Involvement of prelimbic cortex neurons and related circuits in the acquisition of a cooperative learning by pairs of rats

Experiments were carried out with male Lister Hooded rats (3 months old, 250–300 g at the beginning of the experiments) provided by an authorized supplier (Charles River Laboratories, Barcelona, Spain). Upon their arrival at Pablo de Olavide Animal House (Seville, Spain), animals were housed in pairs in Plexiglas® cages until the end of the experiments and were trained through all experimental phases with the same partner. Rats were randomly paired and were kept on a 12-h light/dark cycle (from 8:00 am to 8:00 pm local time) with constant ambient temperature (21.5 ± 1 °C) and humidity (55 ± 8%). Unless otherwise indicated, animals had food and water available ad libitum. All the experiments were carried out following the guidelines of the European Union Council (2010/63/EU) and Spanish regulations (BOE 34/11370–421, 2013) for the use of laboratory animals in chronic experiments. Experiments were also approved by the local Ethics Committee (06/03/2018/025) of Pablo de Olavide University.

Cooperation experiments were carried out in a double Skinner box customized by our team (Fig. 1A, B) and consisting of two adjacent Skinner modules, each measuring 29.2 × 24.1 × 21 cm (MED Associates, St. Albans, VT, USA), separated by a grille partition that allowed animals to see, hear, and smell each other and to have partial physical contact (see Conde-Moro et al. 2019). Each box was equipped with a green platform (4.5 cm in height × 7 cm in width) with five infrared beams that detected when the rat was on it, a LED light, and a food dispenser where food pellets (Noyes formula P; 45 mg; Sandown Scientific, Hampton, Middlesex, UK) were delivered after the rat climbed onto the platform.

Before experiments, and in other to increase the interest of the experimental animals for the selected instrumental task, rats were handled daily for 10 days and food-deprived until they reached an 85% of their free-feeding weight. Once the selected weight was reached, rats were habituated to the test room for two 10-min free-exploration sessions in an empty Plexiglas® box, different from the experimental box. Between use for each pair of animals, apparatus and cages were cleaned with 5% ethanol and dried with paper.

During phase I, pairs of animals were placed in the adjacent Skinner boxes for 20 min and were free to explore the cage (Fig. 1C). Each time an animal stayed on the platform for > 500 ms, a pellet of food was delivered to the feeder, following a fixed-ratio schedule (FR 1:1) until the established criterion was reached—that is, to climb onto the platform ≥ 60 times/session for two consecutive days. A LED light located above the feeder indicated when a pellet of food was delivered.

For phase II (simultaneous cooperative task), within the same set-up, animals had to climb and stay on their platforms simultaneously for > 500 ms to obtain a mutual reward (Fig. 1D). If either of the animals climbed onto the platform and went to the feeder on its own, the trial was considered wrong (for acting individually) and there would be no reward for either of them. Animals were trained daily until reaching the established criterion—that is, to climb simultaneously onto the platform to get the mutual reward ≥ 40 times/session for at least two consecutive days. Conditioning programs, LFP recordings, platform climbs, and delivered reinforcements were monitored and recorded. All room lights, except for dim light, were switched off before each experimental session to improve the rats’ comfort.

Two cohorts of rats participated in this study. The objective with the first cohort, consisting of 40 animals (38 at the end of the experiment that carried out all tests), was to study the effects of the cooperation experiment and the different roles developed by each rat (leader or follower) in the main cell types of the PrL cortex, using immuno-colocalization analysis. The objective with the second cohort, consisting of 18 animals (10 at the end of the experiment with good recordings), was to study the effect of cooperation not only in the PrL but also in subcortical projection areas (NAc and BLA), using in vivo recording and analysis of the LFPs collected in the aforementioned areas.

To prepare animals for the in vivo electrophysiological experiment, rats were anesthetized with 1–2.5% isoflurane delivered by a rat anesthesia mask (David Kopf Instruments, Tujunga, CA, USA). Isoflurane was supplied from a calibrated Fluotec 5 (Fluotec-Olmeda, Tewksbury, MA, USA) vaporizer, at a flow rate of 1–3 L/min oxygen (AstraZeneca, Madrid, Spain).

For LFP recordings, and following the Paxinos and Watson atlas (2007), animals were chronically implanted with two sets of recording electrodes aimed at the right PrL cortex (3.24 mm anterior, 0.5 mm lateral to bregma, and 2.5 mm from brain surface) and the right NAc Core (2.0 mm anterior and 1.5 mm lateral to bregma, and 6.5 mm from brain surface) and one set of recording electrodes aimed at the BLA (2.28 mm posterior and 5 mm lateral to bregma, and 7.5 mm from brain surface).

All electrodes were handmade from 50 μm, Teflon-coated, tungsten wire (Advent Research, Eynsham, UK). Each electrode set consisted of two tungsten wires with a separation between tips of ≈ 0.3 mm. The Teflon coating was removed from the first 200 μm of each cable tip for better wire surface exposure. A bare silver wire was affixed to the bone as ground. All the implanted wires were soldered to sockets (RS Amidata) that were fixed to the skull with six small bone anchor screws (Stoelting Co., Woodale, IL, USA) and dental cement.

Brain collection and preparation for immunofluorescence. To collect the brain tissue at key moments of the cooperation acquisition, three groups started the cooperation experiment and rats from each group were sacrificed at a different point of the experiment (Fig. 2A).

Group 1 was the individual group, which performed the protocol until reaching the criterion for the individual phase (i.e., getting ≥ 60 pellets of food for two consecutive days). Group 2 was the cooperation group, which performed all phases of the protocol until they reached the criterion for the cooperation phase (i.e., ≥ 40 pellets for two consecutive days), and Group 3 completed the whole experimental protocol (i.e., they completed the 10 sessions of the cooperation phase even after reaching criterion). The control group, which underwent the same protocols as the remaining rats but did not perform the cooperation experiments, were sacrificed in a scattered way, distributed along with the perfusion of the other groups (i.e., 3 rats perfused along with group 1, 3 rats with group 2, and 4 rats with group 3, always counterbalancing the order with those from experimental groups). Rats from the control group were handled 2 min per day on the same days that the other animals were performing experiments and were perfused 100 min after the first opening of the cage for handling. The order in which each pair started the experiments every day was counterbalanced.

As explained before, rats performed the cooperation experiment until they reached the criterion for their group. The day after reaching criterion, animals performed a reminder session of 15 min, and 85 min later were anesthetized (i.p.) with pentobarbital (150 mg/kg) and sacrificed by transcardial perfusion using 0.9% saline solution.

After perfusion, rats’ brains were carefully removed and kept in PFA 4% at 4 ºC for one day and put into a sucrose solution (30%) at 4 ºC for 3 days or until brains sank in the solution. When brains were ready, they were frozen with liquid nitrogen and stored at -80 ºC. Afterward, brain areas of interest were serially cut at the cryostat (Leica, CM3050 S) in coronal Sects. (30 µm thick) and kept in a cryoprotectant solution (30% glycerol, 30% ethylene glycol, and 0.2 M phosphate buffer) at − 20 ºC.

Free-floating sections were labeled for c-FOS, DAPI, and one of the antibodies against the target cell types of the PrL cortex. Sections were washed three times in 0.1 M phosphate buffer and blocked with 0.1 M phosphate buffer containing 10% Triton (VWR, M143-1L) and 5% normal donkey serum (Merck Millipore S30-100 mL). After that, sections were incubated for 40 h at 4 ºC with rabbit anti-c-FOS (Synaptic Systems, 226,003) or goat anti-c-FOS (Santa Cruz, sc-52-G) and one of the following primary antibodies: mouse anti-GAD67 for staining GABAergic cells (Merck & Co., MAB5406), goat anti-Substance P for staining D1-containing cells (Santa Cruz, sc-9758), and rabbit anti-Met Enkephalin for staining D2-containing cells, (Abcam, ab22620). The sections were then incubated for 2 h in one of the secondary antibodies: donkey anti-mouse Alexa 647 (Thermo Fisher Scientific, A-31571) for GAD67, donkey anti-goat Alexa 488 (Thermo Fisher Scientific, A-11055) for Substance P, donkey anti-rabbit Alexa 568 (Thermo Fisher Scientific, A-10042) for Met-Enkephalin, and donkey anti-rabbit Alexa 568 (Thermo Fisher Scientific, A-10042) or donkey anti-goat Alexa 488 (Thermo Fisher Scientific, A-11055) for c-FOS. After washing in 0.1 M phosphate buffer, the sections were incubated for 10 min in DAPI (Sigma Aldrich), rinsed, and mounted with Fluoromount-G (Southern Biotech). Images were taken with a confocal microscope (Zeiss, LSM700) using a 20 × objective and were captured at the same coordinates for each animal (see Hollis et al. 2015).

For immunofluorescence quantification, we used the imageJ/FIJI software. We calculated each channel background by measuring four random background areas and calculating the mean, which was subtracted from each channel. Cells were delineated with a Huang threshold to label those cells stained with DAPI within 50–200 pixels. Once we had this number of cells, we counted how many of them were also labeled with c-FOS and the antibody of interest, and converted it to a percentage.

To find out behavioral traits that could predict which rats from the groups that did not complete the cooperative phase of the experiment (control group and group 1) would develop a leader or follower role in the future (denominated here as predicted leaders and predicted followers), all animals of the immunofluorescence experiment, including the control group, underwent a battery of tests to measure their basal levels of anxiety, locomotor behavior, and social dominance before the cooperation experiment.

Water and Food competition tests (WCT, FCT). To measure social dominance, we used the water and food competition tests, in which animals compete for resources after a time of deprivation (Cordero and Sandi 2007). The water competition test (WCT) is used to assess social dominance between rats. The experiments took place in the housing cages where rats lived in cohabitation for at least 1 week before the test. Animals were water-deprived for 6 h before the test and marked on the back for identification. A single bottle of water was placed in an accessible part of the cage at the beginning of the test, and the rats’ behavior recorded for 10 min after the bottle is introduced.

For the food competition test (FCT), which also took place in the home cages, the protocol was similar to the water competition test, but instead of water, we placed 10 pellets of palatable food (Noyes formula P; 45 mg; Sandown Scientific, Hampton, Middlesex, UK) in the middle of the cage. Rats were food-deprived for 12 h before the test.

The WCT and FCT sessions were also recorded and social behaviors such as aggression or displacements from the feeder or water bottle were computed. Social dominance was determined by the summation of the total duration of water consumption and the number of food pellets eaten by each rat in the pair.

Data for quantification of animal performance in the Skinner boxes were selected and analyzed offline using Spike 2 software (Cambridge Electronics Design) and statistically analyzed afterwards using Sigma Plot 11. LFP power spectra, spectrograms, coherence spectra, and coherograms from different groups and conditions were statistically compared using Chronux customized scripts (Mitra and Bokil 2008; Bokil et al. 2010) to obtain the jackknife estimates of the variance and of Z-statistics (for details refer to Bokil et al. 2007).

LFP spectral decomposition and representation. The computational tools used for neurophysiological signal processing and analysis were customized MATLAB scripts (version 9.4, R2018a. The MathWorks, Natick, MA, USA) based on Chronux, (version 2.12, 2016. Website: http://​chronux.​org/​) by Mitra and Bokil (2008). Chronux is a spectral analysis toolbox for MATLAB specialized in the processing of brain signals, which has been validated and widely used for experimental procedures, including LFPs.

Analyses in the frequency domain were carried out according to the following frequency bands: delta (3–6 Hz), low theta (6–9 Hz), high theta (9–12 Hz), beta (12–32 Hz), and gamma (32–100 Hz). A high-pass filter was applied to remove low-frequency (0–2 Hz) movement artifacts. A band-pass filter (0–200 Hz) was also applied to remove artifacts produced by the animal’s chewing.

Analyses in the frequency and time–frequency domains and coherence described in this section were computed following the multitaper spectral estimation method developed by Thomson (1982) and implemented in the Chronux toolbox. The spectral power estimates the magnitude of Fourier transforms in the frequency domain, highlighting the frequencies with higher energy (power) within the signals. Spectral power was computed for the epochs selected for each phase and condition. Multitaper spectrograms were also computed for all phases and conditions to represent Fourier coefficients in the time–frequency domains, which allowed us to inspect the moment at which significant changes in power took place. Spectrograms were computed using a moving time window (T) with a length of 500 ms (shifted in 10-ms increments) and bandwidth (W) of 6 Hz, resulting in a bandwidth product of 3 and K = 5 tapers. These parameters verify the number of selected tapers (i.e., K = 2 × T × W – 1 taper or windowing functions). The multitaper estimates of the spectrum with NT trials and K tapers were based on computing NT × K Fourier transforms that determined an appropriate number of degrees of freedom; dof = 2 × NT × K for all the computations (see details in Conde-Moro et al. 2019).

To study the functional connectivity between the LFPs recorded from different brain structures, we computed the LFP-LFP phase coherence and coherograms, revealing the levels of oscillatory synchrony between electrodes from different structures at certain frequencies.

Before the cooperation experiment, the four groups of rats (n = 38) underwent a battery of validated behavioral tests in an attempt to assess behavioral traits, such as anxiety (EPM test), locomotor activity (OFT), and social dominance (WCT and FCT).

Although leader rats, in general, spent more time in the center area of the arena than followers, they presented a high variability, and no significant differences between leaders and followers were found in the percentage of time spent in the center—and more anxiogenic—area of the OFT (Supplementary Fig. 2A, One-way ANOVA, F = 2.58, p = 0.12). No significant differences were found either in the total distance traveled by leaders and followers during this test (Supplementary Fig. 2B, One-way ANOVA, F = 1.95, p = 1.81). After 10 min in the open field arena, a novel object was inserted in the center for 5 min. Again, there were no significant differences between the percentage of time that leader and follower rats spent in the object zone (Supplementary Fig. 2C, One-way ANOVA, F = 1.10, p = 0.30).

However, leader rats spent significantly more time in the open arms of the elevated plus-maze (Fig. 3D, One-way ANOVA, F = 6.71, p = 0.02), showing lower levels of anxiety. No significant differences were found in the latency to enter the open arms (Fig. 3E, One-way ANOVA on ranks, H = 0.24, p = 0.62). According to the scale validated by Sandi et al. (2008), rats can be classified by three different levels of anxiety based on the percentage of time spent in the open arm of the EPM: low-anxious (LA, more than 20% of the time in the open arms), intermediate-anxious (IA, between 5 and 20% in the open arms), and high-anxious (HA, less than 5% in the open arms). Most leader rats in our study were classified as low-anxious (Fig. 3F, LA = 66%, IA = 33%); no rat in the leader group was classified as high-anxious. In the follower group, most of the rats were classified as intermediate-anxious (Fig. 3F, LA = 33%, IA = 61%, HA = 5%). These results indicate that leader rats were in general less anxious than follower rats.

For this experiment, we used WCTs and FCTs, in which animals compete for resources (water or food) after a time of deprivation. To measure social dominance, we calculated the percentage of time that rats spent drinking water during the water competition test and the percentage of food intake during the food competition test. Leader and follower rats spent a similar percentage of time drinking water in the water competition test (Fig. 3G, One-way ANOVA, F = 0.18, p = 0.67), while follower rats ate a significantly higher percentage of food pellets during the food competition test (Fig. 3H, One-way ANOVA, F = 8.69, p = 0.01).

To have a deeper insight of the hierarchical dynamics taking place, we summed the number of successful displacements and aggressions to the social dominance index based on the work of Costa, et al., (2021). We found that follower rats showed a significantly higher level of social dominance (Fig. 3I, One-way ANOVA, F = 5.29, p = 0.03) than leader rats during these tests.

To determine the level of activation of putative neuronal groups involved in cooperation behaviors (D1- and D2-receptor containing cells and GABAergic cells), 3 groups of rats participated in the cooperation experiment; they were sacrificed at key moments and their brain tissue was collected. Group 1 participated in the protocol until reaching criterion for the individual phase, group 2 participated in the protocol until reaching criterion for the cooperation phase, and group 3 completed the whole protocol (10 sessions of individual training and 10 of cooperative training). A control group participated in the previous behavioral tests (i.e., open field, elevated plus) but not in the cooperation experiment (Fig. 2A).

Analysis of the percentage of c-FOS expression in PrL cortex dopaminergic receptor expressing cells revealed a significantly higher activation of D1-containing cells of leader rats from group 2 during the cooperation phase (Fig. 4A, B, One-way ANOVA, F = 16.95, p = 0.007), while non-significant differences were found between leaders and followers from group 3 (Fig. 4C, D, One-way ANOVA, F = 5.33, p = 0.08), although the activation level of D1-containing cells was higher for leader rats.

The levels of c-FOS expression in D2-containing cells were lower than those in D1-containing cells (Fig. 4A, B). The highest levels of activation of D2-containing cells were also observed for leader rats during the cooperation phase (Fig. 4A, B), although the difference between leader and follower rats was not significant (Fig. 4B, One-way ANOVA, F = 1.63, p = 0.24; D, One-way ANOVA, F = 1.44, p = 0.29).

During the individual phase, and contrarily to what happened during cooperation, the rats predicted to be followers (predicted followers) presented higher percentages of c-FOS expression than the ones predicted to be leaders (predicted leaders), but this difference was non-significant (Supplementary Fig. 3C, D, One-way ANOVA, F = 0.68, p = 0.42). The same happened in the control group (Supplementary Fig. 3A, B, One-way ANOVA on ranks, H = 0.02, p = 0.93). Non-significant differences were found between the level of activation of D2 cells for predicted leaders and predicted followers during the individual phase (Supplementary Fig. 3D, One-way ANOVA on ranks, H = 2.41, p = 0.11) and in the control group (Supplementary Fig. 3B, One-way ANOVA, F = 1.85, p = 0.20).

Regarding the percentage of c-FOS activation in GABAergic cells during the cooperative phase, although leader rats presented higher levels of activation than followers, non-significant differences were found in the levels of c-FOS expression for GABAergic cells between leaders and followers of groups 2 (Supplementary Fig. 4A, B, One-way ANOVA, F = 2.50, p = 0.12) and 3 (Supplementary Fig. 4C, D, One-way ANOVA, F = 1.62, p = 0.22).

In GABAergic cells, the levels of c-FOS expression were higher in predicted follower rats during the individual phase and in the control group, but, again, the differences between predicted leaders and predicted followers were not significant (Supplementary Fig. 5A, B, One-way ANOVA, F = 0.49, p = 0.49. Supplementary Fig. 5C, D, One-way ANOVA, F = 2.15, p = 0.16).

As described in Methods, during the cooperation experiment we analyzed LFPs recorded from three different brain regions: the PrL cortex, the NAc, and the BLA. A total of 10 rats were successfully trained pair-wise in the adjacent Skinner boxes for the individual phase (phase I) and the cooperation phase (phase II).

During phase I, rats were trained to climb independently—regardless of their partner’s behavior—onto the platform to obtain a pellet of food at a fixed 1:1 ratio. As illustrated in the acquisition curve (Fig. 5A), rats improved their performance across sessions. Compared with session 1, the average number of correct responses increased significantly from session 4 on (RM-One-way ANOVA, multiple comparisons: S04, p < 0.05; S05, S06, and S07, p < 0.01), showing an even higher increase in the last three sessions (S08, S09, and S10, p < 0.001).

During phase II, rats were trained to climb onto their respective platforms and stay on them simultaneously for at least 500 ms to mutually get a reward (a food pellet for each rat). As shown in the acquisition curve in Fig. 5B, rats learned to climb simultaneously onto the platform for the mutual reward, and the number of cooperation trials also increased significantly in session 4 (RM-One-way ANOVA, multiple comparisons, S04, p < 0.01), and from sessions 7 to 10 (RM-One-way ANOVA, multiple comparisons, S07, S08, S09, and S10, p < 0.001).

To examine the oscillatory activity occurring in these brain areas during the two phases, 2-s epochs were selected from the continuous recordings obtained during two different situations: “BEFORE-platform”, comprising the 2 s before the animals climbed onto the platform, and “ON-platform”, which refers to the 2 s after the animal climbed on the platform.

Figure 5C–E shows a visualization of normalized and averaged spectral-power results for all frequency bands during the individual and cooperative phase. In general, when rats were cooperating on the platform the PrL cortex presented higher power in delta, low theta, and high theta bands. These bands also showed increases in power when rats were individually on the platform, but to a lower degree (Fig. 5C). The highest powers for the NAc were also observed during the cooperative phase, when high theta increased before the climb onto the platform to cooperate, while delta and low theta did so after the climb onto the platform (Fig. 5D). The BLA showed less activation than the aforementioned areas, having the maximum activity when rats were individually on the platform, especially in the delta band (Fig. 5E).

To know more about the functional connectivity between the brain areas of interest, we calculated the LFP-LFP coherence in the frequency domain between electrodes from the three areas of interest: PrL-NAc, PrL-BLA, and NAc-BLA at different moments (BEFORE- and ON-platform) and for both phases (individual and cooperative). The epochs analyzed were the same as in the previous spectral power analysis (2-s epochs, NT = 70 per condition).

Figure 5F–H shows a visualization of averaged LFP-LFP coherence results for all frequency bands and phases. In general, the connectivity between the PrL and NAc showed the highest coherence value in the theta band before the rats climbed onto the platform individually, followed by an increase in delta band before they climbed onto the platform to cooperate. During cooperation the connectivity in the delta band decreased, increasing in theta (low and high) (Fig. 5F). For the PrL-BLA, the highest coherence values were found in the low theta band when rats were individually ON-platform and in high theta when rats were cooperating ON-platform (Fig. 5G). The NAc-BLA connectivity increased in delta and theta bands before the climb to cooperate. Delta also showed high coherence when rats were individually ON-platform (Fig. 5H).

Comparisons of the spectral power (Fig. 6A–C) and spectral dynamic analysis (spectrograms, Fig. 6D–I) of LFPs recorded from the two conditions—BEFORE- and ON-platform—during the individual phase revealed significantly higher power (jackknife estimates of the variance, p < 0.05) in the three selected brain areas before the animals climbed individually onto the platform as compared with the 2 s on the platform. In the PrL cortex, a significant increase in power was observed in the bandwidth in the range 8–11 Hz (Fig. 6A; jackknife estimates of the variance, p < 0.05). The dynamic analysis represented in Fig. 6D indicates that the differences were particularly of note around 1 s after the rats climbed onto the platform, while the significant differences found (Fig. 6G) indicate higher activity for the 15–20 Hz band during the first second ON-platform and for the 10–20 Hz band 2 s after the climbing individually onto the platform. These differences were not detected by the comparison of average spectral power in Fig. 6A. In the NAc, the spectral power BEFORE-platform was significantly higher from 2 to 6 Hz and from 9 to 20 Hz (Fig. 6B), and the dynamic analysis confirmed these findings, revealing that the power values in the NAc were higher from 2 to 1 s BEFORE-platform. The power values in the BLA were also significantly higher BEFORE-platform for the 2–5 Hz and 11–17 Hz bands (Fig. 6C). The spectrograms in Fig. 6F, I, revealed that the power increase in the range 2–5 Hz observed in Fig. 6C was greater around 1 s BEFORE-platform, while the differences found from 11 to 17 Hz were greater 2 s and 500 ms BEFORE-platform.

Comparisons of the spectral power (Fig. 7A–C) and spectral dynamic analysis (spectrograms, Fig. 7D–I) of LFPs recorded for the two conditions—BEFORE- and ON-platform—during the cooperative phase showed a significant increase in the spectral power of PrL cortex in frequencies of 3–5 Hz, 7–10 Hz, and 14–19 Hz (jackknife estimates of the variance, p < 0.05) when rats were cooperating ON-platform compared with the seconds BEFORE-platform to cooperate (Fig. 7A). The dynamic analysis in Fig. 7D–G confirmed these findings and revealed that the spectral power was significantly higher, particularly at 0–10 Hz during the first second ON-platform and at 10–20 Hz after 500 ms ON-platform (jackknife estimates of the variance, p < 0.05). The spectral power observed in the NAc (Fig. 7B) was significantly higher BEFORE-platform to cooperate at 7–11 Hz (jackknife estimates of the variance, p < 0.05). The comparison of multitaper spectrograms in Fig. 7E–H showed that the predominant power BEFORE-platform was especially high for the band ranging between 7 and 15 Hz 2 s and from 4 to 15 Hz 1 s BEFORE-platform (jackknife estimates of the variance, p < 0.05). A significant increase in the power of the NAc at 3–4 Hz around 1.5 s after the climbing onto the platform was revealed in Fig. 7H. The spectral power of BLA was significantly higher at 3–20 Hz (Fig. 7C) and the spectrogram comparison shown in Fig. 7F–I indicated a significant decrease of power when rats were ON-platform for practically the whole time window analyzed.

After comparing the spectral powers from the structures under study, we compared the averaged spectral coherence during the two conditions (BEFORE- and ON-platform). After that, to obtain deeper knowledge about the time in which these changes in coherence were taking place for each area, we computed multitaper coherograms for each condition and compared them with the jackknife estimates of the variance method.

As shown in Fig. 8A, during the individual phase the coherence between the PrL cortex and the NAc was quite similar for the two different moments analyzed (BEFORE- and ON-platform) except for the range 8–11 Hz, where the coherence was significantly higher before rats climbed individually onto the platform (jackknife estimates of the variance, p < 0.05). The dynamic analysis illustrated in Fig. 8D–G revealed that the highest coherence magnitudes between the PrL cortex and the NAc were found BEFORE-platform, especially from 3 to 4 Hz 1.5 s BEFORE- and from 13 to 16 Hz, at around 1.5 s BEFORE-platform. The coherence between PrL cortex and the BLA was also similar for the two conditions (Fig. 8B), except around 6 Hz, which was significantly higher ON-platform (jackknife estimates of the variance, p < 0.05). The opposite happened at 10 Hz, where the coherence between the PrL cortex and the BLA increased BEFORE-platform. The coherograms in Fig. 8E–H revealed significantly higher coherence clusters BEFORE-platform, particularly 2 s BEFORE- in the range 6–9 Hz, and at 5 Hz and 15 Hz 1 s BEFORE-platform and 6 Hz to 9 Hz and 16 Hz 500 ms BEFORE-platform. The coherence between the NAc and the BLA was significantly higher when rats were individually ON-platform from 4 to 6 Hz and 7 Hz to 10 Hz (Fig. 8C, jackknife estimates of the variance, p < 0.05) and the coherograms in Fig. 8F–I confirmed these differences, indicating significantly higher coherence from 3 to 6 Hz 1 s after rats climbed individually onto the platform and from 7 to 9 Hz 500 ms after the climbing ON-platform (jackknife estimates of the variance, p < 0.05). Some small clusters were significantly higher 500 ms before the climbing onto the platform at 8 Hz and 15 Hz.

As shown in Fig. 9A, during the cooperative phase the coherence between the PrL cortex and the NAc was significantly higher when rats were cooperating ON-platform at 8–10 Hz and at 18–19 Hz (Fig. 9A, jackknife estimates of the variance, p < 0.05). This result was confirmed by the dynamic analysis illustrated in Fig. 9D–G, revealing the highest coherence magnitudes between the PrL cortex and the NAc 500 ms to 1 s after rats climbed ON-platform at 10–15 Hz, and another significantly higher cluster at 17–20 Hz right after they climbed ON-platform (jackknife estimates of the variance, p < 0.05). It is worth mentioning that the peak of coherence around 8–10 Hz observed during the individual phase before the climb onto the platform (Fig. 8A) disappeared during cooperation (Fig. 9A). In addition, coherence in this band during cooperation increased.

The coherence between PrL cortex and the BLA was similar for the two conditions (Fig. 9B), except at 18–19 Hz. The peak coherence was found around 10 Hz, for the ON-platform condition, although the differences were not significant. The coherograms in Fig. 9E–H were very similar, revealing only small clusters of higher coherence at 17–20 Hz 2 s before the climbing onto the platform. The coherence between the NAc and the BLA was significantly higher when rats were cooperating ON-platform at 8–10 Hz (Fig. 9C) and significantly higher BEFORE-platform at 15–17 Hz (jackknife estimates of the variance, p < 0.05). The coherograms in Figurre 9F–I situated these differences in small clusters, showing the highest coherence at 9–14 Hz 200 ms BEFORE-platform and around 14 Hz 200 ms after the climbing onto the platform. The coherence between the PrL cortex and BLA and between NAc and BLA showed high coherence in the band at 10–15 Hz 1 s after the climbing ON-platform, but the differences with the BEFORE-platform conditions were not significant.

In this work, before the cooperation experiment, most leader rats presented low levels of anxiety, while most follower rats presented intermediate levels of anxiety (Fig. 3D–F). According to Herrero et al. (2006) and Venero et al. (2004), higher levels of trait-anxiety in rodents are related to impaired learning and memory. This could explain why follower animals in our experiments—presenting higher trait-anxiety than leaders—had more difficulty in grasping the requirements to complete the cooperation trials. In this sense, leaders would likely be less anxious animals and present a better cognitive performance. The locomotor activity was similar for both leaders and followers, indicating that the differences observed in brain activity are not related to differences in locomotion. Regarding social competition, we hypothesized that leader rats would present higher levels of dominance than followers, but to our surprise, followers scored higher in the social dominance index. These results were consistent with the Larrieu et al. (2017) study, in which animals classified as dominants also presented higher trait-anxiety.

Results from the co-localization analysis in the PrL cortex indicated that leader rats participating in the cooperation phase presented higher c-FOS activation in neurons containing D1 receptors than follower ones. This activation was also higher than that presented by leaders and followers during the individual phase and by the control group, suggesting the involvement of these medium-spiny neurons in the acquisition of the cooperative task. Previous neuroanatomical studies found that most prefrontal cortex (PFC) neurons express either D1 or D2 receptors, rather than expressing both in the same neurons (Vincent et al. 1993; Gaspar et al. 1995; Santana et al. 2009). This anatomical differentiation separates PFC dopaminergic receptor containing neurons in two populations, which would project to different subcortical areas.

In this work, c-FOS activation of cells containing D2 receptors and participating in the cooperation phase also increased in comparison with the control and the individual groups, although the differences between leaders and followers were not significant, suggesting that these cells could also play a role in the cooperation process, and be worth further analysis.

After showing the involvement of the PrL cortex in cooperation and identifying cell types that increased their activation during this task, we recorded LFPs from other subcortical structures, such as the NAc and the BLA, known to receive and send projections to PrL neurons and to have an important role in social behaviors.

In this study, rats completed the cooperation task successfully, but rats' strategies to cooperate were more homogeneous than in previous experiments (Conde-Moro et al. 2019), and there was not a clear difference between leaders and followers. This might be due to the adaptations of the cooperation experiment or to differences between rat batches, as some might contain less anxious rats, which, as mentioned before, should present fewer cognitive impairments and thus show better performance in the task (Venero et al. 2004; Herrero et al. 2006).

The analysis of the averaged spectral power and time–frequency decomposition showed that the highest spectral power was found in the PrL (delta band) when rats were cooperating on the platform. This increase was greater during the first second after the climbing onto the platform. When rats were cooperating on the platform, the power in low and high theta bands also increased significantly compared with moments of resting individually on the platform or before climbing onto it. These findings support those found in our previous work (Conde-Moro et al. 2019) suggesting the involvement of the PrL cortex in the acquisition of a cooperative task. LFPs recorded in the NAc showed interesting power dynamics as well. Delta and low theta ranges were selectively more active during the first second of cooperation, while the high theta band had a considerable dip at this same moment. As previously noted, dopaminergic receptors containing neurons located in the rat PrL cortex project to the NAc (Ongür and Price 2000) and to other subcortical structures that, in conjunction, can modulate social behavior (Grossman, 2013; Gunaydin et al. 2014). According to Goto and Grace (2015), a dopamine release from the PFC can modulate the activity of the NAc in goal-directed behaviors. The co-localization results in this study showed that the activity of D1-containing cells in the PrL is increased in leader rats during cooperation. Jenni et al. (2017) found that dopamine in PFC D1 receptors reinforces responses, providing larger rewards through connections to the NAc, while dopamine on PFC D2 receptors facilitates adjustments in decision-making through connections to the BLA. Following this argument, it would be reasonable to assume that the D1 neurons that were more active during the cooperation task in our rats are those likely projecting to the NAc, rather than to the BLA.

Additionally, in the connectivity analysis, we observed that when rats were cooperating on the platform, there was higher phase coherence between the LFPs recorded in the PrL cortex and the NAc in the 9–15 Hz frequency range (high theta) than before they climbed onto the platform, especially 1.5–1 s on the platform. BEFORE-platform, the highest coherence was found in the delta and theta bands 0.5–1 s before the climbing onto the platform. These findings add more evidence to suggest an involvement of populations of cells in the NAc in the decision of climbing onto the platform to cooperate or in the prediction of a mutual reward by cooperating with a conspecific. Additionally, the electrophysiological experiments included in this study and our previous one (Conde-Moro et al. 2019) showed an increase in the spectral power of LFPs recorded in the PrL due to cooperation, suggesting that the increased power observed in the NAc was related to the dopamine release from the PrL and not only to endogenous activity from the NAc.

According to Liu et al. (1994), Haber and McFarland (1999), and Goto and Grace (2015), the NAc could act as a gatekeeper, controlling what information is important enough to get access to basal ganglia nuclei, such as the amygdala. In this work, the highest power found in the BLA was in the delta, and low and high theta bands, when rats were individually on the platform. The connectivity analysis also indicated higher levels of coherence between the NAc and the BLA when rats were individually on the platform. The highest coherence during the cooperation phase was found between the NAc and the BLA in the theta band before the climbing onto the platform, which is consistent with the accentuated decrease in BLA power in all the bands and across the whole time window observed during the cooperation phase. The amygdala has been proposed to contribute to social cognition (Adolphs 2010; Bickart et al. 2014) and it is known that the BLA sends strong projections to the PRL cortex and the NAc (Groenewegen et al., 1990; Haber et al. 1995), structures known to modulate emotional and motivational processing, including the motivational aspects of predicting cues (Davis 1922; Everitt et al. 2000). Thus, we consider that the BLA might be involved in the prediction and anticipation of the task outcome.

The fact that during the individual phase the activity of almost all areas studied did not differ much before and after the climbing onto the platform, and that all the changes we observed took place during the cooperative phase, leads us to conclude that the changes in brain activity observed in this work are due to the cooperative aspect of the task and not to the rewarding effect that other task requirements could have for the rats (such as climbing onto the platform) or for the mere social component of it (i.e., being in the apparatus with a cagemate), as all of these conditions were also met during the individual phase.