Mechanisms of Experience-Driven Synaptic Modification
A central question in neuroscience concerns the nature of nerve cell changes underlying learning and memory. The objectives of the current project are to elucidate the molecular changes occurring in the coupling between nerve cells taking part in a learning process. In particular, the project is concerned with the existence and maintenance of place cells in the hippocampus and dentate gyrus. In order to achieve these objectives we will employ a newly developed multi-tetrode recording technology that enables online monitoring of cellular activity in awake, behaving mice. We intend to employ this technique in transgenic mice with restricted and controllable genetic deletions of different glutamate receptor subtypes. The molecular alterations of these nerve cells lie at the foundation for new ways to diagnose, treat and prevent learning and memory defects.
Several different theories of hippocampal function have been proposed, notable among them the theory of the hippocampus as a cognitive map (O'Keefe and Nadel, 1978) and the episodic memory hypothesis of the hippocampus (Cohen and Eichenbaum, 1993). In addition, the hippocampus has been declared necessary for working memory (Olton and Samuelson, 1976), for spanning temporal discontiguities (Rawlins, 1985), and for novelty recognition (Gaffan, 1972). Although both the spatial map theory and the episodic memory theory have been highly influential, they are not without their weaknesses. Specifically, the spatial map theory cannot readily explain non-spatial consequences of hippocampal lesions (e.g. Good and Honey, 1997). The episodic memory theory, on the other hand, fails to explain how episodes or events are stored. In general, however, there is consensus that hippocampal cells do respond to spatial events and that temporary memory storage is, at least partly, provided by networks of hippocampal pyramidal cells.
The hippocampus is known to contain cells that are strikingly activated in particular locations. These 'place cells' can be defined as cells that fire when an animal is in a specific place (O'Keefe and Dostrovsky, 1971) and such cells have been found in CA1 and CA3 pyramidal cells (O'Keefe, 1979) as well as in dentate granule cells (Jung and McNaughton, 1993). Place fields, the boundaries of the area in which a given place cell is active, have been found to be stable across sessions separated by hours, weeks, or months (Muller et al., 1987; Thompson and Best, 1990). Place fields have been recorded from small, enclosed arenas (Muller and Kubie, 1987; Sharp et al., 1990), in the T-maze (O'Keefe, 1976), in the radial maze (Olton et al., 1978; Miller and Best, 1980) and on linear tracks (O'Keefe and Recce, 1993). Furthermore, place cells remain in place after distal cues are removed (O'Keefe and Conway, 1978) but at the same time, when cues are not removed but altered, place cells change their firing properties proportionately (Muller and Kubie, 1987; O'Keefe and Speakman, 1987). The loss of long-lasting synaptic changes, such as those measured by LTP, does not necessarily disrupt the formation of place fields, but does interfere with their stability (Nakazawa et al., 2002).
The relation between hippocampal LTP and hippocampus-dependent learning and memory functions is therefore not unequivocal. There exists a correlation between the time course of dentate LTP, the rate of forgetting a spatial task (Barnes, 1979) and the ability to learn new tasks (Barnes, 1988). Furthermore, blocking LTP in dentate perforant path synapses with 2-amino-5-phosphono-pentanoate (AP5) abolishes learning in the most widely used test of hippocampus-dependent spatial memory, the hidden platform task in the Morris water-maze (Morris et al., 1986). However, although spatial training may give LTP-like changes in dentate field potentials, the effect lasts less than 30 minutes (Moser et al., 1994). Moreover, attempts to interfere with spatial learning by saturation of LTP have produced contradicting results (Bliss and Richter-Levin, 1993; Korol et al., 1993; Jeffery and Morris, 1993; Sutherland et al., 1993; Cain et al., 1993; Moser et al., 1998). In addition, Bannerman et al. (1995) showed that the sensitivity of spatial learning to AP5 administration depends on task familiarisation and the environment. Finally, the relative importance of CA1 and dentate LTP for learning is not satisfactorily understood.
The introduction of genetic engineering techniques to manipulate genes that code for particular proteins has brought a new level of accuracy to the study of hippocampal plasticity. Early studies in this field developed mice with malfunctioning NMDA receptors, either globally (e.g. Silva et al., 1992) or restricted to specific brain areas (Tsien et al., 1996; McHugh et al., 1996; Wilson and Tonegawa, 1997; Huerta et al., 2000). In mice whose GluR-A subunit has been selectively deleted, AMPA subunits, which would normally be recruited to the synapse (Takahashi et al., 2003), are missing; only the lower-conductance AMPA receptors remain. These GluR-A knockout mice (GluR-A-/- mice) do not show normal hippocampal LTP (Zamanillo et al., 1999, but see Hoffman et al., 2002). Despite this, GluR-A-/- mice show normal acquisition of the water-maze hidden platform task. This result is surprising because it appears to break the link between normal hippocampal LTP and normal spatial memory performance, which had been intially inferred from studies of manipulation of the NMDA receptor. This problem formed the bulk of the work for the applicant's doctoral thesis.
A clear prediction of the LTP hypothesis is that animals that are deficient in hippocampal LTP should be impaired on hippocampus-dependent behavioural tasks. The observation that GluR-A-/- mice are capable of solving a spatial reference memory task in the water-maze whilst demonstrating a block of CA1 LTP (Zamanillo et al., 1999) is therefore at odds with the hippocampal LTP/spatial learning hypothesis. However, hippocampus dependent memory can be assessed in other ways than the water-maze. Discrete trial, rewarded alteration on the elevated T-maze is a working memory task, which can be used as a powerful alternative to the water-maze for measuring hippocampal dysfunction (Rawlins and Olton, 1982). While the spatial reference water maze task requires the animal to retain spatial information over several days, performance on the T-maze requires only temporary storage of spatial information. In this task, mice are forced left or right down one arm of the T-shaped maze where they receive a food reward. Then, they are immediately given a free choice of either arm of the maze, and are rewarded for choosing the previously unvisited arm (i.e. for alternating). We found that whereas wildtype mice perform well on the T-maze rewarded alternation task (over 80% correct over 40 trials); GluR-A-/- mice perform at chance (49% correct) (Reisel et al., 2002). Hippocampal lesions performed in the GluR-A-/- mice abolished their performance in the water-maze. It is not immediately clear why such a dissociation between adequate water-maze performance and impaired T-maze performance arises in these animals. One possibility is that the difference may reflect differential task sensitivity (the T-maze may be a more sensitive task for hippocampal dysfunction). Another possibility, far more sweeping, is that the GluR-A subunit may be a necessary component of a molecular mechanism that underlies working memory whilst, at the same time, not being required for some types of reference memory.
The differential reinforcement of low rates task (DRL) is a non-spatial test of working memory. In this task, subjects are required to space their responses: responses that meet a criterion are rewarded; early responses are not. Early reports suggested that mice perform poorly on this task (Richelle, 1972) but in fact they learn it well with a nose-poke response and liquid reward. In rats, damage restricted to the hippocampus alone is sufficient to impair performance on this task; the extent of impairment increases with hippocampal lesion size, but does not depend on intra-hippocampal lesion site (e.g. Sinden et al., 1986; Bannerman et al., 1999). Performance is also impaired by AP5 administration (Tonkiss et al., 1989). In a DRL schedule where one response in 15 seconds is rewarded (DRL-15), we found that GluR-A-/- mice were deficient compared to wildtype controls (Reisel et al., in preparation). On this task, the mice must make remember and integrate their responses from trial to trial, depending on information given mere seconds before each choice. This type of working memory may be a closer analogue of human episodic memory than is offered by the conventional hidden platform task in the water-maze, and seems to require shorter-term and more flexible use of spatial memory. Preliminarily, we can therefore conclude that there are two, separable forms of memory that both depend on the hippocampus, but may depend on heterogenous underlying mechanisms.
To monitor the plasticity of place fields, we will record extracellular action potentials from CA1 and dentate gyrus cell layers. We will track the animal using a pair of infrared diodes attached to its back, which will enable us to precisely identify its location and orientation. Using this technique, we will train mice on a variety of spatial learning tasks, as well as tasks that require non-spatial learning. Analysis of the spike patterns will be carried out by looking at three interrelated features: (1) the strength of the place fields, (2) the size of the place fields, and (3) the rate of change of a place field to accommodate a change in the immediate environment of the animal. To analyse place fields, the spatial environment will be divided into pixels. Mean pixel firing rates will be calculated by dividing the total number of spikes detected at a pixel location by the total occupancy time within that pixel. Place field peaks will be identified as locations with all surrounding pixels having lower values. Spatial tuning of individual cells will be assessed both by mean firing rate (defined as the average firing rate across all pixels with the mean pixel firing rate exceeding 10% of the cell's peak firing rate) and place field size (defined as the area of pixels whose firing rate exceeds 10% of the peak firing rate).
The majority of the work will be conducted in Bert Sakmann's laboratory at the Max Planck Institute for Medical Research in Heidelberg. Recently, his laboratory has constructed a state-of-the-art in vivo multi-tetrode recording suite and, in collaboration with Nick Rawlins and David Bannerman at Oxford, several mazes (T-maze, radial maze) have been produced. A recent pilot study has demonstrated adequate performance in tethered wildtype animals on the T-maze and that they learn as fast as untethered controls (Reisel and Hahn). The pilot study also successfully obtained spike signals from the hippocampus. Several modifications and, in particular, miniturisation of the implanted drive, are needed before experiments can begin in earnest.
Genetic mutants of interest. The project timetable (see Appendix, p. 10) contains a systematic set of experiments to investigate the essential molecular changes responsible for the plasticity of place cells in a set of mice with modified AMPA and NMDA subunits. We will use mutant lines engineered in the laboratory of Rolf Sprengel at the Max Planck Institute for Medical Research in Heidelberg. The GluR-A-/- mouse, due to its selective impairment in working memory, is a well-suited mutant with which to attack these issues. This mutation was generated by homologous recombination in embryonic stem cells of SV129 mice and then transferred by backcrossing into C57BL6 mice because they more readily learn to perform spatial tasks. Briefly, the mutant is derived from a targeted mouse constructed by using the GluR-A targeting vector pFC II containing 10.7 kb of the mouse GluR-A gene (GRIA I) encoding parts of intron 10, intron 11 and exon 10. Exon 10 is flanked by two loxP elements (GluR-A2lox) and can be removed by Cre recombinase producing a non-functional GluR-A gene lacking exon 10. By making this exon10 deletion in the germline of mice the GluR-A mutant was generated.
The same laboratory has also made transgenic mice harbouring a bidirectional operon for a GFP-tagged GluR-A subunit and for the indicator gene lacZ. Both genes are under the common control of minimal promoters with an intervening set of seven binding sites for the tetraycyline repressor (tet operators). The promoters are thus linked to the binding of the tetracycline repressor-VP16 transactivation domain fusion (tTA) at the tet operators, and such activation can be suppressed in the presence of tetracycline and certain derivatives such as doxycycline (Gossen and Bujard 1992). These transgenic mice will be crossed with transgenic mice expressing tTA in specific regions of the brain, such as the line described by Mayford et al. (1995) where tTA is under the transcriptional control of promoter for the alpha subunit of CaMKII. In double-transgenic mutants, the GFP-tagged GluR-A subunit, and b-Galactosidase, will be expressed in a doxycycline-regulated manner in forebrain principal excitatory neurons. When this double transgenic mutant is combined with the GluR-A knockout genotype, we have a mouse model in which we can regulate CA3/CA1 LTP (Mack et al., 2001). Moreover, we will monitor methods the incorporation of GFP-tagged AMPA receptors into hippocampal CA1 pyramidal cell synapses by fluorescence imaging. In addition to mice that have GFP-tagged GluR-A fusions, additional mouse lines are available to the project that contain instead of the GFP-tagged GluR-A, GluR-A mutants without C-terminal Phosphorylation sites (SA), without PDZ interaction domain (TG*) and with GFP-tagged GluR-B and GFP-tagged GluR-C. These mutants will enable us to investigate whether the wildtype GluR-A subunit can be functionally replaced by modified or altogether different AMPA receptor subunits.
Two NMDA mutants will be available to the project. Since the progenitor mice contain exon 10 flanked by two lox2 elements, exon 10 removal can be limited to those neurons expressing CRE, e.g. CA1 pyramidal cells. The successful application of this approach has recently been demonstrated by a restricted deletion of the NMDA receptor subunit NR1 in granular cells of dentate gyrus (NR1-DG mutant). CA1 LTP in these animals is normal, whereas LTP in the dentate gyrus is abolished. In order to assess the role of the NR1 subunit in the construction and maintenance of place fields, we intend to examine in vivo the effect of the deletions of NR1 specific to the dentate gyrus. A recent pilot study has demonstrated that NR1-DG mutants are incapable of solving the rewarded alternation task on the elevated T-maze (KO: 52% vs. WT: 83% choice accuracy after 40 training trials, n = 6). This raises the possibility that normal glutamate function in the dentate gyrus may be necessary for successful completion of tasks that require intact working memory.
In an additional line of NMDA mutants, the NR2A subunit has been over-expressed, which causes a reduced magnesium block and increases the duration of calcium influx through NMDA ion channels. Extracellular slice recordings performed by Øivind Hvalby and Vidar Jensen in Oslo demonstrate that NR2A M2N/S mutants have increased CA1 LTP (in a four burst paradigm). In a pilot study of massed trials rewarded alternation on the T-maze, we found that the mutants performed better than littermate controls. This raises the fascinating possibility that a specific molecular alteration of the NMDA subunit may improve some types of memory. We intend to test the rate of change in place cell modification in these mutants, and we will attempt to correlate place cell plasticity with potential enhanced behavioural performance. At the same time, we will examine whether, unless coupled with an enhanced capability for de-potentiation, this mutant will show poorer performance on the T-maze rewarded alternation task (which requires flexible working memory). With the multi-tetrode recording technique, we have a unique opportunity to investigate such correlations in vivo.
Behavioural tasks. The Rawlins laboratory has recently developed a reference/working memory version of the radial maze, which constitutes an extension of the working memory version of the radial maze first developed by David Olton and colleagues (Olton and Samuelson, 1976). In this task, only half the arms are baited; after the reward is taken from the end of an arm, re-entry into that arm constitutes a working memory error. Entry into arms that are never baited is considered a reference memory error. This procedure allows a within-subject, within-trial assessment of both spatial working memory and reference memory. GluR-A-/- mice are essentially unimpaired in terms of reference memory performance, however, they made considerably more working memory errors than controls (Schmitt et al., 2002). This study provides a simultaneous demonstration of impaired spatial working memory but spared spatial reference memory in GluR-A-/- mice, using identical spatial cues, and with the same sensorimotor and motivational demands. We plan to use the tetrode recording technique to monitor place cells during this task, and especially the analysis of different firing patterns involved as the animal employs putatively different types of memory. The comparison of different mutant lines would provide a window on the mechanisms that subserve rapid-onset, flexible spatial working memory.
In a second behavioural assey, we plan to examine the effect of partial cue removal on the strength of place fields in the various mutants. To accomplish this, we intend to build a modified radial maze with moveable cues. In this paradigm, the integrity of the distal cues is reduced to assess the degree to which the animal is capable of pattern completion at the behavioural level. Under partial cue removal, NR1-CA3 mutants are severely impaired (Nakazawa, 2002). Moreover, there is evidence that place cells generally follow distal cues (Miller and Best, 1980; Shapiro et al., 1989). Thus, rotating the maze allows a separation between local and distal cues. In such a cue-driven environment, the place fields are rotated when the distal cues are rotated. This experiment seeks to establish time-variant alterations of place cell firing fields following reduced or altered environmental exposure and address the question of which mechanisms are involved in experience-dependent modification of place cell firing patterns.
A third behavioural task is intended to investigate the effect of cue alteration on the stability of place fields. To test this we plane to build a modified version of the spatial divergence task developed by John O'Keefe's group (Lever et al., 2002). In this task, a rat is exposed to two differently shaped environments. In rats repeatedly exposed to either a square or a circular space, the CA1 place cell representations of both environments gradually diverge. The divergence is specific to environmental shape and persists for long periods of time. We intend to test the mutants on a version of this task adapted for mice. It is of particular interest to establish whether GluR-A-/- mice, which are deficient in spatial working memory, also have impairments in this type of place cell plasticity and whether the adjustment of the spatial map occurs at a different rate in the mutants than in control animals.
Finally, a fourth paradigm is designed to probe the role of experience-driven synaptic modification during non-spatial learning. To this end, we will employ the olfactory discrimination task developed by Howard Eichenbaum and colleagues (Wood et al., 1999). In this study, rats were trained to dig for food in small cups of scented sand. Neurons were identified which responded to particular odours, irrespective of their spatial location. Intriguingly, the study found neurons that fired when the present cup smelled like the previous cup. The same cells did not fire when the scents did not match. These results raise the possibility that hippocampal neurons are not only tuned to spatial locations but also selectively respond to non-spatial events in their environment. We intend to train the GluR-A and NR1 mutants on a modified version of this task, which would allow us to experimentally compare hippocampal functioning during non-spatial tasks in mice with specific molecular deletions.
Electrophysiological analysis. In collaboration with Morten Raastad and Øivind Hvalby at the Institute of Physiology, University of Oslo, the applicant will apply a range of complementary physiological tests on the various strains of transgenic mice described above. The Raastad/Hvalby laboratory has extensive experience in cellular recordings (sharp and patch electrodes) from hippocampal neurons, intracellular injections (enzyme fractions, blockers, drugs, fluorescent markers), spine detection, counting and reconstruction. The mechanical stability, direct visual access to brain structures and control of the fluid environment are advantages in slice preparations when compared to experiments with freely moving animals. Both extracellular and intracellular experiments can be performed in a more stable and long-lasting way and pharmacological manipulations are more easily done.
Specifically, the work will include (1) simultanous testing for the presence of LTP in the CA1 region and the dentate gyrus in the mutants, (2) dendritic recordings, and (3) slice calcium imaging. Potentiation will be made by tetanic stimulation, theta burst stimulation or a pairing protocol. The ability of these protocols to distinguish between LTP properties in wild type and mutant mice will be determined. The efficacy of induction will be ensured by experiments with blockers of GABAA-mediated inhibition and by recording the voltage integral of the field and intracellular EPSPs during the tetanus. We will seek to identify specific molecular alterations of AMPA and NMDA receptor molecules and assess their influence on the degree and synaptic location of LTP expression. These experiments will seek to correlate hippocampal functional properties to the in vivo data obtained in the tetrode recordings. The electrophysiological experiments in Oslo are estimated to last for the duration of the third year of the project.
Expected Achievements. At the end of the project period, we expect to have definite answers to some of the key questions of place cell plasticity. Most significantly, we expect to have a systematic profile of the necessary or sufficient molecular mechanisms of place cell formation, maintenance and modifiability. Moreover, we expect that the project will elucidate the respective roles of CA1 and the dentate gyrus in place cell formation. Finally, we expect that our results with in vivo tetrode recordings developed specifically for genetically engineered mice should prove of value for further studies in the area, including the rational development of methods that could benefit industrial pharmacological research.
FOR MEDICINE AND HEALTH IN SOCIETY
POSITION AND INTERNATIONAL COLLABORATORS