Construction and temperature control of the ThermoMaze.

A) Schematic of the ThermoMaze. The floor was built using 25 Peltier elements attached to water cooling block heatsinks (building block). The position of the animal and the temperature of the ThermoMaze can be recorded using a video camera and an infrared camera positioned above the box, respectively. An ‘X’ was taped inside the maze as an external cue below the camera synchronizing LED. Water circulates through the water cooling heatsinks using a water pump submerged in a water tank (one row of heatsinks is attached to one pump). The temperature of the water tank is monitored and recorded using a thermocouple (white symbol inside water tank, DAQ – analog input of the data acquisition system). Peltier elements are connected to a power supply (red and blue dots represent the anode and cathode connection). B) Circuit diagram and schematic of Peltier elements (n = 25), viewed from the top. TTL pulses generated by an AVR-based microcontroller board (Arduino Mega 2560) close a relay switch connected to a variable voltage power source. Each Peltier element can be independently heated (surface temperature depends on applied voltage and temperature difference between hot and cold plate of Peltier element). C) Schematic of the water circulation cooling system, viewed from the bottom of the floor (each Peltier element has its own water-cooling aluminum heatsink, shown in silver, n = 25). Five submerging DC pumps are used to circulate water across 25 heatsinks (dashed lines show the Peltier elements connected to one pump). The temperature of the heatsink is transferred to the Peltier element passively through the silver epoxy resulting in passive cooling of the floor of the ThermoMaze.

Calibration of the ThermoMaze temperature regulation.

A) Side view of the ThermoMaze. Prior to animal experiments, we calibrated the heating and cooling performance of the Peltier elements and temperature measurement. We attached thermocouples (white symbols) to the surface of the Peltier elements serving as the ground-truth for calibrating the infrared camera placed above the ThermoMaze. Different voltage levels were used for the calibration (2.2, 2.4, 2.6, 2.8 and 3V) while the water tank temperature was kept constant. B) Top: four Peltier elements used in later experiments are chosen for calibration (four corners). Bottom: one minute heating was repeated four times at each voltage level. C) Simultaneously recorded temperature by thermocouples (left) and infrared camera (right). Increasing voltages induced increased heating (n = 4 trials per intensity, mean ± SD are shown). While the temporal dynamics yielded similar results between the two systems, we found ∼4 °C offset between infrared and thermocouple-measured signals. D) Temperature changes of four Peltier elements used during an emulated behavioral session (without any animal subject) tracked by thermocouples. E) Temporal dynamics of temperature changes at the four Peltier elements during active heating and following passive cooling. The temperature reaches steady state within 31 ± 10.3 seconds (mean ± SD, n = 4 trials across 4 Peltier elements).

Mice track and stay immobile on hidden warm spots in the ThermoMaze.

A) Five sub-sessions constituted a daily recording session: (1) rest epoch in the home cage, (2) pre-cooling exploration epoch (Pre), (3) Cooling, (4) post-cooling exploration epoch (Post) and (5) another rest in the home cage. B) Schematic of temperature landscape changes when the animal is in the ThermoMaze (top) and example animal trajectory (below). During Cooling, one Peltier element always provided a warm spot for the animal (four Peltier elements in the 4 corners were used in this experiment). Each Peltier element was turned on for 5 minutes in a sequential order (1-2-3-4) and the sequence was repeated four times. C) Session-averaged duration of immobility (speed ≤ 2.5 cm/s) that the animal spent at each location in the ThermoMaze; Color code: temporal duration of immobility (s); white lines divide the individual Peltier elements; n = 17 session in 7 mice). D) Cumulative distribution of animal speed in the ThermoMaze during three sub-sessions from 7 mice). Median, Kruskal–Wallis test: H = 139304.10, d.f. = 2, p < 0.001. E) Animal’s distance from the previously heated Peltier element site. F) Speed of the animal centered around warm spot transitions. G) Animal’s distance from the target warm spot as a function of time (red curve: median; time 0 = onset of heating). *p < 0.05, **p < 0.01, ***p < 0.001. In all panels, box chart displays the median, the lower and upper quartiles. (see Supplementary Table 2 for exact p values and multiple comparisons).

Location-specific distribution of SPW-R in the ThermoMaze

A) Spatial map of the number of SPW-Rs during the Cooling sub-session averaged across all sessions (Color code: average number of SPW-Rs per session at each location). Session-average number of SPW-Rs during Cooling was 627.3 (corresponding to 0.136 Hz). B-D) Boxplots of SPW-R properties in ThermoMaze and in the home cage (n = 19 sessions in n = 7 mice). B) Mean ripple duration in seconds (s; p = 0.108). C) Mean ripple amplitude in μV (p = 0.9). D) Mean ripple peak frequency in Hz (p < 0.001). Dots (females) and diamonds (males) of the same color represent the same animal.

Spikes of CA1 pyramidal neurons during awake SPW-Rs are spatially tuned.

A) Within SPW-R firing rate maps (ThermoMaze divided into four quadrants) of 6 example cells with high within SPW-R spatial tuning score (STS; from left to right, top to bottom, STS= 0.458, 0.639, 0.592, 0.672, 0.655, 0.660 respectively). Color represents within SPW-R firing rate (in Hz) of the neuron in each quadrant of the ThermoMaze. B) Cumulative distribution of spatial tuning scores of pyramidal neurons (top; n = 1150; p < 0.001) and interneurons (bottom; n = 288; p < 0.001) during SPW-Rs. Chance levels were calculated by shuffling the quadrant identity of the SPW-Rs. One-sided Wilcoxon rank sum tests. C) Bayesian decoding of the mouse’s location (quadrant of the ThermoMaze) from spike content of SPW-Rs in an example session (blue: actual ripple location; green: decoded locations; red: locations of the warm spot; session decoding accuracy = 0.65; chance level = 0.26). D) Histogram of session Bayesian decoding accuracies of ripple locations using spiking rate maps constructed during ripples as templates (with a uniform prior and a 100-fold cross-validation; P < 0.001). One-sample t-test. E) Firing rate ratios of pyramidal cells constructed during SPW-Rs and movement are positively correlated (Pearson’s r = 0.321, p < 0.001). The firing rate ratio measures the firing rate of a cell in one quadrant versus the sum of its firing rates in all four quadrants under a specific condition (within-ripple or during movement). F) Matrix of the pairwise correlation coefficient between each pair of firing rate ratio population vectors constructed during SPW-Rs and movements in different quadrants (x and y axes). Color represents Pearson’s r.

Mice sleep at experimenter-defined locations.

A) Schematic of ThermoMaze with warm spot locations (top) and the trajectory of an example animal (bottom; red rectangles correspond to the location of warm spots). During Cooling, one Peltier element was turned on for 20 min followed by another (1-2) and the sequence was repeated two times. B) Session-averaged duration of immobility (speed ≤ 2.5 cm/s) at each location in the ThermoMaze; white lines divide the individual Peltier elements (n = 7 sessions, n = 4 mice). C) Spatial distribution of SPW-R occurrences (color code: average number of SPW-Rs per session at each location, n = 7 sessions, n = 4 mice). Session-average of SPW-Rs during Cooling was 775 (corresponding to 0.16 Hz). D) Long duration of heating allowed for NREM sleep occurrence during Cooling sub-session in an example session. Brain state changes45 are shown together with SPW-Rs (green ticks). Note that NREM sleep occurs in the second half of the 20-min warming. E) Mice spent a larger fraction of time in NREM during 20 min Cooling sub-session compared to the 5 min task variant (p = 0.003, n = 19 sessions in 7 mice and n = 7 sessions in 4 mice). (F). Mice typically spent ∼1000 seconds awake between NREM epochs. G) Box charts of Pearson’s correlation coefficients between population vectors of CA1 pyramidal neurons constructed during awake SPW-Rs, movement, and NREM SPW-Rs. Median, Kruskal–Wallis test: H = 20.7, d.f. = 2, p < 0.001 (pairwise comparison: *p = 0.037 and ***p = 1.6x10-05).