Mechanisms of Neurotoxic Symptoms as a Result of Breast Cancer and Its Treatment: Considerations on the Contribution of Stress, Inflammation, and Cellular Bioenergetics

Cancer therapy can elicit an inflammatory response through several pathways, including direct immune changes in the tumor microenvironment, tumor cell death, and damage to healthy tissue [27‐32]. Associations between symptoms and inflammatory markers such as interleukin (IL)-6, tumor necrosis factor (TNF)-α, and C-reactive protein (CRP) have been observed for every treatment modality, in both cross-sectional and longitudinal designs [13, 33‐38]. Smith et al. [39] observed reduced DNA methylation in blood mononuclear cells in patients who recently completed chemotherapy (compared with treatment-naïve patients) that was associated with higher plasma concentrations of sTNFR2 and IL-6. sTNFR2 in turn was associated with fatigue. Interestingly, these associations were only observed shortly after chemotherapy completion, suggesting a transient effect of chemotherapy on inflammation and subsequent fatigue.

Associations between inflammatory markers and symptom severity are not always driven by cancer treatment. For example, Reed et al. [40] observed an increase in plasma TNF-α as well as in the severity of a cluster of symptoms (including fatigue, nausea, and pain) during cancer treatment. However, although IL-6 did not change during cancer treatment, the symptom cluster was associated with both TNF-α and IL-6, suggesting that only part of the inflammation leading to symptom experience was due to treatment.

In a sample of early stage breast cancer patients followed for 2 years, Starkweather et al. observed an association between increased CRP and reductions in several aspects of cognitive functioning but not between CRP and cancer treatment modality [41]. However, in a follow-up study using the same patient sample, cognitive functioning was associated with treatment modality [42]. Further, CRP was associated with body mass index and education level, both of which also predicted cognitive functioning, thus suggesting that the contribution of CRP to symptom severity was not dependent on treatment, but rather reflective of more stable patient characteristics. This same group also reported that a cluster of symptoms including fatigue was not associated with CRP [41].

Reports are also emerging on associations between genetic polymorphisms of inflammation-regulating genes and symptoms, which could contribute to individual differences in the inflammatory response to cancer treatment. Doong et al. [43••] found that single nucleotide polymorphisms (SNPs) in inflammation-regulating genes IL6, TNFα, and IL13 predicted a combination of severe fatigue and pain in patients prior to breast cancer surgery. In this study, more patients with high symptom severity than with low symptom severity had undergone neoadjuvant chemotherapy. Thus, symptom experience was possibly the result of chemotherapy-induced inflammation. In the same sample, SNPs in the IL1R1 and IL13 genes were associated with breast pain prior to surgery [16], and SNPs in the IL1r2 and IL10 genes were associated with severe and ongoing breast pain after surgery [44]. Kober et al. [45] reported increased prevalence of SNPs in the IL1b and IL10 genes in breast cancer patients who consistently reported high levels of fatigue during a 6-month period. Of note, patients in the high fatigue group more often had undergone neoadjuvant chemotherapy or received chemotherapy during the study. Thus, here again, fatigue was probably caused by chemotherapy-induced inflammation, which might have been exaggerated by a genetic vulnerability for inflammatory responses. In contrast, Chae et al. [46] reported that SNPs in TNFα and IL6 genes were not associated with chemotherapy-related cognitive impairments. We propose that associations between certain SNPs in inflammation-related genes and symptom experience are inconclusive at best. Most likely, higher symptom burden is related to the severity of the cancer treatment. It would be interesting to assess whether a genetic profile fostering a proinflammatory phenotype would identify patients with increased vulnerability to neurotoxic symptom development.

Although inflammation has been associated with symptoms during survivorship, Smith et al. [39] and others suggest that the peripheral mild increase in inflammatory mediators at this time is not the direct result of previous cancer treatment. However, the findings described above on symptoms and SNPs in inflammatory genes during and shortly after treatment might also explain the associations between inflammation and symptoms during survivorship: these SNPs would imply an enhanced vulnerability for inflammatory responses to new challenges experienced after cancer therapy. For example, Bower et al. [47•] showed that a higher genetic risk index (i.e., presence of alleles associated with higher IL-1β, IL-6, and TNF-α expression) was associated with fatigue and memory complaints in breast cancer patients 3 months after primary treatment. This approach of studying a combination of SNPs already linked to increased expression of proinflammatory mediators provides more insights than studying individual SNPs for which the functionality is often unknown. More studies using this approach will eventually provide a better insight into the possibility of a proinflammatory phenotype as a vulnerability for developing neurotoxic symptoms.

Evidence is accumulating that cancer therapy leads to changes in brain structure and metabolism that are associated with inflammation and symptoms. Kesler et al. [48] reported an association between lower left hippocampal volume as measured by MRI and higher levels of circulating TNF-α and lower levels of IL-6 in breast cancer survivors, but not in controls. Verbal memory performance in survivors was predicted by an interaction between hippocampal volume and TNF-α. Most of the breast cancer survivors in this study (>80%) had received a chemotherapeutic cocktail including doxorubicin, which was shown in an animal model to increase TNF-α (and IL-1β) in the hippocampus [49]. Thus, the association between circulating TNF-α and hippocampal volume found in the clinical study possibly reflects associations between central TNF-α and hippocampal volume, leading to cognitive symptoms. A small pilot study that included eight breast cancer patients who underwent chemotherapy during the study [50] found associations of higher sTNFR2 and IL-6 levels with decreased gray matter volume in specific regions. However, fatigue and cognitive functioning were not related to the inflammatory markers. Two preliminary reports from an ongoing longitudinal study showed associations between plasma levels of inflammatory mediators and reduced brain metabolism (as measured by FDG PET-scan) after chemotherapy for breast cancer: Ganz et al. [51] showed an association between plasma sTNFR2 and increased memory complaints, along with diminished resting-state metabolism in the inferior frontal regions in a subsample of 12 breast cancer patients who had undergone chemotherapy. Declines in sTNFR2 over the 12-month study period were associated with fewer memory complaints. Interestingly, associations between plasma sTNFR2 and memory complaints diminished when controlled for fatigue, implying that the report of memory complaints was partly due to inflammation-induced fatigue. Conversely, Pomykala [52], reporting on the same longitudinal study but with a larger sample (23 chemotherapy-treated patients), observed associations between higher plasma levels of inflammatory mediators (IL-6, sTNFRII, IL-1ra, and CRP) and increased resting-state metabolism in prefrontal and anterior temporal cortex shortly after chemotherapy completion as well as 1 year later. Metabolism in these brain regions was associated with patient-reported memory complaints. These apparently contradictory findings might be explained by a compensatory system in which reduced activity in some brain regions is compensated by increased activity in other regions. Zick et al. [53] found that both IL-6 and specific brain metabolites (higher glutamate + glutamine to N-acetyl-aspartate ratio in the posterior insula) predicted fatigue in breast cancer survivors. However, these brain metabolites did not correlate with IL-6, suggesting that not all changes in brain metabolism are inflammation-related, although we note that only one inflammation marker was tested.

Cancer therapy can elicit a peripheral inflammatory response that may reflect central inflammation, and correlations between neurotoxic symptoms and peripheral inflammation markers are consistently found, even long after completion of cancer treatment. Individual genetic differences in the capacity to mount an inflammatory response to environmental challenges might explain who will develop inflammation-related symptoms during and after cancer therapy. However, not all inflammatory mediators associated with symptom development result from cancer therapy, and there is little evidence to suggest that therapy-related inflammatory processes last long after treatment cessation. Thus, other causal mechanisms of inflammation both during and after treatment need to be considered.

Findings on brain structure and metabolism, although sparse and in need of replication, suggest that chemotherapy-induced increases in inflammation, specifically TNF-α or sTNFR2 (and to a lesser extent, IL-6), are associated with changes in brain volume accompanied by decreased brain metabolism in some regions and possibly with compensatory increases in metabolism in other regions. These central changes in volume and metabolism are associated with both cognitive impairment and fatigue.

Distress peaks at the time of a breast cancer diagnosis, with 60–77% of patients reporting moderate to severe distress at this point [54, 55] and declines slowly after completion of treatment [56]. Still, up to 40% of patients experience moderate to severe distress 6 months after diagnosis, and 30% report distress 15 months postdiagnosis [57].

Associations between distress and symptom severity are observed before, during, and after completion of cancer therapy [58]. Xiao et al. [59] reported that stress led to fatigue in breast cancer survivors, mostly via the development of depressive symptoms. These effects were independent of the effects of several inflammatory mediators related to fatigue. Several factors have been identified as contributors to a heightened vulnerability for stress and subsequent neurotoxic symptoms in breast cancer patients. For example, neuroticism is a personality characteristic related to high stress vulnerability. Deimling et al. [60] found that neuroticism was the strongest predictor of cancer-related worry in older adult survivors, and Lo-Fo-Wong et al. [57] reported that high stress at both 6 and 15 months postdiagnosis was associated with more frequent cancer worry and higher neuroticism. Neuroticism has also been related to fatigue in breast cancer survivors [61]. Stress during diagnosis of breast cancer was related to younger age and having experienced emotional problems in the past in a study by Jorgensen et al. [54]. Also prior to cancer treatment, Han et al. found an association between childhood trauma and increased expression of gene transcripts related to inflammatory signaling [62].

Cortisol is released by the hypothalamic-pituitary-adrenal (HPA) axis in response to psychosocial stress. Elevations in cortisol are typically followed by a rapid decline in cortisol production due to feedback mechanisms inhibiting the HPA axis. Stress-related disorders are characterized by sharper acute increases in cortisol in response to a stressor or slower declines after the stressor, indicating a sensitized HPA response or a desensitized feedback loop, respectively [63, 64]. In addition, cortisol concentrations follow a diurnal rhythm, with a sharp increase during awakening followed by a gradual decline during the day. Stress has also been shown to lead to a blunting of the diurnal slope (i.e., reduced decrease in cortisol during the day) [65], which has been associated with, among others, depressive symptoms, fatigue, and pain [66, 67].

Preliminary evidence indicates that tumors can affect endocrine function. Preclinical studies have shown that corticosterone (the primary glucocorticoid in most animals) levels are generally higher in tumor-bearing rodents than in tumor-free controls, and stress-hormone responsivity to stress is reduced in these animals [68•]. Similar observations have been made in patients, although causes of the dysregulations are more difficult to determine and could include a combination of the tumor and psychosocial stress associated with the diagnosis [68•].

Studies of the association between general neuroendocrine functioning and symptoms during cancer treatment suggest that cancer treatment itself does not affect neuroendocrine functioning. For example, Tell et al. [69•] showed that radiation therapy was not a significant predictor of changes in diurnal cortisol rhythm in recently diagnosed breast cancer patients. However, an association between higher fatigue with higher wakening saliva cortisol levels and a flatter slope in diurnal cortisol decline was observed, suggesting that dysregulations in the neuroendocrine stress response occur independently of cancer treatment. Further, Schmidt et al. [70] reported an association between physical fatigue and increased evening cortisol levels as well as higher overall cortisol secretion in a patients at different stages of treatment; the association between cortisol parameters and fatigue was independent of treatment status.

Although primary cancer treatments do not seem to affect neuroendocrine functioning, supportive medication and adjuvant endocrine therapy do. Dexamethasone, a synthetic glucocorticoid often used during chemotherapy, is known to suppress adrenal function [71]. Furthermore, Baumgart et al. [72] reported higher cortisol levels in survivors treated with aromatase inhibitors or tamoxifen. Andreano et al. [73] observed decreased cortisol response to a physical stressor (cold pressor test) in combination with mild impairments in memory performance in nine survivors receiving treatment with the GRnH agonist Lupron. In summary, breast cancer patients show alterations in neuroendocrine functioning that are associated with symptom experience. Whereas primary cancer treatments do not seem responsible for these alterations, supportive medication and endocrine therapy have been shown to alter neuroendocrine function. Observations on alterations occurring very early on, such as reported by Tell et al. [69•], suggest that nontreatment factors such as the tumor itself and stress due to the diagnosis also contribute.

Several studies by Bower et al. indicated that the neuroendocrine system in breast cancer survivors has reduced capacity to regulate inflammatory responses and that fatigue during survivorship is associated with a blunted cortisol response and more-pronounced inflammatory response to psychological stress [74, 75]. Possibly, the lack of cortisol response leads to insufficient control of the inflammatory response. This group also reported a decreased expression of transcripts containing response elements for glucocorticoids on leukocytes, combined with increased expression of genes coding for a proinflammatory reaction [76]. These changes in gene expression would result in a decreased leukocyte sensitivity to the (inhibiting) effects of cortisol, combined with a more pronounced readiness to initiate a proinflammatory response. In line with this, fatigued breast cancer survivors exhibit a flatter diurnal cortisol slope [77], suggesting that the feedback loop for cortisol is also desensitized. A flatter cortisol slope would result in higher daily levels of cortisol, which might ultimately contribute to immune cells being less sensitive to the effects of cortisol. These findings point towards dysregulated neuroendocrine response and decreased capacity of glucocorticoids to regulate the inflammatory response in survivors, both reflecting the effects of chronic exposure to stress hormones induced by stress and cancer treatment.

Stress also activates the sympathetic nervous system (SNS), leading to enhanced levels of catecholamines such as epinephrine. Catecholamines influence immune activity mostly via activating β2-adrenergic receptors on leukocytes, leading to increases in intracellular cAMP and activation of anti-inflammatory cytokines [78]. However, especially norepinephrine has a high affinity for α-adrenergic receptors which are present on monocytes and macrophages that can convey more proinflammatory signaling [79, 80].

Repeated exposure to catecholamines as a result of chronic stress has been linked to poorer breast cancer outcomes [81, 82•]. However, few studies have directly investigated associations between catecholamines and symptoms in breast cancer patients and survivors. Thornton et al. [83] reported moderate positive associations between epinephrine and symptoms of pain and fatigue in patients with advanced breast cancer.

The COMT gene regulates expression of catechol-O-methyltransferase (COMT) enzymes, which metabolize dopamine, norepinephrine, and epinephrine; several SNPs in the COMT gene have been associated with symptoms in breast cancer patients and survivors. Functional outcomes of one of these SNPs (i.e., rs4680 or Val158Met) are well characterized: compared to carriers of the Val allele, homozygous Met (Met/Met) carriers have three to four times lower enzyme activity resulting in slower clearance of catecholamines. The homozygous Met genotype has been associated with increased stress vulnerability [84] but better cognitive performance [85] in non-stressful situations [86]. Breast cancer survivors carrying the Met/Met genotype reported higher fatigue and exhibited higher pain sensitivity and higher cortisol levels compared with survivors carrying the more stress-resilient variants [87, 88]. Showing the protective effect of the Met/Met genotype on cognition, Small et al. [89] reported that Met/Met carriers performed better on several measures of cognitive performance and showed an interaction between the genotype and type of cancer therapy: carriers of the Val allele who had received chemotherapy performed more poorly on tests of attention, compared with Val allele carriers who had received radiation. A polymorphism in a different region of the COMT gene (rs165599) was associated with less decline in memory performance during chemotherapy [90].

One important outcome of increased activity of the sympathetic nervous system, and thus increased secretion of catecholamines, is decreased heart rate variability (HRV). HRV reflects a balance between sympathetic and parasympathetic activity. The sympathetic nervous system affects the heart rate via release of epinephrine and norepinephrine [91]. Higher catecholamine levels lead to lower HRV, which is associated with poorer health outcomes [91], depression, and fatigue. Lower HRV is also a known outcome of chronic psychosocial stress. Treatment with anthracyclines like doxorubicin, a class of chemotherapeutics commonly used in the treatment of breast cancer, are thought to lower HRV [82•]. Several studies reported an association between lower HRV and fatigue in breast cancer survivors [92]. Fagundes et al. showed a lower resting-state HRV and higher norepinephrine levels in association with higher fatigue in breast cancer survivors. Interestingly, changes in HRV in response to an experimental acute stressor were not associated with fatigue, suggesting that basal secretion of catecholamines, but not catecholaminergic responsiveness, is related to fatigue. Underlining the notion that inflammation is not the only pathway via which stress leads to symptom experience, Crosswell et al. [93] observed an association between lower HRV and higher fatigue in breast cancer survivors but did not find a mediating effect of inflammation in this association.

Stress is highly prevalent in breast cancer patients and, especially when prolonged, can lead to dysregulations in neuroendocrine functioning that generally represent as higher overall cortisol levels combined with reduced responsivity to cortisol. Although primary cancer therapies such as chemotherapy and radiation do not seem to alter endocrine function, synthetic glucocorticoids administered during cancer therapy and adjuvant hormonal therapies do have the potential to alter endocrine function. The few studies on catecholaminergic outcomes suggest that increased catecholamine secretion is associated with increased symptom experience.

Several chemoagents have been shown to affect mitochondrial function. Doxorubicin treatment has been associated with impairments in hippocampal mitochondrial enzyme complexes I and II and with redox activity in rats [49]. Cisplatin treatment was shown to decrease neuronal mitochondrial capacity on the level of spare respiratory capacity [97, 98], which translates to reduced capacity to increase energy production when demand increases. Taxanes, often used in breast cancer treatment, were shown to cause structural mitochondrial damage [99]. Other cancer treatment modalities likewise seem to affect mitochondrial function. For example, radiation therapy was shown to increase oxidative stress markers in breast cancer patients with severe acute skin reactions to radiation [100]. Further, maintenance therapeutic agents such as tamoxifen were also shown to negatively affect mitochondrial functioning [101]. Interestingly, in a rodent model of head and neck cancer, chemoradiation seemed to normalize the tumor-related changes in mtDNA expression in the liver while causing severe changes in mtDNA expression in the brain, suggesting that cancer treatment specifically affects cellular energy production in the central nervous system.

Glucocorticoids, released during stress or administered as part of cancer therapy, can affect mtDNA transcription by binding to mitochondrial GRs [95, 102]. Glucocorticoid effects on mitochondrial energy production seem dose-dependent in a u-shaped manner: exposure to several low doses of glucocorticoids is associated with enhancement of neuronal mitochondrial function, whereas exposure to high doses is associated with a reduction in function [103]. Nonetheless, low-dose glucocorticoid exposure can become detrimental when repeated often enough [104, 105]. Catecholamines, released during stress by the SNS, can also affect mitochondrial function, either by increasing the metabolic state [106, 107], or by inducing ROS production leading to cytogenetic damage [108].

Cancer therapy can lead to long-lasting changes in the central nervous system. For example, Kesler et al. [109] showed alterations in cerebral white-matter organization as assessed with diffusion tensor imaging in chemotherapy-treated breast cancer survivors, along with altered functional connectivity in several brain regions [110, 111]. Furthermore, inflammation has been associated with reductions in brain volume and brain metabolism, as discussed above. We propose that these changes are (in part) the result of reduced cellular activity due to mitochondrial damage.

Evidence is emerging that cancer treatment-induced neurotoxicities are related to mitochondrial damage and reduced function. Kober et al. [112], comparing breast cancer patients undergoing chemoradiation who did or did not have evening fatigue, showed increased expression in genes related to mitochondrial dysfunction in patients with evening fatigue. Furthermore, radiation-related fatigue was associated with expression of several mitochondrial genes in prostate cancer patients [113].

Animal models provide insight into effects of mitochondrial dysfunction on neurotoxic symptoms. Xiao et al. [114] observed an increase in neuropathic pain after administration of mitochondrial toxins in rats with taxane-induced neuropathy, but not in rats that had not received taxanes, suggesting that taxanes render the mitochondria more vulnerable to further insult. Our group showed that taxane treatment-induced structural damage in mitochondria in the dorsal root ganglia and peripheral nerves of mice [99]. In addition, Zheng et al. [115] showed that both taxanes and platinum-based chemoagents reduce mitochondrial functioning. These changes might be the cause of the nerve-ending retraction from the epidermis observed in animal and clinical studies of chemotherapy-induced neuropathic pain [99, 116, 117]. Indeed, treatment with pifithrin-μ, a small molecule that protects mitochondria by inhibiting mitochondrial p53 accumulation and thereby preventing activation of mitochondrial damage pathways [118], not only prevented the taxane-induced mitochondrial damage and neuropathic pain in mice, but also nerve fiber loss in the epidermis [99].

Our lab also observed that cisplatin-induced cognitive impairments are accompanied by structural changes in mitochondria and decreased mitochondrial function in brain synaptosomes in mice [98]. Unchallenged (basal) mitochondrial function was comparable to that seen in control mice, suggesting that only the mitochondrial capacity to increase energy production in more demanding situations was impaired. In addition, cisplatin decreased the number of neuronal progenitors in specific brain areas and reduced white-matter coherency, as has also been observed in humans via neuroimaging. The mitochondrial protectant pifithrin-μ prevented not only mitochondrial damage and cognitive impairments, but also cisplatin-induced changes in neuronal precursors and white-matter integrity, suggesting that these changes are downstream consequences of cisplatin-induced mitochondrial dysfunction. Lomelli reported similar findings for cisplatin-treated rats [97].

We propose that cancer therapy-induced mitochondrial damage leads to changes in brain structure and functional integrity, resulting in enhanced energy demands during cognitive challenges. However, this increased demand cannot be met by the mitochondria, leading to impaired performance on challenging cognitive tasks.