Ancestral Circuits for the Coordinated Modulation of Brain State

Abstract

Internal states of the brain profoundly influence behavior. Fluctuating states such as alertness can be governed by neuromodulation, but the underlying mechanisms and cell types involved are not fully understood. We developed a method to globally screen for cell types involved in behavior by integrating brain-wide activity imaging with high-content molecular phenotyping and volume registration at cellular resolution. We used this method (MultiMAP) to record from 22 neuromodulatory cell types in behaving zebrafish during a reaction-time task that reports alertness. We identified multiple monoaminergic, cholinergic, and peptidergic cell types linked to alertness and found that activity in these cell types was mutually correlated during heightened alertness. We next recorded from and controlled homologous neuromodulatory cells in mice; alertness-related cell-type dynamics exhibited striking evolutionary conservation and modulated behavior similarly. These experiments establish a method for unbiased discovery of cellular elements underlying behavior and reveal an evolutionarily conserved set of diverse neuromodulatory systems that collectively govern internal state.

Keywords: Brain state; Calcium imaging; Mice; Neuromodulation; Volume registration; Zebrafish.

Figure 1. Variability in Sensorimotor Reaction Time Reflects the Internal State of Alertness in Larval Zebrafish

(A) Schematic of behavioral apparatus and example data. Tethered zebrafish are presented with a looming dot stimulus, and tail movements are recorded with an infrared (IR) camera. (Right) Trials in an example fish. Black dots indicate escape onset. (B) Histogram of RTs for an example fish. (C) Schematic of combined behavior and heart-rate monitoring with two-photon cardiac imaging in Tg(cmlc2:GFP) fish. (Right) Example two-photon images of GFP+ heart when dilated and constricted (each image is an average of six frames). White brackets indicate ventricle cross section. (D) Example of heart-rate time series together with behavior across two trials. (E) Summary data, n = 6. See Figure S1A for individual fish example. Solid line: linear regression; shaded region: 95% confidence interval. See also Figure S1.

Figure 2. Whole-Brain Cellular-Resolution Registration of Activity and Molecular Features with MultiMAP

(A) Schematic of sample-handling pipeline. (B) Schematic of data-handling pipeline. (C) Orthogonal planes (XY, YZ, and XZ) and x, y zoom of signal overlays from a single fish. White lines indicate location of cross section. The cell types at bottom right are the cholinergic tegmentum (cyan), the serotonergic dorsal raphe nucleus (red), and the noradrenergic locus coeruleus (yellow). Scale bars: 100 μm (left) and 25 μm (right). See Figure S2 and Movie S1 for additional details and examples. (D) Spinal projection neurons (SPNs) in Tg(elavl3:H2B-GCaMP6s) fish are back-labeled through injection of Texas Red Dextran. Scale bar: 50 μm. (E) Schematic of data-handling pipeline, where GCaMP volumes are used for registration, and co-imaged SPN volumes are used to assess accuracy. (F) Z-projections through the midbrain and hindbrain of live (red) and fixed (cyan) SPN samples are overlaid before (top) or after (bottom) registration of GCaMP volumes. Scale bar: 50 μm. (G) Mean distance between matched cell centers (STAR Methods) before and after registration. Bars are mean ± SEM. Two-tailed Wilcoxon signed-rank test, n = 4 fish. ***p < 0.001. See also Figure S2, Movie S1, and STAR Methods.

Figure 3. Diverse Neuromodulatory Cell Types Are Correlated with Alertness States

(A) Example cart+ tegmentum neuron, with pre-stimulus time series and onset to first escape-like tail movement for each trial (black dots), ordered by RT. (B) Example data from individual th+ locus coeruleus neuron (top) and cart+ tegmentum neuron (bottom). (Left) Overlays of antibody label (red) with live GCaMP (green) in example Z plane after volume registration (neurons identified with white arrow). Scale bars: 100 μm. (Right) Scatterplot of mean pre-stimulus neural activity and RT. Solid line: linear regression; shaded region: 95% confidence interval. (C) Summary data from 22 neuromodulatory cell types, displaying correlation coefficients from comparison of RT and pre-stimulus neural activity. Bars are mean ± SEM, n = 34 fish. Significance values are determined by one-sample Wilcoxon signed-rank tests and false discovery rate correction for multiple comparisons. Groups are colored according to their correlations: blue (negative correlation), green (positive correlation), or black (no correlation). See Figure S3 for details and examples of each cell type and Figure S4 for additional functional characterization. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 4. Correlated and Cooperative Activity of Neuromodulatory Cell Types

(A) Schematic of analysis. Correlations are assessed between time series recorded from neurons classified as alertness+ or alertness−. (B) Simultaneous recordings from multiple neuromodulatory cell types in an example fish. d: dorsal, v: ventral. Each line is one neuron. (C) Cumulative density plot of correlations between alertness+/alertness+ pairs (magenta), alertness+/alertness− pairs (cyan), and alertness−/alertness− pairs (gray). n = 6 fish, with bar plot of same data as inset. Two-sided Kolmogorov-Smirnov tests, +/+ versus +/−, D = 0.407, p < 0.01; +/+ versus −/−, D = 0.436, p < 0.005; +/− versus −/−, D = 0.087, p = 0.57. (D) Correlations between cell types measured in trials with fast (bottom 25^th percentile) versus slow (top 25^th percentile) RTs. Mean ± SEM, paired t tests. Color scheme as in (A) and (C). (E) Correlation between true RT and RT predicted from linear regression using simultaneously recorded alertness+ (black) and alertness− (gray) neurons (permutation tests between + and − and corrected for multiple comparisons). (F) Correlation between true RT and RT predicted from linear regression using one to three distinct neuromodulatory cell types as inputs. Alertness+ cells: n = 34, 26, and 6 for 1, 2, and 3 cell types, respectively, one-way ANOVA, F_(2,65) = 8.05, p < 0.001; alertness− cells: n = 32, 25, and 6 for 1, 2, and 3 cell types, respectively, one-way ANOVA, F_(2,62) = 2.75, p > 0.05. Mean ± SEM. Color scheme as in (E). *p < 0.05, **p < 0.01, ***p < 0.001. See also Figures S3 and S4.

Figure 5. Conservation of Cell-Type-Specific State-Related Activity in Mammals

(A) Schematic of mouse auditory RT task. (B) Example recording from GCaMP+ CART neurons in the Edinger-Westphal nucleus of CART-IRES2-Cre mice, with pre-stimulus time series and onset to first lick for each trial, ordered by RT (black dots denote licks). (C) Example recordings from the locus coeruleus of a TH-IRES-Cre mouse (top) and the Edinger-Westphal nucleus of a CART-IRES2-Cre mouse (bottom). (Left) Location of recording region and co-labeling of antibody label with GCaMP expression. (Right) Scatterplot of mean pre-stimulus neural activity and RT. Solid line: linear regression; shaded region: 95% confidence interval. Scale bars: 100 μm. (D) Summary data of recordings from seven cell types displaying correlation coefficients from comparison of RT and pre-stimulus neural activity. Mean ± SEM, n = 29 mice (n = 3, 3, 3, 3, 4, 6, 3, and 4, from left to right). Significance values were determined by permutation tests with the eYFP control group and false discovery rate correction for multiple comparisons. (E) Summary data from zebrafish neuromodulatory cell types homologous to the cell types recorded from mice in (D) (see Figure S5). Cell types are presented in the same order as (D), without the eYFP control group. Data are taken from Figure 3C. *p < 0.05, **p < 0.01, ***p < 0.005. See also Figures S5 and S6.

Figure 6. Behavior-Dependent Effects of Cell-Type Manipulation

(A) Schematic of mouse behavioral task. (B and C) Summary data for RT behavior displaying the mean RT in light ON trials normalized to light OFF trials. Bars are mean ± SEM. (B) n = 27 mice (n = 5, 3, 3, 4, 3, 3, 3, and 3, from left to right). (C) n = 18 mice (n = 4, 3, 4, 3, and 4, from left to right). D) Raw video images and pupil traces for an example CART-IRES2-Cre mouse, with ChR2 stimulation of the Edinger-Westphal nucleus. Mean pupil trace in black. Red dotted line indicates pupil perimeter. (E and F) Summary data for pupil size displaying the mean size in light ON trials normalized to light OFF trials. Bars are mean ± SEM. (E) n = 27 mice (n = 5, 3, 3, 4, 3, 3, 3, and 3, from left to right). (F) n = 18 mice (n = 4, 3, 4, 3, and 4, from left to right). For data in panels (B), (C), (E), and (F), significance values were determined by permutation tests with the eYFP control group and false discovery rate correction for multiple comparisons. *p < 0.05. See also Figures S5 and S6.