The main findings of this study are: (1) that immunostaining with a GRP antibody is present in both primary afferents and non-primary glutamatergic axons in the superficial dorsal horn, (2) that immunostaining is partially blocked by pre-incubation with substance P, but not with NMB, (3) that EGFP-labelled cells in the GRP-EGFP mouse line are Pax2-negative, with most lacking PKCγ, and (4) that many of the axons derived from EGFP-labelled cells in this line express detectable levels of GRP-immunoreactivity, but are not immunoreactive with antibodies against substance P or PPTB.
Sources of GRP-immunoreactive axons in the superficial dorsal horn
Our finding that GRP immunoreactivity could be detected in the majority of CGRP
+ axons in laminae I-IIo is consistent with several previous reports that other antibodies raised against GRP or the related peptide bombesin stain cell bodies of peptidergic primary afferents in the dorsal root ganglion [
5,
57,
63‐
66], and that there is significant depletion of immunostaining in the dorsal horn following dorsal rhizotomy [
8,
57]. However, although GRP mRNA has been identified in dorsal root ganglion cells with
in situ hybridisation and the polymerase chain reaction by the Chen laboratory [
5,
20], studies from two other groups have failed to detect significant levels of the mRNA in dorsal root ganglia using either these methods or quantitative RNA sequencing [
8‐
10]. The question of whether GRP is actually expressed in primary afferents therefore remains controversial. Fleming et al. [
8], noting the high level of expression of NMB in dorsal root ganglia and the close sequence similarity between NMB and GRP, suggested that much of the GRP immunoreactivity identified in primary afferents could result from cross-reaction of the GRP antibodies with NMB. However, we found that relatively high concentrations of NMB (10
−5 M) were unable to block immunostaining with our anti-GRP antibody, even though this was completely blocked by GRP at 100 times this dilution (10
−7 M). It is therefore unlikely that the primary afferent staining that we saw with anti-GRP represents a cross-reaction with NMB. Goswami et al. [
9] suggested that GRP antibodies might cross-react with tachykinin peptides, such as substance P, which are highly expressed in primary afferents, and which show some similarity in their C-terminal amino acid sequence to GRP (−GLM in tachykinin peptides, compared to -GHLM in GRP and -GHFM in NMB). In addition, we had found in preliminary studies that including anti-substance P in the primary antibody mixture reduced the staining for GRP in these sections (MGM and AJT, unpublished data). We therefore tested the effects of pre-incubating with substance P on the staining seen with the GRP antibody, and found that this was suppressed. Although we cannot be certain that the GRP antibodies would bind to substance P in fixed tissue, this observation raises the possibility that the GRP-like immunoreactivity seen in peptidergic primary afferents in this study may represent a cross-reaction with substance P or the closely related peptide NKA. It is also possible that previous reports of primary afferent labelling with different GRP antibodies [
5,
57,
63‐
66] may have resulted, at least in part, from cross-reaction with tachykinin peptides. This could be tested in future studies by analysis of tissue from mice in which the gene for PPTA has been knocked out [
67].
GRP-immunoreactivity was also detected in boutons that were VGLUT2
+ but lacked CGRP, and most of these are likely to originate from local excitatory interneurons [
23,
24]. Consistent with this suggestion, we found that somatostatin, which is expressed by many excitatory interneurons in the superficial dorsal horn [
25], was frequently co-localised with GRP in VGLUT2
+ axons. It is likely that at least some of this represents genuine GRP, as it was sometimes co-localised with EGFP in axonal boutons in the GRP-EGFP mouse, and these cells did not express either of the two preprotachykinins.
Since only ~40% of the GRP
+/VGLUT2
+ boutons in this mouse line were immunoreactive for EGFP, it is possible that the staining in other non-primary (i.e. VGLUT2
+/EGFP
−) axonal boutons resulted from cross-reaction of the GRP antibody with tachykinin peptides, which are expressed by many excitatory neurons in this region [
35‐
37]. However, the failure to detect EGFP in these boutons may alternatively have resulted from weak expression (or lack of expression) of EGFP in some GRP-containing cells in this BAC transgenic line.
GRP-expression by excitatory interneurons
Several studies have reported that large numbers of cells with GRP mRNA are present in the superficial dorsal horn [
8,
10‐
15]. It has also been shown that these are likely to be excitatory neurons, since their survival is dependent on expression of transcription factors that are involved in determining a glutamatergic phenotype [
11,
12,
15]. In addition, it has been reported that most of the GRP-expressing cells are lost in mice lacking the testicular orphan nuclear receptor (TR4), in which there is preferential depletion of excitatory interneurons [
14]. However, Liu et al. [
20] recently claimed that cells with GRP mRNA were restricted to lamina I and state that “to date, little evidence exists to support that Grp mRNA in lamina I of the spinal cord is in fact translated into GRP protein”. This statement was apparently based on the failure of immunocytochemical studies to demonstrate GRP-immunoreactivity within cell bodies in the superficial dorsal horn. However, although some of the neuropeptides expressed by neurons in this region can be identified in their cell bodies with immunocytochemistry [
25,
68‐
70], others, such as substance P, NKB, dynorphin or the enkephalins are either not detected, or else are seen in very few cells. The failure to detect somatic staining for these peptides results from rapid transport of the peptide out of the cell body, and for this reason colchicine was administered in several early studies to prevent axoplasmic transport, therefore elevating the levels of neuropeptides in cell bodies [
71,
72].
Our finding that ~70% of the axon terminals that contained EGFP in the GRP-EGFP mouse were GRP-immunoreactive (and that they lacked tachykinins, with which the GRP antibody can cross-react) strongly suggests that these cells are indeed translating the mRNA into GRP. The lack of immunoreactivity in the other ~30% of EGFP+/VGLUT2+ boutons may reflect variable levels of GRP expression, with some axons containing amounts below the detection threshold of the GRP antibody. Alternatively it could result from ectopic expression of EGFP in this line. However, even if there is ectopic EGFP expression, it is evidently not random, as it seems to be restricted to excitatory (Pax2−) cells.
Excitatory interneurons account for ~70% of the neurons in the superficial dorsal horn [
73], and are thought to include several distinct subpopulations that are likely to have differing roles in the processing of sensory information [
14,
59,
74‐
76]. Several attempts have been made to define these functional populations, using morphological, electrophysiological, neurochemical and/or developmental approaches [
12,
24,
34,
44‐
47,
77‐
79]. Although the neurons in this region are morphologically diverse, certain distinctive classes have been identified among the excitatory cells in lamina II. These include vertical cells, which have dendrites that project ventrally from the soma and an axon that often enters lamina I where it may be presynaptic to projection neurons, and radial cells, which are small and have short radiating dendrites [
24,
34,
44‐
47,
80]. A third morphological group (central cells), has also been described [
34,
44‐
46]. These have relatively short dendritic trees orientated along the rostrocaudal axis. However, central cells are a more diverse set that includes both excitatory and inhibitory interneurons [
24]. A limitation of this scheme is that all morphological studies have found a significant fraction of lamina II interneurons (typically ~25%) that cannot be be assigned to any of these morphological classes [
24,
34,
44‐
46]. Based on the appearance of GRP-EGFP cells in sagittal sections, most did not appear to be vertical or radial cells, and so they are likely to belong to central or unclassified populations.
A neurochemical approach has proved useful in defining functional subsets among the inhibitory interneurons in the superficial dorsal horn. For example, we have identified four largely non-overlapping populations, those that express neuropeptide Y (NPY), galanin, neuronal nitric oxide synthase (nNOS) and parvalbumin, and found that these differ in their responses to noxious stimulation [
27]. Many of the cells that contained NPY or galanin were activated by noxious heat, as well as by subcutaneous injection of formalin or capsaicin. In contrast, the nNOS-containing cells responded to noxious heat and formalin injection, but seldom to capsaicin, while the parvalbumin cells were not activated by any of these stimuli. In addition, it has been shown that the parvalbumin-expressing cells give rise to axoaxonic synapses on myelinated low-threshold mechanoreceptive afferents [
81], while the nNOS and/or galanin populations have been implicated in the suppression of itch by counterstimuli [
33].
Several neurochemical markers are expressed by subsets of excitatory neurons, and it would be of value if we could also define non-overlapping populations among these cells. It had already been shown that there is little co-localisation of PPTB and substance P in axons in laminae I-II [
35,
38], and taken together with the results of this study, it therefore appears that different sets of excitatory neurons express PPTA, PPTB and GRP. Neurotensin is also present in some excitatory neurons in laminae IIi-III [
23,
82]. These are known to be different from those that express PPTB [
35], and we have found little or no co-localisation of neurotensin and substance P (MGM and AJT, unpublished data). Since most of the neurotensin-containing cells also express PKCγ [
41], it is likely that they are largely separate from the cells with GRP, as the EGFP
+ cells seen in this study were generally PKCγ-negative. This suggests that there may be four distinct neuropeptide-expressing populations among the excitatory neurons in laminae I-III: those with PPTA, PPTB, GRP and neurotensin. Since the axonal boutons of these cells can be recognised by co-localisation of the corresponding peptide with VGLUT2, it should be possible to use immunocytochemistry to investigate the postsynaptic targets for each population. Nothing is apparently known yet about the responses of the GRP-expressing interneurons to peripheral stimuli, but as GRP and the GRPR have been implicated in itch [
5,
6], it will be important to determine whether the GRP-expressing neurons in laminae I-II are activated by pruritic stimuli.