Precision cutaneous stimulation in freely moving mice
- Wolfson Institute for Biomedical Research, University College London, United Kingdom
Figures
Figure 1
Closed-loop cutaneous stimulation of mice freely moving in naturalistic environments.
(A) Mice can be remotely targeted with cutaneous stimuli while freely exploring complex environments, such as a maze. (B) Schematic illustrating the closed-loop control workflow. A freely moving mouse is recorded using a camera feed, enabling real-time pose estimation to track multiple body part keypoints. The extracted frame keypoint (x, y) of a selected body part is converted to pre-mapped x, y mirror galvanometer control signals to steer the laser beam paths and pulse the lasers. The movement of the galvanometer mirrors and triggering of the laser are determined by pre-programmed behavioral or environmental conditions, allowing stimulation to depend on behaviorally relevant states: for example, if the mouse was performing specific actions (running, sleeping, grooming, rearing) or making choices (turning right in a maze, exploring a specific area of the environment). Flexible, state-dependent laser targeting was accomplished using an infrared laser for thermal stimulation and a blue laser for optogenetic stimulation of genetically targeted primary afferent neurons, enabling high spatiotemporal control of stimulation to small areas of skin. Schematics in panel B were created using BioRender.com.
Figure 2 with 2 supplements
A system for closed-loop cutaneous stimulation.
(A) Rendering of the system shows camera and stimulation optics 1 m below the glass platform, accommodating a large circular arena (0.5 m diameter) for freely moving mice. (B) Side and aerial views. The blue laser beam (blue) was aligned to the galvanometer mirrors (GM) using mirrors (M1, M2), and lenses (L1, L2, L3) via ND filters. The infrared laser beam (red) is directed through a beam shutter with mirrors (M3, M4) and lens (L4). Converged beams in purple. (C) Average image of the laser across a linear voltage grid (left) and a pixel grid after mapping (left-middle). Pixel-voltage mapping corrects distortion (right-middle to right). (D) A mouse on the platform. (E) Tracking in the arena. (F) Galvanometer mirrors tracking the left hind paw keypoint. (G) 2D histograms of paw keypoints highlight the dwell of the locomotor stance phase compared to tail base motion. (H) Histogram of tail base speed indicating categories from four wild-type mice (16,000 frames). (I) Keypoint traces illustrating the out-of-phase swing-stance during locomotion. The left hind paw trace is shown in pink, while the right hind paw trace is shown in blue. The tail base speed is shown in orange. (J) Traces showing alternating left and right paw movement. (K) Accuracy of the laser targeting the hind paws across speed categories. (L) Error between the ground truth keypoint and laser spot in these same four mice, expressed as mean average Euclidean error (MAE). See also related Figure 2—figure supplements 1 and 2. Renderings were created using Solidworks.
Figure 2—figure supplement 1
Spatial and temporal characterization of the closed-loop optical system.
(A) Uniformity of laser spot area and optical power. Heatmaps of mean measurements taken from triplicate samples at 16 separate locations across the glass platform, and fit with a two-dimensional polynomial (area R2=0.91; power R2=0.75). (B) Galvanometer mirrors directed in a spiral formation across the glass platform. (C) The shift in the spatial calibration map was negligible as shown every week for 10 weeks during intensive use. Automatic remapping takes 30 min to complete. (D) The relative timings of the camera exposures (blue), mirror galvanometers (green), and laser (red). The acquisition frame time is shown in gray with corresponding galvanometer mirror jumps and laser pulse occurring around 80 ms later. (E) Histograms of the dwell time for hind paws spent in the swing and stance phases during locomotion. (F) Accuracy of the laser targeting the fore paws across the speed categories, which was limited by left-right confusion in the tracking of small body parts.
Figure 2—figure supplement 2
Hardware and software information flow design.
Primary computer (C1) runs real-time pose estimation on the camera feed to predict multiple body part keypoints. These are converted to voltage signals at the DAQ device (DAQ1) to control the galvanometer mirrors (GM) to target the laser spot coordinates. C1 also sends a trigger to DAQ1 to trigger the blue light laser or the infrared laser shutter via an Arduino Uno (Arduino 2). To generate blue light pulse trains following a trigger, an Arduino was used (Arduino 1). The second computer (C2) interfaces with another DAQ device (DAQ2) to generate audio during experimental sessions. DAQ1 can also interface with DAQ2 to trigger audio depending on processor class conditions and for analog modulation of the blue light laser. The reward delivery system in Figure 4 is controlled via two Arduinos (Arduino 3 and 4) interfacing with DAQ1. Schematics were created using BioRender.com.
Figure 3
Cutaneous stimulation in large environments drives behavioral responses.
(A) Schematic of the open arena. (B) Protocol for minimal cutaneous stimulation using transdermal optogenetic activation of cutaneous nociceptors (Trpv1::ChR2, n=10 mice). (C) A single frame showing a mouse exploring the open arena (left). Keypoints for the left hind paw for 750 frames prior to and 1750 frames after the frame (1 min 23.33 s duration, middle). The body and head orientation at four time points are shown as orange rhombi connecting snout, left, and right forepaw, and tail base (middle). Keypoint skeletons (right). (D) Representative images of a 10 ms laser pulse spot targeting the plantar surface of the hind paw in littermate (top) and Trpv1::ChR2 (bottom) mice. (E) Representative keypoint traces during stimulation of the left hind paw for two mice (columns): five trials with one trial shown in bold across body parts (rows) for each mouse. (F) Example keypoint skeletons from Trpv1::ChR2 mice showing orienting behavior to hind paw stimulation (indicated by the blue arrow). Schematics in panel A were created using BioRender.com.
Figure 4
Multi-animal stimulation for automatic nociceptive testing.
(A) Concept of the random-access multi-animal stimulation. Motion energy was used to detect idle mice in multiple chambers, randomly selecting and cropping to one chamber for real-time pose estimation and stimulation. A laser spot was targeted to the hind paw of the mouse placed in the chamber. The process looped through each of the chambers, automatically targeting and stimulating the mice. (B) An example camera frame highlighting the chambers with different colors (left). Motion energy and body part keypoints shown for an individual chamber (right). (C) Representative paw responses and body repositioning following thermal stimulation (10 s pulse) and optogenetic stimulation (3 ms pulse). (D) Representative paw responses during thermal stimulation of wild-type mice. Two traces plotted with keypoints are shown. The gray dashed line indicates laser stimulation onset. (E) Raster plot of motion energy during thermal stimulation trials for 18 wild-type mice (315 trials), sorted by response latency. (F) Representative hind paw responses during optogenetic stimulation of Trpv1::ChR2 mice. The gray dashed line indicates stimulation onset. (G) Cumulative distribution of paw response latencies to thermal and optogenetic stimulation. Thermal: 18 wild-type mice, 315 trials. Optogenetic: 9 Trpv1::ChR2 mice, 181 trials (range of 15–24 trials for individual mice). Schematics in panel A were created using BioRender.com.
Figure 5
Closed-loop cutaneous stimulation in mice running through a maze.
(A) Schematic of the maze design. A single trial was defined as the collection of a single reward, indicated by the orange and green arrows. The left and right corridors leading to the reward chambers were paired with stimulation of nociceptors using transdermal optogenetics. (B) Maze renderings from aerial, front, and side views. Mice entered via an entry chamber leading to a corridor and junction, choosing left or right through one-way doors. A sucrose-water reward awaited in the reward chamber, with exit through another one-way door. (C) Total number of rewards collected (left and right-hand side reward ports combined) for each training session. (D) Movement trajectories over an entire session (left) and a single trial (right). Trajectories are shown from one mouse for the first stimulation session. (E) Frame sequences (0.2 s apart) from four trials in four mice show runs along maze corridors toward the reward chamber. Blue arrows indicate targeted stimulation. (F) Relative timings of corridor entry and subsequent reward collection (n=4 mice). (G) Transition matrix showing mice predominately alternate between rewards at the left and right reward ports. (H) Example movement trajectories (tail base) in the left and right corridors from one mouse (left). Bar plots showing path coherence and speed in the high stimulation corridor relative to the low stimulation corridor (right). (I) Plot of speed and coherence in a 2 s window prior to the first stimulation in each trial colored by movement state cluster (Gaussian mixture modeling; 116 trials). (J) Speed and coherence relationships before and after stimulation, where the values of the prior state are colored and the responses are shown in black. The black lines pair values for each trial. Stimulation causes a consistent shift in the fast-direct state towards the slow-assess state, while stimulation of the slow-assess state shifts towards a direct movement. (K) Bar plots showing the change (post–pre) in speed and coherence. Stimulation resulted in changes in fast-direct trials (speed: Welch’s t=–6.90, p=0.006; coherence: Welch’s t=–9.39, p=0.003) and also slow-assess trials (coherence: Welch’s t=–26.9, p<0.0001). (L) Plot of post-pre stimulation speed values across all trials colored by state clusters. Schematics in panel A were created using BioRender.com. Renderings were created using Solidworks.
Tables
Key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Strain, strain background (Mus musculus) | R26-CAG-LSL-hChR2(H134R)-tdTomato (Ai27D) | Jackson Laboratory | Stock #: 012567 RRID:IMSR_JAX:012567 | PMID:22446880 |
| Strain, strain background (Mus musculus) | Trpv1-IRES-Cre (TRPV1-Cre) | Jackson Laboratory | Stock #: 017769 RRID:IMSR_JAX:017769 | PMID:21752988 |
| Strain, strain background (Mus musculus) | Wild-type C57BL/6J | Jackson Laboratory | Stock #: 000664 RRID:IMSR_JAX:000664 | |
| Software, algorithm | Python | http://www.python.org/ | RRID:SCR_008394 | |
| Software, algorithm | NI-DAQmx | https://nidaqmx-python.readthedocs.io | ||
| Software, algorithm | Arduino C++ | https://www.arduino.cc | RRID:SCR_024884 | Version 1.8.18 |
| Software, algorithm | DeepLabCut | https://github.com/DeepLabCut | RRID:SCR_021391 | PMID:30127430 Version 2.2.0.2 |
| Software, algorithm | DeepLabCut-Live! | https://github.com/DeepLabCut/DeepLabCut-live | PMID:33289631 Version 1.0 | |
| Software, algorithm | Basler Pylon | https://www.baslerweb.com/en/software/pylon | Version 6.2.4.9387 | |
| Software, algorithm | Cobalt Monitor | https://hubner-photonics.com/downloads/ | ||
| Software, algorithm | Thorlabs Kinesis | https://www.thorlabs.com/kinesis-software | ||
| Software, algorithm | Original code | https://github.com/browne-lab/closed-loop-somatosensory-stimulation | RRID:SCR_028036 | See Data availability section |
| Software, algorithm | Seaborn | http://www.seaborn.pydata.org | RRID:SCR_018132 | |
| Software, algorithm | Adobe Illustrator | http://www.adobe.com | RRID:SCR_010279 | Version 24.0 |
Additional files
-
Supplementary file 1
Optics.
This table details the optical components for the assembly of the system.
- https://cdn.elifesciences.org/articles/106033/elife-106033-supp1-v1.docx
-
Supplementary file 2
Mounting components.
This table details the parts for mounting optics in the system.
- https://cdn.elifesciences.org/articles/106033/elife-106033-supp2-v1.docx
-
Supplementary file 3
Acquisition and control components.
This table details the parts for acquisition and control.
- https://cdn.elifesciences.org/articles/106033/elife-106033-supp3-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/106033/elife-106033-mdarchecklist1-v1.docx
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Precision cutaneous stimulation in freely moving mice
eLife 14:RP106033.
https://doi.org/10.7554/eLife.106033.3
