Leveraging place field repetition to understand positional versus nonpositional inputs to hippocampal field CA1

  1. William Hockeimer
  2. Ruo-Yah Lai
  3. Maanasa Natrajan
  4. William Snider
  5. James J. Knierim

  • Curated by eLife

    eLife logo

    eLife Assessment

    This is a valuable work that convincingly reveals that place cells in the hippocampus that exhibit repeated firing fields incorporate information about non-positional variables in each firing field. They reveal that individual firing fields of a single place cell can exhibit tuning to different head orientations, suggesting hippocampal neurons are flexible in terms of how they incorporate non-positional inputs.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

The hippocampus is believed to encode episodic memory by binding information about the content of experience within a spatiotemporal framework encoding the location and temporal context of that experience. Previous work implies a distinction between positional inputs to the hippocampus that provide information about an animal’s location and nonpositional inputs which provide information about the content of experience, both sensory and navigational. Here we leverage the phenomenon of “place field repetition” to better understand the functional dissociation between positional and nonpositional inputs to CA1. Rats navigated freely on a novel maze consisting of linear segments arranged in a rectilinear, city-block configuration, which combined elements of open-field foraging and linear-track tasks. Unlike typical results in open-field foraging, place fields were directionally tuned on the maze, even though the animal’s behavior was not constrained to 1-D trajectories. Repeating fields from the same cell tended to have the same directional preference when the fields were aligned along a linear corridor of the maze, but they showed uncorrelated directional preferences when they were unaligned across different corridors. Lastly, individual fields displayed complex time dynamics which resulted in the population activity changing gradually over the course of minutes. These temporal dynamics were evident across repeating fields of the same cell. These results demonstrate that the positional inputs that drive a cell to fire in similar locations across the maze can be behaviorally and temporally dissociated from the nonpositional inputs that alter the firing rates of the cell within its place fields, thereby increasing the flexibility of the system to encode episodic variables within a spatiotemporal framework provided by place cells.

Article activity feed

  1. eLife

    eLife Assessment

    This is a valuable work that convincingly reveals that place cells in the hippocampus that exhibit repeated firing fields incorporate information about non-positional variables in each firing field. They reveal that individual firing fields of a single place cell can exhibit tuning to different head orientations, suggesting hippocampal neurons are flexible in terms of how they incorporate non-positional inputs.

  2. eLife

    Reviewer #1 (Public Review):

    The authors investigate whether during free exploration of an environment with an internal structure of corridors and occasionally fluid-rewarded alleys, rat CA1 place cells generate multiple firing fields in repeating patterns, allowing the investigators to analyze whether firing field positional properties like alley orientation, and non-positional properties like heading, field-rate modulation and other properties are similar or different within and across single place cell place fields. They adopt a standard cognitive map analysis framework, conceiving each cell as an individual map element and characterizing each cell's individual activity independently of the activity of other cells, such that the main unit of analysis is a place field averaged across recording times of many minutes. Despite framing the work as an investigation of a fundamentally-subjective episodic memory system sensitive to hidden cognitive and attentional variables, the experiment and analyses are conceived as if the cells respond to positional and non-positional features of experience as static "inputs" that the investigators infer. These "inputs" are conceptualized as effectively stationary and steady, and they are not manipulated. The authors find that there are many "repeated" firing fields, that they tend to have similar orientation more than expected by chance, and that each field's rate is modulated distinctly by heading direction and other factors, leading them to conclude that each field's nonpositional inputs are "individually addressable." The authors do not consider alternative possibilities for which there are strong indications in the contemporary literature like 1) CA1 activity could be internally generated; 2) that there could be hidden cognitive variables that influence CA1 activity episodically and in non-stationary ways rather than consistently; 3) that CA1 cells exhibit mixed tuning to a variety of environmental and navigational variables; 4) that CA1 activity is better interpreted from the point-of-view of a neural ensemble or a neural manifold of conjoint neural activity that represents multiple information variables, or 5) that stable neural representations of information need not depend on stable stimulus-response properties of individual cells. In fact, the analyses provide evidence consistent with each of these alternatives, but they are not considered. There is a case to be made that the authors are allowed to ignore these alternatives because they properly engage the dogmatic point of view, in which case there is little to adjust in the manuscript, which is both well-conceived and well-executed in the classic (but not contemporary) norms of place cell investigations.

    My comments are focused on improving the manuscript without insisting that the authors adopt alternative (contemporary) points of view, but requiring them to clarify their point of view and explain that there are alternatives.

    (1) The authors define what they mean by "positional" and "non-positional" "inputs" later in the manuscript. Since the experimental apparatus and task have been designed to isolate these "inputs" the authors should in the initial description of the environment and task explain what the task does and does not allow them to analyze. Instead, they have repeatedly asserted that the environment is a hybrid of an open-field and a linear track environment. This may be the case, but so what? The authors need to better explain, up front, why that matters and what they will be able to investigate as a result. As written, this all seems to me rather vague and post hoc.

    (2) The abstract states "Previous work implies a distinction between positional inputs to the hippocampus that provide information about an animal's location and non-positional inputs which provide information about the content of experience." While I understand what the authors mean, I want to point out that it is not straightforward to identify the "positional inputs" and the "non-positional inputs." What are they, how can they be measured? Is it not also possible that hippocampus generates "positional" information rather than receiving it, that is in fact the longstanding view of the cognitive map framework that the authors have adopted, and yet they frame the essential issue as one of differential receipt of positional and non-positional inputs. This seems to me imprecise and hard to defend but demonstrates the authors' opinion in framing this work. In my view a more objective and accurate statement might be "Previous work implies a distinction between hippocampal (positional) activity representing information about an animal's location and (non-positional) activity which represents information about the content of experience." This opinion about "inputs" is found throughout the manuscript over 50 times, starting with the title. While in my view this is not an objective treatment of the experimental design or data (positional and non-positional inputs are never identified or manipulated, they are merely inferred), I accept that the authors can say whatever they want so long as they make it clear to the reader that theirs is an opinion or assumption rather than a measurement. The manuscript is written as if the different inputs are identified and valid, rather than inferred.

    (3) The abstract states "even though the animal's behavior was not constrained to 1-D trajectories" whereas page 13 states "but their trajectories were constrained to orthogonal directions by the city-maze architecture" and page 23 states "but their trajectories were constrained to a rectilinear grid." While I understand what the authors mean, the first statement appears to contradict the others. There are additional examples that I do not identify here. In any case, I would like to have seen examples of the animals' trajectories through the maze. A figure showing the raw trajectories and another after the unwanted behaviors have been filtered out should be given, allowing the reader to understand how much the animals tended to travel through the alleys, how much they turned and lingered within them, etc.

    (4) The abstract ends with "These results demonstrate that the positional inputs that drive a cell to fire in similar locations across the maze can be behaviorally and temporally dissociated from the nonpositional inputs that alter the firing rates of the cell within its place fields, thereby increasing the flexibility of the system to encode episodic variables within a spatiotemporal framework provided by place cells." I don't see the evidence for the "thereby ..." claim. The authors are free to speculate and discuss but they should say they are speculating and/or discussing a possibility, rather than assert as if they have demonstrated a fact.

    (5) The Introduction begins with "All behavior is embedded within a spatial and temporal framework." By this statement, I believe the authors mean to assert, or at least they cause a reader to understand that there is a spatial and temporal framework that is separate from the behaving subject. They will use this point of view to design their experiment around the utility of a city- maze. Since the authors appeal to cognitive map theory so much, I point out that O'Keefe and Nadel write in The Hippocampus as a Cognitive Map that "Space was a way of perceiving, not a thing to be perceived." Sentence number 2 of the book states "We shall argue that the hippocampus is the core of a neural memory system providing an objective spatial framework within which the items and events of an organism's experience are located and interrelated." Consistent with Kant and O'Keefe and Nadel, the present authors might more accurately state "All behavior is embedded within a subjective spatial and temporal framework." but then they will have to explain why they conceive of there being "positional inputs" to which they are measuring CA1 responses. This framing seems to me problematic and not logically self-consistent.

    (6) On page 2 the authors assert "Neurons within the hippocampus respond to a wide array of sensory and otherwise nonspatial cues..." then they go on to list sensory features and "non-positional" features of experience to which CA1 cells respond. It seems to me they leave out a class of features of experience that might be considered "subjective spatial frames" that have been investigated by Gothard and Redish when they were in the McNaughton and Barnes lab, as well the Fenton and Muller labs, amongst others. All of these papers describe non-stationary, multi-stable place cell phenomena that are tied to subjective variables, which have the potential to undermine the premise of the present work's analyses and so they should be considered. I list a sample but certainly not all the work that might be considered.

    Gothard KM, Skaggs WE, Moore KM, McNaughton BL (1996) Binding of hippocampal CA1 neural activity to multiple reference frames in a landmark-based navigation task. J Neurosci 16:823-835.

    Gothard KM, Skaggs WE, McNaughton BL (1996) Dynamics of mismatch correction in the hippocampal ensemble code for space: interaction between path integration and environmental cues. J Neurosci 16:8027-8040.

    Gothard KM, Hoffman KL, Battaglia FP, McNaughton BL (2001) Dentate gyrus and ca1 ensemble activity during spatial reference frame shifts in the presence and absence of visual input. J Neurosci 21:7284-7292.

    Redish AD, Rosenzweig ES, Bohanick JD, McNaughton BL, Barnes CA (2000) Dynamics of hippocampal ensemble activity realignment: time versus space. J Neurosci 20:9298-9309.

    Rosenzweig ES, Redish AD, McNaughton BL, Barnes CA (2003) Hippocampal map realignment and spatial learning. Nat Neurosci 6:609-615.

    Jackson J, Redish AD (2007) Network dynamics of hippocampal cell-assemblies resemble multiple spatial maps within single tasks. Hippocampus 17:1209-1229

    Lenck-Santini PP, Fenton AA, Muller RU (2008) Discharge properties of hippocampal neurons during performance of a jump avoidance task. J Neurosci 28:6773-6786.

    Fenton AA, Lytton WW, Barry JM, Lenck-Santini PP, Zinyuk LE, Kubik S, Bures J, Poucet B, Muller RU, Olypher AV (2010) Attention-like modulation of hippocampus place cell discharge. J Neurosci 30:4613-4625.

    Kelemen E, Fenton AA (2013) Key features of human episodic recollection in the cross-episode retrieval of rat hippocampus representations of space. PLoS Biol 11:e1001607.

    (7) The Introduction asserts that "rate remapping" is a hypothesis. Rate remapping is a phenomenon, something that is observed. The interpretation of the observation as being the substrate of episodic memory is certainly a hypothesis that in my opinion has not been tested and is not being tested in the present work. After making the above statement, the authors go on to describe that firing rates differ across "repeated" firing fields, which seems to be a form of rate remapping, and predicted by the relevant hypothesis that different episodes of experience at the same locations are represented by different firing rates. This is very speculative and there are many other explanations.

    (8) The Introduction ends with the statement "Here, we show that repeating fields of the same neuron do not always display the same nonpositional rate modulation, demonstrating that nonpositional cues are dissociable from, and more flexible than, the positional inputs onto place cells in a given environment." Apart from my concern about using the "input" terminology I which to point out that there is very little novel in this statement. It has been described many times before that on linear tracks CA1 firing fields are directionally modulated such that the field rates for traversals in one direction are different compared to field traversals in the opposite direction. Jackson and Redish (2007) cited above show this to be due to reference frame or map switching. That and other work allow one to state that "Others show that repeating fields of the same neuron do not always display the same nonpositional rate modulation, demonstrating that nonpositional cues are dissociable from, and more flexible than, the positional inputs onto place cells in a given environment." Either the present authors should acknowledge that they are demonstrating what others have already demonstrated, or they should more precisely describe what about their contribution is unique.

    (9) Page 6 Methods - Data Filtering and Pre-processing. How did the authors handle theta cells and others that fired more or less everywhere but with spatial modulation?

    (10) Page 9 Methods - Why was the session-wide activity used to normalize the firing rates for the activity vector input to the random forest classifier? The authors state "The normalized firing rate was computed as discussed above with the change that the session-wide activity in the alley was used." It seems to me better to have used the session-averaged firing rate map because the activity would be normalized by the expected positional firing. I imagine "The classifier used the population vector of firing rates as the input." is incorrect and the authors mean to state "The classifier used the population vector of normalized firing rates as the input."

    (11) What does "spatially-gated" mean? The use of such jargon should be explained, or better avoided.

    (12) Page 12: Since fields tend to have similar orientations, but not repeat at all geometrically similar locations, did they tend to be clustered? Was there a proximity feature to their distribution?

    (13) Page 18 states "Thus, although there was a slight trend for repeating field ..." The authors are reporting a significant effect not a "slight trend." They do something similar in reporting Figure 5's result. Despite significant effects, they seem to think the findings are not large enough so state that repeating-field directionality is not conserved. It is fine to explain that a significant effect was small (for example give the effect size, which would have been welcome throughout) but as in these cases and others, the authors should be more objective in their reporting of the outcomes. Either a statistical test was or was not significant. It is not "a little" or "a lot" significant.

    (14) Page 18: What do the authors mean by "topology?" Might they mean "topography?"

    (15) Figure 6 shows field instability and multi-stability (termed temporal dynamics) as described on page 22. The recording sessions were 60 min. Is this impression simply due to long recording sessions? If 10 or 15 minutes of data were analyzed (which is more the norm), would similar instability be observed/detectable?

    (16) I found the Discussion very confusing. On the one hand, there is an assertion that because the location of firing fields is stable there is a "positional code." How would that actually work? Any neural system has to signal by firing rates or firing coincidences across groups of cells (that are affected by changes in rate) so if there is firing field firing rate instability the authors should explain how position can be accurately decoded on a behaviorally-meaningful time scale. In fact, they should demonstrate such decoding explicitly. Just because there is modulation and instability, it is a rather long leap to assert that this is how episodic experience/memory is encoded (as stated at the end of the abstract and elsewhere for example on page 24: "The present data utilize repeating fields to suggest that, within an environment, the positional inputs are relatively rigid, whereas the nonpositional inputs are more flexible, allowing different repeating fields to show different directional preferences. In other words, fields are individually addressable with respect to the nonpositional inputs they receive; they do not inherit their nonpositional tuning as a global property of the cell." What does it mean that a field is "individually addressable?" How is that achieved by neurons? If the authors want to make such assertions they should explain and demonstrate how their assertions can be valid, given the data and findings. At least they should explain what they are assuming.
    The main findings seem related to the published finding that in large environments place cells have multiple firing fields, with distinct rates in each field, quite similar to what is here described in the city maze. In my opinion, positional representations can only plausibly work in such cases by using the conjoint population activity moment to moment, which necessarily marginalizes the value of individual firing fields, yet the present work focuses the discussion (and analyses) on interpretations of single firing fields (which they assert are individually addressable multiple times). I don't know what that means exactly and the authors should explain why maintaining the standard single-field perspective is appropriate and how position can be represented in such a system, given the data. In fact, I would have thought that the present findings would cause the authors to reject as invalid the framework they have adopted.

    (17) This is a further example, on page 25 which asserts that "Directionality is affected by an animal's experience through the field (Navratilova et al., 2012), so it is possible the difference in experience between sampling fields on the same versus different corridors affects the directional tuning properties between them." I do not understand how "the difference in experience between sampling fields on the same versus different corridors affects the directional tuning properties between them." If I follow the logic then the so-called directionality would depend on experience and so only emerge after a certain time for experience, or else the firing during one traversal would need to be modulated by information about future traversals, which I suppose the authors would agree does not make sense.

    (18) I found it at times confusing to follow the arguments because the terms "route" and "trajectory" and also "direction" and "heading" were used sometimes interchangeably and sometimes in ways that appear distinct.

    (19) Page 25 states "One explanation for these data is that fields sampled along contiguous routes, without interruptions from heading change or reward delivery, are more likely to share their directionality." The authors should consider alternative explanations like reference frame shifts as mentioned in comment 6 above. These alternatives can be rejected based on data, but they should be considered because they seem to offer more parsimonious explanations for the observations than what the authors have offered. For example, what can explain the bimodality reported in Fig. 5G?

    (20) The authors assert on page 15 that "In the present study, turns at the ends of corridors, along with reward deliveries, may be salient task boundaries at which point theta sequences are terminated. Fields active within the same theta sequence (typically same corridor fields) may be functionally coupled, while fields active on opposite sides of a theta sequence termination (different corridor fields) may be uncoupled and their tuning uncorrelated." The authors should check this. They recorded the LFPs. Why speculate when they can evaluate the speculation?

    (21) The authors assert on page 26 "It is important to note that because a Pearson correlation was used, it is possible the fields are related in time with a phase shift, and we did not have the statistical power to test this possibility adequately." I either do not understand this statement or it is untrue. Please clarify.

    (22) The authors continue on page 26, asserting "Thus, although it is clear that the place fields of repeating cells do not change their firing rates in synchrony, as if the cell had a global excitability change that made all its fields wax and wane together, it nonetheless remains an open question as to whether the subfields of repeating cells engage in certain types of competitive interactions or other network dynamics that couple changes in their firing rates in more complex ways." This statement implies that it might even be possible for firing fields in distinct and distant locations to be modulated together. Could the authors please explain how that is possible? A firing field is an observation that requires averaging over minutes and behavioral sampling across minutes. How might one cell be modulated to fire at a low rate during one minute and then at another minute later be modulated to fire at a high rate everywhere in the environment? Perhaps I am again not understanding the assertion - please clarify.

  3. eLife

    Reviewer #2 (Public Review):

    The authors present evidence that free-foraging behavior within an environment having structural regularity in its distribution of obstacles (an internal "city block" configuration) yields multiple place-specific firing fields for CA1 neurons. These fields tend to be aligned to analogous locations within the environment. Aligned fields tend to share direction-biased tuning of place-specific activity. The distribution of in-field firing rates across repeating fields of individual neurons varies and in a reliable enough fashion, that reconstruction of the animal's location in the environment can still be achieved. These results are interpreted as reflecting a combined mapping of environmental position as well as repeating structural features of the environment. The results have strong implications for understanding how navigation and spatial awareness might be represented within environments having such regularities (e.g., a city such as Manhattan). Further, the results suggest that repeating firing fields for CA1 neurons can develop in the absence of regularized path-running behavior. Finally, the authors consider drift in the character of the representation across time to represent the position in time across the foraging session. This last claim lacks evidence for reproducibility and is unnecessarily speculative. Altogether, the work is original and, for the most part, well-evidenced.

  4. Version published to 10.1101/2022.06.01.494227v2 on bioRxiv
  5. Version published to 10.1101/2022.06.01.494227v1 on bioRxiv