Male Sprague-Dawley rats (Janvier), weighing 200 to 250 g (5 to 6 weeks of age) at the beginning of the experiments, were used. They were housed four per cage under standard conditions of light and temperature, for at least one week before and throughout the whole experimental period. Commercial chow pellets and tap water were available ad libitum. Animal-related procedures were approved by the French Ethics Committee in Animal Experiment (Comité d'Ethique pour l'Expérimentation Animale Charles Darwin, n°Ce5/2010/007) and carried out according to the French Standard Ethical Guidelines for Laboratory Animals.

Complete Freund's adjuvant, dexamethasone, collagenase, penicillin/streptomycin, transferrin, sodium selenite, putrescin, progesterone, corticosterone, triiodothyronine, insulin, cytosine arabinoside, proteases and phosphatases inhibitor cocktails, horseradish peroxidase-conjugated secondary antibodies and rat TNF-α and IL-1β cytokines were purchased from Sigma-Aldrich. MMP-9 inhibitor (Inhibitor-I) was purchased from Calbiochem. Lipofectamine RNAiMax and SiRNAs were obtained from Qiagen. F-12 nutrient mixture culture medium with L-glutamine, HBSS, trypsin, horse, goat and donkey sera and secondary antibodies conjugated to Alexa Fluor 488 or 594 were purchased from Invitrogen.

The primary antibodies used in this study were mouse anti-glial fibrillary acidic protein (GFAP) and anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from Chemicon, goat polyclonal anti-Iba1 (Abcam), mouse anti-neuronal nuclei (NeuN, Chemicon), mouse anti-neurofilament 200 (NF200, Sigma-Aldrich), fluorescein isolectin B4 (IB4, Vector) and guinea pig anti-calcitonin gene-related peptide (CGRP, Abcys SA, 27 ). For NOV immunochemical detection and western blotting experiments, we used a homemade affinity-purified rabbit antibody (referred as CT-Mu, anti-mouse NOV antibody). This antibody was raised against a peptide specific to the carboxyl-terminal region of mouse NOV (amino acids 336 to 354, QNNEAFLQDLELKTSRGEI) which was coupled to activated keyhole limpet hemocyanin36 and used to generate a rabbit polyclonal antibody (CT-Mu). The CT-Mu immunoglobulin G (IgG) fraction was purified using standard methods on a CT-Mu peptide affinity column. The flow-through resulting from CT-Mu purification and containing IgG depleted from NOV specific IgG was used as a negative control 26 . Commercial human NOV protein was obtained from R&D Systems. Recombinant human NOV protein was produced using a baculovirus expression system in insect cells and purified as previously described 28 .

Inflammatory pain was induced by injecting subcutaneously an emulsion of 100 μl of CFA 50% dissolved in 9‰ NaCl (saline) into the left hind paw while the rats were under brief isoflurane anesthesia. Sham rats received a 100 μl injection of saline. For characterization of NOV expression in the CFA model, injured and corresponding sham animals (n = 4 rats/group) were sacrificed in the time course experiment at days 1, 3, 5, 7, 14, 21 and 60 after CFA injection. Two independent experiments were performed.

Water soluble dexamathesone (2 mg/kg i.p. dissolved in saline) was daily administered for 4 consecutive days (n = 4 rats/group) starting from day 9 post-CFA or saline injection. The control groups received four injections of saline in the same volume. The experiment was performed twice.

To proceed to small interfering RNA (SiRNA) intrathecal delivery, control non silencing SiRNA or specific NOV SiRNA (see below) were prepared immediately prior to administration by mixing the RNA solution (100 μM in annealing buffer) using two transfection reagents, either i-Fect™ in a ratio of 1:4 (w/v) (Neuromics, 29 ) or Transductin™ complex respecting a molar ratio of 1:10 (IDT, 30 ) following manufacturers protocols. At these ratios, the final concentration of RNA as an RNA/lipid or peptide complex was 2 μg in 10 μl. Ten microliters of NOV siRNA or control were delivered by intrathecal injection. Injection was given daily for 3 consecutive days starting from day 3 after CFA or vehicle injection (n = 3 to 4 rats/group) and tissues were harvested 24 h after the last injection for RNA analysis.

For behavioral and protein analyses, rats were daily injected intrathecally under brief isoflurane anesthesia with 2 μg of NOV siRNA (160 pmoles in i-Fect™ transfection reagent) or control siRNA on the day of CFA injection and then on the two following days (n = 8 rats/group). To test the NOV effect on pain intensity, three micrograms of commercial human NOV protein (in 12.5 μl saline) were delivered for 4 consecutive days, starting on the 9th day after CFA or saline injection. Control animals were injected intrathecally with the same volume of vehicle (n = 8 rats/group). Similarly, CFA rats were injected with 5 μg of MMP-9 inhibitor or vehicle (10% DMSO) from day 9 to day 12 post-CFA injection (n = 6 rats/group). Last intrathecal (i.t) injection was given 6 hours or 24 hours before animals were killed. Tissues were used for protein analysis.

After arrival in the laboratory, animals were left to accustom to the animal care unit for a week. During the seven days before the experiment, rats were weighed daily and placed in the test room for one hour. Mechanical allodynia (painful response to normally innocuous tactile stimuli) was evaluated the last two days preceding the scheduled experiment day (-d2 and -d1) as well as before the CFA-injection (d0) and each day of the experiment. Mechanical allodynia was measured using calibrated von Frey filaments (Bioseb) as described by Chaplan 31 . Animals were acclimated for 20 minutes in inverted individual cages on top of a wire mesh to provide access to the ventral side of the hind paws. The von Frey filaments were presented perpendicularly to the plantar surface of the selected hind paw, and then held in this position for approximately 5 seconds with enough force to cause a slight bend in the filament. A positive response was indicated by an abrupt withdrawal of the paw. A 50% paw-withdrawal threshold (PWT) was determined by increasing and decreasing the intensity and estimated using the Dixon's up-down method 32 that is, 10^{(Xf + kΔ)}/10 000, where Xf = the value of the last von Frey filament employed, k = Dixon value for the p positive/negative pattern, and Δ = the logarithmic difference between stimuli. On treatment days, rats were tested prior to anesthesia followed by i.t administration of drugs. Data are presented as mean standardized values ± SEM and were analyzed by analysis of variance (ANOVA) followed by the Student-Newman-Keuls' post hoc test analysis.

Rats were deeply anesthetized with pentobarbital sodium (150 mg/kg) and perfused transcardially with saline. The lumbar part of the spinal cord and DRG were removed and placed in a refrigerated cold plate (0 to 4°C). The spinal lumbar enlargement (L5-L6) was divided into left (lesioned side) and right parts by sagittal cut, and then into their dorsal and ventral ones by a horizontal cut passing through the ependymal canal. Dorsal quadrants of spinal cord and DRG were rapidly frozen in liquid nitrogen then kept at -80°C until being processed for RNA or protein analysis. Rats used for immunohistochemical study were perfused transcardially with heparin (1,000 U/ml) in saline followed by 500 ml of a freshly prepared solution of 4% paraformaldehyde (PAF) in PBS, pH 7.4. The dissected DRG and lumbar spinal cord were post-fixed overnight in 4% PAF at 4°C then cryoprotected in 15% sucrose (24 hours, 4°C) before being frozen at -40°C in isopentane.

Fourteen- and twenty-micrometer cryostat sections of DRG and lumbar spinal cord, respectively, were cut with a cryostat (Leica CM3050S). Sections were incubated in a 1X PBS containing 0.1% Triton and 6% serum prior to incubation with primary antibodies. Primary antibodies used for this study were CT-Mu anti-NOV murine (1:5000 for spinal cord, 1:1,000 for DRG), anti-GFAP (1:800), anti-Iba1 (1:500) to detect glial markers, anti-NeuN (1:500), anti-NF200 (1:800), fluorescein isolectin B4 (IB4 1:500) and anti-CGRP (1:800) for neuron detection. Bound antibodies were detected using secondary antibodies conjugated to Alexa Fluor 488 or 594. A tyramide signal amplification was used to optimize NOV detection in spinal cord the following manufacturer's protocol (TSA Fluorescein System, PerkinElmer). Sections were finally mounted in the presence of DAPI to counterstain nuclei (Vector).

Thoracic and lumbar DRG were removed from 3 young adult rats and were chemically and mechanically dissociated into single isolated neurons by enzyme treatment of 0.125% collagenase for 2 hours, followed by trypsin and DNAse for 10 minutes at 37°C and by trituration with Pasteur's pipette in F-12 medium. The cell suspension was then passed over a 70 μm filter, centrifuged and the pellet was resuspended in the F-12 nutrient mixture culture medium with L-glutamine containing 5% horse serum, 1% penicillin/streptomycin, glucose (4.5 g/l) and supplemented with 1.5% N3 medium (0.1% BSA, 10 mg/ml transferrin, 500 μg/ml insulin, 5.8 μM sodium selenite, 20 μM putrescin, 4 μM progesterone, 11 nM corticosterone, 3 μM triiodothyronine, in HBSS). Cells were plated on polylysine-coated wells (10 μg/ml). Cytosine arabinoside (1 μM) was added to eliminate mitotic cells including Schwann cells and fibroblasts. Cultures were maintained at 37°C in a water-saturated atmosphere with 5% CO₂ for 5 days before the experiment. On the fifth day of culture, neurons exhibited globular cell bodies and extended axonal processes.

Viral vectors used were either adenoviral vector containing the NOV gene under the control of a cytomegalovirus promoter (Ad-NOV, 1.1 × 10¹¹ infectious particles/ml) kindly provided by Dr C. Haberberger (Fibrogen Inc) or a control virus (Ad-null, 2.2 × 10¹¹ infectious particles/ml). Cells were infected with viruses at the ratio of 500 infectious particles/cell in culture medium containing 1% horse serum for 2 hours at 37°C. After 2 hours, media were replaced by serum deprived medium for an additional 48 hours prior to treatment. Neither the Ad-NOV nor the mock-infection showed deleterious cytotoxicity.

Specific siRNA targeting NOV, integrins β₁, β₃ or β₅ was used as previously described 26 . A siRNA with a non-silencing oligonucleotide sequence (control, Ctr) showing no known homology to mammalian genes was used as a negative control. Cultured DRG cells were transfected with siRNA (1.5 μg, 120 pmoles) using RNAiMax transfection reagent according to the manufacturer's instructions. Forty-eight hours after transfection, cultures were serum deprived and treated with cytokines and/or NOV for 6 to 24 hours. Cells were harvested and subjected to reverse transcription and real-time PCR. Conditioned media were collected for immunoblotting and detection of MMP-2 and -9 enzymatic activities.

Total RNA was extracted from DRG, dorsal horns of spinal cord or cultured DRG neurons using the Nucleospin RNA II kit (Macherey-Nagel). Then, first-strand cDNA was synthesized from 500 ng total RNA using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems) according to the manufacturer's protocol. Quantitative real-time PCR amplifications were performed on the ABI 7300 apparatus (Applied Biosystems). The real time PCR amplification mixture (20 μl) contained 2 μl of 1/10 diluted cDNA template, 10 μl of Kit Power SYBR Green Master Mix (Applied Biosystems), 6 μl of nuclease-free water and 2 μl of primer mix (5 μM). Standard curves for target and housekeeping genes were generated by serial dilution. Specific primers (Sigma-Aldrich) used for amplification of target genes were designed using the online Primer3 Input program and verified for specificity on BLAST. Rat primers sequences are listed below. The comparative Ct method 33 was used to calculate gene expression values. The ribosomal S26 was used as a housekeeping gene 1.

The gelatinases MMP-2 and MMP-9 proteolytic activities from rat tissues and from DRG neurons cultured in serum-deprived medium were detected by zymography. Conditioned media were centrifuged to remove cell debris and then stored at -20°C until further analysis. Frozen tissues were homogenized on ice by sonication in RIPA buffer. Twenty and thirty micrograms of total proteins (from conditioned media and DRG tissues, respectively) were subjected to electrophoresis under non-reducing conditions in 8% SDS-polyacrylamide gels copolymerized with 1 mg/ml gelatin. Gels were washed twice for 30 minutes in 2.5% Triton X-100 to remove SDS, incubated in substrate buffer (50 mM Tris-HCl, 5 mM CaCl2, 1 μM ZnCl2, 0.01% NaN3, pH 7.5) overnight at 37°C, stained in 0.5% Coomassie Blue G in 40% methanol, 10% acetic acid for 30 minutes at room temperature and destained in 40% ethanol, 1% acetic acid. Clear proteolytic zones indicated the presence of gelatinases at their respective molecular weights.

CCL2 concentrations in primary sensory neuron culture supernatants were quantified by using an ELISA test (OptEIA ™, BD Biosciences Pharmingen) following the manufacturer's instructions. At least two independent experiments were performed.

Frozen tissues were homogenized on ice by sonication in lysis buffer (20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM Na₂EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM b-glycerophosphate) (Cell Signaling) supplemented with protease and phosphatase inhibitor cocktails. Protein concentration was determined using the Bradford assay (Bio-rad) with BSA as the standard. Proteins present in 20 μg of lysates were analyzed. For the detection of secreted NOV from primary sensory neuron cultures, protein samples from conditioned medium were collected after precipitation by deoxycholate/trichloracetic acid (DOC/TCA). Protein samples were subjected to electrophoresis in 10% reducing SDS-PAGE before being transferred to polyvinylidene difluouride membranes (Millipore) for immunological detection. The membrane was incubated with the CT-Mu mouse anti-NOV polyclonal antibody diluted to 1:500 for 1 hour at room temperature, followed by horseradish peroxidase-conjugated secondary antibodies. Immunoreactive proteins were detected by Enhanced chemiluminescence (Amersham, GE Healthcare).

Data are presented as mean values ± SEM and were analyzed with student t-test (two groups only) or ANOVA followed by the Student-Newman-Keuls' post hoc test analysis. A value of P < 0.05 was considered statistically significant.

NOV immunoreactivity (-IR) in nociceptive structures was first investigated by immunohistochemistry in the adult normal rat. In the DRG, NOV-IR was mostly neuronal as it displayed an overlapping pattern with the neuronal marker NeuN (Figure 1, a-c). NOV-IR was present in all neurons and was colocalized with both CGRP⁺-peptidergic and IB4⁺-non-peptidergic neurons (Figure 1, d-i) as well as with the large-diameter neuronal marker NF200 (Figure 1, j-l). NOV-IR could be detected in both neuronal cell bodies and processes (Figure 1, m-o).

In the lumbar DHSC, NOV-IR was detected in all lumbar levels and throughout all laminae (Figure 2A, a). NOV-IR was associated with the cytoplasm of spinal neurons of profound laminae and had a characteristic appearance in the ECM/pericytoplasmic region in the superficial laminae of the DHSC (Figure 2A, a-d). NOV-IR showed overlapping staining with CGRP and NF200 neuronal markers (Figure 2B, a-f). NOV-IR did not exhibit an overlapping staining with either the microglial marker Iba1 or the astrocytic marker GFAP (Figure 2B, g-l). Interestingly, a more moderate NOV-IR was also present in the dorsolateral funiculus of the spinal cord (Figure 2A, a asterisk). These data indicate that NOV is present in nociceptive structures and is essentially expressed by neurons in the DRG and DHSC.

To identify whether NOV could play a role during the development of inflammatory pain, we first performed a time-course analysis of NOV expression in the DRG and DHSC after either CFA or saline hind paw injection. From day 1 to 7 after pain induction, no significant variation of NOV mRNA expression was observed in the ipsilateral DRG and DHSC (Figure 3A and 3B, and data not shown). By contrast, in the DRG, NOV mRNA levels were significantly decreased by 40% at day 15 (**P < 0.01, n = 6 to 8) and returned to their basal levels thereafter (Figure 3A). Spinal NOV mRNA expression was significantly downregulated by 58% starting from day 15 post-CFA (**P < 0.01, n = 8) and persisted at low levels for at least 60 days with a 40% decrease (**P < 0.01, n = 6) (Figure 3B). Consistent with the NOV mRNA expression pattern, NOV protein quantification on western blots in the ispilateral DHSC showed a significant (*P < 0.05, n = 5 to 6) decrease by 34%, 31% and 40% at 15, 21 and 60 days post-CFA, respectively (Figure 3C). No variation of NOV distribution was observed in this structure in CFA rats compared to respective shams (data not shown). These results reveal the fact that, in the time course of CFA-induced pain, NOV levels are downregulated in the ipsilateral DRG and DHSC, suggesting a potential functional association between reduced NOV levels at a late phase and the establishment of prolonged pain.

We then asked whether the treatment of CFA rats with the anti-inflammatory and analgesic glucocorticoid dexamethasone 34 could impact NOV expression. As shown in Figure 4A and 4C, 13 days after CFA-induced pain, NOV mRNA levels were significantly decreased in DRG and in DHSC (*P < 0.05, n = 4). Dexamethasone treatment re-established NOV-reduced expression in ipsilateral DRG and DHSC to levels measured in sham rats (*P < 0.05, n = 4). Dexamethasone had no significant effect on NOV expression in sham rats (Figure 4A and 4C). We also investigated mRNA expression of the cytokines IL-6, CCL2, IL-1β and TNF-α in saline and CFA rats ± dexamethasone. Cytokine expression was significantly upregulated in CFA rats, with a mean of 1.5-fold in the DRG and 3.5-fold in the DHSC compared to respective sham (Figure 4B and 4D, *P < 0.05, **P < 0.01, n = 4). Dexamethasone had an inhibitory effect on the induction of IL-6, CCL2, IL-1β and TNF-α in both DRG and DHSC (Figure 4B and 4D, †P < 0.05, ††P < 0.01, n = 4). This finding demonstrates for the first time in vivo a regulatory effect of dexamethasone on NOV expression and is in accordance with ex vivo studies 35 . This result is compatible with a possible involvement of NOV in inflammatory and/or nociceptive processes.

To gain insights into the possible role of NOV downregulation in the persistence of inflammatory pain, we carried out experiments based on cultured neurons from rat DRG. Because of the aforementioned effects of NOV on the regulation of cytokine, chemokine, and MMP expression in non-neuronal cell types, we explored the hypothesis that NOV was involved in such regulation in cultured DRG neurons. We first assessed the impact of siRNA-mediated NOV inhibition on the expression of MMP-9, CCL2, MMP-2 and substance P (SP) in response to cytokines stimulation.

DRG neuron cultures were transfected with a NOV-targeted siRNA (SiNOV) or a control siRNA (Ctr) containing a non-silencing sequence. SiNOV inhibited more than 80% of both NOV mRNA and protein expression at 72 hours (Figure 5A and 5B).

As shown in Figure 5C and 5D, NOV knockdown led to amplified stimulatory effects of the cytokines TNF-α and IL-1β on the mRNA levels of MMP-9 (siNOV × 73.9 ± 1.6 versus Ctr × 27 ± 2.2; siNOV × 312 ± 8.11 versus Ctr × 164 ± 11.5, respectively, †P < 0.05) and CCL2 (siNOV × 31 ± 0.7 versus Ctr × 14 ± 0.4; siNOV × 20 ± 0.75 versus Ctr × 8 ± 0.58, respectively, †P < 0.05). NOV depletion also impacted CCL2 expression under basal conditions (Figure 5D, not treated, NT, †P < 0.05).

As shown in Figure 5E and 5F, under basal condition (NT), MMP-2 and SP mRNA levels showed significant increases with a mean of 2-fold when NOV expression was abrogated (†P < 0.05). These increased MMP-2 and SP expressions were not amplified upon cytokine exposure (Figure 5E and 5F). Comparable results were obtained when DRG cultures were exposed to cytokines for 6 hours (data not shown).

NOV depletion further enhanced cytokine-induced MMP-9 enzymatic activity with a mean of 1.8-fold (Figure 6A, †P < 0.05) and CCL2 protein levels with a mean of 1.6-fold (Figure 6B and 6C, †P < 0.05). NOV knockdown failed to significantly impact MMP-2 activity (Figure 6A).

Collectively, these data indicate that the blockade of endogenous NOV expression in cultured DRG neurons potentiates the induction of specific pro-nociceptive molecules by cytokines in vitro. Thus NOV deficiency leads to a greater response of sensory neurons to inflammatory stimuli thereby participating in the hypersensitization of nociceptors.

We next focused our interest on MMP-9 modulation by NOV. We investigated MMP-9 mRNA levels in DRG neuron cultures exposed to TNF-α (1 ng/ml) for one hour prior to NOV addition (1 μg/ml) in time-course experiments. As shown in Figure 7A, addition of NOV decreased the stimulatory effect of TNF-α on MMP-9 mRNA expression. The NOV inhibitory effect on TNF-α-induced MMP-9 was rapid and transient since inhibition peaked at 5 hours with 50% inhibition (***P < 0.001) and declined thereafter. The specificity of this effect obtained with a home-purified recombinant protein (Figure 7B, bac, ***P < 0.001) was confirmed by using commercial human NOV produced in murine cell line (Figure 7B, mu, **P < 0.01). Similar results were obtained with NOV overexpression by infecting DRG cultures with an adenovirus recombinant for NOV (Ad-NOV, *P < 0.05) (Figure 7C). Thus, NOV counteracts the stimulatory action of TNF-α on MMP-9 expression in primary cultured sensory neurons.

So far, NOV has been shown to interact with the integrin dimers α₅β₁, α₆β₁, α_vβ₃ and α_vβ₅ and with Notch receptor 13 and, more recently, to modulate MMP-13 expression through α_vβ₃/α_vβ₅ engagement 23 . Thus, we wondered if NOV inhibitory action on cytokine-stimulated MMP-9 involved integrin receptors. Integrin subunits α_v and β₁ were expressed at the highest level (α_v mRNA level was approximately 3-fold higher than α₅ and α₆ and, β₁ was 50-fold more expressed than β₃ and β₅, data not shown). Thus, the integrins with which NOV interacts are expressed by cultured sensory neurons. As the β subunits in integrins are essential for signal transduction 36 , we blocked β₁, β₃ or β₅ expression using a siRNA approach. Cultured primary sensory neurons transfection with a β₁-, β₃- or β₅-specific siRNA resulted in an 88%, 75% and 85% decrease of β₁, β₃ and β₅ mRNA levels, respectively (data not shown). While β₃ or β₅ depletion did not prevent NOV from inhibiting the induction of MMP-9 by TNF-α (with respective 61% and 39% inhibitions, **P < 0.01, ***P < 0.001), β₁ knockdown significantly abolished this inhibitory effect (Figure 7D). These data indicate that, in cultured primary sensory neurons, the NOV inhibitory action on TNF-α-induced MMP-9 expression requires interaction with integrin β₁.

It has been reported that following nerve injury, MMP-9 is transiently upregulated in injured DRG and participates in neuropathic pain induction, whereas MMP-2 displays a delayed increase in DRG and DHSC and is involved in a later phase of neuropathic pain maintenance 10 . So far, no characterization of MMPs expression in the CFA model of inflammatory pain has been carried out. Thus, we first examined MMP-2 and MMP-9 expression patterns in the DRG and DHSC during CFA-induced inflammatory pain. As shown in Figure 8A and 8B, MMP-9 mRNA levels were transiently upregulated in DRG on days 15 (×3.95 ± 0.55, **P < 0.01, n = 4) and 21 (×2 ± 0.36, *P < 0.05, n = 4) as well as in the DHSC 15 days (×2.8 ± 0.6, *P < 0.05, n = 4) after CFA injection, and returned to levels of sham rats thereafter. Markedly, MMP-2 expression showed no variation in DRG, but a mean 2-fold persistent upregulation in DHSC at late stages (×3.45 ± 0.66 on day 15; ×2 ± 0.66 on day 21 and × 1.94 ± 0.36 on day 60 post-CFA, *P < 0.05, n = 4). Moreover, dexamethasone treatment at late stages of inflammatory pain repressed MMP-2/-9 expression in DRG and DHSC of CFA rats (Figure 8C and 8D, *P < 0.05, **P < 0.01, n = 4). Thus, the induction of MMP-2 and MMP-9 mirrors reduced NOV expression at late stages in the establishment of inflammatory pain (Figure 3A and 3B).

In an acute model of inflammatory pain, carrageenan-induced mechanical allodynia is partially prevented in MMP-9-null mice [ 10 , suppl. data]. To test whether MMP-9 could impact CFA-induced persistent inflammatory pain, the mechanical allodynia of CFA rats treated with an MMP-9 inhibitor (MMP-9i) was evaluated using the von Frey filament test. As shown in Figure 8E, repeated intrathecal injections of MMP-9 inhibitor from day 9 to day 12 post-CFA injection significantly attenuated mechanical allodynia (5.5 ± 0.8 g in CFA rats treated with MMP-9i compared to 2.7 ± 0.3 g in CFA rats injected with vehicle, **P < 0.01, n = 6). Altogether, these data suggest a role for MMP-9 in the maintenance of CFA-induced inflammatory pain.

We further investigated whether inhibition of NOV endogenous expression could modulate MMPs expression in vivo. As shown in Figure 9A, the intrathecal administration of NOV siRNA (siNOV) at early stages of CFA-induced pain (when neither NOV nor MMPs endogenous expressions are altered) resulted in a 40% inhibition of NOV mRNA expression in the DRG (data not shown) and a 58% reduction of NOV protein expression in the DHSC, compared with control siRNA-treated rats (**P < 0.01, n = 6). NOV downregulation led to a significant increase of MMP-9 mRNA levels in DRG of CFA rats (×1.43 ± 0.08, *P < 0.05, n = 6) without affecting MMP-2 expression (Figure 9B). Conversely, in DHSC, where MMP-9 mRNA levels showed no significant variation when NOV was inhibited, MMP-2 mRNA levels were significantly upregulated (×2 ± 0.17, **P < 0.01, n = 6) (Figure 9C). Consistent with NOV knockdown effects on MMPs mRNA in DRG, NOV siRNA delivery specifically increased MMP-9 proteolytic activity (x 1.59 ± 0.10 **P < 0.01, n = 6) and had no effect on MMP-2 activity in CFA rat DRG (Figure 9D). Due to lower MMP-2 and MMP-9 expressions in DHSC compared to DRG (5-fold and 2-fold less mRNA, respectively, data not shown), their gelatinase activities were below the detection threshold. Thus, in CFA-induced pain, NOV downregulation specifically affects MMP-9 and MMP-2 expressions in DRG and spinal cord, respectively.

In order to test whether endogenously produced NOV could modulate inflammatory pain, we evaluated the mechanical allodynia of CFA rats treated with NOV. As shown in Figure 9E, intrathecal delivery of siNOV resulted in a significant increase of mechanical allodynia compared to rats injected with control siRNA (*P < 0.05, n = 8). These data strongly suggest that endogenously produced NOV influences pain intensity and further support the hypothesis that NOV downregulation could participate in pain processes through upregulation of MMP-2 and MMP-9.

We then wondered if NOV treatment of CFA rats at later stages had an impact on MMPs expression and nociceptive thresholds. To this aim, rats were intrathecally injected with NOV protein (3 μg) for 4 consecutive days starting from day 9 after CFA or saline injection. In accordance with MMPs pattern of expression in the CFA-induced pain model, as shown in Figure 10A and 10B, MMP-9 was upregulated both in the DRG and DHSC of CFA rats compared to sham animals, while MMP-2 was only significantly increased in DHSC. NOV treatment significantly abolished MMP-9 upregulation in ipsilateral DRG and DHSC (*P < 0.05, n = 4 to 5). Notably, NOV treatment resulted in significant inhibition of MMP-2 induction in DHSC (*P < 0.05, n = 4 to 5) and had no impact on its expression in DRG (Figure 10A and 10B). NOV administration to sham rats had no effect on MMPs expression. Consistent with NOV action on MMPs mRNA in DRG, NOV treatment specifically reduced MMP-9 proteolytic activity and had no effect on MMP-2 activity in CFA rats DRG (Figure 10C, *P < 0.05, n = 3 to 4). MMP-2 and -9 gelatinase activities were below the detection threshold in DHSC. Thus, in vivo NOV abolishes MMP-2 induction in the dorsal spinal cord and suppresses MMP-9 upregulation in both DRG and DHSC.

We tested whether concomitant with inhibition of MMPs, exogenous NOV could modulate persistent inflammatory pain. As shown in Figure 10D, CFA injection produced long-lasting mechanical stimulation hypersensitivity as reflected by the drastic decrease of withdrawal threshold to von Frey hairs dropping from 13 ± 0.5 g before CFA injection to 3.6 ± 0.3 g at day 9 post-CFA. Importantly, NOV treatment significantly attenuated the tactile allodynia developed in CFA rats (*P < 0.05, **P < 0.01, n = 8) and had no effect on basal pain perception in sham animals. These findings show NOV as an anti-allodynic molecule that is able to reduce established mechanical allodynia in the persistent phase of inflammatory pain and further suggest that NOV could exert its anti-allodynic effect through downregulation of MMP-2 and MMP-9.