All animal procedures were approved by the Boston Children’s Hospital Animal Care and Use Committee, under animal protocol numbers 15-04-2928R and 16-01-3080R. All experiments were conducted blind to injury and treatment in a quiet room from 09:00 to 18:00. Male and female adult C57BL/6J mice (delivered 6–8 weeks for use at 8–10 weeks) were obtained from Jackson Labs (Maine, United States). Each behavioral test was performed on at least two independent cohorts; data were then merged.

The CST apparatus consists of a galvanized steel wire support from which a chainmail hammock was hung at a 45° angle relative to the floor (Figure 1A). The device is positioned so that a mouse can freely climb up the chainmail or remain on an aluminum floor occupying equivalent space at the base of the CST structure. Mouse movement throughout the assay period was video recorded from above. The dimensions of the CST apparatus are provided (Figure 1B). Two adjacent CST devices are enclosed by Plexiglas walls, allowing the evaluation of two mice per trial (Supplementary Figure S1). The aluminum floor consists of the solid base of a Cold Hot Plate (Bioseb, Vitrolles, France). Each CST arena are separated by a matte black Plexiglas divider to prevent any visual or physical interactions between mice during testing.

Mechanical allodynia was measured via the CST using the apparatus described above. Testing was conducted in a randomized manner, with animals from different groups being tested on the same day. Random number generation was used to allocate the order in which individual mice were tested. Animals were kept in a quiet room at room temperature at least 10 minutes before testing. Operators ensured no loud noises or rapid movements occurred while preparing mice. Cages were not changed or cleaned for at least 24 h leading up to testing. Unless required for drug administration, mice were not handled prior to use of the CST. At the start of each trial, mice were gently lifted and carefully lowered onto the aluminum floor side of the enclosed device, facing the chainmail. The investigator then started the video recording software and vacated the testing room. Animals were allowed to freely roam and explore the chamber for 30 min. Trials were recorded using an overhead EverFocus Ultra 720+ EQ(700) camera with a Tamron lens, mounted 1.524 m (5 feet) above the testing area. The chainmail was purchased from Amazon.com as a 8 by “6” rectangle cast iron cleaner.

The video-recorded CST trials were tracked and saved on the Ethovision XT 11.5 software (Noldus, Leesburg, VA). Trials were recorded at 30 Hz for 30 min. Arenas and zones (for chainmail and aluminum) were delineated using the Ethovision software interface, as indicated in Supplementary Figure S1. Mice were tracked with center-point tracking (the “Differencing” option was selected with sensitivity set to 9). Summary statistics in 30-s time bins were automatically computed and exported as Microsoft Excel files. Position heat maps as seen in Figure 2A were generated using the same software package.

Between trials, the aluminum floor, walls, and chainmail were cleaned with a damp disinfecting wipe (Peroxigard, Virox Technologies Inc., Canada), and were dried with paper towels. After completion of a trial session and at least daily, the chainmail apparatus and enclosing box were thoroughly washed with soap and water, and left to dry. We have found this cleaning protocol reduces experimental variance, especially in the control mice. We assume this variance is related to animal odor building up on the device.

The output files from Ethovision were processed by a custom-made program using Python 3.7. Numerical computations were performed using the NumPy package (Version 1.17.4) (van d et al., 2011). Features computed by the chainmail application for each trial include the total and cumulative amount of time spent on each surface, the average velocity of the animal, the total distance traveled, the number of independent entries onto each surface, and the average visit duration for each surface.

Tactile sensitivity was measured at 7 days post-SNI in C57BL/6J mice using von Frey monofilaments (Touch-Test Sensory Evaluators; North Coast Medical, Inc., Gilroy, CA). Filaments used include 0.04, 0.07, 0.16, 0.4, 0.6, 1.0, and 2.0 g. Beginning with the 0.6 g filament, pressure was applied to the left hind paw three times from below, with 10 s intervals between each application. Depending on whether a response was elicited (brisk paw withdrawal and/or an escape attempt), a stronger or weaker filament was then used, following the up-down method as described in (Gonzalez-Cano et al., 2018).

Paw volume measurements were made 3 h after initial intraplantar injection of either saline or carrageenan. Paw volume was measured using solution displacement quantified by a plethysmometer (Bioseb, Vitrolles, France). Paw volume measurements were repeated twice, with average displacement calculated.

PCA analysis was performed on the entire set of measurements described above using the Python SciPy library. A biplot was generated from this output to show separation of the data and the key contributors to such separation.

The CST uses video tracking software to monitor a mouse’s preference between two equally-sized areas of an enclosed arena over a 30 min testing period. The first area is a plain metal floor which is slightly aversive for wild-type C57BL/6J mice at room temperature (22°C-25°C), as they prefer 33°C (Alexandre et al., 2017). The second area houses a chainmail hammock angled 45° to the floor, which mice may freely explore and climb (Figure 1A). Mice were tested 1 week after nerve injury, when tactile allodynia has fully developed (Cobos et al., 2018). Control mice unfamiliar with the CST are usually eager to climb on the chainmail and remain on it for 44.9% ± 2.4% (n = 29) of the testing period (male and female). In contrast, mice with a spared nerve injury (SNI) spend significantly less time on the apparatus, 31.1% ± 1.3% (n = 31); p < 0.0001, two tailed-test (male and female; Figures 2A, B). We propose that climbing on the free-moving chain links prompts mice to place weight on their injured paw, leading to hypersensitivity and consequently, avoidance. This contrasts with climbing on more rigid structures such as metal grids, where mice guard their injured extremity in an effort to prevent discomfort (Alexandrov et al., 2015).

The time spent on the chainmail showed significant differences between control (sham-injured) and SNI mice in both sexes. When results above are separated by sex, sham male mice spent 42.7% ± 0.4% (n = 16) of the time on chainmail, while the SNI group spent 29.7% ± 0.2% (n = 14) (p = 0.0038, one-way ANOVA, Bonferroni post hoc). In female mice, the sham group spent 47.5% ± 0.6% (n = 13) of the time on chainmail, and the SNI group spent 32.3% ± 0.4% (n = 17) (p = 0.018, one-way ANOVA, Bonferroni post hoc). However, no significant difference in time on the chainmail was observed between the sexes in sham mice or after injury (not significant, one-way ANOVA, Bonferroni post hoc, Figure 2B).

Despite this, female mice exhibited higher activity levels than male mice, moving faster both in the sham (1.48 ± 0.06 cm/s for males vs. 2.11 ± 0.11 cm/s for females; p < 0.0001, one-way ANOVA, Bonferroni post hoc) and SNI (1.43 ± 0.04 cm/s for males vs. 1.91 ± 0.07 cm/s for females; p < 0.0001, one-way ANOVA, Bonferroni post hoc, Figure 2C). Furthermore, female mice traveled greater distances throughout the assay period, regardless of injury type (Supplementary Figure S2A).

Given that the time spent on chainmail was similar between male and female groups, the overall increase in female mice’s activity was accompanied by shorter durations spent on the chainmail per visit (Supplementary Figure S2B).

To determine if time spent on the chainmail correlates with von Frey filament measures of tactile hypersensitivity, SNI and sham control mice were subjected to the CST and von Frey tests 1 week apart. Figure 2D shows that these two data sets are well correlated with p < 0.0001 and an R-squared value for the best fit slope of 0.364 (n = 48). As it is generally accepted that von Frey monofilament thresholds represent an accurate quantification of tactile sensitivity (Chaplan et al., 1994; Bourquin et al., 2006), we conclude that time spent on the chainmail represents a reliable inverse measure of stimulus-evoked tactile allodynia (Mogil et al., 2010).

To determine the behavior of male mice after repeated exposure to the CST apparatus, we placed sham and SNI-injured mice in the CST device weekly for 1 month. Sham-injured mice spend 43.8% ± 1.9% of their first exposure on the chainmail (n = 16), but this significantly decreases over repeated weeks, resulting in 29.2% ± 3.1% by Week 4, p < 0.01, two-way ANOVA (mixed-model, Bonferroni post hoc). SNI mice spend 29.2% ± 3.3% of time on the chainmail during their first exposure and 22.1% ± 2.1% of time by Week 4, a non-significant decrease (n = 16, Figure 2E). Between the SNI and sham groups, two-way ANOVA, mixed-model, with Bonferroni post hoc tests demonstrated a significant time difference in Week 1 (p = 0.0004) and Week 2 (p = 0.0033), but not in Week 3 (p > 0.05) or Week 4 (p > 0.05).

Gabapentin is a first line anti-neuropathic pain agent in humans (Gilron et al., 2015; Finnerup et al., 2018) and multiple studies have shown it is effective in relieving chronic neuropathic hypersensitivity in rodents (von Hehn et al., 2012). Here we tested if gabapentin could reduce the reticence of male mice to explore the chainmail side of the CST device 7 days after SNI. Gabapentin (60 mg/kg) dissolved in saline was IP injected 30 min before testing, while vehicle control mice were injected with saline (Figure 3A). The standard dose range of gabapentin used in von Frey-based rodent assays of neuropathic allodynia is 50–100 mg/kg (Field et al., 2000; Patel et al., 2001; Joshi et al., 2006). SNI-injured mice treated with gabapentin (60 mg/kg) (n = 10), spent 45.7% ± 5.0% of assay time versus 33.26% ± 3.2% for gabapentin-treated sham controls (n = 9), not significant, one-way ANOVA, Bonferroni post hoc, Figure 3B).

To control for the possibility that SNI mice do not climb on the chainmail portion of the CST apparatus because of a lack of motor control brought on by nerve injury, we tested male mice subject to a full sciatic nerve axotomy on the CST. Fully axotomized mice lack any evoked sensation within the denervated hind paw, save the saphenous territory (Hsieh et al., 2000). At 7 days post injury no directly injured sensory axons remain in the lower paw. In addition, motor control is profoundly affected, with little or no movement possible in the lower leg. The saphenous does not contain any motor fibers and remains uninjured. Conversely, mice subject to SNI retain some innervation from the spared sural nerve which, with respect to the sensory system, is responsible for the tactile allodynia present, and within the motor system allows for a minimal level of movement. Mice subject to complete axotomy of the sciatic nerve (n = 13) explored the chainmail 39.9% ± 3.5% of the assay time, compared to 44.9% ± 1.9% for sham control mice (n = 13) and 27.3% ± 2.3% for SNI mice (n = 15). There was no significant difference between sham and axotomy groups, though there was a significant difference between axotomy and SNI group, (p = 0.01, one-way ANOVA, Bonferroni post hoc, Figure 3B).

To explore the possibility that the SNI animals do not move due to pain or anhedonia, we quantified the mean velocity by these mice and compared it to the sham-injured animals (Figure 3C). PBS-injected mice subject to SNI had a velocity of 1.06 ± 0.07 cm/s (n = 15), while sham mice 1.38 ± 0.05 cm/s (n = 13), not significant (one-way ANOVA, Bonferroni pos hoc). At 60 mg/kg of gabapentin, sham mice (n = 9), and SNI groups (n = 10), had a slightly greater level of activity (1.64 ± 0.12 cm/s and 1.62 ± 0.16 cm/s, respectively) than the saline injected controls, though neither change was significant (one-way ANOVA, Bonferroni post hoc). Mice subject to sciatic nerve axotomy had a velocity of 1.49 ± 0.07 cm/s (n = 13), which was also not significantly different from the SNI and sham-injured mice (one-way ANOVA followed by Bonferroni post hoc test). This indicates that the SNI, axotomy and sham-injured males are equally active. Additionally, distance (Supplementary Figure S3A), average time per chainmail visit (Supplementary Figure S3B) and the number of visits to the chainmail (Supplementary Figure S3C) did not differ among the groups above.

Multivariate analysis demonstrates that time spent on the chainmail versus time on the aluminum surface effectively separates SNI from sham injured values (Figure 4A; Supplementary Figure S4A). Interestingly male and female mice separate in a manner consistent with mean velocity and time per visit (Figures 4B, C) as described above. Highlighting SNI mice treated with 60 mg/kg gabapentin on this plot shows that these mice, in the main, separate with sham injured animals (Figure 4A, black points) as was the case with naive mice (Supplementary Figure S4B). The variance in the plotted data is primarily accounted for by the first (PC1) and second (PC2) principal components, which contribute 50% and 24% respectively (Supplementary Figure S4C).

Here we examined quantification of carrageenan-induced hypersensitivity in male mice (Winter et al., 1962; Morris, 2003). Carrageenan solution (50μL, 1% wt/vol in saline) was injected into the plantar surface of the left hind paw 4 hours prior to the CST, while vehicle control mice were injected with saline. Mice treated with ibuprofen at 30 mg/kg were injected 30 min before the assay (Figure 5A). Carrageenan-treated mice (n = 15) spent an average of 30.4% ± 2.6% of the assay time on the chainmail versus 50.5% ± 4.3% for naïve-PBS (saline-injected) mice (n = 12), (p = 0.006, one-way ANOVA, Bonferroni post hoc, Figure 5B). In addition, carrageenan-inflamed mice treated with ibuprofen (n = 11) spent 49.4% ± 7.7% of the assay time on the chainmail, which is not significantly different to naïve-saline controls (one-way ANOVA, Bonferroni post hoc). These data provide evidence of the relationship between the time spent on the chainmail and tactile hypersensitivity due to peripheral inflammation. In order to demonstrate peripheral edema due to carrageenan-induced inflammation, we measured paw volume 3 h after carrageenan injection. Carrageenan mice (n = 15) measured 0.188 ± 0.006 mL, which was significantly higher than the 0.103 ± 0.007 mL measured in naïve-saline mice (n = 14), (p < 0.0001, two-tailed t-test, Supplementary Figure S5A).

By examining the time spent on the chainmail during a trial (Figure 5C), we identified a difference in the behavior of the SNI mice relative to carrageenan-inflamed male animals. This was examined as time spent on the chainmail in the first 15 min of the assay period relative to the final 15 min. In the carrageenan mice (n = 15), there was a significant difference in the time spent on the chainmail between these two periods, with mice spending more time on the chainmail in the first 15 min window (Carrageenan 0–15 min 37.7% ± 2.7%; carrageenan 15–30 min 23.0% ± 3.6%; (p > 0.0001, 2-way ANOVA mixed-model, Bonferroni post hoc). This is in contrast to the SNI and sham mice, where no such difference between these two periods exist (Figure 5D). We assume this difference in CST exploration in acutely inflamed mice is due to sensitization of the peripheral inflamed tissue of the paw, which results in the mice becoming less willing to climb on the chainmail as time progresses.

Neither velocity (Figure 5E), distance traveled (Supplementary Figure S5B), nor time per chainmail visit (Supplementary Figure S5C) were significantly different after carrageenan injection or subsequent ibuprofen treatment. Carrageenan-treated mice (n = 15) visited the chainmail 53.1 ± 3.8 times during the 30-min trials, significantly less than 74.3 ± 4.7 visits for saline-injected controls (n = 12), (p = 0.049, one-way ANOVA followed Bonferroni post hoc, Supplementary Figure S5C).