PD-1/PD-L1 Inhibition Enhances Chemotherapy-Induced Neuropathic Pain by Suppressing Neuroimmune Antinociceptive Signaling

The use of checkpoint inhibitors of the programmed cell death-1 (PD-1) receptor or its ligand PD-L1 has revolutionized the treatment of several malignancies (1). Their application as a monotherapy or in combination with chemotherapy agents, such as taxanes, is now the standard treatment for advanced non–small cell lung cancer and metastatic PD-L1–positive triple-negative breast cancer (2, 3). However, with the increasing clinical use of these agents, novel and/or enhanced drug-related adverse events have been reported, including peripheral neuropathy (4–6).

Although neurons and immune cells are unique cell types, they express various common genes and proteins that are associated with host protection, such as cytokines, chemokines, and Toll-like receptors (TLR; ref. 7). Moreover, dorsal root ganglion (DRG) nociceptive neurons express functional PD-1 and mice lacking Pd1 (Pdcd1) exhibit hyperexcitability in sensory neurons and hypersensitivity to pain (8). In addition, PD-1 blockade suppresses gamma aminobutyric acid (GABA)–mediated neurotransmission and GABA-mediated analgesia and anesthesia (9). Although studies have described the overall tolerability of checkpoint inhibitors in patients with cancer, the combined use of PD-1/PD-L1 inhibitors with chemotherapeutic agents such as paclitaxel is associated with an increased risk of peripheral neuropathy development (6).

Paclitaxel is a chemotherapeutic agent used for the treatment of many types of cancers, including breast and lung cancers. However, paclitaxel-based regimens are often associated with the development of peripheral neuropathy (10). This drug triggers the infiltration of macrophages into the DRG and peripheral nerves and, through the activation of the innate immune receptor TLR4, induces the release of cytokines (TNFα, IL1β, and IL6) that sensitize neurons, inducing pain (11). On the other hand, TLR4 activation also triggers the release of anti-inflammatory mediators, such as IL10, which modulates the response to nerve damage and might attenuate pain behavior (12). Previous studies reported upregulation of PD-L1 expression upon the activation of innate immune signaling by TLR4 in fibroblasts and tumor cells (13, 14). However, the effect of this signaling on chemotherapy-induced peripheral neuropathy has not been investigated.

Thus, we studied the role of the PD-1/PD-L1 axis in the modulation of paclitaxel-induced peripheral neuropathy.

Male and female wild-type (WT), Nav1.8-Cre-tdTomato, Tlr4^−/−, and Tnfr1^−/− mice (6–8 weeks old) on a C57BL/6J background were obtained from the animal facility of the Ribeirao Preto Medical School of the University of São Paulo, São Paulo, Brazil. The mice were housed in barrier cages under controlled environmental conditions (12/12 hours light/dark cycle, 55 ± 5% humidity, 23°C). All animal experiments were approved by the Animal Ethics Committee of the Ribeirao Preto Medical School, University of Sao Paulo, São Paulo, Brazil (protocol number: 200/2018).

To induce neuropathic pain, WT mice were treated with paclitaxel (Libbs, #ONTA_V.16–21) diluted in saline (NaCl 0.9%, Samtec) at doses of 2, 4, and 8 mg/kg, i.p. every other day for a total of 4 injections. The control group was treated with saline (NaCl 0.9%, Samtec). Tactile allodynia and cold hypersensitivity, which are manifestations of peripheral neuropathy, were assessed using the von Frey test and acetone test, respectively. The sciatic nerve, DRG, and spinal cord tissue were collected from WT mice treated with paclitaxel (8 mg/kg, i.p. every other day for a total of 4 injections) or saline (NaCl 0.9%) to analyze expression of Cd274 or Pdcd1 mRNA by RT-PCR on days 3, 7, and 14 or 3 and 7 after the initiation of treatment, respectively. On day 7, we collected DRGs and sciatic nerve from Nav1.8-Cre-tdTomato and WT mice to analyze PD-1 and PD-L1 by immunofluorescence, and PD-L1 by flow cytometry. WT mice were treated with recombinant PD-L1 (rPD-L1, 10 μg, i.p., Abcam, #ab130039) or 1x PBS as control (Thermo Scientific, #AM9624) 1 hour before treatment with paclitaxel (8 mg/kg, i.p.) or formalin (Synth, #01F2558.01.BJ) in the paw (20 μL of formalin 1%) and nociceptive behavior was analyzed by von Frey, acetone test, or licking time. WT, Tlr4^−/−, or Tnfr1^−/− mice were treated with paclitaxel (8 mg/kg, i.p.) or saline (NaCl 0.9%), and the mechanical withdrawal threshold was tested using von Frey filaments at 2, 4, and 6 hours after the treatment. The DRGs and sciatic nerves were collected 6 hours after the treatment to analyze the levels of TNFα, IL1β, and IL6 by ELISA. WT mice were treated with anti–PD-L1 (200 μg, i.p., on day 1 and day 5, a total of 2 injections, BioxCell, Clone, 10F.9G2, #BE0101), paclitaxel (2 mg/kg, i.p., every other day for a total of 4 injections), or anti–PD-L1 plus paclitaxel, the group control was treated with 1x PBS (i.p., on day 1 and day 5, a total of 2 injections) and saline (NaCl 0.9%, i.p., every other day for a total of 4 injections). The nociceptive behavior was analyzed by von Frey and acetone test. The sciatic nerves and DRGs were collected to analyze the expression of inflammatory markers by RT-PCR. Nav1.8-Cre-tdTomato mice were treated with anti–PD-L1 (200 μg, i.p., on day 1 and day 5, a total of 2 injections), paclitaxel (2 mg/kg, i.p., every other day for a total of 4 injections), or anti–PD-L1 plus paclitaxel. The group control was treated with 1x PBS (i.p., on day 1 and day 5, total of 2 injections) and saline (NaCl 0.9%, i.p., every other day for a total of 4 injections). The intraepidermal nerve fibers (IENF) density was analyzed on day 7 after the beginning of treatment.

Mechanical allodynia in mice was tested using an ascending, calibrated series of von Frey filaments (Exacta Precision & Performance – Touch Test, #58011) with logarithmically increasing stiffness (–2.35 to 2.65 log of force, g), as described by Vivancos and colleagues (15). In brief, mice were placed in individual acrylic cages with a wire grid floor 60 minutes before the beginning of the tests in a quiet room. The filaments were perpendicularly applied to the plantar surface of the foot with a gradual increase in pressure. The weakest filament able to elicit a response was the mechanical threshold (g). The results are reported as D log of force (g) which was calculated by subtracting the value obtained after from that obtained before the treatment.

Mice were placed on an elevated wire grid and allowed to habituate for 60 minutes prior to testing. Then, 50 μL of fluid acetone (Synth, #01A1017.01.BJ) was sprayed on the plantar surface of the left hind paw using a 1-mL syringe. An increase in the frequency of foot withdrawals in response to acetone was interpreted as cold allodynia, as described by Yoon and colleagues (16).

Acute inflammatory pain was induced by intraplantar injection of 20 μL of formalin (1%) into the dorsal surface of the right hind paw of mice followed by analysis using the von Frey filaments test and by the time in seconds spent flinching or licking the injected paw. The licking/flinching behavior was monitored for 50 minutes. The observational session was divided into two phases: the early phase (0–10 minutes) and late phase (10–40 minutes) after formalin injection. Cumulative licking/flinching for each phase, i.e., sum of flinches from 1 to 10 minutes for phase I and from 10 to 40 minutes for phase II, was compared between the groups. The nociceptive behaviors induced by formalin in the paw were assessed in a blinded manner.

L929 cell–conditioned medium was used as a source of macrophage colony-stimulating factor. L929 cells (ATCC #CCL-1), a mouse fibroblast cell line, were grown for three passages from cryogenic storage before seeding for supernatant collection. L929 cells were seeded in cell culture 150-cm² flask (Corning, #430825) seeded in 50 mL of RPMI1640 (Corning, #15–040-CV) containing 10% (vol/vol) FBS (Gibco, #12657–029), l-Glutamine (2 mmol/L, Corning, #25–015-CI), penicillin (100 U/mL, Sigma-Aldrich, #P4333), streptomycin (100 μg/mL Sigma-Aldrich, #P4333), and amphotericin B (2.5 μg/mL, Gibco #15290–018) and incubated at 37°C and 5% CO₂. The medium was carefully removed after L929 cells reached 80% confluence and replaced with 75 mL of fresh RPMI media for a subsequent 10 to 14 days. The supernatant collections were centrifuged, sterile filtered, and tested for Mycoplasma (Lonza, #LT07–318) before aliquoting into 50-mL falcon tubes and stored at –20°C.

Murine bone marrow was isolated from 6- to 10-week-old WT and Tlr4^−/− mice by flushing both femurs. After euthanasia mice were soaked with 70% ethanol. The lower limb skin was removed, and we cut the major muscles near the base of the lower limb to expose the hip joint. After that, we remove the femur. Then, we remove the epiphyses of the femur so that the bone marrow can be accessed from the ends with a 23G needle. The bone marrow was flushed into a 15 mL tube by slowly injecting approximately 2 to 3 mL RPMI1640 without FBS per bone. The bone marrow suspension was centrifuged at 450×g for 5 minutes at 4°C and resuspended in 20-mL bone marrow culture medium—RPMI medium containing 20% FBS (v/v), penicillin (100 U/mL), streptomycin (100 μg/mL), amphotericin B (2.5 μg/mL), and 20% L929 cell culture supernatant (v/v). Bone marrow cells were seeded in four nontreated Petri dishes (Corning #430591) and incubated at 37°C and 5% CO₂. Four days after seeding, fresh bone marrow culture medium was added, and the cells were incubated for an additional 3 days. At the end of this period, bone marrow–derived macrophages (BMDM) were seeded in 96-well plates at a density of 2×10⁵ cells per well or 12-well plate with 4×10⁵ cells per well and stimulated with lipopolysaccharide (LPS; 100 ng/mL, Sigma-Aldrich, #L2630), or paclitaxel (30 μmol/L, Sigma-Aldrich, #T7402), unstimulated cells were used as control. After 48 hours of stimulation, culture supernatant was analyzed by ELISA to measure the levels of TNFα, IL1β, IL6, and IL10 or the cells were analyzed by flow cytometry for expression of PD-L1, respectively.

At the indicated times, mice were anesthetized with ketamine (100 mg/kg, i.p., Agener, #ketaminaagener10) and xylazine (10 mg/kg, i.p. Agener, #calmiunagner) and then perfused with PBS pH 7.4 (4°C). The sciatic nerve, DRG, and spinal cord were collected from the region correspondent to lumbar segments (L3–L5) and homogenized in TRIzol Reagent (Life Technologies Corporation, #15–596–018) reagent at 4°C. Then, total cellular mRNA was purified from the homogenized whole tissues according to the manufacturer's instructions. Afterward, the mRNA was isolated using RNeasy micro kit (Qiagen, #74004). One microgram purified RNA was converted into cDNA using high-capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, #43–688–13). RT-PCR was performed using SYBR Green PCR Master Mix (Promega, M7123) in the StepOnePlus detection system (Applied Biosystems). The level of each gene was normalized to the levels of the mouse Gapdh gene, and the results were analyzed by the method of quantitative relative expression 2^−ΔΔCT. There were two replicates per sample. Primer pairs were as follows: Gapdh: R- GTGGTTCACACCCATCACAA and F- AGGAGCGAGACCCCACTAAC; Tnf: R- TGGGAGTAGACAAGGTACAACCC and F-CATCTTCTCAAAATTCGAGTGACAA; Il6: R-CCTTCTGTGACTCCAGCTTATC and F-TTCCTACCCCAATTTCCAAT; Il1b: R-TTGGAAGCAGCCCTTCATCT and F-TGACAGTGATGAGAATGACCTGTTC; Cd274: R-CCCTTCAAAAGCTGGTCCTT and F-GCATTATATTCACAGCCTGC; Pdcd1: R-TGGTGTCTTCTCTCGTCCCT and F- GAACATCCTTGACACACGGC. Cx3cr1: R-ATGGTGTCCAGAAGAGGAAGAA and F-TCACAAAGAGAAAGGACAACGA; Ccl2 - R-TGCAGCAGTCAACACAAATTG and F-AAGACAGATGTGGTGGGTTT; Aif1 - R- GCTTCAAGTTTGGACGGCAG F-TGAGGAGCCATGAGCCAAAG.

Mice were anesthetized with ketamine (100 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.) and then perfused PBS pH 7.4 (4°C), the DRGs were collected and immediately place in DMEM medium (corning, #15–013-CV) supplemented with 10% FBS, 500 μL per sample. We then added 500 μL/sample of collagenase type II solution (1 mg/mL, Worthington Biochemical Corporation, #LS004177) and incubated for 40 minutes at 37°C under constant agitation 400×g. after incubation, 1 mL of DMEM (10% of FBS) was added and the samples passed through a 70-μm cell strainer (corning #352350). The samples were centrifuged (450×g, 5 minutes at 4°C). DRG samples were stained with fixable viability dye (eBioscience, #65–0865–14). Then, the cells were stained with CD45 (BV421, BD, clone 30F11, #563890), CD11b (PE, BioLegend, clone M1/70, #101208), F4/80 (FITC, Thermo Fisher Scientific, clone – BM8, #11–4801–82), and PD-L1 (APC, BioLegend, clone 10F.9G2, #124312). Flow cytometry was performed on a FACS-canto (BD Bioscience).

BMDMs were cultured in a 12-well plate at 4×10⁵ cells/well and stimulated with LPS (100 ng/mL) or paclitaxel (30 μmol/L), unstimulated cells were used as control. After 48 hours, the cells were harvested in cold PBS (4°C). The samples were centrifuged (450×g, 5 minutes at 4°C) and stained with fixable viability dye (and a cocktail of antibodies (all diluted 1:300): F4/80 (BV421, BD, clone T452342, #565411), and PD-L1 (PE, BD, clone MIH5, #558091). Flow cytometry was performed on a FACS-Verse, BD Bioscience. FlowJo software version 10.8.1 (BD) was used to analyze the data.

The Mouse DuoSet ELISA kits for measuring TNFα, IL1β, IL6, and IL10 were purchased from R&D Systems (#DY410, #DY401, #DY406, and #DY417, respectively). BMDM culture supernatant, DRG, and sciatic nerve were analyzed as per the manufacturer's instructions. To obtain the DRG and sciatic nerve samples, animals were anesthetized with ketamine (100 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.). Then, intracardiac perfusion with PBS pH 7.4 (4°C) solution was performed. DRG segments between L3 and L5 and the sciatic nerve were removed 6 hours after treatment with paclitaxel or saline. Samples were homogenized in PBS (pH 7.4) containing 1x protease inhibitor cocktail (Roche, #11697498001) for 30 seconds using a basic homogenizer (Polytron, #PT3100). Homogenates were centrifuged at 15,000 g for 30 minutes at 4°C and supernatants were collected for use in the cytokine assays.

Nav1.8-Cre-tdTomato mice were treated with saline (NaCl 0.9%) or paclitaxel (8 mg/kg, i.p. every other day for a total of 4 injections). On day 7 after the initiation of treatment, mice were anesthetized with ketamine (100 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.) and then perfused transcardially with PBS pH 7.4 (4°C) solution, followed by fresh 4% paraformaldehyde (PFA, Synth, #00P1005.01.AH) in 1x PBS (pH 7.4). After the perfusion, segments of the sciatic nerve and DRGs correspondent to L3 to L5 were dissected out. The dissected tissue was fixed for 4 hours in PFA, and then the PFA was replaced with 30% sucrose (Synth, #01S1002.01.AH) in 1x PBS (pH 7.4) for 2 days, at 4°C. Samples were embedded in Tissue-Tek O.C.T compound (Sakura, #4583) and sectioned into 20 μm in a cryostat (Croppa Technology, #CM1850–7-1). Subsequently, the sections were incubated overnight at 4°C with anti–PD-L1 (1:200, Bio X Cell, clone 10F.9G2, #BE0101), anti–IBA-1 (1:400, Wako Chemicals, #016–20001), or anti–PD-1 (1:100, Abcam, clone EPR206665, #ab214421). The sections were then incubated with a secondary antibody (donkey anti-rabbit 1:200: Alexa fluor 488, Invitrogen, #A21206 or donkey anti-mouse 1:200: Alexa fluor 594, Invitrogen, #A21203) for 2 hours. The slides were assembled using VectaShield with DAPI (Vector, #H-1200). Images were acquired using a Zeiss fluorescence microscope (AxioObserver LSM780). Three images were obtained per animal. A qualitative assessment of the localization of PD-1 and PD-L1 was performed according to the markers Nav1.8tdtomato and IBA-1.

For IENF analysis, Nav1.8-Cre-tdTomato mice were treated with anti–PD-L1 (200 μg, i.p., on day 1 and day 5), paclitaxel (2 mg/kg, i.p., every other day for a total of 4 injections), or anti–PD-L1 plus paclitaxel, the group control was treated with 1x PBS (i.p., on day 1 and day 5) and saline (NaCl 0.9%, i.p., every other day for a total of 4 injections). On day 7 after the initiation of treatment the animals were anesthetized as described above and then perfused with 1x PBS and PFA 4%. The hind paw footpad skin was excised, fixed in PFA 4%, and then transferred to a 30% sucrose solution overnight at 4°C. Samples were embedded in Tissue-Tek O.C.T compound and sectioned into 16 μm in a cryostat. All sections were blocked with 1% BSA (Sigma-Aldrich, #A7906) in 0.1% Triton X100 (Thermo Fischer, #HFH10) for 1 hour at room temperature and incubated overnight at 4°C with an antibody specific for collagen IV (1:500, goat anti-type IV Collagen; SouththernBiotech, #1340–1;) followed by a secondary antibody conjugated with AlexaFluor-488 (1:1,000, anti-goat; Invitrogen, #A11055). Images were acquired using a Zeiss fluorescence microscope (AxioObserver LSM780). The ImageJ software version 2.0.0-rc-69/1.52p were used to measure the number of nerve fibers crossing the collagen-stained dermal/epidermal junction into the epidermis and the epidermis length within each field. IENF density was expressed as the total number of fibers/length of the epidermis (IENFs/mm).

The gene read counts of the RNA sequencing genotype-tissue expression (GTEx) version 8 data set (GTEx_Analysis_2017–06–05_v8_RNASeQCv1.1.9_gene_reads.gct.gz) were downloaded from the GTEx Portal (https://gtexportal.org/home/datasets), along with the de-identified sample annotations (GTEx_v8_Annotations_SampleAttributesDS.txt). The read counts were normalized using the VST feature in DESeq2 version 1.22.1 in R version 3.6.1. (https://www.r-project.org). The 56,202 genes were then filtered on the basis of their mean expression (mean transformed count > 5) to reduce noise and lessen the inflated effect of low expressing genes on correlations. The number of patients per type of tissue analyzed is shown in Supplementary Table S1.

The data and statistical analysis comply with the recommendations on experimental design and analysis. All experiments were designed to generate groups using randomization and blinded analysis. Statistical analysis was undertaken only for experiments where each group size was at least n = 5 of independent values and performed using these independent values. The data are reported as the means ± SEM obtained from at least two independent experiments. The means of different treatments were compared using Student t test, one-way or two-way ANOVA, followed by a Tukey test, as appropriate. Spearman rank-order correlation coefficient (r) was calculated to describe correlations. A value of P < 0.05 was considered significant. All statistical analyses were performed using GraphPad Prism software version 8.0.1 for Windows (GraphPad Software, USA).

Previously published data analyzed in this study were obtained from GTEx Portal (https://gtexportal.org/home/datasets), along with the de-identified sample annotations. All data generated in this study are available within the article and its Supplementary Data files or from the corresponding author upon reasonable request. A list of resources and reagents is on the Supplementary Table S2.

We first analyzed expression of PD-L1 and PD-1 in neuronal tissues to assess the role of the PD-1/PD-L1 system in the development of paclitaxel-induced neuropathic pain. Using a GTEx bank, we identified a positive correlation between the expression of PD-L1 and PD-1 in different human tissues. Neural tissues, including the tibial nerve and spinal cord, expressed the genes encoding PD-L1 and PD-1, CD274 and PDCD1, respectively (Fig. 1A).

We next investigated the dynamics of PD-L1 and PD-1 expression in the peripheral nervous system in a mouse model of paclitaxel-induced peripheral neuropathy. Consistent with previous studies, peripheral neuropathy was induced in mice by administering paclitaxel (11). To demonstrate the development of paclitaxel-induced peripheral neuropathy, tactile allodynia, which is a manifestation of peripheral neuropathy, was assessed using the von Frey test. We observed a decrease in the mechanical withdrawal threshold until Day 7, which was sustained through at least Day 21 (Supplementary Fig. S1). In parallel to the development of neuropathic pain, Cd274 mRNA expression was increased in the mouse DRG (Fig. 1B–E). The PD-L1 was found mainly in DRG-macrophages (IBA-1⁺; Fig. 1E). The treatment with paclitaxel did not alter Pdcd1 mRNA expression levels in these tissues (Fig. 1F and G). However, immunofluorescence staining revealed that PD-1 was expressed in most DRG neurons (Fig. 1H). By using Nav1.8-Cre-tdTomato mice, we confirmed that PD-1 expression was in primary sensory neurons (Fig. 1H). Furthermore, PD-1 immunoreactivity was also detected in the sciatic nerve (Fig. 1I).

In physiologic homeostasis, the interaction of PD-L1 with PD-1 is essential for the development of immune tolerance. However, functional PD-1 is also expressed by nonimmune cells. According to Chen and colleagues (8), PD-1 is expressed at high levels in DRG neurons, and mice lacking Pdcd1 exhibit thermal and mechanical hypersensitivity. Consistent with these findings, we observed that PD-L1 expression was increased by paclitaxel treatment in DRG tissue and that primary sensory neurons expressed PD-1, providing a neuronal basis for a possible effect of PD-1/PD-L1 on the development of paclitaxel-induced neuropathic pain.

Next, we investigated which cells from the DRG tissue expressed PD-L1. Using flow cytometry, we showed that PD-L1 was mainly expressed by CD45⁺ cells and that paclitaxel treatment increased PD-L1 expression only in CD45⁺ cells (Fig. 2A; Supplementary Fig. S2). Furthermore, macrophages (CD45⁺CD11b⁺F4/80⁺) from the DRG were the main immune cells expressing PD-L1, and treatment with paclitaxel increased the frequency and number of these cells (Fig. 2B and C; Supplementary Fig. S3).

Previous studies showed that stimulation of TLRs induces PD-L1 expression on fibroblasts and cancer cells (13, 14). We previously showed that paclitaxel exerted an LPS-mimicking effect on TLR4 activation (17). Thus, we investigated whether paclitaxel modulated the expression of PD-L1 on macrophages via TLR4. BMDMs from WT and Tlr4^−/− mice were cultured with LPS or paclitaxel. Like LPS, paclitaxel increased PD-L1 expression in BMDMs in a TLR4-dependent manner (Fig. 2D–G; Supplementary Fig. S4).

Because nonneuronal cells, including macrophages, microglia, and astrocytes, influence neuronal excitability and consequently participate in the establishment of neuropathic pain (18), these data reinforce the hypothesis that DRG macrophages might modulate the nociceptive response via PD-L1. To test the effect of PD-L1 on the development of neuropathic or inflammatory pain, we treating mice with exogenous recombinant PD-L1 before paclitaxel or formalin injection and then analyzed nociceptive behaviors. First, we observed that PD-L1 treatment reduced mechanical and cold hypersensitivity after paclitaxel treatment (Fig. 2H and I). Next, we tested the effect of rPDL-1 on the inflammatory pain model induced by an intraplantar injection of formalin (2%). The von Frey filament test revealed that rPD-L1 treatment inhibited the reduced mechanical withdrawal threshold and licking indicating pain behavior compared with the control group (Fig. 2J and K). These results indicate the antinociceptive role of the PD-1/PD-L1 axis.

TLR4 activation induces the production of proinflammatory mediators and anti-inflammatory signals, such as IL10, that contribute to attenuating inflammation and protecting the tissues (17, 19). In our study we showed that BMDMs stimulated with LPS or paclitaxel produced the inflammatory cytokines TNFα, IL1β, and IL6 (Fig. 3A–C) as well as the regulatory mediator IL10 (Fig. 3D) and PD-L1 (Fig. 2D–G). Consistent with previous findings (11), we observed that Tlr4^−/− mice were resistant to paclitaxel-induced acute pain and that the absence of Tlr4 reduced the levels of inflammatory mediators in the DRG (Fig. 3E–H) and sciatic nerve (Fig. 3I–K). Furthermore, Tnfr1^−/− mice were also protected from paclitaxel-induced acute pain (Fig. 3L). However, as described above, TLR4 activation by paclitaxel also increased the expression of PD-L1, which has antinociceptive properties (Fig. 2H–K). Together, these data reinforce the algesic role of TLR4/TNFα signaling but indicate that TLR4 activation also induces compensatory mechanisms, such as PD-L1 expression, to attenuate pain development.

We next examined the nociceptive response of mice treated with anti–PD-L1, paclitaxel, or the combined therapy to determine the effect of PD-1/PD-L1 checkpoint inhibition on the development of neuropathic pain (Fig. 4A). The combined therapy induced increased levels of cold hypersensitivity compared with monotherapy and vehicle control (Fig. 4B). In addition, mice treated with paclitaxel showed a significant decrease in the mechanical withdrawal threshold until Day 7, which was sustained through at least Day 21 (Fig. 4C). Similarly, mice treated with anti–PD-L1 also showed a decrease in mechanical withdrawal threshold by Day 1 after treatment that became more pronounced over time, achieving a significant difference from baseline by Days 7, 14, and 21. Mice treated with the combination of paclitaxel plus anti–PD-L1 presented increased mechanical allodynia and neuropathy compared with those treated with single agents (Fig. 4C).

We examined the nociceptive response of male and female mice to determine if the effect of PD-1/PD-L1 checkpoint inhibition on the development of neuropathic pain could be affected by sex. We observed that in both sexes, treatment with paclitaxel or combined therapy induced a similar increase in the mechanical allodynia and cold hypersensitivity compared with controls (Supplementary Fig. S5). Likewise, others have reported no gender difference in paclitaxel-induced neuropathic pain or analgesic response in rats (20). In another study, cold allodynia was more robust in female mice treated with paclitaxel, whereas no differences were observed for paclitaxel-induced mechanical allodynia (21). It has been demonstrated that the LPS-induced activation of spinal TLR4 mediates mechanical allodynia in male but not in female mice, but this sex-specific response is limited to the spinal cord. LPS administered to the brain or hind paw produces equivalent allodynia in both sexes (22). These finds suggest that TLR4 activation by paclitaxel in DRGs, located in the peripheral nervous system could be associated with the development of neuropathic pain in both sexes.

We further characterized the inflammatory response by measuring mRNA levels of proinflammatory cytokines in the sciatic nerve, DRG, and spinal cord collected on Day 7 after the initiating treatments. Compared with the single therapies (paclitaxel or anti–PDL-1), significant increases in Tnf, Il6, and Cx3cr1 mRNA levels were observed in the sciatic nerves of animals treated with the combined paclitaxel plus anti–PD-L1 therapy (Fig. 4D; Supplementary Fig. S6). Chemotherapy-induced peripheral neuropathy is associated with the decreased density of sensory nerve fibers in the periphery (23). Thus, to analyze the effect of combined therapy on IENF, Nav1.8-Cre-tdTomato mice, which express the red fluorescent protein (Td-tomato) on voltage-gated sodium channel 1.8 (Nav1.8)-expressing peripheral sensory neurons, were treated with paclitaxel, anti–PD-L1, or the combined therapy. We observed that both paclitaxel and anti–PD-L1 reduced IENF density. A similar effect was observed for the combined treatment (Fig. 4E and F). These results suggest that the nociceptive behavior induced by combined treatment may be associated with the blockage of an inhibitory effect of PD-L1 on the DRG. Consistent with this hypothesis, previous findings have demonstrated that Cd274 knockout promoted pro-inflammatory polarization of macrophages/microglia and more serious pathologic pain compared with WT mice after traumatic spinal cord injury (24).

In patients with metastatic urothelial carcinoma treated with anti–PD-1 (pembrolizumab) and metastatic melanoma treated with anti–PD-1 (nivolumab) plus anti-CTLA-4 (ipilimumab), cases of severe pain crises (4, 5) were observed. In addition, a meta-analysis of 17 clinical trials with 10,500 patients found that PD-1/PD-L1 inhibitors increased the risk of peripheral neuropathy when combined with chemotherapy (6). Consistent with the clinical evidence, we observed that the inhibition of the PD-1/PD-L1 axis by anti–PD-L1 evoked hyperexcitability in the basal (naïve) state. Furthermore, the combination of anti–PD-L1 with paclitaxel increased mechanical allodynia and chronic neuropathy development compared with paclitaxel or anti–PD-L1 alone.

In summary, the present study describes a crucial role for the PD-1/PD-L1 axis in regulating paclitaxel-induced neuropathic pain and suggests that PD-1/PD-L1 blockade exacerbates this adverse event. The understanding of the underlying mechanisms associated with this finding may aid in the development of new strategies to prevent the development of peripheral neuropathy associated with cancer treatment.

J.M. Mota reports personal fees and other support from Zodiac, Bayer, Ipsen, AstraZeneca, Amgen; personal fees, nonfinancial support, and other support from Janssen, Astellas, Pfizer, Bristol Myers Squibb, Merck; nonfinancial support from Libbs; and nonfinancial support from Roche outside the submitted work. R. Barroso-Sousa reports personal fees and other support from AstraZeneca, Daiichi Sankyo, Eli Lilly, Merck; personal fees from Pfizer, Novartis; and personal fees from Libbs outside the submitted work. No disclosures were reported by the other authors.

C.W.S. Wanderley: Conceptualization, data curation, formal analysis, writing–original draft, project administration, writing–review and editing. A.G.M. Maganin: Data curation, formal analysis, validation, investigation, methodology. B. Adjafre: Data curation, formal analysis. A.S. Mendes: Data curation, formal analysis, investigation, methodology. C.E.A. Silva: Formal analysis, investigation, methodology. A.U. Quadros: Data curation, formal analysis, validation, methodology. J.P.M. Luiz: Data curation, formal analysis, validation, investigation, methodology. C.M.S. Silva: Data curation, formal analysis, validation, investigation, visualization, methodology. N.R. Silva: Data curation, formal analysis, methodology. F.F.B. Oliveira: Data curation, formal analysis, investigation, methodology. F.I.F. Gomes: Data curation, formal analysis, investigation, visualization, methodology. J.L.J. Restrepo: Formal analysis, validation, investigation, methodology. C.A. Speck-Hernandez: Data curation, formal analysis, validation, investigation, visualization, methodology. F. Turaça: Data curation, formal analysis, methodology. G.V.L. Silva: Data curation, validation, investigation. G.R. Pigatto: Data curation, investigation. H.I. Nakaya: Data curation, supervision, investigation. J.M. Mota: Conceptualization, supervision, visualization, writing–original draft, writing–review and editing. R. Barroso-Sousa: Conceptualization, supervision, writing–original draft. J.C. Alves-Filho: Conceptualization, resources, funding acquisition. T.M. Cunha: Conceptualization, supervision, investigation, writing–review and editing. F.Q. Cunha: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.

This work was supported by grants from the São Paulo Research Foundation (FAPESP) under grant agreement no. 2013/08216–2 (Center for Research in Inflammatory Diseases) and no. 2017/25308–9.

We are grateful to Juliana Abmansur, Marcerlla Grando, Marco Antônio, Ana Kátia dos Santos, Ieda Regina, Sergio Rosa, and Diva A. M. Souza for providing technical assistance.