Neuronal Dynamics Regulating Brain and Behavioral State Transitions

Abstract

Prolonged behavioral challenges can cause animals to switch from active to passive coping strategies to manage effort-expenditure during stress; such normally adaptive behavioral state transitions can become maladaptive in psychiatric disorders such as depression. The underlying neuronal dynamics and brainwide interactions important for passive coping have remained unclear. Here, we develop a paradigm to study these behavioral state transitions at cellular-resolution across the entire vertebrate brain. Using brainwide imaging in zebrafish, we observed that the transition to passive coping is manifested by progressive activation of neurons in the ventral (lateral) habenula. Activation of these ventral-habenula neurons suppressed downstream neurons in the serotonergic raphe nucleus and caused behavioral passivity, whereas inhibition of these neurons prevented passivity. Data-driven recurrent neural network modeling pointed to altered intra-habenula interactions as a contributory mechanism. These results demonstrate ongoing encoding of experience features in the habenula, which guides recruitment of downstream networks and imposes a passive coping behavioral strategy.

Figure 1:. Active-to-passive coping behavioral state transition in larval zebrafish.

(A) BC rig tracks fish position and delivers electric shocks (B and C) The behavioral response of two fish: 0V control fish (B) and 5V shocked fish (C). The path during the pre-shock period (purple) and the post-shock period (orange) (upper left, scale bar = 10 mm). The position along the short (upper right) and long (lower left) sides of tank over the entire protocol (pink shading indicates shock period). Speed of fish in 60 s windows (lower right) shows that BC results in reduced movement. (D) Shocked fish (blue, n=33) increased their speed (AC) in response to the first five minutes of BC (p=5.15 × 10⁻⁶) then transitioned to a reduced mobility state (PC) compared to controls (black, n=14; p=7.03 × 10⁻⁸ at final time point). See also Figure S1A. (E) Fish context was changed post-BC to assess recovery rate (transition indicated with dashed lines). Fish moved to the neutral context (blue) recovered to the level of control fish kept in the same tank (black; p=0.11, n.s.; One-way ANOVA with Holm-Sidak post-hoc comparison; data also in panel D), while fish returned to the aversive context (green, n=40) or the partially-neutral context (yellow, n=29) did not recover compared to shocked fish in the neutral environment (p=2.12 × 10⁻⁸ and p=.003, respectively). (F) Exposure to ketamine one hour before BC (n=16) reduced and delayed the onset of PC compared to controls (n=8, p<0.05 between 7 and 17 minutes post shock), but did not fully prevent the transition when fish were exposed to a longer BC protocol. See also Figure S1B and S1C. In all figures, shaded area and error bars indicate standard error of the mean (s.e.m.) and all statistical tests are Student’s t-tests, unless otherwise indicated (*P<.05; **P<.01; ***P<.001).

Figure 2:. Whole-brain Ca²⁺ imaging using LFM: unique vHb hyperactivity following BC.

(A) LFM was used to measure GCaMP6s Ca²⁺ signals during a 45-minute post-BC period. (B) Orthogonal maximum intensity projections of a LFM volume (scale bar = 50µm). Hb: habenula; OT: optic tectum; Fore: forebrain; Oe: olfactory epithelium; Ce: cerebellum; Hind: hindbrain. See also Figure S2. (C) Orthogonal min-max projections of volumes showing the change in fluorescence over the post-BC period in a representative fish from the control group (top), shocked group (middle); and ketamine group (bottom). Red and blue indicate an increase and decrease in fluorescence, respectively (arbitrary units; black arrows show the location of the vHb; scale bar = 50µm). (D) Change in average fluorescence over the post-BC period. Shocked fish showed an increase in vHb activity (blue, n=6) compared to control fish (black, n=4; p=0.0002, two-way repeated-measures ANOVA with Tukey HSD post-hoc comparison). This effect was reduced in shocked fish that were previously treated with ketamine (pink, n=8; p=0.0014).

Figure 3:. 2P imaging of cellular-resolution brainwide responses during BC reveal the dynamics of experience-encoding activity in the vHb.

(A) Head-fixed volumetric 2P calcium imaging during BC while monitoring tail movement. (B) Diagram of imaged brain area (yellow rectangle). (C) Head-fixed fish exhibit a reduction in tail movement rate as a result of BC (black: control, n=6; blue: shocked, n=8; p=0.030, one-sided Student’s t-test). Prior exposure to acute ketamine eliminates this response (pink, n=4, p=0.283, n.s.). (D) Regressor consisting of linearly rising activity following the onset of shocks was used to search for cells with activity correlated to the extent of BC. (E) Example of the fit between the regressor and the ∆f/f Ca²⁺ trace for four neurons (bottom; locations at top; black dashed line indicates start of shocks). (F) ROI of each neuron colored by the correlation (r) between its ∆f/f Ca²⁺ trace and the regressor (scale bar = 25 µm). See also Figure S3. (G) Average baseline-subtracted ∆f/f response for all neurons in each of several brain regions (see also Figure S4). In the vHb, the average activity over the final two minutes was increased in shocked fish (n=8) compared to control (n=6) and ketamine-treated (n=4) fish (p=0.0189 and p=0.0432, respectively; two-way repeated-measures ANOVA with Tukey HSD post-hoc comparison). Shocked fish also showed a decrease in raphe activity (p=0.012 and p=0.007) and dorsal thalamus activity (p=0.007 and p=0.006) compared to controls and ketamine-treated fish. (H) Example of MultiMAP registration to identify serotonergic neurons in the superior raphe (see methods). (I and J) Raster plot of baseline-subtracted ∆f/f activity of all 5-HT+ neurons (I; n = 158 neurons from 3 fish) and 5-HT– neurons (J; n = 332 neurons) sorted based on hierarchical clustering with top 8 flat clusters demarcated by coloring scheme of left. 5-HT+ neurons, on average, exhibited a significantly larger response to BC onset compared to 5-HT– neurons (60 second window following onset; p = 0.026, paired Student’s t-test) and compared to 5-HT+ neurons in non-shocked controls (n=4 non-shocked control fish; p = 0.027, Student’s t-test). 5-HT– neurons did not have a significant response to BC onset (p = 0.37, Student’s t-test).

Figure 4:. Staggered recruitment of individual vHb neurons temporally tiles BC epoch.

(A) Percentage of neurons that showed a significant increase in activity (top) or decrease in activity (bottom) as a result of BC (black: control, n=6; blue: shocked, n=8; pink: ketamine, n=4; data from these fish also presented in Figure 3). A higher percentage of vHb neurons were excited in shocked fish (p=0.037), while more raphe neurons were inhibited (p=0.048), compared to controls. (B) ∆f/f Ca²⁺ traces (purple) for all vHb neurons that showed a significant increase in activity were smoothed (dark purple) and then fit to a sigmoid (black). 4 example neurons with varying response durations (time between 10^th and 90^th percentile of the sigmoid) and latencies (time to the 50^th percentile) are shown (dashed line is onset of shocks). (C) Histogram of response durations. (D) Histogram of response latencies (excluding neurons with response durations longer than 30 minutes). (E) Raster plot of ∆f/f traces for neurons in panel D from a representative fish sorted by response latency. (F) ROIs of neurons in panel E colored according to their response latency (6µm between sections; scale bar = 25 µm). (J) Histogram of difference in response latency between neighboring neurons and neurons with random spatial locations. The latency difference between neighbors is smaller than expected by chance (n=979 neurons from 8 fish; p=.0005, nonparametric bootstrap, see Methods). (H, I) Analogous to panels C and D for cells showing significantly decreased activity in the raphe. (J) Analogous to panel E for raphe neurons.

Figure 5:. Prior exposure to BC enhances behavioral and neural passive-state responses.

(A) Fish were exposed to free-swimming BC, allowed to recover for 3 hours, and then imaged during head-fixed BC. (B) Change in tail movement rate relative to baseline as a result of head-fixed BC for control fish (black, n=6), shocked fish (blue, n=8) and re-exposed fish (teal, n=12) (dashed line indicates start of shocks; data from shocked and control fish also presented in Figure 3). Re-exposed fish have reduced movement rates compared to baseline starting 3 minutes after shock (p<0.05 at all time points between 4 and 15 minutes, except for minute 8; one-sided Student’s t-test), while shocked fish did not show a significant reduction until 20 minutes after the start of shock. (C) Average baseline-subtracted ∆f/f Ca²⁺ response for all neurons in the vHb across fish. During the final two-minutes of the BC protocol, re-exposed fish exhibited higher vHb activity compared to fish experiencing the protocol for the first time (p=0.0267; two-way repeated-measures ANOVA with Tukey HSD post-hoc comparison). (D) Scatter plot comparing vHb activity during final two minutes of protocol to the change in tail movement rate during the protocol reveals a significant negative correlation (gray line: linear regression; shaded area: 95% confidence interval; r=−.56, p = 0.003). See also Figure S5A.

Figure 6:. Optogenetic activation of vHb causally evokes passive-state behavior and brainwide activity dynamics.

(A) Rig used to activate ChR in free-swimming fish. (B) Max-projection of confocal stack of Hb in 15dpf Tg(ppp1r14ab:Gal4-VP16; UAS:ChR2-mCherry;elavl3:h2b-GCaMP6s) showing expression pattern of pp1r14ab (red) on elavl3 background (green; scale bar = 25 µm; see also Figure S6A). (C) Inward photocurrents in an in vivo voltage-clamped ChR2-mCherry+ vHb neuron (left) and ChR2-mCherry– dHb neuron (right), evoked by 5ms pulses of 475-nm light (blue bars; average response indicated by solid black line, n=6 ChR+ pulses, 8 ChR– pulses). (D) Baseline-subtracted speed of ChR+ fish, Tg(ppp1r14ab:Gal4-VP16; UAS:ChR2-mCherry) (blue, n=10), and ChR– clutch mates (black, n=11) in response to two minutes of 20Hz 460-nm light pulses (5 ms duration; 1 mW/mm²; 4-minute baseline). During stimulation, the ChR+ fish showed reduced speed compared to ChR– fish (p=0.0038) and compared to baseline (p=0.048, paired Student’s t-test). In contrast, the visible blue light caused ChR– fish to show an increase in speed (p=0.040, paired Student’s t-test). See also Figure S6B. (E) Rig used to activate NpHR in free-swimming fish. (F) Max-projection of 2P stack of superior raphe in 13 dpf Tg(tph2:Gal4ff; UAS:NpHR-mCherry) (scale bar = 25 µm). (G) Baseline-subtracted speed of NpHR+ fish, Tg( tph2:Gal4ff; UAS:NpHR-mCherry) (blue, n=11), and NpHR– clutch mates (black, n=10) in response to 2 minutes of continuous 545-nm light (5 mW/mm²; 4-minute baseline). NpHR+ fish showed lower speed than NpHR– fish during the second minute of inhibution (p=0.0039) and also than baseline (p=0.039, paired Student’s t-test). In contrast, the visible yellow light caused NpHR– fish to show an increase in speed (p=0.013, paired Student’s t-test). (H) Max-projection of confocal stack of Hb in 15dpf Tg(dao:Cre-mCherry; vglut2a:loxP-DsRed-loxP-GFP; elavl3:h2b-GCaMP6s) showing expression dao (red) on elavl3 background (green) (scale bar = 25 µm). (I) Speed of NpHR+ Tg(dao:Gal4-VP16; UAS:NpHR-mCherry) fish (blue, n=8) and NpHR– clutch mates (black, n=10) in response to BC followed by two minutes of continuous 545-nm light (5 mW/mm²). NpHR+ fish showed a significantly larger increase in speed than NpHR– fish during the second minute of inhibition and the two minutes following stimulation (p=0.034). Both the NpHR+ and NpHR– cohorts showed a significant increase in speed compared to the 4-minute baseline period prior to light onset (p=0.009 and p=0.027, respectively; paired Student’s t-test). (J) 2P Ca²⁺ imaging of the response to unilateral one-photon laser scanning stimulation of the vHb. (K) Max projection of ChR+ and ChR– fish (site of stimulation indicated by white arrows). ROIs of the neurons with a significant response to simulation (p<0.05, see methods) are colored red if activated and blue if inhibited. (L) Fold-change in the fraction of neurons that showed a significant excitatory response in the two seconds following stimulation in ChR+ fish (n=6) compared to ChR– fish (n=4) in several regions both ipsilateral and contralateral to the site of stimulation (top). The analogous plot for inhibitory responses (bottom). The ipsilateral vHb showed a significant activation (p=0.00002). See also Figure S6D. (M) Analogous to panel L for the period 8–10s following stimulation. The ipsilateral raphe showed a significant increase in the number of inhibitory responses during this period (p=0.004), as did the contralateral dHb (p=0.048). See also Figure S6E.

Figure 7:. RNN model of cellular-resolution imaging data reveals candidate circuit mechanisms for Hb recruitment and action.

(A) For each fish an RNN model was composed of N_wb model neuronal units (green spheres), equal to the number of neurons recorded. Connectivity is represented by the weight matrix J (lower right; green lines represent specific weights, J_ij; noise and shock inputs not schematized). All weights are subject to plasticity (learning rule, right panel) based on the difference between each model unit’s activity and a target function derived from experimentally measured neural activity. Signals from fish in different conditions (control, ctrl; shocked, shk; re-exposed, reshk) and from different time periods within the experiments (baseline period, denoted e1 for epoch 1; challenge period, e2; passive period, e3) were used to train separate models, each resulting in a connectivity matrix, denoted by J^{M, condition, epoch} (e.g. J^{M, shk, e3}). Training was performed using the recursive least squares learning rule. (B) Mean square error between RNN activity and target functions converges with training. (C) Snippets of activity from individual model neurons (red) compared to experimental data (blue) for control (left) and BC (right). (D) Principal component analysis was performed on both experimental data from the challenge period (all Hb and raphe neurons from representative control and shocked fish) and on the output from the associated models (i.e. networks with connectivity J^{M, ctrl, e2}, J^{M, shk, e2}). The population activity was projected on to the largest 3 principal components for control (top two panels) and shocked (bottom two panels; shock times indicated with orange dots) fish. (E) Log distribution of connectivity in a representative shocked (J^M,shk,e3, blue) and control (J^M,ctrl,e3,black) model. Analogous results for models trained to match data from the optogenetic imaging experiments are in Figure S7. (F) Analogous to panel E, but for sub-matrices within J representing specific intra- and inter-region connections (see lower right subpanel in A). (G) Effect of BC on projection-specific connectivity. The percentage change in the standard deviation of projection-specific connectivity strength distribution (see panel F) between models trained on the passive period and the baseline period. This change was measured for all control (black, n=5), shocked (blue, n=5) and re-exposed (teal, n=5) fish presented in Figure S3D (vertical grey lines indicate std across fish). Intra-Hb connectivity undergoes a significant change in shocked and re-exposed fish compared to control (p = 0.007 and 0.002, respectively), as does the raphe-to-Hb projection (p = 0.008 and 0.003, respectively); see also Figure S7A. Analogous analysis based on the 1^st and 3^rd moments (mean and skewness) showed no significant changes.