Acute Low Alcohol Disrupts Hippocampus-Striatum Neural Correlate of Learning Strategy by Inhibition of PKA/CREB Pathway in Rats

The hippocampus and striatum guide place-strategy and response-strategy learning, respectively, and they have dissociable roles in memory systems, which could compensate in case of temporary or permanent damage. Although acute alcohol (AA) treatment had been shown to have adverse effects on hippocampal function, whether it causes the functional compensation and the underlying mechanisms is unknown. In this study, rats treated with a low dose of AA avoided a hippocampus-dependent spatial strategy, instead preferring a striatum-dependent response strategy. Consistently, the learning-induced increase in hippocampal, but not striatal, pCREB was rendered less pronounced due to diminished activity of pPKA, but not pERK or pCaMKII. As rats approached the turn-decision area, Sp-cAMP, a PKA activator, was found to mitigate the inhibitory effect of AA on intra- and cross-structure synchronized neuronal oscillations, and rescue response-strategy bias and spatial learning deficits. Our study provides strong evidence of the critical link between neural couplings and strategy selection. Moreover, the PKA/CREB-signaling pathway is involved in the suppressive effect of AA on neural correlates of place-learning strategy. The novel important evidence provided here shows the functional couplings between the hippocampus and striatum in spatial learning processing and suggests possible avenues for therapeutic intervention.

ensure that the cluster boundaries were well separated and waveform shapes were consistent with action potentials (Fig. S4A). Interspike interval histograms were additionally examined for ensuring single unit activity (inserts of Fig. S4). Using methods described elsewhere (1), units were then graded for quality and classified as putative medium spiny neuron (MSN), fast spiking interneuron (FSI), or tonically active neuron (TAN) subtypes form striatal recording and pyramidal neuron and putative interneuron from hippocampal recording, respectively. A pyramidal neuron or MSN was further classified as "related-to-task'' if its firing rate in any ±300 ms perievent window was more than 2 standard deviations (SD) above its baseline firing rate for three consecutive 20 ms bins (2). Units not classified as task-responsive were deemed ''unrelated-to-task.''

Neural activity in behavioral task
Behavioral correlates of neural activity changes were assessed by constructing perievent time histograms synchronized during task-related event periods (i.e., 2 sec before trial start as baseline, 0.5 sec before to 1.0 sec after trial start, 0.5 sec before to 1.0 sec after turn start, 0.5 sec before to 1.5 sec after trial end, 1.0 sec after rewarding).
Firing rates in 500-ms bins were each compared against the firing rates from the basal recording (from 90 sec pre-trial initiation to 30 sec pre-trial initiation). Significance was tested by using ANOVA analyses (P<0.05). The neuron was related to the occurrence of the event only if one test bin from it significantly differed from all baseline bins (3). Neurons were classified into four categories based on the task period in which they exhibited significant firing-rate changes ( Fig. S4A and S5A; for HPC: change in relation to turn (start to end), movement in the aim arm or reward; for DS: change in related to trial start, turn start or reward). The time windows of task event were based on visual inspection of the data. Similarly, z-score (subtracting firing rate averages and dividing by the SD) was compared to ensure that a minority of neurons with exceptionally high-firing rates did not dominate the average.
To compare the alternation of neural activity between HPC and DS following alcohol exposure, the distribution of the ratio of HPC activity to DS activity was calculated. Briefly, total number of cells and total number of turn-related cells from each recording day were collected and grouped by strategy that rats used. Percentage of each brain area activity was calculated as turn-related cell number divided by total cell number. Then the characteristic of neural activity between HPC and DS when the strategy was applied were obtained (Fig. S5E). The ratio of HPC activity to DS activity was calculated as the percentage of HPC activity divided by the percentage of DS activity. The distribution was determined by the learning strategy obtained from behavioral task under alcohol or saline condition. Each single data was collected by each recording day, thus the mean and S.E.M. was obtained through the whole recording days (Fig. S5E insert).

Spectral analysis of LFPs
Spectral analysis was used to assess the dominant frequencies in the LFPs during the task-related event periods (as above). The analysis of LFP power was performed by using Neuroexplorer (Nex Technologies, Littleton, MA). The power spectra were calculated using Welch's method (1024 frequencies between 1 and 200 Hz, smoothed with a Gaussian Kernel with bin width 3). Then the low-theta (4-7 Hz), high-theta (7-12 Hz), low-gamma (30-48 Hz), high-gamma (52-120 Hz) and high-frequency (120-200 Hz) were obtained by band-pass filtering. The mean power spectral density (PSD) in each given band was calculated.   S2. The effect of AA on using learning strategy immediately after a three-session training or 30 min before probe test. The percentage of animals that used a place strategy was calculated and presented. Rats received a probe trial starting from the opposite start arm. Compared with AA-D1PRE group (treated 30 min before training), a higher number of rats with place strategy was found when AA was treated immediately after training (AA-D1POST; Chi-square test: X 2 =5.63, df=1; P<0.05) or 30 min before probe test on the 2 nd day (AA-D2PRE; Chi-square test: X 2 =5.63, df=1;  ###P<0.001, AA-Sp2.1-intraCA1 vs. CON or AA-Sp0.21-intraCA1). n=6 for each group.

and DS
Neuron spike trains were electrophysiologically recorded and classified by waveform shape and spiking patterns (see SI Materials and Methods for explanation). Examination of the recording tracks confirmed that HPC units were recorded both in CA1 region of dorsal hippocampus. Striatal units were confined to dorsal-medial and dorsal-lateral striatum. Data were combined across areas since we did not detect regional specificity of the responses. The fact that many more striatal neurons were recorded in DS than HPC was likely due to the failure of a few electrodes on HPC drives, and did not reflect a meaningful difference in information representation.
To understand the effect of alcohol on neural response to strategy behavior, we then sought to examine how the firing pattern of individual neurons changed in HPC and DS regions. Since the rats was well-trained before recording (average 615 trials), the neural firing patterns of HPC ( The neural activities were collected and analyzed during the event periods as illustrated in Fig. S1A. The turn, but not reward or aim arm, related HPC pyramidal neurons recorded in the alcohol-treated group fired at markedly lower rate than those in the control one (P<0.001; Fig. S5B left). Similar results were also found in z-score analysis during turn-decision period (P<0.01; Fig. S5C left). However, no difference was observed in the spike frequency or z-score during the baseline or trial start period between two groups, although firing rate at turn, aim arm and reward moments were higher than their baseline (P<0.001, for all comparisons). The striatal MSNs had higher firing rates in alcohol group during start, turn and reward periods compared with its baseline (P<0.0001, for all comparisons; Fig. S5B right). Higher firing rates were also found during event periods, including start (P<0.001), turn-decision (P<0.001), and reward (P<0.001) in control group. However, their activity was equally enhanced in both groups. Normalized firing rates by z-score analysis also found firing frequency of control group and alcohol group elevated during events (P<0.001, for all comparisons; Fig. S5C right). Consistent with previous reports, we found the selective inhibitory effect of alcohol on HPC, but not DS, cerebellum, lateral septal nucleus or prefrontal cortex (4,5). The firing rate and its normalized z-score in each strategy have also been considered within each condition (control or alcohol). However, no statistical difference was found between the strategies (data not shown). Since the magnitude of the neuronal activity did not depend on the use of place or response behavioral strategy, we combined the results of two strategy groups in each condition.
The distribution of the turn-related pyramidal neurons in alcohol group was statistical different (Chi-square test: X 2 =4.69, df=2; P<0.05, Fig. S5D left), which was tended to decrease. There is no change in the proportion of other task-related cells or cells were not related to task events. However, the population of turn-related MSNs did not appear differently (Chi-square test: X 2 =0.07, df=2; P>0.05, Fig. S5D right).
The characteristic of activity pattern between HPC and DS was shown in Fig. S5E.
There was no significant correlation between strategy and HPC activity (Pearson's r = 0.021, P>0.05), between strategy and HPC activity (Pearson's r = 0.016, P>0.05), or between strategy and the ratio of HPC activity to DS activity (Pearson's r = 0.08, P>0.05). The activity ratio of HPC to DS in CON groups was higher than in AA groups (P<0.001, for all comparisons; Fig. S5E insert).

Effect of alcohol on power oscillatory activity of HPC and DS during strategy task.
We recorded LFPs simultaneously in the HPC and DS as rats performed the session III of learning period (Fig. S7A and Fig. S7B). Both HPC and DS low-theta (4-7 Hz) remained nearly constant. The power of both HPC and DS low-gamma (30-48 Hz) declined abruptly as the rats left the start chamber, but gradually rose as the reward zone was approached. By contrast, the high-frequency (120-200 Hz) oscillation of HPC and DS reached to peak near the turn area and then declined to baseline level slowly. ANOVA analysis indicated main effects of event on low-gamma (F (4,112) =3.87, P<0.01) and HF (F (4,112) =4.80, P<0.01), without effects of treatment. For high-theta turn, aim arm and reward, P<0.001). There was no difference in normalized power in HPC region (Fig. S7B insert of the fourth image). Overall, although alcohol inhibited HPC, but not DS, theta and gamma power during behavioral task, while functional HPC power of alcohol-treated rats kept a similar neural correlate with behavioral events as controls. The strategy-specific increasing power at DS high-theta band suggested the essential role of DS in response strategy task, particularly at turn-decision period.
These findings indicated that the suppressive effect of AA treatment on high-theta and high-gamma oscillations from HPC should not be attributed to the strategy bias.
Moreover, DS high-theta power was not affected in AA-treated rats that used response strategy. gray hash: CON-R vs. AA-R.***P<0.001; black asterisks: CON-P vs. CON-R; gray asterisks: AA-P vs. AA-R). n=7 for each group.

Effect of alcohol on synchronized neuronal oscillations during strategy-learning task.
Since both LFP and neural activity were related to learning task while no strategy-specific alternation was observed in these analyses (Fig. S5), we therefore examined the relationship between the firing of individual neurons and the simultaneously recorded LFPs in each brain area during the turn-decision period (Fig.   S8A). As shown in Fig. S8B, the mean vector length were in a similar level in low-theta, low-gamma, and high-gamma of HPC and DS. Significant effects of strategy (F (1,30) =41.89, P<0.001) and treatment (F (1,30) =67.02, P<0.001) on high-theta band was found. Tukey's test revealed there was a significant enhanced phase-locking compared place-strategy to response-strategy (P<0.001, CON-R vs. CON-P; P<0.001, AA-R vs. AA-P). The vector length of AA-treated group (AA-R), which tend to apply response strategy, was even lower than control response-strategy group (CON-R; P<0.05). The vector length AA-treated place-strategy group (AA-P) was also smaller than control place-strategy group (CON-P; P<0.01). The mean phase of HPC neurons tended to fire at the peak of the high-theta cycle in CON-P, CON-R and AA-P groups, but not in AA-R group (Fig. S8C left). These results indicated a powerful phase-locking of HPC high-theta oscillation was essential for a place strategy but weakened in response-strategy AA condition. Interestingly, the prominent firing phase advanced alcohol-treated animal to apply a place but not response strategy. DS neurons also had an unimodal population distribution, which also occurred at the peak ( Fig. S8C right). Therefore, the phase-locking of DS high-theta oscillation either was not essential for a response-strategy or played a critical role for both place and response strategies.
Since DS and HPC were considered to be functionally related (6), the phase-amplitude coupling between two regions were calculated (Fig. S9A). Analysis of the mean peak coherence values between the HPC and DS revealed a significant main effect of strategy across high-theta frequency during the tuning period (F (1,30) =56.60, P<0.001). However, no effect of response strategy was observed within the other bands. Tukey's test revealed that the coherence of CON-P was higher than CON-R (P<0.001). Similar result was also found between AA-P and AA-R (P<0.001).
No main effect on treatment was observed across frequency bands investigated. These results indicated place-strategy was associated with a higher coherence between the HPC and DS, while response-strategy was associated with a lower coherence.
Since the preferred low-frequency rhythm modulating high-gamma amplitude was dependent on the task-specific modality (7), we tested whether the phase-amplitude coupling was associated with the strategy selection. The phase-amplitude couplings between the DS high-theta phase and the amplitude of HPC fast-frequency oscillations were very prominent in place strategy groups (Fig. S9A, first-row, first-image and third-image). However, this effect was much weaker in response strategy groups (Fig. S9A, first-row, second-image and fourth-image). ANOVA analysis indicated a main effect of strategy (Fig. S9B middle), but no effect of treatment on the mean peak of modulation index (MI). This different did not resulted from specific band effects, since the statistical differences were identified in two common selected bands, 120-160 Hz (F (1,30) =83.17, P<0.001) (8) and 120-200 Hz (F (1,30) =94.05, P<0.001) (6). Although the DS theta modulation of HPC 55-120 Hz (high-gamma) oscillations was found (Fig. S9B, first-row), there was no statistical different among groups (Fig. S9B, left). No significant difference in the phase-amplitude couplings between the HPC theta phase and the amplitude of DS high-gamma oscillations was found among groups (Fig. S9A second-row; Fig S10B   right).  (***P<0.001; black asterisks: CON-P vs. CON-R; gray asterisks: AA-P vs. AA-R).
n=7 for each group.   for striatal MSNs of CONP-Sp0.21. Infusion of Sp-cAMP into the HPC did not affect neural activity of place strategy rats. All data are presented as mean ± S.E.M..