Paclitaxel Chemotherapy Induces Long-Term Memory Impairment and Neuroinammation in a Mouse Model of Breast Cancer Survivorship

Cancer-related cognitive impairment (CRCI) has been reported in cancer survivors 20 years or more after cancer treatment, and has been associated with sustained increases in circulating inammatory biomarkers. One of the major risk factors for CRCI is chemotherapy, and preclinical studies typically examine the impact of chemotherapy in cancer naïve mice to evaluate potential mechanisms However, clinical evaluation of the long-term effects of chemotherapy cannot avoid the potential cumulative impact of preceding factors on the brain including the cancer itself and cancer surgery. Methods To evaluate the cumulative impact of cancer-related factors on cognitive impairment and hippocampal cytokine expression, we evaluated the effect of paclitaxel chemotherapy vs. placebo on a background of 67NR mammary carcinoma and surgical resection of the primary tumour in mice. Memory was assessed using the Y maze test and novel object/novel place recognition test. Changes in hippocampal proinammatory and anti-inammatory cytokines, microglia and neuron markers were assessed using qRT-PCR. Paclitaxel prolonged spatial memory impairment survivor mice with the sustained changes in hippocampal-dependent memory, we assessed neuroinammation and neuronal activity-related gene expression in the hippocampus of paclitaxel and placebo treated mice, untreated 8-week-old (pre-cancer baseline age), and untreated age-matched 19-week-old control groups (baseline for experimental end point). All the groups were normalized to the 8-week-old control group for general comparison. qRT-PCR was used to assess gene expression changes in 100 ng RNA with iScript™ One-Step RT-PCR Kit for Probes (Bio-Rad, USA). Taqman probes targeted Actb, Il1b, Il6, Il10, Tnfa, Iba1/Aif1, Bdnf, and Rbfox3/NeuN. The RNA and probe mix was subjected to 40 PCR amplication cycles. Each sample was run in triplicate and the genes of interest were normalized to the housekeeper gene β-actin (Actb) via the 2 − ∆∆ Ct method. Details provided in Supplementary Material.


Abstract Background
Cancer-related cognitive impairment (CRCI) has been reported in cancer survivors 20 years or more after cancer treatment, and has been associated with sustained increases in circulating in ammatory biomarkers. One of the major risk factors for CRCI is chemotherapy, and preclinical studies typically examine the impact of chemotherapy in cancer naïve mice to evaluate potential mechanisms However, clinical evaluation of the long-term effects of chemotherapy cannot avoid the potential cumulative impact of preceding factors on the brain including the cancer itself and cancer surgery.

Methods
To evaluate the cumulative impact of cancer-related factors on cognitive impairment and hippocampal cytokine expression, we evaluated the effect of paclitaxel chemotherapy vs. placebo on a background of 67NR mammary carcinoma and surgical resection of the primary tumour in mice. Memory was assessed using the Y maze test and novel object/novel place recognition test. Changes in hippocampal proin ammatory and anti-in ammatory cytokines, microglia and neuron markers were assessed using qRT-PCR.

Results
Cancer and cancer surgery was su cient to induce long-term memory impairment and sustained increases in hippocampal pro-in ammatory cytokines. Paclitaxel prolonged spatial memory impairment in the Y maze test and exacerbated hippocampal Il6 and Tnfa mRNA expression compared with placebo treatment.

Conclusions
These ndings suggest that cancer and cancer surgery can sensitise the brain to an exaggerated neuroin ammatory response to chemotherapy, and may contribute to sustained chemotherapy-induced cognitive impairment observed in cancer survivors.

Background
As the survival rate of cancer improves, the long-term impact of cancer and cancer treatment on breast cancer survivors becomes a signi cant public health concern. After cessation of cancer treatment, up to 50% of breast cancer survivors have long-term cognitive impairment in verbal short-term memory, attention, concentration, psychomotor function, information processing speed, or executive function (1)(2)(3)(4)(5). These cognitive de cits have been associated with altered brain activity and structural damage to brain areas including white matter tracts, and reduced hippocampal volume (2,3,(6)(7)(8), and with changes in systemic and brain glucose metabolism (9)(10)(11)(12).
Treatments including chemotherapy, hormonal therapy and radiation are well known to contribute to these cognitive problems. It is now also accepted that factors that precede adjuvant treatment -such as the cancer itself and cancer surgery -may also contribute to cognitive impairment in cancer patients by inducing in ammation in the body and brain (4,(13)(14)(15). While clinical evaluation of the long-term effects of chemotherapy cannot avoid the potential cumulative impact of these preceding factors on the brain, most preclinical studies to date have investigated the impact of chemotherapy on the brain and behaviour in the absence of other factors (16,17). That approach enables greater precision in isolating the mechanisms of cognitive impairment that occur from chemotherapy. The resulting studies in nontumour bearing mice have demonstrated that chemotherapy enhances (neuro)in ammation and, neural damage, reduces neurogenesis, activates physiological stress pathways, and contributes to cognitive impairment and behavioural changes (18-20). However, evaluating the effect of chemotherapy on cognitive impairment in isolation ignores the contribution of cancer and other treatments such as surgery, which may plausibly compound the effects of adjuvants therapies like chemotherapy on the brain. The contention that multiple events during the cancer journey can induce a synergistic or cumulative impact on the brain is supported by Winocur et al. who demonstrated worse memory performance in cancerbearing mice treated with chemotherapy (methotrexate and 5-uorouracil) compared to either cancer or chemotherapy alone (21). The sustained impact of chemotherapy in cancer free survivors was not evaluated in that study due to the aggressive nature of the metastatic transgenic model of breast cancer (MMTVneu FVB) used [19]. Therefore, to address the long-term impact of paclitaxel chemotherapy on memory in a clinically relevant setting, here we used a less aggressive murine model of mammary carcinoma, and evaluated the effect of adjuvant chemotherapy after primary tumour resection.

Materials And Methods
Animals BALB/c female mice (6-8 weeks old) (Monash University, Australia) were single caged in a temperature and humidity-controlled environment with a 12/12-h reverse dark-light cycle (lights on at 19:00). Food and water were available ad libitum. All procedures involving mice were carried out under protocols approved by the Monash University Animal Ethics Committee (protocol number MIPS2015.01 and 2015.16). Animals were monitored daily for distress. Humane endpoints based on body weight loss, body condition scoring and metastatic progression were in place. Mice were euthanized with CO 2 .

Breast cancer model
To generate syngenic mammary tumours, 1 x 10 6 67NR cells in 20 µL PBS were injected into the 4th left mammary gland of anaesthetized BALB/c mice. Once primary tumours were palpable, tumour growth was monitored by caliper measurement and tumour volume determined using the following formula: (length x width 2 )/2. Cell culture details are provided in Supplementary Material.

Breast cancer surgery
The primary tumour, left inguinal mammary fatpad and adjacent left inguinal lymph node were resected under sterile conditions and 3% iso urane. The wound site was then sutured. Buprenorphine (1 µg/100 µL, s.c.) was administered prior to surgery and every 12 h after surgery for 3 doses in total. All mice underwent the same anesthesia procedure and received buprenorphine. Body weight and burrowing activity and body condition scoring were recorded daily to assess sickness behaviour and activity in response to surgery.

Paclitaxel treatment
Mice were randomly allocated to placebo or paclitaxel treatment groups while ensuring equal distribution of mice to each group based on performance in a NOPR and Y maze test after tumour cell injection and primary tumour burden (Fig S1A-C). Mice were injected intraperitoneally with 10 mg/kg paclitaxel (Assay Matrix and Sigma-Aldrich, USA) solubilized in vehicle (Cremorphor EL:ethanol:saline, 1:1:10) or an equal volume of placebo every second day for 2 weeks as in (22). Within the 13 mice of the placebo group, to con rm there was no side effect of Cremorphor EL on behaviour (23), 8 mice were administered saline and the other 5 mice were administered vehicle (Cremorphor EL:ethanol:saline, 1:1:10). No signi cant differences were observed between saline and vehicle treated mice for any measure, therefore these were collapsed and treated as a single placebo group.

Memory and sickness behaviour assays
Attention and episodic memory were tested using a novel object/novel place recognition (NOPR) test as previously described (22). Spatial memory was tested using the Y maze. Behaviour was video recorded and analyzed using tracking software (Viewer 3 , Biobserve, Germany). Locomotor activity was assessed as total movement (cm) in both tests. Daily body weights and home cage burrowing activity of mice were recorded to assess sickness behaviour and activity. Details are available in Supplementary Material.

RNA extraction and quantitative reverse transcriptase polymerase chain reaction (qRT-PCR)
To explore the association between in ammation in our cancer-free survivor mice with the sustained changes in hippocampal-dependent memory, we assessed neuroin ammation and neuronal activityrelated gene expression in the hippocampus of paclitaxel and placebo treated mice, untreated 8-week-old (pre-cancer baseline age), and untreated age-matched 19-week-old control groups (baseline for experimental end point). All the groups were normalized to the 8-week-old control group for general comparison. qRT-PCR was used to assess gene expression changes in 100 ng RNA with iScript™ One-Step RT-PCR Kit for Probes (Bio-Rad, USA). Taqman probes targeted Actb, Il1b, Il6, Il10, Tnfa, Iba1/Aif1, Bdnf, and Rbfox3/NeuN. The RNA and probe mix was subjected to 40 PCR ampli cation cycles. Each sample was run in triplicate and the genes of interest were normalized to the housekeeper gene β-actin (Actb) via the 2 −∆∆Ct method. Details provided in Supplementary Material.

Statistical analysis
Paired student's t-tests were used to assess tumour-induced memory performance from baseline in the Y-Maze and NOPR test. Repeated two-way analyses of variance (ANOVAs) were used to analyze behaviour in the NOPR test and Y maze, locomotor activity, burrowing activity, and body weight for paclitaxel-vs placebo-treated groups. Unpaired Student's t-test or between subjects one-way ANOVAs were used to analyze data from cytokine expression assays and spleen mass. For within group planned comparisons between baseline and other time points, Dunnett's multiple comparison test was used. For between group planned comparisons independent t-tests were used with α adjusted based on number of comparisons. Statistical signi cance was set at α less than or equal to 0.05.

Results
Paclitaxel exacerbates cancer-and cancer surgery-induced Tnfa and Il6 gene expression in the hippocampus To investigate the effect of cancer and surgery on the brain, mice with 67NR tumours underwent surgical resection and we investigated the long-term effect on cytokine expression in the brain (Fig. 1A). 58 days after tumour resection, mice that previously had tumours showed a signi cant increase in expression of hippocampal pro-in ammatory cytokines including Il1b (F (3,26) = 16.15, p < 0.0001), Tnfa (F (3,26) = 66.70, p < 0.0001), Il6 (F (3,26) = 85.31, p < 0.0001), and anti-in ammatory Il10 (F (3,26) = 7.29, p < 0.001) compared to cancer naïve mice. Paclitaxel treatment after tumour resection signi cantly exacerbated expression of Tnfa and Il6 (p < 0.05 for both) (Fig. 1B). To determine if increases in hippocampal cytokines were associated with changes in hippocampal resident microglia/macrophages or changes in the neuron population, we assessed expression of Iba1, Rfox3 (NeuN) and brain derived nerve growth factor (Bdnf). Mice that had previously had tumours resected had increased Iba1 expression, suggesting elevated numbers or activation of brain microglia/macrophages (F (3,26) = 47.40, p < 0.0001) (Fig. 1C). Rfox3 (NeuN) is a gene expressed in neural tissue and was signi cantly increased in mice that previously had tumours compared to age-mated untreated controls (F (3,26) = 42.32, p < 0.0001) such that they had equivalent expression to 8 week old untreated control mice (p > 0.05) (Fig. 1D). Increased hippocampal Bdnf mRNA expression was also observed (F (3,26) = 47.40, p < 0.0001) (Fig. 1E), which could be an early response to behaviour tests that was performed 24 h prior to sample collection and is consistent with previous studies (24). Paclitaxel did not signi cantly impact Iba1, Rfox3 or Bdnf.
To con rm that these changes were not due to aging given the length of the experiment, we assessed hippocampal gene expression for all genes in 8 week old and 19 week old untreated mice coinciding with the ages of mice at the start and end of the experiment. No age-related changes hippocampal mRNA expression was observed for cytokine, Iba1 or bdnf expression. Older mice showed a decrease in Rfox3 (NeuN) (F (3,26) = 42.32, p < 0.0001), which has been reported for the protein upon aging (25).
To determine if hippocampal in ammation was associated with peripheral in ammation, we assessed spleen weight -a measure of myeloid expansion and in ammation in 67NR tumour bearing mice (26- 29). No signi cant differences were observed in relation to spleen weight by 37 days after treatment with paclitaxel (58 days after 67NR primary tumour resection) (Fig. 1F).

Cancer-related memory impairment is and prolonged by paclitaxel treatment
We examined whether paclitaxel exacerbated or prolonged memory impairment in mice that previously had cancer. Consistent with previous ndings (28), prior to resection mice with tumours showed memory impairment in the Y-maze test (t (26) = 3.28, p < 0.01) and NOPR test (t (26) = 3.60, p < 0.01) (Fig S1A and B).
To assess the impact of treatment with paclitaxel on memory, mice were tested in the NOPR test and Y maze for 6 weeks starting 24 h after the end of paclitaxel or placebo treatment. Performance in the Y maze test changed over time (F (3.774,87.56) = 3.31, p = 0.016) ( Fig. 2A). Multiple comparisons revealed that memory was signi cantly worse than pre-cancer memory only 2 days after treatment for mice placebotreated mice (p < 0.05). However, memory was signi cantly worse than pre-cancer memory at 48 h (p < 0.05) 9 days (p < 0.01) and 37 days (p < 0.05 ) after cessation of treatment with paclitaxel. Paclitaxeltreated mice performed worse than placebo-treated controls at each time point following baseline, which reached signi cance at 37 days after treatment cessation (t (23) = 2.60, p = 0.05) ( Fig. 2A). Performance in the in the NOPR test also changed over time (F (4.90, 118.4) = 4.63, p = 0.00007) (Fig. 2B). Memory was worse compared to the pre-cancer baseline at all time points regardless of paclitaxel or placebo treatment (p < 0.001). Paclitaxel did not signi cantly affect memory in the NOPR test compared to placebo-treated mice (p > 0.05) (Fig. 2B).
To con rm that memory impairment did not re ect sickness, cancer-induced fatigue or inactivity, we assessed sickness responses and burrowing activity. While surgery induced a temporary reduction in body weight due to both the removal of the tumour mass and a transient impact of surgery, all mice increased in body weight over the course of the experiment and there were no differences between mice treated with paclitaxel versus placebo (Fig. 3A). Paclitaxel did not signi cantly affect locomotor activity in the Y maze or NOPR test ( Fig. 3B and C). Burrowing activity declined from baseline until the end of the experiment (F (5.08, 144.4) = 4.48, p = 0.0004). Treatment with paclitaxel also signi cantly affected burrowing activity, vs. placebo (F (1,29) = 9.62, p = 0.004). Planned comparisons revealed that the placebo treated mice burrowed signi cantly less than paclitaxel treated mice in response to tumours and between day 8 and day 29 after chemotherapy or placebo treatment (all p < 0.05) (Fig. 3D).

Discussion
Paclitaxel is a chemotherapeutic agent that is commonly used to treat diverse solid tumours and has been associated with cognitive decline (30). However, the effect of paclitaxel in the context of cancer and cancer resection surgery has not been described. To determine if there is a cumulative long-term impact on cognition, we developed a mouse model of cancer survivorship to evaluate long-term memory in mice after the resection of a cancer and treatment with adjuvant chemotherapy. Evaluation of the model demonstrated a sustained impact of cancer, surgery and adjuvant paclitaxel on the brain. 37 days after cessation of chemotherapy, mice showed memory impairment. This was found in tests that examine spatial memory (Y Maze, NOPR) and recognition memory (NOPR). Previous studies by us and others show that memory and in ammation is impacted by each of these single events: having a cancer (21,28,(31)(32)(33)(34), cancer surgery (35,36), and treatment with paclitaxel (18-20, 22). The ndings of the current study extend these observations by demonstrating a cumulative effect of cancer and chemotherapy on memory and brain in ammation that results in long-term impairment.
Using this model of paclitaxel chemotherapy in cancer naïve mice, we previously observed no change in brain cytokine mRNA or protein (22). The ndings here indicate that cancer and surgery sensitise the brain to chemotherapy-induced neuroin ammation, which persists long after chemotherapy treatment has ended. The discovery that paclitaxel exacerbated Tnfa and Il6 mRNA expression in the hippocampus, supports the contention of a synergistic effect of cancer and chemotherapy on brain in ammation. These ndings are supported by previous research. Consistent with this, using a mouse model of mammary carcinoma, Pyter et al. (2014) demonstrated that peripheral tumours can prime or sensitise the brain to in ammation which resulted in an exacerbated in ammatory response to lipopolysaccharide (32). Here, we con rm the capacity for tumours and cancer surgery to prime the brain and extend these ndings to a clinically relevant context by demonstrating that a peripheral tumour and cancer surgery can prime the brain for an exaggerated neuroin ammatory response to chemotherapeutic agents such as paclitaxel long after the tumour has been resected. To control for the impact of tumour burden on in ammation, in this study we ensured uniform tumour sizes between paclitaxel and placebo treatment groups, and resected all tumours at the same time. Future studies should examine whether cancer stage or tumour size in uences the magnitude of brain sensitization to chemotherapy.
The ndings suggest that brain-resident macrophages may contribute to neural in ammation in longterm cancer survivor mice. While paclitaxel speci cally exacerbated Tnfa and Il6 mRNA, all cytokines assessed were elevated in mice that previously had tumours, and this was sustained 58 days after tumour resection compared with cancer-naïve mice. Iba1 was similarly increased following cancer and cancer treatment, suggesting increased proliferation or activation of macrophages in the hippocampus consistent with ndings demonstrating a synergistic effect of mastectomy and cancer in another preclinical breast cancer model (35). Spleen weights did not differ by experimental endpoint suggesting that peripheral myeloid cell expansion and in ammation caused by 67NR tumours may have subsided following primary tumour resection. Comparison between 19 week and 8 week old cancer naïve mice indicated that age did not contribute to neuroin ammation. Taken together, these ndings suggest that brain resident macrophages remain activated and may drive sustained elevations in hippocampal cytokines thus creating a new "baseline" level of brain cytokines.
In future studies it will be important to understand the role of hippocampal TNFa and IL6 in cancerrelated CI (Fig. 1). Both TNF-α and IL-6 or their receptors have been reported to still be elevated in the blood of breast cancer survivors 1-12 years after chemotherapy, and are associated with reduced hippocampal volume and cognitive impairment (2). Those clinical ndings suggest that the sustained increase in circulating TNF-α and IL-6 in breast cancer survivors may serve as biomarkers for elevated hippocampal in ammation long after cancer treatment. TNF-α and IL-6 play important roles in hippocampal-dependent memory by impeding long-term potentiation (LTP). IL-6 can inhibit LTP by reducing MAPK/ERK through prolonged activation of STAT3 (37,38). High levels of TNF-α can affect neural transmission by causing an imbalance between excitatory and inhibitory transmission by enhancing PI3K-dependent AMPA tra cking to increase Ca 2+ in ux and decreasing incorporation of GABA A receptor to reduce inhibitory transmission. This can lead to apoptosis of hippocampal neurons (20,(38)(39)(40). In the current study neither of the neuronal and neurogenesis related genes Rfox3 or Bdnf were reduced in paclitaxel treated mice, suggesting that the elevations in Tnfa and Il6 may not be severe enough to cause long-term neuronal death. However, future studies should con rm these mRNA ndings using protein homogenate and immunohistochemistry techniques. Nevertheless, these ndings support the evaluation of TNF-α and IL-6 neutralizing antibodies as potential therapeutics for cancer survivors who suffer paclitaxel chemotherapy-induced cognitive impairment. In support of this contention, treatment with the TNF-α inhibitor thalidomide during paclitaxel treatment has been shown to prevent spatial memory impairment and neuron apoptosis in the hippocampus (20).
In contrast to clinical studies, which use cognitive function after cancer diagnosis or cancer surgery as the baseline to determine dysfunction induced by chemotherapy, here, we compared changes to a true pre-cancer baseline, in addition to between-subjects analysis of paclitaxel treated mice with placebotreated controls. Paclitaxel treatment prolonged spatial memory impairment measured by the Y maze test whereas placebo treated mice eventually returned to baseline function. However, regardless of chemotherapy treatment, cancer and cancer treatment caused memory impairment in the Y maze test for several weeks and impairment in the NOPR test was sustained in both paclitaxel and placebo treated mice. In conjunction with the widespread hippocampal in ammatory response caused by tumours, these memory ndings suggest that cancer and cancer surgery contribute to the profound long-term changes in brain function in addition to chemotherapy. However, the ndings of this study may be confounded by additional risk factors including 3 months of ageing across the course of the experiment, and the stress of multiple injections. For instance, studies have shown that prior stresses amplify subsequent in ammatory responses: mice exhibited higher in ammatory responses to stress (induced by injection of saline or tail suspension) 14 days after primary tumour resection compared with their control groups (35,36). Nevertheless, the ndings of this study suggest that longitudinal clinical studies that assess cognitive changes in cancer patients or survivors after adjuvant chemotherapy compared to a prechemotherapy baseline, may fail to observe quanti able cognitive decline relative to cognition at cancer diagnosis. This bias may lead some studies to underestimate or underdiagnose cognitive impairment in cancer survivors after adjuvant chemotherapy.

Conclusion
This study demonstrates long-term cognitive impairment and a cumulative impact of cancer treatment and the cancer itself on the brain in a mouse model of cancer survivorship. For the rst time, we demonstrate that tumour-induced neuroin ammation and memory impairment persists for months in mice, even after tumours are resected, and that chemotherapy can exacerbate tumour and surgeryinduced neuroin ammation and prolong memory impairment.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests. Authors' contributions NC, EKS and AKW were involved in all aspects of this study including conception, design, data acquisition, analysis, interpretation, and manuscript preparation. AC was involved in data acquisition, interpretation, and manuscript preparation. RDG was involved in data acquisition and manuscript preparation. All authors have approved this submission and take responsibility for the accuracy and integrity of the manuscript.   Between group comparisons revealed no signi cant differences between paclitaxel and placebo treated groups but both groups signi cantly increased in body weight over time (p < 0.0001) (A), and in NOPR test locomotor activity (p < 0.001) (C). Burrowing activity declined in both groups over the course of the experiment (p < 0.001) (D).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.