Enhancing spinal cord stimulation-induced pain inhibition by augmenting endogenous adenosine signalling after nerve injury in rats

Abstract

Background

The mechanisms for spinal cord stimulation (SCS) to alleviate chronic pain are only partially known. We aimed to elucidate the roles of adenosine A1 and A3 receptors (A1R, A3R) in the inhibition of spinal nociceptive transmission by SCS, and further explored whether 2′-deoxycoformycin (dCF), an inhibitor of adenosine deaminase, can potentiate SCS-induced analgesia.

Methods

We used RNAscope and immunoblotting to examine the distributions of adora1 and adora3 expression, and levels of A1R and A3R proteins in the spinal cord of rats after tibial-spared nerve injury (SNI-t). Electrophysiology recording was conducted to examine how adenosine receptor antagonists, virus-mediated adora3 knockdown, and dCF affect SCS-induced inhibition of C-fibre-evoked spinal local field potential (C-LFP).

Results

Adora1 was predominantly expressed in neurones, whereas adora3 is highly expressed in microglial cells in the rat spinal cord. Spinal application of antagonists (100 μl) of A1R (8-cyclopentyl-1,3-dipropylxanthine [DPCPX], 50 μM) and A3R (MRS1523, 200 nM) augmented C-LFP in SNI-t rats (DPCPX: 1.39 [0.18] vs vehicle: 0.98 [0.05], P=0.046; MRS1523: 1.21 [0.07] vs vehicle: 0.91 [0.03], P=0.002). Both drugs also blocked inhibition of C-LFP by SCS. Conversely, dCF (0.1 mM) enhanced SCS-induced C-LFP inhibition (dCF: 0.60 [0.04] vs vehicle: 0.85 [0.02], P<0.001). In the behaviour study, dCF (100 nmol 15 μl⁻¹, intrathecal) also enhanced inhibition of mechanical hypersensitivity by SCS in SNI-t rats.

Conclusions

Spinal A1R and A3R signalling can exert tonic suppression and also contribute to SCS-induced inhibition of spinal nociceptive transmission after nerve injury. Inhibition of adenosine deaminase may represent a novel adjuvant pharmacotherapy to enhance SCS-induced analgesia.

Editor's key points.

• •Spinal cord stimulation (SCS) has been used for decades for severe refractory pain, but optimisation of its efficacy has been hampered by the poor understanding of its mechanisms of action.
• •Among potential neurochemical mechanisms, spinal adenosine A1 and A3 receptors (A1R, A3R) might play a leading role by modulating inhibition of spinal nociceptive transmission by SCS.
• •In this experimental study conducted in nerve-injured rats, activation of A1R and A3R in the spinal cord are shown to be important mechanisms of SCS-induced analgesia.
• •Combining an inhibitor of adenosine deaminase as adjuvant pharmacotherapy with SCS may represent a new strategy to enhance SCS-induced analgesia.

The treatment of neuropathic pain remains challenging and impacts millions of individuals across the world.¹ Spinal cord stimulation (SCS) has been used for the treatment of radiculopathy and complex regional pain syndromes for decades. However, the conventional SCS (50–60 Hz, 0.2 ms, paraesthesia intensity) often provides suboptimal efficacy for a subset of patients.^1,2 The clinical advance that would enhance its efficacy and broaden indications has been hampered because of our limited understanding of the mechanisms underlying its pain inhibitory effects.^2,3

Recent insights from mechanistic studies suggest that multiple neurochemical mechanisms including GABA, serotonin, acetylcholine, cannabinoid, and adenosine are be involved in the therapeutic effects of SCS.1, 2, 3 Among them, the G_i/o-coupled adenosine A1 and A3 receptors (A1R, A3R) are promising targets for pain control.4, 5, 6 However, clinical translation has been delayed as the roles of adenosinergic signalling pathways are more complex than once thought and can either promote or inhibit pain, depending on the site of action and specific receptor involved.7, 8, 9 Importantly, the details of spinal adenosinergic signalling in SCS-induced analgesia remain partially known. Most nociceptive afferent fibres (C-fibres) terminate in the superficial dorsal horn, and measuring local field potential (LFP) in this region that corresponds to C-fibre inputs (C-LFP) has been a well-established approach to examine spinal nociceptive transmission in vivo.^10,11 Accordingly, we aimed to delineate the roles of spinal A1R- and A3R-signalling in modulating the inhibition of spinal nociceptive transmission by SCS, by recording of spinal LFPs in a rat model of neuropathic pain.

Combining adjuvant pharmacotherapy with SCS is a promising strategy to improve pain treatment,¹² and intrathecal administration of adenosine was shown to increase pain inhibition from conventional SCS.¹³ However, the beneficial effect of adenosine was short-lived and often associated with dose-limiting side-effects.^8,14 As SCS may modulate pain through delicate neuronal networks, we speculate that combining exogenous agonists with SCS might ‘flood’ different neuronal circuitries, and thus cause adverse effects by affecting circuitries that serve other biological functions. Intriguingly, adenosine is released into the spinal cord during SCS.¹⁵ Endogenously produced adenosine might alleviate pain without causing notable side-effects,^4,16 presumably because it is mainly released into ‘pain circuitry’. In light of this, we further tested whether fine-tuning spinal adenosinergic signalling by combining SCS with an adjuvant drug 2′-deoxycoformycin (dCF) that inhibits the degradation of endogenous adenosine,¹⁷ may enhance SCS-induced inhibition of C-LFP and neuropathic pain-like behaviour in nerve-injured rats.

Methods

Additional methods and details are presented in the supplementary material.

Animals

All studies were performed on adult male Sprague–Dawley rats (6–8 weeks old, Charles River Laboratories, Wilmington, MA, USA). All procedures were approved by the Johns Hopkins University Animal Care and Use Committee (Baltimore, MD, USA).

Tibial-spared nerve injury model

The tibial-spared nerve injury (SNI-t) model of neuropathic pain was performed as previously described.¹⁸ Briefly, the common peroneal and sural nerves were cut, but the tibial nerve was left intact.

Behavioural tests

Mechanical pain test using electronic von Frey (e-VF) was used to detect mechanical hypersensitivity induced by nerve injury, and paw withdrawal threshold (PWT) was measured as previously described.¹⁹ Mechanical allodynia test using Von Frey monofilament fibres was used to examine the inhibitory effects of SCS on mechanical pain sensitivity, using the up-down method as previously described.²⁰ Hargreave's test was conducted to examine heat pain sensitivity using the Plantar Analgesia Meter (IITC, Life Science, Woodland Hills, CA, USA) as described.²¹

Electrophysiologic recording of spinal local field potential

The experimental setup for in vivo spinal LFP recording was similar to our previous study.¹¹ Briefly, the lumbar spinal cord was exposed in rats anaesthetised with urethane (1.2 g kg⁻¹, i.p.) and the dura mater was partially removed at the recording segments (L4). The parylene-coated tungsten microelectrode (3 mΩ, Frederick Haer Company, Brunswick, ME, USA) was inserted into the superficial dorsal horn (200–500 μm below the surface) at the L4 spinal segment. The ipsilateral sciatic nerve was exposed and placed on the hook electrode to deliver the test stimulus (5 mA, 0.2 ms, 1 test 1 min⁻¹), evoking the spinal LFP. A real-time, computer-based data acquisition and processing system (CED Spike2, Cambridge Electronics Design, Cambridge, UK) was used to collect analogue data. Raw data were collected at a sampling rate of 1000 Hz. The data stream was amplified and then filtered (0.1–100 Hz, model DAM80; World Precision Instruments, Sarasota, FL, USA), and artifacts of stimulation were removed online by a notch filter (IIR filters of the CED 1401 data acquisition system).

LFP was examined before SCS (baseline, 10 min) and at 0–30 min after SCS. Based on the conduction velocity and activation threshold, LFPs that correspond to activated A- and C-fibres can be distinguished. In comparison to the large A-LFP, the C-LFP exhibits a longer latency (90–130 ms), smaller amplitude, and a higher activation threshold. Offline modulus measurements of Spike2 software were used to calculate the peak amplitude of A-LFP and the area under curves (AUC) of C-LFP. This method has been used in our previous studies and also by others in the analysis of spinal LFPs.^11,22,23 A comparison was made between pre-SCS and post-SCS conditions within and between the groups after normalising LFP to pre-SCS baseline values in each animal.

Spinal cord stimulation for in vivo electrophysiology

The methodology was described in our previous studies.^11,22 The bipolar SCS electrode, shortened from the quadripolar lead used in the behavioural study, was positioned after the laminectomy surgery for spinal LFP recording. The rats were anaesthetised with urethane (1.2 g kg⁻¹, i.p.). A laminectomy was performed and then the bipolar SCS electrode was placed epidurally over the T13-L1 spinal level which was identified based on the corresponding T11 vertebrate. The electrode was connected to an external stimulator (model 2100, A-M, Sequim, WA, USA) and provided SCS (50 Hz, 0.2 ms, 50% or 100% Ab1, 5 min) as in previous studies.^11,22,24

Statistical analysis

Graphing and statistical analysis were undertaken with GraphPad version 8.0 (San Diego, CA, USA). The methods for statistical comparisons in each study were indicated in the figure legends. The criterion for statistical significance was set to P<0.05. The sample size was determined based on similar previous studies of SCS.^11,20,25

Results

Distributions of adora1 and adora3 in spinal neurones and glial cells

Using the highly selective RNAscope in situ hybridisation combined with co-immunostaining of cell-type specific markers, we examined the distribution of adora1 and adora3 mRNA expression in the spinal cord dorsal horn (Fig. 1a–d). Adora1 was expressed in a high proportion in both Pax2-labeled inhibitory interneurones (86.5%) and VGlut2-labelled excitatory neurones (97.1%, Fig. 1a and b). Comparatively, a much smaller percentage of neurones in each subpopulation expressed adora3 (Pax2: 26.7%, VGlut2: 10.4%). In glial cells, adora1 was expressed at low levels in both astrocytes labelled with glial fibrillary acidic protein (GFAP) (<10%, Fig. 1c) and microglial cells labelled with Iba1 (12.8%, Fig. 1d). Strikingly, adora3 was highly expressed in microglial cells (73.9%), but not in astrocytes (6.1%).

Fig 1.

Distribution of adenosine A1 receptor (A1R) and A3 receptor (A3R) messenger RNAs in the spinal cord dorsal horn and changes in their protein expression in rats after SNI-t. (a–d) Representative images showing co-staining and quantification of adora3 (red), adora1 (magenta), and immunoreactivities of Pax2 (a, an inhibitory interneurone marker), VGlut2 (b, an excitatory neurone marker), GFAP (c, an astrocyte marker), and Iba1 (d, a microglia marker). (a–c) A high-power view of the box region in a–d, with arrowheads indicating double-labelled cells. Right: Quantification of the percentage of double-labelled cells in the total number of cells identified by each cell-type marker. Scale bar: a–e (100 μm), a–c (50 μm). N=3 rats. (e) The schematic for SNI-t surgery (left) and timeline of pain behavioural tests. (f and g) The decreases of paw withdrawal threshold to mechanical stimulation (f) and paw withdrawal latency to heat stimulation (g) in the ipsilateral hind paw at Day 14 after SNI-t. N=12. Two-way mixed model analysis of variance with Bonferroni post hoc test. (h–k) Representative immunoblotting images and quantification of protein levels of A1R (h), A3R (i), adenosine deaminase (ADA, j), and GFAP (k) in the ipsilateral lumbar spinal cord on Day 14 after SNI-t or sham surgery (control). N=8/group. GFAP, glial fibrillary acidic protein; SNI-t, tibial-spared nerve injury. Unpaired t-test, ∗P<0.05. ∗∗P<0.01, ∗∗∗P<0.001 vs baseline, ^##P<0.01, ^###P<0.001 vs contralateral side.

A1 receptor and A3 receptor protein levels were decreased in the spinal cord after tibial-spared nerve injury

The ipsilateral PWT to mechanical stimulation and paw withdrawal latency to radiant heat stimulation were significantly decreased in rats at Day 14 after SNI-t, compared with baseline and the contralateral side (Fig. 1e–g), suggesting the development of mechanical and heat hypersensitivity. Accordingly, immunoblotting experiments were conducted at this time point when neuropathic pain has reached a peak level in this model.^18,26,27 Both A1R and A3R expression was decreased in the ipsilateral lumbar spinal cord, compared with that after sham surgery (Fig. 1h and i). The adenosine deaminase (ADA) level was also decreased (Fig. 1j), whereas GFAP expression was upregulated after SNI-t (Fig. 1k).

Activation of A1 receptors exerted tonic inhibitory effects and contributed to suppression of C-fibre-evoked spinal local field potentials by spinal cord stimulation after nerve injury

Functionally, we first examined whether endogenous A1R signalling may exert a tonic inhibition of spinal nociceptive transmission after nerve injury. Spinal LFP was evoked by a high-intensity test stimulus (5 mA, supra-C-fibre activation threshold) applied at the sciatic nerve in rats at 2–3 weeks after SNI-t (Fig. 2a), and can be separated into an earlier A-fibre component and a later C-fibre component (Fig. 2b). Group comparison showed that spinal topical application of an A1R antagonist 8-cyclopentyl-1,3-dipropylxanthine (DPCPX) (50 μM, 100 μl) significantly augmented the AUC of C-LFP (Fig. 2d–f, average C-LFP post-DPCPX vs post-vehicle: 1.39 (0.18) vs 0.98 (0.05), P=0.046), and also increased the amplitude of A-LFP in SNI-t rats (Fig. 2g, average A-LFP post-DPCPX vs post-vehicle: 1.20 (0.03) vs 1.04 (0.04), P=0.004).

Fig 2.

Activation of adenosine A1 receptor (A1R) exerted tonic inhibition and contributed to spinal cord stimulation (SCS)-induced inhibition of C-local field potential (C-LFP) in SNI-t rats. (a) The schematic diagram for recording of spinal LFP to electrical test stimulation at the sciatic nerve, and for recording the antidromic sciatic compound action potential (CAP) evoked by epidural SCS in rats on Day 14 after SNI-t. (b) Representative trace for evoked LFP to A-fibre inputs (A-LFP) and C-fibre inputs (C-LFP). (c) Representative trace of sciatic antidromic CAP evoked by increasing intensity of SCS (single pulse). We determined online the current thresholds of SCS that elicited the first detectable Aα/β-waveform (Ab0), and the peak Aα/β-waveform (Ab1) without inducing the Aδ-waveform. (d) Timeline for LFP recording with spinal topical application of the vehicle, followed by DPCPX (50 μM, 100 μl). (e) Representative trace of spinal LFP before and 30 min after vehicle or DPCPX application. (f and g) The area under the curve (AUC) of C-LFP (f) and the amplitude of A-LFP (g) during each 5-min period after vehicle and DPCPX application was averaged for analysis. N=8 per group. (h) Protocol of examining the effects of SCS on LFP with spinal topical application of vehicle or DPCPX (50 μM, 100 μl, 10 min pretreatment). (i) Representative trace of LFP before and after SCS (50 Hz, 100% Ab1, 5 min) with the vehicle and DPCPX pretreatment. (j and k) The AUC of C-LFP (j) and the amplitude of A-LFP (k) during each 5-min period after SCS were averaged for analysis. N=8 per group. f–k: Two-way repeated measures analysis of variance with Bonferroni post hoc test. DPCPX, 8-cyclopentyl-1,3-dipropylxanthine; SNI-t, tibial-spared nerve injury. ∗P<0.05, ∗∗P<0.01 vs pre-drug and pre-SCS (time 0), ^#P<0.05, ^###P<0.001 vs vehicle.

We next examined whether activation of A1R also contributes to SCS-induced inhibition of C-LFP. To select the intensity of SCS, SCS that results in a peak Aα/β-waveform without inducing an Aδ-waveform (Ab1) was first determined in each experiment by recording the sciatic compound action potentials (Fig. 2c), as shown previously.^11,28 Compared with the pre-SCS baseline, the first SCS (50 Hz, 100% Ab1, 5 min, with vehicle pretreatment) significantly decreased C-LFP at 5 min and 10 min post-SCS, but did not affect A-LFP (Fig. 2h–k). However, inhibition of C-LFP by the second SCS with DPCPX pretreatment (50 μM, 100 μl) was diminished (Fig. 2i and j, average C-LFP post-SCS, DPCPX vs vehicle: 1.06 (0.04) vs 0.79 (0.06), P=0.046). In separate experiments, we confirmed that the inhibition of C-LFP by the second SCS with vehicle pretreatment was comparable to that from the first SCS (Supplementary Fig. S1). Collectively, these findings suggest that DPCPX blocked the inhibition of C-LFP by SCS.

Activation of A3 receptors also exerted tonic inhibition of spinal C-fibre-evoked spinal local field potential after nerve injury and contributed to suppression of C-fibre-evoked spinal local field potentials by spinal cord stimulation

Compared with the vehicle, spinal topical application of an A3R antagonist MRS1523 (200 nM, 100 μl) significantly augmented the AUC of C-LFP (Fig. 3a–c, average C-LFP post-MRS1523 vs post-vehicle: 1.21 (0.07) vs 0.91 (0.03), P=0.002), but did not affect the amplitude of A-LFP in SNI-t rats (Fig. 3d, average A-LFP post-MRS1523 vs post-vehicle: 0.96 (0.02) vs 0.92 (0.05), P=0.506). In separate experiments, the inhibition of C-LFP by SCS (50 Hz, 100% Ab1, 5 min) was blocked by MRS1523 (200 nM, 100 μl, 10 min pretreatment), compared with the vehicle (Fig. 3e–g, average C-LFP post-SCS, MRS1523 vs vehicle: 1.07 (0.13) vs 0.74 (0.04), P=0.002). Whereas A-LFP amplitude was not significantly changed (Fig. 3h, average A-LFP post-SCS, MRS1523 vs vehicle: 1.03 (0.04) vs 0.95 (0.07), P=0.387).

Fig 3.

Activation of adenosine 3 receptor (A3R) also contributed to tonic inhibition and spinal cord stimulation (SCS)-induced inhibition of C-local field potential (C-LFP) in SNI-t rats. (a) Timeline for LFP recording with spinal topical application of the vehicle, followed by MRS1523 (200 nM, 100 μl) in rats on Day 14 after SNI-t. (b) Representative trace of LFP before and 30 min after vehicle or MRS1523 application. (c-d) The area under the curve (AUC) of C-LFP (c) and the amplitude of A-LFP (d) during each 5-min period after vehicle and MRS1523 application was averaged for analysis. N=8 per group. (e) Protocol of examining the effects of SCS on LFP with the pretreatment of the vehicle or MRS1523 (200 nM, 100 μl, 10 min pretreatment). (f) Representative traces of LFP before and after SCS (50 Hz, 100% Ab1, 5 min) with vehicle and MRS1523 pretreatment. (g–h) The AUC of C-LFP (g) and the amplitude of A-LFP (h) during each 5-min period after SCS were averaged for analysis. N=8 per group. c–h: Two-way repeated measures analysis of variance with Bonferroni post hoc test. SNI-t, tibial-spared nerve injury. ∗P<0.05, ∗∗P<0.01 vs pre-drug and pre-SCS (time 0), ^#P<0.05, ^##P<0.01 vs vehicle.

Downregulation of spinal A3 receptor expression attenuated inhibition of C-fibre-evoked spinal local field potentials by spinal cord stimulation

In addition to the pharmacological approach, we further utilised AAV-U6-shRNA(adora3)-GFP-mediated gene knockdown strategy to determine the role of A3R in SCS-induced inhibition of C-LFP (Fig. 4a).

Fig 4.

Knockdown of adora3 in the spinal cord blocked inhibition of C-local field potential (C-LFP) by spinal cord stimulation (SCS) in SNI-t rats. (a) The schematic diagram of the experimental protocol. AAV-U6-shRNA(adora3)-GFP or the control vector AAV-U6-shRNA (SCRM)-GFP was injected into the spinal dorsal horn after conducting a small laminectomy in naive rats. Rats then received SNI-t surgery on the same side (left) 14 days after the virus injection. Spinal LFP was recorded at 14 days after SNI-t. (b) A representative image showing the colocalisation of GFP (green) with adenosine A3 receptor (A3R) immunoreactivity (red) in the dorsal horn at 14 days after virus vector injection in a naive rat. [b] A high-power view of the box region in [a], with arrowheads denoting double-labelled cells. Scale bar: 100 μm. (c–f) Quantification of messenger RNA and protein levels of adora3 (c–d) and adora1 (e–f) in the spinal cord of SNI-t rats 28 days after intraspinal injection of AAV-U6-shRNA(adora3)-GFP (AAV-shAdora3), compared with rats that received the control vector (AAV-SCRM). N=4–5 per group. Unpaired t-test. (g) Representative trace of spinal LFP before and after SCS (50 Hz, 100% Ab1, 5 min) in SNI-t rats pretreated with intraspinal injections of AAV-shAdora3 or AAV-SCRM. (h and i) The area under the curve (AUC) of C-LFP (h) and the amplitude of A-LFP (i) during each 5 min after SCS in each group was averaged for analysis. N=8 per group. Two-way mixed model analysis of variance with Bonferroni post hoc test. SNI-t, tibial-spared nerve injury. ∗P<0.05 vs pre-SCS (time 0), ∗∗P<0.01, ^#P<0.05 vs AAV-SCRM control.

In naive rats at Day 14 after intraspinal injections of AAV-U6-shRNA (SCRM)-GFP (control vector), the co-localisation of A3R immunoreactivity and GFP signal was broadly observed in the spinal cord dorsal horn (Fig. 4b[a]), suggesting the successful transfection of A3R-expressing cells by the virus vectors. Double-labelled cells include both Iba1-negative and Iba1-positive cells (Fig. 4b[b]). Thus, the virus vectors non-selectively infected neuronal and non-neuronal cells expressing adora3. Importantly, quantitative polymerase chain reaction (qPCR) and immunoblot studies showed that both adora3 and A3R protein levels were significantly decreased in the lumbar spinal cord of SNI-t rats at 28 days after AAV-U6-shRNA(adora3)-GFP injection, compared with the control vector (Fig. 4c and d). In contrast, adora1 and A1R levels were not significantly changed (Fig. 4e and f). These findings demonstrated the efficiency of selective downregulation of A3R expression by intraspinal injection of AAV-U6-shRNA(adora3)-GFP in vivo.

Functionally, at Day 14 post-SNI-t (28 days after virus injection), inhibition of C-LFP by SCS (50 Hz, 100% Ab1, 5 min) was diminished in rats that previously received AAV-U6-shRNA(adora3)-GFP, compared with control (Fig. 4g and h). The A-LFP was not significantly changed by SCS in either group (Fig. 4i).

An adenosine deaminase inhibitor enhanced inhibition of C-fibre-evoked spinal local field potentials by spinal cord stimulation in tibial-spared nerve injury rats

Adenosine deaminase is a critical enzyme that irreversibly converts adenosine to inosine, and thus reduces adenosine level and terminates adenosine signalling. Next, we tested whether limiting the degradation of endogenous adenosine by dCF (Pentostatin, Fig. 5a), an FDA-approved selective and potent ADA inhibitor, can enhance the effect of SCS. At Day 14 after SNI-t, dCF (0.01, 0.1 mM, 100 μl) concentration-dependently inhibited C-LFP, but not A-LFP, from 5 to 10 min after spinal topical application, compared with vehicle (Fig. 5b–d). Notably, a pretreatment of dCF 0.1 mM markedly enhanced and prolonged the inhibition of C-LFP by SCS (Fig. 5e–h). A-LFP was, however, not significantly changed after SCS in both groups. In this experiment, to avoid a potential ceiling effect, SCS was applied at a low intensity (50% Ab1) which produced minimal inhibition of C-LFP in rats that received vehicle (Fig. 5g and h).

Fig 5.

Spinal application of dCF enhanced inhibition of C-local field potential (C-LFP) by spinal cord stimulation (SCS) inSNI-trats. (a) a. The schematic diagram shows that dCF blocks the degradation of adenosine into inosine by inhibiting adenosine deaminase (ADA). b. Timeline for LFP recording before and after spinal topical application of the vehicle or dCF in rats on Day 14 after SNI-t. (b) Representative trace of spinal LFP before and after vehicle and dCF (0.01, 0.1 mM, 100 μl) treatment. (c and d) The area under the curve (AUC) of C-LFP (c) and the amplitude of A-LFP (d) during each 5-min period after vehicle and dCF (0.01, 0.1 mM, 100 μl) treatment was averaged for analysis. N=6 per group. Two-way mixed model analysis of variance with Bonferroni post hoc test. ∗P<0.05 vs pre-drug (time 0); ^#P<0.05, ^##P<0.01 vs vehicle. (e) The schematic diagram illustrates that dCF may enhance SCS-induced inhibition by preventing adenosine from degrading into inosine. (f) Timeline for LFP recording in response to SCS with vehicle or dCF pretreatment. (g) Representative traces of LFP before and after SCS with the vehicle and 0.1 mM dCF pretreatment (10 min). (h and i) The AUC of C-LFP (h) and the amplitude of A-LFP (i) during each 5-min period before and after low-intensity SCS (50 Hz, 50% Ab1, 5 min) with vehicle or dCF (0.1 mM) pretreatment. N=8 per group. Two-way repeated measures analysis of variance with Bonferroni post hoc test. dCF, 2′-deoxycoformycin; SNI-t, tibial-spared nerve injury. ∗P<0.05, ∗∗P<0.01, ∗∗∗P<0.001 vs pre-SCS (time 0), ^#P<0.05, ^##P<0.01, ^###P<0.001 vs vehicle.

Intrathecal injection of 2′-deoxycoformycin enhanced inhibition of mechanical hypersensitivity by spinal cord stimulation

We further conducted animal behavioural tests to determine whether dCF can also enhance pain inhibition by low-intensity SCS (Fig. 6a and b). At 2–3 weeks after nerve injury, rats first received an intrathecal injection of vehicle or low-dose dCF (100 nmol, 15 μl) which did not significantly change PWT at 30 min post-injection, compared with pre-drug level. Rats then received SCS (50 Hz, 50% motor threshold [MoT], 0.2 ms) for 30 min. Compared with the vehicle, dCF pretreatment potentiated the inhibition of mechanical hypersensitivity by low-intensity SCS (Fig. 6c). The percentage of maximum possible effect of SCS was significantly higher in the dCF-treated group than that in the vehicle-treated group (Fig. 6d).

Fig 6.

Intrathecal injection of dCF enhanced pain inhibition by low-intensity spinal cord stimulation (SCS) in nerve-injured rats. (a) The schematic diagram shows the experimental timeline. (b) Paw withdrawal threshold (PWT) to mechanical stimulation was measured before injury (baseline), on Day 7 (pre-lead implantation), and on Day 12 post-injury (Day 4 post-lead implantation). N=7. One-way analysis of variance with Bonferroni post hoc test. ∗∗∗P<0.001 vs pre-injury. (c) PWTs were measured before injury (baseline), before and 30 min after intrathecal injection of vehicle or dCF (100 nmol, 15 μl) at 2 weeks post-injury, followed by low-intensity SCS (50 Hz, 50% motor threshold, 0.2 ms, 30 min). PWTs were then measured at 15 min during SCS (intra-SCS), and at 5 and 30 min after completing the SCS (post-SCS). N=8 per group. Two-way repeated measures analysis of variance with Bonferroni post hoc test, ∗P<0.05, ∗∗∗P<0.001 vs pre-injury baseline, ^&P<0.05 vs pre-drug, ^#P<0.05 vs vehicle. (d) The maximum possible effect (%) of SCS in each group. N=8 per group. Unpaired t-test, ∗P<0.05. (e) Working hypothesis. a. Endogenous A1R and A3R signalling may exert a tonic inhibition of spinal nociceptive transmission under neuropathic pain conditions. SCS increases the release of adenosine into the spinal cord, which further attenuates spinal nociceptive transmission by activating A1R and A3R expressed in both neurones and glial cells. At the spinal level, activating A1R and A3R on both presynaptic nerve terminals and postsynaptic dorsal horn neurones can attenuate synaptic transmission. Yet, it remains unclear how A1R and A3R expressed on spinal glial cells might serve in pain modulation by SCS. b. Inhibition of adenosine deaminase (ADA) activity by dCF would reduce adenosine degradation into inosine and thus increase/prolong adenosine signalling and enhance SCS-induced pain inhibition. dCF, 2′-deoxycoformycin; SNI-t, tibial-spared nerve injury.

Discussion

Spinal cord stimulation is a promising therapeutic approach in pain medicine, but the neurochemical mechanisms involved in pain inhibition by SCS are partially understood. In this study, we demonstrated that activations of A1R and A3R in the spinal cord are important to SCS-induced inhibition of C-LFP in SNI-t rats. Moreover, the inhibition of spinal C-LFP and mechanical hypersensitivity by SCS in nerve-injured rats was potentiated by dCF, which increases endogenous adenosine signalling through inhibiting ADA. Current findings suggest that combining an ADA inhibitor as adjuvant pharmacotherapy with SCS could present a new strategy to enhance SCS-induced analgesia (Fig. 6e).

Activation of A1 receptor and A3 receptor contributed to tonic inhibition of spinal C-fibre-evoked spinal local field potential after nerve injury

A1R and A3R are encoded by the adora1 and adora3 genes, the expression profiles of which remain unclear in the spinal neurones and non-neuronal cells. Our RNAscope in situ hybridisation study showed that adora1 is highly expressed in spinal neurones, including both VGlut2-labelled excitatory neurones and Pax2-labeled inhibitory interneurones, but only in a small percentage of glial cells. Compared with adora1, a much smaller portion of neurones express adora3. Notably, adora3 was highly expressed in microglial cells, but not astrocytes. These findings support recent observations.²⁹

Functionally, both A1R and A3R predominantly activate heterotrimeric G proteins belonging to the Gα i/o family, and activation of these receptors often suppresses pain transmission.⁴ Yet, our findings suggested that the A1R and A3R expressions were decreased in the lumbar spinal cord at Day 14 after SNI-t. In line with this finding, a recent study also showed that spinal adora1 and adora3 levels were downregulated after nerve injury.³⁰ Besides transcriptional changes, post-translational alterations (e.g. ubiquitination, proteolytic cleavage) might also contribute to the decreases of A1R and A3R expression after nerve injury, which warrants further investigation. These findings suggest that the endogenous pain inhibition mediated by A1R and A3R may be compromised under neuropathic pain conditions. Nevertheless, spinal application of DPCPX and MRS1523, which block A1R and A3R activation, respectively, augmented C-LFP in SNI-t rats. Thus, A1R and A3R signalling still exerted a tonic inhibition of spinal nociceptive transmission after nerve injury. Intriguingly, the levels of ADA, an enzyme involved in adenosine degradation,³¹ also decreased after SNI-t, which may partly compensate for the decreased endogenous adenosinergic signalling attributable to receptor downregulation. The A1R and A3R antibodies were selected based on previous findings,^32,33 but their specificities may warrant further validation. Future research is also needed to examine whether changes in A1R and A3R expression may be different under other neuropathic pain models, females, and at different post-injury time points.

Activation of A1 and A3 receptors contributed to inhibition of spinal C-fibre-evoked spinal local field potential by spinal cord stimulation

A previous study suggested that adenosine is released into the spinal cord during SCS.¹⁵ Here, a pretreatment with DPCPX or MRS1523 diminished the inhibition of C-LFP by SCS in SNI-t rats, suggesting that both A1R and A3R signalling may be involved. It is unclear why both drugs completely abolished SCS-induced inhibition of C-LFP. Similarly, previous studies showed that blocking either A1R or A3R eliminated adenosine-induced effects.^34,35 A3R was highlighted as a promising target for pain control and A3R agonists may have a good translational potential,^29,36 because of a lack of cardiovascular side-effects known to A1R agonists. In addition to A3R antagonists, selective downregulation of A3R expression in the lumbar spinal cord by using AAV-U6-shRNA(adora3)-GFP, which did not affect A1R level, also blocked the inhibition of C-LFP by SCS.

A1R and A3R are also expressed on the dorsal root ganglion neurones and at their central terminals in the dorsal horn.⁴ Adenosine can activate postsynaptic A1R and A3R to reduce dorsal horn neurone excitability, and presynaptic ones to decrease neurotransmitter release.^4,37 The relative contributions of presynaptic and postsynaptic A1R and A3R activation in spinal pain transmission and SCS-induced analgesia need to be further explored. We recently discovered an astrocyte-mediated spinal pain-gating mechanism through A1R signalling, which highlighted the roles of adenosinergic signalling in neuronal-glia interaction during pain transmission and neuromodulation.⁶ Adenosine also acts via A2R including A2AR and A2BR subtypes. Intriguingly, activation of spinal A2AR can induce long-term inhibition of allodynia in a neuropathic pain condition, possibly through inhibition of glial activation.³⁸ Roles of glial A2AR, A2BR, and A3R which is highly expressed in microglia cells, in spinal non-neuronal pain gating and SCS-induced analgesia warrant further investigation.

Enhancing adenosine signalling increased inhibition of spinal C-fibre-evoked spinal local field potential by spinal cord stimulation

Adenosine deaminase is important to adenosine metabolism and can be active both intracellularly and on the cell surface by forming ecto-ADA (eADA) via interacting with adenosine receptors. Its enzymatic activity is essential for terminating adenosine signalling through catalysing the irreversible deamination of adenosine to inosine and thus regulating extracellular adenosine concentration.³⁹ dCF has been found useful and safe in treating hairy cell leukaemia and some indolent lymphomas. Although increasing tonic adenosinergic transmission by dCF may induce pain inhibition, previous studies showed that intrathecal injection of dCF (100–300 nmol) did not inhibit carrageenan-induced thermal hyperalgesia.^4,40 Similarly, dCF (100 nmol) did not inhibit neuropathic mechanical hypersensitivity in the current study. It is possible that dCF alone cannot produce significant pain inhibition by inhibiting the metabolism of adenosine at the basal/resting level, which may be very low in the synapse. However, SCS increased adenosine levels in the spinal cord,¹⁵ and dCF might thus potentiate the inhibitory actions of adenosine on C-LFP and pain hypersensitivity by SCS.

Applying exogenous agonists often produces transient pharmacologic actions and causes dose-limiting side-effects. For example, adenosine is a short-acting drug and causes headache, lower back pain, cardiovascular side-effects (e.g. cardiac arrhythmia, decreasing blood pressure), and paralysis.^8,14 In contrast, endogenously produced adenosine may alleviate pain without notable side-effects.^4,16,17 Our findings suggest that potentiating endogenous adenosinergic signalling by using dCF as adjuvant pharmacotherapy may augment the inhibition of spinal nociceptive transmission and neuropathic pain by SCS.

Conclusions

Continuing preclinical research on the biological basis of neuromodulation pain therapies is crucial, providing solid scientific evidence for mechanism-oriented treatment and rationales for future translational studies.² Our findings demonstrated the important roles of spinal A1R and A3R signalling in the inhibition of spinal C-LFP by spinal cord stimulation after nerve injury. Importantly, intrathecal deoxycoformycin in combination with spinal cord stimulation might represent an innovative drug repurposing strategy to improve spinal cord stimulation analgesia and speed translation. This unique use-dependent feature of deoxycoformycin (i.e. in the presence of increased adenosine after spinal cord stimulation) might enhance pain inhibition more selectively than using exogenous agonists, limiting adverse effects.

Authors’ contributions

Study conception and design: YG, XC

Conduct of experiments: XC, JL, AU, QX, CZ, GZ, FY

Data analysis: XC, JL

Drafting of the manuscript: XC, YG

Editing and revision of the manuscript: XC, ES, BL, SNR, YG

Approval of the final version of the manuscript: all authors

Acknowledgements

This study was conducted at the Johns Hopkins University School of Medicine. The authors thank Medtronic Inc. for providing the miniature lead for spinal cord stimulation in rats. We are grateful to Qin Zheng and Neil C. Ford from Johns Hopkins University School of Medicine and Xuewei Wang from the University of South Florida Morsani College of Medicine for their constructive insights and discussion of data.

Declarations of interest

YG and SNR received research grant support from Medtronic, Inc., Minneapolis, MN, USA. JL is a consultant for Medtronic Inc., Minneapolis, MN, USA; St. Jude Medical, Austin, TX, USA; Boston Scientific, Marlborough, MA, USA, and Elekta AB, Sweden. However, none of the authors has a commercial interest in the material presented in this paper. The authors declare that they have no conflicts of interest. There are no other relationships that might lead to a conflict of interest in the current study.

Funding

US National Institutes of Health (Bethesda, MD, USA) (NS110598 and NS117761 to YG), and the Blaustein Pain Research Fund from the Johns Hopkins University (to YG). Funders had no role in study design, data collection, data interpretation, or the decision to submit the work for publication.

Handling Editor: Nadine Attal

Appendix A. Supplementary data

The following are the Supplementary data to this article: