Tumour brain: Pretreatment cognitive and affective disorders caused by peripheral cancers

People that develop extracranial cancers often display co‐morbid neurological disorders, such as anxiety, depression and cognitive impairment, even before commencement of chemotherapy. This suggests bidirectional crosstalk between non‐CNS tumours and the brain, which can regulate peripheral tumour growth. However, the reciprocal neurological effects of tumour progression on brain homeostasis are not well understood. Here, we review brain regions involved in regulating peripheral tumour development and how they, in turn, are adversely affected by advancing tumour burden. Tumour‐induced activation of the immune system, blood–brain barrier breakdown and chronic neuroinflammation can lead to circadian rhythm dysfunction, sleep disturbances, aberrant glucocorticoid production, decreased hippocampal neurogenesis and dysregulation of neural network activity, resulting in depression and memory impairments. Given that cancer‐related cognitive impairment diminishes patient quality of life, reduces adherence to chemotherapy and worsens cancer prognosis, it is essential that more research is focused at understanding how peripheral tumours affect brain homeostasis.

Recently, the CNS has been shown to be a critical contributor to tumour initiation and progression, via both direct and indirect modulation of the tumour micro-environment (TME) (Faulkner et al., 2019;Mauffrey et al., 2019). The relatively new field of Cancer Neuroscience aims for a better understanding of the complex interactions between the TME and the nervous system, specifically the influence of peripheral neuronal activity in both tumour initiation and cancer progression (Monje et al., 2020). Neurotransmitters and their receptors are also being investigated as novel and promising therapeutic targets to halt tumour growth. However, the reciprocal effects of tumour growth on the CNS and brain are being largely overlooked. The existence of pretreatment cognitive and affective disorders in people with non-CNS cancers indicates a bidirectional crosstalk between the TME and CNS, which is likely to occur indirectly, via modulation of the innate immune system, and/or directly, through neurotransmitter release (Monje et al., 2020;Olson & Marks, 2019). However, even though cancer-associated neurological disturbances have been reported in a variety of pretreatment peripheral cancers, the mechanisms underlying these phenomena are still unknown (Baekelandt et al., 2016;Berman et al., 2014). A better understanding of how extracranial tumour growth leads to neurological complications is critical to enhancing patient survival. Cancer-associated neurological disturbances often decrease the quality of life of people diagnosed with cancer, and this can have knock-on effects, such as reducing patient adherence to chemotherapy drug regimens and decreasing overall survival rates (Bender et al., 2014;Janelsins et al., 2014). Here, we review how different brain regions are involved in tumour progression and how they, in turn, can be adversely affected by peripheral tumour growth (Table 1).

| BLOOD-BRAIN BARRIER
The blood-brain barrier (BBB) is formed by microvascular endothelial cells, which line the cerebral capillaries of the brain and spinal cord.
The BBB plays a critical role in protecting the brain parenchyma from blood-borne molecules and cells (Kadry et al., 2020). Peripheral tumours are thought to affect brain function via disruption of the BBB and peripheral immune cell infiltration (Olson & Marks, 2019) ( Figure 1). The BBB is highly plastic and is known to undergo significant modification in response to physiological and pathological stimuli (Erickson & Banks, 2018). Impairment of the BBB has been reported in various CNS pathologies and particularly those initiated by inflammation (Wardill et al., 2016). The neuro-damaging capabilities of immune cell-derived pro-inflammatory cytokines are well characterised, and they are gaining more and more attention as the primary drivers of chemotherapy-and cancer-associated cognitive impairment. Although the neurotoxic side effects of chemotherapy were thought to be the main cause of BBB disruption in patients with cancer, the inflammatory environment created by the tumour itself is now being investigated as the initial source of BBB dysfunction. For example, some cytokines, such as IL-1β alter the paracellular barrier via breakdown and translocation of tight junction proteins, whereas other cytokines, such as TNF-α, mainly target the intracellular caveolae. Moreover, peripheral tumours could activate MMPs, via the release of pro-inflammatory cytokines, which further disrupt the basement membrane and tight junctions of the BBB (Kadry et al., 2020;Wardill et al., 2016). When BBB integrity is compromised, solutes and immune cells from the periphery can cross over to the brain more easily, where they can disrupt cognitive function (Geng et al., 2018).
Tumour-induced peripheral inflammation was also shown to induce astrogliosis in the brains of mice bearing Lewis lung carcinoma (Demers et al., 2018). Astrocytes are intimately linked with endothelial cells of the BBB via the direct contact of astrocyte endfeet with pericyte cells that wrap around CNS vasculature. Therefore, astrocytes are acutely tuned to sense changes in BBB permeability (Heithoff et al., 2021), although the exact contribution of astrocytes to the maintenance of BBB integrity remains disputed (Kubotera et al., 2019). Demers et al. (2018) evaluated BBB integrity and observed increased accumulation of fibrin levels around cerebral vasculature and subsequent endothelial cell activation, potentially via endothelial granule release of von Willebrand factor (VWF) (Demers et al., 2018). More recently, in a mouse model of pancreatic ductal adenocarcinoma, increased circulating immune cell infiltration was observed in conjunction with extensive neuroinflammation and cancer-associated cachexia (Burfeind, Zhu, Norgard, Levasseur, Huisman, Buenafe, et al., 2020a). Here, a distinct population of non-CNS neutrophils (CD45 high CD11b + ) expressing the chemokine receptor CCR2 were shown to accumulate at the meninges surrounding the hippocampus (known as the velum interpositum) of tumour-bearing mice. Pharmacological blockade of CCR2 prevented the circulating neutrophils from entering the brain and reduced the symptoms of cancer-associated cachexia. The authors further showed that tumour-derived factors induced the expression of the chemokine CCL2 in brain macrophages, providing a potential mechanism underlying the recruitment of CCR2 expressing neutrophils to the brain (Burfeind, Zhu, Norgard, Levasseur, Huisman, Buenafe, et al., 2020a). BBB integrity is also known to be disrupted by the release of extracellular vesicles (EVs) in systemic inflammatory diseases (Saint-Pol et al., 2020) (Figure 1). EVs are defined as membranederived vesicles (including apoptotic bodies, microvesicles and exosomes) and have been identified as important mediators of cell communication via the exchange of lipids, nucleic acids and protein products. Peripherally derived EVs can easily cross the BBB via the circumventricular organs and endothelial transcytosis (Balusu et al., 2016). Given the fact that a heterogeneous population of EVs are being released during tumour growth, their potential involvement in disrupting brain function in people with extracranial cancers is now receiving much attention (Koh et al., 2020) systemic circulation and CNS parenchyma (Morad et al., 2019). In a subsequent study, the group worked on elucidating the functional consequences of tumour-derived EV transport across the BBB (Morad et al., 2020). Astrocytes were found to be the main recipients of peripheral EVs within the brain via the non-canonical Cdc42-dependent, clathrin-independent carriers (CLIC)/GPI-anchored protein-enriched early endosomal compartment (CLIC/ GEEC) endocytosis pathway.
Upon tumour-derived EV infiltration, astrocytes were found to express significantly less tissue inhibitor of MMP-2 (TIMP2), which was hypothesised to promote a suitable micro-environment for growth of brain metastases. The authors further determined the EV factors that were driving the decrease in TIMP2 in astrocytes by evaluating a number of microRNAs (miRNAs) with the ability to target the 3 0 -untranslated region (3 0 -UTR) within the TIMP2 mRNA. A miRNA known as miR-301a-3p was found to be significantly increased in vitro and in vivo (Morad et al., 2020). Although the authors did not investigate the potential neurological effects of this EV-mediated decrease in TIMP2 expression and BBB disruption, it is likely that this could cause cognitive impairment. In the context of brain ageing, a decline in plasma and hippocampal TIMP2 levels is thought to be associated with age-related cognitive dysfunction. Systemic administration of TIMP2 was shown to increase hippocampal synaptic plasticity and to improve behaviour and memory function in aged mice (Castellano et al., 2017). Moreover, several miRNAs have already been identified for their involvement in neuropathological diseases. For example, cerebral miR-30 directly impairs hippocampal synaptic transmission and plasticity, and miR-301b was shown to accelerate microglial activation and cognitive impairment (Song et al., 2019;Tang et al., 2019). These findings suggest that tumour-derived EVs could contribute to tumour-associated neurological disorders through the shuttling of miRNAs that disrupt BBB integrity and/or directly alter neuronal function. Further investigation of tumour-derived brain-infiltrating EVs and their contents is likely to uncover more mediators capable of causing affective and cognitive disorders in cancer patients (Koh et al., 2020).

| HIPPOCAMPUS
Although there is little evidence of direct crosstalk between the periphery and the hippocampus, this limbic structure is one of the main brain regions investigated in studies of CRCIs and mood disorders in pretreatment cancer patients. This is due to the critical role of the hippocampus in the processing of sensory stimuli and in memory formation, consolidation and retrieval (Opitz, 2014). Moreover, the dentate gyrus of the hippocampus is one of only a few brain regions that display lifelong neurogenesis, a process particularly  (Xu et al., 2020) Abbreviations: 5-HIAA, 5-hydroxy indoleacetic acid (serotonin metabolite); BDNF, brain-derived neurotrophic factor; DA, dopamine; DOPAC, 3,4-dihydroxyphenylacetic acid (dopamine metabolite); EVs, extracellular vesicles; GFAP, glial fibrillary acidic protein; HO, hypocretin/orexin; mPFC, medial prefrontal cortex; NA, noradrenaline; TIMP-2, tissue inhibitor of MMP-2.
important in the positive mood-enhancing mechanism of action of several antidepressant medications (Kang et al., 2015). To date, most neuroimaging studies in humans with non-CNS tumours have focused on chemotherapy-related brain abnormalities, such as the diffuse decrease in grey and white matter volumes, the decrease in neuronal stem cell (NSC) proliferation in the dentate gyrus and subtle inward deformation of the hippocampus, which correlates with self-reported impairments in episodic memory (Deprez et al., 2018;Sim o et al., 2013). However, there is also evidence for the presence of structural and functional hippocampal changes in chemotherapy-naïve cancer patients, including hippocampal atrophy and the inability to deactivate regions within the posterior cingulate, which receives input from the hippocampus (Berman et al., 2014;Cimprich et al., 2010;Perrier et al., 2020;Scherling et al., 2012). Moreover, the cognitive domains that are commonly affected in people with non-CNS cancer before treatment, such as working memory, explicit memory, processing speed and executive function, all point towards the presence of hippocampal aberrations (Hermelink et al., 2007;Lange, Léger, et al., 2019;Meyers et al., 2005;Wefel et al., 2004). Given that increased levels of circulating inflammatory factors have been positively associated with cognitive dysfunction in people with cancer, prior to chemotherapy (Cheung et al., 2015;Lyon et al., 2016;Patel et al., 2015), and that the hippocampus is particularly vulnerable to stress and inflammation (Conrad, 2008), researchers have suggested a role for tumour-induced inflammation in pretreatment hippocampal dysfunction (Santos & Pyter, 2018;Schrepf, Lutgendorf, & Pyter, 2015). However, confounding factors, such as tissue traumarelated stress responses, anaesthesia-mediated immune suppression and opportunistic infection-driven systemic inflammation after surgery, can complicate the interpretation of these findings and make it difficult to attribute cognitive dysfunction directly with tumour burden-associated inflammation (Emmer et al., 2019;Joly et al., 2015;Kim, 2018;Lange, Léger, et al., 2019). Although the direct effects of peripheral tumour growth on the hippocampal formation is unclear in humans, several rodent studies support tumour-associated inflammation as a potential mechanism for hippocampal dysfunction (Norden et al., 2015;Pyter et al., 2009;Santos et al., 2019;Yang et al., 2014).
Tumour-associated systemic inflammation has been a major focus in trying to decipher the causes of pretreatment CRCI in people diag- F I G U R E 1 Peripheral tumour growth can affect blood-brain barrier (BBB) integrity via the secretion of pro-inflammatory mediators, immune cell activation and extracellular vesicles (EVs). Under normal circumstances, the BBB is constituted by endothelial cells, pericytes (not depicted), a basal membrane and astrocytic endfeet. The endothelial cells of the CNS are held together by tight junctions, which are formed by transmembrane molecules including claudins and occludins. These transmembrane adhesion complexes are linked to the cytoskeleton via cytoplasmic adaptor proteins, such as zonula occludens-1 (ZO-1). In response to peripheral tumour growth, BBB integrity is thought to be compromised via the release of pro-inflammatory cytokines (e.g., TNF-α and IL-1β) and MMPs. Increased fibrin deposition, immune cell infiltration and the transcytosis of tumour-associated EVs could cause further tight junction and basal membrane disruption. However, the exact effects of tumour-induced disruption of BBB integrity on cognitive and affective brain function require further elucidation. Figure was created with BioRender.com TNF-α and IL-10 expression in rats with N-nitroso-N-methylurea (NMU)-induced mammary tumours. Here, tumour-bearing mice showed depressive-like behaviour based on the Porsolt forced swim test, which was not observed by NMU exposure alone (i.e., before tumour development), indicating that tumour growth itself was responsible for the central inflammatory response (Pyter et al., 2009).
The increased production of pro-inflammatory cytokines in response to tumour growth is thought to signal to the brain via humoral and neural signalling (Dantzer, 2018). IL-10 is classically known as an antiinflammatory cytokine. However, the role of IL-10 in cancer is ambiguous, and it teeters between immune stimulation and immune suppression (Zhao et al., 2015). Yang et al. (2014) assayed the serum levels of circulating cytokines in a BALB/c mouse model 2 weeks after inoculation with a colon carcinoma cell line (CT26) (Yang et al., 2014).
In this study, an increase in the levels of serum IL-6 was observed, as well as increased mRNA levels of hippocampal IL-6 and TNF-α, which correlated with a decrease in the numbers of proliferating hippocampal neurons and a decrease in the mRNA levels of brain-derived neurotrophic factor (BDNF) and COX-2 in the hippocampus. In this study, tumour-inoculated mice showed depressive-like behaviour in the tail suspension test. This study suggests that peripheral tumour growth alone can cause neurological dysfunctions through IL-6-mediated inflammatory signalling, a reduction of hippocampal neurogenesis and decreased levels of hippocampal BDNF and COX-2 (Yang et al., 2014). However, although this study by Yang et al. (2014) suggests that depressive-like behaviour and memory impairments may be attributed to a decrease in hippocampal COX-2 expression, contrasting findings have been observed in tumour-bearing mice treated with low dose-aspirin. Aspirin, an anti-inflammatory nonselective COX inhibitor, prevented tumour-induced memory impairments, although it did not affect tumour-induced sickness or tumour growth, in 4T1 or EO771 mammary tumour mouse models . Moreover, human studies suggest that using aspirin in the year prior to receiving a cancer diagnosis reduced the risk of developing depression, anxiety and stress-related disorders in the year following a cancer diagnosis (data were adjusted for sociodemographic factors, co-morbidities and cancer characteristics).
Interestingly, aspirin was more protective than other non-steroidal anti-inflammatory drugs (NSAIDs) (Hu et al., 2020). Furthermore, the decrease in hippocampal BDNF mRNA that was observed in CT26-induced colon carcinoma by Yang et al. (2014) was not extended to other tumour models, including 4T07 and 4T1 mammary tumour models (Walker Ii et al., 2017) and NMU-induced ductal mammary tumours (Pyter et al., 2010). These observations highlight that tumour-induced changes in the hippocampus are highly cancer specific.
Using the C26 colon adenocarcinoma in vivo model, Norden et al. (2015) observed an increase in fatigue and depressive-like behaviour in the voluntary wheel running activity paradigm and in the sucrose preference and forced swim tests (Norden et al., 2015). They report that the increase in cancer-related fatigue and depression was associated with an increase in pro-inflammatory IL-1β and IL-6 mRNA expression in the cortex and hippocampus. Microglial immunoreactivity was also significantly increased in the cortex, although no changes were observed in the hippocampus (Norden et al., 2015). This suggests an increase in neuroinflammation and microglial priming in the brain in response to peripheral colon cancer.
More recently, Santos et al. (2019) examined changes in neuroinflammatory markers in response to mammary tumour growth and in response to an additional LPS-mediated immune challenge . Moreover, the extent to which mammary tumour resection could attenuate the tumour-induced neuroinflammatory response was also investigated. An in vivo mouse model of nonmetastatic mammary carcinoma (67NR cell line in BALB/c mice) was used. This was important to differentiate between the effects of peripheral tumour burden and blood-borne metastatic cancer cells that can invade and grow in CNS tissue. The protein levels of circulating TNF-α, IL-1β and IL-6 markedly increased in the plasma of both control and tumour-bearing mice following LPS stimulation. However, in the brain, the tumour-bearing cohort showed an attenuated neuroinflammatory response following peripheral immune challenge, compared with control mice. Although the mRNA levels of TNF-α, IL-1β and IL-6 increased in the hippocampus, hypothalamus and frontal cortex of control mice, this inflammatory response was much weaker in the brains of tumour-bearing mice following LPS stimulation. This is interesting in the context of the large body of literature describing how peripheral cancer growth is largely immunosuppressive and that immune modulation is a mechanism employed by tumours to evade destruction by cells of the innate immune system (Nguyen & Spranger, 2020). Interestingly, tumour resection restored the neuroinflammatory response to a peripheral immune challenge, measured as raised levels of hippocampal IL-6 and IL-1β, but not TNF-α. Their results indicate that peripheral tumour growth can significantly dampen neuroinflammatory pathways in the brain. However, no baseline mRNA levels (i.e., prior to the LPS challenge) were measured in the tumour-bearing mice and they did not investigate if the attenuated neuroinflammatory response had any effect on cognitive or affective functioning . Therefore, it is difficult to assess whether tumour-mediated attenuation of neuroinflammation has beneficial or detrimental effects on cognition and behaviour. In another study, 67NR mammary tumours were surgically resected 14 days prior to LPS-mediated immune challenge (Emmer et al., 2019). No changes in circulating inflammatory protein markers were observed between control mice (vehicle injection + sham surgery without removal of the mammary gland), mastectomy surgery mice (vehicle injection + removal of the mammary gland) and tumour mastectomy mice (tumour injection + removal of the mammary gland). However, mRNA expression of pro-inflammatory markers -IL-1β, TNF-α, CD68 and the IL-4 receptor α-subunit -was significantly increased in the hippocampus of tumour-bearing mice that had mastectomy surgery, in comparison with control mice. Of note, at the time of mastectomy, there were no significant differences in the hippocampal mRNA expression of the markers IL-1β, TNF-α, IFN-γ and IL-6 between tumour-and vehicle-injected animals (Emmer et al., 2019). Overall, these studies indicate that the hippocampus is particularly vulnerable to the presence of a peripheral tumour and is likely very susceptible to off-target actions of tumour-induced systemic inflammation. Therefore, much of the cognitive and affective disturbances in people with non-CNS tumours could be the result of a chronic low-level systemic inflammation that affects key brain structures involved in memory and the regulation of mood, such as the hippocampus ( Figure 2).

| HYPOTHALAMUS
The hypothalamus maintains essential physiological processes, including sleep, growth, metabolism, reproduction, body temperature, stress response, reward, feeding and circadian rhythms, mainly via the secretion of neurotransmitters. The disruption of many of these processes is linked to tumour initiation and progression (Masri & Sassone-Corsi, 2018). The hypothalamus is divided into three main regions (supraoptic, tuberal and mammillary) and three areas (paraventricular, medial and lateral). Each zone within the hypothalamus contains specific subsets of neurons, also known as nuclei, with spe-

| HO neurons in the lateral nuclei
Sleep disturbances, especially insomnia, are a common problem in people with extracranial cancers prior to, during and after antineoplastic therapy (Fiorentino & Ancoli-Israel, 2007). Unfortunately, poor sleep has been correlated with impaired 'quality-of-life' scores, metastatic spread and even increased mortality in cancer patients (Collins et al., 2017;Palesh et al., 2014). Thus, it appears that tumour growth promotes unhealthy sleep patterns, and reciprocally, poor sleep can further stimulate cancer progression (Walker & Borniger, 2019).
Moreover, sleep disruption has also been associated with weight gain, metabolic dysfunction and the development of a chronic inflammatory state, which could further contribute to cancer progression (Walker & Borniger, 2019). However, the underlying mechanisms remained unclear until Borniger et al. (2018) investigated the involvement of the brain in cancer-associated sleep disruption and metabolic dysfunction (Borniger et al., 2018).

Borniger et al. (2018) observed a marked increase in the circulat-
ing levels of IL-6 in mammary tumour-bearing mice (67NR), in conjunction with a disrupted sleeping pattern. These tumour-induced changes were associated with aberrant activity of HO-producing neurons (Borniger et al., 2018). Lateral hypothalamic HO neurons project throughout the brain to regulate a variety of functions, including F I G U R E 2 Peripheral tumour growth may affect the hippocampus, primarily via systemic inflammation. Increased levels of circulating inflammatory cytokines (e.g., IL-6, TNF-α and IL-1β), in response to peripheral tumour growth, appear to indirectly trigger an inflammatory response in the hippocampus. Because peripheral tumour growth is capable of altering various organ systems (e.g., hepatic metabolism, gut microbiome and spleen myeloid immune cell populations), tumour-derived signals are likely to be amplified and propagated to the brain via modulation of peripheral tissue physiology. In addition, tumour-induced systemic inflammation has been linked to a reduction in hippocampal neurogenesis, as well as the decreased expression of brain-derived neurotrophic factor (BDNF) and COX-2. Whereas the communication between the periphery and the brain is thought to occur via neural and humoral signalling, the exact mechanism is still unclear. Figure was (Sakurai, 2007). Inhibition of HO neuron signalling, using a dual HO receptor antagonist, almorexant, ameliorated the observed sleep abnormalities, but not the IL-6-mediated peripheral inflammation (Borniger et al., 2018). These results indicate a link between peripheral tumour growth and dysregulation of central neuromodulators involved in maintaining healthy sleep/wake cycles (Figure 3b). The hypothalamic-pituitary-adrenal (HPA) axis, which originates in the paraventricular nucleus of the hypothalamus, is activated by tumour-secreted factors, such as corticotropin-releasing hormone (CRH) and IL-6, as well as input from activated neuronal populations, such as HO neurons. In the presence of neoplastic growth, the HPA negative feedback loop is disrupted, which increases circulating levels of cortisol (CORT) and decreases dendritic branching and hippocampal volume, likely causing neuropsychological and neurocognitive disturbances. (d) Disruption of circadian rhythms, regulated by the suprachiasmatic hypothalamic nuclei, is frequently reported in people with extracranial tumours. In rodents, gene expression changes have been observed in core clock-regulating genes (e.g., Clock and Per1), which caused endocrine and immune rhythm dysfunction. Circadian clock dysfunction is likely to cause fatigue in people with non-CNS tumour growth and potentially affective and cognitive disturbances. Figure was created with BioRender.com and cognitive reserve (Krause et al., 2017 (Sakurai, 2007). HO neurons not only regulate neurotransmitter release but also modulate calcium-mediated synaptic plasticity via their G protein-coupled receptors (GPCRs), OX 1 and OX 2 (Kukkonen & Leonard, 2014). Moreover, HO neurons are strongly connected to the autonomic output nuclei in the brainstem, the sympathetic nervous system (SNS) and, most importantly, the HPA axis (Sakurai, 2007). Indeed, optogenetic stimulation of HO neurons was shown to activate the HPA axis, resulting in increased levels of circulating glucocorticoids (Bonnavion et al., 2015). This suggests that Here, microglia were shown to respond to tumour-derived factors and infiltrate the medio-basal hypothalamus, where they assumed an activated state. Interestingly, microglial depletion, or prevention of their activation using an antagonist against the colony stimulating factor 1 receptor, worsened cachexia. Moreover, antagonist-treated animals had significantly higher levels of circulating cortisol, which indicates activation of the HPA-axis. Overall, this study demonstrates a protective role for microglial activation within the hypothalamus in the context of cancer-associated cachexia (Burfeind, Zhu, Norgard, Levasseur, Huisman, Michaelis, et al., 2020b). However, the effects of localised microgliosis on cognitive and affective functioning in patients with cancer cachexia are yet to be determined.

| HPA axis in the paraventricular nuclei
As previously mentioned, many hypothalamic processes are influenced by the HPA axis, which is primarily driven by the paraventricular nucleus of the hypothalamus (DeMorrow, 2018). The HPA axis is known to be activated by internal and external stimuli, which are perceived as a threat by higher order cognitive centres.
These regions identify the threat to homeostasis and send excitatory signals to the hypothalamus to initiate the synthesis and release of corticotrophin-releasing hormone (CRH) and arginine vasopressin Psychological distress, which can begin with a cancer diagnosis and continue during treatment, is commonly experienced by people with cancer (Annunziata et al., 2011). Stress hormones, both cortisol and noradrenaline, have been shown to have a severe impact on disease progression and prognosis (Volden & Conzen, 2013). Thaker et al. (2006 were one of the first to show that elevated circulating levels of catecholamines resulted in increased tumour burden, angiogenesis and invasiveness of ovarian carcinoma cells in an orthotopic mouse model (Thaker et al., 2006). These findings were later corroborated in other cancer mouse models, which showed accelerated tumour progression in the presence of stress hormones, caused by disrupting anti-tumour immunity and other endocrine effects on tumour metabolism, angiogenesis, oncogenesis and DNA repair (Flaherty et al., 2017;Zhang et al., 2019). In humans with adrenocortical carcinoma, glucocorticoid excess was associated with low amounts of tumour-infiltrating lymphocytes and poor overall survival (Landwehr et al., 2020). In turn, altered cortisol responses and aberrant adrenergic signalling in cancer patients have also been correlated with neurological issues, such as cancer-related fatigue, anxiety, depression and potentially CRCI (Andreotti et al., 2015;Schmidt et al., 2016;Weinrib et al., 2010).
The HPA axis not only regulates the immune system, but the reverse also happens. For example, immune-neuroendocrine feedback has been shown to play a role in viral infections. Immune cells can dampen their own activity via the production of pro-inflammatory cytokines and the subsequent release of glucocorticoid hormones, which, in turn, attenuates the production of pro-inflammatory mediators by immune cells (Silverman et al., 2005). A similar mechanism could be happening in people with cancer prior to receiving treatment.
For example, the pro-inflammatory cytokine, IL-6, is a potent inflammatory cytokine released by various cells within the TME, and increased serum levels of IL-6 have been demonstrated in various cancers (Masjedi et al., 2018). IL-6 was shown to be a potent activator of the adrenal axis following LPS-mediated immune activation (Bethin et al., 2000). Moreover, in colon or pancreatic tumour-bearing mice, an IL-6-mediated decrease in hepatic ketogenesis was associated with a glucocorticoid-mediated stress response, decreased anti-tumour immunity and decreased immunotherapy efficacy (Flint et al., 2016).
In patients undergoing glucocorticoid therapy, serum levels of IL-6 are significantly reduced (Fujio et al., 2016). However, in the presence of neoplastic cell growth, the HPA negative feedback loop appears disrupted, which is likely to lead to the observed neurocognitive and neuropsychological disturbances in cancer patients. Interestingly, IL-6 and hyperactivity of the HPA axis also play a crucial role in the aetiology of depression. Jehn et al. (2010) found a strong correlation between increased plasma levels of IL-6 and HPA-axis dysfunction in cancer patients diagnosed with depression (Jehn et al., 2010). This faulty HPA feedback system is thought to be due to desensitised glucocorticoid receptors in the brain. Pro-inflammatory cytokines, including those produced in response to tumour growth, contribute to this desensitisation of glucocorticoid receptors by impairing the nuclear translocation and/or transcriptional function of these receptors (Silverman & Sternberg, 2012). Tumour cells, in conjunction with recruited immune cells, not only produce inflammatory cytokines but also secrete various other bioactive peptides that can affect the HPA axis.
For example, intra-tumour production of CRH has been reported in human ovarian cancer (Minas et al., 2007) and in human melanoma cells (Sato et al., 2002). Prolonged activation of the HPA axis and

| Suprachiasmatic nuclei
The master regulator of the circadian system is located within the hypothalamus, known as the paired SCN. The SCN regulate neuronal activity, body temperature, hormone release, immune function, sleep and feeding in a 24-h cycle based on light-dark input from the retinal ganglion cells (Dumbell et al., 2016). Cancers have been linked to disruptions in the circadian rhythm (Sulli et al., 2019). Moreover, disruption of the circadian rhythm is strongly associated with brain dysfunction, suggesting that tumour-induced disruption of the circadian rhythm could alter brain function in cancer patients (Cash et al., 2015;Logan & McClung, 2019). After quantification of clock-related gene expression in the hypothalamus, two core clock-regulating genes (Clock and Per1) and four additional clock-related genes (Aanat, Camk2a, Creb1 and Mntr1a) were found to be differentially expressed in mammary tumourbearing mice (67NR). Alongside hypothalamic gene expression changes, the researchers observed a significantly higher percentage of running wheel locomotor activity during the light phase of a 14:10 light-dark cycle in tumour-bearing mice, in comparison with non-tumour-bearing controls and tumour-resected animals. Moreover, the normal 24-h rhythm in circulating corticosterone and circulating neutrophils (CD11b + /Ly6G + ) was absent in tumour-bearing mice (Sullivan et al., 2019). Their findings confirm that peripheral tumour growth, independent of harsh cancer treatments, can disrupt the central molecular clock, as well as the physiological rhythms of the endocrine and immune systems (Figure 3d). They provide a potential explanation for the behavioural co-morbidities seen in cancer patients. However, the exact mechanism of how tumour growth causes circadian rhythm dysregulation is still unclear but is likely to be due to tumour-secreted factors, such as inflammatory cytokines.
For example, abnormal diurnal cortisol rhythms in people with epithelial ovarian cancer prior to treatment were associated with increased levels of the pro-inflammatory cytokine, IL-6, and overall decreased survival (Schrepf, Thaker, et al., 2015) (Figure 3d).

| MIDBRAIN VENTRAL TEGMENTAL AREA
The midbrain ventral tegmental area (VTA) is commonly known for its role in motivational behaviours and in the brain's reward system.
However, more recently, the VTA has gained attention because of its ability to modulate tumour progression via the production of the neurotransmitter, dopamine (Ben-Shaanan et al., 2018;Xu et al., 2020).
The VTA is one of the primary sources of dopamine within the brain.
Dopamine is important in regulating motivation, mood and cognition and also exerts various actions in the periphery, such as blood pressure regulation and modulation of both the innate and adaptive immune systems . Dopamine primarily mediates its effects via dopamine receptors, which are GPCRs and can be divided shown to enhance phagocytic activity of splenic dendritic cells and macrophages (Ben-Shaanan et al., 2016). In turn, immune dysfunction is likely to affect dopaminergic signalling in both the CNS and periphery (Mackie et al., 2018).
Until now, cancer biologists have mainly focused on peripheral dopamine and its receptors , and therefore, the central dopaminergic system has received little attention for its ability to modulate tumour progression. Several epidemiological studies have suggested that people with schizophrenia, associated with hyperreactive dopaminergic systems, are relatively protected from developing cancer (Li et al., 2018). However, gender and the type of cancer studied are two important variables because men showed a significant decrease in the incidence of colorectal cancer, whereas women with schizophrenia may have a higher risk of developing breast cancer (Wu Chou et al., 2017). Similarly, the development of Parkinson's disease, marked by declining dopamine levels, was associated with a lower risk of smoking-related cancers and a higher risk of developing malignant melanoma and breast cancer (Driver et al., 2007).
Smaller primary tumours, reduced angiogenesis and a reduced metastatic potential of the tumour were observed following mammary tumour implantation (MADB106) in Wistar APO-SUS rats with high dopaminergic reactivity, compared with those in APO-UNSUS rats with low dopaminergic reactivity (Teunis et al., 2002). More recently, chemogenetic activation of VTA-dopaminergic neurons was shown to reduce primary tumour volumes in two lung cancer mouse models (LCC and B16 melanoma) (Ben-Shaanan et al., 2018). To study how activation of VTA-dopaminergic neurons within the brain could signal to the primary tumour, the SNS was ablated using the neurotoxin 6-hydroxydopamine (6-OHDA). In mice that received 6-OHDA, acti-  (Xu et al., 2020).
Given the known roles of the VTA in regulating mood, the studies discussed above have uncovered a potential physiological mechanism underlying how a patient's psychological status may modulate antitumour immunity and cancer progression. However, could modulation of tumour progression by the VTA-dopaminergic axis and the SNS have physiological consequences that act back on the brain? In chronic neurological conditions, such as depression, the SNS is known to be continuously activated without the balancing counteractions of the parasympathetic nervous system. This results in immune system activation and increased levels of systemic pro-inflammatory cytokines, which is often reported in patients with depression (Won & Kim, 2016). Therefore, chronic activation of the SNS by the presence of a peripheral tumour could prime the immune system and increase peripheral inflammation that, over time, may lead to depression in people with untreated cancers (Sforzini et al., 2019). Indeed, inflammation-driven alterations to metabolic pathways, such as the indoleamine 2,3-dioxygenase (IDO)-kynurenine and tetrahydrobiopte rin (BH4) pathways, have been shown to cause substantial alterations in the synthesis of dopamine, 5-HT and noradrenaline and are likely to be the cause of depressive symptoms in cancer patients (Vancassel et al., 2018). In a mouse model inoculated with B16F10 melanoma cells, for example, the tumours induced depressive-like behaviour, as assessed by the tail suspension test, and led to increased levels of IL-6 and TNF-α in the brain (Lebeña et al., 2014). This corresponded with a decrease in dopaminergic activity in the striatum and a decrease in 5-HT in the PFC, with both regions receiving dopaminergic inputs from the VTA (Lebeña et al., 2014).
These alterations in monoamine turnover have been ascribed to tumour-induced dysfunction of metabolic processes within the brain.
With this in mind, Kovalchuk et al. (2018) performed direct flow injection/MS (DI-MS) to assess the effects of extra-cranial tumour growth (including lung cancer, pancreatic cancer and sarcoma) on the brain metabolome in mice (Kovalchuk et al., 2018). The authors demonstrated that non-CNS tumour growth affected protein biosynthesis, and amino acid and sphingolipid metabolism. Of five different pathways that were commonly affected in all three cancer types, protein biosynthesis was significantly up-regulated in all three groups. Moreover, amino acid metabolism (including phenylalanine and tyrosine), as well as valine, leucine and isoleucine degradation, was commonly affected (Kovalchuk et al., 2018). Decreased amino acid levels or the deregulation of their metabolic machinery in the brain can cause neuronal death and apoptosis. Moreover, the observed alterations in catecholamine biosynthesis and increased levels of phenylalanine and tryptophan are known to cause neurotransmitter imbalances in levels of 5-HT, dopamine and noradrenaline (Xu et al., 2016). These findings further support the notion that tumour-induced inflammation can cause dysfunctions in monoaminergic systems, which in turn can lead to altered neural signalling in the affected brain regions (Figure 4).

| BRAIN STEM
Peripheral tumours are also likely to modulate the CNS by influencing the vagus nerve. The vagus nerve is the main component of the parasympathetic nervous system and relays inflammatory, satiety and metabolic cues to the brain. The vagal afferent fibres terminate within the nucleus tractus solitarius (NTS) in the brain stem, which further projects the information to higher brain regions including the hypothalamus, locus coeruleus and amygdala (Breit et al., 2018). The vagus nerve has been implicated in various diseases, including cancer.
Elevated vagal tone, measured by heart rate variability (HRV), is thought to exert a protective effect on cancer prognosis via the inhibition of oxidative stress, inflammation and sympathetic activity.
However, rodent studies investigating the effects of vagal denervation have uncovered both protective and detrimental effects on tumour initiation, progression and metastasis (Partecke et al., 2017;Zhao et al., 2014).
In response to peripheral cancers, the vagus nerve is thought to inform the brain about tumour growth via inflammatory signals.
In response, the brain can modulate tumour growth via the neuroendocrine and immune systems (Giese-Davis et al., 2015). Activation of the vagus nerve has been recorded electrophysiologically following peripheral administration of the pro-inflammatory cytokines, IL-1β and TNF-α (Tsaava et al., 2020). On one hand, vagal nerve stimulation (VNS) has been shown to be predominantly neuroprotective and ameliorates LPS-induced cognitive dysfunction in mice (Huffman et al., 2019). On the other hand, stress, which is common in cancer patients, inhibits the vagus nerve and has deleterious effects on the gastrointestinal tract and gut microbiota (Bonaz et al., 2018). As such, the vagus nerve also serves as a connection between the enteric nervous system (ENS) and the CNS, also known as the gut-brain axis. The gut-brain axis is responsible for monitoring the body's physiological homeostasis by connecting emotional and cognitive brain areas to peripheral functions, such as immune activation and enteroendocrine signalling.
Neuroactive compounds, such as ACh and 5-HT, released by the gut microbiota reach the CNS via the blood and circumventricular organs or via the vagus nerve (Bonaz et al., 2018). Altered compositions of the gut microbiota have been linked to neurological complications, including depression and anxiety (Breit et al., 2018).
Given that dysbiosis of the gut microbiome has been reported in cancer patients, gut-brain signalling via the vagus nerve could be another potential pathway for neurological dysfunction in F I G U R E 4 Peripheral tumours may disrupt monoaminergic neurotransmitter synthesis in the midbrain ventral tegmental area (VTA). The midbrain VTA is not only modulated by tumour progression, via the production of dopamine, but is also reciprocally affected by peripheral tumour growth. Tumour-induced inflammation and immune cell activation caused alterations to metabolic pathways (e.g., indoleamine 2,3-dioxygenase [IDO]-kynurenine), which has a marked effect on the synthesis of monoaminergic neurotransmitters, such as dopamine and 5-HT. Enhanced tryptophan (Trp) breakdown, which is reflected by an increase in circulating kynurenine (Kyn), is observed in various peripheral cancers. Trp breakdown is mainly mediated by IDO-1, which is expressed by antigen-presenting cells (e.g., macrophages and dendritic cells) in response to inflammatory stimuli (e.g., IFN-γ, TNF-α, COX-2 and PGE 2 ). Quinolinic acid (QUIN), also a tryptophan by-product, is a selective agonist of the NMDA glutamate receptor. QUIN accumulation results in excitotoxicity, oxidative stress (via accumulation of ROS), neuronal cell death and disruption of glutamatergic neurotransmission (Lugo-Huitr on et al., 2013). As QUIN cannot cross the blood-brain barrier (BBB), it can only be synthesised by resident microglia or infiltrated monocytes/macrophages (Lugo-Huitr on et al., 2013). Increased levels of inflammatory markers, ROS and reactive nitrogen species (RNS) also contribute to the oxidation of tetrahydrobiopterin (BH4). Limited bioavailability of BH4 for phenylalanine hydroxylase (PAH), TH and tryptophan hydroxylase (TrypOH) activity drastically reduces monoamine synthesis, including 5-HT and dopamine. Although disruption of monoamine synthesis has only been demonstrated in rodent studies, this mechanism is a likely cause of depressive symptoms and cognitive dysfunction in people with extra-cranial tumours. Figure

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article because no new data were created or analysed in this study.