mPFC single units have task-related firing patterns in the standard EPM
To characterize the activity of mPFC single units in the EPM, 79 well-isolated cortical single units were recorded from the deep layers of the prelimbic cortex in 17 129/SvevTac mice during exploration of a standard cross-shaped EPM under dim (200 Lux) illumination. The mean firing rate of these units was 2.05 ± 0.64 Hz. Units with fewer than 100 spikes (n = 10) were excluded from further analysis. Spatial firing maps revealed that many of the single units tended to fire in specific subcompartments of the EPM (). For example, the unit shown in fired preferentially in the two closed, or “safe” arms, while the unit in fired preferentially in the two open, or “aversive” arms.
mPFC single units have task-related firing patterns in the EPM
To further characterize firing patterns across the entire population of recorded mPFC units, normalized firing rates (% difference from mean firing rate) were calculated in the five compartments (each open arm; each closed arm; and the center) of the EPM (). Units with task-related firing patterns would be expected to have similar firing rates in arms of the same type (open/safe vs. closed/aversive), and negatively correlated firing rates in arms of opposite type. In line with this prediction, firing rates in both closed arms (r=+0.38, p<0.0001, ) and both open arms (r=+0.25, p<0.04, ) were positively correlated, while firing rates across arms of different types were inversely correlated (r= −0.64 p<0.0001, ). Note that with the presence of a center compartment, the inverse correlation between arms of different types is not an automatic consequence of the normalization technique (Figure S1
mPFC single units differentiate between open and closed arms in the EPM
Negative correlations between one open and one closed arm were present after only 90 seconds of exploration of the EPM (r=−0.47, p<0.001), demonstrating that single unit representations of EPM arms arise quickly and do not require extensive exploration of the maze. The results were not due to novelty, as similar results were found during a second exposure to the EPM 24 hours later (). Moreover, the results were not due to differences in locomotion between the open and closed arms, as velocity and acceleration profiles were similar across arms (), and firing rates did not correlate with either measure (r2=0.03, p=0.6 for velocity and r2=0.02, p>0.72 for acceleration).
mPFC firing patterns in the EPM are not due to novelty or locomotor differences
Correlations of firing rates between different arms indicate that the population of mPFC single units is capable of representing anxiety-related task components. However, such correlations do not quantify the extent to which the firing pattern of any given single unit is paradigm-related. To address this issue, we first binned each spike train into three-second segments, and calculated the influence of arm type (open vs. closed) on firing rate by ANOVA. 29/69 (42%) of the recorded neurons fired significantly differently (p<0.05) to the closed and open arms by ANOVA. Next, to confirm that the observed frequency of task-related firing patterns in the population of single units was not due to chance, an EPM score was calculated for each unit. The EPM score is a normalized ratio of the average difference in firing rates across arms of the same type, compared to the average differences in firing rates across arms of different types (see Experimental Procedures). The resultant measure, which varies from −1 to 1, indicates the degree to which that unit’s firing pattern represents the “open vs. closed” structure of the EPM. Units with positive EPM scores closer to 1 represent this structure well; units with EPM scores near or below zero fire do not. Accordingly, the correlation of firing rates across arms of the same type was higher in units with positive EPM scores than in units with negative EPM scores (). Furthermore, single units with a significant effect of arm type on firing in the ANOVA had higher EPM scores than other units (mean score =0.3±0.06 and 0.064 ±0.04 for units with and without significant main effects of arm type), demonstrating the utility of the EPM score as a quantification of the strength of paradigm-related activity.
mPFC units with anxiety-related firing patterns are over-represented in the population
We next examined whether the distribution of EPM scores obtained in our sample () could have been obtained by chance, using a bootstrap method. Briefly, 500 simulated spike trains were generated for each unit. The location of each spike was assigned randomly from the actual path of the animal in the maze when that spike was recorded, and EPM scores were computed from these simulated spike trains. The distribution of simulated EPM scores (, red line) was significantly different from the experimental distribution (p<0.0001, Wilcoxon’s rank-sum test), due to the presence of a greater fraction of units with positive (i.e., paradigm-related) EPM scores in the experimental distribution. These results confirm that the paradigm-related firing patterns seen in our sample in the standard EPM were unlikely to have arisen by chance.
mPFC unit firing changes prior to leaving or entering the closed arms
In cognitive tasks, mPFC unit activity predicts future choice behavior (Fujisawa et al., 2008
; Peters et al., 2005
; Rich and Shapiro, 2009
). To examine whether this predictive capacity is seen in during anxiety-related behavior, peri-event time histograms were calculated for each unit across 10-second segments centered at transitions in which the animal exited or entered a closed arm (). Binned firing rates were then converted to z-scores and averaged across all cells with positive EPM scores units and all such transitions. As expected, units that fired preferentially in the closed arms had higher firing rates prior to leaving the closed arm (, upper panel). Consistent with predictive firing patterns, closed arm preferring unit firing rates began to decrease approximately 2.5 seconds before the mouse left the closed arm. Similarly, firing rates of open arm-preferring units were low in the closed arms, and began to increase several seconds before the transition point (, middle panel). During transitions back to the closed arms, firing rates of these neurons demonstrated complementary profiles (). In both types of transitions, units with negative (non-paradigm-related) EPM scores did not display consistent changes in firing rates.
Changes in unit activity precede transitions across compartments in the EPM
To quantitatively demonstrate predictivity, the time bins at which firing rates began to change were identified using a change point analysis (Gallistel et al., 2004
). This method identifies the point at which the slope of the cumulative sum of the time series of interest changes significantly (Kolmogorov-Smirnov test, p<0.01). The identified change points are indicated by arrows in . Note that in each case, mPFC single unit activity began to change 1.5–2.7s prior to the exit from or entry into the closed arm, demonstrating that firing rates are not simply passively reflecting the location of the animal but rather foreshadowing behavior a few seconds into the future.
To confirm these firing patterns using an unbiased approach, we used principal component analysis (Chapin, 2004
) on firing rates of all units during arm transitions (). As predicted from the firing patterns described above, the first principal component (PC1) during each transition type appeared to closely follow the patterns of closed- and open-arm preferring units, with PC1 value switching sign at or just prior to the transition point. Closed arm- and open arm-preferring units loaded inversely onto the PC1 for each transition type.
Firing patterns do not depend on arm location or specific sensory cues
The above data demonstrate that mPFC single units fired differently in closed and open arms of the EPM. However, firing patterns shown in could be induced by differences between the closed and open arms that are unrelated to anxiety. One such confound is the geometric arrangement of the arms. It is possible, for example, that a cell that is active preferentially in the open arms is actually firing not because the animal is in the open arms, but rather, because it is walking in the north-south direction. To exclude this possibility, 18 single units were recorded from five additional mice while they explored an altered EPM in which the open arms were adjacent to each other rather than across from each other (). Similarly to the results obtained in the standard EPM, firing rates in the altered EPM were positively correlated between arms of the same type (, respectively, for the closed arms (r=+0.71, p<0.0003) and for the open arms (r=+0.67, p<0.001 ). Furthermore, firing rates between closed and open arms were negatively correlated, as in the standard EPM (r= −0.54, p<0.002). To examine the relationship of firing across the two mazes, the same units were recorded while mice were exposed to a standard EPM after a 1-hour delay. Strikingly, firing rates between arms of the same type were positively correlated across the two configurations (, r=+0.43, p<0.04 for the closed arms and r=+0.53, p<0.01 for the open arms, n=18 units). The correlations between firing across the two mazes show that individual mPFC neurons follow arm type (open vs. closed) as opposed to arm location.
Paradigm-related firing patterns do not depend on the geometric arrangement of the arms
A second potential confound is the sensory experience used to induce avoidance. We reasoned that if the firing patterns of mPFC units are indeed associated with anxiety, units should differentiate between safe and aversive arms regardless of the particular anxiogenic cues used. To this end, we characterized the response of mPFC single units to openness and brightness, as both are anxiogenic, despite providing different sensory input. Anxiety induced by openness was studied in a standard EPM with two open and two closed arms, in the dark (closed/open maze). Reponses to anxiety caused by brightness were explored in an EPM with four closed arms, where two arms were brightly lit (dark/bright maze). These behavioral paradigms were both anxiogenic, as mice avoided the aversive (open or bright) arms in both conditions (% time spent in open arms and bright arms was 21.4 ± 5.3 and 20.3 ± 2.5, respectively, n = 5 naive mice; see ).
mPFC single units respond similarly to different aversive stimuli
An additional 8 implanted mice were exposed to both modified mazes. 105 single units were recorded in both mazes. As in the standard EPM, normalized firing rates were inversely correlated between aversive (bright or open) and safe (dark or closed) arms in each maze (r= −0.51, p<0.001 for closed/open and r= −0.55, p<0.001 for dark/bright correlations; ), demonstrating that under these conditions, mPFC neurons continue to represent the task-related features of the mazes. Crucially, firing rates in the aversive (open and dark) arms in the closed/open maze correlated with rates in the aversive (closed and bright) arms in the dark/bright maze (r=0.21, p<0.05; ), even though completely different stimuli were used to induce aversion. The positive correlation between firing rates on arms made aversive through the use of different anxiogenic cues argue strongly that that mPFC single units represent the anxiety-related features of the maze, rather than appearance or configuration of the arms.
Anxiety-related firing patterns are associated with vHPC input
The above results suggest that the mPFC may encode aspects of the environment related to anxiety. We reasoned that since the vHPC and mPFC are required for and synchronize during anxiety (Adhikari et al., 2010b
), mPFC single units with more robust anxiety-related firing patterns might be more strongly influenced by vHPC activity. Indeed, EPM scores were higher in units significantly phase-locked to vHPC theta (Rayleigh’s p<0.05) compared to other units (, mean score = 0.31 ± 0.07 and 0.17 ± 0.04, for phase-locked and other units, respectively, p<0.05, n=69 units). Importantly, this result is not due to differences in firing rates, as EPM scores and phase-locking to vHPC theta were correlated, even when phase-locking values were calculated on a subsample of 100 spikes from each unit (r=+0.25, p<0.03; Figure S2
). These results demonstrate that cells that receive vHPC input have stronger anxiety-related firing patterns. Consistent with previous results (Adhikari et al., 2010b
), this effect was specific for the theta-frequency range, as EPM scores did not differ with phase-locking to vHPC delta- (1–4 Hz) or gamma-frequency (30–80 Hz) oscillations (data not shown). Furthermore, phase locking of mPFC single units to dHPC theta oscillations was not related to EPM scores (), in agreement with lesion (Kjelstrup et al., 2002
) and physiology (Adhikari et al., 2010b
) studies suggesting that the dHPC is not required for normal anxiety-related behavior in the EPM.
mPFC units phase-locked to vHPC theta oscillations have stronger task-related firing
The above results suggest that mPFC single units with robust anxiety-related firing patterns are preferentially recruited into a circuit involving the vHPC. The projection from the vHPC to the mPFC is unidirectional (Parent et al., 2009
; Verwer et al., 1997
), and hippocampal theta-range activity has been shown to lead the mPFC (Adhikari et al., 2010a
; Siapas et al., 2005
; Sigurdsson et al., 2010a
). We reasoned that if the vHPC input plays a role in the generation of anxiety-related firing patterns, mPFC single units that follow vHPC theta should have stronger paradigm-related firing patterns compared to units that do not. To find which cells follow hippocampal theta activity, MRL values were calculated after shifting the spike train of each mPFC single unit in time, relative to the vHPC theta-filtered LFP (see Experimental Procedures). Consistent with the known anatomy and previous results, the overall mean lag for maximal phase-locking was negative, indicating that on average, mPFC unit activity followed vHPC activity (mean lag = −13.8 ± 8.1 ms). However, units with positive lags relative to hippocampal theta were also found, similarly to previous reports (Adhikari et al., 2010b
; Siapas et al., 2005
; Sigurdsson et al., 2010a
). Positive lag units may result from chance, or may be involved in polysynaptic modulating of hippocampal activity. Consistent with our prediction, cells that followed the vHPC had significantly higher EPM scores than other units (, mean score = 0.24 ± 0.047 and 0.07 ± 0.05 for units that follow vHPC theta and other units, p<0.05, Wilcoxon’s test), consistent with the notion that information from the vHPC plays a role in generating anxiety-related firing patterns. As expected, there was no difference in EPM scores comparing units that followed dHPC to those that did not ().
mPFC units that follow vHPC theta oscillations have more robust task-related firing patterns
mPFC single unit activity is correlated with behavioral display of anxiety
mPFC single units appear to differentiate between safe and aversive locations in the EPM. However, it is unclear whether this feature of mPFC activity is related to behavioral measures of anxiety in the EPM. In order to investigate this hypothesis, the mean EPM score for each animal was calculated for all mice with at least three simultaneously recorded single units in the EPM. Mean EPM scores per animal were significantly positively correlated with open arm exploration (r=+0.65, ). Thus, in animals that display behavioral avoidance of the open arms (dark grey points in ), mPFC single units show less differentiation between open and closed arms.
EPM scores and mPFC single unit activity are correlated with anxiety-related behavioral measures in the EPM
To strengthen this association of EPM scores with anxiety-like behavior, we calculated EPM scores in serotonin 1A receptor knockout (5-HT1AR KO) mice. 5-HT1AR KO mice have a robust phenotype of increased anxiety, as well as increased strength of vHPC and mPFC theta oscillations, when exposed to the EPM (Gross et al., 2002
; Klemenhagen et al., 2006
; Ramboz et al., 1998
; Adhikari et al., 2010
). In agreement with the unexpected result that lower EPM scores are associated with higher avoidance, 5-HT1AR KO mice had lower EPM scores than WT mice (). Indeed, the distributions of EPM scores of avoidant WT mice (those that spent <50% time in the open arms) and 5-HT1AR KO mice were not significantly different from the chance distribution of EPM scores generated after randomly shuffling spike location (Wilcoxon’s test, p<0.79), suggesting that these mice fail to form appropriate representations of the EPM in the mPFC. This result is consistent with the notion that the failure to represent the EPM is related to anxiety.
Why would mice that avoid the aversive arms fail to develop mPFC representations of aversiveness? One clue comes from overall firing rates. Mean absolute firing rates in the EPM tended to be higher in avoidant WT and 5-HT1AR KO mice compared to WT mice that failed to avoid the open arms (mean+/− s.e.m. firing rate= 2.8 ± 0.58 and 2.94 ± 0.80 Hz for avoidant WT and 5-HT1AR KO mice, respectively, compared to 1.57 ± 0.3 Hz for non-avoidant WT mice). There were no significant differences in the firing rates between these groups in recordings obtained in a control, non-anxiogenic familiar environment. Thus, the elevated firing rates in the EPM of avoidant mice are a consequence of greater increases in rate relative to the familiar environment (). These increases are significant only in avoidant WT and 5-HT1AR KO mice (Wilcoxon’s test, p<0.05). This result suggests the intriguing possibility that aversion-preferring mPFC units in these animals generalize across open and closed arms of the maze, raising overall firing rates and signaling anxiety regardless of maze location.