Ibotenic acid, but not vehicle, targeted to mPFC caused lesions that included the cingulate cortex, prelimbic and infralimbic areas (Figure 1B; Figure 1—figure supplement 1 and 2). To assess the effect of mPFC lesion on behavior, we first examined locomotion during the pretraining sessions (Figure 1C). There is no effect of lesion on the total distance walked across the session (Figure 1D; F_1,16 = 0.04, p=0.85, η²=0.002). There is a main effect of trial (F_1,16 = 15.31, p=0.001, η²=0.09) but there is no effect of the group × trial interaction (F_1,16 = 0.14, p=0.71, η²=10^–3). Similarly, locomotion does not differ between the groups or across trials, nor is there a significant group × trial interaction during initial training on days 2–3 (df = 3.17/50.77) or during conflict training on days 4–5 (df = 4.97/79.58; F’_S ≤ 1.96, p≥0.09, η²≤0.03).

We then examined the acquisition of active place avoidance memory during initial training using the time to first enter the shock zone, the clearest estimate of between-trial memory. Both groups learn to increase the latency to first enter the shock zone (Figure 1E). There is no effect of group (F_1,16 = 0.1, p=0.76, η²=0.001) but clear effects of day (F_1,16 = 69.61, p=10^–7, η²=0.17) and trial (F_{5.33, 85.33} = 12.15, p=10^–9, η²=0.12). There are also no significant interactions (group × day: F_1,16 = 0.94; p=0.35, η²=10^–3; group × trial: F_{5.33, 85.33} = 1.83, p=0.11, η²=0.02; day × trial: F_{4.78, 76.43} = 1.93, p=0.10, η²=0.03; group x day × trial: F_{4.78, 76.43} = 0.26, p=0.93, η²=10^–3). Neither was there a difference in avoidance learning between the groups measured by the number of entrances into the shock zone (Figure 1F). The effect of group is not significant (F_1,16 = 0.002, p=0.96, η²=10^–5) but the effects of day (F_1,16 = 26.34, p=10^–3, η²=0.19) and trials (F_{3.22, 51.50} = 36.76, p=10^–8, η²=0.1) are significant. There are no significant interactions (group × day: F_1,16 = 0.003, p=0.96, η²=10^–5; group × trial: F_{3.22, 11.50} = 0.74, p=0.54, η²=10^–3; group × day × trial: F_{4.05, 64.81} = 0.35, p=0.74, η²=10^–3). But the interaction between day and trial is significant (day × trial: F_{2.26, 36.08} = 13.75, p=10^–5, η²=0.11). The two groups are also indistinguishable on day 4 during which retention of 24 h memory is assessed by their times to first enter the shock zone (t₁₆=1.01, p=0.3, d=0.49) and the number of entrances (t₁₆=0.95, p=0.4, d=0.002).

To increase the cognitive challenge, we were motivated by evidence that mPFC manipulation can affect cognitive flexibility as assessed by reversal learning tasks (Ragozzino et al., 1999; McDonald et al., 2007). Accordingly, we trained the rats to avoid the shock zone after relocating it 180° to assess the impact of mPFC lesion on cognitive flexibility with the additional challenge to distinguish between the current and previously learned location of shock, under the cognitive control challenge (Rich and Shapiro, 2007). Both groups learn the conflict task variant, measured by the number of entrances. There is no effect of group (F_1,16 = 0.21, p=0.65, η²=10^–3), but the effects of day (F_1,16 = 8.99, p=0.01, η²=0.05) and trials (F_{1.89, 30.12} = 18.55, p=10^–9, η²=0.28) are significant. The group interactions are not significant (group × day: F_1,16 = 0.55, p=0.47, η²=10^–3; group × trial: F_{1.88, 30.12} = 1.38, p=0.27, η²=0.02; day × trial: F_{2.30, 36.87} = 2.13, p=0.11, η²=0.02, group × day × trial: F_{2.30, 36.87} = 1.16, p=0.33, η²=10^–3). Also, the time to enter the new shock zone, which was low in both groups, consistent with poor between-day memory for the novel location of shock (Figure 1, days 4–5 conflict training). There is no effect of group (F_1,16 = 0.21, p=0.65, η²=10^–3), but the effects of day (F_1,16 = 20.46, p=0.01, η²=0.04) and trials (F_{1.89, 30.12} = 11.52, p=10^-3, η²=0.13) are significant. The group interactions are not significant (group × day: F_1,16 = 0.03, p=0.86, η²=10^–3; group × trial: F_{1.88, 30.12} = 0.78, p=0.54, η²=0.01; group × day × trial: F_{2.30, 36.87} = 0.47, p=0.76, η²=10^–2). Also, the interaction between day and trial is not significant (day × trial: F_{2.30, 36.87} = 1.29, p=0.28 η²=0.02). We more closely examined conflict learning by analyzing the initial experience of the new shock zone location when the rats are first confronted with the change in the location of shock. We compared the number of entrances during the first 5 min on the first trial of each day to estimate savings, measured as each animal’s difference in performance during initial training (Figure 1F, left inset) and during conflict training (Figure 1F middle inset). During initial training, both sham (paired t₇=4.5, p=0.003, d=1.2) and lesion (paired t₉=4.7, p=0.001, d=1.5) rats improved. During conflict training, sham rats improved across the pair of first conflict trials of each day (Figure 1F, middle inset; paired t₇=3.8, p=0.006, d=1.4) but not lesion rats (paired t₉=0.87, p=0.4, d=0.27). Indeed, while all sham rats improved across the first two conflict trials of each day only 6/10 lesion rats improved (test of proportions z=2.3, p=0.03). We also quantified the improvement for each subject during conflict as a savings index (Figure 1F, right inset). The sham rats improved, avoiding more than the lesion rats in the first 5 min of the second conflict day compared to the first conflict day (t₁₆=2.4, one-tailed p=0.03, d=0.85). This difference on the second day of conflict training was observed despite equivalent performance across eight trials on the first day. We therefore find no evidence that mPFC lesion causes a deficit in active place avoidance learning to avoid the initial location of shock, nor do we observe a deficit in memory retention up to 24 h for the initial shock location, both of which require cognitive control. However, the lesion is sufficient to impair conflict learning after a 24 h delay.

CO, a sensitive metabolic marker for neuronal function (Wong-Riley, 1989), was used to evaluate whether lesion effects were restricted to the mPFC. In a subset of rats (lesion n=8, sham n=8), CO activity was evaluated in 14 functionally-related brain regions (Figure 2A). CO activity in the central amygdala, but not elsewhere, is altered by the lesion, with activity increased in lesion rats (Supplementary file 1).

Pair-wise interregional correlations of CO activity were evaluated to assess whether extra-mPFC functional relationships are altered after mPFC lesion. After FDR correction (0.01), 15 of the 91 correlations are significant in the sham sample but only 7 in the lesion sample (Figure 2B, Figure 2—figure supplement 1B; test of proportions: z=2.26, p=0.03). The correlations within the ventral hippocampus are preserved after mPFC lesion, but some correlations are lost within the dorsal hippocampal formation (Figure 2B). Correlations between the nucleus reuniens and both the dorsal and ventral hippocampus are also lost after mPFC lesion. Correlations between the basolateral amygdala and nucleus reuniens and between the basomedial amygdala and ventral hippocampus are lost after mPFC lesion, and a correlation between dorsal CA1 and the central nucleus of the amygdala appears after mPFC lesion. These data indicate that the mPFC lesion causes functional changes beyond mPFC.

When the discharge of principal cells of the hippocampus is location-specific to so-called place fields, the cells are called place cells. Place fields are disrupted by molecular manipulations that disrupt spatial learning and memory and changes in place cell firing tend to track environmental changes like geometry and other spatial features that often accompany spatial learning and memory challenges (review Jeffery and Hayman, 2004). In contrast, spatial learning and memory tasks that depend crucially on hippocampus and hippocampal synaptic plasticity, like active place avoidance or task changes, such as those introduced by the conflict task variant, cannot be assessed by simple changes in place fields and their quality (Chung et al., 2021; van Dijk and Fenton, 2018; Talbot et al., 2018; Fenton, 2024; Jeffery et al., 2003; Levy et al., 2019; Pavlowsky et al., 2017; Park et al., 2015; Pastalkova et al., 2006). This is in part because dynamic cognitive variables, like adjustments of attention and cognitive control, operate during learning and memory tasks. The dynamics of such internal cognitive variables modulate place cell firing, creating extra-positional signals (Chung et al., 2021; Kelemen and Fenton, 2010; Talbot et al., 2018; Kelemen and Fenton, 2013; Fenton et al., 2010; Kelemen and Fenton, 2016). The presence of these extra-positional signals can be detected as momentary discharge deviations from the expectations of firing-field-based spatial discharge (Johnson et al., 2009). On the timescale of the few seconds it takes to walk across a firing field, these noise-like deviations can be measured as overdispersion (Olypher et al., 2002; Olypher et al., 2003). We can also measure the momentary positional information in the collective neuronal population discharge dynamics, which during active place avoidance alternates between encoding locations in the room and locations in the arena (Kelemen and Fenton, 2010; Olypher et al., 2003). The hippocampal and entorhinal cortex neuronal populations, each fluctuate between representing space in these two distinct spatial frames every few seconds (Park et al., 2019; Fenton, 2024; Levy et al., 2023). These representational dynamics are purposeful and can be summarized as a spatial frame ensemble preference (SFEP) that is biased to room-frame locations near the room-defined shock zone and is biased to arena-frame locations far from the shock zone (Chung et al., 2021; van Dijk and Fenton, 2018).

In light of these cognitive dynamics in neuronal activity representations of position and the ability to decode task information from neuronal population discharge, it is powerful to characterize cognitive control by investigating neuronal population discharge. Cognitive control requires the subject to process and use a class of information purposefully at the expense of other competing information. Accordingly, we previously demonstrated such a cognitive control signal in the population dynamics of dorsal hippocampus spatial discharge representations of room and arena positions as rats and mice navigate on a rotating arena (Chung et al., 2021; van Dijk and Fenton, 2018; Kelemen and Fenton, 2010; Kelemen and Fenton, 2016). Consequently, we examined the SFEP, the representational dynamics of hippocampus CA1 discharge on the rotating arena during pretraining, as well as retention of the conditioned place avoidance of the initial and conflicting shock zone locations (Figure 3A). Note that the physical conditions were identical in these three trials because the shock was off.

We first examined basic discharge properties of individual CA1 principal cells, which do not differ between the sham and mPFC lesion groups (Figure 3B, Supplementary file 2). Nor does mPFC lesion alter thalamic firing rates, but it reduces the likelihood of bursting activity (Supplementary file 3) complementing the CO evidence that the mPFC lesion affected activity in areas beyond the lesion sites.

We then examined overdispersion of place cell firing, which is known to be reduced by prefrontal inactivation (Hok et al., 2013). In contrast, overdispersion is expected to increase with cognitive control and other extra-positional processes that increase discharge non-stationarity (Fenton et al., 2010). We find that during pretraining overdispersion of hippocampal discharge is reduced by a factor of two in lesion compared to sham rats, consistent with an effective lesion (Figure 3C,, left). Might this also indicate reduced cognitive control in the lesion rats? Remarkably, during conditioned place avoidance the overdispersion in lesion rats increases to the level observed in sham rats, consistent with the increased demand for cognitive control (Figure 3C, middle and right). These findings confirm the effectiveness of the prefrontal lesion and also provide electrophysiological evidence consistent with intact cognitive control in the lesion rats.

We next examined the SFEP, the electrophysiological signature of cognitive control of place representations, during conditioned place avoidance. CA1 ensemble discharge alternates between preferentially representing room locations and arena locations (Figure 4A; runs tests for all recordings were significant; z’s≥12.3, p’s≤10^–33). During pretraining, SFEP is equally likely to signal the current room location or alternatively, the current arena location, in both the sham and lesion rats (Figure 4A). Importantly, during avoidance of the initial shock location, SFEP favors the arena frame because the rat avoids the room frame location of shock (Figure 4B) and SFEP is room-preferring near the shock zone and arena-preferring far from the shock zone (Figure 4C and D), as previously demonstrated (Chung et al., 2021; van Dijk and Fenton, 2018). Relocating the shock zone 180° for conflict training changes SFEP, but only in the lesion group such that it is equally probable to be room- or arena-preferring, consistent with needing to decide between avoiding the current and previous shock location throughout the environment (Figure 4B; Dvorak et al., 2018). The ability to manipulate SFEP of hippocampal representational dynamics by the presence and location of shock directly demonstrates cognitive control at the level of the hippocampus. Despite an influence of mPFC lesion on cognitive control and the non-stationarity of hippocampal representations of space in hippocampal discharge, the lesion does not impair cognitive control of place representations in the place avoidance task.

Thirty adult, male, Long–Evans rats were purchased from Charles River to arrive at New York University at approximately 40 days old. The rats were given at least 1 week to acclimate to the facility and were housed two per cage on a 12:12 light:dark cycle with free access to food and water. Fifteen rats were randomly assigned to the lesion group and 14 to the sham group. The experimenters working with the rats were blind to group identity. One animal was used to confirm the coordinates, by injecting fluorogold (Fluorochrome, Denver, CO, 2.5% w/v in distilled water) at the same rate and for the same duration as for the lesions. Three of the sham and three of the mPFC lesions rats were used for the electrophysiology recordings. The behavior of these rats was not included in the behavioral assessment due to the requirement for extended training sessions to record neuronal activity.

Between the ages of 48–56 days, the rat received PFC or sham lesion under sodium pentobarbital (50 mg/kg, i.p.) anesthesia. Bilateral lesion targeted two sites in each hemisphere at the following stereotaxic coordinates: from Bregma, Site 1: A.P +2.5, ML ± 0.6, DV - 5.0, relative to the skull surface, Site 2: AP +3.5, ML ± 0.6, DV - 5.2, relative to skull surface. Ibotenic acid (0.06 M in PBS) was used to create an excitotoxic lesion following published work (Birrell and Brown, 2000). A 0.2 µl volume was injected at each site with a micro-infusion pump at a flow rate 0.2 μl/min through a stainless-steel cannula (0.25 mm outer diameter); the cannula was left in place for 4 min (for a total of 5 min per injection site) before being slowly withdrawn. Rats in the sham group underwent the same surgery procedure, but only PBS was injected. The rats were given 1 week to recover.

A commercial rotating arena and tracking software (Tracker, Bio-Signal Group, Acton, MA) was used for the active place avoidance task, which has been described previously (Pavlowsky et al., 2017; Pastalkova et al., 2006; O’Reilly et al., 2016). Briefly, the rats were placed on a stainless steel 82 cm diameter arena that could rotate about its center at 1 rpm under computer control. The arena was in the center of a 3 m × 3 m space that was surrounded by black curtains with a distinctive orienting cue (striped fabric). The arena floor was made of parallel stainless-steel rods organized into five electrical poles. A constant current source could be triggered by the software to deliver shock that was scrambled across the five poles to ensure it was unlikely that shock could be avoided by a fortuitous posture or by feces shorting the bars. A 50 cm high transparent PETG wall ensured the rats remained on the arena and allowed them to see into the room as the arena rotated. The position of the rat in the spatial frame of the room was tracked 30 times a second at 3.2 mm resolution by analysis of the video image from an overhead camera. An infrared LED that rotated with the arena was tracked similarly from the video image, and the rat’s arena-frame position on the arena surface was computed relative to the arena LED (Fenton et al., 1998; Bures et al., 1998). The room-frame and arena-frame position time series, along with the delivery of shock, were stored in data files for off-line automated software analysis of end-point measures (TrackAnalysis, Bio-Signal Group, Acton, MA). Three end-point measures are reported. (i) The distance walked measures how much the rat walked on the arena surface. (ii) The time to first enter the shock zone increased with training; it estimates the between-session active place avoidance memory by estimating how well the rat can remember the previous room-frame locations of shock. (iii) The number of entrances into the shock zone decreased with training; it estimates place learning and can decrease due to between-session memory as well as within-session memory of the locations of shock.

After recovery from surgery, the rats were brought to the laboratory in their home cages that were placed on a rack to the side of the experimental room. The rats were each handled for 5 min per day for 5 days. Behavioral training consisted of 10 min trials with an intertrial interval of 10 min during which the rat was returned to its home cage in the experimental room. On the first day of behavior training, animals had a pretraining session that consisted of two trials during which the rats were allowed to explore the stationary arena to habituate to the environment. The following two days the rats underwent eight training trials per day on the rotating arena, learning to avoid entering a 60° sector shock zone in which the rats received a mild foot shock (500 ms, 60 Hz, 0.4 mA). This shock zone was defined by the stationary cues in the room and was the same for all rats. On the fourth day of the behavioral training, the rats had a single trial with the shock on to test retention of the training. The rats were then trained on days 4 and 5 in a conflict training session in which the rats received eight trials per day with the shock zone located 180° from the initial training position. All trials were 10 min with 10 min intertrial intervals. The total distance walked during the session measured locomotion, the time to first enter the shock zone location measured avoidance memory, and the number of entrances into the shock zone was used to measure place learning. Savings in place learning on the conflict trials when the shock zone was relocated 180° was assessed for each rat by a savings index, which compares the number of entrances during the first 5 min of the first and second days of conflict training, conflict trials 1 and 9, respectively. The difference is normalized by the number of entrances on the first conflict trial and reported as a percentage:Savings Index=(NumberofEntrancesconflict1−NumberofEntrancesconflict9)NumberofEntrancesconflict1X100\begin{document}$$\displaystyle Savings\ Index=\frac{\left (Number\, of\, Entrances_{conflict1}-Number\, of\, Entrances_{conflict9}\right)}{Number\, of\, Entrances_{conflict1}}X100$$\end{document}

The following procedures have been previously described in detail (Park et al., 2019). Briefly, rats were anesthetized (pentobarbital 50 mg/kg) and received mPFC or sham lesions as described above. The Neuropixel recording system was used (Putzeys et al., 2019). A Neuropixel electrode array was implanted in the brain to traverse Bregma coordinates AP –3.8 mm, ML 2.5 mm for recording dorsal hippocampus and the tip targeted DV 9.3 mm, which allowed recording from the underlying posterior and lateral posterior thalamus. The Neuropixel was stabilized and protected by mounting it in a custom 3-D printed appliance. The appliance and five bone screws were cemented to the skull (Unifast, GC America Inc, IL). The rats were allowed at least 1 week to recover before further manipulation.

The rats received active place avoidance training as described above. In addition, to permit neural ensemble recordings, after the last pretraining, training, and conflict training sessions, shock was turned off and the session was extended for 10–20 min during which CA1 neural ensemble discharge was recorded using SpikeGLX. Signals were filtered between 300 Hz and 10 kHz and sampled at 30 kHz for single unit recording. Because of these extended sessions, the behavior from these animals was not included in our behavioral assessment.

Single units were automatically sorted using Kilosort 2 (Pachitariu et al., 2016). Units were only studied if the estimated contamination rate was <20% with spikes from other neurons. This was estimated from refractory period violations relative to expected. Units with non-characteristic or noisy waveforms were also excluded. We computed a ‘burst ratio’ to characterize burstiness of discharge as number of spikes with ISI ≤30 ms divided by number of spikes with 100 ms ≥ ISI ≤ 130 ms.

Hippocampus single units were classified as complex-spike or theta cells as in prior work (Kelemen and Fenton, 2010; Fenton et al., 2008; Ranck, 1973). Complex-spike cells appear to be pyramidal cells, having long-duration waveforms (>250 µs), low discharge rate (<5 AP/s), and a tendency to fire in bursts. Theta cells are likely local interneurons (Fox and Ranck, 1975) and had short-duration waveforms (<250 µs), high discharge rate (>2 AP/s), and did not tend to fire in bursts. Only hippocampal units classified as pyramidal cells were studied. Thalamic units, recorded from the posterior and lateral posterior thalamic nucleus, which lies directly beneath the hippocampus, fired at high rates and were less likely to discharge in bursts (Supplementary file 3).

The representational dynamics of CA1 discharge alternating between preferentially signaling the current location in the stationary room or the current location on the rotating arena was investigated to directly evaluate cognitive control during place avoidance behavior. Spatial-frame-specific momentary positional information (I_pos) is used to estimate the location-specific information from the activity of a cell during a brief interval (Δt=133 ms).

Ipos(t)=pi|xlog2⁡(pi|xpi),pi∣x\begin{document}$I_{\text{pos}}(t) = p_{i|x} \log_2 \left( \frac{p_{i|x}}{p_i} \right), p_{i \mid x}$\end{document} is the probability of observing i activity at location x, and pi\begin{document}$p_i$\end{document}

is the overall probability of observing i activity. I_pos(room) estimates the information about the current location x in the room, whereas I_pos(arena) separately estimates the information about current location x on the rotating arena (Olypher et al., 2003). Ipos(t)\begin{document}$I_{\text{pos}}(t)$\end{document} can be positive or negative, but the absolute value is large whenever the cell’s activity at the current location is distinct or ‘surprising’ compared to the location-independent probability of observing the same activity. The value Ipos(t)\begin{document}$\left |I_{pos}\left (t\right)\right |$\end{document} is abbreviated Ipos\begin{document}$I_{pos}$\end{document}. A unit’s spatial-frame preference is estimated at each Δt moment as the difference:

ΔIpos=Ipos(room)−Ipos(arena)\begin{document}$\Delta I_{\text{pos}} = I_{\text{pos}}(\text{room}) - I_{\text{pos}}(\text{arena}) $\end{document}, where positive values indicate a momentary preference for signaling the room location, and negative values indicate a preference for signaling the arena location. ΔIpos\begin{document}$\Delta I_{pos}$\end{document} time series were evaluated for significance compared to chance fluctuations around the mean using the z value from a runs test. The SFEP at each Δt was computed as the average ΔIpos\begin{document}$\Delta I_{pos}$\end{document} over all cells. This ensemble ΔIpos\begin{document}$\Delta I_{pos}$\end{document} time series (assessed for significance by runs test) was averaged over all times and locations to estimate an overall SFEP as the proportion of time the ensemble activity was room (or arena) preferring (Chung et al., 2021; van Dijk and Fenton, 2018). To evaluate if SFEP was purposeful, the arena was divided into approximately two halves based on where individual rats avoided during a retention test with shock off. The ‘near’ half was the largest sector that includes at least 50% of the recording time and surrounds the least visited 20° sector of the room frame (within the shock zone). The ‘far’ half is the remaining sector. The probability of observing room-preferring ΔIpos\begin{document}$\Delta I_{pos}$\end{document} was computed at each location to make a representational preference map and the average probability in the near and far halves were compared by statistical evaluation.

Place cell overdispersion measures discharge non-stationarity by estimating how much neuronal discharge deviates from what is expected if the discharge only signals position (Fenton and Muller, 1998). Increased overdispersion indicates an extra-positional signal like attention or cognitive control is modulating discharge up and down in addition to position, whereas decreased overdispersion indicates such modulation is dampened, possibly to sustained attention (Fenton et al., 2010). Overdispersion of hippocampal discharge was computed as previously described (Fenton et al., 2010). Briefly, The spike and position time series is divided into 5 s episodes and the expected firing (exp) is computed assuming an inhomogeneous Poisson process. The standardized rate is computed as the normalized difference between the observed (obs) and expected firing: z=obs-expexp\begin{document}$z=\frac{obs-exp}{\sqrt{exp}}$\end{document}. Episodes with exp =0 are undefined, and only episodes in which exp is greater than the cell’s mean firing were analyzed because this criterion selects episodes during which the rat passes through the central region of the firing field. The variance of the distribution of z values quantifies overdispersion, which can be compared between conditions using a F test.