Discordance of Spatial Representation in Ensembles of Hippocampal Place Cells

Transcription

1 Discordance of Spatial Representation in Ensembles of Hippocampal Place Cells Heikki Tanila, 1 Matthew L. Shapiro, 2 and Howard Eichenbaum 3 * 1 Department of Neuroscience and Neurology, University of Kuopio, Kuopio, Finland 2 Department of Psychology, McGill University, Montreal, Canada 3 Department of Psychology, Boston University, Boston, Massachusetts HIPPOCAMPUS 7: (1997) ABSTRACT: The extent to which small ensembles of neighboring hippocampal neurons alter their spatial firing patterns concurrently in response to stimulus manipulations was examined in young adult rats as well as in aged rats with and without memory impairment. Recordings from CA1 and CA3 cells were taken as rats performed a spatial radial-maze task that employed prominent distal visual stimuli attached to dark curtains surrounding the maze and local cues on each maze arm provided by inserts with distinctive visual, tactile, and olfactory stimuli. To test the influence of the different stimulus subsets, the distal and local cues were rotated 90 in opposite directions (a Double Rotation). In response to this manipulation, place fields could maintain a fixed position to room coordinates, rotate with either the local or the distal cues, disappear, or new fields could appear. On average 79% of the cells within an ensemble responded in the same way, but only 37% of all ensembles were fully concordant. Typically discordant ensembles had place fields that rotated with one set of cues, whereas the other fields disappeared or new fields appeared. Ensembles in which the place fields rotated in two opposite directions were less frequent in young rats than would be expected by the occurrence of the individual responses, indicating selective competition between directly conflicting representations and ultimate suppression of one. These findings indicate that hippocampal neurons independently encode distinct subsets of the cues in a complex environment, although processing within the hippocampal network may actively reduce the simultaneous representation of conflicting orientation information. This kind of population activity might reflect the higher-order organization of new memories within an established knowledge framework or schema. Concordance was higher in aged memory-impaired rats than in young rats, and the suppression of conflicting representations was absent in these rats. These findings suggest that age-related memory impairment is at least in part associated with a decrease in the scope of information coded and in the coordination of encoded representations. Hippocampus 1997; 7: Wiley-Liss, Inc. KEY WORDS: aging hippocampus; place cells; spatial learning; memory; Grant sponsor: NIA; Grant sponsor: N ational Institute of Mental H ealth; Grant sponsor: Academy of Finland; Grant sponsor: MRC of Canada; Grant sponsor: NSERC of Canada. *Correspondence to: H oward Eichenbaum, Ph. D., Laboratory of Cognitive N eurobiology, D epartment of Psychology, Boston U niversity, 64 Cummington Street, Boston, MA hbe@bu.edu Accepted for publication 15 May 1997 INTRODUCTION An intact hippocampus is necessary for successful spatial learning and memory in humans, monkeys, and rats (Morris et al., 1982; Parkinson et al., 1988; Smith and Milner, 1989). Perhaps the strongest evidence for a link between hippocampal function and spatial information processing is the finding that most hippocampal pyramidal cells fire selectively when the animal occupies a restricted portion of the environment (O Keefe and Dostrovsky, 1971). The existence of these place cells may reflect the maintenance of an internalized or cognitive map of the current environment (O Keefe and Nadel, 1978; McNaughton et al., 1996). According to cognitive map theory the active subset of place cells should all reliably maintain the same spatial selective firing pattern as long as the same map or reference frame is being used for navigation. Experimental evidence thus far supports this notion. In a simple environment such as a gray cylinder with a single polarizing cue card, rotation of the cue usually results in equal rotation of firing fields without a change in the field shape or size (Muller and Kubie, 1987), and the place fields of all cells within an ensemble recorded simultaneously rotate concordantly (Knierim et al., 1995). In addition, simultaneous recordings of thalamic head direction cells and hippocampal place cells indicate that these cells are strongly coupled so that a rotation for either set of cells is always accompanied by corresponding rotation for the second set (Knierim et al., 1995). It is not known, however, whether an ensemble of hippocampal place cells responds corcordantly to selective rotation of a subset of cues in an environment that includes the multiple cues, and it is not clear whether or how hippocampal cell assemblies encode relations among distal and local cues. While most conceptions of place cells assume that the spatial maps primarily use distal cues, Young et al. (1994) showed that place cells can be controlled by salient local stimuli as well as distal cues. In 1997 WILEY-LISS, INC.

2 614 TANILA ET AL. our own recent study, sets of local and distal cues that served as landmarks were rotated to opposite directions (Double Rotation) so that they gave conflicting information about orientation to the rat in a radial arm maze (Shapiro et al., 1997). In this setting the location of place fields for individual hippocampal neurons was controlled not only by local or distal cues but also by fixed room cues or a combination of different types of cues. These findings suggest that some hippocampal cells encode only a subset of the available cues and their spatial relationships, and that the cognitive map is composed of the collection of many distinct codings that reflect particular cue relations. However, this result could appear if all the place cells responded the same way during each Double Rotation, but if the response of the cells followed one subset of cues on some trials, and a different subset on other trials, leading to the observed overall mixture of responses. Such a full concordance is more consistent with the view that each cell reflects one location within a unified map that cannot be decomposed. The present study was also designed to identify any changes in the extent of concordance in hippocampal ensembles associated with aging and loss of spatial memory capacity. Both anatomical and physiological findings indicate that up to one-third of perforant path inputs to dentate granule cells are lost with age (Geinisman et al., 1986; Barnes, 1988). It has been suggested that the role of the dentate gyrus is to transform potentially highly correlated patterns in the entorhinal cortex into much sparser and much less correlated patterns (Marr, 1971; Skaggs and McNaughton, 1992). Loss of specific inputs is likely to compromise spatial processing in aged animals. Furthermore, there is evidence for an age-related increase in electrotonic coupling between principal cells in all hippocampal subfields (Barnes et al., 1987). To the extent that discordance in place cell ensembles occurs in young animals, it should be less prevalent in aged animals due to these changes. Using tetrode recordings which allowed simultaneous recording of two to ten place cells in a freely moving rat we were able to address the following questions: 1. When local and distal cues rotate to opposite directions, firing fields of individual place cells were shown to rotate with either set of cues or to remain in the original location. Do all the place fields of simultaneously recorded cells rotate concordantly? 2. Occasionally completely new place fields appear in Double Rotation trials indicating recoding of the altered environment. Does this recoding happen in all recorded place cells simultaneously? 3. Do ensembles of place cells in aged rats differ from those of young rats in their concordance? If they do, does the altered concordance correlate with spatial learning ability and memory? A preliminary version of these results have been presented in abstract form (Tanila et al., 1996). Subjects MATERIALS AND METHODS Six young (4 6 months, weighing g) and seven aged (25 29 months, weighing g) male Long-Evans rats served as subjects. The animals were individually housed from the beginning of pretraining, maintained on 12:12 h light:dark cycle, and given ad libitum access to water. Food consumption was controlled to prevent weight increase during the experiment. The health of the aged rats was followed by monitoring their food and water consumption and general health. When the brains of the rats were removed for histological verification of the electrode marks, all brains were carefully inspected for tumors. The aged rats were prescreened by M. Gallager (Dept. Psychology, UNC, Chapel Hill, NC) in the Morris water maze. Three of the aged rats had their learning index (Gallagher et al., 1993) within the range of young rats, and were therefore considered memory-intact. Four of the aged rats that had their learning indices outside the range of young rats were considered memory-impaired. The experimenter conducting the recordings was blind to the learning index of the aged rats. Electrodes, Surgery, and Data Acquisition The recording system has been described in detail earlier (Hetherington and Shapiro, 1997). The recording electrodes were tetrodes made of 30- m Formvar-coated nichrome wires, and the stimulation electrodes were a twisted pair of 100- m tefloncoated stainless steel wires. The animals were anesthetized with ketamine (50 mg/kg) and xylazine (7.5 mg/kg i.m.). Using asceptic surgical procedures, the electrodes were implanted stereotactically at the following coordinates: the stimulation electrodes were aimed at the lateral hypothalamus 0.5 mm posterior and 1.5 lateral to bregma, and 7.7 mm below the pial surface (tip of the longer wire); the tetrode was aimed at dorsal CA1 3.3 mm posterior and 2.0 lateral to bregma, and mm below the pial surface. The microdrive and the connector were attached to the surface of the skull by dental cement and four stainless steel screws, two of which served as the electrical ground. The data were digitized and analyzed using Enhanced Discovery and Autocut software (DataWave Technologies Inc.). The position of the rat in the maze was determined by a video camera following system (DataWave Technologies Inc.) that tracked two incandescent light bulbs mounted on the headstage assembly. Location was digitized in the form or x- and y-coordinate pairs at the rate of 60 Hz. Only complex spike cells (Ranck, 1973) with a duration of the negative spike more than 300 s and a signal-to-noise ratio more than 3:1 were sampled for this study. Behavioral Apparatus The apparatus consisted of a four-arm radial maze elevated 70 cm above the floor. The maze had a central octagonal platform 12 cm on each side and four maze arms each of which was 45 cm long, 10 cm wide, and with 6-cm angled sides at the end of each arm. The apparatus was surrounded by four 175-cm-wide dark blue curtains each of which supported a distinct 30- to 90-cmwide, complex contrasting pattern that served as a distal cue. The local cues were surfaces that overlaid the maze arms and were composed of coarse plastic mesh, sandpaper, fine wire mesh, or coarsely ridged rubber. In addition each arm was sprayed daily

3 DISCORDANT PLACE CELL ENSEMBLES 615 with an aerosol of a common food odor (anise, coconut, strawberry, or peppermint). The maze was illuminated by four 12 VDC lights located symmetrically on a ceiling panel above the maze. White noise was delivered by two speakers on the ceiling panel. The curtained enclosure could be entered from either of two opposite corners. Behavioral Procedures While the tetrode was slowly moved toward CA1 over a period of 1 to 2 weeks, the rats were trained to visit the ends of the arms and return to the center by rewarding them with electrical stimulation to the lateral hypothalamus (0.5 trains/s of 0.5-ms pulses at 100 Hz, A). The stimulus current was adjusted to the minimal level that kept the rat constantly moving. When this was achieved a working memory contingency was introduced such that rats received rewarding stimulation only when they ran to the end of an arm not previously visited on that trial. When all four arms had been visited, a new trial began immediately. All the rats adopted a stereotypic movement pattern that involved runs into adjacent arms, occasionally changing the direction of rotation. During each recording session place fields of the isolated cells were first mapped with all the cues in the arrangement used during training; this is referred to below as the Standard condition. This was followed by a Double Rotation condition during which all the distal cues were rotated 90 clockwise and all the local cues were rotated 90 counterclockwise, or vice versa. Finally all the cues were arranged back to their original position (Standard condition 2). Data was collected continuously over a 5- to 10-min period for each condition (which contained eight to 16 trials), and the rat was placed in a round covered bucket between the conditions. Prior to the first trial of each condition the bucket was spun slowly to disorient the rat. In addition, the start arm was alternated pseudorandomly. Furthermore, white noise was presented from two speakers above the maze to mask auditory cues from the computers which could help the rat s orientation to the room. Histology At the end of the recordings the rats were deeply anesthetized with pentobarbiturate (60 mg/kg i.p.). The electrode tips were marked by passing 15 A anodal current for 10 s, which plated out iron from the tips of nichrome electrodes. The rat was perfused with saline followed by 4% formalin, 4% potassium ferrocyanide, and 4% glacial acetic acid. The brains were removed, carefully inspected for any gross pathology, and soaked in formalin followed by 30% sucrose. Standard 30- m frozen sections were cut, and the slices were stained with cresyl violet. The location of electrode tips were identified by the Prussian blue reaction, and the recording sites were determined by estimating distances along the electrode track associated with microelectrode position at the time of the recording. Statistical Analysis For each condition the spatial distribution of firing rates was calculated by dividing the maze into 3-cm 3-cm pixels and computing firing rate for each pixel as the total number of spikes divided by the total time spent in that pixel. To maintain a relatively constant behavioral condition, firing rates were calculated only for periods when the rat was moving at least 2 cm/s. A place field was defined as an area of at least three adjacent pixels each having a firing rate at least three times the grand mean rate (total number of spikes/total time spent moving in the maze), and a mean within-field (infield) firing rate at least five times the overall mean firing rate for that neuron. Unit responses to manipulations of the environmental stimuli in Double Rotation were assessed by comparing the firing rate and location of each place field to the preceding Standard condition. The place field was considered to be fixed if a place field appeared in the same arm with an axial shift less than one-quarter of the arm length. The place field was considered to be rotated if a place field appeared in a different arm with an axial shift less than onequarter of the arm length. New place fields were identified as firing meeting the criteria of a place field but not localized within the preceding constraints. The place field was considered to have disappeared if no field could be found in the original or rotated location. Using these criteria, four major types of responses to cue manipulations were scored as follows: 1) No change if the place field was fixed, 2) Rotation with distal cues if the location of the place field corresponded to the new location of one or more distal cues, 3) Rotation with local cues if the location of place field corresponded to the new location of one or more local cues, and 4) New representation if the place field disappeared or any new place fields appeared in a location that did not correspond to the positions of fixed, distal, or local cues. With one tetrode per animal it was possible to record simultaneously two to ten complex spike cells with distinct place fields. These simultaneously recorded cells are hereafter referred to as ensembles. The mean concordance for the ensemble was calculated as the percentage of cells that responded in the same way as did the majority of cells within the ensemble. For example, if the place fields of four of the cells in a six-neuron ensemble rotated with the distal cues and the place fields of two of the cells disappeared, the mean concordance would be four out of six, or 67%. In cases where one cell had two subfields that responded differently, each subfield was given the weight of half a cell. For example, if the place field of one cell in a three-neuron ensemble rotated with the distal cues, two place fields of the second cell disappeared, and one subfield of the third cell disappeared and its other subfield rotated with the distal cues, the mean concordance would be out of 3, or 50%. The distribution of the four major response types in ensembles of the three experimental groups was analyzed separately for the concordant and discordant ensembles using Pearson s chi-square. An over- or under-representation of a particular response type in a group was estimated by the adjusted residuals in the cross tabulation. An adjusted residual with an eigenvalue larger than 1.9, corresponding to two standard deviations, was considered statistically significant. When an ensemble was discordant such that two or more response types were present during Double Rotation, a further analysis was performed to determine whether particular response

4 616 TANILA ET AL. combinations appeared as frequently as expected by the occurrence of individual responses. For this purpose each ensemble was assigned binary codes for each response category (1 present, 0 absent), and the data was analyzed using Model Selection Log Linear Analysis of the SPSS for Windows. For any combination of two responses it was ascertained whether their occurrence could be represented by a log-linear model that did not have any interaction terms, that is, that the responses appeared independently of each other. Statistical significance was set to Pearson s chi test at the level of P.05. RESULTS Data were collected from 137 complex spike cells in 35 ensembles from six young rats, 97 cells in 24 ensembles from three aged rats who had performed as well as young rats in spatial learning, and 105 cells in 24 ensembles from four aged rats characterized as impaired in spatial learning ability. A detailed description of the firing properties of the individual cells has been compiled into another report (Tanila et al., 1997). The number of major response types encountered in ensembles from each experimental group are summarized in Table 1. Overall the mean concordance was relatively high, such that an average of 79% of the cells responded in the same way as did the majority of cells within an ensemble. The mean concordance tended to be higher in ensembles of aged memory-impaired rats than in memory-intact aged rats or young rats (ANOVA, F(2, 82) 2.6, P.08; Fig. 1). Even though the overall effect of aging was only marginally significant, a post-hoc Duncan s test indicated that the concordance of aged memory-impaired rats differed significantly from that of young rats (P.05). Despite the relatively high proportion of cells that were concordant within the ensembles, only a minority of the ensembles were entirely concordant in response to the Double Rotation. On average 37% of the ensembles were fully concordant, with 31.4% in young rats, 37.5% in aged memory-intact rats, and 45.8% in aged memory-impaired rats showing this pattern. A typical concordant ensemble is shown in Figure 2. The distribution of major response types in fully concordant ensembles is shown in Table 2. Subject groups differed significantly TABLE 1. Number of Ensembles With 1, 2, or 3 4 Major Response Types Number of response types Young Aged intact Aged impaired FIGURE 1. Mean ( SEM) ensemble concordance in the three experimental groups. The asterisk indicates significantly different concordance from that in the group of young rats (Duncan s test, P.05). in that all four major response types were represented in the fully condordant ensembles of young rats, whereas only distal and local rotation response types were represented in the fully condordant ensembles of aged rats (Pearson chi (6) 20.8, P.01). The distribution of combinations of major response types in ensembles that were at least partially discordant are shown in Table 3. In these ensembles the distribution of response types did not differ between the groups (Pearson chi (16) 19.3, P.25). The most common response combination in ensembles of all groups was rotation with the distal cues combined with new representations. In young rats, the new representations usually appeared as new place fields that represented a particular conjunction between distal and local cues unique to the Double Rotation (Fig. 3). In aged memory-impaired rats all new representations involved the disappearance of place fields that had been present during the Standard condition (Fig. 4). In young rats relatively fewer ensembles had place fields that followed the distal cues while others followed the local cues (Fig. 5). In aged memory-impaired rats the most prominent observation was ensembles composed of some place fields that rotated with the distal cues while others developed a new representation. The major observation of this study is that ensembles of hippocampal neurons can be discordant in their response to stimulus manipulations that result in ambiguous spatial orientation cues. Moreover, all possible types of spatial firing responses can coexist in the same ensemble. In order to further determine whether any combination of two response types coexist as frequently as predicted by their overall occurrence, we tested possible interactions between all six two-way combinations of the four general response types using a log-linear model. In the group of young rats our examination showed that five combinations out of six did not involve a significant interaction between the response types (Table 4). Only the combination of a rotation with the distal cues together with a rotation with the local cues appeared less frequently than expected by the occurrence of

5 DISCORDANT PLACE CELL ENSEMBLES 617 FIGURE 2. A typical example of a concordant ensemble of four cells recorded in an aged memory-intact rat. The schematic at the top of the figure illustrates the experimental manipulation. Local cues are depicted as numbered patterns on the four maze arms, and distal cues are represented by alphabetic symbols. Left column: Place fields from four cells recorded in the first Standard condition. Middle column: Place fields of the same four cells recorded in the Double Rotation condition. Right column: Superimposed waveforms for each cell recorded in the four channels of the tetrode. In this ensemble the place fields of all the cells followed the distal cues as they rotated clockwise (arrows). individual response types, indicating a significant interaction between these two response types (Pearson chi (1) 4.1, P.05). The interaction between these two response types was marginally significant in aged memory-intact rats (Pearson chi (1) 3.2, P.08) and not significant in aged memory-impaired rats (Pearson chi (1) 0.14, P.70). The observed combination of some place fields rotating with the distal cues and other fields rotating with the local cues suggests some degree of uncertainty about orientation in the environment. One possible explanation is that the cellular ensemble flips between two states representing these two orientations. Therefore the most obvious case (Figure 5) was analyzed trial by trial to determine whether both directions of rotation are present on identical trials. Place fields with opposite rotations indeed were observed to coexist during the same trial as shown in Figure 6.

6 618 TANILA ET AL. TABLE 2. Distribution of Major Response Types Among Concordant Ensembles* Response type Young Aged Memory-intact DISCUSSION Discordance of Spatial Representations by Hippocampal Place Cells Aged Memory-impaired NC RD RL NR *NC, no change; RD, rotation with the distal cues; RL, rotation with the local cues; NR, new representation;, more than 2 standard deviations above the mean;, more than 2 standard deviation below the mean. In the present study simultaneously recorded hippocampal place cells responded differently to selective alterations of cue relationships in an environment with multiple orienting cues. Indeed most of the small hippocampal ensembles in the present study demonstrated discordance of the spatial representation. Some of the observed discordance might be attributable to the disappearance of place fields for cells whose spatial firing was initially activated by particular cues and after the rotation falls below a threshold when some cues are moved. However, such an explanation would not account for place fields that are still present after the rotation but rotate in different directions for distinct sets of cues. These findings contrast with earlier observations on ensembles recorded as rats explored a plain cylindrical testing environment with a single vertical stripe that provided the orienting cue, and where changes in place field representations occurred when the rats were disoriented between repeated trials. Based on recordings in this test environment Knierim et al. (1995) reported that a rotation of a place field or appearance of a new field ( remapping ) of one hippocampal place cell was always accompanied by a corresponding change in the fields of other simultaneously recorded place cells or thalamic head direction cells. In another experiment using the same type of environment, it was also found that all simultaneously recorded cells showed concordant responses when the color of the orienting cue was changed (Bostock et al., 1991). However, in situations in which the running patterns of the rat were changed by altering the reward loci, at least one example of discordance was observed (see Fig. 9 in Markus et al., 1995). In none of these situations were the alterations of different subsets of the spatial cues in direct conflict. Thus, it is not clear whether the frequent observation of discordance in the present study is a consequence of the complexity of the environment or the explicit ambiguity that follows the cue manipulations. O Keefe and Speakman (1987) examined the effects of rotation of controlled distal cues against a fixed background and found that place fields of simultaneously recorded cells were controlled to differing extents by distal and fixed environmental cues. Collectively across these studies, various hints of discordance noted in previous studies were confirmed and extended by the present observations of considerable discordance within a complex environment composed of many controlled cues. Sources of Cues Employed in Discordant Spatial Representations Some place cells responded to the Double Rotation by corresponding rotations of the place fields with different subsets of the cues, whereas other place fields either maintained their location with respect to room coordinates or recoded the environment as reflected by a cessation or unpredictable change in the spatial firing pattern. With regard to cells that maintained their place fields, there were several potential sources of consistent orienting cues that could have been encoded. Thus, even though rats were spun in a bucket before each Double Rotation and the start arm was varied pseudorandomly, the experimenter always left the curtained enclosure from the same corner providing a consistent transient cue. Also, the animal may have been able to hear the background noise from nearby recording equipment or other sources that maintained a consistent position throughout the Double Rotation. With regard to cells that recoded the environment, the simplest explanation is that these cells had encoded spatial relations among the distal, local, and fixed cues in the Standard environment and that these relations were disrupted by the Double Rotation. In particular, the appearance of a new place field was confirmed to reflect the novel spatial configuration between distal and local cues that appeared after Double Rotation, as verified by an analysis of further probe trials (Tanila et al., 1997). TABLE 3. Distribution of Response Types Among Discordant Ensembles* Response type Young Aged memoryintact Aged memoryimpaired NC RD NC NR RD RL RD NR RL NR NC RD NR NC RL NR RD RL NR NC RD RL N *NC, no change; RD, rotation with the distal cues; RL, rotation with the local cues; NR, new representation;, more than 2 standard deviations above the mean;, more than 2 standard deviation below the mean.

7 DISCORDANT PLACE CELL ENSEMBLES 619 FIGURE 3. A typical example of a discordant ensemble of four cells recorded in a young rat (see Fig. 2 legend for explanation of the figure organization; subfields are distinct place fields recorded from the same single cell). In the Standard condition only cell 2 had a place field, whereas the other cells were almost silent. In response to the Double Rotation cell the place field of cell 2 rotated counterclockwise with the distal cues. In all the other cells place fields appeared after the Double Rotation, indicating a recoding of the reconfigured environment. All the place fields were directional: firing associated with place fields located on arm 4 in the Double Rotation condition occurred only during outward runs, whereas firing associated with the place field on arm 3 occurred only during inward runs. Hippocampal Spatial Representations: A Unified Cognitive Map or a Collection of Codings of Specific Cues and Their Spatial Relations? These combined findings show that small ensembles of simultaneously recorded hippocampal cells incorporate distinct subsets of the cues and their spatial configuration, providing a strong challenge to the view that hippocampal place cells encode loci or pointers within a unified cohesive map of the environment (O Keefe and Nadel, 1978; McNaughton et al., 1996). Our findings are similar to those of Gothard et al. (1996), who examined the firing patterns of multiple simultaneously recorded cells as rats explored an environment with two prominent local landmarks and a start box, both of which were moved against the static background cues. They found cells whose firing patterns were determined by one of those cues ( reference frames ) independently, as well as some cells that had separate firing fields

8 620 TANILA ET AL. FIGURE 4. A combination of rotated place fields and place fields that disappeared in response to Double Rotation in the ensemble of an aged memory-impaired rat (see Fig. 2 legend for explanation of the figure organization). One of the place fields of cell 1, as well as the place fields of cells 2 and 3, rotated counterclockwise, whereas the other subfield of cell 1 disappeared and the place field of the cell 4 disappeared. in more than one reference frame. The authors concluded that the representation of the environment as a cognitive map can consist of several submaps, each bound to a different reference frame (i.e., the start box, local landmarks, and background cues), but that only one reference frame can be represented simultaneously. The present data calls the latter conclusion into question. If in the present study the configurations of the fixed, distal, and local cues are considered as different reference frames, our findings suggest that several reference frames can be represented simultaneously within the resolution of a single trial. One possibility is that the hippocampus switched reference frames within trials, for example between going into and out of arms (Gothard et al., 1995). However, this interpretation would imply that rats do not employ a consistent cognitive map of the environment to guide their continuous approach toward a goal within a single trial. Furthermore, our data showing the same directionality in adjacent but discordant place fields (e.g., Fig. 3) indicate that movement direction cannot be taken to reflect what reference frame is used, as Gothard et al. (1995) suggested. Instead, the present findings support the more parsimonious notion that rats do indeed

9 DISCORDANT PLACE CELL ENSEMBLES 621 FIGURE 5. An ensemble from aged memory-impaired rat in which the place fields rotated in opposite directions. Cell 1 had three subfields in the Standard condition. In the Double Rotation condition subfield 1 rotated clockwise with the local cues, whereas subfield 2 rotated counterclockwise with the distal cues, and subfield 3 disappeared. Cell 4 had a large place field in the Standard condition that appeared to split, and its subsequent subfields rotated in opposite directions in response to the Double Rotation. The summed firing volume of the two fields in the Double Rotation was 231 (subfield 1: area 5 mean rate ; subfield 2: area 21 mean rate ), which is less than the firing volume of the large field in the Standard condition (area 38 mean rate ). have and use a constant map of the environment continuously, and that the firing patterns of single hippocampal neurons capture relationships between select subsets of the available cues. By such a view then, the cognitive map is the sum of these independent associations (Eichenbaum et al., 1992). The apparent cohesion or unification of the map, and corresponding capacities for navigational inferences (Eichenbaum et al., 1990), would arise through the combination of overlapping associations, consistent with recent evidence that rats with hippocampal damage are selectively impaired in guiding behavior based on indirect associations (Bunsey and Eichenbaum, 1996; Dusek and Eichenbaum, 1997). The coding for relationships between environmental cues by hippocampal place cells seemed not entirely independent, however. In young rats the coexistence of cells whose place fields followed the distal cues combined with others that followed the local cues was less frequent than expected by the occurrence of

10 622 TANILA ET AL. TABLE 4. Co-Occurrence of Major Response Types in Ensembles of Young Rats RD RL NR NC RD 2.03* 0.64 RL 0.49 The figures show the adjusted standardized residuals (observed cases minus expected cases). Note that most of the residuals are negative due to the fact that mean ensemble concordance was above 50%. *P.05, Pearson s chi. each individual response type. This finding might reflect network processing that is attempting to resolve conflicting information regarding orientation to the environment. This process may be based on competitive interactions among the specific representations, and the consequent suppression of some losing ones. This kind of alignment process to reconcile head direction (postsubiculum), path integration (pre/parasubiculum), and place information (hippocampus proper) has been proposed to be executed by the subiculum (McNaughton et al., 1996; Redish and Touretzky, 1997). Through its projections to the entorhinal cortex, the subiculum could enhance the activity of a subset of hippocampal place cells which, combined with competitive inhibitory processes in the hippocampus, would result in suppression of place cells with opposite directional information (Redish and Touretzky, 1997). By contrast, coexistence of rotated fields with fixed or new fields would not reflect the explicit conflict, because the fixed fields can be viewed as containing information about the past (Standard) environment and the new fields can be viewed as reflecting the encoding of newly appearing features within the novel environment (Double Rotation). Indeed the coexistence of rotated fields with fixed or new fields in the active population of hippocampal pyramidal cells could be employed within the hippocampal network to make cross-temporal comparisons that analyze how the arrangements of cues in the current environment differ from the arrangement in the past, rather than information specifically about orientation in the environment. This kind of mismatch function of the hippocampal neurons has been well documented in olfactory delayed non-match-to-sample task (Otto and Eichenbaum, 1992) and other situations in which expectations and observations about cues mismatch (Ranck, 1973; O Keefe, 1976). Through such a process, only generally implicated here, the hippocampus could participate in the coherent organization of new and old stimulus relations that would provide the animal with a sense of both consistencies and distinctions between the two related situations. The finding of increased ensemble concordance selectively in aged rats with spatial memory impairment indicates that this parameter is related to information processing capacity of the brain. Increased concordance is consistent with the notion of age-related increase in electrotonic coupling between hippocampal principal cells (Barnes et al., 1987). If indeed such a cellular mechanism underlies the increased concordance, the decrease in capacity for pattern separation and sparse coding of information could be interpreted as the source of the diminished memory capacity of the system. Another feature characteristic of ensembles in memory-impaired aged rats was the increased coexistence of place cells that reflected conflicting orientation information (cells with fields rotating in opposite directions during the Double Rotation). This finding also suggests a diminished capacity of the integrative network computations that might suppress such conflicting codings as suggested above. Our findings with only few cells of this kind are only suggestive but warrant further studies on the issue. FIGURE 6. A trial-by-trial analysis of the Double Rotation condition for Cell 4 from the ensemble shown in Figure 5. Data are shown for ten trials (numbers to the left of each panel) each consisting of a visit to each of the four maze arms. The stippled pattern represents positions of the rat at 60 Hz as it explores the four-arm radial maze. Cell firing rate is depicted by circles the diameter of which corresponds to the number of spikes in a firing episode. Note that during trials 2, 3, 5, 8, and 9 the cell fires at the location of both subfields of the cell.

11 DISCORDANT PLACE CELL ENSEMBLES 623 Acknowledgments We thank Ms. J. Niedermair for preparation of histological material for confirmation of electrode locations and Ms. Pirjo Halonen, M.Sc., for assistance in the statistical analysis. M. Shapiro received salary support from McGill University during his sabbatical. REFERENCES Barnes CA. Aging and the physiology of spatial memory. Neurobiol Aging 1988;9: Barnes CA, Rao G, McNaughton BL. Increased electrotonic coupling in aged rat hippocampus: a possible mechanism for cellular excitability changes. J Comp Neurol 1987;259: Bostock E, Muller RU, Kubie JL. Experience dependent modifications of hippocampal place cell firing. Hippocampus 1991;1: Bunsey M, Eichenbaum H. Conservation of hippocampal memory function in rats and humans. Nature 1996;379: Dusek JA, Eichenbaum H. The hippocampus and memory for orderly stimulus relations. Proc Natl Acad Sci USA 1997;94: Eichenbaum H, Stewart C, Morris RGM. Hippocampal representation in place learning. J Neurosci 1990;10: Eichenbaum H, Otto T, Cohen NJ. The hippocampus what does it do? Behav Neural Biol 1992;57:2 36. Gallagher M, Burwell RD, Burchinal M. Severity of spatial learning impairment in aging: development of a learning index for performance in the Morris water maze. Behav Neurosci 1993;107: Geinisman Y, de-toledo-morrell L, Morrell F. Loss of perforated synapses in the dentate gyrus: morphological substrate of memory deficit in aged rats. Proc Natl Acad Sci USA 1986;83: Gothard KM, Skaggs WE, McNaughton BL. Interaction between multiple spatial reference frames in the rat hippocampus. Soc Neurosci Abstr 1995;21:941. Gothard KM, Skaggs WE, Moore KM, McNaughton BL. Binding of hippocampal CA1 neural activity to multiple reference frames in a landmark-based navigation task. J Neurosci 1996;16: Hetherington PA, Shapiro ML. Hippocampal place fields are altered by the removal of single visual cues in a distant-dependent manner. Behav Neurosci 1997;111: Knierim JJ, Kudrimoti HS, McNaughton BL. Place cells, head direction cells, and the learning of landmark stability. J Neurosci 1995;15: Markus EJ, Qin Y-L, Leonard B, Skaggs WE, McNaughton BL, Barnes CA. Interactions between location and task affect the spatial and directional firing of hippocampal neurons. J Neurosci 1995;15: Marr D. Simple memory: a theory for archicortex. Proc R Soc Lond [Biol] 1971;262: McNaughton BL, Barnes CA, Gerrard JL, Gothard K, Jung MW, Knierim JJ, Kudrimoti H, Qin Y, Skaggs WE, Suster M, Weaver KL. Deciphering the hippocampal polyglot: the hippocampus as a path integration system. J Exp Biol 1996;199: Morris RG, Garrud P, Rawlins JN, O Keefe J. Place navigation impaired in rats with hippocampal lesions. Nature 1982;297: Muller RU, Kubie JL. The effects of changes in the environment on the spatial firing of hippocampal complex-spike cells. J Neurosci 1987;7: O Keefe JA. Place units in the hippocampus of the freely moving rat. Exp Neurol 1976;51: O Keefe J, Dostrovsky J. The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res 1971;34: O Keefe J, Nadel L. The hippocampus as a cognitive map. Oxford: Clarendon, O Keefe J, Speakman A. Single unit activity in the rat hippocampus during a spatial memory task. Exp Brain Res 1987;68:1 27. Otto T, Eichenbaum H. Neuronal activity in the hippocampus during delayed non-match to sample performance in rats: evidence for hippocampal processing in recognition memory. Hippocampus 1992; 2: Parkinson JK, Murray EA, Mishkin M. A selective mnemonic role for the hippocampus in monkeys: memory for the location of objects. J Neurosci 1988;8: Ranck JB, Jr. Studies on single neurons in the dorsal hippocampal formation and septum in unrestrained rats. Exp Neurol 1973;41: Redish DA, Touretzky DS. Cognitive maps beyond the hippocampus. Hippocampus 1997;7: Shapiro ML, Tanila H, Eichenbaum H. Cues that hippocampal place cells encode: Dynamic and hierarchical representation of local and distal stimuli. Hippocampus 1997;7: Skaggs WE, McNaughton BL. Computational approaches to hippocampal function. Curr Opin Neurobiol 1992;2: Smith ML, Milner B. Right hippocampal impairment in the recall of spatial location: encoding deficit or rapid forgetting? Neuropsychologia 1989;27: Tanila H, Shapiro ML, Eichenbaum H. Discordance within ensembles of CA1 and CA3 cells in coding for spatial relationship. Soc Neurosci Abstr 1996;22:430. Tanila H, Shapiro ML, Gallagher M, Eichenbaum H. Brain aging: changes in the nature of information coding by the hippocampus. J Neurosci 1997;17: Young BJ, Fox GD, Eichenbaum H. Correlates of hippocampal complexspike cell activity in rats performing a nonspatial radial maze task. J Neurosci 1994;14: