High-speed imaging of evoked rodent mechanical behaviors yields variable results that are not predictive of inflammatory injury

1. Introduction

Animal models are widely used to study nociception and test experimental analgesics. However, despite promising results in pre-clinical trials, few analgesics originally characterized in rodents have successfully translated to clinical use [10,37]. This may be due to fundamental differences in gene expression between species [18], but another possibility – and one that deserves critical appraisal in experimental design – is that our field is over reliant on reflexive, unidimensional paw withdrawal behaviors as a primary measure of pain [26]. A recent review found that over 60% of preclinical papers published in the journal Pain between 2016–2020 used mechanically evoked behavior assessments; more specifically, over 50% of studies used the up-down von Frey method to calculate mechanical withdrawal thresholds as a behavior correlate of touch hypersensitivity [26]. Despite the widespread use of von Frey filament-based mechanical assessments, there is considerable debate as to whether the percept induced by von Frey stimulation of an uninjured animal hind paw can be considered painful or reflects tickle, itch or other non-painful tactile detection. In most laboratories, responses to von Frey filament stimulation are reported in a binary manner; animals either withdraw their paw during filament stimulation or they do not. By coding responses in this manner, we are limiting the number of behavior parameters on which putative analgesics may exert their effect. As such, affective aspects of the withdrawal response may be important measures of nociception that are typically not incorporated into the determination of mechanical and thermal hypersensitivity.

To this end, several groups have used high-speed videography to characterize nuanced aspects of mechanically evoked paw withdrawals [1,9,19,31]. Recent work suggests that specific features of the paw withdrawal sequence, mainly paw height, paw velocity and pain score – a composite score that accounts for the presence of paw-specific (e.g., paw hover, paw flutter) and body-wide (e.g., jumping, grimace) behaviors – can be used to distinguish behavioral responses to innocuous and noxious stimuli [1]. These specific quantitative measurements should increase the reproducibility of evoked, reflexive behavior assessments between laboratories. However, a key impetus for the current manuscript was anecdotal differences reported by members of our laboratory as they began using high-speed videography in their independent projects. Analysis of naïve and vehicle control animals’ withdrawal measurements seemed to vary widely from person-to-person. Thus, we set out to systematically assess if experimenter identity was a critical source of variability in high-speed imaging responses. Furthermore, if these measures reflect the percept evoked by innocuous and noxious stimuli, exaggerated responses to both lower intensity and frankly noxious stimuli should be observed in diverse injury models based on the definitions of allodynia and hyperalgesia respectively. To test these hypotheses, we had 12 members of our laboratory stimulate the same group of C57BL/6 mice with various mechanical probes to first, test the reproducibility of the aforementioned paw withdrawal parameters between experimenters, and second, determine if the paw height, paw velocity, and pain score extracted from high-speed recordings could be used to objectively quantify the mechanical allodynia and hyperalgesia that develop in the Complete Freund’s Adjuvant model of inflammatory pain.

2. Methods

2.1. Experimenters:

Twelve experimenters participated in this study (Table 1). Experimenter experience was classified by the timeframe over which the experimenter actively conducted mouse pain behavior experiments within the previous five years; gaps in experience (e.g., PI who last performed behavior tests > 5 years ago) were not considered in classification. Experimenters were classified as novice (<3 months experience), intermediate (>3 months and <3 years) or expert (>3 years).

Table 1:

Individual experimenter information.

ID	Sex	Experience level	Primary species experimenter works with	Initial training occurred in/by:
1	Male	Expert	Mouse	Stucky lab, experimenter 7
2	Female	Novice	Mouse	Stucky lab, experimenter 7
3	Female	Novice	Mouse	External
4	Female	Expert	Mouse	External
5	Female	Intermediate	Mouse	External
6	Female	Intermediate	Rat	External
7	Female	Expert	Mouse	External
8	Female	Novice	Mouse	Stucky lab, experimenter 7
9	Male	Intermediate	Rat	Stucky lab, experimenter 7
10	Male	Novice	Mouse	Stucky lab, experimenter 7
11	Male	Intermediate	Rat	External
12	Female	Intermediate	Mouse	Stucky lab, experimenter 7

2.2. Animals.

8-week-old male and female C57BL/6J mice were purchased from The Jackson Laboratory and housed on a 14:10 light/dark cycle with ad libitum access to food and water. Animals were acclimated to the MCW vivarium for a week prior to start of experiments. Animals used for all the studies were group-housed (n=4–5 mice per cage), all cages were housed in the same room, and all cages contained same bedding (P.J. Murphy Coarse Grad Sani-Chips). All experimental protocols were done in accordance with the National Institute of Health guidelines and approved by the Medical College of Wisconsin’s Institutional Animal Care and Use Committee (Milwaukee, WI; Protocol #383). Equally sized cohorts of male and female mice were used in all experiments.

2.3. Behavioral assays for high-speed videography recordings.

Animals were habituated to small rectangular testing chambers (9 cm length, 5.6 cm height, 5.2 cm width) with opaque plexiglass short sides and translucent plexiglass long sides for ~30 minutes over 5 consecutive days prior to experimental testing. After experiment initiation, each animal was tested once daily for 4 or 5 days each week. The same animals were tested by all experimenters to account for inter-animal behavior variability. Experimenters tested each mouse once with each stimulus by order of escalating stimulus intensity and the testing order was counterbalanced across mice. On each testing day, animals were placed in testing chambers on an elevated wire-mesh for 1 hour prior to experimenter acclimation. All experimenters then acclimated with mice for 30 minutes prior to beginning hind paw stimulations [29]. Each mechanical stimulus was applied once to the center of the hind paw in the following order, waiting ~2 minutes between stimuli application: von Frey filaments (0.6g, 1.4g, 4.0g), paintbrush, and needle. All experimenters used the same set of calibrated von Frey filaments, paintbrush, and needle to perform stimulations. Mechanical stimuli were applied to the hind paw as in traditionally used behavioral assessment: von Frey filaments were applied to center of the hind paw and bent to a 45-degree angle, brush was angled 45-degrees from the paw and moved from the heel to the toes in one fluid motion, and needle was applied to the center of the hind paw to induce a deformation of the plantar surface without puncturing the skin.

2.4. High-speed Videography.

We performed high-speed videography similar to Abdus-Saboor and Fried -Saboor, Fried, and colleagues[1]. Hind paw movement was recorded before, during, and for ~0.2 sec after each stimulus application at 2000 frames per second (fps) using a high-speed camera (FASTCAM Mini UX50 160K-M-32GB) equipped with a macro lens (Zeiss Planar T* 85mm f/1.4 ZF.2 Lens Nikon). Camera was mounted on a tripod and placed ~3 feet away from the chambers in which mice were located. Data were collected and annotated on an HP EliteBook 840 using Photron Fast Cam Viewer 4 (PFV4) software. Videos were trimmed to include 500–600 frames (~0.2s), which captured the total motion of the paw.

2.5. Scoring high-speed behavior features:

Paw height, velocity, and pain score were extracted from the high-speed video recording using Photron FastCam software and analyzed by a singular experimenter in the following manner [1]: 1) paw height was measured as the distance from the mesh to the highest point of the paw arc during the initial lift; 2) paw velocity was calculated by dividing the distance the paw travelled during the linear acceleration phase of the withdrawal (i.e., once paw was raised off the mesh until reaching apex of withdrawal arc) by the time that elapsed during that movement; 3) pain score was determined by the presence (1 point) or absence (0 points) of four behavioral features: orbital tightening, paw flutter, paw hover, and jump. Orbital tightening refers to the animals’ eyes going from fully open to partially or completely closed after stimuli application. Paw flutter was defined as multiple, fast paw shakes following stimulus application. Paw hover was defined as paw guarding or avoiding placing the paw flat on the mesh following stimulus application. Lastly, a jump was defined as an action during which two or more paws were raised off the mesh at the same time during the response to stimulus.

2.6. Mechanical force tester:

A force sensor created by the MCW engineering core was used to measure the force applied by individual experimenters during von Frey testing. The force sensor was connected to a Dell Inc OptiPlex 760, using LabChart7. Upon force application to the sensor, a voltage proportional to the force applied was generated and recorded using LabChart7. The force sensor was placed inverted on the top of the elevated wire mesh behavior testing apparatus and experimenters applied von Frey filaments (0.6g, 1.4g, and 4.0g) to the sensor pad in triplicate. Peak force was determined as the highest point of the force wave for each stimulation.

2.7. Complete Freund’s Adjuvant (CFA) Injury Model.

Unanesthetized animals were restrained then received a 20 mL intraplantar injection of undiluted CFA or sterile PBS into the right hind paw 1 week after being delivered to the animal facility [25]. Animals were randomly assigned to CFA and vehicle groups. All animals housed in each cage received the same treatment to account for social spread of pain behaviors. Behavior testing started on day 7 post CFA or vehicle injection and was completed by day 40 following CFA or vehicle injection[36]. Equal sized male/ female cohorts were used in each treatment group. The same cohort of CFA or vehicle injected mice was tested by all experimenters to account for inter-animal behavior variability. Experimenter testing sessions were counterbalanced so that every experimenter had an early and late testing session following CFA/ vehicle injections. These studies were done in parallel with naïve cohort studies.

2.8. Traditional evoked behavioral tests:

For all traditional evoked behavioral assessments, animals were habituated to testing chamber (1 hour) and experimenter presence (30 minutes). An experimenter from each experience group (Experimenters 1, 2, and 5 in Table 1) performed the behavioral assessments listed below on a separate cohort of mice injected with CFA or vehicle. Experimenters tested mice such that the intermediate experimenters went first, followed by the advanced experimenter and the novice experimenter, with at least two days in between testing. This cohort of mice was tested between 7- and 16-days following injection of CFA or vehicle.

2.8.1. von Frey mechanical allodynia:

Hind paw sensitivity to punctate mechanical stimulation was assessed using calibrated von Frey filaments (0.02 to 1.4g) and the up-down assessment method. The 50% withdrawal threshold was determined as previously described [8].

2.8.2. Dynamic paintbrush test:

Hind paw sensitivity to dynamic light touch was assessed via paintbrush stimulation. A fine paintbrush was swept across the center of the plantar surface from heel to toe. Speed and force were kept constant between the 10 stimulations of each paw. Response frequency and characterization were reported as null (no response), innocuous (simple paw withdrawal) or nocifensive (withdrawal along with liking, biting, flicking and/or additional paw tending behaviors) [27].

2.8.3. Noxious needle test:

Hind paw sensitivity to noxious punctate stimuli was assessed using a needle. A 25-gauge needle was applied to the center of the paw with enough force to indent, but not puncture the skin. Each paw was stimulated 10 times and the responses were characterized as described in dynamic paintbrush test [27].

2.9. Statistics:

Data were analyzed using GraphPad Prism 9 and results were considered significantly different when P<0.05. The z-scores and first principal component of all data was calculated as a “PC”-score as presented by Abdus-Saboor, Fried and colleagues [1] using R version 4.3.1 (Fig 1j). Data from each sex were first independently analyzed but given that no significant differences were noted for any experiment, data from both sexes were combined. Statistical comparisons were guided and analyzed by biostatisticians and authors Ulrich Kemmo Tsafack and Aniko Szabo.

Figure 1: High-speed imaging measures differentiate behavior responses to needle and innocuous mechanical stimulation.

High-speed imaging was used to assess paw height, paw velocity, and pain score following hind paw stimulation with various mechanical stimuli. (a.) Paw height was defined as the distance between the base and peak height of a given spot on the paw during the withdrawal. (b.) Paw velocity was calculated by dividing the distance over which the paw traveled during the linear movement phase of the withdrawal by the time over which that movement occurred. (c.) Pain score was a cumulative measure coded by the presence (+1) or absence (+0) of the following behaviors: paw hover, paw flutter, grimace, and jumping. (d.) Withdrawal height varied as a factor of mechanical stimulus applied to the paw; needle-induced responses were higher than those elicited by the 1.4 g von Frey filament (1-way ANOVA main effect of stimulus F_(4,47)=1.828, P=0.0309; all experimenters tested same mice, average score for each of n=12 experimenters). (e.) Withdrawal velocity also varied by stimuli; on average, needle stimulation resulted in faster withdrawals than stimulation with any von Frey filament (1-way ANOVA main effect of stimulus F_(4,48)=7.722, P<0.0001). (f.) Mechanical stimulus also had an overall effect on pain score, but when averaged across experimenters, significant differences were not observed between individual stimuli (Kruskal-Wallis test P=0.0275). Raw paw height, paw velocity and pain score were transformed into Z scores. (g.) Although normalized paw height were not different across stimuli (1-way ANOVA no main effect of stimulus F (_{1.893, 17.04})=3.187, P=0.0690), (h.) normalized paw velocity (1-way ANOVA main effect of stimulus F(_{2.068, 18.62})=8.286, P=0.0025) and (i.) normalized pain score (1-way ANOVA main effect of stimulus F(_{1.986, 17.88})=4.692, P=0.0232) varied across stimuli applied. (j.) The first principal component of paw height, paw velocity, and pain score z score values was calculated. The PC1 of z scores varied across stimuli (1-way ANOVA main effect of stimulus F(_{1.571, 14.14})=7.967, P=0.0071).

2.9.1. High-speed videography measurement analysis:

Paw height comparisons were made using a two-way ANOVA with a Tukey’s post hoc test. For paw velocity data, comparisons were made using a two-way ANOVA followed by a Tukey’s post hoc test. Pain scores were compared using a Kruskal Wallis test.

2.9.2. Variance component analysis:

A mixed effects variance components model was fitted for all continuous outcomes with stimulus, treatment, and their interaction as fixed effects, and animal, experimenter, and session within experimenter as random effects. Method-of-moments estimation was used that does not make assumptions about the distribution of the outcome.

2.9.3. Traditional behavioral assays:

Withdrawal thresholds were compared using a two-way ANOVA. Response classifications to paintbrush and needle stimuli were analyzed using Chi-square test with Fisher’s exact tests.

2.9.4. Treatment groups mean difference analysis:

Paw height, velocity, and pain score were compared between the groups using Student’s t-test whenever at least 3 observations were available in both groups. The effect was quantified as mean difference with 95% confidence interval.

3. Results

3.1. Paw velocity, but not paw height or pain score, distinguish behavioral responses to innocuous versus noxious mechanical stimuli.

Using high-speed videography, Abdus-Saboor, Fried, and colleagues identified the following three parameters as critical for distinguishing behavior responses to innocuous and noxious mechanical stimuli in naïve mice [1]: paw height, paw velocity, and pain score. They further suggested that stimulation with a 4.0g von Frey filament induces a pain-like withdrawal reflex that is similar to paw withdrawal following needle or pin prick stimulation [1]. To determine if these same measurements could be used by an independent laboratory to achieve similar results, we performed the following experiment: high-speed videography was used to record paw height (Fig. 1a), paw velocity (Fig. 1b), and pain score (Fig. 1c) measurements exhibited by naïve C57BL/6 mice during hind paw stimulation with 0.6 g, 1.4 g, and 4.0 g von Frey filaments, brush, and punctate needle. The same cohort of 10 mice was tested by 12 members of our laboratory who have varying levels of behavior experience (Table 1). Average withdrawal parameters were calculated across mice for each experimenter, so that the effects of stimulus intensity could be investigated across a varied group of participants. Although there was an overall effect of stimulus on paw height, the only significant differences that persisted following post hoc analyses were between responses to the needle and the 1.4 g von Frey stimulus; as expected, paw height was increased following needle stimulation relative to 1.4 g von Frey filament stimulation (Fig. 1d). Notably however, the 1.4 g filament was the only innocuous stimulus that induced responses that differed from those induced by needle. Paw height evoked by 4.0 g filament stimulation – reported previously as a noxious stimulus – did not differ from the withdrawal height induced by supposedly innocuous stimuli (0.6 g, 1.4 g, brush).

Assessments of paw velocity yielded a clearer distinction between noxious and innocuous stimuli (Fig. 1e). Similar to paw height, paw velocity changed as a function of stimulus, but in post hoc analysis, only needle responses differed from those generated by innocuous stimuli; needle-evoked paw withdrawals were significantly faster than those exhibited following stimulation with the 0.6 g, 1.4 g, or 4.0 g von Frey filament. Like paw height, paw velocity induced by 4.0 g filament stimulation did not differ from responses exhibited following innocuous stimulation. Lastly, we assessed pain score, a cumulative measure of binary affective measures including orbital tightening, jumping, paw hover, and paw flutter [1]. Although there was an overall effect of stimulus on pain score, post hoc analyses did not reveal specific differences between any stimuli across the 12 experimenters (Fig. 1f).

We further analyzed these results by transforming paw height, paw velocity and pain score values into Z scores, as performed by Abdus-Saboor and Fried et. al. [1]. Although normalized paw height did not vary across stimuli (Fig. 1g), normalized paw velocity (Fig. 1h) and pain score (Fig. 1i) values were different across stimuli, with responses to brush and needle being significantly larger than responses to von Frey filaments. Finally, the data was transformed through a PC-score (first principal component of paw height, paw velocity, and pain score (Fig. 1j). Similar to raw measurements of paw velocity and pain score, we detected a significant main effect of stimulus on PC-scores (Fig. 1j). On post hoc analysis, brush and needle had higher PC-scores compared to 0.6 g and 1.4 g von Frey filaments. In conclusion, when a group of diverse experimenters attempted to validate paw height, paw velocity, and pain score as fundamental parameters for distinguishing naïve C57BL/6 mouse behavioral responses to innocuous and noxious mechanical stimuli, we failed to observe the same robust differences originally reported by Abdus-Saboor, Fried, and colleagues. Despite observing exaggerated responses to needle stimulation, we failed to observe a force-dependent increase in response to von Frey filament application. Furthermore, responses to brush stimulation appeared more similar to those observed following needle simulation than those observed following innocuous mechanical stimulation.

3.2. Hind paw withdrawal frequency depends on experimenter identity whereas nuanced withdrawal parameters vary widely within and between individuals.

To determine if the lack of correlation between stimulus intensity and behavioral measures resulted from a lack of reproducibility between experimenters, we independently analyzed the hind paw withdrawal parameters elicited by each stimulus for all 12 experimenters. Before assessing individual withdrawal parameters, the percentage of mice that responded to a given stimulus was compared across experimenters. Response frequency to the 0.6 g filament varied widely between experimenters (Fig. 2a). Despite testing the same cohort of animals, in the hands of some experimenters, the 0.6 g filament elicited a response in 70% mice (e.g., Experimenter 6), whereas in others, 0% of mice responded to this stimulus (e.g., Experimenters 5 and 11). We next determined whether objectively quantifiable aspects of the paw withdrawal varied in a similar fashion. Paw withdrawal height (Fig. 2b) and velocity (Fig. 2c) were not statistically different between experimenters following application of the 0.6 g filament. However, pain score did vary in an experimenter-dependent fashion (Fig. 2d). Notably, Experimenter 7, an individual who trained six of the other participants in this study, observed paw flutters with every 0.6 g filament-induced withdrawal; the other experimenters trained by Experimenter 7 did not observe the same phenomenon.

Figure 2: High-speed withdrawal measures are variable within and between experimenters.

(a.) Although the percentage of mice that responded to hind paw stimulation with a 0.6 g von Frey filament varied between experimenters (χ2=33.32, P=0.0005; all experimenters tested same n=10 mice), there was no effect of experimenter identification on (b.) the height (1-way ANOVA no main effect of experimenter F_(9,24)=0.4251, P=0.9085) or (c.) the velocity (1-way ANOVA no main effect of experimenter F_(9,25)=1.373, P=0.2522) of the paw withdrawal as observed in high-speed videography. (d.) The pain score exhibited following 0.6 g von Frey stimulation did vary between experimenters (Kruskal-Wallis test P=0.0187). (e.) The percentage of mice that responded to 1.4 g von Frey stimulation (χ2=19.03, P=0.0606), (f.) the paw height (1-way ANOVA no main effect of experimenter F_(10,56)=1.490, P=0.1675), (g.) paw velocity (1-way ANOVA no main effect of experimenter F_(11,60)=0.6580, P=0.7715), and (h.) pain score elicited by this stimulation did not vary as a function of experimenter (Kruskal-Wallis test P=0.1478). Similarly, no effect of experimenter identification was observed on (i.) the percentage of mice that responded to 4.0 g von Frey filament stimulation (χ2=17.35, P=0.0047), (j.) paw height (1-way ANOVA no main effect of experimenter F_(11,81)=1.490, P=0.1675), (k.) paw velocity (1-way ANOVA no main effect of experimenter F_(10,78)=0.9651, P=0.4803), or (l.) pain score (Kruskal-Wallis test P=0.2657) induced by 4.0 g filament stimulation. (m.) The percentage of mice that responded to dynamic brush stimulation varied across experimenters (χ2=26.91, P=0.0047), but the brush-induced withdrawal (n.) height (1-way ANOVA no main effect of experimenter F_(11,82)=1.837, P=0.0607), (o.) velocity (1-way ANOVA no main effect of experimenter F_(11,81)=1.439, P=0.1716), and (p.) pain score (Kruskal-Wallis test P=0.0642) did not differ between experimenters. (q.) The percentage of mice that responded to needle hind paw stimulation varied by experimenter (χ2=28.96, P=0.0023), as did the (r.) height with which animals withdrew their paw following needle stimulation (1-way ANOVA no main effect of experimenter F_(11,93)=3.405, P=0.0005); no effect of experimenter was observed on needle-induced (s.) withdrawal velocity (1-way ANOVA no main effect of experimenter F_(11,93)=1.810, P=0.0630) or (t.) pain score (Kruskal-Wallis test P=0.1980).

The percentage of mice that responded to the 1.4 g filament did not differ between experimenters (Fig. 2e). Notably, all participants except Experimenter 5 observed ~50% of mice respond to 1.4 g von Frey filament stimulation. Given that the average 50% withdrawal threshold for mice reported in Pain publications is 1.36 g [26] we can conclude that 11 of our 12 experimenters stimulate the hind paw in a manner consistent with other laboratories. Similarly, neither paw height (Fig. 2f), paw velocity (Fig. 2g), nor pain score (Fig. 2h) elicited by the 1.4 g filament varied as a function of experimenter. Conversely, the percentage of responses to the 4.0 g filament varied by experimenter (Fig. 2i). However, paw height (Fig. 2j), paw velocity (Fig. 2k), and pain score (Fig. 2l) evoked by stimulation with the 4.0 g filament were statistically equivalent across experimenters.

Similar to the variability in response rate following von Frey filament stimulation, the percentage of animals that responded to brush application varied as a function of experimenter (Fig. 2m). Although brush-induced paw height, paw velocity, and pain score were not statistically different from those observed following 0.6 g and 1.4 g von Frey filament stimulation (Fig. 1), the percentage of mice that responded to brush stimulation, was closer to that observed following 4.0 g filament or needle stimulation. Wide variability was observed in brush-induced paw height (Fig. 2n), paw velocity (Fig. 2o), and pain score measurements within a given experimenter (Fig. 2p), and thus, there was no overall statistical effect of experimenter on these measures.

Needle stimulation response frequency also varied across experimenters (Fig. 2q) as did paw withdrawal height (Fig. 2r). Although paw velocity (Fig. 2s) did not vary across experimenters, needle responses were, on average, significantly faster than those induced by all other stimuli (Fig. 1e). Lastly, there was relatively equal distribution of pain scores across experimenters (Fig. 2t). Taken collectively, these data illustrate the widespread variability observed in response frequency, paw withdrawal height, paw velocity, and pain score when the same cohort of mice are stimulated by various experimenters.

3.3. Inter-experimenter variability in mechanical withdrawal frequency may result from the manner in which individuals apply mechanical stimuli.

Given the inter-experimenter differences in mechanically evoked paw withdrawal frequency, we next wanted to determine which experimenter qualities contributed to this variability. The effects of experimenter sex have been well-documented in the preclinical pain literature. Specifically, acute exposure to a male experimenter leads to stress-induced analgesia in both male and female mice [29]. Despite allowing animals to acclimate to experimenter presence for a time that should negate potential sex effects, we compared average withdrawal parameters collected by members of each sex. Experimenter sex did not affect paw height (Fig. 3a), paw velocity (Fig. 3b), or pain score measures (Fig. 3c).

Figure 3: High-speed withdrawal measurements are not affected by experimenter sex or experience with pain behavior but may differ due to variability in the application of mechanical stimuli.

Experimenter sex did not affect (a.) paw height (2-way ANOVA no main effect of experimenter sex F_(1,42)=2.642, P=0.1115, main effect of stimulus F_(4,42)=6.979, P=0.0002; all experimenters tested same n=10 mice), (b.) withdrawal velocity (2-way ANOVA no main effect of experimenter sex F_(1,42)=0.5503, P=0.4623, main effect of stimulus F_(4,42)=3.300, P=0.0194), or (c.) pain score (2-way ANOVA no main effect of experimenter sex F_(1,45)=2.544, P=0.1177, main effect of stimulus F_(4,45)=4.294, P=0.0050). (d.) Experimenters stimulated a force-transducer with von Frey filaments to assess stimulus application accuracy and variability. (e.) Experimenters applied discrete forces with each von Frey filament regardless of experience level (2-way ANOVA no main effect of experience level F_(2,27)=1.647, P=0.2114, main effect of stimulus F_(2,27)=81.05, P<0.0001). (f.) The actual force applied by individual experimenters varied when using each of the von Frey filaments (1-way ANOVA main effect of experimenter identification for 0.6 g filament F_(11,24)=6.559, P<0.0001; 1.4 g filament F_(11,24)=5.901, P=0.0001; 4.0 g filament F_(11,24)=12.39, P<0.0001). (g.) Stimulus application velocity did not differ between experimenters with varying experience levels or between von Frey filaments (2-way ANOVA no main effect of experience level F_(2,23)=1.746, P=0.1968, no main effect of stimulus F_(2,23)=0.1754, P=0.8402). (h.) However, the velocity with which each von Frey filament was applied did vary between individual experimenters (1-way ANOVA main effect of experimenter identification for 0.6 g filament F_(11,49)=0.8857, P=0.5600; 1.4 g filament F_(10,42)=2.659, P=0.0130; 4.0 g filament F_(11,56)=8.548, P<0.0001).

We next reasoned that experimenters may exhibit variability in the application of mechanical stimuli, particularly those individuals who have limited evoked behavior experience. To accurately measure the peak force each experimenter applied with von Frey filaments, we engineered a force gauge that could be positioned upside-down on the behavior testing mesh, thus allowing experimenters to stimulate the gauge as if they were stimulating a mouse hind paw. Resulting force waveforms from each von Frey filament were collected and analyzed for each participant (Fig. 3d). Experimenter differences in peak force and application duration time were observed. When the average peak force applied by individuals was analyzed by experience level, no effect of previous experience was found (Fig. 3e). However, the average peak force applied by the 0.6 g, 1.4 g and 4.0 g filaments did vary across individual experimenters (Fig. 3f). Although most experimenters applied a force relatively close to the expected value, peak force measurements from Experimenters 6 and 11 were exceedingly low; when stimulating the force gauge with the 4.0 g filament, these two individuals only applied ~1.5 g of force.

In addition to applied force, the rate of mechanical stimulus application and removal affects subsequent perception; steeper application ramps (i.e., step application) result in higher mechanical thresholds (i.e., lower mechanical sensitivity) in human subjects relative to slower application ramps [23]. To this end, we compared the rate with which von Frey filaments were applied to animal hind paws in the original high-speed video recordings. As with application force, application velocity was not affected by experimenter experience level (Fig. 3g). When the application velocity of individual filaments was analyzed independently, effects of experimenter identity were observed for the 1.4 g and 4.0 g filaments (Fig. 3h). These effects may be driven by the high velocity – or ‘jab-like’ nature – with which experimenter 11 applied these filaments. In summary, we observed variability in the speed and force with which experimenters applied von Frey filaments. These inter-individual discrepancies may contribute to the variability observed in hind paw withdrawal behaviors.

3.4. Paw height, paw velocity and pain score variability is primarily driven by unidentified factors.

Our work so far suggests that features of experimenter identity, particularly the way in which they apply mechanical stimuli, may drive variability in mouse behavior responses. We next performed a variance component analysis to determine the relative contribution of experimenter identity against other controlled variables including animal identity, testing time of day, or type of mechanical stimulus. Paw height variability could be equally, albeit limitedly, explained by experimenter identification, stimulus-dependent effects within each experimenter, the individual session during which a given experimenter tested the mice, animal identification, and stimulus-dependent effects within each animal (Fig. 4a). Surprisingly, about 80% of the variability observed in paw height could not be explained by any of the variables that we controlled for or identified, as noted by the residual component. Like paw height, variability in paw velocity measurements was largely (87%) dictated by unidentified or uncontrolled factors; animal identity was the second largest factor, driving 8.5% of variability (Fig 4b). Lastly, pain score variability was primarily driven by unexplained factors (81% of variability), although session within experimenter (i.e., time) accounted for 13.1% of variability (Fig. 4c). Importantly, these response variables also did not change with repeated exposure to stimuli (Supp Fig.1). Based on these variance component analyses, variability in paw height, paw velocity and pain score is primarily driven by factors that have yet to be considered in preclinical pain behavior testing.

Figure 4: Majority of variability in high-speed withdrawal measurements cannot be accounted for by known variables.

A mixed effect variance component model indicates that variability of (a.) paw height (b.), paw velocity (c.) and pain score values is primarily driven by residual, while factors controlled for such as experimenters, stimulus, session, and stimulus account for under 15% of measurement variability.

3.5. Paw withdrawal measurements cannot be used to differentiate between vehicle and CFA injected mice.

In order for high-speed videography to be widely adopted in preclinical pain research, this technique must be able to reliably detect allodynia and hyperalgesia that often occur following injury. To this end, we injected mouse hind paws with Complete Freund’s Adjuvant (CFA), a compound that is widely used to model inflammatory pain (reviewed in [26]). Before performing high-speed video analysis, we confirmed that experimenters with varying levels of experience could detect mechanical allodynia and hyperalgesia phenotypes in CFA injected mice using traditional reflexive behavior assays. Encouragingly, novice, intermediate, and expert experimenters all observed decreased withdrawal thresholds in CFA injected mice using traditional up-down von Frey assessments (Fig 5a), and an increase in nocifensive responses in CFA injected mice following stimulation with a brush (Fig 5b) and noxious needle (Fig 5c).

Figure 5: Inflammation-induced mechanical allodynia and hyperalgesia are not detected using high-speed imaging methods.

Hind paw injection of Complete Freund’s Adjuvant (CFA) resulted in mechanical hypersensitivity as measured by (a.) up-down von Frey assessment (2-way ANOVA main effect of treatment F_(2,48)=15.07, P<0.0001, main effect of experimenter F_(1,48)=34.02, P<0.0001; all experimenters tested same mice, n=9 mice per treatment group), (b.) real-time experimenter characterization of responses to brush stimuli (χ2=121.3, P<0.0001; Fisher’s exact test *P<0.05, **P<0.01 vehicle vs. CFA, ^##P<0.01 vs. novice vehicle), and (c.) real-time experimenter characterization of responses to needle stimulation (χ2=95.86, P<0.0001; Fisher’s exact test *P<0.05, **P<0.01 vehicle vs. CFA, ^{^}P<0.05, ^{^^}P<0.01 vs. expert vehicle). When high-speed imaging was used to assess mechanical behavior responses, (d.) no difference in paw height was observed between vehicle and CFA treated mice (2-way ANOVA no main effect of treatment F_(1,106)=3.259, P>0.05, main effect of stimulus F_(4,106)=8.183, P<0.0001; average score for each of n=12 experimenters). An overall effect of CFA treatment was detected in (e.) high-speed paw velocity measurements (2-way ANOVA main effect of treatment F_(1,105)=4.730, P=0.0319, main effect of stimulus F_(4,105)=17.83, P<0.0001) and (f.) high-speed pain score measurements (2-way ANOVA main effect of treatment F_(1,106)=7.990, P=0.0056). However, following post-hoc analyses, the only statistically significant difference between vehicle and CFA treated animals was the 1.4 g von Frey filament pain score. Unless otherwise stated, Tukey’s multiple post-hoc comparisons *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Paw height, paw velocity and pain score values from mice injected with vehicle and CFA were normalized using Z-score. (g.) No difference between vehicle and CFA injected mice was observed in normalized paw height (2-way ANOVA main effect of treatment F (_{1, 90}) = 2.132, P=0.147. (h.) An overall effect of stimulus and CFA was detected in normalized paw velocity measurements (2-way ANOVA main effect stimulus F(_{1, 9})=8.177, P=0.0188, main effect of treatment F(_{4, 36})=23.20, P<0.0001). (i.) Similarly, an effect of stimulus and CFA was detected in normalized pain score measurements (2-way ANOVA main effect of stimulus F(_{2.036, 18.33})=4.172, P=0.0316, main effect treatment F(_{1.000, 9.000})=8.868, P=0.01550). (j.) Principal component (PC1) from normalized paw height, paw velocity, and pain score noted no difference between vehicle and CFA injected mice (2-way ANOVA, main effect of stimulus, F(_4,90)=14.72, P<0.0001).

Next, we performed high-speed imaging on CFA and vehicle injected mice. Given the definition of allodynia and hyperalgesia, we expected injured animals to have increased paw height, paw velocity and pain scores following stimulation with innocuous and noxious stimuli respectively. Our group’s collective data noted no differences in paw height between vehicle and CFA injected mice (Fig 5d). Although an overall effect of CFA treatment was noted for paw velocity, no within-stimulus statistical differences were noted following post-hoc analyses (Fig 5e). Lastly, an overall effect of CFA treatment was noted for pain score, but specific post-hoc differences were only noted for the 1.4 g von Frey filament; CFA injected mice had increased pain scores in response to 1.4 g filament stimulation compared to vehicle injected mice (Fig 5f).

Further analysis of these results was done through normalization of paw height, paw velocity and pain score values into Z scores and compared through a principal component analysis as in Abdus-Saboor and Fried, et. al. [1]. An overall effect of treatment was noted in normalized paw height; however, post hoc analyses showed no within-stimulus statistical differences (Fig 5g.). Analysis of normalized paw velocity (Fig 5h.) and pain score (Fig 5i) indicated overall effects of treatment and stimulus, and post hoc analyses further revealed that CFA mice had decrease paw velocities following stimulation with 4.0 g filament and brush while pain scores were higher in CFA-injected mice following stimulation with 1.4 g von Frey filament. Lastly, normalized data was transformed through a principal component analysis to plot the principal component 1 (PC1) (Fig. 5j). PC-score analysis noted to no overall effect of CFA. In conclusion, normalization of paw height, paw velocity, and pain score reveals a limited dynamic range to identify response differences between CFA and vehicle injected mice.

To further dissect whether paw height, paw velocity, and pain score could be used by individual experimenters to differentiate between naïve and injured mice, mean differences between vehicle and CFA injected mice were calculated. The number of mice that responded to 0.6g filament stimulation varied between experimenters, and surprisingly, did not always suggest an injury phenotype; several experimenters observed more vehicle-injected mice respond to the 0.6 g filament than CFA-injected mice (Fig. 6a). Overall, no treatment group differences were noted for paw height (Fig. 6b), paw velocity (Fig. 6c) or pain score (Fig. 6d) following 0.6 g von Frey stimulation, further supporting the inability of high-speed videography to detect a mechanical allodynia phenotype in our CFA model. Likewise, experimenters observed variable response rates to 1.4 g von Frey filament stimulation (Fig. 6e) and failed to observe effects of CFA on the mean differences in paw height (Fig. 6f), paw velocity (Fig. 6g) or pain score (Fig. 6h). As with the 0.6 g and 1.4 g von Frey filaments, the number of mice that responded to 4.0 g von Frey stimulation varied across experimenters (Fig. 6i). The mean difference in paw height (Fig. 6j) and pain score (Fig. 6l) was also similar across experimenters. However, a single experimenter (Experimenter 2) observed a statistically significant difference in paw velocity when they compared values from CFA and vehicle treated mice; specifically, they observed that, on average, vehicle injected mice withdrew their paws faster than CFA injected mice following 4.0 g filament stimulation (Fig. 6k).

Figure 6: Individual experimenters cannot detect differences in mechanical withdrawal behaviors between injured and sham animals using high-speed videography.

(a). Individual experimenter 0.6g von Frey filament response rate difference between CFA and vehicle injected mice. Additionally, no differences were noted in treatment group difference in (b.) paw height (mixed effect linear regression, p=0.148), (c.) paw velocity (mixed effects linear regression, p=0.879) nor (d.) pain score (mixed linear regression, p=0.763) for 0.6g von Frey stimulations. (e.) Response rate differences between CFA and vehicle injected mice following 1.4g von Frey filament across experimenters. No differences were noted on treatment (f.) mean difference paw height (mixed effect linear regression, p=0.367), (g.) paw velocity (mixed effect linear regression, p=0.135) or (h.) pain score (mixed effect linear regression, p=0.017). (i.) Experimenter response number differences between CFA and vehicle injected mice upon stimulation with 4.0g von Frey filament. No differences were noted between experimenter (j.) paw height (mixed effect linear regression, p=0.446), (k.) paw velocity (mixed effect linear regression, p=0.088) or (l.) pain score (mixed effect linear regression. P=0.159). (m.) Individual experimenter brush response rate differences between CFA and vehicle injected mice. Overall (o.) paw velocities were different across experimenters (mixed effect linear regression, p=0.026), but no differences were noted in (n.) paw height (mixed effect linear regression, p=0.378) or (p.); pain score (mixed effect linear regression, p=0398). (q.) Experimenter response rate differences between CFA and vehicle injected mice following needle stimulation. (r.) Paw height (mixed effect linear regression, p=0.787), (s.) paw velocity (mixed linear regression, p=0.615) nor (t.) pain score (mixed effect linear regression, p=0.929) from needle stimulation were different across experimenters.

Experimenters also obtained varying response rates to paintbrush stimulation. Five experimenters observed more CFA injected mice respond to paintbrush, whereas three experimenters observed more vehicle injected mice responding to this stimulus, despite testing the exact same group of animals (Fig. 6m). Paw height did not differ between treatment groups following paintbrush stimulation (Fig. 6n). However, three experimenters observed statistically significant differences in paintbrush-induced paw withdrawal velocity; all three individuals reported that, on average, vehicle injected mice remove their paw faster than CFA-injected mice following paintbrush stimulation (Fig. 6o). Additionally, three experimenters reported significantly different pain scores between CFA and vehicle injected mice following paintbrush stimulation; two experimenters found that CFA injected mice had higher paintbrush-induced pain scores than vehicle-treated mice, whereas another experimenter found the opposite (Fig. 6p). Lastly, needle stimulation resulted in different proportions of mice responding for each experimenter; three experimenters observed more CFA-injected mice respond to needle stimulation, whereas four experimenters reported more vehicle-injected mice respond to this stimulus (Fig. 6q). Needle-induced paw withdrawal height (Fig. 6r) and velocity (Fig. 6s) did not differ between treatment groups when averaged across experimenters, but pain score values did, with one experimenter in particular observing a higher average pain score in CFA injected mice relative to vehicle injected mice. Altogether, high-speed paw height, paw velocity, and pain score measurements did not reliably or reproducibly differ between vehicle and CFA injected mice.

4. Discussion

Here, we set out to determine whether high-speed imaging could be used by individuals with varying levels of pain behavior testing expertise to differentiate animal responses to noxious and innocuous stimuli. Additionally, we asked whether this method could differentiate evoked mechanical behavior responses from injured and non-injured animals. This study failed to replicate the original findings on which this study was based [1]; we were unable to differentiate withdrawal responses to graded innocuous and noxious mechanical stimuli using paw withdrawal height, paw withdrawal velocity, and pain score. We found that the response of naïve mice to brush and von Frey filament stimulation, regardless of filament force, could not be distinguished using paw height, paw velocity, or pain score. However, needle stimulation induced faster, and higher withdrawal motions compared to other stimuli (Fig. 1d, 1e). Further normalization using principal components analysis allowed for the distinguishing of responses to needle and brush versus responses induced by 0.6 g and 1.4 g von Frey filaments (Fig. 1j). This force intensity-dependent effect was limited, however, as brush and needle could not be distinguished from one another, von Frey filaments could not be distinguished from one another, and suprathreshold (4.0 g) von Frey filament testing could not be distinguished from any other stimulus (Fig. 1j). Most surprisingly, CFA treatment had no effect on paw withdrawal height and had minimal effects on paw withdrawal velocity and pain score when averaged across all 12 investigators (Fig. 5d–f, Fig. 6). Additionally, normalization of paw withdrawal measurements revealed no effect of CFA treatment on paw height, and minimal effects on paw velocity and pain scores (Fig 5g–i), while the compilation of all three measures through a principal component analysis was unable to differentiate between CFA and vehicle injected mice (Fig. 5j).

4.1. Regardless of application method, blunt, mechanical stimuli do not induce a painful percept in uninjured rodents.

Despite being the most used tools to assess ‘pain’ in rodents, there is debate as to whether von Frey filaments induce a painful perception in animals. Previous high-speed videography analyses show rats do not exhibit nocifensive behaviors following von Frey hind paw stimulation but do so after hind paw stimulation with a noxious needle [5,31]. Similar differences were observed in a learned avoidance paradigm; naïve rats exhibit avoidant behaviors when their hind paw is stimulated with a needle, but not von Frey filaments [39].

Although not delineated as ‘pain-like’ in these or previous studies [1], the percept induced by a brush stimulus also remains abstract. Most experimenters in our laboratory observed that >75% of C57BL/6 mice respond to hind paw stimulation with a paintbrush (Fig. 2m). Similarly high response levels to brush stimulus were observed previously [13]. Why might this innocuous brushstroke across the paw induce approximately the same “aversive” responses as punctate needle in both high-speed (Fig. 1b, 1c, 2d, 2m vs. 2q) and traditional (Fig. 5b, vs. 5c) behavior testing? One explanation may be that brush activates more receptive fields with each application compared to focal stimulation. Additional explanations are that this form of touch is not innocuous and is instead equivalent to knismesis, an aversive tickle mediated by Aδ fibers [33] or that this type of touch causes mechanical itch mediated by pruriceptors [2,16,21]. Future studies should explore the extent to which dynamic and punctate mechanical stimuli of various intensities engage signaling in affective central circuits.

4.2. The primary factors that drive hind paw withdrawal variability in uninjured mice remain unknown.

The impetus for this manuscript was anecdotal differences reported by members of our laboratory as they began using high-speed videography in independent projects. Withdrawal properties of naïve and sham control animals seemed to vary greatly from person-to-person. Thus, we set out to systematically assess if experimenter identity was a critical source of variability in high-speed imaging responses similar to other evoked behavioral assays [7].

Surprisingly, even though the force and velocity with which experimenters applied mechanical stimuli varied widely between individuals (Fig. 3d–h) and the proportion of mice that responded to a given stimulus varied between experimenters (Fig. 2a, 2e, 2i, 2m, 2q), experimenter identity accounted for <15% of variability in paw height, paw withdrawal, or pain score measurement (Fig. 4). To this end, it is possible that the nature of stimulus application has no effect on the resulting withdrawal reflexes. This is unlikely, however, given that spinalized animals exhibit enhanced reflexive responses following peripheral injury [14,24]. Another factor we explored was ‘session within experimenter’. Experimenters tested half of the vehicle and CFA treated mice in each of two sessions. These sessions were separated by 7–14 days, a time period during which animals were being tested in the same environment by additional experimenters.

Over 80% of variability in paw height, paw withdrawal, and pain score measurements could not be explained by factors controlled or accounted for in our variance component analysis. In attempting to identify additional factors that vary between sessions, animals, and experimenters, we landed on animal internal state. Despite being housed and tested in highly controlled, homogeneous environments, daily changes to cage hierarchy [32][20], variation in caloric intake [15][11], and sleep quality [3] may drive variability in pain behaviors. In conclusion, the measures that are used to distinguish behavioral responses to innocuous and noxious mechanical stimuli are not impacted by differences in experimenter application of said stimuli to the extent that was originally predicted, but rather by factors the field has yet to explore.

4.3. CFA induced mechanical allodynia can be detected using traditional behavioral assays, but not with high-speed mechanical withdrawal measurements.

We originally hypothesized that CFA treatment would increase paw height, paw velocity, and pain score regardless of the stimulus tested. Overall effects of CFA injection were noted on paw velocity and pain score (Fig. 5d–f), but only two experimenters detected statistically significant differences between CFA and vehicle injected mice supporting this hypothesis (Fig. 6p, 6t). Three experimenters observed that withdrawal responses to brush stimuli were significantly slower in CFA injected mice than vehicle injected animals (Fig. 6o). Surprisingly, many experimenters did not observe increased proportions of CFA injected animals responding to single application of a given filament (Fig. 6a, 6e, 6i, 6m, 6q). Experimenters with varying experience level did however observe CFA-induced mechanical allodynia and hyperalgesia using traditional mechanical behavior assessments (Fig. 5a–c).

The most obvious difference between the current study and traditional behavior assessments is that stimuli are repeatedly presented to an animal in the traditional up-down von Frey assessment and in our paintbrush and needle protocols. Repeated presentation of a stimulus, particularly one capable of activating C fiber nociceptors, can lead to temporal summation and a spinal phenomenon known as wind-up [38]. In CFA treated animals, there may be ongoing C fiber activity driven by the release of pro-inflammatory mediators in the hind paw. Factors in this inflammatory milieu may also recruit silent nociceptors or lower the mechanical thresholds of high-threshold mechanoreceptors by sensitizing mechanically gated ion channels (e.g., Piezo2) or conveying mechanical sensitivity to ion channels that are mechanically insensitive under naïve conditions (e.g., TRPV1 [12,28,30,34]). As such, repetitive stimulation of the hind paw with innocuous mechanical stimuli – like von Frey filaments – may lead to activation of novel populations of C fiber mechanoreceptors and subsequent spinal summation.

Caveats of this study and reflection on the continued utility of high-speed videography assessments in evoked behavior tests.

Our group firmly believes that traditional, preclinical behavior tests fail to measure many aspects of chronic pain, and thus, we are extremely supportive of new methodologies designed to address this question [26]. To this end, one of our lab members recently used high-speed video analysis to measure changes in mechanical withdrawal behaviors of mice lacking the mechanically gated ion channel PIEZO1 in epidermal keratinocytes [19] and found that conditional PIEZO1-knockout mice exhibit less intense mechanical withdrawal behaviors than wildtype littermates. However, when others in our group began using this same procedure in other mouse strains and injury models, we received anecdotal reports of high inter-experimenter variability in naïve/sham animal responses, and, furthermore, lab members observed less robust differences between injured and non-injured mice than we routinely observe using blinded, traditional mechanical tests.

After completing this systematic analysis, our results do not support the use of paw height, paw velocity and pain score from high-speed videography as assessment for nociceptive features. Instead, we believe that the traditional up-down von Frey assessment and assays—where stimuli are repeatedly applied to the hind paw—currently provide more rapid and robust observations of mechanical allodynia and hyperalgesia as compared to the current high-speed imaging paradigm, at least in the CFA model of inflammatory pain in C57BL/6 mice. von Frey filament testing is not translationally defunct as this procedure has been used to measure changes in mechanical sensitivity of humans following an inflammatory insult, namely ultraviolet B burn, mild thermal burn, and freezing injury [4,6,22,35]. While we acknowledge there are differences between these models in humans and the use of CFA in rodents, it is important to note that von Frey filament testing detects mechanical hypersensitivity in both human and rodent models of inflammatory pain. Our inability to observe strong injury-related changes in paw height, paw velocity, or pain score may be because injury alters mechanical withdrawal responses in a manner that does not mirror naïve animal responses to noxious stimuli, but rather, reflects the incorporation of additional protective behaviors such as paw guarding. Unbiased machine learning data recently published by Bohic et al. directly support this hypothesis [17].

When originally developing this method, only naïve animal withdrawal responses to light touch and noxious pin prick were used [1]. It is possible that additional behavior sequences that were infrequently observed in naïve animals may be more frequently observed in diverse injury models, and thus not included in the parameter list. As this list may be distinct for species, strain, and behavioral test, we encourage future investigators to employ unbiased machine learning-based approaches for the global identification of withdrawal behaviors captured with high-speed videography. Additionally, this methodology should be tested in reflexive measures of heat (e.g., Hargreaves radiant heat test) and cold (e.g., plantar dry ice) hypersensitivity. Other types of acute and chronic tissue or nerve injury should also be tested. An open repository of the withdrawal parameters for diverse models of pain may facilitate analgesic compound screening, which has the potential to provide greater specificity compared to traditional evoked behavior assays. Altogether, we believe the incorporation high-speed videography may one day reduce bias and improve the robustness, reproducibility, and translatability of evoked withdrawal behaviors. However, our results indicate that this approach is not yet suitable for widespread use in pain research. Further attempts at refining this approach should identify novel parameters that are able to differentiate between innocuous and noxious stimuli in naïve and injured animals that are reproducible between laboratories, individual experimenters, and pain conditions.