- Top of page
- Materials and methods
To test potential parallels between hippocampal and anterior thalamic function, rats with anterior thalamic lesions were trained on a series of biconditional learning tasks. The anterior thalamic lesions did not disrupt learning two biconditional associations in operant chambers where a specific auditory stimulus (tone or click) had a differential outcome depending on whether it was paired with a particular visual context (spot or checkered wall-paper) or a particular thermal context (warm or cool). Likewise, rats with anterior thalamic lesions successfully learnt a biconditional task when they were reinforced for digging in one of two distinct cups (containing either beads or shredded paper), depending on the particular appearance of the local context on which the cup was placed (one of two textured floors). In contrast, the same rats were severely impaired at learning the biconditional rule to select a specific cup when in a particular location within the test room. Place learning was then tested with a series of go/no-go discriminations. Rats with anterior thalamic nuclei lesions could learn to discriminate between two locations when they were approached from a constant direction. They could not, however, use this acquired location information to solve a subsequent spatial biconditional task where those same places dictated the correct choice of digging cup. Anterior thalamic lesions produced a selective, but severe, biconditional learning deficit when the task incorporated distal spatial cues. This deficit mirrors that seen in rats with hippocampal lesions, so extending potential interdependencies between the two sites.
- Top of page
- Materials and methods
Rats with lesions of the anterior thalamic nuclei were trained on a series of biconditional discriminations. These discriminations revealed a contrasting pattern of spared and impaired learning. Whether the biconditional task required the rats to approach a covered magazine following food delivery in a chamber (experiment 1) or to dig within cups containing different media (experiment 2E), the surgery did not appear to affect task acquisition when the appropriate choice behaviour was signalled by local contextual cues. That is, rats with anterior thalamic lesions could learn the biconditional rule ‘if A do Y, but if B do X’ when signalled by cues such as hot versus cold floors, dotted versus checkerboard walls (experiment 1), or by dark versus transparent walls and floors (experiment 2E). In contrast, the same surgeries prevented biconditional learning when the conditional cues consisted of different room locations, i.e., if place A do Y, if place B do X (experiments 2D, 2F and 4). For this same reason, the mild deficit in the anterior thalamic lesion group in experiment 2B (Fig. 7A) presumably reflected their ability to use local context cues, but not distal spatial cues, when both were present to help solve the same biconditional problem.
For these biconditional experiments, the conditional response comprised either nose-poking (experiment 1) or digging in different media (experiments 2, 3 and 4) for food. These particular behaviours were selected as there was no prior evidence that these responses should be affected by the anterior thalamic lesions. This assumption was borne out by the rats’ intact performance on the contextual nose-poking biconditional task (experiment 1) and by the normal acquisition of the digging media discrimination (experiment 2A; see also experiment 2E). Consequently, the critical feature in determining whether the anterior thalamic lesions disrupted biconditional learning in these experiments appears to be the nature of the conditional signal.
A number of previous studies have explored the impact of anterior thalamic lesions on biconditional learning, with varying outcomes. Based on the present findings it would seem appropriate to divide these previous studies into two categories. The first category comprises conditional stimuli or responses that are not defined by reference to distal spatial cues, e.g. that entail the use of local context cues, auditory signals or egocentric-based responding. The prediction is that anterior thalamic lesions will spare biconditional tasks confined to these elements. This prediction builds on the knowledge that anterior thalamic lesions typically spare the discrimination of simple, elemental stimuli and do not impair egocentric spatial tasks (Aggleton et al., 1996, 2009; Warburton et al., 1997; Sziklas & Petrides, 1999; Mitchell & Dalrymple-Alford, 2006; Wolff et al., 2008). The second category comprises biconditional tasks where the signal stimuli or the conditional responses are distinguished by location cues. This grouping is underpinned by evidence that anterior thalamic lesions impair a range of tasks thought to rely on distal spatial cues (Sutherland & Rodriguez, 1989; Aggleton et al., 1995; Byatt & Dalrymple-Alford, 1996; Warburton & Aggleton, 1999; Wilton et al., 2001; van Groen et al., 2002; Loukavenko et al., 2007).
Examples of the first category include when an animal forms a biconditional association between an item and a left–right position that can be defined egocentrically (Chudasama et al., 2001; Ridley et al., 2002). For instance, when presented with one of two visual stimuli on a screen, rats with anterior thalamic lesions learnt to nose-poke to the left or to the right according to the particular stimulus (Chudasama et al., 2001). Similarly, when marmoset monkeys (Callithrix jacchus) were presented with two copies of the same object (e.g. A1, A2) they were rewarded for selecting the item on the left, but when two copies of object B (B1, B2) were presented, they were rewarded for selecting the item on the right (Ridley et al., 2002). Lesions confined to the anterior thalamic nuclei did not impair performance on this task (Ridley et al., 2002). Likewise, rats with anterior thalamic lesions learnt to turn to the right or left depending on which object was placed at the choice point (Sziklas & Petrides, 1999, 2004). Consequently, tasks in this first category seem insensitive to anterior thalamic damage, as found in the present study.
Examples of the second category include evidence that anterior thalamic lesions impair the formation of spatial–visual biconditional associations (Sziklas & Petrides, 1999; Henry et al., 2004). In these experiments, rats chose one of two items depending on whether they were located at the north or south of an open-field. Likewise, anterior thalamic lesions impaired an odour-location biconditional task in a circular arena (Gibb et al., 2006). Rather like the present task, rats were rewarded for digging in one location when signalled by a medium with a particular odour, but when in a different location the rats were rewarded for digging in a medium with a different odour. Clearly there are strong parallels with the present results (experiments 2D, 2F and 4). There does, however, appear to be an exception to this general pattern when location cues appeared without effect. Rats with anterior thalamic lesions were able to select a particular location in a cross-maze depending on which item was present at the choice point of the maze (Sziklas & Petrides, 2007). Because the task involved approaching the choice point from different directions it was assumed that the rats had successfully used allocentric spatial information to determine the appropriate response (Sziklas & Petrides, 2007). Although this finding might indicate an asymmetry, such that anterior thalamic lesions disrupt biconditional tasks when the signal stimulus is location-specific but not when the conditional response is location-specific (Sziklas & Petrides, 2007), this interpretation is inconsistent with the findings by Gibb et al. (2006).
As noted earlier, the present study had two related goals. The first was to compare the profile of anterior thalamic lesion deficits with that following hippocampectomy. Lesions in these two sites seemingly have many of the same effects on tests of spatial and contextual learning (Beracochea et al., 1989; Sutherland & Rodriguez, 1989; Aggleton et al., 1995; Byatt & Dalrymple-Alford, 1996; Aggleton & Brown, 1999; Warburton et al., 2001; Law & Smith, 2012), leading to the notion that they have integrated functions (Aggleton & Brown, 1999). An apparent exception concerns some biconditional tasks as it has been found that hippocampal lesions can impair an egocentric conditional task that is spared by anterior thalamic damage (Sziklas & Petrides, 2004). Likewise, hippocampal, but not anterior thalamic, lesions can impair learning to go to a given location depending on the identity of an object at the choice point (Sziklas & Petrides, 2002, 2007). These findings suggest that anterior thalamic lesions and hippocampal lesions have different profiles of effect when considering spatial biconditional tasks.
The present study strongly supports the opposite view, i.e., lesions in the anterior thalamus and hippocampus have very similar consequences on biconditional discriminations. The present findings could be matched with previous biconditional tasks that used the same test chamber stimuli (hot/cold, checks/spots) as those in the present study. These experiments found that hippocampal lesions, like anterior thalamic lesions, spare acquisition of the biconditional rule (Coutureau et al., 2002). Likewise, prior experiments using the same biconditional digging task as that in the present study found that hippocampal lesions spared the ability to use local contextual cues to learn in which cup to dig for food rewards (Albasser et al., 2013). Then, just as in the present study, hippocampectomy impaired the ability to use room location cues to determine in which cup to dig (Albasser et al., 2013). The similarity between the impact of anterior thalamic and hippocampal lesions extended to the ability to learn a go/no-go place discrimination when trained by running in one direction (Albasser et al., 2013; experiment 3B present study), but then performing poorly on a subsequent biconditional that used this same spatial information. Consequently, the profiles of spared and impaired performance following anterior thalamic and hippocampal lesions on the tasks used in the present study appear extremely similar. Furthermore, others have also demonstrated that rats with hippocampal lesions are unable to form object-location and odour-location associations using distal room cues (Gilber & Kesner, 2002); however, these same rats were able to solve a non-spatial biconditional problem that involved object–odour associations. These results are also consistent with those reported in the present communication following anterior thalamic damage, once again highlighting the similarity between hippocampal and anterior thalamic function for biconditional learning. This conclusion closely accords with two disconnection studies showing that the hippocampus and anterior thalamic nuclei function together to solve the visuospatial biconditional problem, select object A at the north of an arena but select object B at the south end (Henry et al., 2004; Dumont et al., 2010).
The matching patterns of learning found after hippocampal (Albasser et al., 2013) and anterior thalamic lesions for the biconditional discriminations that involved either local contextual cues (spared) or distal location information (impaired) highlight the need to understand what qualitatively separates contextual cues from spatial cues. One potential difference concerns their proximity. Local cues, including visual cues, can be regarded as being available by direct exploration. That is, the cues are within the rat's ‘working space’ and, hence, not further than the tip of the nose or the vibrissae (Parron et al., 2004). There are, however, shortcomings with this proximal–distal distinction when trying to explain the present set of results. One problem is that the rats with anterior thalamic lesions could still acquire a spatial go/no-go task that relied on distal cues (experiment 3B, Fig. 8D). Furthermore, rats with anterior thalamic lesions can readily solve visual discriminations in a water maze in which the stimuli were selected from a distance (Aggleton et al., 2009; see also Ridley et al., 2002). There also remains the problem of deciding a priori when a cue is ‘distal’ and when it is ‘proximal’ (Good et al., 1998).
A different explanation focuses on the nature of the stimuli used in the various experiments. For the biconditional problems in the automated test chamber (experiment 1) and the test boxes with different appearances (experiment 2E), which the ATNx1 rats could readily solve, the critical stimuli could be discriminated by their individual salient features (e.g. different wall patterns, floor temperatures and floor coverings). As a consequence, any stimulus ambiguity from overlapping or common elements was kept low. In contrast, in those tasks where anterior thalamic lesions impaired performance (experiments 2D, 2F, 3A, 3B and 4), the rats had to use distal room cues to identify locations where presumably there would be overlap of common cues. This potential requirement to disambiguate common cues closely relates to the notion that the hippocampus is required for contextual learning when it inherently involves configural learning, reflecting the need to distinguish overlapping cues and utilise pattern separation (Gaffan & Harrison, 1989; Gilbert et al., 1998; Holland & Bouton, 1999; Rudy, 2009; Iordanov et al., 2011). Such functions might be expected to depend on the integrity of the extended hippocampal system, including the anterior thalamic nuclei, and so help to explain the present pattern of results. One problem with this account, however, concerns the variable effects of hippocampal lesions on configural tasks. Given the previous account, it might be supposed that visual configural tasks are consistently hippocampal-dependent, yet this is sometimes not the case (Rudy & Sutherland, 1995; Sanderson et al., 2006; Saksida et al., 2007).
One possible solution is to suppose that the hippocampal–anterior thalamic axis is important for a subset of configural problems. These problems require configural learning, but also involve determining the relative spatial positions of the common cues. Such configural problems require ‘structural learning’ as the animal has to learn not only which elements are combined in a given scene but also how these elements are positioned with respect to each other, i.e., how they are structured (George et al., 2001; George & Pearce, 2003). There is appreciable evidence that the hippocampus is required for this form of learning (Save et al., 1992; Aggleton & Pearce, 2001; Sanderson et al., 2006; Barker & Warburton, 2011; Albasser et al., 2013). While there is also evidence that the anterior thalamic nuclei are required for tests that should tax structural learning (Parker & Gaffan, 1997; Wilton et al., 2001), anterior thalamic lesions failed to impair a formal test of this form of learning (Aggleton et al., 2009).
An alternative, closely related, proposal concerns the role of the hippocampus for pattern separation (Gilbert et al., 1998; Hunsaker & Kesner, 2013; Leutgeb et al., 2007; Rolls & Treves, 1994). This process would help the animals to distinguish room cues when they overlap, a situation presumably more prevalent in the bidirectional than unidirectional tasks. Consequently, this account would explain the different results for the two types of go/no-go tasks. This account would also have to assume that pattern separation depends on anterior thalamic interactions with the hippocampus, an assumption that is not implausible given the results of disconnection studies (Dumont et al., 2010; Henry et al., 2004; Warburton et al., 2001). It is not clear, however, how a pattern-separation account will explain the failure to learn the biconditional problem (experiment 4) once the relevant location cues had been distinguished. While it could be argued that this additional biconditional deficit reflects the conflict between competing similar demands on the rats in the biconditional task, exaggerated by the use of spatial stimuli, this account remains largely post hoc in nature. A further possibility is that the deficit reflects the combination of pattern separation demands and a closely related mnemonic component (Hunsaker & Kesner, 2013).
This consideration of spatial learning brings us to the second goal of the present study, namely, to identify the nature of any observed biconditional learning deficits associated with anterior thalamic damage. As already noted, the use of distal spatial cues appears to be a common factor in many examples of biconditional tasks sensitive to anterior thalamic damage (Gibb et al., 2006; Henry et al., 2004; Sziklas & Petrides, 1999). One apparent exception, seemingly unaffected by anterior thalamic lesions, involved using object identity to signal whether to go to place A or place B for reward (Sziklas & Petrides, 2007). The goal of the final test (experiment 4) was, therefore, to understand more precisely what prevents rats with anterior thalamic lesions from solving most biconditional discriminations involving distal spatial cues. Consequently, rats were first trained on a spatial go/no-go task (experiment 3B). When measured by latency scores, the anterior thalamic lesions disrupted performance, resulting in more rapid response times overall. There is, however, a concern that this thalamic surgery can induce hyperactivity (Jenkins et al., 2004; Poirier & Aggleton, 2009) and so latency ratios might be more appropriate. It was, therefore, notable that the ATNx2 group seemed unimpaired when location discrimination performance was measured as a ratio of go/no-go latencies (Fig. 8D). However, despite learning this go/no-go location task, the thalamic-lesioned rats were still unable to use that same spatial information to guide a subsequent biconditional learning task (experiment 4). In fact, the rats with anterior thalamic lesions remained at chance (see Fig. 9). This dissociation clearly questions the parsimonious notion that it is the discrimination of distal location information per se that accounts for the pattern of anterior thalamic lesion deficits in the present study. Rather, the added burden of the biconditional problem left the ATNx2 rats at chance.
The present results strengthen the notion that the anterior thalamic nuclei and hippocampus work together to resolve spatial problems, including biconditional discriminations (see also Warburton et al., 2000, 2001; Henry et al., 2004). Given the dense, direct fornical projections from the hippocampus to the anterior thalamic nuclei it might naturally be supposed that fornix lesions would, therefore, match the impact of anterior thalamic damage on biconditional learning tasks. There are, however, problems with this prediction. Not only has it been found that anterior thalamic and fornix lesions can both spare spatial biconditional problems (Sziklas & Petrides, 2002, 2007) that are sensitive to hippocampectomy (Sziklas & Petrides, 2002), but of more concern is the finding that fornix lesions spare a spatial biconditional task (Dumont et al., 2007; Sziklas et al., 1998) that is sensitive both to anterior thalamic lesions (Sziklas & Petrides, 1999) and to crossed anterior thalamic–hippocampal lesions (Henry et al., 2004). Such findings (see also Aggleton et al., 2009; Warburton & Aggleton, 1999) either suggest the importance of indirect routes linking the hippocampus with the anterior thalamus, e.g. via the retrosplenial cortex (Vann et al., 2009), or indicate that critical thalamic contributions involved in biconditional learning emanate from the diencephalon and then target the hippocampus and, hence, do not require the fornix (Taube, 2007; Vann, 2009; Vann & Albasser, 2009). Both interpretations could be correct. In order to test the former notion, combined lesions were placed in the retrosplenial cortex and fornix, resulting in impaired learning of a spatial biconditional task otherwise spared by fornix lesions alone and by retrosplenial lesions alone (Dumont et al., 2007, 2010; St-Laurent et al., 2009; Sziklas et al., 1998). Such findings highlight the need to uncover the various pathways by which the anterior thalamic nuclei and the hippocampus might conjointly support forms of biconditional learning.