Characterisation of neurotensin-expressing interneurons in the mouse spinal dorsal horn

Abstract

The spinal dorsal horn contains numerous excitatory interneurons, which can be assigned to functional classes based on morphological, electrophysiological and neurochemical criteria. One population consists of neurons that express neurotensin, and these belong to a larger group defined by the presence of protein kinase Cγ (PKCγ). It has been proposed that PKCγ neurons form part of a circuit that can convey low-threshold mechanoreceptive information to nociceptive projection neurons in lamina I, forming a pathway that could underlie mechanical allodynia in pathological pain states. However, despite their potential importance, relatively little is known about the properties of the neurotensin-expressing cells. Here we have used a neurotensin-Cre line, together with intraspinal injection of AAVs coding for Cre-dependent constructs, to characterise the morphological and electrophysiological properties of these cells. Reconstruction of their dendritic trees revealed that they were morphologically diverse, although many could be assigned to a class known as central cells. All cells examined received synaptic contacts from putative A- and C-low-threshold mechanoreceptors (identified by expression of VGLUT1 and VGLUT3, respectively). However, these only accounted for a minority of their excitatory synapses. Around 40% of their synapses were from VGLUT2-immunoreactive boutons, which are likely to have originated mainly from local excitatory interneurons. Electrophysiological analysis revealed similarities to, and differences, from other neurochemically-defined excitatory interneuron populations. Our findings are compatible with the proposed role of neurotensin cells in mechanical allodynia, but suggest additional functions for these cells.

Introduction

Somatosensory information is conveyed to the spinal dorsal horn by primary afferents, processed through complex circuits involving local interneurons, and transmitted to the brain by spinal projection neurons.^1–3 Pathological changes within the dorsal horn are thought to contribute to persistent pain states, and it is therefore important to understand the neural circuitry of this region.

Primary afferent input to the spinal cord follows a laminar pattern. Fine diameter afferents, most of which are nociceptors, project mainly to the superficial dorsal horn (SDH; laminae I-II), while cutaneous low-threshold mechanoreceptors arborise in a region extending ventrally from the inner part of lamina II (IIi). ⁴ The dorsal horn contains a diverse set of neurons. Projection neurons, defined as those with axons that reach the brain or the lateral cervical nucleus, ⁵ account for only a very small proportion (<1%) of dorsal horn neurons in lumbar spinal cord, ¹ with the remaining ∼99% being interneurons. Laminae I-III contain a high density of neurons, most of which are small compared to projection neurons,^6,7 and 25-40% of these (depending on lamina) are GABAergic. ⁸ The remaining 60-75% of neurons in these laminae are glutamatergic excitatory interneurons.

Transcriptomic studies have revealed several populations among both excitatory and inhibitory interneurons^9–11 and have shown that the chronotopic arrangement of developing excitatory interneurons underlies the laminar organisation of the dorsal horn. ¹² Based partly on these studies, we have identified seven largely non-overlapping populations among SDH excitatory interneurons.^13–19 Five of these are defined by expression of neuropeptides: neurotensin, neurokinin B (NKB), cholecystokinin (CCK), substance P and neuropeptide FF (NPFF), and one by the expression of the gastrin-releasing peptide receptor (GRPR). The final population (GRP-GFP cells) was identified based on expression of green fluorescent protein (GFP) in a BAC transgenic line (GRP::GFP) in which GFP is expressed under the control of the gastrin-releasing peptide (GRP) promoter. We have also shown that some of these classes correspond to populations that had been recognised based on anatomical and/or electrophysiological criteria. ²⁰

Excitatory interneurons expressing neurotensin or CCK are present in lamina IIi-III, and show considerable overlap with a class of neurons previously identified by their expression of protein kinase Cγ (PKCγ).^16,21–23 Neurons expressing PKCγ have been implicated in development of neuropathic pain. ²¹ It has been shown that they are innervated by myelinated low-threshold mechanoreceptive primary afferents (A-LTMRs),^24,25 and suggested that they form part of a synaptic circuit that allows this input to access nociceptive projection neurons in lamina I following nerve injury, thus contributing to mechanical allodynia. ²⁶ Relatively little is known about the neurotensin-expressing neurons, although we reported that they account for ∼7% of neurons in lamina II and ∼8% of those in lamina III. ¹⁶ The aim of this study was to characterise the anatomical and electrophysiological features of these cells, to allow comparison with the properties of other populations of dorsal horn excitatory interneurons, and to investigate whether their location in synaptic circuits is consistent with their proposed role in mechanical allodynia. ²⁶

Materials and methods

All animal 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. The study was carried out in compliance with the ARRIVE guidelines.

Animals and surgical procedures

For all of the experiments we used a neurotensin-Cre (Nts^Cre) knock-in line (Jackson Laboratories strain #017525), which had been crossed with a Flp-deleter mouse to remove a FRT-flanked Neo casette. These mice were maintained on a C57BL/6J background, were heterozygous for the Cre allele, and were aged between 5 and 11 weeks at the time of surgery.

In all cases, mice received intraspinal injections of AAVs that encoded Cre-dependent fluorescent proteins into the dorsal horn of the L3 and/or L5 segments. The experimental procedure was the same as that used in previous studies.^17,27 Briefly, mice were given perioperative analgesia (buprenorphine 0.1 mg/kg and carprofen 10 mg/kg, s.c.), were then anaesthetised with isoflurane (∼1.5%) and placed in a stereotaxic frame. After surgical exposure, the T12 and L1 vertebrae were clamped and injections were performed through glass micropipettes (outer tip diameter ∼60 µm) attached to a 10 µL Hamilton syringe. Injections into the L3 and L5 segments were made through intervertebral spaces on either side of the T13 vertebra. Each injection consisted of between 300-500 nL and was administered at a rate of 30-40 nL/minute. The pipette was left in place for 5 minutes to minimise leakage of the injectate. Mice survived for between 1-5 weeks and were then either fixed by intracardiac perfusion with 4% freshly depolymerised formaldehyde (for anatomical studies) or were perfused with ice-cold dissection solution (for electrophysiology). Mice used for anatomical studies were aged between 7 and 13 weeks at the time of perfusion fixation. Details of the viruses used in each type of experiment are described below.

Tissue processing and immunohistochemistry

Following perfusion fixation, spinal cord segments were post-fixed for 2 hours in the same fixative, and transferred to 30% sucrose in phosphate buffer. The tissue was then rinsed in phosphate bufer, and transverse or parasagittal 60 µm thick sections were cut from the L3 and L5 segments with a vibrating blade microtome (Leica VT1200 or VT1000). The sources and concentrations of antibodies are listed in Table 1. Sections were incubated for 3 days at 4°C in primary antibodies diluted in phosphate buffered saline that contained 0.3M NaCl, 0.3% Triton X-100 and 5% normal donkey serum, and then overnight in species-specific secondary antibodies that were raised in donkey and conjugated to Alexa 488, Alexa 647, Rhodamine Red or biotin (Jackson Immunoresearch). All secondary antibodies were diluted 1:500 in the same diluent, apart from those conjugated to Rhodamine Red, which were diluted 1:100. Biotin was revealed by subsequent incubation in avidin conjugated to Pacific Blue (1:1000; Life Technologies). Following immunoreaction, sections were mounted in anti-fade medium and stored at -20°C. Sections were scanned with either a Zeiss LSM710 confocal microscope with Argon multi-line, 405 nm diode, 561 nm solid state and 633 HeNe lasers, or with a Zeiss LSM900 Airyscan confocal microscope with 405, 488, 561 and 640 nm diode lasers. Scans were obtained through a 40× or 63× oil-immersion lens (numerical aperture 1.3 and 1.4, respectively) with the aperture set to 1 Airy unit or less.

Table 1.

Antibodies used in this study.

Antigen	Species	Dilution	Source	Catalogue #	Rrid
Neurotensin	Rat	1:1000	P Ciofi	​	RRID:AB_2314928
Neurotensin	Rabbit	1:5000	Immunostar	20072	RRID:AB_572254
PKCγ	Rabbit	1:1000	M Watanabe	​	RRID:AB_2571824
NeuN	Guinea pig	1:1000	Synaptic Systems	266 004	RRID:AB_2619988
Homer1	Goat	1:1000	M Watanabe	​	RRID:AB_2631104
PSD-95-Alexa 647*	Alpaca	1:100	Synaptic Systems	N3702-AF647-L	RRID:AB_2936216
mCherry	Chicken	1:5000	Abcam	Ab205402	RRID:AB_2722769
mTFP	Rat	1:500	Kerafast	EMU103	​
tagRFP	Guinea pig	1:500	Kerafast	EMU107	​
VGLUT1	Guinea pig	1:5000	Millipore	ab5905	RRID:AB_2301751
VGLUT2	Guinea pig	1:5000	Millipore	ab2251	RRID:AB_1587626
VGLUT3	Guinea pig	1:100	M Watanabe	​	RRID:AB_2571856

^*The PSD-95 antibody is a camelid single-domain antibody (nanobody) conjugated to Alexa 647.

Characterisation of antibodies

The rabbit antibody against neurotensin was raised against synthetic human neurotensin conjugated to bovine thyroglobulin, and immunostaining is abolished by pre-incubation with 10 µg neurotensin per ml of diluted antibody (manufacturer’s specification). Staining with the rat polyclonal neurotensin antibody is blocked by pre-incubation with the peptide. ²⁸ The PKCγ antibody was raised against amino acids 684-697 of the mouse protein, and gives an identical staining pattern to other well characterised PKCγ antibodies. We have found that the guinea-pig NeuN antibody labels the same cells as the mouse antibody, ¹³ which has been shown to detect only neurons in the spinal dorsal horn. ²⁹ The affinity-purified Homer1 antibody was raised against amino acids 1–175 of mouse Homer1 and detects a band at 43–45 kDa in immunoblots of mouse brain extracts. Since the first 120 amino acids are highly conserved between Homer1, 2 and 3 the antibody is likely to detect all forms of Homer. We have shown that punctate staining with this antibody is associated with glutamatergic boutons in the spinal dorsal horn.^30,31 The PSD95 nanobody was raised against a recombinant protein corresponding to amino acids 68-251 from mouse PSD95 and gives a very similar labelling pattern to that seen with the Homer1 antibody. Labelling with the antibodies against fluorescent proteins is confirmed by the lack of staining in regions that do not contain these proteins. The VGLUT1 and VGLUT2 antibodies were raised against sequences of 19 or 18 amino acids (respectively) from the corresponding rat proteins, and both have been shown to label identical structures to well characterised rabbit antibodies. ³² The VGLUT3 antibody is directed against amino acids 522-588 of the mouse protein, and detects a single band at 60-62 kDa.

Validation of the Nts^Cre line

Three male Nts^Cre mice received injections of 300 nl of AAV1.CAG.Cre^ON.GFP containing 7.5 × 10^-8 gene copies (Addgene 51502) into the L3 and L5 segments on the right side. Transverse sections from the L3 segment from each mouse were immunostained to reveal neurotensin, NeuN and PKCγ. Three sections from each mouse were scanned through their full thickness to include the right dorsal horn and the resulting scans were analysed with Neurolucida. Initially, the lamina II-III border was located based on the appearance of PKCγ-immunoreactivity ³³ and the approximate ventral border of lamina III was outlined based on the pattern of NeuN staining. All cells within laminae I-III that were immunoreactive for either GFP or neurotensin, and for which the entire cell body was contained within the section, were identified and plotted onto an outline of the dorsal horn. The presence or absence of PKCγ was noted for each cell. In this way, we determined the proportion of GFP-positive cells that were neurotensin-immunoreactive, and the percentage of neurotensin containing cells that expressed GFP, together with the relation of these cells to PKCγ.

Morphology of neurotensin cells

Brainbow labelling ³⁴ was achieved by injecting 500 nL of a mixture of AAV9.EF1a.BbTagBY (AAV.BB1, Addgene #45185, 7.55 × 10⁷ gene copies) and AAV9.EF1a.BbChT (AAV.BB2, Addgene #45186, 7.44 × 10⁷ gene copies) into the L3 or L5 dorsal horn of 3 female Nts^Cre mice, as described previously.^13,17 Sagittal sections were immunostained to reveal neurotensin, together with 3 of the resulting fluorescent proteins: TagBFP, teal fluorescent protein (TFP) and mCherry.

Sections were initially viewed to allow the selection of labelled neurons. This was based on the following criteria: (1) neurotensin-immunoreactivity in the cell body, (2) relatively strong membrane labelling with at least one of the fluorescent proteins, (3) location of the cell such that most or all of the soma and dendritic tree were contained within the section, (4) the presence of a colour hue that allowed separation of processes from those of nearby cells. Ten neurons were selected from each mouse, and these were reconstructed with Neurolucida software. Morphometric data were obtained from Neurolucida Explorer.

Synaptic inputs to neurotensin cells

Glutamatergic boutons in the dorsal horn can contain one or more of the three vesicular transporters, VGLUT1, VGLUT2 and VGLUT3, and these are differentially distributed among different types of axon.^{24,32,35–37} VGLUT1 is largely restricted to A-low threshold mechanoreceptors, but is also expressed in corticospinal boutons.^24,31 Although VGLUT2 is expressed by spinal excitatory neurons and many fine diameter primary afferents, the levels in primary afferent central terminals are generally very low. ³² VGLUT3 in lamina II-III is largely restricted to C-low threshold mechanoreceptors. ³⁵

We used this pattern of labelling to investigate excitatory synaptic inputs to neurotensin expressing neurons. To do this, we combined a “sparse Brainbow” method, as described previously ³⁰ with immunohistochemical detection of the different VGLUTs. Five Nts^Cre mice (four female, one male) received intraspinal injections of AAV.BB2 (3.76 × 10⁶ – 2.15 × 10⁷ gene copies in 500 nL) into the L3 and L5 segments. Sagittal sections through these segments were initially immunostained to reveal TFP, and viewed to allow identification of sections that contained TFP-labelled neurons that had extensive dendritic labelling. The sections were then reacted with antibodies against neurotensin, one of the VGLUT antibodies (VGLUT1, VGLUT2 or VGLUT3), together with antibodies against either Homer1 or PSD95.

For each of the VGLUT antibodies, we selected a total of 7 cells (obtained from at least 3 different mice) that were TFP-labelled and contained neurotensin-immunoreactivity within the cell body. Sections containing these cells were then scanned to generate z-stacks through as much of the dendritic tree as was present in the section. Scans were analysed with Neurolucida. Initially, the soma and dendritic tree was reconstructed, and then all excitatory synapses (revealed with either Homer1 or PSD95 antibodies) were plotted onto the reconstruction. The channel corresponding to the VGLUT antibody was then revealed, and the presence or absence of VGLUT immunoreactivity was recorded for each synapse. Note that the distribution of VGLUT3 (corresponding to C-LTMRs) varies in different parts of the lumbar dorsal horn, as these afferents are largely restricted to hairy skin. VGLUT3 labelling is therefore very sparse in the medial parts of L3 and L5 (which are innervated from glabrous skin). ¹³ For this reason, TFP-labelled cells in sections that had been reacted to reveal VGLUT3 were only analysed if a substantial part of the dendritic tree lay within the plexus of VGLUT3-immunoreactive boutons. Similarly, VGLUT1-immunoreactive boutons are present at high density in the ventral part of lamina II and in deeper laminae. ³² For the VGLUT1 analysis, we therefore selected cells that had the great majority of their dendritic tree within this densely innervated region.

Electrophysiological properties

Electrophysiological recordings were performed in spinal cord slices from 11 Nts^Cre mice (5 female, 6 male; mean age 60 days, range 43-78 days) that had received bilateral injections of 300 nl of AAV1.CAG.Cre^ON.tdTomato containing 1.19 × 10^-8 or 2.37 × 10^-8 gene copies (Zürich VVF v167-1) into the L3 segment. Slices were prepared as described previously.^17,18 Briefly, at least 1 week following AAV injections, mice were deeply anaesthetised with pentobarbitone (20 mg i.p.), perfused with ice-cold dissection solution and the lumbar region of the spinal cord was isolated, embedded in low melting point agar (∼3%; ThermoFisher Scientific, Paisley, UK). Parasagittal (300 μm) slices were cut with a vibrating blade microtome (7000smz-2; Campden Instruments, Loughborough, UK). Immediately after cutting, the slices were incubated in a N-methyl-D-glucamine (NMDG) based recovery solution at 32°C for ∼15 minutes and were then placed in a modified recording solution at room temperature for an additional 1 hour before being transferred to the recording chamber of a fixed stage upright microscope (BX51; Olympus, Southend-on-Sea, UK) equipped with a 40× water-immersion objective, infrared differential interference contrast (IR-DiC) illumination, a 550 nm LED (pE-100; CoolLED, Andover, UK) and a CMOS camera (Prime BSI; Teledyne Photometrics, Birmingham, UK), and were continually perfused with recording solution at a rate of ∼2 ml/min. The solutions contained the following (in mM): Dissection, 3.0 KCl, 1.2 NaH₂PO₄, 0.5 CaCl₂, 7.0 MgCl₂, 26.0 NaHCO₃, 15.0 glucose, 251.6 sucrose; NMDG recovery, 93.0 NMDG, 2.5 KCl, 1.2 NaH₂PO₄, 0.5 CaCl₂, 10.0 MgSO₄, 30.0 NaHCO₃, 25.0 glucose, 5.0 Na-ascorbate, 2.0 thiourea, 3.0 Na-pyruvate, 20.0 HEPES; Modified recording, 92.0 NaCl, 2.5 KCl, 1.2 NaH₂PO₄, 2.0 CaCl₂, 2.0 MgSO₄, 30.0 NaHCO₃,25 glucose, 5.0 Na-ascorbate, 2.0 thiourea, 3.0 Na-pyruvate, 20.0 HEPES; Recording, 125.8 NaCl, 3.0 KCl, 1.2 NaH₂PO₄, 2.4 CaCl₂, 1.3 MgCl₂, 26.0 NaHCO₃,15 glucose. All solutions were bubbled with 95% O₂/5% CO₂.

Targeted whole-cell patch-clamp recordings were made from tdTomato-positive neurons in the SDH using patch pipettes (typical resistance 4 to 8 MΩ) that were filled with an intracellular solution containing (in mM): 130.0 K-gluconate, 10.0 KCl, 2.0 MgCl₂, 10.0 HEPES, 0.5 EGTA, 2.0 ATP-Na, 0.5 GTP-Na, and 0.2% Neurobiotin, pH adjusted to 7.3 with 1.0M KOH. Data were recorded and acquired with a Multiclamp 700B amplifier and pClamp 10 software (both Molecular Devices, Wokingham, UK), and were filtered at 4 kHz and digitised at 10 kHz.

Once a stable whole-cell configuration was achieved, the cells were voltage clamped at -70 mV and a series of voltage steps from -70 to -50 mV (100 ms duration, 2.5 mV increments) was applied to determine the current-voltage relationship, which was used to calculate the resting membrane potential. Any cells that had a resting membrane potential that was less negative than -35 mV were excluded from all analyses. Input resistance was determined by applying a series of -5 mV steps (1 second duration). Action potential firing patterns were assessed in current clamp, in response to a series of 1 second depolarising current steps (5 pA increments) from a membrane potential of around -60 mV, with firing patterns classified using previously published criteria.^{17,18,20,38–40} The action potential properties were determined from the first action potential that occurred at rheobase.

Subthreshold voltage-activated currents were determined by applying a voltage step protocol, where cells were held at -60 mV before stepping to -90 mV for 1 second and then -40 mV for 200 ms, with automated leak subtraction to remove capacitive and leak currents. The currents revealed were classified as rapid (I_Ar) or slow (I_As) A-type potassium currents, low-threshold ‘T-type’ calcium currents (I_Ca,T) or hyperpolarisation-activated currents (I_h), using previously established criteria.^39,41

Input to recorded cells from TRPV1-expressing primary afferents was investigated by recording miniature excitatory postsynaptic currents (mEPSCs), at a holding potential of -70 mV, in the presence of tetrodotoxin (0.5 µM), bicuculine (10 µM) and strychnine (5 µM), prior to and during the application of capsaicin (2 µM).

All chemicals were obtained from Sigma, except sucrose, glucose, NaH₂PO₄ (Thermo Fisher Scientific, Paisley, UK), NaCl, KCl, HEPES (VWR, Lutterworth, UK), tetrodotoxin (Alomone, Jerusalem, Israel) and bicuculline (Biotechne, Abingdon, UK).

Quantification and statistical analysis

mEPSC frequencies before and after application of capsaicin were compared with a Wilcoxon matched-pairs signed rank test. Comparison of electrophysiological properties of the neurotensin cells with those of other excitatory interneuron populations was carried out with the Kruskall-Wallis test, followed by Dunn’s multiple comparisons test. Numerical data are shown as mean ± standard deviation.

Results

Relationship between cre-positive cells and neurotensin immunoreactivity

Intraspinal injection of AAV1.CAG.Cre^ON.GFP in Nts^Cre mice resulted in GFP labelling of many cells in the dorsal horn (Figure 1). These were most numerous on either side of the border between laminae II and III and scattered in deeper laminae, but were rare in laminae I and the outer part of lamina II (lamina IIo). Neurotensin immunoreactivity showed the same pattern as described previously, ¹⁶ with a dense axonal plexus that was largely restricted to lamina II, with slight extension into the dorsalmost part of lamina III. No axonal labelling was seen in the lateral spinal nucleus (LSN). Cell body labelling consisted of granular deposits in the perikaryal cytoplasm, and this was seen in some neurons in lamina IIi and the dorsal half of lamina III. On average 210 neurotensin-immunoreactive cells (189-226) were identified in laminae I-III across the 3 mice, and 97.7-99.2% of these were GFP-positive. In these sections, 283.3 GFP-positive cells in laminae I-III were identified per mouse and 71.4-75.5% contained neurotensin immunoreactivity (Figure 1). Neurotensin was detected in few of the GFP-positive cells in the ventral part of lamina III, or deep to this. We have previously reported that 90% of neurotensin-immunoreactive cells in laminae I-III are also PKCγ-immunoreactive, ¹⁶ and consistent with this we found that 84.1-91% of neurotensin-immunoreactive cells showed PKCγ-immunoreactivity (Figure 1(b) inset). As shown in Figure 1(c), the axonal plexus overlaps the location of the cell bodies but is displaced slightly dorsally, since it is mainly located in lamina II.

Figure 1.

Labelling following injection of AAV1.Cre^ON.GFP into the dorsal horn of a Nts^Cre mouse. (a), A transverse section showing GFP labelling in the dorsal horn from the L3 segment of a Nts^Cre mouse that had received an injection of AAV1.CAG.Cre^ON.GFP. GFP-positive cells are concentrated in the inner part of lamina II and in lamina III, with scattered cells dorsal and ventral to this region. (b), Immunostaining for neurotensin (NTS) in the same section. The labelling is particularly dense in lamina II, and consists mainly of axonal boutons. (c), A merged image, with the outline of the dorsal horn (solid line), the approximate positions of laminar borders (dashed lines) and the lateral spinal nucleus (asterisk) shown. The box indicates the region shown in the inset in b. The inset shows three cell bodies containing GFP (green), together with immunoreactivity for neurotensin (red) and PKCγ (blue). Two of the cells (arrows) are surrounded by PKCγ-immunoreactivity and contain neurotensin, while the other (arrowhead) lacks PKCγ and neurotensin. Main images are maximum intensity projections through the entire section thickness, while the insets are projections of two optical sections at 1 μm z-spacing. Scale bar = 100 μm.

Morphology of neurotensin-expressing cells in laminae II-III

As reported previously, the AAV Brainbow technique resulted in membrane-limited immunoreactivity for TagBFP, TFP, and mCherry in Cre-positive cells.^13,17 The variation in the relative intensities of labelling for these 3 fluorescent proteins, together with its restriction to the plasma membrane, allowed reconstruction of the dendritic trees of cells, including dendritic spines. Although axons could often be seen emerging from the cells, these were not followed in detail. Thirty cells (10 from each of 3 mice) were selected, and in each case the presence of neurotensin immunoreactivity in the soma was confirmed (Figure 2). The cell bodies of these neurons were located in lamina II (20 cells, Figure 3(a)) or lamina III (10 cells, Figure 3(b)), and gave rise to between 2 and 5 primary dendrites. Dendritic trees were generally elongated along the rostrocaudal axis, but in rare cases extended mainly in dorsal and/or ventral directions. Seven of the 20 lamina II cells had dendrites that extended at least 50 µm dorsal to the cell body. Since the combined dorsoventral extent of laminae I and II at its widest part in the mouse mid-lumbar cord is ∼80-100 µm ⁸ these cells presumably had dendrites that entered lamina IIo. All cells had dendritic spines, and the density of these averaged 26.8 (±4.3, SD) per 100 µm. Comparison with previously published data showed that this was higher than the mean spine densities on GRPR-expressing, GRP-GFP or substance P-expressing cells (15.9, 17.2 and 18.6 per 100 µm, respectively), but lower than that for NPFF-expressing cells (30.7 per 100 µm).^13,17,18

Figure 2.

Brainbow labelling in the Nts^Cre mouse. (a), A maximum intensity projection through the full thickness of a sagittal section from the dorsal horn of a Nts^Cre mouse that had received intraspinal injections of the two Brainbow AAVs. The section has been scanned to reveal tagBFP (BFP, blue), teal fluorescent protein (TFP, green) and mCherry (red). Two labelled cells in lamina III are visible and their cell bodies are marked (1, 2). The insets show single optical sections through the cell bodies scanned to reveal TFP (green) and neurotensin (NTS, magenta). There is granular labelling for neurotensin in both of the cell bodies. (b), Neurolucida reconstructions of the cell bodies and dendritic trees of the two cells. Note that positions of dendritic spines are shown, although the sizes of spine heads and the shapes of spine necks on the drawings do not represent the actual sizes and shapes of these structures. The orientation is shown (D, dorsal; V, ventral; RC, rostrocaudal). Scale bar = 50 μm.

Figure 3.

Dendritic morphology of neurotensin-expressing neurons in laminae II-III revealed with the viral Brainbow technique. (a), Reconstructions of the cell bodies and dendritic trees of the 20 Brainbow-labelled cells that had cell bodies located in lamina II. (b), Reconstructions of 8 of the 10 labelled cells with cell bodies in lamina III (the remaining two are shown in Figure 2). Note that positions of dendritic spines are shown, although the sizes of spine heads and the shapes of spine necks on the drawings do not represent the actual sizes and shapes of these structures. As stated in the text, all of these cells had detectable neurotensin-immunoreactivity in their cell bodies. All cells were drawn from sagittal sections, and the orientation is shown (D, dorsal; V, ventral; RC, rostrocaudal). Scale bar = 50 μm.

Excitatory synaptic input to neurotensin-expressing cells

For each vesicular glutamate transporter, 7 neurons labelled with the “sparse Brainbow” technique were examined, and in each case the presence of neurotensin immunoreactivity in the cell body of the neuron was confirmed. Quantitative data are provided in Table 2, and examples of immunostaining are shown in Figure 4.

Table 2.

Synaptic input from boutons immunoreactive for VGLUT1, VGLUT2 or VGLUT3.

VGLUT	Number of cells analysed	Length of dendrite examined (µm)	Number of PSD95/Homer1 puncta identified	% of PSD95/Homer1 puncta with VGLUT contact
VGLUT1 (PA)	7	1171.2 (671-1660)	402 (206-495)	10.9 (9.06-13.28)
VGLUT1 (CST)	7	1171.2 (671-1660)	402 (206-495)	1.37 (0.87-2.14)
VGLUT2	7	995 (670-1464)	383 (248-525)	42.01 (20.08-63.44)
VGLUT3	7	828.2 (469-1162)	324 (99-556)	16.64 (12.12-22.12)

Ccolumns 3-5 show average values with the range in brackets. VGLUT1-immunoreactive boutons were classified as originating from the corticospinal tract (CST) if they were associated with one PSD95/Homer1 punctum, and as primary afferent (PA) if they were associated with more than one punctum.

Figure 4.

Synaptic inputs to neurotensin neurons from boutons containing VGLUT1, VGLUT2 or VGLUT3. These confocal images show examples of synapses onto Brainbow-labelled neurotensin-immunoreactive cells from boutons containing VGLUT1 (VG1, (a)-(c)), VGLUT2 (VG2, (d)-(f)) and VGLUT3 (VG3, (g)-(i)). In each case, the left image ((a), (d), (g)) shows immunostaining for TFP (green) in the dendritic membrane of the cell, the middle image ((b), (e), (h)) shows labelling for the VGLUT (red) together with that for PSD95 (blue), while the right image ((c), (f), (i)) shows merged immunostaining for the 3 antigens. (a)-(c): a VGLUT1-immunoreactive bouton (asterisk) is adjacent to a PSD95-labelled punctum in the membrane of a TFP-positive dendritic spine belonging to the labelled neuron (marked with arrow). Note that this VGLUT1 bouton is associated with other PSD95 puncta. Another VGLUT1 bouton (arrowhead) is partially surrounded by PSD95 puncta. This appearance arises because these VGLUT1 boutons form the centres of synaptic glomeruli, which are a feature of some A-low threshold mechanoreceptor central boutons. (d)-(f): a TFP-labelled dendritic spine belonging to a Brainbow-labelled neuron is adjacent to a VGLUT2-immunoreactive bouton, with a small PSD95 punctum at the point of contact (arrow). (g)-(i): A TFP-labelled dendritic spine is in contact with a VGLUT3-immunoreactive bouton (asterisk), with a small PSD95 punctum at the point of contact (arrow). The VGLUT3 bouton is associated with several other PSD95 puncta, and again, this reflects a glomerular arrangement. All images are obtained from single confocal optical sections. Scale bar = 5 μm.

VGLUT1-immunoreactive boutons in the dorsal horn can originate from two sources: low-threshold mechanoreceptive afferents (A-LTMRs) and corticospinal axons.^24,31 The great majority of VGLUT1-immunoreactive boutons in this region are associated with several Homer1/PSD95 puncta, and these are likely to be the central terminals of myelinated low-threshold mechanoreceptors, which are often associated with synaptic glomeruli.^42–45 We therefore classified VGLUT1 boutons in this analysis into two types – those associated with a single Homer1/PSD95 punctum were assumed to belong to corticospinal axons, while those associated with 2 or more puncta were classified as originating from A-LTMRs. The great majority of VGLUT1-immunoreactive boutons contacting the cells were associated with more than 1 Homer1/PSD95 punctum (Figure 4(b)), and were therefore classified as originating from A-LTMRs. These accounted for 10.9% of the synaptic puncta on the dendrites of the neurotensin cells (Figure 4(a)–(c); Table 2). VGLUT1 boutons forming only a single synapse were much rarer, and accounted for 1.4% of the synapses identified on these cells.

We previously reported that in the rat, most VGLUT2-immunoreactive boutons formed a single synapse. ⁴² Consistent with this, most VGLUT2-immunoreactive boutons were associated with a single Homer1/PSD95 punctum (Figure 4(e)). The proportion of Homer1/PSD95 puncta on the neurotensin cells that were associated with VGLUT2-containing boutons varied between 20-63%, with a mean of 42% (Figure 4(d)–(f); Table 2).

VGLUT3-immunoreactive boutons were particularly numerous in the inner part of lamina II and the dorsal part of lamina III, and were restricted to the lateral part of the dorsal horn, which receives input from afferents innervating hairy skin 13, 35 and 46. As for VGLUT1, these boutons were commonly associated with several Homer1/PSD95 puncta (Figure 4(h)), consistent with the formation of glomerular central endings by C-LTMRs.^36,47 For the 7 cells analysed, each of which had a substantial part of its dendritic tree within the VGLUT3 plexus, VGLUT3 boutons were apposed to 17% of Homer1/PSD95 puncta (Figure 4(g)–(i); Table 2).

Electrophysiological properties

Whole-cell patch-clamp recordings were made from a total of 36 tdTomato-positive cells (13 from female mice and 23 from male mice). No sex differences were observed, therefore all data described are a combination of recordings from female and male tissue. Since we were not able to verify that the recorded cells contained detectable levels of neurotensin, some of these cells are likely to have belonged to other populations. However, our anatomical findings suggest that this would apply to fewer than 30% of the recorded cells. For convenience, the tdTomato-positive cells that we recorded from are referred to as NTS cells.

Action potential firing patterns were characterised in response to depolarising current steps, and this revealed that transient firing was the predominant firing pattern in NTS cells (19/35; 54.3 %; Figure 5(a)). The remaining cells exhibited single spike (7/35; 20.0 %), tonic (4/35; 11.4 %), delayed (4/35; 11.4 %), or reluctant (1/35; 2.9 %) firing patterns. In a subset of cells, action potential firing patterns were also determined at a more hyperpolarised potential. When depolarising steps were applied at a membrane potential of around -85 mV, this dramatically increased the incidence of delayed firing (13/22; 59.1% vs. 2/22; 9.1 %) and reduced the incidence of both transient (4/22; 18.2% vs. 13/33; 59.1%) and tonic (0/22; 0.0 % vs. 3/22; 13.6%) firing patterns. A small increase in the number of neurons that displayed single spike firing was also seen (5/22; 22.7% vs. 4/22; 18.2%). These findings are consistent with the presence of a slow A-type potassium current in many of these cells (see below).

Figure 5.

Electrophysiological properties of Nts^Cre-positive neurons. (a), The predominant action potential firing pattern seen in Nts^Cre-positive neurons from a membrane potential of ∼ -60 mV was transient firing, with smaller proportions displaying single spike, tonic, delayed or reluctant firing. Example voltage traces show action potential firing at rheobase current and with a suprathreshold current injection, with the corresponding current injection traces shown below. For the reluctant cell only the response to maximal current injection is shown. Note that for each cell the upper voltage trace has been off-set for clarity. In all cases, both traces were recorded with the same initial membrane potential, and this is marked with a dashed line, with the value shown to the right, on the lower trace. Scale bars apply to all traces. (b), Example of the current response (top trace) to the voltage-step protocol (bottom trace) used to investigate subthreshold voltage-activated currents in NTS cells. The trace shows an average of 5 sweeps and demonstrates the presence of I_h during the hyperpolarising step, from -60 to -90 mV, and I_As during the depolarising step from -90 to -40 mV. (c), Quantification of the incidence of subthreshold voltage-activated currents in NTS cells demonstrated that most cells exhibited I_As, and a small proportion also displayed I_h. I_Ca,T was seen in few cells, while I_Ar was not present in any of the cells. A comparison of the incidence of subthreshold voltage-activated currents in NTS cells to those previously reported in NPFF, Substance P (SP), GRPR and GRP-GFP neurons is shown in (d).

The resting membrane potential of NTS cells was -53.0 ± 10.4 mV, their input resistance was 910.9 ± 438.3 MΩ and their capacitance 7.0 ± 2.8 pF. The rheobase current of NTS cells was 30.0 ± 20.5 pA, with the following parameters measured from the first action potential at rheobase; action potential threshold (defined as the point where the rate of rise exceeded 10 mV/ms) was -32.8 ±7.9 mV, the latency between the onset of the depolarising step to the first action potential was 111.6 ± 113.9 ms, action potential width was 1.9 ± 0.3 ms, action potential height (measured as the difference between voltage threshold and the action potential peak) was 58.1 ± 11.4 mV and after-hyperpolarisation was -29.0 ± 8.0 mV. Few NTS cells (4/36; 11.1 %) exhibited spontaneous action potential firing (defined as ≥1 AP/min), with the mean firing frequency being 0.1 ± 0.1 Hz.

The instances of subthreshold voltage-activated currents in NTS cells were investigated using a voltage step protocol that can reveal the presence of 2 types of transient outward current and two types of inward current, which are consistent with rapid (I_Ar) or slow (I_As) A-type potassium currents (I_A), and the low-threshold “T-type’ calcium current (I_Ca,T) or the hyperpolarisation-activated current (I_h), respectively. Almost all cells tested exhibited I_As (10/11; 90.9 %; Figure 5(b) and (c)), and in these instances the currents displayed a slow onset (mean time to peak = 41.5 ± 4.5 ms, range 30.8 to 46.7 ms) and were not fully inactivated by the end of the depolarising step. Of those cells that displayed I_As two also displayed I_h (18.2 %). I_Ca,T was seen in only one cell (9.1 %) and no cells exhibited I_Ar. The peak amplitude of the I_As was 233.7 ± 82.2 pA, and the amplitude of the I_h (measured as the mean of the final 200 ms of the hyperpolarising step) was -12.0 ± 6.8 pA.

Excitatory synaptic input to NTS cells was investigated by recording spontaneous (sEPSC) and miniature (mEPSC) excitatory postsynaptic currents, at a holding potential of -70 mV, the latter in the presence of tetrodotoxin (0.5 µM), bicuculline (10 µM) and strychnine (1 µM). The frequency of the EPSCs was 0.42 ± 0.80 and 0.06 ± 0.12 Hz for sEPSCs and mEPSCs, respectively (Figure 6(b) and (c)). In five instances, both sEPSCs and mEPSCs were recorded in the same cell, and while the mEPSC frequency was lower this difference was not significant (0.18 ± 0.21 vs. 0.07 ± 0.13 Hz, P = 0.063, Wilcoxon matched-pairs signed rank test, Figure 6(d)).

Figure 6.

Excitatory synaptic input to Nts^Cre-positive neurons. (a), Example electrophysiological traces of sEPSCs, and mEPSCs recorded prior to (black trace) and during capsaicin application (red trace) in the same cell. Quantification of EPSC frequency is displayed in (b), (sEPSC) and (c), (mEPSC). (d), In 5 instances both sEPSCs and mEPSCs were recorded in the same cell. While the frequency of mEPSCs was lower in all cells, this difference was not significant. (e), Application of the TRPV1 agonist, capsaicin, did not alter the mEPSC frequency in these cells, indicating that NTS cells receive limited (or no) input from TRPV1-expressing primary afferents.

Input to NTS cells from TRPV1-expressing primary afferents was investigated by recording mEPSCs in response to the TRPV1 agonist, capsaicin (2 µM). Capsaicin did not alter mEPSC frequency (0.07 ± 0.13 vs. 0.03 ± 0.03 Hz, n = 5, P = 0.625, Wilcoxon matched-pairs signed rank test, Figure 6(e)), suggesting that NTS cells do not receive monosynaptic input from TRPV1-expressing primary afferents, which include peptidergic nociceptors.^48,49

Comparisons with previously reported electrophysiological data obtained from four other classes of excitatory interneuron, those expressing substance P, GRPR or NPFF, as well as the GRP-GFP cells,^13,17,18 revealed significant differences between the NTS cells examined in this study, and each of these other populations (Figure 5(d), Table 3). For example, the resting membrane potential was more depolarised than that for the GRPR or NPFF cells, while the rheobase was lower than that for GRPR cells but higher than that for the GRP-GFP cells. In addition, the sEPSC and mEPSC frequencies were considerably lower than was the case for cells expressing substance P, GRPR or NPFF, but were comparable to those of the GRP-GFP cells. In addition, the NTS cells were the only group that lacked I_Ar and had a high incidence of I_As.

Table 3.

Comparison of electrophysiological data with those obtained from other classes of dorsal horn excitatory neuron.

​	NTS	GRP-GFP	SP	GRPR	NPFF
Capacitance (pF)	7.1 ± 2.7 (n = 36)	5.1 ± 1.6 *** (n = 323)	7.1 ± 3.0 (n = 83)	9.9 ± 2.7 **** (n = 256)	10.6 ± 2.2 **** (n = 31)
Input resistance (MΩ)	911 ± 438 (n = 36)	1588 ± 1297 ** (n = 323)	836 ± 487 (n = 82)	767 ± 473 (n = 223)	750 ± 307 (n = 31)
Resting membrane potential (mV)	-53.0 ± 10.4 (n = 36)	-52.9 ± 11.8 (n = 230)	-55.9 ± 8.2 (n = 82)	-58.9 ± 8.9 ** (n = 218)	-59.0 ± 8.2 * (n = 31)
Rheobase (pA)	30.0 ± 20.5 (n = 34)	18.3 ± 13.4 ** (n = 155)	33.4 ± 14.8 (n = 22)	68.7 ± 43.4 **** (n = 111)	26.9 ± 20.5 (n = 26)
Action potential threshold (mV)	-32.0 ± 7.9 (n = 34)	-22.5 ± 6.0 **** (n = 155)	-33.6 ± 8.5 (n = 22)	-30.0 ± 7.3 (n = 111)	-35.3 ± 5.4 (n = 26)
Action potential latency (ms)	111.6 ± 113.9 (n = 34)	137.1 ± 77.6 (n = 155)	519.3 ± 338.2 **** (n = 22)	349.5 ± 380.0 (n = 111)	321.8 ± 235.8 ** (n = 26)
Action potential width (ms)	1.9 ± 0.3 (n = 34)	2.4 ± 1.6 (n = 155)	1.6 ± 0.3 * (n = 22)	1.4 ± 0.4 **** (n = 111)	1.4 ± 0.5 **** (n = 26)
Action potential height (mV)	58.1 ± 11.4 (n = 34)	49.4 ± 12.9 ** (n = 155)	54.4 ± 14.7 (n = 22)	45.2 ± 10.4 **** (n = 111)	64.8 ± 10.2 (n = 26)
After-hyperpolarisation (mV)	-29.0 ± 8.0 (n = 34)	-32.9 ± 7.6 * (n = 155)	-27.3 ± 3.9 (n = 22)	-25.9 ± 4.9 (n = 111)	-28.42 ± 5.2 (n = 26)
sEPSC frequency (Hz)	0.43 ± 0.82 (n = 22)	0.20 ± 0.50 (n = 120)	4.72 ± 4.65 **** (n = 27)	4.34 ± 5.26 **** (n = 189)	6.86 ± 6.15 **** (n = 27)
mEPSC frequency (Hz)	0.07 ± 0.13 (n = 5)	0.05 ± 0.18 (n = 32)	3.24 ± 2.21 *** (n = 11)	1.02 ± 1.34 * (n = 42)	2.35 ± 1.99 *** (n = 15)

Data for GRP-GFP and substance P (SP) -expressing cells was obtained from reference 13, that for GRPR-expressing cells from reference 17 and that for NPFF cells from reference 1.

Significant differences from the neurotensin (NTS) cells are indicated in bold, and are shown as follows: *P<0.05, **P<0.01, P<0.001, ****P<0.0001, Kruskal-Wallis test, followed by Dunn’s multiple comparison test vs. NTS data.

Discussion

Our main findings are: (1) that neurons containing detectable levels of neurotensin are morphologically diverse, although (like many other dorsal horn neurons) they generally have dendrites that are most extensive along the rostrocaudal axis, (2) that although these cells receive primary afferent input from A- and C-LTMRs, this accounts for less than half of their excitatory synapses, and (3) that Cre-expressing cells in the NTS^Cre line show similarities and differences in terms of electrophysiological properties when compared to other neurochemically defined populations of excitatory spinal interneurons.

Neurotensin-expressing cells as a distinct population among excitatory interneurons

Häring et al. ⁹ reported that neurotensin was expressed in 12 of their 15 transcriptomic populations of excitatory dorsal horn neurons, although it was seen at high levels in only one population, Glut4. These cells were located on either side of the lamina II-III border, closely matching the distribution of neurotensin-immunoreactive cells. ¹⁶ It is therefore likely that the virally-labelled (Cre+) neurotensin-immunoreactive cells seen in the present study correspond to the Glut4 population. Those Cre+ cells that did not contain neurotensin may belong to other populations with lower levels of expression. ⁹

The great majority of the neurotensin-expressing cells were PKCγ-immunoreactive, and we have previously shown that neurotensin-positive cells account for nearly half of the neurons with high levels of PKCγ, while most of the remainder express CCK. ¹⁵ The CCK-expressing neurons presumably correspond to the Glut2-3 populations of Häring et al., ⁹ based on laminar location and PKCγ expression. Cells with strong PKCγ-immunoreactivity in lamina II/III can therefore be divided into two main subpopulations: those in Glut4 (expressing high levels of neurotensin) and those in Glut2/Glut3 (CCK-expressing).

Perl and colleagues^20,50–52 combined anatomical and electrophysiological approaches to identify distinct classes of interneuron in rodent SDH and investigate their locations in synaptic circuits. They defined 3 main classes of excitatory interneuron, vertical cells, with ventrally-directed dendrites, radial cells, with radiating dendritic trees, and a population of cells with relatively short rostrocaudally-orientated dendrites and a transient firing pattern, which they named “transient central” cells. Alba-Delgado et al. ⁵³ identified two morphological types among PKCγ-immunoreactive neurons in the spinal trigeminal nucleus: central and radial cells. Our morphological reconstructions show that many of the neurotensin-containing cells had dendritic trees that were elongated in the rostrocaudal axis, and could therefore be classified as central cells. ²⁰ Grudt and Perl reported that radial cells were typically found in the middle part of lamina II, ²⁰ whereas neurotensin-expressing cells are clustered around the lamina II/III border, making it unlikely that they correspond to the radial population. Electrophysiological comparisons are more difficult to make, as we were not able to restrict our recordings to cells belonging to the Glut4 population. ⁹ However, the majority of our cells showed transient firing (also described as “initial burst” or “phasic”), which is commonly associated with excitatory interneurons in this region,^39,54–56 and this is consistent with the report by Lu et al. that PKCγ-immunoreactive neurons also showed transient firing. ²⁶ Despite their resemblance to central cells and their transient firing pattern, the neurotensin cells are apparently distinct from the transient central population identified by Perl and colleagues, as Lu et al. did not assign any of their PKCγ-immunoreactive cells to this class. ²⁶

Synaptic inputs to neurotensin cells

There are substantial differences in the density of different types of excitatory bouton in different laminae and between areas innervated by hairy and glabrous skin. This will be reflected in differences in input to cells with dendrites in these regions. For the VGLUT1 and VGLUT3 analyses, we selected cells based on the location of their dendritic trees, and the percentages of input that we calculated will not be representative of all neurotensin-expressing neurons. Nonetheless, our findings suggest that between them, A- and C-LTMRs account for no more than 30% of the excitatory synapses on the neurotensin cells in areas innervated from hairy skin, and the proportion may be lower in glabrous skin territory. Several studies have examined synaptic inputs to PKCγ-immunoreactive dorsal horn neurons.^24,25,36,57 Three studies provided evidence for A-LTMR input,^24,25,57 although the proportion of excitatory synapses formed by LTMRs was 55% in one study, ²⁴ but only 15% in another. ⁵⁷ Also, while two of the studies found numerous synapses from C-LTMRs,^24,36 one found no evidence for such an input. ⁵⁷ Differences between our findings and those reported for PKCγ-immunoreactive neurons could reflect differences in synaptic input to the neurotensin- and CCK-expressing subsets. In addition, it is likely that distal dendrites of Brainbow-labelled cells can be followed further than those of PKCγ-immunoreactive cells, allowing a more complete analysis of synaptic input.

The largest source of input that we identified consisted of VGLUT2-immunoreactive boutons, which accounted for ∼40% of the excitatory synapses. Since VGLUT2 expression is low in primary afferents, ³² these boutons probably originate mainly from local excitatory interneurons, with potential contributions from axon collaterals of projection neurons or axons descending from the brainstem. Surprisingly, the types of bouton that we analysed (VGLUT1, VGLUT2, VGLUT3) only accounted for ∼70% of excitatory synapses on the neurotensin cells. Part of the deficit may be accounted for by variability between cells, resulting from the limited sample of neurons examined. However, given the size of the deficit it may be that axons containing detectable levels of these transporters do not account for all of the excitatory synapses on the neurotensin cells. Other candidates could be nociceptive primary afferents, which often contain very low levels of VGLUT2 and lack the other transporters. ³² It is unlikely that neurotensin neurons receive significant input from peptidergic nociceptors, as these terminate in lamina I and the outermost part of lamina II and express TRPV1, and none of the cells tested showed a capsaicin-induced increase in mEPSC frequency. However, non-peptidergic nociceptors may account for some of the input, as these arborise in lamina II. Neumann et al. ²⁵ reported that there was no overlap between the band of IB4-labelling (which corresponds to non-peptidergic nociceptors) and the PKCγ plexus, and Peirs et al. ⁵⁷ did not find synapses between PKCγ dendrites and central axons of type I synaptic glomeruli, ⁴⁵ which originate from non-peptidergic nociceptors. However, some neurotensin cells have dendrites that extended dorsally into the outer half of lamina II, and may therefore receive input from these afferents on their dorsal dendrites.

Involvement of neurotensin cells in dorsal horn circuits

Lu et al. ²⁶ performed paired recordings from ex vivo preparations and proposed a serial circuit involving excitatory synaptic connections between Aβ low-threshold mechanoreceptors, PKCγ interneurons, transient central cells, vertical cells and lamina I projection neurons. They suggested that this was normally kept closed by feedforward glycinergic inhibition acting on the PKCγ cell, but that disinhibition occurring in neuropathic states opened the circuit, allowing low-threshold mechanical stimuli to gain access to nociceptive lamina I projection neurons, leading to mechanical allodynia.

We have integrated our neurochemical classification scheme ⁹ with the populations defined by Perl and colleagues,^13,17–19 and the transcriptomic classes of Häring et al. ⁹ In this way, we have identified two distinct classes among the vertical cells, defined by expression of GRPR (likely Glut12) or NPFF (Glut9), we have shown that many lamina II neurons expressing substance P (Glut10-11) correspond to radial cells, and have found that the GRP-GFP cells (probably belonging to Glut8) correspond to transient central cells.^{13,17–19,30} Based on this, we predict that at least some of the synapses linking transient central to vertical cells are between GRP-GFP and GRPR-expressing neurons, and may therefore be more important for itch than for mechanical allodynia.^17,58,59

Hachisuka et al. ⁶⁰ used the same NTS^Cre line in a study investigating windup in lamina I ALS neurons. They showed that Cre-expressing cells innervate ALS lamina I neurons both directly and via polysynaptic circuits, and that they contributed to windup of these cells, probably via a reverberatory circuit. However, as they noted, their use of reporter lines to manipulate Cre-expressing cells means that they are likely to have targeted a broader group of cells than the Glut4 population.

At present we have some limited information concerning the postsynaptic targets of neurotensin cells. In the present study, we found that their axons terminate mainly in lamina IIi, with some extension into laminae I-IIo and III. However, unlike GRPR- NPFF- and substance P-expressing cells, they do not enter the LSN.^14,17,61 We have previously reported that they account for 8% of excitatory synapses on GRPR cells, thus providing a direct link from PKCγ neurons to vertical cells. ⁶² In addition, neurotensin-immunoreactive axons selectively target substance P-expressing radial cells, providing ∼20% of their excitatory synapses.

Overall, our results are consistent with the proposed role of PKCγ cells in conveying LTMR input as part of an allodynia circuit. However, synapses from LTMRs account for only a minority of the excitatory input to these cells. Taken together with our previous findings they suggest additional roles for the neurotensin cells, for example through their direct input to vertical cells, as well as their input to substance P-expressing radial cells, which were not part of the allodynia circuit proposed by Lu et al. ²⁶ Future studies, in which molecular genetic approaches are used to inhibit, silence or ablate these cells should reveal their roles in normal and pathological pain states.

Appendix.

List of abbreviations

BFP

Blue fluorescent protein

CCK

Cholecystokinin

GFP

Green fluorescent protein

GRP

Gastrin-releasing peptide

GRPR

Gastrin-releasing peptide receptor

LSN

Lateral spinal nucleus

NKB

Neurokinin B

NPFF

Neuropeptide FF

PKCγ

Protein kinase Cγ

SDH

Superficial dorsal horn

TFP

Teal fluorescent protein.

ORCID iDs

Allen C. Dickie https://orcid.org/0000-0002-6339-2801

Andrew J Todd https://orcid.org/0000-0002-3007-6749

Data Availability Statement

The datasets generated and analysed during the current study are available from the corresponding authors on reasonable request.*