Behavioural Effects of Using Sulfasalazine to Inhibit Glutamate Released by Cancer Cells: A Novel target for Cancer-Induced Depression

Despite the lack of robust evidence of effectiveness, current treatment options for cancer-induced depression (CID) are limited to those developed for non-cancer related depression. Here, anhedonia-like and coping behaviours were assessed in female BALB/c mice inoculated with 4T1 mammary carcinoma cells. The behavioural effects of orally administered sulfasalazine (SSZ), a system xc− inhibitor, were compared with fluoxetine (FLX). FLX and SSZ prevented the development of anhedonia-like behaviour on the sucrose preference test (SPT) and passive coping behaviour on the forced swim test (FST). The SSZ metabolites 5-aminosalicylic acid (5-ASA) and sulfapyridine (SP) exerted an effect on the SPT but not on the FST. Although 5-ASA is a known anti-inflammatory agent, neither treatment with SSZ nor 5-ASA/SP prevented tumour-induced increases in serum levels of interleukin-1β (IL-1β) and IL-6, which are indicated in depressive disorders. Thus, the observed antidepressant-like effect of SSZ may primarily be attributable to the intact form of the drug, which inhibits system xc−. This study represents the first attempt at targeting cancer cells as a therapeutic strategy for CID, rather than targeting downstream effects of tumour burden on the central nervous system. In doing so, we have also begun to characterize the molecular pathways of CID.


Glutamate release through system x c
− . Glutamate release through system x c − was quantified using the cellular uptake of radiolabeled 14 C-cystine. In cancer cells, cystine uptake is coupled with glutamate release mediated by a 1:1 exchange via system x c −31 . Therefore, measuring the uptake of radiolabeled 14 C-cystine provides a proportional measure of glutamate release. Furthermore, the specificity of this assay in assessing system x c − activity is supported by xCT knockdown results. An approximate 2-fold knockdown of xCT mRNA expression in human breast cancer cells (MDA-MB-231) causes an approximate 2-fold decrease in 14 C-cystine uptake (unpublished data).
The cystine uptake protocol was adapted from previous reports 32,33 . Briefly, 250,000 cells were seeded in 6-well plates. All wells were incubated with 14 C-cystine, either in the presence or absence of 200 μ M of SSZ, for 30 minutes. Cells were then lysed, and the lysate was analyzed using a scintillation counter to quantify radioactivity. The lysate was also used to quantify the amount of protein for each well, and cystine uptake was normalized to total protein. Assay procedures are further detailed in "Supplementary Materials and Methods".
Mice. Fifty-nine female BALB/c mice aged 4-6 weeks were obtained from Charles River Laboratories (St. Constant, QC, Canada). Mice were single-housed in cages maintained at 24 °C with a 12-h light/dark cycle, and were provided ad libitum access to food and water. Single housing eliminates the protective effect of group housing on depressive-like behaviours observed in tumour-bearing mice 34  Tumour cell inoculation. Tumour cell inoculations were performed as previously described 8 . Briefly, mice were inoculated with 15,000 4T1 cells in sterile phosphate buffered saline (PBS) subcutaneously above their right flank. Control mice were sham-inoculated with sterile PBS. Mice were weighed weekly (Supplementary Materials and Methods, Figure S1) and tumour size was monitored using a digital caliper every 3-4 days once tumours became palpable at day 8 (Supplementary Materials and Methods, Figure S2). All mice were euthanized 28-29 days after cancer cell inoculation.
Drug treatments. All drugs were administered ad libitum to the mice in place of their normal drinking water. This method of drug treatment was chosen to minimize repeated stressful injections and excessive handling, which might confound the behavioural results. Drug treatments were initiated on the day of tumour cell inoculation and continued until endpoint.
FLX hydrochloride (Sigma-Aldrich) was dissolved in drinking water at a concentration of 100 μ g/mL to obtain a target mean drug dose of approximate 20 mg/kg/day, which is the highest chronic dose reported to be effective in mice 35,36 . SSZ was dissolved in a small volume of 1 M NH 4 OH, which was then adjusted to pH 8.5 using 1 M HCl. The concentrated SSZ solution was then diluted 1:30 in drinking water to a final concentration of 0.3 mg/ Scientific RepoRts | 7:41382 | DOI: 10.1038/srep41382 mL. In humans, only approximately 12% of orally ingested SSZ escapes colonic cleavage and is absorbed into systemic circulation 29 . Based on these bioavailability data, mean water consumption, and mean weight of mice, it was estimated that 0.3 mg/mL of SSZ would translate into a dose of approximately 8 mg/kg/day. This target dose was chosen to be comparable to a dose that we have previously demonstrated to effectively reduce cancer-induced bone pain 37 . The SSZ metabolites, 5-ASA and SP (Sigma-Aldrich), were prepared the same way as SSZ to a final concentration of 0.09 mg/mL for 5-ASA and 0.15 mg/mL for SP. These target concentrations were chosen to match the expected concentrations of metabolites produced from oral treatment with 0.3 mg/mL of SSZ. As such, a liberal estimation of 80% colonic cleavage of SSZ into equal parts 5-ASA and SP was assumed 28 and concentrations were adjusted based on the molecular weights of each metabolite.
FLX and SSZ are known to have an aversive taste. Therefore, all vehicles included 2% orange extract (McCormick Canada Inc., London, ON, Canada) to increase the palatability of drug solutions. Sweetening the water was avoided in order to maintain sucrose preference test (SPT) sensitivity at post-treatment measurement. All drugs were administered in opaque bottles for light protection.
Behavioural analyses. Behavioural tests were performed as previously described 8 . Anhedonia was evaluated using the sucrose preference test (SPT) and active versus passive coping behaviour was evaluated using the forced swim test (FST) and the tail suspension test (TST). For the SPT, mice were first habituated to sucrose solution and a 2-bottle setup in their home cages for 72 h. For baseline and post-treatment testing, mice were presented with a two-bottle option of water and 3% sucrose solution for 48 h. At 24 h, the positions of the two bottles were switched to reduce location bias. Bottles were weighed before and after testing, and preference was calculated as the percentage of sucrose solution intake relative to total fluid intake 38 .
The FST and TST were performed using automated systems (BioSeb, Vitrolles, France). For the FST, mice were individually placed into beakers of water. For the TST, mice were suspended by their tails, and strain gauge input was used to assess mobility. For both tests, the total duration was 6 minutes, with the first minute of the test discounted from analysis to allow for behavioural stabilization 39,40 . Immobility time was the primary outcome measure, and power of movement (PM) was reported as a secondary measure for the TST. Although this study utilizes female mice, it has been demonstrated that estrous cycle does not impact sucrose preference or coping behaviours in healthy or tumour-bearing female mice 34 .
Euthanasia and tissue harvesting. Euthanasia was performed by isoflurane anesthesia and cardiac puncture exsanguination for endpoint blood collection, followed by decapitation. Blood samples were collected intracardially using sterile 25 G needles and 1 mL syringes. Blood samples were allowed to coagulate for 30-60 minutes at room temperature, and then centrifuged to extract serum 41 . Brains were immediately removed, rinsed and placed in sterile Hanks' Balanced Salt Solution (HBSS) for further processing (see Brain Metastases below).
Brain Metastases. Brain metastases were investigated using a clonogenic assay as outlined by Pulaski and Ostrand-Rosenberg (2001). This assay exploits the fact that, unlike mouse host cells, 4T1 tumour cells are resistant to 6-thioguanine (6-TG). Briefly, brains were minced and dissociated in a cocktail of collagenase type IV and elastase in HBSS. Samples were cultured in complete media containing 60 μ M 6-TG for 10-14 days. Cells were then fixed with methanol and stained with methylene blue, and the number of colonies were quantified. Clonogenic assay procedures are further detailed in "Supplementary Materials and Methods". Serum Glutamate. Serum proteins were precipitated from samples using 1 volume of chilled 100% (w/v) trichloroacetic acid (TCA) to 3 volumes of serum. Samples were centrifuged and the supernatant removed and neutralized with potassium hydroxide (KOH). Protein pellets were quantified for each sample using the BioRad assay as described in "Supplementary Materials and Methods". Serum glutamate in the supernatant was analyzed using the AMPLEX Red glutamic acid assay kit (Invitrogen/Molecular Probes, Eugene, OR, USA) and analyzed on a CytoFluor Series 4000 Fluorescence Multi-Well Plate Reader (PerSeptive Biosystems, Framingham, MA, USA). Serum glutamate was normalized to total serum protein for each sample. For each comparison, results from 3 independent experiments (each performed in triplicates) are represented as mean fold-changes in counts per minute (CMP)/mg protein relative to controls. All pairwise comparisons were evaluated using one-tailed Student t-tests.
One-way analysis of variance (ANOVA) was used to analyze the behavioural data (SPT, FST, and TST), serum glutamate, and serum cytokines. For these data, 4 planned comparisons were analyzed. Control mice were compared with tumour mice to establish benchmark tumour-induced changes for each assay. Tumour mice were compared with each intervention group (FLX, SSZ, and 5-ASA) to explore drug effects on each assay. The Holm-Šidák correction was used to correct for multiple comparisons.
In this investigation we used a peripheral tumour model for CID. Therefore, we are primarily interested in attenuating depressive behaviours by targeting peripheral tumour cells. However, 4T1 cells are capable of metastasizing to the brain 42 . Therefore, as a secondary measure for all behavioural data, linear regression analyses were performed to investigate whether brain metastases were predictive of tumour-induced or drug-induced behavioural changes.
For all analyses, the significance level was set at α = 0.05. Grubb's test was used to identify and exclude any outliers. The Shapiro-Wilk test of was used to determine normality. Data that was not normally distributed was subsequently transformed to achieve approximate normality prior to further statistical considerations. Data are presented as mean ± the SEM. Statistical analyses were performed using GraphPad Prism software version 6.0 for Macintosh (GraphPad Software, Inc., La Jolla, CA, USA). All graphs were prepared on Microsoft Excel 2011 version 14.1.0 (Microsoft, Redmond, WA, USA).
Behavioural analyses. Mice 29 . Mice in the tum+ 5-ASA/SP group consumed a mean 5-ASA dose of 20.32 ± 0.54 mg/kg/day, and a mean SP dose of 33.86 ± 0.89 mg/kg/day. In the SPT, lower preference for sucrose solution indicates anhedonia-like behaviour, which is a core symptom of depression characterized by the loss of ability to experience pleasure. Our results revealed a significant main effect of treatment group (F (4,53) = 3.84, P = 0.008). As expected, untreated tumour mice showed lower sucrose preference compared to control mice (P = 0.026) ( Fig. 2A). Treatment with FLX, SSZ, and 5-ASA/SP were all associated with higher sucrose preference compared to untreated tumour mice (P = 0.040, P = 0.002, P = 0.035, respectively).
In the FST and TST, longer immobility times indicate passive coping behaviour, a depressive feature that is sensitive to antidepressant treatment. Our results revealed a significant main effect of treatment group on FST immobility (F (4,52) = 4.36, P = 0.004). As expected, untreated tumour mice had higher immobility time compared to control mice (P = 0.033) (Fig. 2B). FLX and SSZ were associated with lower immobility times compared to untreated tumour mice (P = 0.033, P = 0.019, respectively). 5-ASA/SP treatment did not prevent the tumour-induced increase in immobility time.
Two measures for the TST are reported: immobility time (Fig. 2C) and PM (Fig. 2D). PM did not follow a normal distribution. Log 10 transformation of these data achieved approximate normality. There was no main effect of treatment group on TST immobility (F (4,54) = 1.70, P = 0.163) or log 10 PM (F (4,54) = 1.32, P = 0.276). Here we independently replicated our previous finding that the TST is not sensitive in detecting passive coping behaviour in the CID model 8 .
Approximately 35% of tumour-bearing mice developed brain metastases 4 weeks after inoculation, which is consistent with previous observations 42 . Among mice that developed brain metastases, the number of metastatic colonies was not normally distributed. Log 10 transformation of this data achieved approximate normality. Subsequently, regression analyses revealed that brain metastases did not predict sucrose preference in any direction for any of the tumour groups (Fig. 3A). Brain metastases were moderately predictive of increased immobility time on the FST for the tumour group treated with SSZ (P = 0.019) (Fig. 3B). A possible explanation for this is that mice with brain metastases were less responsive to SSZ treatment on this measure. In the TST, metastases were moderately predictive of lower immobility time and higher PM for untreated tumour mice (P = 0.005 and  (Fig. 3C). This result was counterintuitive, but may help explain the consistent lack of sensitivity of this test in detecting tumour-induced passive coping behaviour and drug effects. In addition, brain metastases were moderately predictive of higher PM for the mice treated with 5-ASA/SP (P = 0.013), but did not predict PM for mice treated with SSZ or FLX. Serum Glutamate. There was no main effect of treatment group on serum glutamate (F (4,47) = 1.87, P = 0.132). However, trends observed in serum glutamate levels are consistent with the expected glutamate release by 4T1 cells in tumour-bearing mice, and with the hypothesized inhibition of this glutamate release by SSZ (Fig. 4).

Serum Cytokines.
There was a significant main effect of treatment group on IL-1β serum level (F (4,51) = 2.62, P = 0.046). FLX-treated mice had a significantly lower level of serum IL-1β compared to untreated tumour mice (P = 0.022) (Fig. 5A). Serum levels of IL-1β did not differ for mice treated with SSZ or 5-ASA/SP compared to untreated tumour mice. There was a significant main effect of treatment group on IL-6 serum level (F (4,44) = 7.79, P < 0.0001). Untreated tumour mice had higher levels of serum IL-6 compared to control mice (P = 0.015) (Fig. 5B). Serum levels of IL-6 did not differ for mice treated with FLX, SSZ or 5-ASA/SP compared to untreated tumour mice. IL-17A and TNF-α levels were not different between groups (Fig. 5C and D).

Discussion
In the present study, murine 4T1 mammary carcinoma cells had greater cystine uptake, and consequently greater glutamate release, compared to non-tumourigenic TM40A mammary epithelial cells derived from the same mouse strain (Fig. 1A). Furthermore, we demonstrated that SSZ inhibits cystine uptake in both TM40A cells and 4T1 cells (Fig. 1B), suggesting that system x c − is primarily responsible for glutamate release. These results are consistent with previous studies from our group showing that human breast and prostate cancer cell lines release large amounts of glutamate into their microenvironment through the glutamate/cystine antiporter system x c −16,17 .
We also found that the use of SSZ (a system x c − inhibitor) and FLX (used as a positive control) in tumour-bearing In the sucrose preference test (SPT) preference is calculated as (sucrose water consumption/total fluid intake) × 100, where 50% represents equal consumption of sucrose water and regular water. Tumour mice showed a lower sucrose preference compared to control (P = 0.026). Treatments with FLX, SSZ, and 5-ASA/SP were associated with higher sucrose preference compared to untreated tumour mice (P = 0.040, P = 0.002, P = 0.035, respectively). (B) In the forced swim test (FST), tumour mice had higher immobility time compared to control mice (P = 0.033). FLX and SSZ were associated with lower immobility times compared to untreated tumour mice (P = 0.033, P = 0.019, respectively). (C) In the tail suspension test (TST), no significant changes were observed. Trends indicated a lower immobility time and higher log 10 PM associated with FLX and SSZ treatments compared to untreated mice. Data are presented as mean ± SEM. Results were analyzed using one-way ANOVA. The Holm-Šidák correction was used to correct for multiple comparisons. *P < 0.05, **P < 0.01.
Scientific RepoRts | 7:41382 | DOI: 10.1038/srep41382 mice exerted antidepressant-like effects, namely increased sucrose preference and decreased FST immobility. Results from all behavioural tests, serum glutamate, and serum cytokine analyses are summarized in Table 1.
To establish a benchmark for antidepressant-like efficacy in the present study, we used the selective serotonin reuptake inhibitor (SSRI) FLX. FLX effectively attenuates depressive-like behaviours in animal models when administered chronically 35,36 . SSRIs are useful for treating acute depressive episodes and preventing subsequent episodes in MDD subjects 43 . The rapid progression of tumour growth in our CID model was more suitable for testing the prophylactic efficacy of FLX, SSZ, and 5-ASA/SP, rather than reversal of depressive-like behaviours. In our CID model, chronic treatment with SSZ was at least as effective as FLX in preventing depressive-like behaviours on both the SPT and FST. Treatment with the SSZ metabolites, 5-ASA/SP, prevented the development of anhedonia-like behaviour on the SPT, but did not prevent passive coping on the FST. Since the FST is widely considered to be the gold standard in vivo assay for antidepressant detection, this suggests that the observed antidepressant-like effect of SSZ was likely primarily due to intact SSZ, and therefore system x c − inhibition. However, given the mixed behavioural results observed for 5-ASA/SP, further investigation is warranted to clarify the differential antidepressant effects of SSZ and its metabolites.
Serum cytokine analysis suggested that neither 70 mg/kg/day of SSZ nor 20 mg/kg/day of 5-ASA were effective at preventing the increase in IL-6 that was induced by tumour burden (Fig. 5A and B). A previous study found that 5-ASA, which is used to treat inflammatory bowel disease, decreased intestinal inflammation at doses as low as 8.3 mg/kg/day in rats 44 . The lack of effect we observed on serum cytokines may be due to the fact that only about 10% of 5-ASA is absorbed systemically, where it is rendered largely inactive by extensive acetylation 45 . Although anti-inflammatory agents, such as celecoxib, can exert antidepressant effects in a subset of individuals with MDD 46 , 5-ASA produced mixed behavioural results in our CID model, perhaps due to its limited absorption. Furthermore, it is worth noting that the present study focused on inflammatory markers that have previously been associated with depression. In light of the behavioural effect of 5-ASA/SP treatment on the SPT despite failure to prevent tumour-associated increases in IL-1β and IL-6, an investigation of a wider array of inflammatory markers might help clarify the observed behavioural effects of these drugs in our model. By contrast, FLX administration was associated with lower serum IL-1β compared to untreated tumour mice (Fig. 5A). This result  is consistent with meta-analysis results in MDD showing that FLX reduces serum levels of IL-1β but not TNF-α, with mixed results for IL-6 47 .
Results from the serum cytokine analysis also provide plausibility for a direct depressive mechanism of peripherally released glutamate on the CNS. In our CID model, tumour burden was associated with increased IL-6, and possibly IL-1β . These cytokines have been demonstrated to reduce BBB integrity, thereby increasing permeability 24,25 . Future analyses of BBB integrity and systemic administration of radiolabeled glutamate would be useful in better characterizing the impact of peripherally released glutamate on the CNS.
Intraperitoneal (i.p.) administration of SSZ inhibits tumour growth in a glioma model 48 , and affects CNS extracellular glutamate levels in healthy rats 49 . These data suggest that SSZ is able to cross the BBB. Therefore, we cannot rule out the possibility that the observed antidepressant-like behavioural effects of SSZ in the present study are partially attributable to a direct central mechanism. However, it is interesting to note that Lutgen et al. did not observe an effect on the FST with up to 16 mg/kg of i.p. SSZ in healthy rats 49 . Therefore, for the present study, it is likely that the observed antidepressant-like effect of SSZ is primarily attributable to peripheral inhibition of tumour-derived glutamate. This is consistent with our observation that intact SSZ, but not 5-ASA/ SP, prevented the development of passive coping behaviour on the FST. Furthermore, it has been demonstrated that the anti-tumour effects of SSZ are entirely attributable to system x c − inhibition, rather than its reported anti-inflammatory and apoptotic influence through nuclear factor-κ B (NF-κ B) inhibition 50 .
A limitation in our study is that the serum glutamate analysis did not yield significant results, which may in part be attributable to regulatory mechanisms, or due to the technique we used. Glutamate scavenging by glutamate-oxaloacetate transaminase (GOT) and glutamate-pyruvate transaminase (GPT) would make it difficult to detect dynamic glutamate dysregulation at a single time point 51 . Nevertheless, trends in serum glutamate were consistent with SSZ inhibition of glutamate release by cancer cells (Fig. 4). Although not statistically significant, the slight increase in serum glutamate in tumour mice, and decrease in SSZ treated mice, may be biologically relevant considering the tight regulation of this molecule, both peripherally and centrally. Further investigation into serum glutamate would benefit from the inclusion of more blood collection time points, as well as a more direct method of glutamate measurement, such as high performance liquid chromatography (HPLC). Caution should also be taken when interpreting the results of this study due to the lack of specificity of SSZ to system x c − . Oral SSZ was chosen for the current investigation due to the clinical relevance of this model. SSZ is a clinically useful and well-tolerated system x c − inhibitor 52 . Valuable insight may be derived from further investigation of more specific system x c − inhibitors, such as S-4-carboxy-phenylglycine (S-4-CPG) 53 . However, the pharmacokinetics of orally administered S-4-CPG has not been described, and this model would likely require daily intraperitoneal injections or surgically implanted osmotic pumps. Stress induced by either of these modes of drug delivery would need to be carefully considered. Genetically modulating system x c − activity represents a more specific method of investigating the impact of cancer cell-derived glutamate on CID. To this end, we are currently developing an xCT knockdown 4T1 cancer cell line to be used in future in vivo studies. This cell line will be crucial in determining whether glutamate secretion by peripheral cancer cells is necessarily and/or sufficient to mediate a depressive-like phenotype.
In summary, we have demonstrated that chronic oral administration of SSZ is at least as effective as FLX at preventing the development of depressive-like behaviours in a preclinical model of CID. This finding carries clinical significance, particularly considering recent evidence that SSZ inhibits tumour progression 48,50 and cancer-induced bone pain 37 . Our results are consistent with the hypothesis that the primary mechanism of antidepressant-like action of SSZ may be through inhibition of glutamate release via system x c − at the peripheral tumour site. However, this hypothesis warrants further investigation, as previously discussed. FLX was also efficacious in preventing the development of depressive-like behaviours in tumour-bearing mice. It may, therefore, be interesting to investigate the effects of standard antidepressants augmented with SSZ in the treatment of CID. While this investigation focuses on peripheral modulation of glutamate signaling in CID, further assessment of central glutamate signaling, using ketamine and other recently described glutamatergic modulators, would further enhance our understanding of CID. Overall, our results represent a potential new pathway for drug development in the treatment of CID. This approach can target the effects of cancer cells directly to prevent depressive  Table 1. Summary of results for behavioural and cytokine analyses. "− "Denotes no differences between groups. One arrow denote a mean difference between groups with 0.05 > P > 0.01. Two arrows denote a mean difference between groups with P < 0.01. The direction of the arrows represents the direction of change of mean for that particular measure. SPT, sucrose preference test; TST, tail suspension test; FST, forced swim test; FLX, fluoxetine; SSZ, sulfasalazine; 5-ASA, 5-aminosalicylic acid; SP, sulfapyridine; Glu, glutamate; IL, interleukin; TNF-α , tumour necrosis factor-α .