With improved treatment of colorectal cancer comes the potential for long-term side effects in survivors, a population that is set to continue growing for the foreseeable future . Although clinical data highlight peripheral neuropathy as a major side-effect in cancer survivors  and the fact that the pro-inflammatory milieu generated by tumors can affect neuronal function [4, 6], the extent to which colorectal cancer is directly responsible for neuropathy symptoms remains under-explored. Therefore, we set out to determine whether inflammation related to colorectal cancer was capable of inducing neuronal dysfunction/damage. Our study demonstrates colon tumor-initiated neuronal damage in a preclinical setting, by the application of orthotopic colon cancer models in two different mouse strains. Local (trans-anal) inoculation of both MC38 and CT26 cancer cells led to continuous tumor development without overt changes in measures of overall activity, shifts in pain sensitivity or cachexia (Fig. 2, Fig. 8c-d). Only four weeks post-tumor injection did tumor-bearing mice begin to develop a trend toward abdominal hypersensitivity and reduced/stalled weight gain (Figs. 2b, 8c). The most typical symptoms of neuropathic pain, cold and mechanical hypersensitivity were also investigated throughout the 4 week time period . Based on our findings, tumor development did not lead to significant cold or mechanical hypersensitivity in this time period (Figs. 2d-f, Fig. 9d).
Despite the lack of behavioral symptoms consistent with neuropathy, we assessed IENF density in paw skin by immunohistochemistry, as a decrease in IENF density is among the most typical indicators of peripheral neuropathy [62, 63]. Damage of peripheral neurons was detected as decreased IENF density in hindpaw skin of tumor-bearing mice three weeks after cancer cell injection, which persisted into the fourth week. Since the dermatomes of the forepaw are innervated by ganglia outside of the thoracolumbar innervation of the lower GI tract , a similar reduction in forepaw skin IENF density suggested to us that tumor growth systemically reduces IENF density, rather than injuring DRG neurons that innervate both the colorectum and epidermis. This association between tumor growth and IENF loss is supported by the observation that colon cancer patients commonly show symptoms of peripheral neuronal damage when tested (IENF density loss; minor sensory deficits in the extremities), despite not presenting with overt symptoms of pain or sensory loss [60, 63]. That said, it is important to underline that there is not always a direct association of IENF loss with neuropathic symptoms, and further work is needed to understand the degree to which tumor-induced neuropathy modifies lifetime risk of neuropathic pain [63, 64].
We then determined the extent to which indicators of nerve injury were present at the level of the DRG, namely macrophage infiltration  and ATF3 expression . Consistent with the lack of alteration in pain behaviors, we did not detect increased macrophage density or ATF-3 expression in the DRG, both of which are reported features of overt neuropathic pain hypersensitivity seen following induction of CIPN or traumatic nerve injury. Finally, one of the main factors underlying neuropathy and neuropathic pain is mitochondrial dysfunction [17, 55, 66]. Therefore, we analyzed the mitochondrial function of lumbar DRGs from tumor-bearing mice, 3 weeks after tumor induction (first timepoint of IENF loss). The Seahorse assay revealed substantial mitochondrial dysfunction indicated by both OCR and ECAR (Fig. 5a-c, Fig. 9). Deficits of this magnitude (approximately 30% reduction in OCR) are less severe than those seen in models of cisplatin-induced CIPN (typically 50–60% reduction in OCR), changes that are associated with significant pain hypersensitivity and IENF loss . This suggests that the relationship between mitochondrial dysfunction and pain hypersensitivity may be non-linear, that the onset of mitochondrial dysfunction occurs prior to the development of pain hypersensitivity, or that mitochondrial dysfunction induced by tumor growth is uncoupled from pain hypersensitivity. At this point, we cannot rule out mitochondrial dysfunction/reduced ATP production in tissues outside the DRG. Indeed, such effects have been reported in the skeletal muscle of patients with cancer [67, 68], though the energy demands and post-mitotic nature of neurons makes them particularly vulnerable [69, 70]. As such, our data are consistent with observations that energy availability and metabolic derangement are common features of cancer and cancer-related fatigue [71–73].
To further address the nature of the tumor-induced neuronal dysfunction, live-cell in vivo calcium imaging was performed on DRG neurons. Our analysis revealed reduced intracellular Ca2+ levels ([Ca2 + i]) in DRG neurons from tumor-bearing mice. Though this can appear counterintuitive when compared with chronic pain states which tend to show hyperexcitability , this finding is consistent with prior reports showing low [Ca2 + i] after neuronal damage [75–79]. For example, Andreas Fuchs and co-workers showed that spinal nerve ligation decreased resting [Ca2 + i] levels in rat DRGs . Reduced neuronal [Ca2 + i] is known to precipitate cell loss, a feature in different forms of neuropathy [77, 80]. However, it is unclear as yet if reduced [Ca2 + i] is directly tied to neuropathy and IENF loss, or the extent to which this is due to a systemic hypocalcemic state, as has been reported for hematological and colorectal cancers .
Mitochondria play a prominent role in in neuronal Ca2+ signaling , and abnormal mitochondrial function can lead to axonal degeneration as well as disturbances in Ca2+ homeostasis, which can manifest in low [Ca2 + i] levels and neuronal damage . The specific contribution of plasma membrane and organelle Ca2+ pumps, such as the sarco-endoplasmic reticulum Ca2+-ATPase (SERCA), should be investigated, since they may contribute to reduced [Ca2 + i] levels, mitochondrial dysfunction and/or ER stress ). In summary, colon tumor growth leads to asymptomatic, but prominent neuronal dysfunction, which may have clinical implications for any future neurological insult incurred as a side-effect of cancer treatment.
In search of circulating factors that could underlie the systemic effects on sensory neurons, cytokine arrays of plasma samples from tumor bearing and control mice revealed robust systemic inflammatory changes (Figs. 6, 10). Interferon-γ was notable for its detection both 1 week and 3 weeks after tumor injection. Notably, several chemokines (CCL2, CXCL1, CXCL2, CXCL10) were shown to be increased in plasma samples as well as to be released by both investigated cancer cell lines (MC38, CT26), many of which are classed as ‘interferon-stimulated genes’ , an assertion borne out by our DRG RNA sequencing data (Fig. 7). Prior studies have found increases in chemokine expression in tissues from colorectal cancer patients that are associated with disease progression [85, 86]. The pathway enrichment analysis showed the most significant increase in the ‘Role of Hypercytokinemia/hyperchemokinemia in the Pathogenesis of Influenza’ pathway, further underlining the likely significant role of chemokines in the observed alterations.
In addition to pro-inflammatory signaling, the other main pathways highlighted by the RNAseq data relate to hemostasis/hypercoagulability (e.g. ‘role of tissue factor in cancer,’ ‘intrinsic/extrinsic prothrombin activation pathway,’ ‘coagulation system’). Colorectal cancer is often associated with hypercoagulability in patients, to the extent it substantially increases the risk of thrombosis . It is plausible that impaired endoneurial blood flow results from such clotting events, contributing to mitochondrial dysfunction and neuropathy, though this needs to be addressed experimentally.
There is a growing body of literature showing the multifaceted roles of chemokines in neuropathic pain development. Literature data indicates that the expression of chemokines and their receptors, such as CCL2/CCR2, CXCL1/CXCR2, (among others) are altered in CIPN . CXCL2 has been also shown promote neuropathic pain in a recent study on a rat model of trigeminal neuropathic pain . In addition, a large number of studies suggests the key role of CXCL10/CXCR3 signaling in the pathophysiology of different neuropathies [88–90]. Our results from the bulk RNA sequencing on DRGs from MC38 tumor bearing mice also revealed a significant increase in the expression of CXCL10 (Fig. 7). The extent to which the inflammatory signatures at the level of the DRG are a cause or consequence of DRG mitochondrial dysfunction remains to be established. NLRP3 inflammasome activation and subsequent production of inflammatory mediators could be downstream of ROS production from dysfunctional mitochondria , or inflammatory mediator signals originating from the circulation could be responsible for inducing DRG mitochondrial dysfunction.
It is well-established that pre-existing neuropathy of various etiologies is a risk factor for subsequent development of neuropathic pain . In this context, tumor-induced neuropathy may have major implications for CIPN. Oxaliplatin treatment is still a cornerstone of colon cancer therapy and is well described to cause CIPN (in 90% of patients). However, a subset of patients (about 40%) develop chronic, intractable CIPN symptoms (such as mechanical allodynia or cold hypersensitivity) . The risk factors leading to this detrimental chronic condition are still not understood. However, based on our observations, tumor induced inflammation may represent a crucial risk factor in chronic CIPN development, which should be further investigated in the future.
In order to further confirm our observations on peripheral neuronal dysfunction, and to increase the generalizability of our findings, we employed the same model on a different genetic background, engrafting CT26 cells into BALB/c mice. Without significant alterations in behavior, the CT26 model showed a similar decrease in IENF density, as well as indicators of mitochondrial dysfunction in DRG neurons (at the same time point as in the MC38 model – 3 weeks after engraftment, Fig. 8). The proteome profiler arrays of serum 3 weeks after engraftment depicted a similarly robust systemic inflammation, with substantial overlap with the MC38 model (i.e. chemokines), as described above (Fig. 10).
Taken together, our study depicts colon cancer-induced peripheral neurological dysfunction in a preclinical setting (by the application of orthotopic colon cancer models), as indicated by: IENF density decrease in the skin of fore and hind paws; mitochondrial dysfunction of lumbar DRG neurons and reduced resting [Ca2 + i] in DRG neurons. The observed neurological dysfunction was not accompanied by behavioral changes, or any significant changes in the investigated parameters of animal well-being. Additionally, tumor development led to robust systemic inflammatory changes. The chemokines, CCL2, CXCL1, CXCL2, CXCL10 were shown to be increased across both tumor models. CXCL10 expression was also increased in the DRGs of MC38-inoculated mice.
It remains to be seen if the pro-inflammatory and hypercoagulable state associated with other cancer types elicits similar neuronal dysfunction, but subclinical peripheral neuropathy is also known to occur in patients with lung cancer or multiple myeloma [15, 92], suggesting this phenomenon may extend beyond colorectal cancer. Our observations indicate that the tumor induced systemic changes include peripheral neuronal dysfunction, without any pharmacological or other intervention. The increased mediators might be crucial in the induction of chronic, intractable CIPN, developing in a subset of colon cancer patients. This and the exact mechanism by which colon tumor induces systemic, peripheral neuronal dysfunction in mice warrants further investigation in future studies.