ErbB1-dependent signalling and vesicular trafficking in primary afferent nociceptors associated with hypersensitivity in neuropathic pain.
Effective analgesic treatment for neuropathic pain remains an unmet need, so previous evidence that epidermal growth factor receptor inhibitors (EGFRIs) provide unexpected rapid pain relief in a clinical setting points to a novel therapeutic opportunity. The present study utilises rodent models to address the cellular and molecular basis for the findings, focusing on primary sensory neurons because clinical pain relief is provided not only by small molecule EGFRIs, but also by the anti-EGFR antibodies cetuximab and panitumumab, which are unlikely to access the central nervous system in therapeutic concentrations. We report robust, rapid and dose-dependent analgesic effects of EGFRIs in two neuropathic pain models, matched by evidence with highly selective antibodies that expression of the EGFR (ErbB1 protein) is limited to small nociceptive afferent neurons. As other ErbB family members can heterodimerise with ErbB1, we investigated their distribution, showing consistent co-expression of ErbB2 but not ErbB3 or ErbB4, with ErbB1 in cell bodies of nociceptors, as well as providing evidence for direct molecular interaction of ErbB1 with ErbB2 in situ. Co-administration of selective ErbB1 and ErbB2 inhibitors produced clear evidence of greater-than-additive, synergistic analgesia; highlighting the prospect of a unique new combination therapy in which enhanced efficacy could be accompanied by minimisation of side-effects. Peripheral (intraplantar) administration of EGF elicited hypersensitivity only following nerve injury and this was reversed by local co-administration of selective inhibitors of either ErbB1 or ErbB2. Investigating how ErbB1 is activated in neuropathic pain, we found evidence for a role of Src tyrosine kinase, which can be activated by signals from inflammatory mediators, chemokines and cytokines during neuroinflammation. Considering downstream consequences of ErbB1 activation in neuropathic pain, we found direct recruitment to ErbB1 of an adapter for PI 3-kinase and Akt signalling together with clear Akt activation and robust analgesia from selective Akt inhibitors. The known Akt target and regulator of vesicular trafficking, AS160 was strongly phosphorylated at a perinuclear location during neuropathic pain in an ErbB1-, ErbB2- and Akt-dependent manner, corresponding to clustering and translocation of an AS160-partner, the vesicular chaperone, LRP1. Exploring whether neuronal ion channels that could contribute to hyperexcitability might be transported by this vesicular trafficking pathway we were able to identify Na < sub > v < /sub > 1.9, (Na < sub > v < /sub > 1.8) and Ca < sub > v < /sub > 1.2 moving towards the plasma membrane or into proximal axonal locations - a process prevented by ErbB1 or Akt inhibitors. Overall these findings newly reveal both upstream and downstream signals to explain how ErbB1 can act as a signalling hub in neuropathic pain models and identify the trafficking of key ion channels to neuronal subcellular locations likely to contribute to hyperexcitability. The new concept of combined treatment with ErbB1 plus ErbB2 blockers is mechanistically validated as a promising strategy for the relief of neuropathic pain.
Contents lists available at ScienceDirect Neurobiology of Disease journal homepage: www.elsevier.com/locate/ynbdi ErbB1-dependent signalling and vesicular trafficking in primary afferent nociceptors associated with hypersensitivity in neuropathic pain Rory Mitchella, Marta Mikolajczaka, Christian Kerstenb, Sue Fleetwood-Walkera,⁎ a Centre for Discovery Brain Sciences, Edinburgh Medical School, College of Medicine Veterinary Medicine, University of Edinburgh, Edinburgh, UK b Center for Cancer Treatment, Sorlandet Hospital, Pb 416, 4604 Kristiansand, Norway A R T I C L E I N F O Keywords: Neuropathic pain EGF receptor ErbB1 ErbB2 Cetuximab Src Akt Vesicular trafficking Nav1.7/1.8/1.9 Cav1.2 A B S T R A C T Effective analgesic treatment for neuropathic pain remains an unmet need, so previous evidence that epidermal growth factor receptor inhibitors (EGFRIs) provide unexpected rapid pain relief in a clinical setting points to a novel therapeutic opportunity. The present study utilises rodent models to address the cellular and molecular basis for the findings, focusing on primary sensory neurons because clinical pain relief is provided not only by small molecule EGFRIs, but also by the anti-EGFR antibodies cetuximab and panitumumab, which are unlikely to access the central nervous system in therapeutic concentrations. We report robust, rapid and dose-dependent analgesic effects of EGFRIs in two neuropathic pain models, matched by evidence with highly selective antibodies that expression of the EGFR (ErbB1 protein) is limited to small nociceptive afferent neurons. As other ErbB family members can heterodimerise with ErbB1, we investigated their distribution, showing consistent coexpression of ErbB2 but not ErbB3 or ErbB4, with ErbB1 in cell bodies of nociceptors, as well as providing evidence for direct molecular interaction of ErbB1 with ErbB2 in situ. Co-administration of selective ErbB1 and ErbB2 inhibitors produced clear evidence of greater-than-additive, synergistic analgesia; highlighting the prospect of a unique new combination therapy in which enhanced efficacy could be accompanied by minimisation of side-effects. Peripheral (intraplantar) administration of EGF elicited hypersensitivity only following nerve injury and this was reversed by local co-administration of selective inhibitors of either ErbB1 or ErbB2. Investigating how ErbB1 is activated in neuropathic pain, we found evidence for a role of Src tyrosine kinase, which can be activated by signals from inflammatory mediators, chemokines and cytokines during neuroinflammation. Considering downstream consequences of ErbB1 activation in neuropathic pain, we found direct recruitment to ErbB1 of an adapter for PI 3-kinase and Akt signalling together with clear Akt activation and robust analgesia from selective Akt inhibitors. The known Akt target and regulator of vesicular trafficking, AS160 was strongly phosphorylated at a perinuclear location during neuropathic pain in an ErbB1-, ErbB2- and Akt-dependent manner, corresponding to clustering and translocation of an AS160-partner, the vesicular chaperone, LRP1. Exploring whether neuronal ion channels that could contribute to hyperexcitability might be transported by this vesicular trafficking pathway we were able to identify Nav1.9, (Nav1.8) and Cav1.2 moving towards the plasma membrane or into proximal axonal locations – a process prevented by ErbB1 or Akt inhibitors. Overall these findings newly reveal both upstream and downstream signals to explain how ErbB1 can act as a signalling hub in neuropathic pain models and identify the trafficking of key ion channels to neuronal subcellular locations likely to contribute to hyperexcitability. The new concept of combined treatment with ErbB1 plus ErbB2 blockers is mechanistically validated as a promising strategy for the relief of neuropathic pain.
1. IntroductionChronic hypersensitive pain states, especially those due to nerve injury (neuropathic pain) are difficult to treat with current medications, which show limited efficacy and frequently treatment-limiting side effects (Finnerup et al., 2010). There is a clear unmet need to develop new, safe and efficacious analgesics for neuropathic pain through novel targeting strategies. One such approach originated from our serendipitous observation of remarkable pain relief in a rectal cancer patient treated with the anti-epidermal growth factor receptor (EGFR) https://doi.org/10.1016/j.nbd.2020.104961 Received 14 April 2020; Received in revised form 26 May 2020; Accepted 8 June 2020 ⁎ Corresponding author at: Centre for Discovery Brain Sciences, Edinburgh Medical School: Biomedical Sciences, University of Edinburgh, Hugh Robson Building, George Square, Edinburgh EH8 9XD, UK. E-mail address: firstname.lastname@example.org (S. Fleetwood-Walker). Neurobiology of Disease 142 (2020) 104961 Available online 10 June 2020 0969-9961/ © 2020 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/). T antibody, cetuximab, despite tumour progression (Kersten and Cameron, 2012), which initiated a reverse-translational programme to elucidate the role of the EGFR in pain processing. Clinical follow-up studies on a small case series and then a larger study of patients with neuropathic pain of different origins (both cancer and non-cancer) showed similarly promising results (80% clinically meaningful response) with cetuximab, panitumumab and the small molecule EGFR inhibitors, gefitinib and erlotinib (Kersten et al., 2013; Kersten et al., 2015). A small-scale randomised proof-of-concept trial of cetuximab in neuropathic pain patients (nerve compression and complex regional pain syndrome), corroborated the idea of a clinically meaningful analgesic signal (Kersten et al., 2019). Importantly, many of the patients responding to EGFR inhibitors had pain that had proved refractory to all other interventions, emphasizing the potential of this novel strategy. Some mechanistic insight arises from the efficacy displayed by cetuximab and panitumumab (which are unlikely to penetrate the bloodbrain barrier in therapeutic concentrations) and therefore implicates primary sensory afferents as a likely site of action. Independent support for the idea of EGFR involvement in chronic pain has come from molecular genetics analysis (Martin et al., 2017; Verma et al., 2020) of patients with temporomandibular disorders – a group of conditions associated with chronic pain from joint, muscle or central origins (Cairns, 2010). A number of single nucleotide polymorphisms (SNPs) in the genes for epiregulin (an EGFR ligand) and EGFR showed significant differential association with TMD patient and control groups (Martin et al., 2017, Verma et al., 2020). The minor allele of the epiregulin SNP most strongly associated with reduced odds of chronic TMD was matched to lower epiregulin mRNA levels in leukocytes (Martin et al., 2017). However, the minor allele was also associated with increased numbers of acute pain sites in TMD patients (Verma et al., 2020). Rodent chronic pain models have provided further support for a role of the EGFR in pain processing. The hypersensitivity due to nerve injury, inflammation or direct compression injury to DRG (CCD) was attenuated by EGFR-selective inhibitors (although across a range of potencies) and also (in the case of CCD) by EGFR siRNA (Martin et al., 2017; Wang et al., 2019). Furthermore, mice expressing a constitutively activating EGFR mutation show enhanced responses to formalin, while Drosophila larvae with EGFR knockdown in peripheral nociceptors display impaired thermal nociception (Martin et al., 2017). The EGFR (ErbB1) is one of a family of 4 receptor tyrosine kinases (RTKs), which display sequence homology and a common structural organisation (Olayioye et al., 2000). ErbB2 is thought to lack a physiological ligand yet act as the preferred heterodimerisation partner for all other ErbB receptors (Graus-Porta et al., 1997; Tzahar et al., 1997) and can enable constitutive functional activity (Garrett et al., 2003), while ErbB3 has ligands but displays minimal levels of tyrosine kinase catalytic activity (around 1000-fold less than ErbB1) (Shi et al., 2010). The functional importance of ErbB1:ErbB2 heterodimers is emphasised, for example, by the elevated basal phosphorylation of both ErbB1 and ErbB2 in co-transfected cells and the potent inhibition of ErbB2-overexpressing tumour and cell line proliferation by highly selective ErbB1 inhibitors (Moasser et al., 2001). ErbB dimerisation generally comes about as a result of ligand-induced conformational changes stabilising an extended configuration of extracellular domains (Dawson et al., 2005). The chimeric human/murine monoclonal antibody, cetuximab (Kim et al., 2001; Harding and Burtness, 2005) binds to an extracellular domain of ErbB1 that partly occludes the ligand binding site and also inhibits transition to the extended configuration necessary for dimerisation (Li et al., 2005; Berger et al., 2011) but importantly results in down-regulation of EGFR expression (Perez-Torres et al., 2006). Although the classical paradigm for ErbB receptor activation centres on ligand-induced initiation of dimerisation, there is evidence for ligandindependent formation of homo- and hetero-dimers (including ErbB1:ErbB2) and in some cases for their initial assembly in the endoplasmic reticulum (Verveer et al., 2000; Moriki et al., 2001; Sawano et al., 2002; Yu et al., 2002; Kumagai et al., 2003; Tao and Maruyama, 2008; Junttila et al., 2009). ErbB receptors undergo constant internalisation, intracellular shuttling and recycling (Wiley, 2003). Although ErbB1 is predominantly localised at the cell surface in fibroblasts, other cell types maintain large intracellular pools of unliganded receptor (Burke and Wiley, 1999; Kim et al., 2001). In the case of cetuximab binding to ErbB1, the internalised complex is resistant to dissociation in the acidified environment of endosomes and is therefore targeted for lysosomal degradation rather than recycling, thereby quickly depleting the shuttling pool of ErbB1 receptor protein (Jimeno et al., 2005; Li et al., 2005). Neuronal ErbB1 expression has been reported in both the CNS and dorsal root ganglia (DRG) (Gomez-Pinilla et al., 1988; Ferrer et al., 1996; Xian and Zhou, 1999). In DRG, expression has been reported to be preferentially in smaller neurons (Huerta et al., 1996; Xian and Zhou, 1999), whereas other studies have suggested a widespread neuronal localisation (Andres et al., 2010; Martin et al., 2017) or preferential expression in medium/large myelinated neurons and minimal co-expression with markers of small unmyelinated neurons (Wang et al., 2019). Some studies also report ErbB1 expression in glial cells within DRG (Andres et al., 2010; Wang et al., 2019). Expression of other ErbB family members has also been reported in DRG neurons (Pearson Jr. and Carroll, 2004). ErbB receptors achieve diverse signalling outcomes by trans- or autophosphorylation of Tyr residues in their intracellular carboxyterminal domains, which act as high affinity docking sites for various signalling and adapter proteins (Olayioye et al., 2000). ErbB1 autophosphorylation at Tyr1068 primarily leads to the recruitment of signalling adapters, Grb2 (to regulate the ERK MAP kinase pathway) and Gab1 (to enable PI 3-kinase binding and activation, then downstream activation of Akt) (Lowenstein et al., 1992; Downward, 1994; HolgadoMadruga et al., 1996; Rodrigues et al., 2000; Mattoon et al., 2004; Cao et al., 2009). In endothelial cells stably transfected with ErbB1 or ErbB1 plus ErbB2, there is evidence for ligand-independent activation as shown by significant basal phosphorylation of ErbB1, ErbB2 and ERK (Berger et al., 2011). In this model, cetuximab markedly reduced not only EGF-induced responses, but also basal, unstimulated ERK phosphorylation, clearly impacting on ligand-independent downstream signalling from ErbB1/2. In view of the clinical findings pointing to the potential of EGFR inhibition as a strategy for alleviation of chronic hypersensitive pain, we sought to corroborate this in a controlled laboratory setting and explore some of the underlying mechanisms. Although the humanised/ human monoclonal antibodies, cetuximab and panitumumab, would not be useful tools in rodent studies due to immune incompatibility, many small molecule inhibitors of different ErbB family receptors have been developed because of their therapeutic potential in oncology (Roskoski Jr., 2014; Kavuri et al., 2015). A number of these tyrosine kinase inhibitors (TKIs) are very highly selective for particular ErbB receptors, against other ErbB family members and panels of other protein kinases (Fabian et al., 2005; Anastassiadis et al., 2011; Davis et al., 2011).
2.1. AnimalsAll animal breeding, maintenance and experimental procedures complied with ARRIVE guidelines and were carried out in accordance with the UK Animals (Scientific Procedures) Act 1986, with approval from the University of Edinburgh's Local Ethical Review Board. Animals were housed under a 12 h light-dark cycle and given access to food and water ad libitum. Most experiments were carried out on adult SpragueDawley rats (200–350 g in weight, which were male unless otherwise indicated). The sciatic nerve chronic constriction injury (CCI) preparation was used as a model of chronic neuropathic pain in rats (Bennett and Xie, 1988). Animals were anaesthetised with a 4% isoflurane/oxygen mixture (Zeneca, Cheshire, UK) before exposure of the sciatic nerve, proximal to the trifurcation at mid-thigh level. Four loose ligatures of chromic catgut (SMI AG, Hunningen, Belgium) were tied around the nerve, separated by 1mm. CCI rats consistently developed ipsilateral thermal reflex hypersensitivity, which peaked between days 8 and 12 post-surgery, when experiments were carried out. Some experiments were carried out in adult male C57/Bl6J mice (25–35 g in weight), using an oxaliplatin model of chemotherapy-induced peripheral neuropathy (CIPN) (Zhao et al., 2012). Oxaliplatin (Abcam) in 5% glucose vehicle was injected intraperitoneally (under light isoflurane anaesthesia) at a dose of 7mg/kg (100 μl/mouse) and experiments carried out 3–7 days later when mechanical hypersensitivity was fully developed.
2.2. Quantitative sensory testing in vivoMechanical and thermal nociceptive sensitivity were assessed using standard behavioural tests, during which, the investigator was blinded to drug treatment. Mechanical nociception was assessed as Paw Withdrawal Threshold (PWT; in g) from force-calibrated von Frey nylon filaments (Stoelting, Illinois). Nociceptive heat sensitivity was measured as Paw Withdrawal Latency (PWL; in sec) using Hargreaves' infrared apparatus (Linton Instrumentation), set to a maximum temperature of 52 °C and a cut-off time of 20s. Testing was always separated by at least 5 min to avoid sensitisation of responses. Nociceptive cold sensitivity was measured as Suspended Paw Elevation Time using a shallow water bath at the temperature of iced water (~4 °C) (Garry et al., 2003). Animals were initially habituated to the sensory testing environments to establish consistent responses.
2.3. Drug administration in vivoSmall molecule agents were obtained from MedChem Express, Selleckchem, Key Organics or Axon MedChem. Murine EGF, purified from submaxilliary gland and recombinant human neuregulin1-β1 EGF domain (NRG1) were obtained from Sigma-Aldrich and R D Systems, respectively. For intraperitoneal administration (200 μl/rat and 100 μl/ mouse) agents were dissolved in 10% dimethylsulphoxide, 40% polyethylene glycol-400 and 50% propylene glycol. For intraplantar administration (50 μl/rat) agents were dissolved in normal saline with 0.3% dimethylsulphoxide. Vehicle controls showed no discernible change in pain-associated behavioural responses. Doses of pharmacological agents were selected on the basis of those showing clear efficacy/target coverage (producing at least 80–90% maximal effects) in literature reports, wherever possible, or their reported potency in vitro compared to that of the benchmark, erlotinib.
2.4. Immunofluorescence histochemistryDRGs from spinal segments L4–6 were dissected, rapidly embedded in cryo-sectioning medium (Thermo Scientific) and frozen on dry ice. 18 μm sections were cut by cryostat and mounted onto poly-L-lysine coated slides (Thermo Scientific). Sections were washed in TrisBuffered Saline (TBS) pH 7.60 and then incubated for 1 h at room temperature in blocking buffer (10% normal donkey serum, 4% fish skin gelatin, 0.2% Triton X-100 in TBS) prior to overnight incubation at 4 °C with a combination of primary antibodies in buffer (4% normal donkey serum, 4% fish skin gelatin and 0.2% Triton X-100 in TBS). The primary antibodies used are listed in Table 1. Sections were washed in TBS and incubated for 1 h at room temperature with a combination of extensively cross-adsorbed, fluorescent secondary antibodies or Alexa Fluor-488 conjugated isolectin IB4 (Invitrogen; I21411; 1:3000) in buffer (4% normal donkey serum, 4% fish skin gelatin in TBS). Donkey anti-mouse, rabbit, chicken or guinea pig secondary antibodies, labelled with 633/647, 568, 488 or 405 nm-emitting fluorophores in appropriate permutations, and all highly cross-adsorbed against other relevant species, were obtained from Biotium via Sigma-Aldrich and were used at a dilution of 1:600. Sections were washed three further times in TBS and then mounted in ProLong® Gold Antifade (Life Technologies). Standard primary antibody omission or blocking peptide controls, wherever possible, confirmed that non-specific staining was minimal. Both concentration-dependence of staining and association with peripherin-positive or NF-200-positive cellular profiles, where appropriate, corroborated specificity.
2.5. Confocal microscopy and image analysisFluorescence signals were acquired using a Nikon A1R confocal microscope at 1024×1024 pixel size frame, 12 bits per pixel images, with a pinhole size 1 AU calculated for 488, using objectives 10× Plan Fluor/NA0.3, 20× Plan Apo VC/NA 0.8, or 60× Plan Apo VC/NA1.4 (oil immersion). In all cases, emissions for each fluorophore were obtained sequentially to avoid channel bleed-through. Z-stacks were acquired covering the whole thickness of each section at a low power magnification (acquired with 10× or 20× objectives; 5 μm and 2 μm step size between planes) and maximum intensity projections were generated using ImageJ/Fiji. All cells positive for markers of interest were blind-counted manually (using ImageJ/Fiji software plugin), usually versus ErbB1 or peripherin across the whole section in each case. Mean percentage expression values were generated over sections from at least 4 separate animals. Antigen distribution within single cells (collected at a high power with optical zoom at 0.09×0.09 pixel size) was analysed by measuring fluorescence intensity using Plot Profile tool in ImageJ/Fiji in single plane images (at the centre of each cell) and drawing a transect from the nuclear perimeter to the neuronal apex, that was then divided into 50 bins. Visualization and three-dimensional reconstruction of images was performed using IMARIS (Bitplane). In the case of phospho-AS160 (Thr642), which was clustered tightly in a perinuclear location, only the proximal part of the transect (covering all the discernible fluorescence signal) was analysed by division into 50× 0.09 μm bins starting at the nuclear perimeter, and fluorescence intensity recorded at each. In order to facilitate comparisons between transect profiles, values were normalised. If the mean intensity from the naïve and CCI datasets differed by more than 40% (as might be anticipated for newly evoked changes such as target phosphorylation or clustering of previously dispersed target proteins into aggregates), normalisation was carried out to the overall mean intensity of the entire data set. This enabled detection of changes in relative intensity as well as the location of any changes. If this value was less than 40% (consistent with simple redistribution of the target protein), normalisation was carried out to the mean for the individual cell sampled, which enabled clearer analysis of changes in fine subcellular localisation. Mean values were calculated generally across 8–9 individual cell transects originating from 4 separate animals in each case.
2.6. Proximity ligation assayEvidence for in situ protein:protein interactions was obtained using the Duolink Proximity Ligation Assay (Sigma-Aldrich) (Fredriksson et al., 2002; Soderberg et al., 2006; Liu et al., 2013; Rivera-Oliver et al., 2019). Rabbit or mouse primary antibodies for potential protein partners are outlined in Table 1. Oligonucleotide-derivatised secondary antibody probes were donkey anti-rabbit PLUS and donkey anti-mouse MINUS. Probe ligation, rolling circle amplification (RCA) and detection of the RCA product by Duolink Orange fluorophore (emission 576 nm) were all carried out according to the manufacturer's instructions. An optimised procedure was developed to integrate conventional immunofluorescence counterstaining (with chicken anti-peripherin and donkey anti-chicken 488) into the protocol. 2.7. Non-linear curve-fitting and statistical analysis All data analysis was carried out using GraphPad Prism. Non-linear curve-fitting used a sigmoidal dose-response (variable slope) model. Data in two-group format were analysed statistically by Student's t-test. Comparisons between more than two groups were made by One-Way ANOVA with Tukey's or Dunnett's test, or by Two-Way ANOVA with Bonferroni's test.
3.1. ErbB1 involvement in hypersensitive pain behaviours and nociceptor activation following nerve injuryUsing the CCI neuropathic pain model, marked nociceptive hypersensitivity was observed in the ipsilateral (but not contralateral) hindlimb 8–12 days following nerve injury. This was manifest as a reduction in Paw Withdrawal Threshold (PWT) from von Frey filament mechanical stimuli (Fig. 1A), a reduction in Paw Withdrawal Latency (PWL) from Hargreaves' noxious thermal stimuli (Fig. 1B) and an increase in suspended paw elevation time from 4 °C noxious cold stimuli (Fig. 1C). In each case, intraperitoneal administration of the highly selective inhibitor of ErbB1 tyrosine kinase function, erlotinib (Moyer et al., 1997) at a dose of 10mg/kg ip, produced marked rapid reversal of ipsilateral hypersensitivity, which was complete at peak effect and remained statistically significant for 140–160min. No discernible effects were observed on responses from the contralateral limb or from vehicle injection. This time course is consistent with evidence that plasma levels of erlotinib increase rapidly after administration and begin to decline steeply after around 2 h (Pollack et al., 1999). The reversal of CCI-induced mechanical hypersensitivity by erlotinib was dose-dependent (with a statistically significant effect at as little as 1mg/kg and an IC50 of 1.4 [1.3/1.5] mg/kg ip (mean [95% confidence interval]), (Table 2), corresponding to the dose range (100–150mg) found to provide effective pain relief in patients (Kersten et al., 2015). The effect of erlotinib here was also replicated by further highly selective ErbB1 inhibitors, gefitinib (Moasser et al., 2001, Pedersen, Pedersen et al., 2005) and AG 1478 (Traxler et al., 1999), with effects similar to erlotinib at corresponding doses (Table 2). Reversal of nerve injury-induced mechanical hypersensitivity by erlotinib was similarly observed in female CCI rats (mean percentage reversal over 20–140min post-injection of 88.9 ± 9.2%, n=4, p .01, at a dose of 10mg/kg ip). As most selective ErbB1 inhibitors are based around a common quinazoline core, we tested two further, structurally distinct selective agents, falnidamol (BIBX 3182) (Solca et al., 2004) and EGFRi 324,674 (Zhang et al., 2006). These contain distinct pyrimidopyrimidine and 4,6-substituted-pyrimidine cores, respectively and at appropriate doses they also produced robust reversal of hypersensitivity through 20–140min post-administration, respectively (n=4, in each case), confirming target-specific efficacy (Table 2). Similar results (demonstrating dose-dependence in each case) were observed in a mouse model of oxaliplatin-induced peripheral neuropathy (Zhao et al., 2012), with IC50 values [95% CI] of 3.2 [3.0/3.3] mg/kg ip for erlotinib, 2.4 [2.3/2.6] mg/kg ip for gefitinib and 3.0 [2.7/3.4] mg/kg ip for AG 1478 (n=4 at 4–5 different doses in each case). In both rat CCI and mouse CIPN models, animals showed no discernible changes in general locomotion or motor function, or evidence of sedation, following administration of ErbB1 inhibitors. The doses of erlotinib producing analgesia in the neuropathic pain models here are similar to, or less than, those necessary to reverse EGF-induced autophosphorylation of ErbB1 in liver and tumour xenografts of athymic mice, consistent with appropriate target-associated potency and coverage (Moyer et al., 1997; Pollack et al., 1999). Repeated testing with erlotinib (10mg/kg ip) showed no significant attenuation of its ability to reverse CIPN-induced hypersensitivity over 7 daily treatments, similar to observations in the ErbB1:⁎no cross reaction with ErbB2, 3, 4; ErbB2:⁎⁎no cross reaction with ErbB1, 3, 4; ErbB3: †no cross reaction with ErbB1, 2, 4; ErbB4: ††no cross reaction with ErbB1, 2, 3. CCD (chronic DRG compression) model (Wang et al., 2019). In order to provide an unbiased biomarker of activity in nociceptive afferents we measured levels of phospho-CaM kinase IIα (Thr286) in DRG cells. This marker reflects Ca2+-dependent autophosphorylation of the enzyme to an autonomous form of the enzyme, but both activity and Thr286-phosphorylation diminish rapidly, within minutes, following stimulus removal (Lou et al., 1986; Schworer et al., 1986; Lengyel et al., 2004; Chang et al., 2017). Fig. 1D shows that phospho-CaMKII (Thr286) staining in small, peripherin-positive DRG cells was significantly increased ipsilateral to CCI and that this was almost completely reversed by erlotinib (10mg/kg ip, 1 h), indicating that CCIinduced activity in nociceptors is highly dependent on ErbB1 functional activity.
3.2. ErbB1 is selectively localised in small, peripherin-positive DRG neuronsAs there was some disparity in previous reports over ErbB1 expression in subpopulations of DRG cells (Huerta et al., 1996; Xian and Zhou, 1999; Andres et al., 2010; Martin et al., 2017; Wang et al., 2019), we addressed this using extensively characterised ErbB1-specific antibodies. Fig. 2A shows results using two antibodies documented to label ErbB1, but not other ErbB family members, revealing very high correlation between staining for ErbB1 and peripherin in small DRG cells that are likely to represent unmyelinated nociceptors (Goldstein et al., 1991; Fornaro et al., 2008). In DRG from naïve and CCI animals similarly, both ErbB1-specific antibodies stained 80–90% of peripherinpositive small cells with minimal staining outwith this population. A third ErbB1 antibody (mouse monoclonal 8G6.2); raised against a distinct intracellular epitope in ErbB1), confirmed these results, staining 84.8 ± 1.2% of peripherin-positive DRG cells from naïve animals (CCI not tested; mean ± SEM, n=4). In contrast, co-staining with the myelinated neuronal marker, NF-200 (Goldstein et al., 1991, Fornaro et al., 2008), the satellite glial cell marker, glutamine synthetase (Miller et al., 2002) and the microglial/recruited macrophage marker, CD68 (Hu and McLachlan, 2003) was minimal (Fig. 2B). The pattern and levels of ErbB1 expression in DRG were not significantly altered following CCI. Within the population of ErbB1-expressing small DRG cells, both presumed peptidergic C-fibres, staining for TRPV1 (Cavanaugh et al., 2009) and presumed non-peptidergic fibres, labelled with isolectin, IB4 (Dong et al., 2001) were strongly represented, with 87.1 ± 2.1% of TRPV1 (and peripherin)-positive cells, and 88.6 ± 2.1% of IB4 (and peripherin)-positive cells co-staining for ErbB1 (means± SEM, n=12). Similar values were found in CCI animals. These observations are consistent with the idea of ErbB1 playing an important role in nerve injury-induced hypersensitivity throughout the whole C-fibre nociceptor population. As ErbB1 may heterodimerise with other ErbB family members (Olayioye et al., 2000), we also investigated the expression in DRG of ErbB2–4, using highly characterised, isotype-specific antibodies. We found ErbB2 to be widely and selectively expressed in small, peripherin-positive DRG cells, like ErbB1, with around 90% of peripherinpositive cells expressing ErbB2 in both naïve and CCI animals (Fig. 2C). Similar results were obtained for ErbB2 expression in ErbB1-positive cells (Fig. 2C) and were confirmed with a second highly characterised ErbB2-specific antibody (mouse monoclonal, UMAB36) which showed staining in 76.2 ± 6.3% and 78.5 ± 4.4% of ErbB1-positive cells in naïve and CCI DRG, respectively (means± SEM, n=4). ErbB3- and ErbB4-specific antibodies showed minimal staining in DRG of naïve and CCI animals; with any staining rarely present in small, peripherin-positive cells, only 3–8% of the population in each case (Fig. 2C). These observations suggested that ErbB2, in addition to ErbB1, could potentially represent an analgesic target in nociceptors; acting either independently or in concert with ErbB1 as a heterodimer; an idea supported by evidence that ErbB2 is the preferred heterodimerisation partner for all of the other ErbB family members (Graus-Porta et al., 1997; Tzahar et al., 1997). 3.3. Functional evidence for a role of ErbB2 within C-fibre nociceptors in CCI-induced hypersensitivity Using paw withdrawal threshold in the von Frey test, we showed that several ErbB2 blockers caused marked reversal of CCI-induced mechanical hypersensitivity. The highly selective ErbB2 inhibitors, mubritinib and tucatinib (ARRY-380), both of which show very low affinity for ErbB1 (Nagasawa et al., 2006; Moulder et al., 2017), each at a dose of 15mg/kg ip, produced marked reversal of CCI-induced hypersensitivity, with complete reversal at peak in each case and both remaining statistically significant for 180min. The mean ± SEM percentage reversals of hypersensitivity from 20 to 140min following administration were 91.5 ± 4.7% (n=6) for mubritinib and 87.1 ± 6.9% (n=4) for tucatinib, respectively (Fig. 3A). This validated ErbB2, in addition to ErbB1, as a potential analgesic target for neuropathic pain. Two dual inhibitors of both ErbB1 and ErbB2, lapatinib (a reversible blocker targeting ErbB1ErbB2ErbB4 (Rusnak et al., 2001) at a dose of 20mg/kg ip, and afatinib (an irreversible covalent blocker targeting ErbB1ErbB2ErbB4 (Li et al., 2008)) at a dose of 2.5mg/kg ip, were both similarly effective, reversing hypersensitivity completely at peak effect and remaining statistically significant for 180 and 140min, respectively. The mean ± SEM percentage reversals of hypersensitivity from 20 to 140min following administration were 94.4 ± 6.9% (n=6) for lapatinib and 75.5 ± 9.0% (n=4) for afatinib, respectively (Fig. 3B). Animals treated with inhibitors of ErbB1, ErbB2 or combinations of these were closely observed and showed no apparent changes in locomotor activity, motor co-ordination or any evidence of sedation.
3.4. Evidence for ErbB1:ErbB2 interaction in CCI-induced hypersensitivityIn order to evaluate any direct interaction between ErbB1 and ErbB2 in small DRG cells, we used a Proximity Ligation Assay, which relies on hybridisation between DNA-tagged secondary antibodies to produce an amplified fluorescent signal if the epitopes recognised by two, species-distinct, primary antibodies are in close proximity (Fredriksson et al., 2002). Fig. 3C shows that ErbB1:ErbB2 proximity ligation in peripherin-positive DRG cells from naïve animals yielded a low fluorescent signal, which was strongly and significantly increased following CCI. Control experiments using rabbit ErbB1 and mouse ErbB3 or ErbB4 antibodies showed minimal proximity ligation signals from DRG cells of naïve or CCI rats. These findings provide explicit evidence for close interaction of ErbB1 and ErbB2 in nociceptive Cfibres following nerve injury. In the classical paradigm of ErbB receptor activation, ligand binding induces receptor dimerisation, which leads to activation of the intrinsic tyrosine kinase function of the monomers and autophosphorylation or transphosphorylation of intracellular tyrosine residues (Weiss and Schlessinger, 1998). ErbB2 lacks any known direct ligand, but readily heterodimerises with each of the other ErbB monomers (Graus-Porta et al., 1997, Tzahar et al., 1997), so may act as a subservient, signalling co-receptor in ErbB1:ErbB2 heterodimers. ErbB2 shows prominent auto−/trans-phosphorylation at Tyr1221/1222 upon activation (Ricci et al., 1995). Using a phospho-specific antibody that specifically recognises ErbB2 phosphorylated at Tyr1221/1222, we showed a marked increase in staining in small, ErbB1-positive DRG cells following CCI, which was reversed by erlotinib treatment (at a dose of 10mg/kg ip, 1 h) (Fig. 3D). This indicates that ErbB2 auto−/trans-phosphorylation occurs through an ErbB1-dependent mechanism following nerve injury; most likely representing the function of ErbB1:ErbB2 heterodimers. Given the accumulating evidence for ErbB1:ErbB2 co-engagement following nerve injury, we investigated whether ErbB1 and ErbB2 blockers might interact functionally in producing analgesia. We selected very low doses of erlotinib (0.33 mg/kg) and mubritinib (1mg/ kg) that produced just discernible analgesia, (only 9.9 ± 3.1% and 11.4 ± 3.7% reversal of mechanical hypersensitivity when averaged over 20–140min, n=4 in each case). Administration of these in combination produced clearly greater than additive analgesia (52.9 ± 11.1% reversal of hypersensitivity, p .01, n=4, with complete reversal at peak effect and statistically significant effects for up to 100min, Fig. 3E). Formal assessment of synergy using Bliss Additivism effect-based modelling (Berenbaum, 1981; Borisy et al., 2003), for combinations of erlotinib through the range 0.1–3.3mg/kg with mubritinib 1.0mg/kg, showed consistently greater observed responses than those predicted from combining individual effects, with more than 3.8-fold reduction in EC50 for observed compared to predicted combination dose-response curves (p .001 by Extra Sum of Squares F-test; Fig. 3F). Two further, structurally distinct selective ErbB1 and ErbB2 inhibitors, were administered separately at low dose and then in combination, to independently confirm the concept of synergistic analgesic Experiments were carried out in animals that displayed marked ipsilateral hypersensitivity to von Frey filaments following CCI carried out 8–12 days previously. Drugs were injected (in 0.2ml per animal) under light isoflurane anaesthesia and after a 20min delay for recovery, Paw Withdrawal Threshold (PWT) testing was carried out at 20min intervals up to 3 h post-injection. The percentage reversal of hypersensitivity (ipsilateral compared to contralateral pre-drug PWT) was meaned over 20–140min post-drug administration. A range of highly selective ErbB1 inhibitors showed robust reversal of hypersensitivity, displaying both concentration-dependence and efficacy from structurally diverse agents. ⁎⁎ indicates significant reversal of hypersensitivity (One-Way ANOVA with Dunnett's post-hoc test or Student's t-test, as appropriate). No significant changes in contralateral PWT were observed for any of the drugs or on either side following vehicle. effects from dual ErbB1/ErbB2-targeting. Falnidamol (0.33mg/kg) and tucatinib (1.0 mg/kg) individually produced 9.0 ± 1.6% and 13.8 ± 3.8% reversal of mechanical hypersensitivity, whereas in combination they produced 54.9 ± 6.1% reversal over 20–140min following administration, clearly much greater than the 21.6 ± 4.1% predicted from Bliss Additivism modelling (p .01 by unpaired t-test, n=4, Fig. 3G). These observations fully support the idea that ErbB1 and ErbB2 in C-fibre nociceptors co-operate in leading to nerve injuryinduced hypersensitivity and suggest that conjoint administration of agents to block both ErbB1 and ErbB2 may represent a favourable strategy for analgesia.
3.5. Evidence that peripheral terminals of nociceptors can be sensitised by an ErbB1-dependent process following CCIIn naïve rats, intradermal injection of EGF or heparin-binding EGF (another ErbB1-selective ligand) was reported to have no effect, whereas GDNF or NGF caused lasting hypersensitivity in a paw pressure withdrawal test (Andres et al., 2010; Ferrari et al., 2010). Nonetheless, as we had found ErbB1 abundantly expressed in the cell bodies of small, nociceptive DRG cells, some deployment to peripheral nerve terminals would be anticipated. To investigate whether peripherally localised ErbB1 might play a role in nerve injury-induced sensitisation of nociceptors, we carried out intraplantar injections in both naïve and CCI rats and assessed mechanical hypersensitivity by measuring von Frey PWT scores. Table 3 shows that intraplantar injection of EGF amplified the ipsilateral hypersensitivity in CCI animals but not in naïve controls. No changes were seen in contralateral PWT values and contralateral injection of EGF had no effect on either contralateral or ipsilateral PWT values. These findings suggest that the effect observed upon ipsilateral injection was a local, not systemically mediated, action and that some event brought about by nerve injury (perhaps local neuroinflammatory processes) was required to prime the nociceptors before an active role of ErbB1 was revealed. The effect of EGF was not mimicked by neuregulin1-β1 EGF domain (NRG1; a selective ErbB3/ErbB4 ligand, matching our evidence for minimal expression of these proteins in small nociceptive DRG cells. The EGF-evoked amplification of hypersensitivity was however prevented by intraplantar co-injection of erlotinib, confirming the anticipated targeting of ErbB1, and also by mubritinib, consistent with our other evidence for ErbB1 acting by way of ErbB1:ErbB2 heterodimers. Pilot experiments in CIPN mice showed similar marked amplification of hypersensitivity lasting from 20 to 80min (caption on next page) following intraplantar injection of EGF (80 ng), but no discernible effect in naïve controls.
3.6. Evidence for ErbB1 transactivation by intracellular signalling from other receptors following CCIErbB1 can act as a hub for transactivation by signals from diverse G protein-coupled receptors (GPCRs), chemokine-, and orphan-receptors (Wang, 2016; Kose, 2017). Gq-, Gi- and Gs-linked GPCRs can lead to ErbB1 transactivation by a mechanism involving Src and the recruitment of signalling adapter molecules including Grb2 and Gab1 (Daub et al., 1997; Maudsley et al., 2000; Drube et al., 2006; Qian et al., 2016). While activation of Gi-linked GPCRs, such as for lysophosphatidic acid, is reported to lead to Src-dependent tyrosine phosphorylation of ErbB1, ErbB2 phosphorylation was not directly increased (Luttrell et al., 1997). This is consistent with the idea of Srcdependent targeting of ErbB1, rather than ErbB2, representing the primary step in the transactivation of ErbB1:ErbB2 heterodimers by diverse upstream regulators. Src phosphorylation of ErbB1 specifically at Tyr845 leads to activation of ErbB1 signalling (Tice et al., 1999; Baumdick et al., 2018). Similarly, another Src family kinase, Yes, phosphorylates ErbB1 at sites including Tyr845 (but not Tyr1068), causing activation of downstream signalling, while the receptor is located intracellularly in endosomes (Su et al., 2010). Multiple candidate activators; inflammatory mediators, chemokines and cytokines are released during the local neuroinflammation following nerve injury (White et al., 2005; Uceyler et al., 2007; Sacerdote et al., 2008; Ji et al., 2014; Zhu et al., 2014). GPCRs for peptides (such as bradykinin), lipids (such as prostaglandins), proteases (such as thrombin) and chemokines (such as CCL2), as well as receptors for cytokines, such as interleukin1β and TNF-α, can all activate Src-family kinases (Viviani et al., 2003; van Vliet et al., 2005; Davis et al., 2006; Hardyman et al., 2013; Ghosh et al., 2016; Kose, 2017; Zhang et al., 2017; Yao et al., 2019); a signalling event reported to occur in primary afferents during neuropathic pain (Chen et al., 2014; Lam et al., 2018). To explore whether Src-dependent transactivation of ErbB1 was important in nerve injury-induced hypersensitivity, we firstly assessed the effect of selective inhibitors of Src on CCI-induced hypersensitivity in paw withdrawal thresholds from von Frey filaments. The potent Src inhibitor, dasatinib (O'Hare et al., 2005), at a dose of 3mg/kg ip, and the Src family-specific kinase inhibitor, A419259 (Wilson et al., 2002), Fig. 3. ErbB2 inhibitors reverse nerve injury-induced mechanical hypersensitivity and act synergistically with selective ErbB1 inhibitors, matching evidence for direct ErbB1:ErbB2 interaction in small nociceptive DRG neurons. A) and B) show effects of the highly selective ErbB2 inhibitors, mubritinib (15mg/kg, ip) and tucatinib (15mg/kg, ip) in A), and the dual ErbB1/ErbB2 inhibitors, lapatinib (20mg/kg, ip) and afatinib (2.5mg/kg, ip) in B), on mechanical hypersensitivity following CCI, as assessed by von Frey filament PWT values (n=4–6). Ipsilateral hypersensitivity was significantly reversed for up to 140–180min in each case; †− ††p .05–0.01 by One-Way Repeated Measures ANOVA with Dunnett's post-hoc test, while no significant effects were observed on contralateral responses. C) shows typical images from a Proximity Ligation Assay, with specific antibodies for ErbB1 (ab30) and ErbB2 (06–562). Close target proximity is reported as orange fluorescence. Sections were counterstained for peripherin (green) using conventional immunofluorescence. Nerve injury induced a clear ( 2-fold) increase in the percentage of peripherin-positive cells that showed ErbB1:ErbB2 proximity fluorescence compared to naïve. Typical numbers of positive cells counted per section were 117 for peripherin and 40 for ErbB1:ErbB2 proximity ligation, n=4 in each case. The statistical significance of changes was assessed by unpaired two-tailed ttest; ⁎⁎p .01, showing a significant CCI-induced increase in ErbB1:ErbB2 heterodimerisation compared to naïve. Scale bars represent 50 μm. D) shows typical images of immunofluorescence staining for a prominent auto−/trans-phosphorylation site in ErbB2, phospho-ErbB2 (Tyr1221/2), identified by a specific rabbit monoclonal antibody [6B12], compared to ErbB1, identified using ab30, in DRG from naïve, CCI and CCI animals treated with erlotinib, 10mg/kg ip, 1 h. The percentage of ErbB1-positive cells showing phospho-ErbB2 (Tyr1221/2) staining was more than 2.5-fold increased following CCI and this was reversed by erlotinib. One-Way ANOVA with Tukey's post-hoc test revealed a significant increase due to CCI (⁎⁎⁎p .01) and its significant reversal following erlotinib treatment (††p .01), n=4 in each case. Numbers of positive cells counted per section were typically 189 for ErbB1 and 53 for phospho-ErbB2 (Tyr1221/2). Scale bars represent 50 μm. E) shows effects of selective ErbB1 and ErbB2 inhibitors, in combination, on mechanical hypersensitivity following CCI, as assessed by von Frey PWT scores. Low doses of erlotinib (0.33mg/kg, ip) and mubritinib (1.0mg/kg, ip), selected to produce just discernible levels of analgesia alone, were tested in combination. One-Way Repeated Measures ANOVA with Dunnett's post-hoc test indicated statistically significant attenuation due to erlotinib at 20 and 40min following administration, for mubritinib at 20min and for the combination throughout the 20–100min period; ††p .01; n=4 in each case. F) shows formal assessment of analgesic synergy between mubritinib (1.0 mg/kg, ip) and erlotinib across a range of erlotinib doses (ip), using Bliss Additivism effect-based modelling to predict expected combination outcomes. Comparison of observed versus expected combination dose-response curves by Extra Sum of Squares F test indicated a significant difference, reflected in a more than 3.8-fold reduction in EC50 value (mean [95% CI] from 1.09 [1.06/1.14]) to 0.28 [0.25/0.31]mg/kg, †††p .001 (n=4–5 for each point). G) shows the effects of individual and combined treatment with further (structurally distinct) selective ErbB1 and ErbB2 blockers, falnidamol (0.33mg/kg, ip) and tucatinib (1.0 mg/kg, ip), respectively, against nerve injury-induced hypersensitivity. The observed analgesic effect was significantly greater than the predicted combination effect according to Bliss Additivism modelling (p .001, Student's t-test, n=4). Experiments were carried out 8–12 days following CCI in animals that displayed marked ipsilateral hypersensitivity to von Frey filaments. Drugs were injected intraplantarly into the interdigital skin (in 50 μl per animal) under light isoflurane anaesthesia, and Paw Withdrawal Threshold (PWT) testing recommenced after a 20min delay for recovery and then repeated at 10min intervals up to 70min post-injection. The mean percentage reduction in PWT score over 20–40min post-drug administration was calculated and the statistical significance of changes in ipsilateral or contralateral PWT scores was assessed by One-Way ANOVA with Dunnett's post-hoc test. Injection of EGF (100 ng) ipsilateral to CCI caused significant exacerbation of injury-induced hypersensitivity (⁎⁎p .01) from 20 to 40min following administration. This effect was reversed by co-injection of the ErbB1 inhibitor, erlotinib (2.2 ng) or the ErbB2 inhibitor, mubritinib (7.0 ng), which had no discernible effects alone. The ErbB3/4-selective agonist, neuregulin1-β1 EGF domain (NRG1, (100 ng) did not mimic the effect of EGF. Contralateral injection of EGF or NRG1 had no effect on ipsilateral PWT values in CCI animals. Either ipsilateral or contralateral injection of EGF had no effect on contralateral PWT values in CCI animals and was also without effect on baseline PWT values in naïve controls. Administration of vehicle alone had no discernible effect. at a dose of 15mg/kg ip, both caused clear reversal of CCI-induced hypersensitivity, with complete reversal of hypersensitivity at peak and statistically significant reversal for 60min and 50min in each case (Fig. 4A). The amplification of mechanical hypersensitivity caused by intraplantar injection of 100 ng EGF ipsilateral to CCI (48.1 ± 14.0% reduction in PWT, Table 3) was also prevented by local co-administration of 0.73 ng dasatinib (2.8 ± 13.4% reduction in PWT, n=4). These observations strongly support the idea of a key role for Src throughout primary nociceptive neurons in bringing about neuropathic hypersensitivity (and are consistent with previous reports). In future studies, outwith the scope of current work, it would be of interest to assess the possibility of synergistic analgesic effects arising from combined delivery of Src- and ErbB1 inhibitors. As Src activity is normally restrained due to constitutive phosphorylation at Tyr527 by Csk and disinhibited by intracellular signals leading to dephosphorylation at this locus (Hunter, 1987), we investigated the status of phosphorylation at this site, using an antibody specific for phospho-Src (Tyr527). We showed significantly decreased staining in small, ErbB1-positive DRG cells following CCI, which would be predicted to reflect increased Src activity (Fig. 4B). This was unaffected by erlotinib, exactly as would be expected for a signalling process upstream of ErbB1, that could drive its transactivation or priming sensitisation by neuroinflammatory signals. We further investigated phosphorylation of ErbB1 at the Src-target site, Tyr845. Fig. 4C shows that CCI induced a marked increase in staining with a phospho-ErbB1 (Tyr845)-specific antibody in small, peripherin-positive DRG cells. This was blocked, as expected, by dasatinib treatment, but was unaltered by erlotinib, consistent with this step representing an upstream regulation of ErbB1 that is independent of ErbB1's own catalytic activity.
3.7. Downstream signalling by ErbB1 activated following CCIFollowing ErbB1 activation, Tyr1068 is among the most prominent receptor autophosphorylation/transphosphorylation sites, which forms the major recruitment site for the adapter proteins, Grb2 and Gab1; leading to signalling through ERK MAP kinase and PI 3-kinase/Akt (Rojas et al., 1996; Olayioye et al., 2000; Rodrigues et al., 2000; Nishida and Hirano, 2003; Mattoon et al., 2004). Fig. 5A shows that ErbB1 phosphorylation at Tyr1068 was clearly increased following CCI and this was reversed by treatment with erlotinib (10mg/kg, for 1 h), suggesting that Tyr1068 phosphorylation was maintained by an active, rapidly turning-over process following nerve injury. The potent Src inhibitor, dasatinib (3mg/kg) was similarly effective, consistent with an upstream role of Src in bringing about ErbB1 autophosphorylation (Fig. 5A). Fig. 5B displays the subcellular distribution of phospho-ErbB1 (Tyr1068) immunoreactivity in DRG neurons from naïve, CCI and erlotinib-treated CCI animals. Both high power images of individual cells and fluorescence intensity profiles of transects from the nuclear perimeter to the plasma membrane at the cell apex, show that phosphoErbB1 (Tyr1068) was clearly increased following CCI and this increment was reversed by erlotinib treatment; n=8 in each case. The increase in ErbB1 phosphorylation at Tyr1068 following nerve injury was however, predominantly at an intracellular (perinuclear) site, as opposed to plasma membrane; consistent with its induction by an intracellular signalling process rather than an external ligand. The close molecular engagement of Gab1 with ErbB1 was explicitly demonstrated by Proximity Ligation Assay, which shows ErbB1:Gab1 proximal interaction was significantly increased in peripherin-positive DRG cells following CCI and that this increment was reversed by in vivo treatment with erlotinib or mubritinib (Fig. 5C), consistent with roles of the tyrosine kinase function of both ErbB1 and ErbB2 in Gab1 recruitment. The subcellular distribution of Gab1 immunoreactivity itself was not studied in detail. Trials for antibody specificity were only carried out with naïve tissue and examined at low power, where staining appeared to be distributed throughout the cytoplasm. Whether this distribution might be altered in CCI tissue was not assessed. Two downstream signalling cascades are activated following ErbB1 phosphorylation at Tyr1068; the PI 3-kinase/Akt and ERK MAP kinase pathways (Olayioye et al., 2000). Akt is activated as a consequence of PIP3 generation by PI 3-kinase, which may occur as a result of its Gab1dependent recruitment to phospho-ErbB1 (Tyr1068). The activation state of Akt can be monitored by its obligatory phosphorylation at Ser473 (Sarbassov et al., 2005). Fig. 6A shows that phospho-Akt (Ser473) staining in small, ErbB1-positive DRG cells was significantly increased following CCI and that this increment was fully reversed by erlotinib (10mg/kg ip, 1 h), the highly selective Akt inhibitor, ipatasertib (GDC-0068) (Lin et al., 2013) (20mg/kg ip, 1 h) or the Src inhibitor, dasatinib (3mg/kg ip, 1 h). Fig. 6B shows that two selective Akt inhibitors of distinct structure, ipatasertib and afuresertib, both reversed CCI-induced mechanical hypersensitivity in the von Frey paw withdrawal test, at doses providing effective target coverage in vivo (20mg/kg ip and 3.3mg/kg ip, respectively) (Lin et al., 2013; Dumble et al., 2014). Both agents caused significant reversal (p .01) of hypersensitivity for at least 160min following administration, with mean ± SEM percentage reversal of hypersensitivity from 20 to 140min of 87.6 ± 6.8%, n=6 and 75.5 ± 8.5%, n=5, respectively. These observations indicate that Akt is activated in ErbB1-expressing small DRG cells following CCI and that this is prevented by erlotinib blockade of ErbB1 signalling or dasatinib inhibition of the upstream ErbB1 regulator, Src. Highly selective Akt inhibitors also cause marked and sustained reversal of CCI-induced mechanical hypersensitivity. These findings suggest that activation of Akt by ErbB1 plays a key role in hypersensitivity following nerve injury. It is notable that the functional lifetime of ErbB1 and its co-immunoprecipitation with Gab1 and the p85 subunit of PI 3-kinase are reported to be increased by the coexpression of ErbB2 (Hartman et al., 2013). Whether co-administration of ErbB1/ErbB2 inhibitors with blockers of Akt might lead to further synergistic enhancement of analgesia would be interesting to explore in follow-on studies. ERK MAP kinase is also activated upon ErbB1 autophosphorylation at Tyr1068, through recruitment of the adapter Grb2, or potentially indirectly through a Gab1/SHP-2 complex (Nishida and Hirano, 2003), to trigger a kinase cascade leading to MEK and ERK. Activation of ERK was monitored by its obligatory phosphorylation by MEK at Thr202 and Tyr204 (Robinson and Cobb, 1997). While the majority of DRG cells positive for phospho-ERK (Thr202/Tyr204) were also ErbB1-positive, a sizeable fraction of this ERK phosphorylation occurred in cells that did not stain for ErbB1 (25.2 ± 5.3%, n=4), and this proportion was unaltered following CCI, or CCI plus erlotinib treatment. Correspondingly, highly selective blockers of ERK MAP kinase, (ravoxertinib, GDC0994) (Blake et al., 2016) or its upstream regulator, MEK, (AZD-8330) (Wallace et al., 2005), produced only modest reversal of CCI-induced mechanical hypersensitivity. The mean ± SEM values for percentage reversal of CCI-induced hypersensitivity in the von Frey test were 9.7 ± 2.6% (n=4) and 35.8 ± 4.6% (n=11), over 20–140min following administration, for 30mg/kg ravoxertinib and for 1.25mg/kg AZD-8330 (doses that have been documented to provide effective target coverage in vivo). In comparison with the Akt inhibitor results, these observations suggest a relatively minor involvement of ERK in driving the hypersensitivity measured here. This corresponds to reports of rather transient ERK activation in DRG and dorsal horn neurons, following nerve injury (Obata et al., 2004; Zhuang et al., 2005; Ji et al., 2009). While EGF treatment of cultured DRG neurons does cause activating phosphorylation of Shc (an intermediary adapter) for the ERK MAP kinase pathway (Ganju et al., 1998), ERK itself is reported to be only minimally activated (Andres et al., 2010). 3.8. ErbB1-dependent activation of Akt targets downstream vesicular trafficking Among the range of Akt substrates discovered in an unbiased search using an antibody raised against the consensus Akt phosphorylation motif (Kane et al., 2002) was AS160 (TBC1D4). AS160 is now known to act as a Rab-GAP and participate (upon its phosphorylation and inhibition by Akt at Thr642) in insulin-induced translocation of GLUT4 storage vesicles in adipocytes (Sano et al., 2003). We hypothesised that following nerve injury, ErbB1 activation of Akt might similarly regulate trafficking of targets relevant to neuronal excitability in sensory neurons. Fig. 7A shows that the percentage of ErbB1-positive DRG cells staining for phospho-AS160 (Thr642) was markedly increased following CCI and that this increment was reversed by treatment for 1 h with either erlotinib (10mg/kg ip) or ipatasertib (20mg/kg ip), n=4 in each case; fully consistent with our hypothesis. Levels of pan-AS160 staining in ErbB1-positive DRG cells were unaltered following CCI (69.7 ± 3.6% in naïve, and 71.6 ± 2.4% in CCI, n=4 in each case). Even in low power images of phospho-AS160 (Thr642) staining, it was obvious that the marked CCI-induced increase was intracellularly localised in a prominent perinuclear ring of fluorescence. High power images and transect fluorescence intensity analysis of phospho-AS160 (Thr642) staining in individual ErbB1-positive DRG cells confirmed this, with a marked (3-fold) increase in maximal intensity following CCI that occurred entirely within 2 μm of the nuclear perimeter (Fig. 7B, n=8 in each case). The CCI-induced increment was reversed by erlotinib, mubritinib (15mg/kg ip, 1 h) or ipatasertib treatment, confirming the involvement of ErbB1, ErbB2 and Akt in this response. AS160 and its Akt-phosphorylated form have been identified at both intracellular and plasma membrane locations in various cell types (Larance et al., 2005; Ng et al., 2010; Zheng and Cartee, 2016), although here, in nociceptive afferent neurons, as in muscle for example (Zheng and Cartee, 2016), the location is almost entirely intracellular. AS160 has been shown to interact directly with LRP1, a chaperone/ cargo protein in GLUT4 storage vesicles, that is essential for GLUT4 trafficking to the plasma membrane (Jedrychowski et al., 2010; Brewer et al., 2014). Interestingly, LRP1 has been identified as a key factor in the trafficking of multiple cargoes to the plasma membrane, including β1 integrin and voltage-sensitive ion channels (Salicioni et al., 2004; Lillis et al., 2008; Parkyn et al., 2008; Kadurin et al., 2017; Au et al., 2018). As such a role could potentially play a part in increased nociceptor excitability following nerve injury, we investigated the subcellular distribution of LRP1 immunoreactivity in DRG cells after CCI. Fig. 7C shows example high power images and transect fluorescence intensity analysis of LRP1 staining in individual ErbB1-positive DRG cells. LRP1-immunoreactivity was found in small clusters, consistent with a vesicular localisation, which in naïve animals was limited to the inner, perinuclear half of the transect range. In CCI animals, a significant increase in LRP1 staining (with greater clustering) was observed in the outer portion of the transect range and this increment was reversed by treatment with erlotinib or ipatasertib, again defining ErbB1 and Akt involvement in the response (n=9 in each case).
3.9. CCI-induced ion channel trafficking in nociceptors through ErbB1 activation of AktNumerous ion channels could contribute to altered excitability and action potential generation in nociceptors if they show increased localisation at the plasma membrane following injury. The voltage-sensitive Na+ channels, Nav1.7/1.8/1.9, are selectively expressed in nociceptors and are thought to play key roles in threshold setting, and in the generation and repetitive firing of action potentials (Dib-Hajj et al., 2015; Dib-Hajj et al., 2017; Hoffmann et al., 2017). Molecular genetics of human hereditary channelopathies with altered pain sensation has directly implicated each of these channels in pain processing, including hypersensitive neuropathic pain states (Dib-Hajj et al., 2017; Huang et al., 2017). They each show a predominantly intracellular location under basal conditions, suggesting that their trafficking to the plasma membrane could well lead to increased excitability (Amaya et al., 2000; Okuse et al., 2002; Bao, 2015). Analysis of Nav1.7/1.8/1.9 expression using specific antibodies showed very high proportions of expression in ErbB1-positive small DRG cells (97.1 ± 5.8%, 71.6 ± 6.3% and 90.3 ± 6.6%, respectively, n=4 in each case). While CCI caused no discernible change in Nav1.7 or Nav1.8 expression levels, Nav1.9 expression and subcellular deployment were significantly altered. We observed a slight reduction in the number of ErbB1-positive DRG cells co-expressing Nav1.9 to 75.3 ± 7.9% of ErbB1-positive cells (p .05, t-test), but a marked redeployment of the channel towards the plasma membrane (Fig. 8A). A partial reduction in expression of Nav1.9 following sciatic, but not trigeminal nerve i njury has been reported previously (Dib-Hajj et al., 1999; Luiz et al., 2015), while expression is increased in a bone cancer model, which likely incorporates a nerve injury component (Qiu et al., 2012). Nerve injury-induced trafficking of Nav1.9, which has a long-lasting impact on nociceptor firing threshold (Hoffmann et al., 2017), has not previously been reported. Fig. 8A shows high power images of individual ErbB1-positive DRG cells and transect fluorescence intensity analysis indicating marked recruitment of Nav1.9 to sites adjacent to the apical plasma membrane (transect bins 43–49 out of 50) following CCI and its efficient reversal by treatment with erlotinib or ipatasertib (n=8 in each case). It was not possible to readily assess whether Nav1.9 blockade would reverse CCI-induced pain hypersensitivity in the current experiments, as no highly selective small molecule Nav1.9 inhibitors are available (Dib-Hajj et al., 2015; Lin et al., 2016). Similar (but less marked) results were observed with Nav1.8, indicating significant CCI-induced recruitment to transect bins 46–50 and reversal by erlotinib treatment (n=9 in each case), while no significant changes were detected in Nav1.7 distribution (Supplementary Fig. 1). Fig. 8B shows that Proximity Ligation Assay using specific Nav1.9 and LRP1 antibodies, identified a close molecular interaction of these two proteins in peripherin-positive DRG cells, which was significantly increased following CCI, and reversed by treatment with erlotinib. This fully supports the idea of LRP1 driving Nav1.9 translocation to the plasma membrane following CCI, through an ErbB1, Akt and AS160-dependent process. Several voltage-sensitive Ca2+ channel types are also expressed in DRG neurons, particularly L-type, generally involving Cav1.2 or Cav1.3 α1-subunits, and also N-type, involving Cav2.2, which between them contribute the majority of voltage-dependent Ca2+ current (Heinke fluorescence were seen in a broadly perinuclear location. Nerve injury induced the significant appearance of LRP1 clusters in more distal parts of the transect and this change was reversed by treatment with selective ErbB1 or Akt inhibitors. et al., 2004; Woodall et al., 2008; Bourinet et al., 2014). Cav1.2 and Cav2.2 have both been shown to undergo PI 3-kinase/Akt-dependent trafficking to the plasma membrane of transfected fibroblasts and DRG neurons (Viard et al., 2004), so could potentially play a part in the ErbB1-dependent hypersensitivity identified following nerve injury. We focused here on Cav1.2, as intrathecal administration of L-type Ca2+ channel blockers (affecting DRG as well as spinal cord (Cao et al., 2016)) produces analgesia in chronic pain states (Vanegas and Schaible, 2000; Yaksh, 2006), while knockdown of Cav1.2 specifically reverses neuropathy-associated mechanical hypersensitivity (Fossat et al., 2010). In contrast, Cav1.3 knockout mice lack any pain phenotype (Clark et al., 2003). Cav1.2 is also of particular interest in relation to nociceptor excitability due to its subcellular localisation, not only at the plasma membrane and intracellularly, but also in the Axon Initial Segment (AIS), where it may play a role in excitability setting (Brandao et al., 2012). A large proportion of Cav1.2-positive cells were also ErbB1-positive, and this was unaltered following nerve injury (74.1 ± 3.7% naïve, and 66.5 ± 3.9% CCI, n=4 in each case). The subcellular distribution of Cav1.2 however, was strikingly altered following nerve injury. Fig. 8C shows high power images of individual cells and fluorescence intensity analysis of cell transects. This indicates remarkable trafficking of Cav1.2 to the apical plasma membrane and proximal stem axon following CCI, which was reversed by treatment with erlotinib or ipatasertib (n=8 in each case). Nerve injury-induced translocation of Cav1.2 to such sites in nociceptors could contribute importantly to increased excitability. The reversal of this response by erlotinib and ipatasertib indicates its dependence on the ErbB1/Akt pathway we have elucidated above.
4. DiscussionOur results provide powerful evidence that highly selective small molecule inhibitors of the tyrosine kinase function of ErbB1 (TKIs) reverse hypersensitivity in rodent neuropathic pain models. This represents robust corroboration of previous clinical findings with both ErbB1 antibody reagents, such as cetuximab or panitumumab, and small molecule ErbB1-selective TKIs, such as erlotinib and gefitinib (Kersten and Cameron, 2012; Kersten et al., 2013; Kersten et al., 2015). It was not possible to test cetuximab or panitumumab here due to their humanised/human immune origins. Our results using the CCI model of neuropathic mechanical pain hypersensitivity in rats (von Frey paw withdrawal threshold) gave a mean IC50 value of 1.4 mg/kg ip for erlotinib, similar estimated potencies for gefitinib and AG 1478, and robust analgesic effects of several structurally distinct ErbB1-selective TKIs (Table 2). These observations were reproduced in the mouse CIPN model, which showed mean IC50 values of 3.2, 2.4 and 3.0mg/kg ip for erlotinib, gefitinib and AG 1478, respectively. The effective analgesic doses here are in the range of approved human dosing. Similar analgesic effects of small molecule ErbB1-selective TKIs have been reported using various inflammatory and neuropathic pain models in mice and the CCD, chronic DRG compression, model in rats (Martin et al., 2017; Wang et al., 2019). In many cases though, the mean IC50 values reported were notably higher, both in inflammatory pain models (formalin phase-2 and Complete Freund's Adjuvant), and particularly in neuropathic pain models (Spared Nerve Injury and CCI), where IC50s for gefitinib and AG 1478 ranged from 77 to around 300mg/kg (Martin et al., 2017). The basis for this disparity is unclear, although the analgesic signal is fully corroborated. The present study provides substantive evidence for the highly selective expression of ErbB1 in small unmyelinated nociceptors. In the CCI model here, ErbB1 expression is unaltered following nerve injury, in contrast to findings from the CCD (DRG compression) model (Wang et al., 2019). We found a high degree of co-localisation of ErbB1 with ErbB2, but not ErbB3 or ErbB4, in small unmyelinated nociceptors. Our study further delivers fundamental new insights by showing that ErbB1 is likely to act as a heterodimer with ErbB2 in driving neuropathic pain hypersensitivity. The key evidence arises from converging experimental approaches; proximity ligation showing direct ErbB1:ErbB2 interaction in DRG, analgesia induced by either ErbB1 and ErbB2 inhibitors (both systemically and peripherally), crossed blockade of an activation marker by inhibiting the partner receptor, and importantly, definitive synergism in the analgesia produced by co-administration of ErbB1 and ErbB2 inhibitors. This synergistic co-operation of ErbB1 and ErbB2 inhibitors in delivering analgesic efficacy against neuropathic pain identifies clear translational potential; highlighting a strategy for enhanced pain relief with lesser side effects than might be achieved by an EGFR inhibitor alone. Our evidence that Src could well play a key role in triggering proexcitatory ErbB1 processes in nociceptive afferents is consistent with reports of Src activation in DRG (including in TRPV1-positive afferents) in both neuropathic and inflammatory pain models (Alessandri-Haber et al., 2004; Jin et al., 2004; Zhang et al., 2005; Liu et al., 2015). Src is readily activated in DRG by cytokines (IL-1β and CXCL12) (Igwe, 2003; Rivat et al., 2014), which would support the hypothesis that ErbB1 activation in CCI here may result from intracellular transactivation, potentially as a result of nerve injury-induced neuroinflammation. Src phosphorylation of ErbB1 at Tyr845 leads to facilitated or direct activation of ErbB1 signalling (Tice et al., 1999; Baumdick et al., 2018), thereby obviating the need for increased availability of an extracellular EGF-like ligand. Nevertheless, genetic association of a SNP in the gene locus for epiregulin (an ErbB1 and ErbB4 receptor ligand) in patient cohorts with chronic temporomandibular disorder (TMD) pain has been identified (Martin et al., 2017), although acute pain seems to be inversely correlated (Verma et al., 2020). Both inflammatory and neuropathic pain models in mice result in increased blood levels of epiregulin and its intrathecal administration produces modest hypersensitivity in baseline thermal and mechanical responses as well as a dose-dependent increase in licking/biting behaviour in the second phase of the formalin test (Martin et al., 2017). However, the facilitation of formalin responses was not replicated by any of 4 other ErbB1 agonist-ligands, despite the binding affinity of epiregulin to ErbB family receptors (including ErbB1 homodimers and ErbB1:ErbB2 heterodimers) being lower than that of all other EGF family ligands (Toyoda et al., 1995; Shelly et al., 1998; Jones et al., 1999; Sato et al., 2003; Freed et al., 2017). The application of epiregulin to dorsal nerve roots in rats showed disparate effects on the spontaneous or C-fibre-evoked activity of dorsal horn neurons (Kongstorp et al., 2019), matching the dichotomous molecular genetics findings (Martin et al., 2017; Verma et al., 2020). Similarly, while systemic infusion of an epiregulin-neutralising antibody speeded the recovery from hypersensitivity in a nerve injury model, this had mixed effects in an inflammatory model and exacerbated capsaicin-evoked pain responses (Verma et al., 2020). With respect to the specific role of ErbB1, although attenuation of epiregulin facilitation of phase-2 formalin responses by the ErbB1-selective TKI, AG 1478, was reported (Martin et al., 2017), the fact that this itself reduces formalin responses, complicates interpretation. Similarly, although epiregulin facilitation of formalin responses was absent in (heterozygous) mutant mice with a large deletion of the ErbB1 extracellular domain, the mutants showed significantly altered basal formalin responses. Overall, while systemic epiregulin levels increase in chronic pain models and may indeed contribute to hypersensitivity, there is limited explicit evidence that this relates specifically to ErbB1 in primary nociceptive afferents. To explore the role of ErbB1 in the peripheral terminals of nociceptors, we carried out intraplantar injections of EGF and assessed mechanical paw withdrawal thresholds (Table 3). EGF showed no discernible effect in naïve animals, in agreement with previous observations (Andres et al., 2010; Ferrari et al., 2010; Araldi et al., 2018), but a clear amplification of ipsilateral hypersensitivity after nerve injury. This suggests that ErbB1 plays a role in nociceptive processing only in chronic injury-induced hypersensitive states. The EGF response was specific to the injured limb, reversed by local erlotinib, mubritinib or dasatinib (indicating ErbB1, ErbB2 and Src involvement) and was not mimicked by neuregulin-1-β1 EGF domain (indicating lack of ErbB3/4 involvement). Interestingly, the prolongation of intradermal PGE2-induced mechanical hypersensitivity caused by repeated administration of μ-opioid receptor agonist, is reported to be attenuated by inhibitors of both ErbB1 and Src (Araldi et al., 2018), emphasizing a key role of ErbB1 in GPCR-induced pain hypersensitivity. In that model, intradermal EGF did not on its own affect pain responses, as described here. Responses to PGE2 were unaltered 5 days after EGF administration, which would align with our observations that concurrent nerve injury is required to reveal a sensitising effect of ErbB1 activation and support the hypothesis that intracellular signalling plays a key role in this connection. A number of reports have described the activation of Akt in DRG neurons in chronic pain hypersensitivity induced by nerve injury (Xu et al., 2007; Shi et al., 2009), chemotherapeutic-evoked neuropathy (Jiang et al., 2016; Li et al., 2016), inflammation (Liang et al., 2013), bone cancer (Guan et al., 2015) and intradermal injection of capsaicin, ephrin or formalin (Sun et al., 2007; Guan et al., 2010; Martin et al., 2017). In many cases, administration of PI 3-kinase or Akt inhibitors attenuated pain hypersensitivity, matching our observations here with the second generation highly selective Akt inhibitors, ipatasertib and afuresertib. Various reports have also outlined changes in mTOR and protein translational machinery (significant downstream targets of Akt) in neuropathic, inflammatory and formalin-induced pain models, with analgesic effects seen due to blockers of different elements (Xu et al., 2010; Obara et al., 2011; Liang et al., 2013; Khoutorsky et al., 2015; Martin et al., 2017). Epiregulin-induced enhancement of formalin-induced nocifensive behaviour was attenuated by inhibitors of PI 3-kinase (but not MEK) and accompanied by increased Akt phosphorylation in DRG (Martin et al., 2017), potentially relating to ErbB1 activation. Epiregulin-induced pain hypersensitivity was also diminished by inhibitors of mTOR and in mutant mice with disrupted elements of the protein translational machinery, indicating a clear functional role. In the current study we investigated whether alternative rapid Akt-dependent processes such as ion channel trafficking may be key to the role of ErbB1 in neuropathic pain and help explain the rapid analgesic effects of ErbB1 inhibitors. Akt has important cellular roles in the regulation of vesicular trafficking, for example in insulin-induced rapid translocation of the GLUT4 glucose transporter to the plasma membrane of adipocytes (Hill et al., 1999). The substrates for Akt phosphorylation include AS160 (TBC1D4), a Rab-GAP (GTPase-activating protein), which signals GLUT4 translocation by terminating vesicle retention upon its inactivating phosphorylation by Akt (Kane et al., 2002; Sano et al., 2003; Fujita et al., 2010). AS160 has a predominantly peri-nuclear subcellular localisation, associated with GLUT4 storage vesicles, from which it dissociates following insulin stimulation (Larance et al., 2005). Aktphosphorylated AS160 (at Thr642) has been identified at both intracellular and plasma membrane sites after insulin stimulation of adipocytes (Ng et al., 2010), although in muscle its localisation is entirely intracellular (Zheng and Cartee, 2016), as we found here associated with ErbB1-dependant nociceptive hypersensitivity following nerve injury. An AS160 paralogue, TBC1D1, and AS250 (a Ral-GAP) may also contribute to Akt regulation of GLUT4 trafficking (Sakamoto and Holman, 2008; Peck et al., 2009; Chen et al., 2011; Leto and Saltiel, 2012). We further showed that the AS160-interacting protein and vesicular chaperone/co-cargo, LRP1 (Jedrychowski et al., 2010), which participates in the plasma membrane trafficking of various surface-deployed proteins, such as Cav2.2 (Kadurin et al., 2017), undergoes clustering and localisation closer to the plasma membrane following CCI. Crucially these changes were prevented by selective blockers of ErbB1 or Akt. In addition, we demonstrated increased ErbB1- and Akt-dependent trafficking of Nav1.9 and Nav1.8 (but not Nav1.7) to DRG somata plasma membrane following CCI (Fig. 8A, Supplementary Fig. 1), a process that potentially could also occur in peripheral nociceptor endings. This matches human molecular genetics evidence associating these channels with hereditary painful neuropathies, and their disruption with insensitivity to pain (Dib-Hajj et al., 2017; Bennett et al., 2019). The electrophysiological characteristics of Nav1.8 point to a role in rapid re-priming and maintenance of repetitive firing (Dib-Hajj et al., 2017, Bennett et al., 2019). The activation of Nav1.9 by weak stimuli at hyperpolarised voltages and very delayed time-course of inactivation are consistent with a role in threshold setting of excitability (Dib-Hajj et al., 2017, Bennett et al., 2019). Many, but not all, knockdown, knockout and pharmacological studies indicate a role for Nav1.8 in both inflammatory and neuropathic pain (Porreca et al., 1999; Kerr et al., 2001; Lai et al., 2003; Nassar et al., 2005; Joshi et al., 2006; Dong et al., 2007). Differences in findings may relate to the characteristics of particular mutant mouse lines and compensatory changes (Leo et al., 2010). Although no truly selective pharmacological selective agents are yet available (Dib-Hajj et al., 2015; Lin et al., 2016), knockout/deletion studies have implicated Nav1.9 in inflammatory pain, but also in neuropathic and basal pain (Porreca et al., 1999; Priest et al., 2005; Amaya et al., 2006; Maingret et al., 2008; Leo et al., 2010; Hockley et al., 2014; Osorio et al., 2014; Lolignier et al., 2015; Luiz et al., 2015; Hoffmann et al., 2017). Very little is known however, of the processes involved in their deployment at the plasma membrane of DRG neurons. There is evidence that contactin and FHF1B interact directly with Nav1.9 and could play a role (Liu et al., 2001a, 2001b), although FHF1B and contactin are also implicated in the function/trafficking of Nav1.5/1.3 (Liu et al., 2003; Shah et al., 2004). The heterologous expression of functional Nav1.9 has proved difficult to achieve and is likely to require Navβ subunits as well as probably additional, unknown factors (Goral et al., 2015; Lin et al., 2016). Similarly, little is known of Nav1.8 trafficking and its control, although interaction with annexin light chain p11 has been reported as a key factor (Okuse et al., 2002; Foulkes et al., 2006). Annexin light chain p11 is also implicated in the plasma membrane trafficking of epithelial Na+ (ENaC) channels and related ASIC1a channels (Donier et al., 2005; Cheung et al., 2019), both of which can be driven by PI 3-kinase/Akt signalling (Markadieu et al., 2004; Duan et al., 2012). The question of why Nav1.9 deployment at the plasma membrane is relatively low under basal conditions was addressed in a recent study that compared key trafficking motifs in Nav1.9 with those in Nav1.7 (Sizova et al., 2020). Evidence for distinct trafficking signals between Nav1.9 and Nav1.7 fits well with our observations that the trafficking of Nav1.9 (and Nav1.8), but not Nav1.7 to DRG cell plasma membrane can be upregulated by an ErbB1-dependent mechanism in neuropathic pain. The ErbB1-dependent trafficking of Nav1.9 (and Nav1.8) to key locations for neuronal excitability may be crucial to neuropathic hypersensitivity. We further demonstrated striking, ErbB1- and Akt-dependent trafficking of Cav1.2 to the plasma membrane and axon initial segment/ proximal axon of small peripherin-positive DRG cells following CCI (Fig. 8C). This matches evidence for the preferential expression of Cav1.2 (together with known protein partners) in the soma, plasma membrane and axon initial segment/proximal axon of small, peripherin-positive C-fibre afferents (Brandao et al., 2012) and may, together with the well documented clustering of Nav channels in this region (Zhou et al., 1998), contribute to increased excitability. Indeed, selective Cav1.2 blockers or Cav1.2 knockdown produce analgesia in a number of chronic pain models (Vanegas and Schaible, 2000; Yaksh, 2006; Fossat et al., 2010) (Vanegas and Schaible, 2000; Yaksh, 2006; Fossat et al., 2010). Furthermore, numerous reports describe growth factor-induced trafficking of channels (including TRP, KCa and both Land N-type Ca2+ channels) to the plasma membrane in various cell types including neurons and the crucial role of PI 3-kinase/Akt signalling in this (Blair and Marshall, 1997; Kanzaki et al., 1999; Bezzerides et al., 2004; Viard et al., 2004; Chae et al., 2005). In addition to the Akt-dependent targeting of the Rab-GAP AS160 and disinhibition of LRP1-associated trafficking identified here (as also identified in GLUT4 translocation (Leto and Saltiel, 2012), trafficking of Cav1.2 and 2.2 has been shown to depend on both β and α2δ channel subunits, the latter of which interact directly with LRP1 (Viard et al., 2004; Hoppa et al., 2012; Kadurin et al., 2017; Nieto-Rostro et al., 2018). Nerve injury-induced trafficking of Cav1.2 to functionally important subcellular locations in DRG neurons, including the plasma membrane, proximal axon and potentially, even synaptic release sites (Hoppa et al., 2012) could play a crucial part in nociceptive hypersensitivity. Other channelinteracting proteins, such as CRMP2, are reported to modulate trafficking of channels, including Nav1.7 and Cav2.2 (Brittain et al., 2009; Chi et al., 2009; Chew and Khanna, 2018). The pseudo-polar morphology of DRG neurons is thought to be responsible for an increased probability of action potential failure at the bifurcation, especially for high frequency firing (Luscher et al., 1996; Nascimento et al., 2018). Peripheral nerve injury leads to a reduction in this low-pass filtering and may thereby contribute to neuropathic pain hypersensitivity (Gemes et al., 2013). Ion channel recruitment to a potential axon initial segment-like zone in the proximal stem axon and soma plasma membrane, particularly in small diameter DRG neurons, may thus exert an important influence over spike propagation (Nascimento et al., 2018). Both Na+ and Ca2+ currents contribute to action potentials in DRG cells (Kostyuk et al., 1981; Heyer and Macdonald, 1982) and intracellular Ca2+ levels may additionally facilitate spike propagation (Luscher et al., 1996), so the nerve injuryassociated translocation of Na+ and Ca2+ channels here could potentially play a part in enhanced action potential generation along the axon of nociceptors. Overall, the work presented here corroborates and provides a mechanistic basis for previous clinical observations in neuropathic pain patients. Increased recruitment of Nav1.9, Nav1.8 and Cav1.2 to the apical plasma membrane and proximal stem axon of primary afferent nociceptive neurons following nerve injury may be crucial to the increased excitability and excessive firing that is likely to underlie pain hypersensitivity. These events, in small diameter nociceptive DRG neurons, are driven by ErbB1, acting in synergy with ErbB2, and their downstream signalling through Akt to regulate vesicular trafficking, emphasizing the potential value of targeting ErbB1 and ErbB2 to suppress these processes and achieve analgesia for neuropathic pain. Supplementary data to this article can be found online at https:// doi.org/10.1016/j.nbd.2020.104961. Declaration of competing interest None.
We are grateful to Martin Michaelis for his consistently wise advice, to Anisha Kubasik-Thayil from the IMPACT Confocal Imaging Facility for expert imaging assistance and to the staff of Biomedical Research Resources (BRR) for their ongoing technical support. Role of funding source This work was supported by a scientifically unrestricted grant (Ref RB0435) from Merck KGaA, who played no role in the design of experiments or the analysis and interpretation of data.