A microfluidic-induced C. elegans sleep state

Abstract

An important feature of animal behavior is the ability to switch rapidly between activity states, however, how the brain regulates these spontaneous transitions based on the animal's perceived environment is not well understood. Here we show a C. elegans sleep-like state on a scalable platform that enables simultaneous control of multiple environmental factors including temperature, mechanical stress, and food availability. This brief quiescent state, which we refer to as microfluidic-induced sleep, occurs spontaneously in microfluidic chambers, which allows us to track animal movement and perform whole-brain imaging. With these capabilities, we establish that microfluidic-induced sleep meets the behavioral requirements of C. elegans sleep and depends on multiple factors, such as satiety and temperature. Additionally, we show that C. elegans sleep can be induced through mechanosensory pathways. Together, these results establish a model system for studying how animals process multiple sensory pathways to regulate behavioral states.

Fig. 1

Quiescence dynamics are strongly affected by microfluidic environments. a Characteristic worm quiescence in a microfluidic chamber. (Left) 1-h activity trace from a worm swimming in a microfluidic chamber, calculated by subtracting consecutive frames to quantify movement (see “Methods” section). Quiescence is hallmarked by a clear drop in animal activity to near zero. (Right) 1-s overlap of frames shows the animal swimming during the wake period (W) and immobile during quiescence (Q). b All quiescent bouts recorded from animals in a small microfluidic chamber (50 μm width), large microfluidic chamber (500 μm width), WorMotel multi-well PDMS device, and large microfluidic device during a 30 min heat shock at 30 °C (see inset images). Raster plots show the bouts recorded from each animal during a 4 h imaging period (n = 47 animals for each condition). Experiments represented in the raster plots are sorted by the onset of the first detected sleep bout. c–f Quantitative sleep metrics. Sμ small microfluidic chambers, Lμ large microfluidic chambers, WM WorMotel, HS first hour of heat shock. c The total fraction of time each animal spent in the quiescent state. d Onset time of the first quiescent bout. e Length of individual quiescent bouts. f Length of individual wake bouts, excluding the first period of wake after animals are loaded into the chamber. (n = 47 animals for each condition; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, *****p < 0.00001; Kruskal–Wallis with a post hoc Dunn–Sidak test). Source data is available as a Source Data file

Fig. 2

Sleep states show stereotypical posture, reversibility, and a decreased response to weak stimuli. a C. elegans show increased body curvature during sleep. (Left) Example of behavioral activity and normalized body curvature for a single animal. Dotted line indicates sleep threshold. (Right) Average body curvature during wakefulness and sleep across all animals (from Large microfluidic data in Fig. 1, n = 47 animals, ****p < 0.0001, unpaired t-test). b The sleep state is reversible. (Left) Heatmaps of behavioral activity (top) and nose speed (bottom) for quiescent animals that received a 5 s light pulse at t = 30 s. (Right) Mean behavioral activity (top) and nose speed (bottom) of each animal before and after stimulation. Dotted line indicated sleep threshold. (n = 13 worms, *****p < 0.00001, paired t-test). c–e Quiescent animals have a decreased response to weak sensory stimuli. c Heatmaps of activity from awake and sleeping animals that received strong or weak mechanical stimuli from microfluidic valves (Supplementary Movie 4, 5). For each condition, only the 25 trials with the highest mean activity post stimulation are shown. “Wake” indicates trials in which animals were awake but have a below-average activity before stimulation. “Sleep” indicates trials in which animals were quiescent before stimulation. Heatmaps contain an ~2 s long stimulation artifact beginning at 10 s due to movement from microfluidic valves. d Mean behavioral activity before and after mechanical stimulation from the trials in c. Dotted line indicates sleep threshold. In all cases, average activity significantly increases after the stimulation (largest p-value = 0.002, paired two-sided t-test). However, the behavioral activity after stimulation is significantly different when quiescent animals received weak mechanical stimuli (Error bars are sem; Strong-Wake n = 108, Strong-Sleep n = 51, Weak-Wake n = 176, Weak-Sleep n = 29; ****p < 0.0001, ns not significant, Kruskal–Wallis with a post hoc Dunn–Sidak test). e Fraction of animals awake following mechanical stimuli (Error bars are standard deviation, calculated by bootstrapping each data set with 5000 iterations; ns not significant, *****p < 0.00001; significance was calculated by data resampling 5000 iterations and a post hoc Bonferroni correction). Source data is available as a Source Data file

Fig. 3

Microfluidic-induced sleep behavior depends on ALA and RIS neurons. a Raster plots of detected sleep bouts from WT, ceh-17(lf) and aptf-1(lf) animals. Only the top 50 animals showing the most total microfluidic-induced sleep are shown. b Both ceh-17(lf) and aptf-1(lf) show less total sleep than WT. The data suggest that microfluidic-induced sleep is strongly dependent on the ALA and RIS neurons. (WT n = 57, ceh-17(lf) n = 57, aptf-1(lf) n = 60; *****p < 0.00001 compared with WT, Kruskal–Wallis with a post hoc Dunn–Sidak test). Source data is available as a Source Data file

Fig. 4

A global brain state transition governs C. elegans microfluidic-induced sleep. a Adult animal immobilized in a microfluidic chamber tailored for whole-brain imaging. (Inset) Single-plane epifluorescence image of an animal with pan-neuronal expression of GCaMP6s. b Representative animal shows behavioral quiescence correlates with less neural activity. (Top) Behavioral activity trace quantified by tracking the motion of ten individual neurons (see “Methods” section). (Middle) Average fluorescence across the whole worm head ganglia. (Bottom) Activity of ten individual neurons show a clear brain-state transition and less neural activity during microfluidic-induced sleep. Arrow indicates the circled neuron in a, which increases in activity during sleep. c Using only behavioral activity, we identified quiescent bouts then quantified neural activity during sleep and wake. During microfluidic-induced sleep, animals exhibited a large-scale downregulation of neural activity across both the entire ganglia and most individual neurons. The neurons chosen for analysis were randomly selected from the field-of-view and are not necessarily the same neurons across every animal. (n = 7 animals; ten neurons were tracked per animal; ***p < 0.001, *****p < 0.00001, paired t-test). Source data is available as a Source Data file

Fig. 5

The RIS neuron is more active during microfluidic-induced sleep. a Fluorescent micrograph of the mKate channel, showing HBR1361 animals immobilized in 50 μm-wide chambers and expressing mKate and GCaMP3 in the RIS neurons. Inset is a zoom-in on a single RIS neuron (scale bar 50 μm). b Representative traces of animal behavioral activity (top) and RIS calcium activity (bottom). RIS activity dramatically increases during microfluidic-induced sleep bouts (shaded regions), opposed to the majority of worm brain activity (see Fig. 4). c RIS is more active during microfluidic-induced sleep. We automatically detected sleep bouts across animals, calculated mean RIS activity during wake and microfluidic-induced sleep, and quantified mean RIS activity in each behavioral state. Dashed line shows the average RIS activity for animals that did not display a sleep bout. (Data points represent individual animals; n = 48 animals total, n = 31 animals exhibited at least one sleep bout, n = 17 animals did not sleep; *****p < 0.00001, paired t-test). d RIS ∆R/R activity negatively correlates with animal behavior. This correlation was not seen in shuffled ∆R/R data or when the mKate channel only was compared with behavior, indicating that RIS correlation with behavior is not a result of movement artifacts. Source data is available as a Source Data file

Fig. 6

Microfluidic-induced sleep is regulated by satiety and multiple sensory circuits. a Detected sleep bouts for WT animals in several experimental conditions. Raster plots show detected sleep bouts during a 2 h imaging period. “Baseline” indicates the standard experimental conditions: 500 μm chamber width, no food in the buffer, and a 22 °C temperature. “Starved” indicates animals that were starved prior to the assay. “+Food” indicates conditions in which E. coli OP50 was added into the buffer during recordings. “+Heat” indicates imaging conditions where the temperature was raised to 25 °C. “+Compression” indicates that animals were partially immobilized in 50 μm-wide chambers. See micrographs under d for chamber geometries. In all cases, the sleep phenotype varies dramatically from the baseline. Only the 55 animals that displayed the most sleep are plotted for clarity. b Total WT sleep under varying satiety conditions. As satiety increases from Starved to +Food, animals exhibit less microfluidic-induced sleep (from left to right on the plot the number of animals n = 55, 68, and 67). c Total microfluidic-induced sleep under varying temperature conditions. Increasing temperature increases total microfluidic-induced sleep for WT animals. Thermosensory-defective mutants show the same microfluidic-induced sleep phenotype as WT at 18 °C, but significantly less sleep at 22 and 25 °C, indicating that thermosensory input can act to drive or suppress microfluidic-induced sleep (from left to right on the plot the number of animals n = 37, 68, 71, 41, 67, and 60). d Total sleep under different confinement conditions. Micrographs show chamber geometries. When WT animals are confined in smaller chambers, they only show an increase in total microfluidic-induced sleep when slightly compressed in 50 μm-wide chambers. mec-10(tm1552) mutants show an identical phenotype to WT in 500 μm and 110 μm chambers, but dramatically reduced sleep compared with WT when compressed. These results suggest that nociceptive input to mechanosensory neurons regulates microfluidic-induced sleep (from left to right on the plot the number of animals n = 68, 64, 69, 66, 64, and 57). (ns = not significant, **p < 0.01, ***p < 0.001, ****p < 0.0001; Kruskal–Wallis with a post hoc Dunn–Sidak test). Source data is available as a Source Data file