An amygdalar neural ensemble that encodes the unpleasantness of pain

1Pain is both a sensory and affective experience. The unpleasant percept that dominates the affective dimension of pain is coupled with the motivational drive to engage protective behaviors that limit exposure to noxious stimuli. Although previous work has uncovered detailed mechanisms underlying the sensory detection of noxious stimuli and spinal processing of nociceptive information, how brain circuits transform emotionally inert information ascending fromthe spinal cord into an affective pain percept remains unclear. Attaining a better understanding of the mechanisms underlying pain affect is important, because it could lead to novel therapeutic strategies to limit the suffering of chronic pain patients.

2The amygdala critically contributes to the emotional and autonomic responses associated with valence coding of neural information, such as responses during fear or pain. Damage to the basolateral amygdala (BLA) can induce a rare phenomenon in which noxious stimuli remain detected and discriminated but are devoid of perceived unpleasantness and do not motivate avoidance. Conversely, impairment of somatosensory cortex function reduces the ability to both localize noxious stimuli and describe their intensity, without altering aversion or avoidance. Thus, BLA affective neural circuits might link nociceptive inputs to aversive perceptions and behavior selection.

3Patients with chronic pain often suffer allodynia, a pathological state in which an intense unpleasant percept arises in response to innocuous stimuli such as light touch. Notably, the BLA displays heightened activity during chronic pain, and longitudinal functionalmagnetic resonance imaging studies in humans and rodents show that neural hyperactivity and altered functional connectivity in the amygdala parallel the onset of chronic pain, suggesting that the BLA might play a critical role in shaping pathological pain perceptions. However, it remains unclear how the BLA influences the unpleasant aspects of innate acute and chronic pain perceptions, while the role of nociceptive circuits in the central amygdala are better understood. Previous studies attempting to define pain affect mechanisms recorded the acute nociceptive responses of single amygdalar neurons in anesthetized animals. However, recent work has shown that the BLA encodes information via the coordinated dynamics of neurons within large ensembles; it is therefore important to resolve how the BLA processes pain affect at the neural ensemble level in awake, freely behaving animals.

4We first performed fluorescence in situ hybridization studies and used the immediateearly gene marker of neural activity, c-Fos, to determine that c-Fos+ neurons activated by nociceptive stimuli comprised a population of mid-anterior BLA Camk2a+ principal neurons (fig. S1). To identify how the BLA encodes nociceptive information, we used a head-mounted miniature microscope to track the somatic Ca2+ dynamics of individual BLA Camk2a+ principal neurons in freely behaving mice presented with diverse noxious and innocuous stimuli (Fig. 1, A to D, and figs. S2 and S3). We monitored pain-related behaviors by measuring each animal’s locomotor acceleration, which allowed us to track both reflexive withdrawal and affectivemotivational behaviors that include attendance to the stimulated tissue and escape (Fig. 1, A and E, and fig. S4).

5Noxious heat, cold, and pin prick stimuli elicited significant Ca2+ responses in 15 ± 2% (SEM), 13 ± 2%, and 13 ± 2% of active BLA neurons, respectively [3397 neurons (117 ± 8 neurons per session)] (Fig. 1, F to H, and table S1). Innocuous light touch induced Ca2+ activity in a smaller subset of neurons (7 ± 1%) (Fig. 1, F and I, and fig. S5E). Alignment of all stimulusevoked ensemble responses to the noxious heat trials revealed an overlapping population of principal neurons that encoded nociceptive information across pain modalities (i.e., noxious heat, cold, pin), which we refer to here as the BLA nociceptive ensemble (24 ± 2% of active BLA neurons) (Fig. 1, F to I).

6This ensemble was composed of multimodal responsive neurons, as well as a unique population that appeared to encode nociception selectively and no other sensory information (6 ± 1% of all imaged neurons) (Fig. 1K and fig. S5G). Pain behavioral responses evoked by noxious stimuli closely mirrored the activity of this nociceptive neural ensemble (Fig. 1, E and G, and fig. S4, D and E). The nociceptive ensemble contained a subset of neurons that maintained their noxious stimulus response properties for more than a week (11% of 3223 cross-day–aligned neurons) (fig. S6). Increasingly salient stimuli, from light touch (18 ± 3% of the nociceptive ensemble) to mild touch (31 ± 4%), activated larger subsets of the nociceptive ensemble (Fig. 1, G and I, fig. S5, D and E, and table S1) and induced heightened behavior (Fig. 1E and fig. S4). Expectation of stimulus contact (“approach/no contact” trials) also evoked sparse BLA activity (7 ± 2% of the total population) (fig. S5, A to E, and table S1). BLA activity did not correlate with exploratory locomotion (fig. S7, A to E).

7To determine whether the BLA nociceptive ensemble broadly encodes stimulus valence, we presented mice with an appetitive stimulus (10% sucrose). Sucrose consumption was encoded by a distinct ensemble (18 ± 3% of all neurons) that only overlapped with a subset of neurons in the nociceptive ensemble (7% of total neurons) (Fig. 1J and fig. S5E). Similar to conditioned responsive valence networks, neurons encoding unconditioned nociceptive and appetitive information were spatially intermingled (fig. S5, F, H, and I). Consistent with these results, nociceptive c-Fos+ Neuronsexpressed the negative valence marker gene Rspo2 but not the positive valence marker gene Pppr1b (fig. S1, D and E).

8We next determined if the nociceptive ensemble was engaged during aversive experiences other than pain by presenting a panel of sensory, but nonsomatosensory or nonnaturalistic, aversive stimuli, including repulsive odor, bitter taste, loud tone, facial air puff, andelectric shock. We found that while there was overlap between the neural ensembles that encode nociceptive, aversive, and electric shock stimuli (~10% of all imaged neurons), there remained a subset of BLA neurons (~6% of imaged neurons) that responded only to naturalistic nociceptive stimuli (Fig. 1K and fig. S8).

9By analyzing the neural ensemble dynamics with pattern classification methods, we were able to classify and distinguish with high accuracy noxious stimuli from other aversive stimuli (fig. S8E), supporting the finding that noxious stimuli induce a distinct mode of BLA activation (supplementary text S1). Moreover, sensory stimuli of different valences, intensities, and modalities are represented by unique activity codes. Noxious stimuli were encoded distinctly from one another and could be distinguished with even higher fidelity from innocuous, nonnociceptive aversive, and appetitive stimuli (Fig. 1L and fig. S9, A and B), indicating that there is a core set of BLA neurons that encodes nociceptive stimuli via specific dynamic neural codes. One crucial finding was that greater activation of this BLA nociceptive ensemble was predictive of increased pain behaviors, suggesting that BLA nociceptive processing influences the magnitude of pain behaviors (Fig. 1M and fig. S7, H and I).

10To test the causal role of the BLA nociceptive ensemble for pain behaviors, we expressed a Cre-dependent inhibitory DREADD neuromodulator (hM4-mCherry) in mutant TRAP mice (FosCreERT2) by applying noxious pin pricks that induced activity-dependent, spatially, and temporally controlled DNA recombination and hM4- mCherry expression (noci-TRAPhM4 mice) (Fig. 2, A to C, and fig. S10). Since the BLA encodes multiple modalities of nociceptive stimuli within a core ensemble (Fig. 1H), we hypothesized that silencing the neurons activated by noxious pin prickwould alter behavioral responses to all types of noxious stimuli. Indeed, the hM4 agonist clozapine-N-oxide (CNO; 10 mg/kg) significantly reduced both attending and escape behaviors, but not stimulus detection and withdrawal, for both mechanical and thermal noxious stimuli (Fig. 2, D to G, and fig. S11, A and B). CNO alone had no effect on pain behaviors in control mice (fig. S11C). To test operant pain behavior, we next allowed noci-TRAPhM4 mice to explore a thermal gradient track in which the polar ends were set at noxious cold (5 to 17°C) and hot (42 to 48°C) temperatures (Fig. 2H). The noci-TRAPhM4 mice injected with control saline rapidly acquired an adaptive avoidance strategy of the noxious zones. In contrast, noci-TRAPhM4 mice treated with CNO visited the noxious zones more frequently and for prolonged periods (Fig. 2, H to J, and fig. S12). Similarly, inhibition of the BLA nociceptive ensemble eliminated pain affectivemotivational behaviors induced by the optogenetic activation of peripheral primary afferent nociceptors (fig. S13).

11Whether pain and anxiety rely on common or distinct BLA ensembles is unknown; therefore, we placed noci-TRAPhM4 mice within an elevated plus maze, in which anxiety drives avoidance of the open arms (Fig. 2K). The noci-TRAPhM4 mice given either saline or CNO displayed equivalent visits to and occupancy of the open arms (fig. S14, A and B). Since nociceptive and sucrose reward-related information were encoded in divergent networks (Fig. 1J), we tested the contribution of the nociceptive ensemble to appetitive motivational drive during sucrose preference training. CNO enhanced sucrose reward in sucrose-naïve conditions but had no retarding effects on preference development or on lick rates, relative to controls (Fig. 2L and fig. S14C). Thus, this BLA nociceptive ensemble transforms emotionally inert nociceptive information into an affective signal that is necessary for the selection and learning of motivational protective pain behaviors.

12We next investigated the contribution of BLA neural ensemble activity to chronic pain. A hallmark of chronic neuropathic pain is the appearance of allodynia and hyperalgesia, both pathological perceptual states in which aversion is ascribed to innocuous somatosensory stimuli and exacerbated in response to noxious stimuli, respectively (Fig. 3A). We hypothesized that this pathological perceptual switch might result from maladaptive transformations in BLA coding. We tracked the longitudinal dynamics of BLA ensembles before and after the development of neuropathic pain induced by sciatic nerve injury (17,396 neurons, n = 17mice) (Fig. 3). Throughout the development of chronic neuropathic pain, a subset of neurons stably encoded the nociceptive ensemble for both noxious mechanical and cold stimuli (fig. S6). Nerve injury did not significantly increase the spontaneous activity of the nociceptive ensemble and overall BLA population (fig. S15, A and B). However, BLA neural activity elicited in response to light touch displayed a significant expansion within the nociceptive ensemble in neuropathic (291 ± 88% increase) but not in uninjured mice (38 ± 14% decrease) (Fig. 3, D to G, and fig. S15, C to E). The emergence of this neuropathic coding schema was accompanied by the development of reflexive paw withdrawal hypersensitivity and by enhanced affective-motivational pain behaviors (Fig. 3, B and C, and fig. S4, C to F). The magnitudes of the behavioral responses and the BLA nociceptive ensemble Ca2+ activity were significantly correlated before and after injury (Fig. 3H and fig. S15F). These results suggest a role for the BLA in the emergence of allodynia in chronic pain states.

13We next asked if we could prevent the neural transformation of light touch sensory information into an aversive signal and eliminate chronic pain unpleasantness by gaining genetic access to the nociceptive ensemble with innocuous stimuli in neuropathic TRAP mice. At 21 days post–nerve injury, when allodynia had fully developed (fig. S16, B to E), we delivered a light touch TRAP protocol to express hM4-mCherry in the BLA nociceptive ensemble (neuropathic TRAPhM4 mice) (Fig. 4, A and B, and fig. S16). At day 42 postinjury, neuropathic TRAPhM4 mice displayed significant allodynia and hyperalgesia, for both reflexive and affective-motivational pain responses, relative to uninjured mice (Fig. 4, C to E).While the injection of CNO in neuropathic TRAPhM4 mice did not alter reflexive hypersensitivity (Fig. 4D), we observed a profound decrease in neuropathic affective-motivational behaviors, regardless of stimulus intensity or modality (Fig. 4E and fig. S17, A and B). Uninjured TRAPhM4 mice given the light touch TRAP protocol expressed levels of hM4-mCherry in the BLA that were similar to those of nonstimulated control mice (Fig. 4B and fig. 2C), presumably because the nociceptive ensemble does not strongly encode innocuous information under normal conditions (Fig. 1I). We observed neither CNO-mediated changes in affectivemotivational pain behaviors in these uninjured mice nor CNO effects on neuropathic reflexive or affective-motivational behaviors in the absence of hM4 expression (Fig. 4, C to E, and fig. S17, A and B). In addition to tactile allodynia, patients with neuropathic pain often report intense pain in response to cold temperatures (cold allodynia). We therefore ran neuropathic TRAPhM4 mice through a two-chamber thermal escape-avoidance assay in which the floor of one chamber was cooled (from 30° to 10°C) (Fig. 4F). Uninjured TRAPhM4 mice avoided the cold chamber, while mice with nerve injury showed enhanced avoidance, consistent with allodynia (Fig. 4, F and G). Notably, CNO administration to neuropathic TRAPhM4 mice generated a near-total indifference between cold and neutral temperature chambers (Fig. 4, F and G). Together, these results indicate that the BLA nociceptive ensemble is also necessary for the pain aversion associated with allodynia and hyperalgesia during chronic pain states.

14Thus, disrupting neural activity in a nociceptive ensemble in the BLA is sufficient to reduce the affective dimension of pain experiences, without altering their sensory component. The unconditioned nociceptive ensemble described here is a stable network of amygdalar principal neurons that is responsive to a diverse array of noxious stimuli.Within this ensemble, combinatorial neural ensemble codes distinguish the various thermal and mechanical nociceptive stimuli. These codes likely represent stimulus modality, intensity, salience, and valence to provide additional qualitative information that enriches individual pain affect percepts. The presence of a purely nociceptive-specific subpopulation of neurons within the larger BLA nociceptive ensemble, distinct from general aversion-encoding populations, suggests the capacity for computing and assigning an accompanying “pain tag” to valence information. This categorical signal could prioritize the negative valence of intense noxious stimuli and scale the selection of conative pain protective behaviors. It is thought that hierarchical pathways transform low-level sensory inputs into higher-order affective responses. Our chemogenetic manipulations suggest that this critical node in the nociceptive brain circuitry plays a critical role in shaping pain experiences, by providing an evaluation of nociceptive information that, in turn, intrinsically motivates protective behaviors associated with pain.

15The phenomenological description of a pain experience is normally that of a complex but unified sensory and emotional perception that can neither exist alone as an unanchored aversive state nor stand merely on its emotionally inert sensory qualities. Though activity within the BLA nociceptive ensemble cannot account for the instantiation of the entire pain experience, we propose that the BLA nociceptive ensemble transmits abstracted valence information to the central amygdala, striatal, and cortical networks. For example, BLA neurons projecting to the CeA may send a “pain tag” that helps select for appropriate defensive responses to impending or immediate threats (supplementary text S2). In parallel, connected cortical regionsmight coalesce BLA affective signals with sensory-discriminative information to process them against prior experiences and internal states for further evaluation at cognitive levels, all of which contribute to the construction of a pain experience.

16During chronic pain states, BLA ensemble coding of innocuous somatosensory information changes to engage the nociceptive ensemble, leading to perceived aversion and protective behavioral responses when encountering normally nonpainful stimuli, such as light touch. Whether this change in ensemble activity results from peripheral or central sensitization, amygdalar input, or intra-amygdala plasticity remains an open question. Chronic pain is not simply a sensory disorder but a neurological disease with affective dysfunction that profoundly impacts the mental state of millions of pain patients. Clinical management of chronic pain remains a staggering challenge, given the heterogeneity of underlying causes, and the overreliance on opioid analgesics has contributed to the opioid epidemic. Comprehensive strategies that provide substantive relief across pain types are urgently needed. To make progress along this translational path, we have identified in the BLA a critical neural ensemble target that mediates chronic pain unpleasantness. This finding may enable the development of chronic pain therapies that could selectively diminish pain unpleasantness, regardless of etiology, without influencing reward, and importantly, preserving reflexes and sensorydiscriminative processes necessary for the detection and localization of noxious stimuli. Collectively, our findings begin to refine the neural basis and coding principles underlying the multiple dimensions and complexity of the pain experience for developing more effective analgesic therapies.