Do rats with retrosplenial cortex lesions lack direction?

Transcription

1 European Journal of Neuroscience European Journal of Neuroscience, Vol. 28, pp , 2008 doi: /j x BEHAVIORAL NEUROSCIENCE Do rats with retrosplenial cortex lesions lack direction? Helen H. J. Pothuizen, John P. Aggleton and Seralynne D. Vann School of Psychology, Cardiff University, Tower Building, Park Place, Cardiff CF10 3AT, UK Keywords: allocentric, cingulate, egocentric, spatial memory, T-maze Abstract The retrosplenial cortex is seen as a convergence point for different classes of spatial cue, yet aside from allocentric processing, little is known about other cue types that depend on the integrity of this area. Rats with bilateral retrosplenial cortex lesions were, therefore, trained on a sequence of reinforced spatial alternation tasks designed to isolate different spatial strategies. Using a standard T-maze alternation procedure, which could be solved using multiple strategies, only a marginal lesion effect was observed. Next, by using two T-mazes set side-by-side in the light, and then the dark, it was possible to examine alternation around a fixed bearing (direction alternation). Retrosplenial cortex lesions only disrupted the latter (direction alternation) condition. Direction alternation is of particular interest as it presumably taxes head-direction information, and so provides a way of behaviourally assessing the contribution of this navigation system. Finally, rats were tested on a spatial working memory task in a radial-arm maze. A retrosplenial lesion deficit appeared when the maze was rotated mid-trial, as repeatedly found in previous studies. The pattern of findings in the present study strongly indicates that retrosplenial cortex lesions impair the use of direction cues for alternation, in addition to previously established impairments for allocentric-based navigation and path integration. Introduction The contribution of the retrosplenial cortex to spatial learning remains poorly defined. One reason is that this region receives multiple classes of spatial information (Mizumori et al., 2001; Whishaw et al., 2001; Burgess, 2002). Consequently, it has proved difficult to determine the impact of retrosplenial cortex manipulations unless the behavioural tasks isolate specific forms of spatial information. Current evidence indicates that the region is required for the flexible use of allocentric information (Sutherland et al., 1988; Warburton et al., 1998; Harker & Whishaw, 2002; Vann & Aggleton, 2002, 2004; Van Groen et al., 2004) and for path integration (Cooper et al., 2001; Whishaw et al., 2001). As a consequence, the present study focussed on two other aspects of spatial processing: direction heading and egocentric information. Rats with retrosplenial cortex lesions were, therefore, trained on a spatial alternation task in which various cues types (allocentric, intramaze, egocentric, directional) could be systematically removed. While egocentric alternation could solve all test phases, directional alternation was only useful for specific stages. In a classic study, Douglas (1966) discovered that T-maze alternation is supported by multiple spatial strategies. This discovery arose from testing rats in two T-mazes placed side-by-side (one maze for the sample, the other for the alternation run). To demonstrate direction alternation, Douglas (1966) oriented the two mazes at 90 Correspondence: Dr S. D. Vann, as above. VannSD@Cardiff.ac.uk Received 14 August 2008, revised 15 October 2008, accepted 26 October 2008 to each other, rather than parallel. The decline in performance compared with the use of two parallel mazes was seen to reflect an inability to use bearing ( direction ) alternation information (Douglas, 1966; Dudchenko & Davidson, 2002). With this procedure it should, therefore, be possible to test the importance of selected brain regions for various spatial cue types, including heading direction. The retrosplenial cortex could be particularly informative as it contains head-direction cells (Chen et al., 1994; Cho & Sharp, 2001). Furthermore, it is densely connected with other regions (e.g. cortical areas 17 and 18b) able to provide visual information (Van Groen & Wyss, 1992). Consequently, the retrosplenial cortex could be important for integrating visual and idiothetic information (Mizumori et al., 2001). Even so, previous attempts to examine heading performance after retrosplenial lesions have found either no obvious deficit (Zheng et al., 2003; Futter & Aggleton, 2006) or only evidence of a mild impairment (Vann & Aggleton, 2004). The second goal stems from the fact that while retrosplenial cortex lesions impair various tests of spatial memory (Aggleton & Vann, 2004; Harker & Whishaw, 2004), T-maze alternation provides a striking anomaly (Neave et al., 1994; Aggleton et al., 1995). It has been variously argued that the failure to observe alternation deficits reflects incomplete retrosplenial lesions (Vann et al., 2003) or the choice of rat strain (Harker & Whishaw, 2004). The present study addressed these two goals by examining rats with extensive retrosplenial cortex lesions. For comparison purposes, rats were finally tested in a radial-arm maze using protocols that are reliably sensitive to retrosplenial damage in this strain (Vann et al., 2003; Vann & Aggleton, 2004).

2 Retrosplenial cortex and direction 2487 Materials and methods Experimental design Rats were first trained on T-maze alternation using a testing regime that should initially permit the use of multiple strategies to solve the task (Experiment 1). These strategies were: allocentric (place alternation); intra-maze (alternating away from an intra-maze cue, e.g. scent trail); egocentric (turning left and right with reference to body position); and direction [turning in opposite directions (e.g. East then West) with reference to a known bearing (e.g. North)]. Following initial training, the task was systematically modified so that fewer and fewer strategies could help solve the problem, and so isolate specific strategies. Experiment 2 (radial-arm maze) helped to confirm whether or not the retrosplenial lesions were sufficient to impair spatial memory tasks. Subjects Subjects were 22 male pigmented rats (Dark Agouti strain; Harlan Bicester, UK) weighing g at the time of surgery. Animals were housed in pairs under diurnal light conditions (14 h light 10 h dark) and testing was carried out during the light phase. Animals were given free access to water throughout the study. The animals were handled daily for at least 1 week before surgery, when they were randomly assigned to one of the two surgical groups: receiving complete retrosplenial cortex lesions ( Rspl, n = 12) or sham operations ( Sham, n = 10). All experiments were carried out in accordance with UK Animals (Scientific Procedures) Act, 1986 and associated guidelines. Surgery Animals were deeply anaesthetized by an intraperitoneal (i.p.) injection (60 mg kg) of 6% sodium pentobarbital (Sigma Chemical Company, Poole, UK; freshly dissolved in saline). The rats were each placed in a stereotaxic headholder (David Kopf Instruments, Tujunga, CA, USA) with the nose bar at The scalp was then cut and retracted to expose the skull. A craniotomy was performed to expose the cortex. The lesions were made by injecting a solution of 0.09 m N-methyl-d-aspartic acid (NMDA; Sigma) dissolved in phosphate buffer (ph 7.2). Retrosplenial cortex lesions were made by injecting eight sites per hemisphere using a 1-lL Hamilton syringe (Bonaduz, Switzerland) at an infusion rate of 0.1 ll min. The needle was left in place for 2 min after each infusion. The anterior-posterior (AP) coordinates were measured (in mm) with respect to bregma, the medio-lateral (ML) coordinates (in mm) with respect to the sagittal sinus, and the dorsoventral (DV) coordinates (in mm) with respect to the surface of the cortex. The stereotaxic co-ordinates of the eight injections were: (1) 0.15 ll at)1.7 (AP), ± 0.6 (ML), )1.6 (DV); (2) 0.15 ll at)1.7 (AP), ± 0.6 (ML), )1.5 (DV); (3) 0.15 ll at)3.2 (AP), ± 0.6 (ML), )1.6 (DV); (4) 0.15 ll at)3.2 (AP), ± 0.6 (ML), )1.5 (DV); (5) 0.15 ll at)4.9 (AP), ± 0.8 (ML), )2.2 (DV); (6) 0.30 ll at)4.9 (AP), ± 1.0 (ML), )1.2 (DV); (7) 0.30 ll at)6.0 (AP), ± 1.1 (ML), )1.6 (DV); and (8) 0.30 ll at)6.8 (AP), ± 1.3 (ML), )1.0 (DV). At completion of the surgery, the skin was sutured and an antibiotic powder (Acramide; Dales Pharmaceuticals, Skipton, UK) applied topically. Animals also received subcutaneous injections of glucose saline (5 ml), and were given paracetamol and sucrose in their drinking water for at least 4 days post-surgery. The surgically operated controls received the same procedure and drugs as the lesioned animals, with the exception of lowering the needle into the brain and injection of NMDA. The rats were allowed 3 weeks to recover before the start of behavioural testing (Experiment 1). Two weeks after completion of Experiment 1 (and 3 months post-surgery), the animals were trained on the radial-arm maze task (Experiment 2). Experiment 1: T-maze alternation Throughout Experiment 1, cross-mazes were used to create T-mazes. This change was achieved by blocking off the arm opposite the start arm on the sample run and again on the choice run. Rats were trained on five stages in which the cues available to perform the task were systematically removed. Apparatus Two identical cross-mazes were used during the entire experiment. The mazes were placed side-by-side (with the end of the two arms touching each other) on a table (74 cm high) in the same experimental room. The walls of the experimental room were hung with distinctive sets of salient visual cues. The floor of each cross-maze was made from wood and painted white. Each arm was 45.5 cm long and 12.0 cm wide. The side walls (32.5 cm high) were made from black Perspex. At the end of each arm was a sunken food well (2 cm in diameter and 0.75 cm deep). Access to an arm could be prevented by placing an aluminium barrier at the entrance of the arm. Lighting during initial habituation and training in the light stages was provided by two standard ceiling lights giving a mean light intensity of lx (measured at the centre of the two mazes). During training in the dark stages, lighting was provided by a standard lamp, equipped with a 60 W red light bulb (giving a mean light intensity of 1.4 lx at the centre of the mazes), which was placed on the floor of the test room under the centre of the table. During the dark stages, the visual cues were removed from the wall. Procedure One week prior to the start of the experiment, the animals were placed on a food-restricted diet. The weight of the animals was monitored every other day and was not allowed to drop below 85% of their freefeeding weight. Water remained available ad libitum throughout the experiment. One day before the start of the experiment, the rats were familiarized with the sucrose reward pellets (45 mg; Noyes Purified Rodent Diet, UK) in their home cage. Pretraining. Pretraining began with 5 days of habituation to the maze. On day 1, four sucrose pellets were scattered down each of the arms and two in each food well of both mazes, and the rats were allowed to explore each maze for 5 min. Rats were habituated in cage pairs on day 1 only. On days 2 and 3, the number of pellets was increased to four in each well and the rats were habituated to both mazes for 5 min each. On days 4 and 5, the number of pellets was reduced to two in each well and the aluminium barrier was placed at the entrance of one of the arms. The rats were again allowed to explore both mazes for 5 min each. During each habituation session, the reward pellets were continuously replaced so that no arm was found to be empty on return. Training on the forced-choice alternation rule commenced on the following day. Basic experimental procedure for stages 1 5. Throughout the entire experiment the rats received eight trials per daily session. Each trial consisted of a forced sample run followed by a choice run (Fig. 1).

3 2488 H. H. J. Pothuizen et al. Fig. 1. Examples of test protocols used for the forced-choice alternation experiment that was run in one maze (Stage 1) and then two mazes (Stages 2 5). Solid lines indicate the forced sample phase while dashed lines indicate the correct response in the choice phase. Access to an arm could be blocked by placing a barrier at the entrance of the arm (effectively turning the cross-maze into a T-maze configuration). Initial training on the task (A) permitted the use of multiple cue information, i.e. allocentric, intra-maze, egocentric and direction (with reference to a known bearing). The task was then systematically modified (Stages 1 5) so that fewer and fewer spatial strategies could help alternation. These manipulations included changing starting direction of the choice trial (B and G J), using two mazes instead of one (C J), or running in the dark as illustrated with the dark-grey hatched background (E H). Sample runs could start from the south or the north arm (but for simplicity only the south trials are depicted); diff. place, different place. Although all testing was carried out in a cross-maze, it was modified by placing a barrier blocking access to the arm directly in line with the start arm (effectively turning it into a T-maze configuration). During the forced sample run, one of the side arms of the maze was blocked by an aluminium barrier. After the rat turned into the preselected arm, it was allowed to eat two reward pellets that had been previously placed in the food well. The rat was then picked up from the maze and immediately returned to the preselected start arm, either in the same maze (Stage 1) or in the second maze (Stages 2 5; Fig. 1). Approximately 3 s after the end of the sample phase the choice phase

4 Retrosplenial cortex and direction 2489 began. The rat was allowed to run up the stem of the maze and was now given a free choice between the left and the right turn arms. The rat received two reward pellets only if it turned in the direction opposite to that in the sample run (i.e. non-matching). At the start of each session, four rats were taken from the holding room to the experimental room in an enclosed, opaque carry-box made of aluminium. Each rat was placed in a separate compartment in the box. Groups of either three or four rats were tested together, with each rat having one trial in turn, so that the inter-trial interval was approximately 3 min. In order to control systematically the strategies supporting alternation (allocentric, intra-maze, egocentric, direction), the alternation training consisted of five sequential stages (Fig. 1). Stage 1 (allocentric, intra-maze, directional and egocentric cues all available) In Stage 1 (10 sessions), both the sample and choice run were carried out in the same maze (Fig. 1A and B). Both mazes were used equally and all sample runs started from the South. The animals, however, received two distinct trial types, that were intermingled within each session. One of the alternation trial types ( standard ) closely followed the way that T-maze alternation had been tested in previous studies (e.g. Neave et al., 1994; Aggleton et al., 1995). The other trial type ( constant direction place ) also tested alternation, but now the problem could also be solved by heading in a constant direction place on the choice run (i.e. North; Fig. 1B). On standard alternation trials (Fig. 1A) the sample and choice runs started from the same point (i.e. South). For half of the standard alternation trials the rat was forced to the West in the sample run, and rewarded for going into the East arm on the choice run (Fig. 1A). The remaining standard alternation trials consisted of the rat being forced to go East in the sample run, and rewarded for going West on the choice run. For the standard alternation trials all cue types could be used to support performance (i.e. allocentric, intra-maze, egocentric and direction). Testing was counterbalanced with four trials per session, half to the West and half to the East. The remaining 50% of the alternation trials used the constant direction place design. Now, the start direction always changed between the sample and choice run, as all sample runs started from the South but the test runs could start from either West or East (Fig. 1B). Like the standard alternation condition, both allocentric and intramaze cues remained valid (i.e. alternate by avoiding previously encountered cues places), though now the rat had to avoid re-entering what had been the start arm on the sample run. Egocentric alternation would again also solve all trial conditions. Furthermore, the correct response for all choice runs in the constant direction place condition was to run North (Fig. 1B). As a consequence, the rats could learn to go to a constant place direction, in addition to using the full array of alternation cues. Each rat received four trials of each trial type per session; trial types and left vs. right turns were counterbalanced across sessions. Stage 2 (allocentric, directional and egocentric cues available) In Stage 2 (eight sessions), the sample and choice runs were conducted in different mazes to prevent the use of intra-maze cues. The number of sample and choice runs was counterbalanced across the two mazes. Again, the eight trials per session were equally divided across two distinct trial types (Fig. 1C and D). One of the trial types ( different place ; Fig. 1C) was identical to the standard trial type from Stage 1 (Fig. 1A), except it was run in two mazes. For the other trial type ( same place ; Fig. 1D), the rats were rewarded for choosing the arm leading to a goal that was located very close to the absolute position of the food well in the choice run (Fig. 1D). Thus, for different place trials all but intra-maze cues should be available, while for same place trials allocentric cues could potentially interfere. In contrast to Stage 1, equal numbers of sample runs (and choice runs) began from the North and the South start point (note Fig. 1C and D only shows trials starting from the South). However, to promote the use of direction alternation, the same start point (North or South) was always used for the sample and choice run on any given trial (only true for 50% of trials in Stage 1). Stage 3 (directional and egocentric cues available) Stage 3 (10 sessions) was identical to Stage 2 except it was run in the dark. This change was designed to block the use of visual allocentric cues (Fig. 1E and F). Stage 4 (egocentric cues available) The four sessions of Stage 4 were the same as Stage 3 but with one critical difference now the sample and choice runs not only occurred in two mazes but always involved different start directions (Fig. 1G and H). Hence, no intra-maze, visual or direction cues were available to successfully perform these trials (i.e. rats should now be reliant on just egocentric cues). Sample runs always began in either the South or the North start point, resulting in a combination of eight different trial types (two of which are depicted in Fig. 1G and H). Because two start points were used for the sample runs, the task could not be solved by going to a constant direction place on the choice run (unlike the constant condition in Stage 1). Stage 5 (egocentric cues available) Because performance of the control group had declined to almost chance levels in Stage 4, the animals were run in Stage 5 (also four sessions) on identical trials as in Stage 4 except under normal light conditions (Fig. 1I and J). It should be noted that unlike the Stage 1 constant direction place condition, which this stage superficially resembles, the sample and choice runs were in two separate mazes and the sample runs were in one of two directions. As a consequence, allocentric cues could not readily support performance. Experiment 2: radial-arm maze Apparatus Testing was carried out in an eight-arm radial maze. The maze consisted of an octagonal central platform (34 cm diameter) with eight equally spaced radial-arms (87 cm long, 10 cm wide). The floor of the central platform and the floors of the eight arms were made of wood, while clear Perspex (24 cm high) formed the walls of the arms. Close to the furthest end of each arm was a sunken food well (2 cm in diameter and 0.5 cm deep). At the start of each arm was a clear Perspex guillotine door (12 cm high) that controlled access in and out of the central platform. Each door was attached to a pulley system enabling the experimenter to control access to the arms. The maze was in a rectangular room ( cm) that contained salient visual cues such as geometric shapes and high contrast stimuli on the walls. Procedure The animals remained on a food-restricted diet and their weight remained at or above 85% of their free-feeding body weight. Water remained available ad libitum. Pretraining involved three habituation

5 2490 H. H. J. Pothuizen et al. sessions where the animals were allowed to explore the maze freely for 5 min. All of the guillotine doors were raised and reward pellets were scattered down the arms. This was followed by formal training that lasted for 20 sessions and consisted of two stages. Acquisition The Acquisition phase (sessions 1 12) was the standard working memory version of the radial-arm maze task (Olton et al., 1978) where the animals optimal strategy was to retrieve the reward pellets from all eight arms without re-entering any previously entered arms. At the start of a trial all eight arms were baited with a single reward pellet. The animal would make an arm choice and then return to the central platform and all the doors were closed for about 10 s before they were opened again, permitting the animal to make another choice. This continued until all eight arms had been visited. Rotation Previous studies have shown that retrosplenial lesioned rats are particularly sensitive to rotation of the maze (Vann et al., 2003; Vann & Aggleton, 2004). Maze rotation (sessions 13 20) was, therefore, introduced after the rat had made its first four choices. For the rotation, the rat was confined to the centre of the maze while the maze was rotated one arm position (either clockwise or anticlockwise), and the arms rebaited so that the spatial locations of the remaining pellets were unchanged with respect to the room cues. The trial then continued until all eight arms had been visited. Statistical analysis The number of correct responses per session was noted in all stages of Experiment 1 (forced-choice alternation). For the stages that comprised two different trial types (i.e. Stages 1 3), the number of correct responses per trial type per session were also noted. The percentage correct was then calculated and averaged across two sessions to create two-session blocks. This measure was subjected to an anova with the between-subject factor of group (Rspl and Sham), a within-subject factor of trial type (only for Stages 1 3) and the repeated measure of block. As mentioned previously, the different stages of the T-maze experiment were designed in such a way that cue information available to the rats was systematically eliminated. In order to assess the effects of this cue elimination across the stages, the overall mean per stage was calculated. For this comparison, only the results for the standard alternation task from Stage 1 were used as this provides a start point that can be compared across studies. These measures were then subjected to a 2 2 anova with the between-subject factor of group and the repeated measure of stage for the Stages 1 and 2, 2 and 3, 3 and 4, and 4 and 5, respectively. For Experiment 2 (radial-arm maze), the number of errors and the number of correct visits out of the first eight arm choices were noted per session. The data were then collapsed into two-session blocks. Both measures were subjected to an anova with the between-subject factor of group and the repeated measure of block. The data for acquisition and rotation performance were analysed separately as the latter included a rotation of the maze after the first four correct choices. The number of sequential arm choices, defined as successive choices involving immediately adjacent arms in a constant direction, was also calculated for the Acquisition stage. Sequential choice responses were measured by giving the rat a score of +1 (clockwise) or )1 (counter clockwise) if the arm entered was immediately adjacent to the previous choice and 0 for any other arm choice. A higher absolute score reflects the use of a sequential response strategy (Olton & Samuelson, 1976). This measure was also subjected to an anova with the between-subject factor of group and the repeated measure of block. All data were subjected to a parametric anova performed using the statistical software spss implemented on a PC running the Microsoft Windows XP operating system. Statistical significance was set at a probability level of P < 0.05 for all tests. Newman Keuls post hoc tests were used and simple effects were evaluated when appropriate. Histology At the end of behavioural testing, the rats were deeply anaesthetized with sodium pentobarbital (60 mg kg, i.p.; Euthatal; Merial Animal Health, Harlow, UK) and then transcardially perfused with 0.1 m phosphate-buffered saline (PBS) at room temperature for approximately 2 min (flow rate 35 ml min), followed by 4% solution of depolymerized paraformaldehyde in 0.1 m phosphate buffer for approximately 10 min at a flow rate of 35 ml min. The brains were removed and post-fixed for 4 h in the same fixative and then cryoprotected in 25% sucrose solution overnight. Four adjacent series of coronal sections (40 lm) were cut on a freezing sliding microtome. Two series were collected in a PBS sodium azide solution and stored at 5 C until further use. One of the two remaining one-in-four series was directly mounted onto gelatine-coated slides and stained with Cresyl violet the following day. After staining, the sections were dehydrated through an alcohol series, cleared with xylene and coverslipped with the mounting medium DPX. Neuronal nuclei protein (NeuN) immunostaining The remaining one-in-four series was processed for the immunohistochemical demonstration of the NeuN protein to enable a more accurate evaluation of the extent of the excitotoxic lesions (Jongen- Rêlo & Feldon, 2002). Free-floating sections were rinsed in 0.1 m PBST (PBS with 0.2% Triton X-100) and treated with 0.3% H 2 O 2 (hydrogen peroxide) in 0.1 m PBST for 10 min to suppress endogenous peroxidase activity. Sections were rinsed four times with 0.1 m PBST for 10 min each. Sections were subsequently incubated for 48 h at 4 C in the monoclonal anti-neun serum (1 : 5000; Chemicon, Temecula, CA, USA) diluted in PBST. After rinsing four times in 0.1 m PBST (10 min each), sections were incubated for 2 h at room temperature in biotinylated horse anti-mouse serum (5 ll ml PBST; Veclab, UK) in secondary diluent consisting of normal horse serum (15 ll ml PBST; Veclab, UK) in 0.1 m PBST. After four rinses in 0.1 m PBST (10 min each), sections were incubated for 1 h in an avidin-biotinhorseradish peroxidase complex (1 : 200; ABC-Elite, Vector Laboratories, Orton Southgate, Peterborough, UK) in PBST. Following four rinses in 0.1 m PBST and two rinses in 0.05 m Tris buffer, sections were placed for 1 2 min in a chromagen solution consisting of 0.05% diaminobenzidine (Sigma), buffer solution and 0.01% H 2 O 2 (DAB substrate kit; Vector Laboratories). The reaction was visually monitored and stopped in rinses of cold 0.1 m PBS. The sections were mounted and dried on gelatine-coated slides, and then prepared for cover-slipping as described before. Results Histological evaluation of the lesions One rat in the lesion group was excluded because the lesion extended bilaterally into the hippocampus. A coronal section of a

6 Retrosplenial cortex and direction 2491 Fig. 2. Location and extent of retrosplenial cortex lesions. Photomicrograph of a coronal section immunostained for NeuN from a case with a mid-sized retrosplenial lesion (A) and from a sham control (B). (C) Coronal sections illustrating the smallest (black) and largest (grey) retrospinal cortex lesions from the cases included in the behavioural analyses. The numbers indicate the distance (in mm) from bregma (adapted from Paxinos & Watson, 1997). Distance from bregma in (A) also applies to the section in (B). The dashed lines in (A) and (B) illustrate the lateral border of the retrosplenial cortex ( Rspl crtx ). The scale bar in (A) represents 500 lm and applies to all photomicrographs. medium-sized lesion from the remaining animals is depicted in Fig. 2A (NeuN). The lesions were characterized by marked cell loss and intense gliosis in a discrete area around the injection site. The cell loss induced by the NMDA injection resulted in shrinkage of the entire retrosplenial cortex (Fig. 2A). The retrosplenial lesions were either complete anterior to the splenium or there was a small amount of sparing in the most lateral part of the dysgranular cortex (Fig. 2A). In six cases, some sparing was observed in the most caudal retrosplenial cortex in the granular and or dysgranular subregions. In two of these six cases, the sparing was bilateral. In three cases, the lesion extended in just one hemisphere into a very limited part of the dorsal subiculum immediately anterior to the back of the splenium (Fig. 2). In six cases, a unilateral patch of atrophy was observed in a small portion of the most dorsal medial tip of the CA1 hippocampal subregion. According to the rat brain atlas of Paxinos & Watson (1997), the largest retrosplenial lesion extended from )2.56 to )7.30 mm from bregma, and the smallest lesion )2.56 to )6.72 mm from bregma (Fig. 2C). The final number of subjects was 11 in the retrosplenial lesion and 10 in the Sham group. Behavioural results Experiment 1: T-maze alternation Stage 1 (allocentric, intra-maze, directional and egocentric cues all available) Performance of both groups (Fig. 3) improved with training as demonstrated by a main effect of session block (F 4,76 = 17.41, P < ). When performance was separated into standard and constant direction place conditions it becomes apparent that the constant direction place trials were more difficult for all rats (F 1,19 = 9.53, P < 0.007). This fall in performance accords with the removal of direction cues around a fixed axis along with the increased ambiguity of the allocentric cues for alternation. Although the retrosplenial lesion rats appear to be relatively more impaired on the constant direction place than on the standard trials, the group trial type interaction was not significant (F 1,19 = 0.77, P = 0.39). Further analyses, restricted to the two trial types separately, yielded no clear group effects in either analysis ( standard : F 1,19 = 3.49, P = and constant direction place : F 1,19 = 3.69, P = 0.070).

7 2492 H. H. J. Pothuizen et al. Fig. 3. Stage 1. Acquisition of the forced-choice alternation rule in Stage 1. The line graphs show the mean percentage correct in two-session blocks (10 sessions) for the standard (left) and constant direction place trials (right), respectively. The insets in both line graphs depict an example of the corresponding trial type. Error bars refer to ± SEM of the corresponding mean values. Rspl: retrosplenial cortex lesions (n = 11); Sham: surgical controls (n = 10). Stage 2 (allocentric, directional and egocentric cues available) In Stage 2, intra-maze cues were removed by using two adjacent mazes, and the use of allocentric cues became ambiguous on the same place subset of trials. Overall performance across this stage of training revealed no differences between the retrosplenial lesion group and shams (Fig. 4A; F 1,19 = 0.02, P = 0.88). Performance levels for both groups were, however, appreciably lower than in Stage 1 as shown by the 2 2 anova comparing the means per group across Stages 1 and 2 (Stage: F 1,19 = 57.31, P < ). There was no evidence, however, of an improvement with additional training on Stage 2 (Fig. 4A; block effect: F 3,57 = 1.75, P = 0.17). Fig. 4. Stage 2. Performance on the forced-choice alternation rule using two adjacent mazes (in the light). The line graph in (A) shows the mean percentage correct in two-session blocks (eight sessions) across both trial types for Stage 2. The histogram in (A) depicts the mean percentage correct across the four blocks. (B) The mean percentages correct for the different place (left) and same place trials (right), respectively. The insets in both line graphs in (B) depict an example of the corresponding trial type. Error bars refer to ± SEM of the corresponding mean values. Rspl: retrosplenial cortex lesions (n = 11); Sham: surgical controls (n = 10).

8 Retrosplenial cortex and direction 2493 As expected, both groups tended to make more errors on the same place trials than on the different place trials (Fig. 4B). This observation was supported by a close to significant effect of trial type (F 1,19 = 4.19, P = 0.055). As can be seen in Fig. 4B, the pattern of performance across the two trial types appeared the same for both groups (group type: F 1,19 = 0.01, P = 0.92). Stage 3 (directional and egocentric cues available) Stage 3 was run in the dark with two different mazes. As a consequence, vestibular direction and egocentric cues were available, but not intra-maze or visual allocentric cues. Performance remained above chance (i.e. 50%) for both groups (one-sample t-tests: Rspl: P < 0.003; Sham: P < ), but now in contrast to Stage 2 the Sham group very clearly outperformed the retrosplenial lesion group (Fig. 5A). This observation was confirmed by a significant effect of group (F 1,19 = 9.85, P < 0.006). There was no evidence that the groups improved with training (effect of training block: F 4,76 = 0.63, P = 0.65), nor was there a group training block interaction (F 4,76 = 1.46, P = 0.22). The comparison between same place and different place trials was informative as it would be predicted that performance levels would now be equivalent if the rats could not detect that on same place trials they were running to adjacent locations (i.e. they could not use visual allocentric cues; Fig. 1F). This prediction was confirmed by the lack of any trial type difference (Fig. 5B; F 1,19 = 0.77, P = 0.78). While inspection of Fig. 5B suggests that the retrosplenial deficit was relatively greater for same place trials, this was not supported by the group trial type interaction (F 1,19 = 2.51, P = 0.13). Stage 4 (egocentric cues available) It was anticipated that rats would rely heavily on egocentric cues during Stage 4, and previous studies have shown that rats find this form of information very difficult to use for delayed alternation (Baird et al., 2004; Futter & Aggleton, 2006). Consistent with our prediction, the overall performance of both groups was significantly lower than in Stage 3 (F 1,19 = 24.42, P < ) and was at or close to chance level (Fig. 6A). One-sample t-tests conducted to assess whether the mean level of performance differed significantly from 50% (chance) confirmed this observation (Rspl: P = 0.06; Sham: P < 0.05). No group differences were observed (group: F 1,19 = 0.17, P = 0.68; group block: F 1,19 = 0.002, P = 0.97). Stage 5 (egocentric cues available) Despite the fact that Stage 5 was conducted under normal light levels, only egocentric cues would unambiguously solve the task. The overall performance of both groups was again close to, or at, chance level (one-sample t-test, Rspl: P < 0.05; Sham: P = 0.077; Fig. 6B). Fig. 5. Stage 3. Performance on the forced-choice alternation rule in the dark. The line graph in (A) shows the mean percentage correct in two-session blocks (10 sessions) across both trial types. The histogram in (A) depicts the mean percentage correct across the five blocks. (B) The mean percentages correct for the different place (left) and same place trials (right), respectively. The insets in both line graphs in (B) depict examples of the corresponding trial types. Note that the trials are identical to Stage 2 with the exception that they were run in the dark. The asterisk indicates P < Error bars refer to ± SEM of the corresponding mean values. Rspl: retrosplenial cortex lesions (n = 11); Sham: surgical controls (n = 10).

9 2494 H. H. J. Pothuizen et al. effect of stage (F 1,19 = 14.74, P < ). As can be seen in Fig. 7, the drop in performance was most evident in the Sham group. This observation was confirmed by a significant group stage interaction (F 1,19 = 5.37, P < 0.033). Further evaluation of the simple effects showed that this interaction was due to the drop in performance of the Sham group (stage: F 1,19 = 18.95, P < ) as the effect of stage was not significant in the retrosplenial lesion group (F 1,19 = 1.16, P = 0.30). Neither groups performance differed between Stages 4 and 5 (group stage: F 1,19 = 0.54, P = 0.47). Fig. 6. Stage 4 (A) and Stage 5 (B). Performance on the forced-choice alternation rule in the dark (Stage 4) and the light (Stage 5) when direction alternation is excluded. The line graphs in (A) and (B) show the mean percentage correct in two-session blocks (four sessions) for Stages 4 and 5, respectively. The histograms in (A) and (B) depict the mean percentage correct across the two blocks of Stages 4 and 5, respectively. The insets in the line graphs in (A) and (B) depict an example of the trial type used. Note that the trials for Stages 4 and 5 were identical except that Stage 4 was in the dark. The dashed lines represent chance level of performance. Error bars refer to ± SEM of the corresponding mean values. Rspl: retrosplenial cortex lesions (n = 11); Sham: surgical controls (n = 10). Similarly to Stage 4, no group differences were observed (group: F 1,19 = 0.02, P = 0.89; group block: F 1,19 = 0.03, P = 0.86). Comparison of mean performance across all five stages Figure 7 depicts the mean performance levels for each of the five stages for the two groups in order to show better how performance changed across stages. The analysis conducted to compare the mean performance of Stage 1 ( standard trials) and Stage 2 yielded a clear significant effect of test procedure (F 1,19 = 67.58, P < ), but no group stage interaction (F 1,19 = 0.29, P = 0.59), which confirms the observation that performance of both groups was lower in Stage 2 compared with Stage 1 (see also Fig. 7). Comparing the mean performance of Stage 3 with that of Stage 2 showed that whereas performance of the retrosplenial lesion rats decreased across the two stages, performance of the shams appeared to be unaffected by the introduction of running in the dark. This observation was confirmed by a significant group stage interaction (F 1,19 = 5.69, P < 0.03). Further evaluation of the simple effects confirmed an effect of stage in the retrosplenial lesion group (F 1,19 = 6.50, P < 0.03), but not in the Sham group (F 1,19 = 0.68, P = 0.42). The transition from Stage 3 to Stage 4 generally reduced performance, as indicated by a significant Experiment 2: radial-arm maze Acquisition (sessions 1 12) Initial acquisition by the retrosplenial lesion rats appeared normal. For both groups the number of errors made in the standard version of the working memory task decreased as training progressed (Fig. 8). This observation was confirmed by a significant effect of block (F 5,95 = 12.62, P < ). No differences in the number of errors were, however, observed between the retrosplenial lesion and Sham group (group: F 1,19 = 2.13, P = 0.16; group block: F 5,95 = 1.02, P = 0.41). An analysis of the numbers of correct entries in the first eight choices revealed a similar pattern of results and is, therefore, not depicted graphically. Whereas overall performance increased as training progressed (block: F 5,95 = 8.13, P < ), the two groups did not differ (group: F 1,19 = 0.02, P = 0.88; group block: F 5,95 = 0.66, P = 0.65). The analysis of the sequential choice responses revealed that the lack of a lesion effect during acquisition could not be attributed to the fact that the retrosplenial lesion group might be using a sequential response strategy (i.e. choosing adjacent arms in a constant direction) to solve the task as neither the group effect (F 1,19 = 0.36, P = 0.56) nor the group block effect (F 5,95 = 0.93, P = 0.46) attained significance. The mean sequential choice scores for both groups were low (Rspl = 1.17 ± 0.6 and Sham = 1.26 ± 0.8, mean ± SEM), showing that the rats were not using a simple response strategy. Rotation (sessions 13 20) Following rotation of the maze mid-way through a trial, the shams maintained or even improved on the number of errors made at the end of initial training (i.e. acquisition; Fig. 8). In contrast, the retrosplenial lesion rats remained at a relatively poor level of performance throughout the entire four blocks of rotation. This led to a significant group effect (F 1,19 = 6.12, P < 0.03), but no group block interaction (F 3,57 = 1.05, P = 0.38). Again, the numbers of correct entries in the first eight choices (not presented graphically) revealed a similar pattern of results. The overall level of performance did not increase as training progressed (block: F 3,57 = 1.10, P = 0.36). However, while the control rats continued to improve, the retrosplenial lesion rats remained at a poor level throughout. This difference was reflected in a significant group block interaction (F 3,57 = 4.05, P < 0.02). Although the main effect of group failed to attain significance (F 1,19 = 3.17, P = 0.09), additional post hoc pair-wise comparisons on the successive blocks indicated that the two groups significantly differed on the last block of rotation (P < ). Discussion The present study examined reinforced spatial alternation in rats with retrosplenial cortex lesions to answer two related questions. The first

10 Retrosplenial cortex and direction 2495 Fig. 7. Effects of retrosplenial cortex lesions on performance in the T-maze forced-choice alternation experiment across all five stages to illustrate the transition between stages. Data are expressed as mean percentage correct (± SEM) per stage. Only the mean percentage from the standard trials in Stage 1 is depicted. The dashed line represents chance level of performance. The asterisk indicates P < Rspl: retrosplenial cortex lesions (n = 11); Sham: surgical controls (n = 10). Fig. 8. Radial-arm maze (Experiment 2). The line drawings depict the mean number of errors in two-session blocks during acquisition (left, 12 sessions) and maze rotation (right, eight sessions). Error bars refer to ± SEM of the corresponding mean values. Rspl: retrosplenial cortex lesions (n = 11); Sham: surgical controls (n = 10). was whether retrosplenial cortex lesions impair spatial alternation, and so resolve a striking anomaly in the current literature on retrosplenial cortex function (Aggleton & Vann, 2004; Harker & Whishaw, 2004). The second was to determine the importance of the retrosplenial cortex for heading (bearing) information. While it is known that the retrosplenial cortex is involved in spatial memory in both rodents (Sutherland & Hoesing, 1993; Cooper et al., 2001; Whishaw et al., 2001; Vann & Aggleton, 2002, 2004) and humans (Maguire, 2001; Epstein et al., 2007; Iaria et al., 2007), there remains much uncertainty about its relative importance for different classes of spatial information. In view of the strategic position of the retrosplenial cortex in the head-direction system (Chen et al., 1994; Taube, 1998; Mizumori et al., 2001) the present study focussed on the use of direction information in alternation. Rats were also tested on a spatial working memory task in a radial-arm maze. The sensitivity of retrosplenial cortex lesions to rotation of the radial-maze mid-way through a trial has been shown repeatedly (Vann et al., 2003; Vann & Aggleton, 2004), and so was used to test the effectiveness of the present retrosplenial lesions. Extensive retrosplenial cortex lesions produced only a borderline (non-significant) effect on standard T-maze alternation ( standard trials, Stage 1). This finding leaves the apparent anomaly that retrosplenial cortex lesions impair other tests of spatial working memory (Vann et al., 2003; Vann & Aggleton, 2004; Lukoyanov et al., 2005; Cain et al., 2006), while clear deficits on T-maze alternation remain to be observed (Neave et al., 1994; Aggleton et al., 1995). The explanation that ceiling effects masked any clear lesion effect is possible, but comparisons of performance on just the initial sessions still failed to show a significant group difference. Previous failures to find a T-maze alternation deficit have been interpreted in various ways. One suggestion (Harker & Whishaw, 2004) is that these null results reflect peculiar properties of the particular rat strain (Dark Agouti) used in previous T-maze studies. That suggestion seems most improbable as the Dark Agouti rats in the present study displayed

11 2496 H. H. J. Pothuizen et al. radial-arm maze deficits, which also taxes non-matching-to-place, along with significant alternation deficits in Stage 3. A counterproposal was that previous null results reflect incomplete lesions that spare caudal retrosplenial cortex (Vann et al., 2003; Vann & Aggleton, 2004). The lesions in the present study extended more caudally than in those studies failing to observe an alternation deficit (Neave et al., 1994; Aggleton et al., 1995) and, while there is a suggestion of a lesion effect, there still remains the issue of why any effect was not more clear-cut. One explanation is that T-maze alternation can be solved using an array of spatial strategies and that retrosplenial lesions have a more selective effect, being most disruptive for those strategies that are typically not the most critical. This account receives direct support from the second question examined by the alternation studies. The second question was whether retrosplenial cortex lesions would disrupt T-maze alternation when solving the task relied on direction alternation (Douglas, 1966; Dudchenko, 2001). Direction alternation refers to turning left then right from a particular environmental alignment or axis (e.g. the direction of the long arm of a T-maze), and is distinct from egocentric alternation where the rat alternates around its body position. The difference between these two strategies becomes clear when rats are tested in two adjacent T-mazes, as in the present study. While egocentric alternation can support all alternation conditions across Stages 1 5, direction alternation can only help when the two T-mazes are in the same alignment, i.e. parallel (Dudchenko, 2001). Furthermore, by using two T-mazes it is possible to control the other types of cues (allocentric, intra-maze) that rats normally use to solve alternation problems (Dember & Fowler, 1958; Douglas, 1966; Dudchenko & Davidson, 2002; Futter & Aggleton, 2006). A potential concern is that head-direction cells might reorient to the 90 change in start direction when using two mazes (Dudchenko & Zinyuk, 2005), presumably reflecting the familiarity of the two start directions. For this reason, new start positions were used in Stage 4, providing the critical contrast with Stage 3. Each of the successive stages in the T-maze was designed to remove or add a specific cue type. The key condition was Stage 3 (two mazes in the dark) as only egocentric alternation and direction alternation based on interoceptive cues should be able to support performance. A clear retrosplenial lesion deficit emerged on Stage 3, in contrast to the preceding stages in the light. While performance on Stage 3 could, in theory, be supported by egocentric as well as directional information, the notion that the retrosplenial lesion deficit on Stage 3 reflects a selective loss of egocentric alternation can be rejected on several grounds. First, previous studies have found that rats find it exceptionally difficult to use egocentric information to solve spatial alternation tasks (Dudchenko, 2001; Baird et al., 2004; Futter & Aggleton, 2006). The same difficulty could be seen in the present study as both groups barely performed above chance on Stages 4 and 5, where an egocentric strategy was the only unambiguous solution. Furthermore, control performance dropped markedly from Stage 3 (direction plus egocentric) to Stage 4 (egocentric), again consistent with the view that normal performance on Stage 3 did not just reflect egocentric alternation. Taken together, this pattern of results strongly supports the view that the retrosplenial lesions most clearly impaired alternation when it was heavily reliant on heading direction in the dark. In apparent contrast, a previous study found that retrosplenial cortex lesions spare the ability to swim in a fixed direction (Zheng et al., 2003). There are, however, a number of important differences with the present study. These differences include the fact that in the previous study the trajectory was learnt before surgery (Zheng et al., 2003) and that the required swim direction was constant across all trials. In the present study, the initial heading direction (long arm of T-maze) remained constant for sample and choice runs (Stages 2 and 3), but could be from the South or the North. Furthermore, this information now had to be used flexibly (i.e. for alternation). A second study, also in a swim maze, again failed to find a clear retrosplenial lesion effect when rats were trained to swim in a fixed direction (Vann & Aggleton, 2004), though rats with extensive retrosplenial lesions did make significantly longer swim paths before finding the submerged platform. A third study also used pairs of cross-mazes with the specific goal of training egocentric working memory (Baird et al., 2004). The training procedure was very similar to that of Stage 4 in the present study. Although there was some evidence that the rats also used direction information to help solve some of the task conditions, performance was barely above chance (always less than 60%), leaving the task insensitive to lesion effects (Baird et al., 2004). The present task differed as the training procedures in Stage 3 were more likely to encourage direction alternation (e.g. constant North South axis for the start of all runs). As a result, there were much higher levels of performance on the critical Stage 3 (over 70%) than that seen in the previous study (Baird et al., 2004). Other potential information about the impact of retrosplenial damage on heading comes from more detailed analyses of the earlier stages of training in the present study (Stages 1 and 2). In Stage 1 there were two trial types ( standard and constant direction place ; Fig. 1). Their comparison is of interest as the constant direction place trials could be solved by heading in a fixed direction (i.e. North) or to a fixed location on the choice runs. In fact, these test conditions ( constant direction place ) did not reveal a clear retrosplenial lesion effect. A limitation is that although this particular trial type may tax heading information it can also be solved using other classes of information (allocentric and intra-maze). The possibility that the rats learnt a fixed rule (always go North or go to the arm located North) does, however, receive support from the initially steep learning curve for this subset of trials (Fig. 3). Evidence that the retrosplenial cortex receives multiple classes of spatial information (Mizumori et al., 2001; Whishaw et al., 2001; Burgess, 2002) serves to underline the need to focus on tasks that can only be solved using very specific strategies. One of the classes of information that reaches the retrosplenial cortex is allocentric. Previous evidence that retrosplenial cortex lesions produce allocentric deficits includes that from studies using the Morris water-maze. The overall pattern is typically for relatively mild impairments such that the rats show retarded acquisition (Sutherland et al., 1988; Warburton et al., 1998; Harker & Whishaw, 2002; Vann & Aggleton, 2002, 2004; Van Groen et al., 2004) but nevertheless do eventually learn the general location of the submerged platform (but see Whishaw et al., 2001). This pattern of swim maze (allocentric) performance can be equated with the mild deficit on Stage 1 of T-maze alternation that appears to vanish with extended training (Stage 2). The consequent matching of performance levels by the end of Stage 2 then helped to highlight the particular sensitivity of Stage 3 to retrosplenial damage. The radial-arm maze performance in the present study is again consistent with the idea that retrosplenial lesions only produce a mild allocentric deficit. On initial task acquisition there was no clear lesion effect, though during this period rats could use various strategies including allocentric, intra-maze and body turn (always turn clockwise or anticlockwise). Our sequential arm choice analysis clearly revealed, however, that this last strategy was not used. The present study reinforces the notion that retrosplenial cortex loss disrupts multiple spatial strategies, i.e. not just allocentric. Other support comes from the finding that inactivation of the retrosplenial cortex with tetracaine is most disruptive to radial-arm maze