Tumor resection ameliorates tumor-induced suppression of neuroinflammatory and behavioral responses to an immune challenge in a cancer survivor model

Breast cancer survivors display altered inflammatory responses to immune challenges relative to cancer-naive controls likely due to previous cancer treatments, stress associated with cancer, and/or tumor physiology. Proper inflammatory responses are necessary for adaptive sickness behaviors (e.g., fatigue, anorexia, and fever) and neuroinflammatory pathways are also implicated in mental health disturbances (e.g., cognitive impairment, depression) suffered by cancer patients and survivors. Rodent cancer models indicate that tumors are sufficient to exacerbate neuroinflammatory responses after an immune challenge, however primary tumors are not usually present in cancer survivors, and the behavioral consequences of these brain changes remain understudied. Therefore, we tested the extent to which mammary tumor resection attenuates tumor-induced neuroinflammation and sickness behavior following an immune challenge (i.p. lipopolysaccharide [LPS] injection) in mice. Tnf-α, Il-1β, and Il-6 mRNA decreased in multiple brain regions of LPS-treated tumor-bearing mice relative to LPS-treated controls; tumor resection attenuated these effects in some cases (but not Tnf-α). Tumors also attenuated sickness behaviors (hypothermia and lethargy) compared to LPS-treated controls. Tumor resection reversed these behavioral consequences, although basal body temperature remained elevated, comparable to tumor-bearing mice. Thus, tumors significantly modulate neuroinflammatory pathways with functional consequences and tumor resection mitigates most, but not all, of these changes.

changes in rodent cancer models 4,8 , similar to neuroinflammatory "priming" observed in other health contexts (e.g., chronic stress) 17 . In these contexts, priming of brain immune cells can lower the threshold for subsequent activation of inflammatory pathways resulting in exaggerated or pathological neuroinflammation and negative behavioral consequences 18 .
In the cancer field, the majority of the clinical research testing this "two-hit" hypothesis has focused on endocrine responses to acute stressors 11 . However, more recent work examines immune responses. Studies using acute psychological stressors to probe peripheral cytokine or leukocyte responses among subsets of cancer survivors (fatigued, smokers) intimate that cytokine responsivity may be somewhat increased or delayed among cancer survivors 19,20 . In contrast, two studies in current cancer patients have used immune challenges and observed either no alterations in cytokine responses (after interferon-α administration 21 ) or elevated cytokine responses (after surgery 22 ). Thus, alterations in inflammatory responses relevant to behavior may vary between cancer patients and survivors.
As a first step in understanding how tumors may alter immune responses in the brain (as opposed to at rest), our previous study compared neuroinflammatory responses to an acute peripheral challenge between tumor-bearing and tumor-free rats. This study indicated that carcinogen-induced mammary tumors exacerbate neuroinflammatory responses to a peripheral infection mimetic (lipopolysaccharide, LPS) 4 . Corroborating evidence of select exacerbated neuroinflammatory markers and sickness behavior is observed in a murine head and neck tumor model following LPS administration 8 . To advance this work, the current study aimed to determine: (1) the extent to which these neuroinflammatory changes are attenuated by tumor resection in a novel mouse non-metastatic breast cancer survivor model, and (2) the potential functional consequences of altered neuroinflammation on sickness behaviors. Understanding the extent to which neuroimmune responsivity may be altered by tumor biology has potential relevance to the mental and physical health of cancer patients and survivors.

Results
Tumors suppress adaptive sickness behaviors. The baseline measures of body mass and food intake, taken 24 h prior to the LPS injection, were not different among all groups (p > 0.05; see Fig. 1 for experimental design and samples size). LPS treatment significantly reduced body mass, driven primarily by the data from 24 h post-LPS (LPS main effect: F 1,37 = 123.5, p < 0.0001; Fig. 2A). Tumor treatments did not affect changes in body mass at any time point (p > 0.05). LPS induced anorexia 4 h and 24 h post-LPS relative to PBS-treated controls (LPS main effect: F 1,37 = 16.2, p < 0.0005 and F 1,37 = 34.3, p < 0.0001, respectively; Fig. 2B). Time since LPS injection influenced food intake (time main effect: F 1,54 = 54.8, p < 0.0001), such that anorexia receded 24 h post-LPS relative to 4 h post-LPS. Within LPS-treated mice, tumor manipulations did not alter food intake at either 4 or 24 h post-injection (p > 0.05), although a trend (p = 0.09) towards attenuated anorexia was observed in tumor-bearing mice relative to tumor-free mice at 4 h post-LPS.
No baseline differences in locomotion were observed among groups (p > 0.05; Fig. 3A and B), although a trend (p = 0.06) towards increased locomotion was observed in tumor-resected mice relative to tumor-bearing mice (Fig. 3B). LPS significantly decreased locomotor activity throughout the 24 h after LPS injection for all groups compared to the surgical controls injected with PBS (LPS main effect: p < 0.0001 for tumor-free and  Fig. 3C). For body temperature before LPS treatment, repeated measures ANOVA revealed an effect of tumor treatment (F 2,27 = 4.3, p < 0.05): baseline body temperature was elevated in tumor-bearing mice relative to tumor-free mice and tumor-resected mice during both the light and dark phases (p < 0.0001 in all cases; Fig. 4A). Tumor-resection reversed the tumor-induced increase in basal body temperature relative to tumor-free mice only during the light phase; body temperature remained elevated in tumor-resected mice relative to tumor-free mice during the dark phase (tumor treatment effect: F 2,27 = 4.1, p = 0.005).
To control for differing baseline body temperature among tumor treatments, data after LPS were analyzed as a percent change from baseline (Fig. 4B). LPS injection induced percent changes in body temperature over time relative to PBS-treated controls (p < 0.0001 in all cases), causing fever during the light phase and hypothermia during the dark phase in tumor-free control mice. Of note, LPS injection acutely decreased body temperature for tumor-free and -resected groups during the first hour post injection ( Fig. 4B; LPS main effect: p < 0.0005 and p < 0.005 respectively) and tended to do the same for tumor-bearing mice (p = 0.06). However, the kinetics of body temperature over the following hours differed among the tumor treatments groups. Among LPS-treated mice, a repeated measures ANOVA revealed that tumor and time interacted (F 2,18 = 2.0, p < 0.05), such that tumors induced an increase in body temperature throughout the 24 h after LPS injection (tumor treatment effect: F 1,24 = 1.6, p < 0.05) relative to tumor-free mice and, in part, an attenuation of LPS-induced hypothermia in the dark phase (Fig. 4B). During the initial light phase (2-8 h post-LPS), hyperthermia peaked earlier (3 h post-LPS) in tumor-bearing mice than in surgical controls (5 h post-LPS). Furthermore, tumor resection attenuated this hyperthermic period relative to surgical controls (tumor treatment effect: F 2,120 = 2.1, p < 0.05). The LPS-induced changes in body temperature receded 17 h after injection (beginning of the light phase) in control and tumor-resected mice, but body temperature remained increased for tumor-bearing mice relative to tumor-free (tumor treatment effect: p < 0.005 for PBS and p < 0.0005 for LPS-treated) and -resected mice (tumor treatment effect: p < 0.005).
Tnf-α. LPS treatment significantly increased Tnf-α gene expression in the hippocampus and frontal cortex at 4 h post-injection for all tumor treatment groups relative to PBS controls (p < 0.05 in all cases); in the hypothalamus, Tnf-α was only increased in the surgical control group (U = 14.5, p < 0.005). At 24 h post-injection, Tnf-α remained elevated in the hippocampus of all LPS-treated groups relative to PBS-treated controls (t 53 = 9.0, p > 0.005), but Tnf-α in the hypothalamus was elevated only in the surgical control (U = 19, p < 0.05) and tumor-resect (U = 54, p < 0.05) groups. Tnf-α was not elevated in the frontal cortex (p > 0.05) at this time point.
The relationships among tumor mass and neuroinflammatory responses were investigated using correlational analyses for both the tumor-bearing and tumor-resected mice. At 4 h post-LPS, current tumor mass predicted microglial marker Cd11b gene expression in all three brain regions examined (p < 0.05 in each case; Supplementary Fig. 2A), such that larger tumors resulted in greater Cd11b gene expression. Among tumor-resected mice, prior tumor mass predicted the resolution of neuroinflammation 24 h post-LPS, such that greater prior tumor mass was negatively correlated with Tnf-α and Cd11b responses, akin to tumor-bearing responses at this time point. This negative correlation between tumor mass and neuroinflammation 24 h post-LPS was statistically significant for hippocampal and hypothalamic Tnf-α and hypothalamic Cd11b (p ≤ 0.05 in each case; Supplementary Fig. 2B). While Cd11b mRNA did not mirror cytokine gene expression responses to LPS, understanding the effects of tumor treatments on the activation or polarization of microglia 23 require further investigation with additional microglial markers. However, Cd11b expression in the brain during peak LPS response (4 h) was related to tumor burden, indicating that Cd11b may be influenced by the presence of a tumor during an inflammatory event. The correlations observed in the current study might potentially be associated with the inflammatory profile of the microglia/monocytes expressing Cd11b, since this marker does not decrease in tumor and tumor-resected mice, independently of the neuroinflammatory status and warrant future investigations. The correlations between all other genes and tumor mass were not statistically significant at 4 or 24 h post-injection (p > 0.05).

Discussion
In the present study, using rodent models of breast cancer patients and survivors, peripheral tumors consistently attenuated neuroinflammatory gene expression and associated sickness behavior in response to an LPS peripheral immune challenge. These remarkable effects of tumors were primarily reversed by tumor removal, although select neuroinflammatory responses to LPS and basal body temperature changes persisted. The latter findings extend our previous observations that tumor removal partially reverses the basal inflammatory and physiological consequences of tumors 5 . Although two prior studies report that peripheral tumors exacerbate peripheral and central inflammatory responses, weight loss, and lethargy to an LPS injection 4,8 , the present conflicting data indicate that tumor type, species, and strain significantly modulates these responses. Notably, the present findings were derived from four treatment-balanced experiments, validating the significant repeatability of the results. Taken together, the basic science literature regarding the extent to which peripheral tumors can induce changes in the activity of innate peripheral and central immune pathways to influence behavioral changes remains relatively scarce and conflicting. Of note, neutropenia and lymphopenia are common side effects of cancer treatment (e.g., chemotherapy) which have immunosuppressive consequences (as well as inflammation), but treatments were absent in the current study. Furthermore, tumors play a dual role in modulating immune function, depending on the arm of the immune system (innate/inflammation vs. adaptive) and the stage and type of tumor 24 . In the clinical and preclinical literature, tumor-induced behavioral changes are most often correlated with increases in pro-inflammatory cytokines, which we reviewed in 25 . Indeed, myeloid-derived suppressor cells (MDSCs) are increased in the primary tumor site, as well as in the spleen and blood in this model, which increase peripheral and central pro-inflammatory cytokines and decrease T-cell function 5 .
Despite the reduced neuroinflammatory cytokine mRNA responses of the tumor group, LPS injection induced anorexia and decreased body mass to the same extent for all treatments (surgical controls, tumor-bearing, tumor-resected). Similarly, previous research using different approaches to evaluate neuroinflammatory signaling demonstrates that suppressed neuroinflammation does not prevent LPS-induced decreases in body mass or food intake after an immune challenge 26 . One potential explanation may be the relatively insensitive tools used to measure these sickness behaviors. For example, manual food intake measurement may omit small pieces of chow that have fallen into the bedding. In addition, a recent study implicates a nontraditional neuroinflammatory pathway as a key mediator of LPS-induced anorexia and weight loss, as well as in pancreatic cancer cachexia 27 . Genetic deletion of the adaptor protein TIR-domain-containing adaptor inducing interferon-β (TRIF) decreases microglial activation and attenuates anorexia and weight loss. Thus, the canonical inflammatory players examined in this study may be less important for these particular sickness behaviors.
In contrast, tumors significantly modulated minute-by-minute thermoregulatory and locomotor activity responses to LPS in the present study. Indeed, tumors altered body temperature both before and after the acute immune challenge relative to tumor-free mice. The observed elevations in baseline body temperature of tumor-bearing mice are comparable to previous findings in rodent tumor models and cancer patients [28][29][30] , and may, in fact, reflect an underlying disruption in circadian rhythm circuitry or temperature homeostasis. Notably, tumor resection was not able to reverse the tumor-induced disruption in body temperature regulation under basal conditions during the dark (active) phase. Thus, mammary tumor biology may have lasting effects on homeostatic pathways which persist >2 weeks after complete tumor resection. Indeed, among breast cancer survivors, thermal discomfort symptoms, such as hot flashes and night sweats, are highly prevalent (65-85%) [31][32][33][34] , although cancer therapies also likely contribute to these thermoregulatory changes.
The LPS-induced thermoregulatory response in mice is multiphasic and involves febrile and hypothermic phases with distinct physiological mechanisms and adaptive value 35  exhibited a transient phase of fever, but failed to mount the hypothermic response observed in LPS-treated controls during the dark (active) phase of the daily light cycle. The observed suppressed neuroinflammatory response likely contributed to this impaired thermoregulatory response, as central proinflammatory cytokines (IL-1β, IL-6 and TNF-α) regulate LPS-induced fever and hypothermia 36,37 . For example, central administration of IL-6 restores the thermoregulatory response to LPS or IL-1β injection in IL-6-deficient mice 38 . Of note, our cancer-related findings are similar to the impaired sickness response (e.g., suppressed thermoregulatory response and decreased peripheral proinflammatory cytokines) observed in colitic rats after an LPS immune challenge 39 . Taken together, peripheral chronic inflammation, like tumors or colitis, may cause a compensatory inhibition of subsequent neuroinflammatory responses via humoral or neural inflammatory signaling pathways 39 . In tumor-resected mice, body temperature responses to LPS returned to control levels, demonstrating the causal role of the peripheral tumor. However, the hyperthermic responses to LPS remained partially impaired in the tumor-resected mice, but only during the light phase. Overall, these temporally-biased results suggest a potential disruption in circadian rhythmicity with tumor and tumor resection and warrant further investigation. The moderate disruption in thermal response of tumor-bearing mice was accompanied by markedly decreased hippocampal and hypothalamic Tnf-α mRNA levels 4 h after LPS injection relative to tumor-free mice. Indeed, in a cancer cachexia rodent model, anti-TNF-α treatment partially reverses tumor-induced decreases in body temperature 40 . However, this robust decrease in Tnf-α persisted in tumor-resected mice without obvious consequences for body temperature responses to LPS. Thus, TNF-α is unlikely to be the sole mediator of thermoregulatory differences in these models.
Indeed, convergent lines of evidence indicate that various other pathways are critical for fever and hypothermic responses in immune-challenged mice 35,39,41 . Specifically, neuroendocrine pathways, such as the hypothalamic-pituitary-adrenal (HPA) axis and neurotransmitter systems, modulate several aspects of both responses (reviewed by 42 ). Here, elevated circulating corticosterone concentrations were observed in tumor-bearing mice relative to tumor-free and -resected mice 24 h after immune challenge. This altered recovery in the corticosterone response to LPS may, therefore, be related to the coincident sickness behavioral changes observed in tumor-bearing mice. Indeed, chronic unpredictable stress in rats increases basal body temperature and hypothermic responses to a cold room challenge, relative to stress-free controls. Treatment with a corticosterone synthesis inhibitor resolves this increased cold challenge thermoregulatory response, although it does not attenuate the elevation in basal body temperature 41 .
As expected, LPS injection reduced locomotor activity in all treatments relative to the PBS-treated mice. However, the relatively attenuated magnitude of LPS-induced lethargy in tumor-bearing mice, compared to LPS-treated tumor-free and -resected mice, is consistent with their attenuated thermoregulatory and neuroinflammatory responses 43 . Indeed, a caspase-1 inhibitor centrally administered to rats, which significantly reduces the brain concentration of IL-1β (pre-frontal cortex, hypothalamus, and hippocampus), attenuates LPS-induced lethargy 44 . Our data contrast with previous work showing exacerbated deficits in locomotor activity in human papilloma virus (HPV)-related head and neck tumor-bearing mice 24 h after an LPS-induced challenge 8 . Although these discrepancies might reflect differences in the tumor model and in the methods used, the locomotor activity assessment in the current study was more sensitive and reliable (data collected every minute for 48 h) than the single 5-min open field test used to assess locomotion in the head and neck tumor model. Overall, more details about the mechanisms by which tumors alter thermoregulatory and other immune responses are warranted.
While prior studies suggest that peripheral tumors might prime the microglia and exacerbate neuroinflammatory responses to subsequent innate peripheral immune challenges 4,8 , our current data indicate that some peripheral tumor models, in fact, attenuate the LPS-induced gene expression of proinflammatory cytokines in the brain. Additionally, in this study, tumor mass positively predicted Cd11b microglial gene expression from brain tissue collected 4 h post-LPS in mice that retained their tumors, even though absolute values of this marker were not statistically different among tumor treatments at this time. This consistent correlation among various brain regions suggests a potential relationship between early microglial activation/number and tumor burden. Furthermore, former tumor mass of tumor-resected mice negatively predicted Tnf-α and Cd11b 24 h post-LPS, indicating that larger former tumor masses predicted more tumor-like (attenuated) later neuroinflammatory responses (24 h post-LPS injection). Therefore, tumor pathology in breast cancer survivors may help predict altered inflammatory signaling of activated pathways and associated behavior during survivorhood.
Indeed, the reversal of the majority of the tumor-induced behavioral and physiological responses to LPS by tumor resection suggests that cancer treatments and/or stress (and their interactions with tumor biology) may play key roles in the lasting behavioral consequences of cancer associated with inflammation 45 . Systematic investigation into these interactions is necessary. All mice received some type of surgery two weeks prior to the LPS challenge, indicating that wound healing or anesthetic exposure do not explain differences among treatment groups. Tumor-resected mice and sham controls received an additional earlier surgical experience (sham or tumor inoculation). However, the instances in which tumor-resected mice were more comparable to tumor-bearing mice than controls (Tnf-α and basal body temperature) indicate that these changes were likely due to tumors and not the number of prior surgical experiences.
In addition, the tumor model used in the current study is non-metastatic and did not induce overt cachexia or sickness behaviors during the tumor development, which could have confounded the evaluation of the sickness response to the immune challenge. Finally, besides the estrous cycle was not tracked in the present study, recent evidence indicates that male and female rodents exhibit comparable variability in a broad range of physiological and behavioral traits 46,47 , regardless of the estrous phase. Taken together, our findings support the hypothesis that tumors decrease the amplitude of the baseline body temperature diurnal rhythm, as well as attenuate neuroinflammation and associated sickness behavioral responses to subsequent innate immune challenges, and that tumor resection reverses most, but not all of these responses. Thus, these data suggest that cancer patients and survivors may be predisposed to some impairment of host defense responses against secondary diseases or challenges (e.g., surgery, infection, stressors). Although several studies corroborate the hypothesis that peripheral solid tumors are sufficient to induce baseline behavioral and neuroinflammatory changes in rodent tumor models (reviewed by 25 ), very little research had focused on how cancer may affect these pathways upon activation. Sickness behaviors in response to an immune challenge are an important strategy for the survival of infectious diseases, and both exacerbated or decreased neuroinflammatory responses may be maladaptive [48][49][50] . Indeed, inflammatory pathways are also related to psychiatric symptoms in cancer patients 25 , and therefore, alterations in pathway responses may have implications for behavioral comorbidities beyond sickness behavior. A better understanding of the mechanisms underlying these complex tumor-induced changes may provide insight into therapeutic targets and alternative treatments that allow for better health outcomes and quality-of-life for cancer survivors.

Methods
Animals. Nulliparous female 8-to 9-week old Balb/c mice (Charles River, Wilmington, MA, USA; see Fig. 1 for samples size) were single-housed and acclimated to the temperature-controlled ( Tumor Induction. Prior to and after tumor induction, mice were acclimated to handling twice/week (n = 108). Under anesthetization (isoflurane vapors), a 5 mm incision was made medial to the 4 th nipple, 5 × 10 6 cells (in matrigel) were injected into the 4 th mammary fat pad, and incisions were closed with a wound clip 5 . This procedure results in an in situ primary mammary carcinoma 53 that does not metastasize 54 . This is a validated orthotopic and syngeneic breast cancer model, eliminating the need to use immunocompromised mice 5 . Mice assigned to the tumor resection group (n = 53) were inoculated with tumor cells 2 weeks prior to the other groups, so that the timing of their tumor resection surgeries corresponded to the time of tumor/control induction surgeries in the other groups. This allowed for simultaneous behavioral and physiological assessments, as well as the same interval between surgery and behavioral testing for all groups. Ear notches were made at tumor inoculation for individual identification purposes. Body mass was measured twice/week. Tumor induction was unsuccessful in 9 mice and they were removed from the analyses.
Tumor resection. A modified radical mastectomy procedure was used to completely remove primary tumors in the "survivor" group 5 . These mice were anesthetized (isoflurane), and tumors with intact membranes were surgically removed along with mammary tissue, fat, and lymph nodes. Skin was closed with wound clips. Buprenorphine (0.05 mg/kg; s.c.) was administered immediately after surgery and again 12 h later. Complete tumor resections were verified at necropsy, and any mice with recurring tumors (n = 8) were removed from the study. The surgical controls were given a sham resection surgery at this time.

Telemetry.
A subset of mice was singly-housed for a week and randomly assigned to the following experi- Sickness Behavior. Body mass, food intake, locomotion, and body temperature (the latter two from the implanted transmitters) were recorded for a subset of mice at 4 and/or 24 h after LPS injection. The baselines for these parameters were measured 24 h prior to injection.
Tissue collection. Following rapid decapitation, blood was collected through heparin-lined natelson blood tubes, kept on ice and then centrifuged (20 min at 2,000 rpm), and plasma was stored at −80 °C until assayed for cytokine and corticosterone quantification. Brain regions relevant to sickness behaviors (hippocampus, hypothalamus, frontal cortex) were dissected out and frozen in RNA Later preservative for later neuroinflammatory gene expression assessment. Tumors and spleens were removed aseptically and weighed.
Plasma corticosterone concentrations. Glucocorticoids are potent innate anti-inflammatory agents and are produced in response to a peripheral LPS injection 59 . Corticosterone was measured in duplicate in plasma samples via EIA according to the manufacturer's instructions (Enzo Life Sciences, Plymouth Meeting, PA, USA) after 1:40 dilution. The threshold of detection was 27 pg/ml. Intrassay and interassay variations were each <10%.
Quantitative RT-PCR. Total RNA was extracted from the brain hippocampus, hypothalamus, and frontal cortex using Qiagen RNeasy mini kits (Valencia, CA, USA). RNA concentrations were measured and 260/280 ratios were determined to be 1. Statistical analysis. Statistical comparisons of body mass, food intake, tissue mass, gene expression, circulating cytokines and corticosterone were analyzed using 2-way ANOVAs followed by post-hoc Fisher's LSD or Student's t-tests with Statview version 5.0.1 software (Scientific Computing, Cary, NC, USA) when variance was normal. Nonparametric Mann-Whitney U tests were used when variance was not normally distributed. Repeated measures ANOVAs were used to compare changes in body temperature and activity responses over time. Pearson's correlations were used to assess the relationships among variables. Due to differences in the baseline measures for body temperature between the surgical control and tumor/tumor-resect groups, the influence of LPS injection on temperature was calculated as a percent change relative to the baseline recordings (i.e., calculated within individual relative to the 24 h prior to the LPS injection). Data were determined to be statistically significant when p ≤ 0.05 and are presented as mean ± standard error of the mean (SEM).