|Home | About | Journals | Submit | Contact Us | Français|
Peripheral tissue injury is associated with changes in protein expression in sensory neurons that may contribute to abnormal nociceptive processing. We used cultured dorsal root ganglion (DRG) neurons as a model of axotomized neurons to investigate early changes in protein expression following nerve injury. Comparing protein levels immediately after DRG dissociation and 24 h later by proteomic differential expression analysis, we found a substantial increase in the levels of the neurotrophin-inducible protein VGF (non-acronymic), a putative neuropeptide precursor. In a rodent model of nerve injury, VGF levels were increased within 24 h in both injured and uninjured DRG neurons, and the increase persisted for at least 7 days. VGF was also upregulated 24 h following hind-paw inflammation. To determine whether peptides derived from proteolytic processing of VGF participate in nociceptive signaling, we examined the spinal effects of AQEE-30 and LQEQ-19, potential proteolytic products previously shown to be bioactive. Each peptide evoked dose-dependent thermal hyperalgesia that required activation of the mitogen-activated protein kinase (MAPK) p38. In addition, LQEQ-19 induced p38 phosphorylation in spinal microglia when injected intrathecally and in the BV-2 microglial cell line when applied in vitro. In summary, our results demonstrate rapid upregulation of VGF in sensory neurons after nerve injury and inflammation and activation of microglial p38 by VGF peptides. Therefore, VGF peptides released from sensory neurons may participate in activation of spinal microglia following peripheral tissue injury.
Sensory neurons respond to peripheral tissue damage, such as nerve injury or inflammation, with changes in protein expression and functional properties that contribute to altered nociceptive processing in the spinal cord (Woolf and Costigan, 1999; Campbell and Meyer, 2006; White et al., 2007). Spinal microglia participate in the development of inflammatory and nerve injury-induced hypersensitivity through mechanisms mediated in part by the mitogen-activated protein kinase (MAPK) p38 (Jin et al., 2003; Svensson et al., 2003; Raghavendra et al., 2004; Tsuda et al., 2004; Tsuda et al., 2005; Boyle et al., 2006). Within sensory neurons, changes in expression levels have been reported for a number of signaling mediators, including neuropeptides, neurotrophins, and chemokines (for example, Fukuoka et al., 1998; Woolf and Costigan, 1999; Karchewski et al., 2002; Zhang and De Koninck, 2006). The upregulation of signaling mediators released from sensory neurons in spinal cord may contribute to microglial activation within the dorsal horn after peripheral tissue damage.
In this study we employed cultured adult dorsal root ganglia (DRG) as a model system of axotomized neurons to investigate early changes in protein expression following nerve injury. Relative quantitation of protein levels was done using iTRAQ, a mass spectrometry-based method for proteomic differential expression analysis (Li et al., 2007; Lund et al., 2007). Among the proteins identified as upregulated in this analysis, we selected for further study the neurosecretory protein VGF (non-acronymic) based on evidence that its expression is induced by neuronal lesions and neurotrophins (Levi et al., 2004). VGF, a 617-amino acid protein related to the chromogranins, has the characteristics of a neuropeptide precursor, but the precise identities of bioactive VGF-derived peptides and the putative receptors that mediate their action are unknown. VGF has been implicated in neuroplasticity associated with depression as well as learning and memory, and in cultured hippocampal neurons, VGF-derived peptides potentiated synaptic activity (Alder et al., 2003; Hunsberger et al., 2007; Thakker-Varia et al., 2007; Bozdagi et al., 2008). In this study we addressed the hypotheses that VGF is upregulated in sensory neurons after peripheral tissue injury and that VGF-derived peptides have pronociceptive spinal effects. Our results demonstrate rapid VGF upregulation following nerve injury and inflammation and show that VGF-derived peptides evoke p38-dependent thermal hyperalgesia and p38 activation in microglial cells.
All procedures conform to guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of the University of Minnesota.
Primary cultures of rat DRG neurons were prepared using methods similar to those previously described (Khasabova et al., 2004). Briefly, DRG were isolated from all spinal levels of four adult male rats (Sprague-Dawley, Harlan, Indianapolis, Indiana, 125–200 g), enzymatically dissociated in collagenase-D (1.5 mg/mL, Roche Diagnostic, Indianapolis, IN; two 45 min incubations) and triturated through fire-constricted Pasteur pipets. For proteomic and Western blot analysis, all cells from one rat were divided in half and plated in two 10 cm glass Petri dishes coated with poly-L-lysine (10 µg/mL; Sigma-Aldrich, St. Louis, MO) and laminin (10 µg/mL; Sigma-Aldrich) in F12/Ham’s medium containing 10% fetal bovine serum (plating density 50,000–80,000 cells/dish). Cells from one of the dishes were harvested approximately 1.5 h after plating (t0) and cells from the other dish were harvested 24 h later (t24). For immunohistochemistry, glass coverslips were placed into 10 cm glass Petri dishes before coating with poly-L-lysine and laminin and removed either 1.5 h (t0) or 24 h (t24) after plating, rinsed twice with PBS, incubated for 20 min in fixative (4% paraformaldehyde and 0.2% picric acid in 0.1 M phosphate buffer, pH 6.9) and rinsed in three changes of PBS before proceeding with immunohistochemistry.
Cells were washed with PBS and detached by gentle pipeting of the buffer and agitation at 4 °C. Recovery of neurons was verified visually. The cells were collected by centrifugation at 3000×g for 15 min, frozen in liquid nitrogen and stored until sample preparation. The cells were separated into subcellular fractions using ProteoExtract™ Subcellular Proteome Extraction Kit (Calbiochem, San Diego, CA) and the proteins in the cytoplasmic and nuclear fractions were prepared for mass spectrometric analysis by precipitation using cold acetone. Proteins from the cytoplasmic and nuclear fractions of both time points were denatured, reduced and cysteine alkylated prior to digestion with trypsin, and labeled with the iTRAQ® reagents (Applied Biosystems, Foster City, CA) following manufacturer’s directions. The labeled tryptic peptides in the four samples (114:t0 cytoplasmic fraction, 115:t24 cytoplasmic fraction, 116:t0 nuclear fraction, 117:t24 nuclear fraction) were combined and separated by two-dimensional liquid chromatography (2D LC) using strong cation exchange on a Magic 2002 HPLC system (Michrom BioResourses Inc., Auburn, CA) followed by C18 reverse-phase LC (Dionex/LC Packings capillary LC system; Sunnyvale, CA) online with a QSTAR Pulsar i Mass Spectrometer (Applied BioSystems) as described in Lund et al., 2007. The tandem mass spectrometry (MS/MS) data were analyzed using ProteinPilot Software 2.0.1 (Applied Biosystems/MDS Sciex, Foster City, CA) with the Paragon Algorithm (Shilov et al., 2007) and searched against a non-redundant database of human, mouse, and rat proteins compiled from NCBI on March 9th, 2007 that included common contaminants (59,817 total proteins) for the identification and quantification of proteins. ProteinPilot assigns confidence of protein identification based on the highest confidence peptide used to identify the protein, and peptide confidence is a measure of sequence accuracy determined by MS/MS sequence coverage and the number of other potential sequence matches. The resulting set of identified proteins, which included corresponding gene identifiers and expression values, was uploaded into Ingenuity Pathways Analysis (Ingenuity Systems) and annotated. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. Further analysis was focused on proteins with iTRAQ® ratios greater than 2. For proteins with ratios that showed more than a 2-fold increase but were identified with less than 95% confidence, the identification was confirmed by manual inspection of the peptide spectra. In addition, ratios with P > 0.05 (assigned by Protein Pilot) were verified by manual inspection of the iTRAQ® reporter regions fragmented from the peptides.
Cells were homogenized for 5 s on ice in extraction buffer (TBS pH 7.4, 1% Triton X-100, 10 mM EDTA, 10 mM EGTA) in the presence of protease inhibitors (Complete Mini, EDTA-free, Roche, Palo Alto, CA) using a glass homogenizer. The supernatant was collected and mixed with 4X loading buffer (3 parts sample, 1 part loading buffer; NuPage, Invitrogen), heated for 10 min at 70° °C and loaded into wells of 4–12% Bis-Tris gels (NuPage, Invitrogen, Carlsbad, CA). The gels underwent electrophoresis at constant voltage (160 V) for 1 h, and proteins were transferred from the gel onto PVDF membranes for 1 h using constant amperage (250 mA). The membranes were incubated as follows: (1) in blocking buffer (TBS pH 7.4, containing 0.2% I-Block reagent (Applied Biosystems) and 0.1% Tween-20) overnight at 4 °C; (2) in anti-VGF primary antisera (1:10,000; generated in house as described below) overnight at 4 °C; (3) washed 4×30 min in blocking buffer; (4) in horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch; 1:10,000) at room temp for 2 h; (5) washed 4×30 min in TBS containing 0.1% Tween-20 and 2×30 min in TBS; (6) in chemiluminescent substrate (Pierce Supersignal, Pierce, Rockford, IL) for 5 min. The membranes were then exposed to Kodak X-O-Mat Blue film.
SNL was performed on adult male Sprague-Dawley rats (Harlan, 150–200 g) similar to previously described procedures (Kim and Chung, 1992). Under isoflurane anesthesia, the left L5 transverse process was exposed and removed and the L5 spinal nerve was tightly ligated with 6-0 silk suture. Sham surgery consisted of removal of the transverse process and visualization of the L5 spinal nerve without touching it. Induction of hypersensitivity was measured using von Frey filaments and the up-down method as described (Chaplan et al., 1994). Animals were sacrificed by perfusion fixation 1-, 3-or 7 days after surgery.
Complete Freund’s Adjuvant (CFA) inflammation was induced as previously described (Wenk et al., 2006). Under isoflurane anesthesia, 100 µL of CFA (1:1 emulsion in saline, Sigma, St. Louis, MO) was injected subcutaneously into the plantar surface of the right hind paw of adult male Sprague-Dawley rats (Harlan, 150–200 g). The animals were sacrificed by perfusion fixation 24 h after CFA injection.
The following VGF-derived peptides were used in this study: AQEE-30, corresponding to the last thirty amino acids (588-AQEEADAEER RLQEQEELENYIEHVLLHRP-617) of the VGF protein; AQEE-11, corresponding to the first 11 amino acids of AQEE-30; LQEQ-19, corresponding to the last 19 amino acids of AQEE-30. The peptides were individually synthesized and purified by HPLC (Biomedical Genomics Center, University of Minnesota).
VGF peptides were administered intrathecally (i.t.) in conscious adult male mice (20–25 g) as described (Hylden and Wilcox, 1981). A 30-gauge, 0.5 inch needle connected to a 50-µL Luer-hub Hamilton syringe was used to deliver 5 μL of either saline (vehicle), AQEE-30, LQEQ-19, or AQEE-11. Baseline tail flick (TF) latency data in response to water immersion (49 °C) was collected on all mice prior to the injection and at selected time points post-injection. In some cases, LQEQ-19 and saline injected mice were sacrificed by perfusion fixation immediately following behavioral testing. TF latency data was analyzed by first calculating delta TF values (Experimental TF latency - Baseline TF latency) and using those values in the formula for Percent Inhibition for the mean of each dose group (Control Delta TF Latency - Experimental Delta TF Latency)/(Control Delta TF Latency)*100. In some experiments the following inhibitors were administered as a 5 min pre-treatment: 7-NI, 10 nmol (Tocris Cookson Biosciences, Ellisville, MO); L-NAME, 100 nmol (Sigma-Aldrich); MK801, 10 nmol (Sigma-Aldrich); GF109203X, 1 and 10 nmol (Tocris); KT5720, 1.7 nmol (Sigma-Aldrich); U0126, 2.5 nmol (Tocris); SB202190, 0.1, 1, and 2.5 nmol (Tocris); SB600125, 2.5 nmol (Tocris). The selected doses were similar to doses previously used in similar experimental paradigms (Inoue and Ueda, 2000; Sakurada et al., 2002; Wu et al., 2006; Gabra et al., 2007; Komatsu et al., 2007). 7-NI, L-NAME, MK01, GF109203X, and KT5720 were dissolved with 0.9% normal saline, pH 7.2. U0126, SB202190, and SB600125 were dissolved in DMSO and diluted to a final concentration of 5% DMSO with 0.9% normal saline, pH 7.2.
For preparation of DRG or spinal cord tissue sections, rats or mice were deeply anesthetized with isoflurane and perfused via the heart with calcium-free Tyrode’s solution (in mM: 116 NaCl, 5.4 KCl, 1.6 MgCl2 6H20, 0.4 MgSO4 7H2O, 1.4 NaH2PO4, 5.6 glucose, and 26 Na2HCO3) followed by fixative (described above), and finally with 10% sucrose in PBS. DRG and spinal cords were removed and incubated in 10% sucrose overnight at 4 °C. Tissues were cryostat-sectioned (14 µm) and thaw-mounted onto gelatin-coated slides. Coverslips with cultured cells or tissue sections were preabsorbed in blocking buffer (PBS containing 0.3% Triton-X 100, 1% BSA, 1% normal donkey serum) for 30 min, incubated in primary antisera listed below (coverslips for 4 h at room temperature; tissue sections overnight at 4 °C), rinsed with PBS 3×10 min, incubated in Cy3- or Cy2-conjugated secondary antisera (1:300 and 1:100 respectively, Jackson ImmunoResearch, West Grove, CA) for 1 h at room temperature, washed with PBS and mounted onto slides or coverslipped using PBS/glycerol containing 0.1% p-phenylenediamine (Sigma). In some cases, DRG sections were counterstained with YOYO-1 (Molecular Probes, Eugene, OR) prior to coverslipping.
At the onset of the immunohistochemical experiments several commercial VGF antisera (Santa Cruz Biotechnology, Santa Cruz, CA) were evaluated for their ability to label cultured DRG neurons. Adequate labeling in the neuronal cell bodies was obtained with the N-terminal antiserum anti-VGF (D-20) diluted at 1:300. This antiserum was used for quantitative image analysis in DRG cultures and DRG tissue sections (see below). Subsequently it was noted that this antiserum did not label the processes of cultured neurons and that the labeling in superficial dorsal horn of spinal cord sections was very weak. Therefore, we generated antisera to the VGF C-terminus using previously described methods (Vulchanova et al., 1997). Briefly, the peptide AQEE-30 was conjugated to bovine thyroglobulin (Sigma, St. Louis, Missouri) using glutaraldehyde. The peptide conjugate (1 mg/mL) was emulsified with an equal volume of Freund’s adjuvant (Difco, Detroit, MI) and injected into female guinea pigs (n = 4, Harlan, Indianapolis, IN) at two week intervals (0.5 mg peptide for initial and 0.25 mg peptide for subsequent immunizations). Staining with anti-VGF antisera was blocked by preabsorption of the antisera with the cognate peptide (10 µg/ml). In addition, the specificity of the antisera was ascertained by Western blot analysis, which showed bands consistent with previous reports (Fig. 1) as well as by double-labeling with the N-terminal antisera in DRG, which showed an overlapping pattern of staining (data not shown). The C-terminal VGF antisera (1:1000) were used for quantitative image analysis in spinal cord. Additionally, the quantification in DRG cultures was repeated with the C-terminal antisera, yielding the same results as the N-terminal antisera. Other primary antiseraThe following other primary antisera were used for immunohistochemistry: mouse anti-PGP (1:1000; Biogenesis, Poole, UK); rabbit anti-TrkA (1:1000; Advanced Targeting Systems, San Diego, CA); guinea pig anti- P2×3 (1:500; Vulchanova et al., 1997); goat anti-ATF3 (1:300; Santa Cruz Biotechnology); rabbit anti-CGRP (1:1000; ImmunoStar, Hudson, WI); rabbit anti-phospho- p38 (1:100; Cell Signaling, Danvers, MA); mouse anti-NeuN (1:1000; Chemicon, Temecula, CA); mouse anti-GFAP (1:1000; Cell Signaling); rat anti-CD68 (1:100; AbD Serotec, Oxford, UK).
During image analysis, the observer was blinded to the experimental treatments. Cultured DRG neurons. Nine coverslips from three different DRG cultures (3 coverslips /culture) were analyzed for each time point. Stacks of PGP9.5/VGF double-labeled confocal images (3 per coverslip) collected with a Bio-Rad MRC-1024 microscope (Bio-Rad, Hercules, CA, USA) were analyzed using Image J (NIH freeware). Neurons, identified based on their PGP9.5 labeling, were outlined, and the average intensity of VGF immunoreactivity (ir) of each neuron was measured. For each of the three cultures, 18 VGF-negative cells (2 per image) were identified from the t0 images (negative cells were often hard to identify in t24 images). The threshold intensity for identifying VGF-positive neurons for t0 and t24 of each culture was defined as the average intensity of these negative cells + 2 Standard Deviations. DRG Neurons in tissue sections: For each DRG, 5 sections equally spaced throughout its length were triple-labeled with anti-VGF (N-terminal),YOYO-1, and either anti-TrkA or anti-P2×3. Stacks of triple-labeled confocal images (9 per DRG) were analyzed using Image J. Neurons were identified and outlined based on their Nissl-like YOYO-1 labeling, and cell body area and the average intensity of VGF staining of each neuron was measured. Three VGF-negative neurons were identified in each image, and the threshold intensity for identifying VGF-positive neurons was defined as the average of the three unlabeled neurons + 2 Standard Deviations. Similarly, the average intensity of TrkA and P2X3 staining was measured and the threshold intensity for identifying positive neurons was determined based on the average intensity of 3 unlabeled neurons. VGF-ir and CGRP-ir in rat spinal cord: Quantitative analysis was performed at two levels of the L5 spinal segment that were separated by approximately 1 mm. The results from these two levels were similar. In sections double-labeled for VGF and CGRP, confocal images of dorsal horn were analyzed using Image J. For each spinal cord, a threshold intensity, corresponding to the average intensity of unlabeled regions within the dorsal horn, was selected to separate VGF labeling from background. The number of pixels with intensity above threshold (thresholded area) was measured in the superficial laminae of the right and left dorsal horn. To eliminate variability between animals due to tissue processing, labeling on the left side (operated in SNL and sham animals) was normalized to the right side. For each spinal cord, data were collected from 3 sections, averaged, and expressed as percent change (i.e. (ipsilateral–contralateral)/contralateral*100). CGRP-ir in dorsal horn was analyzed as described for VGF-ir. Phosphorylated p38 (phospho-p38)-ir in mouse spinal cord: Dorsal horn images were collected with a confocal microscope (Olympus FluoView 1000 IX2). For each image, a threshold intensity, corresponding to the maximum intensity of an unlabeled region, was selected to separate phospho-p38 labeling from background. The pixel area of phospho-p38-ir above background was normalized to the spinal cord area included in the image and expressed as % thresholded area. For each animal, % thresholded area was determined by adding % thresholded area of both left and right dorsal horns from three spinal cord sections.
The BV-2 microglial cell line (a kind gift from Dr. Jonathan Godbout) was maintained in DMEM supplemented with 10% FBS. Cells were treated with LQEQ-19 or vehicle (serum-free DMEM) for 2 min. For Western blot, treated cells were lysed (20 mM Tris, 150 mM NaCl, pH 8, 1% (v/v) Triton X-100) in the presence of protease inhibitors (Complete Mini, EDTA-free, Roche, Palo Alto, CA) and phosphatase inhibitors (Halt Phosphatase Inhibitor Cocktail, Thermo Scientific) on ice for 1 h. The samples were centrifuged (40,000×g, 4 °C, 30 min) and loaded (60 µg per lane) onto 4– 12% bis-Tris gels (Invitrogen, Carlsbad, CA) and run at 130 V for 90 min. Next proteins were transfered to nitrocellulose membranes and treated as follows: (1) incubated in blocking buffer (Tris-buffered saline, 0.05% (v/v) Tween-20, 5% (w/v) non-fat milk) for 120 min at room temperature; (2) washed with TBS-T (4×10 min); (3) incubated in anti-phospho-p38 (1:700; Cell Signaling) in TBS-T with 5% BSA at 4 °C for 12 h; (4) washed in TBS-T (4 ×10 min); (5) incubated in goat anti-rabbit-IRDye800CW antibody (Li-Cor Biotechnology, Lincoln, NE) for 90 min at room temperature; (6) washed in TBS-T (3×10 min) and TBS (2×10 min). The membrane was then scanned with the Li-Cor Odyssey imager. The fluorescence intensities of the phosphorylated p38 signals were measured, and the membrane stripped by NewBlot Nitro Stripping Buffer (Li-Cor) for 4 min. Removal of antibodies was verified by re-scanning of the membrane. The membrane was then reprobed as above except with rabbit anti-p38 as the primary antibody (1:1000, #2831; Cell Signaling). The phospho-p38 intensities were normalized to their corresponding total p38 intensities and then normalized to the control treatment (0 nM LQEQ-19) values. For immunohistochemistry, cells were grown to confluence on coverslips and, following treatment, were processed as described above for DRG cultures.
We postulated that over the first 24 h following dissociation and plating cultured, adult sensory neurons undergo changes in protein expression that are similar to those occurring in axotomized neurons. Cytoplasmic and nuclear fractions obtained from DRG cultures at t0 and t24 were compared using iTRAQ®-based quantification of relative abundance combined with 2D LC-MS/MS. A database search using spectra detected by 2D LC-MS/MS identified 3569 distinct peptides. These peptides were mapped to 537 proteins (≥ 66% confidence), of which 481 were identified with greater than 95% confidence. Quantification of protein ratios in the four samples was based on the ratios of the iTRAQ® reporter tags for each identified peptide (Suppl. Fig.1). The data were further examined using Ingenuity Pathway Analysis (IPA). Of the 537 identified proteins 436 were annotated in the IPA Knowledge Base. The majority of these proteins (approximately 80%) were annotated as cytoplasmic and only 16% were annotated as nuclear. A large number of cytoplasmic proteins was detected in the nuclear fractions, preventing adequate evaluation of changes in nuclear proteins. Therefore, subsequent analysis was focused on changes in protein levels in the cytoplasmic fractions.
The ratio Cytoplasmic t24/Cytoplasmic t0 indicated that the majority of quantified proteins (approximately 60%) were changed less than 1.5-fold between t0 and t24, indicating consistency of sample preparation at the two time points. The proteins that were increased more than 2-fold after 24 h in culture are shown in Table 1. The potential functional relevance of these proteins was inferred by an IPA network analysis-guided literature review. However, confirmation of upregulation and analysis of actual functional relevance requires validation by molecular and cellular biology methods. We focused on elucidating the role of VGF, which demonstrated the largest change in expression. The neurosecretory protein VGF was identified by a single peptide, and iTRAQ relative quantification showed a 45-fold increase in the levels of this peptide at t24, suggesting upregulation of VGF in the DRG cultures (Suppl. Fig.1).
The increase in VGF levels in cultured DRG neurons at t24 was confirmed by Western blot analysis and immunocytochemistry (Fig. 1). Two protein bands, one of approximately 90 kDa and one of approximately 70 kDa were seen in homogenates from cells maintained in culture for 24 h but not in homogenates from cells that were harvested shortly after plating. Immunocytochemical analysis of the cultures indicated that the number of VGF-positive neurons, expressed as the percent of neurons identified by the pan-neuronal marker PGP 9.5, was significantly increased at t24 compared to t0 (20 ±4 % at t0 compared to, 82 ± 7 % at t24; P < 0.05, t-test). The quantification of the percent of VGF-positive neurons was not confounded by a change in the total number of neurons over the 24 h period because the total number of PGP-positive cells at t0 and t24 (91 ± 7 and 86 ± 15, respectively) was not significantly different. Similar immunocytochemical results were obtained with the C-terminal VGF antiserum and the commercially available N-terminal antiserum.
We predicted that changes in protein expression in cultured neurons would parallel nerve injury-induced changes in the L5 spinal nerve ligation (SNL) model. Therefore, the expression of VGF was examined immunohistochemically in L5 DRG 24 h after SNL. As shown in Figure 2, the number of VGF-positive neurons in L5 DRG of SNL animals was significantly increased compared to sham-operated and naïve animals. One day following SNL, 80% of L5 DRG neurons were VGF-positive, compared to 37% in sham-operated animals and 20% in naïve animals. Analysis of colocalization showed overlap of VGF-ir with both TrkA and P2X3 labeling (Fig. 3). Of the VGF-positive neurons, 51.9±1.4% were TrkA-positive and 52.3±3.6% were P2X3-positive; conversely, 90.8±2.3% of the P2X3-positive and 70.7±3.5% of the TrkA-positive neurons in the injured L5 DRG were also VGF-positive. Size distribution analysis indicated that the subset of VGF-positive cells in injured L5 DRG included small/medum-sized and large (diameter > 25 µm) neurons (Table 2). The significant increase in VGF-ir was maintained 7 days after SNL. Notably, whereas the number of small/medium-sized VGF-positive neurons at that time was decreased relative to day 1, the number of large VGF-positive neurons was slightly increased (Table 2).
In addition to the injured L5 DRG neurons, VGF was upregulated in uninjured L4 DRG neurons after SNL (Fig. 2D). The number of VGF-positive cells was significantly increased in L4 DRG both 1 and 7 days after SNL compared to sham controls and naïve animals. Similar to L5 DRG, at day 1 approximately half (53.3±4.4 %) of the VGF-positive neurons were also TrkA-positive (Suppl. Table 1).
It was possible that in L4 DRG of SNL animals and DRG of sham-operated animals the VGF upregulation was restricted to neurons inadvertently injured during the surgery. To address this possibility, we examined immunohistochemically the colocalization of VGF and the nerve injury-inducible transcription factor ATF3 (Fig. 4). As previously described (Shortland et al., 2006), we found that ATF3-ir was present in virtually all L5 neurons of SNL animals as well as in a limited number of neurons within L4 and sham DRG. In the injured L5 neurons, we observed a nearly complete overlap of VGF- and ATF3-ir. However, in L4 DRG of SNL animals, there were many VGF-positive cells that were ATF3-negative. Similarly, in sham DRG, VGF-ir was not restricted to ATF3-ir neurons. These results show that VGF was upregulated in injured as well as uninjured neurons.
VGF-ir also increased in the superficial dorsal horn of the spinal cord following SNL. VGF immunofluorescence was measured in the superficial dorsal horn of three sections of L5 spinal cord from four different animals three days following SNL surgery. VGF-ir was significantly increased within the dorsal horn ipsilateral to SNL compared to sham surgery and naïve animals (Fig. 5). In the same sections, there was no difference in CGRP-ir in the three experimental groups (data not shown). To determine whether VGF-ir increased in central terminals of primary afferent neurons or in intrinsic spinal cord neurons, we examined the relationship of VGF-ir to CGRP-ir, which in dorsal horn is localized exclusively in primary afferent terminals. The colocalization of VGF-ir and CGRP-ir in the superficial dorsal horn ipsilateral to SNL was increased compared to spinal cord from naïve animals, consistent with elevation in VGF levels in central terminals of primary afferent neurons after SNL (Fig. 5, E –J). It should be noted that the abundance of VGF-ir in spinal cord as revealed in these experiments by our C-terminal antiserum may be underestimated if differential proteolytic processing of VGF (Brancia et al., 2005) results in fragments that are not recognized by this antiserum.
The increased number of VGF-ir neurons in DRG from sham animals suggested that VGF might also be induced in inflammatory conditions since the sham surgery is accompanied by substantial tissue damage and inflammation. Therefore, we examined changes in VGF-ir 24 h after CFA-induced hind-paw inflammation. CFA inflammation resulted in a more modest, but still significant increase in the number of VGF-expressing neurons in L5 DRG (28.2 ± 0.3% in CFA L5 DRG compared to 19.0 ± 2.8% in naïve L5 DRG; P < 0.05, t-test). In L5 DRG from CFA-treated animals, 65.5±5.2 % of the VGF-positive neurons were also TrkA-positive (Suppl. Table 1). Thus VGF expression in sensory neurons appears to be induced by a spectrum of manipulations that includes varying degrees of inflammation and nerve injury.
Previous studies have reported biological activity of the VGF C-terminal peptides TLQP-62, AQEE-30 and LQEQ-19 (Alder et al., 2003; Succu et al., 2005). To determine whether they have a role in nociceptive processing, AQEE-30 and LQEQ-19 were injected intrathecally into naïve adult mice (Fig. 6). In the warm water (49 °C) tail-immersion assay, mice show baseline tail flick latencies in the range of 6–9 seconds (x = 7.77 s ± 0.06 (s.e.m.), n = 156, for data shown in Fig.6 and Fig. 7). Following intrathecal delivery, AQEE-30 evoked a partial thermal hyperalgesia by 10 min, which was fully expressed by 30 min and persisted for at least 60 min and up to 90 min at the highest dose (Fig. 6, Table 3). The induction of thermal hyperalgesia was dose-dependent as evidenced by a significant difference in the magnitude of the hyperalgesia evoked by the different doses at the 30 min time point (one-way ANOVA, F(3,20) = 3.48, p < 0.05, Tukey’s posthoc test). However there was no statistical difference between the magnitudes of the thermal hyperalgesia evoked by the two highest doses (3 nmol: −1.3 s ± 0.19: 0.3 nmol: −1.7 s ± 0.34, p > 0.05, n=6),
LQEQ-19, which corresponds to the last 19 amino acids of AQEE-30, evoked hyperalgesia of similar magnitude (−1.4 ± 0.43, n=6), but with a more rapid onset to full expression and a shorter duration. In contrast, AQEE-11, which corresponds to the leading 11 amino acids of AQEE-30, had no effect on tail withdrawal latency (Fig. 6, Table 3).
Several pharmacological agents known to interfere with pathways involved in pain processing were evaluated for inhibition of the thermal hyperalgesia evoked by AQEE-30 and LQEQ-19. These included inhibitors of nitric oxide synthase (7-NI, 10 nmol; L-NAME, 100 nmol), NMDA receptor (MK801, 10 nmol), protein kinase C (GF109203X, 1 and 10 nmol), protein kinase A (KT5720, 1.7 nmol) and MAPK (U0126, 2.5 nmol; SB202190, 0.1, 1, and 2.5 nmol; SB600125, 2.5 nmol). Inhibition of the spinal effects of the VGF peptides was achieved only by the p38 inhibitor SB202190, which dose-dependently reversed AQEE-30 (0.3 nmol) and LQEQ-19 (1 nmol) evoked thermal hyperalgesia (Fig. 7). The thermal hyperalgesia evoked by these doses was equivalent between the two peptides (AQEE-30: −2.4 ± 0.23 s; LQEQ-19: −2.3 ± 0.37 s; p > 0.05, Student’s t test, n = 6 per group) and was not statistically different from the thermal hyperalgesia evoked in the previous experiment (Fig. 7) with the same doses and time points (Student’s t test, p >0.05). Administration of SB202190 alone had no effect on thermal withdrawal latencies.
Activation of p38 MAPK was examined immunohistochemically in spinal cords obtained at the time of peak hyperalgesia (10 min) from mice injected intrathecally with LQEQ-19 or saline (Fig. 8). In spinal cords from LQEQ-19-injected mice, we observed a nearly 3-fold increase in immunofluorescence for phosphorylated p38, measured as pixel area with staining intensity above background (% thresholded area: 0.36 ± 0.08 in LQEQ-19-treated compared to 0.13 ± 0.04 in saline-treated animals; p < 0.05, one-tailed t-test; n = 3 for each group) (Fig. 8, A–C). Immunoreactivity for phosphorylated p38 did not overlap with labeling for the neuronal marker NeuN and the astrocyte marker GFAP. However, it colocalized extensively with labeling for CD68, a marker for reactive microglia, suggesting that phosphorylated p38 was localized in spinal cord microglia (Fig. 8, D–G). To determine whether LQEQ-19 has direct effects on microglia, we exposed the BV-2 microglial cell line to the peptide. This treatment resulted in the concentration-dependent increase of p38 phosphorylation in the cells (Fig. 9).
The present study contributes to the characterization of a novel signaling system within spinal cord pain pathways. The potential relevance of VGF was first inferred from proteomic differential expression analysis of an in vitro model of injured sensory neurons. Subsequent experiments provided evidence for upregulation of VGF after nerve injury and inflammation and for the functional relevance of VGF peptides to spinal nociceptive processing.
Large-scale expression analysis has been used to study global changes under conditions of chronic pain, leading to identification of genes with uncharacterized functions in pain signaling (Costigan et al., 2002; Wang et al., 2002; Valder et al., 2003; Komori et al., 2007). We employed a proteomic approach to analyze early changes in protein levels within cultured DRG neurons, which were used as a model of axotomized neurons. Several of the proteins that were increased more than 2-fold are associated with oxidative stress, consistent with the hypoxia accompanying preparation of primary cultures. The transcription regulators APEX1 and YBX1 have also been linked to protection from oxidative stress (Kohno et al., 2003; Vasko et al., 2005; Qu et al., 2007). Moreover, IPA network analysis showed that APEX1, YBX1 and prothymosin α are regulated by casein kinase II, suggesting that the upregulation of these proteins may reflect activation of a common network that involves this enzyme. Another group of upregulated proteins is associated with membrane-cytoskeletal signaling at growth cones and presynaptic terminals and possibly with regulation of second messenger availability (Laux et al., 2000; Diviani and Scott, 2001; Sundaram et al., 2004). For example, MARCKSL1 and BASP1/CAP-23 may control the phosphoinositol-(4,5)biphosphate-dependent regulation of channels contributing to sensory neuron sensitization, including TRPV1 and TRPA1 (Akopian et al., 2007; Lishko et al., 2007; Lukacs et al., 2007). Our proteomic analysis also indicated upregulation of two neuropeptide precursors. The upregulation of CGRP precursor in the first 24 h of DRG culture is consistent with increases in CGRP in cultured trigeminal neurons and in some models of nerve injury (Kuris et al., 2007; Zheng et al., 2008). Previous studies have shown that in the SNL model, the number of CGRP-positive neurons in injured ganglia was unchanged 24 h after injury (Lee et al., 2001) and substantially decreased at 7 days (Fukuoka et al., 1998; Lee et al., 2001; Hammond et al., 2004). VGF upregulation in DRG after nerve injury has been previously detected by microarray analysis (Costigan et al., 2002; Wang et al., 2002; Valder et al., 2003) and was recently described in the spared nerve injury model (Moss et al., 2008).
The increase in VGF levels in cultured DRG neurons was confirmed by Western blot and immunocytochemical analysis. This increase was paralleled by VGF upregulation in the SNL model of nerve injury. Within 24 h of SNL, VGF-ir was present in nearly all injured L5 and over half of uninjured L4 DRG neurons. The time course and level of upregulation clearly distinguish VGF from other neuropeptides known to be increased after nerve injury (Zhang et al., 1995; Fukuoka et al., 1998; Shortland et al., 2006). For example, one day after SNL, galanin was present in approximately 40% of injured L5 DRG neurons and NPY was nearly absent, whereas the levels of both peptides were unchanged in uninjured L4 neurons (Shortland et al., 2006). VGF labeling colocalized with both TrkA and P2X3 labeling, suggesting that the factors governing VGF induction are not restricted to the NGF- or GDNF-dependent subsets of sensory neurons. The rapid increase in VGF levels following SNL in both injured and uninjured sensory neurons and its persistence 7 days after the nerve injury are consistent with involvement of VGF in both the development and maintenance of nerve-injury induced hypersensitivity. Within 24 h of SNL, injured myelinated and uninjured unmyelinated fibers develop spontaneous activity (reviewed in Ringkamp and Meyer, 2005). The increased VGF expression in injured large neurons and uninjured small neurons suggests that increased release of VGF peptides from spontaneously active fibers may contribute to abnormal nociceptive processing in dorsal horn during the development of hypersensitivity. Moreover, VGF upregulation in large neurons persists 7 days after injury, consistent with continued contribution of large afferents to the maintenance of hypersensitivity (Ossipov et al., 2000; Sun et al., 2001).
Since VGF is a putative neuropeptide precursor (Levi et al., 2004), VGF-derived peptides are likely transported to the central terminals of DRG neurons. Therefore, VGF upregulation in DRG following SNL was expected to result in increased levels of VGF-ir within the superficial dorsal horn of spinal cord. Quantitative analysis in spinal cord demonstrated a significant increase in VGF-ir ipsilateral to SNL, and the increased colocalization of VGF- and CGRP-ir in superficial dorsal horn is consistent with elevated VGF levels within central terminals of sensory neurons. Therefore, increased spinal release of VGF peptides from sensory neurons may provide an early signal for peripheral nerve injury. Increased VGF expression following nerve injury has also been shown in spinal neurons (Moss et al., 2008), suggesting additional functions of VGF peptides in spinal plasticity.
The number of VGF-ir DRG neurons also increased after sham surgery but to a level significantly lower than SNL. This increase is likely due to tissue damage at the surgical site. The sham surgery includes removal of the vertebral transverse process and overlaying muscles, causing inflammation and injury to fibers innervating these tissues. Consistent with tissue damage following sham surgery, ATF3 expression was reported in 40% of sham L5 DRG 1 day and 14 days after surgery (Shortland et al., 2006). We also observed an inflammation-induced increase in the number of VGF-positive neurons after hind-paw injection of CFA. Taken together, these results suggest that VGF expression in sensory neurons may be regulated by a spectrum of manipulations, which include varying degrees of inflammation and nerve injury.
The 617 amino acid sequence of VGF contains nearly a dozen potential cleavage sites. Functional effects have been reported for several potential proteolytic products contained within the C-terminal 62 amino acid portion of VGF (Alder et al., 2003; Succu et al., 2005). Recently, TLPQ-21, a VGF-derived peptide located immediately upstream from AQEE-30 within the VGF sequence, was shown to modulate inflammatory pain (Rizzi et al., 2008), and TLPQ-62, a peptide that includes both TLPQ-21 and AQEE-30, produced mechanical allodynia when administered intrathecally in rat (Moss et al., 2008). Our observations of dose-dependent thermal hyperalgesia evoked by intrathecal injection of AQEE-30 and LQEQ-19 provide further evidence for a role of VGF peptides in nociceptive processing. Notably, the hyperalgesia evoked by VGF peptides exceeded in magnitude and duration the effect of NMDA in the same experimental paradigm (Kitto et al., 1992; Roberts et al., 2005). Although the tail withdrawal paradigm used in our studies is a measure of acute hyperalgesia rather than hypersensitivity associated with persistent pain, Moss et al., (2008) demonstrated TLPQ-62-induced mechanical and cold allodynia, consistent with a role of VGF peptides in nerve injury-induced hypersensitivity.
The signaling mechanisms of VGF-derived peptides are completely uncharacterized. Therefore we examined several signaling pathways for their involvement in the spinal effects of AQEE-30 and LQEQ-19. The thermal hyperalgesia was inhibited dose-dependently by the potent and selective p38 MAPK inhibitor SB202190. The activation of p38 by intrathecal injection of LQEQ-19 was confirmed by immunohistochemistry. LQEQ-19-dependent p38 phosphorylation was seen predominantly in spinal microglia, and LQEQ-19 induced p38 phosphorylation in the microglial cell line BV-2. These findings suggest that AQEE-30 and LQEQ-19 may act directly on spinal microglia and contribute to microglial activation. It remains to be determined whether the longer C-terminal peptide TLPQ-62 also mediates p38 activation in spinal microglia.
Spinal microglia participate in the development of inflammatory and nerve injury-induced hypersensitivity in part through the activation of p38. Several mediators have been implicated in the mechanisms of p38 activation, including TNFα, MCP-1, and fractalkine (Svensson et al., 2005; Zhang and De Koninck, 2006; Clark et al., 2007; Zhang et al., 2007; Zhuang et al., 2007; Milligan et al., 2008). Our results suggest that VGF peptides released from sensory neurons may also participate in activation of spinal microglia. Under normal conditions VGF levels in sensory neurons are low, suggesting limited contribution to acute pain signaling. Following nerve injury or inflammation, the combination of rapid VGF upregulation and increased excitability of sensory neurons is likely to potentiate the release of VGF peptides within the dorsal horn, where they would be able to activate microglial p38.
It is presently unclear which functionally relevant VGF peptides are produced in vivo in sensory neurons and spinal cord. Candidates include TLPQ-62 and shorter fragments derived from it (i.e. TLPQ-21, AQEE-30, LQEQ-19). The excitatory effects of TLPQ-62 on dorsal horn neurons may be underlined by microglia-dependent sensitization following action of the C-terminal portion of TLPQ-62 (i.e. AQEE-30, LQEQ-19) as well as by activity of its N-terminus (i.e TLPQ-21), analogous to the effects in hippocampal slices, which were not mimicked by the C-terminal peptides (Bozdagi et al., 2008; Moss et al., 2008). The pattern of VGF processing can vary among different cell types and sometimes even within the same cell type (Brancia et al., 2005). Therefore, differential proteolytic processing of VGF within sensory neurons following different types of tissue damage may generate signaling peptides with distinct contributions to spinal pain mechanisms.
In conclusion, this study contributes to the increasing body of evidence for a role of VGF in neuroplasticity and provides the first description of a signaling pathway activated by VGF peptides. Our results suggest a novel action of VGF peptides as potential neuro-glial messengers following nerve injury and inflammation – a discovery driven by proteomic differential expression analysis and validated through anatomical and functional experiments.
The authors would like to thank Galina Kalyuzhnaya and Kristin Krebs for technical assistance, Dr. Jonathan Godbout for his kind gift of BV-2 cells, and Drs. Robert Elde, Alice Larson, Virginia Seybold and Laura Stone for valuable insights. We are also grateful to Dr. LeeAnn Higgins and the Center for Mass Spectrometry and Proteomics for assistance with proteomic data collection. Funding for the project was provided to L. Vulchanova by K01 DA017236, NIDA/NIH; P. D.Braun was supported by T32 DA07234, NIDA/NIH.