Effective Elimination of Lymphoma Cells in Stromal Microenvironment by Double Impact on ROS and ATP Metabolism Using Auranon

Protection of cancer cells by the stromal microenvironment is an important drug-resistant mechanism in vivo, and poses a major challenge in cancer treatment. Based on our recent ndings that auranon (AF), an anti-rheumatic drug known to inhibit thioredoxin reductases (TrxR), was effective against several cancer models in vivo, this study tested the ability of AF to eliminate cancer cells in the tumor stromal microenvironment and investigated the underlying mechanisms. and suggests a possibility to repurpose this anti-rheumatic drug for treatment of lymphoma and overcome drug resistance in stromal microenvironment. vivo activity body Results are mean SD, p values determined by Student’s t test. Schematic illustration of the proposed mechanisms of action of AF against cancer cells.AF induces severe ATP depletion by double impacts on glycolysis and mitochondrial respiration through inhibition of GAPDH and suppression of complexes II and III. The thiol-antioxidant NAC rescued AF-induced cell death by conjugating with AF and preventing the drug uptake by cancer cells.


Introduction
Despite signi cant progress in our understanding of cancer biology and in the development of novel therapeutic agents, many cancers remain incurable. One of the major challenges is the development of drug resistance, which often leads to persistence of residual disease and cancer recurrence. Tumor microenvironment plays a major role in promoting cancer cell survival and drug resistance by providing protective tissue niches for cancer cells (1). For instance, some drug-resistant malignant B cells often remains in tissue niches such as lymph nodes and bone marrow. This residual malignant B cells eventually regrow leading to relapse of disease, which is often no longer responsive to original treatment (2). Thus, identi cation of drugs that are able to effectively kill cancer cells in tumor microenvironment is extremely important to enhance in vivo therapeutic activity and improve clinical outcome.
Aurano n (AF) is a gold-containing drug used in clinical treatment of rheumatoid arthritis (3). Recent studies suggest that this compound also has anticancer activity in both solid tumors and hematologic malignancies (4,5). However, the exact mechanisms responsible for its anti-cancer effect remain poorly understood. Inhibition of TrxRs has been considered as a key pharmacological action of AF, which could result in an increase of reactive oxygen species (ROS) leading to damage to cancer cells (6). However, some studies suggest that oxidative stress induced by AF might not be a key mechanism responsible for its cytotoxic action against cancer cells (7,8). In this study, we found that AF exhibited a potent anticancer activity against various types of cancer cells, and was particularly effective in killing lymphoma cells in the presence of stroma cells mimicking tissue microenvironment, which otherwise would protect cancer cells against traditional anticancer drugs. Further mechanistic study revealed that disruption of cancer metabolism by double hits on ATP generation and direct stimulation of TrxR to produce ROS in cancer cells in the tissue niches by AF, rather than direct inhibition of TrxRs, appear to be the underlying mechanisms of action.
Flow cytometric analysis: The cells were stained with Annexin-V-FITC/PI for analysis of cell viability using ow cytometer (Beckman Counter, Fullerton, CA, USA).Cells were stained with hydroethidine (100 ng/ml, 1 hour) for detection of superoxide, CM-H 2 DCFDA (5 µM, 40 minutes) for detection of general ROS, Rhodamine-123 (1 µM, 40 minutes) for detection of mitochondrial transmembrane potential, or NAO (1 µM, 30 minutes) for detection of mitochondrial cardiolipin, using the manufacturer's recommended conditions and analyzed by ow cytometry.
Cancer-stromal cell co-culture: Stromal cells were seeded in 24-well or 6-well plates, allowed to adhere and grow overnight before addition of tumor cells at a ratio of 3:1 (lymphoma:NKtert cells).
Mitochondrial bioenergetics assay: Suspension cells were immobilized on culture plates coated with Cell-Tak (BD Bioscience, CN) at a density of 2-3 × 10 5 cells/well. Oxygen consumption rate (OCR) and extracellular acidi cation rate (ECAR) were measured in unbuffered DMEM (Sigma-Aldrich) supplemented with 10 mM D-glucose, 2 mM L-glutamine, and/or 1 mM pyruvate as indicated, using a Seahorse XF24 analyzer.
Enzyme activity assays: The cellular TrxR activity was measured using a TrxR Activity Colorimetric Assay kit according to the protocol provided by the manufacturer (Biovision, Milpitas, CA, USA). The activities of mitochondrial respiratory complexes were measured using Complex I/II/III/IV/V Enzyme Activity Microplate Assay Kit (Abcam, Cambridge, MA, USA). GAPDH activity was analyzed using a GAPDH Assay In vivo animal studies: JeKo-1 cells were inoculated subcutaneously (2 × 10 6 cells in matrigel per injection) on the right anks of 8-week old nude mice (n = 10 mice per group). Four weeks after tumor inoculation, one group of mice received 10 mg/kg AF by intraperitoneal (i.p.) injection daily, 5 days per week (Monday-Friday). Normal saline was used as a control. After 1 month, the treatment was stopped and the mice were sacri ced. Tumor weights were recorded. To test the effect of AF on cancer cell metabolism in vivo, 2 × 10 6 JeKo-1 cells were inoculated subcutaneously on the right anks of nude mice (n = 3 mice/group x 2). One group received 10 mg/kg AF by i.p. injections for 3 days after 4 weeks. Another group of mice were treated with normal saline as control. The mice were sacri ced and lymphoma cells were isolated from the tumor tissues for analysis of OCR and ECAR. All animal experiments were conducted in accordance with the institutional guidelines and approved by the Animal Care and Use Committee of Sun Yat-Sen University Cancer Center. The investigators were not blinded to the group allocation during the experiments.
Statistical analysis: Additional experimental procedures are provided under Supplemental Information. Data were representative of at least three independent experiments unless otherwise indicated. Statistical analyses including two-sided Student's t-test and analysis of variance (ANOVA) followed by Newman-Keuls multiple-comparison test were performed using GraphPad Prism software (GraphPad, San Diego, CA, USA). Data are expressed as mean ± standard deviation (SD). A p value of less than 0.05 is considered statistically signi cant. Group sizes for mice were chosen based on literature and prior experience with these mouse models (9).

AF selectively kills cancer cells with minimum toxicity to normal cells
We rst tested the sensitivity of various cancer cell lines of different tissue origins to AF. This drug exhibited potent growth inhibition on all 33 cancer cell lines tested, with IC 50 values of less than 2 µM in most cases (Fig. 1a). AF was particularly effective against lymphoma cells, including mantle cell lymphoma (MCL) JeKo-1 cell line, primary lymphoma cells isolated from patients with MCL and diffused large B cell lymphoma (DLBCL) (Fig. 1b). Notably, AF at low concentrations (0.2-0.5 µM) was highly effective against the malignant cells. In contrast, AF induced minimal cytotoxicity in non-cancerous cells, including primary lymphocytes isolated from the biopsy samples of patients with benign lymphadenitis (

AF is highly effective against lymphoma cells in the presence of stromal cells
Since previous studies showed that stromal cells in tumor microenvironment could protect leukemic cells and promote drug resistance (10-13), we tested the effect of AF against lymphoma cells in the presence of stromal cells using a co-culture system. Surprisingly, instead of seeing a protective effect, we found that AF exhibited greater cytotoxic activity against lymphoma cells (both JeKo-1 and Raji) when they were co-cultured with NKtert stromal cells ( Fig. 2a-b). These data together show a striking ability of AF to kill malignant cells in stromal microenvironment.
Inhibition of TrxRs by AF is not a key mechanism responsible for cytotoxicity It has been suggested that inhibition of TrxRs by AF might lead to ROS-mediated apoptosis (14). We tested the impact of AF on TrxR activity and cell viability in two lymphoma cell lines, and found that AF could indeed inhibit TrxR activity and induced an increase of ROS (Fig. 3a). Pretreatment of cells with an antioxidant N-acetyl-L-cysteine (NAC) seemed able to reduce AF-induced apoptosis ( Fig. 3b and Supplementary Fig. 2a), suggesting a potential role of ROS in drug-induced cell death. However, since NAC is a thiol-containing compound capable of directly conjugating with certain compounds including AF (15,16), we tested if the protective effect of NAC was due to its conjugation with AF to prevent cellular uptake of AF. Quantitative assay of cellular gold (Au) content showed an 80% decrease of intracellular AF in the presence of NAC ( Fig. 3b and Supplementary Fig. 2b). When the pre-incubated NAC was removed by washing before addition of AF to avoid direct NAC-AF binding in the medium, such washing restored cellular uptake of AF ( Fig. 3b and Supplementary Fig. 2b) and almost completely abolished the NAC protective effect against AF cytotoxicity ( Fig. 3b and Supplementary Fig. 2a). Since NAC pre-incubation followed by washing still led to lower cellular ROS (data not shown), its failure to prevent AF-induced cytotoxicity suggested that ROS accumulation might not be a key cytotoxic mechanism.
To further evaluate the role of TrxR inhibition in AF-induced cytotoxicity, we tested if suppression of TrxR expression by siRNA could affect cell viability. A partial silencing of TrxR1 expression by siRNA in Raji cells resulted in approximately 50% decrease of total TrxR enzyme activity, which was comparable to the degree of TrxR inhibition by AF, did not cause any loss of cell viability (Fig. 3c). Further, siRNA silencing of both TrxR1 and TrxR2 in a more readily transfectable cell line (HCT116) with a 75% reduction of total TrxR activity still did not induce any loss of cell viability (Fig. 3c). These data together suggest that inhibition of TrxRs was unlike a key event responsible for the cytotoxic effect of AF.
AF induces severe ATP depletion through double impacts on energy metabolism To explore the potential mechanisms for AF-induced cytotoxicity in lymphoma cells, we rst tested potential impact of AF on cellular energy metabolism, and observed that AF at a relatively low concentration (1 µM) could induce a rapid depletion of ATP in JeKo-1 and Raji cells ( Fig. 4a and Supplementary Fig. 3). We then examined the effect of NAC on AF-induced ATP depletion. A simultaneous incubation of NAC with AF could prevent ATP depletion, whereas pretreatment of cells with NAC followed by washing before adding AF was unable to prevent ATP deletion (Fig. 4b). These results were consistent with the cytotoxicity data shown in Fig. 3b and Supplementary Fig. 2a and suggested that ROS did not play a key role in AF-induced energy collapse.
To further explore the mechanism for AF-induced ATP depletion, we tested the impact of AF on two main ATP generation pathways, oxidative phosphorylation and glycolysis, using a Seahorse extracellular ux analyzer to measure oxygen consumption rate (OCR) and extracellular acidi cation rate (ECAR). A substantial decrease of OCR was detected within minutes after addition of AF to JeKo-1 cells (Fig. 4c). This was consistently observed in other lymphoma cells ( Supplementary Fig. 4). Concomitantly, a 30% decrease of ECAR was also observed after AF treatment in lymphoma cell lines ( Fig. 4d and Supplementary Fig. 4). Silencing of TrxR expression by siRNA did not affect the effect of AF on oxygen consumption, since the control Raji cells and TrxR-silenced cells exhibited similar degrees of OCR inhibition by AF (Fig. 4e).
We then further examined the impact of AF on mitochondrial respiratory chain activity. Functional analysis revealed that AF caused a signi cant suppression of respiratory complex II (Fig. 4f) and a moderate decrease in complex III activity (Fig. 4g), whereas no changes in complexes I, IV and V were detected ( Supplementary Fig. 5a). Western blot analysis showed no signi cant changes in expression of any respiratory chain proteins (Supplementary Fig. 5b). We then used speci c inhibitors of mitochondrial respiratory complexes in combination with relevant respiratory substrates to analyze the detail changes in each mitochondrial respiratory chain activity. Supplement of the complex II substrate (succinate) resulted in an increase of OCR in the control cells but not in the AF-treated cells (Fig. 4h), indicating that the activity of complex II was inhibited by AF. In contrast, complex I inhibitor (rotenone) and complex III inhibitor (antimycin) caused a decrease of OCR in both control and AF-treated cells (Fig. 4h and Supplementary Fig. 5c), suggesting complexes I and III were not signi cantly affected by AF. Since the total mitochondrial oxygen consumption represents the electron transport activities from complex I→CoQ→III→IV and complex II→CoQ→III→IV (17), these data together suggest that the main site inhibited by AF was complex II.
To test if AF inhibition of mitochondrial respiration was responsible for its cytotoxicity, various concentrations of rotenone (a potent mitochondrial respiration inhibitor) were titrated to select a concentration that caused a similar degree of respiratory suppression induced by 1 µM AF (Supplementary Fig. 6a). This concentration of rotenone (0.05 µM), although potently inhibit respiration, did not cause any signi cant ATP depletion or cell death (Supplementary Fig. 6b and c). Importantly, the rotenone-treated cells exhibited a signi cant upregulation of glycolysis whereas the AF-treated cells exhibited a decrease in glycolysis (Fig. 4i), indicating that cells were able to compensate the energy loss from mitochondrial inhibition by rotenone, but not able to do so when they were treated with AF. These data together suggest a possibility that AF might inhibit both mitochondrial respiration and glycolysis leading to severe ATP depletion.
We then tested the impact of AF on several key glycolytic enzymes, including hexokinase (HK), glyceraldehyde phosphate dehydrogenase (GAPDH), pyruvate kinase (PK), and lactate dedydrogenase (LDH). A dose-dependent decrease in GAPDH activity was observed in two lymphoma cell lines treated with AF ( Fig. 4j and Supplementary Fig. 7a). In vitro enzyme assay using puri ed GAPDH showed that AF directly inhibited GAPDH (Fig. 4k), whereas it had no impact on other glycolytic enzymes ( Supplementary  Fig. 7b-d).
Since pyruvate is the product of glycolysis that functions as a precursor for acetyl CoA production in mitochondria (18,19), it might be possible that the suppression of mitochondrial respiration in AF-treated cells could be a consequence of glycolytic inhibition leading to a depletion of pyruvate. To exam this possibility, we tested whether exogenous pyruvate could rescue the AF-induced respiratory inhibition and cytotoxicity. Addition of pyruvate did not prevent AF-induced mitochondrial dysfunction ( Supplementary  Fig. 8a), nor did it suppress apoptosis ( Supplementary Fig. 8b), indicating that AF-induced mitochondrial dysfunction was not a consequence of GAPDH inhibition.

Normal cells survive AF inhibition of mitochondrial respiration through upregulation of glycolysis
We then tested the impact of AF on energy metabolism in normal cells. Interestingly, mitochondrial respiration in normal PBMCs and stromal cells was inhibited by AF (Fig. 5a) to similar degrees as observed in cancer cells (Fig. 4c). However, unlike cancer cells whose glycolysis was also inhibited by AF, normal cells exhibited a signi cant upregulation of glycolysis after AF treatment (Fig. 5b). Such increase in glycolysis seemed able to compensate the decrease of mitochondrial ATP generation, since total cellular ATP did not decrease in the three types of normal cells treated with AF (Fig. 5c). The ability of normal cells to maintain cellular ATP seemed consistent with their ability to survive AF treatment.

Stromal cells promote TrxR1 expression in cancer cells and enhance AF-induced ROS generation
The nding that AF induced severe ATP depletion could explain its cytotoxic effect against cancer cells, but this could not explain why AF was more effective in killing cancer cells in the presence of stromal cells. In our attempt to explore the underlying mechanisms, we observed that in the presence of stromal cells, AF induced higher ROS generation compared to the cells treated with AF without stromal cells ( Supplementary Fig. 9). Such higher ROS was unexpected, since stromal cells normally protect cancer cells by reducing their ROS stress (20). This unexpected observation prompted us to investigate the underlying mechanism. Since AF could inhibit TrxRs and causes ROS increase, we rst examined the impact of stromal cells on TrxR expression in cancer cells. Western blot analysis revealed that TrxR1 expression was consistently up-regulated in lymphoma cells when they were co-cultured with stromal cells (Fig. 6a).
The consistent upregulation of TrxR1 by stromal cells led us to speculate its potential role in AF-induced ROS generation. Since the reactive cysteine and selenocysteine residues in the active site of TrxR1 play key roles in transporting electron and mediating the redox cycle during the reduction of thioredoxin (21), we postulate that a reaction of AF with the cysteine or selenocysteine in the active site of TrxR1 would disrupt the electron transport process, leading to a leak of electrons to be captured by molecular oxygen to form superoxide as illustrated in Fig. 6b. As such, AF would turn TrxR1 into a ROS-generating protein, and the higher expression of TrxR1 in cancer cells in stromal environment would generate more ROS in the presence of AF. To test this possibility, we used a luminescence-based in vitro assay to measure superoxide generation in a cell-free system containing puri ed TrxR1, Trx, NADPH, and AF. Addition of AF to the complete reaction mixture caused a rapid generation of ROS in a time-dependent manner (red color curve). Removal of AF or TrxR1 from the reaction system led to a substantial decrease in ROS generation rate (Fig. 6b), suggesting that both TrxR1 and AF were required for a sustained production of ROS. Collectively, these data suggest that AF could turn the TrxR1/Trx/NADPH redox system into a ROSgenerating machinery, and the increased expression of TrxR1 in cancer cells in stromal environment renders them more prone to AF-induced ROS generation.

AF exhibits potent therapeutic activity in lymphoma models
Since AF could effectively kill lymphoma cells in vitro, we further test its therapeutic activity in vivo. In the lymphoma xenograft mouse model bearing lymphoma cells, AF treatment showed a signi cant therapeutic effect (Fig. 7a). This drug was well tolerated, and the mice exhibited no signi cant loss of body weight (Fig. 7a). Importantly, there was a signi cant decrease in both OCR and ECAR in tumor cells isolated from the lymphoma-bearing mice treated with AF ( Supplementary Fig. 10), indicating that AF was able to inhibit tumor cell energy metabolism in vivo, similar to that observed in vitro. These data together suggest that AF has promising in vivo therapeutic activity in lymphoma.

Discussion
As an anti-rheumatoid drug, AF has been used in clinical treatment of patients with certain rheumatoid arthritis for decades. Recent studies show that this compound also has signi cant anticancer activity, and suggest a possibility to repurpose this old drug for new disease indications. However, repurpose of AF as an anticancer drug has been hindered due in part to unclear mechanisms and unde ned advantages over other anticancer drugs. Our study demonstrated that AF was highly effective in killing lymphoma cells, as well as other cancer cells of different tissue origins. One clear advantage of AF is high therapeutic selectivity without signi cant toxicity to normal cells, and another advantage would be the unique mechanisms of action in impacting energy metabolism leading to severe ATP depletion in cancer cells.
Our study has revealed the unique mechanisms of action responsible for AF's anticancer activity. Inhibition of TrxRs by AF has been known for a long time, and TrxRs are considered the primary targets of AF. It has been suggested that inhibition of TrxR by AF would lead to an increase in ROS, which could cause cellular damage and cancer cell death (5,22). However, our study showed that although AF could indeed inhibit TrxR activity, such inhibition seems unlikely responsible for its cytotoxicity. This conclusion is supported by the following observations: (1) siRNA knockdown of TrxR expression reduced TrxR enzyme activity to the same degree as the inhibition by a toxic concentration of AF, but such TrxR knockdown did not cause any signi cant loss of cell viability; (2) Knockdown of TrxR expression by siRNA did not affect cellular sensitivity to AF. Lymphoma cells with TrxR knockdown remained sensitive to AF; (3) Antioxidants such as NAC (with wash) did not rescue cell death induced by AF, suggesting the ROS increase due to TrxR inhibition may not be a key event responsible for cytotoxicity.
In our mechanistic study, we discovered that AF induced severe ATP depletion through "double hits" on energy metabolism. Our data showed that AF suppressed mitochondrial respiration via inhibiting complex II of the respiratory chain, and inhibited glycolysis via targeting GAPDH. Since energy metabolism has a high degree of plasticity and inhibition of one metabolic pathway (e.g. mitochondrial oxidative phosphorylation) could be compensated by upregulation of another pathway (e.g. glycolysis), thus the double hits on both pathways are important for AF to cause energy collapse in cancer cells. This notion is supported by the observation that rotenone potently inhibited mitochondrial respiration but did not cause severe ATP depletion nor induced cell death, due to up-regulation of glycolysis. Such compensatory mechanism may also provide a basis to understand the therapeutic selectivity of AF. It is important to note that in all three types of normal cells (PBMC, NKtert, HK), AF could inhibit their mitochondrial respiration but did not cause ATP depletion and cell death, largely due to an upregulation of glycolysis in the normal cells. However, it is still unclear why normal cells were able to up-regulate glycolysis when mitochondrial respiration was inhibited, whereas cancer cells were unable to do so. One potential explanation would be that GAPDH in normal cells is insensitive to AF due the presence of a yet unknown factor that interacts with GAPDH and protects this enzyme from being targeted by AF. Alternatively, normal cells could still up-regulate glycolysis by bypassing the GAPDH-catalyzed steps through channeling proper metabolites into the glycolytic pathway at a step downstream of GAPDH.
These possibilities require further investigation in separate studies. Figure 7b provides a schematic illustration of the mechanisms of action of aurano n against cancer cells.
In summary, our study showed that the anti-rheumatic agent AF could effectively kill lymphoma cells through induction of severe ATP depletion by inhibition of mitochondrial respiration and suppression of glycolysis. Our new ndings underscore a possibility to repurpose the anti-rheumatic drug AF as a promising anticancer agent for treatment of multiple cancers, especially lymphoma. Based on the new mechanisms of AF, we expect that AF could be even more effective in combination with other mitochondria-targeting drugs such as Ionidamine and arsenic trioxide (23). These possibilities merit further evaluation in laboratory and clinical setting.  ; n=3). For a-c, data are presented as mean ± SD; *p<0.05, **p<0.01.

Figure 2
Enhanced anticancer activity of AF in the presence of stromal cells. (a and b) Viability of JeKo-1 or Raji cells treated with 0.5 or 1 µM AF for 24 h in the presence or absence of stromal NKtert cells, measured by annexin V-PI assay (n = 3). Representative data were shown. For a-b, data are presented as mean ± SD. washed twice with fresh medium or not as indicated, followed by 1μM AF treatment for 20 h or 2 h. Cell viability was measured by annexin V-PI assay; intracellular Au content was analyzed by ICP-MS(n=3). (c) Western blot analysis of TrxR1 and TrxR2 in Raji cells and HCT116 cells after electroporation with siRNA for48 h (n = 3). Representative data were shown. TrxR1 and TrxR2 knockdown by siRNA led to a decrease in cellular TrxR activity and cell survival, n = 3 independent experiments. Representative data were shown.
The number in each panel shows % of viable cells. For a-c, data are presented as mean ± SD; *p<0.05, **p<0.01. selenocysteine residuals in the active site of TrxR1disrupts the electron transport process, leading to a leak of electrons, which could be captured by oxygen to produce superoxide. Measurement of ROS in a cell-free system, detected by luminescence in the presence of NADPH, NADPH + AF, NADPH + AF+ Trx, NADPH + AF + TrxR1, NADPH + Trx + TrxnR1, or NADPH + AF + Trx +TrxnR1as indicated (n = 3). For a and b, data are presented as mean ± SD. Figure 7