A nociceptive amygdala-striatal pathway modulating affective-motivational pain

Abstract

The basolateral amygdala (BLA) assigns valence to sensory stimuli, with a dedicated nociceptive ensemble encoding the negative valence of pain. However, the effects of chronic pain on the transcriptomic signatures and projection architecture of this BLA nociceptive ensemble are not well understood. Here, we show that optogenetic inhibition of the nociceptive BLA ensemble reduces affective-motivational behaviors in chronic neuropathic pain. Single-nucleus RNA sequencing revealed peripheral injury–induced changes in genetic pathways involved in axonal and presynaptic organization in nociceptive BLA neurons. Next, we identified a previously uncharacterized nociceptive hotspot in the nucleus accumbens shell that is innervated by BLA nociceptive neurons. Axonal calcium imaging of BLA projections to the accumbens and chemogenetic inhibition of this pathway revealed pain-related transmission from the amygdala to the medial nucleus accumbens, facilitating both acute and chronic pain affective-motivational behaviors. Together, this work defines a critical nociceptive amygdala-striatal circuit underlying pain unpleasantness across pain states.

A nociceptive basolateral amygdala–to–nucleus accumbens shell circuit contributes to pain unpleasantness in chronic pain.

INTRODUCTION

Survival depends on the ability to rapidly and appropriately respond to environmental threats or internal states that signal potential harm—a fundamental function of the nociceptive nervous system, which gives rise to the perception of pain (1). The unpleasantness of pain engages negative valence circuits in the brain that coordinate protective behaviors such as recuperation, escape, and avoidance (2). In chronic pain, however, this adaptive system becomes pathologically engaged, resulting in persistent and debilitating pain despite the resolution of the original injury, suggesting maladaptive activation of valence circuits.

The basolateral amygdala (BLA) is a key hub for assigning valence to sensory stimuli (3), but the specific cell types and circuit architectures that encode negative valence—particularly in the context of pain—remain unresolved (4–11). We recently identified a functionally distinct subpopulation of BLA neurons, which encodes nociceptive stimuli across modalities and is required for the affective-motivational component of pain—the BLA^noci ensemble (12). Inhibition of this ensemble selectively attenuated pain-related behaviors such as licking, guarding, escape, and stimulus avoidance (12, 13). Using cross-day single-cell calcium imaging, we further demonstrated that the same individual BLA neurons responsive to acute noxious stimuli become sensitized to previously innocuous stimuli after neuropathic injury (12). These findings suggest that the BLA^noci ensemble contributes not only to acute pain processing but also to the emergence of chronic pain states.

However, the functional stability of this ensemble across the transition from acute to chronic pain has not been investigated beyond our prior imaging studies. Moreover, the brain-wide projection architecture and molecular identity of the BLA^noci ensemble—and how chronic pain alters these features—remain unknown. Understanding the stability of ensemble activity over time is critical to resolving debates over “representational drift” in the amygdala (14) and for identifying durable cellular targets for pharmacological intervention. A more complete understanding of the molecular and circuit-level organization of the BLA^noci ensemble may provide a pathway for developing targeted treatments for persistent pain (1).

Recent single-cell transcriptomic studies have identified distinct neuronal subpopulations in the BLA based on gene expression profiles previously associated with valence-related function (11, 15–20). Other studies have classified BLA populations by their projection targets across the brain (6, 9). Yet, the relationship between transcriptional identity, projection specificity, and pain-related function remains largely unexplored. BLA principal neurons send long-range glutamatergic projections to several affective-motivational brain regions, including the nucleus accumbens (NAc) (21, 22), a ventral striatal structure involved in both appetitive and aversive motivation (23). Although the BLA → NAc circuit has been traditionally linked to reward (6, 22, 24, 25), recent work has challenged this view by showing that it can also encode negative valence (9, 26–28). In parallel, accumulating evidence implicates the NAc—particularly the NAc shell (NAcSh) region (29–31)—in the pathophysiology of chronic pain (32–35), yet the role of BLA → NAcSh projections in encoding nociceptive information remains poorly understood.

While the NAcSh has historically been studied for its role in reward processing (36–40), recent findings suggest that anatomical subregions (41, 42), gene expression profiles (40, 43–45), and neuromodulator content (e.g., endogenous opioids) (46, 47) contribute to divergent behavioral outputs. Both clinical imaging and preclinical electrophysiology have implicated the NAcSh in pain-related dysfunction (23, 29, 30, 33, 43). However, the precise location and identity of nociception-responsive neurons within the NAcSh—and whether they are downstream targets of the BLA^noci ensemble—have not been determined.

In this study, we sought to define how acute pain and chronic pain reshape the BLA^noci ensemble in terms of behavior, gene expression, and projection targeting. We show that this ensemble is a distinct subpopulation of BLA neurons required for both acute and chronic pain–related affective-motivational behaviors. Using single-nucleus RNA sequencing (snRNA-seq), we identified transcriptomic changes in the BLA following nociceptive stimulation and chronic neuropathic injury, including up-regulated genes associated with synaptic signaling and pain modulation. Anatomical tracing revealed multiple projection targets of the BLA^noci ensemble, including a previously undescribed nociceptive hotspot in the posterior medial NAcSh (mNAcSh). This region mirrored the valence specificity and stability of the BLA^noci ensemble, suggesting that it may be a critical downstream node in the affective-motivational pain network.

We further demonstrated that BLA^noci axons in the NAcSh transmit nociceptive signals, including those associated with mechanical allodynia after nerve injury. In contrast, Rspo2⁺ BLA neurons—previously considered a negative valence population—showed reduced and nonspecific activity in response to both appetitive and aversive stimuli, consistent with our transcriptomic findings that Rspo2 expression does not define a uniform nociceptive population. Last, chemogenetic inhibition of NAcSh-projecting BLA^noci neurons suppressed both acute and chronic pain–related behaviors, establishing this circuit as a key pathway in the affective-motivational pain network. Together, these findings provide a multilevel framework for understanding how affective-motivational pain-related processing is encoded within a genetically defined BLA ensemble and transmitted to the ventral striatum. They also identify the BLA → NAcSh pathway as a promising target for future therapeutic strategies aimed at reducing the affective-motivational component of acute and chronic pain experiences.

RESULTS

Acute and neuropathic nociception engages a shared BLA valence ensemble

We previously found a nociceptive ensemble of BLA neurons (BLA^noci ensemble) that encodes acute and chronic nociceptive stimuli regardless of sensory modality but does not encode non-nociceptive aversive nor appetitive stimuli (12). However, it remains unknown how the stability of the BLA^noci ensemble across pain states translates to function. Thus, our goal was to determine the cellular stability and valence selectivity of nociception within individual BLA neurons across time, valence assignment, and different pain models. To do this, we leveraged the updated TRAP2 mouse (Fos-FOS-2A-iCreER^T2) crossed with the Ai9 tdTomato fluorescent reporter mouse line (TRAP2:tdTomato) to capture functional ensembles across the brain (48). We genetically captured functionally nociceptive neurons using a nociceptive targeted recombination in active population (nociTRAP) protocol: We applied noxious 55°C water droplets to the left hindpaw [20 drops (~50 μl per drop), once per 30 s, over 10 min] and administered 4-hydroxytamoxifen (4-OHT; 40 mg/kg, subcutaneously). This nociTRAP protocol initiates genetic recombination and permanent expression of tdTomato in nociceptive cells throughout the nervous system (Fig. 1, A and B). Two weeks later, we exposed the nociTRAP2:tdTomato mice to a second noxious 55°C water stimulus to drive the expression of the immediate early gene (IEG) FOS (nociFOS) and examined the colocalization of nociFOS with nociTRAP neurons (Fig. 1, A and B). We found that nociTRAP and nociFOS neurons spread across the anterior-posterior (AP) axis of the lateral and basal subnuclei of the BLA, with 43.92% colocalization, suggesting a level of stability of the BLA^noci ensemble to repeatedly encode nociception (Fig. 1, C to E). This degree of neural reactivation between two temporally distant, similar stimuli is consistent with previous colocalization reports using the TRAP2 mouse (49, 50), although these prior TRAP2 studies were conducted in the cortex and the bed nucleus of the stria terminalis (BNST), which contain distinct cell types with unique gene expression and brain-wide connectivity. Such differences may account for the observed variations in TRAP/FOS overlap compared to our BLA results. Furthermore, we did not find that basal, non–sensory-evoked FOS levels (home-cageFOS) (34.51%) colocalize to the same degree as a second nociceptive stimulus (Fig. 1F). We next tested the valence specificity of nociTRAP by comparing the colocalization of neurons captured during an appetitive mating opportunity (mateTRAP), wherein male TRAP2:tdTomato mice mounted an unfamiliar female mouse and intromissions occurred (7, 51, 52). Compared to nociTRAP, we observed significantly less colocalization between mateTRAP neurons and nociFOS neurons (16.22%), confirming previous reports that valence processes are partially segregated between distinct BLA ensembles (Fig. 1G) (4, 6, 7, 9, 11, 12, 25, 52).

Fig. 1. Inhibition of the BLA^noci ensemble reduces pain aversion in acute and chronic pain.

(A) nociTRAP protocol. (B) Representative images of neurons captured by nociTRAP tdTomato (red) and nociFOS (blue) with quantification of (C) nociTRAP (D) and nociFOS across the AP axis of the BLA, lateral amygdala (LA), and basal amygdala (BA). Scale bars, 250 μm. (E) Quantification of the colocalization of nociTRAP and nociFOS in neurons across the LA, BA, and BLA. (F) Representative images of nociTRAP colocalization with nociFOS (blue) versus home-cageFOS (gray) and quantification. Scale bars, 100 μm. (G) Colocalization of mateTRAP (purple) and nociFOS (blue). Scale bar, 100 μm. Quantification of colocalization of nociFOS with neurons captured by mateTRAP relative to nociTRAP. (H) Schematic and timeline of SNI. (I) Colocalization of uninjured nociTRAP (red) and light touch after SNI FOS (orange). Scale bar, 100 μm. Quantification of colocalization of light-touchFOS in animals than underwent SNI relative to a second acute nociceptive stimulus in uninjured mice [n = 4 nociTRAP nociFOS (2 male), n = 3 nociTRAP home-cageFOS (all female), n = 3 mateTRAP nociFOS (all male), and n = 3 nociTRAP SNI light-touchFOS (2 male)]. (J) Timeline and (K) schematic of experiment. (L) Histological confirmation of bilateral expression of stGtACR2 in the BLA^noci ensemble with bilateral fiber optics. Scale bars, 500 μm (left image) and 250 μm (right image). (M) Behavioral testing setup. (N) Optogenetic inhibition of the BLA^noci ensemble decreases the behavioral response to noxious stimuli compared to a prestimulus baseline [n = 23 StGtACR2 (12 male) and n = 20 tdTomato (10 male)]. (O) The increased response to innocuous and noxious stimuli was reduced during inhibition of the BLA^noci ensemble compared to prestimulus baselines [n = 11 StGtACR2 SNI (6 male), n = 10 tdTomato uninjured (5 male), and n = 10 tdTomato SNI (5 male)].

Previously, single-cell-resolution Ca²⁺ imaging of the BLA revealed that acute pain and chronic pain are encoded by the same BLA neurons. Here, we confirmed this finding by examining the overlap of nociTRAP and FOS activity in a chronic neuropathic pain state. To induce chronic pain, we performed the spared nerve injury (SNI) model of chronic neuropathic pain on nociTRAP:tdTomato mice (Fig. 1H) (53). Three weeks after SNI, we exposed the injured left hindpaw to a static light touch with 0.16-g von Frey filament to evoke FOS (light-touchFOS) as a model of mechanical allodynia. We found that light touch in an SNI state reactivated a similar percentage of nociTRAP neurons (61.69%) compared to nociTRAP mice receiving a second noxious stimulus in an uninjured state. This suggests that the BLA^noci ensemble is stably responsive across the acute-to-chronic pain transition (Fig. 1I).

Acute and chronic pain behaviors require a common BLA ensemble

Our current and previous observations raise the question of whether the same BLA neurons active in acute pain are required for chronic pain. We tested this critical question using an optogenetic approach, expressing the soma-targeted blue-light inhibitory opsin, stGtACR2-FusionRed (54), or a control tdTomato fluorophore, bilaterally in BLA^noci neurons of nociTRAP2 mice. We examined stimulus-evoked reflexive and affective-motivational pain behaviors (i.e., nocifensive) before and during inhibition of BLA^noci neurons (Fig. 1, J to M). We demonstrated that in support of our previous chemogenetic results (12), BLA^noci ensemble inhibition did not affect sensory thresholds but decreased affective-motivational behavior to noxious stimuli in uninjured mice (Fig. 1N). This analgesic effect of BLA^noci ensemble inhibition did not display a sex-specific effect (fig. S1A). Chemogenetic activation of the BLA^noci ensemble in the absence of injury caused a conditioned place aversion in female mice but had no effect on nociception-related behavioral responses (fig. S2). This suggests that the activity of the BLA^noci ensemble can drive negative affective-motivational behavior but does not further potentiate nociceptive behavior when this ensemble is already active.

Our next goal was to determine whether these same neurons within the BLA^noci ensemble also causally contribute to chronic neuropathic pain behavior. We performed the same optogenetic inhibition protocols in the same mice 3 weeks after either SNI or no injury. SNI mice demonstrated neuropathic hypersensitivity compared to their preinjury responses and uninjured controls (fig. S1B). While inhibition did not alter reflexive behavior, we observed decreased allodynia and hyperalgesia behavior to dynamic brush, acetone, pin prick, and 55°C water droplets during blue-light inhibition in SNI nociTRAP:stGtACR2 mice but not SNI nociTRAP:tdTomato controls (Fig. 1O). This demonstrates that chronic neuropathic pain relies on the same BLA^noci ensemble circuitry as acute pain and the engagement of this ensemble displays a level of stability across the acute-to-chronic pain transition. There was no effect of sex on this analgesic effect (fig. S1C). No effect was observed on place preference for the SNI nociTRAP:stGtACR2 mice or other groups, suggesting that inhibition of the BLA^noci ensemble is not engaging motivational or reward-related neural circuitry (fig. S3). Together, these results demonstrate that the BLA^noci ensemble displays a level of functional stability and is a valence-specific ensemble of neurons facilitating affective-motivational pain behaviors.

A transcriptomic atlas of the amygdala reveals nociceptive neurons

Our histological and optogenetic data, in addition to our prior single-cell Ca²⁺ imaging of the BLA (12), confirm that the BLA^noci ensemble is a functionally distinct set of neurons capable of dictating specific behavior responses. However, the molecular signatures that distinguish this nociceptive ensemble remain undefined. The genetic identity of the BLA^noci ensemble has not been previously examined beyond a few genes, such as Slc171a7 (VGLUT1), identifying the ensemble as glutamatergic, and Rspo2, a marker for the putative negative valence BLA ensemble (7, 11, 12). Therefore, to define the transcriptomic signatures of acute and chronic pain in an unbiased manner, we used snRNA-seq to transcriptionally profile the BLA. We collected amygdalar tissue [right hemisphere only, 1 mm between AP: −0.83 and −1.83 mm from bregma] from uninjured or SNI mice immediately following (~5 min) either no stimulus, a light touch, or a noxious 55°C water droplet stimulation of the left hindpaw (n = 3 male mice per group; Fig. 2A). Affective-motivational pain behaviors were qualitatively confirmed for mice stimulated with noxious 55°C water droplets. We chose this timeline to capture the maximum expression of Fos mRNA in the nucleus on the basis of the transcriptional kinetics of the IEG (fig. S4, A to E) (55). All tissue samples were collected and fully processed on the same day to avoid batch effects. Analysis of nuclei from uninjured groups identified 30 cell type clusters, including 18 neuronal clusters, from 72,125 nuclei (Fig. 2B). Nuclei from all stimulus treatments were equally distributed across the 30 clusters (Fig. 2, C and D). On the basis of known cell class–defining marker genes and differentially expressed genes (DEGs), we parsed the clusters into 9 GABAergic neuron clusters, 6 glutamatergic neuron clusters, 3 neuron clusters that labeled putatively nonamygdalar brain regions (e.g., caudate putamen), and 12 non-neuronal clusters (e.g., microglia, astrocytes, etc.) (Fig. 2, D to F). With our large-scale amygdalar transcriptomic atlas, we next homed in on two subnuclei that have been strongly implicated in pain—the central amygdala (CeA) and the BLA (56, 57). From the classified neuronal pool, on the basis of published markers and the Allen Brain in situ hybridization database (fig. S5), we initially defined BLA subclusters as nuclei in the ≥75th percentile for Slc17a7 and putative CeA as the ≥75th percentile for Gad1 and/or Gad2 (and Slc17a6− and Slc17a7−). Further clustering yielded 13 inhibitory subclusters, 8 putative BLA subclusters, 1 intercalated cell (ITC9) cluster, and 1 endopiriform cortex (Endopiro5) cluster (Fig. 2, G and H, fig. S5, and table S1). Subsequently, we compared several published marker genes for unique cell types involved in valence processes and nociception within the BLA and CeA, namely, Rspo2 and Fezf2 for BLA valence ensembles (7, 9, 11, 12) and Prkcd (PKCδ), Sst (Somatostatin), and Calcrl (CGRP receptor) for CeA valence/nociception ensembles (fig. S5) (18, 58–64). Transcriptional profiles did not adequately distinguish the BLA GABAergic cell types (e.g., somatostatin, parvalbumin, and vasoactive intestinal peptide interneurons) from the CeA GABAergic cell types expressing the same marker genes; thus, we have labeled all Gad1+ and Gad2+ nuclei as “inhibitory cell” clusters rather than CeA, and no further analyses are conducted on these cells within this study.

Fig. 2. The BLA^noci ensemble is a functional subpopulation of Rspo2 BLA neurons.

(A) Experimental design. Mice in three uninjured groups: no stimulus (n = 3), light touch (n = 3), or 55°C water droplet stimulation (n = 3) (all male mice, FISH done on mice of both sexes). (B) UMAP of all nuclei (n = 72,125) from uninjured mice (n = 9 total) captured by amygdala punches in 30 unique clusters. (C and D) Contribution of nuclei from stimulation groups to the total UMAP and cell types. (E and F) Dot plot and heatmap displaying a top gene differentiating the 30 cell clusters. (G) UMAP of amygdalar neuron nuclei (n = 51,775) showing 13 inhibitory (Gad1+ or Gad2+) and 10 BLA (Vglut1+) putative clusters. (H) Heatmap displaying top genes differentiating the 23 neural clusters. (I) Feature plots of genes for cell classes, published gene markers for valence/nociception in the BLA and CeA, and IEGs. (J) Representative 4× and 20× fluorescence images and quantification of FISH. Scale bar, 100 μm. (K) Violin plot displaying expression of genes identified in past BLA RNA sequencing, including the expression of Rspo2, by subcluster. (L) Volcano plot of 350 DEGs between Rspo2+ BLA clusters 0 and 4. Yellow dots indicate genes enriched in the BLA subcluster, and orange dots indicate genes enriched in the BLA4 subcluster. (M) IEG modular activity scores of 10 putative BLA clusters across stimulation conditions segregated by percentile thresholds. (N) Volcano plots of BLA clusters 0 and 4 displaying DEG upper stimulation condition enriched in the IEG+ or IEG− nuclei. Purple dots indicate genes only identified as differentially expressed within that condition, and gray dots indicate shared DEGs across multiple conditions.

Next, we mapped several IEGs as proxies for stimulus-evoked neuronal activity and observed expression across the uniform manifold approximation and projection (UMAP) space including within the Rspo2+/Slac17a+ regional clusters (Fig. 2I). As Rspo2+ BLA neurons have been implicated in negative valence processing (7, 11), we explored the colocalization of Rspo2 in nociTRAP BLA neurons and found that 78.63% of nociTRAP BLA neurons express Rspo2 (Fig. 2J). Therefore, we examined the expression of Rspo2 and other genes among the putative-BLA subclusters and identified three subclusters that express Rspo2: BLA0, BLA4, and a third cluster with low Rspo2 (defined as Endpiro5) (Fig. 2K). Upon further investigation of the Allen Brain in situ hybridization database based on the DEG analysis of these three clusters, we confirmed that genes enriched in Endopir5 largely map to the endopiriform cortex, while additional BLA0- and BLA4-enriched genes specifically localize to the BLA (Fig. 2K, fig. S5, D and E, and table S1). Similar to prior genetic identities for valence ensembles, we do not observe the coexpression of Fezf2 nor Ppp1r1b within our Rspo2+ BLA0 and BLA4 subclusters (fig. S5A) (7, 9). We identified 401 DEGs between the Rspo2+ BLA0 and BLA4 subclusters (Fig. 2L), including genes implicated in negative affective and pain behavior, such as Sorcs3, Fstl4, and Garnl3 (65–67). Using fluorescence in situ hybridization (FISH), we validated the expression of four genes identified as differentially expressed between Rspo2+ BLA0 and BLA4 in male and female mice. Expression of Blnk, Etv1, Rerg, and Col5a2 was detected in the BLA, with little expression outside of the BLA (fig. S6). The largest population of Rspo2+ BLA cells expresses both Blnk and Etv1 mRNAs (69.17%), but we also identified populations of Rspo2+ cells that exclusively express Blnk (11.32%) or Etv1 (10.85%), with an additional subpopulation that expresses neither of these genes (8.65%) at detectable levels (fig. S6, A and B). There was no significant difference between the percentage of Rspo2+ BLA cells that express Blnk or Etv1 and no effect of sex on these results (fig. S6, C and D). To validate that these DEGs are also expressed in BLA^noci cells, we examined colocalization of Blnk or Etv1 with tdTomato mRNA, marking cells captured by the nociTRAP protocol in TRAP2:tdTomato mice. A significantly higher percentage of BLA^noci cells expresses Etv1 mRNA (69.45%) compared to Blnk mRNA (42.47%) with no effect of sex (fig. S6, E to H). To further validate our snRNA-seq results, we examined colocalization of Rspo2, Rerg, and Col5a2 (fig. S6I). We observed populations of Rspo2+ BLA cells that expressed both Rerg and Col5a2 mRNA (27.56%), Rerg only (42.58%), Col5a2 only (4.61%), or neither Rerg nor Col5a2 mRNA (25.25%) when combining males and females. There was a significant effect of sex, with male mice displaying a significantly larger percentage of Rspo2+ cells that express both Rerg and Col5a2 mRNA (41.26%) compared to females (13.87%) and female mice displaying a significantly larger percentage of Rspo2+ cells that express neither Rerg nor Col5a2 mRNA (36.77%) compared to males (13.72%) (fig. S6J). Significantly more Rspo2+ BLA cells express Rerg (70.15%) compared to Col5a2 (32.17%) mRNA. We again observed a significant effect of sex, with males displaying a higher percentage of Col5a2 expression in Rspo2+ BLA cells (48.76%) compared to females (15.58%) (fig. S6, K and L). Together, these data define two Rspo2+ BLA subclusters that are genetically distinguishable, both of which include genes linked to pain and negative affective behavior.

Next, we sought to identify BLA^noci neurons using transcriptional profiles of Rspo2+ BLA cells. Notably, the BLA^noci neurons are a subpopulation of the total Rspo2+ BLA neurons (36.86%) (Fig. 2J). To identify the nociceptive BLA cell types, we examined the modular expression of a panel of 25 IEGs (68) against randomly selected background expression between our three stimulus conditions (no stimulus, light touch, and noxious 55°C water droplets) across BLA subclusters. The BLA subclusters displayed variable IEG activity in response to the stimuli, including to the noxious 55°C water droplets. This suggests that despite genetic similarities that grouped nuclei into the subclusters, there is heterogeneous functional activity of BLA nuclei, even within the Rspo2+ BLA subclusters 0 and 4 (Fig. 2M).

Individual analyses of BLA0 and BLA4 did not provide sufficient cell numbers to confidently identify DEGs between IEG+ and IEG− cells. However, across the two Rspo2+ BLA subclusters combined, we identified the genes that are differentially expressed across stimulation conditions (Fig. 2N; with purple indicating genes only identified as differentially expressed within that condition and gray indicating shared DEGs across multiple conditions). Many genes were uniquely up-regulated under the noxious condition, including Htr2a, Hgf, and Zfp536, and have been implicated in pain, pain relief, or negative affective behavior (69–72). Other genes (Pde10a, Sorcs3, and Fstl4) that are found in multiple IEG+ groups or in response to non-noxious stimuli are implicated in pain and negative affective behavior (65, 66, 73), suggesting that the highly active BLA0/4 neurons express greater levels of genes related to pain processing (Fig. 2N). Overall, these findings define a BLA cell population that is highly responsive to nociceptive stimuli, providing a transcriptomic classification of pain-active neurons in the amygdala.

Chronic pain leaves a transcriptomic imprint on nociceptive BLA cell types

While our behavioral and anatomical data suggest that the BLA is uniquely important for the expression of acute and chronic pain, the genetic changes that accompany pain chronification in the BLA remain unclear. To test this, we used snRNA-seq as above to compare the BLA from uninjured and 3 weeks post-SNI mice given either a light-touch stimulus (0.16-g von Frey filament) or no stimulus. During the stimulation protocol, affective-motivational pain behaviors were qualitatively confirmed by the experimenter. From 73,512 neuronal nuclei across four injury and stimulus conditions, we identified the same 10 glutamatergic cell types (Fig. 3A). These 10 subclusters expressed a similar pattern of marker genes defining the amygdalar cell types from the uninjured-only mouse cohorts in Fig. 2G (Fig. 3, B and C). All four conditions (uninjured + no stimulus, uninjured + light-touch, SNI + no stimulus, and SNI + light-touch) were equally distributed across the 18 neuronal clusters (fig. S7). From the total neuronal pool, we also examined the expression of genes previously associated with pain in clinical and preclinical studies across the 18 clusters (fig. S8 and table S2) to examine future targets for pharmaceutical development.

Fig. 3. Chronic pain imparts a unique transcriptomic signature on the BLA.

(A) UMAP of all neuronal nuclei (n = 73,512) from n = 12 mice across uninjured [no stimulus, n = 3; light touch, n = 3 (all male mice)] and chronic neuropathic pain (SNI) [no stimulus, n = 3; light touch, n = 3 (all male mice)] conditions identifying 36 unique cell type clusters. (B) Feature plots of genes for cell classes and for Rspo2. (C) Violin plot displaying the expression of genes identified in past BLA RNA sequencing, including the expression of Rspo2, by subcluster. (D) Volcano plots of all BLA subclusters combined displaying DEGs unique across injury and stimulation conditions. Red dots indicate genes enriched under that condition, blue dots indicate genes enriched under all other conditions, and gray dots indicate shared DEGs across conditions. (E) Synaptic gene ontology analysis of DEGs from (D) up-regulated under SNI + no stimulus or SNI + light touch conditions indicating the cellular compartments these DEGs are associated with. (F) IEG modular activity scores of 10 putative BLA clusters across stimulation conditions. (G) Volcano plots of Rspo2+ BLA subclusters 0 and 4 displaying DEGs unique between IEG+ and IEG− nuclei in SNI. Red dots indicate genes enriched under that condition, blue dots indicate genes enriched under all other conditions, and gray dots indicate shared DEGs across multiple conditions.

Next, we compared the DEGs unique to each of the four injury and stimulation conditions across all BLA neurons and identified genes that are differentially expressed following SNI (Fig. 3D and table S3). Genes uniquely up-regulated or down-regulated under these conditions, relative to no-injury control nuclei, are colored red or blue, respectively, while the neuropathically altered DEGs common to both SNI + no stimulus and SNI + light touch conditions are colored gray. While Rspo2 is up-regulated under the SNI + no stimulus condition, it is down-regulated under the SNI + light touch condition (Fig. 3D). This indicates that there is differential expression of Rspo2 under baseline versus evoked neuropathic pain conditions. We examined the up-regulated genes under the SNI + no stimulus and SNI + light touch conditions using a synaptic gene ontology analysis and identified that the DEGs up-regulated in chronic pain primarily localize to the synapse, suggesting an important recruitment of brain regions that are downstream of the BLA (Fig. 3E and fig. S9). To further isolate the nociceptive neurons of the BLA, we examined the expression of 25 IEGs (68) in response to no stimulus or a light touch in uninjured or SNI mice to identify the subset of neurons that are active under chronic pain conditions. Similar to an uninjured state, BLA0 and BLA4 neurons displayed variable IEG activity in response to a light touch in an SNI state, suggesting heterogeneous changes in baseline activity following SNI and that subsets of each cluster are active under basal conditions and following light touch in SNI mice (Fig. 3F). While the individual analysis of BLA0 and BLA4 again did not provide sufficient cell numbers to identify DEGs in IEG+ versus IEG− nuclei, the combined Rspo2+ BLA0 and BLA4 subclusters across stimuli in SNI mice identified 22 DEGs under the SNI + no stimulus condition and 58 DEGs under the SNI + light touch condition (Fig. 3G and table S3). One gene up-regulated under the SNI + light touch condition, Alkal2, a ligand for the tyrosine kinase receptor, has been previously demonstrated to regulate the behavioral expression of chronic pain (74). This suggests that Alkal2-related processes may partially mediate the nociceptive negative valence of chronic pain within the Rspo2+ BLA neurons. Together, this transcriptional characterization of the BLA in acute and chronic pain identifies multiple candidate genes that are altered by chronic pain and an enrichment of genes involved in synaptic signaling and transmission (fig. S9). This highlights the necessity of examining the effect of chronic pain on the axonal projections of BLA^noci neurons at distinct downstream brain regions.

BLA^noci neurons project to a NAc nociceptive hotspot

snRNA-seq of the BLA suggests that chronic pain changes the differential expression of genes related to downstream axon regulation and signaling. Therefore, we next sought to examine the brain-wide targets of the BLA^noci ensemble to investigate the routes of nociceptive information flow. To begin, we traced axon outputs of the BLA^noci ensemble using anterograde fluorescence tracing. We transfected the BLA of TRAP2:tdTomato mice with a Cre-dependent viral green fluorescent protein (GFP) and captured nociceptive neurons using our nociTRAP protocol, which labels all nociceptive neurons throughout the brain with tdTomato and fills the BLA^noci ensemble with GFP for axon mapping within downstream regions (Figs. 1A and 4A). The viral injection was primarily localized to the BLA, with some spread to the adjoining piriform cortex. We then visualized and quantified BLA^noci ensemble axon density throughout the brain. We observed BLA^noci ensemble axons in brain regions with known roles in affective processing [e.g., anterior cingulate cortex (ACC), insular cortex, olfactory tubercle (OT), NAc, BNST, and CeA] that contained high numbers of nociTRAP:tdTomato neurons, which we used to guide further detailed quantification (Fig. 4, B and C). Relative to the other downstream brain regions, we observed a nociceptive hotspot in the NAcSh (Fig. 4, D and E). We then mapped the specific location of these nociceptive neurons throughout the NAcSh and found a dense cluster in the medial-dorsal aspect that increased in the posterior coordinates of the NAcSh (Fig. 4, F to H). The NAcSh nociceptive neurons (NAcSh^noci) were located in close proximity to BLA^noci ensemble axons (Fig. 4, G and H). To further examine the cellular stability and valence selectivity of nociception within individual NAcSh^noci neurons across different pain models, time, and valence assignment, we performed similar experiments using the TRAP2:tdTomato mouse as in the BLA (Fig. 1). We found that nociTRAP and nociFOS neurons of the NAcSh displayed 44.79% colocalization, suggesting the stability of the NAcSh^noci ensemble to repeatedly encode nociception. As within the BLA, a higher percentage of nociTRAP neurons is reactivated by a second nociceptive stimulus than nociFOS neurons colocalizing with nociTRAP neurons (15.79%) (fig. S10A). We did not find that basal, non–sensory-evoked FOS levels in the NAcSh^noci (home-cageFOS) colocalize (20.89%) to the same degree as a second nociceptive stimulus (Fig. 4I). We next tested the valence specificity of the NAcSh^noci ensemble by comparing the colocalization of neurons captured during mateTRAP. We observed significantly less colocalization between mateTRAP neurons and nociFOS neurons (27.62%) (Fig. 4J). As the NAcSh has been implicated in the expression of chronic pain, we tested whether the NAcSh has a similar functional architecture of pain processing stability as we observed in the BLA. We found that light touch in an SNI state reactivated a significantly greater percentage of NAcSh^noci ensemble neurons (62.01%) compared to uninjured mice receiving a second nociceptive stimulus. This suggests that the NAcSh^noci ensemble mirrors the functional stability of the BLA^noci ensemble and may show an increased recruitment in chronic pain (Fig. 4, K and L). Our retrograde viral tracing strategy enabled an examination of brain-wide nociceptive and non-nociceptive inputs to the mNAcSh. We observed inputs beyond the BLA, including the paraventricular thalamus (PVT), ventral tegmental area (VTA), ventral hippocampus (vHC), and the lateral hypothalamus (fig. S10, L and M). Last, we examined the genetic identity of the NAcSh^noci neurons, as there is extensive literature implicating the dopamine 2 receptor–expressing neurons of the mNAcSh in chronic pain. We found that this functional population of nociceptive neurons was genetically heterogeneous, with some cells expressing the dopamine 1 receptor mRNA and others expressing the dopamine 2 receptor mRNA (fig. S10N).

Fig. 4. The BLA^noci ensemble projects to a nociceptive hotspot in the dorsomedial NAcSh.

(A) Expression of GFP in the right BLA^noci ensemble. Scale bar, 100 μm. (B) BLA^noci ensemble axons with nociceptive tdTomato neurons. Scale bars, 200 μm. (C) Quantification of BLA^noci axon density by area and tdTomato in ipsilateral ROIs. a.u., arbitrary units. (D) nociTRAP captures a nociceptive hotspot in the dorsomedial NAcSh (NAcSh^noci). Scale bars, 500 μm. (E) BLA^noci ensemble axons and NAcSh^noci neurons throughout the AP axis of the NAc. Scale bar, 200 μm. (F) Quantification of tdTomato NAcSh neurons in the medial versus lateral shell. (G) Images (60×) of BLA^noci ensemble axons and NAcSh^noci neurons. Scale bars, 50 μm. (H) Quantification of NAcSh^noci neurons and BLA^noci ensemble axon density across the ipsilateral NAcSh. (I) Representative images of nociTRAP colocalization with nociFOS versus home-cageFOS. Scale bars, 100 μm. (J) Colocalization of mateTRAP and nociFOS. Scale bar, 100 μm. Colocalization of nociFOS with neurons captured by mateTRAP or nociTRAP. (K) Timeline of SNI. (L) Colocalization of uninjured nociTRAP and SNI light-touchFOS. Scale bar, 100 μm. Colocalization of SNI light-touchFOS versus a second acute nociceptive stimulus in uninjured mice [n = 4 nociTRAP nociFOS (2 male), n = 3 nociTRAP home-cageFOS (all female), n = 3 mateTRAP nociFOS (all male), and n = 3 nociTRAP SNI light-touchFOS (2 male)]. (M) Retrograde tracing and reactivation of afferents in the mNAcSh. (N) Timeline. (O) Representative images of the injection site in the mNAcSh and the BLA. Scale bars, 200 μm. (P) Quantification of nociceptive and non-nociceptive mNAcSh-projecting BLA cells. (Q) Colocalization of light-touchFOS in nociceptive and non-nociceptive mNAcSh-projecting BLA cells [n = 4 (1 male)].

While the BLA^noci ensemble axons appeared relatively sparse near the NAcSh^noci neurons, we confirmed that the BLA^noci ensemble projects to this region by transfecting the mNAcSh of TRAP2 mice with a retrograde color-switch adeno-associated virus (AAV). This technique allowed us to label afferents to the mNAcSh that are nociceptive (nociTRAP) with GFP and non-nociceptive afferents with mCherry (Fig. 4M). We further labeled chronic pain-active neurons by stimulating the injured hindpaw of SNI mice with a light touch to evoke FOS (Fig. 4, N to Q, and fig. S10, B to D). Of all BLA cells, 18.34% project to the mNAcSh; of these projectors, 16.58% are nociceptive, confirming that the BLA → mNAcSh nociceptive projection is a subpopulation of all BLA outputs to the mNAcSh (Fig. 4P and fig. S10, E to J). When we looked at the FOS+ cells, we found that a significantly larger number of BLA^noci → mNAcSh cells were reactivated in chronic pain compared to that of the non-nociceptive projection neurons (Fig. 4Q and fig. S10K). This suggests that the stability of the BLA^noci ensemble across pain states is conserved in the projections to the mNAcSh, indicating that this projection is used for both acute and chronic pain–related information transfer. Furthermore, while we observed that the BLA^noci ensemble encompasses the entirety of the AP axis of the BLA, the BLA^noci → mNAcSh ensemble is primarily localized to the posterior half of the BLA (fig. S10, F and G). We conclude that there is a further subdivision of BLA^noci neurons, which suggests differential roles of pain processing in anterior versus posterior BLA^noci neurons that can be refined on the basis of the downstream projection target. While some axon tracing injections transfected the piriform cortex as well as the BLA (Fig. 4A), we did not observe neurons that project to the mNAcSh in the piriform cortex using retrograde tracing.

The BLA^noci ensemble transmits nociceptive information to the NAc

Ca²⁺ recordings from genetically defined BLA somas retrolabeled from the NAc show increased activity in response to negative valence information, including to noxious foot shock (9). However, given the branching architecture of axonal collaterals and the observation that electric shocks do not fully activate the BLA^noci ensemble (12), it remains unknown whether the BLA axons transmit information about acute nociceptive stimuli to the NAc nor how the presence of chronic pain changes this transmission in response to noxious and innocuous stimuli. No direct BLA axon terminal Ca²⁺ recordings within the NAc have been performed. To synthesize our functional, transcriptomic, and projection-defined characterization of nociceptive processing within the BLA, we recorded BLA axon terminal Ca²⁺ activity in the mNAcSh in response to stimuli across valence and sensory modalities. Because our transcriptomic data suggested that the Rspo2+ BLA neurons display increased IEG activity in response to noxious stimuli, we used Rspo2-Cre mice (7) and implanted a fiber optic above the NAcSh nociceptive hotspot we identified (NAcSh^noci) (Fig. 5A and fig. S11). Unexpectedly, this putative negative valence ensemble displayed decreased Ca²⁺ activity in response to innocuous, aversive, and appetitive stimuli (Fig. 5, B to D, and fig. S12A). Our sequencing of the BLA identified two unique subclusters of BLA neurons that express Rspo2, containing individual nuclei that show heterogeneous IEG responses to noxious stimuli (Figs. 2M and 3F). Our axon Ca²⁺ activity recording along with this transcriptomic observation suggests that focusing on a single genetic marker may miss critical information about valence function within the BLA. Therefore, we recorded the Ca²⁺ activity of BLA^noci ensemble axon terminals in the mNAcSh using the TRAP2 mouse (Fig. 5E). The BLA^noci ensemble axons showed increased Ca²⁺ responses to noxious stimuli across sensory modalities and decreased Ca²⁺ responses to innocuous and appetitive stimuli (Fig. 5, F and H, and fig. S12B). This supports previous findings that valence ensembles in the BLA display decreased activity to opposing valence stimuli (7). Last, we tested the impact of chronic pain on nociceptive transmission from the BLA^noci ensemble to the mNAcSh (Fig. 5I). In a chronic neuropathic pain state, the BLA^noci ensemble axons show increased Ca²⁺ activity in response to a light touch, suggesting a neural signature of allodynia in the BLA to the mNAcSh circuit (Fig. 5, J to L, and figs. S12C and S13). We conclude that there are functional subpopulations of Rspo2+ BLA neurons important for transmitting nociceptive information to the mNAcSh in acute and chronic pain. Furthermore, examining neural populations based only on single genetic markers or single projection targets or not considering neural ensembles at distinct anatomical locations within the BLA misses important information for identifying functional subpopulations related to valence processing.

Fig. 5. The BLA transmits nociceptive information to the dorsomedial NAcSh.

(A) AAV5-hSyn1-FLEX-axon-GCaMP6s in the BLA of Rspo2-Cre mice; expression of axon-GCaMP6s in the NAcSh with the optic fiber above the mNAcSh. Scale bars, 500 μm (left image) and 100 μm (right image). (B) BLA Rspo2 axon terminals in the NAcSh display significantly decreased Ca²⁺ activity in response to innocuous, noxious, and appetitive stimuli compared to prestimulus baselines. (C) Quantification of peak z-score of all stimuli. (D) Quantification of AUC (arbitrary units; calculated as AUC poststimulus (0 to 5 s) minus AUC prestimulus (−10 to −5 s)] of all stimuli [n = 7 (4 males); n = 6 for sucrose]. (E) AAV5-hSyn1-FLEX-axon-GCaMP6s in the BLA^noci ensemble of TRAP2 mice; expression of axon-GCaMP6s in the NAcSh with the optic fiber above the mNAcSh. Scale bars, 500 μm (left image) and 100 μm (right image). (F) BLA^noci ensemble axon terminals in the NAcSh display significantly increased Ca²⁺ activity in response to noxious stimuli compared to prestimulus baselines and decreased activity in response to innocuous and appetitive stimuli. (G) Quantification of peak z-score of Ca²⁺ responses to stimuli. (H) Quantification of AUC of all stimuli [n = 13 (6 males)]. (I) Timeline and stimulation and recording paradigm. (J) BLA^noci ensemble axons in the NAcSh display increased Ca²⁺ activity in response to a light touch compared to uninjured mice. (K) The AUC was elevated in response to a light touch in mice with SNI compared to uninjured mice. (L) Quantification of peak z-score of the Ca²⁺ responses to stimuli between uninjured and SNI mice [n = 13 uninjured (6 males) and n = 5 SNI (2 males)]. The peak z-score is calculated over 10 s poststimulus.

NAcSh-projecting BLA^noci neurons are essential for pain behavior

The transmission of acute and chronic pain–related information from the BLA^noci ensemble to the mNAcSh suggests that the subpopulation of the mNAcSh-projecting BLA^noci ensemble neurons may play a key role in the importance of the total population of pain-related behavior. The BLA^noci ensemble projects to many downstream regions, and the importance of this subpopulation had not been tested. We tested this using an intersectional chemogenetic approach (75), expressing the inhibitory chemogenetic protein hM4D(Gi) or control OScarlet protein in the BLA^noci ensemble neurons that project to the mNAcSh (Fig. 6A). We examined stimulus-evoked reflexive and affective-motivational behaviors before and during inhibition of NAcSh-projecting BLA^noci neurons (Fig. 6, B and C). We demonstrated that similar to inhibiting the total BLA^noci ensemble population, chemogenetic inhibition of this subpopulation did not affect sensory thresholds but decreased affective-motivational pain behavior to noxious stimuli in uninjured mice, specifically to noxious thermal stimuli (Fig. 6D). This effect was not sex-dependent (fig. S14A).

Fig. 6. Inhibition of the mNAcSh-to-BLA^noci ensemble reduces affective-motivational behaviors during acute and chronic pain.

(A) Schematic for the intersectional viral approach to express the inhibitory chemogenetic protein hM4D(Gi) or control OScarlet in BLA^noci neurons that project to the mNAcSh. (B) Timeline of chemogenetic experiment and behavioral paradigms. (C) Histological confirmation of the bilateral expression of hM4D(Gi)-mCherry in the BLA^noci-to-mNAcSh ensemble. Scale bars, 1 mm (left image, 4×) and 250 μm (right image, 20×). (D) Chemogenetic inhibition of the mNAcSh-projecting BLA^noci ensemble decreases the behavioral response to noxious thermal stimuli compared to a prestimulus baseline [n = 16 hM4 (9 males) and n = 20 OScarlet (12 males)]. (E) The post-SNI increased response to thermal stimuli was reduced during inhibition of the mNAcSh-projecting BLA^noci ensemble compared to prestimulus baselines [n = 9 hM4 uninjured (5 male), n = 7 hM4 SNI (4 male), n = 10 OScarlet uninjured (6 male), and n = 10 OScarlet SNI (6 male)]. (F and G) Representative 20× images of the BLA^noci ensemble cells that project to the mNAcSh and their axon collaterals throughout the brain (scale bar, 100 μm). (H) Pattern of axon projections from the BLA^noci ensemble cells that project to the mNAcSh throughout the AP axis of the NAc. Scale bar, 100 μm.

Our next goal was to determine the necessity of this subpopulation on affective-motivational pain behavior in a chronic pain state. We performed the same chemogenetic inhibition protocols in the same mice 3 weeks after either SNI or no injury. SNI mice demonstrated neuropathic allodynia and hypersensitivity compared to their preinjury responses and uninjured controls (fig. S14B). While inhibition did not alter reflexive behavior, we observed decreased allodynia and hyperalgesia behavior to acetone, pin prick, and 55°C water droplets during inhibition in SNI nociTRAP:hM4D(Gi) mice but not SNI nociTRAP:OScarlet controls (Fig. 6E). As with uninjured animals, this analgesic effect of chemogenetic inhibition was not sex-dependent (fig. S14C). Together, these results demonstrate that the subpopulation of NAcSh-projecting BLA^noci ensemble neurons plays an essential role in affective-motivational pain behaviors in acute and chronic pain states.

NAcSh-projecting BLA^noci neurons are interconnected with affective brain regions

Our results demonstrate that the NAcSh-projecting BLA^noci neurons critically contribute to the necessity of the overall BLA^noci ensemble for acute and chronic pain–related behavior. However, the BLA^noci ensemble projects to many downstream regions that may be important for affective-motivational pain behavior (Fig. 4, A to C). We therefore wanted to examine the collateral axon projections of the NAcSh-projecting BLA^noci neurons to test whether the projection pattern of this subpopulation was unique. We tested this by visualizing the axon projections throughout the brain from control mice expressing OScarlet protein in NAcSh-projecting BLA^noci neurons (Fig. 6, A and F). We observed axons in brain regions previously identified [ACC, insular cortex (insula), NAc, OT, and BNST] with additional strong innervation in the CPu (caudate putamen) and the vHC (Fig. 6F). We also examined the pattern of innervation to the NAcSh and core to determine whether the pattern of innervation was different. We observed axons throughout the NAc core and shell, including medial and lateral shell, with potentially less dense innervation in the lateral shell (Fig. 6G). In all, the subpopulation of NAcSh-projecting BLA^noci ensemble neurons projects to many of the same regions that the total ensemble does.

DISCUSSION

The BLA^noci ensemble is essential across pain states

Building on our previous work using single-cell calcium imaging and the TRAP1 mouse model (12, 76), we show that the BLA contains a functionally stable ensemble of nociceptive neurons—here referred to as the BLA^noci ensemble—that remains engaged across both acute and chronic pain states. This ensemble is distinct from those encoding neutral or appetitive valence, consistent with prior demonstrations of nonoverlapping valence-specific neural populations in the BLA (6, 7, 9–11, 25, 52).

Only 27.99% of nociFOS+ neurons were captured by the TRAP-based labeling procedure. This partial overlap may reflect differences in the neural representation of an initial, unexpected exposure to a noxious stimulus compared to a second, familiar exposure—consistent with prior findings that the BLA is involved in encoding both unconditioned and conditioned stimulus responses (4). Our study focuses specifically on the unconditioned, innate nociceptive ensemble, reinforcing the idea that this subpopulation plays a key role in real-time affective-motivational responses to pain. We show that this same BLA^noci ensemble is required for the affective-motivational component of both acute pain and chronic pain. This functional stability highlights the potential of the BLA as a therapeutic target for non-opioid analgesic strategies. Notably, optogenetic inhibition of the BLA^noci ensemble did not induce appetitive or reward-seeking behavior, suggesting that interventions targeting this population may be less prone to abuse compared to traditional opioid-based treatments.

snRNA-seq characterizes nociceptive cell types

To gain molecular insight into the BLA^noci ensemble, we conducted snRNA-seq to examine how acute pain and chronic pain influence gene expression in the BLA. Our tissue punches were centered in the BLA but also encompassed nearby amygdalar regions such as the CeA and intercalated cell masses, both of which have known roles in pain and valence processing. While the current analysis focused on the BLA, future work will explore these adjacent nuclei to build a more complete picture of amygdalar involvement in nociception and affective states (58, 60, 77).

Within the BLA, we identified two transcriptionally distinct Rspo2-expressing subclusters, BLA0 and BLA4. Rspo2 has previously been proposed as a marker of negative valence neurons in the BLA; however, our findings suggest that this population is more heterogeneous than previously thought. Although most BLA^noci ensemble neurons were Rspo2+ (76.63%), only a minority of Rspo2+ neurons was captured by the nociTRAP procedure (36.86%). This asymmetry, along with recent studies challenging the notion of a homogeneous Rspo2+ population (9), prompted us to explore transcriptional activity across these clusters in greater depth.

We examined the expression of a panel of 25 IEGs and found that BLA0 and BLA4 nuclei exhibited divergent IEG profiles in response to nociceptive stimuli. This indicates that only a subset of Rspo2+ neurons participate in pain-related activity, supporting the view that Rspo2+ neurons comprise functionally distinct subpopulations. Notably, Etv1—a gene associated with excitatory signaling—was significantly enriched in the BLA^noci ensemble, particularly within BLA4, suggesting that this subcluster may correspond to the active, pain-encoding subset of Rspo2+ cells. In contrast, Blnk mRNA, which was less represented in the BLA^noci population, may mark less responsive or unrelated Rspo2+ neurons.

Together, our findings highlight the need to integrate functional ensemble identity with molecular profiling to resolve discrepancies in the literature and to better define the cellular architecture of pain processing in the amygdala. This study represents an application of snRNA-seq to characterize nociception-responsive cell types in the BLA, revealing previously unidentified gene targets that may be leveraged in the development of precision analgesic strategies.

Chronic neuropathic pain imparts a transcriptional signature on the BLA

While new analgesics for acute pain would benefit millions, there remains an urgent unmet need for treatments that address the persistent and debilitating nature of chronic pain. Given that both acute pain and chronic pain rely on a shared BLA^noci ensemble for the affective-motivational component of pain, investigating how chronic neuropathic pain alters transcriptional landscapes within the BLA may uncover mechanisms that can be leveraged for new therapeutic strategies. Our snRNA-seq analysis revealed eight transcriptionally distinct BLA subclusters in tissue pooled from both uninjured and SNI mice—mirroring the diversity observed when analyzing uninjured tissue alone. We identified DEGs between SNI and uninjured conditions both in the total BLA population and within the Rspo2+ subclusters BLA0 and BLA4, which are enriched for nociceptive activity.

Specifically, a comparison of IEG+ versus IEG− nuclei within BLA0 and BLA4 under SNI conditions revealed chronic pain–associated transcriptional changes, including the up-regulation of Alkal2, a gene previously implicated in chronic pain (74). This provides strong motivation for future work to functionally interrogate the role of these genes in modulating affective pain behavior. Furthermore, gene ontology analysis of DEGs uniquely altered under the SNI condition—particularly among mice exposed to innocuous stimuli (e.g., light touch)—revealed enrichment for genes involved in synaptic signaling and postsynaptic organization. These findings suggest that chronic pain may reshape synaptic architecture and information flow from the BLA to downstream targets. Previous studies have demonstrated that nerve injury alters the excitability and synaptic function of BLA neurons projecting to the prefrontal cortex (78, 79).

Our work expands on these findings by identifying injury-induced transcriptional changes within a specific BLA projection pathway—namely, the BLA → NAcSh circuit. Together, these results highlight the importance of understanding the BLA^noci ensemble as part of a distributed network of brain regions mediating the affective dimensions of chronic pain. Continued investigation of this ensemble’s molecular adaptations and long-range projections will be essential for identifying durable targets for chronic pain relief.

The BLA^noci ensemble projects to a NAcSh nociceptive hotspot

Our snRNA-seq data suggested that some of the most prominent molecular changes in the BLA following chronic pain may occur within axons projecting to downstream targets. To investigate this possibility, we mapped the projection patterns of the BLA^noci ensemble using anterograde tracing. Although we observed minor off-target viral expression in the piriform cortex, we cross-referenced our tracing results with published anatomical studies and confirmed that the primary projection sites we identified were previously reported. Among these, the most notable target was a previously undescribed nociceptive ensemble in the mNAcSh, which we term the NAcSh^noci. These neurons are localized primarily in the posterior mNAcSh—a region previously associated with aversive processing (37, 41, 42, 47)—and were surrounded by dense axonal terminals from the BLA^noci ensemble. The NAcSh^noci neurons were uniquely responsive to noxious stimuli but showed no activation at the baseline or in response to appetitive stimuli, suggesting that they may serve a pain-specific role.

Similar to observations in the BLA, the NAcSh^noci neurons exhibited reactivation upon repeated exposure to nociceptive stimuli. However, only a small fraction (15.79%) of nociFOS+ NAcSh neurons was captured by nociTRAP, again indicating that the neural representation of an initial, unexpected pain exposure differs from that of a learned or conditioned nociceptive experience. This parallel between the BLA and NAcSh raises the possibility that both regions contain distinct ensembles for encoding unconditioned and conditioned pain responses. Using FISH, we further characterized the molecular identity of the NAcSh^noci population and found it to be a heterogeneous mix of Drd1+ and Drd2+ neurons. This is particularly intriguing given prior reports that Drd2+ NAcSh neurons become hyperactive in chronic pain states (29, 30). These findings highlight the need to disentangle the respective contributions of Drd1+ and Drd2+ subtypes within the NAcSh^noci ensemble to affective-motivational pain behaviors.

The NAcSh receives converging nociceptive and non-nociceptive inputs from other pain-relevant regions, including the VTA and PVT, both of which have been implicated in nociceptive and aversive processing (80, 81). Future studies should systematically evaluate how these inputs interact with BLA-derived projections to shape NAcSh^noci activity. Our anatomical tracing revealed that most BLA^noci axons in the NAcSh terminated outside the region occupied by NAcSh^noci neurons. This spatial mismatch suggests that the BLA^noci ensemble may modulate pain affective-motivational behavior not by directly exciting the NAcSh^noci population but perhaps via indirect mechanisms—such as disinhibition through intermediate GABAergic neurons. Further circuit-level studies will be necessary to clarify the precise synaptic and microcircuit relationships between BLA^noci terminals and the NAcSh^noci ensemble. Together, these findings identify the NAcSh^noci as a downstream target of the BLA’s nociceptive ensemble and provide a foundation for understanding how amygdalar output influences striatal circuits in pain.

Rspo2+ BLA projections to the NAcSh are functionally heterogeneous

Although BLA projections to the NAc have historically been studied in the context of positive valence and reward, recent work suggests that a subset of BLA → NAc neurons may also encode negative valence (9, 26). In the present study, we examined the functional identity of Rspo2⁺ BLA neurons projecting to the mNAcSh, where we identified the nociception-responsive NAcSh^noci ensemble. Unexpectedly, fiber photometry recordings revealed that Rspo2⁺ BLA axon terminals in the mNAcSh showed decreased calcium activity in response to a wide range of stimuli—including innocuous, noxious, and appetitive inputs. This broadly suppressed activity profile suggests that Rspo2⁺ BLA neurons projecting to the NAcSh do not uniformly encode negative valence. This observation aligns with our snRNA-seq data, which showed that although most BLA^noci ensemble neurons express Rspo2, only a subset of Rspo2⁺ neurons was captured by nociTRAP. This indicates substantial functional heterogeneity within the Rspo2⁺ population.

In support of this, recent findings have shown that Rspo2 can also be expressed in BLA neurons associated with appetitive valence, particularly in the posterior BLA (9). Consistent with these reports, our anatomical tracing data suggest that most BLA neurons projecting to the mNAcSh originate from posterior regions of the BLA. Together, these findings indicate that Rspo2⁺ BLA neurons projecting to the NAcSh likely comprise a mixed-valence population rather than a dedicated negative valence circuit. These results underscore a broader challenge in assigning valence function based solely on single-gene expression profiles. While Rspo2 remains a useful marker for dissecting BLA subpopulations, our findings caution against overinterpreting its functional role in isolation. To fully resolve the valence coding properties of BLA → NAc neurons, future work should combine genetic access strategies with behaviorally defined ensemble targeting and projection-specific manipulations.

The BLA^noci ensemble transmits nociceptive information to the NAcSh

To integrate our behavioral, transcriptomic, and anatomical findings, we recorded calcium activity from BLA^noci ensemble axons in the mNAcSh in response to a range of stimuli, including innocuous, noxious, and appetitive inputs. We found that these axon terminals exhibited suppressed Ca²⁺ activity in response to innocuous and appetitive stimuli, suggesting that BLA^noci projections to the NAcSh are selectively disengaged under non-noxious or emotionally neutral conditions. This suppression may reflect a broader network mechanism in which different BLA valence ensembles are differentially activated depending on stimulus context. Prior studies have reported antagonistic activity between BLA ensembles encoding opposing valence—such that activation of one is correlated with inhibition of the other (7, 10). In contrast, BLA^noci axons displayed robust increases in Ca²⁺ activity in response to noxious stimuli across multiple sensory modalities. This indicates that the BLA^noci → NAcSh projection conveys pain-specific information and is selectively engaged during aversive sensory states. The increase in terminal activity suggests downstream activation of NAcSh neurons, supporting our model in which the BLA^noci ensemble contributes to affective-motivational pain processing via its projections to striatal targets. Together, these findings provide functional evidence that the BLA^noci ensemble transmits nociceptive signals to the NAcSh in a modality-general but valence-specific manner. Further work will be needed to define the circuit-level mechanisms underlying the mutual inhibition of BLA valence ensembles and to determine how opposing motivational states are balanced within shared projection targets like the NAcSh.

The BLA^noci ensemble transmits allodynia information to the NAcSh

Our previous work using single-cell calcium imaging demonstrated that the BLA^noci ensemble remains stable throughout the development of chronic pain, yet its sensory tuning becomes broader over time. Specifically, in the context of SNI, this ensemble begins to respond to innocuous mechanical stimuli—such as light touch—that it does not encode in the uninjured state. This gain of function at the neural level correlates with the behavioral manifestation of mechanical allodynia. In the current study, we extended these findings by recording calcium activity from BLA^noci axon terminals in the mNAcSh. Following SNI, a light touch elicited increased terminal Ca²⁺ activity in the NAcSh, mirroring the expanded response profile of the parent ensemble in the BLA and paralleling the allodynia behavior observed in SNI mice. These results provide further support that this projection conveys a chronic pain–specific nociceptive signature to downstream regions. Coupled with the increased activation of NAcSh^noci neurons under chronic pain conditions, these findings suggest that the BLA^noci → NAcSh circuit may be particularly important for mediating the affective-motivational component of chronic pain. Future studies will be necessary to determine whether direct manipulation of this projection—and the downstream NAcSh^noci population—can attenuate allodynia and whether this circuit represents a viable target for chronic pain therapeutics.

BLA^noci to NAcSh neurons are essential for pain-related behavior

Calcium imaging of BLA^noci axons in the NAcSh revealed that this projection carries information about both acute and chronic pain states. To test whether this pathway is necessary for pain-related behavior, we used an intersectional INTRSECT-based chemogenetic approach to selectively inhibit the NAcSh-projecting subset of BLA^noci ensemble neurons. Silencing this pathway produced behavioral effects that closely mirrored those observed with optogenetic inhibition of the entire BLA^noci ensemble, indicating that this subpopulation is both necessary and sufficient to drive affective-motivational responses to acute and chronic pain.

The strongest behavioral effects were observed in response to noxious thermal stimuli, suggesting a preferential role for the BLA^noci → NAcSh projection in modulating thermal pain. In contrast, responses to mechanical stimuli were less affected, raising the possibility that other BLA^noci subpopulations—potentially projecting to distinct downstream targets—may preferentially mediate mechanical hypersensitivity. Future studies should dissect these parallel output channels and define their molecular and functional specializations.

We also observed variability in behavioral outcomes across some figures, particularly in uninjured cohorts and between sexes. For example, light-evoked analgesia in uninjured animals was more robust in some cohorts (e.g., Figs. 1N and 6D) than in others (e.g., Fig. 1O and 6E). These discrepancies likely reflect cohort-specific differences, subtle variations in behavioral testing protocols accumulated over the multiyear duration of the study, and technical variability in light delivery. While initial pooled analyses suggested no sex-specific effects, stratified reanalysis revealed that light-induced analgesia was more consistently observed in males. We also found evidence of behavioral sensitization with repeated testing: Control animals often showed increased affective-motivational responses upon reexposure to noxious stimuli. This sensitization was absent in animals receiving optogenetic inhibition of BLA^noci neurons or chemogenetic inhibition of their NAcSh-projecting subpopulation. This suggests that the BLA^noci → NAcSh circuit not only contributes to baseline pain affective-motivational behavior but also facilitates sensitization across repeated noxious exposures. Disruption of this pathway effectively blocks the development of these sensitization-like responses. Future studies with larger, sex-stratified cohorts will be essential to resolve individual variability and further elucidate the role of this ensemble in dynamic pain processing.

Last, although our chemogenetic manipulation was targeted to NAcSh-projecting BLA^noci neurons, tracing data revealed that these neurons also send axon collaterals to other brain regions, consistent with known BLA projection patterns. Therefore, we cannot fully exclude contributions from collateral pathways. This anatomical complexity reinforces broader conceptual insight: Rather than operating through simple one-to-one connections, BLA circuits likely function via distributed, ensemble-based outputs that influence multiple brain regions in parallel. Nonetheless, our study centers on defining the role of a genetically and behaviorally identified nociceptive ensemble within the BLA rather than isolating the contribution of any single downstream projection. By demonstrating that a projection-defined subpopulation of this ensemble is necessary for modulating affective-motivational pain behavior, we provide a framework for understanding how distributed BLA outputs collectively shape pain-related responses. Future studies using projection-specific disconnection or synapse-level silencing tools will be critical for dissecting the distinct contributions of individual downstream targets. Our findings complement and extend prior work on BLA → ACC and BLA → PFC circuits in chronic pain–related affective and sensory processing (78, 82, 83) while uniquely revealing the collateralization of BLA projections and establishing a direct link between ensemble-level neural identity and behavioral relevance. Rather than a limitation, we view this distributed projection architecture as a key insight into the integrative role of BLA circuitry in encoding and modulating the affective-motivational component of pain.

In all, we identify a valence-specific, nociceptive projection from the BLA to the mNAcSh that is selectively activated by noxious stimuli across sensory modalities and remains stable from acute to chronic pain states. This work challenges the prevailing view that BLA → NAcSh projections are primarily involved in appetitive valence signaling and instead positions this pathway as a critical substrate for encoding pain-related affective-motivational processing. We show that this projection is not only active during pain states but is also necessary for both acute and chronic pain–related behaviors—even when segregated from the broader BLA^noci ensemble. These findings highlight the BLA → NAcSh circuit as a potential target for novel, non-opioid analgesic interventions. Looking forward, a deeper understanding of the downstream connectivity and synaptic mechanisms engaged by the BLA^noci ensemble will be essential for developing circuit-specific strategies to treat the affective-motivational dimension of pain.

MATERIALS AND METHODS

Animals

All experimental procedures were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania and performed in accordance with the US National Institutes of Health guidelines. All studies were approved by the University of Pennsylvania Institutional Care and Use Committee (protocol no. 806675: Neural circuits and molecular mechanisms of pain and opioids). Two to five male and female mice aged 2 to 5 months were housed per cage and maintained on a 12-hour reverse light-dark cycle in a temperature- and humidity-controlled environment. All experiments were performed during the dark cycle. Mice had ad libitum food and water access throughout experiments. For behavioral, Ca²⁺ imaging, and anatomical experiments, we used Fos-FOS-2A-iCreER^T2 or “TRAP2” mice (Fos^{tm2.1(icre/ERT2)Luo)Luo}/the Jackson Laboratory, stock no. 030323) (48) bred to homozygosity [inhibitory optogenetics: n = 12 StGtACR2 uninjured (6 male), n = 11 StGtACR2 SNI (6 male), n = 10 tdTomato uninjured (5 male), n = 10 tdTomato SNI (5 male); inhibitory chemogenetics: n = 9 hM4 uninjured (5 male), n = 7 hM4 SNI (4 male), n = 10 OScarlet uninjured (6 male), and n = 10 OScarlet SNI (6 male); photometry: n = 13 uninjured (6 males), and n = 5 SNI (2 males)]. Additional anatomical experiments used TRAP2 mice crossed with Ai9 [B6.Cg-Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}/J] reporter mice that express a tdTomato fluorophore in a Cre-dependent manner (“TRAP2:tdTomato”) purchased from the Jackson Laboratory [stock no. 007909, bred to homozygosity for both genes; n = 4 nociFOS (2 male), n = 3 home-cageFOS (all female), n = 3 mateTRAP (all male), and n = 3 SNI light-touchFOS (2 male)]. RNA sequencing experiments used male TRAP2:tdTomato mice (n = 3 uninjured no stimulus, n = 3 uninjured light touch, n = 3 uninjured 55°C water droplet, n = 3 SNI no stimulus, and n = 3 SNI light touch). FISH experiments used either C57BL/6 mice (Drd1, Drd2, and Fos) (n = 4 males) or TRAP2:tdTomato mice (n = 5 males and n = 5 females). Additional Ca²⁺ imaging experiments used Rspo2-Cre mice from S. Tonegawa’s lab [n = 7 (4 males)] (7). Rspo2-Cre mice were bred to heterozygosity, with breeding cages consisting of one Rspo2-Cre+/− male mouse and one C57BL/6 female mouse.

Viral vectors

All viral vectors were aliquoted and stored at −80°C until use, diluted in 4°C sterile saline, and then stored at 4°C for a maximum of 4 days. For optogenetic inhibition experiments, we intracranially injected 400 nl of AAV1-hSyn1-SIO-stGACR2-FusionRed [Addgene 105677-AAV1; titer: 1.0 × 10¹² viral genomes (vg)/ml] or AAV1-CAG-FLEX-tdTomato (Addgene 28306-AAV1; titer: 1.0 × 10¹² vg/ml) into bilateral BLA at coordinates AP: −1.20 mm, mediolateral (ML): ±3.20 mm, and dorsoventral (DV): −5.20 mm. For anatomical tracing experiments, we intracranially injected 400 nl of AAV5-hSyn-DIO-EGFP (Addgene 50457-AAV5; titer: 1.3 × 10¹² vg/ml) into the right BLA at coordinates AP: −1.20 mm, ML: 3.20 mm, and DV: −5.20 mm. For Ca²⁺ imaging in fiber photometry experiments, we intracranially injected 400 nl of AAV5-hSyn1-FLEx-axon-GCaMP6s (Addgene 112010-AAV5, titer: 2.2 × 10¹² vg/ml) into the right BLA at coordinates AP: −1.20 mm, ML: 3.20 mm, and DV: −5.20 mm. For retrograde tracing of inputs to the NAcSh, we intracranially injected 300 nl of pOTTC1032–AAVrg-EF1a-Nuc-flox(mCherry)-EGFP (Addgene 112677-AAVrg; titer: 2.5 × 10¹³ vg/ml) into the right NAcSh at coordinates AP: 1.20 mm, ML: 0.60 mm, and DV: −4.55 mm. For INTRSECT chemogenetic inhibition experiments (75, 84), we intracranially injected 400 nl of AAV5-nEF-Con/Fon-hM4D(Gi)-mCherry (Deisseroth lab; titer: 7.4 × 10¹² vg/ml) or AAV8-hSyn-Con/Fon-OScarlet (Deisseroth lab; titer: 2.2 × 10¹² vg/ml) into bilateral BLA at coordinates AP: −1.55 mm, ML: ±3.2 or 3.4 mm, and DV: −4.75 or −5.2 mm and 300 nl of AAVretro-EF1a-FlpO (Addgene 55637-AAVrg; titer: 1.6 × 10¹³ vg/ml) into bilateral mNAcSh at coordinates AP: 1.2 mm, ML: ±0.6 mm, and DV: −4.55 mm.

Stereotaxic surgery

Adult mice (~8 weeks of age) were anesthetized with isoflurane gas in oxygen (initial dose, 5%; maintenance dose, 1.5%) and fitted into WPI or Kopf stereotaxic frames for all surgical procedures. Nanofil Hamilton syringes (10 μl; WPI) with 33G beveled needles were used to intracranially infuse AAVs into the BLA or NAcSh at a rate of 50 nl/min over 4 to 5 min. The following coordinates were used, on the basis of the Paxinos mouse brain atlas, to target these regions of interest (ROIs): BLA (from bregma, AP: −1.20 mm, ML: ±3.20 mm, and DV: −5.20 mm) or NAcSh (from bregma, AP: 1.20 mm, ML: ±0.60 mm, DV: −4.55 mm). Mice were given a 3- to 8-week recovery period to allow ample time for viral diffusion and transduction to occur. For fiber photometry studies, following viral injection into the right BLA, we placed a fiber optic implant (5.2-mm fiber, 400-μm diameter, Doric Lenses) ~0.2 mm above the DV coordinate of the injection site for the NAcSh. After setting the fiber optic in position, Metabond (Parkell) and Jet Set dental acrylic (Lang Dental) were applied to the skull of the mouse to rapidly and firmly fix the fiber optic in place. Briefly, after exposure of the skull, the bone was scored with a scalpel blade and a small skull screw (~1.7-mm diameter and 1.6-mm length) was placed to provide additional adhesion points for the Metabond reagent. The Metabond reagent was liberally applied over the skull and up and along the fiber optic cannula. Once dried, the Metabond was then covered with a layer of Jet Set acrylamide to create a reinforced headcap, as well as to cover the exposed skin of the incision site. Mice were then given a minimum of 3 weeks to recover and allow for optimal viral spread and transduction along axon terminals before beginning in vivo Ca²⁺ recordings. For all surgical procedures in mice, meloxicam (5 mg/kg) was administered subcutaneously at the start of the surgery, and a single 0.25-ml injection of sterile saline was provided upon completion. All mice were monitored for up to 3 days following surgical procedures to ensure the animals’ proper recovery and to provide additional daily subcutaneous meloxicam.

Chronic neuropathic pain model

As described previously (12), to induce a chronic pain state, we used a modified version of the SNI model of neuropathic pain (68). This model entails surgical section of two of the sciatic nerve branches (common peroneal and tibial branches) while sparing the third (sural branch). Following SNI, the receptive field of the lateral aspect of the hindpaw skin (innervated by the sural nerve) displays hypersensitivity to tactile and cool stimuli, eliciting pathological reflexive and affective-motivational behaviors (allodynia). To perform this peripheral nerve injury procedure, anesthesia was induced and maintained throughout surgery with isoflurane (4% induction and 1.5% maintenance in oxygen). The left hind leg was shaved and wiped clean with alcohol and Betadine. We made a 1-cm incision in the skin of the mid-dorsal thigh, approximately where the sciatic nerve trifurcates. The biceps femoris and semimembranosus muscles were gently separated from one another with blunt scissors, thereby creating a <1-cm opening between the muscle groups to expose the common peroneal, tibial, and sural branches of the sciatic nerve. Next, ~2 mm of both the common peroneal and tibial nerves was transected and removed without suturing and with care not to distend the sural nerve. The leg muscles are left uncultured, and the skin was closed with tissue adhesive (3M Vetbond), followed by Betadine application. During recovery from surgery, mice were placed under a heat lamp until they achieved normal balanced movement. Mice were then returned to their home cage and closely monitored over the following 3 days for well-being.

Targeted recombination in active population (TRAP)

nociTRAP

nociTRAP induction was performed similarly to the previously described procedure (12). We habituated mice to a testing room for two to three consecutive days. During these habituation days, no nociceptive stimuli were delivered and no baseline thresholds were measured (i.e., mice were naïve to pain experience before the TRAP procedure). We placed individual mice within red plastic cylinders (~9-cm diameter), with a red lid, on a raised perforated, flat metal platform (61 cm by 26 cm). The experimenter sat in the testing room for 30 min of habituation; this was done to mitigate potential alterations to the animal’s stress and endogenous antinociception levels because of the presence of the experimenter. To execute the TRAP procedure, we placed mice in their habituated cylinder for 30 min, and then a 55°C water droplet was applied to the central-lateral plantar pad of the left hindpaw once every 30 s over 10 min (20 applications total) using a 1-ml syringe. Following the water stimulations, the mice remained in the cylinder for an additional 60 min before injection of 4-OHT (40 mg/kg in vehicle; subcutaneously). After the injection, the mice remained in the cylinder for an additional 4 hours to match the temporal profile for c-FOS expression (fig. S4A), at which time the mice were returned to the home cage.

home-cageTRAP

home-cageTRAP induction was performed without habituation. At least 2 hours into the dark cycle, mice were gently removed from their home cages. Mice were then injected with 4-OHT (40 mg/kg in vehicle; subcutaneously) and immediately returned to their home cages.

mateTRAP

The process of mateTRAP began with habituation of mice to “home away from home” (HAFH) cages for 1 hour a day for four consecutive days beginning at least 2 hours into the dark cycle. During these habituation sessions, animals were isolated in HAFH cages and no stimuli were delivered. On the fifth day, mice were placed alone in HAFH cages for 1 hour and then were administered 4-OHT (40 mg/kg in vehicle; subcutaneously) before being gently placed back in HAFH cages. Immediately after injection, an age-matched, mating-receptive (estrus or proestrus phase of estrus cycle) female was introduced to the HAFH cage for 4 hours to match the temporal profile for 4-OHT. During the social interaction, a video recording was taken to allow for post hoc manual quantification of mounting behaviors. After that, mice were returned to their home cages.

Immunohistochemistry

Animals were anesthetized using FatalPlus (Vortech Pharmaceuticals) and transcardially perfused with 0.1 M phosphate-buffered saline (PBS), followed by 10% normal buffered formalin solution (NBF; Sigma-Aldrich, HT501128). Brains were quickly removed and postfixed in 10% NBF for 24 hours at 4°C and then cryoprotected in a 30% sucrose solution made in 0.1 M PBS until it sank to the bottom of their storage tube (~48 hours). Brains were then frozen in Tissue Tek optimal cutting temperature compound (Thermo Fisher Scientific) and coronally sectioned on a cryostat (CM3050S, Leica Biosystems) at 30 or 50 μm, and the sections were stored in 0.1 M PBS. Floating sections were permeabilized in a solution of 0.1 M PBS containing 0.3% Triton X-100 (PBS-T) for 30 min at room temperature and then blocked in a solution of 0.3% PBS-T and 5% normal donkey serum (NDS) for 2 hours before being incubated with primary antibodies [primary antibodies included the following: chicken anti-GFP (1:1000, Abcam, ab13970), guinea pig anti-FOS (1:1000, Synaptic Systems, 226308), rabbit anti-FOS (1:1000, Synaptic System, 226008), and rabbit anti-DsRed (1:1000, Takara Bio, 632496); prepared in a 0.3% PBS-T and 5% NDS solution for ~16 hours at room temperature]. Following washing three times for 10 min in PBS-T, they were incubated with secondary antibodies [secondary antibodies included the following: Alexa Fluor 647 donkey anti-rabbit (1:500, Thermo Fisher Scientific, A31573), Alexa Fluor 488 donkey anti-chicken (1:500, Jackson Immuno, 703-545-155), Alexa Fluor 555 donkey anti-rabbit (1:500, Thermo Fisher Scientific, A31572), and Alexa Fluor 647 donkey anti–guinea pig (1:500, Jackson Immuno, 706-605-148); prepared in a 0.3% PBS-T and 5% NDS solution applied for ~2 hours at room temperature], after which the sections were washed again three times for 5 mins in PBS-T and then again three times for 10 min in PBS-T and then counterstained in a solution of 0.1 M PBS containing 4′,6-diamidino-2-phenylindole (DAPI; 1:10,000, Sigma-Aldrich, D9542). Fully stained sections were mounted onto Superfrost Plus microscope slides (Thermo Fisher Scientific) and allowed to dry and adhere to the slides before being mounted with Fluoromount-G Mounting Medium (Invitrogen, 00-4958-02) and covered with coverslips.

Fluorescence in situ hybridization

Animals were anesthetized using isoflurane gas in oxygen, and the brains were quickly removed and fresh frozen in optimal cutting temperature compound using Super Friendly Freeze-It Spray (Thermo Fisher Scientific). Brains were stored at −80°C until they were cut on a cryostat to produce 16-μm coronal sections of the NAcSh. Sections were adhered to Superfrost Plus microscope slides and immediately refrozen before being stored at −80°C. Following the manufacturer’s protocol for fresh frozen tissue for the V2 RNAscope manual assay (Advanced Cell Diagnostics), slides were fixed for 15 min in ice-cold 10% NBF and then dehydrated in a sequence of ethanol serial dilutions (50, 70, and 100%). Slides were briefly air dried, and then a hydrophobic barrier was drawn around the tissue sections using a Pap Pen (Vector Labs). Slides were then incubated with hydrogen peroxide solution for 10 min, washed in distilled water, and then treated with the protease IV solution for 30 min at room temperature in a humidified chamber. Following protease treatment, C1 and C2 cDNA probe mixtures specific for mouse tissue were prepared at a dilution of 50:1 using the following probes from Advanced Cell Diagnostics: Fos (C1, 316921), Drd2 (C2, 406501-C2), Drd1 (C3, 406491-C3), Rspo2 (C1, 402001), Rerg (C2, 534251-C2), Blnk (C2, 534251-C2), Etv1 (C3, 557891-C3), Col5a2 (C3, 557891-C3), and tdTomato (C4, 3170410C4). Sections were incubated with cDNA probes (2 hours) and then underwent a series of signal amplification steps using FL v2 Amp 1 (30 min), FL v2 Amp 2 (30 min), and FL v2 Amp 3 (15 min). Two minutes of washing in 1× RNAscope wash buffer was performed between each step, and all incubation steps with probes and amplification reagents were performed using a HybEZ oven (ACD Bio) at 40°C. Sections then underwent fluorophore staining via treatment with a series of TSA Plus horseradish peroxidase (HRP) solutions, and Opal 520, 570, and 620 fluorescent dyes (1:5000, Akoya Biosystems, FP1487001KT, FP1495001KT) or TSA Vivid 520, 570, or 650 fluorescent dyes at a dilution of 1:3000. All HRP solutions (C1-C2) were applied for 15 min and Opal dyes for 30 min at 40°C, with an additional HRP blocker solution added between each iteration of this process (15 min at 40°C) and rinsing of sections between all steps with the wash buffer. Last, sections were stained for DAPI using the reagent provided by the Fluorescent Multiplex Kit. Following DAPI staining, sections were mounted and covered with coverslips using Fluoromount-G mounting medium and left to dry overnight in a dark, cool place. Sections from all mice were collected in pairs using one section for incubation with the cDNA probes and another for incubation with a probe for bacterial mRNA (dapB, ACD Bio, 310043) to serve as a negative control.

Imaging and quantification

All tissue was imaged on a Keyence BZ-X all-in-one fluorescence microscope at a 48-bit resolution using the following objectives: PlanApo λ ×4, PlanApo λ ×20, and PlanApo λ ×40. All image processing before quantification was performed with Keyence BZ-X analyzer software (version 1.4.0.1). Quantification of neurons expressing fluorophores was performed via manual counting of TIFF images in Photoshop (Adobe, 2021) using the Counter function or using HALO software (Indica Labs), which is a validated tool for automatic quantification of fluorescently labeled neurons in brain tissue (85–87). Counts were made using 20× magnified z-stack images of designated ROIs. For axon density quantification, immunohistochemistry was performed to amplify the GFP signal and visualize ipsilateral BLA^noci axons throughout the brain in 50-μm tissue free floating slices, as described above. To avoid quantification of contralateral axon projections, viral injections to label BLA axons were unilateral. Areas with dense axon innervation were identified using 4× imaging. Areas implicated in emotion and nociception were selected for additional 20× imaging with z-stacks. These ROIs were initially visualized at 20× to determine the region with the highest fluorescence. The exposures for fluorescein isothiocyanate and cyanine 3 were adjusted to avoid overexposed pixels for the brightest area. This exposure was kept consistent for all slices for an individual mouse. For an individual ROI, one slice per mouse was included. ROIs were drawn in ImageJ, and the threshold intensity was measured. The maximum intensity was set to a value of 1.0, and all ROIs are reported as intensity compared to the maximum.

Single-nucleus RNA sequencing

Nucleus preparation

A single punch of the right side of the whole amygdala measuring 2 mm in width and 1 mm in depth was used to prepare the nucleus suspensions. Slices were taken at coordinates AP: −0.83 to −1.83 mm from bregma. Nucleus isolation was performed using the Minute single-nucleus isolation kit designed for neuronal tissue/cells (catalog no. BN-020, Invent Biotechnologies). Briefly, the tissue was homogenized using a pestle in a 1.5-ml LoBind Eppendorf tube. Subsequently, the cells were resuspended in 700 μl of cold lysis buffer and ribonuclease inhibitor and incubated on ice for 5 min. The homogenate was then transferred to a filter within a collection tube and incubated at −20°C for 8 min. Following this, the tubes were centrifuged at 13,000g for 30 s, the filter was discarded, and the samples were centrifuged at 600g for 5 min. The resulting pellet underwent one wash with 200 μl of PBS and 5% bovine serum albumin and then resuspended in 60 μl of PBS and 1% bovine serum albumin. The concentration of nuclei in the final suspension was assessed by staining with trypan blue and counting using a hemacytometer. The suspension was diluted to an optimal concentration of 500 to 1000 nuclei/μl.

Single-nuclei gene expression assay

The single-nuclei gene expression assay (snRNA-seq) was conducted following the instructions provided by 10x Genomics. A total of 20,000 nuclei was loaded into the 10x Genomics microfluidics Chromium controller, with the aim of recovering ~10,000 to 12,000 nuclei per sample. Subsequently, sequencing libraries were constructed following the manufacturer’s protocol for the Chromium Next GEM Single Cell 3′ version 3.1 kit. Libraries containing individual and unique indexes were pooled together at equimolar concentrations of 1.75 nM and sequenced on the Illumina NovaSeq 6000 using 28 cycles for Read 1, 10 cycles for the i7 index, 10 cycles for the i5 index, and 90 cycles for Read 2.

Data analysis

Preprocessing of scNuc data

Paired-end sequencing reads were processed using 10x Genomics Cellranger version 5.0.1. Reads were aligned to the mm10 genome optimized for single-cell sequencing through a hybrid intronic read recovery approach (88). In addition, the genome was modified to include a custom transgene incorporating tdTomato and the full-length 3′ untranslated region (www.addgene.org/104105/). In short, reads with valid barcodes were trimmed by the template switch oligonucleotide sequence and aligned using STAR version 2.7.1 with MAPQ adjustment. Intronic reads were removed, and high-confidence mapped reads were filtered for multimapping and unique molecular identifier (UMI) correction. Empty GEMs (Gel Bead-in-Emulsions) were also removed as part of the pipeline. Initial dimensionality reduction and clustering were performed before processing to enable batch correction and the removal of cell-free mRNA using SoupX (89). Raw expression matrices with counted, individual nucleus UMIs and genes were used for subsequent steps and filtering by quality control metrics.

Clustering and merging by condition and comparison

Raw matrices for each individual replicate per condition were converted to Seurat objects using Seurat 5.0.1 and filtered to remove UMIs with thresholds of >200 minimum features, <5% mitochondrial reads, and <5% ribosomal reads. Replicates were merged to generate objects per condition for the subsequent steps. Each dataset was normalized (NormalizeData) using the default scale factor of 10,000, and variable selection (FindVariableFeatures) was performed using 2000 features and then scaled and centered (ScaleData) using all features without regressing any variables. Dimensionality reduction with principal components analysis (RunPCA) used the first 30 principal components, and the nearest-neighbor graph construction (FindNeighbors) used the first 10 dimensions. Clustering (FindClusters) was next performed using a resolution of 0.4 before layers corresponding to each replicate were integrated (IntegrateLayers) using CCAIntegration with a k weight of 60 and then rejoined (JoinLayers). The dataset per condition was then dimensionally reduced using the integrated canonical correlation analysis (CCA) at with 30 dimensions (RunUMAP) and the same resolution of 0.4. ScType (90) was used for automated, de novo cell type identification of the clusters followed by manual curation for clusters with low confidence scores. For all comparisons, the objects per condition were merged and processed using the same integration methodology above to scale and normalize between all incorporated samples. Non-neuronal cell types were excluded from all downstream analysis steps after initial classification. Nuclei expressing the top 75th percentile of Slc17a7 (VGLUT1) expression were marked as the BLA. The CeA was marked by the top 25th percentile of either Gad1 or Gad2 expression and no detected Slc17a7 or Slc17a6 (VGLUT2) transcripts. Each independent dataset was then subclustered with a resolution of 0.4. BLA- and CeA-separated objects were then merged, normalized, and dimensionally reduced to generate the BLA-CeA dataset. Wilcoxon rank sum DEGs were identified using FindMarkers with Bonferroni correction for multiple testing with a minimum percent of 0.25 and a fold change threshold of 1.2. Pseudobulk DEGs were identified using DElegate (https://github.com/cancerbits/DElegate?tab=readme-ov-file), a wrapper for EdgeR on single-nucleus data. Counts were aggregated by individual mice per condition under the orig.ident identity, and then pairwise comparisons were computed using quasi-likelihood dispersion with glmQLFit through the findDE function. Genes with an adjusted P value of <0.01 were considered differentially expressed for both methodologies.

Modular activity scoring and subsetting

The BLA is well known for its role in associative/contextual learning related to aversive events, which can alter gene transcription. To avoid inducing “fear-related recall” transcriptomic cascades that could obscure our goal of identifying nociceptive IEGs, we minimized repeated exposure to similar environments, including the testing room, von Frey apparatus, human experimenter, and noxious stimuli. On the day of noxious stimulation and tissue collection, the experimental room’s physical setup (a small, sound-controlled isolation chamber with ~10 lux red light) did not permit video recordings. Therefore, behavior was not recorded, but hypersensitivity responses were observed and documented by the experimenter. We prioritized precise stimulus timing and accurate delivery to the hindpaw, ensuring consistency in our protocol (one stimulus application every 30 s over 5 min), as the study was not powered to stratify mice by behavioral responses.

Modular activity scores were calculated for glutamatergic neuron and BLA datasets using AddModuleScore with the list of the 25 putative IEGs (Arc, Bdnf, Cdkn1a, Dnajb5, Egr1, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1, Junb, Nefm, Npas4, Nptx2, Nr4a1, Nr4a2, Nr4a3, Nrn1, Ntrk2, Rheb, Scg2, Sgsm1, Syt4, and Vgf) against a control feature score of 5. Nuclei in the top 90th activity score percentile were isolated as IEG+, while those under the 50th percentile threshold were IEG−. The merged object was aggregated and dimensionally reduced with a resolution of 0.2. FindMarkers was used to perform all IEG+/IEG− pairwise comparisons.

Gene ontology analysis

Gene ontology analysis was performed via Synaptic Gene Ontology (SynGO) (91) on genes that were up-regulated under the SNI + no stimulus or SNI + light touch condition compared to the other three conditions (uninjured + no stimulus, uninjured + light touch, and the other SNI conditions) as displayed in Fig. 3D using a P value cutoff of 0.05. The list of DEGs was assessed for cellular components and biological processes.

Behavioral tests

On test days, mice were brought into procedure rooms ~1 hour before the start of any experiment to allow for acclimatization to the environment. Mice were provided food and water ad libitum during this period. For multiday testing conducted in the same procedure rooms, animals were transferred into individual HAFH cages ~1 hour before the start of testing and were only returned to their home cages at the end of the test day. All testing and acclimatization were conducted under red light conditions (<10 lux). Equipment used during testing was cleaned with a 70% ethanol solution before starting and, in between, each behavioral trial to mask odors and other scents. Behavioral tests were conducted by experimenters blinded to the condition.

Sensory testing for pain affective-motivational and nociceptive reflex behavioral assays

To evaluate mechanical reflexive sensitivity, we used a logarithmically increasing set of eight von Frey filaments (Stoelting), ranging in gram force from 0.07 to 6.0 g. These filaments were applied perpendicular to the plantar hindpaw with sufficient force to cause a slight bending of the filament. A positive response was characterized as a rapid withdrawal of the paw away from the stimulus within 4 s. Using the up-down statistical method, the 50% withdrawal mechanical threshold scores were calculated for each mouse and then averaged across the experimental groups (12).

To evaluate affective-motivational responses evoked by mechanical stimulation, we used a sharp 25G syringe needle (pin prick) (12). The pin prick was applied as a subsecond poke to the hindpaw, and the duration of attending behavior was collected for up to 30 s after the stimulation.

To evaluate affective-motivational responses evoked by thermal stimulation (12), we applied either a single, unilateral 55°C drop of water or acetone (evaporative cooling) to the left hindpaw, and the duration of attending behavior was collected for up to 30 s after the stimulation. Only one drop stimulation was applied on a given testing session. For the tests conducted for chemogenetic activation studies, mice were administered clozapine-N-oxide (CNO) 30 min before the start of behavior testing to allow for complete absorption of the drug (84). For tests conducted for chemogenetic inhibition studies, mice were administered deschloroclozapine (DCZ) 30 min before the start of behavior testing (92).

Escape/attending behavior was quantified either during the behavioral test or post hoc via videos recorded during testing. The following behaviors constituted “escape” behaviors: rearing, jumping, and climbing up the walls of the behavior chamber. The following behaviors constituted “attending” behaviors: licking the hindpaw that was stimulated, directing attention to the stimulated hindpaw, guarding the hindpaw, flicking the hindpaw, and shaking the hindpaw.

Conditioned place avoidance

Fifteen-minute pre- and postconditioning test sessions and three twice-daily 30-min conditioning sessions consisting of alternating side pairings of CNO [3.0 mg/kg, intraperitoneally (ip)] and saline (0.9% saline, ip) were performed to determine conditioned place preference (93). During the pretest session, mice were placed in a two-chambered place preference chamber (Med Associates Inc.) inside a sound-attenuated chamber (Med Associates Inc.) and allowed to explore both sides for 15 min (900 s). The amount of time spent on each side was recorded, and data were used to assign the animals to be conditioned in the nonpreferred chamber. Mice were conditioned twice daily for 3 days following the pretest, with the saline-paired box conditioning sessions in the a.m. and the CNO-paired conditioning sessions in the p.m., separated by 4 hours. Each group received CNO (3 mg/kg) on one side and saline (0.9% sodium chloride) on the other side 30 min before the conditioning session. Locomotor activities for each conditioning session were measured using two consecutive beam breaks. One day after the last conditioning session, animals were allowed to explore freely between the two sides for 15 min (900 s), and the time spent on each side was recorded. The preference score (time spent on the drug-paired side minus time in the saline-paired side on the postconditioning day minus the preconditioning day) was calculated for each mouse.

In vivo fiber photometry calcium recordings and data analysis

Recordings of the BLA^noci or Rspo2 BLA axons were chosen to be done unilaterally to capture the unilateral projections from the BLA to the NAc and as is standard in the field when recording axon calcium activity. Optical recordings of axon-GCaMP6s fluorescence were acquired using an RZ10x fiber photometry detection system (Tucker-Davis Technologies) consisting of a processor with Synapse software (Tucker-Davis Technologies) and optical components (Doric Lenses and ThorLabs). Excitation wavelengths generated by light-emitting diodes (LEDs; 460-nm blue light and 405-nm violet light) were relayed through a filtered fluorescence minicube at spectral bandwidths of 460 to 495 nm and 405 nm to a prebleached, low-autofluorescence, mono fiber optic patch cord connected to the implant on top of each animal’s head. The power output for the primary 460-nm channel at the tip of the fiber optic cable was adjusted to ~100 μW. The signal in both 460- and 405-nm channels was monitored continuously throughout all recordings, with the 405-nm signal used as an isosbestic control for both ambient fluorescence and motion artifacts introduced by the movement of components in the light path. Wavelengths were modulated at frequencies of 210 to 220 Hz and 330 Hz, respectively. All signals were acquired at 1 kHz and low-pass filtered at 4 Hz. On testing days, mice were connected to the photometry system, and following a 10-min habituation period, recording sessions began. Each stimulus type was applied five times, with an intertrial interval (ITI) of 90 s. Behavior was not quantified, but experimenters verified that the stimulus touched the animal’s foot each time and that the animal noticed the stimuli (and showed appropriate affective-motivational pain behaviors to noxious stimuli). Between different stimulus types (0.07-g von Frey filament, acetone droplet, 25G needle pick prick, and 55°C water droplet), the LED was turned off for 5 min to minimize photobleaching. During testing, behavior was qualitatively examined, but no behavior quantification was performed, as the researcher was focused on the timing and accurate placement of the stimulus. To assess appetitive-related neural activities, sucrose (10%, w/v) was presented via a calibrated lickometer spout positioned within reach of the mouse’s snout while it was connected to the photometry rig. Recordings were synchronized with the onset of sucrose availability to capture calcium activity during active licking and ingestion. Following testing, all mice were perfused, and the tissue was assessed for proper viral targeting and transduction efficacy, as well as optic fiber placement via immunohistochemistry (see the above section for details).

Analysis of the GCaMP signal was performed with the use of the open-source, fiber photometry analysis MATLAB software suite pMAT (94). Using pMAT, bulk fluorescent signals from both the 460- and 405-nm channels were normalized to compare differences in calcium-mediated event metrics for the total duration of a recording (frequency) and at select events using peri-event time histogram analyses locked to specific behaviors designated by the application of an external transistor-transistor logic input. The amplitude and area under the curve (AUC) were determined from peri-event time histogram analyses across groups. The MATLAB polyfit function was used to correct for the bleaching of signal for the duration of each recording using the slope of the 405-nm signal fitted against the 460-nm signal. Detection of GCaMP-mediated fluorescence is presented as a change in the 460-nm/fitted 405-nm signal over the fitted 405-nm signal (ΔF/F). From ΔF/F values, robust z-scores were calculated for analysis (94).

In vivo optogenetic inhibition

Optogenetic inhibition was chosen to be bilateral because of the observed presence of BLA^noci axons in the contralateral hemisphere. At the time of bilateral virus injection, mice were implanted with bilateral optic fibers (200-μm diameter, Prizmatix) above the BLA. The mice were left to recover for 2 weeks before nociTRAP and 3 weeks after nociTRAP to ensure ample virus expression. Optogenetic stimulation was delivered with a blue LED (455 nm) via a bifurcated fiber optic cable. The power output at the tip was measured to be ~5 mW per side. For sensory testing, mice were placed on the elevated wire-floor von Frey rack in small plexiglass containers for 30 min. Sensory testing was conducted before LED stimulation, during a 30-s stimulation, and after stimulation with a 90-s ITI.

In vivo chemogenetic activation and inhibition

Chemogenetic manipulations were chosen to be bilateral because of the observed presence of BLA^noci axons in the contralateral hemisphere. Following bilateral virus injections, mice recovered for 2 weeks before nociTRAP and 5 to 8 weeks after nociTRAP to ensure ample virus expression. Behavioral testing was conducted at the baseline (when reported) and 30 min following intraperitoneal injection of CNO (3.0 mg/kg body weight) or DCZ (0.3 mg/kg body weight).

Drugs

4-OHT (Hello Bio, catalog no. HB6040) was dissolved in 100% ethanol (for a 10-mg bottle, 250 μl of ethanol) on the morning of use. The solution was heated at 50°C for up to 30 min to encourage dissolving. The 4-OHT solution was further diluted in Kolliphor EL (Sigma-Aldrich, catalog no. C5135-500G; 500 μl for a 10-mg bottle) and finally 1× PBS (Calbiochem, catalog no. 524650; 1.75 ml for a 10-mg bottle) when it was delivered subcutaneously at a dose of 40 mg/kg body weight. CNO (water soluble; Hello Bio, HB6149) was diluted in sterile 0.9% saline and delivered intraperitoneally at a dose of 3.0 mg/kg body weight. DCZ (water soluble) was diluted in sterile 0.9% saline and delivered intraperitoneally at a dose of 0.3 mg/kg body weight. To note, chemogenetic manipulations were performed using either CNO or DCZ, depending on the timing of the experiment. hM3Dq activation studies conducted in 2019 used CNO (3 mg/kg, ip), which was the standard agonist at that time. Subsequent experiments, including all hM4Di inhibition studies using the INTRSECT approach, used DCZ (0.3 mg/kg, ip) on the basis of its superior pharmacokinetic properties, higher receptor affinity, and reduced off-target effects (87, 92). The transition to DCZ in 2022 reflects both emerging best practices in the field and validation from internal pilot studies confirming its efficacy and selectivity without motoric side effects.

Inclusion criteria

Following all experiments, brains were collected from all mice to visualize viral expression and fiber placements when applicable. Final inclusion in the data analysis required that viral injections were considered “on target”—for the BLA studies, this required expression in the BLA and not the CeA and with minimal expression elsewhere (such as the piriform cortex). In addition, when manipulations were performed bilaterally, mice with unilateral expression were excluded from analysis (12 stGtACR2 TRAP2 mice excluded, 3 hM3 TRAP2 mice excluded, and 11 hM4 TRAP2 mice excluded). Fibers for optogenetic and photometry studies were confirmed to be above the ROI (BLA or NAc) and above fluorescently labeled cells or axons. The absence of either of these resulted in exclusion from the study (five Rspo2-Cre mice excluded and zero TRAP2 mice excluded). If mice lost a headcap during the experiment, they were euthanized and excluded from analyses. Data points are missing from particular behavior testing experiments when errors in the experimental procedure or failure to record a value occurred (three stGtACR2/tdTomato mice excluded, zero hM3/mCherry mice excluded, and three hM4/OScarlet mice excluded). One male mouse data point is missing from the FISH experiment visualizing tdTomato and Etv1 mRNAs in the BLA in fig. S6, as the experimental procedure destroyed the tissue and visualization was not possible in the BLA. To reduce our animal usage and to account for reported sex differences in pain in humans, male and female mice were used throughout all experiments. We aimed for an equal distribution of males and females, with some variability because of natural variation of sex in the litters produced by our breeding colonies.

Statistics and data presentation

Sample sizes for each experiment were informed by effect sizes reported in the literature, pilot data, and feasibility constraints associated with the specific technique, including behavior, fiber photometry, histology, circuit mapping, and snRNA-seq. Statistical approaches were selected on the basis of the experimental design and are indicated in each figure legend. Representative histological images were selected to illustrate key anatomical features, including viral spread and fiber placements. Image adjustments, where applied, were limited to uniform linear changes in brightness and contrast. Sex as a biological variable was considered in all behavioral experiments. When both male and female mice were used, we performed statistical testing for sex differences using two- or three-way analyses of variance (ANOVAs). When no significant effect of sex or sex interaction was observed, data were pooled for presentation. All experiments included both sexes, except for the snRNA-seq experiments, which used male tissue because of financial limitations; key findings from these experiments were validated using in situ hybridization on tissue from both sexes. Data are reported as the means ± SEM. Statistical analyses were conducted using GraphPad Prism versions 9 and 10. Tests used include unpaired and paired t tests, two-way ANOVAs, and three-way ANOVAs, as appropriate. When ANOVAs were followed by multiple comparisons, Bonferroni corrections were applied. Statistical significance was defined as P < 0.05. All statistically significant comparisons in figures are denoted with an asterisk (*), and corresponding exact P values are reported in table S4. Figures were generated in Prism and finalized in Adobe Illustrator.

Supplementary Materials

The PDF file includes:

Figs. S1 to S14

Legends for tables S1 to S3

Table S4

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S3