Microglia Are Indispensable for Synaptic Plasticity in the Spinal Dorsal Horn and Chronic Pain

Abstract

Spinal long-term potentiation (LTP) at C-fiber synapses is hypothesized to underlie chronic pain. However, a causal link between spinal LTP and chronic pain is still lacking. Here, we report that high-frequency stimulation (HFS; 100 Hz, 10 V) of the mouse sciatic nerve reliably induces spinal LTP without causing nerve injury. LTP-inducible stimulation triggers chronic pain lasting for more than 35 days and increases the number of calcitonin gene-related peptide (CGRP) terminals in the spinal dorsal horn. The behavioral and morphological changes can be prevented by blocking NMDA receptors, ablating spinal microglia, or conditionally deleting microglial brain-derived neurotrophic factor (BDNF). HFS-induced spinal LTP, microglial activation, and upregulation of BDNF are inhibited by antibodies against colony-stimulating factor 1 (CSF-1). Together, our results show that microglial CSF1 and BDNF signaling are indispensable for spinal LTP and chronic pain. The microglia-dependent transition of synaptic potentiation to structural alterations in pain pathways may underlie pain chronicity.

Keywords: brain-derived neurotrophic factor; calcitonin gene-related peptide; chronic pain; colony-stimulating factor 1; high-frequency stimulation; long-term potentiation; microglia.

Figure 1.. HFS that does not injury the stimulated sciatic nerve induces spinal LTP and chronic pain hypersensitivity in mice.

(A) Spinal LTP at C-fiber synapses was induced by HFS at 10 and 20 V but not at 1V. The representative traces of C-fiber evoked fEPSPs were recorded before (i) and 3 h (ii) after high frequency stimulation (HFS) of the sciatic nerve at different intensities. Amplitude of C-fiber evoked fEPSPs (red vertical line) was determined automatically by parameter extraction software of WinLTP. Scale bars: x =100 ms, y = 0.2 mV. Summary data of the C-fiber responses expressed as mean ± SEM plotted vs. time were shown below (n = 6 mice/group). The arrow indicates the time point at which HFS was delivered. (B, C) Representative images and summary data show the expression of ATF3 in L4 DRG neurons 7 days after HFS at different intensities and 7 days after spared nerve injury (SNI). Scale bar, 100 μm. n = 3–4 mice/group, 3 slices/mouse, ***P < 0.001, compared with sham group, ^###P < 0.001 vs. 20V HFS, one-way ANOVA with Tukey’s test. (D) The latencies to fall in rotarod test in different groups at 7 days after HFS at different intensities or SNI are shown. n = 5–12 mice for each group, ***P < 0.001, compared with sham group, one-way ANOVA with Tukey’s test. (E) Immunofluorescence of Iba1 (a marker for resident macrophage) and NIMP-R14 (a marker for neutrophil) in the stimulated sciatic nerve from different groups at 7days after HFS or SNI are shown (n = 3 mice/group, 2–3 sections/mouse). Scale bar: 50 μm. (F) Quantification of Iba1 expression in the sciatic nerve. ***P < 0.001, compared with sham group, ^###P < 0.001 vs. 20V HFS, one-way ANOVA with Tukey’s test. (G, H) HFS at 10V and 20V significantly decreased 50% paw withdrawal threshold (PWT, G) to von Frey filaments and paw withdrawal latency (PWL, H) to radiant thermal stimuli, compared with the sham group or 1V HFS (n = 12 mice for 10V HFS group, n = 5–6 mice for other groups, **P < 0.01, ***P < 0.001 vs. sham group, two-way ANOVA with Fisher’s LSD test). (I) The decrease in mechanical thresholds induced by 10 V HFS was prevented by local application of 2% lidocaine (50 μl) at the stimulated sciatic nerve 15 min before HFS (n = 5–7 mice/group). ^###P < 0.001 vs. HFS group, two-way ANOVA with Fisher LSD’s test. (J) Intrathecal injection of NMDA receptor antagonist D-AP5 (50 μg/ml, 5 μl) but not vehicle (Vehi, PBS) 30 min before HFS abolished HFS-induced mechanical hypersensitivity (n = 6–8 mice/group). ^##P < 0.01, ^###P < 0.001 vs. vehicle group, two-way ANOVA with Fisher LSD’s test.

Figure 2.. HFS induces long-lasting synaptic potentiation and enhances CGRP terminals in SDH in a NMDA receptor-dependent manner.

(A) Original representative traces of fEPSP and the input/output curves (stimulation intensity/C-fiber response) demonstrated that synaptic efficacy was enhanced at 3–5 h or 7–14 days after 10 V HFS (n = 12 mice in sham and in HFS 3–5 h groups, n = 8 mice in HFS 7–14d group, two-way ANOVA with Tukey’s test). Scale bars: x = 50 ms, y = 0.2 mV. (B) Western blot assay illustrates the time course of changes in CGRP levels in the ipsilateral SDH following 10V HFS (n = 3–4 mice/group). **P < 0.01, ***P < 0.001 vs. sham group, one-way ANOVA with Fisher’s LSD test. (C) Representative confocal images of whole ipsilateral SDH (left) and magnified images (right) show the distributions of CGRP⁺ terminals 7 days after 10V HFS or sham operation. The laminae of SDH were shown by white lines. Representative higher-magnification images from dotted boxed regions in larger images show CGRP positive varicosities. Scale bars: 100 μm (left), 20 μm (right), 5 μm (right insert). (D) The histograms show the increased expression of CGRP⁺ terminals in different laminae of the ipsilateral SDH at 3 or 7 days after HFS (n = 3–4 mice/group, 2–3 sections/mouse). **P < 0.01, ***P < 0.001 compared with sham group, one-way ANOVA with Fisher’s LSD test. (E-F) Electron micrographs and the histograms show the CGRP-immunoreactive varicosities and synapses in the different laminae of the ipsilateral SDH in sham (n = 2 mice, 2–3 micrographs/laminae/mouse) and HFS groups (n = 3 mice, 2–3 micrographs/laminae/mouse). The black circles indicate CGRP⁺ varicosities and the white arrows CGRP⁺ synapses pointing from presynaptic to postsynaptic membrane. Scale bar: 200 nm. *P < 0.05, **P < 0.01, ***P < 0.001 compared with sham group, one-way ANOVA with Tukey’s post-hoc test. (G, H) Double fluorescent images and the scatter diagrams show the expression of CGRP and IB4 in SDH at 7 days after 10V HFS with or without NMDA antagonist D-AP5 (AP5). Scale bar: 50 μm (left). n = 3 mice/group, n = 3–4 images/mouse. ***P < 0.001 compared with sham group, ^###P < 0.001 vs. vehicle HFS group, one-way ANOVA with Tukey’s post-hoc test.

Figure 3.. Microglia ablation blocks the spinal LTP, the increase in CGRP terminals and mechanical allodynia in induced by HFS.

(A) Western blots show the time course of Iba1 expression level after 10V HFS (n = 3 mice/ group). ***P< 0.001 vs. sham group, one-way ANOVA with Fisher’s LSD test. (B, C) Representative images of Iba1 immunostaining and the histogram show that the number of Iba1⁺ microglia were increased in the ipsilateral SDH at 1, 3 and 7 days after 10V HFS, and 7 days after SNI. Iba1⁺ cells were increased only in ipsilateral but not in contralateral SDH 7 days after HFS (7dI and 7dC). Scale bar: 50 μm. n = 3 mice/group, 2–3 sections/mouse. ***P < 0.001, ****P < 0.0001vs. sham group, ^###P < 0.001 vs. HFS 7d group, one-way ANOVA with Fisher’s LSD test. (D) Experiment designs for control or microglial ablation (MG abl) at 3 or 21 days after vehicle (vehi, corn oil) diphtheria toxin treatment in CX3CR1^CreER/+: R26^iDTR/+mice. DT: diphtheria toxin, TM: Tamoxifen, Pain Beha: pain behavior test, EP: electrophysiology. In DT 3d HFS or DT 21d HFS group, DT was injected 3 weeks after TM and HFS was delivered at 3 or 21 days after last application of DT. In control group, vehicle (Vehi) but not DT was injected following TM injection. (E) The immunofluorescent staining showed that microglia were largely depleted at 3 days after DT treatment (DT 3d), and then fully repopulated within 21 days after DT treatment (DT 21d) in CX3CR1^CreER/+: R26^iDTR/+ mice. Boxed areas are magnified in the right panel. Scale bars: 100 μm (left) and 25 μm (right). (F) HFS was able to induce LTP at 21 days but not at 3 days after DT injection (n = 12 mice for DT 3 d group, n = 6–7 mice for other groups, two-way ANOVA with Tukey’s test). Representative traces of evoked C-fiber fEPSP are shown above before (i, grey line) and 3 h after 10V HFS (ii, colored line). Scale bars: x = 100 ms, y = 0.15 mV. The arrow indicates the time point at which HFS was delivered. (G) The input-output curves of fEPSP measured 7 d after HFS in mice treated with DT (for microglial ablation) and with vehicles are shown (n = 5 mice for sham group, n = 5 or 7 mice for MG ablation groups). (H, I) The increase in CGRP⁺ terminals induced by HFS was substantially attenuated in microglia ablation mice (n = 3 mice/group, 2–3 sections/mouse). Scale bar: 20 μm. ***P < 0.001 vs. control sham group; ^#P < 0.05, ^##P < 0.01 vs. control HFS 3d or 7d group, one-way ANOVA with Fisher’s LSD test. (J) HFS induced mechanical allodynia in mice with microglia full repopulation (DT 21d), but not in mice with microglia ablation (DT 3d). n = 12 mice in MG ablation DT 3d group, n = 8 mice for other groups. **P < 0.01, ***P < 0.001 vs. Control HFS, two-way ANOVA with Fisher’s LSD test.

Figure 4.. Ablation of microglial BDNF prevented LTP induction, the increase in CGRP terminals and mechanical allodynia induced by HFS.

(A) Schematic overview of experiments with BDNF^−/−, CX3CR1^CreER/+: BDNF^flox/flox mice, in which BDNF can be specifically deleted from microglia by TM injection. (B) 10V HFS induced spinal LTP in BDNF^+/+ mice but not in mice with microglial BDNF deletion (BDNF^−/−). n = 6–7 mice/group. Examples of C-fiber field potentials are shown above (i, grey lines) and 3 h after 10V HFS (ii, color lines). Scale bars: x = 100 ms, y = 0.2 mV. The arrow indicates the time point at which conditional stimulation was delivered. (C, D) Representative images and summary data show that HFS-induced CGRP upregulation was inhibited by microglial BDNF ablation (n = 3–4 mice/group, 3–4 sections/mouse). ***P < 0.001 vs BDNF^+/+ sham group, ^###P < 0.001 vs. BDNF^+/+ HFS group, one-way ANOVA with Fisher’s LSD test. Scale bars: 100 μm (top), 20 μm (bottom). (E) Unlike SNI model, HFS induced mechanical allodynia was significantly reversed in male and female mice deficient of microglial BDNF compared with control mice. ***P < 0.001 vs. female BDNF^+/+ HFS or SNI group, two-way ANOVA with Fisher’s LSD test.

Figure 5.. HFS upregulates CSF1 in CGRP terminals and CSF1 receptors in microglia, leading to spinal LTP and pain hypersensitivity.

(A, B) The representative images and statistical analysis show the expression of CSF1 in the ipsilateral L4 DRG from 1 to 7 days after HFS and 7 days after SNI. The triple staining images showed the colocalization of CSF1, CGRP and ATF3. The arrowheads indicate that the injured DRG neurons (ATF3⁺) that strongly express CSF1 does not express CGRP. The arrows indicate the uninjured neurons (ATF3⁻) that weakly express CSF1 express CGRP following HFS or SNI. Scale bar: 50 μm. n = 3 mice/group, n = 2–3 images/mouse, *P < 0.05, ***P < 0.001 vs sham group; ^###P < 0.001, compared with HFS 7d group, one-way ANOVA with Tukey’s post-hoc test. (C) Triple staining show CSF1 in SDH was co-localized with CGRP but not with IB4 at 3 days after HFS. Scale bar: 20 μm. (D) qRT-PCR analysis for CSF1 mRNA in the ipsilateral L4–5 DRGs from different groups (n = 7–12 mice/group). Values are presented as means ± SEM. *P < 0.05, **P < 0.01, ****P < 0.00001 vs. sham group; ^###P < 0.001, compared with HFS 7d group, one-way ANOVA with Fisher’s LSD test. (E) Western blot and statistical analysis show that treatment of cultured DRG neurons with 40 mM KCl for 24 h increased CSF1 in culture media. n = 6 samples/group, **P < 0.01 vs Vehi group, unpaired Student’s t-test. (F, G) Immunostaining shows CSF1R in microglia was upregulated in the ipsilateral dorsal horn at 1 and 3 days after HFS (n = 3 mice/group, 2–3 sections/mouse). Scale bar: 50 μm. ***P < 0.001, compared with sham group, ^###P < 0.001, compared with HFS 7d, one-way ANOVA with Tukey’s post-hoc test. (H) Western blots illustrate the increased expression of CSF1R and Iba1 protein levels at 3 days after 10V HFS, and the changes were blocked by intrathecal injection of CSF1 neutralizing antibody (anti-CSF1, 40 ng/μl, 5 μl) 30 min before HFS (n = 3–4 mice/group). **P < 0.01, ***P < 0.001 vs. sham group. ^###P < 0.001 vs. control HFS group, one-way ANOVA with Fisher’s LSD test. (I) Local application of anti-CSF1 (40 ng/μl, 20 μl) onto the dorsal surface of spinal cord at recording segments 30 min before HFS did not affect baseline of C-fiber responses, but blocked late-phase of the spinal LTP (n = 5–6 mice/group). Original traces of C-fiber fEPSPs from different groups, recorded at the baseline (i), 30 min before (ii) and 3 h (iii) after HFS, are shown on the top. Scale bars: x = 100 ms, y = 0.2 mV. The arrow indicates the time point at which conditional stimulation was delivered. (J) Intrathecal injection of anti-CSF1 antibody 30 min before HFS prevented the development of mechanical hypersensitivity (n = 6 mice/group, ***P < 0.001 vs. vehicle group (Vehi), two-way ANOVA with Fisher’s LSD test.

Figure 6.. Microglial BDNF is required for CSF1-induced increase in CGRP terminals and pain hypersensitivity.

(A) HFS upregulated BDNF and CGRP levels in SDH, and the effects were prevented by intrathecal injection of anti-CSF1 (40.0 ng/μl) 30 min before HFS (n = 3–6 mice/group). ***P < 0.001 vs. control sham group. ^##P < 0.01; ^###P < 0.001 vs. control HFS group; ^$P < 0.05 vs. anti-CSF1 sham group, one-way ANOVA with Fisher’s LSD test. (B-D) In cultured spinal cord slices, CSF1 (1 μg/ml) markedly upregulated microglial markers (CD11b and Iba1) and CSF1R, and the effects were dose-dependently suppressed by p38 MAPK inhibitor SB203580 (10 μM) applied 1 hour prior to CSF1. Scale bar: 100 μm. n = 4–5 samples/group,* P < 0.05, ***P < 0.001 vs control group; ^###P < 0.001 vs CSF1 group, one-way ANOVA with Tukey’s post-hoc test. (E-F) In culture spinal cord slices, treatment with CSF1 for 6 hours upregulated CSF1R, the phospho-p38 MAPK (p-p38) and BDNF, as well as increased BDNF in culture media. Pretreatment with p38 inhibitor SB203580 (10 μM) abolished the effects of CSF1. n = 3–5 samples/group, **P < 0.01, ***P < 0.001 vs control group; ^##P < 0.01 vs CSF1 group, one-way ANOVA with Tukey’s post-hoc test. (G) Intrathecal CSF1 upregulated CGRP expression in BDNF^+/+ mice but not in mice deficient of microglial BDNF (BDNF^−/−), as measured 3 days after injection (n = 3–6 mice/group). ***P < 0.001 vs. BDNF^+/+PBS group, ^###P < 0.001 vs. BDNF^+/+ CSF1 group, one-way ANOVA with Fisher’s LSD test. (H) The mechanical allodynia induced by intrathecal CSF1 was significantly attenuated in microglial BDNF deletion mice compared with control mice (n = 5–6 mice/group, ^###P < 0.001 vs. BDNF^+/+ CSF1 group), two-way ANOVA with Fisher’s LSD test.

Figure 7.. Schematic diagram depicting the mechanism of HFS of sciatic nerve induced chronic pain without the stimulated sciatic nerve injury.

HFS of sciatic nerve (1) depolarizes DRG neuron (2) to release CSF1 from the central terminals (3); CSF1 activates microglial CSF1R/p38 MAPK to release BDNF (4); BDNF causes the increase of CGRP terminal via TrkB/CREB signaling (5) Microglia activation and CGRP terminal increase underlie prolonged LTP (6) and chronic pain (7).