Gossypol Increases Expression of the Pro-apoptotic BH3-only Protein NOXA through a Novel Mechanism Involving Phospholipase A2, Cytoplasmic Calcium, and Endoplasmic Reticulum Stress*

Background: The mechanism through which gossypol mediates anti-cancer activity is poorly understood. Results: Gossypol rapidly activates phospholipase A2, increases cytoplasmic calcium, endoplasmic reticulum stress, and NOXA, and sensitizes cells to apoptosis. Conclusion: Phospholipase A2 is a novel target of gossypol. Significance: This pathway can explain many reported activities of gossypol including sensitization to the BCL2 inhibitor, ABT-199. Gossypol is a putative BH3 mimetic proposed to inhibit BCL2 and BCLXL based on cell-free assays. We demonstrated previously that gossypol failed to directly inhibit BCL2 in cells or induce apoptosis in chronic lymphocytic leukemia (CLL) cells or platelets, which require BCL2 or BCLXL, respectively, for survival. Here, we demonstrate that gossypol rapidly increased activity of phospholipase A2 (PLA2), which led to an increase in cytoplasmic calcium, endoplasmic reticulum (ER) stress, and up-regulation of the BH3-only protein NOXA. Pretreatment with the PLA2 inhibitor, aristolochic acid, abrogated the increase in calcium, ER stress, and NOXA. Calcium chelation also abrogated the gossypol-induced increase in calcium, ER stress, and NOXA, but not the increase in PLA2 activity, indicating that PLA2 is upstream of these events. In addition, incubating cells with the two products of PLA2 (lysophosphatidic acid and arachidonic acid) mimicked treatment with gossypol. NOXA is a pro-apoptotic protein that functions by binding the BCL2 family proteins MCL1 and BFL1. The BCL2 inhibitor ABT-199 is currently in clinical trials for CLL. Resistance to ABT-199 can occur from up-regulation of other BCL2 family proteins, and this resistance can be mimicked by culturing CLL cells on CD154+ stroma cells. We report here that AT-101, a derivative of gossypol in clinical trials, overcomes stroma-mediated resistance to ABT-199 in primary CLL cells, suggesting that a combination of these drugs may be efficacious in the clinic.

The evasion of apoptosis is an established hallmark of cancer and is frequently mediated by the deregulation of BCL2 3 proteins (1). Intrinsic apoptosis is regulated at the mitochondrial membrane by a balance between anti-apoptotic (e.g. BCL2, BCLXL, MCL1, BFL1) and pro-apoptotic (BAX and BAK) BCL2 family members. In addition, BH3-only proteins (e.g. NOXA, PUMA, BIM, BAD) respond to a variety of cellular stress by inhibiting the anti-apoptotic BCL2 family members, thus tipping the cell toward apoptosis. The up-regulation of anti-apoptotic BCL2 family members gives cancer cells a survival advantage and is a frequent event in leukemias such as chronic lymphocytic leukemia (CLL).
Many compounds termed BH3 mimetics have been developed to inhibit anti-apoptotic BCL2 family members by occupying the BH3 binding pocket, with the goal of selectively killing cancer cells. The BH3 mimetic, ABT-737, is a potent inhibitor of BCL2 and BCLXL, but not of other anti-apoptotic BCL2 family members. A related, orally bioavailable compound, navitoclax (ABT-263), has completed phase I clinical trials in CLL (2) and small cell lung cancer (3). Although this compound has demonstrated efficacy, resistance can occur when cancer cells rely on alternative BCL2 family members, such as MCL1 and BFL1 (4). Therefore, additional compounds are needed which inhibit MCL1 and BFL1.
Gossypol is a naturally occurring polyphenol first isolated from the cotton plant (the genus Gossypium) in 1899 by Marchlewski (5). While early studies focused on the toxic effects of gossypol in animals, a correlation was noted between male infertility and increased use of cotton oil for cooking. This finding led to the first in-human use of gossypol as a male contraceptive (6). However, due to toxicity and persistent infertility after cessation of treatment, gossypol has fallen out of favor as a male contraceptive (7). In addition, gossypol has been reported to inhibit topoisomerase I and II, various steroid dehydrogenases, telomerase, calcineurin phosphatase, lactate dehydrogenase, aromatase, ribonucleotide reductase, arachidonate 5-and 12-lipoxygenases, adenylate cyclase, and catechol-Omethyltransferase (8 -19). There are also many reports claiming that gossypol can function as a BH3 mimetic. Specifically, gossypol is reported to inhibit BCL2 and BCLXL (20,21). Gossypol has demonstrated anti-tumor efficacy in a variety of systems, including cancer cell lines, mouse xenograft models, and ongoing clinical trials (22,23). In addition, gossypol can overcome resistance to the BCL2 inhibitor ABT-737 in a clinically relevant CLL-stroma co-culture system (24). Although gossypol is a proposed BCL2/XL inhibitor, these conclusions are based on results from cell-free assays, and whether this mechanism is involved in the anti-tumor effects of gossypol has not been established.
We reported previously that several putative BH3 mimetics, including gossypol, activate the integrated stress response (ISR) rather than directly inhibiting the anti-apoptotic BCL2 proteins (25). The ISR involves phosphorylation of eIF2␣ followed by the induction of ATF4, ATF3, and ultimately the BH3-only protein NOXA, which in turn can inhibit the BCL2 family members MCL1 and BFL1. Given the efficacy of gossypol in the aforementioned CLL-stroma co-culture system, we elucidated the mechanism by which gossypol induces ISR and NOXA. Here we identify phospholipase A2 (PLA2) as a critical and rapidly activated target of gossypol, which leads to an increase in cytoplasmic calcium, activation of ER stress, and the induction of NOXA.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-The leukemia cell lines NB4, THP1, and U937 were used as described previously (25). Blood from patients with CLL or healthy individuals was obtained from consenting donors at the Norris Cotton Cancer Center (Lebanon, NH). Lymphocytes were purified using Ficoll-Paque PLUS as described previously (26) and immediately incubated with the experimental compounds. NB4 and CLL cells were cultured in RPMI 1640 medium supplemented with 10% (v/v) inactivated fetal bovine serum and 100 units/ml antibiotic and antimycotic. For stroma co-culture experiments, CLL cells were plated on a confluent monolayer of CD154 ϩ stroma cells for 24 h and then incubated with compounds for 6 h as published previously (24). ABT-737 was obtained from AbbVie (Abbott Park, IL), ABT-199 was obtained from Selleck Chem (Houston, TX). BAPTA-AM and thapsigargin were purchased from Enzo Life Sciences (Lausen, Switzerland). AT-101 was purchased from Sigma-Aldrich. Lysophosphatidic acid (LPA) and arachidonic acid (AA) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Bortezomib was obtained from LC Laboratories (Woburn, MA). All compounds were dissolved in dimethyl sulfoxide except LPA, which was dissolved in water.
Cytoplasmic Calcium Assay-Cells were incubated with 1 M INDO-1 AM (Molecular Probes) for 1 h and then analyzed for calcium concentrations using a SpectraMax M2 plate reader. INDO-1 AM has an absorption/emission optimum of 330 nm/400 nm, respectively, when bound to calcium, and 356 nm/478 when calcium is not bound. The cytoplasmic calcium concentration was calculated based on the florescent ratio of INDO-1 as first described by Grynkiewicz et al. (28). PLA2 Activity Assay-The EnzChek Phospholipase A2 Assay Kit (Molecular Probes) was used according to the manufacturer's protocol. Briefly, cells were incubated for 5 min with fluorescent PLA2 substrate (1-O-(6-BODIPY 558/568-aminohexyl)-2-BODIPY FL C5-sn-glycero-3-phosphocholine) reconstituted in liposomes consisting of dioleoylphosphatidylcholine and dioleoylphosphatidylglycerol. PLA2 activity was analyzed in a 96-well plate with a SpectraMax M2 plate reader, and activity was quantified based on a shift in emission fluorescence of the substrate from 575 to 515 nm following cleavage by PLA2. The PLA2 activity was expressed as a FRET ratio (515/ 575). The excitation wavelength used was 460 nm, and the background fluorescence was subtracted for each run. After a 30-s baseline reading, the run was halted, gossypol was added, and the run resumed resulting in a 10-s delay following gossypol addition (represented by the dotted line).
Chromatin Condensation-One of the earliest detectable events in apoptosis is condensation of nuclear chromatin, and this is particularly evident in leukemia cells. To quantify apoptosis, cells were incubated with 2 g/ml Hoechst 33342 for 10 min at 37°C. Cells were analyzed by microscopy, and apoptosis was calculated as the percentage of cells with condensed chromatin as published previously (24).

NOXA Is Required for Gossypol-mediated Sensitization to
ABT-737-We have reported previously that gossypol induces NOXA and sensitizes the acute pro-myelocytic leukemia line NB4 to ABT-737 (25), and these findings are extended here (Fig. 1). A 6-h incubation of NB4 cells with gossypol was sufficient to induce NOXA protein, with maximal induction occurring at 20 M (Fig. 1A). Using caspase-mediated cleavage of PARP as a read-out for apoptosis, 20 M gossypol significantly sensitized NB4 cells to ABT-737 (Fig. 1B). This occurred at concentrations of gossypol and ABT-737 which did not induce apoptosis as single agents. We have also observed NOXA induction following treatment with gossypol in additional leukemia cell lines (THP1, U937, K562, AML3, AML4, KMS12, and RPMI 8266 medium) and primary CLL cells (24,25). We used siRNA to determine whether NOXA is required for the observed apoptosis. The siRNA completely prevented the gossypol-mediated induction of NOXA and prevented apoptosis induced by the combination of gossypol and ABT-737 (Fig. 1C). Hence, the induction of NOXA is a critical event in the induction of apoptosis by this drug combination.
Gossypol Triggers an ER Stress Response and Induces NOXA in an ATF4/ATF3-dependent Manner-We and others have found previously that many putative BH3 mimetics trigger the ISR, which involves the transcription factors ATF4 and ATF3 leading to the induction of NOXA (25,29,30). We also established that one such BH3 mimetic, S1, triggers the ISR pathway through the canonical ER stress, which included activation of the eIF2␣ kinase, PERK (30). Upon incubation with gossypol, we observed activation of the ER stress pathway, including autophosphorylation of PERK (Thr-981) and eIF2␣ followed by induction of ATF4 and ATF3 (Fig. 1D), suggesting that PERK is a mediator of this response.
We next determined whether ATF4 and ATF3 are required to induce NOXA following incubation with gossypol. It has been shown that both ATF4 and ATF3 must be silenced to prevent the induction of NOXA through the ER stress response (30,31). Therefore, we co-transfected NB4 cells with siRNA against ATF4 and ATF3 and found that the induction of NOXA was inhibited (Fig. 1E). Bortezomib also induces NOXA through ATF4/3 (30,31) and was therefore used as a positive control.
Gossypol Increases Cytoplasmic Calcium, Which Is Required for NOXA Induction-Many cellular stresses can activate the ER stress response including proteasome inhibition and increases in reactive oxygen species (31,32). Because the activation of ER stress following treatment occurs rapidly (Ͻ2 h), whatever stress links gossypol to this response must also occur rapidly. Several reports have demonstrated that gossypol causes a rapid increase in cytoplasmic calcium as a consequence of release of calcium from ER stores, and subsequent influx through calcium release-activated channels (33)(34)(35). In addition, the release of calcium from ER stores and increases in cytoplasmic calcium are known activators of the ER stress response (36). Although these initial reports elucidated part of the mechanism of action of gossypol, we determined whether this effect on calcium is responsible for triggering ER stress and NOXA induction. Using INDO-1 AM as an indicator of cytoplasmic calcium, we observed an increase in calcium concentration almost immediately following addition of gossypol to the cells ( Fig. 2A). This increase in calcium was larger and more persistent than thapsigargin, which was used as a positive control.
Next, we used the intracellular calcium chelator BAPTA AM to test whether NOXA induction by gossypol is calcium-dependent. Indeed, using a titration of BAPTA AM, we observed a concentration-dependent decrease in ATF4, ATF3, and NOXA induction by gossypol (Fig. 2B). Taken together, these data show that gossypol rapidly increases cytoplasmic calcium, followed by activation of the ER stress response, and induction of NOXA.
Gossypol Rapidly Increases Phospholipase A2 Activity, Leading to Increased Cytoplasmic Calcium, ER Stress, and NOXA Induction-A previous report demonstrated that aristolochic acid, an inhibitor of PLA2, prevented the increase in cytoplasmic calcium induced by gossypol, but whether gossypol actually increases PLA2 activity was not assessed (33). Moreover, many reports have demonstrated multiple levels of cross-talk between PLA2 and calcium, and in the majority of cases, an increase in cytoplasmic calcium is required for the increase in PLA2 activity (37). In contrast, two products of PLA2, LPA and AA, have been shown to trigger the release of calcium from the ER (via LPA receptors) and the influx of extracellular calcium (via the arachidonate-regulated calcium channel), respectively (38,39). We used a FRET-based PLA2 activity assay to test whether gossypol increases PLA2 activity in cells (40). After a 30-s baseline read, gossypol was added, resulting in a 10-s delay (dotted lines), after which the PLA2 assay was resumed. Gossypol triggered an immediate and concentration-dependent increase in PLA2 activity (Fig. 3A). Repeated measurements made following a 30-s incubation with gossypol are summarized in Fig. 3B. The rapid kinetics are similar to the kinetics of additional PLA2 enzymes (41). Importantly, the increase in PLA2 activity occurs even more rapidly than the increase in cytoplasmic calcium, suggesting PLA2 might lie upstream in this pathway. Gossypol also increased PLA2 activity and cytoplasmic calcium concentrations in THP1 and U937 cells, confirming that these events are not unique to NB4 cells (Fig. 3G).
To confirm the functional link among gossypol, PLA2, increased cytoplasmic calcium, and NOXA induction in NB4 cells, we determined whether aristolochic acid prevents the gossypol-induced PLA2 activity. We observed a concentrationdependent decrease in gossypol-induced PLA2 activity, which achieved significance at 50 M aristolochic acid (Fig. 3C). In addition, pretreatment with aristolochic acid also suppressed the increase in cytoplasmic calcium induced by gossypol (Fig.  3D). To confirm that PLA2 lies upstream of NOXA induction, we incubated NB4 cells with aristolochic acid for 0.5 h, followed by incubation with gossypol (6 h). Indeed, there was a decrease in NOXA induction by gossypol with increasing concentrations of aristolochic acid (Fig. 3E). Taken together, these data show that gossypol rapidly increases PLA2 activity, which leads to an increase in cytoplasmic calcium and NOXA induction.
The PLA2 family of enzymes contains many members, which can be divided into 14 different groups based on functional similarity (37). The expression levels of these different groups vary greatly among species and cell type. We were interested in determining which PLA2(s) might be activated by gossypol and which might serve as the effector molecule(s) that link gossypol to increased calcium, ER stress, and NOXA. Upon reanalysis of prior microarray data from NB4 cells (25), we observed detectable mRNA expression of PLA2 genes from groups IB, IIA, III, IV, V, VI, and X. Although this reduces the number of PLA2 genes that might be targeted by gossypol, we used additional experiments to narrow the search further. Specifically, all of the PLA2 groups require calcium for activity except groups VI, VII, and VIII (37). Because PLA2 activation appeared to occur upstream of the increase in cytoplasmic calcium, we hypothesized that gossypol might function by activating a calcium-independent PLA2. We chelated extracellular calcium, intracellular calcium, or both with EGTA, BAPTA AM, or EGTA plus BAPTA AM, respectively, and found that gossypol still increased PLA2 activity in all cases (Fig. 3F), suggesting that the observed increase in PLA2 activity following gossypol treatment is due to activation of calcium-independent PLA2. These data are consistent with PLA2 activation lying upstream of the gossypol-induced increase in calcium.

Lysophosphatidic Acid and Arachidonic Acid Mimic the Effects of Gossypol by Increasing Cytoplasmic Calcium and
Inducing NOXA Expression Independent of PLA2-Cleavage of phospholipids by PLA2 generates LPA and AA. To further test whether PLA2 activation can increase cytoplasmic calcium, induce ER stress, and NOXA, we tested whether treatment with LPA, AA, or co-treatment with both mimics treatment with gossypol. Indeed, LPA, AA, or LPA plus AA triggered an increase in cytoplasmic calcium, with slightly different kinetics and magnitudes (Fig. 4A). Importantly, this increase in cytoplasmic calcium was accompanied by an increase in NOXA, which was most evident following co-treatment with LPA plus AA (Fig. 4B). In addition, co-treatment of NB4 cells with LPA and AA also induced ER stress, which further mimics treatment with gossypol (Fig. 4C). Finally, to rule out the possibility of off-target effects of aristolochic acid, we tested whether the combination of LPA and AA would increase cytoplasmic calcium and NOXA independently of PLA2. Pre-treatment with aristolochic acid did not prevent the increase in cytoplasmic calcium or increase in NOXA induced by the LPA plus AA FIGURE 3. Gossypol increases PLA2 activity, which is required for increased calcium and NOXA. A, NB4 cells were incubated with PLA2 substrate for 5 min and then analyzed for PLA2 activity. After a 30-s baseline read, the run was halted, gossypol was added, and the run was restarted after a 10-s delay (dotted line).  combination (Fig. 4, D and E). These data further support the hypothesis that gossypol rapidly increase PLA2 activity, which in turns leads to increases in cytoplasmic calcium and NOXA via LPA and AA.
Gossypol Derivatives Also Increase PLA2, Cytoplasmic Calcium, and NOXA and Overcome Stroma-mediated Resistance to ABT-199-Many derivatives of gossypol have been generated and tested for anti-cancer properties. We have reported previously that another gossypol derivative, apogossypol, also induces ER stress and NOXA (25). In addition, the R-(Ϫ)-gossypol racemer (AT-101) is currently in clinical trials (42). To determine whether this mechanism of action of gossypol is a class effect, we determined whether apogossypol and AT-101 increase PLA2 activity, cytoplasmic calcium, and NOXA. Indeed, we observed activation of all of these markers following treatment with each compound, albeit at different concentra-tions (Fig. 5, A-C). These findings demonstrate that this mechanism is a class effect of several gossypol family members and further call into question claims that these gossypol derivatives function as BH3 mimetics in cells.
We have demonstrated previously that gossypol can overcome stroma-mediated resistance to a BCL2 inhibitor (ABT-737) in a co-culture model system (24). Although these results are promising, gossypol has yet to be combined with a BCL2 inhibitor in a clinical setting. Furthermore, the clinical focus has shifted from ABT-737 to the BCL2-specific inhibitor, . Therefore, we tested whether AT-101 would overcome stroma-mediated resistance to ABT-199. Similar to results with ABT-737 and gossypol, AT-101 overcame stromamediated resistance to ABT-199 (Fig. 5D). These promising results suggest that co-treatment with ABT-199 and AT-101 may have pronounced efficacy in a clinical setting. In summary, these data show that multiple gossypol derivatives exhibit similar signaling involving PLA2, calcium, and NOXA and may provide therapeutic benefit when combined with the BCL2 inhibitor ABT-199.

DISCUSSION
The naturally occurring polyphenol, gossypol, has been tested for a variety of indications ranging from male contraception to chemotherapy. We have identified a novel pathway by which gossypol activates PLA2 enzymes, increases cytoplasmic calcium, activates the ER stress response, and induces NOXA. Importantly, this induced NOXA is required to sensitize cells to the BCL2 inhibitor ABT-737. In addition, the products of PLA2 (LPA and AA) mimic treatment with gossypol, leading to an increase in calcium and NOXA.
The mechanism(s) of PLA2 activation are complex, vary depending on the specific class of PLA2, and in some cases are poorly understood (37,44). This multistep activation process involves the formation of an "active" PLA2, association of the enzyme with a lipid bilayer, recognition of the phospholipid substrate, cleavage of substrate at the sn-2 position, release of products, dissociation of the enzyme-lipid complex, and regeneration of the active enzyme (45). Many factors have been shown to impact activation of PLA2, including phosphorylation by kinases such as p38 and ERK, increased cytoplasmic calcium, and lipid membrane composition (40,44,46). Therefore, there are many steps in this process where gossypol may act to increase PLA2 activity. The fact that the activation occurs in seconds and is not inhibited by calcium chelation suggests that it is unlikely that gossypol increases activity via activation of an intermediate kinase (such as ERK) or through changes in calcium. Although the exact mechanism remains elusive, we hypothesize that gossypol may serve as a "tether," which facilitates the binding of PLA2 to the lipid membrane. In this model, the lipophilic portion of gossypol inserts into the lipid bilayer, whereas the opposing face of the molecule (which contains two aldehyde and six hydroxyl groups) interacts with PLA2 via hydrogen bonding. Support for this model includes the fact that gossypol has been shown to increase the binding of another PLA2 (sPLA2) to membranes (47) and has also been shown to alter lipid membrane composition (48,49). Further studies are required to identify definitively the mechanism by which gossypol activates PLA2.
Gossypol is still commonly described as a BCL2/BCLXL inhibitor (50). We and others have generated substantial data showing that gossypol does not directly inhibit BCL2 proteins in cells. For instance, we have shown that gossypol does not disrupt the binding of BIM or BAD to BCL2 in cells (25). Furthermore, gossypol does not induce apoptosis as a single agent in either CLL cells or platelets, which are dependent on BCL2 or BCLXL, respectively, for survival (24). In addition we observed PLA2 activation, increased calcium, ER stress, and NOXA following treatment with two additional gossypol derivatives (apogossypol and AT-101). This suggests that activation of this pathway is common to many gossypol derivatives. Strikingly, AT-101 was the most potent activator of this pathway and overcame stroma-mediated resistance to ABT-199. This is of clinical significance due to the ongoing trials of both ABT-199 and AT-101 and suggests that a combined trial with both drugs may demonstrate additional efficacy.
There are many reported effects of gossypol, and we performed a literature search to determine how many of these effects might be explained by either PLA2 activity, increased cytoplasmic calcium, or both. Gossypol has been reported to inhibit many enzymes as described in the Introduction. Many of these reports are based on results in cell-free assays, and only steroid dehydrogenases, telomerase, and aromatase have been validated in cells. Inhibition of these enzymes can be explained by either PLA2 activity or increased cytoplasmic calcium. For example, AA has been shown to inhibit 3␣-hydroxysteroid dehydrogenase (51), whereas an increase in calcium concentration has been shown to inhibit telomerase and aromatase (52,53).
Several cell signaling pathways and cellular processes are increased by gossypol, including oxidative stress, autophagy, expression of death receptor 5 (DR5), and c-Jun N-terminal kinase (JNK) activity (54 -57). Many of these events can also be explained by PLA2 activity, increased calcium, or both. Specifically, AA or increased calcium can induce oxidative stress, autophagy, and JNK activity (58 -63). In addition, calcium signaling has been shown to increase DR5 expression (29).
Gossypol has also been shown to inhibit NFB signaling, the electron transport system, and glucose transporter 1 (GLUT1) in cells (64 -67). However, an increase in PLA2 activity may also explain some of these responses. For instance, AA has been shown both to suppress NFB activity by stabilizing the inhibitor of NFB (IB) and interfere directly with the electron transport chain (68,69). Conversely, gossypol was shown to interact directly with GLUT1 at the endofacial surface (67). This interaction between gossypol and GLUT1 at the cell membrane is similar to the proposed interaction between gossypol and PLA2 at the membrane. Hence, although many of the effects of gossypol may be mediated by activation of PLA2, some may be the consequence of its recruitment of other proteins to the membrane.
In summary, we have identified a novel mechanism of gossypol action involving activation of PLA2, increased cytoplasmic calcium, activation of the ER stress pathway, and induction of NOXA. These effects, coupled with prior findings refuting the function of gossypol as a true BH3 mimetic, suggest that gossypol should be repurposed as a potent activator of PLA2. Importantly, the PLA2-mediated induction of NOXA provides a novel therapeutic strategy to overcome resistance to the BCL2 inhibitors ABT-737 and ABT-199. In addition, these findings support the hypothesis that activation of PLA2 and the subsequent increase in calcium may be the primary mechanism through which gossypol functions in cells. Both activation of PLA2 and increased cytoplasmic calcium can impact a broad range of cellular signaling pathways and processes. Therefore, we speculate that many of the published effects of gossypol treatment are merely a consequence of the rapid effects on PLA2 and calcium. Perhaps after over a century of research, PLA2 is the long sought target of gossypol which explains its plethora of cellular effects.