Distinctive Neurophysiological Signatures of Analgesia after Inflammatory Pain in the ACC of Freely Moving Mice

GCaMP6f is expressed in excitatory but not inhibitory neurons in GP5.17 mice

To determine the utility of neural ensembles in the ACC toward preclinical assessments of pain, we utilized the C57BL/6J-Tg(Thy1-GCaMP6f)GP5.17Dkim/J mouse line, allowing for the stable expression of the genetically encoded calcium indicator GCaMP6f in neurons over months (Dana et al., 2014). Similar to other Thy1 lines, the GP5.17 transgene is prominently expressed in L2/3 and L5 pyramidal neurons of several cortical regions, including the ACC (Dana et al., 2014). However, there is evidence to suggest that Thy1 lines can in fact express their transgenes in inhibitory interneurons in some cases, albeit at lower levels (Wang et al., 2007; Proskurina and Zaitsev, 2021). To evaluate if GCaMP6f is expressed in excitatory or inhibitory neurons of the ACC in GP5.17 mice, we performed triple fluorescent in situ hybridization with RNAscope using probes for the GCaMP6f, vGluT1, and vGAT genes (Fig. 1). GCaMP6f (GFP) mRNA signals were detected in the entire ACC cortical area with higher expression in deeper layers (Fig. 1A,F) as reported previously (Dana et al., 2014). Though we observed variations in the levels of GCaMP6f and vGluT1 among cells, GCaMP6f signals were colocalized well with vGluT1 (Fig. 1A–C,F–H,K–M). However, GCaMP6f signals were rarely colocalized with vGAT signals, supporting the notion that GCaMP6f expression is absent or at least severely limited in inhibitory neurons (Fig. 1A,D–F,I,J,N–P). Together, these results suggest that the detected GCaMP6f fluorescent signals in the GP5.17 ACC area are primarily derived from excitatory neurons.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

C57BL/6J-Tg(Thy1-GCaMP6f)GP5.17Dkim/J, in which GCaMP6f is predominantly expressed in excitatory neurons, displayed essentially the same pain responses as WT littermates. A–E, RNAscope images of the ACC in a Thy1-GCaMP6f (GP5.17) mouse with low magnification, stained with three probes. The probes and colors used in the images are shown above each panel, including those for the genetically encoded calcium indicator GCaMP6f (green), the vesicular glutamate transporter vGluT1 (magenta), and the vGAT (magenta). F–J, Each panel corresponds to those in A–E but magnified. K–P, High magnification images of the same area of interest (white squares) in F and G, with the addition of the nuclear DNA stain DAPI. Magenta was used to visualize both vGluT1 and vGAT for optimal contrast to compare excitatory and inhibitory populations. K,L, Filled white arrowheads: the cells coexpressing both GCaMP6 and vGluT1. Magenta hollow arrowheads: vGluT1 positive but GCaMP6f negative. Green hollow arrows: GCaMP6f positive but vGluT1 negative. N,O, Magenta hollow arrows, vGAT positive but GCaMP6f negative. Asterisk: GCaMP6f positive, and vGAT may be weakly positive. Q, Von Frey filament series used to assess mechanical sensitivity in WT, HET, and HOM, Thy1-GCaMP6f GP5.17 littermates before and after a 20 µl CFA injection into the right hindpaw. Paw withdrawal was measured once per day across 10 d (3 d uninjected, 7 d CFA injected) for 30 mice (10 per genotype). R, Time course of PWT testing for the uninjected paw, where each tick on the x axis represents 1 d of testing. PWT values were averaged across recording conditions for each mouse for statistical comparison (Tukey’s HSD test, N = 10 mice per genotype). S, Comparisons of the mean PWT values within each recording condition (naive or CFA) between each genotype pair for the uninjected (left) paw are shown. No Tukey’s HSD test comparison between genotypes yielded statistical significance. T, As in R, but for the CFA-injected paw. U, As in S, but for the CFA-injected paw (Tukey’s HSD test). *p < 0.05, **p < 0.01. See Extended Data Table 1-1 for additional details.

Mechanical sensitivity in Thy1-GCaMP6f GP5.17 is indistinguishable from WT littermates

To evaluate the feasibility of GP5.17 mice for behavioral assessments with mechanical stimuli, we assessed PWT using von Frey filament stimulations (Chaplan et al., 1994; Fig. 1Q). We bred GP5.17 and its parental line C57BL/6 and measured PWT once daily for HOM, HET, and WT littermates (10 mice for each genotype), with genotypes blinded to the experimenter. PWT testing was performed on the left and right hindpaws under naive conditions for the first 3 d, followed by 7 additional days of testing after CFA was injected to induce inflammatory pain in the right hindpaw. In the uninjected (left) paw, no significant differences in the mean PWT values could be found between any of the genotypes in either the naive or CFA conditions (Fig. 1R, Extended Data Table 1-1). Furthermore, no significant differences in these values were observed between the naive and CFA conditions within any genotype (Fig. 1S, Extended Data Table 1-1). No significant differences in PWT between genotypes in the naive or CFA conditions were observed in the CFA-injected (right) paw (Fig. 1T, Extended Data Table 1-1), while significant decreases between the naive and CFA conditions were observed within each genotype (Fig. 1U, Extended Data Table 1-1). Taken together, these results suggest that HET and HOM GP5.17 mice exhibit comparable CFA-induced mechanical allodynia to their WT counterparts.

To determine neural correlates of nociception in the ACC, we utilized one-photon excitation from head-mounted miniaturized microscopes to image calcium transients in freely moving HOMO GP5.17 mice across pain and analgesic conditions (Fig. 2A–C). Mice were first habituated 1 h/d for 3 d to a clear cylindrical chamber sitting atop a metal grate, through which thermal or mechanical stimuli were delivered. We used two concurrent stimulation paradigms for testing, including a standard “up–down” PWT assay (Dixon, 1980; Chaplan et al., 1994; performed on both hind paws) with mechanical stimuli (0.04, 0.07, 0.16, 0.4, 1, 2, 4, 6 g von Frey filaments) and a multimodal stimulus paradigm (right hindpaw only) consisting of noxious and innocuous thermal and mechanical stimuli (0.07, 0.04, 2 g von Frey filaments, pinprick, 4, 30, 55°C water droplets, approach without touch) to facilitate our understanding of stimulus encoding and nociception in the ACC (Fig. 2D). To determine how behavioral responses and ACC activity dynamics are shaped by pain and analgesia, we performed these assays across separate days when mice were naive (Day 4), received intraplantar injections of CFA 24 h prior to induce inflammatory pain (right paw only, Day 6), and received a 75 mg/kg oral (gavage) dose of GBP 1 h before the recording session (Day 7; Fig. 2E). PWT measurements, which have long been a standard for behavioral measurements of allodynia and hyperalgesia, were used as a benchmark to determine if CFA and GBP exerted their effects in accordance with previously published studies (Chincholkar, 2018; Zhu et al., 2021). We did not find significant differences in PWT between paws under naive conditions. After CFA injections, however, we observed robust and significant decrease in PWT in the CFA-injected paw compared with that in the uninjected paw (Fig. 2F). This difference was mitigated after GBP treatment, with both CFA-injected and uninjected paws exhibiting similar withdrawal thresholds (Fig. 2F, Extended Data Table 2-1). For the un-injected paw, no significant differences in PWT could be found between any recording condition (naive, CFA, GBP; Fig. 2G, Extended Data Table 2-1). However, a significant reduction in the mean PWT was observed for the CFA-injected paw compared with the naive and GBP conditions (Fig. 2G, Extended Data Table 2-1). No significant difference was found between the naive and GBP conditions of the CFA-injected paw, indicating that GBP restored mechanical sensitivity to its original state (Fig. 2G, Extended Data Table 2-1). When comparing PWT between HOM GP5.17 mice implanted with GRIN lenses and those without, no significant differences could be found within the naive or CFA conditions for either paw, suggesting that the surgical procedure itself did alter mechanical sensitivity (Fig. 2H).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Paw withdrawal testing and imaging of GCaMP6f signals in the ACC of GP5.17 mice across days under pain and analgesia. A, Schematic of the experimental design. A miniaturized microscope was used to image calcium transients in the ACC of awake adult (3–8 months) male HOM Thy1-GCaMP6f (GP5.17) mice on a metal grate within a clear cylindrical chamber. B, Left: stereotaxic atlas image indicating the intended imaging location targeting layer five excitatory neurons in the dorsal ACC (circled in red). Right: coronal brain section of a recorded mouse showing the GRIN lens track location and approximate locations of the dorsal (Cg1) and ventral (Cg2) ACC. C, Example cell contours overlayed onto maximal projection images (ΔF/F) from three recordings made on separate days and under separate conditions from a single mouse. Yellow contours correspond to cells that were longitudinally tracked across all three recording sessions, while other colors (cyan, magenta, green) indicate cells detected within single sessions that did not meet the criteria for longitudinal tracking across all three recording sessions. D, Top: range of mechanical stimuli used to test PWT. Bottom: range of thermal and mechanical stimuli used for multimodal stimulation during imaging sessions. E, Experimental time course for PWT testing and multimodal stimulation across pain and analgesic conditions. All animals were recorded once per day in an untreated condition (“naive”), an inflammatory pain condition (“CFA”), and an analgesic condition (“GBP”). Figure notations: “hb,” 30 min habituation; “PWT,” paw withdrawal threshold assay. F, A point plot indicating mouse-averaged PWT for three PWT measurements per condition for uninjected and CFA-injected paws. Naive, N = 12 mice; CFA, N = 12 mice; GBP, N = 12 mice. *p < 0.05; **p < 0.01. G, Left panel: a bar graph showing cross-condition comparisons of PWT for the uninjected paw (Tukey’s HSD test). Right panel: as in the left panel, but for the CFA-injected paw (Tukey’s HSD test). See Extended Data Table 2-1 for additional details. H, Left panel: a bar graph showing comparisons of PWT between HOM GP5.17 mice with and without implanted GRIN lenses within the naive and CFA conditions for the uninjected paw. Right panel: as in the left panel, but for the CFA-injected paw. No significant differences could be found between normal and lens implanted mice (Mann–Whitney U test of mean PWT; GRIN implanted mice, N = 12; normal mice, N = 10). I, Left: example heatmap showing mean activity (z-scored ΔF/F) across time for all recorded cells in response to pinprick stimulation in the naive condition (12 mice). Z-scored ΔF/F normalized to the prestimulus period (−5 to 0 s) is shown as a heatmap. Right: schematic representation of the classification of single-cell responses. The raw ΔF/F values of the prestimulus (−5 to 0 s) and poststimulus (0–5 s) periods were compared across trials using a two-tailed Wilcoxon rank–sum test, resulting in cells exhibiting “si,” “sd,” or “ns” in activity (threshold p value, 0.05). J, Subpopulations of the cells shown in J, separated according to their responder classifications as “si”, “ns,” and “sd” cells. K, The cell-averaged z-score time course of the classified responders populations shown in K. Shading represents SEM. L, Top: point plot indicating averaged percentages of each responder classification across mice for each stimulus during naive recordings. Bottom: statistical grid plot indicating p values of comparisons of percentage “si” between stimuli (Tukey’s HSD test). See Extended Data Table 2-2 for additional details. M, Alluvial plot of all ACC cells (1,680 across 12 mice) recorded in the naive condition across each stimulus. Each horizontal line indicates a unique pattern of responder classifications across stimuli for one or more cells.

Population responses in the ACC largely reflect stimulus intensity while single-cell responder classifications vary across days

During multimodal imaging experiments, we imaged the ACC of the left hemisphere while stimulating the contralateral (right) hindpaw. In total, we recorded calcium transients from 1,680 neurons (12 mice) in the naive condition, 1,625 neurons (12 mice) after intraplantar CFA injections 24 h prior to recording, and 1,284 neurons (12 mice) after a 75 mg/kg oral (gavage) dose of GBP 1 h before the recording session. After the delivery of a stimulus, we observed increased or decreased calcium transients for subsets of ACC neurons from trial-averaged (z-scored ΔF/F) activity heatmaps, where the peak periods of excitation or inhibition occurred within the first 5 s after stimulation (referred to herein as the “peak response period”) and subsequently decayed across time (Fig. 2I,K). We then categorized these neurons based on the statistical significance of their cross-trial increases or decreases in activity, providing a snapshot of the response polarity (positive, negative, or neutral) and the robustness of single-cell activity across recording conditions (naive, CFA, and GBP) and somatosensory stimuli (Fig. 2I). Each neuron was assigned a responder classification based on whether it displayed significantly increased (“si”), significantly decreased (“sd”), or no significant (“ns”) change in activity when comparing raw ΔF/F values in prestimulus (−5 to 0 s) versus poststimulus (0–5 s) windows across trials (1 s bin size) for each stimulus within each recording condition (Wilcoxon rank–sum test, p < 0.05; Fig. 2I,J). This resulted in most neurons being classified as “ns” (68.3% for the pinprick), while those classified as “si” (20.4% for the pinprick) and “sd” (11.3% for the pinprick) comprised the fewest across the population (Fig. 2J–L). Higher-intensity stimuli evoked larger percentages of responsive cells (“si” or “sd”) and smaller percentages of unresponsive cells (“ns”) across the population (Fig. 2L, Extended Data Table 2-2). However, alluvial plots across stimuli suggested that single cells were highly variable in these classifications (Fig. 2M).

To evaluate the stability of responder classifications across somatosensory stimulations, we first classified ACC cells into “si,” “ns,” or “sd” cell groups in response to a single stimulus (pinprick) and then assessed the responder classification stability and population response magnitude of these groups across all other stimulus modalities in naive mice (Fig. 3A). Aligning single cells across trial-averaged (z-scored ΔF/F) activity heatmaps to each stimuli revealed nonuniform single-cell response dynamics between stimuli but demonstrated a tendency for the majority of cells (pinprick “si” or pinprick “sd” groups in particular) to be classified comparably in response to other stimulation modalities as they were in the pinprick (Fig. 3A,B). In support of this observation, the population responses (mean z-score across the peak response period) between the pinprick-defined cell groups remained significantly separated from one another for each stimulus (Fig. 3C, Extended Data Table 3-1). The response magnitudes of these cell groups (“si” and “sd” in particular) largely scaled with stimulus intensity, with innocuous stimuli (0.07, 0.4, 2 g, approach) eliciting lower magnitude differences on average compared with noxious stimuli (4 and 55°C and pinprick—though biased by initial selection criteria). Interestingly, approaching (but not touching) the hindpaw with a 2 g filament yielded low-magnitude but time-locked responses, perhaps indicating a visual sensory or contextual component of stimulus delivery. Additionally, the seemingly innocuous water stimuli (30°C) evoked average peak response magnitudes that were comparable to noxious (4, 55°C) water stimuli, perhaps as a result of either the large receptive fields activated by water droplets or a negative valence assigned to water stimuli in the context of our experiments. For most stimuli, cell groups displayed significantly higher percentages of responder classifications that matched their original responder classifications (defined by the pinprick) when compared with other cell groups (Fig. 3D–F, Extended Data Tables 3-2–3-4). When single “si” or “sd” cells deviated from their original classifications, they were more likely to be classified as “ns” than they were to switch identities between “si” and “sd,” suggesting that although the participation of single cells in the response to a particular stimulus varies, the response polarities (increased or decrease activity) of those cells are relatively stable when they are actively engaged (Fig. 3B,D–F, Extended Data Tables 3-2–3-4). The mean z-scores and percentages of “sd” cells scaled to stimulus intensity similarly to “si” cells, likely indicating the increased inhibition across the engaged ensemble necessary to counterbalance excitation (Fig. 3C–F). When we assessed the responder fractions of “si,” “ns,” and “sd” across the full populations of cells recorded under each condition (naive, CFA, GBP), only minimal differences were observable (Fig. 3G–I, Extended Data Tables 3-5–3-7). Though the percentages of “si” and “sd” cells classified for each mouse were larger after CFA compared with those after naive recordings for each stimulus and the percentages of “ns” cells were smaller, these differences did not reach statistical significance across animals. Surprisingly, however, GBP engaged a significantly larger proportion of “si” cells in response to noxious water stimuli (4, 55°C) and a significantly lower proportion in response to approach, with an opposite trend for “ns” cells (Fig. 3G,H). To compare the stability of the responder classes over days with and without pain or analgesia, we made a separate set of multimodal recordings in a different group of mice across 3 d with unchanging experimental conditions (no pain or analgesia). Among cells longitudinally tracked across recording conditions (naive–CFA–GBP or Day 1–Day 2–Day 3), we then determined the percentage of cells that changed their responder classifications across conditions for each stimulus (e.g., si–si–si, unchanged; si–ns–si, changed; Fig. 3J). For most stimuli (with the exception of 55°C water), we observed a similar extent of changes in responder classification between days (Day 1–Day 2–Day 3) and conditions (naive–CFA–GBP), suggesting intrinsic variability that is independent from pain or analgesia (Fig. 3J). Next, we assessed the fidelity of single-cell responder classifications across all stimuli. To do this, we categorized cells as “multimodal” if they exhibited an “si” or “sd” response to more than one stimulus, “unimodal” if they exhibited “si” or “sd” response to a single stimulus but “ns” to others, and “unresponsive” if classified as “ns” to all stimuli. Though 0.07, 0.4, and 2 g and approach might all be considered “innocuous” stimuli, they were all considered separately given that the relative engagement of each single-cell responder classification for these stimuli was unique (Fig. 3B). This analysis revealed that most ACC neurons recorded under the naive condition were multimodal (∼60%), while unimodal and (∼21%) and unresponsive (∼19%) cells comprised the remainder (Fig. 3K–M). Interestingly, we found increases in the percentage of multimodal cells comprising the population on average after CFA and GBP (with naive vs GBP reaching significance) and decreases in unimodal and unresponsive cells (though not significant), suggesting a trend toward a decreased fidelity of responder classifications across stimulation modalities after pain and analgesia (Fig. 3K–M).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Single-cell and population stability of response directionality across sensory stimuli in the ACC. A, Heatmaps of z-scored trial-averaged evoked GCaMP6f responses across each stimulus for each cell. Single cells within each cell group are ordered identically across each stimulation modality. Cells were initially segregated into groups based on responder classifications to pinprick stimulation and then tracked across all other stimuli (black, all cells; pink, pinprick “si” cells; violet, pinprick “ns” cells; light blue, pinprick “sd” cells). The population time course of these pin-defined cell groups are shown at the bottom for all stimuli. Each line color corresponds to the averaged responses of cell groups originally defined by the pinprick, with the shading indicating SEM. Single cells are aligned across the horizontal axis of the plot. B, As in A, but showing responder class stability across the stimulation modalities for responder classes defined by the pinprick. In this case, the “heat” of the maps corresponds to the responder classifications of each individual cell (bright red, “si”; gray, “ns”; dark blue, “sd”). C, Top: the averaged evoked GCaMP responses (the mean z-score 0–5 s after stimulation) in cell groups originally classified as “si” (pink), “ns” (violet), or “sd” (light blue) for pinprick stimulation, across all other stimuli. Error bars represent SEM. Bottom: grid plots indicate the p values for Tukey–Kramer comparisons between pinprick-defined cell classifications for each stimulus. For example, the top-right square (ns vs si, approach) indicates that the p value of the comparison of the z-scores of the pinprick-defined “ns” cell group versus the pinprick-defined “si” cell group in response to the approach stimulus is <0.01, while one square to the left (“ns” vs “si,” 30°C) indicates a p value <0.01 when comparing the same sets of cells but in response to 30°C water stimulation. See Extended Data Table 3-1 for additional details. D–F, Top panels: point plots showing the percentages of individual “si” (red color in panel B), “ns” (gray color in panel B), or “sd” (dark blue color in panel B) cells for each stimulus, within cell groups originally classified as pinprick “si” (pink color in panel B), pinprick “ns” (violet color in panel B), or pinprick “sd” cells (light blue color in panel B). Error bars indicate SEM. For example, in panel D the percentage of individual “si” (red color from panel B) cells for the pinprick stimulus is 100% for the pinprick-defined “si” cell group (pink) and 0% for the pinprick-defined “ns” (violet) and “sd” (light blue) cell groups, considering that the pinprick was the stimulus that originally defined the cell groups. In response to 4°C water, however, the pinprick “si” group (pink) is now composed of ∼60% “si” cells, while the pinprick-defined “ns” group (violet) is composed of ∼19% “si” cells, and the pinprick-defined “sd” group (dark blue) is composed of ∼8% of “si” cells. Percentage values across the point plots (D–F) sum to 100% for each individual group of cells [e.g., for the pinprick “si” cell group (pink) in response to 4°C water, percentage, si + % ns + % sd = 100%]. Bottom panels: grid plots indicating the p values from Tukey’s HSD comparisons between pinprick-defined cell classifications for each stimulus, corresponding to the point plots above each. See Extended Data Tables 3-2–3-4 for additional details. G–I, Point plots indicating the percentages of cells classified into “si,” “ns,” or “sd” for each recording condition and statistical grid plots indicating p values from Tukey’s HSD comparisons of these values between conditions for each stimulus. See Extended Data Tables 3-5–3-7 for additional details. J, Point plots indicating the percentage of cells that changed their responder classification across conditions (naive–CFA–GBP, 12 mice) or across recording days with unchanged experimental parameters (Day 1–Day 2–Day 3, 4 mice; Welch's t test; conditions vs days; 55°C, p = 0.0014). K, Left panel: heatmap indicating the responder classifications of single cells (bright red, “si,”; gray, “ns”; dark blue, “sd”) across stimuli, determined to be multimodal in their responses. Right panel: a point plot indicating the percentages of cells in each mouse characterized as multimodal (Tukey’s HSD test; naive vs GBP, p < 0.05). L, As in K, but for unimodal cells. M, As in K, but for unresponsive cells.

Flexible stimulus preferences of single ACC neurons across days

Next, we assessed the degree to which single ACC neurons exhibit stimulus-specific selectivity and whether that selectivity changes under pain or analgesic conditions. For each cell recorded within each condition (naive, CFA, GBP), we defined a preferred stimulus and computed a selectivity index based on the mean response magnitudes to all other stimuli (see Materials and Methods, Analysis of stimulus selectivity; Simmons et al., 2007). Activity heatmaps of mean z-scores during the peak response period revealed subpopulations of cells selective for individual stimuli via this definition (Fig. 4A). Most selective neurons were classified as “si” or “ns,” though in rare cases an “sd” cell could be considered selective so long as the trial-averaged response to the selective stimulus was greater than for all other stimuli (Fig. 4A). Some cells could be found that were highly selective for a particular stimulus, though these were quite rare with only ∼5–6% of cells per recording condition having a selectivity index >0.7 (Fig. 4B). Across recording conditions of pain and analgesia (naive–CFA–GBP) or days with unchanged experimental parameters (Day 1–Day 2–Day3), longitudinally tracked single cells exhibited substantial flexibility in their selectivity preferences (Fig. 4C,D). Only 2.7% (naive–CFA–GBP) to 2.9% (Day 1–Day 2–Day3) of longitudinally tracked cells maintained their stimulus preferences across these recording conditions, with no significant differences observed between mice recorded across pain/analgesia and those recorded across unchanged days (Fig. 4E). This highly variable selectivity is consistent with previous reports suggesting dynamic stimulus representations in single cells across time in the ACC (Acuna et al., 2023). When characterizing the magnitude of selectivity across recording conditions, only a mildly significant increase was observed for 4°C water stimuli after CFA compared with those after naive recordings (Fig. 4F). GBP also significantly decreased selectivity magnitude in response to the pinprick compared with naive recordings while significantly increasing it for noxious water stimuli (4, 55°C; Fig. 4F, Extended Data Table 4-1). We did not find significant differences in the percentages of cells selective to any stimulus after CFA compared with that after naive recordings (Fig. 4G, Extended Data Table 4-2). After GBP treatment, however, significant increases in the percentages of cells selective for 4 and 55°C were observed compared with the naive and CFA (55°C only) conditions (Fig. 4G, Extended Data Table 4-2). A trend of decreased selectivity was observed for all other stimuli after GBP, with significant decreases observed for 0.07 g and pinprick stimuli compared with naive and 0.07 and 2 g stimuli compared with CFA (Fig. 4G, Extended Data Table 4-2). Next, we utilized a naive Bayes decoder to assess the discriminability of multimodal stimuli across all recorded cells for each condition. We first tested broad classifications of innocuous (0.07, 0.4, 2 g, 30°C water) or noxious (4, 55°C water, pinprick) stimuli, finding reasonable decoding accuracies above chance levels. These decoding accuracies were significantly higher than shuffled data for all recording conditions (Fig. 4H,J). The decoding of single stimuli yielded lower accuracy classifications that remained above the chance level (Fig. 4I,K), with lower accuracies overall than other recent examples in the ACC (Acuna et al., 2023). This may partially be explained by limiting our stimulations to the right hindpaw, whereas Acuna et al. stimulated different body parts, that is, cold, pinprick, and touch were applied to the plantar surface, heat was applied to the proximal portion of tail, and air puff was applied to either side of the animals’ face. Similar decoding performance to ours was reported in the BLA ensembles for the essentially the same stimulation paradigms used here (Corder et al., 2019). The reported decoding performance by Acuna et al. was slightly, but significantly, reduced under the SNI condition and then restored by GBP treatment. In our study, changes in the decoding performance under the pain condition were not observed (Fig. 4J,K). These results demonstrate the capability of the ACC to encode a variety of multimodal stimuli at the population and single-cell level. However, response polarity and stimulus selectivity are highly stochastic across recording days, suggesting population encoding that relies on continuously flexible, rather than fixed, single-cell representations of sensory stimuli.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Stochastic tuning to sensory stimuli in ACC neurons. A, Left panel: example heatmap of mean z-scored values within the poststimulus response period (0–5 s) for cells that were preferentially selective to 2 g mechanical stimuli. Right panel: responder classifications across the stimuli for 2 g selective cells, ordered identically to the left panel (bright red, “si”; gray, “ns”; dark blue, “sd”). B, Point plots of mean z-scored values within the poststimulus response period (0–5 s), for single-cell examples with high and low selectivity indices. C, An alluvial plot demonstrating changes to stimulus preference across recording conditions (naive, CFA, GBP) for longitudinally tracked cells. D, As in C, but for cells tracked across days with unchanged experimental parameters (Day 1, Day 2, Day 3). E, Pie charts indicating percentages of longitudinally tracked single cells that changed their preferred stimulus across all recording conditions (naive–CFA–GBP, 696 cells, 7 mice) and across days with unchanged experimental parameters (Day 1–Day 2–Day 3, 516 cells, 4 mice). F, Left panel: a point plot of selectivity magnitude among all cells considered selective for each stimulus and condition. Right panel: a statistical grid plot of p values from Tukey–Kramer comparisons between conditions for each stimulus (cyan, naive 1,680 cells; magenta, CFA 1,625 cells; green, GBP 1,284 cells, where N = subpopulations of these selective for each stimulus). See Extended Data Table 4-1 for additional details. G, As in F, but comparing the percentage of cells selective for each stimulus (within each mouse) for each recording condition with a Tukey’s HSD test (N = 12 mice/condition). See Extended Data Table 4-2 for additional details. H, Top panels: confusion matrices indicating decoding accuracies across recording conditions for innocuous (0.07, 0.4, 2 g and 30°C water) and noxious (pinprick, 55 and 4°C water) stimulus categories trained using a Gaussian naive Bayes classifier. Bottom panels: as in top, but for shuffled labels. I, As in H, but decoding individual stimuli. J, Decoding accuracies of the original and shuffled data for each condition, using the broad decoding categories of innocuous versus noxious (Mann–Whitney U test; N = 12 mice; naive, p = 3.644 × 10⁻⁵; CFA, p = 0.0001; GBP, p = 3.65 × 10⁻⁵). K, As in J, but for single stimulus decoding (Mann–Whitney U test; N = 12 mice; naive, t = 144, p = 3.644 × 10⁻⁵; CFA, t = 144, p = 3.658 × 10⁻⁵, GBP, t = 144, p = 3.644 × 10⁻⁵).

Limited excitability changes and nociceptive engagement in the ACC after CFA

The use of conscious calcium imaging recordings in preclinical pain research necessitates robust neurophysiological hallmarks of pain and analgesia. The prevailing view of pain as a condition of hyperexcitability would suggest generalized increases in activity, and indeed small but significant increases in firing rates and calcium transients have been reported in the ACC of rat CFA models and mouse SNI models, respectively (Zhao et al., 2018; Zhou et al., 2018). To search for similar changes in our recordings, we first assessed the raw ΔF/F recorded during both the prestimulus baseline (−5 to 0 s) and the poststimulus (evoked) response period (0–5 s) for all recorded cells compared across conditions (Fig. 5A). In addition to the naive, CFA, and GBP recording conditions, we included an additional set of recordings (in five mice) where GBP was treated 4 d prior to naive recordings (GBP-Ctrl, 75 mg/kg oral gavage) to assess the effects of GBP without CFA. Given this experimental time course, residual GBP (GBP-Ctrl) before subsequent recording sessions (naive, CFA, GBP) was negligible (Radulovic et al., 1995) and allowed us to compare excitability changes driven by GBP before and after pain. Though the magnitude was low, we found surprisingly that the prestimulus baseline activity (ΔF/F) across all cells was significantly lower after CFA compared with that after naive recordings for all stimuli except 30 and 55°C water, significantly lower compared with that after GBP for pin and 0.4 g, and significantly lower compared with that after GBP-Ctrl for 55 and 4°C stimuli (Fig. 5A, Extended Data Table 5-1). In the poststimulus period, we found no significant differences in evoked ΔF/F after CFA compared with those after naive but found significant increases for several stimuli after GBP or GBP-Ctrl compared with those after naive or CFA recordings (Fig. 5A, Extended Data Table 5-2). When restricting the analysis to exclusively “si” cells, we did not find significant differences in ΔF/F for any stimuli for naive versus CFA recordings but found significant increases after GBP and GBP-Ctrl for the pinprick and 4°C stimuli (Fig. 5B, Extended Data Table 5-3). Similarly to the analysis of all cells, “si” cells showed significant increases in evoked activity for several stimuli after GBP and GBP-Ctrl, with only 4°C showing significantly increased activity after CFA compared with that after naive (Fig. 5B, Extended Data Table 5-4). Next, we assessed the relative rather than absolute response strength of the poststimulus response period [mean z-score (0–5 s) across all recorded cells; Fig. 5C, Extended Data Table 5-5]. Compared with naive recordings, CFA caused low-magnitude but significant increases in evoked activity in response to all stimuli except the pinprick, with increased responses to approach perhaps indicating increased pain expectation (Fig. 5C, Extended Data Table 5-5). GBP yielded mixed results, with significant increases in response to 0.07 g (vs naive) and all water stimuli (vs naive, CFA, and GBP-Ctrl) and significant decreases in response to approach (vs naive and CFA) and 2 g (vs CFA) stimuli (Fig. 5C, Extended Data Table 5-5). GBP-Ctrl recordings were largely indistinguishable from naive with the exception of stronger pinprick responses and yielded significantly lower magnitude responses to several stimuli compared with CFA recordings (0.07 and 2 g, 4°C, approach). A similar analysis performed on exclusively “si” cells did not yield significant differences in mean response strength between the GBP-Ctrl and naive or CFA conditions for any stimulus, and only 55°C water evoked a significant increase after GBP compared with that after GBP-Ctrl (Fig. 5D, Extended Data Table 5-6). After CFA, we found small but significant increases in “si” cell responses to 4°C water and 2 g stimuli compared with naive recordings, with 2 g responses also significantly larger in CFA compared with those in GBP (Fig. 5D, Extended Data Table 5-6). After GBP, both 55 and 4°C water evoked significantly larger responses than in naive recordings, with the response to 55°C also being significantly larger than for CFA (Fig. 5D, Extended Data Table 5-6). Approach stimuli evoked significantly weaker responses after GBP compared with those after both naive and CFA recordings, perhaps due to a disruption in visual or contextual information by GBP-induced drowsiness (Fig. 5D, Extended Data Table 5-6). Next, we performed a search for longitudinally tracked cells that ramped up their activity after CFA and subsequently ramped down their activity after GBP (“up_down” cells) by comparing the response magnitudes of single cells within the peak response period (Fig. 5E). Doing so revealed the existence of cell groups with this activity pattern for all stimuli, where response magnitudes are significantly increased after CFA with no significant differences between naive and GBP recordings (Fig. 5F, Extended Data Table 5-7). However, similar groups of “up_down” cells could also be found in control recordings across days without pain or analgesic manipulations (Fig. 5G,H, Extended Data Table 5-8). Furthermore, direct comparisons of the proportions (percentage of “up_down” cells) and response magnitudes (z-score 0–5 s) of “up_down” cells observed across conditions (naive–CFA–GBP) and days (Day 1–Day 2–Day 3) yielded no significant differences for all stimuli except one (approach), suggesting that the “up_down” activity patterns driven by pain and analgesia cannot be disambiguated from the intrinsic variability observed across recording days (Fig. 5I,J).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Comprehensive search for changes in ACC activity after CFA-induced inflammatory pain. A, Left panel: point plots showing the mean ΔF/F during the prestimulus baseline (−5 to 0 s) and poststimulus response period (0–5 s) for all recorded cells compared across conditions (orange, GBP-Ctrl; cyan, naive; magenta, CFA; green, GBP) for each stimulus. Right panel: statistical grid plots indicating the p values from Tukey–Kramer comparisons of z-score between recording conditions for each stimulus. See Extended Data Tables 5-1 and 5-2 for additional details. B, As in A, but considering “si” cells only. See Extended Data Tables 5-3 and 5-4 for additional details. C, Left panel: a point plot showing the mean z-score during the poststimulus response period (0–5 s) for all recorded cells compared across conditions (orang, GBP-Ctrl; cyan, naive; magenta, CFA; green, GBP) for each stimulus. Right panel: a statistical grid plot indicating the p values from Tukey–Kramer comparisons of z-score between recording conditions for each stimulus. See Extended Data Table 5-5 for additional details. D, As in C, but considering “si” cells only. See Extended Data Table 5-6 for additional details. E, Example heatmaps of z-scored trial-averaged GCaMP responses evoked by pinprick stimulation for longitudinally tracked cells identified by increases in activity from naive to CFA and decreases in activity from CFA to the GBP (“up_down” cells). All cells are ordered identically within each heatmap across the recording conditions. F, Left panel: a point plot showing the mean z-score during the poststimulus response period (0–5 s) for “up_down” cells across conditions (cyan, naive; magenta, CFA; green, GBP) for each stimulus. Right panel: a statistical grid plot indicating the resultant p values for Tukey–Kramer comparisons of z-score between conditions for each stimulus. See Extended Data Table 5-7 for additional details. G, As in E, but across days with unchanged experimental parameters. H, As in F, but across unchanged days (Day 1–Day 2–Day 3). See Extended Data Table 5-8 for additional details. I, A point plot comparing the percentages of cells classified as “up_down” cells between conditions and days. No significant differences were observed (Welch's t test). J, As in I, but comparing z-scores of “up_down” cells (Welch's t test, conditions vs days: approach, p = 0.031). K, Nociceptive ensembles were defined in each recording condition as a union of sets of cells with significantly increased fluorescence responses (p < 0.01) to either pinprick, noxious heat (55°C water), or noxious cold (4°C water). L, Nociceptive ensembles defined within each recording condition. The percentages of longitudinally tracked cells (7 mice, 696 cells) are indicated for each portion of the Venn diagrams M, Venn diagrams indicating nociceptive engagement in response to innocuous stimuli. Light gray: the percentage (out of total cells) of nociceptive cells (“si” to noxious stimuli) not showing significantly increased activity in response to a particular innocuous stimulus. Medium gray: the percentage of cells showing significant increases and also part of the nociceptive ensemble defined in L. Dark gray: the percentage of cells with significantly increased activity and not part of the nociceptive ensemble. Dashed red circle: cells comprising the total nociceptive ensemble defined in L. N, Statistical analysis of the ratio of shared (medium gray)/nociceptive (dashed red circle) cells for various innocuous stimuli across recording conditions. Selected comparisons for Tukey’s HSD multiple comparison of means (2 g stimuli, Naive vs CFA; p = 0.004; CFA vs GBP, p = 0.0159; N = 7 mice for each comparison).

Another strategy to estimate pain among neural populations is to measure the relative engagement of a defined population of cells rather than population response magnitudes. This concept was first demonstrated in the amygdala, where cells demonstrating significantly increased responses to one or more noxious stimuli were considered as a nociceptive ensemble (Corder et al., 2019). We applied the same concept to our ACC recordings, where the unions of “si” cells to the pinprick and 55 and 4°C in each condition (naive, CFA, and GBP) were defined as a nociceptive ensembles (total longitudinally tracked cell number, 696; nociceptive cell counts: naive, 169; CFA, 230; GBP, 224 in seven animals; Fig. 5K,L). Next, we assessed whether the “si” cells evoked by each innocuous stimulation were part of the nociceptive (nox) ensemble and calculated the percentages of the total longitudinally tracked cells that were exclusively nociceptive, stimulus evoked and nociceptive, or stimulus evoked and not nociceptive (Fig. 5M). We observed small increases in the nociceptive engagement of neurons evoked by innocuous stimulations after CFA, which subsequently decreased after GBP (Fig. 5M). We then calculated the fraction of the overlapping “si” cell number evoked by each innocuous stimulations to the cell counts of the noxious ensemble, observing mild but significant increases in the overlapping neuronal populations responsive to only 2 g von Frey stimulation, but not for 0.4 or 0.07 g stimulations (Fig. 5N). This observation was less pronounced than results observed in the BLA using the sciatic nerve injury pain model (Corder et al., 2019). In summary, our results revealed only limited increases in ACC population activity driven by CFA. Furthermore, while we could specifically identify subpopulations of cells that ramped up their activity after CFA and subsequently ramped down after GBP, these activity patterns could not be disambiguated from random variability across recording sessions (Fig. 5E,G).

Temporal sharpening and increases in correlated activity in the presence of analgesia after inflammatory pain

In the population average time course of ACC “si” neuron activity, we noticed a temporal sharpening of responses to certain stimuli after GBP treatment compared with the GBP-Ctrl, naive, or CFA conditions (Fig. 6A). To capture these differences, we computed the total duration that single cells remain 1.5 standard deviations above the prestimulus baseline period, providing a quantitative estimate of sustained excitability. Distributions of these response durations across all stimuli revealed a shift toward shorter durations after GBP compared with those after the naive, CFA, or GBP-Ctrl recording conditions (Fig. 6B). Significant decreases in response duration were observed for most stimuli (except 4°C) after GBP compared with those after the naive recording condition and for most stimuli (except for 0.07, 55°C) compared with CFA (Fig. 6C, Extended Data Table 6-1). No significant differences in duration were observed between the naive and GBP-Ctrl conditions, while duration was significantly lower in GBP compared with GBP-Ctrl for several stimuli (2 g, pinprick, 4 and 55°C; Fig. 6C, Extended Data Table 6-1). Furthermore, response durations after GBP were significantly lower for most stimuli (except 0.07 g) compared with control recordings across days without pain or analgesic manipulations (Fig. 6D,E). To determine if this sharpening of response duration generalized to other mechanistically distinctive analgesics, we imaged a separate set of mice responding to multimodal stimulation that were dosed with ibuprofen (IBP, an NSAID), instead of GBP (a gabapentinoid). Similar to the effects of GBP, IBP caused a shift in the distribution of response durations across stimuli toward smaller values (Fig. 6F). While only 0.07 g stimuli yielded a difference in response durations between the naive and CFA conditions for these experiments, durations were lower after IBP compared with naive for all stimuli, with significant reductions observed for all but two stimuli (2 g, pinprick; Fig. 6G, Extended Data Table 6-2). IBP did not change response durations compared with CFA with a single exception, where 55°C responses were significantly reduced (Fig. 6G, Extended Data Table 6-2). Similar to GBP, we found no significant differences in response durations between naive and IBP-Ctrl recordings, minimal differences between CFA and IBP-Ctrl recordings (significant decrease for 30°C after CFA only), and significantly higher durations for IBP-Ctrl compared with IBP for several stimuli (all but approach; Fig. 6G, Extended Data Table 6-2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Temporal sharpening of response durations driven by analgesia after CFA but not before. A, Time course of “si” cell activity across recording conditions for 0.07 and 2 g stimuli. B, Probability distribution across all stimuli of the total duration that each “si” cell activity trace remains 1.5 standard deviations above the prestimulus baseline period (−5 to 0 s) within a 15 s poststimulus response period (0–15 s) for the GBP-Ctrl, naive, CFA, and GBP conditions. C, The point plot of the duration of thresholded activity corresponding to panel B and statistical grid plot comparing durations across conditions for each stimulus (Tukey–Kramer test). See Extended Data Table 6-1 for additional details. D, Probability distribution of duration as in panel B, but across days with unchanged experimental parameters. E, A point plot comparing the duration of thresholded activity across unchanged days and GBP (Welch's t test, 0.07 g, t = 1.68; p = 0.092; 0.4 g, t = 3.17; p = 0.001; 2 g, t = 9.10; p = 6.486 × 10⁻¹⁹; pinprick, t = 5.53; p = 4.189 × 10⁻⁸; 4°C, t = 2.88; p = 0.004; 55°C, t = 3.17; p = 0.001; 30°C, t = 7.88; p = 8.843 × 10⁻¹⁵; approach, t = 3.02; p = 0.002). F, As in panel B and D, but for recordings made with IBP (40 mg/kg) instead of GBP (IBP-Ctrl, naive, CFA, and IBP). G, A point plot of the duration of thresholded activity corresponding to panel F and statistical grid plot comparing durations across conditions for each stimulus (Tukey–Kramer test). See Extended Data Table 6-2 for additional details.

Next, we asked whether temporal sharpening could be indicative of correlated ACC activity. This was of particular interest in the context of analgesia when considering that correlated activity is thought by some to limit information in the brain, though this idea is debatable (Bartolo et al., 2020; Valente et al., 2021; Hazon et al., 2022). To explore this idea in the context of our experiments, we searched for covariations in activity across ACC neural populations driven by pain or analgesia. Pearson's correlations of ΔF/F were computed across trials between all pairs of simultaneously recorded neurons for each individual stimulus within each animal. To capture correlations relevant to stimulus-evoked activity, we only considered 5 s poststimulus windows (0–5 s) for each trial, where the peak activity is observed to occur in the population activity time course. We then concatenated these windows and correlated ΔF/F across time between all neuron pairs. This resulted in probability distributions of Pearson's correlation coefficients for any particular stimulus that were largely indistinguishable between recording conditions (naive–CFA–GBP) across all neuron pairs (Fig. 7A). However, the probability distributions of coefficients from exclusively “si”–“si” neuron pairs were shifted toward higher values after GBP compared with any other condition (Fig. 7A). The mean Pearson's correlation coefficients of all “si”–“si” pairs after GBP were significantly increased when compared with the naive, CFA, and GBP-Ctrl conditions for all stimuli except 0.07 g (not significant GBP vs GBP-Ctrl) and approach (not significant for GBP vs CFA or GBP vs GBP-Ctrl; Fig. 7B, Extended Data Table 7-1). While small but significant increases in correlation coefficients were observed for select stimuli after CFA compared with those after naive recordings (0.07 g, pinprick, 4 and 30°C, approach), these differences were not maintained when comparing coefficients averaged within each mouse (Fig. 7C, Extended Data Table 7-2). These results indicate, therefore, that GBP robustly increases trial–trial correlations among “si”–“si” cell pairs and between mice after (but not before) CFA injections for all but the most innocuous stimuli (0.07 g GBP vs GBP-Ctrl, approach for all comparisons; Fig. 7C, Extended Data Table 7-2). Additionally, we performed this analysis on the prestimulus period (−5 to 0 s), finding significantly elevated correlation coefficients in the presence of GBP after CFA (but not before) to some stimuli (pinprick, 4, 55, 30°C), while no significant differences were found between comparisons of other recording conditions (naive vs CFA vs GBP-Ctrl; Fig. 7D, Extended Data Table 7-3). However, the magnitude of these coefficients was significantly lower compared with the poststimulus period for all stimuli besides approach (Fig. 7E). To control for any differences driven by repeated sensory experience, we performed the same correlation analysis on a set of animals recorded over 3 d without any pain or analgesic manipulations (Day 1–Day 2–Day 3). In this case, we did not observe a rightward shift in the probability distributions of correlation coefficients for the poststimulus response period among all cell pairs or exclusively “si”–“si” pairs, and only low-magnitude differences in these coefficients were observed among “si”–“si” pairs between days (Fig. 7F,G, Extended Data Table 7-4). Though several significant (but low-magnitude) differences were found among “si”–“si” cell pairs between days, these differences were not maintained when comparing averaged values across mice (Fig. 7H, Extended Data Table 7-5). In the prestimulus baseline period, we found significantly increased correlations for 55°C water stimuli on Day 3 compared with those on Day 1 and Day 2, but did not find differences for any other stimulus or recording day (Fig. 7I, Extended Data Table 7-6). When comparing the prestimulus versus poststimulus periods, we found minor but significant differences (p = ∼0.02) for three stimuli on Day 1 only (Fig. 7J). Finally, we asked whether the increased correlations were specific to GBP or if they generalized to other analgesics. Toward this end, we performed the same analysis on recordings from an experimental set where GBP was replaced with IBP. Surprisingly, we saw a rightward shift in the probability distributions of correlation coefficients in the presence of IBP after (but not before) CFA (Fig. 7K). This shift was reminiscent of that caused by GBP and was also clearly observable across the distribution of all correlation pairs in addition to “si”–“si” pairs alone (Fig. 7K). Similarly, we found low-magnitude differences in the correlation coefficients across cell pairs between the IBP-Ctrl, naive, and CFA conditions which were not maintained when comparing mouse-averaged values (Fig. 7L,M, Extended Data Tables 7-7, 7-8). However, high magnitude and significant increases in correlation coefficients were observable from treatment with IBP after (but not before) CFA which were maintained when comparing mouse-averaged values (Fig. 7L,M, Extended Data Tables 7-7, 7-8). Similarly, recordings during IBP after CFA (but not before) during the prestimulus period showed significantly increased correlations for most stimuli (no significant differences for approach) compared with all other recording conditions (Fig. 7N, Extended Data Table 7-9). Unlike GBP, however, IBP did not show significant increases in these correlation coefficients poststimulus compared with prestimulus (Fig. 7O). These results reveal robustly increased correlations among neurons excited (“si” cells) by a variety of stimuli in the presence of GBP only after pain. Correlations among neural populations are typically very low given the diversity of neural dynamics underlying the encoding of information in the brain, with the ∼0.05 typically seen in our data falling well within the range of commonly reported values (∼0.01–0.20; Cohen and Kohn, 2011; Hazon et al., 2022). Therefore, increases to ∼3–4 times this value after GBP are surprisingly large in this context. Even more surprisingly, these results generalized to another analgesic that exerts its effects through separate mechanisms, potentially suggesting a neurophysiological hallmark of pain attenuation by analgesia in the ACC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Increased correlations of evoked activity driven by analgesia after CFA but not before. A, Left panel: example probability distribution of Pearson's correlation coefficients measuring trial-to-trial covariations (z-score in the 5 s poststimulus period across all trials) in activity for all pairs of recorded cells (Pearson's r for all combinations of “si,” “ns,” and “sd” single-cell pairs in each mouse) for a single stimulus (2 g) and for each recording condition (GBP-Ctrl, 16,832 pairs; naive, 168,188 pairs; CFA, 148,403 pairs; GBP, 101,887 pairs). Right panel: as in the left panel, but for “si” to “si” cell pairs only. B, Left panel: a point plot of mean correlation coefficients across stimuli for “si” to “si” cell pairs for each condition (error bars represent SEM). Right panel: a statistical grid plot indicating p values from cross-condition Tukey–Kramer comparisons of correlation coefficients. See Extended Data Table 7-1 for additional details. C, As in B, but comparing correlation coefficients averaged from pairs within each mouse. See Extended Data Table 7-2 for additional details. D, As in C, but comparing mouse-averaged correlation coefficients from the prestimulus baseline period. See Extended Data Table 7-3 for additional details. E, As in D, but comparing correlation coefficients from the prestimulus (−5 to 0 s) versus poststimulus (0–5 s) periods within each recording condition (naive, CFA, GBP, GBP-Ctrl). Mann–Whitney U test, selected comparisons, GBP pre- versus poststimulus: 0.07 g, t = 95; p = 0.025. 0.04 g, t = 105; p = 0.003. 2 g, t = 115; p = 0.014. Pinprick, t = 119; p = 0.007. 4°C, t = 107; p = 0.046. 55°C, t = 112; p = 0.022. 30°C, t = 108; p = 0.040. GBP-Ctrl, t = 24. Pinprick, p = 0.015. F, Left panel: as in A, but for a separate experimental set where animals were recorded for 3 d without any pain or analgesic manipulations (Day 1, 94,563 pairs; Day 2, 87,788 pairs; Day 3, 72,044 pairs). Right panel: as in the left panel, but for “si” to “si” cell pairs only. G, A point plot of mean correlation coefficients across stimuli for “si” to “si” cell pairs for each recording day (error bars represent SEM). Left panel: a statistical grid plot indicating p values from cross-day Tukey–Kramer comparisons of correlation coefficients. See Extended Data Table 7-4 for additional details. H, As in G, but comparing correlation coefficients averaged from pairs within each mouse. See Extended Data Table 7-5 for additional details. I, As in H, but comparing mouse-averaged correlation coefficients from the prestimulus baseline period. See Extended Data Table 7-6 for additional details. J, As in I, but comparing correlation coefficients from the prestimulus (−5 to 0 s) versus poststimulus (0–5 s) periods within each recording condition (Day 1, Day 2, Day 3). Mann–Whitney U test, selected comparisons, Day 1 pre- versus poststimulus: 0.4 g, t = 16; p = 0.028. 55°C, t = 16; p = 0.028. 30°C, t = 16; p = 0.028. K, Left panel: as in A and F, but for a separate experiment where animals were treated with IBP (40 mg/kg) instead of GBP (IBP-Ctrl, 34,633 pairs; naive, 27,504 pairs; CFA, 34,657 pairs; IBP, 21,974 pairs). Right panel: as in the left panel, but for “si” to “si” cell pairs only. L, Left panel: a point plot of mean correlation coefficients across stimuli for “si” to “si” cell pairs for each condition (error bars represent SEM). See Extended Data Table 7-7 for additional details. Right panel: a statistical grid plot indicating p values from cross-condition Tukey–Kramer comparisons of correlation coefficients. See Extended Data Table 7-7 for additional details. M, As in L, but comparing correlation coefficients averaged from pairs within each mouse. See Extended Data Table 7-8 for additional details. N, As in M, but comparing mouse-averaged correlation coefficients from the prestimulus baseline period. See Extended Data Table 7-9 for additional details. O, As in N, but comparing correlation coefficients from the prestimulus (−5 to 0 s) versus poststimulus (0–5 s) periods within each recording condition (naive, CFA, IBP, IBP-Ctrl). Mann–Whitney U test, selected comparisons, naive pre- versus poststimulus: 55°C, t = 16; p = 0.028. IBP-Ctrl: 4°C, t = 16; p = 0.028.

Locomotion scales with stimulus intensity and is attenuated by GBP

Locomotion and paw withdrawal have long been used as behavioral measures to assess stimulus intensity and pain in rodent models. During multimodal imaging experiments, we made observational assessments as to whether a paw withdrawal was evoked by a particular stimulus. Furthermore, we utilized the accelerometer integrated within the miniaturized microscopes to record total body acceleration simultaneously with the neurophysiology time series. Using these measures, we asked how paw withdrawals and the intensity of escape-like behaviors change across stimulus intensity and pain or analgesic conditions. Considering that locomotion has been shown to decrease correlated neural activity in cortical systems, we also sought to determine if the results of our cross-trial correlation analysis could be explained by differences in locomotion (Dadarlat and Stryker, 2017). Within any recording condition, higher percentages of hindlimb withdrawals were typically associated with higher total acceleration values (Fig. 8A). Noxious cold and hot water (4, 55°C) evoked significantly higher total acceleration than most other stimuli within any recording condition, while no significant differences could be found among innocuous mechanical stimuli (0.07, 0.4, 2 g; Fig. 8B,C, Extended Data Table 8-1). However, acceleration values across animals roughly scaled with stimulus intensity within all recording conditions (Fig. 8D, Extended Data Table 8-2). No significant differences in total acceleration after CFA compared with those after naive recordings could be found for any stimulus (Fig. 8D, Extended Data Table 8-2). GBP, however, facilitated a significant decrease in total acceleration responses to every stimulus compared with naive recordings after CFA (GBP) and nearly every stimulus compared with naive recordings when dosed before CFA (GBP-Ctrl; Fig. 8D, Extended Data Table 8-2). Importantly, no significant differences in acceleration were observed between recordings where GBP was dosed before or after CFA (GBP vs GBP-Ctrl), suggesting that the differences in neural activity correlations between these conditions described previously cannot be explained by decreased locomotion alone (Fig. 8D, Extended Data Table 8-2). Innocuous mechanical stimuli (0.07, 0.4 g) were significantly more likely to evoke a paw withdrawal after CFA compared with that after naive recordings (2 g borderline significant, p = 0.053; Fig. 8E, Extended Data Table 8-3). Additionally, we found significantly lower withdrawal likelihood in the presence of GBP before CFA (GBP-Ctrl, all stimuli except approach), as well as after CFA (GBP, all stimuli except the pinprick, 55°C, and approach) compared with CFA recordings, consistent with concurrent measurements of PWT (Fig. 8E, Extended Data Table 8-3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Escape acceleration and reflexive withdrawal scale with stimulus intensity, while GBP robustly attenuated acceleration. A, Scatterplots showing mean total acceleration values on the X axis (averaged within 5 s poststimulus window for each mouse), against the percentage of reflexive withdraws observed across trials on the Y axis for each stimulus within each recording condition. Errors bars across the X and Y axis indicate SEM. B, Mouse-averaged time course of total acceleration for each recording condition for selected stimuli. Shading indicates SEM. C, Statistical grid plots indicating p values from Tukey’s HSD tests for comparisons of total acceleration between stimuli within each recording condition. See Extended Data Table 8-1 for additional details. D, Left panel: mean acceleration values from the 5 s poststimulus window for each condition and stimulus. Right panel: statistical grid plot indicating p values from Tukey’s HSD tests for comparisons of total acceleration between conditions for each stimulus [GBP-Ctrl (orange), N = 4 mice; naive (cyan), N = 8 mice; CFA (magenta), N = 8 mice; GBP (green), N = 8 mice]. See Extended Data Table 8-2 for additional details. E, Left panel: mean percentage withdrawal values for each condition and stimulus. Right panel: a statistical grid plot indicating p values from Tukey’s HSD tests for comparisons of percentage withdrawal between conditions for each stimulus [GBP-Ctrl (orange), N = 4 mice; naive (cyan), N = 8 mice; CFA (magenta), N = 8 mice; GBP (green), N = 8 mice]. See Extended Data Table 8-3 for additional details.