Remote automated delivery of mechanical stimuli coupled to brain recordings in behaving mice

Abstract

The canonical framework for testing pain and mechanical sensitivity in rodents is manual delivery of stimuli to the paw. However, this approach is time-consuming, produces variability in results, requires significant training, and is ergonomically unfavorable to the experimenter. To circumvent limitations in manual delivery of stimuli, we have created a device called the automated reproducible mechanostimulator (ARM). Built using a series of linear stages, cameras, and stimulus holders, the ARM is more accurate at hitting the desired target, delivers stimuli faster, and decreases variability in delivery of von Frey hair filaments. We demonstrate that the ARM can be combined with traditional measurements of pain behavior and automated machine-learning-based pipelines. Importantly, the ARM enables remote testing of mice with experimenters outside the testing room. Using remote testing, we found that mice habituated more quickly when an experimenter was not present, and experimenter presence led to significant sex-dependent differences in paw withdrawal and pain-associated behaviors. Lastly, to demonstrate the utility of the ARM for neural circuit dissection of pain mechanisms, we combined the ARM with cellular-resolved microendoscopy in the amygdala, linking stimulus, behavior, and brain activity of amygdala neurons that encode negative pain states. Taken together, the ARM improves speed, accuracy, and robustness of mechanical pain assays and can be combined with automated pain detection systems and brain recordings to map central control of pain.

Keywords: brain; mouse; neuroscience; pain; robot; somatosensation; von Frey.

Figure 1.. Mechanical stimulus delivery with the automated reproducible mechanostimulator (ARM).

(A) Comparison between manual stimulus delivery that requires a researcher to aim and deliver stimulus by hand in close proximity to mice vs robotic stimulus delivery via the ARM using motorized linear stages to maneuver and deliver stimulus and a bottom camera to aim. (B) Zoomed-in schematic showing components of the ARM, including the configuration of the linear axis, the aiming camera to the ARM and the stimulus holder. (C) An ARM vs manual stimulus aim comparison was conducted by 5 researchers who delivered 10 instances each of manual and ARM pinprick stimulus to a stationary target. A significant (p<0.0001) 93.3% decrease in distance off-target was observed in ARM stimuli delivery compared to manual delivery.

Figure 1—figure supplement 1.. Automated reproducible mechanostimulator (ARM) stimulus delivery and comparison with manual delivery.

(A) Schematic showing von Frey hair (vFH) wheel mounted on the ARM allowing for seamless switching between full range of vFH filaments and sine wave movement of ARM allowing for full application vFH max force for 2 s. (B) ARM-based application of cotton swab and pinprick stimuli via sine wave motion mimicking manual delivery. (C) Comparison of pinprick stimuli delivered manually and via the ARM, based on max stimulus height measured via high-speed video recordings. Error rate of ±0.152 mm based on resolution. (D) Naïve mice (N=10) were tested using the open field assay; Student’s t-test found no significant difference in distance traveled or (E–G) stress measures including time in center, center entries, or latency to center entry.

Figure 2.. The automated reproducible mechanostimulator (ARM) decreases variability in von Frey hair (vFH) stimulus delivery.

(A) The ARM and external testers each first applied vFH stimulus to a force sensor (1.4 g, 2 g) before applying stimuli to a cohort of mice (n=10) and comparing behavior (0.02 g, 0.07 g, 0.16 g, 0.6 g, 1 g, 1.4 g). (B) Researchers and the ARM user were told to apply stimulus for 2 s to the force sensor for 1.4 g (C) and 2 g vFHs. (D) Stimulus delivery time for 1.4 g and 2 g force sensor trial. (E) Coefficient of variance for vFH (0.6 g, 1 g, 1.4 g, 2 g) on target time as determined by the force sensor was calculated for the ARM and compared to each researcher (p=0.0211), and the combined manual trials (p<0.0001) with a one-way ANOVA. (F) Both researchers and the ARM tested a cohort of wild-type mice (n=10), applying each vFH 10 times to each mouse, producing the expected vFH response curves, including SEM. (G) Comparison between paw withdrawal frequency elicited by Researcher 1 vs Researcher 2 with two-way ANOVA. Significant differences were found in behavior elicited by 0.6 g (p=0.0034), 1 g (p=0.0462), and overall (p=0.0008). (H) Two researchers applied ARM vFH stimulus remotely over 2 days. Two-way ANOVA detected no significant differences.

Figure 2—figure supplement 1.. Stimulus application time decreases and rudimentary withdrawal latency measurement.

(A) Each set of 10 von Frey hair (vFH) applications was timed for both manual and automated reproducible mechanostimulator (ARM) stimulus delivery, with the ARM taking on average 50.9% less time to perform each set of applications (p<0.0001, two-tailed paired t-test). (B) Initial stimulus flexible paw withdrawal latency measurement, made possible by syncing ARM stimulus with high-speed video recordings, separates between responses to cotton swab and pinprick stimuli (p=0.0081).

Figure 3.. Automated reproducible mechanostimulator (ARM) system and integration with pain assessment at withdrawal speeds (PAWS) analysis.

(A) Groups of mice will be tested using the ARM, (B) and for each stimulus delivery, reflexive features including withdrawal latency and mechanical threshold are measured automatically using a force sensor incorporated into the device. (C) Pose analysis of the integrated 500 frames per second (fps) cameras and PAWS high-speed analysis then measures the extent of the reaction, including max paw height, velocity, and behaviors associated with the affective aspect of pain, such as distance traveled, shaking, and guarding. (D) Integration with load cell allows for customizable force ramp stimulus, where force starts low and ramps up over time, and (E) a consistent stimulus that holds at a set force and retracts after duration exceeded or upon paw withdrawal. (F) Test of new PAWS pipeline using carrageenan inflammatory pain model detected significantly higher number of paw shakes at 4 hr compared to baseline (p=0.039), and (G) decreased mechanical threshold with the initial version of the force ramp stimulus (p=0.027). (H) An updated version of the force stimulus was used with a CFA model and found significant decreases compared to control and baseline (p=0.044, 0.025) with the ramp stimulus and a highly significant decrease (p<0.0001) in withdrawal latency in response to a 1 g stimulus.

Figure 3—figure supplement 1.. Testing of automated reproducible mechanostimulator (ARM) 500 frames per second (fps) pain assessment at withdrawal speeds (PAWS) analysis and mouse response to catch trials.

(A) Schematic outlining high-speed recording to pose tracking (DLC or Social LEAP Estimates Animal Poses [SLEAP]) to updated PAWS software pipeline. The blue dotted line denotes the beginning of withdrawal response, and t* denotes the peak of the initial reflexive paw withdrawal response, with reflexive features including max height and max y-velocity measured pre t* and affective features including shaking and paw distance traveled measured post t*. (B) Cohort of male mice (n=10) tested with cotton swab and pinprick stimuli at 500 fps and 2000 fps for comparison. Pinprick was found to elicit significantly increased max paw height (p=0.0352, 0.0352), (C) max y-velocity (p=0.0056, 0.0091), and (D) paw distance traveled (p=0.0340, 0.0692) by a paired t-test. (E) Number of paw shakes was higher for pinprick stimuli but was not found to be significant. (F) Catch trials performed during carrageenan experiments found mice responded 6.6% of the time to catch trials and 95.8% to trials that made contact with the paw, a significant decrease (p<0.0001).

Figure 4.. Remote delivery of mechanical stimuli reveals the effects of researcher presence.

(A) Schematic showing the remote operation of the automated reproducible mechanostimulator (ARM) allowing for researcher-agnostic experiments and flexibility. (B) Male mice (n=10) were habituated either with a researcher present or not for 3 days. Across the 3 days mice rested for the full minute significantly sooner than those with a researcher present (p=0.0217). (C) The number of times each mouse turned, as measured during two 1 min windows 20–30 min each day, normalized by each group's turning behavior during the first 10 min of day 1. On day 2, the remote-habituated mice showed significantly decreased turning behavior compared to those habituated with a researcher present (p=0.024). Only the remote-habituated mice showed significantly decreased turning behavior on day 3 compared to day 1 (p=0.0234). (D) Experimental schematic showing remote ARM stimulus delivery with either a researcher or no researcher in the room. (E–F) A two-way ANOVA found significant differences in max paw height (p=0.0413) and max y-velocity (p=0.0406) in response to cotton swab for male mice when Researcher 2 was present compared to no researcher. (G) Sex-dependent differences were found in response to cotton swab when Researcher 1 was present for distance traveled (p=0.0468). (H–J) Sex-dependent differences were found in response to pinprick stimuli when Researcher 2 was present, but not other conditions for max paw height (p=0.0436), max y-velocity (p=0.0424), and distance traveled (p=0.0038). Male mice showed significant differences in paw distance traveled (p=0.0149) when Researcher 2 was present compared to when none was. (K) To test whether the gender of the experimenter present affects behavior, cohorts of male and female mice (N=15) were tested across 3 weeks with groups of male and female researchers (N=3) and no researcher controls. Male researchers induced significantly decreased mechanical threshold in female vs male mice (p=0.034).

Figure 4—figure supplement 1.. The number of shakes was found to not significantly change based on experimenter presence.

(A–B) Remote experiment comparing mouse response when one of two researchers is present vs none. Nonsignificant sex-dependent differences were found in response to cotton swab (p=0.0818) for Researcher 2 and pinprick (p=0.0842) when Researcher 1 was present.

Figure 5.. Isolating the effect of variation in the application of pinprick stimulus.

(A) Schematic showing how stimulus delivery variation was modeled through changing pinprick intensity by increasing/decreasing pinprick apex and velocity. (B–C) Reflexive features were found to correlate with stimulus intensity based on a simple linear regression, withdrawal latency with a negative correlation, and max paw height with a positive correlation. (D–E) For affective features, paw shaking time showed no significant correlation with stimulus intensity, and paw distance traveled showed a positive correlation.

Figure 5—figure supplement 1.. Isolating the effect of variation in applying pinprick stimulus in female mice.

(A) Based on a simple linear regression, withdrawal latency negatively correlates with stimulus intensity. (B) A piecewise linear regression analysis found that max paw height positively correlates with stimulus intensity for stim apex 1–3 mm and negatively correlates for 3–4.5. (D–E) For affective features, paw shaking time and paw distance traveled showed no significant correlation with stimulus intensity.

Figure 6.. Linking automated reproducible mechanostimulator (ARM) stimulation with behavior and cellular-resolved brain activity in the basolateral amygdala (BLA).

(A) Schematic showing alignment of BLA neural activity recorded by a microendoscope, pain assessment at withdrawal speeds (PAWS) behavioral features, and stimulus facilitated by the ARM. (B) Confirmation of injection of jGCaMP8f virus and insertion of Inscopix mini-scope to the BLA. (C–D) Cell map from processed mini-scope recording with a selection of representative deconvolved cell traces in pseudocolors over a 1000 s window. (E–F) Example traces and cell map of pinprick stimulus aligned up- and downregulated cells based on peri-event analysis. (G) Results of peri-event analysis with up- and downregulated cells based on stimulus, and comparison with random background events. Total regulated cells increased compared to background control for all stimuli (p<0.0001). (H) Percentage of cells registered across multiple days that are regulated during response to mechanical touch and/or pain stimuli. (I) Pearson correlation between the fraction of total peri-event analysis identified mechanical pain-regulated cells with matching regulation for each stimulus event with withdrawal latency (J) and distance traveled in the 1.5 s post-stimulus application.

Figure 6—figure supplement 1.. Correlation of additional pain assessment at withdrawal speeds (PAWS) features with basolateral amygdala (BLA) mechanical pain neuron regulation.

(A) Example cell activity heat map and mean cell trace results of peri-event analysis of representative traces based on either 10 random background or pinprick events. (B) Example mean cell trace results of peri-event analysis of representative traces based on either 10 random background or pinprick events. (C) Fraction of peri-event analysis identified mechanical pain-regulated cells with matching regulation for each stimulus event. Cotton swab events showed decreased down (p<0.0001), up (p<0.05), and total (p<0.0005) regulation of identified mechanical pain cells compared to pinprick or max pinprick. (D) Pearson correlation between the fraction of total regulation of identified mechanical pain cells and paw max height and (E) max paw y-velocity.