions in the contralateral hemisphere (AT-FX Contra). All experiments were conducted in accordance with the United Kingdom Animals (Scienti c Procedures) Act Animals Fifty- ve male rats of the pigmented DA strain (Bantin and Kingman, Hull) were used. Throughout the period of the experiment the animals were housed in pairs under diurnal conditions (14 h light : 10 h dark). At the start of testing the animals were aged 4 months and weighed between 215 and 230 g. Surgery Prior to surgery all animals were deeply anaesthetized by intraperitoneal injection of 60 mg/kg of pentobarbitone sodium (Sagatal) and then placed in a stereotaxic headholder (David Kopf Instruments, Tujunga, California) with the nose bar at The scalp was then cut and retracted to expose the skull. Craniotomies were then made directly above the target regions and the dura cut to expose the cortex. Rats in the disconnection groups received unilateral lesions of the AT combined with a lesion of the FX. For half the animals, the contralateral group, the AT and FX lesions were placed in opposite hemispheres (AT-FX Contra). For the other half, the ipsilateral group, the AT and FX lesions were in the same hemisphere (AT-FX Ipsi). Lesions of the AT were made by injecting 0.15 ml 0.09 MN-methyl- D-aspartic acid (NMDA; Sigma Chemical Company Ltd, UK) dissolved in phosphate buffer (ph 7.2) through a 1 ml Hamilton syringe into three sites in the hemisphere. Each injection was performed gradually over a 5-min period and the needle remained in situ for a further 5 min before withdrawal. The stereotaxic coordinates relative to ear-bar zero (in mm), with the incisor bar set at 5.0 mm above the interaural horizontal plane were: (i) AP, +5.2, LAT, 6 1.0, with DV, 6.2 mm below the top of the cortex; (ii) AP, +5.2, LAT, 6 1.7, with DV, 5.6 mm below the top of the cortex; (iii) AP, +4.5, LAT, 6 2.3, with DV, 5.0 mm below the top of the cortex. Radiofrequency lesions of the FX were made using an RFG4-A Lesion Maker (Radionics, Burlington). The electrode (0.3 mm tip length, 0.25 mm diameter) was lowered vertically and at each site the temperature of the tip was raised to 72 C for 60 s. The coordinates used relative to ear bar zero were: (i) AP +5.3; LAT 6 0.7; DV 3.7 mm (below top of cortex); (ii) AP +5.3; LAT 6 1.7; DV 3.8 mm (below top of cortex). The bilateral FX lesions were made using the same coordinates. Sham control lesions of these structures were made using the identical procedure as described above but in these cases the injection needle and radiofrequency electrode were lowered to the level of the AT and FX, respectively, and then removed. The position of the craniotomies for the anterior thalamic, FX and sham surgeries were the same. At the completion of surgery the skin was sutured and an antibiotic powder (Acramide; Dales Pharmaceuticals, Skipton) applied. All animals then received 5 ml glucose saline (s.c.) containing etamiphylline (Millophyline; Arnold's, Romford; 35 mg/kg, s.c.), a cardiac stimulator. Postoperative care also included systemic analgesia (Temgesic; Reckett and Colman, UK). On completion of the experiment all animals were killed with an overdose of Euthatal and perfused intracardially with saline followed by 10% formol-saline. The brains were then removed and placed in 10% formol-saline for a minimum of two hours. Following xation the brain was transferred to 20% sucrose in 0.2 M phosphate buffer and left overnight. The brain was then cut on a freezing microtome into 60 mm coronal sections. One in two sections were mounted and stained with cresyl violet, a Nissl stain. Apparatus Spontaneous object recognition. The apparatus consisted of an open box ( cm high) made of wood, the inside of which was painted grey. The oor was covered with sawdust. Object recognition. Triplicate copies were made of the complex objects to be discriminated, which were made either of glass, wood, plastic or metal. The height of the objects ranged from 10±18 cm, while their weight ensured that they could not be displaced by the rats. Care was taken to ensure that the pairs of objects being tested were composed of the same material so that they could not readily be distinguished by olfactory cues although they had very different appearances. Object location. In this test four identical copies of the objects were used. In the rst test, the objects were identical empty coffee jars, i.e. glass with a red plastic lid. In the second test, the objects were identical unopened drink cans. Neither set of objects had been used before in any previous test. Spatial forced alternation. All testing for the forced alternation task was carried out in a modi able T-maze. The oors of the maze were 10 cm wide and made of wood, and the walls were 17 cm high and made of clear Perspex. The stem was 70 cm long with a guillotine door located 25 cm from the end of the stem, so creating a start area. The cross-piece was 140 cm long, and at each end there was a food well 2 cm in diameter and 0.75 cm deep. The entire maze was supported by two stands 94 cm high. Lighting was provided by a uorescent light suspended 164 cm above the apparatus. Behavioural procedures Spontaneous object recognition. Rats began training on average 2 weeks after surgery. Prior to the start of the object recognition and object location tests, all rats were given three habituation sessions, in which the rat was placed in the empty test arena for 6 minutes. Forty-eight hours later, testing began. Rats were then given a total of two test sessions, with a minimum interval of 48 h between each session. In none of these tests did the experimenter know the identity of the animals. The time spent exploring objects during the various tests was assessed from video recordings of the sample and test phases. Throughout these tests the rats received food and water ad libitum. Object recognition. Each test session consisted of two phases. In the initial sample phase, two identical objects (A1 and A2) were placed in the far corners of the box arena each 10 cm from the side wall. A rat was then placed in the middle of the arena and the total time spent exploring the two objects was determined from video taped recordings. Exploration of an object was de ned as directing the nose to the object at a distance of less than 2 cm and/or touching it with the nose. Turning around or sitting on the object was not considered as exploratory behaviour. After a delay of 15 min the rat was re-introduced to the arena (`choice phase'). The box now contained a third identical copy of the familiar object (A3) and a new object (B). These were placed in the same locations as the sample stimuli. The location of the two choice objects was counterbalanced

3 1716 E. C. Warburton et al. between rats and across sessions. Further, the objects used were counterbalanced, i.e. for half of the rats object A was presented in the sample phase and object B introduced in the choice phase and for half the rats object B was used in the sample phase and object A introduced in the choice phase. As far as could be ascertained the objects had no natural signi cance for the rats and they had never been associated with a reinforcer. The sample phase ended as soon as the test animal had explored the sample objects for a total of 25 s. This procedure was used in order to equate the sample exposure between animals. The choice phase, however, was held constant at three minutes duration. Each rat was given two sessions with different sets of objects. Object location. In this test the rats were not tested on their ability to distinguish the novel from the familiar object. Instead, we assessed the rat's ability to recognize that an object it had experienced previously had changed location. Two copies of the same object (A) were used in each test. In the sample phase the rat was exposed to objects A1 and A2 which were placed in the far corners of the arena (as in the object recognition test). The animal was allowed to explore both objects during a sample phase of three minutes, and the amount of exploration of each object was recorded by the experimenter. After a delay of 5 min the test phase began. In the test phase, object A3 was placed in the same position as object A1 had occupied in the sample phase. Object A4 was placed in the corner adjacent to the original position of A2, so that the two objects A3 and A4 were in diagonal corners. Thus both objects in the test phase were equally familiar, but one was in a new location. All animals received two test sessions separated by 48 h. The position of the moved object was counterbalanced for each rat. Spatial forced alternation. Three days after the completion of the object location test, all animals were food-deprived to 85% of their free feeding body weight by restricting their daily food intake to approximately 15 g of laboratory diet (Harlan Teklad, Bicester, Oxfordshire). Prior to the start of testing, each animal was then given several days of pretraining in the T-maze apparatus to ensure that the rat would run reliably down the stem of the maze to nd food pellets in the food wells in both arms. At the start of each trial, which consisted of two stages, three food pellets (45 mg Campden Instruments, Loughborough) were placed in each food well and a metal barrier was placed at the neck of the T-maze, so closing off one arm. As a consequence, the animal was forced to enter a preselected arm on each `sample run' and then allowed to eat the food there. The animal was then removed and con ned in the start box for a delay of 10 s, during which the metal barrier was removed. The door to the start box was then opened and the animal allowed a free choice between the two arms of the T-maze. On this `choice run' the criteria for selecting an arm consisted of the rat placing a back foot in one of the arms. No retracing was permitted. If the rat had alternated, i.e. had entered the arm not previously visited on the 'sample run', it was allowed to eat the food reward before being returned to its cage. If the other arm was chosen, i.e. the same arm as visited on the 'sample run', the rat was con ned to that arm for approximately 10 s, and then returned to its cage. The rats were tested in groups of three or four with each rat having one trial in turn, so that the intertrial interval was approximately 4 min. The animals received six trials a day, for a total of 15 sessions. Statistical analysis T-maze. The scores from the T-maze were grouped into ve blocks, each of three sessions. Acquisition of the T-maze task was analysed using a repeated measures analysis of variance (ANOVA) conducted on these blocks of (maximum of 18 correct responses per session). Object recognition: exploration during sample period. The duration of the sample period (i.e. the total length of time it took the animals to explore the objects for 25 s) was analysed using ANOVA. Object recognition during the test period. Total exploration times were calculated for the novel and familiar objects summed across two matching pairs of test sessions. The exploration times for the rst minute and rst 2 minutes of the test session were calculated. The data for the rst 2 minutes were used, as opposed to the data from all three minutes, as it as been shown that discrimination between the novel and familiar objects decreases considerably during the third minute (Dix & Aggleton, 1999). From the results the absolute difference in the time spent exploring the novel and familiar objects was calculated (d1) and the proportion of the total exploration time spent exploring the novel object (d2), i.e. d1 divided by the total time spent exploring the objects. The d2 index takes into account individual differences in the total amount of exploration. Finally, the 95% and 98% con dence limits were calculated to determine whether the group d1 and d2 scores were above zero, and so determine if the individual groups displayed a signi cant preference of the novel stimuli. Expt 2. Crossed lesions of the FX, hippocampal commissure and AT In this experiment the postoperative performance of ve groups of rats was compared. Four lesion groups were the same as those described in Expt 1; however, an additional lesion group in which the rats received a unilateral fornix and unilateral anterior thalamic lesion in the contralateral hemisphere coupled with the transection of the hippocampal commissure (AT-FX-HC) on the same side as the fornix lesion was included. Subjects Sixty-two naive male DA rats, of an age and weight similar to those used in Expt 1 and from the same supplier, were housed under the same conditions. Surgery The general surgical procedures were identical to those described for Expt 1. In the AT-FX-HC group, the FX and anterior thalamic lesions were produced the same way as the AT-FX Contra group, i.e. the anterior thalamic lesions were produced by injection of NMDA (0.09 M) into three sites in the hemisphere, while transection of the FX was achieved using a radiofrequency lesion. The additional lesion of the hippocampal commissure involved lowering the radiofrequency probe caudally at the following additional coordinates on the same side as the fornix lesion: AP +4.5 (from ear bar zero); LAT 0.7; V 3.5 mm (below the top of the cortex) and raising the tip temperature to 72 for 60 s. Seven of the 11 animals in the AT-FX Ipsi group also had a lesion in the hippocampal commissure, but in the same hemisphere as both the anterior thalamic and FX lesions. Although the commissure lesion was placed unilaterally, it was intended to reach the midline and so disrupt all of the tract. The postoperative care and all histological procedures were also identical to those described for Expt 1. Apparatus Spatial forced alternation. See description for Expt 1. Morris water maze. The water maze used in this experiment was a 2 m diameter white breglass pool, 60 cm high and mounted 58 cm above the oor. The pool was situated in a room which contained posters on the wall. Other distal cues included a curtain used to

4 Disconnecting hippocampal-thalamic connections 1717 conceal the experimenter. Lighting was provided by four oormounted spotlights (500 W) placed in each corner of the room. The water in the maze was made opaque by the addition of 2 L of milk. An escape platform was placed 2 cm beneath the water surface and kept in a constant position in the pool during the acquisition trials. The temperature of the water was 25 C at the beginning of each testing period. The swim paths of the rats were tracked using a video camera suspended directly above the pool and all sessions were recorded on video tape. Data were collected and analysed online using an HVS image analyser connected to an Archimedes RISC computer using Watermaze software (Edinburgh, UK). Radial arm maze. The radial maze was constructed from a wooden centre piece measuring 34 cm in diameter, to which were attached eight identical, 90 cm long and 10 cm wide arms. The walls of the centre piece and arms were made of Perspex. The height of the walls was 15 cm, and the height of the wall surrounding the centre piece was 29 cm. Eight guillotine doors, also made of Perspex, separated the arms from the centre piece, and were opened using a pulley system attached to a circular piece of wood above the centre piece to which the doors were attached. The oor of the apparatus was painted white. The room was lit using uorescent overhead lights, and contained several prominent cues, such as posters and the door. The doors were operated by the experimenter, who sat in the same location on every occasion. At the end of each arm was a hidden food well 2 cm in diameter, and 0.5 cm deep. Behavioural procedures Spatial forced alternation. Food deprivation prior to T-maze testing began 2±3 weeks after surgery. The habituation sessions began a week later. Habituation and the subsequent 15 testing sessions were conducted as described for Expt 1. Morris water maze. Following the completion of the T-maze alternation testing, animals were returned to a free-feeding regime and 5 days later, water maze acquisition training began. Eight start positions (N, S, E, W, NE, NW, SE, SW) were allocated relative to the experimenter. Each rat was placed in the pool facing the wall at one of the eight start locations. The rat was required to swim to an underwater platform, which was positioned in the same quadrant throughout the acquisition sessions. Each group of animals was divided into four subgroups, and each subgroup was required to search for the platform in a different quadrant of the maze (SW, SE, NE, NW). Each of the four trials every day was started from a different location on the edge of the pool and the order in which the start locations were used varied from day to day. The start positions for the two groups of animals were also balanced. Thus, if a rat had to locate the platform in the SW quadrant and started the rst trial from the N position at the edge of the pool, a corresponding rat which had to locate the platform in the NE quadrant, started their rst trial from the S position. The testing regimen for all groups was as follows. Animals were initially tested with four trials per day on each of 10 consecutive days. Each acquisition trial was terminated either when the animal located the hidden escape platform or after 120 s had elapsed. If the rat located the hidden platform, it was allowed to remain on the platform for 30 s. Following an intertrial interval of 2 min, the rat was then placed in the pool at the second start location, and so on for four trials. If the rat failed to nd the platform after 120 s it was placed on the platform and allowed to remain on it for 30 s. Following these 10 sessions, animals received a spatial probe trial on day 11. All rats were rst given one normal acquisition trial in which the platform was located in the training quadrant as usual. Then the platform was removed from the maze, and the swim path and distance swum in each of the four quadrants was recorded over 60 s. For this probe trial, each rat was placed at a start position directly opposite to where the platform had been located. For example, if the platform had been located in the SW quadrant, the rat was placed in the water, facing the wall, at the NE position. Radial arm maze. On completion of the water maze probe trial, the animals were again placed on a restricted diet, until they had achieved 85% of their free-feeding weight. Radial arm maze training began 15 days after the water maze probe trial. All animals received two habituation sessions (5 min) in the radial arm maze during which time each rat was placed into the centre of the maze and allowed to move freely around the maze until either it had eaten all available reward pellets or 5 min had elapsed. On the subsequent test trials, all eight arms were baited with two food pellets. At the beginning of a trial the rat was placed in the centre of the maze with the doors to the arms closed, the doors were opened and the rat allowed to chose an arm. When the rat had retrieved the food pellets it returned to the centre of the maze, and the arm doors were closed con ning the rat to the centre of the maze for a 10 s interval. The doors were then re-opened and the rat allowed to visit another arm. If an incorrect choice was made, i.e. an arm chosen which had been previously visited, the animal was con ned to that arm for a 5 s interval without reward before being allowed to return to the centre. The test session ended once the rat had visited all the eight arms, or a maximum of 10 min had elapsed. The rats were tested for 18 days. Statistical analysis Escape latencies were used as a measure of acquisition of the water maze task. These were compared in a repeated measures ANOVA using the means of each set of four trials from the 10 acquisition sessions. In addition the swim speeds of the animals (m/s) during acquisition, and the percentage of time the rats spent within 16.5 cm of the side walls of the maze, were examined. For the probe test, ANOVA was used to compare the percentage of time animals spent in the quadrant in which the escape platform had previously been located. Performance of the radial arm maze task was analysed using a repeated measures ANOVA conducted on (i) the total number of trials taken by the animals to obtain all eight reward pellets; (ii) the number of correct choices made by the animals in the rst eight trials (maximum number, 8). Results Expt 1: histology From the histological analysis the nal group numbers were as follows: Sham, n = 15; AT-FX Contra, n = 7; AT-FX Ipsi, n = 6; and FX, n =9. Fornix The fornix suffered extensive bilateral damage in the nine cases that were retained (one case was rejected as there was sparing of over 50% of the tract in one hemisphere). The bilateral fornix damage was complete in two of the nine cases, while in the remaining seven cases there was sparing of up to 20% of the tract. In all but one case the sparing was at the lateral tips of the tract. In four cases there was additional damage to the most dorsal limit of the AT, and in seven cases the lesion extended dorsally to include parts of the overlying cingulate cortex (three cases bilateral) and the cingulum bundle (four cases bilateral). The largest and smallest lesions are shown in Fig. 1.

5 1718 E. C. Warburton et al. fornical damage did not encroach into the head of the hippocampus, and so did not reach the level of the hippocampal commissure. In one case the very dorsal margin of AV was damaged on the same side as the fornix lesion. All seven cases showed near total cell loss in the AT, and the only region of sparing was in the most medial portion of AM (in two cases). The dorsal and rostral portions of LD also showed cell loss in three cases, and in two cases the lesion reached nucleus reuniens but did not cross the midline. In four cases there was a very restricted patch of cell loss in the granule cells in the medial (outer) blade of the dentate gyrus immediately above rostral LD (Fig. 2). Expt 1: behaviour Object recognition: exploration during sample period The duration of the sample period was considered rst. The analysis revealed that there were no signi cant differences between the lesion groups (F 3,33 < 1.0). The mean sample times were as follows Sham s; AT-FX Contra s; AT-FX Ipsi s; FX s. FIG. 1. Diagrammatic reconstructions showing the cases with the largest (grey) and smallest (solid black) lesions in the bilateral fornix group (FX) from Expt 1. The numbers refer to the plate numbers from Paxinos & Watson (1997). Anterior thalamic + fornix ipsilateral Animals in which there was evidence of bilateral thalamic damage, or in which there was insuf cient damage in either the AT or the fornix were excluded. In the resulting six cases the unilateral fornix lesions were highly selective in that the tract was completely severed in three cases, while approximately 10±20% of the lateral tip was spared in two cases, and 30% of the lateral tip of the tract was spared in one case. No other damage was observed, except that in all cases the tract damage extended caudally so that it just included the head of the hippocampus. In four cases the neurotoxic lesion included all of the AT, and the rostral half of the lateral dorsal nucleus. In the remaining two cases the anterodorsal (AD) and anteroventral (AV) nuclei had largely disappeared but there was sparing of the anteromedial nucleus (AM). In all six cases there was restricted neurotoxic damage to granule cells in the medial (outer) blade of the dentate gyrus (Swanson et al., 1987) immediately above the thalamus. Five of the six cases showed cell loss in the lateral dorsal nucleus (LD). In all ve, this cell loss involved the rostral and dorsal edges of the nucleus, but in two animals the cell loss extended caudally to involve almost all of LD. In one case the rostral third of the medial dorsal nucleus (MD) was damaged. The cases with the largest and smallest combined lesions are shown in Fig. 2. Anterior thalamic + fornix contralateral A total of seven rats had substantial damage in both the AT and the contralateral fornix. As can be seen in Fig. 2, which depicts the largest and smallest of the lesions, the extent of damage was very consistent across cases. The FX was completely severed in four cases, and in three only the very lateral tip of the mbria was preserved. The Object recognition during the test period Analysis of variance using the d1 index for the rst two minutes of the object recognition test sessions revealed no evidence of a difference between the groups (F 3,33 < 1.0). The mean d1 scores were as follows Sham, 19.68; AT-FX Contra, 12.41; AT-FX Ipsi, 21.77; FX, Similar analysis for the d2 index also showed no difference between the groups, F 3,33 = 2.17, P > The mean d2 scores were as follows: Sham, 0.30; AT-FX Contra, 0.17; AT-FX Ipsi, 0.39; and FX Similarly, analyses of the d1 and d2 indices for the rst minute only, when d2 scores are typically highest did not reveal any group differences [for d1, F 3,33 < 1.0 (mean d1 scores: Sham, 14.23; AT-FX Contra, 11.77; AT-FX Ipsi, 22.19; and FX, 13.04) and for d2 F 3,33 = 1.88, P > 0.05 (mean d2 scores: Sham, 0.37; AT-FX Contra, 0.28; AT-FX Ipsi, 0.59; and FX, 0.23)]. Analysis of the con dence limits of the mean d1 and d2 indices showed that all groups were able to discriminate the novel from the familiar object (see Table 1). Object location: exploration during sample period In this procedure the sample period was held constant at 3 min, so the degree of exploration of the sample objects during that period was compared rst. ANOVA showed that there was a signi cant group difference (F 3,33 = 3.47, P < 0.05) as the AT-FX Ipsi group spent longer exploring the objects (means, 19.5 s) compared with the Sham group (mean, 13.8 s). There were no other group differences. (Means of the other groups: AT-FX Contra, 14.5 s; FX, 17.1 s). Object location: recognition during the test period The analysis of the results for the object location test took the same form as those for the object recognition, i.e. the d1 and d2 scores for the rst minute and rst two minutes were compared. Analysis of variance using the d1 index for the rst two minutes showed that the main effect of group just missed signi cance (F 3,33 = 2.65, P = 0.06). A similar result was obtained when the d2 scores were compared (F 3,33 = 2.38, P = 0.09). Post hoc analyses were carried out to compare the AT-FX and FX groups with the Sham group (see Howell, 1997 p.351). Unrelated t-tests (one-tailed) showed that the d1 scores of the AT-FX group differed signi cantly from that of the Sham group (t 20 = 1.97; P = 0.03) as did the d1 scores of the FX group (t 22 = 2.01; P = 0.29). The same pattern of results were obtained when the d2 scores were analysed (AT-FX/Sham t 20 = 2.34; P = 0.015; FX/Sham t 22 = 1.81; P = 0.04).

6 Disconnecting hippocampal-thalamic connections 1719 FIG. 2. Diagrammatic reconstructions showing the cases with the largest (grey) and smallest (solid black) lesions in the anterior thalamic-fornix ipsilateral lesion group (AT-FX Ipsi) and anterior thalamic-fornix crossed lesion group (AT-FX Contra) from Expt 1. The numbers refer to the plate numbers from Paxinos & Watson (1997).

7 1720 E. C. Warburton et al. TABLE 1. Discrimination in the two spontaneous object recognition tasks Sham AT-FX Ipsi AT-FX Contra FX d1 d2 d1 d2 d1 d2 d1 d2 Object recognition ** ** ** ** ** ** * * Object location ** ** ** ** NS NS NS NS Discrimination in the two spontaneous object recognition tasks by the Sham control group, the combined ipsilateral anterior thalamic and fornix lesion group (AT-FX Ipsi), the combined contralateral anterior thalamic and fornix lesion group (AT-FX Contra) and bilateral mbria-fornix lesion group (FX) based on con dence limits (95%, 98%). The discrimination indices, for the rst 2 min of the test period, are calculated as follows; d1, the absolute difference in the time spent exploring the novel and familiar objects was calculated; d2, the proportion of the total exploration time spent exploring the novel object. *P < 0.05, **P < 0.02, signi cant discrimination between the objects; NS indicates no signi cant discrimination. Analysis of the d1 and d2 indices for the rst minute only did not reveal any group differences [for d1, F 3,33 = 1.82 (mean d1 scores: Sham, 8.87; AT-FX Contra, ±1.54; AT-FX Ipsi, 9.51; and FX, 2.93) and for d2, F 3,33 = 1.32, P > 0.05 (mean d2 scores: Sham, 0.30; AT- FX Contra, ±0.02; AT-FX Ipsi, 0.21; and FX, 0.13)]. Examination of the con dence limits (95% and 98%) of the mean d1 and d2 indices for the rst two minutes showed that both the AT- FX Contra and FX groups failed to discriminate the object that had changed location, whereas the Sham and AT-FX Ipsi groups were clearly able to discriminate (see Table 1). Spatial forced alternation Figure 3 shows the postoperative performance of the Sham, AT-FX Contra, AT-FX Ipsi and FX groups in the T-maze alternation test. Analysis revealed a highly signi cant difference between the lesion groups (F 3,33 = 72.37, P < 0.01). Post hoc Newman Keuls analyses showed that the FX group (mean number of correct choices, 10.3) were signi cantly worse than the Sham (mean, 17.4), AT-FX Contra (mean, 16.1) and AT-FX Ipsi groups (mean, 16.0). In addition, the performance levels of the AT-FX Contra and AT-FX Ipsi groups were signi cantly poorer than the Sham group, but not signi cantly different from each other. The analysis of variance also revealed a signi cant lesion by session interaction (F 12,132 = 1.87, P < 0.05). Analysis of the simple main effects showed that the performance of the FX group worsened signi cantly over the training sessions. Expt 2: histology From the histological analysis the nal group numbers were as follows: Sham, n = 12; AT-FX Contra, n = 5; AT-FX-HC, n = 6; AT- FX Ipsi, n = 9; and FX, n = 11. Fornix All animals in the FX group suffered extensive bilateral damage to the fornix, but in no case was the tract completely severed. The extent of sparing ranged between 5% and 30% of the tract, and this sparing typically occurred in the lateral tips of the tract. In three cases there was bilateral damage to the dorsal margin of the anterior ventral thalamic nucleus. The largest and smallest lesions are shown in Fig. 4. Anterior thalamic + fornix ipsilateral A total of nine animals showed clear, but extensive, unilateral damage to the fornix and AT. In one case the fornix lesion extended dorsally to involve the cingulum bundle and cingulate cortex (Fig. 4), FIG. 3. T-maze alternation from Expt 1. The graph shows the mean percentage of correct responses (6 SEM) made by each of the four groups (Sham, AT-FX Contra, AT-FX Ipsi, and FX) in each block of three sessions (each test session comprised 6 trials). SEM values that are not visible on the gure are smaller than the symbol. but in all other cases it was essentially restricted to the tract and the adjacent hippocampus. In six cases, the fornix was completely cut ipsilaterally, while in the remaining three cases the most lateral 10% of the tract was spared. In seven cases the tract damage extended caudally to include the head of the hippocampus and the hippocampal commissure. In ve cases the neurotoxic lesion included essentially all of the AT, and extended into the rostral and dorsal portions of the lateral dorsal nucleus. In the remaining four cases AD and AV had largely disappeared but there was sparing of the middle and caudal portions of AM. All cases showed some additional, neurotoxic damage to the granule cells of the medial (outer) blade of the dentate gyrus, immediately above the thalamus. In two cases there was appreciable, additional thalamic damage, which in the most extensive case involved the rostral two thirds of MD as well adjacent portions of the paracentral nucleus. Anterior thalamus + fornix contralateral In a total of ve cases there was clear unilateral damage to both the fornix and the AT in contralateral hemispheres. The fornix lesions were complete in two cases, and spared 30% of the lateral portion of the tract in the case with the smallest lesion. The fornix lesion was essentially con ned to the tract and stopped at, or just rostral, to the head of the hippocampus. In one case the lesion just touched the dorsal margin of AV. The largest and smallest lesions are shown in Fig. 5. In four cases the neurotoxic lesion involved nearly all of the AT, with some sparing in the most medial AM (two cases). In the remaining case AD had completely disappeared but there was sparing of the middle and caudal portions of AM, and caudal AV. In all cases the lesion extended caudally into the dorsal and rostral parts of LD. The thalamic nucleus MD was spared in all cases, although one animal had limited damage to the paracentral nucleus. Four cases

8 Disconnecting hippocampal-thalamic connections 1721 FIG. 4. Diagrammatic reconstructions showing the cases with the largest (grey) and smallest (solid black) lesions in the bilateral fornix group (FX) and anterior thalamic-fornix ipsilateral lesion group (AT-FX Ipsi) from Expt 2. The numbers refer to the plate numbers from Paxinos & Watson (1997). showed discrete neurotoxic damage to the granule cells of just the medial (outer) blade of the dentate gyrus, immediately above the rostral LD nucleus. This was restricted to the most ventral and rostral portion of the dentate gyrus. Anterior thalamus + fornix contralateral + hippocampal commissure In a total of six cases there was clear unilateral damage to both the fornix, the region of the hippocampal commissure, and to the AT in

9 1722 E. C. Warburton et al. FIG. 5. Diagrammatic reconstructions showing the cases with the largest (grey) and smallest (solid black) lesions in the anterior thalamic-fornix crossed lesion group (AT-FX Contra) and anterior thalamic-fornix-hippocampal commissure lesion group (AT-FX-HC) from Expt 2. The numbers refer to the plate numbers from Paxinos & Watson (1997). the contralateral hemisphere. The fornix lesions were virtually complete in three cases (sparing approximately 5% of the most lateral tip of the mbria), but in the other three between 25% and 30% of the most lateral portion of the tract was spared. The fornix lesion continued into the head of the hippocampus where the additional lesion damaged the medial part of the rostral head of the

10 Disconnecting hippocampal-thalamic connections 1723 hippocampus so cutting the commissure (Fig. 5, sections 25 and 26). In two cases the fornix lesion just touched the very dorsal margin of AV. In four cases the neurotoxic lesion involved nearly all of the AT, the only sparing being in the most medial AM. In the remaining two cases AD had completely disappeared but there was some sparing of the middle and caudal portions of AM and AV (both of which were clearly shrunken in size). In all cases the lesion continued caudally into the dorsal and rostral parts of LD. In two cases almost all of LD disappeared. The rostral portion of MD, along with the adjacent paracentral nucleus, was involved in two cases. All cases showed a very restricted region of neurotoxic damage to the granule cells in just the medial blade of the dentate gyrus. This was con ned to the ventral and rostral portion of the dentate gyrus, immediately above the rostral LD nucleus. In some cases, this cell loss was con ned to a couple of adjacent sections. Expt 2: behaviour Spatial forced alternation The performance of the ve groups of rats is depicted in Fig. 6. Analysis of variance revealed a highly signi cant effect of lesion group (F 4,38 = 11.18, P < 0.01), a signi cant main effect of session (F 4,152 = 7.47, P < 0.01) and a signi cant lesion group by session interaction (F 16,152 = 1.73, P < 0.05). Post hoc analyses showed that the performances of the AT-FX-HC (mean, 13.8) and FX groups (mean, 12.7) were signi cantly worse (P < 0.05) than the Sham (mean, 17.1), AT-FX Ipsi (mean, 15.7), and AT-FX Contra (mean, 15.9), groups. There were no other group differences. Analysis of the simple main effects showed that the performance of the AT-FX Contra and AT-FX-HC groups improved across the training sessions, whereas the performance levels of the other lesion groups remained stable across all training sessions. FIG. 6. T-maze alternation from Expt 2. The graph shows the mean percentage of correct responses (6 SEM) made by each of the ve groups (Sham, AT-FX Contra, AT-FX-HC, AT-FX Ipsi, and FX) in each block of three sessions (each test session comprised six trials). Morris water maze Acquisition. Analysis of variance on the escape latency data, revealed a signi cant effect of lesion (F 4,38 = 4.05, P < 0.01) and a highly signi cant main effect of session (F 9,342 = 64.72, P < 0.01). Post hoc analyses showed that the AT-FX-HC group took signi cantly longer to nd the escape platform (mean, 56.7 s) than any of the other lesion groups (means: Sham, 32.8 s; AT-FX Ipsi, 41.7 s; AT-FX Contra, 37.2 s; and FX, 37.4 s). The data are presented in Fig. 7. A comparison of the swim speeds of the rats just failed to nd a signi cant FIG. 7. Water maze acquisition. The graph shows the mean latency (6 SEM) of the ve groups (Sham, AT-FX Contra, AT-FX-HC, AT-FX Ipsi, and FX) to nd the hidden platform within each session (made up of the mean of four trials).

11 1724 E. C. Warburton et al. difference between the groups [F 4,38 = 2.37, P = 0.07 (means in m/s: Sham, 0.37; AT-FX Ipsi, 0.36; AT-FX Contra, 0.36; AT-FX-HC, 0.33; and FX, 0.33)]. Analysis of the percentage of time the rats spent swimming within 16.5 cm of the side walls of the maze during the acquisition of the water maze task showed a signi cant group difference (F 4,38 = 6.03, P < 0.01). Post hoc analyses revealed that the FX group spent signi cantly less time swimming around the side walls than any of the other groups. While the AT-FX-HC group spent longer at the side walls compared with the FX or AT-FX Contra groups, (means: Sham, %; AT-FX Ipsi, %; AT-FX Contra, %; AT-FX-HC, %; and FX, %). Probe. Analysis of the percentage of time the animals spent in the correct quadrant during the probe test showed no signi cant differences between the performance of the lesion groups [F 4,38 = 1.66, P = 0.26 (means: Sham, 41.4 %; AT-FX Ipsi, 34.3 %; AT-FX Contra, 47.1 %; AT-FX-HC, 31.8 %; and FX 45.1 %)]. However, further analyses comparing the performance levels of the animals in the probe test revealed that the AT-FX-HC group were the only group that failed to spend signi cantly more time in the training quadrant than chance (25%; t 5 < 1.0), and failed to spend signi cantly more time in the training quadrant than in the opposite quadrant (t 5 < 1.0). The mean proportion of time spent in each of the water maze quadrants by the lesion groups is shown in Fig. 8. Radial arm maze. One FX animal was excluded from the analysis of the radial arm maze data as it would not run in the maze. Analysis of the total number of trials taken by the remaining animals to obtain all eight reward pellets revealed a signi cant effect of lesion group (F 4,37 = 5.65, P < 0.01) and a signi cant main effect of session (F 5,185 = 27.66, P < 0.01). Post hoc analyses showed that the AT-FX- HC and FX groups took a signi cantly greater number of trials per session to complete the task (means, 11.9 and 12.8 trials, respectively) than the Sham group (mean, 10.1). There were no other group differences (means: AT-FX Ipsi, 11.3; and AT-FX Contra, 11.4). The data, which have been presented in blocks of three sessions for clarity, are presented in Fig. 9A. Analysis of the number of correct arm choices made by the animal revealed a signi cant effect of lesion (F 4,37 = 3.12, P = 0.03) and a highly signi cant main effect of session (F 5,185 = 27.29). Post hoc analyses again showed that the AT-FX-HC and FX groups made signi cantly fewer correct choices (means, 6.4 and 6.3 trials, respectively) than the Sham group (mean, 6.9). There were no other group differences (means: AT-FX Ipsi, 6.6; and AT-FX Contra, 6.6). The data are presented in Fig. 9B. Discussion Disconnection analyses provide a valuable method of examining whether speci c neural regions are functionally connected in a serial system, although any crossed projections between regions may attenuate the effects of such disconnections. Expt 1 examined the effect of combined contralateral lesions of the AT and fornix on object recognition and on a T-maze alternation task. In the object recognition tests, bilateral lesions of the FX and disconnection lesions of the FX and AT-FX Contra produced a similar pattern of results, i.e. both groups showed evidence of an impairment in their ability to identify the location change of an object, while the spontaneous preference for a novel over a familiar object remained unaffected. In the T-maze alternation task, a somewhat different pattern of results emerged. In this case, the AT-FX Contra and AT-FX Ipsi groups were both mildly impaired, but signi cantly less so than the FX group. FIG. 8. Water maze probe. Mean percentage of time each lesion group spent swimming in each of the four quadrants during the probe trial. The vertical lines represent the standard error of the mean. This latter pattern was repeated in Expt 2 where crossed anterior thalamic and fornix lesions again failed selectively to disrupt T-maze alternation. One explanation for the lack of a de cit in the AT-FX Contra group on the T-maze task is that there are suf cient contralateral projections to support performance. One potential route through which spatial information could have been processed was the spared hippocampal commissure, which provides homotopic and heterotopic connections from the hippocampus on one side of the brain to the hippocampus on the other side (Gottlieb & Cowan, 1973). Research into the development and maintenance of experimentally induced epileptic seizures, produced by hippocampal stimulation has shown that the effects of such stimulation, in the AT is signi cantly reduced by hippocampal commissurotomy (VanLandingham & Lothman, 1991). Such research suggests that information from the hippocampus can pass to the contralateral AT via the hippocampal commissure, possibly via decussating bres. Expt 2 examined the behavioural effects of commissurotomy plus the combined anterior thalamic and fornix lesions (AT-FX-HC) and found that the AT-FX-HC group were severely impaired in the radial arm maze, T-maze and in acquisition of the water maze task. A striking aspect of this result was that the AT-FX-HC group performed signi cantly worse than the AT- FX Contra group even though their surgeries were exactly matched except for the additional commissural damage. While such results are consistent with the hypothesis that the AT, fornix and hippocampus form part of a neural circuit, required for the processing of spatial information, they also indicate that in the rat there are multiple, parallel routes within this functional pathway and the notion of a direct serial system is an oversimpli cation. The AT-FX Contra and FX groups did show evidence of a similar dissociation in their performance in the object recognition tasks, i.e. they were able to distinguish a novel from a familiar object, but not to identify an object in a novel location. The spared object recognition performance is consistent with previous studies showing that bilateral damage to either the FX or the AT does not produce de cits in spontaneous preference tasks or in nonmatching to sample tasks using

12 Disconnecting hippocampal-thalamic connections 1725 FIG. 9. Radial arm maze. (A) The mean number of total arms visited (6 SEM) to retrieve eight reward pellets. The scores are blocked into groups of three sessions. (B) The mean number of correct choices (6 SEM) made in the rst eight trials of each session. The scores are blocked into groups of three sessions. discrete objects as stimuli (Rothblat & Kromer, 1991; Shaw & Aggleton, 1993; Aggleton et al., 1995; Ennaceur et al., 1997; Warburton & Aggleton, 1999). The results obtained in the present study also accord with previous ndings that rats with bilateral lesions of the fornix are impaired on the object location task (Ennaceur et al., 1997). Thus taken together with the results from the present study, it would appear that the AT and FX function conjointly to support judgements based on the spatial position of an object, but not for the familiarity of an object. In view of the results from the object recognition and object location studies, it is somewhat surprising that the animals in the AT- FX Contra group showed only a mild de cit in the T-maze task when compared with the size of the de cit in the FX group. However, a similar pattern of results (impaired object location, intact spatial alternation) has also been found in animals with cingulate cortex lesions (Aggleton et al., 1995; Ennaceur et al., 1997). This is especially noteworthy as the cingulate cortex is extensively and reciprocally connected with the AT (Van Groen et al., 1993). The dissociation between the effects of these lesions on the T-maze and object location tasks may re ect the fact that knowledge of self position or the memory of a location just visited might differ qualitatively from information about the position of an external object, and so may depend on different parts of a complex system. An alternative explanation for these results concerns the in uence of spared connections on tasks with different levels of sensitivity. In the T-maze task, allocentric cues are more readily visible as the apparatus is transparent, while the arena used for the recognition tasks is solid and so restricts the salience of extramaze cues. Thus when the neural system necessary for processing such information is compromised, as in the animals in the AT-FX Contra groups, performance will be poorer in the object location task. Furthermore, the spared pathways linking the hippocampus with the AT may carry enough spatial information to support performance in the T-maze, but not in the object location task. The results from Expt 2 support this account. In Expt 2, transection of the hippocampal commissure combined with contralateral lesions of the AT and fornix produced a greater de cit in each of the spatial tasks than the contralateral lesions alone. The degree to which the animals were impaired was, however, task dependent. In the T-maze and radial arm maze tasks the AT-FX-HC group were as impaired as animals with bilateral fornix lesions. Remarkably in the water maze task, however, disconnection of the AT by unilateral FX and hippocampal commissure lesions produced greater impairments than bilateral lesions of the FX itself. In fact, a more severe impairment was observed following bilateral damage to the AT compared with the fornix, for exactly the same water maze task (Warburton & Aggleton, 1999; Warburton et al., 1999). A part of the explanation for this difference is that fornix lesioned animals more readily swim into the middle of the maze, so reducing the latency to nd the platform. This behavioural pattern was found both in the present study and in previous studies of bilateral fornix lesions (Whishaw & Jarrard, 1995; Warburton & Aggleton, 1999). Bilateral anterior thalamic lesions do not have the same effect (Warburton & Aggleton, 1999). Consequently, it is not unexpected that a crossed lesion involving the AT, could have a greater effect than a bilateral fornix lesion. Finally it is important to note that the greater de cits in the AT-FX-HC group when compared with the AT-FX Contra group were not due to the impact of cutting the hippocampal commissure on its own. Seven of the nine AT-FX Ipsi animals in Expt 2 also had a hippocampal commissure lesion highlighting the fact that the impact of the tract section depends critically on the location of the other pathology in the same brain. That disconnection of the AT and fornix combined with transection of the hippocampal commissure produces either greater de cits or the same pattern of results as those observed after bilateral fornix lesions highlights the importance of those hippocampal projections to the AT, which go through the fornix. However, the contribution of cutting the hippocampal commissure, also suggests that there are multiple, indirect routes by which information from the hippocampus can