The most widely accepted explanation for the inflammation-induced spinofugal positive feedback loop resulting in enhanced peripheral neurotransmitter and peptide release is based on dorsal root reflex (DRR) involvement (Cervero and Laird 1996b; Schmidt 1971; Willis 1999). This theory posits that the maintained, increased afferent activity towards the spinal cord engendered by peripheral inflammation triggers a primary afferent depolarization (PAD) in the central terminals of nociceptive primary afferent fibers. Low levels of PAD result in presynaptic inhibition of afferent activity; this is thought to occur, in part, through low-threshold afferent fiber (large diameter) activation of γ-aminobutyric acid (GABA)-containing inhibitory interneurons that in turn have collaterals ending on the nociceptive afferent terminal. We now know that the primary afferent fibers and DRG neurons are slightly more positive (less polarized) than neurons in the central nervous system. When GABA_A receptors on the central terminals of primary afferent nociceptors are activated, Cl⁻ channels open as they do at all GABA_A receptors, but the Cl⁻ flux is out of the terminal rather than inward, resulting in a small PAD and decreased central release of afferent neurotransmitters, i.e., PAD. This large fiber inhibition of small-fiber throughput is the central tenet of Melzack and Wall’s gate control theory (Melzack and Wall 1965). Following sustained tissue injury, or a high-frequency afferent barrage in nociceptive afferent fibers, sensitized nociceptors can activate the inhibitory interneurons and hijack the PAD pathway (Cervero and Laird 1996b). This hypothesis is substantiated by the anatomy, which shows that GABAergic synapses are often postsynaptic as well as presynaptic to primary afferent terminals (Barber et al. 1978; Carlton and Hayes 1990), many of which are identified or presumptive nociceptors (Alvarez et al. 1992; Bernardi et al. 1995). In addition, tissue inflammation induces an increase in spinal dorsal horn GABA release (Castro-Lopes et al. 1992, 1994). This enhanced presynaptic GABAergic activity on primary afferent terminals leads to a greater level of depolarization, which, if threshold is reached, results in antidromic activity in the afferent fibers. These peripherally directed action potentials are known as DRRs (Schmidt 1971). In addition, at this point orthodromic nociceptive throughput is restored due to the loss of the inhibitory PAD, resulting in enhanced activation of the ascending system (Cervero and Laird 1996a, b; Willis 1999). Thus, drugs or manipulations that alter DRRs elicit covariant changes in inflammation and inflammatory pain. It would be expected that agents that cause presynaptic inhibition via hyperpolarization of the terminal or blockade of neurotransmitter release would reduce both the DRRs and nociception (Cervero and Laird 1996b). In rodent skin, DRR firing in response to intradermal capsaicin has been demonstrated in Aδ and C, but not in Aβ fibers (Lin et al. 2000; Wang et al. 2004; Willis 1999). This antidromic activity results in primary afferent neurotransmitter release in the skin, which extends beyond the injection site. However, in joint afferents of cats and monkeys, DRRs have been demonstrated in group II, III, and IV fibers, indicating that they are generated in large myelinated fibers as well (Sluka et al. 1995). This difference is thought to be due to variations between cutaneous and joint fiber systems rather than to a species difference, although the specific underlying dorsal horn anatomy is not known and the intensity/specificity of the afferent stimuli (kaolin/carrageenan for the joint and capsaicin for the skin) may also be factors. These data conflict with those of an earlier study indicating that electrical stimulation of the distal end of a sectioned posterior articular nerve at group IV, but not group III fiber strength, was sufficient to elicit plasma extravasation and blood cell infiltration into the knee synovia (Ferrell and Russell 1986).

Pharmacology: A second class of acute joint inflammation models is initiated by systemic immunization followed in a few weeks by injection of an immuno-active agent into the knee capsule resulting in a monoarthritis. The model has been tested with a variety of pharmacological treatments applied to the spinal cord prior to knee injection. In the acute phase (up to 6 h after the knee injection), spinal administration of the NMDA receptor antagonist ketamine prevented vascular leakage (edema), infiltration of mononuclear leukocytes, and preserved joint integrity (Boettger et al. 2010a). Spinal pretreatment with tumor necrosis factor (TNF) antagonists (TNF-neutralizing antibody or thalidomide, which restricts TNF synthesis by increasing its mRNA degradation (Moreira et al. 1993)) reduced vascular leakage and infiltration of polymorphonuclear leukocytes into the synovial fluid (Bressan et al. 2010, 2012) and helped to maintain joint integrity (Boettger et al. 2010a). The TNF-neutralizing antibody, but not thalidomide, also blocked infiltration of mononuclear leukocytes into the knee (Bressan et al. 2010, 2012). In the acute phase, i.t. administration of the TNF-neutralizing antibody not only reduced joint swelling but also blocked the shift to a sympathetically dominated autonomic nervous system as measured by increased heart rate and heart rate variability (Boettger et al. 2010b). Spinal pretreatment with furosemide, which inhibits DRRs, or with glial (fluorocitrate and minocycline) or fractalkine inhibitors also successfully blocked edema and leukocyte infiltration into knee synovial fluid during the acute phase of this model (Bressan et al. 2012). These results were unaffected by corticosteroid synthesis inhibitors, indicating a lack of participation of the HPA axis. Coadministration of furosemide with either glial inhibitor produced no additional effect, indicating, perhaps, a common pathway. Thus, as seen for the kaolin/carrageenan model of knee inflammation, DRRs appear to be a necessary element in the efferent leg of spinal modulation of peripheral inflammation. Taken together the pharmacological results indicate that, unlike the kaolin/carrageenan model of joint inflammation, there is an intraspinal cord NMDA and glial dependence of the DRRs in the early stages of adjuvant-induced arthritis (AA). Spinal administration of morphine prevented vascular leakage (edema), infiltration of mononuclear leukocytes, and preserved joint integrity (Boettger et al. 2010a). This last result could be due to the opiate-induced reduction of primary afferent glutamate release, which would block the incoming signal and reduce activation of postsynaptic neurons and glia. Alternatively, presynaptic actions of the opiate could reduce primary afferent excitability, thus, affecting a reduction in DRR magnitude (Cervero and Laird 1996b).

Pharmacology: Another experimental variant designed to demonstrate spinal contributions to modulation of the acute inflammatory response utilizes an intradermal or subcutaneous carrageenan paradigm that measures either vascular leakage (edema) or the mechanistically different neutrophil infiltration (measured as myeloperoxidase activity, MPO) as outcome measures indicative of inflammation. Intrathecal pretreatment with indomethacin, a prostaglandin synthesis inhibitor, reduces the edema in this model (Daher and Tonussi 2003) as does i.t. morphine (Brock and Tonussi 2008). Unexpectedly, neutrophil infiltration is not blocked following i.t. morphine. If this result holds, it points to different spinal feedback mechanisms in the control of plasma extravasation and immune cell chemotaxis as i.t. morphine should undoubtedly have blocked the DRRs. The morphine effect on edema apparently involves a spinal nitric oxide/cyclic-guanosine monophosphate (c-GMP) pathway (Brock and Tonussi 2008), which is likely to be within the afferent terminals. In similar experiments, spinal pretreatment with either 5HT1 receptor agonists or 5HT2 receptor antagonists also reduced paw swelling (Daher and Tonussi 2003). Significantly, spinal administration of adenosine A1, but not adenosine A2-specific agonists, dose-dependently inhibited dermal neutrophil accumulation due to intradermal carrageenan (Bong et al. 1996; Sorkin et al. 2003). The adenosine effect was mimicked by a spinal NMDA antagonist and reversed by NMDA agonists. This latter experiment demonstrated that the NMDA linkage is downstream of the adenosine effect. What was unexpected, given the data on the kaolin/carrageenan knee joint model, was that spinal non-NMDA receptor antagonists had no affect on the intradermal carrageenan-induced neutrophil accumulation (Bong et al. 1996). Unfortunately, spinal GABAergic agents and glial antagonists were not tested in this model (Pinter et al. 2002).

Paw carrageenan elicited a massive reduction in peripheral adenosine levels, which was temporally linked with the neutrophil infiltration (Bong et al. 1996). This adenosine loss, as well as the increase in neutrophil accumulation, is blocked by spinal pretreatment with an NMDA antagonist or an adenosine A1 agonist. Basal levels of peripheral adenosine in the tissue act on A2 receptors to inhibit peripheral neutrophil infiltration; lower concentrations of adenosine, in contrast, preferentially activate A1 receptors, which enhance neutrophil chemotaxis (Cronstein et al. 1992; Nolte et al. 1992). Thus, maintenance of basal peripheral adenosine concentration is thought to be the basis of the anti-allodynic activity of both the spinal adenosine A1 agonists and NMDA receptor antagonists. Surprisingly, although both substance P and TNF increased at the site of inflammation, neither of the spinal pretreatments that reduced neutrophil infiltration in this model affected either of these agents. Thus, these data appear to conflict with the well-known chemotactic effects of substance P and pro-inflammatory cytokines. There have to be other, as yet, unidentified factors that can block neutrophil infiltration despite the permissive influence of the tachykinin and cytokine. Levels of sympathetic nervous system mediators were not investigated in these studies. Taken together, these data indicate that intradermal carrageenan-induced inflammation, like that induced by intradermal capsaicin, is enhanced by a spinal NMDA-mediated linkage (Sorkin et al. 2003).

Pathways: Chemical elimination of the capsaicin-sensitive small afferent sensory fibers had no effect on neutrophil infiltration in the skin carrageenan model or on the ability of intrathecal adenosine A1 agonists to modify the inflammatory response (Sorkin et al. 2003). In agreement with the negative capsaicin results, Chen et al. (2007) recently observed no effect of capsaicin pretreatment and resulting loss of the peptidergic sensory afferents on acute paw swelling induced by paw carrageenan or complete Freund’s adjuvant, but did find that loss of peptidergic innervation reduced increases in paw thickness induced by injection of bee venom. These authors argue that the role of small peptidergic afferents varies according to unidentified aspects of the initiating inflammatory stimulus. Chemical sympathectomy resulted in a strong, but nonsignificant, trend towards decreased carrageenan-induced neutrophil infiltration. However, despite the reduction of basal inflammation due to loss of the sympathetic terminals, intrathecal adenosine A1 agonists resulted in a further reduction in the number of neutrophils found in the injected skin (Sorkin et al. 2003). This implies that the spinal adenosine mediated modulatory effect did not require engagement of the sympathetic nervous system.

It is curious that chemical sympathectomy was ineffective in altering neutrophil infiltration as the sympathetic neurotransmitters norepinephrine and adenosine are known to change in acute inflammation and are well known as leukocyte chemotactic agents (Rose et al. 1988; Straub et al. 2000), although loss of the sympathetic efferents also had no effect on the DRR-dependent knee joint inflammation (Sluka et al. 1994b). The most surprising result of the intradermal carrageenan study was that bilateral carrageenan injections in animals with unilateral rhizotomies resulted in symmetrical infiltration of neutrophils into the injection sites (Sorkin et al. 2003). These data strongly suggest that unlike the kaolin/carrageenan knee joint model, neutrophil trafficking in skin inflammation is independent of endogenous spinal influences. Loss of spinal adenosine modulation of the neutrophil infiltration following rhizotomy points to a role for sensory fibers in this process. Given the previous lack of effect of spinal adenosine agonists on peripheral substance P in the inflamed tissues (Bong et al. 1996), small peptidergic afferents are not the relevant ones, neither are the sympathetic efferents (Sorkin et al. 2003). Thus, the spinal adenosine modulation of the peripheral inflammation is mediated by some combination of large myelinated fibers, non-peptidergic small fibers, or an ascending system with an efferent leg not found within the peripheral nerve.

Animal models of chronic inflammation that last for weeks are used to study spinal cord control of peripheral inflammation. These model rheumatoid arthritis (RA) and include adjuvant-induced arthritis (AA) and, less frequently, collagen-induced arthritis (CIA).

Pharmacology: Continuous i.t. administration of cyclohexyladenosine (CHA), an adenosine A1 agonist, greatly ameliorated the symptoms of AA including joint swelling, bone changes (demineralization, erosions, and heterotopic bone formation), cartilage destruction, synovial integrity/narrowing of the joint space, hyperemia, and expression of activator protein 1 (AP-1) in the joint. This spinal treatment was efficacious if it began as late as 8 days postimmunization when animals first presented with clinical signs, but had a much smaller, nonsignificant effect on paw swelling when the course of treatment began after clinical signs were well established (day 14) (Boyle et al. 2002). Despite the more than 80 % decrease in AA-induced presentation of clinical symptoms and paw volume that was observed with i.t. CHA treatment, CHA only reduced the AA-induced c-Fos expression in the superficial dorsal horn by 22 % (Boyle et al. 2002). Reduction of c-Fos expression in other spinal laminae was unaffected by the A1 agonist. Similar anti-inflammatory results were observed with either i.t. administration of a TNF-neutralizing antibody or SB203580, a p38 inhibitor, with the addition that these agents also suppressed synovial infiltration of immune cells and expression of the pro-inflammatory cytokines interleukin (IL)-1β, IL-6, and TNF and MMP3, a key gene involved with extracellular matrix degradation (Boettger et al. 2010b; Boyle et al. 2006). In addition, i.t. etanercept blocked the arthritis-induced shift in the autonomic nervous system towards sympathetic dominance (Boettger et al. 2010b).

Interestingly, the p38 antagonist did not affect peripheral T cell proliferation. Continuous administration of morphine or of the NMDA antagonist ketamine throughout a 3-week course of AA caused a major decrease in joint swelling and infiltration of the synovia by pro-inflammatory immune cells (Boettger et al. 2010a). This is likely due to a loss of DRRs due to presynaptic actions on μ-opiate and NMDA receptors, respectively. The opiate-associated reduction in swelling was maintained over the entire period and did not develop tolerance. Rats with continuous administration of a catecholamine reuptake inhibitor or a β2-adrenergic receptor antagonist that began prior to disease induction also showed delayed presentation of clinical signs and less severe joint damage than controls (Levine et al. 1988). Smaller, yet significant, protective effects of these agents on joint injury were obtained when treatment was confined to the period either before or after presentation of clinical disease indicating that reduction in endogenous catecholamines was beneficial throughout the entire 28-day time course. Subsequent studies from other groups have confirmed the role of early β2 receptor activation as contributory to joint damage and report that administration of the β2-agonist salbutamol after disease onset was of clinical benefit and reduced inflammation and joint damage (Lubahn et al. 2004).

Pathways: Much as in the acute inflammatory models, section of major lumbar nerve trunks, greatly diminished the development of arthritis in the formerly innervated limb (Courtright and Kuzell 1965; Kane et al. 2005). In AA, dorsal rhizotomy in lumbar segments decreased the time to presentation of clinical symptoms, while in cervical segments magnitude of the radiographic symptoms was greatly exacerbated (Levine et al. 1986). Thus, the more limited sensory lesion has opposite effects from severing the entire nerve trunk including the sensory terminals and the sympathetic efferents. When i.t. capsaicin pretreatment was used to eliminate only the central peptidergic sensory fibers, rather than severing the entire dorsal root, severity of the inflammation was consistently ameliorated (Colpaert et al. 1983; Cruwys et al. 1995; Hood et al. 2001; Levine et al. 1986). The obvious differences between rhizotomy and central capsaicin are the maintained connection provided to the spinal cord via the non-capsaicin-sensitive afferents (myelinated and non-peptidergic C fibers). Both treatments preserve the sympathetic efferent fibers, and it is thought that both rhizotomy and intrathecal capsaicin preserve the DRGs and peripheral peptidergic fibers (Holzer 1991). Interestingly, capsaicin pretreatment also reduced AA-induced T cell infiltration into the synovium (Hood et al. 2001). In a combined lesion study, Levine pretreated animals with capsaicin, to produce a bilateral loss of the peptidergic afferent connection in conjunction with a unilateral rhizotomy. The rats displayed increased disease severity compared to the capsaicin only side, but not compared to control arthritic animals (Levine et al. 1986). The results outlined above point to the complexity of the wiring and the necessity to involve additional autonomic and endocrine feedback loops to explain the system.