Dynamics of Neural Microstates in the VTA–Striatal–Prefrontal Loop during Novelty Exploration in the Rat

Microstate maps are reproducible across different days. Simultaneous LFP signals were recorded on six different days (each day with a new object) in PFC, STR, and VTA. Microstate maps from each region on the first day were chosen as reference maps. Maps are rotated counterclockwise by 90° for PFC and STR and rescaled for visualization purposes. A₁, In the PFC, the squared correlations among the four maps on each day and average microstate maps were computed, and are shown with red, black, blue, and green lines for microstates 1–4, respectively. Lower correlation values on certain days can be loosely interpreted as the absence of the Microstate on that day (in reference to the average Microstate). A₂, Microstate (MS) maps are reordered based on their best match (A₁). Most microstate maps retained the spatial features on each recording day. (A₁ and A₂ are from animal 1 in A₃). A₃, The fraction of maps that were repeated over time for each of four animals. B, C, These rows show the same results as described above, but for data from the striatum and VTA.

Microstate maps represent the spatiotemporal structure of LFP activity. A₁, Surrogate LFP datasets with conserved local structure were created by pivoting the data around a random time point (independently selected per trial; blue, before pivot point; red, after pivot point). A₂, The data were swapped around these pivot points to generate a surrogate dataset. A₁ and A₂ illustrate the principle of the shuffling method; the actual recordings were ∼20 min long. The shuffling strategy disrupts the spatial structure while imposing minimal disruption to the time series within each channel (vertical line at 300 ms as an illustration). B, Microstate analysis was performed on these surrogate datasets, and GEV was computed. The corresponding GEV distribution for one of the recordings (C, red circle) showed a significant decrease in GEV in the surrogate data, indicating that the original temporal alignment between channels is underlying the microstate maps explanatory value. C, GEVs of the empirical data were always higher than the corresponding median GEV from surrogate datasets in all three brain regions, showing statistical significance of the microstate results.

Automated movement tracking to segregate behavioral states. A₁, An example video frame during the experiment. A₂, Animal movement is accurately tracked using DeepLabCut, shown here using the same frame as in A₁. Animal movement was tracked using five body markers (blue, black, red, green, and yellow dots represent, respectively, snout, tail start, left ear, right ear, and head center). B, Each experimental recording consisted of four sessions (∼5 min each session). Sessions 1 and 3 were open field recordings, while session 2 included a novel object (different each day) in the middle of the box. In session 4, the same object was present. Rats spend less time interacting with familiar objects than novel objects (each dot represents snout location in each frame; red dots show time points labeled as interaction). A small number of erroneous object locations were filtered out before analysis. C, The distribution of temporal coverage of each behavioral state (rest, movement, object interaction) across sessions was dominated by periods of rest. Object interaction can take place during rest as well as during movement, represented as an overlap in behavioral states. D₁, Median temporal coverage from all recordings captures the trends over all sessions, recordings, and animals (N = 4). D₂, Median distribution of animal speed movement from all the recordings. The dashed line shows the motion speed threshold separating resting from movement.

Microstate parameters are affected by behavioral state. A, Example results from the recording in PFC (rat 1, recording day 3). A₁, Temporal coverage fraction for all four microstates differed in the two behavioral states (statistics are in C₁). A₂, Occurrence rate during the two behavioral states (solid bars on the left refer to stationary periods, and lighter bars on the right refer to movement periods; the four numbers refer to the different microstates). A₃, Similarly, microstate durations were modulated by behavioral state. B₁–B₃, The modulation of microstate properties by behavioral state (expressed as stationary minus movement) was consistently in the same direction (PFC, 76 ± 13%) across 6 different recording days (each dot corresponds to a day; only microstates with r² > 0.5 repeatability are shown here). C₁–C₃, Similar to B but for data pooled across 27 novel objects in four rats. Significance was assessed by comparison against surrogate datasets (thin horizontal lines indicate a significance threshold of z = ±1.96). There was a high fraction of significant results for all three microstate properties in all three brain regions.

Microstate activation time series are rhythmically correlated within and between brain areas. A–E₂, Illustrative results from one recording session (rat 1, recording day 2). A, Microstate time series cross-correlations show damped oscillatory dynamics across microstates both within regions and across regions. B, Example cross-correlation between activation time series of microstate₂ from STR and microstate₂ from VTA (this is a zoomed-in version of the boxed plot in A). The typical cross-correlation structure is absent in shuffled data (red line). C, The z-normalized power spectrum of the cross-correlation (cross-spectral density) shows a peak at ∼8 Hz. The shuffled data do not exhibit a dominant peak (red line). D, The z-normalized peak power from all combinations of microstate activation cross-correlations illustrates varying degrees of correlation within and between areas. E₁, E₂, The z-normalized power spectra from all combinations of microstates within and between areas were clustered into two groups, one with a lower-frequency peak (∼2-4 Hz; E₁, black lines,) and one with a higher-frequency peak (∼8 Hz; E₂, blue lines). Thick lines indicate cluster centers, and thin lines are individual power spectra. The red lines indicate CSD power for statistical significance calculated from permutation testing (p < 0.01). F₁–F₂, Results from all recordings in all animals show a consistent grouping of power spectra based on the frequency of the peak in power. The red lines indicate CSD power for statistical significance calculated from permutation testing (p < 0.01).

Power spectra of microstate activation cross-correlations are grouped into two clusters that are differentially modulated by behavior. A₁–C₂, Illustrative data from one recording (rat 1, recording day 2). A₁–A₃, Temporal coverage by each microstate during longer resting periods (left bars) and object interaction (right bars shown in light colors) indicates the dominant microstate during each behavioral state. B₁, Dominant microstates (yellow) during longer stationary periods are shown across areas. All dominant microstates were grouped for further cross-correlation spectrum analysis. B₂, Power spectra of cross-correlation of relevant microstates (shown in B₁) during longer rest periods show a peak at ∼8 Hz (thick line shows the mean power spectrum). C₁, Same as B₁ for object interaction microstates. C₂, Cross-correlation power spectra of dominant microstates (C₁, yellow) during object interaction show an earlier peak at ∼2–4 Hz (thick line shows the mean power spectrum). D, Mean power spectra from all recordings show higher affinity to one of the spectral cluster centers (Fig. 6F1,F2), except very few spectra from object interaction microstate combinations that show relatively greater similarity to the cluster more populated by rest microstates.

Distributions of microstate properties in all three regions in all recording sessions provide insights on temporal and spatial features of microstates (Table 1). A, Temporal coverage fraction shows temporal area covered by each microstate. B, Occurrence rate shows that microstates in all three areas appeared approximately three to four times per second on average. C, The top four microstates exhibited a high global explained variance (60–80%) in all three brain regions. D, The top four microstates exhibited high spatial correlation (Pearson) with the electrode topography (∼0.5). E, State duration of microstates followed power law distributions in PFC and STR, and had a characteristic peak of ∼100 ms in the VTA.