Sulforaphane inhibits proliferation and invasive activity of everolimus-resistant kidney cancer cells in vitro

Although the mechanistic target of rapamycin (mTOR) inhibitor, everolimus, has improved the outcome of patients with renal cell carcinoma (RCC), improvement is temporary due to the development of drug resistance. Since many patients encountering resistance turn to alternative/complementary treatment options, an investigation was initiated to evaluate whether the natural compound, sulforaphane (SFN), influences growth and invasive activity of everolimus-resistant (RCCres) compared to everolimus-sensitive (RCCpar) RCC cell lines in vitro. RCC cells were exposed to different concentrations of SFN and cell growth, cell proliferation, apoptosis, cell cycle, cell cycle regulating proteins, the mTOR-akt signaling axis, adhesion to human vascular endothelium and immobilized collagen, chemotactic activity, and influence on surface integrin receptor expression were investigated. SFN caused a significant reduction in both RCCres and RCCpar cell growth and proliferation, which correlated with an elevation in G2/M- and S-phase cells. SFN induced a marked decrease in the cell cycle activating proteins cdk1 and cyclin B and siRNA knock-down of cdk1 and cyclin B resulted in significantly diminished RCC cell growth. SFN also modulated adhesion and chemotaxis, which was associated with reduced expression of the integrin subtypes α5, α6, and β4. Distinct differences were seen in RCCres adhesion and chemotaxis (diminished by SFN) and RCCpar adhesion (enhanced by SFN) and chemotaxis (not influenced by SFN). Functional blocking of integrin subtypes demonstrated divergent action on RCC binding and invasion, depending on RCC cell sensitivity to everolimus. Therefore, SFN administration could hold potential for treating RCC patients with established resistance towards everolimus.


INTRODUCTION
Renal cell carcinoma (RCC) is the most common kidney tumor with more than 330,000 diagnosed cases and more than 140,000 patients succumbing to it every year [1]. Approximately one third of these patients have metastases at diagnosis, and 30-70% of patients with localized disease relapse within 5 years of surgery [2].
During the last decade several targeted drugs have been developed and approved as standard care for patients with metastasized RCC. Licensed target agents include the tyrosine kinase inhibitors, sunitinib and sorafenib, and the mechanistic target of rapamycin (mTOR) inhibitors, temsirolimus and everolimus. Compared to cytokine therapy with interferon a or interleukin-2 [3], these target agents have substantially improved patient outcome, but are not curative since resistance inevitably develops during tyrosine kinase or mTOR inhibitor therapy.
Dissatisfaction, along with strong side effects caused by conventional treatment have driven cancer patients to seek "alternative" and/or "complementary" care options to conventional treatment. "Complementary"

Research Paper
Oncotarget 85209 www.impactjournals.com/oncotarget is regarded as a non-mainstream approach used in addition to conventional medicine, whereas "alternative" refers to a non-mainstream approach instead of conventional medicine [4]. Complementary and alternative medicine (CAM) has become popular among tumor patients with more than 40% utilizing it worldwide [5]. In Europe, CAM application ranges from 15 to more than 70%, whereby herbal medicine is mostly employed [6]. In recent years, isothiocyanate sulforaphane (SFN), found in cruciferous vegetables such as broccoli and cabbage, has received increasing attention. SFN has been shown to inhibit tumor development and progression by modulating cancer related cell signaling and gene transcription [7]. Inhibition is accomplished by activating apoptosis, by cell cycle arrest, and by preventing metastatic processes [8]. Several epidemiologic studies have revealed a correlation between high intake of SFN rich vegetables and reduced cancer risk [9]. A clinical trial has provided evidence that SFN down-regulates the prostate-specific antigen (PSA) level in men with prostate cancer and decreases biochemical recurrence after radical prostatectomy [10]. Since so many RCC patients turn to CAM once resistance towards conventional therapy occurs, the present study was directed towards investigating whether SFN might indeed hold potential for patients with acquired drug resistance. Accordingly, three RCC cell lines were driven to non-responsiveness towards the mTOR-inhibitor everolimus and the influence of SFN on tumor growth, proliferation, and motility of everolimus-resistant and everolimus-sensitive tumor cells was compared.

Growth of everolimus-sensitive and -resistant RCC cells after everolimus application
Everolimus application caused a dose dependent significant reduction in the number of everolimus-sensitive RCC cells generated over 48 h in all three cell lines, Caki-1 par , KTCTL-26 par , and A498 par cells. The reduction was already apparent at 1 nM everolimus ( Figure 1A). Chronic, previous everolimus exposure induced resistance to acute everolimus application. An everolimus induced reduction in the growth rate greater than 20% was only achieved in KTCTL-26 res and A498 res cells when they were exposed to an acute dose of 500 or 1000 nM everolimus.

Growth of everolimus-sensitive and -resistant RCC cells after SFN application
SFN dose-dependently suppressed growth of all three everolimus-sensitive cell lines. It also diminished growth of RCC cells with everolimus-resistance, although the efficacy of SFN was lower in the everolimus-resistant cells, compared to that of the everolimus-sensitive cells ( Figure 1B). SFN did not induce early or late apoptosis, as evidenced by the Annexin V-FITC Apoptosis Detection kit.

Cell cycle shifts during SFN and acute everolimus application
Cell cycle analysis revealed an increased percentage of Caki-1 res cells in the G2/M-phase and an increased number of KTCTL-26 res and A498 res cells in the G2/M-and S-phases, compared to respective everolimus-sensitive cells ( Figure 2). In both Caki-1 par and Caki-1 res cells more G2/M-and S-phase cells and less G0/G1-phase cells were apparent when tumor cells were treated with 20 µM SFN for 24 h ( Figure 3A). Proliferation was elevated in Caki-1 res cells compared to Caki-1 par cells, as evidenced by BrdU incorporation ( Figure 3B). SFN significantly lowered proliferation of both Caki-1 par and Caki-1 res cells, as demonstrated by reduced BrdU incorporation (compared to untreated controls).

SFN alters expression of cell cycle regulating proteins
The influence of SFN on Caki-1 res differed from that on Caki-1 par in as much as total and activated Cdk1 and Cdk2 were elevated in the everolimus-resistant tumor cells. This was also true with respect to cyclin B, p27, pAkt, pRaptor, and pRictor ( Figure 3C). In contrast, p19 was diminished in Caki-1 res compared to Caki-1 par . SFN led to down-regulation of Cdk1, pCdk2, and Cdk2 in both Caki-1 par and Caki-1 res . Cyclin A, cyclin B (cyclin B > cyclin A) and p27 were lowered as well. p19 was up-regulated by SFN. pAkt and pRictor were down-regulated by SFN in Caki-1 res , whereas pRaptor was down-regulated by SFN in Caki-1 par .

Cyclin B-cdk1 knock-down
Since cyclin B and cdk1 were strongly down-regulated by SFN in both Caki-1 par and Caki-1 res cells, this downregulation could be responsible for the SFN evoked RCC cell growth reduction observed with the MTT assay. Figure  4 shows that treating Caki-1 par and Caki-1 res with cyclin B or cdk1 specific siRNA reduced the cdk1 and cyclin B content in both cell types. ( Figure 4B) This was associated with significant suppression of tumor cell growth ( Figure 4A).

Influence of SFN on tumor cell adhesion and motility
Caki-1 par cells rapidly attached to HUVEC with no difference in adherence after 30 and 120 min ( Figure 5A). More Caki-1 res than Caki-1 par cells attached to HUVEC in the initial phase of HUVEC-tumor cell interaction (30 min) but subsequently lost binding capacity, which was lowest after 120 min. SFN significantly down-regulated Oncotarget 85210 www.impactjournals.com/oncotarget Caki-1 par cell binding to HUVEC (30 and 60 min values, compared to the untreated control). Compared to untreated Caki-1 res controls, SFN reduced the number of HUVEC bound Caki-1 res cells after 30 min but binding was enhanced after 60 and 120 min ( Figure 5A). More Caki-1 res than Caki-1 par cells adhered to collagen ( Figure 5B) and more Caki-1 res than Caki-1 par crawled underneath the Transwell membrane ( Figure 5C). SFN increased the number of Caki-1 par but decreased the amount of Caki-1 res , which attached to immobilized collagen (compared to untreated controls; Figure 5B). SFN also significantly diminished chemotaxis in Caki-1 res but not in Caki-1 par cells ( Figure 5C).

Integrin α5, α6, and β4 blockade differentially influence Caki-1 res and Caki-1 par adhesion and migration
Since SFN strongly altered integrin α5, α6, and β4 expression, the relevance of these integrin types for the adhesive and motile behaviour of the everolimus-sensitive and everolimus-resistant Caki-1 cells was investigated. Integrin blockade differentially influenced adhesion and chemotaxis of Caki-1 par compared to Caki-1 res cells ( Figure 7). Blocking α5 triggered enhanced adhesion ( Figure 7A) and chemotactic ( Figure 7B) behavior of Caki-1 par , but diminished adhesion and chemotactic behavior in Caki-1 res cells. A similar phenomenon was observed when Caki-1 cells were treated with α6 or β4 function-blocking antibodies, with the exception that adhesion of Caki-1 res cells was not altered by integrin β4 blockade and chemotaxis of Caki-1 par was not modulated by integrin α6 blockade.

DISCUSSION
Resistance to everolimus in RCC cells was associated with increased mitotic activity, loosening of RCC-HUVEC contacts, increased interaction with a collagen matrix and elevated chemotactic movement. Earlier investigations have shown that everolimusresistance drives RCC cells towards high proliferation Growth was compared to untreated controls, set to 100%. Experiments were done in triplicate and repeated 5 times. * indicates significant difference to controls, # indicates significant difference between everolimus-resistant and -sensitive cells.
The effect was most prominent with a concentration of 20 µM. Whether this concentration can be attained under therapeutic conditions in tumor patients remains to be seen, since no pertinent studies are available. Cipolla et al., in treating prostate cancer patients with SFN, employed an oral dose of 60 mg/day, but provided no further details about SFN metabolism or bioavailability [10]. The urine concentration of dithiocarbamates, a group of SFNmetabolites, has been evaluated in clinical trials with healthy volunteers to be about 20 µM after consuming 50 g/day [13] or 70 g/day broccoli sprouts [14].
In the presence of SFN cell cycle analysis revealed tumor cell accumulation in the S-and G2/M-phase, with accumulation in the G2/M-phase being greater than in the  Oncotarget 85212 www.impactjournals.com/oncotarget S-phase in the parental (everolimus-sensitive) sublines. SFN induced G2/M phase-arrest has also been observed in osteosarcoma [15], bladder [16], colon [17], ovarian [18], prostate [19], and breast cancer cells [20]. This G2/M phase-arrest corroborates the SFN induced decrease in RCC proliferation found in the present investigation. An SFN induced down-regulation of the cyclin B-cdk1 axis also occurred in both everolimus-sensitive and -resistant RCC. The G2/M phase arrest, coupled with down-regulation of the cyclin B-cdk1 axis, caused by SFN indicates that SFN inhibits mitotic processes.
Though SFN has been shown by other investigators to reduce cdk1 [21,22], it is not the rule. An increase in cyclin B has been documented in bladder and prostate cancer cells [23,24] but no change has been reported in colon cancer-derived tumors [25], indicating that the influence of SFN may depend on the tumor entity. The action of SFN on cyclin A also demonstrates dependence on the tumor entity. In the present investigation cyclin A was down-regulated in RCC cells. It has also been reported to be down-regulated in osteosarcoma [26] and colon carcinoma cells [27]. However, down-regulation has not been reported in oral carcinoma cells [28].
The cell cycle regulator, p19, was up-regulated in both Caki-1 par and Caki-1 res by SFN. This observation is corroborated by several other investigations demonstrating a negative association between p19 expression and proliferative activity in a panel of solid tumor types [29][30][31].  Oncotarget 85213 www.impactjournals.com/oncotarget It, therefore, seems likely that SFN triggered elevation of p19, along with cdk-cyclin down-regulation, contributes to the inhibition of RCC cell growth and proliferation.
Since p27 serves as a tumor suppressor, it was expected that SFN would induce enhanced expression. However, this expectation was foiled since application of SFN caused diminished p27 expression in both everolimussensitive and -resistant tumor cells. Other investigators have reported that applying SFN to several tumor cell lines (but not to RCC cells) leads to p27 induction [32,33]. Another investigator found no modification of p27 by SFN [34]. Mans et al. concluded that p27 induces tumor cell senescence in RCC patients [35]. This is in line with other investigators showing loss of p27 to correlate with RCC recurrence and cancer-related patient death [36]. Meanwhile, newer interpretation has been advanced associating high cytoplasmic p27 expression with worse cancer-specific survival in RCC [37]. Presumably, shifting p27 from the nuclear to the cytoplasmic compartment could be predictive for poor outcome in RCC [38]. A further aspect should also be considered. Besides cell growth regulation, p27 also modifies cell-cell interaction,  Oncotarget 85214 www.impactjournals.com/oncotarget driving RCC cell invasion and metastatic progression forward [39]. Speculatively, SFN induced loss of p27 might hinder RCC cells in crossing the endothelial cell barrier and settling as secondary tumors. The alteration of adhesive and invasive behavior of RCC cells by SFN seen in our in vitro assays, at least partially, corroborates the assumption that diminished p27 expression not only contributes to cell growth regulation but also prevents RCC migration.
SFN exerted different influences on Caki-1 par and Caki-1 res . The total tumor cell number was reduced more when SFN was applied to everolimus-sensitive RCC cells, compared to everolimus-resistant cells. Sensitive tumor cells were mainly driven into the G2/M phase, whereas the resistant cells were equally driven into the G2/M and S-phases. The SFN induced inhibition of proliferation was more intense in everolimus-resistant than in -sensitive cells. The differences in tumor cell biology were accompanied by differences in cell cycle protein modifications. SFN caused diminished pCdk1 expression in everolimus-resistant but not in -sensitive RCC cells, whereas Cyclin A was reduced in the everolimus-sensitive but not in the -resistant RCC cells. SFN suppressed pRaptor in the everolimus-sensitive and pRictor in the everolimus-resistant RCC cells. Since resistance is characterized by altered cell signaling machinery, it is not surprising that molecules within the signaling cascade are altered differently in Caki-1 res and Caki-1 par cells when SFN is applied. With respect to the mTOR sub-members Rictor and Raptor, these protein complexes individually modify cell cycle progression and proliferation [40,41]. Divergent regulation of pRictor and pRaptor, depending on everolimus sensitivity, might therefore account for the different influence of SFN on the Caki-1 cell lines.
Different responses of Caki-1 res and Caki-1 par cells to SFN were also apparent with respect to adhesion and chemotaxis. Only a slight reduction in Caki-1 par cells bound to HUVEC was induced by SFN, attachment to collagen was enhanced and motile behavior was not influenced at all. In strong contrast, SFN prevented Caki-1 res from becoming highly adhesive or highly motile. More Caki-1 res cells attached to HUVEC after 120 min incubation in the presence of SFN, fewer cells bound to collagen and only a few cells migrated. These effects on the everolimus-resistant tumor cells open the possibility that SFN might be a treatment option once tumors have become non-responsive to conventional drug treatment. SFN has recently been shown to reduce the metastatic potential of drug-resistant breast cancer cells in vitro [42]. There is also evidence that SFN might overcome chemoresistance towards adriamycin, cisplatin [43], doxorubicin [44], and paclitaxel [45]. The current results demonstrate high efficacy of SFN in reducing the metastatic potential of everolimus-resistant RCC cells in vitro. Whether SFN can re-establish everolimus sensitivity is currently under investigation. SFN caused down-regulation of the integrin adhesion receptors α5, α6, and β4 subtypes in both Caki-1 par and Caki-1 res cells. This consistent down-regulation was accompanied by enhanced binding of Caki-1 par cells, but blocked binding and chemotaxis of Caki-1 res , indicating different integrin function, depending on the drug responsiveness of the tumor cells. For example blocking of α5 was accompanied by elevated adhesion and chemotaxis of Caki-1 par , but by reduced adhesion and chemotaxis of Caki-1 res cells. Loss of β4 integrin enhanced motile activity of Caki-1 par , but reduced that of Caki-1 res cells. Blockade of integrin α6 led to up-regulation of Caki-1 par adhesion, but to a down-regulation of Caki-1 res adhesion. Divergent activity of a particular integrin is not uncommon and we have recently reported that integrins undergo functional changes during the process of resistance development, driving tumor cells to become highly invasive [11,46]. This disparity in integrin function may explain why SFN treatment caused enhanced binding of everolimussensitive RCC cells to collagen but reduced binding in everolimus-resistant cells. Diminished α5, α6, and β4 expression after SFN application could be responsible for the diminished chemotactic behavior of Caki-1 res . Chemotaxis of Caki-1 par cells was not influenced by SFN. Possibly, integrin α6 reduction is irrelevant to motile spreading of everolimus-sensitive cells and/or because integrin α2 suppression, induced by SFN exclusively in Caki-1 par , might correlate with decreased chemotaxis [47], (thereby counteracting the pro-chemotaxic effects induced by α5 and β4 reduction). However SFN actually functions, it triggers firm attachment of everolimus-sensitive RCC to the collagen matrix, thereby indirectly preventing invasive progression. In contrast, everolimus-resistant cells are modified by SFN through direct inhibition of both adhesion and chemotaxis. The direct action of SFN on adhesion and penetration could be of potential importance in treating patients with everolimus-resistant cancer.
Overall, SFN has been demonstrated to act on a panel of RCC cell lines by reducing tumor growth and proliferation in vitro. Verification in animals with acquired everolimus resistant cancer must reveal whether the potential of SFN seen in vitro can be substantiated in vivo. If verified, then SFN might be of value in supporting conventional therapy with an mTOR inhibitor in patients with acquired resistance.

Cells and cell culture
A panel of three RCC cell lines were initially investigated (Caki-1, KTCTL-26, and A498). The effects of everolimus and L-sulforaphane on cell growth and the cell cycle were tested on all three of these cell lines in an everolimus-sensitive and an everolimus-resistant state. Furthergoing investigation including cell proliferation, Oncotarget 85215 www.impactjournals.com/oncotarget apoptosis, cell cycle regulating proteins, the mTOR-akt signaling axis, adhesion to human vascular endothelium and immobilized collagen, chemotactic activity, and influence on surface α and β integrin receptor expression were carried out only on Caki-1 cells.

Drug treatment
Everolimus (Novartis Pharma AG, Basel, Switzerland) was dissolved in DMSO as a 10 mM stock solution and stored in aliquots at −20°C. Prior to experiments, everolimus was diluted in cell culture medium. Cell growth experiments were carried out in the presence of 1 -1000 nM everolimus. Resistance towards everolimus was induced by subjecting Caki-1, KTCTL-26, or A498 cells with stepwise ascending concentrations from 1 nM up to 1 μM. The tumor cells were further exposed to 1 μM everolimus twice weekly for over a year. Tumor cells, resistant to everolimus, were designated Caki-1 res , KTCTL-26 res , and A498 res . Control cells, sensitive to everolimus, were designated Caki-1 par , KTCTL-26 par , and A498 par . L-Sulforaphane was provided by Biomol, Hamburg, Germany. Concentrations from 1.25 -20 µM SFN were applied to cell cultures to evaluate effects on the growth of both everolimus-resistant and everolimussensitive tumor cells. Since 20 µM SFN showed the greatest growth inhibitory effect, all further experiments were carried out with 20 µM SFN. Controls remained untreated. To evaluate toxic effects of everolimus and SFN, cell viability was determined by trypan blue (Gibco/ Invitrogen, Darmstadt, Germany).

Apoptosis
To detect apoptosis the expression of Annexin V/ propidium iodide (PI) was evaluated using the Annexin V-FITC Apoptosis Detection kit (BD Pharmingen, Heidelberg, Germany). Tumor cells were washed twice with PBS-buffer and then incubated with 5 μl of Annexin V-FITC and 5 μl of PI in the dark for 15 min at room temperature. Cells were analyzed on a FACScalibur (BD Biosciences, Heidelberg, Germany). The percentage of vital, necrotic, and apoptotic cells (early and late) in each quadrant was calculated using Cell-Quest software (BD Biosciences).

Measurement of tumor cell growth and proliferation
Cell growth was assessed using the 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) dye reduction assay (Roche Diagnostics, Penzberg, Germany). RCC cells (50 µl, 1 × 10 5 cells/ml) were seeded onto 96-well tissue culture plates. After 24, 48, and 72 h MTT (0.5 mg/ml) was added for an additional 4 h. Subsequently, cells were lysed in a buffer containing 10% SDS in 0.01 M HCl. The plates were incubated overnight at 37°C, 5% CO 2 . Absorbance at 550 nm was determined for each well using a microplate ELISA reader. Each experiment was done in triplicate. After subtracting background absorbance, results were expressed as 24 -72 h cell growth rate calculated in percentage increase compared to controls set at 100%.
Cell proliferation was measured using a BrdU cell proliferation enzyme-linked immunosorbent assay (ELISA) kit (Calbiochem/Merck Biosciences, Darmstadt, Germany). Tumor cells, seeded onto 96-well microtitre plates, were incubated with 20 µl BrdU-labeling solution per well for 8 h, and then fixed and detected using anti-BrdU mAb according to the manufacturer's instructions. Absorbance was measured at 450 nm.

Cell cycle analysis
Cell cycle analysis was performed with everolimusresistant and -sensitive subconfluent RCC cultures. Tumor cell populations were stained with propidium iodide using a Cycle TEST PLUS DNA Reagent Kit (BD Pharmingen, Heidelberg, Germany) and then subjected to flow cytometry with a FACScan flow cytometer (Becton Dickinson). 10,000 events were collected from each sample. Data acquisition was carried out using Cell-Quest software and cell cycle distribution, calculated with ModFit software (BD Biosciences). The number of gated cells in the G1-, S-, or G2/M-phases was expressed as % of total cells.

RCC cell adhesion
To analyze Caki-1 adhesion, HUVECs were seeded onto six-well multiplates (Falcon Primaria; BD Biosciences, Heidelberg, Germany) in complete HUVEC medium. When confluency was reached, Caki-1 cells (everolimus-resistant and -sensitive, SFN treated and non-treated controls) were detached from their culture flasks by Accutase treatment (PAA Laboratories, Cölbe, Germany). Tumor cells (0.5 × 10 6 ) were then added to the HUVEC monolayer for 30, 60, or 120 minutes. Subsequently, non-adherent cells were washed off using warmed (37°C) M199. Adherent cells were fixed with 1% glutaraldehyde and counted in five different fields, each 0.25 mm 2 , using a phase-contrast microscope. Mean cellular adhesion in the five fields was calculated.

Attachment to a collagen matrix
Six-well plates (Falcon Primaria) were coated with collagen G [extracted from calfskin, consisting of 90% collagen type I and 10% collagen type III; diluted to 400 μg/ml in phosphate-buffered saline (PBS); Seromed, Berlin, Germany] overnight. Plastic dishes served as the background control. Plates were washed with 1% BSA in PBS to block nonspecific cell adhesion. Caki-1 cells (0.5 × 10 6 ) were then added to each well for 60 minutes. Subsequently, non-adherent tumor cells were washed off, and the remaining adherent cells were fixed with 1% glutaraldehyde and counted under a microscope. Mean cellular adhesion, defined by adherent cells coated well − adherent cells background , was calculated from five different observation fields (5 × 0.25 mm 2 ).

Tumor cell chemotaxis
Serum-induced chemotactic movement was investigated using six-well Transwell chambers (Greiner Bio-One, Frickenhausen, Germany) with 8-μm pores. Caki-1 cells (0.5 × 10 6 /ml) were placed in the upper chamber in serum-free medium. The lower chamber contained 10% serum. After 20-hour incubation, the upper surface of the Transwell membrane was gently wiped with a cotton swab to remove cells that had not migrated. Cells, which had moved to the lower surface of the membrane, were stained using hematoxylin and counted under a microscope. Mean chemotaxis was calculated from five different observation fields (5 × 0.25 mm 2 ).