Chemogenetic modulation of posterior insula CaMKIIa neurons alters pain and thermoregulation

Abstract

The posterior insular cortex (PIC) is well positioned to perform somatosensory-limbic integration; yet, the function of neuronal subsets within the PIC in processing the sensory and affective dimensions of pain remains unclear. Here, we employ bidirectional chemogenetic modulation to characterize the function of PIC CaMKIIa-expressing excitatory neurons in a comprehensive array of sensory, affective, and thermoregulatory behaviors. Excitatory pyramidal neurons in the PIC were found to be sensitized under inflammatory pain conditions. Chemogenetic activation of excitatory CaMKIIa-expressing PIC neurons in non-injured conditions produced an increase in reflexive and affective pain- and anxiety-like behaviors. Moreover, activation of PIC CaMKIIa-expressing neurons during inflammatory pain conditions exacerbated hyperalgesia and decreased pain tolerance. However, Chemogenetic activation did not alter heat nociception via hot plate latency or body temperature. Conversely, inhibiting CaMKIIa-expressing neurons did not alter either sensory or affective pain-like behaviors in non-injured nor under inflammatory pain conditions, but it did decrease body temperature and decreased hot plate latency. Our findings reveal that PIC CaMKIIa-expressing neurons are a critical hub for producing both sensory and affective pain-like behaviors and important for thermoregulatory processing.

Introduction

Pain is a multidimensional experience comprised of both emotional-affective and sensory-discriminatory components⁵² resulting from the activation of a network of brain structures including the insular cortex (IC)^{3, 55, 72}. Electrical stimulation of the IC evokes pain-like sensations^{1, 47, 54}, and human brain functional imaging studies demonstrate IC activation in response to noxious somatosensory stimuli^{8, 14, 19}, including in a manner correlated with stimulus intensity^{6, 18, 29}.Some clinical case reports suggest that patients with insular lesions exhibit intact sensory-discriminative processing but disrupted pain unpleasantness^{11, 37}, while other reports showed that patients with IC lesions retained the ability to rate the unpleasantness of nociceptive stimuli but exhibited increased pain sensitivity⁶³. These contrasting results from gross lesions suggest that the insula may process selective dimensions of pain dependent on finer anatomical loci or specific cell subgroups within the cortex. Currently, the understanding of cellular functions in the IC that mediate the sensory and/or affective dimensions of pain remains poorly resolved.

The IC is commonly divided into anterior and posterior portions based on cytoarchitecture and connectivity^{31, 34, 36}. While there is evidence that both the anterior IC (AIC) and posterior IC (PIC) are involved in pain processing^{14, 60, 72}, nociceptive responses within the IC originate in PIC and are then observed in the AIC, suggesting that signaling proceeds from PIC to AIC²⁸. Moreover, the PIC receives direct projections from the posterior thalamus, which encodes nociceptive, pruriceptive, and innocuous tactile somatosensory information^{22, 24, 25, 32, 44}. The PIC also has reciprocal connections to areas underlying emotional processing such as the medial prefrontal cortex and the amygdala^{30, 49, 58, 68} and sends outputs to regions involved in affect, reward, and the regulation of motivated behavior, including the nucleus accumbens and the ventral bed nucleus of the stria terminalis^{33, 35}. The PIC is also suggested to integrate interoceptive signals with external stimuli to drive appropriate behaviors in response to a disrupted homeostatic or physiological state^{2, 33, 34, 45, 62}. Thus, the PIC is well positioned to perform multisensory integration.

Modulation of insular alpha calcium/calmodulin-dependent protein kinase II (CaMKIIa)-expressing neurons, which are primarily excitatory neurons^{10, 40}, has been shown to produce changes to pain-related behavior. For example, optogenetic inhibition of CaMKIIa-expressing neurons in the PIC suppressed capsaicin-evoked mechanical hypersensitivity⁶⁵, and optogenetic activation produced a real-time place aversion in non-injured mice³³, suggesting that CaMKIIa-expressing neurons in the PIC can modulate both pain-related sensory and affective behaviors. However, the extent to which sustained modulation of excitatory PIC neurons can produce bidirectional effects on a range of sensory and more complex affective behaviors, and the influence of these neurons over interoception remains unknown. We tested this in non-injured and inflammatory conditions using excitatory and inhibitory chemogenetics in a comprehensive array of behavioral paradigms. Our findings reveal that PIC CaMKIIa-expressing neurons control both hyperalgesia, aversion, complex pain behaviors, and thermoregulatory processing.

Methods

Animals

All experimental protocols were approved by the Institutional Animal Care and Use Committee of the University of Cincinnati and conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. 8–16 week old C57 BL/6J mice were used in this study. Mice were purchased from The Jackson Laboratory and housed four to five per cage under diurnal cycle of 14 h light and 10 h dark. Food and water were provided ad libitum. Both male and female mice were used for immunohistochemistry and male mice were used for electrophysiology and behavioral tests.

Stereotaxic surgery

Mice were deeply anesthetized with 2% isoflurane gas and skull was leveled on stereotaxic apparatus. Using a Hamilton 5uL syringe (model #87931), both the left and right posterior insula were injected with 300–400nl of AAV5-CaMKIIa-hM3D(Gq)-mCherry (Addgene #50476), AAV5-CaMKIIa-hM4D(Gi)-mCherry (Addgene #50477), AAV5-CaMKIIa-mCherry (Addgene #114469), or AAV5-CaMKIIa-eGFP (Addgene #50469). Vectors were injected at a rate of 50nl/min. Stereotaxic coordinates for PIC were ML +/− 4.05 mm, AP −0.6 mm, DV −2.05 mm relative to bregma. After virus infusion, mice were given meloxicam (subcutaneous, 2mg/kg). Behavior experiments started 3 weeks after surgeries to allow for recovery and viral expression. Animals with offtarget injection sites were excluded from the data analysis.

Immunohistochemistry (IHC)

Mice were anesthetized with 2% isoflurane and transcardially perfused with ice-cold 0.01 M phosphate-buffered saline (PBS) and then 4% paraformaldehyde (PFA) in PBS. Brains were extracted and post-fixed in 4% PFA overnight then placed in sucrose for 24 hours prior to sectioning. To detect cFos, mice were euthanized 1.5 hours after Complete Freund’s Adjuvant (CFA, Sigma-Aldrich) injection into left hind paw. 40 μM thick coronal free-floating sections were collected with a cryostat and washed with PBS three times for 20 minutes then blocked for two hours in 0.3% triton-X-100 in 0.01 M PBS and 10% normal goat serum. The tissues were then incubated overnight at 4°C in primary antibody solution rabbit anti-cFos (1:5000) (Abcam ab90289) in 3% normal goal serum (Vector Labs S-1000–20) and 0.3% PBS-TX. The following day, tissues were washed 3X for 20 minutes in 0.3% PBS-TX before being incubated in secondary antibody solution Goat anti-rabbit Alexafluor 555 at 1:1000 (Invitrogen) for 2 hours at room temperature. After final washes the tissue was mounted onto slides and coverslipped using Prolong gold antifade with DAPI. Images were taken as stitched Z-stacks using a 20x objective on the Keyence BZ-X800 (Osaka, Japan). Blinded cFos counts were done in ImageJ. Cells were counted based on their intensity and size threshold. Number of cFos-positive cells over area (mm²) was calculated by taking the number of cFos-positive cells in the PIC region of interest (ROI) and dividing it by the total area of the ROI which was calculated on ImageJ.

Behavior tests

Behavior tests were performed in non-injured and inflammatory pain conditions similar to our previous studies⁵⁹. Hind-paw inflammation was induced by CFA. 10uL of CFA was injected in left plantar surface of the hind paw using a 30 gauge needle 24 hours before behavior testing. CNO (2mg/kg) or saline vehicle was administered via intraperitoneal (i.p) injection 30 to 45 minutes prior to testing. Compound 21 or saline vehicle was injected i.p at 1 mg/kg for GqDREADD expressing mice and 3 mg/kg for GiDREADD ~20 minutes prior to testing. Mice were randomly assigned to agonists or vehicle injections which was counterbalanced the next day. Vehicle and agonist treatments were randomly coded by an outside experimenter to ensure blinding for behavioral tests.

Hargreaves Test

Mice habituated to behavior room in home cages for 30 – 60 minutes before being placed in ventilated clear Plexiglas testing boxes on an elevated glass surface. Following an additional 45 minutes of habituation to the Plexiglas boxes, the radiant heat source was applied to the hind paw. The heat source was set to a constant intensity of 25% (IITC) which, in pilot studies produced reliable withdrawal at approximately 10 seconds. The time to paw withdrawal was recorded, there were a total of three testing sessions per paw each separated by at least 3 minutes. The average of three latencies were taken from both hind paws.

Von Frey Test

Mice habituated to the behavior room for 30 – 60 minutes before being placed in Plexiglas von Frey boxes atop of a mesh elevated platform (IITC). Using the SUDO method¹³, the von Frey filament of 0.6g was applied to the hind paw, if there was no withdrawal, the next higher gram force filament was applied. If there was a withdrawal, a lighter filament was used. There was a total of 5 trials with 3 minutes between each trial. 50% withdrawal threshold was calculated using an online algorithm developed by Christensen et al¹⁷.

Conditioned Place Aversion or Preference

The CPA/CPP apparatus consisted of two distinct chambers with a connecting neutral chamber (Stoelting Co.). The first two days consisted of 30 minute pre-exposures during which the mice had full access to the apparatus. On day 3, the pre-exposure day was 15 minutes and time spent in either chamber was recorded. Next, mice underwent 2–4 days of conditioning during which one chamber was paired with the vehicle i.p injection in the morning, then 4 hours later, the other chamber was paired with the agonist i.p injection. Each conditioning session was 30 minutes. On the final test day, mice had full access to the apparatus for 15 minutes and time spent in each chamber was recorded. Time spent in the agonist side on test day was compared to the final pre-exposure day. The experiment was conducted without investigators in the room, video-recorded and analyzed offline using ANY-maze software (Stoelting Co.).

Open Field Test

Mice habituated in home cages to behavior room for 30–60 minutes before starting. Mice were then placed into a square open field apparatus (40×40 cm) to explore freely for 10 minutes. The experiment was video-recorded and tracked using ANY-maze software (Stoelting Co.).

Temperature preference

The temperature preference apparatus consists of two chambers side by side with temperature-controlled aluminum floors (each 30.5 × 15.2cm) connected to a peltier controller. Mice habituated to behavior room for 30–60 min before starting then had a 15-minute training session with both floors set to 30°C. On the test days, one floor was set to 25°C and the other to 35°C. Mice were tested for 15 minutes after agonist or vehicle i.p injection and mice tracked with ANY-maze software (Stoelting Co.).

Hot plate

Mice were placed on a 50°C plate and were removed upon the first lick of the hind paw. Latency to withdraw and lick the hind paw was recorded. All mice were removed no later than after 60 seconds to avoid tissue injury.

Temperature probe

A rectal temperature probe connected to a SomnoSuite (Kent Scientific) was used to measure body temperature. Mice were i.p injected with vehicle 30–45 minutes before body temperature was recorded. This process was then repeated after agonist i.p injection.

Operant Plantar Thermal Assay (OPTA)

The OPTA is a test of thermal pain tolerance⁵⁹. Two chambers each with a temperature-controlled floor are connected to by a gate allowing free movement between the chambers. One chamber has an empty water bottle spout (the null zone) and the other has a water bottle spout with access to 4% sucrose (the reward zone). Mice train in the apparatus for two days at 30 minutes while both chambers are set to set to a neutral 30°C. On the third training day, the OPTA is set to mimic test day conditions: the floor of the null zone chamber is set to 30°C and the floor of the reward zone chamber is set to a noxious 45°C. The next two days consists of 20-minute testing sessions. The mice are given the DREADD agonist or vehicle injection 30–45 minute before testing. Amount of time spent attending to the reward or null zones was measured with AnyMaze and investigators out of the room.

Brain slice preparation

Mice were deeply anesthetized with isoflurane and perfused with ice-cold carbogenated (95% O2, 5% CO2) N-N-Methyl-D-glucamine (NMDG)-artificial cerebrospinal fluid (aCSF) (in mM: 92 NMDG, 2.5 KCl, 1.25 NaH2PO4, 30 NaHCO3, 20 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES), 25 glucose, 2 thiourea, 5 Na-ascorbate, 3 Na-pyruvate, 0.5 CaCl2 and 10 MgSO4), pH 7.4 adjusted with HCl. Mice were then decapitated and brains were quickly extracted and mounted onto a specimen tube and sectioned into NMDG-aCSF using a VF-300-0z compresstome (Precisionary Instruments) at a thickness of 300 μm. Brain slices were then incubated for recovery in NMDG-aCSF for 15 min at room temperature then transferred to a holding solution aCSF-HEPES (in mM: 92 NaCl, 2.5 KCl, 1.25 NaH2PO4, 30 NaHCO3, 20 HEPES, 25 glucose, 2 thiourea, 5 Na-ascorbate, 3 Na-pyruvate, 2 CaCl2, and 2 MgSO4, 300–310 mOsm adjusted with sucrose, pH 7.4) prior to recording.

Electrophysiology

Whole cell current clamp recordings of AAV expressing neurons were obtained from visually identified mCherry-expressing CaMKIIa PIC neurons. Brain slices were continuously perfused with external recording solution (in mM: 145 NaCl, 3 KCl, 2.5 CaCl2, 1.2 MgCl2, 10 HEPES, 7 glucose, adjusted to pH 7.4 with NaOH). Recording pipettes had an open tip resistance between 4 to 6 MΩ and were filled with intracellular solution: in mM: 130 K-gluconate, 5 KCl, 5 NaCl, 2 MgCl2, 0.3 Ethelyene glycol-bis(2-aminoethylether)-NNN’N’-tetraacetic acid (EGTA), 10 HEPES, 2 Na ATP, adjusted to pH 7.3 with KOH and 294 mOsm with sucrose. Electrophysiological data was collected using the Multiclamp 700B patch-clamp amplifier and Digidata 1550 acquisition system with pCLAMP 10 software (Axon instruments). Recordings were sampled at 20kHz and filtered at 10 kHz. After whole-cell access, the cell was held at −60 mV for a minimum of 2 minutes before recording. Neurons resting > −40 mV were excluded. For DREADD validation, membrane properties were recorded under vehicle conditions, then CNO (10μM) or C21 (10μM) was bath applied while resting membrane potential was monitored. Three minutes post CNO or C21 application, membrane properties were recorded again.

Statistical analysis

Sample sizes were determined based on prior power analyses and from our observations in our previous study⁵⁹. All statistical analyses were performed as recommended using GraphPad Prism 9 (Graphpad Software, San Diego). Data are represented in the figures as mean ± standard error (SEM). Student’s t-test, two-way ANOVA or mixed-effects ANOVAs were used. Sidak or Tukey’s post-hoc tests were applied when appropriate. For all statistical analyses, significance was defined as p < 0.05. * p < 0.05, ** p < 0.01, *** p < 0.001.

Results

cFos expression in the PIC is increased following inflammatory injury.

To investigate whether the PIC performs nociceptive processing, we compared expression of the immediate early gene and marker for neuronal activation, cFos, in the PIC of mice in naive and CFA inflammatory conditions. The insula is subdivided into three layers by cellular cytoarchitecture: the dorsal granular layer, the middle dysgranular layer, and the ventral agranular layer³⁶, as shown in Fig 1A. cFos-positive neurons were primarily located in the granular layer of the PIC (Fig 1B and C). cFos-positive neurons were quantified throughout the rostral-caudal axis of the contralateral PIC. The CFA condition expressed more cFos-positive neurons throughout the PIC compared to naïve condition (Fig 1D, injury p = 0.0192, mixed-effects ANOVA). cFos positive neurons were typically seen in layers 2/3 and 5 of the PIC. These layers contain pyramidal neurons that could be activated by peripheral inflammation.

Figure 1.

cFos expression in the PIC is increased following inflammatory injury. (A) Schematic of the layer subdivision in the insula as indicated in the Paxinos and Franklin’s mouse brain atlas. (B) Representative image of the granular layer of the PIC in naive conditions and (C) CFA conditions. (D) Number of cFos-positive neurons in the PIC throughout the rostral-caudal axis of the PIC (Injury p = .0192, Naive N = 5, CFA N = 7, mixed-effects ANOVA).

Layer 2/3 PIC pyramidal neurons become sensitized in CFA conditions.

Since inflammatory injury produced greater neuronal activation in the PIC, we next determined whether inflammatory injury changed the membrane excitability of PIC neurons. Whole-cell current clamp recordings were performed in contralateral layer 2/3 pyramidal neurons of mice 24 hours after sham (saline) or CFA hind paw injection. Pyramidal neurons were identified by their pyramid-like morphology (Fig 2A). PIC neurons from CFA-treated mice fired significantly more action potentials compared to sham (Fig. 2B), and more action potentials were produced at 2x rheobase compared to 1x (Fig 2C, injury p = 0.04, rheobase p = 0.03, two-way ANOVA). Resting membrane potential was significantly more depolarized in the CFA condition compared to sham (Fig 2D, p = 0.012, paired Student’s t-test). We measured the threshold of APs fired at 1x and 2x rheobase in both groups and revealed that PIC neurons from CFA conditions had a more depolarized AP threshold compared to sham (Fig 2E, injury p = 0.023, mixed-effects ANOVA), while rheobase did not differ between groups (Fig 2F). Overall, these data indicate that layer 2/3 PIC pyramidal neurons were sensitized by peripheral inflammation.

Figure 2.

Layer 2/3 PIC pyramidal neurons are sensitized in CFA conditions. (A) Representative image of PIC pyramidal neuron. White arrows pointing to pyramidal neuron and morphology. (B) Represented image of APs evoked from a current-clamp step protocol, top is sham, bottom is CFA. (C) Number of APs produced at 1x and 2x rheobase (injury p = 0.04, rheobase p = 0.03, two-way ANOVA, N = 13 CFA, N = 17 sham). (D) Resting membrane potential of recorded PIC pyramidal neurons from mice in sham and CFA conditions (p = 0.012, unpaired Student’s t-test N = 16,). (E) Action potential threshold at 1x and 2x rheobase of PIC cells in sham or CFA conditions (injury p = .023, N = 11–19, mixed-effects ANOVA). (F) Action potential rheobase.

Bidirectional chemogenetic modulation of excitatory PIC neurons.

We next employed designer receptors exclusively activated by designer drugs (DREADDs) to bidirectionally modulate the activity of PIC. AAV5-CaMKIIa-hM3Dq-mCherry virus was injected bilaterally to activate the excitatory neurons in the PIC, an AAV5-CaMKIIa-hM4Di-mCherry virus was injected to inhibit excitatory neurons in the PIC, or a control virus, AAV5-CaMKIIa-mCherry, was injected in adult male mice (Fig 3A–C). To confirm whether CaMKIIa is expressed specifically in excitatory neurons in PIC, we injected AAV5-CaMKIIa-eGFP into the PIC of vGAT-cre x tdTomato (ai9) mice, which expresses the tdTomato fluorescent protein in inhibitory GABAergic cells. No overlap was observed between CamKIIa and tdTomato in the neurons of the PIC (Fig 3D). Next, whole-cell patch clamp recordings from mice expressing the excitatory or inhibitory DREADDs was used to confirm the functional output of the DREADDs which were visualized by the mCherry fluorescent tag (Fig 3E). Bath application of the agonists CNO (10μM) and C21 (10μM) increased rheobase (Fig 3F, p = 0.04, paired Student’s t-test) and hyperpolarized resting membrane potential (Fig 3F, p= 0.0002, paired Student’s t-test) of GiDREADD expressing neurons in the PIC, demonstrating inhibition. Conversely, bath application of these agonists decreased rheobase (Fig 3G, p = 0.0072, paired Student’s t-test) and depolarized resting membrane potential (Fig 3G, p = < 0.0001, paired Student’s t-test) of GqDREADD expressing neurons in the PIC, demonstrating increased excitability. These experiments provided validation of the bidirectional chemogenetic strategy for the following behavioral assays.

Figure 3.

Bidirectional chemogenetic modulation of PIC excitatory cells. (A) Schematic of viral constructs. (B) Representative image of mCherry identified DREADD injection site in PIC. (C) Schematic of viral spread. (D) Representative image of AAV5-CaMKIIa-eGFP expressing in PIC section of vGAT-cre x ai9 mouse. (E) Representative image of mCherry identified DREADD expressing neurons in the PIC. (F) Rheobase and resting membrane potential of GqDREADD-expressing neurons in vehicle external then bath applied CNO (10μM) and C21 (10μM) (p = 0.0072, N = 11 and p = < 0.0001, N = 8, respectively, paired Student’s t-test). (G) Rheobase and resting membrane potential of GiDREADD-expressing neurons in vehicle external then bath applied CNO (10μM) or C21 (10μm) (p = 0.04, N = 9, and p = 0.0002, N = 9 respectively, paired Student’s t-test). The change in resting membrane potential in Gi- or GqDREADD-expressing neurons was not statistically different between C21 and CNO (GqDREADD p = 0.33, GiDREADD p = 0.96, unpaired Student’s t-test).

Chemogenetic activation of the PIC produces thermal and mechanical hyperalgesia.

Given that inflammatory injury increased cfos expression in PIC and pyramidal PIC neurons showed signs of hyperexcitability, we hypothesized that sustained modulation of neuronal activity via chemogenetic control of the PIC would alter thermal and mechanical sensory pain-related behavior. We first determined whether activation of excitatory pyramidal neurons in the PIC could recapitulate inflammatory pain conditions. This was done by chemogenetic activation of the excitatory CaMKIIa-expressing neurons of the PIC during sensory threshold tests in mice in non-injured conditions and inflammatory pain conditions. After initial behavior tests were performed in non-injured mice, the left paw was injected with CFA, and is thereafter referred to as “ipsilateral”. GqDREADD activation produced a decrease in thermal threshold between vehicle and agonist conditions in of the ipsilateral paw (Fig 4A, agonist p = 0.0044, injury p < 0.0001, two-way ANOVA), and contralateral paw (Fig 4B, agonist p <0.0001, injury p = 0.0389, two way ANVOA) in both non-injured and CFA conditions. GqDREADD activation also produced a significant reduction in mechanical threshold in the von Frey test in both the ipsilateral (Fig 4C, agonist p = 0.0127, injury p < 0.0001, two-way ANOVA) and contralateral paw (Fig 4D, agonist p = 0.0029, injury p < 0.0001, two-way ANOVA) in both non-injured and CFA conditions.

Figure 4.

Chemogenetic activation of the PIC produces thermal and mechanical hyperalgesia. (A) and (B) Withdrawal latency of ipsilateral and contralateral paw, respectively, of GqDREADD-expressing mice in non-injured and CFA conditions (A, agonist p = 0.0044, injury p < 0.0001, N = 14–21, two-way ANOVA. B, agonist p <0.0001, injury p = 0.0389, N = 14–21, two way ANVOA). (C) and (D) 50% withdrawal threshold of GqDREADD expressing mice in the ipsilateral paw (C, agonist p = 0.0127, injury p < 0.0001, N = 8–14, two-way ANOVA) and contralateral paw (D, agonist p = 0.0029, injury p < 0.0001, N = 8–14, two-way ANOVA). (E, F) Withdrawal latency of ipsilateral paw (injury p < 0.0001, N = 21, two-way ANOVA) and contralateral paw (injury p < 0.0001, N = 21, two-way ANOVA) of GiDREADD-expressing mice in non-injured and CFA conditions. (G, H) 50% withdrawal threshold of ipsilateral paw of GiDREADD-expressing mice in non-injured and CFA conditions (G, injury p < 0.0001, N = 22, two-way ANOVA). (I, J) Withdrawal latency of ipsilateral and contralateral paws of mCherry PIC expressing mice in non-injured and CFA conditions (I, injury p < 0.0001, N = 9, two-way RM ANOVA). (K, L) 50% withdrawal threshold of ipsilateral and contralateral paws of mCherry PIC expressing mice in non-injured and CFA conditions (K, injury p < 0.0001, N = 9, two-way RM ANOVA).

Since activation of the CaMKIIa excitatory PIC neurons produced thermal and mechanical hyperalgesia in the absence of injury, we next asked whether chemogenetic inhibition of these PIC neurons could attenuate the behavioral sensitization produced by inflammatory injury to determine therapeutic potential. Surprisingly, GiDREADD inhibition of the PIC did not change thermal or mechanical threshold in either non-injured or CFA conditions in both the ipsilateral and contralateral hind paw (Fig 4E–H). To confirm that behavioral results were not due to the agonists themselves, we tested sensory thresholds on mCherry control mice and did not see a change in thermal or mechanical threshold between agonist and vehicle (Fig 4I–L) but did exhibit thermal hypersensitivity (Fig 4I, injury p < 0.0001) and mechanical hypersensitivity (Fig 4K, injury p < 0.0001) in the ipsilateral paw from CFA.

Chemogenetic activation of excitatory PIC neurons produced place aversion and increased anxiety-like behavior in non-injured mice.

To determine whether excitatory CaMKIIa-expressing PIC neurons are involved in affective behavior, we first performed the conditioned place aversion/preference assay (Fig 5A). GqDREADD activation of the CaMKIIa-expressing PIC neurons during naïve conditions produced a place aversion (Fig 5B, p < 0.0001, paired Student’s t-test), thus the mice spent significantly less time in the agonist-paired chamber while only spending 28.6 ± 1.7% of the time in the neutral chamber and the rest of the time in the vehicle-paired chamber. On the other hand, inhibiting the CaMKIIa-expressing neurons in the PIC in non-injured conditions produced no change in place preference nor, surprisingly, did inhibition during CFA conditions produce a change in place preference (Fig 5C and 5D). MCherry control mice did not exhibit a place preference or aversion (Fig 5E). We next performed the open field test (OFT) and measured time spent out of the center or in the periphery as a proxy for anxiety-like behavior (Fig 5F). GqDREADD activation of the CaMKIIa-expressing PIC neurons in non-injured mice resulted in a significantly reduced time spent in the center (Fig 5G, p = .046, unpaired Student’s t-test) without changing distance traveled (Fig 5H). Inhibition of CaMKIIa-expressing neurons in neither non-injured nor CFA conditions changed time spent in center (Fig 5I and 5K), nor was locomotion affected (Fig 5J and L). MCherry control mice did not differ in the amount of time spent in the center or distance (Fig 5M and N). Together the results indicate that activation of CaMKIIa-expressing PIC neurons leads to strong aversion and anxiety-like behaviors, while inhibiting these same neurons did not affect either aversion or anxiety like behaviors, even during conditions of inflammatory pain.

Figure 5.

Chemogenetic activation of excitatory PIC neurons produces a place aversion in non-injured mice. (A) Representative heat map image from a GqDREADD-expressing mouse in the CPA. (B) Time spent in agonist side post-conditioning/test day of non-injured GqDREADD mice (p < 0.0001, N = 18, paired Student’s t-test). Mice spent 28.6 ± 1.7% of time in the neutral chamber. (C, D) Time spent in agonist side on post conditioning/test day compared to pre-exposure day of GiDREADD mice in non-injured and CFA conditions. (E) Time spent in agonist side post conditioning of non-injured mCherry mice. (F) Representative heat map image of the open field test of GqDREADD-expressing mice with the agonist (top) or vehicle (bottom) i.p injection. (G, H) Time spent in the center (p = 0.045, N = 9, unpaired Student’s t-test) and distance traveled, respectively, of GqDREADD mice in non-injured conditions. (I, J) Time spent in the center and distance traveled, respectively, of mCherry-expressing mice mice. (K, L) Time spent in the center and distance traveled of non-injured GiDREADD mice. (M and N) Time spent in the center and distance traveled of GiDREADD mice in CFA conditions.

Thermal pain tolerance and heat nociception with hot plate.

Due to its connectivity with motivational and reward regions³³, we considered whether the PIC may be involved in processing more complex behaviors related to pain such as pain tolerance. We used the operant plantar thermal assay (OPTA) to measure the amount of time mice chose to spend in a noxious 45°C zone for a sucrose reward as a measure of thermal pain tolerance (Fig 6A)⁵⁹. In the GqDREADD-expressing group, pain tolerance was not altered in non-injured conditions, however, in CFA conditions, mice spent significantly less time in the reward zone, suggesting that enhanced activation of CaMKIIa-expressing PIC neurons decreased pain tolerance in the inflammatory condition (Fig 6B, injury x treatment p = 0.0126, Non-injured: agonist vs CFA:agonist p = 0.0005, CFA: vehicle vs CFA: agonist p = 0.0002, two-way ANOVA, Sidak’s post hoc). To confirm that modulation of CaMKIIa-expressing PIC neurons did not alter baseline motivation for sucrose, we measured the amount of time mice spent in the reward zone when both chamber floors were set to a neutral 30°C. Under these conditions, activation of PIC CaMKIIa neurons produced no changes in sucrose seeking behavior (Fig 6C). The opposite manipulation with Gi revealed that inhibiting the CaMKIIa-expressing neurons in the PIC did not change time spent in the reward zone in either non-injured or CFA conditions (Fig 6D). Together, these data suggest that PIC activation of excitatory neurons selectively alters higher order decision-making processing under pain conditions.

Figure 6.

Thermal pain tolerance and heat nociception on hot plate. (A) Representative track plot of a GqDREADD-expressing mouse in CFA conditions with vehicle (top) or agonist (bottom) on board in the OPTA. (B) Time spent in the reward zone by GqDREADD expressing mice (injury × treatment p = 0.0126, Non-injured: agonist vs CFA:agonist p = 0.0005, CFA: vehicle vs CFA: agonist p = 0.0002, N = 21, two-way ANOVA, Sidak’s post hoc). (C) Time spent in the reward zone by the GqDREADD expressing mice when both sides of the apparatus are set to 30°C. (D) time spent in the reward zone by GiDREADD-expressing mice. (E) Hot plate latency of non-injured GqDREADD-expressing mice. (F, G) Hot plate latency of non-injured (p = 0.0289, N = 10, paired Student’s t-test) and CFA (p = 0.0495, N = 7, paired Student’s t-test) GiDREADD-expressing mice, respectively. (H) Hot plate latency of mCherry control mice (N = 8).

Next, the hot plate test was used to assess heat nociception and pain processing beyond thermal sensory threshold. We measured the latency to lick the paw which may be a measure of both affective and sensory dimensions, as it has been interpreted as a self-soothing act and a form of coping behavior³⁹. Interestingly, activation of the CaMKIIa-expressing neurons in the PIC did not alter the latency to lick on the hot plate (Fig 6E). However, inhibition of the CaMKIIa-expressing neurons in the PIC increased the latency to lick on the hot plate in non-injured (Fig 6F, p = 0.0289, paired Student’s t-test) and CFA conditions (Fig 6G, p = 0.0495, paired Student’s t-test). MCherry control mice did not show a difference in hot plate latency (Fig 6H).

The posterior insula regulates body temperature and thermal preference.

The insular cortex is considered an interoceptive and homeostatic regulator^{64, 74}. We therefore asked whether PIC modulation could alter homeostatic control of body temperature. Activation of CaMKIIa-expressing PIC neurons did not alter body temperature (Fig 7A), however, body temperature significantly decreased when the PIC CaMKIIa-expressing neurons were inhibited (Fig B, p = 0.0058, paired Student’s t-test). MCherry control mice did not show a difference (Fig 7C). Moreover, this disruption in homeostatic thermoregulation led to changes in floor temperature preference in a two-choice thermal preference assay (Fig 7D–7F). Despite not changing body temperature, activation of CaMKIIa-expressing PIC neurons shifted preference to the cooler 25°C floor (Fig 7Dii, 25°C:agonist vs 35°C:agonist p = 0.0026, two-way RM ANOVA, Tukey’s post hoc). This was further demonstrated by comparing the difference in time spent in the 25°C to time spent in the 35°C side. Activation of the CaMKIIa-expressing PIC neurons resulted in mice spending significantly more time in the 25°C compared to vehicle control (Fig 7Diii, p = 0.0155, paired Student’s t-test). In contrast, inhibition of CaMKIIa-expressing neurons shifted temperature preference to the warmer floor for both groups (Fig 7Eii, temperature × agonist p = .0050, 25°C: vehicle vs 35°C: vehicle p = 0.0084, 25°C:agonist vs 35°C:agonist p < 0.0001, two-way RM ANOVA), and mice spent significantly more time in the 35°C side during PIC CaMKIIa inhibition compared to vehicle control (Fig 7Eiii, p = 0.0050, paired Student’s t-test). This may likely be to compensate for lowered body temperature. MCherry mice did not exhibit a change in preference (Fig 7Fii, iii).

Figure 7.

The posterior insula regulates body temperature and thermal preference. (A) Body temperature reading of GqDREADD expressing mice in vehicle and agonist conditions. (B) Body temperature reading of GiDREADD expressing mice in vehicle and agonist conditions (p = 0.0058, N = 15, paired Student’s t-test). (C) mCherry expressing mice after vehicle and agonist injection (N = 9). (Di) Representative heat map image of the thermal preference paradigm with a GqDREADD-expressing mouse with agonist i.p. (Dii) Thermal preference of non-injured GqDREADD-expressing mice (25°C:agonist vs 35°C:agonist p = 0.0026, N = 23, two-way RM ANOVA, Tukey’s post hoc). (Diii) Difference between time spent in 35°C and 25°C side between vehicle and agonist in GqDREADD-expressing mice (p = 0.0155, N = 23, Student’s t-test). Ei, Representative heat map image of the thermal preference paradigm with a GiDREADD-expressing mouse with agonist i.p. Eii, thermal preference of non-injured GiDREADD-expressing mice (temperature × agonist p = 0.0050, 25°C: vehicle vs 35°C: vehicle p = 0.0084, 25°C:agonist vs 35°C:agonist p < 0.0001, N = 25, two-way ANOVA with Tukey’s post hoc). (Eiii) Difference between time spent in 35°C and 25°C side between vehicle and agonist in GiDREADD-expressing mice (p = 0.0050, N = 25, paired Student’s t-test). (Fi) Representative heat map image of the thermal preference paradigm with an mCherry-expressing mouse with agonist i.p. (Fii) Thermal preference of mCherry control mice (N = 8). (Fiii) Difference in amount of time mCherry-expressing mice spent in the 35°C and 25°C side with vehicle i.p on board and agonist i.p on board.

Discussion

This study demonstrated that PIC neurons are activated during inflammatory injury and exhibit sensitization. Sustained chemogenetic activation of PIC excitatory neurons in non-injured mice produces to thermal and mechanical hypersensitivity and produces negative affect in the absence of injury. In the presence of inflammatory pain, activation of the PIC exacerbates sensory, affective, and motivational pain behaviors, including producing an intolerance to thermal pain. In contrast, inhibition of excitatory PIC neurons did not reverse behavioral hypersensitivity in animals with inflammatory pain. However, we found that inhibiting these neurons had a critical function in maintaining body temperature and increased hot plate latency. Together, the data suggests that CaMKIIa-expressing neurons in the PIC function selectively as a generator of negative experience, modulate heat and mechanical nociception, and control thermoregulation.

Expression of cFos in the PIC revealed that peripheral inflammation produced significant activation of neurons throughout the rostral-caudal extent, particularly within the granular layer. The insular granular layer is known as a target of visceral sensory afferents⁶¹, suggesting that this region of the PIC has some capacity of processing interoception and somatosensation. These data extend the human imaging literature that shows the IC to be consistently activated upon various noxious stimuli^{3, 14, 18, 43} by demonstrating at cellular resolution the cortical locus of PIC neuronal activity. We obtained further insight on neuronal activity in the PIC from layer 2/3 pyramidal neurons which exhibited increased excitability from mice that had inflammatory pain. This is consistent with our previous report showing that layer 2/3 neurons in anterior cingulate cortex show sensitization after inflammatory and neuropathic injury¹⁶. Changes in neuronal excitability in layer 2/3 pyramidal neurons has also been identified in other brain regions important for pain processing. For example, inflammatory or neuropathic injury produced sensitization of layer 2/3 pyramidal neurons in the prelimbic medial prefrontal cortex (PL)^{20, 48} and the somatosensory cortex²⁷. Yet, it has also been reported that layer 2/3 neurons of the PL can exhibit decreased excitability following CFA⁷³, which may be due to species differences. Overall, these studies emphasize that changes in neuronal excitability from a peripheral injury occurs across the cortex, although the direction of these changes may not be uniform and a more complete mapping of cortical sensitization after injury is needed. Central sensitization can generate and maintain abnormal responses to noxious and innocuous stimuli even after the apparent recovery of the injury site⁴².

To better understand the consequences of insular neuronal activity, we activated or inhibited the PIC bilaterally in non-injured and pain conditions. We separated behavioral paradigms into sensory and affective categories to determine how the insula influences each dimension of pain. Both Gi- and GqDREADD function was validated using slice electrophysiology to demonstrate that each approach decreased or increased CaMKIIa neuron excitability, respectively. We then showed that chemogenetic activation of the excitatory neurons of the PIC produced both mechanical and thermal hyperalgesia in both non-injured and pain conditions. No study has yet revealed that ongoing chemogenetic PIC activation produces hyperalgesia. This pro-nociception may be due to an increase in signaling from the insula to pain facilitatory areas such as the raphe magnus nucleus in the brainstem⁶⁵. The insula also possesses descending excitatory inputs to the trigeminal spinal subnucleus caudalis which has a pro-nociceptive role⁵⁰. While there is evidence that the insula communicates with the periaqueductal gray to facilitate descending pain inhibition, surprisingly, we did not see the engagement anti-nociceptive behaviors^{57, 67}. In addition to reflexive tests of nociception and hypersensitivity, activation of the CaMKIIa-expressing neurons in the PIC also produced a place aversion and an increase in anxiety-like behaviors. These findings are consistent with studies showing the PIC processes aversiveness³³ and is involved in generating negative emotion^{5, 26}. Our use of chemogenetics extends prior optogenetic studies that applied transient modulation of PIC CaMKIIa-expressing neurons during aversive and avoidance behaviors³³ or during capsaicin-evoked pain⁶⁵. Here we applied sustained modulation of the PIC throughout the course of the various different sensory and affective behavior paradigms under both non-injured and inflammatory conditions allowing for broader investigation of the role of the PIC in pain.

Because chemogenetic activation of the excitatory neurons of the PIC produced both nociceptive hypersensitivity and negative affect, we hypothesized that inhibition of these neurons would produce the opposite effect and result in analgesia. Despite effective functional inhibition of PIC neurons, we did not see a change in reflexive withdrawal or affective behaviors when inhibiting the CaMKIIa-expressing neurons in the PIC of mice in non-injured or inflammatory pain, except for in the hot plate test (discussed below). These findings differ from expectations based on a prior report that archaerhodopsin-mediated inhibition of CaMKIIa-expressing neurons in the PIC relieved capsaicin-induced mechanical allodynia⁶⁵. This may be due to differences between capsaicin and CFA pain models which vary in mechanism of action and time course, or the difference in method of producing inhibition. Other manipulations that reduce function of the PIC such as lesion, have produced a decrease in long-term mechanical hyperalgesia from neuropathic injury^{7, 9}. But this kind of injury likely includes many other cell types in the PIC. The lack of change in affective behaviors we observed from PIC inhibition is consistent with other studies^{7, 65}. Since pain is an evolutionarily important and protective experience, the lack of behavioral change from PIC inhibition in our experiments may signify the presence of parallel pathways involved in pain processing. Other areas that may compensate for lost PIC or processing the sensory and affective components of a nociceptive stimulus independently include the anterior cingulate cortex^{38, 71} and somatosensory cortices^{12, 56, 70}. Indeed, insular lesions appear to increase the demand of other areas involved in pain processing and produce an increase in activation of the somatosensory cortex and dorsal lateral pre-frontal cortex during nociceptive stimulation⁶³.

The PIC has a distinct modulatory role in nociception, thermal tolerance, and thermoregulatory processing. We revealed that inhibition of the PIC decreased body temperature, similar results were observed by PIC inactivation with bupivacaine¹⁵. The insula has been suggested to be a key region in regulating homeostasis and thermosensation^{21, 45, 69}, yet this is the first evidence that sustained chemogenetic inhibition of excitatory neurons in the PIC maintains a lower body temperature and alter thermal homeostatic behavioral choices. Indeed, inhibition of the PIC in mice shifted thermal preference to a warmer chamber. And excitation shifted preference to a cooler chamber (although without a change in body temperature).

In addition to lowering body temperature, inhibition of the PIC in both non-injured and CFA conditions increased hot plate licking latency. The hot plate test may reflect affective and sensory dimensions of pain depending on the behavior that is measured, in this case, the latency to lick and attend the paw. It has been suggested that the action of licking from a tonic noxious stimuli is a reflection of an interoceptive behavior⁴⁶. However, the increased latency to lick could also have been a result of the decrease in body temperature. Interestingly, inhibition of PIC did not alter thermal tolerance of noxious heat to obtain a sucrose reward in the OPTA. EEG in humans has shown that thermal stimuli evoked potentials in the PIC are observed only at and above pain threshold²⁹ thus we speculate that the borderline noxious temperature of 45°C for the OPTA may have been insufficient to engage the PIC. Moreover, GqDREADD activation of PIC, which did not alter hot plate latency, did decrease thermal pain tolerance, but only during inflammatory pain. These data suggest that the PIC may require sufficiently strong activation to engage higher level behavioral coping mechanisms to noxious somatosensory stimuli. The posterior insula has reciprocal connections with the ventromedial prefrontal cortex, a key area involved in decision-making^{4, 53}, which may have contributed to these findings.

In our study we used both agonists C21 and CNO. Initially, C21 was used to mitigate the side effects associated with CNO back metabolism into clozapine⁶⁶. Our results (not shown) revealed that C21 and CNO produced the same effects in both in vitro electrophysiology experiments as well as in behavioral assays, so results were combined. A limitation of this work is the use of only male mice in our behavioral assays. We acknowledge the presence of sex differences in pain processing, notably there are reported differences in the degree of insular activation or the functional connectivity between the insula and other brain areas in the pain matrix in males and females with migraine or visceral pain^{23, 41, 51}. And we have previously identified sex differences in pain tolerance⁵⁹. Despite this limitation, our bidirectional manipulation of the excitatory neurons suggests that the PIC generates hypersensitivity, may attach negative valence to incoming nociceptive information, and is involved in maintaining body temperature. The insula possess many additional cell types and organization which are likely to be important contributors to myriad pain-related and somatosensory behaviors.

Highlights.

• The posterior insula (PIC) has increased activation in inflammatory pain conditions

• Activation of PIC CaMKIIa-expressing neurons produces hyperalgesia in naïve mice

• Activation of PIC CaMKIIa-expressing neurons produces negative affect in naïve mice

• Bidirectional modulation of PIC CaMKIIa-expressing neurons alters thermoregulation

Perspective.

The present study reveals that activation of the posterior insula produces hyperalgesia and negative affect, and has a role in thermal tolerance and thermoregulation. These findings highlight the insula as a key player in contributing to the multidimensionality of pain.