Sensory neuron dysfunction in orthotopic mouse models of colon cancer

The MC38 colon adenocarcinoma cell line (C57BL/6 origin) was obtained from Kerafast, Boston, USA (catalog # ENH204-FP). Cells were maintained in 10 cm TC dishes at 37 °C, 5% CO₂. Cell culture media: Dulbecco's modified MEM (DMEM) with 10% fetal bovine serum (FBS), 2 mM glutamine, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 10 mM HEPES, 1 × penicillin/streptomycin. Firefly luciferase-expressing CT26 colon carcinoma cells (Balb/c origin) were obtained from Cellomics (catalog #: SC-1298) and maintained at 37 °C and 5% CO₂ in RPMI-1640 Medium with 10% FBS and 1 × penicillin/streptomycin.

In order to non-invasively measure tumor growth in C57BL/6 mice, we transduced MC38 cells with a lentivirus (catalog # FCT160; Kerafast, Boston, USA) containing the firefly luciferase (CMV) and neomycin resistance genes. Our procedure was based on the provider’s protocol [18]. Briefly, cells were plated in six well cell culture plate in 2 ml growth medium. After cell growth reached ~ 80% confluence (48 h), growth medium was aspirated and 2 ml transduction medium was added, containing 7.5 µl (approx. 7.5 × 10⁵ CFU) of the virus. 24 h later, cells were selected for infection by adding 400 µg/ml G418 to the growth medium [19]. This concentration was the minimum required to kill all uninfected MC38 control wells within 7 days. All clones were expanded and stocks cryopreserved. Firefly luciferase expression was verified in vitro with the addition of 150 µg/ml luciferin (catalog #: LUCK-1G; Goldbio) immediately prior to reading luminescence with a Biotek Synergy plate reader.

All experiments and procedures involving the use of mice were approved by the University of Texas MD Anderson Cancer Center Animal Care and Use Committee at MD Anderson Cancer Center (Houston, TX, USA) in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals. Every effort was made to minimize the number of mice used and their suffering. Animals were randomly assigned to individual experimental groups. 8- to 14-week-old male mice (weighing 22–30 g) were used for each experiment. Male C57BL/6J (catalog #: 000664) and BALB/c (catalog #: 000651) mice were obtained from Jackson Laboratories and kept under 12:12 light:dark cycle (07:00 to 19:00 h) with access to food and water ad libitum. All mice were habituated to the housing environment for at least 4 days before experiments began.

Colon tumors were induced using minimally invasive transanal rectal injection, based on the methods of Donigan et al [20]. Under isoflurane anesthesia (3% isoflurane in oxygen), 2 × 10⁵ luciferase-expressing MC38 or 1 × 10⁵ luciferase-expressing CT26 cells in 50 µl growth medium were injected using a 0.3 ml (U30) insulin syringe, into the colorectal wall of C57BL/6 and BALB/c mice, respectively. Control mice were anesthetized and injected with 50 µl vehicle (cell growth medium) in the same manner. No animals died or had to be euthanized as a result of tumor development during the 4-week experimental period.

Tumor growth was followed by in vivo non-invasive bioluminescent imaging (BLI) using an IVIS XR Lumina (PerkinElmer), as described previously, with slight modifications [21]. Mice were injected with D-luciferin (150 mg/kg, i.p. dissolved in sterile PBS). After 10 min, a 1-min exposure of the ventral aspect was captured. Luminescence was quantified within a circular ROI overlying the colorectal region for each animal and expressed as average radiance (p/s/cm²/sr) by the Living Image software (PerkinElmer). Tumor- and vehicle-injected mice were measured weekly. Animals not developing tumors (as indicated by elevated signal intensity of at least 500 p/s/cm²/sr on the second week after injection) were excluded from experiments and did not exhibit visible tumor development upon necropsy (less than 5% of mice, data not shown).

All behavioral assays were performed in the same room as which mice were housed. Before the start of any experiment, mice were habituated to the testing apparatus for 2 days prior to initiating testing. Experimenters were blinded to tumor/vehicle injection to the fullest extent possible, however, 4 weeks post-injection, swelling of the tumor mass in the perianal area was often readily apparent. All behavioral testing was carried out during the light cycle, between 08:00 and 12:00.

Dorsal root ganglia and 3-mm plantar punch biopsies of hind paws and fore paws were harvested after perfusion with 4% paraformaldehyde (PFA) in phosphate buffer (PB), pH 7.4, as described previously [31, 32]. At least 3 biological replicates/groups were measured from at least two different cohorts of mice for all tissues.

Plantar tissues were post-fixed in 4% PFA overnight before being transferred into 15% sucrose and 10% EDTA in PBS for 48 h for decalcification of bones, as described previously [33, 34]. After further incubation in 30% sucrose overnight, 40 µm lateral sections (skin to skin) were prepared with a Leica CM3050S cryostat and collected onto microscope slides. Immunohistochemistry was performed as previously described [31, 32]. After three 5-min washes with 0.1 M PB, slides containing sections were blocked and permeabilized with 10% goat serum, and 0.3% Triton X-100 in 0.1 M PB for 1 h at 4 °C. Sections were then incubated overnight at 4 °C in primary antibodies diluted in blocking buffer (rabbit anti-PGP9.5: Thermo Fisher Scientific, catalog #38-1000; dilution: 1:500; rat anti-CD68, Bio-Rad, catalog #MCA1957GA; dilution: 1:300). The following day samples were washed 3 times in blocking buffer and then incubated with secondary antibody (1:1000 IgG-Alexa Fluor 488 goat anti-rabbit and goat anti-rat 568 antibody in blocking buffer and 1 µg/ml DAPI for 4 h, protected from light at 4 °C. Slides were then washed 3 times in 0.1 M PB, mounted with ProLong Gold anti-fade reagent, cover-slipped, sealed with nail polish and kept at − 20 °C until imaging with a Nikon Ti2 confocal microscope. Images were taken at 20 × magnification. Images are a composite of 10 focal planes in a 40-µm z-stack at 4 µm increments. Tissue samples from at least 3 mice/experimental group were imaged; 3–4 different sections were imaged from the same animal. Following image acquisition, intraepidermal nerve fiber (IENF) density was determined in the plantar skin as described previously [35‐37] with the Nikon NIS-Elements Imaging Software and presented as number of IENFs/100 µm. CD68 signal intensity of the same tissue samples was determined in the hind paw skin by the application of ROIs from the edge of the plantar skin to 200 µm below the epidermis.

DRG samples were harvested and handled similarly to plantar punch tissue samples, as described previously [21, 31, 33]. Lumbar DRGs were harvested under a dissection microscope into 4% PFA, then placed into PBS containing 15% and 30% sucrose solutions in PBS, each overnight. 25-µm sections were stained using the same procedure as for plantar skin samples. For the stains performed using NF200, 1% BSA was added to the blocking buffer. After blocking for 1 h, sections were washed with PBS-Tween 20 (0.05%) and blocked with Goat F(ab) anti-mouse IgG H&L (Abcam, catalog#ab6668; dilution 1:50) in PBS for 1 h. The sections were rinsed again with PBS-Tween20 before the primary antibody was added in blocking buffer. The following primary antibodies were used for DRG immunofluorescence: rabbit anti-activating transcription factor 3 (ATF-3; Novus Biologicals, catalog #NBP1-85816; dilution: 1:300), rat anti-CD68 (Bio-Rad, catalog #MCA1957GA; dilution: 1:300), rabbit anti-CXCL10 (Thermo Fisher Scientific, catalog #701225; dilution 1:200), rabbit anti-CXCR3 (Thermo Fisher Scientific, catalog #PA5-23104; dilution 1:200), and mouse anti-NF200 (Sigma-Aldrich, catalog #N0142; dilution 1:200). Secondary antibodies: IgG-Alexa Fluor 594 goat anti-rabbit, IgG-Alexa Fluor 488 goat anti-rat (dilution: 1:1000), IgG-Alexa Fluor 488 goat anti-rabbit (dilution: 1:500), and IgG-Alexa Fluor 555 goat anti-mouse (dilution: 1:500). DAPI (1 µg/ml) was added together with the secondary antibodies.

To determine tumor-related effects on neuronal mitochondrial bioenergetics, DRGs of tumor-bearing and control animals were investigated by the XF^e24 Extracellular Flux Analyzer (Seahorse Bioscience), as described previously [35]. Immediately after collection of lumbar (L2–L5/6) DRG from tumor-bearing and control mice into Ham’s F10 media (Thermo Fisher Scientific, USA), connective tissue was removed, ganglia were enzymatically dissociated, and an enriched neuronal fraction was obtained [38]. Cells were plated into poly-DL-polyornithine (0.01%)/laminin (2 µg/mL)-coated XF24 plates (Seahorse Bioscience, Santa Clara, CA) in Ham’s F10 medium containing N2 supplement without insulin at approximately 1,000 cells/well (Invitrogen). 24 h after plating, oxygen consumption rate (OCR) and the extracellular acidification rate (ECAR) was analyzed by the application of an ATP synthase inhibitor (2 µM oligomycin), a protonophore that dissipates the proton gradient across the inner mitochondrial membrane (4 µM carbonyl cyanide p-trifluoromethoxyphenylhydrazone, FCCP), and the combination of a complex I and II inhibitor (2 µM rotenone and antimycin A, respectively). This enables determination of the portion of basal OCR coupled to ATP synthesis, the maximal respiratory capacity, and the amount of non-mitochondrial respiration [35, 39]. Dissociated DRGs were measured in an assay cycle of [3 min mixing, 2 min pause, 3 min measurement], repeated 3 times. Data were normalized to total protein levels, measured by BCA Protein Assay according to the manufacturer’s instructions (Thermo Scientific; USA).

The live-cell calcium imaging (Ca²⁺ imaging) experiments were performed on cultured mouse DRG neurons as previously described [21, 40]. DRGs were collected from C57BL/6 mice as detailed in [41]. Isolated DRGs were digested with collagenase and pronase (2 mg/ml and 1 mg/ml in DMEM, respectively). Before plating, the prepared cell suspension was centrifuged through a 1:1 28% and 12.5% Percoll density gradient (3900 rpm for 10 min), in order to remove non-cellular debris [42]. Thereafter, cells were plated onto poly-L-ornithine- and laminin-coated 12 mm circular glass coverslips and incubated in 1:1 TNB media supplemented with protein–lipid complex (Biochrom) and DMEM + 10% FBS at 37 °C in 5% CO₂ for 1 to 2 days. Calcium imaging was conducted using Fura 2-AM, as described previously [21, 43]. Briefly, DRG neuron coverslips were incubated with the Ca²⁺-sensitive dye Fura-2- AM (Invitrogen, 2 μM) at room temperature, for 20 min, dissolved in HEPES-buffered HBSS (‘HH buffer’) containing the following (in mM): 140 NaCl, 5 KCl, 1.3 CaCl₂, 0.4 MgSO₄, 0.5 MgCl₂, 0.4 KH₂PO₄, 0.6 NaHPO₄, 3NaHCO₃, 10 glucose, and 10 HEPES adjusted to pH 7.4 with NaOH and adjusted to 310 mOsm with sucrose. Coverslips were then placed in an MS-512SP recording chamber (ALA scientific Instruments) on the stage of an inverted Nikon Ti2 microscope and continuously superfused for 5 min with HH buffer, fed through a valve-controlled gravity perfusion system with an approximate flow rate of 1 ml/min. The applied parameters during imaging: alternated Fura-2 excitation at 340 and 380 nm, 2 nm band-pass, 50 ms exposure) at 1 Hz using a Lambda LS Xenon lamp (Sutter Instruments, Novato, CA, USA) and a 10x/NA 0.5 objective. Emitted fluorescence was collected at 510 nm (sCMOS pco.edge camera). The ratio of fluorescence (F340:F380) was calculated. The baseline recording of F340:F380 ratio without stimulus was performed to obtain an indication of resting intracellular calcium concentration ([Ca²⁺_i]) under continuous superfusion of HH buffer at 37 °C for 60 s. Thereafter, neurons were stimulated by the application of 25 mM KCl solution in HH buffer, superfused for 15 s, followed by superfusion with HH buffer for another 60 s to wash out KCl. 50 mM KCl was applied at the end of the recording to identify all viable, excitable cells. Only neurons responding to 50 mM KCl were included in the data analysis. Imaging was analyzed with Nikon NIS Elements software.

DRGs were collected from C57BL/6 mice, digested and cleared of non-neuronal cells and debris as detailed in [41, 42]. Thereafter, cells were plated in 10 µl droplets (approximately 10⁴ neurons per well) of 1:1 TNB media supplemented with protein–lipid complex (Biochrom) and DMEM + 10% FBS onto six wells of a 24-well CytoView MEA plate (Axion Biosytems). After 90 min, 365 µl of media was added to each well. Four hours after plating, a 15 min recording of spontaneous activity each well was carried out on an Axion Maestro Edge at 37 °C and 5% CO₂. Continuous data were acquired simultaneously at 12.5 kHz per electrode (16 electrodes per well). Individual spikes were detected by filtered continuous data crossing of a 6-µV static threshold. Spike number and firing rate were calculated using Axion Neural Metrics Tool. Immediately after the 4-h recording, 375 µl of MC38 cell culture medium or MC38-conditioned media was added to each well. Conditioned media were collected from 2 × 10⁶ MC38 cells grown overnight in a 10-cm tissue culture dish and centrifuged at 3900 rpm for 10 min to pellet any cellular debris. Subsequent 15-min recordings were carried out at 8 h and 20 h in vitro (4 h and 16 h of exposure to MC38 media).

Lumbar DRG from control and MC38 tumor-bearing mice were collected after perfusion with PBS, and RNA was extracted using the RNeasy MinElute Cleanup Kit (Qiagen, Hilden, Germany), based on the manufacturer’s protocol, as described previously [44]. The prepared samples were then shipped on dry ice and the bulk RNA sequencing (mRNA library preparation) was performed by Novogene (California, USA; California Clinical Laboratory License No: 05D2146243), after ensuring the samples met or exceeded the proper quality control criteria (Agilent 2100 RNA integrity number for all samples > 7). Raw reads in FASTQ format were analyzed using the FastQC tool for quality control [45]. The STAR package was used for the reference genome mapping with the mouse genome (mm10) [46]. The gene counts were estimated from uniquely mapped reads using the feature Counts program from the Subread tool [47]. Further, the gene counts were normalized and differentially expressed genes (adjusted p-value < 0.05) between the control and tumor-bearing mice were calculated using the DEseq2 program [48]. The pathway enrichment analysis was performed using the IPA tool with default parameters.

Sera were isolated by centrifugation of blood samples collected by terminal cardiac puncture on control and tumor-bearing mice. Serum samples were analyzed semi-quantitatively using a mouse XL Cytokine Array Kit (R&D Systems, USA) to assess circulating tumor-associated inflammatory mediator levels 1 and 3 weeks after tumor injection. To investigate factors released directly by tumor cells, conditioned media samples were generated by collecting growth media after 24 h from 10 cm dishes at 80% confluence. Samples were incubated with spotted nitrocellulose membranes according to the manufacturer’s instructions [49]. Protein concentration of samples was determined by a BCA assay prior to loading samples, to enable dilution of samples to normalize total protein content (~ 200 µg total protein). Membrane HRP luminescence intensities were visualized using an Amersham ImageQuant 800 biomolecular imager (Cytiva, USA), with an exposure time of 5 min and images were saved as high-resolution TIFF files. The registered intensity of each of the dots on the membranes was individually determined in duplicates by Image Studio software (version 5.2), and the fold-increase was calculated compared to the average value of control samples. Negative control-, and reference spots were also analyzed in each experiment to validate the analyses. Three biological replicates were measured per condition.

Following colorectal injection of MC38 cells, C57BL/6 mice were monitored weekly for tumor growth using bioluminescent imaging (BLI; Fig. 1a). MC38-injected mice developed continuously increasing luciferase signal at the site of injection, indicating increasing tumor burden over time (Fig. 1b, c). Altogether, 90% of MC38-injected mice developed tumors by the 1st week after tumor inoculation (n = 108/120), as indicated by a robust increase in the bioluminescence signal. In order to identify tumor-associated changes in pain sensitivity and/or general well-being, MC38-injected and vehicle-injected mice were assessed for their nestlet shredding behavior and abdominal and hindpaw mechanical sensitivity, along with hindpaw cold sensitivity and nestlet shredding behavior. We found no alterations in nestlet shredding behavior at 3 weeks post-MC38 injection (n = 10/group), suggesting there is no debilitating pain sensitivity or disruption to spontaneous behavior at this timepoint (Fig. 2a). This observation was also underlined by no significant differences in the increase in animal weights compared to controls throughout the testing period (Fig. 2b; n = 8–12/group). MC38 tumor burden also did not elevate abdominal mechanical sensitivity in the first 3 weeks. However, 4 weeks after injection, MC38 tumor-bearing mice showed a strong trend toward increased sensitivity compared to vehicle-injected controls (Fig. 2c; n = 7–11/group). Collectively, these data suggest MC38 tumor burden does not cause overt debilitation or augment visceral pain sensitivity during the 4-week experimental period. We also did not detect any statistically significant changes in hindpaw von Frey withdrawal thresholds (Fig. 2d; n = 11–17/group), latency to detect/remove adhesive tape (Fig. 2e; n = 8–12/group), or withdrawal latencies in response to cooling (Fig. 2f; n = 5–8/group), suggesting that no obvious shifts in tactile or cold sensitivity are induced by MC38 tumor growth.

Having seen no clear evidence of debilitation or sensorimotor dysfunction in MC38-injected mice, we next assessed MC38-induced changes to IENF density in the plantar skin, a prominent symptoms of peripheral neuronal damage [37, 39]. Plantar punch biopsies of hindpaws from tumor-bearing and control animals (n = 7–16 samples/group from ≥ 3 biological replicates) were analyzed by immunohistochemistry weekly after MC38 cell injection. We found no change in hindpaw IENF density (nerve endings/100 µm of epidermis) in the first 2 weeks post-injection (Fig. 3a). However, at 3 and 4 weeks, there was a significant decrease in IENF density in tumor-bearing animals compared to vehicle-injected controls (Fig. 3b). To investigate whether this phenomenon was systemic, or restricted only to sensory innervation that overlaps with the thoracolumbar (T10–L1) and lumbosacral (L6–S1) innervation of the colon/rectum [50], we also analyzed forepaw skin, which is innervated by levels C6–T2, and as such does not innervate the colon/rectum [51]. We also found a significant decrease in IENF density in forepaw skin at the third week of MC38 tumor development (Fig. 3c, d; n = 11/group from ≥ 2 biological replicates), indicating that reduced IENF density is not confined to those ganglia that also provide sensory innervation to the colon. This significant alteration suggested that peripheral nerve endings experience injury/damage in MC38-injected mice, despite finding no significant changes in mouse well-being or pain-related behaviors. Since prior studies of nerve injury have reported increased macrophage density in the skin [21], we determined macrophage density in skin biopsies three weeks after MC38 injection, using the macrophage marker CD68. We found no alterations in CD68 density in the hind paw skin of tumor-bearing mice (Fig. 3e, f; n = 17–20/group from ≥ 3 biological replicates), indicating that the IENF loss associated with MC38 injection does not elevate skin macrophage density.

Since MC38 tumor growth was associated with systemic reductions in IENF density, we investigated well-established indicators of neuronal damage at the level of the DRG: elevated macrophage density and increased expression of ATF-3 [52‐54]. We did not observe any significant changes in lumbar DRG CD68 density or ATF-3 expression in samples from tumor-bearing vs control mice (Fig. 4a–d; n = 7–12/group from ≥ 3 biological replicates). In contrast, repeat dosing with oxaliplatin (cumulative dose: 12 mg/kg i.p.; n = 3) as a positive control did significantly elevate ATF-3 expression (Fig. 4c, d). This suggests that MC38 tumor-induced IENF loss is not associated with conventional markers of neuronal damage in the DRG.

A well-described mechanism of neuropathic pain is mitochondrial dysfunction in sensory neurons [55]. Therefore, we tested whether MC38 tumor growth was associated with mitochondrial dysfunction in mouse DRGs. Despite not detecting elevated ATF-3 expression or CD68⁺ macrophage density, MC38 tumor development led to significant mitochondrial dysfunction in the DRGs of tumor-bearing animals, by the third week after MC38 injection (Fig. 5a–c; n = 4 biological replicates/group in duplicates). This is indicated by significant changes in the oxygen consumption rate (OCR; Fig. 5a); the extracellular acidification rate (ECAR) approached, but did not reach statistical significance (Fig. 5b). Further analysis revealed reduced basal and maximal respiratory capacity and ATP production in DRG cultures from MC38 mice (Fig. 5c). Importantly, these effects were observed on DRGs pooled from every level between C4 and L5, again suggesting this phenomenon is unlikely to be confined to those DRG neurons innervating the colon.

Since mitochondrial function is a major contributor to calcium dynamics in DRG neurons [56], we next performed live-cell calcium imaging of DRGs from control and MC38 tumor-bearing mice (Fig. 5d, e; n = 78–138 cells/group from ≥ 3 biological replicates). DRG neurons from MC38 tumor-bearing mice showed a small but significant decrease in baseline 340:380 nm Fura-2 ratio (Fig. 5d), indicating decreased resting [Ca²⁺_i] levels. Lumbar and non-lumbar DRG neurons were cultured and tested separately (data not shown), but, since no differences were observed (similar [Ca²⁺_i] decrease at both the lumbar and upper levels, see Additional file 1: Fig. S1), the data were pooled. Despite this decrease in baseline [Ca²⁺_i], the response to a depolarizing stimulus (25 mM KCl) [43] did not show any difference in response amplitude (Fig. 5e). Finally, we wanted to establish whether exposure to MC38-derived factors could be altering DRG neuron activity. Naïve C57BL/6 DRG neurons were dissociated and cultured on multi-electrode array plates (n = 8–10 cell culture wells/group from ≥ biological replicates). Four hours after plating, 50% of growth media was exchanged for fresh MC38 media (‘control media’) or media in which MC38 cells had been grown for 48 h (‘MC38-conditioned media’). At 20 h in vitro, DRG neurons exposed to MC38-conditioned media displayed a significantly lower spontaneous firing rate (Fig. 5f). Collectively, this suggests that MC38-derived factors can influence DRG mitochondrial function and calcium homeostasis, which may underlie the reduced activity of cultured DRG neurons.

In order to assess the secreted factors that could be driving MC38-induced DRG neuron dysfunction, we analyzed MC38 secretory output using Proteome Profiler assays (Mouse XL Cytokine Array Kit; R&D Systems [49]). Culture media in which MC38 cells were grown (2 × 10⁶ cells, 48 h) was compared with uncultured media (n = 3 different MC38 cell cultures in duplicates). Factors showing ≥ 1.2-fold increase compared to control are depicted in Fig. 6a. An additional table shows all measured cytokines (see Additional file 2: Tables S2 and S3). Abundant secretion of chemokines (CXCL1, CCL2, CX3CL1, CCL5, CXCL10, CCL11) and growth factors (GDF-15, VEGF, M-CSF, FGF-21, amphiregulin) was detected in MC38-conditioned media. We also collected serum from MC38 tumor-bearing mice and vehicle-injected controls, 1 week and 3 weeks after MC38 injection (n = 3 biological replicates/group in duplicates). At week 1, elevated levels of cytokines (IL-1α, IL-6), chemokines (CXCL2, CCL2, CCL3/4, CCL20) and growth factors (EGF, FGF acidic, PDGF-BB) were detected (Fig. 6b). 3 weeks after MC38 injection, the following mediators were shown to be increased by more than a 1.2-fold (in decreasing order): GM-CSF, G-CSF, GDF-15, HGF, Osteoprotegerin, IL-23, Proliferin, Pentraxin-2, CCL6, CXCL10, BAFF, Angiopoietin-2, Osteopontin, Lipocalin-2, CXCL1, IL-28A, Gas6, Proprotein Convertase 9, Serpin E1, CXCL16, P-Selectin (Fig. 6c). IFN-γ, IGFBP-1, IL-23, Endoglin and Serpin E1 were elevated in tumor-bearing serum at both timepoints (Fig. 6d). Altogether, seven factors were detected both in MC38-conditioned media and MC38 tumor-bearing serum: CCL2, GDF-15, Proliferin, CXCL10, Osteopontin, PCSK9 and Serpin E1. In agreement with our data, the chemokines shown to be increased here were also found to be elevated in different previous studies investigating cancer or neuropathic pain. Given their extensive role in the development of different neuropathies (including CIPN [57]), it is plausible that some of these factors play a significant part in the neuropathy observed here as well.

In order to better understand the signaling events and gene expression changes that MC38 tumors may be inducing to elicit functional changes in neuronal physiology, we subjected lumbar DRGs from MC38 cancer cell injected mice and control (n = 3/group) to bulk RNA-sequencing (Fig. 7). 134 differentially expressed genes were identified, 39 of which were significantly upregulated (p < 0.05; Fig. 7a and Additional file 1: Table S1). Functional and disease enrichment analysis on upregulated genes from the tumor-bearing mice showed cancer as a top enrichment (number of genes: 23, see Additional file 4: Fig. S2a, b). Consistent with the elevated levels of multiple cytokines and chemokines in tumor-bearing serum, the pathway analysis showed the most robust increase in the ‘Role of hypercytokinemia/hyperchemokinemia in the Pathogenesis of Influenza’ pathway, along with numerous other inflammation-related pathways (Fig. 7b). Several pathways related to thrombosis/coagulation (hemostasis) are also consistent with the hypercoagulable state attributed to colon cancer [58]. Collectively, these data indicate that the tumor induces transcriptional changes due to systemic inflammation. Of particular note is the significant upregulation of CXCL10 expression (Log₂ fold increase: 0.843, p = 0.012). This increased expression in bulk RNA is likely derived from DRG neuron somata, as they are the chief CXCL10-immunoreactive cell type in immunohistochemistry (Fig. 7c). It is notable that the proteome profile data also identified elevated expression of this chemokine in MC38 tumor-bearing mouse serum.

We next set out to determine if the disruption of DRG neuron function seen in MC38 tumor-bearing mice would generalize to other mouse strains and cancer cell lines. Similar to the MC38 model, we carried out orthotopic injection of luciferase-expressing CT26 cells into Balb/c mice, followed in vivo by bioluminescence measurements (Fig. 8a). All mice injected with CT26 cells developed colon tumors (n = 16). When investigating the weekly tumor growth in experimental animals, we found a significant increase in bioluminescence signal at the 1st week after cell injection, which continuously further elevated during the 3-week time period (Fig. 8b; n = 6/group), indicating the development of colon cancer. Again, similarly to the MC38 model, we found no significant differences in animal weight (Fig. 8c; n = 7–10/group). Mechanical sensitivity of mice was assessed 1, 2 and 3 weeks post-tumor injection. Similar to the MC38 model, we found no significant change in hindpaw mechanical sensitivity throughout the 3-week time period (Fig. 8d; n = 5–6/group). Crucially, 3 weeks after CT26 tumor cell injection, a significant decrease in the IENF density of samples from the tumor group was observed (Fig. 8e, f; n = 9–16 samples/group from at least 3 biological replicates), without any change in macrophage density (Fig. 8 g, h; n = 6–7/group from at least 2 biological replicates), mirroring the sensory neuron dysfunction seen in C57BL/6 mice with MC38 cells. This IENF loss was again accompanied by mitochondrial dysfunction in the DRGs of tumor-bearing animals, as indicated by significant changes in both OCR and ECAR (Fig. 9a–c; n = 3 biological replicates/group in duplicates).

To investigate the tumor-induced inflammatory alterations in BALB/c mice, a cytokine array experiment was performed on mouse plasma samples 3 weeks after CT26 tumor cell injection. An additional table shows all measured cytokines (see Additional file 2: Tables S2 and S3). Similar to the MC38 model, we found robust alterations in secreted factor output in CT26-conditioned media (Fig. 10a; n = 3 cell culture replicates). The following mediators showed at least a 1.2-fold signal increase compared to un-conditioned media samples: cytokines (IL-1α/β, IL-2, IL-11, IL-12 p40, IL-27 p28, LIF,) chemokines (CXCL1, CXCL2, CCL2, CX3CL1, CCL5, CCL17, CCL19, CXCL10, LIX) and growth factors (Osteoprotegerin, IGFBP-1/3/6, M-CSF, VEGF) predominate (Fig. 10a; n = 3 biological replicates/group in duplicates). In serum of CT26 tumor-bearing mice, there was extensive overlap with the factors detected in CT26-conditioned media: following mediators identified in conditioned media showed at least a 1.2-fold increase in CT26 tumor-bearing mice: IL-1α, IL-2, IL-13, CCL2/5/19, LIX, CX3CL1, CXCL1/2/10 and CCL19 (Fig. 10b).

Across serum, conditioned media and both tumor types, four factors are consistently detected: CCL2, CXCL1, CXCL10 and Serpin E1. Collectively, these data indicate that similar pro-inflammatory pathways across two colorectal cancer models are associated with subclinical mitochondrial dysfunction and IENF loss in sensory neurons.