Expression of cholecystokinin by neurons in mouse spinal dorsal horn

Abstract Excitatory interneurons account for the majority of dorsal horn neurons, and are required for perception of normal and pathological pain. We have identified largely non‐overlapping populations in laminae I‐III, based on expression of substance P, gastrin‐releasing peptide, neurokinin B, and neurotensin. Cholecystokinin (CCK) is expressed by many dorsal horn neurons, particularly in the deeper laminae. Here, we have used immunocytochemistry and in situ hybridization to characterize the CCK cells. We show that they account for ~7% of excitatory neurons in laminae I‐II, but between a third and a quarter of those in lamina III. They are largely separate from the neurokinin B, neurotensin, and gastrin‐releasing peptide populations, but show limited overlap with the substance P cells. Laminae II‐III neurons with protein kinase Cγ (PKCγ) have been implicated in mechanical allodynia following nerve injury, and we found that around 50% of CCK cells were PKCγ‐immunoreactive. Neurotensin is also expressed by PKCγ cells, and among neurons with moderate to high levels of PKCγ, ~85% expressed CCK or neurotensin. A recent transcriptomic study identified mRNA for thyrotropin‐releasing hormone in a specific subpopulation of CCK neurons, and we show that these account for half of the CCK/PKCγ cells. These findings indicate that the CCK cells are distinct from other excitatory interneuron populations that we have defined. They also show that PKCγ cells can be assigned to different classes based on neuropeptide expression, and it will be important to determine the differential contribution of these classes to neuropathic allodynia.

Dorsal horn excitatory interneurons are functionally diverse, and there have been many attempts to define distinct populations among these cells (Todd, 2017). Most recent studies have used neurochemical criteria, in particular the expression of neuropeptides. We have identified four largely non-overlapping populations in laminae I-II, based on the expression of neurotensin, neurokinin B (NKB), substance, P and gastrin-releasing peptide (GRP), and have shown that between them, they account for over half of the excitatory interneurons in this region (Gutierrez-Mecinas et al., 2017;Gutierrez-Mecinas, Furuta, Watanabe, & Todd, 2016). While the neurotensin and NKB cells can be identified by immunocytochemistry, we revealed the substance P cells by spinal injection of adeno-associated viruses (AAVs) coding for Cre-dependent expression cassettes into mice in which Cre recombinase was knocked into Tac1 (the gene for substance P). The GRP cells were identified by the presence of enhanced green fluorescent protein (eGFP) in a BAC transgenic mouse line (GRP::eGFP) in which eGFP is expressed under control of the GRP promoter (Gutierrez-Mecinas, Watanabe, & Todd, 2014;Mishra & Hoon, 2013;Solorzano et al., 2015). Neurotensin and NKB neurons, which are concentrated in inner lamina II (lamina IIi) and lamina III, partly correspond to cells that express protein kinase Cγ (PKCγ; Gutierrez-Mecinas et al., 2016). PKCγ-expressing cells are of particular interest, because they are thought to form part of a circuit that is responsible for mechanical allodynia (Lu et al., 2013;Malmberg, Chen, Tonegawa, & Basbaum, 1997;Miraucourt, Dallel, & Voisin, 2007).
A recent transcriptomic study by Häring et al. (2018) identified 15 classes of excitatory dorsal horn neuron, and their results are broadly consistent with the findings described above, as their classes include cells with high levels of the mRNAs for neurotensin, NKB, and substance P. Another peptide that featured in their classification scheme was cholecystokinin (CCK), which was expressed in three separate populations of excitatory neurons. Early reports had indicated that neurons with CCK mRNA were most numerous in lamina III and scattered through the other laminae (Abelson & Micevych, 1991;Cortes, Arvidsson, Schalling, Ceccatelli, & Hokfelt, 1990;Schiffmann, Teugels, Halleux, Menu, & Vanderhaeghen, 1991), while a recent study found that CCK defined a specific population of excitatory neurons in the deeper laminae of the dorsal horn (Abraira et al., 2017).
The first aim of the present study was to estimate the proportion of excitatory neurons in laminae I-III that expressed CCK. This was achieved with two approaches: immunocytochemistry with an antibody against the precursor pro-CCK (Booker et al., 2017) and multiple-labeling fluorescent in situ hybridization. We also examined the extent of overlap with the four populations of excitatory interneurons that we had identified in the superficial dorsal horn. Our preliminary studies indicated that some pro-CCK-positive neurons in laminae II-III showed strong PKCγ-immunoreactivity, and we therefore compared expression of neurotensin, CCK, and PKCγ, because neurotensin is also expressed by cells with high levels of PKCγ in this region (Gutierrez-Mecinas et al., 2016). Finally, because Häring et al. (2018) reported that one of the CCK populations could be identified based on expression of thyrotropin-releasing hormone (TRH), we looked for cells that contained mRNAs for both CCK and TRH and tested whether these corresponded to PKCγ-expressing neurons.

| Animals
All experiments were approved by the Ethical Review Process Applications Panel of the University of Glasgow, and were performed in accordance with the UK Animals (Scientific Procedures) Act 1986.
Five wild-type C57BL/6 mice (either sex, 18-27 g) and 3 GRP:: eGFP mice (Gutierrez-Mecinas et al., 2014;either sex, 19-27 g) were deeply anesthetized with pentobarbitone (20 mg i.p.) and perfused through the left cardiac ventricle with a fixative consisting of 4% freshly depolymerized formaldehyde in phosphate buffer. Mid-lumbar spinal cord segments (L3-L5) were removed and postfixed for 2 hr at 4 C in the same fixative before being processed for immunocytochemistry. Spinal cord tissue was also obtained from three male Tac1 Cre mice (19-22 g;Harris et al., 2014) that had received intraspinal injections of AAV (serotype 1) coding for Cre-dependent eGFP (AAV.flex.eGFP; Penn Vector Core, Philadelphia, PA) into the lumbar enlargement, and had been used in a previous study . These mice had survived 8 days after the spinal injection before being perfused with the same fixative. Transverse sections from segments that included the injection sites were processed for immunocytochemistry.
Three wild-type C57BL/6 mice (either sex, 18-20 g) were used for in situ hybridization experiments. The mice were decapitated under deep isoflurane anesthesia and spinal cords were removed and rapidly frozen on dry ice.

| Immunocytochemistry, confocal microscopy, and analysis
Transverse spinal cord sections 60 μm thick were cut from the perfusion-fixed tissue with a vibrating blade microtome and immersed in 50% ethanol for 30 min to enhance antibody penetration. These were then reacted for multiple-labeling immunofluorescence staining as described previously (Gutierrez-Mecinas et al., 2016). The primary antibodies used in this part of the study are listed in Table 1. Unless stated otherwise, sections were incubated for 3 days at 4 C in primary antibodies diluted in PBS that contained 0.3 M NaCl, 0.3% Triton X-100 and 5% normal donkey serum, and then overnight in appropriate species-specific secondary antibodies (Jackson Immunoresearch, West Grove, PA) that were raised in donkey and conjugated to Alexa 488, Alexa 647, Rhodamine Red, Pacific Blue or biotin. All secondary antibodies were used at 1:500 (in the same diluent), apart from those conjugated to Rhodamine Red or Pacific Blue, which were diluted to 1:100 and 1:200, respectively. Biotinylated secondary antibodies were detected with either Pacific Blue conjugated to avidin (1:1,000; Life Technologies, Paisley, UK) or with a tyramide signal amplification (TSA) method (TSA kit tetramethylrhodamine NEL702001, PerkinElmer Life Sciences, Boston, MA). The TSA reaction was used to detect the NKB precursor preprotachykinin B (PPTB), as this method can be used to reveal cell bodies of NKBexpressing dorsal horn neurons (Polgár, Furuta, Kaneko, & Todd, 2006). Following the immunocytochemical reaction, sections were mounted in antifade medium and stored at −20 C.
Sections from the GRP::eGFP mice were reacted with primary antibodies against eGFP, pro-CCK, and NeuN (mouse antibody), and following incubation in secondary antibodies, they were counterstained with 4 0 6-diamidino-2-phenylindole (DAPI) to label cell nuclei.
Sections from the wild-type mice were reacted with three different combinations of primary antibodies: (a) pro-CCK, neurotensin, PKCγ, and NeuN (mouse antibody); (b) pro-CCK, PPTB, NeuN (mouse antibody); (c) pro-CCK, Pax2, NeuN (mouse antibody). For the third reaction, a sequential staining protocol was used, because both the pro-CCK and Pax2 antibodies are raised in rabbit. In this case, sections were initially incubated in pro-CCK and NeuN antibodies, which were revealed with fluorescent secondary antibodies conjugated to Alexa 488 and Alexa 647, respectively. They were then treated with unconjugated Fab' donkey anti-rabbit IgG fragment for 6 hr to saturate binding sites for rabbit IgG, before being incubated for 2 days in the Pax2 antibody, which was subsequently revealed with secondary antibody conjugated to Rhodamine Red. Tissue from the Tac1 Cre mice that had received intraspinal injections of AAV.flex.eGFP was incubated in antibodies against pro-CCK and NeuN (guinea pig antibody).
Sections were scanned with a Zeiss LSM 710 confocal microscope with Argon multiline, 405 nm diode, 561 nm solid state, and 633 nm HeNe lasers. Confocal image stacks were obtained through a 40× oil immersion lens (numerical aperture 1.3) with the confocal aperture set to 1 Airy unit. In each case, the entire mediolateral width of laminae I-III was scanned to generate z-series of at least 20 μm (and in many cases the full thickness of the section), with a zseparation of 1 μm. Unless otherwise stated, for each antibody combination, two sections from each of three mice were selected for scanning. To avoid bias, this selection was made before the pro-CCKimmunoreactivity was viewed. Confocal scans were analyzed with Neurolucida for Confocal software (MBF Bioscience, Williston, VT).
The lamina II-III border was identified from the distribution of NeuN immunoreactivity, based on the relatively low neuronal packing density in lamina IIi. The lamina I-II border was assumed to be 20 μm from the dorsal edge of the dorsal horn (Dickie et al., 2019). The lamina III/IV border was identified by the somewhat lower packing density of neurons in lamina IV (Molander, Xu, & Grant, 1984).
To determine the proportion of neurons in laminae I-III that are pro-CCK-immunoreactive, we used a modification of the optical dissector method (Polgár, Gray, Riddell, & Todd, 2004)  To test whether any of the CCK cells were inhibitory interneurons, we identified pro-CCK-positive neurons in laminae I-III through a depth of at least 24 μm in sections reacted with the Pax2 antibody from each of three mice. We then viewed the Pax2 channel and determined whether any of the selected cells had Pax2-immunoreactive nuclei.
To look for evidence of coexpression of CCK and NKB, we analyzed sections reacted for pro-CCK and the NKB precursor PPTB.
Pro-CCK-immunoreactive neurons were identified throughout the full thickness of sections from three mice, and then the presence or absence of PPTB was noted.
To determine whether there was overlap between CCK and substance P populations, we analyzed sections from the three Tac1 Cre mice that had been injected with AAV.flex.eGFP. All pro-CCK cells were identified throughout the full thickness of sections from three mice, and we then looked for the presence of eGFP (representing Tac1 expression) in these cells. Because we found some overlap between these populations, we then reversed the procedure by identifying all eGFP-positive cells and determining whether they were pro-CCK-immunoreactive.
Sections reacted with antibodies against pro-CCK, neurotensin, and PKCγ were used to look for co-expression of pro-CCK and neurotensin, and also to determine the relationship of both peptides to PKCγ. In this case, three sections from each of three mice were examined through their full thickness. Initially, all pro-CCK cells were selected and these were then tested for the presence of neurotensin and PKCγ. In a separate analysis, we viewed the PKCγ channel and identified all of the neurons that showed strong or moderate levels of PKCγ immunoreactivity and had cell bodies that were completely contained within the section. We then switched on the pro-CCK and neurotensin channels and recorded the presence or absence of these peptides in each of the selected PKCγ cells. probes were also tested on other sections (Table 2).

| Fluorescent in situ hybridization histochemistry
Sections were scanned with a Zeiss LSM 710 confocal microscope through the 40× oil-immersion lens with the confocal aperture set to 1 Airy unit. In each case, a single optical plane through the middle of the section was scanned (Häring et al., 2018)

| Antibody characterization
Details of the primary antibodies that were used in this study are given in Table 1. Specificity of the pro-CCK antibody has been demonstrated by showing that it stains identical cells to those detected by fluorescent in situ hybridization with a probe against CCK mRNA in mouse cortex and hippocampus (Booker et al., 2017). The mouse monoclonal antibody NeuN, which was generated against cell nuclei extracted from mouse brain, was found to react with a protein specific to neurons (Mullen, Buck, & Smith, 1992), and this has subsequently been identified as the splicing factor Fox-3 (Kim, Adelstein, & Kawamoto, 2009). We have found that the guinea-pig NeuN antibody stains the same cells as the mouse antibody (Dickie et al., 2019).
Staining with the neurotensin antibody is identical to that seen with a well-characterized rabbit neurotensin antibody, and is blocked by preincubation with 10 −4 M neurotensin (Porteous et al., 2011). The PKCγ antibody recognizes a band of the appropriate molecular weight on western blots from wild-type, but not PKCγ −/− mice (Yoshida et al., 2006). The PPTB antibody recognizes PPTB, but not PPTA, NKB, or substance P on dot blots, and staining is blocked by preincubation with the immunizing peptide. The Pax2 antibody recognizes bands of the appropriate molecular weight on western blots of mouse embryonic kidney (Dressler & Douglass, 1992). Staining with the eGFP antibody shows an identical distribution to that of the native fluorescent protein in tissues.

| Specificity of fluorescent in situ hybridization
All neurons were labeled with the positive control probes (which detect mRNAs for ubiquitously expressed proteins), while none were labeled with the negative control probe (which detects a bacterial enzyme). 3 | RESULTS

| Expression of CCK by excitatory neurons in the dorsal horn
In transverse sections immunostained for pro-CCK, labeling was seen in the perikaryal cytoplasm of neurons. These immunoreactive neurons were particularly numerous in the deep dorsal horn (laminae III-V), and were scattered at lower density in laminae I and II (Figure 1).
Quantitative results obtained with the disector method from three mice are shown in Table 3. Pro-CCK-immunoreactive cells accounted for~5% of all neurons in laminae I-II and 15% of those in lamina III. To test whether CCK cells in laminae I-III were excitatory or inhibitory, we compared Pax2 and pro-CCK immunostaining in sections from three wild-type mice. Altogether, 247 pro-CCK cells were identified (76-95 per mouse) and none of these had Pax2 immunoreactivity in the nucleus (Figure 2), indicating that they are all excitatory neurons.
We have previously determined the proportions of neurons in laminae I-III of the mouse dorsal horn that lack immunoreactivity for GABA and glycine, and are therefore presumably excitatory neurons (Polgár et al., 2013). Based on these results, we estimate that pro-CCK can be detected in 6.9, 6.7, and 24.7% of the excitatory neurons in laminae I, II, and III, respectively (Table 3). Laminae I-II are often considered together as the superficial dorsal horn, and pro-CCK-immunoreactive cells would account for 6.8% of the excitatory neurons in this region.
Quantitative results for the in situ hybridization analysis of VGLUT2 and CCK in sections processed with probes directed against Slc17a6 (VGLUT2) and Cck mRNAs are shown in Table 4, and an example of the labeling is shown in Figure 3. As expected, the vast majority of cells that were positive for Cck mRNA also contained VGLUT2 mRNA (93 and 98% for laminae I-II and lamina III, respectively). These accounted for 7% of the VGLUT2+ cells in laminae I-II and for 36% of those in lamina III. Although we did not formally analyze the dorsal horn below lamina III, we also noted that~30% of the VGLUT2+ cells in the region corresponding to lamina IV were also CCK+.
The in situ hybridization analysis was based on profile counts (rather than stereology) and there may therefore be a bias toward larger neurons, which are more likely to have been included in the optical section analyzed (Guillery, 2002). However, the close agreement between the estimates of excitatory neurons that are CCK-positive in the superficial dorsal horn (6.8% with immunocytochemistry, 7.3% with in situ hybridization) suggests that this did not have an impact on our estimates in this region. The proportion of lamina III excitatory neurons in which Cck mRNA was detected was somewhat higher than our estimate of the proportion that were pro-CCK-immunoreactive (36% compared to 25%). This discrepancy could have resulted from a size bias, but may have been due to our failure to detect pro-CCK in some of the CCK-expressing neurons in this lamina.

| CCK is not expressed by PPTB-positive or GRP-eGFP neurons
In the sections from the three GRP::eGFP mice, we identified an average of 129 (116-152) pro-CCK-immunoreactive cells in laminae I-III and all except two of these cells (both located in lamina III) were eGFP-negative (Figure 4), indicating that CCK is very seldom expressed in GRP-eGFP neurons.
In sections from three wild-type mice that had been immunoreacted for PPTB and pro-CCK, we identified a mean of 95 pro-CCKimmunoreactive cells (range 80-115) in laminae I-III in sections from In each case, the mean values for the three mice are shown, with the range in brackets. a The percentages of neurons in each lamina that lack immunoreactivity for GABA and glycine are taken from four wild-type C57BL/6 mice examined by Polgár et al. (2013).

| Partial overlap between CCK and substance P populations
In sections from Tac1 Cre mice that had been injected with AAV.flex. eGFP, we found a similar distribution of eGFP-labeled (Tac1-expressing) cells to that described previously (Dickie et al., 2019;Gutierrez-Mecinas et al., 2017). These were concentrated in laminae I and II, with labeled cells scattered through the deeper dorsal horn. Although most pro-CCK-immunoreactive cells were different from those that were eGFPlabeled, we found some overlap. In laminae I-II, double-labeled neurons corresponded to 29% of the pro-CCK-immunoreactive cells, and 3% of those that were eGFP-labeled. In lamina III, the corresponding proportions were 13 and 10%, respectively (  Figures 6e and 7). The much higher proportion of Tac1+ cells with Cck mRNA in lamina III may have resulted from our failure to detect pro-CCK-immunoreactivity in some CCK-expressing cells in this region (see above).
Because we had previously found that many neurotensin-expressing cells also showed moderate-strong labeling for PKCγ (  (21-28%) of these were TRH-positive. In these sections, Trh mRNA was present in 46% (42-53%) of those CCK+ cells that had 10 or more Prkcg mRNA transcripts. To further assess the relationship between TRH and PKCγ, we analyzed sections reacted with this probe combination, and also sections reacted with probes against Prkcg, Trh, and Nts mRNAs. We identified 60.7 (58-64) Trh mRNA-positive cells,

| Technical considerations
We used two different approaches as surrogate measures for gene expression in dorsal horn neurons: in situ hybridization with RNAscope that detects mRNA transcript numbers, and immunocytochemistry, which reveals the proteins or peptides that are generated from these mRNAs. Although the levels of mRNAs and their corresponding proteins are correlated within individual cells, there is not a direct relationship between them, for example due to variations in post-transcriptional regulation and protein stability (Vogel & Marcotte, 2012). It is, therefore, not possible to predict how much mRNA in a particular neuron will lead to a specific level of protein expression in that cell. Comparing these two techniques, the levels of protein or peptide, as detected by immunocytochemistry, are presumably more important in terms of function. However, knowledge of mRNA production forms the basis for transcriptomic methods to define neuronal populations (Häring et al., 2018), and also provides valuable information that is needed for molecular/genetic approaches in which specific neurochemical populations are to be targeted.
Although it is possible to combine immunocytochemistry with in situ hybridization, and therefore confirm expression in the same cell (Booker et al., 2017), we did not attempt to do this in the present study, as we wished to retain optimal sensitivity offered by the RNAscope technique. Nonetheless, the consistency between patterns of colocalization that we observed with the two techniques strongly suggests that in general, they were detecting equivalent populations.
However, as the RNAscope method allows identification of individual mRNA molecules, it is likely to reveal cells with very low levels of gene expression, which may be missed with immunocytochemistry.

| Neurochemical populations among dorsal horn excitatory interneurons
Classification schemes for dorsal horn neurons have been based mainly on the laminar location, morphology, electrophysiological properties, and neurochemistry of the cells (Abraira et al., 2017;Graham, Brichta, & Callister, 2007;Grudt & Perl, 2002;Yasaka, Tiong, Hughes, Riddell, & Todd, 2010). Among the four neurochemical populations that we had previously identified, the neurotensin and NKB cells had certain features in common, as they were both concentrated on either side of the lamina II-III border, and both included cells that expressed PKCγ. We have recently provided evidence that the other two populations differed from each other in both morphological and electrophysiological properties with those that express substance P corresponding to radial cells (which generally show delayed firing) and the GRP-GFP cells resembling a population that had been defined by Grudt and Perl (2002) (Dickie et al., 2019;Solorzano et al., 2015), and this suggests that eGFP is restricted to a subset of the GRP-expressing cells. We have found that at least some of these GRP-eGFP cells contain mRNA for the neuromedin U receptor 2 (Nmur2; AMB, FIGURE 7 Quantitative data for coexpression of cholecystokinin (CCK) and substance P in the dorsal horn. The bar chart shows the mean numbers of cells in laminae I-II and lamina III that were defined as expressing CCK, Tac1 (the gene coding for substance P), or both peptides, based on immunocytochemistry (ICC) or in situ hybridization (ISH). The immunocytochemical detection of pro-CCK was performed on 3 Tac1 Cre mice that had received intraspinal injections of AAV.flex. eGFP, and eGFP-containing cells were defined as Tac1-positive. In situ hybridization was performed with RNAscope and single optical sections from three wild-type mice were analyzed. Cells containing four or more transcripts for CCK or Tac1 mRNAs were defined as positive for the respective peptide. For further details, see Table 5 [Color figure can be viewed at wileyonlinelibrary.com] unpublished data), and these may therefore correspond to the Glut8 population, in which Nmur2 mRNA is enriched. indicating that CCK is seldom, if ever, expressed by substance P radial The merged image shows that two of these are neurotensin-immunoreactive (arrowheads), while the other three are pro-CCK-immunoreactive (arrows). In each case, immunostaining for the neuropeptide/precursor can be seen in the perikaryal cytoplasm just inside the ring of PKCγ-immunostaining, which represents the plasma membrane. (e-g) In situ hybridization in a section that had been reacted with probes for CCK, neurotensin, and PKCγ mRNAs and counterstained with NucBlue. Several PKCγ+ cells are present in this field, and six of these are marked. Three (arrows) are positive for CCK mRNA and two (arrowheads) for neurotensin mRNA, while one (asterisk) contains mRNAs for both CCK and neurotensin. The in situ hybridization images were obtained from a single optical section. (h) Quantitative analysis of the proportions of PKCγ cells that were positive for CCK and/or NTS detected with either immunocytochemistry (ICC) or in situ hybridization (ISH). The bar chart shows the results for cells with strong to moderate PKCγimmunoreactivity (ICC) or that had 10 or more Prkcg mRNA transcripts (ISH). In each case, these have been divided into those that were positive for CCK only (green), neurotensin only (red), or both CCK and neurotensin (orange). Cells that were negative for both peptides are represented as gray bars. In each case, data from three animals were pooled. Scale bars for (a-d) and for (e-g) = 20 μm [Color figure can be viewed at wileyonlinelibrary.com] cells in lamina II (Dickie et al., 2019). The cells that express both CCK and Tac1 in this region could belong to either Glut9 or Glut10, because both peptides are significantly expressed in these clusters, and they include cells in the superficial dorsal horn. CCK and Tac1 showed a higher frequency of coexpression in lamina III and based on similar criteria, these could belong to either Glut10 or Glut14. Peirs et al. (2015) have described a population of deep dorsal horn cells that transiently express VGLUT3, and have shown that these are required for mechanical hypersensitivity in both neuropathic and inflammatory pain states. It will, therefore, be of interest to know whether these "transient VGLUT3" neurons overlap with the CCK cells, and also how they fit into the classification scheme of Häring et al.
Anatomical studies have shown that PKCγ cells are innervated by myelinated low-threshold mechanoreceptive afferents, and have provided evidence that they respond to innocuous mechanical stimuli (Neumann et al., 2008). Taken together with electrophysiological and pharmacological data (Lu et al., 2013;Miraucourt et al., 2007;Torsney & MacDermott, 2006), this has led to the suggestion that PKCγ cells form part of a polysynaptic pathway that links low-threshold mechanoreceptive input to lamina I projection neurons. This pathway is normally closed due to feedforward inhibitory mechanisms, but these may become ineffective following nerve injury, resulting in mechanical allodynia (Lu et al., 2013). Interestingly, there is evidence that the PKCγ cells do not contribute to mechanical hyperalgesia in inflammatory pain states Peirs & Seal, 2016).
We had previously shown that many of the cells with moderate-  The relationship of thyrotropin-releasing hormone (TRH)-expressing dorsal horn neurons to cholecystokinin (CCK), protein kinase Cγ (PKCγ), and neurotensin. (a-c) A scan through laminae II and III from a section reacted with probes against TRH, CCK, and PKCγ mRNAs, and counterstained with NucBlue. Three TRH+ cells are marked with arrows, and all of these are positive for both CCK and PKCγ mRNAs. Asterisks indicate two PKCγ cells that lack message for TRH and CCK, while the arrowheads show two CCK cells that are negative for TRH and PKCγ. (d-f) A scan through the equivalent region from a section reacted with probes against TRH, neurotensin, and PKCγ mRNAs, and counterstained with NucBlue. Two TRH+ cells that contain mRNA for PKCγ, but not for neurotensin, are indicated with arrows. The arrowheads point to two cells that are positive for neurotensin and PKCγ mRNA, but negative for TRH mRNA. All images were obtained from single optical sections. Scale bar = 20 μm [Color figure can be viewed at wileyonlinelibrary.com] Around half of the CCK+/PKCγ+ cells contained mRNA for TRH, and as this is virtually restricted to the Glut3 cluster of Häring et al. (2018), these cells presumably belong to this population. As PKCγ was

| Role of CCK-expressing cells in the dorsal horn
The principal function that has been attributed to CCK at the spinal level is a tonic reduction of the antinociceptive action of opioids, and this is thought to be mediated mainly by CCK B receptors (Wiesenfeld-Hallin, Xu, & Hokfelt, 2002). It has also been reported that CCK contributes to the development of tolerance to the analgesic effects of transcutaneous electrical nerve stimulation in a rat model of inflammatory arthritis, again involving an anti-opioid mechanism (DeSantana, da Silva, & Sluka, 2010). A recent report suggests that the anti-opioid action of CCK may involve hetero-dimerization of μ-opioid and CCK B receptors (Yang et al., 2018). As the highest density of μ-opioid receptors is in the superficial dorsal horn (Arvidsson et al., 1995), it is surprising that CCK-expressing cells are far more numerous in laminae III-IV. However, CCK is presumably released via a nonsynaptic mechanism ("volume transmission") and may diffuse from release sites in the deeper laminae to the superficial dorsal horn.
Because the CCK cells are glutamatergic, it is likely that their main action will be mediated by synaptically released glutamate. Two recent studies have tested the effects of ablating CCK-expressing dorsal horn neurons, in both cases using a mouse line in which Cre recombinase was knocked into the CCK locus (Abraira et al., 2017;Liu et al., 2018).
Abraira et al. reported that mice in which CCK cells had been ablated could not distinguish between objects with different textures, and failed to show prepulse inhibition in response to a light air puff applied before an acoustic stimulus. Liu et al. showed that ablation of these cells reduced mechanical allodynia in the spared nerve injury model of neuropathic pain, and significantly decreased the number of dorsal horn neurons that showed Fos in response to brush stimulation in these mice.
Together, these findings indicate that dorsal horn CCK cells have important roles in the perception of tactile stimuli in naïve animals, and that they are also involved in mechanical allodynia in neuropathic pain states.
The CCK cells represent a large subset of excitatory interneurons, accounting for around a third of those in deeper laminae, and it is therefore not surprising that ablation results in a dramatic behavioral phenotype. Our findings, together with those of Häring et al. (2018), suggest that CCK cells include functionally distinct subsets, such as those that co-express PKCγ. The extensive overlap between CCK and PKCγ that we observed, taken together with earlier studies that have implicated PKCγ cells in neuropathic pain, raise the possibility that the CCK+/PKCγ+ cells may be particularly important for the development of mechanical allodynia. In future studies, it will be important to determine the contribution of the different neuropeptide-expressing subsets of PKCγ neurons to the development of allodynia in neuropathic pain states.