The present results confirm and extend previous evidence of a convergence of trigeminal and GON afferents in the TCC (reviewed in [26]), and show that GON stimulation exerts different modulatory effects on trigeminal input in the TCC of intact animals and in animals that display pain and allodynia induced by the CCI of the IoN. When paired at intervals <300 ms, GON stimulation facilitates, in the TCC, neuronal responses to subsequent innocuous tactile stimuli to the vibrissal pad in control cases, whereas the same stimulation evokes an inhibition of vibrissal responses in CCI-IoN animals, an effect that is mediated by GABAergic and Glycinergic mechanisms. Vibrissal and GON stimulation-evoked spike firing of TCC neurons involve NMDA glutamatergic receptor activation, since it was reduced by the application of a NMDA-receptor antagonist, as also shown for responses to occipital muscle input in the TCC [33]. These findings may help to explain the beneficial effects of GON stimulation in treating some refractory craniofacial pain syndromes that involve trigeminal territories [16, 47, 48].
Primary afferents putatively driving the responses observed in TCC
Impulses elicited by mild mechanical stimuli applied to the whisker pad are likely to emerge from all the mystacial, non-mystacial and intervibrissal fur low-threshold mechanoreceptors (LTMR) innervated by thickly- (Ab) and thinly-myelinated (Ad) fibers [49-52]. In the spinal cord, Ab fibers distribute mainly in lamina III, with terminal and en passant boutons. Most boutons form the central component of type IIb glomeruli [53]), receiving axo-axonic contacts from presynaptic axons that express both GABA and Glycine (55-75%, depending on the type of afferent), GABA only (25-40%), or only Glycine (0-10%), and being in turn presynaptic to dendrites, which only in a small fraction of cases express either or both of these inhibitory transmitters [54-57]. In deeper laminae afferent boutons from Ab fibers are replaced by simpler boutons, many of which still display triadic contacts with presynaptic axons that simultaneously synapse on the afferent bouton and a postsynaptic dendrite [56, 57]. Ad fibers ending in laminae IIi-III mainly arise from LTMR in hair follicles [58, 59]. A similar pattern of LTMR fibers was described in the caudal Sp5C in the cat [60, 61]. In the rat, large boutons in terminal arbors of myelinated, Ab fibers, occupy the same layers with a somatotopic pattern [62-64].
The shortest-latency responses to GON electrical stimuli clearly fall within the range of Ab afferents. Thick fibers with abundant large- and medium-sized terminal and en passant boutons are distributed in the lateral one-third of laminae IIi-IV of the upper cervical segments and, more sparsely, along the lateral two-thirds of Sp5C and other lower brain stem structures [27]. Moreover, while the GON stimulation parameters (single pulse, twice the short latency response threshold) are unlikely to recruit unmyelinated C afferents, longer latency responses appearing within the 100 ms interval that mediate the GON and vibrissal stimulations likely represent Ad fiber activation. Fine myelinated fibers, including Ad and low conduction velocity Ab afferents [65] from spinal nerves, distribute mainly in laminae I-IIo and V [66], but may also reach intermediate laminae [67]. Myelinated afferents from GON in laminae II and IV form dense meshworks with terminal and en passant boutons of assorted sizes, while those in lamina I mainly consist of long thin axons decorated with abundant en passant varicosities, most of them small [27]. In deep laminae, boutons from finely myelinated fibers often make the central element of Type IIa glomeruli, which are postsynaptic to axonal boutons expressing GABA and/or Glycine, as well as GABA-expressing dendrites [53, 54]. In laminae I-IIo these fibers establish simpler axo-dendritic synapses or make the central element of Type I glomeruli, which receive only GABAergic axo-axonic contacts [53, 54].
TCC neurons driven by vibrissal and/or GON input
All neurons in laminae I-VI had ipsilateral orofacial mechanoreceptive fields, with a predominance of those responding to low-threshold tactile input in laminae III-IV [68]. Laminae IIi-IV of the spino-medullary dorsal horn, where most recordings were made, are the main target for LTMR afferents and a key node for early processing of tactile input. In these laminae in the spinal cord up to seven types of excitatory and four types of inhibitory LTMR interneurons have been described on the basis of molecular-genetic, morphological and electrophysiological profiling [69-72]. These cells account for 98% of all neurons in the region, with a 2.3-to-1 ratio between excitatory and inhibitory cells, and just 2% of ‘projection’ neurons, which send their axons to supraspinal levels through the dorsal or lateral columns [71]. Most, if not all can be monosynaptically driven by primary afferent input to these laminae. Both nociceptive and non-nociceptive input from Ad afferents reach at least some excitatory (such as PKCg neurons) and inhibitory (such as islet cells) neurons placed in lamina II and superficial part of III, as well as deeply-placed projection neurons with dendrites extending to superficial laminae [73]. The latter may also show convergence of all kinds of low- and high-threshold afferents [74]. Although comparably less thoroughly investigated, the corresponding laminae in the medullary dorsal horn contain similar neuronal populations and afferent input [75-77].
Although we cannot be sure of the cell type from which we obtained the recordings, indirect data may shed some light on this issue. Spontaneous firing is very low in both WDR and low-threshold mechanosensory-responsive (LTM) neurons of Sp5C under control conditions; following CCI-IoN, however, spontaneous activity markedly increased in the WDR, but not in the LTM [39]. Moreover, it was recently found that excitatory neurons in laminae I-II in rat lumbar spinal cord showed fast adaptation to light tactile stimuli and very low spontaneous firing, whereas inhibitory neurons with a variety of non-adapting responses had much higher spontaneous activity ([78]; see also [38]). Should these findings be applicable to the neurons recorded in TCC, it would indicate that these neurons are likely to be excitatory interneurons or projection neurons.
Responses in TCC to GON and vibrissal stimulation are differentially affected by CCI-IoN
Injured peripheral nerves exhibit increased ectopic firing, originating in the nerve itself [79] and/or medium-sized neurons in the Ab and Ad range in the affected spinal ganglia [80-82]. Both low- and high-threshold mechanosensory ganglion neurons become hyperexcitable and exhibit increased activity as attested by changes in several electrophysiological parameters, and thus may transfer as nociceptive messages normally innocuous tactile stimuli [83]. This abnormal input is responsible for all or most of the increased level of spontaneous activity in dorsal horn WDR neurons, because the conduction block of a constricted nerve proximal to the constriction abolishes it [79]. Nevertheless, the injured nerve causes additional effects, not only bringing about an altered drive on dorsal horn neurons. The sensory neurons affected by nerve lesions also undergo rapid and profound transcription changes [84, 85], so that large afferents change their expression of many transmitters, peptides, and other factors which contribute to drive sensitization and nociceptive responses in DH neurons (reviewed in [86]).
In addition to increased spontaneous activity, we found that the response in TCC to light tactile stimulation of the vibrissal pad increased significantly under CCI-IoN, as previously reported for somewhat more rostral levels of Sp5C [39]. A similar finding had also been reported in the lumbar spinal cord after sciatic CCI or spinal nerve ligation [38, 87]. While this increased response could be attributed to both an excess of incoming signals through the IoN and to hyperresponsive neurons in the TCC, it must be noted that such heightened response to peripheral stimuli was not found upon GON stimulation (Fig. 7A). When a spinal nerve is injured, there is an increased expression of peptides involved in central sensitization in small and medium-sized DRG neurons contributing to a neighboring spared spinal nerve [88], as well as electrophysiological features of sensitization in Ab, Ad and C nociceptors [52]. These uninjured neurons show a notable overlap of thin afferents in the superficial laminae of the dorsal horn. However, while the GON and trigeminal terminals also show a degree of overlap in the TCC, the corresponding ganglia and peripheral course of these nerves are quite apart and GON ganglia and nerve are undamaged. The fact that responses to GON stimulation did not vary between the controls and CCI-IoN cases suggests that neurons in a TCC sensitized territory only appear hyperresponsive when driven from an injured (IoN) nerve, but not when activated from a converging, but intact (GON) nerve (Fig. 7A). Remarkably, this apparently unchanged response to GON stimulation was ensued by different effects on succeeding responses to facial tactile stimuli depending on whether IoN is, or is not injured, as discussed below.
A conditioning stimulus to GON increases the TCC response to vibrissal stimulation in controls, but reduces it in CCI-IoN cases
Despite the expanding use of GON stimulation to treat a variety of craniofacial pain disorders (reviewed in [89, 90]), scarce attention has been paid to the basic neural mechanisms that may underlie this connection. Early electrophysiological and anatomical findings, mostly in cats, showed a convergence of GON and trigeminal afferents on upper cervical or medullary dorsal horn (see [27]). More recently, the existence was proved of a functional convergence in Sp5C of nociceptive and non-nociceptive input from different trigeminal domains, such as the supratentorial dura mater and superficial territories of the first and second trigeminal branches [91]), as well as of a direct functional coupling in TCC neurons of dural and cervical afferents conveyed by the GON [26, 28]. However, the interaction between low-threshold afferents and the GON input to TCC, under control and neuropathic conditions, still remained unexplored.
In controls, light vibrissal stimulation elicited a stronger response in the TCC when preceded by a brief electric shock to the GON. This facilitatory effect was due to the activation of NMDA receptors since the effect was blocked by AP-5. Since we could only ascertain that it was the same unit that responded to vibrissal and GON stimulation in some cases, the possibility exists that the GON effects on vibrissal responses are mediated by mono- or polysynaptic connections in the TCC. Those neurons receiving synapses from different sources may show heterosynaptic facilitation, whereby activation of some synapses on a target neuron may potentiate other inactive synapses on the same neuron. This has been demonstrated for low frequency stimulation of C and Ad fibers in a dorsal root enhancing responses in a motor neuron to input from an adjacent intact root [92], for C fiber stimulation by topical mustard oil unmasking (potentiating) low-threshold A fiber input onto nociceptive-specific and wide dynamic range neurons at superficial or deep dorsal horn laminae [93], or for C or Ad stimulation-mediated presynaptic potentiation of GABAergic synapses on lamina I neurons [94]. In absence of any direct proof, we think it unlikely that Ad (or Ab) input on TCC neurons could potentiate the response to the vibrissal input in the same neurons by a similar mechanism. Moreover, the GON responses in the TCC were all but eliminated by blocking GABA transmission, and enhanced by blocking Glycine transmission (Fig. 7B), suggesting a release of an interposed Glycinergic inhibition probably through a GABAergic neuron, and therefore indicating a predominance of a multisynaptic enhancing effect of the GON activation on the vibrissal responses.
In the CCI-IoN cases, GON stimulation reduced the response to vibrissal stimulation to a level comparable to that seen in the control cases without prior GON stimulation (Fig. 7A). This decrease occurred within a significant increment in the spontaneous activity of the same TCC neurons, consistent with the overall state of GABA- and Glycine-dependent disinhibition in the affected spinal or medullary territory after a neuropathic injury ([39, 95]; see below). Given the essentially unchanged TCC response to GON stimulation compared to controls, it is therefore likely that input from the GON was able to ‘rescue’ local inhibitory circuits that were down-regulated by the nerve injury, thus reducing the TCC response to incoming vibrissal input. To summarize, under CCI, vibrissa-responding TCC units display heightened excitability revealed by their increased basal spontaneous discharges. Hence, GON conditioning is unlikely to affect directly the excitability level of TCC neurons but rather seems to temporally disinhibit local inhibitory GABAergic neurons, thus resetting presynaptic GABAergic and postsynaptic Glycinergic inhibition to normal levels.
Inhibitory transmission in theTCC and its involvement in the effects of GON stimulation on vibrissal responses
Inhibitory interneurons account for 30-40% of the neurons in laminae I-III of the rat spinal cord, most of which are enriched in GABA, fewer in Glycine, and an undetermined fraction expressing both transmitters in different combinations with other transmitters and neuropeptides [96-98]. In Sp5C about 33% of the neurons express GABA and/or Glycine; of these immunolabeled neurons up to 52% co-express GABA and Glycine, and 17% and 32% express only GABA or only Glycine, respectively [44]. A population of the calcium-binding protein parvalbumin-expressing cells, morphologically assigned to the islet and central types in laminae IIi and dorsal III, co-express GABA and Glycine and have as their predominant synaptic output most of the axo-axonic synapses on boutons from Ab and Ad fibers within the same laminae [99, 100].
In addition to exerting presynaptic control on primary afferents, all the inhibitory interneurons make axodendritic and/or axosomatic contacts [71] and, through finely tuned mono- and polysynaptic effects, regulate the transmission of innocuous somatosensory input to other neurons that send both nociceptive and non-nociceptive signals to supraspinal levels (reviewed in [69, 101, 102]. In control conditions, these postsynaptic effects are aimed at controlling the excitability of the dorsal horn neurons, and blocking the flow of excitatory signals to nociceptive-specific projection neurons [103]. Dual simultaneous recordings of synaptically linked interneuron pairs in laminae III-IV have shown that inhibitory synapses outnumber the excitatory ones by 2:1 [104]. In these deeper laminae inhibitory interneurons expressing GAD67 receive predominantly Glycinergic synaptic contacts and are also under tonic extrasynaptic Glycinergic control, in contrast to similar interneurons in laminae I-IIo, which are mostly targets for GABAergic synaptic and extrasynaptic modulation [105]. The importance of Glycinergic inhibition was also found in patch clamp studies on spinal slices, which showed that, although 55% of the neurons in lamina II receive GABA and Glycine synaptic input and all the neurons in laminae III-IV receive GABA and/or Glycinergic input, the inhibitory synaptic transmission is characterized in all cases by a dominant role of Glycinergic inhibition [106]. Moreover, GABA would be acting on presynaptic GABAB receptors, giving a negative feedback signal on the inhibitory afferent (reviewed in [107]). A similar GABA-mediated negative feedback has been shown to occur on glutamate release by Ab terminals in laminae III-IV [108].
The separate or combined blockade of GABA or Glycine transmission revealed a complex involvement of inhibitory circuits in the effects of GON stimulation on vibrissal responses, and strongly suggests that the synaptic weight of GABAergic and Glycinergic inputs modulating GON and vibrissal inputs is not the same, under either control or CCI-IoN conditions. This complexity was compounded by the uncontrolled degree of penetration of the drugs, a variable in the design intended to mimic the effects of applying agonists or antagonists of inhibitory neurotransmitters intrathecally or intracisternally in both basic research [109-113] and clinical settings [114, 115]. Yet, our findings may be supported by known local connections in the dorsal horn and provide new data on the possible involvement of these circuits under conditions of neuropathic pain.
The GABAA receptor blockade with Bic reduces the response to GON stimulation in both controls and CCI-IoN cases, probably due to a disinhibition of Glycinergic neurons (see above). The effects on the vibrissal responses were negligible in both cases (Fig. 6), but these responses differed when preceded by GON stimulation: The facilitatory effect of GON pre-conditioning shown in controls disappears, consistent with the diminished response to GON stimulus; by contrast, under CCI-IoN, GON stimulation still facilitates the response to the successive facial stimulus, probably because of the local state of disinhibition caused by the CCI [39]. Glycinergic blockade, however, results in a quite different effect on the spontaneous activity, which increases in controls and even more under CCI, and the response to GON, which also increases in controls, but is unaffected by CCI. This would be consistent with the dominant role of Glycinergic inhibition [106], whose removal would disinhibit units responding to the vibrissal input in controls, but would also disinhibit the down-regulated GABAergic transmission under CCI.
As expected, the combined removal of GABA- and Glycinergic transmission produces a marked increase in the responses to the GON and to the vibrissal inputs, but an intriguingly contrasting effect of the conditioning effect of GON stimulus on vibrissal input in controls and in CCI-IoN cases (Fig. 7B). In controls, a strong response to facial stimulation follows the strong response to GON stimulation, consistent with a global state of disinhibition. The heightened response to GON stimulation persists under CCI-IoN, but now the responses to vibrissal stimulation are somewhat decreased by a preceding GON stimulus. Although lacking experimental proof, it may be speculated that the strong response to GON input might have stimulated the release of GABA from local interneurons. With the GABAA receptors blocked by Bic, GABA could still activate GABAB receptors by the extrasynaptic diffusion of the transmitter, not just on a cell postsynaptic to the interneurons, but on other neurons as well [116]. These cells thus become less excitable by the activation of postsynaptic GABAB receptors [117], as will the vibrissal afferents themselves, since presynaptic GABAB receptors are expressed in terminals of Aβ fibers in the spinal laminae III-IV [108]. The circuit diagram sketched in Fig. 7C gives grounds to the proposal that GON-driven alteration of the TCC responses to low-threshold input from the face can be explained, at least in part, by the interplay of inhibitory circuits that modulate the activation of presumed excitatory neurons.