NMDA Receptor Activation Underlies the Loss of Spinal Dorsal Horn Neurons and the Transition to Persistent Pain after Peripheral Nerve Injury.

Peripheral nerve lesions provoke apoptosis in the dorsal horn of the spinal cord. The cause of cell death, the involvement of neurons, and the relevance for the processing of somatosensory information are controversial. Here, we demonstrate in a mouse model of sciatic nerve injury that glutamate-induced neurodegeneration and loss of γ-aminobutyric acid (GABA)ergic interneurons in the superficial dorsal horn promote the transition from acute to chronic neuropathic pain. Conditional deletion of Grin1, the essential subunit of N-methyl-d-aspartate-type glutamate receptors (NMDARs), protects dorsal horn neurons from excitotoxicity and preserves GABAergic inhibition. Mice deficient in functional NMDARs exhibit normal nociceptive responses and acute pain after nerve injury, but this initial increase in pain sensitivity is reversible. Eliminating NMDARs fully prevents persistent pain-like behavior. Reduced pain in mice lacking proapoptotic Bax confirmed the significance of neurodegeneration. We conclude that NMDAR-mediated neuron death contributes to the development of chronic neuropathic pain.

INTRODUCTION

A key characteristic of pain caused by a lesion or disease of the nervous system (neuropathic pain) is its persistence (Colloca et al., 2017). Preclinical studies have revealed that multiple molecular changes in primary sensory neurons, plasticity of central nociceptive connections, and neuroinflammation collectively contribute to the development of neuropathic pain (von Hehn et al., 2012). Many of these investigations focus on the onset of pain during the first one or two weeks following nerve injury. It is increasingly clear though that the involvement of individual pain mechanisms changes with time. Microglia, for example, promote the onset of neuropathic pain through cytokine and growth factor release in the spinal cord (Peng et al., 2016). In contrast, astrocytes respond to nerve injury with a delay of several days or weeks and appear to support the maintenance of pain rather than its initial development (Ji et al., 2014; Scholz and Woolf, 2007). Although insight into the short- and long-term changes of the nociceptive system after nerve injury has improved, the mechanisms driving the transition from acute to chronic pain remain to be resolved.

One process potentially linked to the emergence of persistent pain is the apoptosis of dorsal horn neurons. A loss of dorsal horn neurons after nerve injury has been found in independent studies (Yowtak et al., 2013; Scholz et al., 2005). Conflicting reports may be explained by inadequate statistical power because they relied on the analysis of a single section per spinal cord and an unconventional stereological design that has not been validated (Polgár et al., 2004, 2005). However, the mechanisms responsible for the induction of apoptosis are unknown and its functional significance has been disputed (Polgár et al., 2005). Clarifying the etiology and relevance of nerve-injury-induced neurodegeneration is essential because neuroprotection may offer a disease-modifying treatment strategy for neuropathic pain.

We have previously shown that afferent input from the injured nerve promotes the apoptosis induction (Scholz et al., 2005). The major transmitter of primary sensory neurons at central nociceptive synapses of the dorsal horn is glutamate. Postsynaptic insertion of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid-type glutamate receptors (AMPARs), voltage-dependent recruitment of N-methyl-D-aspartate receptors (NMDARs), and changes in the composition of postsynaptic densities strengthen glutamatergic transmission following a nerve lesion (Kuner, 2010; Latremoliere and Woolf, 2009). Activation of NMDARs is critical for this process of central sensitization because it leads to enhanced Ca2+ influx and the engagement of intracellular signaling pathways that amplify the response to nociceptive input. However, excess Ca2+ increases may cause neuron death, and extrasynaptic NMDAR activation may suppress survival-promoting genes (Hara and Snyder, 2007; Hardingham et al., 2002). To test whether nerve injury produces an excitotoxic challenge in the spinal cord, we selectively eliminated NMDARs in dorsal horn neurons of adult mice. Unlike pharmacological inhibition, this targeted genetic approach did not interfere with presynaptic glutamatergic transmission or NMDAR activity in supraspinal pathways. We found that the conditional deletion of NMDARs at dorsal horn synapses provided effective neuro-protection and blocked the transition from acute to persistent neuropathic-pain-like behavior without affecting physiological nociception. Mice lacking proapoptotic Bcl2-associated × (Bax) showed a similar reduction of chronic pain after nerve injury, confirming the functional impact of neurodegeneration.

RESULTS

Eliminating Functional NMDARs Abolishes Nerve-Injury-Induced Apoptosis

We examined the induction of apoptosis after spared nerve injury (SNI), a partial sciatic nerve lesion associated with persistent pain hypersensitivity in rodents (Figure 1A; Decosterd and Woolf, 2000). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) revealed cell profiles with chromatin changes indicative of apoptosis in the dorsal horn of C57BL/6 mice 7 days after SNI (Figure 1B). Most of these profiles were located in the medial dorsal horn of spinal cord segment L4 (Figure 1C), in a distribution corresponding to the central projections of the lesioned common peroneal and tibial branches of the sciatic nerve (Corder et al., 2010). Apoptosis induction continued for 2 weeks and was limited to the ipsilateral dorsal horn (Figures 1D and 1E).

Figure 1.

NMDAR-Mediated Glutamatergic Transmission Causes Nerve-Injury-Induced Apoptosis

(A) Schematic illustration of SNI and the relevant neuroanatomy. DRG, dorsal root ganglion.

(B) Apoptotic cell profiles 7 days after SNI. H33342, Hoechst 33342 (bisbenzimide). Scale bar, 10 μm.

(C) Distribution of apoptotic profiles within the dorsal horn, shown in overlays of 10 sections per mouse (n = 8 mice).

(D) Time course of apoptosis induction. Apoptotic profiles were counted in 10 sections per mouse (n = 8). p < 0.001 for surgery and time in a two-way ANOVA. ***p < 0.001 in Sidak’s test following the ANOVA.

(E) Apoptosis in uninjured mice and the ipsilateral and contralateral dorsal horns of mice 7 days after SNI (n = 8). p < 0.001 in a one-way ANOVA. ***p <0.001 in Dunnett’s test following the ANOVA. Apoptosis in the white matter was not significantly different in Student’s t test.

(F and G) Stereotaxic injection of AAV8-GFP-Cre into the spinal cord of Grin1 mice eliminated NMDAR function.

(F) GFP expression (left) and in situ hybridization of Grin1 mRNA (right) 3 weeks after injection of the vector into the left dorsal horn of the lumbar (L4) spinal cord. Scale bars, 100 μm (left) or 200 μm (right).

(G) Excitatory currents evoked by NMDA (100 μM) in Grin1 mice injected with AAV8-GFP or AAV8-GFP-Cre. SNI was performed 2 or 3 weeks after the vector injection. Two weeks after the nerve injury, we compared current amplitudes in the dorsal horn of mice injected with AAV8-GFP (n = 5 neurons) or AAV8-GFP-Cre (n = 11). ***p <0.001 in Student’s t test. Currents recorded in Grin1 mice injected with AAV8-GFP did not differ from those in uninjured C57BL/6 mice (n = 4) or C57BL/6 mice after SNI (n = 6).

(H) Apoptotic profiles 7 days after SNI in the dorsal horn of Grin1 mice injected with AAV8-GFP or AAV8-GFP-Cre (n = 6 mice). ***p < 0.001 in Student’s t test.

Error bars indicate SEM. See also Figures S1 and S5.

To test the involvement of NMDAR-mediated glutamatergic transmission, we utilized transgenic mice with a floxed sequence of the Grin1 gene (Tsien et al., 1996). Grin1 encodes NMDAR subunit GluN1, which is required for the assembly of functional receptors. Stereotaxic injection of an adeno-associated virus (AAV8) with a plasmid encoding a fusion protein of Cre recombinase and GFP eliminated Grin1 expression and NMDAR activity in the spinal cord within 2 or 3 weeks (Figures 1F, 1G, and S1A). The loss of NMDAR-mediated transmission was a direct result of transgene recombination and not related to AAV8 infection or GFP expression. NMDA-evoked currents in Grin1 mice in jected with a GFP-encoding control vector did not differ from those in uninjured C57BL/6 mice or C57BL/6 mice after SNI (Figure 1G). To examine the effect of conditional Grin1 deletion on apoptosis induction, we first injected Grin1 mice with AAV8-GFP-Cre and waited 3 weeks to ensure complete recombination before performing SNI. Seven days after the nerve lesion, we found that apoptotic profiles in the dorsal horn were markedly reduced compared to mice injected with the control vector AAV8-GFP (Figure 1H).

Grin1 Deletion Prevents Neuron Loss

AAV8 infection and Cre expression were restricted to dorsal horn neurons (Figures S1B–S1F). To examine the efficacy of neuro-protection provided by NMDAR elimination, we conducted a stereological analysis of neuron survival. Because apoptosis induction was concentrated within the distribution of central terminals of the injured sciatic nerve branches, we counted neurons immunostained for neuronal nuclei protein (NeuN) in the medial half of the dorsal horn. We performed these counts 4 weeks after SNI, assuming that neurodegeneration would be complete at this time. SNI led to a decrease in the number of dorsal horn neurons by 25% (95% confidence interval [CI] 16%–35%; p < 0.001) in laminae I+II and 14% (95% confidence interval 4.1%–24%; p <0.01) in laminae III+IV (Figures 2A and 2B). Grin1 mice in jected with AAV8-GFP-Cre were fully protected against the loss of neurons. The number of dorsal horn neurons in these mice after SNI did not differ from that in uninjured C57BL/6 mice. In contrast, neurons in Grin1 mice injected with the control vector AAV8-GFP were reduced by 19% in both laminae I+II (95% CI 4.4%–34%; p < 0.05) and laminae III+IV (95% CI2.6%–36%; p < 0.05; Figures 2C and 2D).

Figure 2.

Grin1 Deletion Protects against the Loss of Dorsal Horn Neurons

(A) Representative photographs of the L4 dorsal horn of uninjured C57BL/6 mice and 28 days after SNI. Neurons were immunostained for NeuN. Scale bar, 100 μm.

(B) Stereological counts of neurons (NeuN) in laminae I+II and laminae III+IV of the dorsal horn of uninjured mice (n = 6) and after SNI (n = 8).

(C) Representative photographs of the L4 dorsal horn of Grin1 mice 28 days after SNI. Scale bar, 100 μm.

(D) Stereological counts of neurons in the dorsal horn of Grin1 mice injected with AAV8-GFP (n = 6) or AAV8-GFP-Cre (n = 5). Neurons were counted 28 days after SNI.

*p < 0.05; **p < 0.01; ***p < 0.001 in Student’s t tests. Error bars indicate SEM. Schematic drawings indicate the counting ROIs (gray). See also Table S1.

NMDAR-Mediated Neuron Death Impairs Spinal Inhibition

γ-aminobutyric acid (GABA)ergic and glycinergic inhibition is essential for the processing of nociceptive information (Peirs and Seal, 2016). Pharmacological blockade of inhibitory signal transmission, experimental ablation, or silencing of inhibitory dorsal horn neurons provoke spontaneous pain-like behavior and exaggerated reactions to painful stimuli (Cui et al., 2016; Foster et al., 2015; Petitjean et al., 2015; Duan et al., 2014; Lu et al., 2013). Whether nerve-injury-induced neurodegeneration in the dorsal horn involves inhibitory neurons is controversial (Yowtak et al., 2013; Scholz et al., 2005; Polgár et al., 2003).

To address this issue, we first determined whether SNI caused a decrease in GABAergic neurons. For this purpose, we used mice that express GFP under the promotor for glutamate decarboxylase 1 (Gad1), a marker of GABAergic neurons (Figure 3A; Tamamaki et al., 2003). Four weeks after SNI, we found GABAergic neurons in laminae I+II of the L4 spinal cord reduced by 10% (95% CI 0.4%–20%; p < 0.05). The number of GABAergic neurons in laminae III+IV did not change (Figure 3B). Lamina III and lower contain the majority of glycinergic dorsal horn neurons. Most of these neurons are also GABAergic, but those that produce glycine only would not be fluorescently labeled in Gad1-GFP mice (Zeilhofer et al., 2012). To account for these neurons, we utilized mice that express the yellow fluorescent protein Venus under the promoter for vesicular inhibitory amino acid transporter (Viaat), a marker of both GABAergic and glycinergic neurons (Figure 3C; Wang et al., 2009). Stereological analysis of neurons expressing Venus confirmed that the number of inhibitory neurons in laminae III+IV did not decline after SNI (Figure 3D).

Figure 3.

Neuroprotection Preserves Spinal Inhibition

(A) GFP expression in GABAergic dorsal horn neurons of Gad1-GFP mice. Some GABAergic neurons also produced glycine. Scale bars, 100 μm (large panels) or 10 μm (small panels).

(B) Stereological counts of GABAergic neurons in the L4 dorsal horn of uninjured Gad1-GFP mice and 28 days after SNI (n = 6 mice).

(C) Viaat-Venus mice expressed the fluorescent protein in both GABAergic and glycinergic neurons. Scale bars, 100 μm (large panels) or 10 μm (small panels).

(D) Stereological counts of inhibitory interneurons in laminae III+IV of uninjured Viaat-Venus mice (n = 5) and 28 days after SNI (n = 6).

(E–I) Miniature IPSC recordings.

(E) Representative traces of total mIPSCs in C57BL/6 mice. Bar graphs show current frequency and peak amplitude in uninjured mice (n = 17 neurons) and 2 weeks after SNI (n = 25).

(F) Comparison of total mIPSCs in the spinal cord of Grin1 mice injected with AAV8-GFP (n = 8 neurons) or AAV8-GFP-Cre (n = 12). Currents were recorded 2 weeks after SNI.

(G) Frequency and peak amplitude of GABAergic mIPSCs in uninjured C57BL/6 mice (n = 12) and 2 weeks after SNI (n = 12).

(H) GABAergic mIPSCs in Grin1 mice injected with AAV8-GFP (n = 5) or AAV8-GFP-Cre (n = 4). GABAergic currents were recorded in the presence of strychnine.

(I) Glycinergic mIPSCs, recorded in the presence of bicuculline, did not differ between uninjured C57BL/6 mice (n = 9) and 2 weeks after SNI (n = 11). *p < 0.05; **p < 0.01; ***p < 0.001. Student’s t test was used for all comparisons. Error bars indicate SEM. See also Figure S2 and Table S1.

Next, we tested whether neuroprotection by Grin1 deletion maintained the functional integrity of inhibitory neurons. To this end, we analyzed miniature inhibitory postsynaptic currents (mIPSCs) from neurons in lamina II and the border to lamina III in slices of the L4 spinal cord of C57BL/6 mice. Following SNI, the frequency of mIPSCs decreased by 69% (p < 0.001). The current amplitude did not change, indicating a presynaptic cause of the inhibitory deficit (Figure 3E). Patch-clamp recordings from the spinal cord of Grin1 mice with intact NMDARs showed a similar low frequency of total mIPSCs after SNI. However, eliminating functional NMDARs prevented the loss of mIPSCs. Frequency and amplitude of mIPSCs in the dorsal horn of Grin1 mice injected with AAV8-GFP-Cre were indistinguishable from currents recorded in uninjured C57BL/6 mice, demonstrating that Grin1 deletion preserved spinal inhibition (Figure 3F).

Fast synaptic inhibition in the spinal cord is mediated through GABAA or glycine receptors (Zeilhofer et al., 2012). For a differential evaluation of these components, we recorded GABAergic mIPSCs in the presence of the glycine receptor inhibitor strychnine (1 μM) and found that, after SNI, these mIPSCs were reduced by 56% (p < 0.05). The isolation of GABAergic currents further revealed a moderate decrease in amplitude by 25% (p <0.05; Figure 3G). Grin1 deletion protected against the changes in both frequency (p < 0.01) and amplitude (p < 0.05) of GABAergic mIPSCs (Figure 3H). Glycinergic mIPSCs, recorded in the presence of the GABAA receptor antagonist bicuculline (10 μM), were not affected by the nerve injury, indicating that the inhibitory deficit after SNI was primarily caused by a loss of GABAergic currents (Figure 3I).

We also tested whether excitatory currents decreased. However, neither frequency nor amplitude of miniature excitatory postsynaptic currents (mEPSCs) recorded from neurons in the L4 spinal cord of C57BL/6 mice changed after SNI (Figure S2).

The Deficit in Inhibition Is Not Explained by a Downregulation of GABA Synthesis or Vesicular Transport

The decrease in mIPSCs was greater than the proportion of inter-neurons lost after SNI. This may reflect the dense connectivity and high number of synapses formed by GABAergic neurons or suggest a deficit in transmitter synthesis and release by surviving interneurons. Gad1 and Gad2 catalyze GABA synthesis, whereas glycine is primarily recycled from the extracellular space by glycine transporter 2 (Glyt2). Viaat concentrates both GABA and glycine in presynaptic vesicles (Zeilhofer et al., 2012). To determine whether changes in the expression of these proteins contributed to the loss of inhibition, we used real-time qPCR and western blotting. Despite the reduced number of GABAergic neurons after SNI, expression of all four genes in the ipsilateral dorsal horn remained stable (Figures 4A–4D). To examine whether gene transcription was upheld by epigenomic modulation, we used chromatin immunoprecipitation (ChIP) for acetylated (ac) lysine (K) residues 9 and 27 of histone 3 (H3) and trimethylated (me3) H3K4. These histone modifications are commonly associated with active gene transcription (Kouzarides, 2007). Enrichment of H3K27ac at the start sites for Gad1 (p < 0.05) and Gad2 (p = 0.08) suggested enhanced transcription of these genes (Figure 4E). Consequently, Gad1 and Gad2 may have been upregulated in surviving GABAergic neurons without fully compensating for the effects of neurodegeneration. Other histone modifications appeared to reflect less specific changes in the dorsal horn transcriptome as they were also found at promoter regions of unrelated genes, such as actin (Actb) or activity-regulated cytoskeletal-associated protein (Arc) (Figure S3).

Figure 4.

Surviving Neurons Maintain the Expression of Gad, Glyt2, and Viaat

(A–D) Transcript and protein levels of (A) Gad1, (B) Gad2, (C) Glyt2 (Slc6a5), and (D) Viaat (Slc32a1) were unchanged in the L4 dorsal horn of mice 28 days after SNI (n = 3 or 4 biological replicates). qPCR results are presented as mean fold change of ΔCT ± SD after normalization to Gapdh and protein levels as mean fold change ± SEM after normalization to β3-tubulin (Tubb3). Images show representative western blot results.

(E) ChIP analysis of H3K27ac at the promoter regions for Gad1, Gad2, Actb, and Arc (n = 4). His-tone modifications 28 days after SNI are shown, compared to uninjured mice.

*p < 0.05. Student’s t test was used for all comparisons. Error bars indicate SEM. See also Figure S3 and Table S2.

Eliminating NMDAR Signaling Stops the Transition to Chronic Neuropathic Pain

Stereotaxic injection of AAV8 into the dorsal horn or GFP expression alone did not alter nociceptive behavior except for a slightly longer response to the cooling sensation evoked by acetone evaporation (Figure S4).

To examine the impact of NMDAR elimination on pain-like behavior after SNI, we compared the responses to calibrated von Frey filaments and acetone evaporation in Grin1mice injected with AAV8-GFP-Cre or AAV8-GFP. Withdrawal responses in both groups of mice increased within 7 days (p <0.001 compared to sham-operated mice), equivalent to the development of mechanical and cold allodynia in patients with neuropathic pain (Figure 5). Cold hypersensitivity after SNI was initially higher in mice lacking Grin1 (p < 0.05 compared to Grin1 mice injected with AAV8-GFP). However, the responses of Grin1 mice injected with AAV8-GFP-Cre to both mechanical and cold stimulation returned to pre-injury levels, whereas neuropathic-pain-like behavior in mice with intact NMDARs persisted throughout the testing period of 8 weeks. Interestingly, the time to full recovery differed between stimulation modalities. Mechanical pain hypersensitivity in Grin1 mice injected with AAV8-GFP-Cre was completely reversed 3 weeks after SNI (p < 0.001; Figures 5A and 5B). Four weeks later, the mice no longer exhibited pain-like responses to cold stimulation (p < 0.01; Figures 5C and 5D).

Figure 5.

Eliminating Functional NMDARs in the Dorsal Horn Blocks the Transition from Acute to Chronic Neuropathic Pain

(A and C) Withdrawal responses to (A) mechanical (von Frey filaments) or (C) cold stimulation (acetone evaporation) after sham surgery (n = 7 mice) or SNI (n = 10) in Grin1 mice injected with AAV8-GFP-Cre, and SNI in Grin1 mice injected with AAV8-GFP (n = 10). Behavioral outcomes were compared using twoway ANOVAs. p < 0.001 for treatment and time in the responses to stimulation with von Frey filaments; p < 0.01 for treatment and p < 0.001 for time in the acetone test. Asterisks and pound signs indicate the results of Bonferroni’s tests following the ANOVAs. (B and D) Areas under the curve (AUCs) were first compared for the total test duration after SNI (p < 0.001 for Frey filaments, B, and acetone test, D, in one-way ANOVAs). Separately, we compared AUCs in the acute phase of the first 7 days after SNI (p < 0.001 for both test modalities), during the transition from acute to persistent pain (p < 0.001 for the stimulation with von Frey filaments and p < 0.05 for the acetone test), and for persistent pain after 21 or 49 days, respectively (p <0.001 for von Frey filaments and p < 0.01 for the acetone test). Asterisks and pound signs indicate the results of Tukey’s tests following the ANOVAs.

*p < 0.05, **p < 0.01, and ***p < 0.001 for the comparison of AAV8-GFP-Cre + SNI and AAV8-GFP + SNI; #p < 0.05, ##p < 0.01, and ###p < 0.001 for the comparison of AAV8-GFP-Cre + Sham and AAV8-GFP-Cre + SNI. Error bars indicate SEM. See also Figure S4.

Spinal Disinhibition and Neuropathic Pain Are Attenuated in Bax-Deficient Mice

Increased Ca2+ entry in response to NMDAR activation may trigger neuron death through the intrinsic apoptosis pathway. To test the role of this pathway for the loss of spinal inhibition and chronic pain after nerve injury, we used mice with a constitutive deletion of Bax (Knudson et al., 1995). Bax is one of the key proteins involved in intrinsic apoptosis and regulates the distribution of Ca2+ between cytosol and endoplasmic reticulum (D’Orsi et al., 2015).

Apoptosis after SNI was almost completely blocked in Baxdeficient mice (Bax−/−) (p < 0.001; Figure 6A). Miniature IPSCs in uninjured Bax−/− mice were less frequent and had a lower amplitude compared to C57BL/6 mice but did not differ from wild-type mice (Bax+/+). After SNI, the frequency of total and GABAergic mIPSCs decreased by 64% (p < 0.05) and 56% (p = 0.06), respectively, in wild-type mice, similar to the loss of inhibition in C57BL/6 mice (Figures 6B and 6C). In contrast, synaptic inhibition in Bax−/− mice remained intact (p < 0.01 for total mIPSCs and p < 0.05 for GABAergic currents compared to Bax+/+ mice; Figures 6B and 6C).

Figure 6.

Mice Lacking Proapoptotic Bax Are Protected against Nerve-Injury-Induced Disinhibition

(A) Apoptotic profiles in Bax+/+ and Bax−/− mice 7 days after SNI (n = 6).

(B) Representative recordings of total mIPSCs. Bar graphs below show current frequency and peak amplitude in uninjured Bax+/+ mice and Bax+/+ mice 2 weeks after SNI (n = 18 neurons) and in uninjured Bax−/− mice (n = 13) and Bax−/− mice 2 weeks after SNI (n = 19).

(C) GABAergic mIPSCs in uninjured Bax+/+ mice (n = 11) and Bax+/+ mice 2 weeks after SNI (n = 14) and in uninjured Bax−/− mice (n = 8) and Bax−/− mice 2 weeks after SNI (n = 9).

*p < 0.05 and ***p < 0.001 in Student’s t tests comparing uninjured Bax+/+ mice and Bax+/+ mice after SNI; #p < 0.05, ##p < 0.01, and ###p < 0.001 for the comparison between Bax+/+ and Bax−/− mice after SNI. Error bars indicate SEM. See also Figure S5.

Nociceptive behavior of Bax−/− and Bax+/+ mice at baseline did not differ. However, painful hypersensitivity to mechanical (Figures 7A and 7B) or cold stimulation (Figures 7C and 7D) following SNI was mitigated in Bax−/− mice (p < 0.001 in a two-way ANOVA). This reduced neuropathic-pain-like behavior was observed throughout the testing period of 8 weeks. Taken together, the involvement of proapoptotic Bax in spinal disinhibition and persistent pain after SNI provided additional evidence of neurodegeneration contributing to chronic neuropathic pain (Figure S5).

Figure 7.

Disrupting the Intrinsic Apoptotic Pathway Reduces Neuropathic-Pain-like Behavior

(A and C) Withdrawal responses to (A) mechanical (von Frey filaments) or (C) cold stimulation (acetone evaporation) after SNI (n = 9). p < 0.001 for genotype and p < 0.05 for time when responses to the stimulation with von Frey filaments were compared in a two-way ANOVA. p < 0.001 for both genotype and time in the acetone test. Pound signs indicate the results of Bonferroni’s tests following the ANOVAs.

(B and D) AUCs for the withdrawal responses to (B) mechanical or (D) cold stimulation were compared for the total test duration after SNI. Pound signs indicate the results of Student’s t tests.

#p < 0.05, ##p < 0.01, and ###p < 0.001 in Student’s t tests. Error bars indicate SEM.

DISCUSSION

Targeted deletion of functional NMDARs protected against the nerve-injury-induced loss of dorsal horn neurons, suggesting strongly that glutamate excitotoxicity caused the degeneration of these cells. Eliminating NMDAR-dependent transmission at dorsal horn synapses had surprisingly little impact on physiological nociception or the onset of pain hypersensitivity after SNI but disrupted the transition to chronic neuropathic pain. Mechanical and cold allodynia in mice lacking functional NMDARs began to resolve one week after the nerve lesion, whereas pain hypersensitivity in mice with intact glutamatergic transmission persisted for the entire testing period of 2 months. These findings suggest that NMDAR activation plays a fundamentally different role in neuropathic compared to inflammatory pain. Footpad injections of inflammatory irritants, such as capsaicin or formalin, produce NMDAR-dependent central sensitization to nociceptive stimuli within minutes (South et al., 2003), whereas NMDAR activity in the dorsal horn is not required for the maintenance of inflammatory pain (Weyerbacher et al., 2010). Nerve-injury-induced cell death is likely to involve both inhibitory and excitatory dorsal horn neurons. Here, we show that the degeneration of GABAergic neurons has major functional consequences. Neuroprotective NMDAR elimination prevented the loss of GABAergic mIPSCs and chronic pain after SNI. A constitutive deletion of proapoptotic Bax provided a similar safeguard against nerve-injury-induced disinhibition and attenuated pain-like behavior, supporting the conclusion that neurodegeneration promotes persistent neuropathic pain by weakening the inhibitory control of nociceptive signal transmission in the superficial dorsal horn of the spinal cord.

NMDAR activation requires depolarization of the cell membrane to release a voltage-dependent Mg2+ block from the receptor ion channel. Enhanced nociceptive input and ectopically generated activity in injured and neighboring intact nerve fibers increase the recruitment of NMDARs at dorsal horn synapses (Kuner, 2010; Latremoliere and Woolf, 2009). An excitotoxic challenge may result directly from heightened glutamate release by primary afferents (Amir et al., 2005; Wu et al., 2001) or follow the activation of excitatory interneuron circuits within the dorsal horn (Cheng et al., 2017; Peirs et al., 2015; Duan et al., 2014; Lu et al., 2013). Insufficient glutamate uptake and presynaptic facilitation of transmitter release further increase the risk of glutamate accumulation after nerve injury (Yan et al., 2013; Inquimbert et al., 2012). Even a moderate buildup of glutamate may cause neuro-degeneration through the activation of extrasynaptic NMDARs (Bao et al., 2009; Hardingham et al., 2002). Ultimately, excitotoxic cell death is triggered by excess Ca2+ influx through NMDARs and subsequent activation of neuronal nitric oxide synthase (Hara and Snyder, 2007). Our results indicate a downstream engagement of the intrinsic apoptotic pathway, which involves mitochondrial translocation of Bax or Bcl2-antagonist killer 1 (Bak1) followed by cytochrome c release into the cytoplasm and caspase activation (Youle and Strasser, 2008). Bax may further promote excitotoxicity by regulating the Ca2+ exchange between cytosol and endoplasmic reticulum (D’Orsi et al., 2015).

Spatially restricted elimination of NMDAR-mediated transmission was necessary to clarify the functional significance of gluta-mate-induced neurodegeneration in the dorsal horn. NMDARs are expressed at excitatory synapses throughout the CNS. Pharmacological studies based on the use of NMDAR antagonists cannot distinguish between analgesic effects achieved by blocking postsynaptic receptors in spinal or supraspinal pathways involved in the processing of pain or by inhibiting presynaptic receptors at the central terminals of primary afferents (Yan et al., 2013; Suzuki et al., 2002). Interpreting pain-like behavior in mice treated with NMDAR antagonists is further complicated by motor side effects and the attenuation of stress, fear, and depression. Targeted Grin1 deletion in dorsal horn neurons enabled us to reveal the role of neurodegeneration in the spinal cord for the transition from acute to chronic neuropathic pain. Pain after nerve injury appeared to evolve in three phases. (1) In the acute period of approximately 7 days, hypersensitivity to mechanical and thermal stimulation emerged independently of NMDAR expression. The initial increase in pain sensitivity must therefore be mediated primarily by glutamate signaling through AMPARs and group I mGluRs or signaling pathways involving peptide transmitters. In mice with intact NMDARs, withdrawal responses remained high as previously described (Decosterd and Woolf, 2000). (2) In the absence of NMDAR-mediated glutamatergic transmission, pain sensitivity returned to baseline levels within 3–7 weeks, demonstrating that NMDAR activity is required for the establishment of persistent neuropathic pain. The time to complete recovery may delineate the period during which neuropathic pain normally progresses to (3) the final phase of chronic pain. It is unclear why this transition period differed for mechanical and cold allodynia. One possible explanation is the high proportion of cold-sensitive C fiber afferents with increased activity after nerve injury, which may slow the resolution of cold-evoked pain (Jänig et al., 2009; Wu et al., 2001). Loss of GABAergic neurons in laminae I+II will directly affect the inhibitory control of input from these afferents (Peirs and Seal, 2016). In contrast, the survival of inhibitory neurons in laminae III+IV may facilitate the recovery of normal thresholds for mechanically evoked pain, because these interneurons control polysynaptic connections between touch-sensitive A fiber afferents and neurons in lamina I that convey nociceptive input to the brain (Cui et al., 2016; Foster et al., 2015; Petitjean et al., 2015; Duan et al., 2014; Lu et al., 2013). It should be noted that NMDAR deletion may have interfered with mechanisms unrelated to neurodegeneration, for example, synaptic plasticity. This would offer one explanation for the difference between the behavioral phenotypes of Grin1 mice lacking functional NMDARs and Bax− mice. Alternatively, Bax-independent cell death pathways may be responsible for the incomplete prevention of persistent pain-like behavior in Bax− mice (Youle and Strasser, 2008).

The anatomical distribution of nerve-injury-induced neurode-generation suggests differences in the risk of NMDAR-mediated cell death. Apoptosis induction was concentrated in the medial dorsal horn, which receives afferent input from the injured sciatic nerve branches (Corder et al., 2010). Furthermore, a larger proportion of neurons (25%) were lost in the superficial dorsal horn compared to laminae III+IV (14%). Laminae I+II receive afferent input from nociceptive C and Ad nerve fibers. Ectopic activity in these nerve fibers is common after nerve injury (Jänig et al., 2009; Wu et al., 2001). Moreover, non-nociceptive Aβ fibers gain access to superficial dorsal horn layers as normally suppressed polysynaptic pathways open (Cheng et al., 2017; Peirs et al., 2015; Duan et al., 2014; Lu et al., 2013). This additional input involves NMDAR activation, heightening the risk of glutamate toxicity (Torsney and MacDermott, 2006). A lower exposure of inhibitory neurons in laminae III and deeper to such increases in glutamatergic input may explain the preserved integrity of glycinergic inhibition. Neurons releasing glycine, either in conjunction with GABA or as their sole neurotransmitter, reside in this region of the dorsal horn. Their number did not decline after SNI (Lu et al., 2013; Zeilhofer et al., 2005). Our findings correspond to the recently reported survival of parvalbumin-expressing inhibitory neurons at the border between inner lamina II and III (Petit-jean et al., 2015) and the pattern of neurodegeneration in the dorsal horn observed after spinal nerve ligation (Yowtak et al., 2013).

Similar to immunohistochemical studies of the rat dorsal horn (Todd, 2010), we identified 35% of neurons in laminae I+II and 40% in laminae III+IV of uninjured mice as inhibitory neurons based on the expression of fluorescent reporter proteins under the Gad1 or Viaat promoter. Degeneration of these important gatekeepers of nociceptive pathways has substantial functional consequences, as the decrease in inhibitory currents after SNI demonstrates. It is not too surprising that the reduction of mIPSCs was proportionally larger than the loss of GABAergic neurons. Stereological neuron counts do not reflect the extent of changes in synaptic connections or their functional implications. They do not reveal the decay of dendrites, dendritic spines, or axon terminals associated with neurodegeneration (Lorenzo et al., 2014; Bao et al., 2009). Moreover, excess gluta-mate may have impaired the functioning of surviving neurons.

The neuroprotective effect of α-phenyl-N-tert-butylnitrone, a compound that traps free radical species and inhibits their synthesis, suggests that oxidative stress is one consequence of NMDAR-mediated excitotoxicity in the dorsal horn (Yowtak et al., 2013). Other mechanisms are likely to contribute to the decrease in synaptic inhibition after nerve injury. Brain-derived neurotrophic factor (Bdnf) released from microglia provokes downregulation of K+-Cl− cotransporter 2 (KCC2). Reduced KCC2 expression, phosphorylation, and cleavage of the transporter by the Ca2+-dependent protease calpain lead to a rise in intracellular Cl− that weakens the hyperpolarizing efficacy of GABAA receptor activation (Kahle et al., 2016; Zhou et al., 2012; Coull et al., 2005). The functional significance of these mechanisms changes with time. Microglial activity, for example, is high during the onset of neuropathic pain. Consequently, KCC2 regulation by Bdnf is probably less relevant for chronic pain (Peng et al., 2016). In contrast, the NMDAR-dependent degeneration of GABAergic neurons contributes specifically to the persistence of neuropathic pain.

We conclude that chronic neuropathic pain resembles, in a key aspect of its pathophysiology, a neurodegenerative disorder. Cell death induction after nerve injury appears to be a protracted process in which a relatively low level of apoptosis leads to a sizable decrease in the number of dorsal horn neurons over time. This slow progression is comparable to neurodegeneration in other diseases associated with chronic glutamate toxicity, such as amyotrophic lateral sclerosis (Lewerenz and Maher, 2015). The irreversible loss of GABAergic neurons causes a major shift in the normally carefully regulated balance between inhibitory and excitatory modulation of pain processing. This long-term deficit in the control of nociceptive signaling is likely to contribute to the difficulty of treating neuropathic pain. On the other hand, neuroprotection, which we accomplished through targeted elimination of NMDAR-mediated glutamatergic transmission and deletion of proapoptotic Bax, may provide an opportunity for the prevention of chronic pain. Mechanism-based treatment approaches for neuropathic pain are urgently needed (Colloca et al., 2017). Our findings strengthen the rationale for strategies aiming to restore GABAergic inhibition through gene transfer (Hao et al., 2005) or the transplantation of GABAergic neuron precursors (Bráz et al., 2012).

EXPERIMENTAL PROCEDURES

In compliance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines, mice were randomly allocated to the experimental conditions and investigators blind to the allocation. A detailed description of the Experimental Procedures is provided in the Supplemental Information.

Animals

This study was conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committees (IACUCs) of Columbia University and Boston Children’s Hospital.

Experiments were performed with male adult (8–12 weeks old) animals. C57BL/6 and Bax− mice (Knudson et al., 1995) were purchased from Jackson Laboratory. Grin1, Gad1-GFP, and Viaat-Venus mice have been described (Wang et al., 2009; Tamamaki et al., 2003; Tsien et al., 1996).

Surgery

For SNI, we ligated and transected the common peroneal and tibial branches of the sciatic nerve (Decosterd and Woolf, 2000). Mice received 2 unilateral (left) stereotaxic injections of AAV8 vectors expressing GFP or GFP-Cre into the dorsal horn, cranial, and caudal of spinal cord segment L4, as previously described (Inquimbert et al., 2013).

TUNEL Assay

We used ApopTag In Situ Apoptosis Detection Kits (EMD Millipore) for TUNEL on cryosections (10 μm) of the dorsal horn. Only profiles exhibiting chromatin changes indicative of apoptosis (pyknosis, fragmentation, and marginalization) were counted.

Stereology

For the stereological analysis of total neurons, we cut transverse cryosections (50 μm) of the spinal cord. From a random start, we selected every 4th section and stained them for NeuN (EMD Millipore). Neurons were counted in laminae I+II and laminae III+IV of the medial half of the dorsal horn. Dorsal and ventral boundaries of these regions of interest (ROIs) were defined based on cytoarchitecture. The number of neurons was determined using the optical fractionator (West et al., 1991), with counting spaces distributed in a systematic-random fashion (Table S1). The analysis was performed with Stereo Investigator software (MBF Bioscience).

Inhibitory neurons expressing GFP or Venus were counted in sections of 30-μm thickness. We selected every 5th section and immunostained them for NeuN and the γ isoform of protein kinase C (PKCγ) (Santa Cruz Biotechnology), using fluorescently labeled secondary antibodies for detection. We delineated the same ROIs as for the counts of total neurons based on cytoarchitecture and PKCγ expression, which demarcates the boundary between laminae II and III (Table S1).

qPCR

RNA was purified from the ipsilateral dorsal quadrants of the L4 spinal cord. qPCR was performed with a SYBR Select Master Mix (Thermo Fisher Scientific) except for Glyt2 (Slc6a5). For Slc6a5, we utilized a TaqMan Universal Mastermix II (Thermo Fisher Scientific) and FAM dye-labeled minor groove-binding (MGB) probes. Targets were normalized to glyceraldehyde-3-phosphate dehydrogenase (Gapdh). Table S2 lists the forward and reverse primers. All reactions were performed on a StepOnePlus System (Thermo Fisher Scientific).

ChIP

For ChIP assays, we combined the left (ipsilateral) dorsal quadrants of the L4 spinal cord of two mice into one sample. ChIP was performed as previously described (Korb et al., 2015). Three technical replicates were averaged for each biologically independent sample and DNA levels normalized to input. Table S2 lists the forward and reverse primers.

In Situ Hybridization

Riboprobes of 911 bp length were generated from a Grin1 cDNA template and labeled with digoxigenin (Roche Life Science). Probes bound to spinal cord cryosections (10 μm) were immunostained with antigen-binding fragments (Fabs) conjugated to alkaline phosphatase (Roche Life Science). Probes in sense orientation served as negative controls.

Electrophysiology

Recordings were performed on acute spinal cord slices as described (Inquim-bert et al., 2012; Tong and MacDermott, 2014). We obtained voltage-clamp whole-cell patch recordings from dorsal horn neurons in lamina II and the border between laminae II and III. We recorded mIPSCs at a holding potential of −70 mV in the presence of 0.5 μM tetrodotoxin (TTX) and ionotropic glutamate receptor blockers, either 2 mM kynurenic acid or 50 μM D-2-amino-5-phosphonovalerate (AP5) and 10 μM 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX). To isolate GABAergic or glycinergic currents, we used 1 μM strychnine HCl or 10 μM bicuculline methiodide, respectively. Miniature EPSCs were recorded at −70 mV in the presence of 0.5 μM TTX, 1 μM strychnine, and 10 μM bicuculline. Currents evoked by 100 μM NMDA were recorded at +40 mV in the presence of 25 mM NBQX. We used pClamp10 software (Molecular Devices) for data acquisition and Mini Analysis software (Synaptosoft) or the Win EDR and Win WCP programs of Strathclyde Electro-physiology Software for analysis.

Behavioral Tests

All behavioral evaluations were replicated in two independent groups of animals. We applied calibrated von Frey filaments in the territory of the spared sural nerve to determine the mechanical withdrawal threshold. To test for cold sensitivity, we recorded the withdrawal duration after applying a drop of acetone to the skin, which produces a cool sensation upon evaporation (Inquimbert et al., 2012). Technical details of the tests for mechanical hyperalgesia and heat-induced pain are provided in the Supplemental Experimental Procedures.

Statistics

We used an unpaired two-sided Student’s t test for comparing two experimental conditions, a one-way ANOVA followed by Dunnett’s or Tukey’s test for comparing more than two conditions, and a two-way ANOVA followed by Sidak’s or Bonferroni’s test for comparing experimental conditions over time. We utilized Prism software (GraphPad) for these tests. Sample sizes provided ≥90% power for statistically significant results at α = 0.05, as calculated with G*Power software (Heinrich Heine University Düsseldorf). Unless otherwise mentioned, data are represented as mean ± SEM.