Movement magnitude is represented by rapid HON dynamics

A, Schematic depicting the photometry system and voluntary wheel-running apparatus with video capture (left) and representative expression (right) of AAV1-hORX-GCaMP6s in HONs. LHA; lateral hypothalamic area; 3V; third ventricle; VMH; ventral medial hypothalamus. B, Processing pipeline to generate a “movement metric” from video recording of voluntary movement. C, Whole-session correlation of HON photometry with simultaneously recorded behavior metrics from n = 15 mice: running (r = 0.50 ± 0.03), the movement metric (r = 0.68 ± 0.04), the movement metric in epochs where locomotion was < 1 cm·s−1 (r = 0.54 ± 0.03). Metrics given as mean ± SEM, connected lines represent individual mice. Asterisks indicate comparison between running and movement metric (paired t-test: t14 = 6.676, **** p < 0.0001). D, Traces from an example experiment. From top to bottom: running on the wheel, the video-derived movement-metric, HON GCaMP6s photometry, and isosbestic control. Shaded beige regions indicate non-locomotor epochs where the movement metric still reported strong positive correlations with HON activity.

HON dynamics represent movement magnitude across classified behaviors

A, Diagram for classifying behaviors from video recordings via deep learning. B, Example experiments showing movement (upper, black) and color-coded photometry (lower) after behavioral classification. C, Average normalized HON GCaMP6s signal across classified behaviors in n = 15 mice (rmANOVA: F4,56 = 90.556, **** p < 0.0001). D, Average value of the normalized HON GCaMP6s signal from n = 15 mice plotted against the average value of the movement metric in each classified behavior. Thick bars represent SEM across both metrics. The dotted line represents an idealized 1:1 linear relationship between movement and photometry with an intercept at zero. E, Comparison of three linear mixed effects models using movement and/or behavioral classification to predict HON GCaMP6s signal, quantified by Akaike information criterion. F, Average behavioral transition matrix for n = 15 mice. G, Average movement metric aligned −4 to +4 seconds from a behavioral transition. The six most frequent behavioral transitions are plotted. Lines and shaded region represent mean and SEM. H, Same as G, but with hORX-GCaMP6s signal.

Movement magnitude is phase-aligned to HON activity across the frequency domain

A, Representative empirical mode decomposition (see methods) of HON GCaMP6s activity from one experiment. The top green trace displays a summation of all IMFs. Below are the first 10 IMFs sorted by characteristic frequency in Hz. Relative power is reported as a percentage of the total. Dashed grey box (middle right) represents a cutout of the 0.031 Hz IMF in which movement epochs (shaded beige regions) are clearly phase-aligned in the 0-π period. B, Average power plotted against characteristic frequency of n = 229 GCaMP6s-derived IMFs derived from 27 experiments using 15 mice. Thick line and shaded region represent a local regression and 95% CI. Boxplots represent the maximum power IMF from each experiment. C, Same as B, using n = 214 movement metric-derived IMFs. D, Left: Preferred phase of both active (black) and quiescent (grey) epochs of the movement metric plotted against characteristic frequency. Dots represent n = 229 GCaMP6s-derived IMFs derived from 27 experiments using 15 mice. Thick line and shaded region represent a circular regression and 95% CI. Right: average absolute value of the z-scored movement metric plotted on the radial axis against the phase of the maximum-power IMF. Lines represent mean ± SEM from n = 15 mice. E, Coupling strength of the movement metric to GCaMP6s IMFs in both active (black) and quiescent (grey) epochs plotted against the IMF’s characteristic frequency. Dots represent n = 229 GCaMP6s-derived IMFs derived from 27 experiments using 15 mice. Thick line and shaded region represent a local regression and 95% CI.

HONs are not required to maintain movement profiles

A HON ablation in HON-DTR+ mice. B, Brain slice histology of a wild-type mouse (WT, left, image representative of n = 11 mice) and a HON-ablated mouse (DTR+, right, image representative of n = 12 mice) stained for OrexinA (red). 3V = 3rd ventricle. C, Average power plotted against characteristic frequency of movement metric-derived IMFs (WT: n = 217 IMFs from 30 experiments using 11 mice. DTR+ n = 267 IMFs from 34 experiments using 12 mice). D, Diagram of K-Means clustering to generate two clusters of movement representing voluntary initiation motion following > 4s of no motion, see methods. E, Proportion of bouts which were assigned “small” or “large” were not significantly different across genotypes (Unpaired t-test, t = 1.1167, p = 0.2767).

Movement-related orexin peptide release is non-homogenous across projection targets

A, Schematic representing targeting of the orexin sensor OxLight1 to the LC and SNc, representative histology from n = 7 mice. B, Example experiment displaying the movement metric and OxLight1 signal in the LC (blue) and SNc (purple) in the same mouse. Shaded regions represent movement epochs. C, Mouse-averaged whole-session correlation of movement with OxLight1 photometry in the SNc and LC (paired t-test: t6 = 2.068, p = 0.0841). D, Preferred phase of individual EMD-derived OxLight1 IMFs to active epochs of the movement metric plotted as a function of the IMF’s characteristic frequency. n = 317 IMFs from 21 experiments using 7 mice. Thick line and shaded region represent a circular regression and 95% CI. E, Coupling strength of individual EMD-derived OxLight1 IMFs to active epochs of the movement metric plotted as a function of the IMF’s characteristic frequency. n = 317 IMFs from 21 experiments using 7 mice. Thick line and shaded region represent a local regression and 95% CI. F, Photometry baselined −3 to −1 seconds before movement initiation in two clusters: small movements (left) and large movements (right). Lines and shaded regions represent mean ± SEM of n = 6 mice. Asterisk represents a paired statistical comparison of mean signal 3 to 4 seconds after movement initiation (Wilcoxon test, D6 = 1.0, * p = 0.03125).

HONs multiplex movement and blood glucose across frequency domains

A, Correlation of movement with HON photometry across three states of fasting (n = 15 mice). Fed: ad-libitum access to chow. Re-fed: 18-hour fasted and then food returned 2 hours before experiment. Fasted: 18-hour fasted mice (rmANOVA: F2,28 = 1.848, p = 0.176). B, Scheme of experimental setup including photometry, glucose telemetry (DSI), video capture, and intragastric catheter infusions 23. C, Example HON GCaMP6s photometry trace (top), three derived IMFs (middle), and simultaneously recorded physiological variables of blood glucose (plotted as the first derivative) and movement magnitude (lower). A vertical dashed orange line represents an intragastric infusion of glucose. D, A linear model was fit to predict n = 94 IMFs using the glucose derivative and movement. Fitted coefficient weights are plotted as a function of the IMF’s characteristic frequency. E, Per-mouse (n = 6) binned coefficient weights. For the higher frequency bin (left, 10−1 to 10−2 Hz), coefficient weights were tested against a population mean of zero (Bonferroni-corrected one-sample t-test: glucose derivative t5 = 1.965, p = 0.2131; movement t5 = 14.102 p < 0.0001). Coefficient weights of glucose derivative were compared to movement (paired t-test: t5 = 8.331, p = 0.0004). For the lower frequency bin (right, < 10−3 Hz), coefficient weights were similarly tested against a population mean of zero (Bonferroni-corrected one-sample t-test: glucose derivative t5 = −5.631, p = 0.0049; movement t5 = 1.963, p = 0.2139). Coefficient weights of glucose derivative were compared to movement (paired t-test: t5 = 5.385, p = 0.0030). Bar plots display mean ± SEM. F, Alternate visualization of E with coefficient weights plotted against each other. G, Left: correlations of HON GCaMP6s signal and the movement metric across blood-glucose quartiles in an example mouse. Right: Regression predicting HON GCaMP6s activity using movement and blood-glucose percentile, performed per-mouse (interaction effect; one-sample t-test: t5 = 0.6250, p = 0.5594). Bar plot displays mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 and ns, not significant by two-tailed tests.

Disentangling contributions of arousal vs movement to HON activity

A, Scheme of pupillometry processing pipeline. B, Example aligned recording of HON GCaMP6s photometry (green), movement metric (black), pupil size (dark blue), and ocular movement (light blue). Right inset shows correlation of pupil size and ocular movement with HON photometry (n = 7 mice). C, Partial least squares regression fit to predict HON GCaMP6s photometry from the movement metric, pupil size, and ocular movement. Percent contribution of each feature is quantified using drop-one feature analysis (see methods). D, Empirical mode decomposition of HON GCaMP6s IMFs to quantify the frequency of HON activity containing the maximum correlation with each feature. Movement metric maximally correlated with HON frequencies that were higher than that of pupil size, but lower than that of ocular movements (Bonferroni-corrected paired t-test: against pupil size t6 = 3.4199, * p = 0.0283; against ocular movements t6 = 4.6059, ** p = 0.0073). E, Local regression displaying coefficient weights from a partial least squares regression fit to individual HON GCaMP6s IMFs, plotted as a function of the IMF’s characteristic frequency.

Representation of movement across genetically defined neurons

A, C57BL/6J-derived strains of mice permitted Cre-dependent photometry recordings from four separate neural populations (MVeGLUT, n = 6; VMHSF-1, n = 8; LCNA, n = 20; and SNcDA, n = 8 mice). B, Representative histology from each recording site. LVe; lateral vestibular nucleus; MVe; medial vestibular nucleus; Pr; nucleus prepositus; 4V; fourth ventricle; VMH; ventral medial hypothalamus 3V; third ventricle; LC; locus coeruleus; VTA; ventral tegmental area; SNc; substantia nigra pars compacta; SNr; substantia nigra pars reticulata. C, Whole-session correlation of movement with photometry across neural subtypes. Asterisks represent one sample t-tests against mean of zero, after Bonferroni correction. Additional comparison is shown between LHA (HONs, n = 15 mice) and MVe neurons (unpaired t-test: t19 = 0.209, p = 0.8366). D, Cross correlation displaying lag time of each recorded population. Negative values imply photometry precedes movement. Lines and shaded regions represent mean ± SEM. E, Photometry baselined −3 to −1 seconds before movement initiation in two clusters: small movements (left) and large movements (right). Lines and shaded regions represent mean ± SEM. F. Correlation of movement and photometry during initiation of large movements. Asterisks represent one sample t-tests against mean of zero, after Bonferroni correction. Additional comparison is shown between HONs (LHAORX) and MVeGLUT (unpaired t-test: t19 = 0.877, p = 0.3915). G. Correlation of movement and photometry during initiation of small movements. Asterisks represent one sample t-tests against mean of zero, after Bonferroni correction. Additional comparison is shown between LHAORX test: t19 = 2.864, p = 0.0099). (HONs) and MVe GLUT (unpaired t- *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 and ns, not significant by two-tailed tests.