In this study, we demonstrate the presence in lamina I of a significant number of boutons originating from non-peptidergic afferents immunoreactive for P2X3 or binding the lectin IB4. We also provide evidence at the confocal microscopy level that IB4+ boutons are in apposition to lamina I projection neurons immunoreactive for the NK-1r. Lastly, we provide ultrastructural evidence of synapses between P2X3-IR boutons and lamina I dendritic profiles immunoreactive for the NK-1r.
Because most of this study was carried out using confocal microscopy, we could not fully ensure that boutons from non-peptidergic afferents were presynaptic to neurons in lamina I. For this reason, we carried out an ultrastructural study using antibodies against P2X3 and provided direct evidence of synapses between P2X3-IR axonal boutons and dendrites and cell bodies in lamina I. We also performed at the ultrastructural level a double labelling for P2X3 and the NK-1r that demonstrated that some of the structures innervated by these boutons expressed the NK-1r, in agreement with what was assumed from the confocal data. We could observe unequivocal evidence of synaptic contacts, in spite of the fact that we had to use a fixative without glutaraldehyde and pre-treatment of the sections with a detergent for a short period to obtain P2X3 staining at the ultrastructural level. We were unable to use IB4 binding for electron microscopy because of the bad penetration of the lectin in the tissue. Conversely, we were unable to use P2X3 immunoreactivity for the study of the innervation of the neurons at the confocal level because of antibody incompatibilities when performing the required triple labeling. It should be pointed out that Naim et al.
 examined at the ultrastructural level ultrathin sections recut from thicker sections previously examined by confocal microscopy. Naim et al.'s study showed that immunoreactive varicosities seen in close apposition to the membrane of NK-1r-IR under the confocal microscope actually formed synapses when viewed under the electron microscope. Therefore, we are confident that the IB4+ varicosities in apposition to NK-1r-IR lamina I projection neurons should establish synapses in a high proportion of cases.
It is known that in rat there is a small proportion of peptidergic sensory fibers in lamina I and II that colocalize CGRP and somatostatin, do not respond to NGF and bind IB4
[19, 20]. This obliged us to investigate the co-localization of IB4 binding and P2X3 immunoreactivity, our markers of non-peptidergic nociceptive C fibers, with CGRP immunoreactivity. Indeed, as predicted, we found a limited level of co-localization of either marker of non-peptidergic afferents with CGRP immunoreactivity. However, most varicosities that we observed in lamina I which were IB4+ or P2X3-IR did not co-localize CGRP immunoreactivity, what reassured us regarding the validity of our findings regarding the overall innervation of lamina I by non-peptidergic afferents. But it was still possible that a subpopulation of these afferents innervating NK-1r-IR lamina I projection neurons would represent exactly this minor subpopulation that is simultaneously CGRP-IR and IB4+. Therefore, we carried out a quadruple labeling in which we detected CTb (the retrograde label), NK-1r, CGRP and IB4 binding. Unfortunately, because of antibody incompatibilities, we could not perform P2X3 staining and had to use the lectin IB4 conjugated to a fluorochrome. Although our confocal microscope can identify 4 separate signals reliably using the multi-track approach, it was technically impossible to perform a quantitative analysis of the innervation of lamina I cells using more than 3 signals (IB4, CTb and NK-1r) because of photobleaching during the performace of the Z-stacks. However, we examined enough cells using the quadruple labeling to ensure that the proportion of axonal boutons colocalizing IB4 binding and CGRP immunoreactivity in apposition to lamina I neurons was low. An example of the quadruple labeling is given on Figure
Innervation of lamina I projection neurons by non-peptidergic C fibers
The main objective of this study was to investigate the issue of the innervation of NK-1r-IR lamina I projection neurons by non-peptidergic C fibers because in our knowledge a systematic study combining labeling of these sensory fibers with labeling of these lamina I neurons had never been done. This is an important issue for the reasons given below. Previous work by Lu and Perl
, using simultaneous whole-cell recordings from pairs of neurons, provided some evidence that input from primary afferent C fibers terminating in inner lamina II may reach lamina I projection neurons via interposed interneurons. Since the great majority of C fibers innervating inner lamina II are non-peptidergic, it has been later suggested that the above pathway may be important for pain-related information conveyed by non-peptidergic C fibers to reach lamina I projection neurons which then project to supraspinal levels. If this is true, lamina I nociceptive projection neurons would receive direct input from peptidergic afferents and polysynaptic input from the non-peptidergic afferents
Alternatively, it has been proposed that the non-peptidergic afferents are part of a distinct and parallel pathway from that of their peptidergic counterpart. This view obtained some support from a study by Braz et al.
 in a transgenic mouse, which demonstrated that the termination of non-peptidergic afferents on lamina II would be on excitatory interneurons, which in turn would synapse on deep lamina V projection neurons with ascending connections to the amygdala, hypothalamus and bed nucleus of stria terminalis. As they found minimal connection with lamina I neurons expressing the NK-1r, the above group proposed the involvement of the non-peptidergic afferents in the affective/emotional component of pain. Alternatively, other studies also in the mouse, proposed that distinct subsets of primary sensory afferents selectively mediate responses to different stimulus modalities. These studies provided some evidence suggesting that, in the mouse, non-peptidergic afferents play a particular role in transmitting mechanical pain, as opposed to the peptidergic which would be involved in conveying thermal pain
[26, 27]. This idea of divergent pain pathways can be criticized since all the studies supporting it were performed in mice, which have been shown to demonstrate an explicit dichotomy between the C fiber populations that does not apply to rats and higher order species such as primates. In particular, the vanilloid TRPV1 receptors are localized only in the peptidergic fibers in mice
[28, 29] whereas they are present in both peptidergic and non-peptidergic afferents in the rat and higher species
Our data provides evidence that non-peptidergic primary afferents establish direct connections with NK-1r-IR lamina I projections neurons, in addition to the possible indirect connections via an interposed interneuron suggested in previous studies
. Although the non-peptidergic primary afferent termination in lamina I may seem as a minor contribution to the lamina I synaptic circuitry when compared to their termination in inner lamina II, they may prove to have a significant role in the transmission of nociceptive signals. Indeed, the terminals of these fibers in inner lamina II represent the central bouton of type Ia glomeruli, and are involved in complex modulatory circuits involving GABAergic presynaptic inhibition
. A direct termination of non-peptidergic afferents on lamina I cells would bypass such modulation. Since non-peptidergic afferents have a more extensive distribution in the epidermis than peptidergic afferents
, show larger and more sustained responses to capsaicin than peptidergic afferents
 and signal mainly via glutamate without the presence of neuropeptide co-transmitters, the activation of NK-1r-IR lamina I projection neurons by these non-peptidergic afferents would result in a different signal transmitted than when activated by the peptidergic afferents.
Our results revealed that, on average, there were 6.7 appositions from non-peptidergic boutons per 100 μm of length of NK-1r-IR lamina I projection neuron membrane. How does this compare to the peptidergic innervation of these cells as shown in other studies? Although such comparison is not fully legitimate because of methodological differences with our study, we report it here to give the reader an idea of the order of magnitude of the two types of innervation. Indeed, Todd at al.
 have detected a density of peptidergic innervation of 16.2 contacts/100 μm of NK-1r-IR lamina I projection neuron membrane. Therefore, the non-peptidergic innervation is of lamina I NK-1R-IR neurons is substantial, although less abundant than the peptidergic. Although the focus of this study was mainly on NK-1r-IR projection neurons in lamina I, there is also a population of NK-1r-positive neurons in deep laminae III-V, which also project supraspinally and send processes to laminae I-II
. These neurons have been shown to receive on average 18.8 contacts per 100 μm of dendrite length from substance P-IR boutons in these superficial layers
. Another study has reported that these deeper neurons receive contacts from IB4-binding non-peptidergic afferents in laminae I-II, although in much lower density (just 2 appositions per 100 μm of length of NK-1r-IR dendrite membrane) than for the peptidergic afferents
. These values of non-peptidergic innervation of laminae III/IV neurons are considerably lower than those we found in the current study.