Toll-Like Receptor 4 Signaling Contributes to Paclitaxel-Induced Peripheral Neuropathy - ScienceDirect

Paclitaxel is the front-line chemotherapeutic agent used to treat many of the most common solid tumors, including those of the breast, ovary, and lung.³⁴ Peripheral neuropathy is the major dose-limiting side effect of paclitaxel and can force dose reduction or even discontinuation of therapy, thus affecting survival in cancer patients.²⁰ In addition, chemotherapy-induced peripheral neuropathy (CIPN) often persists long after cancer treatment is completed and is commonly refractory to current treatment strategies, thus affecting rehabilitation, the return to productivity, and quality of life in cancer survivors.9, 10 Neuropathic pain in general is considered to involve an important role of glial cells and proinflammatory immune responses in the underlying basic pathophysiology,⁶² and evidence implicates similar mechanisms in CIPN.8, 7 Of special interest in this regard is the observation that paclitaxel engages the same signaling pathway via toll-like receptor 4 (TLR4) as the very well-known proinflammatory agent lipopolysaccharide (LPS).¹² Paclitaxel binds to and activates TLR4 in macrophages resulting in activation of the nuclear factor-kappa B (NF-κB) signal path and the induction of proinflammatory cytokine expression identical to that produced by LPS.¹²

TLR4 is expressed on the surface of innate immune cells, small primary afferent neurons,⁶³ and central nervous system cells, including microglia and astrocytes.¹¹ Thus, there is a plausible link between TLR4 and the production of proinflammatory cytokines in neural tissue, which could contribute to behavioral hypersensitivity following exposure to chemotherapy drugs. A previous study indicated that TLR4 in the central nervous system plays a role in the development of behavioral hypersensitivity in a rodent model of neuropathic pain.⁶⁰ Rats that lacked functional TLR4 and those that received intrathecally administrated TLR4 antisense oligonucleotides showed attenuated nerve injury–induced behavioral hypersensitivity in both the initiation and maintenance phases.⁶ There are no data at present regarding the potential role of TLR4 signaling in paclitaxel-induced neuropathic pain. This report explores the effects of paclitaxel chemotherapy on the expression of TLR4 in neural tissue and the effects of antagonists to TLR4 and its immediate downstream signals, myeloid differentiation primary response gene 88 (MyD88) and toll/interleukin 1 receptor domain–containing adapter-inducing interferon-β (TRIF), in reducing paclitaxel-induced behavioral hypersensitivity.

Methods

Animals

Male Sprague-Dawley rats weighing 250 to 300 g (Harlan, Houston, TX) were used to establish the neuropathic pain model. Rats were housed in temperature- and light-controlled (12-hour light/dark cycle) conditions with food and water available ad libitum. All 169 rats used in the study were included in the behavioral analysis portions and then used in follow-up pharmacologic, immunohistochemistry, or Western blot analysis. The numbers of rats in each of these studies are detailed in the relevant sections. All experimental protocols were approved by the Institutional Animal Care and Use Committee at the University of Texas MD Anderson Cancer Center and were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. Every procedure was designed to minimize discomfort to the animals and to use the fewest animals needed for statistical analysis.

Paclitaxel-Induced Neuropathy Model

Rats were treated with paclitaxel (TEVA Pharmaceuticals, Inc, North Wales, PA) as previously described⁶⁹ based on the protocol of Polomano et al.⁴⁸ In brief, pharmaceutical-grade Taxol was diluted with sterile saline from the original stock concentration of 6 mg/mL (in 1:1 Cremophor EL:ethanol) to 1 mg/mL and given at a dosage of 2 mg/kg intraperitoneally every other day for a total of 4 injections (days 1, 3, 5, and 7), resulting in a final cumulative dose of 8 mg/kg. Control animals received an equivalent volume of the vehicle only, which consisted of equal amounts of Cremophor EL and ethanol diluted with saline to reach a concentration of vehicle similar to the paclitaxel concentration. No abnormal spontaneous behavioral changes were noted during or after paclitaxel or vehicle treatment.

TLR4 and MyD88 Antagonist Administration

To assess the role of the TLR4–MyD88 signaling pathway in maintaining paclitaxel-induced neuropathic pain, 20 μg of the TLR4 antagonist LPS derived from Rhodobacter sphaeroides (LPS-RS) in 20 μL phosphate-buffered saline (PBS; InvivoGen, San Diego, CA) or 500 μM of MyD88 homodimerization inhibitory peptide (MIP) in PBS (Imgenex, San Diego, CA) was injected intrathecally by L5 puncture at day 14 following confirmation of paclitaxel-induced mechanical hypersensitivity. The rats were briefly anesthetized with 3% isoflurane and flexed over a tube and a 27-gauge needle inserted between the L5-S1 vertebrae, with a deflection of the tail indicating entry to the subarachnoid space. The dose of LPS-RS was chosen based on previously published studies, whereas the dose of MIP was based on the results of pilot studies wherein rats that received a dose of 100 μM MIP showed no effect on paclitaxel CIPN and rats injected with 1 mM MIP showed pronounced motor impairment. PBS (20 μL) and 500 μM MyD88 control peptide (CP, also in 20 μL PBS) were used separately as controls. To test whether LPS-RS may have an effect in preventing paclitaxel-induced CIPN, rats were treated with LPS-RS beginning 2 days before and then daily through day 2 after paclitaxel treatment. It was not possible to test the role of TRIF signaling in maintaining paclitaxel-induced neuropathic pain because there is no inhibitor available.

Mechanical Withdrawal Test

Mechanical withdrawal threshold was tested before, during, and following paclitaxel treatment by an experimenter (Y.L.) blinded to treatment groups during the mid-light hours (10 am–2 pm). The 50% paw withdrawal threshold in response to a series of 8 von Frey hairs (.41–15.10 g) was examined by the up–down method, as described previously beginning with a filament with a bending force of 2.0 g.¹⁹ Animals were placed under clear acrylic cages atop a wire mesh floor. The filaments were applied to the paw just below the pads with no acceleration at a force just sufficient to produce a bend and held for 6 to 8 s. A quick flick or full withdrawal was considered a response.

Rotarod Test

Rotarod performance was evaluated in the rats that were treated with intrathecal drugs to ensure a lack of treatment effects on motor function. Briefly, the rats were trained on the rotarod apparatus for 3 days prior to any intrathecal drug applications. Acceleration of the rotarod was set to increase from 0 to 40 r.p.m. over 5 minutes. Each rat was tested 3 times at 5-minute intervals, and the average of the latency to drop from the rod for all trials was recorded. The rats were again tested before and then 3 hours after treatment with MIP or CP.

Immunohistochemical Analysis

Rats were deeply anesthetized with sodium pentobarbital (Nembutal, 100 mg/kg, intraperitoneally; Lundbeck, Inc, Deerfield, IL) and perfused through the ascending aorta with warm saline followed by cold 4% paraformaldehyde in .1 M PBS. The L4 and L5 dorsal root ganglia (DRGs) were removed, fixed in 4% paraformaldehyde for 6 hours, and then cryoprotected in 30% sucrose solution. The L4 and L5 spinal cord segments were also removed, fixed in 4% paraformaldehyde for 12 hours, and then cryoprotected in 30% sucrose solution. Transverse spinal cord sections (15 μm) and longitudinal DRG sections (8 μm) were cut in a cryostat, mounted on gelatin-coated glass slides (Southern Biotech, Birmingham, AL), and processed for immunofluorescent staining. After blocking in 5% normal donkey serum and .2% Triton X-100 in PBS for 1 hour at room temperature, the sections were incubated overnight at 4°C in 1% normal donkey serum and .2% Triton X-100 in PBS containing primary antibodies against the following targets: TLR4 (rabbit anti-rat, 1:200; Abcam, Cambridge, MA), MyD88 (rabbit anti-rat, 1:500; Abcam), TRIF (rabbit anti-rat, 1:200 Abcam), glial fibrillary acidic protein (GFAP, mouse anti-rat, 1:1,000; Cell Signaling Technology, Beverly, MA), OX-42 (mouse anti-rat, 1:1,000; Serotec, Raleigh, NC), NeuN (mouse anti-rat, 1:1,000; Millipore, Billerica, MA), IB4 (1:1,000 BS [Bandeiraea simplicifolia]–isolectin B4 fluorescein isothiocyanate conjugate; Sigma-Aldrich, St. Louis, MO), and calcitonin gene-related peptide (CGRP, mouse anti-rat, 1:1,000; Abcam). After washing, the sections were incubated with Cy3-, Cy5-, or fluorescein isothiocyanate–conjugated secondary antibodies overnight at 4°C. Sections were viewed under a fluorescence microscope (Eclipse E600; Nikon, Tokyo, Japan). For a given experiment, all images were taken using identical acquisition parameters and experimenters (A.K.K., A.B.J.) blinded to treatment groups. To measure cell size, each individual neuron, including the nuclear region, was graphically highlighted. For each positive staining of TLR4, MyD88, and TRIF with IB4 or CGRP colocalization, the numbers of total and positive neurons from 3 sections of DRG of 3 rats were counted; data from 3 sections of the same rat were averaged and then mean values used for Mann-Whitney U test comparisons. The percentages of positive neurons to total neurons were calculated and statistically analyzed. To determine whether intrathecal injection could reach and affect signaling in both the DRG and the spinal cord level, 20 μL of Alexa488-labeled mismatch TLR4 oligonucleotide (Invitrogen, Carlsbad, CA) was injected by lumbar puncture and the tissues removed, sectioned, and mounted as described above. All images were analyzed using NIC Elements imaging software (Nikon).

Western Blot Analysis

L4 and L5 DRGs and L4–L5 spinal cord segments were collected from rats that were deeply anesthetized with sodium pentobarbital (Nembutal, 100 mg/kg, intraperitoneally). The samples were snap-frozen in liquid nitrogen. Tissues were disrupted in radioimmunoprecipitation assay lysis buffer (20 mM Tris–HCl, 150 mM NaCl, 1 mM disodium ethylenediaminetetraacetate [Na₂EDTA], 1 mM ethylene glycol tetraacetic acid [EGTA], 1% NP-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na₃VO₄, and 1 μg/mL leupeptin) mixed with 1 mM dithiothreitol, protease inhibitor cocktail (P8340; Sigma-Aldrich), and phosphatase inhibitor cocktails (P0044 and P5726; Sigma-Aldrich) on ice. The supernatant was then collected and denatured with sample buffer (X5) consisting of .25 M Tris–HCl, 52% glycerol, 6% sodium dodecyl sulfate, 5% β-mercaptoethanol, and .1% bromophenol blue for 10 minutes at 70°C. Lysates (total protein: 20 μg) were separated on sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels and transferred to polyvinylidene fluoride membranes (Bio-Rad, Hercules, CA). After blocking with 5% fat-free milk in Tris-buffered saline with Tween (TBST; 137 mM sodium chloride, 20 mM Tris, .1% Tween-20) for 1 hour at room temperature, membranes were then incubated with different antibodies: TLR4 (rabbit anti-rat, 1:1,000; Proteintech, Chicago, IL), MyD88 (rabbit anti-rat, 1:1,000; Imgenex), TRIF (rabbit anti-rat, 1:4,000; Abcam), and β-actin (mouse anti-rat, 1:10,000; Sigma-Aldrich) in 5% fat-free milk in TBST overnight at 4°C. After being washed with TBST, the membranes were incubated with goat anti-rabbit antibody (labeled with horseradish peroxidase; Calbiochem, San Diego, CA) diluted with 5% fat-free milk in TBST for 1 hour at room temperature, and TLR4, MyD88, and TRIF were detected with enhanced chemiluminescence reagents (GE Healthcare, Little Chalfont, UK). The blots were scanned with Spot Advanced and Adobe Photoshop 8.0 (Adobe, Inc, San Jose, CA), and the band densities were detected and compared using ImageJ (NIH, Bethesda, MD). The data from 3 rats per treatment group were averaged for group comparisons.

Data Analysis

Data were expressed as mean ± standard error of the mean and analyzed with GraphPad Prism 5. Behavioral data were analyzed with 2-way analyses of variance (treatment × time) followed by a Bonferroni post hoc test. The cell counts for immune-positive neurons and the percentages of neurons that colocalized with IB4 or CGRP were analyzed using 1-way analysis of variance. Western blot data were analyzed using the Mann-Whitney U test. P <.05 was considered statistically significant.

Results

Changes in Expression of TLR4, MyD88, and TRIF in Rats With Paclitaxel CIPN

Rats treated with paclitaxel showed a decrease in mechanical withdrawal threshold by day 1 following treatment that became more pronounced over time, achieving a significant difference from baseline by day 7 of treatment (Fig 1A). The decrease in threshold became maximal at day 14, and this was sustained through day 21. In contrast, although rats treated with vehicle showed a small decrease in withdrawal threshold at day 1, this returned to the baseline level by day 3 and remained stable throughout the experiment (Fig 1A).

Figure 1. TLR4, MyD88, and TRIF are increased in the DRG in paclitaxel CIPN. The scatter and line plots in A show the mean (and standard error) mechanical withdrawal threshold (in grams) for vehicle-treated (n = 16) (open circles) and paclitaxel-treated (n = 16) (filled circles) rats. A decrease in withdrawal threshold was observed at day 1 after chemotherapy that became more pronounced and significantly different from that in the vehicle-treated group by day 7. Withdrawal threshold remained significantly lower than in the vehicle group over the remainder of the time frame observed. The representative Western blot images shown in B, C, and D illustrate that the expression of TLR4 (B) and MyD88 (C) increased by day 1 of chemotherapy, whereas TRIF (D) increased at day 7 in the DRG but then fell back to the baseline level by 2 weeks after treatment and then remained at this level over the time frame observed. The bar graphs summarize the grouped data and indicate that the level of expression of TLR4 (B), MyD88 (C), and TRIF (D) in the DRG was significantly increased in the paclitaxel-treated rats (filled bars) compared to the vehicle-treated rats (open bars). n = 3 for each group in B through D. Abbreviations: β-act, beta-actin; V, vehicle; P, paclitaxel. **P < .01, ***P < .001. F = 14.38(4, 127) in A.

All paclitaxel-treated animals (positive controls) that were advanced to the Western blot and immunohistochemistry experiments had confirmed mechanical hypersensitivity. The expression of TLR4 was significantly increased in the L4-5 DRG beginning by day 1 after paclitaxel treatment, and this was sustained through day 7 in comparison to rats that received vehicle (Fig 1B). The expression of TLR4 then fell below the baseline level by day 14 and remained as such through day 21 (Fig 1B). The expression of MyD88 and TRIF in the DRG paralleled that of TLR4. MyD88 showed a pronounced and significant elevation by day 1 that was sustained through day 7 after paclitaxel and then dropped back to baseline when measured at day 14 and then below baseline at day 21 (Fig 1C). TRIF showed a robust elevation by day 7 after paclitaxel and then returned back to baseline when measured at days 14 and 21 (Fig 1D).

The expression of TLR4 in the spinal cord showed both similarities and differences to that observed in the DRG. As observed in the DRG, spinal TLR4 expression was significantly increased by day 1 and this was sustained but also somewhat lower at day 7 after paclitaxel treatment. However, unlike that observed in the DRG, the expression of TLR4 showed an apparent resurgence in the spinal cord at the later time points observed at days 14 and 21 (Figs 2A, 2B). Surprisingly, the expression of MyD88 and TRIF in the spinal cord showed no changes over time following paclitaxel in treated rats versus vehicle controls (Figs 2A, 2C, 2D).

Figure 2. TLR4 but not MyD88 or TRIF are increased in the L5 spinal dorsal horn in paclitaxel CIPN. The representative Western blot images shown in A illustrate that the expression of TLR4 in the spinal cord was increased by day 1 of chemotherapy, decreased some by day 7, but then showed increasing expression at days 14 and 21. The representative Western blots shown in A also illustrate that the expression of MyD88 and TRIF did not change over the time frame observed. The bar graphs summarize the grouped data and indicate that the level of expression of TLR4 (B) in the spinal cord was significantly increased in the paclitaxel-treated rats (filled bars) compared to the vehicle-treated rats (open bars), but the expression of MyD88 (C) and TRIF (D) was not significantly increased. n = 3 for each group. Abbreviations: β-act, beta-actin; V, vehicle; P, paclitaxel. *P < .05, **P < .01.

Cellular Localization of TLR4, MyD88, and TRIF in the DRG

Immunohistochemistry was used to define the cellular localization of TLR4, MyD88, and TRIF in the DRG and spinal cord at the peak of increased expression at day 7 after paclitaxel treatment. As shown in Fig 3, TLR4 was increased in small neurons after paclitaxel treatment compared with the vehicle group (Figs 3A, 3B). More specifically, TLR4 was found to be colocalized in both CGRP- and IB4-positive small DRG neurons (Figs 3C–3F). Interestingly, however, MyD88 was colocalized only in CGRP-positive and not IB4-positive neurons (Figs 4A–4D). TRIF was found to be colocalized in both CGRP- and IB4-positive small DRG neurons and also localized to medium- and large-size DRG neurons (Figs 5A–5F). No colocalization was observed for TLR4 in either the NF200- or GFAP-positive neurons or cell profiles in DRG, respectively (data not shown).

Figure 3. TLR4 is increased and colocalized to subsets of DRG neurons following paclitaxel chemotherapy. The representative immunohistochemistry images in A and B show that the expression of TLR4 (red) in the DRG is normally quite low in vehicle-treated rats (A) and in naïve rats (data not shown) but becomes quite pronounced by day 7 following paclitaxel treatment (B). The bar graphs in F show that the TLR4-positive neurons are predominantly small sized with a diameter less than 30 μm. Double immunohistochemistry images shown in C and D indicate that TLR4 expression (red) is found in subsets of CGRP-positive (blue) neurons (C, colocalization indicated in purple) as well as in IB4-positive (green) neurons (D, colocalization indicated in yellow). The merged image in E and the bar graph shown in F indicate that TLR4 was expressed in a larger percentage of IB4-positive neurons than CGRP-positive neurons as well as in a substantial proportion of DRG neurons that were neither IB4 nor CGRP positive. The bar graphs also illustrate that there was no change in the proportions of IB4-positive or CGRP-positive neurons, or in the combinations of DRG neurons following paclitaxel treatment. Scale bar = 100 μm. ***P < .001. F = 14.69(2, 48).

Figure 4. MyD88 is increased and colocalized to subsets of DRG neurons following paclitaxel chemotherapy. The representative immunohistochemistry images in A and B show that the expression of MyD88 (red) in the DRG is normally quite low in vehicle-treated rats (A) and in naïve rats (data not shown) but becomes quite pronounced by day 7 following paclitaxel treatment (B). Double immunohistochemistry images shown in C and D indicate that MyD88 expression (red) is found in subsets of CGRP-positive (blue) neurons (C, colocalization indicated in purple, shown using arrows) but not in IB4-positive (green) neurons (D, colocalization in yellow).

Figure 5. TRIF is increased and colocalized to subsets of DRG neurons following paclitaxel chemotherapy. The representative immunohistochemistry images in A and B show that the expression of TRIF (red) in the DRG is normally quite low in vehicle-treated rats (A) and in naïve rats (data not shown) but becomes quite pronounced by day 7 following paclitaxel treatment (B). The bar graphs in G show the quantification of this observation. The bar graphs in F show that TRIF-positive neurons are in large-, medium-, and small-size neurons. Double immunohistochemistry images shown in C and D indicate that TRIF expression (red) is found in subsets of CGRP positive (blue) neurons (C, colocalization indicated by purple) as well as in IB4-positive (green) neurons (D, colocalization indicated by yellow). The merged image in E and the bar graph shown in H indicate that TRIF was expressed in a substantial proportion of DRG neurons that were neither IB4 nor CGRP positive. Scale bar = 100 μm. *P < .05, **P < .01.

Distribution of TLR4 After Paclitaxel Treatment in the Dorsal Horn of the Spinal Cord

Consistent with our Western blot results, immunohistochemistry showed a significant increase in the expression of TLR4 at day 7 after paclitaxel treatment in the spinal cord dorsal horn that was present at only very low levels in rats treated with vehicle (Fig 6A, 6B). TLR4 was found coexpressed in GFAP-positive cells (Figs 6C, 6D) but not in NeuN- or OX42-positive cells (Figs 6E, 6F and Figs 6G, 6H, respectively).

Figure 6. Spinal cord expression of TLR4 is increased in astrocytes but not microglia or neurons in paclitaxel-treated rats. TLR4 staining is relatively low in the spinal dorsal horn in vehicle-treated rats (A) but becomes quite prominent at day 7 following paclitaxel treatment (B). Double immunohistochemistry reveals that TLR4 colocalizes to GFAP-positive cells (C and higher magnification in D, indicated by arrows) but is not found to colocalize with NeuN-positive (E and F) or OX42-positive neurons (G and H). Scale bar in A, B, C, E, and G is 100 μm and in D, F, and H is 50 μm.

Reversal and Prevention of Paclitaxel-Induced CIPN by LPS-RS and MIP

The contribution of TLR4 and MyD88 to paclitaxel-induced neuropathic pain was assessed by testing the effects of intrathecally administered LPS-RS and MIP on both preestablished paclitaxel CIPN and the induction of paclitaxel CIPN (Figs 7A–7C). LPS-RS was tested to determine whether blockade of TLR4 signaling might be useful in preventing paclitaxel CIPN. LPS-RS (20 μg in 20 μL) or PBS (20 μL) was given by intrathecal injection every 12 hours beginning 2 days before and continuing through 2 days after the chemotherapy drug. LPS-RS had no effect on baseline mechanical withdrawal threshold and showed no interaction with the paclitaxel vehicle over time (Fig 7A). The paclitaxel-PBS–treated rats (n = 8) showed the expected decrease in mechanical withdrawal threshold that was significantly different from the vehicle–treated rats by day 7 (Fig 7A). In contrast, the paclitaxel-LPS-RS–treated rats showed only a partial development of mechanical hypersensitivity that was significantly less when compared with the paclitaxel-PBS–treated rats (Fig 7A).

Figure 7. Reversal and prevention of paclitaxel-induced neuropathic pain by intrathecal injection of a TLR4 antagonist (LPS-RS) and MyD88 homodimerization inhibitory peptide (MIP). After a baseline (BL) behavioral test, rats received intraperitoneal injection of paclitaxel (P) or vehicle (V). In A, the rats were also treated every 12 hours with 20 μg TLR4 antagonist (LPS-RS) or PBS beginning 2 days before and continuing for 2 days after paclitaxel or vehicle treatment. The paclitaxel-LPS-RS group (n = 8) rats showed a significant partial prevention of mechanical hypersensitivity compared with the paclitaxel-PBS group rats (n = 8) (A). Vehicle-treated rats receiving intrathecal PBS (n = 5) or LPS-RS (n = 5) showed no changes from the baseline measures (A). In B and C, paclitaxel-induced mechanical hypersensitivity was confirmed at 14 days after treatment, and then rats were treated with 20 μg of the TLR4 antagonist LPS-RS (n = 7) or PBS (n = 4) (intrathecal, B) or 500 μM MyD88 inhibitor peptide (n = 8) (MIP, C) or 500 μM MyD88 control peptide (n = 8) (CP, C). LPS-RS (black circles) transiently reversed the mechanical hyperresponsiveness with peak effect at 3 hours (B). Similarly, MIP (black circles) reversed mechanical hyperresponsiveness with a peak effects at 3 hours, but the duration in effect was shorter as mechanical withdrawal threshold returned to baseline by 6 hours after the injection (C). No significant difference was observed in the rotarod test of the rats between MIP group 500 μM (n = 4) and CP 500 μM group (n = 4) (D) (*P < .05; **P < .01; ***P < .001; 2-way analysis of variance followed by Bonferroni post hoc test). Asterisks indicate significant differences between the paclitaxel-LPS-RS groups and paclitaxel-MIP groups versus the paclitaxel-PBS and paclitaxel-CP groups. Crosses indicate significant decrease in mechanical withdrawal threshold in paclitaxel-treated groups to baseline measures (+++ and ***P < .001; **P < .01, *P < .05). F = 4.98(12, 115), 4.18(7, 72), 8.84(21, 176), and .01(1, 12) for A, B, C, and D, respectively. The representative photographs show a low-power view of the DRG (E) and a high-power view of a pair of DRG neurons (F) demonstrating that the Alexa488-labeled oligonucleotide reached the L5 DRG and was found in neurons 3 days following intrathecal injection by lumbar puncture (20 μL). Scale bar in E and F is 100 and 10 μm, respectively.

In the second experiment, wherein LPS-RS was used to reverse paclitaxel-induced CIPN, 2 groups were first treated with paclitaxel, and hypersensitivity to mechanical stimuli was confirmed in each group such that there was a statistical difference between the baseline measurement and that at day 14 for both, whereas neither treatment group was different from the other (Fig 7B). Rats in both groups were then given a single intrathecal dose of either 20 μg LPS-RS (n = 7) in 20 μL PBS or 20 μL PBS alone (n = 4). The LPS-RS group showed an increase in mechanical withdrawal threshold that was evident by 2 hours after treatment, and this became significantly different from the PBS group at 3 hours after treatment. The effect of LPS-RS then subsided over time such that no difference from the PBS-treated group remained at 24 hours after injection (Fig 7B).

The effect of inhibiting MyD88, the immediate downstream signal of TLR4, was tested in a similar fashion. Rats were treated with either paclitaxel or vehicle and then tested to verify establishment of the CIPN model at day 14 after treatment (Fig 7C). The 2 paclitaxel-treated groups of rats showed a mechanical withdrawal threshold significantly lower than that of the vehicle-treated group of rats but were not different from each other. The rats were then treated with MIP (n = 8 in the paclitaxel group, n = 5 in the vehicle group) or CP (n = 8 in the paclitaxel group, n = 5 in the vehicle group), and the mechanical withdrawal threshold was reevaluated. MIP-treated rats showed a significant increase in mechanical withdrawal threshold that was evident by 1 hour and significant from the CP-treated paclitaxel group by 2 hours after injection (Fig 7C). The peak effect of MIP in reversing paclitaxel CIPN was observed at 3 hours after injection and then waned, with complete loss of effect by 6 hours after drug delivery (Fig 7C). Finally, none of the doses of intrathecal MIP used here had any effect on motor performance assessed in the rotarod test (Fig 7D).

To determine whether the intrathecal route of drug administration would reach and so affect signaling in the DRG as well as the spinal cord, 20 μL of Alexa488-labeled mismatch TLR4 oligonucleotide (Invitrogen) was injected and the DRG removed 3 days later, sectioned, and mounted for visualization. Representative photographs showing labels in both the DRG and the spinal cord are shown in Figs 7E and 7F.

Changes in Expression of MyD88 and TRIF in Rats After LPS-RS Treatment

Our data showed that intrathecal injection of the TLR4 antagonist LPS-RS prevented the induction of paclitaxel-induced pain (Fig 7A). Because MyD88 and TRIF are downstream signaling molecules of TLR4, we performed experiments to verify whether LPS-RS treatment could block the paclitaxel-induced increase of MyD88 and TRIF in DRG. The expression of MyD88 was significantly decreased in the L4-5 DRG at day 7 after LPS-RS treatment in paclitaxel-treated rats in comparison to rats that received vehicle (Fig 8A). The increase of expression of TRIF at day 7, however, was not affected by LPS-RS treatment (Fig 8B).

Figure 8. The increase of MyD88, but not TRIF, is prevented in the DRG in paclitaxel CIPN under LPS-RS treatment. The representative Western blot images shown in A illustrate that the increasing expression of MyD88 in the DRG was blocked by day 7 of LPS-RS. The representative Western blots shown in B illustrate that the expression of TRIF did not change. The bar graphs summarize the grouped data and indicate the levels of expression of MyD88 (A) and TRIF (B) in the DRG in the paclitaxel-LPS-RS–treated rats (filled bars) compared to the paclitaxel-PBS–treated rats (open bars). n = 3 for each group. Abbreviations: β-act, beta-actin; P, paclitaxel-PBS; L, paclitaxel-LPS-RS. **P < .01.

Discussion

A possible mechanism of nerve damage in CIPN is derived from several lines of evidence that show that proinflammatory cytokines produce sensory fiber dysfunction and pain in many diverse models of neuropathic and inflammatory pain,45, 24, 51, 56, 54, 55, 58, 53, 66, 25 and that these are also induced by the major chemotherapy drugs.65, 49 The cytokines interferon-α/γ, tumor necrosis factor alpha, interleukin (IL)-1, and IL-6 are increased in vitro following exposure of macrophages to paclitaxel68, 42 or cisplatin,3, 32, 46 and the chemokine CCL2 is increased in the DRG and spinal cord in the early stages of taxol chemotherapy in subsets of DRG neurons and spinal astrocytes.⁶⁹ Loss of function studies using anti-CCL2 antibodies protected animals from taxol-induced hyperalgesia and loss of peripheral skin innervation density. A key observation is that the pattern of cytokine gene induction, synthesis, and release by chemotherapeutics is very similar to that induced by LPS.4, 26 This strongly implicates a role for TLR4 in CIPN given that TLR4 is the receptor in the innate immune system activated by LPS. TLR4 is expressed on the cell surface of neurons and glial cells in human brain,¹¹ and the data shown here indicate that TLR4 is expressed in the dorsal horn and subsets of DRG neurons and glial cells. TLR4 antagonists reduce nerve injury–induced hyperalgesia in mice and rats,6, 36 and knockout of either TLR4 or its signaling cofactor CD14 results in abbreviated postinflammatory hyperalgesia, reduced spinal glial responses to inflammation, and reduced neuropathic pain.22, 14, 60 The chemotherapeutic cisplatin increased expression of TLR4 on peritoneal macrophages in vitro and primed their cytokine response to L929 cells,²¹ and the data shown here indicate a similar effect in vivo in both the dorsal horn and the DRG following paclitaxel. Finally, TLR4-mediated induction of hyperalgesia by LPS includes effects of damage-associated molecular pattern components including heat shock protein 90 (HSP90)³⁵ and high-mobility group box-1 (HMGB1) protein.18, 61 HSP90 and/or HMGB1 may be directly induced from DRG neurons or glia by damage to cell organelles or DNA by chemotherapy drugs,³⁹ and both directly induce hyperalgesia when exogenously administered.50, 30 An increase in HSP90 is induced also following proteasome inhibition with bortezomib most likely due to accumulation of misfolded intracellular proteins.⁶⁴ Thus, TLR4 activation and/or damage-associated molecular pattern signaling may be a common entry point of pathophysiology shared among chemotherapeutics that ultimately results in the generation of the shared clinical CIPN phenotype.

Signal transduction following TLR4 activation occurs through 2 distinct pathways. One pathway results in the induction of a MyD88-dependent cascade leading to early NF-κB activation and subsequent increased synthesis and release of multiple cytokines and chemokines.43, 47 The second cascade is TRIF dependent and leads to delayed NF-κB activation and interferon-β production.43, 47 There is growing evidence indicating that TLR4, along with other toll-like receptors, shares the same signaling cascade as the IL-1 receptor.12, 44 MyD88 is recruited to the receptor as an adaptor protein, and this recruitment is followed by activation of IL-1 receptor–associated kinases and tumor necrosis factor receptor–associated factor 6, which leads to NF-κB activation.¹² Alternatively, LPS also increases mitogen-activated protein kinase function and NF-κB nuclear translocation in MyD88-knockout cells, indicating that non-MyD88 signaling also contributes to the biological response to LPS via TLR4.³⁸ The MyD88 activation pathway is recruited by all toll-like receptors except TLR3, and it leads to activation of NF-κB and production of proinflammatory cytokines such as tumor necrosis factor alpha and IL-1.³⁸

Our results showed a transient analgesic effect of intrathecal LDP-RS and the MyD88 inhibitor at day 14 following paclitaxel, yet there was no increase in TLR4 expression in the DRG and no elevated expression of MyD88 in the DRG or spinal cord at this time point. Indeed, MyD88 was not observed in the spinal cord. These results raise a number of interesting questions that will form the basis for several lines of interesting follow-up investigation. The results with both antagonists suggest there is some means for tonic stimulation at TLR4 that is separate from that provided at early time points where paclitaxel is present and presumably providing this stimulus. One logical candidate would seem to be reactive oxygen species that have been implicated by others in the pathophysiology of CIPN29, 31 and that also stimulate TLR4.¹ Alternatively, chemotherapeutics stimulate the release of damage-associated molecular pattern proteins including HMGB1⁴⁰ that both stimulates TLR4 in neurons² and promotes neuropathic pain.³⁰ Indeed, it is intriguing to speculate that these supplemental players could account for both the coasting often seen in patients and the transition to a chronic pain condition given that each could support a self-sustaining positive feedback signaling loop. The lack of MyD88 in the spinal cord and its presence in only a subset of DRG neurons also indicates that parallel signal pathways contribute to CIPN. TLR4 and TRIF were shown to be expressed on both IB4- and CGRP-positive neurons in the DRG, but MyD88 expression was increased only in CGRP-positive neurons. No TLR4 expression was found in large neurons, but somewhat surprisingly, TRIF expression was found in large neurons. TRIF is shared by both TLR3 and TLR4 signaling, and it produces type I interferon.33, 43 Hence, TLR3 may also be involved in paclitaxel-induced CIPN. In that a key role of IB4-positive DRG neurons in CIPN has been demonstrated,³⁷ it would seem likely that there is also a prominent role for mitogen-activated protein kinase in this subset of cells in the pathophysiology of CIPN that remains to be defined.

It has become widely recognized that glial cells in the peripheral nerve, DRG, and spinal cord react following peripheral inflammation or nerve injury and contribute to the pathophysiology of the resulting hyperalgesia.41, 13, 70, 59 Moreover, a key component in the glial response is mediated by TLR4.22, 57 Microglia has arguably received the maximum attention as the key glial element involved in nerve injury–related pain.5, 23 Yet, a striking feature of the CIPN glial phenotype is an apparently major role of spinal astrocytes and a much less prominent or complete lack of recruitment of microglia.72, 71 Spinal astrocytes become hypertrophic and show increased expression of GFAP in both paclitaxel- and oxaliplatin-related CIPN.71, 67 In addition, astrocytes in CIPN rats show increased expression of the gap junction protein connexin 43 as well as downregulation of glutamate transporters.15, 71 Patients complain not only of ongoing pain with CIPN, but also, when pain is aggravated by peripheral stimuli, complain that pain supersedes their ongoing pain and, importantly, it may last for minutes, hours, or sometimes days.9, 10, 16, 28, 27 Experimental animals similarly show exaggerated nocifensive behavioral reactions to peripheral stimuli in models of chemoneuropathy.15, 17 Given that skin innervation density is depleted in all models of CIPN studied thus far,8, 7, 52 there must be some mechanism for augmentation of signaling by the residual fibers. One potential mechanism is downregulation of glutamate transporters in spinal astrocytes. The data shown here raise the question as to the relationship between TLR4 activation and the change in spinal astrocyte phenotype. Given that the chemotherapeutics that cause CIPN poorly penetrate to the central nervous system, it is of further keen interest as to what the potential signal or link may be between peripheral and central TLR4 activation.