Switch of serotonergic descending inhibition into facilitation by a spinal chloride imbalance in neuropathic pain

Our first aim was to clearly establish the role of acute stimulation of 5-HT neurons of the RMg in nociceptive transmission. We used a specific mouse strain expressing cre-recombinase under the control of a PET1 (polyethylene terephthalate pheochromocytoma 12 ETS (E26 transformation-specific) promoter specific to 5-HT neurons (26). We first crossbred ePet::Cre mice with Ai9::tdtomato reporter mice. We observed a high level of colocalization of cre-recombinase with tryptophan hydroxylase 2 (TPH2), a specific marker of 5-HT neurons (fig. S1). We also observed spinal projections of cre(+) fibers in the dorsal horn in two main regions, the superficial and deep laminae (Fig. 1A). Higher magnification showed clear putative synaptic bouton in the deep layer of the DHSC (Fig. 1A, inset). Next, we used a viral-based transduction approach to express the excitatory opsin channelrhodopsin 2 (ChR2) in 5-HT neurons. Three weeks after injection of an adeno-associated virus (AAV) containing ChR2 and enhanced yellow fluorescent protein (EYFP) [AAV-EF1a-DIOhChR2(H134R)-EYFP-WPRE-pA] in the RMg, we found EYFP expression in both cell bodies of 5-HT neurons of the RMg and their projections to the DHSC (fig. S1B and Fig. 1B). We used slices of RMg to perform whole-cell patch clamp recordings to determine the validity of opsin expression and light stimulation protocol (Fig. 1C). We used a stimulation protocol that activates 5-HT neurons in a physiological range (5 Hz/5 ms) (27), and we observed that neurons fired action potentials that faithfully followed optogenetic stimulation (Fig. 1C, inset). In the next step, we used high-performance liquid chromatography (HPLC) measurements of the DHSC after optogenetic stimulation of 5-HT neurons, and we observed that our protocol elicited a significant increase in the level of 5-HT in the DHSC, but not of the other monoamines measured (Fig. 1D).

Then, we performed optogenetic stimulation of 5-HT neurons in both male and female animals using our stimulation protocol. ChR2 activation with optic fibers implanted above the injection site (i.e., RMg) (see Materials and Methods) elicited a strong mechanical and thermal analgesia (Fig. 2, A and B). We show that (i) paw withdrawal threshold (PWT) was significantly increased (fig. S2A1), (ii) five repetitive stimulations (RSs) with a suprathreshold Von Frey filament failed to induce reproducible behavioral responses (fig. S2A2), and (iii) thermal latency (TL) to paw withdrawal was significantly increased (Fig. 2B). These effects were equivalent in males and females. To focus on descending pathways projecting to the DHSC, we performed the same experiments but implanted the optic fiber above the lumbar spinal cord to specifically target 5-HT projecting fibers (28). As shown in Fig. 2 (C and D), we observed the same results as in Fig. 1 (A and B), suggesting that 5-HT–induced analgesia is due to 5-HT fibers projecting in DHSC. As a control, optogenetic stimulation in ChR2(−) mice remained ineffective regardless of the location of the fiber implantation (fig. S2).

To confirm that such analgesia is due to direct modification of nociceptive transmission in DHSC, we performed in vivo electrophysiological recordings. We first evaluated the global consequence of optogenetic stimulation of 5-HT neurons by recording local field potential after peripheral electrical stimulation of the paw (Fig. 3, A to C). Nociceptive field potentials (NFPs) were decreased during optogenetic stimulation in mice expressing ChR2, but not in mice expressing green fluorescent protein (GFP) only. This effect was equivalent in males and females and confirmed that 5-HT–induced analgesia occurs through a modification of the nociceptive transmission in the DHSC. To determine whether the decrease in nociceptive information is localized at the level of the DHSC, we recorded deep dorsal horn neurons that project outside DHSC. We restricted recordings to wide dynamic range neurons (WDRs) that are highly convergent and exhibit a clear C-component, i.e., a delayed response following suprathreshold electric stimulation of the corresponding receptive field, which can be used as a readout of nociceptive transmission (29) (Fig. 3, D to G). During optogenetic stimulation, we observed a decrease in the number of action potentials in the C-fiber range (Fig. 3, E and F) that was not present with mice expressing the GFP tag only. This decrease was equivalent in males and females (Fig. 3G).

Subsequently, we worked to determine the spinal microcircuit involved in this analgesia. 5-HT receptors are widely expressed in the DHSC and exert excitatory or inhibitory influences on either excitatory or inhibitory spinal neurons. Although it is known that 5-HT can activate inhibitory neurons (30, 31), we cannot exclude that analgesia would be due to a direct inhibitory action of 5-HT fibers on excitatory afferent fibers or projection neurons. We first aimed at determining if 5-HT neurons could also express inhibitory neurotransmitters. Most of the inhibitory interneurons in the RVM including RMg express both glycine and GABA (32), and we used GAD67::GFP mice to determine whether GFP fluorescence could be colocalized with TPH2 staining in the RMg and found no colocalization (fig. S3A). We also used ePet::Cre-Ai9::tdtomato mice (see above) and GAD65 immunostaining and found numerous putative synaptic contacts of GABA terminals on 5-HT neurons and fibers in the RMg but no colocalization between GABA and 5-HT fibers/terminals neither in the RMg nor in the DHSC (fig. S3, B and C). Then, we sought for 5-HT targets in the DHSC and performed immunostaining of both excitatory and inhibitory interneurons in spinal slices of epet::cre mice injected in the RMg with an AAV-CAG-Flex-GFP. We compared the apposition of GFP fluorescence in 5-HT fibers with a marker of excitatory (tlx3) or inhibitory neurons (Pax2) (Fig. 4, A to C). We found that 5-HT fibers are in closer contact with Pax2 than Tlx3 in both males and females (Fig. 4C), suggesting connections between 5-HT fibers and local inhibitory interneurons. To confirm this hypothesis, we used GAD67::GFP mice and performed TPH2 immunostaining. We found putative 5-HT(+) boutons onto GABAergic interneurons on both cell bodies and fibers (Fig. 4D). We injected an AAV-DIO-Synaptophysin-myc in the RMg of ePET::cre and found synaptic terminals in the deep lamina of the DHSC (Fig. 4E). Last, we crossbred the ePET::cre mouse with the GAD67::GFP mouse to seek for synaptic buttons between 5-HT fibers originating in the RMg and GABAergic neurons in the DHSC. We injected an AAV-DIO-Synaptophysin-myc in the RMg and performed a co-immunohistochemistry of Myc tag and GFP. We observed synaptic contacts between 5-HT fibers and GABAergic neurons on both cell bodies and fibers (Fig. 4F). To determine whether our optogenetic protocol activates inhibitory interneurons in the DHSC, we performed optogenetic stimulation of 5-HT descending fibers in anesthetized mice. One hour after optogenetic stimulation, mice were euthanized and quickly perfused, and early gene cFos immunostaining was performed (fig. S4). To identify inhibitory interneurons, an immunostaining against Pax2 was also conducted. We found, in both superficial and deep layer of the DHSC, c-Fos (+) neurons that were also Pax2(+), showing that optogenetic activation of 5HT fibers activate spinal inhibitory interneurons. Last, to confirm that the analgesic effect was due to the microcircuit linking 5-HT neurons of the RMg and inhibitory interneurons in the DHSC, we evaluated mechanical and thermal pain sensitivity during optogenetic stimulation of 5-HT fibers after intrathecal injection of picrotoxin to block GABA_A and glycine (Gly) receptors. Picrotoxin suppressed the analgesic effect of optogenetic stimulation in both males and females (Fig. 5, A and B). To confirm that this blockade altered nociceptive transmission in DHSC, we performed in vivo electrophysiological recordings of WDR neurons, and we observed that blockade of GABA/Gly receptors suppressed the decrease in WDR neuron responsiveness (Fig. 5, C and D). Together, these results show that 5-HT fibers come from the RMg synapse onto spinal inhibitory interneurons to, in turn, inhibit spinal nociceptive transmission.

In neuropathic pain models, disinhibition due to chloride dysregulation appears to be a crucial mechanism responsible for neuronal hyperexcitability and pain hypersensitivity (33–35). We then investigated whether we could reverse the effect of the optogenetic stimulation of 5-HT fibers coming from the RMg by changing chloride balance in the DHSC. We performed intratumoral injections of two different blockers of chloride transporters, the nonselective inhibitor furosemide (fig. S5) and the selective K⁺-Cl⁻-transporter KCC2 inhibitor VU043271 (Fig. 6, A and B). In both cases, optogenetic stimulation of 5-HT fibers induced a significant increase in pain sensitivity in both males and females. PWT was significantly decreased (Fig. 6A1 and fig. S5), subthreshold Von Frey filament became suprathreshold (RS, Fig. 6A2 and fig. S5), and TL was significantly decreased (Fig. 6B and fig. S5). In vivo electrophysiological recordings showed that VU043271 caused a switch in the response of WDR neurons to optogenetic manipulation of 5-HT neurons. Under VU043271 superfusion, but not vehicle, WDR neurons increased their response to peripheral stimulation when RMg 5-HT fibers are optogenetically stimulated (Fig. 6, C and D). These results show that altering chloride balance could switch serotoninergic inhibition into excitation.

We then studied the role of the 5-HT descending pathway in a pathological context using the spared nerve injury (SNI) model of painful mononeuropathy (36). We first confirmed that 2 weeks after SNI, mice expressed a strong mechanical and thermal hypersensitivity (Fig. 7, A and B) that was equivalent in both males and females. Using epet::cre-Ai9::Tdtomato, we first confirmed that 5-HT descending fibers are not modified in SNI as their density is equivalent to the one quantified in naïve animals, and their contacts are mostly on inhibitory interneurons again as in naïve animals (fig. S6). Therefore, we used epet::cre SNI mice to manipulate 5-HT descending fibers from the RMg spinal optogenetic stimulation of RMg 5-HT fibers to induce a reinforcement of mechanical and thermal hypersensitivity in both males and females. PWT and TL were lowered during optogenetic stimulation of 5-HT fibers (Fig. 7, C and D). In vivo electrophysiological recordings showed that these mechanical and thermal hypersensitivities were associated with an increased responsiveness of the DHSC (NFP; Fig. 8, A and B). Last, WDR neurons increased the number C-fiber spikes in response to peripheral stimulation during optogenetic activation of spinal RMg 5-HT fibers (Fig. 8C). These effects were equivalent in both males and females (Fig. 8D). These results strongly suggest that chloride imbalance in spinal neurons could modify descending pain controls by switching the impact of descending RMg 5-HT input from inhibition to facilitation. To address that possibility, we proposed to pharmacologically enhance KCC2 chloride transporters in SNI mice using the molecule CLP290 (37, 38). We first confirmed that CLP290 induces an increase in membrane levels of KCC2. We performed immunohistochemical analysis on DHSC slices from SNI mice treated with either CLP290 or vehicle, and we found that the treatment in SNI mice (i) significantly increased KCC2 immunoreactivity in the spinal dorsal horn (fig. S7), (ii) significantly increased membrane KCC2 levels in individual dorsal horn neurons (Fig. 9, A to C), and (iii) induces a slight but significant increase in PWT and TL (fig. S7). Then, we compared pain behavior in SNI mice treated with either saline or CLP290 during optogenetic stimulation of 5-HT fibers (Fig. 9, D and E). In SNI mice treated with CLP290 but not vehicle, optogenetic stimulation of 5-HT fibers promoted mechanical and thermal analgesia as shown by a significant increase in PWT and TL in both males and females. Using in vivo electrophysiology recordings, we also found that, after CLP290 treatment, WDR neurons decreased their responsiveness to peripheral stimulations in both males and females upon optogenetic stimulation of RMg 5-HT fibers (Fig. 9, F and G). Together, these results show that KCC2-mediated chloride balance control descending influence: Normal chloride extrusion maintains effective RMg 5-HT descending input net inhibitory, while KCC2 hypofunction after nerve injury, causing chloride dysregulation, promotes RMg 5-HT descending input to become net facilitatory.

This mechanism may explain why SSRIs are not capable of relieving pain in patients with chronic pain (18, 39). To test this hypothesis, we combined SSRI treatment with the KCC2 enhancer (Fig. 10). First, we confirmed in SNI mice that the SSRI fluoxetine injection alone induced a slight mechanical and thermal hypersensitivity (Fig. 10, in red). On the other hand, CLP290 induced a slight but significant decrease in mechanical and thermal hypersensitivity (Fig. 10, in yellow). By contrast, the combination of fluoxetine and CLP290 strongly relieved mechanical and thermal hypersensitivity (Fig. 10, in green). This analgesic effect was durable and still present 24 hours after a single injection of the combined molecules (Fig. 10, C and F). This effect was comparable in males and females (fig. S8). Last, to go beyond evoked pain responses that involve sensory-motor reflexes, we designed a conditioned place preference (CPP) paradigm to address the effect of our drug combination on supraspinal pain integration (fig. S9). SNI mice were first placed in a two-chamber maze of the same size but with an opaque and a transparent chamber and remained free to explore the whole maze (fig. S9A). Then, they were alternately locked in one or the other chamber; one was associated with either our drug combination (fluoxetine + CLP290; group 1) or fluoxetine alone (group 2). Two days after, mice were again left free to move in the maze and spent significantly more time in the chamber associated with the combination, but not in the chamber associated with fluoxetine alone (fig. S9B). These results show that SNI mice associate the chamber and the combination of fluoxetine and CLP and that this combination improves the painful state of the animal.

Male and female mice, between 6 and 12 weeks of age, were maintained under a standard 12-hour light/12-hour dark cycle with food and water available ad libitum. All experiments followed European Union (Council directive 86/609EEC) and institutional guidelines for laboratory animal care and use. Institutional license for hosting animals was approved by the French Ethical Committee (APAFIS#3751). Male and female mice exhibit the same responses; thus, we pooled results from both sexes in all figures. We used different transgenic mice strains derived from C57BL6/J genetic background: ePet::Cre (strain 012712; RRID: IMSR_JAX:012712), ePet::CreXAi9::tdtomato (strain 007909, RRID: IMSR_JAX:007909), Gad67::GFP, and ePet::CreXGad67::GFP. ePet::Cre: It enables specific expression of light-activated opsins and molecular or fluorescent tags. ePet::CreXAi9::tdtomato: It allows the expression of a tomato fluorescent tag in cre-expressing cells, used to verify the expression of CRE in targeted neurons and, thus, 5-HT–expressing cells. Gad67::GFP (strain 007677, RRID: IMSR_JAX:007677): Mouse strain expressing GFP fluorescent tag in Gad67-positive neurons mainly GABA inhibitory neurons. ePet::CreXGad67::GFP: CRE expression in 5-HT neurons and GFP expression in GABA neurons. In each experimental group, mice were sex-, weight-, and age-balanced. Animals with no trace of viral infection in the RMg, those whose optic fibers were removed after implantation, those with no exaggerated pain behavior after SNI procedure, and those with a loss of body weight > 15% were excluded from the study.

We performed surgical procedures following (61).

Behavioral experiments were carried out blind and performed following procedures previously described in (61). For optogenetic experiments, the experimenter did not know the virus (ChR2 or GFP) injected into the tested mice. For pharmacology, the experimenter was not aware whether the drug or vehicle was injected.

In vivo electrophysiology was performed following procedures previously described in (61). Mice were anesthetized with urethane 20% and placed on a stereotaxic frame (M2E, Asnières, France). A laminectomy was performed on lumbar vertebrae L1 to L3, and segments L4 to L5 of the spinal cord were exposed. Extracellular recordings of WDR dorsal horn neurons were made with borosilicate glass capillaries (2 megohms, filled with 684 mM NaCl) (Harvard Apparatus, Cambridge, MA, USA). For NFPs, the signal was low pass–filtered. For single-unit recordings, the signal was high pass–filtered, and the criterion for the selection of a neuron was the presence of an A-fiber response (0 to 80 ms) followed by a C-fiber response (80 to 300 ms) to electrical stimulations of the ipsilateral paw with bipolar subcutaneous stimulation electrodes. Threshold for C-fiber–evoked response was evaluated. Trains of electrical stimulation (at 0.067 Hz) at two times the threshold for C-fibers were performed before, during, and after optogenetic stimulation, with an optic fiber placed above the recording site.

epet::cre mice expressing ChR2 in the RMg (see above) were anesthetized with urethane 20% and placed on a stereotaxic frame (M2E, Asnières, France). Surgical procedure was performed as previously described in (61). Briefly, a laminectomy was performed on lumbar vertebrae L1 to L3, and segments L4 to L5 of the spinal cord were exposed. An optic fiber was placed above the spinal cord to allow light illumination of 5-HT descending fibers. After 1 hour of stable anesthesia, an optogenetic stimulation at 5 Hz/5 ms for 2 min was performed. One hour later, mice were quickly euthanized and perfused through the left ventricle with 4% (w/v) paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (PBS). Spinal cords were extracted. dissected out, post-fixed for 3 hours at 4°C in the same solution, then cryoprotected in a solution of 0.1 M PBS and 25% sucrose overnight, and stored at −80°C. c-Fos immunostaining with c-Fos antibodies (rabbit anti-c-Fos (1/500; 9F6, mAB Cell Signalling Technology 2250 S) was performed as below (see the “For molecular identifications of spinal network” section).

Tissue samples from optogenetically stimulated and control mice were deposited in previously weighed Eppendorf tubes and placed back in the freezer at −80°C. The tubes containing the samples were cautiously weighed again the day of the biochemical analysis, and 200 μl of perchloric acid (HClO₄ 0.1 N, 4°C) was added. The samples (weighing approximately between 16 and 34 mg with no significant differences between the stimulated and the non-stimulated group) were homogenized with ultrasound for about 8 s and centrifuged at 13,000 rpm for 30 min at 4°C. A volume of 10 μl of the supernatant was injected into an HPLC system.

The tissue concentrations of monoamines were measured by HPLC coupled to electrochemical detection. The mobile phase of the HPLC system was composed of methanol (7%), NaH₂PO₄ (70 mM), triethylamine (100 μl/liter), EDTA (0.1 mM), and sodium octyl sulfate (100 mg/liter) diluted in deionized water (pH 4.2, adjusted with orthophosphoric acid) as previously reported (65). It was filtered (0.22 μm) before its installation in the system. The mobile phase was delivered through the HPLC column (Hypersyl, C18 at a flow rate of 1.2 ml/min using an HPLC pump (LC10Ad Vp, Shimadzu, France). The column was protected by a Brownlee-Newgard precolumn (RP-8, 15 × 3.2 mm, 7 μm; C.I.L.). The injection of the samples (10 μl) was carried out by a manual injection valve (Rheodyne, model 7725i, C.I.L., Sainte-Foy-La-Grande, France) equipped with a loop of 20 μl. The monoamines exit the column at different retention times (approximately NA: 3′38; DA: 7′43; 5-HT: 17′35) and passed into the coulometric detection cell (Cell 5011, ESA, Paris, France) equipped with two electrodes. The potential of these two electrodes was fixed via the coulometric detector (CoulochemII, ESA) at +350 mV (oxidation) and −270 mV (reduction), respectively. Only the signals from the oxidation channel were monitored. The coulometric detector was connected to a computer through an interface (Ulyss, Azur system, Toulouse, France). The calibration curves were performed once the peaks in a standard solution (1 ng/10 μl) were well separated in the chromatogram. The calibration curves were adapted to the expected and heterogeneous quantities of monoamines (0.1 to 1 ng/10 μl for 5-HT and NA; 0.01 to 0.1 ng/10 μl for DA) (66). Using a timeline method, the gains used ranged from 10 nA (for DA) to 200 nA (for NA and 5-HT). Standard solutions were used before each series of samples to verify the chromatographic conditions. The overall sensitivity for the compounds ranged from 2.4 pg/10 μl for DA to 7 pg/10 μl for 5-HT with a signal-to-noise ratio of 3:1.

Patch-clamp recordings were performed following procedures previously described in (61), on brain slices from mice expressing either AAV-EF1a-DIOhChR2(H134R)-mCherry-WPRE-pA or AAV-CAG-Flex-GFP. Briefly, 8- to 12-week-old mice were intracardially perfused during euthanasia (exagon/lidocaine: 300/30 mg/kg, i.p.) with ice-cold N-methyl-d-glucamine (NMDG) solution [containing the following: 1.25 mM ascorbate, 3 mM Na-pyruvate, 2 mM thiourea, 93 mM NMDG, and 93 mM HCl 37% (pH 7.3 to 7.4); osmolarity: 305 to 310 mosM]. Brains were quickly removed, and 250-μm slices containing the RMg were prepared with a VT1100S Leica vibratome in ice-cold oxygenated NMDG solution before recovery for 12 to 15 min at 34°C in oxygenated NMDG solution. Slices were then transferred at room temperature into artificial cerebrospinal fluid (aCSF) solution [containing the following: 124 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 2 mM MgCl₂, 2.5 CaCl₂, 227 2.5 mM d-glucose, and 25 mM NaHCO₃ (pH 7.3 to 7.4); osmolarity: 305 to 310 mosM] for at least 1 hour. Slices were placed in the recording chamber under a microscope (Nikon EF600) outfitted for fluorescence and IR (infrared)-DIC (digital image correlation) video microscopy and perfused with oxygenated aCSF at 2 to 3 ml/min. Viable RMg 5-HT neurons were visualized with a fluorescence video camera (Nikon). Borosilicate pipette (4 to 6 megohms; 1.5-mm optical density; Sutter Instrument) was filled with an intracellular solution (containing the following: 128 mM K gluconate, 20 mM NaCl, 1 mM MgCl₂, 1 mM EGTA, 0.3 mM CaCl₂, 2 mM Na₂–adenosine 5′-triphosphate, 0.3 mM Na–guanosine 5′-triphosphate, 0.2 mM adenosine 3′,5′-monophosphate, and 10 mM Hepes; 280 to 290 mosM, pH 7.3 to 7.4). Recordings were made using a Multiclamp 700B amplifier, digitized using the Digidata 1440A interface, and acquired at 2 kHz using pClamp 10.5 software (Axon Instruments, Molecular Devices, Sunnyvale, CA). Pipette and cell capacitances were fully compensated, but junction potential was not corrected. RMg 5-HT neurons were recorded in whole-cell current-clamp mode. Brain slices expressing ChR2 or GFP were opto-stimulated at 470 nm (5 Hz, 5 ms pulse width) for 10 s every 30 s. Laser intensity was set at 5 mW/mm².

Drug injections were performed as previously described in (61).

As already described in (61) and after completion of experiments, mice were euthanized with urethane and perfused through the left ventricle with 4% (w/v) PFA in 0.1 M PBS. Brains and spinal cords were extracted, dissected out, post-fixed for 3 hours at 4°C in the same solution, then cryoprotected in a solution of 0.1 M PBS and 25% sucrose overnight, and stored at −80°C.

For verification of both viral expression and location of optic fiber in RMg, a series of 20-μm thin slices containing the RMg were incubated free-floating in 0.1 M PBS containing Triton X-100 (0.3%), bovine serum albumin (1%; Sigma-Aldrich), and chicken anti-GFP antibody (1:1000; Averlabs) overnight at 4°C. After washing with 0.1 M PBS, secondary antibodies, such as Alexa Fluor 488–conjugated goat anti-chicken (1:500), were added in 0.1 M PBS for 2 hours at room temperature. Sections were finally viewed on epifluorescence microscope (axiophot2, Zeiss) fitted with a 20× dry objective and both 40× and 63× oil immersion 1.3 numerical aperture (NA) objective.

For molecular identifications of the spinal network, spinal cord sections were incubated free-floating in 0.1 M PBS containing Triton X-100 (0.3%), bovine serum albumin (1%; Sigma-Aldrich), guinea pig anti-TLX3 antibody (1:1000; a gift from T. Müller), goat anti-Pax2 antibody (1:300; Bio-techne), rabbit anti-Myc antibody (1:500; Euromedex), rabbit anti-TPH2 antibody (1:1000; Bio-techne), and rabbit anti-GAD65/67 (G5163, Sigma-Aldrich) in combination with chicken anti-GFP antibody (1:1000; Averlabs) overnight at 4°C. After washing with 0.1 M PBS, secondary antibodies were added in 0.1 M PBS for 2 hours at room temperature. Depending on the primary antibody species used, we used Alexa Fluor 568–conjugated goat anti-guinea pig (1:500, Molecular Probes), Alexa Fluor 568–conjugated donkey anti-goat (1:500; Thermo Fisher Scientific), Alexa Fluor 568– or 647–conjugated goat anti-rabbit (1:500; Invitrogen), and Alexa Fluor 568–conjugated goat anti-mouse (1:500; Invitrogen) associated with Alexa Fluor 488–conjugated goat anti-chicken (1:500; Thermo Fisher Scientific). For c-Fos experiments, rabbit anti–c-Fos (1:500; 9F6, mAB Cell Signalling Technology 2250 S) was combined with goat anti-Pax2 antibody (1:300; Bio-techne). c-Fos primary antibody was used for 24 hours, and then Alexa Fluor 488–conjugated goat anti-rabbit (1:500; Thermo Fisher Scientific) and Alexa Fluor 647–conjugated donkey anti-goat (1:500; Thermo Fisher Scientific) were used for c-Fos and Pax2, respectively.

Reactions using biotinylated antibodies were further incubated with Alexa 568–conjugated streptavidin (1:500; Molecular Probes) for 1 hour at room temperature. Sections were then mounted on gelatin-coated slides with DakoCytomation Fluorescent Mounting Medium (Dako SA) and finally viewed on a confocal microscope (Leica TCS SPE, Mannheim, Germany) fitted with a 20× dry objective and both 40× and 63× oil immersion 1.3 NA objective; confocal image stacks (0.75-μm steps) were acquired for each sample, and sequential acquisition was used to prevent cross-talk between Alexa 488 and Alexa 568. Three-dimensional (3D) reconstructions were performed using imaris software (Oxford Instruments, UK) from z-stack acquisitions. For 5-HT fiber contacts to Pax3- or Tlx3-positive cells, immunofluorescent 5-HT fiber and Tlx3 or Pax2 cell bodies were isolated. Each 5-HT fiber as close as half-diameter of each cell body was considered as a contact. The percentage of contact was evaluated as the ratio between cell bodies contacted by at least one 5-HT fiber, and the total amount of cell bodies was identified in one 3D image. Statistics were performed with three different 3D images of three different lumbar spinal slices from separate animals.

The analysis method presented in this study is based on an already published algorithm used to quantify receptor membrane internalization (68). This technique was designed to reduce conscious or unconscious biases that can arise from user interventions. Before the analyses, the intensity of the immunostaining background noise was defined from white matter of the spinal dorsal horn, a region where KCC2 is known to be absent. For each image, the background average intensity was subtracted to the whole image to exclusively calculate the KCC2(+) immunostaining contribution. All neurons, identified by a NeuN(+) labeling, present in the immunocytochemistry confocal image were considered. The distance of each neuron from the gray/white matter border was first estimated. This provides for each neuron a measure of anatomical position in the dorsal horn and will be used for pooling KCC2 intensity profiles. To measure the subcellular KCC2 intensity profiles, the membrane regions of neuronal cells were delineated. For each pixel in the region of interest, the distance to the closest membrane segment was calculated. Using this distance map, the mean pixel intensity and SD of KCC2 fluorescence signal were quantified as a function of the distance to the neuronal membrane (defined as zero). Positive axis values correspond to neuronal intracellular space (Fig. 9, A and B). KCC2 subcellular intensity profiles were averaged as a function of the distance from gray/white matter border (Fig. 9C). The average pixel intensity profiles for each analyzed neuron were pooled together (distance bins from gray/white matter border of 20 μm, from 20 to 120 μm) to generate a graph of the KCC2 expression as a function of the anatomical position in the dorsal horn (Fig. 9G). A ratio paired t test was used to compare KCC2 membrane immunostaining intensities between vehicle and CLP290 at the different distance ranges from the gray matter surface. A total of 542 neurons (N = 6 mice) for the vehicle condition and 534 neurons (N = 8 mice) for the CLP290 condition were considered. The KCC2 membrane intensity profile for the 100- to 120-μm bin (68 neurons from 6 vehicle mice and 68 neurons from 8 CLP290 mice) was shown to illustrate the KCC2 membrane intensity profile in Fig. 2G. A ratio paired t test was used to compare KCC2 membrane immunostaining intensities between vehicle and CLP290 at 0, 1, 2, and 3 μm from the membrane.