In this study, the time-dependant changes in the number of GAD65-IR inhibitory terminals were examined in the superficial dorsal horn of the rat spinal cord following polyethylene cuff induced CCI of the sciatic nerve. Using specific markers, and restrictive quantification techniques, we show conclusively that cuff application results in a significant but transient loss of GAD65 immunoreactivity, corresponding to inhibitory terminals, within the area of loss of IB4+ boutons in LII and, dorsal to it, in LI. The early time-course of the loss of GAD65-IR terminals correlated well with the loss of IB4+ terminals in LII. Furthermore, the pattern of alterations in terminal densities paralleled the changes in thresholds to both mechanical and thermal stimuli.
It is well known that loss of GABAergic or glycinergic inhibition in the dorsal horn leads to allodynia in naive rats , and that loss of inhibitory tone in the dorsal horn might underlie symptoms of neuropathic pain . In light of these results, many studies have explored the mechanisms which might contribute to this altered neurotransmission. Of particular interest was the question of whether or not nerve injury resulted in a loss of inhibitory interneurons. Conflicting evidence quickly emerged. Some studies suggested that GABAergic interneurons were lost after nerve injury as indicated by a loss of GABA immunoreactivity in the dorsal horn [23, 24, 27] and the presence of apoptotic markers [3, 21, 22] while other studies using stereological counts opposed these results by showing that neurons in LI-III were not lost after either CCI [2, 26] or spared nerve injury (SNI) . Independent of a change in the number of GABAergic cells following injury, it is likely that they either retract or extend their distal processes, a phenomenon that has been well described in hippocampal neurons after stress .
In this study, GAD65 was used to visualize inhibitory terminals by immunofluorescence and to clarify the effect of nerve injury on inhibitory neurons in the dorsal horn. GAD65 was chosen for a number of reasons. 1. It primarily localizes to nerve terminals [17–19]. 2. In laminae I and II, where this study was focused, inhibitory profiles with high levels of GAD65 are relatively common; this is not the case in more ventral laminae where profiles with high levels of GAD67 dominate . 3. GAD65 knockout mice have impaired GABA synaptic function and a lower baseline pain threshold indicating a direct role for this isoform in both inhibition and pain . Using software parameters that specifically allowed puncta the size of GABAergic boutons with a minimal staining intensity to be counted as inhibitory terminals, we have clearly shown that the density of GAD65+ inhibitory terminals was reduced in the ipsilateral superficial dorsal horn of the spinal cord after cuff injury. The loss of GAD65-IR inhibitory terminals was not constant over time or space. The loss of inhibitory terminals was most significantly reduced at 3–4 weeks and recovered partially by 56 days. Furthermore, the loss of GAD65-IR terminals was greatest in LII, the region where the majority of IB4+ fibres terminate and are lost after nerve injury.
The finding that GAD65-IR density is temporally altered by nerve injury is consistent with a number of complementary findings. Meisner et al.  showed that GAD65/67 immunoreactivity in the dorsal horn is reduced at 8 weeks after spinal cord injury (SCI), as are the protein levels of GAD67 and GAD65 at 6 weeks. Moore et al.  showed using Western blot that GAD65 protein is transiently reduced in the ipsilateral dorsal horn by CCI and SNI; however, unlike us, they saw a full restoration of GAD65 levels at 4 weeks post CCI. These immunoblotting approaches used only provide a global assessment of GAD protein levels in the entire ipsilateral dorsal horn and likely do not reflect what is happening in LI and LII, regions important in nociceptive transmission. Finally, Eaton et al.  show that GAD67-IR cell bodies in LI-III initially decrease in number 3 days post CCI but begin to increase, eventually surpassing levels in intact animals at 8 weeks after nerve injury. The different time-course seen in this study is most likely attributable to the fact that cell bodies, rather than terminals, were counted. The major advantage of our study over the previous studies is that we were able to specifically quantify the number of inhibitory terminals at various time points, within the specific area in the spinal dorsal horn where the afferents affected by the nerve injury terminate.
It is possible that the loss of GAD65 immunoreactivity does not necessarily reflect a loss of inhibitory terminals, but instead simply a down regulation of the protein. Because it has been shown that some boutons in the superficial dorsal horn express high levels of GAD65 and low levels of GAD67 , it is possible that these boutons would be unable to synthesize GABA after GAD65 downregulation, effectively rendering these synapses non-functional (as far as GABAergic transmission is concerned). One limitation of the current study is that we investigate the changes in GAD65 immunoreactivity only, and do not consider GAD67. While some studies suggest that GAD65 is more dramatically reduced , others report that GAD65 and 67 are similarly decreased by nerve injury . If GAD67 is unchanged or even up regulated, some inhibitory synapses could remain functional. Contrary to this, the frequency of inhibitory postsynaptic currents (mIPSCs) is reduced by CCI  and in GAD65 knockout animals , indicating that inhibitory neurons release less GABA in these models. To more globally assess the entire GABAergic inhibitory terminal population in neuropathic pain, antibodies against GAD67 could be used in combination with those against GAD65 or even by using antibodies that indiscriminately recognize both isoforms. Alternatively, the vesicular GABA transporter (VGAT), which acts as an anatomical marker for inhibitory terminals [34, 35], could be used.
One of the most interesting and relevant findings of this study is that the behavioural results matched well with the loss and recovery of both IB4+ and GAD65-IR terminals. As the density of both IB4+ fibre terminals and GAD65-IR inhibitory terminals decreased from 5–21 days, sensitivity to both mechanical and thermal stimuli increased. As GABAergic inhibitory terminals are known to be presynaptic to the central boutons of type Ia glomeruli [9, 36], it is hypothesized that the GAD65-IR terminals might retract and become lost after glomerular IB4+ central terminals undergo degenerative atrophy. This is supported by the fact that we found that loss of GAD65-IR terminals was significantly correlated with the loss of IB4+ terminals in LII from 5–14 days after cuff placement. Between 14 and 28 days, a time at which both IB4+ and GAD65-IR terminals were maximally depleted, hypersensitivities were greatest. It is possible that the disrupted synaptic organization in the dorsal horn at this time might drive the development of neuropathic pain. In normal animals, nociceptive information is forwarded to the spinal cord via two distinct populations of C fibres: the peptidergic and non-peptidergic. Evidence suggests that nerve injury produces more profound and longer lasting changes to both the central and peripheral terminals of non-peptidergic C fibres in comparison to the peptidergic population, which appears to be more resilient [13, 37–39]. We suggest that nociceptive information following nerve injury might be preferentially transmitted through the less affected peptidergic population. Given that peptidergic afferents do not receive axo- or dendroaxonic synapses from GABAergic neurons and therefore undergo less direct inhibition , it is conceivable that nociceptive signals are more easily transmitted from these remaining peptidergic fibres to projection neurons. It might be argued as well that as the loss of IB4+ terminals and of GAD65-IR is parallel, the loss of inhibition would be on terminals that disappeared with the lesion. However, such reasoning would assume that the major synaptic targets of GAD65-IR boutons in LIIi are the central boutons of Type Ia synaptic glomeruli, which is far from the truth. Indeed, it should be kept in mind that the great majority of GABA- and glycine-containing terminals are presynaptic to spinal neurons, not to primary afferent terminals [36, 40].
In LI, despite the few IB4 fibres that terminate there , the loss and recovery of GAD65 terminals followed the same time course as within LII, albeit less severe. These data from LI suggest that it is mostly the inhibitory terminals that contact dorsal horn neurons that are lost after CCI. In this way, the loss of inhibitory terminals could result in impaired dorsal horn inhibition and altered nociception.
Finally, we suggest that the amelioration in behaviour seen at later time points might be in part due to a recovery of the normal synaptic architecture of type Ia glomeruli and the reestablishment of an equilibrium between the peptidergic and non-peptidergic populations. In line with this idea, regenerative proliferation of IB4+ terminals is observed after cuff , and 40 days after crush injury growth cones of primary afferent terminals have been shown to re-establish synapses with dendritic processes from substantia gelatinosa neurons . Future studies using EM would be required to determine if GAD65-IR terminals resume proper connectivity and presynaptic inhibition to non-peptidergic C fibres. Recently it was shown that when immature GABAergic interneurons were transplanted and integrated into spinal cord circuitry, the mechanical hypersensitivity associated with SNI was reversed . These results suggest a potential therapeutic role for restoring normal inhibitory function in the dorsal horn in chronic pain conditions.