A Central Amygdala Input to the Parafascicular Nucleus Controls Comorbid Pain in Depression.

While comorbid pain in depression (CP) occurs at a high rate worldwide, the neural connections underlying the core symptoms of CP have yet to be elucidated. Here, we define a pathway whereby GABAergic neurons from the central nucleus of the amygdala (GABACeA) project to glutamatergic neurons in the parafascicular nucleus (GluPF). These GluPF neurons relay directly to neurons in the second somatosensory cortex (S2), a well-known area involved in pain signal processing. Enhanced inhibition of the GABACeA→GluPF→S2 pathway is found in mice exhibiting CP symptoms. Reversing this pathway using chemogenetic or optogenetic approaches alleviates CP symptoms. Together, the current study demonstrates the putative importance of the GABACeA→GluPF→S2 pathway in controlling at least some aspects of CP.


INTRODUCTION
Recently, chronic pain has emerged as the leading cause of years lived with disability, while major depressive disorder is the fifth (GBD 2016 Disease and Injury Incidence and Prevalence Collaborators, 2017). Pain symptoms occur with an estimated prevalence of up to 65% in patients with major depression, which is much higher than that observed in nondepressed cohorts (Bair et al., 2003;Robinson et al., 2009;Rudy et al., 1988). Furthermore, comorbid pain in depression (CP) symptoms can worsen depressed mood and are associated with poor responses to both depression and pain treatment (Bushnell et al., 2013;Tracey and Mantyh, 2007), creating a cycle of depression and pain that is often difficult to break (Chapman and Gavrin, 1999). The proper treatment for CP represents a major challenge in the field, as the underlying mechanisms of CP are poorly understood.
Chronic pain varies in etiology, with different conditions exhibiting distinct profiles of pain symptomatology (Basbaum et al., 2009;Todd, 2010;Woolf and Ma, 2007). Alterations in serotonergic (Morita et al., 2015) and noradrenergic (Hermans et al., 2011) systems are considered the common pathological roots of depression and chronic pain, and they share a clinical pattern of persistence beyond the precipitant. However, approximately one-third of depression patients do not respond well to treatment by targeting these systems (Hieronymus et al., 2016). Notably, the acute use of serotonin-selective serotonin reuptake inhibitors (SSRIs) produces early adverse events in some patients, which may exacerbate pain symptoms (Hieronymus et al., 2016). Together, the current evidence indicates that the neuroanatomical and molecular substrates that underlie CP may be different from those that underlie other various somatic pain syndromes (Bair et al., 2003), which could be beyond the serotonergic and noradrenergic systems. Many brain regions, such as the amygdala (Zhou et al., 2019), anterior cingulate cortex (ACC) (Barthas et al., 2015), prefrontal cortex (PFC) (Moda-Sava et al., 2019), insular cortex (IC) (Gehrlach et al., 2019), and the core of the nucleus accumbens (NAc) (Descalzi et al., 2017), are involved in the central mechanisms of depression. However, little is known about whether, or how, depression may influence the brain neural system that would modify the central mechanisms of chronic pain. In addition, the brain regions involved in chronic pain closely mirror those involved in depression regulation, such as the amygdala, ACC (Bliss et al., 2016), IC (Zhuo, 2008), and PFC . Therefore, chronic pain and depressive symptoms appear to be associated. However, the mechanism, especially with regard to how the two are linked on the level of neuroanatomical and molecular substrates, is unknown. For example, it is unclear whether CP results from maladaptive changes in depression-related brain circuits.
The amygdaloid complex is well known to be relevant to fear learning, anxiety, reward, and pain (Duvarci and Pare, 2014;Li et al., 2013;Tovote et al., 2015;Tye et al., 2011;Zhuo, 2008). Even the PFC  and IC (Berret et al., 2019) have been reported to interact with the amygdala in processing cognition, emotion, and pain. This raises the possibility that the amygdala could be an important site for processing the depressive symptoms associated with pain (Senn et al., 2014). The central nucleus of the amygdala (CeA), referred to as the ''nociceptive amygdala,'' serves as the main output nucleus for amygdala functions (Janak and Tye, 2015). Highly processed, polymodal information reaches the CeA indirectly from the thalamus and cortical areas or direct nociceptive inputs from the spinal cord and brainstem. Imaging studies have shown significant alterations in the activity of the CeA in chronic pain. Furthermore, the CeA forms widespread connections with forebrain areas (Penzo et al., 2015) and the brainstem (Tovote et al., 2016), which has been implicated in mediating fear and mood disorders. These studies suggest that the CeA is a probable convergent point of chronic pain and depression. However, the cause-and-effect relationship between the adaptation of CeA circuits and the pathology of CP is unknown.
The thalamus is a key hub that regulates pain signal transmission (Vartiainen et al., 2016) and processing of depressive emotions (Greicius et al., 2007). It shares anatomical connections with other brain areas, such as the medial PFC (Ährlund-Richter et al., 2019), superior colliculus (SC), brainstem, basal ganglia (Assous et al., 2017;Bieler et al., 2018), andamygdala (LeDoux et al., 1990). The precise organization and the function of the CeA-PF circuit remains unknown. Based on evidence linking the amygdala with both pain and depression, the current study sought to examine the pathological causes of the CP by defining the neurocircuitry, architectural rules, and plasticity of the CeA.
Combining viral tracing, optogenetic and electrophysiological methods, we dissected the functional organization of the CeA→thalamus pathway and explored this adaptation of this pathway in CP mice. Collectively, we describe an inhibition GABA CeA →Glu PF →S2 pathway, the alteration of which is both necessary and sufficient for manifesting CP.

An Inhibitory GABA CeA →Glu PF Pathway
First, the H129-G4 virus, used as an anterograde polysynaptic tracer (Zeng et al., 2017), was stereotaxically delivered into the CeA of a wild-type mouse (Figures 1A and 1B,left). Thirty-six hours later, the GFP + signal was observed in many regions, such as S2 (Bushnell et al., 2013;Navratilova and Porreca, 2014), zona incerta (ZI) (Masri et al., 2009), and ventrolateral periaqueductal gray (vlPAG) (Tovote et al., 2016), which are involved in the pathophysiology of pain ( Figure S1). Interestingly, GFP-labeled neurons were identified only in the PF of the thalamus ( Figure 1B, right), which were predominantly co-localized with the glutamate antibody ( Figure 1C). These results suggest that Glu PF neurons may receive CeA projections.
This finding is interesting because the thalamus is a well-known ''pain nucleus'' in mammals (Tracey and Mantyh, 2007). To characterize the CeA-Glu PF organization, a celltype-specific retrograde trans-monosynaptic tracing system was employed. Cre-dependent helper viruses (AAV-EF1α-DIO-TVA-GFP and AAV-EF1α-DIO-RVG) were injected into the PF of Ca 2+ /calmodulin-dependent protein kinase II (CaMKII, an enzyme in glutamatergic neurons)-Cre mice ( Figure 1D). After 3 weeks, the rabies virus (RV) (EnvApseudotyped RV-∆G-DsRed) was injected into the same site ( Figure 1E, left). The presence of these helper viruses facilitated monosynaptic retrograde RV spread (Wickersham et al., 2007). In addition to multiple inputs to the PF, such as the PFC, bed nucleus of the stria terminalis (BNST), and lateral hypothalamus (LH; Figure S2), we also identified intensely DsRed-labeled neurons in the CeA ( Figure 1E, right). The DsRed signal was co-localized with the GABA antibody ( Figure 1F), which is consistent with previous research suggesting that the CeA consists of 95% GABA neurons (Janak and Tye, 2015;Martin-Fernandez et al., 2017). By contrast, the same tracing strategy conducted in glutamic acid decarboxylase 2 To examine the functional connections of the GABA CeA →Glu PF pathway, an adenoassociated virus expressing Cre-dependent channelrhodopsin-2 (AAV-DIO-ChR2-mCherry) was infused into the CeA of GAD2-Cre mice ( Figure 2A). We observed mCherry + (GABA) cell bodies in the CeA (Figures 2B and 2C) and numerous mCherry + fibers in the PF ( Figure  2D) and other regions, such as ZI and vlPAG but not S2 ( Figure S3). By whole-cell recording in brain slices, brief light stimulation of ChR2-containing GABA CeA terminals in the PF reliably elicited inhibitory postsynaptic currents (IPSCs,121.75 ± 13.26 pA) in Glu PF neurons, which were eliminated by the GABA receptor antagonist bicuculline (12.98 ± 10.05 pA; Figure 2E). Post hoc immunostaining showed that the recorded neurons labeled with neuronbiotin-488 in pipette solution were glutamatergic ( Figure 2F). In addition, in vivo electrophysiology recordings were performed in the PF of freely moving mice. AAV-DIO-ChR2-mCherry viruses were injected into the CeA in GAD2-Cre mice. After 3 weeks, multi-tetrode recording and optic fibers were implanted into the PF ( Figure 2G). Multiple channel recordings of the firing rate of the PF before, during, and after light photostimulation ( Figures 2H and 2I) were conducted. We found that blue light decreased the firing rate in PF neurons. To confirm the CeA-PF Glu projection, we injected anterograde monosynaptic AAV2/1-Cre virus into the CeA of Ai14 (Cre-dependent tdTomato reporter) mice (Zingg et al., 2017) (Figures 2J and 2K). After 3 weeks, we observed tdTomatoexpressing neurons in the PF ( Figure 2L), which were co-localized with the glutamate antibody ( Figure 2M). These data confirm that GABA CeA neurons send afferents to the PF that synapse on Glu PF neurons exerting inhibition.

An Enhanced Inhibition of the GABA CeA →Glu PF Pathway in CP
To determine the role of the GABA CeA →Glu PF pathway in CP, we used the chronic restraint stress (CRS)-induced depression model in mice (Nestler and Hyman, 2010) ( Figure 3A). After 3 weeks of CRS (CRS 3W), the mice displayed multiple depressive-like behaviors in routine assays: forced-swim test (i.e., increased immobility), tail suspension test (i.e., increased immobility), open-field test (i.e., reduced time and entries into the center zone), and elevated plus maze test (i.e., reduced open arms time and entries; Figure 3B; Figures  S4A and S4B). The total travel distance was unchanged, suggesting that locomotor ability was not affected by CRS 3W ( Figure S4C). Of note, after 1 and 2 weeks of CRS (CRS 1W and CRS 2W, respectively) mice did not display pain sensitization or depressive-like behaviors ( Figures S4D-S4H). However, at 3 weeks, CRS mice displayed significant pain sensitization ( Figure 3C). Furthermore, traditional analgesic agents, which were effective in Complete Freund's adjuvant (CFA)-induced inflammatory ( Figures 3D, 3E, and 3H) and spared nerve injury (SNI)-induced neuropathic pain ( Figures 3F, 3G, and 3I), had no effect on CRS-induced pain sensitization ( Figure 3J).
Next, we investigated the activity of PF-projecting GABA CeA neurons in CRS 3W mice. To visualize GABA neurons, GAD2-Cre mice were used to cross with Ai14 mice to produce transgenic mice with red tdTomato-expressing GABA neurons (GAD2-tdT; Figure S5A). Whole-cell recordings were performed in acute brain slices on PF-projecting GABA CeA neurons by PF infusion of the retrograde tracer CTB-488 ( Figures 4A and 4B). We found an increase in the spike number ( Figure 4C) and a decrease in rheobase ( Figure 4D Based on this result, we subsequently investigated whether GABA levels were increased in the PF. Using in vivo microdialysis and high-performance liquid chromatography (HPLC) measurements by implanting a microdialysis probe into the PF (Cheng et al., 2009) ( Figure  4E), we found that the concentration of GABA in the PF was increased in the CRS 3W mice (18.94 ± ng) but not CRS 2W mice (16.75 ± 1.67 ng) relative to control mice (14.44 ± 0.82 ng; Figure 4F; Figure S6A). These results suggest enhanced PF-projecting GABA CeA neuronal activity in CP states.

Necessary and Sufficient Role of the GABA CeA →Glu PF Pathway for CP
Given the increased GABA CeA →Glu PF inhibitory input in the presence of CRS, we infused Cre-dependent expression of chemogenetic inhibitory hM4Di in the CeA and intraperitoneally injected its ligand clozapine-N-oxide (CNO), to selectively inhibit GABA CeA neurons of GAD2-Cre mice ( Figure 4G; Figures S6E-S6G). We found that the CRS-induced pain sensitization ( Figure 4H) and depressive-like behaviors ( Figure S6I) were alleviated and that the GABA concentration in the PF decreased (mCherry, 21.57 ± 2.18 ng; hM4Di, 4.37 ± 1.68 ng) after chemogenetic inhibition of GABA CeA neurons (Figures 4I and 4J; Figure S6J). Chemogenetic inhibition of GABA CeA neurons did not affect distance traveled in the open-field test ( Figure S6H).
In addition, reduced pain sensitization effects ( Figure 5E) but not locomotion ( Figure 5D) or depressive-like behaviors (Figures S7A-S7D) were obtained by injecting Cre-dependent AAV carrying eNpHR (AAV-DIO-eNpHR3.0-EYFP) into the CeA and optical inhibition of eNpHR-containing GABA CeA terminals in the PF of GAD2-Cre mice ( Figures 5A-5C). It should be noted that the optogenetic inhibition had no effect on CFA-induced inflammatory and SNI-induced pain sensitization ( Figure 5F). In naive mice, optical activation of GABA CeA terminals in the PF through CeA injection of Cre-dependent AAV-DIO-ChR2-mCherry in GAD2-Cre mice caused the animals to display pain sensitization and did not affect distance traveled ( Figures S7E-S7G). In addition, brief (5 min) light stimulation of ChR2-containing GABA CeA terminals in the PF ( Figure S7H) elicited a significant decrease in the spike number of the Glu PF neurons, and this effect could last 15 min after light withdrawal (Figures S7I and S7K). Meanwhile, the spike number of the Glu PF neurons did not show a change after GAD2-Cre mice with CeA infusion of AAV-DIO-mCherry ( Figures   S7I and S7J). These results suggest that the GABA CeA →PF pathway is necessary and sufficient for the development of CP symptoms.
If Glu PF neurons are inhibited by GABA CeA inputs, the increase of the inputs under CRS conditions should cause inhibitory effects. We next investigated Glu PF neuronal activity by whole-cell recordings performed in visualized glutamate neurons in slices from CaMKII-tdT mice, which were generated by CaMKII-Cre mice crossed with Ai14 mice ( Figure S8A). A decrease in the spike number in PF CaMKII-tdT neurons from CRS 3W mice ( Figure 5G) but not CRS 2W ( Figure S8B) was observed when compared with control mice. In addition, optical activation of Glu PF neurons through PF injection of Cre-dependent AAV-DIO-ChR2-mCherry in CaMKII-Cre mice relieved CRS-induced pain sensitization (Figures 5H and 5I;Figures S8C and S8D), while, in naive mice, chemogenetic inhibition of Glu PF neurons induced severe pain sensitization ( Figures S8E-S8J). Furthermore, we investigated the neuronal excitability of GABA neurons in the PF (GABA PF ) via whole-cell recordings in visualized GABA neurons in slices from GAD2-tdT mice ( Figure S9A). We found that the firing rate ( Figure S9B) and rheobase ( Figure S9C) were unchanged in CRS 3W mice when compared to control mice. In addition, optogenetic or chemogenetic inhibition of GABA PF neurons in GAD2-Cre mice could not relieve CRS-induced pain sensitization ( Figures S9D-S9K). These data suggest that the occurrence of CP is likely caused by an enhanced inhibitory process that involves direct GABA CeA -mediated inhibition of Glu PF neurons.

The S2 Is the Output of the GABA CeA →Glu PF Pathway for CP
We subsequently aimed to identify the GABA CeA →Glu PF output circuitry that may mediate CP behavior. Following injection of AAV-DIO-ChR2-mCherry into the PF of CaMKII-Cre mice ( Figures 6A and 6B), we observed an extremely large number of the mCherry + fibers in the S2 ( Figure 6C), a key brain region purportedly involved in the pathophysiology of pain, in addition to other brain regions ( Figure S10A). Optical activation of Glu PF terminals in the S2 elicited excitatory postsynaptic currents (EPSCs) in brain slices (−140.25 ± 13.08 pA), which were blocked by the glutamate receptor antagonist KOH 6,7dinitroquinoxaline-2,3(1H,4H)-dione (DNQX, 10 μM; −10.24 ± 0.78 pA; Figure 6D). In addition, in vivo electrophysiology recordings were performed in the S2 of freely moving mice. AAV-DIO-ChR2-mCherry viruses were injected into the PF of CaMKII-Cre mice.
After 3 weeks, multi-tetrode recording and optic fiber were implanted into the S2 ( Figure  6E). Multiple channel recordings of the firing rate of the S2 before, during, and after light photostimulation were conducted ( Figures 6F and 6G). We found that blue light increased the firing rate in the S2 neurons. Furthermore, optical stimulation of these fibers in the S2 reduced CRS-induced pain sensitization (Figures 6H-6J) but did not affect locomotion (Figures S10B-S10D). These results suggest that the GABA CeA →Glu PF outputs to the S2 are part of the inhibitory circuit that underlies CP behavior.
To characterize the connectivity of the GABA CeA →Glu PF →S2 pathway, we took advantage of an efficient combination strategy that permits direct visualization of these projections ( Figure 7A). We initially injected retro-AAV (AAV2/2Retro-CMV-bGl-Cre-EGFP) into the S2 that allowed retrograde virus transport to the soma of Glu PF neurons, allowing these neurons to express Cre-GFP (Tervo et al., 2016) (Figure 7B, left). Subsequently, the Credependent retrograde trans-monosynaptic tracing system was used to infect Glu PF neurons that expressed Cre-GFP ( Figure 7B, middle; Figures S11A and S11B). We visualized numerous DsRed + cell bodies in the CeA ( Figure 7B, right), which were identified as GABA neurons by immunofluorescence staining ( Figure 7C).
To examine the functional synaptic connectivity of the GABA CeA →Glu PF →S2 pathway, we injected Fluoro-Gold (FG) into the S2 to label the S2-projecting Glu PF neurons and injected Cre-dependent AAV-DIO-ChR2-mCherry into the CeA of GAD2-Cre mice ( Figure 7D).
Similar to the findings in Glu PF neurons with GABA CeA inputs (see Figure 2E), photostimulation of GABA CeA fibers in the PF elicited an IPSC in FG + neurons, which was reversed by bicuculline ( Figure 7E). Post hoc immunofluorescence by intracellularly labeling with neuronbiotin-488 confirmed the recorded FG + neurons were glutamate neurons ( Figure 7F). In addition, following injection of FG into the S2 and H129-GFP4 into the CeA (Figures S11C and S11D), total overlapping of GFP + and FG + signals in the PF was observed ( Figure S11E), which were identified as glutamate neurons ( Figure S11F). These results demonstrate a direct functional connectivity of GABA CeA →Glu PF →S2 pathway ( Figure S12).

DISCUSSION
This study defines a GABA CeA →Glu PF →S2 pathway through which CP is generated. Central to this process is a circuit mechanism involving increased inhibition of CeA GABAergic neurons on S2-projecting PF glutamatergic neurons under depression conditions to prime pain behavior.
The major problem of why certain depression treatments are ineffective for CP patients is far from understood (Bair et al., 2003). For instance, the SSRI paroxetine has been shown to be effective for treating depression but not effective for CP (Marks et al., 2009). Consistent with this notion, we demonstrate that typical analgesic agents for inflammatory and neuropathic pain had no effect on the symptoms of CP in animals, implicating that a pathway is necessary for the occurrence of CP. In marked contrast, the inhibition of the GABA CeA →Glu PF →S2 pathway relieved the symptoms of CP animals, indicating that the GABA CeA →Glu PF output to the S2 is likely critical for CP treatment. This also suggests that typical analgesic agents are likely inappropriate for treating CP patients. In addition, data supporting pain-reducing effects of selectively serotonergic medications are more limited, presumably due to the role of serotonin in both inhibiting and enhancing pain via descending pathways. In fact, the exact targets for these antidepressants remain elusive. For their analgesic effects, one possible mechanism is the activation of the descending modulatory pathway from the rostral ventromedial medulla and periaqueductal gray to the dorsal horn neurons, via inhibition of serotonin and norepinephrine reuptake. These results raise the possibility that selective activation of this pathway is a possible intervention for CP.
The advantage of studying the neural circuit of a disorder seems to reveal convergent points of drug action and pathologically relevant behavioral consequences. Although the neural circuits underlying major depression (Knowland et al., 2017) and chronic pain (Cheng et al., 2017;Liu et al., 2018) have been extensively investigated, the mechanism underlying CP remains unknown. Based on our present findings, we would like to propose a hypothesis for the neural circuit responsible CP, in which depression enhances the inhibition of the GABA CeA →Glu PF pathway, thus decreasing excitatory outputs to the S2, and resulting in the symptoms observed in CP. Of note, CP occurred in a state-dependent manner, accompanied by a gradual increase of GABA in the PF and a decrease of Glu PF neuronal activity at CRS 3W but not at CRS 2W. Thus, it appears that the duration of stress is important for the emergence of pain symptoms. This complex behavioral outcome can likely be attributed to the dynamic network activity during progressive stress. Adaptation of the GABA CeA →Glu PF pathway at CRS 3W primes the development of CP. In addition, we found that reversing the dysfunction of the GABA CeA →Glu PF pathway alleviated the symptoms of CP but not those of inflammation-or neural-injury-induced pain, suggesting that this pathway is likely specifically recruited under chronic stress conditions. The CeA forms widespread connections with forebrain areas and the hypothalamus, to regulate emotional behavior (Pessoa, 2008;Roozendaal et al., 2009). This is confirmed by our findings that the chemogenetic inhibition of GABA CeA neurons reduced CRS-induced depressive-like behaviors. Although the CeA is part of a descending endogenous pain control system that includes circuits in the brainstem (e.g., periaqueductal gray [Tovote et al., 2016]) and spinal cord (Fields, 2004), how the CeA integrates affective information to increase pain processing through the ascending pain pathway is unknown. Our results show that the PF is a direct target of the CeA that contributes to CP. Moreover, we identified that the S2 receives glutamatergic inputs from the GABA CeA →Glu PF pathway to control the symptoms of CP. As the S2 is the final path for processing polymodal sensory information integration (Keysers et al., 2010;Milligan and Watkins, 2009;Navratilova and Porreca, 2014), cell-type-specific regulation of the tone of the GABA CeA →Glu PF pathway could be one of the specific rules for the development of CP. Interestingly, inhibiting GABA CeA neurons relieved both the pain and depressive symptoms, while inhibiting GABA CeA terminals in the PF only relieved pain symptoms. These data suggest that CRS-induced pain sensitization and depressive-like behaviors are regulated by the CeA through different individual pathways. Although GABA CeA neuronal activity likely contributes to multiple behavioral outputs, including fear, anxiety, and depression (Li et al., 2013), the specific projection of this type of neuron to the PF could be relatively more important for CP in the current animal models. The significance of these findings with respect to treatment-resistant CP is that manipulations that target this neural circuitry would have dual effects on the symptoms of depression and chronic pain.
As pharmacological options for the treatment of CP remain quite limited, these findings raise the possibility of developing optimal treatments that involve the use of dual-acting drugs or non-drug approaches, such as deep-brain stimulation (Mayberg et al., 2005) or transcranial magnetic stimulation (Pascual-Leone et al., 1996), which target the converging pathway.

STAR★METHODS LEAD CONTACT AND MATERIALS AVAILABILITY
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Zhi Zhang (zhizhang@ustc.edu.cn).
This study did not generate new unique reagents.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Animals-In all experiments, C57BL/6J, GAD2-Cre, CaMKII-Cre and Ai14 (RCL-tdT) male mice (purchased from Charles River or Jackson Laboratories) at 8-10 weeks of age were used. Unless cannula surgery was performed, the mice were housed five per cage in a colony with ad libitum access to water and food. They were maintained under a 12-hour light/dark cycle (lights on from 7:00 a.m. to 7:00 p.m.) at a stable temperature (23-25°C).
The animal protocols were approved by the Care Committee of the University of Science and Technology of China. A total of 350 mice were recruited for behavioral and viral tracing experiments in the current study, with 85 mice excluded due to missed targets, including the injection of viruses, or the placement of optic/microdialysis fibers. The viral tracing and behavioral data of these animals were excluded from further analyses.

CRS:
Mice were periodically constrained from moving by placing them in a 50 mL syringe for 6 h, every day, for 3 weeks. Holes were drilled in the ends of the syringes to allow the mice to breathe (see Figure 3A). During the restraint period, control mice were allowed to freely move around in the cage without water or food provided. Syringes were thoroughly washed every day after the restraint period. To exclude the effect of acute stress, mice were allowed to rest one day after CRS before the anatomical and electrophysiological experiments were conducted.
CFA: Complete Freund's adjuvant (CFA, 25 ul) or saline was injected into the left hind paw under brief isoflurane anesthesia for inflammatory pain model and control.

SNI:
The SNI surgery was performed under anesthesia with isoflurane. The skin and muscle of the left thigh were incised to explore the sciatic nerve, consisting of the sural, common peroneal, and tibial nerves. After exploration, nonabsorbent 4-0 chromic gut were used to ligate the common peroneal and tibial nerves, and then the nerves were transected, and about 2 mm sections from the dot were removed. The skin was stitched and disinfected with iodophor (see Figure 3F). For the sham mice, the procedure was the same as for the experimental group except that the nerves were left intact.

METHODS DETAILS
Virus and trace injection-Mice were fixed in a stereotactic frame (RWD, Shenzhen, China) under a combination of xylazine (10 mg/kg) and ketamine (100 mg/kg) anesthesia. A heating pad was used to maintain the core body temperature of the animals at 36°C. A volume of 100-300 nL virus (depending on the expression strength and viral titer) was injected using calibrated glass microelectrodes connected to an infusion pump (micro 4, WPI, USA) at a rate of 30 nl/min. The coordinates were defined as dorsal-ventral (DV) from the brain surface, anterior-posterior (AP) from bregma and medio-lateral (ML) from the midline (in mm).
After three weeks, the rabies virus RV-EnvA-∆G-dsRed (2 × 10 8 IFU/ml, 300 nl) was injected into the same site of the PF. For GABA CeA →Glu PF →S2 triple tracing, the AAV2/2Retro-CMV-bGl-Cre-EGFP virus (AAV-Cre-GFP, 1.83 × 10 13 vg/ml, 100 nl; Taitool, Shanghai, China) was first injected into the S2 (AP: −0.50mm, ML: −4.50mm, DV: −0.90mm with a 20° angle) to allow the virus to spread retrogradely to the PF soma to express Cre-GFP; simultaneously, the Cre-dependent mixed helper virus was infused into the ipsilateral PF. The RV was injected after three weeks. Mice that had been anesthetized with pentobarbital (20 mg/kg, i.p.) were transcardially perfused 7 days after the last injection, and brain slices were prepared (40 μm) for tracing DsRed signal or co-staining with GABA antibody. The retrograde tracer cholera toxin B subunit conjugated to CTB-488 (0.1% mg/mL, 100 nl, Thermolife) was injected into the PF of GAD2-tdT mice for 7 days to label PF projecting GABA CeA neurons for visualized electrophysiological recordings. Retrograde tracer Fluoro-Gold (0.1% mg/mL, 100 nl, Santa Cruz Biotechnology) was injected into the S2 to label S2 projecting PF neurons.
Optogenetic manipulations in vivo-An optical fiber cannula was initially implanted into areas of interest, which included the PF and S2, in the brain of an anesthetized mouse that had been immobilized in a stereotaxic apparatus. The implant was secured to the animal's skull with dental cement. Chronically implantable fibers (diameter, 200 μm, Newdoon, Hangzhou) were connected to a laser generator using optic fiber sleeves. The delivery of a 5-min pulse of blue light (473 nm, 2-5 mW, 15-ms pulses, 20 Hz) or yellow light (594 nm, 5-8 mW, constant) was controlled by a Master-8 pulse stimulator (A.M.P.I., Jerusalem, Israel). The same stimulus protocol was applied to the mice in the control group. Optical activation of CeA GABA neurons induced obvious freezing behavior (Video S1) and reduced movement during light on ( Figures S6B and S6C), which is consistent with previous studies (Janak and Tye, 2015;Li et al., 2013). Locomotion would recover during 5-min light off ( Figure S6D). Thus, it is inappropriate to analyze depressive-like behavior during this time point, because of the deficiency in locomotion during freezing. To avoid the effect of locomotion and unify the behavioral paradigm, we performed behavioral experimentation after light stimulation withdrawal. The location of the fibers was examined in all mice at the conclusion of the experiments, and data obtained from mice in which the fibers were located outside of the desired brain region were discarded.

Whole-cell patch-clamp recordings:
Neurons were visualized using a 40 × water immersion objective on an upright microscope (BX51WI, Olympus, Japan) equipped with interference contrast (IR/DIC) and an infrared camera connected to the video monitor. Whole-cell patch-clamp recordings were obtained from visually identified CeA, PF or S2 cells. Patch pipettes (3-5 MΩ) were pulled from borosilicate glass capillaries (VitalSense Scientific Instruments Co., Ltd., Wuhan, China) with an outer diameter of 1.5 mm on a fourstage horizontal puller (P1000, Sutter Instruments, USA). The signals were acquired via a Multiclamp 700B amplifier, low-pass filtered at 2.8 kHz, digitized at 10 kHz and analyzed with Clampfit 10.7 software (Molecular Devices, Sunnyvale, CA, USA). The data were collected from the neurons with the series resistance < 30 MΩ and input resistance > 100 MΩ. In some experiments, the 0.5% neurobiocytin 488 was included in the intracellular solution, together with immunofluorescence staining, for cell type identification. The current-evoked firing was recorded in current-clamp mode (I = 0 pA). The threshold current for firing was defined as the minimum strength of current injection required to elicit at least one or two spikes. For recording the intrinsic membrane properties, the pipettes (5-7 MΩ) were filled with potassium gluconate-based internal solution resistance containing (in mM): 130 K-gluconate, 2 MgCl2, 5 KCl, 0.6 EGTA, 10 HEPES, 2 Mg-ATP and 0.3 Na-GTP (osmolality: 285-290 mOsm/kg, pH: 7.2). Measurement of the intrinsic membrane properties were conducted under current-clamp mode and data were collected from neurons with a resting membrane potential negative than -50 mV and an overshot of the action potentials. The threshold current of the action potential was defined as the minimum current to elicit an action potential.
Light-evoked response: Optical stimulation was delivered using a laser (Shanghai Fiblaser Technology Co., Ltd. China) through an optical fiber 200 μm in diameter positioned 0.2 mm from the surface of the brain slice. To test the functional characteristics of AAV-DIO-ChR2-mCherry, fluorescently labeled neurons that expressed ChR2 in GAD2-Cre or CaMKII-Cre mice 3-4 weeks after virus injection were visualized and stimulated with a blue (473 nm, 5-10 mV) laser light using 5-Hz, 10-Hz, and 20-Hz stimulation protocols with a pulse width of 15 ms. In some experiments, the function of eNpHR3.0 was assessed by applying sustained yellow (594 nm, 5-10 mV, 200 ms) laser light stimulation.
In vivo electrophysiology recording-For chronic extracellular recordings, a custommade four movable tetrode array was implanted into the PF (AP: −2.10 mm, ML: −0.77 mm, DV: −3.00 mm) and the S2 (AP: −0.50mm, ML: −4.00mm, DV: −1.2mm). Each tetrode was made of four twisted fine platinum/iridium wires (12.5 μm diameter, California Fine Wire). The optrode was constructed by surrounding an optical fiber (200 μm, Newdoon) with tetrodes. The screw-based microdrive scaffolds for lowering the electrodes were cemented onto the skull. The mice were allowed to recover for at least 3 days before recordings were made. Recording electrodes were attached to a 16-channel headstage, and neuronal signals were amplified, filtered at a bandwidth of 300-5000 Hz, and stored using Neurostudio software (Jiangsu Brain Medical Technology, China). Neuronal firings in the PF or S2 were recorded in the absence and presence of blue light stimulation (60 s). Spike sorting was performed with a sorting method involving a T-Dis E-M algorithm built in Offline Sorter 4 (Plexon, USA). The firing rates of sorted units were calculated using Neuroexplorer 4 (Nex Technologies, USA).
Immunohistochemistry and imaging-The mice were deeply anesthetized with pentobarbital sodium (50 mg/kg, i.p.) and sequentially perfused with saline and 4% (w/v) paraformaldehyde (PFA). The brains were subsequently removed and post-fixed in 4% PFA at 4°C overnight. After cryoprotection of the brains with 30% (w/v) sucrose, coronal sections (40 μm) were cut on a cryostat (Leica CM1860) and used for immunofluorescence. The sections were incubated in 0.3% (v/v) Triton X-100 for 0.5 h, blocked with 10% donkey serum for 1 h at room temperature, and incubated with primary antibodies, including anti-GABA (1:500, rabbit, Sigma), anti-Glutamate (1:500, rabbit, Sigma), at 4°C for 24 h, followed by the corresponding fluorophore-conjugated secondary antibodies for 2h at room temperature. Fluorescence signals were visualized using Zeiss LSM710 microscope.
For post hoc immunofluorescence, slices were collected after whole-cell recording, in which the patched neurons were labeled with neuronbictin-488, and were immersed into 4% PFA for post-fixation in 4°C for 1 week. Then, slices were washed in 0.01M PBS for 3 × 30 minutes, followed by incubation overnight with blocking solution [10% (v/v) donkey serum, 1.0% (v/v) Triton X-100 dissolved in 0.01M PBS]. Then, slices were incubated with primary antibody diluted in blocking solution at 4°C for three days, including anti-Glutamate (1:100, rabbit, Sigma). After washing in 0.01M PBS, slices were immersed in Alexa 594-or Alexa 647-conjugated secondary antibody (1:100, Invitrogen) diluted in 0.01M PBS at RT for 8 hours, then washed in 0.01M PBS for 3 × 30 minutes, mounted on the glass slides. Images were captured via confocal microscopy.
Assessment of depressive-like behaviors-Mice were transported to the behavioral testing room in a holding cabinet to habituate at least 2 hours prior to testing. The behavior of the animals during testing was recorded using a video tracking system and was subsequently analyzed offline. Dim light (~20 lux) was used in the room to minimize the anxiety of the animals.
Forced swim test: A mouse was placed in a transparent Plexiglas cylinder (30 cm high and 12 cm in diameter) that contained fresh water (24 ± 1°C) up to a height of 25cm from the bottom for 6 min. The animal's behavior was videotaped from the side, and the duration of immobility of each mouse during the last 5 min of recording was measured offline. Despair behavior was defined as floating or remaining motionless such that the overall immobility time was two s.d. greater than that of the control.

Tail suspension test:
Mice were individually suspended about 50 cm above the surface of a table using adhesive tape that was placed approximately 1 cm from the tip of the tail. Each mouse was tested only once for 6 min; the test was videotaped from the side, and the immobility time of the animal was measured in the last 5 min. Mice were considered immobile without initiated movements; immobility was considered to include passive swaying.
Open field test: Mice were placed in one corner of an open field apparatus that consisted of a square area (25 cm × 25 cm) and a marginal area (50 cm × 50 cm × 60 cm); the mice were allowed to freely explore their surroundings. The animals' movement trajectories were recorded for 5 min using EthoVision XT software, which records the number of entries into and the amount of time spent in the central area. The area was cleaned with 75% ethanol after each test to remove olfactory cues from the apparatus.

Elevated plus maze test:
The elevated plus maze consisted of a central platform (6 × 6 cm), two closed arms (30 × 6 × 20 cm) and two opposing open arms (30 × 6 cm), which was placed 100 cm above the floor. A mouse was placed in the central platform facing a closed arm and was allowed to explore the maze during 5 min session. The time spent in the open arms and the number of entries into the open arms were analyzed using EthoVision XT software. Seventy-five percent ethanol was used to clean the area between tests.
Von Frey filament test-Mice were individually placed in transparent plastic chamber on a wire mesh grid to allow for calibrated Von Frey hairs insertion on the surface of hind paw. Before testing, mice were allowed to acclimate to the testing environment for 30 min. We tested the withdrawal threshold (g) of the planta when stimulating the paw's surface with the calibrated Von Frey hairs. The mean threshold was calculated from five applications. Pain thresholds were measured every 5 min or daily by the paw-withdrawal test on a freely moving animal with the Hargreaves apparatus (IITC Life Science Inc., Woodland Hills, CA, USA) for thermal hyperalgesia. Pain thresholds were tested immediately after 1 min of light stimulation with optogenetic manipulations.
In vivo microdialysis-HPLC-A microdialysis probe (CMA7, CMA, USA), connected to a syringe infusion pump (CMA402, CMA, USA) via polyethylene tubing, was initially implanted into the right PF of deeply anesthetized mice. The tissue was perfused with normal Ringer's solution (145 mM NaCl, 3 mM KCl and 1.3 mM CaCl 2 ) via the pump at a rate of 0.5 ml/min, and the dialysate was collected in a 15 mL quantitative loop through the probe in freely moving mice. Then the dialysate was derived using ophthaldialdehyde (OPA) solution (37 mM OPA, 50 mM Na2SO3, 90 mM H3BO3, and 5% MeOH) for 3 min. After that, dialysate was automatically loaded to the mobile phase (0.1 M NaH2PO4 and 10% MeOH, pH = 3.2) and separated on a 1 mm × 50 mm column (ALF-105, Antec, Netherlands) with a 3 μm particle size at a rate of 0.35 ml/min. Detection was conducted using an Alexys online analysis system (Antec Leyden) that consisted of a DECADE II electrochemical detector and VT-3 electrochemical flow cells. The data were analyzed using Clarity software (ANTEC, Netherlands) based on standard samples.
For chemogenetic-microdialysis experiments, samples were collected for 3 hours before CNO intraperitoneal injection. Then, sample was collected at 1, 2, and 3 hours after CNO injection. The mean threshold of GABA concentration after CNO was calculated from three applications.

QUANTIFICATION AND STATISTICAL ANALYSIS
We conducted simple statistical comparisons using Student's t test. ANOVA (one-way and two-way) and post hoc analyses were used to statistically analyze the data from the experimental groups with multiple comparisons. All data are expressed as the mean ± SEM, and significance levels are indicated as * p < 0.05, ** p < 0.01 and *** p < 0.001. OriginPro 2017 software (Origin Lab Corporation, USA) and GraphPad Prism 5 (Graph Pad Software, Inc., USA) were used for the statistical analyses and graphing. Offline analysis of the data obtained from electrophysiological recordings was conducted using Clampfit software version 10.7 (Axon Instruments, Inc., USA) and MiniAnalysis software version 6.03 (Synaptosoft Inc., USA). All statistical data are presented in Table S1.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.

•
GABAergic neurons from the CeA project to glutamatergic neurons in the PF

•
Enhanced inhibition of the GABA CeA →Glu PF pathway in CRS mice with pain • Optical inhibition of the GABA CeA →Glu PF pathway relieves pain symptoms in CRS mice  (H and I) Representative traces (H) and summarized data (I) showed the firing rate of the PF neurons in the GAD2-Cre mice before, during, and after light photostimulation with multiple channel recordings (n = 14 neurons).