All procedures and behavioral experiments involving animals were approved by the Institutional Animal Care and Use Committee at the University of North Carolina at Chapel Hill. Mice were raised under a 12:12 light:dark cycle and fed Prolab RMH (LabDiet) ad libitum. All mice were backcrossed to C57BL6/J for greater than ten generations. Young adult mice between postnatal day (P)60 to P100 were used in all experiments.

Floxed-COX2 mice were generated as previously described (Yu et al., 2006). Briefly, floxed-COX2 mice were generated by inserting loxP sites, using homologous recombination, into introns 5 and 8 to enable deletion of exons 6–8. The insertion of the two loxP sites did not affect normal COX2 expression and induction (Yu et al., 2006; Vardeh et al., 2009). GFAP-CreERT2 mice were generated previously in our lab (Casper et al., 2007). Briefly, the human 2.2 kB GFAP promoter from pGfa2lac1 (M. Brenner, UAB) was cloned upstream of intron from β-globin. CreERT2 was excised from pCreERT2 (P. Chambon, Collège de France) and cloned behind β-globin and upstream of SV40 pA. Heterozygous GFAP-CreERT mice were bred to homozygous floxed-COX2 mice to produce GFAP-CreERT::floxed-COX2 mice (experimental group) and Cre-negative, floxed-COX2 mice (littermate controls). The GFAP-CreERT::floxed COX2 mice were induced with a course of 16 tamoxifen injections (3 mg/40 g), twice daily I.P., starting at post-natal day (p) 21. Littermate mice were given tamoxifen to control for the potential effect of tamoxifen injections as well.

Another transgenic mouse line, the tetO-IKKdn mice were generated as previously described (Herrmann et al., 2005). Briefly, a transdominant version of human IKK2 (VSV-IKK2-DN; D145N) was cloned into the bidirectional tetracycline-dependent expression vector pBl-5. tetO-IKKdn mice were crossed with mice in which the tTA is driven by a 2.2 kb fragment of the human GFAP promoter (Casper et al., 2007). The GFAP-tTA::tetO-IKK2dn mice (IKKdn mice) were maintained on doxycycline (25 μg/mL in the drinking water) to suppress transgene expression throughout the central and peripheral nervous system while in utero and until p21. At weaning, the mice were taken off doxycycline and given drinking water either without doxycycline (to allow IKKdn expression both central and peripheral GFAP-positive cells) or with Oxy (200 mg/mL) to suppress IKKdn expression only in peripheral GFAP-positive glia. The mice were maintained on this treatment for at least 30 days and until experimental use.

Neuropathic pain was generated using a model of spared nerve injury (SNI) adapted from Dr. Allan Basbaum’s laboratory (Shields et al., 2003). Using isoflurane, mice at least 60 days old were deeply anesthetized, as monitored by toe pinch and eye blink reflexes. The left hind leg was secured with tape, the hair removed with Nair, and the leg was cleaned with iodine and alcohol. The three divisions of the sciatic nerve were exposed. The common peroneal and sural nerves were ligated, and a 1–3 mm portion of the nerve was removed distal to the ligation to prevent regrowth. The musculature was returned to its original position, and the skin closed with wound clips. The area was again cleaned with iodine and alcohol, and anti-bacterial and anti-fungal agents were applied. The animals were housed in cages with soft bedding.

Animals were singly placed in a Plexiglas enclosure on top of a metal grate. Paw withdrawal frequency was used to measure mechanical allodynia in animal hind paws. A Von Frey hair (IITC Life Science, Woodland Hills, CA, United States) calibrated to 3.61 log force was applied to the center of the foot pad (the region innervated by the tibial nerve) ten times at an interval of at least 1 min between applications. Both hind paws were tested; the mechanical sensitivity of the contralateral paw to the surgery side was used as an internal control for the sensitivity of the ipsilateral paw. The number of times the paw was withdrawn due to the stimulus was recorded. The paw withdrawal threshold was measured using the up-down method of von Frey hair application (Chaplan et al., 1994). Mechanical sensitivity was measured for 2 days prior to surgery as the baseline of hind paw mechanical sensitivity. Von Frey tests were repeated post-SNI at 4 days, 7 days, and then every subsequent week through 3 weeks.

Animals were deeply anesthetized with an intraperitoneal injection of urethane (2 mg/kg). Minimally five animals were used in each experimental group as biological replicates. Once deeply anesthetized, as monitored by the toe-pinch and eye-blink reflexes, the animals were transcardially perfused with 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS). Lumbar segments 4–6 (L4–L6) of the spinal cord, the corresponding dorsal root ganglia (DRG) and a 5–8 mm segment of the tibial nerve (at the level of the lesion in the other branches) were removed and post-fixed in ice-cold 4% PFA for 2 h. The tissue was cryoprotected with 30% sucrose in PBS and subsequently frozen in OCT. Fourteen micrometer (14 μm) thick sections were made on a cryostat (Leica, Wetzlar, Germany). Immunohistochemistry was performed with primary antibodies against mouse GFAP (1:500; Dako), chicken GFAP (1:500; Abcam), rabbit Iba-1 (1:500; Wako), mouse NeuN (1:500; Millipore), rabbit peripherin (1:1000; Millipore), rabbit neurofilament 200 (1:500; Sigma), mouse S100β (1:1000; Sigma), rabbit pNFκB (1:400; Cell Signaling) and biotinylated IB4 lectin (1:100; Vector). Secondary antibodies were either anti-mouse, anti-rabbit, or anti-chicken Alexa 568, Alexa 488, Alexa 647, or avidin-conjugated Alexa 568. The slides were cover-slipped with Vectashield containing DAPI. Images were collected on a Zeiss Axioscope fluorescent microscope with a Dage XL16 camera or on a Zeiss LSM 710 confocal microscope. Quantification of immunopositivity was performed manually.

All data are presented as the mean ± SEM. Significant differences were assessed by Student’s t-test or, in the case of behavior assays, a two-way repeated measures analysis of variance followed by a Bonferroni post hoc analysis. Indicators of significance were as follows: P < 0.05 (*), P < 0.01 (^**), and P < 0.001 (^***).

Prostaglandin E2 has been shown to sensitize nociceptive neurons (Ferreira, 1972; Ferreira et al., 1978). In this experiment, we used tamoxifen inducible GFAP-CreERT2 mice to drive the removal of the floxed COX2 gene specifically in GFAP-positive glia. Tamoxifen injections were given post-weaning to avoid potential developmental confounds of transgene expression. To control for possible effects of tamoxifen, a subset of Cre-negative littermate controls was also injected with tamoxifen along with the experimental group.

To assess the role of COX2 in GFAP-positive glial cells on pain behavior, we measured mechanical allodynia and paw withdrawal threshold for 2 days prior to SNI and for 3 weeks following peripheral nerve injury in COX2 cKO, littermate controls, and tamoxifen (Tam)-injected littermate controls (male COX2 cKO n = 10, female COX2 cKO n = 10, male controls n = 10, female controls n = 10, male control + Tam n = 10, female control + Tam n = 10). As shown in Figure 1, prior to SNI, no significant differences in mechanical sensitivity at the baseline were detected among different genotypes. At 4 days post-SNI, both male and female animals developed mechanical allodynia in the paw ipsilateral to the lesion, but not in the contralateral paw, as indicated by increased paw withdrawal frequency and decreased withdrawal threshold. In littermate control and tamoxifen injected littermate control animals, mechanical sensitivity of the ipsilateral paw was maintained through the 3-week testing period. By 2 weeks post-injury, both male and female COX2 cKO mice showed abrupt and profound reductions in mechanical sensitivity (P < 0.001 compared to the ipsilateral paw of the littermate control animals). No significant impact of tamoxifen injections alone was observed on the littermate control animals at any time point. These data suggested that COX2 in GFAP-positive glia contributes to the maintenance of neuropathic pain following peripheral nerve injury.

We next tested the hypothesis that inhibiting NFκB signaling may suppress neuropathic pain in vivo. In its inactive form, NFκB is sequestered in the cytoplasm by binding to inhibitory factors of the IκB family. Upon stimulation by inflammatory mediators, IκB is phosphorylated by the IκB kinase (IKKβ) complex, leading to its ubiquitination and degradation, which allows NFκB to translocate to the nucleus. In the nucleus, NFκB promotes the transcription of cytokines, chemokines and proinflammatory molecules. To suppress NFκB signaling, we used the tet-Off system driven by the hGFAP promoter to induce glial-specific expression of a dominant negative form of IKK. To characterize the potential expression in the GFAP-positive glia, GFAP-tTA mice were first bred to a tetO-eGFP reporter line of mice. We observed strong eGFP expression in the dorsal horn, which colocalized with the astrocyte marker GFAP, but not with markers for microglia (Iba-1) or neurons (NeuN) (Figure 2). In the sciatic nerve, we observed strong eGFP expression in GFAP-positive NMSCs. eGFP did not colocalize with markers for macrophages (Iba-1), large diameter axons (NF200) or small diameter axons (peripherin) (Figure 2). Surprisingly, there was no eGFP expression in GFAP-positive SGCs in the dorsal root ganglia (DRG) nor in any other cells of the DRG. Immunohistochemistry was also performed 2 weeks post-SNI to detect potential changes in eGFP localization after nerve lesion and gliosis; however, the expression pattern of eGFP reveal the same pattern in GFAP-positive spinal astrocytes and NMSCs, but not in SGCs (Figure 2).

To investigate the necessity of NFκB signaling in GFAP-positive glia on pain behavior, we measured mechanical allodynia and paw withdrawal threshold for 2 days prior to SNI, and for 3 weeks following peripheral nerve injury in IKKdn and IKKdn-negative littermate control animals (male IKKdn n = 6, female IKKdn n = 5, male littermate control n = 9, female littermate control n = 9). There was no difference in basal mechanical sensitivity in the male or female groups (Figure 3). Following SNI, all animals developed mechanical allodynia ipsilateral to the nerve lesion, as shown by increased paw withdrawal frequency and decreased withdrawal threshold. In control animals, mechanical sensitivity of the ipsilateral paw was maintained through the 3-week testing period. In IKKdn-positive animals, pain behavior began to subside 1 week post-injury (P < 0.001) and was alleviated by 2 weeks post-injury in both the male and female IKKdn mice (P < 0.001). No change was observed in the contralateral paws of the IKKdn mice or littermate controls.

Persistent pain is associated with ongoing neural plasticity in both primary sensory neurons and spinal dorsal horn neurons. Neurons and glial cells in both peripheral and central sensory nervous system contribute to sensitization underlying persistent pain. Likewise, using the GFAP promoter, our system drives the expression of transgenes in GFAP-positive glia in both CNS (astrocytes) and PNS (NMSCs). The tet-Off system is typically regulated by the tetracycline derivative doxycycline, which, similar to tetracycline, binds the tTA, preventing it from binding to the tetO promoter and blocking transcription of the downstream transgene. Doxycycline is highly bioavailable, and easily crosses the blood-brain barrier. To tease apart the central and peripheral roles of GFAP-positive glia, we used Oxy, a tetracycline derivative that does not cross the blood-brain barrier (Barza et al., 1975) to block the transcription of transgene in peripheral NMSCs but not in central astrocytes.

The GFAP-tTA::tetO-eGFP reporter mice were administered Oxy in their drinking water from weaning until p60. At p60, they were sacrificed to examine transgenic protein expression using immunohistochemistry. We found that Oxy effectively inhibited transgenic eGFP expression peripherally while leaving central expression intact (Figure 4) (n = 3 per group). There was no eGFP detected in the sciatic nerve either in naïve animals or in animals 2 weeks post-SNI. In the spinal cord, eGFP remained colocalized with the astrocyte marker GFAP in both naïve and nerve-injured animals. eGFP was not colocalized with any markers for microglia or macrophages (Iba-1), neurons (NeuN) or large or small peripheral nerves (NF200 or peripherin).

The permeability of the blood-brain barrier increases for 1 week post-nerve injury (Beggs et al., 2010). To determine whether Oxy entered the spinal cord in a great enough quantity to regulate transgene expression, we performed time-course of immunohistochemical staining following SNI (Figure 5). Adult animals on Oxy for at least 1 month prior to surgery underwent SNI and were sacrificed at 4 days, 2 weeks, or 6 weeks post-injury (n = 3 per time point). eGFP expression was suppressed in all sciatic nerve sections, while its central expression in the spinal cord was preserved at every time point (Figure 5). These findings indicate that Oxy was not able to cross the blood-brain barrier in sufficient concentration to block gene expression post-nerve injury during the experimental period.

Next, we evaluated the ability of Oxy to regulate the expression of IKKdn by examining the downstream marker pNFκB. If effective, Oxy should peripherally inhibit IKKdn expression, thus allowing pNFκB activation in NMSCs of the sciatic nerve. Activated phosphorylated NFκB is localized to the cell nucleus, whereas the non-myelinating Schwann cell marker GFAP is often localized to cellular processes, making colocalization of these two markers difficult. S100β is a marker for both myelinating and non-myelinating Schwann cells and fills the cytosol of the cell, allowing for colocalization with nuclear markers. We, therefore, performed immunohistochemical staining with the markers for GFAP and S100β in conjunction with the nuclear marker DAPI and the pNFκB antibody. Only cells positive for both GFAP and S100β were counted as NMSCs, and only cells with pNFκB localized in the nucleus were counted as pNFκB positive.

Quantification was performed in transgenic animals and littermate controls, both on and off Oxy. Prior to nerve injury, animals from all treatment groups displayed little pNFκB staining and very few pNFκB positive NMSCs (cells containing all four colocalized markers) (Figure 6B) (n = 3 animals per group, three sections per animal). To drive the expression of pNFκB, we performed SNI and sacrificed the animals 4 days post-injury. Figure 6A shows representative sections from all four treatments at the 4 days post-SNI time point. Following injury, littermate control animals showed increases in nuclear pNFκB staining in NMSCs both in the drug-free and Oxy groups (P < 0.05). In transgenic animals, expression of IKKdn completely prevented the SNI-induced increase in pNFκB-positive NMSCs observed in littermate control animals. This transgenic suppression of NFκB activation in the sciatic nerve was eliminated by Oxy administration. IKKdn animals on Oxy showed significantly higher numbers of pNFκB positive NMSCs compared to IKKdn animals without Oxy (P < 0.001).

These data showed that Oxy effectively inhibited peripheral transgene expression while leaving CNS glial expression intact. The peripheral expression of IKKdn suppressed NFκB signaling in NMSCs. The administration of Oxy inhibited peripheral expression of IKKdn, allowing NFκB signaling to occur. Therefore, in IKKdn animals on Oxy, IKKdn should only inhibit NFκB signaling centrally in astrocytes.

To tease apart the role of NFκB in astrocytes vs. peripheral NMSCs following nerve injury, we measured mechanical sensitivity following SNI in IKKdn and littermate controls both on and off Oxy (IKKdn Oxy n = 10, IKKdn Off n = 10, littermate controls Oxy n = 9, littermate controls Off n = 8). Male and female mice were tested separately, but the data were pooled because there were no significant sex differences in the previous cohort. We measured mechanical allodynia and paw withdrawal threshold for 2 days prior to SNI and for 3 weeks following peripheral nerve injury. There was no difference in basal mechanical sensitivity in the “off drug” or Oxy groups (Figure 7). Following SNI, all animals developed mechanical allodynia as assessed by paw withdrawal frequency and threshold. In littermate control animals, the mechanical impairment of the ipsilateral paw was maintained through the 3-week testing period. Transgenic IKKdn animals showed alleviation of mechanical sensitivity starting at 1 week following nerve injury (P < 0.001). Interestingly, this effect was completely reversed in transgenic animals administered Oxy, which blocks transgene expression peripherally but not centrally. Transgenic animals on Oxy exhibited the same development of mechanical sensitivity as littermate animals following SNI.

In summary, IKKdn suppressed NFκB signaling in both astrocytes and NMSCs in the transgenic mice without Oxy. This resulted in the alleviation of pain behavior following SNI. Oxy inhibited peripheral NMSC expression of IKKdn but left intact central astrocytic expression of IKKdn (Figure 4). As a result, in transgenic mice on Oxy, NFκB signaling is only suppressed centrally in astrocytes. Our data showed that the suppression of NFκB translocation and signaling in nuclei in astrocytes was not sufficient to relieve pain behavior.