To test the hypothesis that TGF-β1 alleviates behavioural signs of neuropathic pain, we first evaluated the effects of TGF-β1 on the development of mechanical allodynia and thermal hyperalgesia in rats following nerve injury. Rats started to receive TGF-β1 or saline infusion on the day of the surgery and treatment lasted 14 days. As illustrated in Fig 1, shortly after partial ligation of the left sciatic nerve, rats receiving saline infusion showed an exaggerated unilateral response to heat stimuli (Fig. 1A) and a sharp bilateral decrease in response to Von Frey hair stimulation at the plantar surface (Fig. 1B), which confirmed a similar observation reported by Seltzer et al (1990). TGF-β1 did not alter paw withdrawal threshold and latency in animals without injury (data not shown). However, central infusion of TGF-β1 almost completely abolished thermal hyperalgesia 14 days after nerve injury (89% reduction with 5 g TGF-μβ1) (Fig. 1A). Paw withdrawal latency in response to heat stimuli was 11.04 ± 0.93 sec (5 μg TGF-β1) and 8.84 ± 0.18 sec (2 μg TGF-β1) vs. 5.42 ± 0.37 sec in animals receiving saline treatment (P < 0.001 vs. saline treated), whereas pre-surgery paw withdrawal latency was 11.51 ± 0.84 sec. Nerve injury-induced mechanical allodynia was reduced by 38% with a central infusion of TGF-β1 (5 μg) for 14 days (Fig. 1B). Paw withdrawal threshold on the ipsilateral side reached 5.57 ± 0.57 g (5 μg TGF-β1) and 5.54 ± 0.44 g (2 μg TGF-β1) vs. 1.51 ± 0.2 g (saline) 13 days following nerve injury (P < 0.001 vs. saline treated) (Fig. 1B). There was no significant difference between 2 μg and 5 μg doses in paw withdrawal threshold and latency. We assume that 2 μg and 5 μg doses have already reached the maximal effect of TGF-β1 on behavioral outcomes.

We also examined the effects of TGF-β1 on already established hypersensitivity following nerve lesion. Implantation of a mini-osmotic pump containing TGF-β1 (2.5 μg) was performed on day 7 post-injury, where both paw withdrawal threshold and latency in response to mechanical and thermal stimuli, respectively, had already reached their lowest level. Intrathecal infusion of TGF-β1 for 7 days produced a 62% reduction in established thermal hyperalgesia (Fig. 2A) and a 28% reversal in mechanical allodynia at ipsilateral sides (Fig. 2B). TGF-β1 also had a slight effect on contralateral side mirror pain (Fig. 2A–B).

We previously reported that peripheral nerve injury (chronic constriction of sciatic nerve with a polyethylene tube) triggers a striking cell proliferation within the spinal cord at the ipsilateral side of the injury. Cell proliferation peaks at day 3 post-injury and persists at high levels at least for 14 days, with most newly generated cells being microglia 7. In this study, using bromodeoxyuridine (BrdU) as an index of DNA synthesis, we again detected a large number of dividing cells in the lumbar spinal cord, ipsilateral to the sciatic nerve with partial ligation (Fig 3A). The majority of BrdU-incorporated cells were microglia, as they are immunopositive for the microglial marker ionized calcium-binding adaptor molecule 1 (Iba-1) (Fig 3B), indicating that they were newly formed microglia. Intrathecal TGF-β1 treatment successfully inhibited spinal microgliosis following nerve injury, as shown by a significant reduction in the number of BrdU⁺ cells at the ipsilateral side of the spinal cord (Fig.3A, 3C), and a reduction in density of total Iba-1+ microglial cells (Fig. 3D–E). No significant differences were observed for two different doses (2 μg and 5 μg) of TGF-β1, which suggests the effect of TGF-β1 on the microglial reaction has reached the limit.

Central TGF-β1 treatment not only inhibited generation of new microglial cells but also prevented activation of existing microglia. In nerve-injured rats receiving saline treatment, there was a strong increase in Iba-1 immunoreactivity (ir) condensed at the ipsilateral side of spinal dorsal horn (DH) and ventral horn (VH) at 14 days post-surgery. This pattern of microglial expression was no longer observed in rats treated with TGF-β1 (Fig. 4A-a). High magnification images revealed that stereotypical changes in microglial morphology during activation were also prevented by TGF-β1. Similar to those in normal animals, Iba-1⁺ cells in the spinal cord of TGF-β1-treated rats with nerve injury had small cell bodies with long and fine ramifications, whereas microglia in animals without TGF-β1 treatment displayed large cell bodies and enlarged branches following nerve injury (Fig. 4A-b). We quantified the intensity of Iba-1 immunofluorescence 14 days after nerve injury in both groups of rats with 2 μg and 5 μg, and without TGF-β1. Both 2 μg and 5 μg TGF-β1 significantly reduced the effects of nerve injury on Iba-1 ir at the ipsilateral sides DH and VH, with the exception of 2 μg TGF-β1 on the ventral horn, where the effect was not statistically significant (Fig. 4B). We also quantitatively evaluated the effects of TGF-β1 on Iba-1 protein levels at day 7 post-injury with Western Blot. Seven days after nerve injury, the quantity of Iba-1 protein at the ipsilateral side was 1.48 times higher than their contralateral counterpart. Infusion of 1 μg TGF-β1 for one week, similar to the paradigm we used for immunohistochemistry and behavioural studies with 2 μg TGF-β1 infusion for two weeks, significantly reduced the difference between ipsi and contralateral dorsal horns to 1.1 times (Fig. 4C–D).

Next, we examined the effects of TGF-β1 on astrocyte activation in the spinal cord following nerve injury. Spinal astrocytes were labelled with an antibody against glial fibrillary acid protein (GFAP). Changes in GFAP ir and in the morphology of astrocytes within the spinal cord after peripheral nerve injury are depicted in Fig 5A. Partial ligation of sciatic nerve dramatically increased GFAP ir and altered the morphology of astrocytes with enlarged branches at the ipsilateral side of the spinal cord (Fig. 5A, top panels). These effects were attenuated by TGF-β1 in a dose-dependent manner (Fig. 5A–B, middle and lower panels). Treatment with TGF-β1 at 2 μg for 14 days reduced the effect of nerve injury on GFAP ir by 13.9% in DH and 12.3% in VH, whereas at 5 μg TGF-β1 decreased the up-regulation of GFAP ir by 43.1% in DH and 49.12% in VH (Fig. 5A–B), showing a higher dose (5 μg) is more effective in astrocytic response.

Partial sciatic nerve ligation induced the expression of activating transcription factor 3 (ATF3) in the nucleus of sensory neurons in the dorsal root ganglia (DRG) and in the nucleus of motor neurons in the spinal cord (Fig. 6A–C). The number of ATF3⁺ neurons in the spinal cord was remarkably reduced by 2-week intrathecal TGF-β1 infusion starting on the day of surgery. Fourteen days after sciatic nerve injury, in the ipsilateral side ventral horn, there were 9.36 ± 0.96 ATF3⁺ cells in the group of rats infused with saline, whereas in rats treated with 2 μg TGF-β1, there were only 3.22 ± 0.15 (2 μg) ATF3⁺ cells. Fourteen days after sciatic nerve injury, there were 9.36 ± 0.96 ATF3⁺ cells on the ipsilateral side ventral horn in saline-treated group, whereas in rats treated with 2 μg TGF-β1, there were only 3.22 ± 0.15 ATF3⁺ cells (Fig. 6D). Although TGF-β1 was introduced intrathecally, expression of ATF3 in the ipsilateral DRG was also affected; 4.2 ± 0.5 ATF3⁺ cells in saline treated rats vs. 1.00 ± 0.09 ATF3⁺ cells in TGF-β1 treated rats (Fig. 6E).

As we have previously reported that damaged ATF3⁺ neurons express chemokine MCP-1 9 which is a trigger for surrounding microglial activation 8, we further investigated the regulation of MCP-1 expression by TGF-β1. Similar to that of ATF3 regulation, both in the spinal cord ventral horn and in the ipsilateral DRG, the number of MCP-1⁺ neurons and the intensity of MCP-1 imunoreactivity was significant lower in rats treated with TGF-β1 than those treated with saline (Fig. 6F).

The role of nerve injury-induced spinal inflammation in the initiation and maintenance of chronic pain has been well documented. To explore whether TGF-β1, a potent anti-inflammatory cytokine, could alter local inflammatory response and thereby explain its potential effect on hypersensitivity, we measured levels of pro-inflammatory cytokines Il-1β, TNF-α and IL-6 transcripts in the spinal cord with quantitative real time-PCR. Relative mRNA levels were normalized with the average of three housekeeping genes (atp5o, Hprt1, G6pdx). Fold changes relative to naïve spinal cord were determined using comparative Ct method 25 and is depicted in Figure 7. Expression levels of IL-1β and IL-6, but not of TNF-α, increased dramatically at the ipsilateral spinal cord following nerve injury. These effects were significantly reduced by TGF-β1 treatment (Fig. 7).

To gain insight into the potential mechanisms of the pleiotropic effects of TGF-β1 in this specific neuropathic condition, we examined the expression of endogenous TGF-β1, the cellular localization of its receptors-TGFβRI/II and the phosphorylation states of receptor-regulated Smad2/3, as well as the expression of the common mediator Smad 4. By using in situ hybridization method, TGF-β1 mRNA was not detected within the spinal cord of animals suffering nerve injury from day 0 (naïve) to day 14 (Fig. 8A). Western Blot analysis revealed that endogenous TGF-β1 protein was not detectable in the spinal cord of rats following nerve injury. However, chronic intrathecal infusion of recombinant TGF-β1 resulted in a stable and similar level of the protein at both dorsal and ventral horns (Fig. 8B). This expression pattern confirms that the pleiotropic cellular effects of TGF-β1 and the subsequent impact on behavioural outcomes are essentially mediated through exogenous TGF-β1, excluding the interference of endogenous TGF-β1.

TGFβRI mRNA was up-regulated in the spinal cord and DRG, ipsilateral to the nerve injury (Fig. 8C). Positive signals (silver grains) for TGFβRI mRNA were observed on the ventral horn motor neurons and DRG sensory neurons (identified based on the position, appearance and size of cells) and on some glial cells in both dorsal and ventral horns, which were heavily stained with thionine (Fig. 8C). TGFβRII mRNA was not detected by in situ hybridization within the spinal cord parenchyma following nerve injury but was found in the plexus and ventricle in the normal brain (data not shown).

The binding of ligand TGF-β1 to a heteromeric receptor complex consisting of TGFβRI and TGFβRII, leads to activation of the Smads signalling pathway, including receptor-regulated Smads (R-Smads)-Smad2/3 and the common mediator (Co-Smad)-Smad 4. This intracellular cascade is critical to the pleiotropic action of TGF-β1 superfamily responses. To determine whether Smad signalling was activated in the spinal cord of rats with nerve injury following TGF-β1 challenge, we measured protein levels of phosphorylated Smad 2/3 and Smad 4 in the spinal cord with Western Blot. Both p-Smad 2/3 (Fig. 8D) and Smad 4 (Fig. 8E) were up-regulated at the ipsilateral spinal cord after TGF-β1 treatment.

Adult male Sprague-Dawley rats (Charles River, Quebec, Canada) were used and weighed 250–275 g at time of surgery. Before surgery, they were acclimatized to standard laboratory conditions (14-h light, 10-h dark cycle) and given free access to rat chow and water. All protocols were performed in accordance with guidelines from the Canadian Council on Animal Care and were approved by the McGill University Animal care Committee.

Rat recombinant TGF-β1 (Peprotech, RockyHill, New Jersey, U.S.A.) was delivered into nerve injured and sham-operated rats with an intrathecal catheter driven by a mini-osmotic pump (Alzet, Palo Alta, CA). The catheter and pump were implanted according to the procedure originally described for chronic catheterization of the rat spinal cord 46. Animals were anaesthetized with a mixture of ketamine/xylazine (100 mg/kg, intraperitoneally). A midline incision was made to expose the atlanto-occipital membrane. The membrane was pierced and the catheter was passed intrathecally to a distance of 8.5 cm, i.e., caudal to the level of the atlanto-occipital junction. The coiled part of the system with the pump attached was then implanted subcutaneously in a pouch to lie behind the shoulders. Rats showing any signs of motor impairment were excluded from the protocol. Alzet osmotic mini-pumps were filled with 0.9% saline or TGF-β1 and primed in a 37°C bath over night. All intrathecal tubings and pumps were verified at the end of experiments during tissue collection. Animals having a defective intrathecal catheterization (either disconnected or clogged) were eliminated from the study.

The thymidine analog bromodeoxyuridine (BrdU) was used to label proliferating cells in the spinal cord after sciatic nerve injury. BrdU was injected intraperitoneally (100 mg/Kg body weight) at day 3 post-injury; the time when peripheral nerve injury-induced cell proliferation within the spinal cord reached its peak 7. Animals were perfused transcardially with 4% paraformaldehyde (PFA) for tissue collection. To allow for detection of cells incorporated with BrdU, free-floating sections were pre-treated with 50% formamide in 2 × SSC for 2 h at 65°C, followed by 15 min in 2 × SSC at room temperature, and 30 min in 2N HCl at 37°C for 10 min in 0.1 M borate buffer at room temperature. A polyclonal goat anti-rat antibody against BrdU (1:250; Accurate Chemicals, Westbury, NY) was incubated with tissue sections for 48 h at 4°C. After primary antibody incubation, sections were incubated in Alexa 488-conjugated goat anti-rat IgG (1:250; Invitrogen, Carlsbad, CA) for 1 h. Sections were mounted onto slides and coverslipped with Vectashield mounting medium (Vector Laboratories, Burlingame, CA).

Regular immunofluorescent staining was performed to characterize the spinal glial cell reaction to peripheral nerve injury and to TGF-β1 chronic infusion. Free-floating sections were incubated overnight at 4°C with the following antibodies: rabbit anti-ionized calcium-binding adaptor molecule 1 (Iba-1) polyclonal antibody (for microglia and macrophages, 1:1000; Wako Chemicals, Richmond, VA); and rabbit anti-glial fibrillary acid protein (GFAP) polyclonal antibody (for astrocytes, 1:1000; DakoCytomation, Carpinteria, CA); followed by a 60-min incubation at room temperature in fluorochrome-conjugated goat secondary antibody. 4',6-Diamidino-2-phenylindole dihydrochloride (Dapi) was also used as a nuclear counterstain (1:10000, Sigma).

Some spinal cord and DRG sections were processed by avidin-biotin method using peroxidase as a substrate to detect damaged neurons labelled by antibodies against activating transcription factor 3 (ATF3) and monocyte chemoattractant protein-1 (MCP-1). Briefly, sections were rinsed with TBS and incubated with goat anti-rabbit ATF3 antibody (1:500, Santa Cruz biotechnology, USA,) or goat anti-rabbit MCP-1 antibody (1:2000, Peprotech, RockyHill, New Jersey, U.S.A) over night at 4°C and incubated sequentially with a biotinylated secondary antibody (Vector Laboratories, Canada), followed by an avidin-biotin-peroxidase complex (Vectostain ABC Elite Kit, Vector Laboratories). After several washes in TBS, tissue sections were reacted in 0.05% diaminobenzidine and 0.003% hydrogen peroxide and some sections were counterstained with thionin.

Detection of mRNAs encoding TGF-β1, TGFβRI and TGFβRII was performed on lumbar spinal cord and DRG sections using 35S-labeled riboprobes. Riboprobe synthesis and hybridization were performed as per a previously described protocol 47. Briefly, plasmids were linearized and sense and anti-sense cRNA probes were synthesized with appropriate RNA polymerase as described in Table 1. Sections were postfixed in 4% PFA and digested by proteinase K (10 μg/ml, at 37°C for 20 min), after which spinal cord sections were rinsed in water and by a solution of 0.1 M triethanolamine (TEA, pH 8.0), acetylated in 0.25% acetic anhydride in 0.1 M TEA and then dehydrated. Hybridization of the sections by riboprobe involved a 10⁶ cpm/ml/slide of hybridization mixture and incubation at 55°C overnight in a slide warmer. Slides were rinsed in standard saline citrate (1 × SSC: 0.15 M NaCl, 15 mM trisodium citrate buffer, pH 7.0) and digested by RNase A at 37°C (20 μg/ml), rinsed in descending concentrations of SSC, and dehydrated through graded concentrations of ethanol. Sections were exposed to x-ray film (BioMax, Kodak, Rochester, NY) for 2–3 days and dipped in NTB2 nuclear emulsion (diluted 1:1 with distilled water, Kodak). Slides were kept at 4°C for 3–5 weeks safe from light and developed in D19 developer (Kodak).

Tissues were homogenized in a lysis buffer containing a cocktail of proteinase and phosphatase inhibitors (Roche, Indianapolis, USA). Protein concentrations were determined by BCA Protein Assay (Pierce) and 15 μg of proteins were loaded and separated on SDS-PAGE gel (12%). After the transfer, blots were incubated overnight at 4°C with the following antibodies as per manufacturer's instructions: goat anti-p-Smad2/3 (1:500, Santa Cruz); goat anti-Smad4 (1:250, Santa Cruz); rabbit anti-Iba-1 (1:500, Wako); and goat anti-TGF-β1 (1:500, Santa Cruz). For loading control, blots were probed with β-actin antibody (1:10000, Sigma). Density of specific bands from Western Blotting was quantified with a computer-assisted imaging analysis system (Image Pro Plus).

Total RNA was extracted from spinal cords using Qiazol (Invitrogen) followed by digestion with RQ1 RNase-Free DNase RNeasy kit (Qiagen) to remove DNA contamination. Synthesis of cDNA from total RNA was performed with SuperScript II Reverse Transcriptase (Invitrogen). Primers (see Table 2) were chosen using the GeneTools software (Biotools Inc.) and their specificity was verified using the GenBank database from the NCBI website. Nineteen samples were analysed (5 TGF-β1-treated animals; 3 saline-treated animals; 3 naive animals). Experiments were performed in duplicate using the LightCycler 480 Real-Time PCR Detection System (Roche Diagnostics). Levels of target mRNAs were normalized to the average of the three housekeeping genes; Atp50, Hpt1 and G6pdx. Fold changes versus naive animals in their respective ipsi and contralateral sides were analyzed using the comparative Ct (dCT) method 25.

Images were acquired using either an Olympus BX51 (Tokyo, Japan) microscope equipped with a colour digital camera (Olympus DP71) or an Olympus confocal laser-scanning biological microscope (Fluoview 1000). Colocalization was ensured with confocal Z stacks at 0.8 μm intervals and visualization in three-dimensional orthogonal planes. Quantitative analysis of the immunofluorescence intensity was performed on images digitized using a constant set of parameters (exposure time, gain, and post-image processing) with special attention to avoid signal saturation. We measured the intensity of Iba-1 and GFAP immunofluorescence as the average fluorescence intensity within four areas of interest (AOI) defined as: 2 rectangles (100 × 300 μm/each) on the dorsal horn (DH) (lamina I–IV) and 2 rectangles (261 × 370 μm/each) on the ventral horn (VH) (lamina IX), on both sides relative to the side of injury. Fluorescence intensity was measured using Image Pro Plus 6.2 for Windows (Media Cybernetics Inc.). The total number of BrdU⁺ cells and Iba-1⁺ microglial cells was counted by blinded investigators in four different regions: ipsilateral DH (DHi), contralateral DH (DHc), VHi, and VHc. Only uniformly BrdU⁺-labelled nuclei and Iba-1⁺ cells showing a positive nucleus stained with Dapi were considered for quantification. Samples from 4–6 sections per rat and 6 rats per group were included for each quantitative analysis.

All data are presented as means ± SEM. Statistic significance was determined using: 1) for behavioural analysis in Fig 1 and Fig 2, one way ANOVA followed by Dunnet's test for the changes of all time points vs. pre-surgery baseline; unpaired t-test for the difference between groups (TGF-β1-treated vs. saline-treated at each time point); 2) paired t-test for the difference between ipsi- and contra-lateral sides within the same group; and 3) unpaired t-test for the difference between groups (TGF-β1-treated vs. saline-treated in their respectively DH and VH region. The criterion for statistical significance was p < 0.05.