Introduction

Chemotherapy for cancer can cause significant gut toxicity leading to a breakdown of the selectivity of the mucosal barrier. This results in bacteria passing into the systemic circulation and causing serious symptomology such as oral mucositis, increased infection rates, and localized gut toxicity (diarrhea) [1, 2]. Consequently, chemotherapy-induced gut toxicity (CIGT) is a significant burden on patients’ quality of life and causes substantial additional health costs [3]. Research has focused on the mechanisms by which this toxicity occurs and has highlighted key roles for apoptosis [4], the microbiome and matrix metalloproteinases [2]. More recently an appreciation has been gained regarding the profound impact that the innate immune system has in creating CIGT, specifically via Toll-like receptors (TLRs) [5]. The gut is the largest immunological organ in the body, capable of generating immense immunological responses through activation of resident mucosal associated lymphoid tissue [6]. In the late 1990s, the pattern recognition receptor Toll was discovered, [7, 8] which led to the further characterization of the TLRs. Normal gut bacteria are tightly regulated via TLR signaling [9, 10]. TLRs are expressed within the epithelial cells of the small and large intestine [11], with research implicating TLRs as playing key roles in the maintenance of gut epithelial homeostasis [12]. Our previous research has demonstrated that upregulation of nuclear factor kappa B (NF-κB) [13] and pro-inflammatory cytokines, [14] and alterations in epithelial cell proliferation [15] occur in the gut following chemotherapy. This implicates the role of TLR pathways being pivotal in the pathogenesis of CIGT and may offer new therapeutic targets to investigate. Further, TLR4 recognizes exogenous factors from invading microbes, endogenous danger signals, and some chemical structures. Thus, TLR4 may be uniquely positioned to be able to detect both the tissue damage generated as a consequence of chemotherapy, and the invading pathogens resulting from gut toxicity.

Aside from the substantial gut toxicity following chemotherapy, pain induced by chemotherapy is another adverse effect commonly experienced in the clinical setting [16, 17]. It can be one of the most important dose limiting complications, particularly in patients receiving taxane-, vinca alkaloid-, and/or platin-based chemotherapy [16, 17]. It is characteristically reported as a tingling, burning pain often accompanied with significant sensory-motor impairments [18, 19], is refractory to treatment in a large number of patients, and can persist as a chronic condition significantly impacting on the patients’ quality of life [20].

The underlying mechanisms of the chemotherapy-induced pain are ill defined. It has been proposed that some cytotoxic agents may damage neurons through binding to axonal microtubules to subsequently alter axonal transport as shown with vincristine [21, 22]. However, chemotherapeutic agents are generally unable to cross the blood brain barrier and therefore should theoretically be unable to damage central neurons, strongly implying chemotherapy induces pain via other mechanisms. In addition to the direct damage to neurons as a result of chemotherapy exposure, the damage itself generates a “damaged tissue” signal, which would in turn result in localized glial activation and the subsequent secretion of pro-inflammatory cytokines to further exacerbate neuronal responses to potentiate pain [23].

The enteric nervous system is comprised of neurons and glial cells [24] and it is located within the gastrointestinal tract. Traditionally it has been thought enteric glia, which similarly to astrocytes in the central nervous system express increasing amounts of glial fibrillary acidic protein (GFAP) when activated, provide critical support and nutrition to the enteric neurons [24]. However, this has recently been expanded with research suggesting enteric glia regulate the intestinal epithelial barrier [25–27].

Our research [14, 28] and that of others [29] has strongly implicated the transcription factor NF-κB and pro-inflammatory cytokines including tumor necrosis factor (TNF-α), interleukin (IL)-1 beta (IL-1β), and IL-6, as central regulators of CIGT. Although it has been reported that taxol, which causes peripheral pain, also induced changes in the synthesis of pro-inflammatory cytokines including TNF-α and IL-1β [17], it remains unknown whether these same factors are key players in pain induced by other cytotoxics. Despite this common mechanism of these chemotherapy-induced toxic effects, to date, there has been no established link between the timing of pain and gut toxicity. Therefore, the aim of this study was to determine if there is a link between pain assessed by facial pain scoring [30], and gut toxicity assessed by diarrhea and weight loss. Our secondary aims were to determine if altered expression of TLRs, and specifically TLR4, is a key driver of these effects; and also to determine whether enteric nervous system GFAP expression is altered in response to chemotherapy.

Materials and methods

Animals and ethics

Female Dark Agouti (DA) rats, weighing between 150 and 170 g were used for this study. Rats were housed in Perspex cages at a temperature of 22 ± 1 °C and subject to a 14 h light/10 h dark cycle. Animals had ad libitum access to autoclaved chow and water. Experimental design was approved by the Animal Ethics Committees of the Institute of Medical and Veterinary Science (IMVS), and The University of Adelaide, and complied with the National Health and Medical Research Council (Australia) Code of Practice for Animal Care in Research and Teaching.

Experimental design

Forty rats were randomly assigned to receive either irinotecan (n = 5–8 per time point, total of 34) or vehicle (n = 5). All rats received breast cancer inoculum as described previously [15]. In brief, rats were implanted with 4 × 106 cells in 0.2 mL phosphate-buffered saline (PBS) s.c. into each flank and tumors were allowed to grow for 1 week prior to administration of chemotherapy. All rats received 0.01 mg/kg s.c. atropine (to reduce the cholinergic reaction) immediately prior to administration of either 175 mg/kg i.p. irinotecan (kindly supplied by Pfizer, administered in a sorbitol/lactic acid buffer: 45 mg/mL sorbitol/0.9 mg/mL lactic acid, pH 3.4), or vehicle (sorbitol/lactic acid buffer: 45 mg/mL sorbitol/0.9 mg/mL lactic acid, pH 3.4, previously shown to have no gut toxicity effects [31]) at time 0 h. Groups of rats were killed using 3 % isofluorane in 100 % O2 anesthesia and cervical dislocation at times 6, 24, 48, 72, 96, and 120 h post-irinotecan treatment. The entire gastrointestinal tract (from the pyloric sphincter to the rectum) was dissected out and separated into the small intestine (pyloric sphincter to ileocecal sphincter) and colon (ascending colon to rectum). The small intestines and colons were flushed with chilled, sterile saline (Baxter Healthcare), and 1 cm samples dissected from 50 % of the length of each, and fixed in 10 % neutral-buffered formalin, processed and embedded in paraffin for immunohistological analyses.

Gut toxicity assessment

Gut toxicity was assessed through weight loss and occurrence of diarrhea as previously published as being associated with histological damage in the small and large intestine [32]. These measures were recorded on a daily basis including baseline. Animals were weighed daily at the same time, and total weight loss/gain recorded. Further, diarrhea occurrence and severity was recorded four times daily according to previous grading [31]: 0, no diarrhea; 1, mild diarrhea (staining of anus); 2, moderate diarrhea (staining spreading over top of legs); and 3, severe diarrhea (staining over legs and abdomen, often with continual anal leakage). All gut toxicity assessments were conducted in a blinded fashion (RJG).

Chemoneuropathy: facial pain assessment

Chemoneuropathy was assessed four times daily using the established facial pain model that scored orbital tightening, nose bulge, cheek bulge, whisker change, and ear position [30]: 0, not present; 1, moderate; and 2, severe. All facial pain assessments were conducted in a blinded fashion (RJG).

Immunohistochemistry

Toll-like receptors

Sections of jejuna and colon were cut from paraffin blocks at 4 μm thickness and mounted onto silane-coated slides (Lomb Scientific Pty Ltd). The jejuna was chosen as a representative region of the small intestine as in our model, irinotecan causes equivalent damage along the entire length of the small intestine as well as the large intestine [31, 32]. Tissue samples were dewaxed in xylene (Merck Pty Ltd) and rehydrated through a graded series of ethanol (Chem Supply). Sections were immersed in 10 mM citrate buffer (pH 6.0, Sigma-Aldrich Pty Ltd) and antigen retrieval performed by heating sections in a microwave (LG) on high (800 W) until boiling and then on low (160 W) for 10 min. Sections were allowed to cool and endogenous peroxidise activity subsequently blocked with 3 % H2O2 in methanol (Chem Supply). Non-specific antibody binding was blocked with 20 % normal goat or horse serum (NGS or NHS) as appropriate (Sigma-Aldrich Pty Ltd) in PBS for 30 min at room temperature (RT). Avidin-Biotin Blocking Kit (Vector Laboratories) was used to block endogenous avidin-biotin activity. Sections were incubated overnight with primary rat polyclonal antibodies (diluted in 5 % NGS or NHS) directed at TLR2 at 1:50 (Novus Biologicals, #NB200-536), TLR4 at 1:50 (Abcam, #ab13556), TLR5 1:50 (Abcam, #ab62460) and TLR9 at 1:50 (Abcam, #ab12121) at 4 °C. Sections were then incubated with the appropriate secondary antibody (20 min at RT), ABC labeling reagent (30 min at RT) and developed with 3,3′-Diaminobenzidine. Sections were counterstained with Lillie-Mayer’s hematoxylin (HDS Scientific Supplies Pty Ltd), dehydrated and cleared in xylene (Merck Pty Ltd) before being mounted.

GFAP

Sections of jejuna and colon were cut from paraffin blocks at 4 μm thickness and mounted onto silane-coated slides (Lomb Scientific Pty Ltd). Tissue samples were dewaxed in xylene (Merck Pty Ltd) and rehydrated through a graded series of ethanol (Chem Supply). Sections were immersed in 10 mM citrate buffer (pH 6.0, Sigma-Aldrich Pty Ltd) and antigen retrieval performed by heating sections in a microwave (LG) on high (800 W) until boiling and then on low (160 W) for 10 min. Sections were allowed to cool and endogenous peroxidise activity subsequently blocked with 3 % H2O2 in methanol (Chem Supply). Non-specific antibody binding was blocked with 20 % normal horse serum (NHS) (Sigma-Aldrich Pty Ltd) in PBS for 30 min at room temperature (RT). Avidin-Biotin Blocking Kit (Vector Laboratories) was used to block endogenous avidin-biotin activity. Sections were incubated overnight with a primary goat polyclonal antibody (diluted in 10 % NHS) directed at GFAP at 1:400 (Santa Cruz Biotechnology, #SC-6171) at 4 °C. Sections were then incubated with an appropriate secondary antibody (20 min at RT), ABC labelling reagent (30 min at RT) and developed with 3,3′-Diaminobenzidine. Sections were counterstained with Lillie-Mayer’s hematoxylin (HDS Scientific Supplies Pty Ltd), dehydrated and cleared in xylene (Merck Pty Ltd) before being mounted.

Analysis

Semi-quantitative immunohistochemistry was performed, with staining intensity graded as follows [33]: 0, no staining; 1, weak staining; 2, moderate staining; 3, strong staining; and 4, intense staining. All assessments were done in a blinded fashion (RJG).

Statistics

All data for diarrhea, weight loss, facial pain, TLR2/4/5/9, and GFAP expression was compared to baseline. Comparisons of diarrhea severity scores (a single score in each 24 h period), percent weight loss, facial pain scores, TLR2/4/5/9, and GFAP expression over time was performed using the Kruskal–Wallis or one-way ANOVA tests as appropriate. Comparisons between facial pain scores in rats with and without diarrhea, and in rats with and without TLR expression were performed using the Mann–Whitney test. GraphPad Prism v 5.01 (GraphPad Software Inc, La Jolla, USA) was used for all statistical analysis. All data are presented as either median (range) or mean ± SEM unless otherwise stated and P < 0.05 was considered statistically significant.

Results

Irinotecan caused significant gut injury and pain

All rats receiving irinotecan developed gut injury evidenced by significant diarrhea and significant weight loss. Diarrhea occurred in a biphasic response, with symptoms appearing at 6 h firstly before resolving and then maximal symptoms observed at 72 h post-irinotecan (Fig. 1). Rats receiving vehicle control did not develop diarrhea at any time point.

Fig. 1
figure 1

Percentage of rats with grade mild, moderate, and severe diarrhea following irinotecan (175 mg/kg i.p.) administration

Full size image

Peak weight loss post-irinotecan compared to baseline occurred at 72 h (mean ± SD = 11.1 ± 6.6 %, P < 0.0001) before recovery at 120 h (mean ± SD = −0.25 ± 6.7 %) (Fig. 2). Rats receiving vehicle control continued to gain weight over the course of the experiment (Fig. 2).

Fig. 2
figure 2

Percentage change in weight from baseline to 120 h in rats following vehicle control or irinotecan (175 mg/kg i.p.). **P < 0.01 versus 120 h, ***P < 0.0001 versus 24 h and 120 h

Full size image

Facial pain scores (Fig. 3), used as an indicator of pain, varied over time with peak pain occurring 72 h post-irinotecan, median (range) = 5 (0–5) versus 0 (0–0) at baseline, P < 0.0001. Recovery occurred by 120 h after irinotecan (0 (0–0)). Rats receiving vehicle control only exhibited facial pain immediately after anesthetic and i.p. injections; these resolved after 6 h and did not reoccur over the course of the experiment. Importantly, there was also an association between the occurrence of facial pain and gut injury, such that rats with diarrhea had significantly higher facial pain scores compared to those without diarrhea: median (range) of 2.0 (0 to 5) versus 0 (0 to 5) (P = 0.01).

Fig. 3
figure 3

Percentage of rats with facial pain scores of 0, 1, 2, and ≥ 3 between 6 and 120 h following irinotecan (175 mg/kg i.p.) administration

Full size image

Irinotecan decreased TLR4 and TLR5 expression in the jejuna, with TLR4 expression associated with pain

There was no significant change in epithelial cell expression of TLR2 or TLR9 following irinotecan in either the jejuna or colon regions (P = 0.06–0.76). In contrast, both TLR4 and TLR5 expression was significantly decreased following irinotecan in the jejuna. More specifically, although TLR4 and TLR5 expression was unchanged across all time points in the epithelial cells of the villi of the jejuna, expression significantly decreased in jejunal crypts at 96 and 120 h after irinotecan (P < 0.001, Fig. 4a TLR4, and P < 0.008, Fig. 4b TLR5). In contrast, TLR4 and 5 expression in the apical and basal colon remained unchanged across all time points (P = 0.16–0.57). Staining was cytoplasmic throughout the study. Comparison of toxicity measures and TLR4 and TLR5 expression in the jejuna revealed that those rats with no TLR4 expression experienced less pain than those that expressed TLR4: median (range) = 0 (0–5) versus 2 (0–5), P = 0.012. No other significant associations were observed.

Fig. 4
figure 4

Immunohistochemical staining intensities for TLR4 (a) and TLR5 (b) expression in jejuna of rats 6 to 120 h following irinotecan (175 mg/kg i.p.) administration compared to vehicle control. *P < 0.05 versus control. Data are median (range). Representative images of TLR4 (c) and TLR5 (d) immunohistochemical staining (×200 original magnification, scale bar length = 50 μm)

Full size image

Irinotecan increased GFAP expression in jejuna and colon

GFAP was analyzed as a marker of altered enteric nervous system function. In the jejunum, GFAP expression was not detected in any control animal (Fig. 5). Further, there was no GFAP expression seen in the jejuna of animals killed 6 and 24 h after irinotecan administration. Mild GFAP expression was seen in animals killed 48 h and this continued to increase at 72 h peaking and remaining high at 96 and 120 h after irinotecan (P = 0.017). In contrast, in the colon, GFAP expression was detected in all animals, although expression was significantly higher 24 and 96 h after irinotecan expression (P = 0.04). GFAP staining was localized to the lamina propria and the submucosa.

Fig. 5
figure 5

Immunohistochemical staining intensities for GFAP expression in jejuna of rats 6 to 120 h following irinotecan (175 mg/kg i.p.) administration compared to vehicle control. *P < 0.05 versus control. Data are median (range). Representative images of immunohistochemical staining in vehicle control and 120 h irinotecan treated rats are also given (×200 original magnification, scale bar length = 50 μm)

Full size image

Discussion

Both CIGT and pain are frequent, debilitating and dose-limiting adverse effects of anti-cancer cytotoxic therapies [17, 20] affecting a large number of cancer patients. The present study has demonstrated that both toxicities have a similar time course of presentation, with peak gut toxicity and pain occurring at 72 h following chemotherapy. This implies that the underlying pathobiology of the two conditions may be linked in this preclinical model. In further support of this, we have also demonstrated that GFAP expression within the jejuna increases following chemotherapy, peaking at the later time points and remaining high during the healing phase of chemotherapy-induced gut toxicity. This increase in GFAP expression, indicative of glial activation, is in accordance with previous research that suggests enteric glia enhance epithelial restitution [26].

The currently accepted pathogenesis of CIGT is an intertwined five-phase model, characterized by dynamic biochemical interactions between chemotherapeutic agent(s) and the cellular constituents of the mucosa [34]. More specifically, NF-κB is a key transcription factor that plays a critical role in the pathogenesis. Our previous studies in this DA rat model of chemotherapy-induced gut toxicity have clearly shown NF-κB to be up-regulated by over twofold in the gut mucosa [13]. Further, as NF-κB modulates gene expression of pro-inflammatory cytokines, including TNF, IL-6, and IL-1β, it is not surprising that we have also observed elevated levels of TNF, IL-1β, and IL-6 in the serum and tissue with our model [35].

Previous research has shown the enteric nervous system plays a key role in the maintenance of intestinal barrier homeostasis [36]. Enteric glial cells of the gut are specifically activated by pro-inflammatory cytokines, specifically IL-1β, and IL-6 [24], which are known to be elevated early after the administration of chemotherapy, and are key players in development of chemotherapy-induced gut toxicity [13]. Enteric glial cells are also thought to produce IL-1β however the underlying mechanisms remain unknown [24]. Thus, we suggest pro-inflammatory cytokines, elevated early after chemotherapy and involved in the development of CIGT, also activate enteric glia, represented here by an increase in GFAP expression, which then play a role in restoring the damaged epithelium.

Chemotherapy-induced pain is one of the key dose limiting factors for many chemotherapy patients [17], leading to a reduced chance of cure. To date, there is no clear mechanism through which cytotoxic agents induce pain [17], and importantly hyperalgesia. However, chemotherapeutic agents have been demonstrated to cause the upregulation of IL-1β [37], a crucial mediator of neuropathic pain [38–40]. Thus, it is tangible to suggest that chemotherapy-induced pain may have the same key immune mediators underlying its pathogenesis as CIGT. However, to date, the link joining these two toxicities has been missing.

The observations from this current study indicate that this link may be involvement of TLRs, as evidenced by relationships between gut toxicity and pain, and TLR4 expression and pain. These data are similar to that reported in previous studies. For example, rats receiving cisplatin had a significant weight loss accompanied by changes in pain from mechanical and thermal tests [16]. TLR3, TLR4, and myeloid differentiation primary response 88 (MyD88, signaling mediator for TLR4) knockout mice had lower levels of cisplatin-induced pain from mechanical tests compared to wild-type mice, and there was a total absence of pain in double knockout mice (TIR-domain-containing adapter-inducing interferon-β (Trif, signaling mediator for TLR4/MyD88)) [41]. While taxol treated macrophages from TLR4 and MyD88 knockout mice secreted minor amounts of pro-inflammatory mediators compared to macrophages from wild-type mice [42]. TLR4 signaling is known to activate NF-κB, and therefore could be the key initiating factor in the underlying pathogeneses of both chemotherapy-induced gut toxicity and pain. Although in the present study TLR4 expression did not change during peak gut damage, it was completely absent during recovery phase in agreement with the inverse relationship between TLR/NF-κB and the Wnt/β-catenin signal transduction pathway. It is known that Wnt/β-catenin is “switched off” during times of pro-inflammation [43], but activated during periods of intestinal growth [44], such as during intestinal recovery following chemotherapy which we have shown to be a time of rapid cell division [31], hence explaining TLR4 absence during the healing phase. Further, as TLR4/5 expression was absent from the jejuna during recovery from gut toxicity, pharmacological inhibition of TLRs may be of clinical use to increase healing in the jejuna following chemotherapy.

In conclusion, this is the first study to show a close link between chemotherapy-induced pain and diarrhea, indicating a possible common underlying trigger and mechanism; here we provide evidence for altered TLR signaling pathways and increases in enteric nervous system GFAP expression. Furthermore, we have shown that TLR4 and TLR5 are specifically downregulated during gastrointestinal tissue recovery and pain resolution, whereas GFAP expression is significantly upregulated during gastrointestinal recovery. This suggests that feedback on these receptors occurs to facilitate healing and tissue restoration. Further research is now required to investigate the impact of TLR modulation in the development and resolution of chemotherapy-induced toxicity.