Chronic Oxaliplatin Treatment Induces CIPN in Mice via Activation of the TXNIP Pathway

Abstract

Background:

Oxaliplatin is a platinum-based alkylating chemotherapeutic agent primarily used for colorectal cancer treatment. While oxaliplatin is an effective chemotherapeutic agent, it is commonly associated with severe and often long-lasting neurotoxic side effects, including chemotherapy-induced peripheral neuropathy (CIPN). Because neurotoxicity develops with cumulative dosing, oxaliplatin-induced peripheral neuropathy represents a major dose-limiting side effect. The underlying mechanisms of chronic oxaliplatin-induced neuropathy remain poorly understood, and behavioral models of CIPN in mice have been challenging to reproduce. As a result, the precise pathophysiology of CIPN has yet to be elucidated.

Objectives:

The aim of this study was to investigate the underlying mechanisms of oxaliplatin-induced peripheral neuropathy using a chronic mouse model treated with a clinically relevant, human-equivalent dose.

Methods:

C57BL/6 male mice were treated with intraperitoneal oxaliplatin (0, 5 mg/kg, and 10 mg/kg) once a week for eight weeks. During treatments, mice were followed with measurements of spontaneous pain behaviors, induced pain behaviors and nerve conduction velocity. Electron microscopy of the sciatic nerve and histology and bulk RNA sequencing of the dorsal root ganglia (DRG) were performed post-mortem.

Results:

Oxaliplatin-treated mice that received 10 mg/kg/weekdose over 8 weeks replicated clinically significant spontaneous and evoked pain behaviors in our mouse model and induced strong changes in redox and inflammatory profile in the DRG. Histological post-mortem evaluation of the sciatic nerves identified strong effects on mitochondria. KEGG analysis, Non-Negative Matrix Factorization, and correlation analysis on the differentially expressed genes in DRG pointed to the potential role of thioredoxin interacting protein (TXNIP) in regulating oxidative stress pathways. Upregulation of Txnip in the DRG was verified through qPCR, western blotting, and immunohistochemistry and connected to the upstream and downstream pathways. Overall, the results provided insights into the pathophysiology and the mechanism of CIPN with oxaliplatin and highlighted potential drug targets for treatments.

Conclusions:

Oxaliplatin at high accumulative dosage leads to behavioral and physiological symptoms in mice resembling those in cancer patients undergoing chemotherapy. Our findings suggest TXNIP may serve as a central regulator of oxidative stress and inflammation in CIPN caused by oxaliplatin, presenting a potential therapeutic target for alleviating neuropathy.

INTRODUCTION

Cancer is increasing globally, with the annual number of cancer cases in the United States expected to increase driven by the aging population and other risk factors (1). Due to the development of new early diagnostics and new treatments, cancer survival rates are also rising, leading to a projected 19 million cancer survivors in 2024 in the US (2). Cancer survival often comes with a cost. Even when treatment is effective, cancer patients suffer from debilitating acute and chronic pain during and long after treatment is complete. One condition, known as chemotherapy-induced peripheral neuropathy (CIPN) can be life altering. Symptoms of CIPN include dull and sharp pain, cold-, hot- and mechanical allodynia, and burning or tingling in the extremities. These symptoms often appear within minutes to hours after chemotherapy treatment. Over time, patients develop loss of touch sensation, difficulty performing fine motor tasks, inability to tolerate cold, cramping, constipation, muscle weakness, and balance problems (3). The severity of CIPN can be so intense that it forces the discontinuation of life-saving cancer treatments (4–6). Although these symptoms are common with many chemotherapy drugs, they are especially severe in patients treated with oxaliplatin and oxaliplatin-based combinations like FOLFOX commonly used for colorectal cancer (7) and other malignancies (8). Various neuroprotective strategies investigated in experimental studies and clinical trials (9) have not addressed this problem and currently there is no effective preventive strategy or treatment for oxaliplatin induced CIPN with only limited effect achieved with Duloxetine (10,11). Identifying new therapeutic targets to alleviate the painful symptoms of CIPN is an unmet clinical need to improve the quality of life in cancer patients and survivors (9).

Oxidative stress and inflammation have long been implicated in the development and continuation of chronic pain (12). Reactive oxygen species (ROS) serve as a protective mechanism against toxins or mutations, generated in response to external stimuli and injury (13). ROS also serve as signaling molecules and play a role in a variety of neural pathways, synaptic plasticity, and neuronal maturation (14). While this response is beneficial for cellular defense by protecting the body from potentially cancerous or infected cells, unchecked ROS can be harmful in other contexts. In the presence of injury or toxins, excessive ROS activates signaling pathways that trigger the release of pro-inflammatory cytokines (e.g., IL-1β, TNF-α) (15), driving inflammation and potentially leading to neuronal death.

In the context of cancer treatment, many chemotherapy drugs induce acute mitochondrial oxidative stress in DRG neurons, leading to increased ROS production and subsequent inflammation with cellular damage and axonal degeneration, likely contributing to the development of chronic CIPN (16,17). Cells exhibit variability in their production of ROS due to distinct cellular mechanisms and antioxidant capacities. For example, connective tissue fibroblasts and intestinal stem cells are relatively stable in response to ROS due to efficient detoxification mechanisms (18). In contrast, neurons show much higher and rapid sensitivity to oxidative stress due to the combination of high metabolic activity, limited antioxidant defenses, high polyunsaturated lipid content (from myelin), calcium sensitivity, and mitochondrial vulnerability (19–21). Both peripheral neuronal processes and cell bodies in the dorsal toot ganglia (DRG) are particularly susceptible to ROS compared to other neuronal subtypes (22). Extensive peripheral neuronal projections and high metabolic demands make these neurons highly vulnerable to oxidative stress leading to loss of sensitivity, persistent chronic pain, or neurodegeneration.

TXNIP, also known as thioredoxin binding protein-2 (23), plays a central role in regulating glucose and lipid metabolism, apoptosis, cell death, and inflammation via manipulation of ROS levels (24–26). As an endogenous inhibitor of the thioredoxin (TRX) redox system, TXNIP modulates oxidative stress in response to various stresses (27). Recent studies have shown that TXNIP can induce oxidative stress and excessive ROS production through inhibition of TRX activity (28), while reduced TXNIP expression has been associated with protection against endothelial dysfunction in various conditions (29). Importantly, TXNIP has been indicated in various neurodegeneration conditions including diabetic-induced neuropathy, Alzheimer’s disease and Parkinson’s disease (30–32). Thus, it is promising that TXNIP may contribute to development of CIPN and could present a potential therapeutic target for this condition. We hypothesized that chronic oxaliplatin exposure in mice induces Txnip upregulation in DRG, contributing to oxidative stress and CIPN-like phenotypes.

In this study, we investigated the role of oxaliplatin in the development of CIPN in healthy wild type mice using a dosage equivalent to that administered in humans. Male mice were used in this study to reduce biological variability. We acknowledge the potential for sex-related differences in CIPN and future studies will address these differences to ensure comprehensive understanding. First, we demonstrated that oxaliplatin induces cold allodynia, reduced nerve conduction velocity, and gait changes resembling those observed in human patients at clinically relevant cumulative doses. Next, we conducted histological analysis of the sciatic nerve and performed both histological studies and RNA-seq on the extracted dorsal root ganglia (DRG). Differential gene analysis using Non-Negative Matrix Factorization (nNMF) identified Txnip as a key driver of oxidative stress and inflammation. The upregulation of Txnip in the DRG was further validated through Western blotting and imaging of DRG, pointing to TXNIP as a potential target in oxaliplatin-triggered CIPN.

METHODS

Animal models

C57BL/6 mice (8–10-week-old) were purchased from Charles River Laboratory and housed in a central animal care facility with food and water ad libitum. We chose male C57BL/6 mice due to their well-characterized genetic background and widespread use in preclinical studies, facilitating reproducibility. However, we acknowledge the potential for sex-related variability in CIPN as female mice may exhibit different sensitivity or responses to oxaliplatin-induced neuropathy. The follow up study will address these differences to ensure comprehensive understanding. Animals were kept in standard cages at room temperature on a 12:12 day/night light cycle. Prior to the injection, clinical grade oxaliplatin (Sandoz) was diluted in 5% dextrose. Mice were weighed and an appropriate oxaliplatin dose was delivered for that animal’s mass. To reproduce CIPN conditions in animal models we treated mice over 8 weeks with a human equivalent dosage of oxaliplatin. The dose of 10 mg/kg oxaliplatin administered weekly in mice was calculated based on the human dosage of 85 mg/m² intravenously repeated every two weeks (see Supplemental Information). The treatments consisted of control (5% dextrose), and oxaliplatin (either 5 mg/kg, accumulated dose 40 mg/kg, or 10 mg/kg, accumulated dose 80 mg/kg). The rationale of selecting the dosage was previously reported (33). To avoid circadian effects the mice were treated at the same time of day at 2 - 3 pm. Within one week after the experiments, mice were anesthetized via isoflurane induction (2% at 1.5 L O₂/min) and euthanized by cervical dislocation.

Nerve conduction velocity (NCV)

Nerve conduction velocity (NCV) was measured on mouse tails by an automated functional assessment station (FASt System, Red Rock Laboratories (RRL), St. Louis, MO) and its RRL software. Mice were anesthetized with a constant flow of 2% (v/v) isoflurane. During measurements, the mouse body temperature was maintained at 37°C with a heating pad and monitored with a thermal camera. To measure NCV, stainless steel needle electrodes (sanitized with isopropanol) were used to deliver current pulses (1 mA, 0.2 ms) distally, using an analog stimulus isolator (Model 2200, A-M Systems) controlled by RRL software. Positive and negative recording electrodes were inserted in the tail 3 cm away from the stimulating electrodes and the reference electrode was inserted under the back skin. The post-stimulus latency of the compound actin potential was obtained and analyzed automatically with RRL software. The NCV was calculated by dividing the distance between stimulating and recording electrodes by the latency. The measurement was performed at least 3 times for each mouse.

Cold allodynia

Cold allodynia was assessed using a hot/cold plate (Ugo Basile), which allows regulated plate temperature manipulation. Mice were habituated to the room and the plate (15 min per mouse) at room temperature prior to the tests. The plate was then cooled to 4°C, and the mice were placed on it individually. The time taken to elicit nociceptive behaviors such as jumping, flinching, or hind-paw licking or lifting was recorded. Cutoff time was set to 60 seconds to avoid tissue damage. Unlike cold allodynia, heat hypersensitivity is not a common symptom of oxaliplatin-induced CIPN and is not a prominent feature in animal models (34,35) and was not measured in this study.

Behavioral study

Gait data were collected with the DigiGait system (Mouse Specifics Inc), using a procedure similar to previously published methods (36). Due to the logistics of the experiments, gait analysis was conducted after 6 weekly treatments (n = 10 for each group). This technique utilized ventral plane videography of the paws of each mouse at 150 frames/sec as it ambulated on a translucent treadmill. Mice were given a habituation trial before actual testing was initiated. This involved placing a mouse in the apparatus and allowing it 3–5 minutes to explore the apparatus while the treadmill remained stationary. After this brief acclimation period, movement of the treadmill belt was started at a speed of 10 cm/s, which was increased to 20 cm/s. Actual testing began at least 1 h after the habituation trial and involved collecting data from each mouse as it ambulated on the treadmill. Changes in the area of contact for each paw as it was placed onto and lifted from the belt during a step were calculated and analyzed by system software. Approximately 5s of video were collected from each mouse to provide an adequate number of sequential strides for quantification of many stride-related variables at a treadmill belt speed of 30 cm/s, including swing and stance phases and braking and propulsion.

Electron microscopy

For electron microscopy, tissue samples from sciatic nerves were harvested and stored in 3% glutaraldehyde (Polysciences Inc.). The tissue samples were post-fixed in 1% osmium tetroxide and serially dehydrated in ethanol and propylene oxide. Stained and dehydrated nerve samples were then embedded in epoxy (Polysciences), sectioned on an ultra-microtome into 1 μm cross sections, mounted on Cu/Rh mesh grids and stained in series with 3% uranyl acetate and 3% lead citrate. Images were taken with a transmission electron microscope JEM-1400 (Joel) at 120KV and analyzed in ImageJ to calculate mitochondria area and number of vacuolated mitochondria.

RNA sequencing of DRG

All bilateral lumbar DRGs (L1–L6) from euthanized animals were collected and quickly immersed in ice-cold RNAlater solution (ThermoFisher Scientific). RNA was extracted using the magnetic bead-based Maxwell RSC 48 Instrument (Nordic Biolabs) with the Maxwell RSC SimplyRNA Tissue Kit. RNA quality was assessed using the TapeStation 4200 system (Agilent). All RNA samples used for RNA sequencing had an RNA integrity number greater than 8.

Library was prepared according to manufacturer’s protocol with Clontech SMARTer method, indexed, pooled, and sequenced on a NovaSeq 6000 (Illumina). Basecalls and demultiplexing were performed with Illumina’s bcl2fastq software and a custom Python demultiplexing program with a maximum of one mismatch in the indexing read. RNA-seq reads were then aligned to the Ensembl release 76 primary assembly with STAR version 2.5.1a. Aligned reads were uploaded to Partek Flow software (Partek Inc.) for downstream analysis. Gene counts were derived from the number of uniquely aligned unambiguous reads and then normalized using the median ratio method. Differentially expressed genes (DEGs) were identified with DESeq2 package implemented in Partek Flow.

RT-qPCR

Total RNA (200 ng) was reverse transcribed using SuperScript IV VILO Master Mix (ThermoFisher Scientific) according to manufacturer’s instructions. RT-qPCR was performed with SYBRGreen Master Mix and QuantStudio 3 Real-Time PCR system (Applied Biosystem). 18S was used as an internal control and relative quantification was calculated by 2^−ΔΔCt method. Primer sequences used were as below: 18S forward AAGTTCCAGCACATTTTGCGAGTA; 18S reverse TTGGTGAGGTCGATGTCTGCTTTC; Txnip forward AGTGATTGGCAGCAGGTC; Txnip reverse GGTATCTGGGATGTTTAGG; Nlrp3 forward ATTACCCGCCCGAGAAAGG; Nlrp3 reverse TCGCAGCAAAGATCCACACAG. 18S was used as an internal control.

Immunofluorescence

Extracted DRG were fixed in 10% neutral buffered formalin (NBF) overnight at 4°C, washed with PBS and 30%, 50%, 70% alcohol, and embedded into paraffin. DRG were then sectioned at 5μm with a microtome. Slides were deparaffinized, and rehydrated with 95%, 70%, and 50% alcohols followed by rinsing with DI water. Slides were then blocked with blocking buffer made with TBS containing 0.1% Triton X-100 (TBST) and 10% normal donkey serum (Millipore Sigma, D9663) for 2 hours at room temperature. DRG slides were incubated with primary antibodies diluted in blocking buffer (anti-TXNIP: 1:500, D5F3E, Cell Signaling Technology; anti-NeuN: 1:500, MAB377, Millipore Sigma) overnight at 4°C. The following day, sections were washed 3x10min with TBST and then incubated with secondary antibodies (donkey anti-rabbit 488: 1:500, 711-545-152, Jackson ImmunoResearch; donkey anti-mouse 555: 1:500, A32773, Thermofisher Scientific) in TBST for 1 hour at RT. The slides were washed again 3x10min in TBST, followed by incubation with DAPI for 10 min at RT. Finally, DRG slides were mounted with Fluoromount mounting medium (Millipore Sigma) and imaged with a fluorescence microscope using 10x and 40x objective lens.

Western blot

Total protein was extracted from DRG of control and oxaliplatin-treated mice using a radioimmunoprecipitation assay buffer (RIPA buffer; Cell Signaling Technology) with 1% ice-cold phenylmethanesulfonylfluoride. Each well of a 10% Tris-Glycine gel (Invitrogen) was loaded with 25 μg total protein for electrophoresis. After electrophoresis, the gel was transferred onto a nitrocellulose membrane using the iBlot 2 Gel Transfer Device (ThermoFisher Scientific). The membrane was then blocked with 5% non-fat milk in TBS with 1% Tween-20 at room temperature for 1h, followed by incubation with rabbit anti-TXNIP (1:1000; D5F3E, Cell Signaling Technology) and HRP conjugated mouse anti-β-actin (1:1000; sc-47778, Santa Cruz Biotechnology) antibodies in 5% milk TBST overnight at 4 °C on a low-speed shaker. The next day, the membrane was incubated with HRP conjugated goat anti-rabbit secondary antibody (1:2000, 7074, Cell Signaling Technology) at room temperature for 2 hours. Chemiluminescent substrate (Immobilon, Millipore Sigma) was then applied to the membrane for 1 minute and the protein bands were visualized with the ChemiDoc digital imager (Bio-Rad). Intensity of the bands was measured using ImageJ software, where the blots were given a greyscale value. β-actin was used for normalization.

IL-1β measurements

Sciatic nerves were collected from mice after the last week of injections and stored in ice-cold EDTA-treated tubes. Samples were then homogenized and centrifuged (10,000 rpm, 4°C, 10 min) to separate cellular debris. The supernatant was stored at −80°C. The IL-1β level in the supernatant was assessed using a commercially available, customized magnetic bead multiplex kit (Mouse IL-1 beta Simplex ProcartaPlex Kit, ThermoFisher). Cytokine analysis was performed using Mouse 10-Plex Panel for the Luminex^™ Platform according to the manufacturer recommended protocol, with the modification of performing an overnight 4°C incubation for the multiplex cytokines. Milliplex Analyst 5.1.0.0 software was used to calculate the level of cytokines from the median fluorescence intensity (MFI) of each bead.

Bioinformatics tools

RNA-seq data were processed by Partek Flow online software (now part of Illumina). Correlation analysis was performed using Inter Variability Cross-Correlation Analysis (IVCCA) (37) software developed in house using MATLAB 2024a. Non-Negative Matrix Factorization (nNMF) package was also developed in house using MATLAB 2024a, its code is available through Github. KEGG analysis and the heatmap representation was performed using Partek Flow.

RESULTS

Oxaliplatin induces strong alterations in treadmill running, NCV, and cold allodynia

Long-term behavioral changes corresponding to chronic CIPN are elusive in animal models, and isolating pain-related behavior from overall toxicity is challenging. We applied the DigiGait system, which utilizes ventral plane videography, to gather more information on behavioral changes during the chronic phase of pain induced by oxaliplatin. The procedure involves having mice walk or run on a translucent treadmill that controls the speed of ambulation. Among several measurements, swing/stance ratio and brake phase percentage showed significant treatment effects in the control vs 10 mg/kg oxaliplatin-treated group (Figure 1). For swing/stance ratio, there was a significant difference between the control group and 10 mg/kg oxaliplatin-treated group (p < 0.01), with increased time in the swing phase (% of stride) observed in the 10 mg/kg oxaliplatin-treated mice particularly in the hind paws. Pain or weakness are likely contributors to the increased time in the swing phase as has been earlier determined with paclitaxel (38). Reduced braking duration was also observed in 10 mg/kg group, suggesting impaired willingness or ability to decelerate before stance likely due to pain or discomfort. Other DigiGait parameters associated with pain-related behavior and showed significance in 10 mg/kg group were shown in Supplementary Figure S1 (a–e).

Figure 1.

DigiGait analysis of mice treated with oxaliplatin at different dosage. (a) Swing/Stance Ratio Hind and (b) % Brake Stride Hind (percentage of the total stride cycle that hind limb spends in the brake phase), were significantly decreased in the 10 mg/kg oxaliplatin-treated mice but remained unchanged in 5 mg/kg group. Data are presented as mean ± SD, n=6-10 with individual data points plotted. One-way ANOVA with multiple comparison, **p < 0.01, *p < 0.05.

Our data show an increase in % Swing Stride (Supplementary Figure S1c) and a decrease in % Stance Stride (Supplementary Figure S1d) in the hind paws of 10 mg/kg oxaliplatin-treated mice (Supplementary Figure S1 and Gait Data). The reduction in paw contact time suggests the animal is spending less time bearing weight on its limbs, potentially reflecting pain with mechanical contact. While indirect, these measures can be therefore used as a proxy for the von Frey test (39) indicating the mice developed mechanical allodynia after oxaliplatin treatment.

Notably, there was no significant change in any of the gait parameters, including Swing/Stance Ratio in hind paws (Figure 1a) and % Brake Stride in hind paws (Figure 1b), in the 5 mg/kg oxaliplatin-treated group compared to controls (Figure 1, Supplementary Figure S1 and Gait Data), suggesting that this lower dose of oxaliplatin did not induce observable neuropathy symptoms. Thus, subsequent behavioral and transcriptomic analyses focused on the 10 mg/kg group.

CIPN in humans is characterized by sensory and motor deficits, including a reduction in NCV. Patients with CIPN exhibit decreased sensory NCV and sensory nerve action potential (SNAP) amplitudes, which correlate with the severity of neuropathy (40). Patients with more severe grades of CIPN showed progressively lower NCV and SNAP values, indicating a dose-dependent relationship between chemotherapy exposure and nerve conduction impairment.

Similar to the clinical studies, our animal models also showed a substantial and highly statistically significant decrease in NCV values with oxaliplatin treatments (Figure 2a), suggesting a sensory deficit. The significant decrease in NCV in human patients was correlated with a strong increase in cold sensitivity, known as cold allodynia. According to the American Society of Clinical Oncology (ASCO) guidelines, oxaliplatin-induced neuropathy is characterized by cold sensitivity in the feet and hands, muscle cramps, and discomfort when swallowing cold liquids (10). These symptoms typically peak 2 to 3 days after each dose of oxaliplatin and worsen with subsequent treatment cycles (10).

Figure 2. NCV and the cold allodynia test of mice treated with oxaliplatin after 8 weeks of injection.

(a) NCV of oxaliplatin-treated mice was significantly lower than control animals. (b) Latency of oxaliplatin-treated mice, which represents the time that first pain-related behavior appears, decreased significantly. Number of control mice, n = 8; number of mice treated with 10 mg/kg dose, n=7. Data are presented as mean ± SD. Unpaired two-tailed t-test, ****p < 0.0001.

In mice, cold allodynia is typically measured using the cold plate test or the acetone drop test, which involve exposing the mice to cold stimuli (41,42). In our experiments, cold allodynia was assessed by observing the behavior of mice within 1 minute of placement on a cold plate at 4°C. As shown in Figure 2b, oxaliplatin significantly decreased the latency period indicating increased sensitivity of the mice to cold stimuli.

Electron microscopy reveals signficant mitochdrial damage from oxaliplatin

Transmission electron microscopy (TEM) images of the sciatic nerve from oxaliplatin-treated mice revealed substantial changes in mitochondrion morphology (cross sectional area/per mitochondria and vacuolated ratio) (Figure 3). Significantly enlarged and vacuolated mitochondria were observed in the sciatic nerves of oxaliplatin-treated mice (Figure 3a). In addition, abnormality in the structure of cristae, irregular mitochondrial outer membranes, swelling and vacuolation in the sciatic nerves of oxaliplatin-treated mice indicated substantial damage to the mitochondria (Figure 3b–3c). Similar changes in mitochondria have also been noticed in the sciatic nerves of cisplatin-treated mice, suggesting a common mechanism across platinum-based and potentially other drugs (43).

Figure 3. Oxaliplatin leads to abnormal mitochondrial morphology in axons.

(a) Representative TEM image of the sciatic nerve from control and oxaliplatin-treated mice. Enlarged and vacuolated mitochondria were seen in oxaliplatin group. Mitochondria are highlighted with red arrows, scale bar: 2μm. (b) Mitochondrial cross-sectional area of oxaliplatin-treated nerves were significantly larger compared to controls. 48 (control) and 58 (10 mg/kg) mitochondria from n=3/group mice were measured. (c) The number of vacuolated mitochondria in sciatic nerves increased significantly after oxaliplatin treatment. n=3. Data are presented as mean ± SD with individual data points plotted. Unpaired two-tailed t-test, *p < 0.05, **p < 0.01.

RNA-seq analysis highlights oxidative stress response

Our analysis of DRG genes via RNA-seq identified Txnip as a significantly upregulated gene, validated through qPCR, Western blotting, and immunohistochemistry, establishing its potential role in oxaliplatin-induced CIPN.

Following the establishment that animal models develop oxaliplatin-induced neuropathy similar to humans, we then investigated the differential expression of genes in DRG between the control and oxaliplatin-treated mice using bulk RNA-seq of DRG. From 13,145 identified genes in DRG, we found 820 differentially expressed genes (DEGs) (FDR<0.05) with 271 genes that were differentially expressed with threshold |FC| ≥1.5 and 91 genes with |FC| ≥ 2 (shown in Figure 4a). A heatmap shown in Figure 4b evidences an excellent separation between the control and oxaliplatin-treated group of mice, suggesting a significant impact of oxaliplatin on the gene expression in DRG.

Figure 4. Effect of oxaliplatin on the RNA transcription in DRG.

(a) Volcano plot of genes identified in the DRG tissue: vertical dashed lines are the FC ≤ −2.0 and ≥ 2.0, horizontal dashed line is the false discovery rate (FDR) threshold < 0.05. Total number of identified genes in DRG =13,145, total number of DEGs with the specified criteria 91. Among them, upregulated genes = 66, downregulated genes = 25. (b) Heatmap of 91 DEGs in DRG show clear separation between the treatment groups: control – upper 5 rows, oxaliplatin treated - bottom 5 rows. n=5 for each group. (c) Top 15 statistically significant KEGG pathways affected by the oxaliplatin (p<0.05). Most of the top affected pathways represent an inflammatory phenotype (performed by Partek software using 91 DEGs).

Analysis of the DEGs with the highest response (by fold change) span a range of functions, including oxidative stress (Txnip, Mt2), inflammation (Cd74, Ly86), response to hypoxia (Hif3a), lipid metabolism (Plin4, Etnppl) and tissue remodeling (Adm, Timp4). Although these genes have diverse roles, their simultaneous differential expression suggests a complex biological scenario where multiple cellular processes are affected. Interestingly, the set of DEGs identified in our experiments from the 10 mg/kg 8 week oxaliplatin-treated group (cumulative dose of 80 mg/kg) was substantially different from the published set of DEGs reported after a single dose of oxaliplatin of 40 μg (equivalent to a 2 mg/kg dose for a 20 g mouse) (44) with no overlap between the DEG profiles. Apparently, there is a strong difference between the chronic (our approach) and the acute form of CIPN described in reference (44) .

Pathway Enrichment Analysis (PEA) (45) of 91 genes that met the fold change |FC| ≥ 2.0 requirement identified KEGG pathways with the top affected pathways, not surprisingly, associated with inflammation and cytokine-cytokine interaction (Figure 4c).

In parallel with PEA, we also employed Non-Negative Matrix Factorization (nNMF), often used to uncover novel gene clusters and latent patterns in the gene expression data (46). While PEA provides a top-down approach, using existing biological knowledge to interpret gene expression changes, nNMF offers a bottom-up approach, discovering novel patterns or clusters in the data directly. PEA is primarily used for validating known biological processes affected by treatment (e.g., inflammation, oxidative stress). In contrast, nNMF often reveals unexpected or novel gene associations that are not captured by standard pathway annotations. For nNMF we focused on the 91 genes with a strong differential expression (|FC| ≥ 2), to show clear patterns and biological relevance with minimum noise from weaker DEGs. To facilitate the interpretation of the genes in the cluster, we limited the study to a two-cluster analysis.

We first deconstructed the Original matrix V (10 samples x 91 genes) (Figure 5a) into two matrices: Loading matrix H (10 samples x 2 clusters) (Figure 5b–5c) and Basis matrix W (2 clusters x 91 genes) (Figure 5c–5d). A clear separation between the groups in the Basis Matrix (W) is an indicator that the nNMF analysis captured meaningful biological variation between the two groups (Figure 5c). Cluster 1 was enriched with genes primarily associated with immune system pathways, many of which are significantly downregulated in the treatment group compared to the control group. The top 10 genes from this cluster are highlighted in Figure 5e, including key immune-related proteins such as Cd74 (FC = −4.27), H2-Ab1 (FC = −4.25), H2-Aa (FC = −4.52), and H2-Eb1 (FC = −3.60). These genes are central players in antigen presentation and immune response, and their downregulation suggests a suppression of immune activity in the treatment group.

Figure 5. NMF analysis on gene expression data from DRG of oxaliplatin-treated and control mice.

(a) Original matrix V where individual mice correspond to rows and genes expressions are stored in columns, (b): nNMF decomposes matrix V into the product of two lower rank matrices W and H so that V is approximately equal to W × H; (c) Bar plot of the Basis matrix W, showing how strongly each mouse associates with the identified clusters, with each bar representing a mouse; (d) Bar plot of the Loading matrix H, showing the loading coefficients of genes associated with each latent factor, allowing for easy identification of key genes contributing to each cluster; (e) and (f) Bar plots of the top 10 genes for Cluster 1 and Cluster 2, respectively, ranked by loading coefficients. These plots display the genes most strongly associated with each cluster.

Genes in Cluster 2 appear to represent processes activated specifically by oxaliplatin treatments. This cluster prominently features Txnip (Thioredoxin-Interacting Protein) (FC = +2.25) and Metallothionein 2A (Mt2) (FC = +3.43) (Figure 5f), suggesting a response to a strong oxidative stress. Txnip is known to increase oxidative stress by inhibiting thioredoxins that control ROS in both cytosol and mitochondria (47) (see the Discussion). Mt2 responds to oxidative stress and heavy metal exposure (such as Pt), acting both as an antioxidant and participating in detoxification, which is particularly relevant in the context of oxaliplatin (48). Other top genes in this cluster such as Matrix gla protein (Mgp) (FC= +2.04) and Fibronectin 1 (Fn1) (FC = +2.00) are associated with extracellular matrix (ECM) regulation and tissue remodeling. Their activation reflects structural adaptations in response to cellular damage (49).

Protein level of TXNIP is increased in DRG of oxaliplatin treated mice

The observed upregulation of Txnip in DRG from RNA-seq after oxaliplatin treatment was validated with qPCR (Figure 6a), Western blot (Figure 6b–6c, also Figure S4) and imaging studies using anti-TXNIP and fluorescently labelled secondary antibodies (Figure 6d–6f). All three methods demonstrated significant upregulation of TXNIP at the transcriptional and protein levels. Colocalization of NeuN, a neuronal marker, and TXNIP suggested that TXNIP remains neuron specific after the oxaliplatin treatment (Figure 6d–f) making it a specific target for quenching oxaliplatin-induced peripheral neuropathy.

Figure 6.

Oxaliplatin lead to up-regulation of TXNIP in DRG. (a) qPCR shows higher Txnip RNA level in DRG from oxaliplatin treated mice. The amount of RNA was normalized to control group. n=5. (b) Western blot of TXNIP protein in DRG (the full Blot is shown in Figure S4). (c) Intensity of the bands was measured in ImageJ and normalized to control group. n=5. (d) IF staining of DRG slides from the control group with obj 10X, and (e) obj 40X, (f) IF staining of DRG slides from the oxaliplatin group with obj 10X, and (g) obj 40X. TXNIP staining was brighter in oxaliplatin group. (h) Quantification of fluorescence signal from TXNIP antibody of the IF images. Mean grey value of the green channel was measured with ImageJ. 2–4 slides from each mouse, total 6–10 slides were measured for each group. Individual data points are the average for each mouse. Data are presented in mean ± SD. Unpaired two-tailed t-test, *p < 0.05, **p < 0.01.

Inflammatory response

The acute inflammatory response in the injured DRG is generally characterized by elevated levels of pro-inflammatory cytokines, like IL-1β and TNF-α (15), which contribute to the initiation of pain hypersensitivity (50). In contrast, chronic pain is associated with a more sustained but less intense inflammatory response (51). Over time, the number of acutely acting inflammatory cells in the DRG are expected to decrease, although a baseline level of inflammation persists, contributing to the maintenance of chronic pain.

Relatively mild changes in the expression of inflammatory cells’ markers aligned with only a few differentially expressed cytokines including downregulated Ccl2 (FC = −2.32) and upregulated Ccl6 (FC =+2.91) and Ccl9 (FC =+2.23). Ccl2 is a key chemokine for recruiting monocytes and macrophages (52,53) and its downregulation likely reduced the influx of these cells into the DRG, suppressing the acute inflammatory response and associated pain hypersensitivity. Conversely, the upregulation of Ccl6 and Ccl9 facilitated the recruitment of immune cells, such as mast cells, and the release of pro-inflammatory cytokines, contributing to continued neuroinflammation and chronic pain (54). This complex interplay reflects a transitional stage between acute inflammation induced by initial oxaliplatin injections and the sustained chronic neuroinflammation state.

The observed increased level of the pro-inflammatory cytokine IL-1β in the sciatic nerve from mice treated with oxaliplatin is in line with the release of IL-1β in vitro in the oxaliplatin-treated satellite glial cell cultures derived from DRG (55). IL-1β is well known to be associated with pain: direct intraneural injection of IL-1β at a low level of 2.5 pg/mL into the rat sciatic nerve led to pain-related behavior (56). In our work, IL-1β levels (1.5–3 pg/mg) were observed in the sciatic nerves of the oxaliplatin-treated mice (Figure S2a). The mechanism of IL-1β generation, however, remains unclear. The inflammasome complex composed of NLRP3 proteins is considered a critical step in IL-1β generation (57). However, our RNA-seq data show barely detectable levels of Nlrp3 in the DRG, a finding also confirmed by qPCR (Figure S2b) with no changes in in expression of Nlrp3 in response to oxaliplatin treatment. Apparently, inflammasomes do not play an active role in the pain mechanism associated with oxaliplatin treatment. A similar conclusion has been reported in the literature from studies on neuropathic pain in lipopolysaccharide-treated mice (58).

Expression of transient receptor potential (TRP) channels weakly responds to oxaliplatin

The hallmark of CIPN characterized by oxaliplatin-induced cold hypersensitivity has traditionally been associated with alterations in ion channel activity, particularly transient receptor potential (TRP) channels. Oxaliplatin has been previously shown to increase the activity of TRPA1 and TRPV1 (59,60) as well as TRPM8 (61). In our study, none of the genes encoding these TRP channels showed significant differential expressions. Mild differential expressions were observed in some other TRP channels: Trpv2 (FC= −1.22), Trpm2 (FC= −1.21), and Trpc7 (FC= −1.57). While these transcriptional changes are relatively small and negative, they do not necessarily rule out a functional contribution of TRP channels to oxaliplatin-induced CIPN. Post-transcriptional regulation, protein modifications, and channel sensitization likely play critical roles in modulating TRP channel activity, e.g. through oxidation of the TRPA1 channel with ROS that leads to channel opening.

DISCUSSION

Summary of Key Findings

Using a human equivalent dose of systemic oxaliplatin administration injected into mice over 8 weeks, we observed symptoms in mice characteristic of human chronic CIPN. That included increased cold sensitivity, decreased NCV, and gait behavior changes. We also observed significant histological, genetic and molecular changes in many organs including DRG which was the focus of this work. The collected data were critical to propose a new mechanism of oxaliplatin-triggered chronic CIPN based on the critical and potentially central role of Txnip as a regulator of oxidative stress.

Activation of TXNIP by hypoxia

Oxaliplatin treatment has been reported to result in severe loss of red blood cells carrying oxygen to tissues in both preclinical and clinical settings (62,63). The lack of oxygen can lead to hypoxia in many organs including the DRG causing the upregulation of hypoxia-inducible factors. Hypoxia in DRG has been shown to cause ectopic neuronal firing, a phenomenon closely associated with radiculopathy-related pain (64). Studies in animal models (64) and patients (65,66) have demonstrated that hyperbaric oxygen therapy provided therapeutic benefit to neuropathic pain conditions.

Hif3a, a critical regulator of cellular responses to low oxygen levels, is overexpressed in oxaliplatin-treated mice (FC = +4.01) suggesting a strong hypoxic phenotype. Overexpression of the facilitative glucose transporter GLUT1 (encoded by Slc2a1, FC = +1.31) that increases glucose uptake to meet the high metabolic demands of DRG cells under low oxygen conditions also indicates a potential hypoxia condition in DRG. Whether the hypoxia in DRG directly stimulates the overexpression of Txnip is not clear. However, a strong correlation between Hif3a and Txnip expression (p=0.95) indicates hypoxia as a plausible trigger for Txnip upregulation. Given the established role of Txnip in regulating oxidative stress, this correlation provides a mechanistic link between Txnip upregulation, oxidative stress and hypoxia in oxaliplatin affected DRG.

Activation of TXNIP through a p53/p21 pathway

Once oxaliplatin enters the cell, the aquation reaction removes the oxalate ligands, “activating” the platinum (Pt) to form more reactive species (67). The formed reactive, positively charged Pt complex has a higher affinity toward guanine residues in DNA that leads to the formation of Pt-DNA adducts, which are responsible for the cytotoxic effects of the drug (67). The reactive Pt species can also form intra- and interstrand cross-links in DNA that trigger cellular stress response pathways, such as p53/p21 signaling, leading to DNA damage response mechanisms or apoptosis (68). The apoptotic response pathways are central to the cytotoxic effects of platinum-based drugs in cancer treatment but also to leads to the off-target affects in other organs including DRG.

While the expression of p53 (encoded by Trp53) in the DRG was not affected by oxaliplatin (FC = 1.01), we observed a significant upregulation of p21 (encoded by Cdkn1a, FC = +1.8) in line with the previously published effect of oxaliplatin on cancer cells (69). p21 is strongly expressed in DRG compared to other major organs, in part due to the high demand for neuronal survival and differentiation. In non-mitotic cells, such as neurons and other cells in DRG, the increased expression of p21 suggests the induction of repair mechanisms and cell cycle arrest, providing time for DNA repair. The evidence suggests during this time TRX can be activated contributing to the p53 DNA binding, DNA repair, cell cycle arrest, and/or apoptosis (70). A strong correlation between Cdkn1a and Txnip expression (p=0.82) supports the involvement of the p53/p21 pathway in Txnip activation, although details of these mechanisms are currently lacking. One of the missing links to upregulate Txnip could be promyelocytic leukemia zinc finger (PLZF), encoded by Zbtb16 (FC=4.11), associated with the upregulation of Txnip in type 2 diabetic mice (71). This transcription factor has the highest correlation to all other genes based on IVCCA analysis of the DRG of oxaliplatin-treated mice (Figure S3b). Recently Txnip has been identified as the key downstream gene of Zbtb16 assay in atria (71) demonstrating their role in oxidative stress regulation. The strong co-expression of Zbtb16 gene with Txnip (p=0.92) and Cdkn1a (p= 0.91) suggests a plausible Cdkn1a-Zbtb16-Txnip axis underlying Txnip overexpression in DRG cells.

TXNIP regulates oxidative stress

The cellular level of oxidative stress is highly controlled by the thioredoxin system that maintains a reducing intracellular microenvironment by controlling and removing intracellular ROS (72). This system is composed of TXNIP, thioredoxin proteins TRX1,2 (encoded by Txn1,2 genes), thioredoxin reductases and other redoxins (73,74). Among the genes that encoded these biomolecules only Txnip was statistically significantly differentially expressed (Txn1 expression was significantly different, but showed low fold change, FC = −1.17). TRX1 in the cytosol and TRX2 in the mitochondria act as reducing agents, catalyzing the reduction of disulfide (cystine) bonds in oxidized proteins restoring their functions and protecting cells from oxidative stress. TRX1 and TRX2 also remove ROS radicals via two closely located thiol groups forming a cystine bond preventing ROS from oxidizing proteins, lipids, and DNA. TXNIP regulates this process by binding to TRX1,2. When overexpressed, TXNIP inhibits TRX1,2 via binding and prevents their reactions with ROS. This decreased antioxidant capacity within the cell allows ROS to oxidize biological molecules as illustrated in Figure 7.

Figure 7. Suggested mechanism of TXNIP in regulating oxidative stress.

Under normal conditions (left panel), cytosolic thioredoxin 1 (TRX1) and mitochondrial thioredoxin 2 (TRX2) play a critical role in suppressing reactive oxygen species (ROS, shown as star shapes) and maintaining cellular redox balance. Upon oxaliplatin treatment (right panel), oxaliplatin enters the cell through transporters (potentially OCTN1, encoded by Slc22a4 (81) or CTR1, encoded by Slc31a1) (82)) and moves further to nucleus. DNA damage triggers the p53/p21 defense pathway, while reduced oxygen delivery to DRG induces hypoxia, likely due to impaired red blood cell counts. As a result, a transcription factor Zbtn16 upregulates Txnip. Overproduced TXNIP proteins translocate to cytosol and mitochondria, where TXNIP binds and inhibits TRX1 and TRX2 correspondingly, impairing their ability to neutralize ROS. This inhibition leads to ROS accumulation, initiating mitochondrial dysfunction and eventual mitochondrial death. The resulting cellular distress activates inflammatory pathways, promoting the production of pro-inflammatory cytokines (e.g., IL-1β). These cytokines recruit inflammatory cells, ultimately contributing to the development of chronic neuropathic pain and maladaptive plasticity in the DRG.

Txnip is tightly correlated with Mt2 (p=0.95) as well as other ROS related genes such as Sesn1 (p=0.93, FC=1.55) whose product protects cells from oxidative stress (75) and Trp53inp1 (p=0.9, FC=1.78) that encodes a protein that controls the cell’s redox status by regulating mitochondrial homeostasis. Importantly, TRP53INP1 is essential for the removal of damaged mitochondria and the maintenance of mitochondrial integrity (76) and thus its upregulation makes sense given the massive deaths of mitochondria in sciatic nerve (Figure 3). Txnip is also tightly correlated with the cytochrome P450 2D22 protein (encoded by Cyp2d22, p = 0.98, FC = +1.82) and lipoxygenase LOX (Lox, p = 0.93, FC = +2.26) that when overexpressed produce ROS (77,78). Such strong correlations between a variety of ROS related genes reflect a coordinated cellular strategy to respond to oxaliplatin.

Proposed mechanism of chronic pain

How the upregulation of Txnip results in the painful CIPN phenotype of the chronically treated mice is largely unknown. Based on our results and the published literature, we propose the following sequence of events after oxaliplatin enters neuronal cells, as schematically illustrated in Figure 7.

Following venous administration in humans, oxaliplatin reaches the DRG through systemic blood circulation. Due to the incomplete blood-nerve barrier in DRG (79) oxaliplatin can readily penetrate DRG (80). Once in the DRG, oxaliplatin penetrates cells (neurons, satellite glial cells through organic cation transporters, such as OCTN1 (encoded by Slc22a4) (81) or CTR1 (encoded by Slc31a1) (82). Notably, the expression of another transporter, OCT2 (Slc22a2) often mentioned in literature as the key transporter of oxaliplatin (83), was undetectable in the DRG in our mouse models.

Both hypoxia and activated p53/p21 pathway can drive upregulated expression of Txnip in the nucleus. Being overexpressed, TXNIP proteins translocate from the nucleus to other cellular compartments, including the mitochondria (84,85). Here, TXNIP inhibits the antioxidant thioredoxins (TRX1,2), resulting in an increase in ROS level, which causes lipid peroxidation and damages both cellular and mitochondrial membranes perpetuating a cycle of oxidative stress. The produced cytokines such as IL-1β leads to attracting inflammatory cells such as mast cells to the DRG. This cascade of events ultimately contributes to the development of CIPN.

Sustained oxidative stress and the resulting damage to the structural architecture of neurons are expected to change the plasticity in DRG neurons. Indeed, our results show that oxaliplatin leads to the overexpression of Adm (FC=4.01), which encodes adrenomedullin (ADM) - a known mediator in the progression of chronic neuropathic pain that plays a critical role, particularly in the late-phase, in the development of neuropathic pain (86). The strong correlation between Txnip and Adm (p =0.95) highlights a potential link between oxidative stress and maladaptive plasticity within the nociceptive system.

Clinical/Translational Implications: TXNIP pathway as a therapeutic target in CIPN caused by oxaliplatin

Targeting TXNIP through localized therapeutic strategies (e.g., intrathecal delivery of TXNIP inhibitors or gene silencing) could provide a novel, precise approach for CIPN management. Enhanced ROS generation has long been suggested as one of the primary mechanisms of peripheral neuropathy induced by platinum-based chemotherapy drugs (12,87), yet past and current clinical trials targeting ROS have only provided mild pain relief (88). Based on our results, a TXNIP-centered pathway appears to be a promising target for oxaliplatin-induced pain. Recently, several groups found that suppressing the TXNIP pathway benefits diabetic and other neuropathies. For instance, inhibiting TXNIP with verapamil attenuated prediabetic neuropathy in mice fed a high-fat diet (30). Similarly, TXNIP has been identified as a key factor in Schwann cell dysfunction in diabetic peripheral neuropathy, and its knockdown reduced autophagy and lowered apoptosis in high glucose-treated cells (89). Our previous research on the extracted DRG showed the benefits of suppressing TXNIP, which resulted in lower oxidative stress and partially restored DRG overgrowth (90).

However, Txnip is also known as a tumor suppressor gene, with most tumors exhibiting downregulated levels of Txnip (91). Therefore, systemic treatment with TXNIP inhibitors might promote cancer cell proliferation if administered during cancer treatment. To mitigate this risk, alternative strategies such as epineural injections of vectors to knock down Txnip expression could be developed. Although highly invasive, this targeted method promises to provide clinical benefits by confining the treatment to the DRG, thus minimizing potential effects on tumors. Gene silencing using AAVs or non-viral alternatives such as siRNA, antisense oligonucleotides, or CRISPRi offers high specificity and flexibility for this task. For example, recent preclinical studies have demonstrated the potential of AAVs in pain modulation (92). We envision that direct injection of Txnip-silencing vectors into the DRG, or delivery via intrathecal injection (93), may enable targeted therapy localized to TXNIP-related pain pathways without systemic exposure. This will be crucial in the treatment of CIPN, where off-target effects could interfere with oxaliplatin’s action and/or exacerbate cancer progression.

Summary

Our study explores the impact of oxaliplatin, a common chemotherapy drug, on developing chronic CIPN in mice with a focus on oxidative stress pathways within DRG and sciatic nerve. Oxaliplatin, while effective in cancer treatment, frequently results in chronic and debilitating neuropathy in patients, for which effective treatments are currently lacking. In our mouse model, we observed that chronic oxaliplatin administration with a human equivalent dosage at 10 mg/kg over 8 weeks led to CIPN-like symptoms, including increased cold sensitivity, lowered nerve conduction velocity and changed gait. Histological analysis revealed strong mitochondrial damage in the sciatic nerve while RNA sequencing of DRG showed altered gene expression with a shift towards a pro-oxidative state. A variety of bioinformatics tools revealed Txnip may contribute to this process, suggesting that targeting TXNIP can offer a promising therapeutic strategy to reduce oxaliplatin-induced nerve damage and neuropathic pain. These data demonstrate that TXNIP is associated with, and likely contributes to, CIPN. Additional work will be needed to provide more direct evidence, such as using knockout or overexpression models in mice. Future work focusing on validation of this new target in mouse models and human tissues will provide a promising strategy to develop new treatments to prevent oxaliplatin-evoked CIPN without diminishing cancer treatment efficacy.

Availability of data and materials:

1. Complete RNA-seq data for DRG are deposited in the GEO database GSE286387, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE286387; a token for reviewers to visualize data: ilupuceghzkxreh

2. Matrix with 10 mice and 91 DEGs; gene expressions are normalized. See file: DRG 91 genes FDR 0.05 FC more or equal 2 sorted.xls located in https://github.com/MikhailBerezin/NMF-Gene-Clustering

3. Function nnmf_rna.m for running nNMF in MATLAB (tested in 2022A and 2024A versions). https://github.com/MikhailBerezin/NMF-Gene-Clustering