A grant proposal to answer the following question:

Do cells within the entorhinal cortex underlie spatial navigation, with respect to a specific set of environmental cues from place cells?

Motivation

The title question is really question about how animals internally represent their locations within an environment which requires use of cumulatively-learned properties of space, in general, and unique elements of the environment that are learned upon initial exploration.The hippocampus contains neurons that represent unique locations in space and the entorhinal cortex (EC) con- tains neurons that are likely to provide an internal coordinate system to be used for naviga- tional planning (Nadasdy et al., 2017). These cells lie within the hippocampal formation which includes the dentate gyrus, cornu ammonis (CA) subfields, subiculum, presubiculum, para- subiculum and the EC (Bush, Barry, & Burgess, 2014). Additional spatially responsive cells have been discovered within the hippocampal formation. These are head-direction and bound- ary vector/border cells (Woolley, 2017). Anatomical connectivity of cells within these regions apply limits to how the computational problem of cognitive mapping is carried out neuronally. The classification of neurons based on response patterns and stereotypical projection patterns makes the hippocampal formation a stable system to study.

The mechanism of low-level, sensory representation the outside world is rather simple and, thus, relatively well understood. Sections of cortical tissue are organized retino-, somato- and tonotopically to reflect spatial organization of sensory receptors (Moser et al., 2014). Inter- nal representations of external space form at a higher level and require more intrinsic neural computation. Evolutionarily, the ability to navigate in space is a widely shared cognitive trait of many animals. The computation of self-location takes slightly lower-level inputs and transforms it into a more complex representation. Theoretical and experimental researchers, alike, believe that cognitive mapping introduced a new, adaptive type of information processing in the brain. Thus, the study of how network computations taking place in the hippocampal formation give rise to spatial representations may provide insight into higher-lever, cortical processing more generally (Moser et al., 2014).

Clinically, dysfunction in the hippocampal formation causes memory problems and disori- entation, which can be found in early stages of Alzheimers. In Alzheimer’s, hippocampus and EC neurons are the earliest targets of plaques and tangles and cell death. This causes spatial disorientation and spatial memory loss (Nadasdy et al., 2017). Unearthing the mechanism of spatial representation in the hippocampal formation could guide development of targeted intervention for this population.

Background

Place cells

Originally discovered in rat hippocampi, place cells are complex spiking pyramidal cells in the CA3 and CA1 that are activated by an animal’s perceived location within an environment (O’Keefe & Dostrovsky, 1971). Each cell is tuned to a single place field within a space, though will display several tuned locations if the space is large. Place fields are rapidly instantiated in novel environments and are stable across many visits to the same environment (Woolley, 2017). Representation of a space is relatively invariant to the removal of a subset of sensory cues (O’Keefe, 1979), suggesting place fields are not driven by unique combinations sensory input. Yet they are not independent to sensory cues. First, place fields will move with salient cues or landmarks that are visual in nature (Woolley, 2017). Second, boundaries in an envi- ronment control place cell firing, as developed in the boundary vector cell (BVC) model. Place fields tend to maintain a fixed distance to one or more boundaries following changes to the ge- ometry of a familiar environment (O’Keefe & Burgess, 1996), reflecting input from cells tuned to a preferred distance and allocentric direction (boundary cells) (Bush et al., 2014). Addi- tionally, the introduction of a new boundary in a familiar environment will produce secondary firing fields in same position relative to the new boundary as the initial firing field had to familiar environmental boundary (Barry et al., 2006).

Two types of place field re-mapping support the idea that place cell activation is relatively environment-dependent. First, enhancement or suppression in the firing rate of a place cell, occurs with minor changes to the environment, such as a new wall color. This is called rate re-mapping (McNaughton, Battaglia, Jensen, Moser, & Moser, 2006). Second, a place cell can completely relocate its place field when an animal is put in an entirely new environment (Woolley, 2017). This new place field rapidly stabilizes (Barry et al., 2006). This phenomenon is called global re-mapping. Re-mapping subserves episodic memory of particular environments my minimizing similarities between internal representations of unique environments.

Grid cells

Given a constant environment, place cell firing alone has the capacity to localize a body within that space. However, knowledge of the firing relationships between place cells cannot predict the relative or absolute position of an animal in a different environment (McNaughton et al., 2006). Thus, investigation of areas with projections to place cells in the hippocampus revealed another cell types that may support environment-independent cognitive maps. Cells within the medial EC (mEC) called grid cells are spatially-selective cells that respond to discrete lo- cations, at the vertices of a triangular lattice overlaid onto any environment (Hafting, Fyhn, Molden, Moser, & Moser, 2005). Grid cells differ in spacing (distance between adjacent fields), grid orientation (angle of grid relative to some fixed axis) and grid phase (displacement of fields with respect to fixed boundary). Robust grid spacing invariance is a key feature of grid cells and means that, over differently shaped and sized environments, the distance between grid fields of a particular cell remains unchanged (Hafting et al., 2005). Grid cell’s orientation is anchored to environmental cues, such that rotation of a cue in a round room induces a corresponding rotation in grid orientation (Nadasdy et al., 2017). Grid cell activation is driven by an animal’s movement within an environment, suggesting their role in path integration. Collective informa- tion from a suite of grid cells has been modeled to represent/define allocentric self-location, demonstrating the cells capacity to underlie spatial navigation (Moser et al., 2014).

Beyond sensory information, information about changes in position and direction in space can be computed based off of signals encoding the animal’s movement. Grid cells along with other cells in the mEC are positively modulated by running speed (McNaughton et al., 2006). Grid cells have been found to integrate elapsed time and distance run while an animal is running in place (Kraus et al., 2015). Models have utilized a set of universal grid-cell type inputs to conduct look-ahead approaches to goal-oriented navigation (Kubie & Fenton, 2012; Erdem & Hasselmo, 2012; Barry & Bush, 2012). A simple explanation for how this might work is as follows: when a rat is traveling along a single dimension, movement from one location of peak activity to another represents the displacement of an integer multiple of the grid spacing of that cell. With respect to egocentric aspects of the environment, a full suit of grid cells can extract the translational vector of traversal to reach a particular goal state (Barry & Bush, 2012). These models do not operate on grid cell inputs, alone, though. They require additional information from inputs representing other cell types within the hippocampal formation, as navigation is dependent on specific location cues. This would be consistent with the inability of animals with EC lesion to return to their ”home base” on the basis of self-motion alone (Moser et al., 2014).

Other entorhinal cortex cells

Until now, classification of cell types into grid, head-direction and boundary cells left a large population of cells unclassified (Kraus et al., 2015). This has implications in hypothetical mod- els of path integration and navigation behavior that only use information provided by specialized spatially-tuned neurons, without introducing potential interacting variables that further inform the cognitive map encoding. Recent, unbiased approaches to probing the navigational sys- tem in the hippocampal formation have revealed novel coding properties within the mEC. In a general analysis of hundreds of neurons in mice during a foraging task, this group looked for cells that were informative about any subset of navigational variables (position, speed, and head-direction or intrinsic modulators of firing). This approach classified the majority of the test neurons, revealing mixed selectivity and heterogeneity within these cells. Additionally, tuning properties of specific cells were dependent on the speed of the animal at any given moment (Hardcastle, Maheswaranathan, Ganguli, & Giocomo, 2017). These findings explain the dif- ficulty in classifying cells in the entorhinal cortex and seemingly incongruent results among studies of these populations. The speed-dependency of mEC coding is adaptive because, if this structure does in fact support path integration and vector navigation, faster speeds will accumulate greater error with the same initial miscalculations.

Network interactions

In the recent past, theories have been developed on how grid cell firing patterns generate the unitary firing fields of place cells. However, this view has recently been challenged by evidence that place fields in hippocampal place cells are relatively unaffected by absence of reliable grid cell activity (Moser et al., 2014). Also, stable place fields are observed earlier than stable grid firing in pre-weanling rats (Langston et al., 2010). Both of these reached matura- tion after head-direction cells which had adult-like response patterns at birth (Langston et al., 2010). This does not fit with the original, hierarchical model of cognitive mapping and, instead, suggests complementary roles of grid and place cells. Additionally, the full suite of neuronal types within the hippocampal formation has only just been identified and is not completely understood (Hardcastle et al., 2017). Thus, current models of cognitive mapping are largely uninformed when attempting to demonstrate or predict a navigational behavior.

One alternative theory presented by Bush, Barry and Burgess of University College London posits that place cells are determined by environmental sensory inputs while grid cells incorpo- rate or provide information about bodily motion. In this theory, the grid and place cells networks provide complementary spatial representations where place cells encode defining cues at spe- cific locations (as in episodic memory) while grid cells constitute context-independent spatial encoding that supports path integration and vector navigation (Bush et al., 2014). Here, place cells convey sensory cues of a unique environment to inform the estimate of the animal’s allo- centric location and the navigation within it (Barry & Bush, 2012)

Approach and possible outcomes

Here, I would like test the isolated functions of grid and place cell in the cognitive mapping of an environment. More specifically, I am interested in seeing how alterations in place cell firing patterns alter grid cell firing patterns during a navigational planning task.

The experimental paradigm I propose includes electrophysiological recordings of a wide array of neurons in the hippocampal formation, focusing on grid and place cells, but not ignoring the other mEC neurons that display mixed selectivity. These recordings will be made in a way that is similar to the protocol in the Hardcastle 2017 paper (Hardcastle et al., 2017). In this protocol, a chronic recording device will be surgically implanted. These will record activity during all conditions of the behavioral task.

The rats will be trained, using a food reward, to traverse two nearly-identical learned paths in an open environment. The trajectory two is simply a little bit longer, but on the same path, as trajectory one. A single environmental addition towards the end of the path indicates that the rat should continue to past the end of trajectory one along the same dimension and continue a corner to complete trajectory two. Levers will be at both trajectory endpoints and will only be rewarded when the rat pushes the lever of the ”correct” trajectory given environmental cues. Trajectory number one will serve as the control task. The second trajectory will be the experimental condition where the introduction of an environmental cue should produce a different navigational plan (and will be termed the minor change condition for the rest of this paper). A third condition will have a trajectory identical to trajectory two but, at the same point as the cue in the minor change condition, the animal enters a different environment from the first two conditions (and will be termed the major change condition for the rest of this paper). Multiple trials of the different conditions will be run and trials in which the rats do not traverse to the goal position correctly will be eliminated. There will also be a baseline, exploratory phase that will serve as a reference for neuronal activity along both trajectories while the rat is not participating in the goal-directed task.

If the task were purely exploratory, rather than goal-directed, grid cell activity in the envi- ronment would be essentially the same when the minor environmental cue is present as when it is absent. The minor cue will be chosen to be sufficient to induce rate re-mapping in relevant place cells, but not extreme enough to induce global re-mapping. The last experimental con- dition within a different environment would be expected to induce global re-mapping in place cells. Additionally, in the major change condition, the new environment would be created to maintain grid cell firing patterns by exploiting grid cell’s anchoring to environmental cues. This could be done by introducing a landmark external to the environment that is common to all conditions.

The resulting data set for these tasks will represent hippocampal formation encoding dur- ing exploration, spatial navigation, spatial navigation under place cell rate-remapping and spa- tial navigation under global remapping. Additionally, the velocity of the rat will be constantly recorded such that firing patterns may be normalized with respect to speed.

To confirm that rate and global remapping are still present in these goal-directed navigation tasks, recordings from place cells will be analyzed and compared across conditions. It is expected that place cell tuning properties are independent of the type of movement in a space (although temporal analysis may display some degree of vicarious trial and error (Woolley, 2017)).

The first step in the analysis process would be to identify neurons that have the stereotypi- cal firing patterns associated with current classifications of spatially selective cells: place, grid and boundary cells. A first pass analysis will probe the effect of environmental cues on individ- ual grid cell tuning patterns. Here, we will test whether or not individual grid cell firing patterns (including spacing, orientation and phase) in the minor and major change condition are signifi- cantly different from the control condition, after potentially controlling for velocity effects. A null result in this analysis does not necessarily mean that environmental cues/place cells do not play a role in guiding spatial navigation, it only means that the properties of spacing, orientation and phase have not been altered in these different contexts. It is unlikely that these contextual changes will alter the intrinsic properties of individual place cells given previously presented evidence of grid cell constancy across environments, especially when the environments are designed to exploit that constancy as they are here.

The effect of place cell remapping may also take effect on grid cell activity in a temporal way. I will analyze the population responses of grid cells in the time period proximal to en- countering the minor or major change. If there are differences, it suggests that either or both environmental changes that signal the rat to alter its navigation plan. Two kinds of differences may be observed. First, which grid fire cells fire at that point in time may change (i.e. novel firing from neurons that were silent in the control condition or silencing of neurons active in the control condition). These kinds of changes could be indicative of a trial-and-error type mech- anism of grid cells in navigational planning that is influenced by environmental cue. Second, rates of involved grid cells may be altered. If firing rate is enhanced in grid cells that have grid fields along the dimension of traversal and/or firing rate is inhibited in grid cells that do not have grid field axes along the intended trajectory, that indicates that grid cell firing directs or predicts decisions made in spatial navigation, favoring activation in the path of choice. If the resulting change is some intermediate between these two, then grid cells may be conducting other computational processes beyond extracting optimal translation vectors to act upon.

Next, I would conduct an un-targeted analysis of differences between conditions using all neurons that were recorded in the mEC. Comparing response profiles of individuals neurons across conditions may reveal mixed selectivity cells that are dependent upon, and thus may receive input from the place cells encoding environmental cues. Comparing population re- sponses at time of encounter of the minor and major changes in environment may also reveal neurons that underpin goal-directed navigation. This phase of analysis would be exploratory in nature and is inspired by inability to classify many mEC neurons into predefined categories, thus limiting analysis to only classifiable neurons limits the information we have about the hip- pocampal formation. In a more sophisticated analysis, response profiles across the multiple mEC neurons could be used to model estimates of spiking rates given a set of predefined vari- ables. Here, the variables would be position, short term trajectory, long term trajectory, speed, place cell activation pattern and, if measured, head direction and egocentric boundary distance. The un-targeted approach may reveal associations and tuning properties that targeted approaches would miss.

Conclusion

With the experimental setup that I propose here, I beg the question of whether changes in place cell activity influences grid cell activity or activity of cells throughout the mEC. To test the hypothesis that place cells influence grid cell activity in spatial navigation tasks, this experiment introduces new environmental cues in spatially equivalent routes. We hope to observe the way that environmental information may or may not be utilized hippocampal formation during a navigational task. The significance of this question lies in a few domains and this experiment serves to inform two directly.

First, results would inform theories on the neuronal substrates of cognitive mapping. Un- changed grid cell activity co-occuring with changed place cell activity (as in the minor and major change conditions), will provide evidence against the model that unitary place fields are built from converging grid cell input. In contexts designed to maintain grid cell firing across environments, co-occuring changes would suggest that the way general spatial and unique environmental cues is encoded in the interaction between grid and space cells. Knowledge about the mechanism of cognitive mapping may also inspire targeted intervention techniques in populations that have deficits in spatial awareness, such as individuals with Alzheimer’s disease.

Second, results address a more general line of inquiry about intrinsic neuronal computa- tion that begets cognitive ”senses”, or information processing tasks integrating many sensory cues. The latter, un-targetted analysis I propose would be more informative in this respect. As demonstrated in the Hardcastle 2017 report, individual neurons are not necessarily ”tuned” to characteristic qualities of a single parameter (Hardcastle et al., 2017). Instead, a large por- tion of neurons carry out multiplexed encoding that, together, represent space generally and episodically.

The ability to map general and unique environments is a highly adaptive function that relies not only on sensory cues, but also intrinsic computation. It is believe that this function is localized to the hippocampal formation. It is still unclear, though, how this representation is computed. Until recently, it was theorized that spatial representation was instantiated in a hierarchical way. Here, we test that assumption and explore the possibility of intrinsic network dynamics that incorporate sensory cues to guide goal-directed navigation.

Citations

Barry, C., & Bush, D. (2012). From A to Z: a potential role for grid cells in spatial navigation. Neural Syst. Circuits, 2(1), 6. doi: 10.1186/2042-1001-2-6

Barry, C., Lever, C., Hayman, R., Hartley, T., Burton, S., O’Keefe, J., . . . Burgess, . (2006). The Boundary Vector Cell Model of Place Cell Firing and Spatial Memory. Rev. Neurosci., 17(1-2), 71–97. doi: 10.1515/REVNEURO.2006.17.1-2.71

Bush, D., Barry, C., & Burgess, N. (2014). What do grid cells contribute to place cell firing? Trends Neurosci., 37 (3), 136–145. Retrieved from http://dx.doi.org/10.1016/j.tins.2013.12.003 doi: 10.1016/j.tins.2013.12.003

Erdem, U. M., & Hasselmo, M. (2012). A goal-directed spatial navigation model using for- ward trajectory planning based on grid cells. Eur. J. Neurosci., 35(6), 916–931. doi: 10.1111/j.1460-9568.2012.08015.x

Hafting, T., Fyhn, M., Molden, S., Moser, M.-B., & Moser, E. I. (2005). Microstructure of a spatial map in the entorhinal cortex. Nature, 436(7052), 801–806. Retrieved from http://www.nature.com/doifinder/10.1038/nature03721 doi: 10.1038/na- ture03721

Hardcastle, K., Maheswaranathan, N., Ganguli, S., & Giocomo, L. M. (2017). A Multiplexed, Heterogeneous, and Adaptive Code for Navi- gation in Medial Entorhinal Cortex. Neuron, 94(2), 375–387.e7. Re- trieved from http://dx.doi.org/10.1016/j.neuron.2017.03.025 doi: 10.1016/j.neuron.2017.03.025

Kraus, B. J., Brandon, M. P., Robinson, R. J., Connerney, M. A., Hasselmo, M. E., & Eichenbaum, H. (2015). During Running in Place, Grid Cells In- tegrate Elapsed Time and Distance Run. Neuron, 88(3), 578–589. Re- trieved from http://dx.doi.org/10.1016/j.neuron.2015.09.031 doi: 10.1016/j.neuron.2015.09.031

Kubie, J. L., & Fenton, A. A. (2012). Linear Look-Ahead in Conjunctive Cells: An Entorhinal

Mechanism for Vector-Based Navigation. Front. Neural Circuits, 6(April), 1–15. Retrieved from http://journal.frontiersin.org/article/10.3389/fncir.2012.00020/abstract doi: 10.3389/fncir.2012.00020

Langston, R. F., Ainge, J. A., Couey, J. J., Canto, C. B., Bjerknes, T. L., Witter, M. P., . . . Moser, M.-B. (2010). Development of the Spatial Representation System in the Rat. Science (80-. )., 328. doi: 10.1126/science.1172133

McNaughton, B. L., Battaglia, F. P., Jensen, O., Moser, E. I., & Moser, M.-B. (2006). Path integration and the neural basis of the ’cognitive map’. Nat. Rev. Neurosci., 7(8), 663–678. Retrieved from http://www.nature.com/doifinder/10.1038/nrn1932 doi: 10.1038/nrn1932

Moser, E. I., Roudi, Y., Witter, M. P., Kentros, C., Bonhoeffer, T., & Moser, M.-B. (2014). Grid cells and cortical representation. Nat. Rev. Neurosci., 15(7), 466–481. Retrieved from http://www.nature.com/doifinder/10.1038/nrn3766 doi: 10.1038/nrn3766

Nadasdy, Z., Nguyen, T. P., To ̈ro ̈k, A ́., Shen, J. Y., Briggs, D. E., Modur, P. N., & Buchanan, R. J. (2017). Context-dependent spatially periodic activity in the human entorhinal cortex. Proc. Natl. Acad. Sci., 114(17), E3516–E3525. Re- trieved from http://www.pnas.org/lookup/doi/10.1073/pnas.1701352114 doi: 10.1073/pnas.1701352114

O’Keefe, J. (1979). A review of the hippocampal place cells. Progress in Neurobiology, 13(4), 419 - 439. Retrieved from http://www.sciencedirect.com/science/article/pii/0301008279900054 doi: https://doi.org/10.1016/0301-0082(79)90005-4

O’Keefe, J., & Burgess, N. (1996). Geometric determinants of the place fields of hippocampal neurons. Nature Publishing Group.

O’Keefe, J., & Dostrovsky, J. (1971). The hippocampus as a spatial map. preliminary evidence from unit activity in the freely-moving rat. Brain Research, 34(1), 171 - 175. Retrieved fromhttp://www.sciencedirect.com/science/article/pii/0006899371903581 doi: https://doi.org/10.1016/0006-8993(71)90358-1

Woolley, C. S. (2017, May). Spatial orientation. IL.