(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Prefrontal cortical activity predicts the occurrence of nonlocal hippocampal representations during spatial navigation

['Jai Y. Yu', 'Department Of Psychology', 'Institute For Mind', 'Biology', 'Neuroscience Institute', 'University Of Chicago', 'Chicago', 'Illinois', 'United States Of America', 'Loren M. Frank']

Date: 2021-10

The receptive field of a neuron describes the regions of a stimulus space where the neuron is consistently active. Sparse spiking outside of the receptive field is often considered to be noise, rather than a reflection of information processing. Whether this characterization is accurate remains unclear. We therefore contrasted the sparse, temporally isolated spiking of hippocampal CA1 place cells to the consistent, temporally adjacent spiking seen within their spatial receptive fields (“place fields”). We found that isolated spikes, which occur during locomotion, are strongly phase coupled to hippocampal theta oscillations and transiently express coherent nonlocal spatial representations. Further, prefrontal cortical activity is coordinated with and can predict the occurrence of future isolated spiking events. Rather than local noise within the hippocampus, sparse, isolated place cell spiking reflects a coordinated cortical–hippocampal process consistent with the generation of nonlocal scenario representations during active navigation.

Funding: This work was supported by a Jane Coffin Childs Memorial Fund for Biomedical Research postdoctoral fellowship (J.Y.Y.), the Howard Hughes Medical Institute, the Kavli Institute for Fundamental Neuroscience, and University of California Office of the President Lab Fees Award #LF-12-237680 (L.M.F.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Our examination of isolated spiking of place cells revealed that these events reflect the coherent activation of hippocampal representations of physically distant locations and that these events are coordinated with ongoing activity in the PFC. These findings suggest that isolated spikes are a signature of distributed and coherent information processing in the brain.

We therefore examined spiking both within the hippocampus and across the hippocampus and prefrontal cortex (PFC), focusing on activity during movement. PFC is anatomically connected to the hippocampus through both direct and indirect projections [ 31 – 33 ], and coordinated activity across these networks reflects their engagement during memory processing [ 34 – 36 ]. For example, network level coherence between PFC and hippocampus increases during periods when memory retrieval occurs [ 37 – 44 ]. Whether PFC activity differs systematically at the time of isolated spiking in the hippocampus remains unknown.

How can we determine whether isolated spiking outside of a place cell’s spatially and directionally tuned receptive field reflects information processing in the hippocampal circuit as opposed to activity that does not reflect information processing or noise? Spiking due to stochastic cellular events is expected to be local to individual neurons. By contrast, spiking associated with information processing would be expected to covary in a consistent manner across neurons in both local and distributed networks [ 30 ]. Thus, if spiking outside of the classical place field conveys information, we would expect it to (1) be coordinated across multiple hippocampal neurons; (2) contain coherent spatial information; and (3) be coordinated with activity outside the hippocampus.

Noise in neural networks can arise from stochastic cellular events that cause the membrane voltage to occasionally exceed the action potential threshold, even without upstream input [ 23 , 24 ]. While the spatially and directionally selective inputs to a place cell raise the membrane voltage closer to the action potential threshold when an animal approaches the cell’s place field [ 25 , 26 ], stochastic events causing occasional increases in membrane potential could result in spiking outside of a cell’s place field. However, previous observations indicate that at least some spiking outside of a cell’s typical place fields reflect mnemonic processes rather than noise. CA1 and CA3 place cells can emit spikes outside of their place fields as an animal approaches choice points [ 27 , 28 ] and during vicarious trial and error [ 27 ] or when an animal is traveling in the opposite direction over a location with a place field [ 28 ]. These events are hypothesized to reflect noncurrent scenarios, such as simulating possible future scenarios when a decision needs to be made [ 28 , 29 ].

Although the majority of place cell spiking occurs when an animal is moving within the cell’s place field(s), occasional spiking occurs when the animal is at locations outside the field(s) [ 5 , 6 , 15 – 17 ]. These “isolated” spiking events can occur during movement and are distinct from sparse spiking observed during sharp-wave ripples (SWRs) seen during immobility [ 18 ]. Importantly, isolated spikes are not locked to specific locations. As a result, standard analyses that average activity across many passes through the same location [ 13 , 15 , 19 – 22 ] effectively exclude these spikes from further consideration. Whether these spikes reflect unreliable, noisy processes that merit exclusion or whether they instead reflect coherent, meaningful signals remains unknown.

The concept of a receptive field [ 1 – 3 ] provides a fundamental model for how neural spiking can convey information about features in the external environment. In the hippocampus, many cells show spatially tuned receptive fields [ 4 , 5 ]. The spiking rate of these “place cells” rises and then falls as an animal traverses specific locations in an environment. In linear environments, the animal’s movement direction can also modulate spiking [ 6 – 10 ], resulting in location and direction-specific activity. Locations and directions with high spiking rates are defined as a cell’s “place fields” [ 5 , 6 ], and place field–associated spiking of place cells conveys sufficient spatial information to estimate the animal’s location with high accuracy [ 11 – 14 ].

Results

In order to understand the extent of isolated spiking during active behavior and to identify a potential function of this activity, we took an unbiased approach where we surveyed CA1 place cell spiking across all movement periods (animal speed >2 cm/s) as animals performed a spatial navigation task in a complex environment with multiple linear track segments [45,46] (Fig 1A and 1B). In the hippocampus, the temporal structure of spiking during locomotion is strongly influenced by the endogenous approximately 8-Hz theta rhythm [47], and bouts of higher rate spiking corresponding to place field traversals spanned multiple, adjacent cycles of theta (Fig 1C). As expected, we also observed isolated spikes where a neuron would be silent for many theta cycles, emit a small number of spikes on a single theta cycle, and then return to being silent (Fig 1D) [15,19,20,27].

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Isolated and adjacent spiking activity of hippocampal CA1 place cells. (A) Occupancy normalized spiking rate maps for spiking activity during active movement (animal speed >2 cm/s) across behavior sessions for each day for 4 example CA1 cells. (B) Location of spiking (black dots) and animal trajectory (gray) for occupancy maps in A. Spike count shown below each panel. (C) Spike raster and corresponding location for a bout of spiking activity over adjacent theta cycles. Raw (orange) and theta frequency filtered (black) LFP are shown below the spike raster. The corresponding location of the animal’s trajectory (orange line and arrow) and the spikes (black dots) on the maze for the bout are shown on the right. (D) Spike raster and corresponding location for spiking isolated from other spiking activity. The corresponding location on the maze for the bout is shown on the right. (E) Distribution of mean separation between theta cycles with spiking. Separation is defined as the mean cycle count to 3 nearest neighbor cycles with spiking. (F) Population distribution of mean separation between theta cycles with spiking (n = 298 cells). LFP, local field potential. https://doi.org/10.1371/journal.pbio.3001393.g001

The standard approach to defining place field spiking relies on averaging spiking rates across many traversals of a location. This average provides a useful experimental summary of spiking, but information averaged across traversals is not directly available to downstream neurons. We therefore we used a criterion to distinguish between “adjacent” and “isolated” spiking based on the local temporal organization of spiking. Specifically, given the importance of theta in organizing hippocampal activity [47–49], we calculated the interval between neighboring theta cycles with spiking (in cycles, mean of nearest 3) (Fig 1E, S1A Fig). As expected, most spike-containing theta cycles are near another spike-containing cycle. The remaining spike-containing theta cycles are separated from neighboring spike-containing cycles by up to hundreds of cycles, reflecting their temporal isolation. When plotted on a log scale, the underlying distribution was bimodal, and based on this distribution, we chose a threshold of 8 cycles of mean separation to each theta cycle with spiking to define “adjacent” or “isolated” activity (n = 298 cells, Fig 1F). This method identifies spiking on individual theta cycles and also spiking on a small number of nearby cycles that are nevertheless isolated from periods of adjacent spiking.

This separation captured intuitive notions of within- and extra-place field activity: Adjacent activity was spatially concentrated and had high spiking rates, as expected from place-field spiking (Fig 2A). By contrast, isolated activity was spatially sparse and lack the high spiking rates observed for place field activity. Nonetheless, isolated spiking represented 17 ± 1.8% (median ± 95% CI) of spikes included for analysis (S1B Fig). As expected, while individual adjacent spikes tended to occur at locations close to other adjacent spikes (median adjacent-adjacent = 0.49 cm; Fig 2B), isolated spikes tended to occur at more distant locations (median isolated-adjacent = 8.98 cm, p = 6.02 × 10−76; Fig 2B). In cases where isolated spikes occurred at locations close to adjacent spiking, these isolated spikes typically occurred on a different trial (Fig 2C, median adjacent-adjacent = 0.18 seconds, median isolated-adjacent = 28.23 seconds, p = 3.10 × 10−57; trials are approximately 7.5 seconds long [45]). Consistent with previous findings, these isolates spikes most often occurred when the animal was traveling in the opposite direction compared with adjacent spiking (Fig 2D, median adjacent-adjacent = 0.22°, median isolated-adjacent = 158.9°, p = 4.66 × 10−76) [6–9,28]. Importantly, isolated spiking was not well explained by an overall tendency to show more spatially diffuse representations since the spatial distribution properties of adjacent spiking are not correlated with the proportion of isolated spiking (S2 Fig). We also verified that isolated spikes, although sparsely emitted, were very unlikely to be spike cluster assignment errors (S3 Fig).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Spatial and temporal separation between isolated and adjacent spiking. (A) Location of spiking classified as adjacent or isolated activity for the 4 example cells in Fig 1A. Spike count shown below each panel. (B) Mean distance on the maze from one spike (adjacent: blue or isolated: red) to its nearest 5 neighboring adjacent spikes. The distance is the shortest path on the maze between 2 spikes. Wilcoxon rank sum test: p = 6.02 × 10−76. (C) Mean separation in time between one spike and other adjacent spikes that occur at locations on the maze within 1 cm. Wilcoxon rank sum test: p = 3.10 × 10−57. (D) Mean difference in the trajectory vector between one spike and other adjacent spikes within 1 cm. Wilcoxon rank sum test: p = 4.66 × 10−76. (B–D) Histogram shows mean ± SEM for across 247 cells. https://doi.org/10.1371/journal.pbio.3001393.g002

As expected, isolated spiking was also highly concentrated within the later phases of each theta cycle (Fig 3A and 3B). Place field–associated spiking displays strong phase coupling to the hippocampal theta rhythm, where the maximum probability of spiking occurs in earlier phases near the trough of theta [47,50]. Later phases correspond to times where inhibition is lower, and, thus, activity outside the main place field could be generated [26,49]. Isolated spiking was also more tightly phase locked to theta compared with adjacent spiking (Fig 3C). This was true both for isolated spikes that occurred in locations close to locations where adjacent spiking was seen and for isolated spikes that occurred far from those locations (S4 Fig), indicating that isolated spiking has similar network coupling properties irrespective of their spatial proximity to the cell’s adjacent spiking.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Isolated and adjacent spiking activity show distinct phase locking to hippocampal theta oscillation. (A) Theta cycle separation versus mean spike theta phase preference. A separation threshold of 8 cycles between isolated and adjacent is based on Fig 1. Histogram shows the mean spiking phase for each theta cycle. Examples correspond to the 4 cells from Fig 1. Circular median test between the isolated and adjacent distributions: top left: p = 1.4 × 10−5; top right: p = 8.9 × 10−8; bottom: left p = 2.4 × 10−4; bottom right: p = 5.3 × 10−2. (B) Mean theta phase preference distribution for adjacent and isolated spiking for the CA1 cell population (n = 247 cells). Circular median test: p = 0. (C) Mean theta phase concentration distribution for adjacent and isolated spiking for the CA1 cell population (n = 247 cells). Wilcoxon rank sum test: p = 2.09 × 10−28. https://doi.org/10.1371/journal.pbio.3001393.g003

We also ensured isolated spiking was not associated with SWRs, which are transient network oscillations observed in the local field potential (LFP) and are predominantly found when the animal is moving slowly or is immobile [18]. This was done by excluding SWRs from our analyses (see Methods) and independently confirming the isolated spiking events did not have the spectral signature of SWRs. The LFP associated with excluded spiking showed a network spectral signature consistent with SWRs (S5A Fig, left column; S5B Fig), with power in the slow gamma (approximately 30 Hz) and ripple frequencies (approximately 150 to 250 Hz). In contrast, the LFP associated with isolated spiking shows a different network spectral signature, with power in the theta band [50] (S5A Fig, right column; S5B Fig). Indeed, the network spectral signature of isolated spiking is very similar to the LFP associated with adjacent spiking and even has slightly higher theta power (S5A Fig, center column; S5B Fig).

Recent findings from our group indicated that spiking related to possible future locations or opposite directions of travel can occur in animals traveling at high speeds and in the absence of overt deliberative behaviors [28]. We replicated these findings for isolated spikes: Isolated activity was not more frequent around choice point locations (Fig 4A), nor were there differences in the speed (Fig 4B) or angular acceleration (Fig 4C) of the animal at times of isolated as compared to adjacent spiking. Thus, isolated spiking is not restricted to specific active behavioral states or locations, such as path choice points. We next examined the relationship between isolated spiking and task behavior. We reasoned that if these events reflect task-related cognitive processes, we may see differences in the rate of isolated spiking across different phases of a behavior epoch. For the early trials in a behavior session, when the animal explores the environment in order to find the current reward locations, its performance is low. In the last trials of an epoch, when the animal has correctly identified the reward locations, its performance becomes high (Fig 4D). The difference in performance is not a reflection of unfamiliarity with the task since the median trial durations are comparable (Fig 4E). During the first 5 trials, we found a significantly higher rate of isolated spiking compared with the last 5 trials of a behavior session (Fig 4F). This was not the case for adjacent spiking (Fig 4G), and overall place representations are stable across each behavior session. Thus, these observations suggest that isolated spiking is associated with discovering new reward rules during early trials of a new behavior session.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Spatial distribution and behavioral correlates of isolated and adjacent spiking. (A) Normalized spatial distribution of theta cycles with adjacent (left) or isolated (center) spiking. Normalized difference between the spatial distributions (right). (B) Distribution of animal speed (mean ± SEM) at the time of adjacent or isolated activity (top). Significance of the difference (z) between the 2 distributions as determined using a permutation test (bottom). Dotted lines indicate ± 2 z. (C) Distribution of animal angular acceleration (mean ± SEM) at the time of adjacent or isolated activity (top). Significance of the difference (z) between the 2 distributions as determined using a permutation test (bottom). Dotted lines indicate ± 2 z. (D) Mean performance for the first and last 5 trials of each behavior session (n = 38). Wilcoxon signed rank test: p = 7.74 × 10−8. (E) Median duration for the first and last 5 trials of each behavior session (n = 38). Wilcoxon signed rank test: p = 0.89. (F) Mean isolated spiking rate for the first and last 5 trials of each behavior session (n = 247 cells). Wilcoxon signed rank test: p = 9.79 × 10−9. (G) Mean adjacent spiking rate for the first and last 5 trials of each behavior session (n = 247 cells). Wilcoxon signed rank test: p = 0.91. https://doi.org/10.1371/journal.pbio.3001393.g004

Individual hippocampal place cells can be active in cell assemblies [51] that show temporally correlated activity on multiple timescales [52–54]. High spiking rate activity of these cell assemblies, typically associated with place fields, express information about current location of the animal. We therefore asked whether isolated spiking reflects coordinated activity between CA1 cells and sought to identify their corresponding spatial representations. Specifically, we knew that spiking during the late phases of theta is associated with the expression of nonlocal representations, including to be visited locations or locations previously visited [28,55,56]. Given that isolated spikes occur at late phases of theta and are separated from adjacent spiking both in time and space, could isolated spiking reflect the transient activity of cell assemblies with place field activity in another part of the environment? If so, then we would expect that pairs of neurons that are coactive during periods of adjacent spiking, corresponding to cells that are likely to have overlapping place fields, would also be coactive within a theta cycle containing isolated spiking events. We examined this possibility by using an approach that has been used to demonstrate reactivation of nonlocal spatial representations during SWRs, where a pair of place cells is more likely to spike together if their place fields overlap [57,58] (Fig 5A). First, we calculate the likelihood of co-spiking for a pair of place cells that had isolated spiking within the same theta cycle. We then quantified the overlap in their adjacent spiking activity. We found that cells that fired together during periods of adjacent spiking were also more likely to fire together during isolated spiking events. Across the population, lower lags in spiking during adjacent activity were correlated with greater co-spiking during isolated events (Fig 5B, R = −0.28, R2 = 0.077, p = 6.40 × 10−9). These findings support the notion that isolated spikes may reflect the spiking of cell assemblies with spatial representations for locations away from the animal or in direction of travel opposite to the current direction.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 5. Reactivation of spatiotemporal place field activity relationships during theta cycles with isolated spiking activity. (A) Three pairs of CA1 cells with overlapping adjacent activity. The place fields (occupancy normalize spiking rate >5 Hz) for each cell as well as their spatial overlap are shown. Example spiking bouts of adjacent and isolated activity are shown with raw and theta frequency band filtered LFP. The animal’s trajectory (orange line and arrow) on the maze for each bout is shown on the right. (B) Normalized coactivity (z) for CA1 cell pairs during theta cycles with isolated activity (n = 425 pairs) grouped by the mean separation in time (mean lag) between their adjacent activity (R = −0.28, R2 = 0.077, p = 6.40 × 10−9). LFP, local field potential. https://doi.org/10.1371/journal.pbio.3001393.g005

We next asked whether isolated spiking events reflect spontaneous activation of cell assemblies in CA1 or coordinated activity with other functionally connected networks. We examined simultaneously recorded activity in PFC, a region that connected with CA1 by mono- and multisynaptic pathways [31–33]. Given the anatomical connectivity between these regions, evidence of spiking coordination between hippocampus and PFC would provide strongly suggest that these events are not the result of spontaneous activation of local cell assemblies in the hippocampus but instead reflect coherent and structured activity across brain regions. An example of such hippocampal–cortical engagement occurs during SWRs, where hippocampal reactivation is accompanied by the coordinated reactivation of cortical representations [45,46,59–62]. If such coordination is seen around the times of isolated spikes, we should be able to identify PFC neurons that spike differently around times of isolated activity in the hippocampus than at comparable periods where isolated spiking was not observed.

Although isolated spiking does not occur at specific locations, we can use the times of isolated spiking events as reference points to look for coordination with PFC activity. We first selected theta cycles with isolated spiking for a given CA1 cell. Next, we found matching theta cycles from other times when the animal was moving through the same locations in the same direction at a similar speed, but where the CA1 cell was not active (e.g., did not have isolated spiking) (S6 Fig). This was possible because, in our task, the animal traversed a given location multiple times, providing a pool of theta cycles, of which only a subset contained isolated spiking. Importantly, none of the matching cycles contained adjacent spiking, confirming that the isolated spiking events were not simply events on the edge of a place field. We then compared the spiking of simultaneously recorded PFC neurons between cycles with isolated activity and these matched control cycles (S6A Fig). We note that theta coordinates activity in hippocampal–cortical networks [40, 41], allowing us to continue to use theta cycles as the temporal reference to relate activity across structures.

We found multiple PFC cells whose spiking rate differed depending on whether there was an associated period of isolated spiking for a given CA1 cell (Fig 6B). We expect that only a small fraction of PFC cells would show a significant difference in spiking relative to the isolated spiking of a given CA1 cell, but, nonetheless, across the population (n = 2,798 PFC–CA1 cell pairs), the difference in PFC firing rates between isolated and matched control periods was significantly larger than the permutation control (Fig 6C, S7 Fig). This difference indicates coordination between CA1 and PFC around the time of CA1 isolated activity. Interestingly, this coordination was not limited to the specific isolated theta cycle: The difference remained significant even in a window of 8 to 12 theta cycles before the isolated spiking event, indicating that PFC activity could play a causal role in driving isolated spiking events in the hippocampus. In addition, we found significant PFC spiking rate differences remained after the occurrence of CA1 isolated activity (Fig 6D). We also verified that this coordination could not be explained by the higher rate of isolated spiking early in each epoch: Overall, there was no significant difference in the time interval between pairs of isolated cycles and pairs of matched cycles (S6E Fig). Thus, any coordination found cannot be explained by isolated cycles being closer together in time than their matched counterparts.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. PFC activity is coordinated with hippocampal isolated activity. Schematic illustrating potential CA1 and PFC activity around the time of isolated spiking. Changes in PFC spiking around the time of CA1 isolated spiking may reflect coordination between the 2 regions. (A) Example spike raster and spiking rate (mean ± SEM) for pairs of co-recorded hippocampus CA1 and PFC cells. Each raster shows spiking aligned to isolated hippocampal activity (cycle 0) and matched control trials. Spiking is plotted relative to the cycles of the hippocampal theta rhythm. For CA1 cells, red indicates spikes and spiking rate for intervals with isolated spiking at cycle 0. Black indicates control intervals without isolated spiking at cycle 0. For PFC cells, purple indicates spikes and spiking rate for intervals with isolated spiking at cycle 0. Black indicates control intervals without isolated spiking at cycle 0. (B) Violin plots and quantification of spike rate differences between control and actual intervals for PFC–CA1 cell pairs (n = 2,798) in time windows relative to CA1 isolated activity. Rate difference, original data (black) and permuted (gray), is expressed the z-score of the absolute observed difference relative to its own permuted distribution. The Wilcoxon signed rank test (*** p < 0.001) was used to compare the original and permuted groups: p = 4.7 × 10−8, 3.6 × 10−12, and 2.2 × 10−10 for each group, respectively. (C) Violin plots and quantification of spike rate differences between control and actual intervals for PFC–CA1 cell pairs (n = 2,892) in time windows post CA1 isolated activity. The Wilcoxon signed rank test (*** p < 0.001, ** p < 0.01) was used to compare the original and permuted groups: p = 9.5 × 10−7, 1.8 × 10−3, and 2.1 × 10−3 for each group, respectively. PFC, prefrontal cortex. https://doi.org/10.1371/journal.pbio.3001393.g006

If these differences signify coordinated activity, the ensemble activity of PFC neurons should predict the future occurrence of hippocampal isolated activity (Fig 7A). To test that prediction for a given CA1 cell, we used the spiking activity from all simultaneously recorded PFC ensembles (median n = 20, IQR = 8 PFC cells per CA1 cell) to build cross-validated generalized linear models (GLMs) with elastic net regularization. We compared the ability of the models to predict the occurrence of isolated activity related to a permutation control (see Methods; S8 Fig). We then carried out that analysis for CA1 cells (n = 158) with isolated spiking.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. PFC activity predicts the occurrence of hippocampal isolated spiking. (A) PFC activity leading up to isolated spiking is used to predict the future occurrence of isolated spiking in one CA1 place cell. (B) Prediction gain (mean ± SEM) of GLMs where PFC spiking activity is used to predict isolated spiking in the upcoming CA1 theta cycle (n = 158). Pairwise permutation test (** p < 0.05) for mean with multiple comparison correction: p = 0.079, p = 0.0027, p = 0.0006, and p < 0.0002 for each group, respectively. (C) PFC activity after isolated spiking is used to predict the previous occurrence of isolated spiking in one CA1 place cell. (D) Prediction gain (mean ± SEM) of GLMs where PFC spiking activity is used to predict whether CA1 isolated activity had occurred (n = 162). Pairwise permutation test (** p < 0.05) for mean with multiple comparison correction: p = 0.0071, p = 0.0002, and p = 0.0053 for each group, respectively. GLM, generalized linear model; PFC, prefrontal cortex. https://doi.org/10.1371/journal.pbio.3001393.g007

We found that PFC activity can predict the occurrence of isolated spiking in CA1 at above chance levels, even in a window of 4 to 8 theta cycles before isolated spiking (Fig 7B). We also asked whether there was any evidence consistent with isolated spiking in CA1 influencing subsequent PFC activity (Fig 7C). We found that the coordination between hippocampus and PFC persists after the occurrence of isolated activity but is weaker compared to intervals immediately before and during cycles with isolated activity (Fig 7D). We also found a weak but significant increase in prediction gain over the time bins, consistent with the strength of prediction increasing toward the isolated cycle (R data = 0.106, p data = 0.0073; R perm. = 0.0005, and p perm. = 0.990). We noted the average prediction gains were small in magnitude, which has previously been observed for prediction gains relating auditory and hippocampal activity around the times of SWRs [60]. This is not surprising given the relatively small numbers of simultaneously recorded PFC cells that were available to predict the activity of any given CA1 unit. We can therefore regard these cross-validated predictions as lower bounds on the actual values that would be obtained if it were possible to sample the entire PFC population. Indeed, examining the values for individual PFC ensemble—CA1 models revealed several cases with prediction gains between 2.5% and 5% (S8 Fig). Thus, our results demonstrate that information expressed by prefrontal cortical and hippocampal cell populations is coordinated around the time of isolated activity.

Importantly, the predictive PFC activity patterns were specific for individual CA1 cells. We examined the correlation between β coefficients of PFC predictors across predictive models. If the spiking of specific PFC cells was strongly predictive of isolated spiking of a particular CA1 cell but not of other CA1 cells, this β coefficient correlation should be low, indicating that a given PFC cell would predict the spiking in one model (e.g., one CA1 cell) but not another. By contrast, if a subset of PFC cells consistently predicted isolated spiking across CA1 cells, then these correlations would be high, as the same PFC cells would show similarly β coefficients across models. We found that the mean correlation coefficient was not significantly different from 0 (median = −0.021, IQR = 0.16, Wilcoxon rank sum test p = 0.431). This indicates that the PFC ensembles predicting the occurrence of isolated activity for different CA1 cells are distinct and argues for specificity in PFC–CA1 coordination around the occurrence of isolated activity.

[END]

[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001393

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.