Adenosine A¹but Not A^2aReceptor Agonist Reduces Hyperalgesia Caused by a Surgical Incision in Rats: A Pertussis Toxin-sensitive G Protein–dependent Process

These experiments were reviewed and approved by the Institution's Animal Care and Use Committee in Muenster, Germany. The animals were treated in accordance with the Ethical Guidelines for Investigations of Experimental Pain in Conscious Animals as issued by the International Association for the Study of Pain.20 

Experiments were performed in 135 adult (weight, 280–350 g) male Sprague-Dawley rats (Harlan, Indianapolis, IN) housed in pairs before surgery. All persons performing experiments were blinded to the drug administered. Food and water were available ad libitum . Postoperatively, the animals were housed individually. Eight animals were excluded for wound dehiscence or clotted catheters; at the end of the protocol, all animals were anesthetized and killed with an overdose of potassium chloride administered intracardially.

For subarachnoid drug administration, a lumbar intrathecal catheter was placed using a technique21modified from one described previously.22Briefly, after isoflurane anesthesia, the lumbar skin was cleansed and incised. The intervertebral space between L5 and L6 was punctured with a hypodermic needle, and a 32-gauge polyurethane catheter (length 10 cm, 32-PU, OD 0.010 inches, ID 0.005 inches; Micor, Allison Park, PA) was advanced through the needle in the lumbar area of the spinal cord. The distal end was secured, inserted into PE-10 tubing, and tunnelled to the cervical region. The catheter (8 μl dead space) was flushed with saline and sealed. One day after catheter placement, 20 μl lidocaine, 2%, was administered, and only rats with a brief bilateral hind limb paresis were studied. Experiments were begun not less than 3 days after intrathecal catheter placement. At the end of the experiment, the localization of the catheter was verified by injecting 30 μl methylene blue, killing the rat, and dissecting the lumbar spinal cord. In rats with a functional intrathecal catheter, segments of the lumbar spinal cord were dyed with methylene blue.

For paw incisions, all rats were anesthetized with 1.5–2% isoflurane delivered via  a nose cone. As described previously,19a 1-cm longitudinal incision was made through skin and fascia of the plantar aspect of the right hind paw including the plantaris muscle. The skin was apposed with two mattress sutures of 5-0 nylon. After surgery, the animals were allowed to recover in their cages. Sutures were removed approximately 30 h later at the end of postoperative day 1, and the wounds healed well within 5–6 days. The animals did not receive additional analgesics after surgical treatments such as catheter placement or paw incision.

On the day of the experiment, the rats were placed individually on an elevated plastic mesh floor. After adaptation to testing conditions, baseline nonevoked or evoked pain behaviors before paw incision were measured as described.19,23 

Nonevoked (spontaneous) pain behaviors were assessed by a cumulative pain score as described previously.23,24Unrestrained rats were placed on a plastic mesh floor, and the incised foot of each animal was closely observed during a 2.5-min period for 30 min. Depending on the position in which the foot was found during the majority of the 2.5-min scoring period, a 0, 1, or 2 was given. Full weight bearing of the foot (score = 0) was present if the wound was blanched or distorted by the mesh. If the foot was completely off the mesh, a score of 2 was recorded. If the area of the wound touched the mesh without blanching or distorting, a 1 was given. The sum of the 12 scores (0–24) obtained during the 30-min session for the incised foot was obtained.

Withdrawal responses to punctate mechanical stimulation (evoked pain behavior) were determined using calibrated nylon von Frey monofilaments applied from underneath the cage through openings (12 × 12 mm) in the plastic mesh floor to an area adjacent to the wound. Each von Frey filament was applied once starting with 15 mN and continuing until a withdrawal response occurred or 247 mN (the cutoff value) was reached. The median of the lowest force from the three tests producing a response was considered the withdrawal threshold. For this study, 563 mN was recorded as the withdrawal threshold if there was no withdrawal response to the next lowest filament (247 mN). However, using standard von Frey filaments to assess mechanical pain thresholds has several limitations. Most importantly, shape and size of the filaments are critical factors for mechanical pain perception. Von Frey filaments with flat ends bend on the skin at their sharp corners, and different probe angles may cause variable mechanical thresholds. Greenspan et al.  25,26reported that a change in probe angle of 15° produced variable pain thresholds with a stimulus of the same tip size. Furthermore, different von Frey filament strengths vary in their tip diameters up to 100-fold, indicating that there is a trade-off between the force applied and the probe circumference for pain threshold. In addition, Hogan et al.  27further indicate that decreased withdrawal thresholds caused by a peripheral nerve injury assessed by von Frey filaments are not valid measures for neuropathic pain. However, we speculate that this is different to acute postoperative pain. In contrast to neuropathic pain, a surgical incision causes robust mechanical hyperalgesia characterized by decreased withdrawal thresholds to von Frey filaments; mechanical hyperalgesia was blocked by morphine or other analgesics. Therefore, we suggest that besides the limitations of standard von Frey filaments, mechanical hyperalgesia is a robust and important feature of postoperative pain.

Twenty-one rats were pretested for nonevoked pain behaviors as described in the first two paragraphs of the Pain Behaviors section. The incision was made in the plantar aspect of the foot, and after a recovery time of 2 h, spontaneous pain behavior was assessed.

Subsequently, 5 nmol R-phenylisopropyl-adenosine (R-PIA; adenosine A¹receptor agonist), 10 nmol CGS21680, or vehicle was administered intrathecally (n = 6–8/group). Nonevoked pain behavior was measured every 30 min after drug injection for the next 3 h on the day of surgery and postoperative day 1.

Twenty-eight rats were pretested for mechanical withdrawal threshold as described above. The incision was made in the plantar aspect of the foot, and after a recovery time of 2 h, responses to mechanical stimuli were tested.

Subsequently, 0.3, 1, or 5 nmol R-PIA (adenosine A¹receptor agonist) or vehicle was administered intrathecally (n = 6–8/group). Mechanical hyperalgesia was measured at 30 and 60 min and then every hour after drug injection for the next 5 h. In separate groups of animals, administration of an A¹R antagonist (CPT, 242 nmol) or an A^2aR antagonist (DMPX, 50 nmol) was followed by 5 nmol R-PIA (n = 4).

In another group of rats, 2, 10, or 20 nmol CGS21680 (adenosine A^2areceptor agonist) or vehicle was administered intrathecally (n = 6–8/group). Mechanical hyperalgesia was measured at 30 and 60 min and then every hour after drug injection for the next 5 h. In separate groups of animals, administration of an A¹R (CPT, 242 nmol) or A²R antagonist (DMPX, 50 nmol) was followed by 20 nmol CGS21680 (n = 4).

In a separate experiment, rats were pretreated with 0.5 μg PTX 7 days before incision. A greater effect of the toxin at 6 days compared with 2 days to disrupt the inhibitory secondary messenger pathway has been observed.28In the current study, PTX-pretreated rats were tested before and 2 h after incision, with observation of similar mechanical thresholds compared with untreated animals, indicating that 0.5 μg PTX did not affect pain behaviors on its own. In agreement, other investigators demonstrated that low PTX doses (0.5–1 μg) did not produce pain behavior in uninjured animals.28,29On the day of surgery, 5 nmol R-PIA was administered after incision, and mechanical hyperalgesia was measured at 30 and 60 min and then every hour after drug injection for the next 5 h (n = 6). In another group of animals, 5 nmol R-PIA was injected in rats pretreated with 10 μg forskolin, and mechanical hyperalgesia was assessed for 5 h (n = 5). Similar forskolin doses were used by other investigators without showing pain behavior in uninjured animals.30 

In another group of animals, 5 nmol R-PIA was administered after pretreatment with vehicle or 20, 50 or 100 nmol of the adenosine triphosphate–sensitive potassium (K^ATP) channel blocker glibenclamide (n = 6–8/group) or after pretreatment with the nonselective potassium channel blockers 4-aminopyridine (4-AP; n = 5; 5 nmol) and tetraethylammonium (n = 5; 100 nmol); mechanical hyperalgesia was assessed for 5 h as described above. Doses for tetraethylammonium and 4-AP were evaluated by a dose finding experiment. Greater doses of tetraethylammonium and 4-AP cause significant adverse effects, including intense pain behaviors or motor impairment.

In a separate group of animals, R-PIA was administered 10–15 min after intrathecal administration of apamin (n = 4; 3 ng), charybdotoxin (n = 4; 1 ng), or margatoxin (n = 6; 10 ng), and pain behaviors were assessed. None of these selective potassium channel blockers alone produced pain behaviors.31 

All drugs were purchased from Sigma (Hamburg, Germany) or Tocris (Bristol, United Kingdom). To enhance their solubility, R-PIA (0.3–10 nmol, A¹R agonist; Ki A¹/A^2a: 2/860 nm)32was dissolved in a pH-adjusted HCl solution, and CGS21680 (2–20 nmol, A^2aR agonist; Ki A¹/A^2a: 2,600/15 nm)13was dissolved in a pH-adjusted NaOH solution. The pH of all drugs and vehicles administered was 7.0–8.4. CPT (A¹R antagonist, 242 nmol; Ki A¹/A^2a: 10.9/1,440 nm),13DMPX (A^2aantagonist, 50 nmol; Ki A¹/A^2a: 1,200/8.2 nm),33tetraethylammonium (nonselective potassium channel blocker, 100 nmol), 4-AP (nonselective potassium channel blocker, 5 nmol), PTX (0.5 μg),28,30apamin (small-conductance calcium-activated potassium channel inhibitor, Kd: 60 pm–100 nm; 3 ng),34charybdotoxin (large- and intermediate-conductance calcium-activated potassium channel blocker, Kd: 1.7–5 nm, voltage-gated potassium channel inhibitor 1.3, Kd: 0.17 nm, 1 ng),31and margatoxin (voltage-gated potassium channel inhibitor Kv1.3 and Kv1.6; Kd: 0.03 and 5 nm; 10 ng)31were dissolved in saline; forskolin (cAMP activator, 10 μg)30was dissolved in 10% ethanol, and glibenclamide (K^ATPchannel inhibitor, 10–100 nmol)35was dissolved in 50% dimethyl sulfoxide. Peng et al.  36demonstrated that up to 50% of dimethyl sulfoxide did not change the activity of dorsal horn neurons, and greater doses of dimethyl sulfoxide were administered intrathecally in in vivo  experiments by others.31Intrathecal injection volumes for all substances were 10 μl followed by a 10-μl flush of preservative-free saline. All drugs were prepared on the day of the experiment.

The results are presented as median for ordinal data, and all data were compared using nonparametric analyses. The Friedman test for within-group comparisons and the Kruskal-Wallis test and Wilcoxon–Mann–Whitney test for between-groups comparisons were used. Multiple comparisons after the Friedman test and Kruskal-Wallis test were performed using a Dunnett test or Dunn test for nonparametric analysis.37 P < 0.05 was considered significant.

In all groups of animals, similar nonevoked pain behavior 2 h after surgery was observed (figs. 1A–C). Intrathecal administration of R-PIA (5 nmol) but not of CGS21680 (10 nmol) produced a decrease of the median pain score from 18.5 (0 h) to 7 during the first hour after drug administration (P < 0.05 vs.  0 min; figs. 1B and C). Decreased nonevoked pain behaviors lasted for more than 2 h. The next day, baseline pain scores were assessed, and similar results were obtained (P < 0.05 vs.  0 min; fig. 1D).

In vehicle-treated rats, the median withdrawal threshold to von Frey filaments decreased from 563 mN (pre) before surgery to 36.9 mN 2 h (0 min before intrathecal drug administration) after incision. Hyperalgesia was persistent; median withdrawal thresholds were 36.9 mN or less throughout the day of surgery (fig. 2A). Intrathecal administration of 0.3, 1, and 5 nmol R-PIA produced a dose-dependent increase of the reduced withdrawal thresholds after incision for 30 and 60 min, respectively (P < 0.05 vs.  0 min; figs. 2B–D), indicating an A¹R-induced reduction of hyperalgesia. Intrathecal pretreatment with CPT (242 nmol), a specific A¹R antagonist, but not with DMPX (50 nmol), a specific A^2aR antagonist, blocked R-PIA (5 nmol)–induced antinociception (figs. 2E and F). Similar results were observed on postoperative day 1 (fig. 2G).

In separate groups of animals, intrathecal administration of a vehicle or 2 or 10 nmol CGS21680 did not modify pain behaviors (figs. 3A–C), but intrathecal injection of 20 nmol CGS21680 caused a brief increase in withdrawal thresholds 30 min after injection (P < 0.05 vs.  0 min; fig. 3D). This short-lasting antihyperalgesic effect was blocked by intrathecal pretreatment with CPT (A¹R antagonist) but not with the corresponding A^2aR antagonist DMPX, indicating that the CGS21680-induced increase of withdrawal thresholds was due to an unspecific activation of the A¹R (figs. 3E and F). Similar results were observed on postoperative day 1 (fig. 3G).

Intrathecal pretreatment with 0.5 μg PTX 7days before administration of 5 nmol R-PIA blocked R-PIA–induced increase of withdrawal thresholds (fig. 4A). Similarly, in rats pretreated with forskolin, the subsequent intrathecal administration of R-PIA did not modify mechanical withdrawal thresholds after incision (fig. 4B). In addition, pretreatment of PTX or forskolin followed by a vehicle injection (data not shown) did not increase hyperalgesia in rats after an incision, indicating that the inhibition of R-PIA–induced antinociception by PTX or forskolin pretreatment was not caused by a baseline shift of withdrawal thresholds.

Intrathecal pretreatment with the nonselective potassium blocker tetraethylammonium (100 nmol) or 4-AP (5 nmol) reduced R-PIA–induced antinociception, indicating that potassium channels were involved in this process (P < 0.05 vs.  0 min; figs. 5A and B). Intrathecal pretreatment with glibenclamide produced a dose-dependent inhibition of R-PIA–induced antinociception (P < 0.05 vs.  0 min; figs. 5C–F). Intrathecal administration of glibenclamide (100 nmol) and subsequent vehicle did not modify persistent hyperalgesia after incision (data not shown). Intrathecal administration of neither apamin nor charybdotoxin modified hypoalgesia mediated by R-PIA (P < 0.05 vs.  0 min; figs. 6A and B); a small but not statistically significant inhibition of R-PIA–induced hypoalgesia was observed with a pretreatment of margatoxin (P < 0.05 vs.  0 min; fig. 6C). Higher intrathecal doses of apamin, charybdotoxin, or margatoxin produced severe pain behaviors and were not used in this study. Therefore, K^ATPchannels but not calcium-activated potassium channels and the voltage-gated potassium channels Kv1.3 and Kv1.6 are partially involved in A¹R-induced antinociception.

In the current study, we demonstrated that the intrathecal administration of an A¹R agonist (R-PIA) but not of an A^2aR agonist (CGS21680) decreased nonevoked pain behaviors and evoked mechanical hyperalgesia after a surgical incision. Furthermore, A¹R-induced spinal antinociception is mediated by interactions with PTX-sensitive G proteins and due to the inhibition of adenylate cyclase. In addition, the opening of K^ATPchannels but not of calcium-activated potassium channels or Kv1.3 or 1.6 channels contribute to the antinociceptive effect of adenosine receptor agonists.

Several investigators have examined the effects of spinally administered A¹R agonists in models of persistent pain and hyperalgesia. In vitro  and in vivo  electrophysiologic experiments demonstrated that activation of A¹Rs hyperpolarize dorsal horn neurons in lamina II and inhibit the C fiber–induced windup phenomenon.5,8,11,38Intrathecal administration of selective A¹R agonists reduced pain behaviors after chemical irritation,12nerve injury,32,39spinal cord injury,40and inflammation.13Furthermore, mice lacking the A¹R exhibited increased nociceptive responses under normal conditions and enhanced heat hyperalgesia after inflammation and nerve injury.6,7In agreement, results from the current study revealed that spinal administration of an A¹R agonist decreased dose dependently mechanical hyperalgesia after an incision, indicating that activation of the A¹R affects the maintenance of exaggerated pain behaviors in this model for postoperative pain. Accordingly, previous experiments demonstrated a dose-dependent antihypersensitivity effect of the positive allosteric adenosine receptor modulator T62 after intrathecal administration in the incisional model for postoperative pain.18,41 

In contrast, the role of the spinal A^2aR for pain processing is uncertain.8,11Recently, Yoon et al.  12demonstrated that intrathecal administration of the A^2aR agonist DPMA inhibited both phases of formalin-induced pain behaviors. Poon et al.  13revealed only a modest antinociceptive effect of the spinally administered A^2aR agonist CGS21680 on inflammation-induced thermal hyperalgesia, arguing that this might be due to an unspecific activation of A¹Rs. However, other investigators did not observe antinociceptive effects after spinal A^2aR activation. In electrophysiologic studies of spinal cord slices, application of CGS21680 (A^2aR agonist) was without effects on neuronal transmission11or exhibited a mix of excitatory and inhibitory effects.2,8In the current study, intrathecal administration of CGS21680 did not decrease mechanical hyperalgesia after incision. Moderate effects on incision-induced hyperalgesia after application of higher doses of CGS21680 were due to an unspecific activation of A¹Rs and therefore blocked by an A¹R antagonist but not by an A^2aR antagonist.

Although there is ample evidence that adenosine-induced spinal antinociception is largely mediated by activation of A¹Rs, little is known about the underlying mechanisms. A¹Rs are coupled to PTX-sensitive Gi proteins reducing cAMP levels by inhibiting adenylyl cyclase.2Because increased cAMP levels facilitate the release of neurotransmitters,42enhance the excitability of dorsal horn neurons or spinothalamic tract cells,43and result in hyperalgesia,44it has been suggested that modifying the cAMP transduction cascade represents one downstream mechanism by which activation of A¹Rs produces antinociception.45In agreement, presynaptic A¹Rs located on sensory afferent terminals and dorsal root ganglion neurons decreased cAMP production1,2and inhibited the release of neurotransmitters such as CGRP46and glutamate.47,48Furthermore, a decrease of spinal glutamate release caused by presynaptic A¹R activation was observed after capsaicin injection and nerve injury47,48but not after formalin injection,49indicating that characteristic mechanisms of A¹R activation occur after different types of tissue injuries. Although not investigated in the current study, we speculate that activation of presynaptic A¹Rs is crucial for A¹R-induced antinociception.

Pertussis toxin ribosylates the α subunit of inhibitory G proteins linked to adenylate cyclase and disrupt this inhibitory secondary messenger pathway, leading to increased levels of cAMP.50Previous studies demonstrated that PTX prevented A¹R-mediated inhibition of adenylate cyclase50and reduced effects of A¹R agonists on normal nociception in uninjured animals.30In the current study, pretreatment with PTX inhibited the hypoalgesic effect of intrathecal R-PIA, implicating that A¹R-induced spinal antinociception is mediated by interactions with PTX-sensitive inhibitory Gi proteins. In addition, we demonstrated that a brief pretreatment with forskolin attenuated the hypoalgesic effect of intrathecal R-PIA, suggesting that the cAMP transduction cascade is involved in A¹R-induced antinociception. Accordingly, several lines of evidence demonstrated that spinal PTX-sensitive G proteins seem to be involved in the antinociceptive effect of various agents, including α²agonists,28baclofen, and noradrenaline.51 

Furthermore, there is experimental evidence of a possible interaction between A¹R and PTX-resistant stimulatory Gs proteins or inhibitory Gz proteins.52,53However, the role of PTX-insensitive G proteins for A¹R-induced antinociception in the spinal cord is not clear. In in vitro  experiments, Cordeaux et al.  52,53demonstrated that after an increased expression of A¹Rs, the application of different A¹R agonists (NECA ≫ CPA = R-PIA) caused an accumulation of cAMP mediated by PTX-insensitive G proteins. If this pathway occurs in the spinal cord, we would expect that the intrathecal administration of an A¹R agonist in PTX-pretreated animals causes pain behaviors due to inhibition of adenylate cyclase and subsequently an accumulation of cAMP. In contrast, in the current study, the R-PIA–induced antinociceptive effect was largely blocked by PTX pretreatment without producing increased pain behaviors, suggesting an interaction between A¹R and PTX-sensitive Gi proteins but not PTX-insensitive Gs or Gq proteins. However, a pronociceptive effect mediated by PTX-resistant Gs or Gq proteins using higher doses of R-PIA, other A¹R agonists, or a different animal pain model cannot be discarded.

Among various Gi proteins, Gz is the only member of this subfamily with a PTX-resistant α^zsubunit.54The PTX-resistant Gz protein is located predominantly in the central nervous system and retina and interacts with several Gi-coupled receptors, including A¹R and opioid receptors.55To date, the only known signaling function of Gz is inhibition of adenylyl cyclase. Although a role of Gz proteins for A¹R-induced antinociception cannot be discarded, we demonstrated that a blockade of PTX-sensitive G protein receptors prevented completely the antinociceptive effect of R-PIA, speculating that the PTX-resistant G protein receptor Gz will be of limited importance for A¹R-mediated antihyperalgesia.

Although similar results were obtained for normal nociception in uninjured animals by using the tail flick and hot plate tests,30results of the current study demonstrate for the first time that A¹R-induced spinal antinociception after an incision is mediated by interactions with PTX-sensitive G proteins and due to the inhibition of adenylate cyclase.

There is ample evidence that potassium channels are involved in nociception and that the opening of specific potassium channels is critical for G protein-coupled receptor–mediated antinociception including opioid receptors, γ-aminobutyric acid type B receptors, α²receptors, and A¹Rs.8,56Ocana and Baeyens57revealed in uninjured animals that different intracerebroventricularly administered K^ATPchannel blockers antagonized A¹R-induced supraspinal antinociception. In the current study, we assessed for the first time the importance of different potassium channels for A¹R-mediated antinociception in animals after a tissue injury (incision). In agreement with Ocana et al. ,57we demonstrated that R-PIA–induced hypoalgesia was attenuated by intrathecal administration of a K^ATPchannel blocker. Similarly, intrathecal administration of the nonselective potassium channel blockers tetraethylammonium and 4-AP decreased the antinociceptive effect of R-PIA on incisional pain, indicating a possible role of other potassium channels for A¹R-induced hypoalgesia.

Calcium-activated potassium channels are subdivided in small (SK), intermediate (IK), or large (BK) conductance channels and are activated by an increased intracellular calcium concentration. The bee venom toxin apamin (high-affinity SK channel inhibitor; Ki: 60 pm–100 nm)58and the scorpion toxin charybdotoxin (IK, BK inhibitor; Ki: 1.7–5 nm)59have been used to demonstrate that spinal SK channels seem to be involved in antinociception mediated by cannabinoid and Δ-opioid receptors.60,61In the current study, the intrathecal application of neither apamin nor charybdotoxin modified R-PIA–induced hypoalgesia, indicating that calcium-activated potassium channels are not involved in A¹R-mediated antinociception.

Besides its effects on BK and IK conductance channels, charybdotoxin inhibits also the voltage-gated potassium channel Kv1.3 (Ki: 0.17 nm).62Margatoxin blocks Kv1.3 and Kv1.6 (Ki: 0.03 and 5 nm) but not calcium-activated potassium channels.59,62Both scorpion toxins have been used to demonstrate that Kv1 channels are important for α²-adrenoceptor–, μ-opioid receptor–,56and nonsteroidal antiinflammatory drug31–induced antinociception. In the current, study neither charybdotoxin nor margatoxin decreased R-PIA–induced hypoalgesia, indicating that Kv1.3 and Kv1.6 channels are not involved in A¹R-mediated antinociception.

Therefore, the opening of K^ATPbut not of calcium-activated potassium channels and of Kv1.3 and 1.6 channels is important for antinociception mediated by A¹R agonists. The role of other potassium channels for A¹R-induced antinociception must be clarified.

In the current study, we demonstrated that the intrathecal administration of an adenosine A¹R agonist (R-PIA) but not of an adenosine A^2aR agonist (CGS21680) decreased mechanical hyperalgesia after a surgical incision. Furthermore, A¹R-induced spinal antinociception is mediated by interactions with PTX-sensitive G proteins and due to the inhibition of adenylate cyclase. The opening of K^ATPchannels but not of calcium-activated potassium channels or the voltage-gated channels Kv1.3 and Kv1.6 may contribute to the antinociceptive effect of A¹R agonists.