Antagonistic effects of nitric oxide in a glioblastoma photodynamic therapy model:mitigation by BET bromodomain inhibitor JQ1

Endogenous nitric oxide (NO) generated by inducible NO synthase (iNOS) promotes glioblastoma cell proliferation and invasion, and also plays a key role in glioblastoma resistance to chemotherapy and radiotherapy. Non-ionizing photodynamic therapy (PDT) has anti-tumor advantages over conventional glioblastoma therapies. Our previous studies revealed that glioblastoma U87 cells upregulate iNOS after a photodynamic challenge and that resulting NO not only increased resistance to apoptosis, but rendered surviving cells more proliferative and invasive. These findings were largely based on the effects of inhibiting iNOS activity and scavenging NO. Demonstrating now that iNOS expression in photostressed U87 cells is mediated by NF-κB, we hypothesized that (i) recognition of acetylated lysine (acK) on NF-κB p65/Rel A by bromodomain and extra-terminal (BET) protein Brd4 is crucial, and (ii) by suppressing iNOS expression, a BET inhibitor (JQ1) would attenuate the negative effects of photostress. The following evidence was obtained: (i) Like iNOS, Brd4 protein and p65-acK levels increased several fold in photostressed cells; (ii) JQ1 at minimally toxic concentrations had no effect on Brd4 or p65-acK upregulation after PDT, but strongly suppressed iNOS, survivin, and Bcl-xL upregulation, along with the growth and invasion spurt of PDTsurviving cells; (iii) JQ1 inhibition of NO production in photostressed cells closely paralleled that of growth/invasion inhibition; (iv) At 1% the concentration of iNOS inhibitor 1400W, JQ1 reduced post-PDT cell aggressiveness to a far greater extent. This is the first evidence for BET inhibitor targeting of iNOS expression in cancer cells and how such targeting can markedly improve therapeutic efficacy.


INTRODUCTION
Most established malignant tumors exist under moderate inflammatory conditions, which foster tumor cell survival, proliferation, and metastatic expansion (1)(2)(3). Low level reactive oxygen species (ROS) as well as nitric oxide (NO) generated by inducible nitric oxide synthase (iNOS/NOS2) play important roles in many of these processes (3)(4)(5)(6). There is now compelling evidence that endogenous iNOS/NO not only supports growth and progression of many tumors, but also plays a key role in pro-tumor immunosuppression (7,8) as well as resistance to chemotherapeutic and radiotherapeutic interventions (9)(10)(11). A less common intervention for solid tumors is photodynamic therapy (PDT), which employs non-ionizing radiation. PDT was introduced about 45 years ago as a minimally invasive modality involving a photosensitizing agent (PS), PS-exciting visible-to-near infrared light, and molecular oxygen (12)(13)(14). All three components (PS, light and O 2 ) must be engaged concurrently for PDT to lethally damage tumor cells, which often occurs via formation of the cytotoxic ROS, singlet oxygen ( 1 O 2 ). Lightindependent PS effects are usually negligible, and little, if any, damage to normal tissue occurs during PDT, which is not the case for many chemotherapeutic agents. Another advantage of PDT is site-specificity, i.e. limitation of photodynamic action to the tumor site at which light is directed, typically via fiber optic transmitters (13,14). An oligomeric hematoporphyrin preparation, now known as Photofrin ® , was the first PS to receive FDA approval for PDT ca. 20 years ago, and it is now used for a variety of solid tumors (13,14). 5-Aminolevulinic acid (ALA)-based PDT is a more recently developed alternate in which ALA (or an ALA ester) is administered as a pro-PS. ALA is metabolized to the actual PS, protoporphyrin IX (PpIX), via the heme biosynthetic pathway, PpIX accumulating initially in mitochondria (15,16). Since heme synthesis is enhanced in tumor cells, they can attain much higher levels of ALAinduced PpIX than surrounding normal cells (17), which for this type of PDT, provides a further element of tumor site specificity. The potential interference of NO with PDT was discovered by showing that Photofrin ® -PDT (18,19) or ALA-PDT (20) cure rates for various mouse-borne tumors could be significantly increased by administering NOS inhibitors, particularly for tumors with relatively high basal NO outputs. The proffered explanation was that NO-mediated dilation of tumor micro-vasculatures was acting in opposition to the vasoconstrictive effects of PDT (19,20). However, until relatively recently, many questions remained unanswered, e.g. the NOS isoform(s) involved and its/their cellular source(s).
In previous work, we showed that NO from endogenous iNOS in various human cancer lines (breast, prostate, glioblastoma) subjected to an ALA-PDT-like challenge elicited the following negative responses: (i) increased resistance to apoptotic photokilling, and (ii) increased proliferative, migratory, and invasive aggressiveness for cells surviving the challenge (21)(22)(23)(24)(25)(26). Most of this evidence was based on the strong counteractive effects of iNOS enzyme inhibitors such as 1400W and GW274150 (27,28) or the NO scavenger cPTIO (29). Using human glioblastoma cells in the present study, we determined that basal and photostress-induced iNOS is regulated by nuclear factor-kappa B (NF-κB). Knowing this and projecting from recently published evidence (30,31), we hypothesized that bromodomain and extra-terminal (BET) protein recognition of ε-N-acetylated lysine residue(s) (acK) on the NF-κB p65/RelA subunit played a key role in iNOS expression. BET family proteins (Brd2, Brd3, Brd4, and Brdt) 'read' acK residues on histones and transcription factors (32,33). Brd4 is important in cancer progression (30,31,33), but the role of iNOS in Brd4-mediated cancer progression has not been described previously. In testing our hypothesis, we found that Brd4 is a key co-activator of photostress-augmented iNOS expression. Blocking Brd4 with the BET bromodomain inhibitor JQ1 (34,35) substantially reduced acquired cell aggressiveness, and to a much greater extent than an iNOS enzymatic inhibitor at many times greater concentration. These and related findings demonstrate for the first time that suppressing iNOS expression via Brd4 inhibition can markedly increase the efficacy of an anti-tumor therapy, in this case PDT. We demonstrate this using a model system for glioblastoma PDT.

JQ1 Suppression of iNOS/NO Anti-PDT Effects
PDT by itself produced about the same level of apoptosis as 0.5 µM JQ1, but when combined with PDT, JQ1 enhanced apoptosis significantly such that a synergistic effect was apparent. For example, PDT combined with 0.3 µM JQ1 caused 50% apoptosis, whereas PDT alone and JQ1 alone caused 30% and 10% apoptosis, respectively (Fig.  1B). We chose 0.3 µM JQ1 for all subsequent experiments, since this was minimally cytotoxic by itself, thus emphasizing the ability of JQ1 to enhance PDT cytotoxicity. In a previous study (26), we showed that 25 µM 1400W, a competitive inhibitor of iNOS enzymatic activity, increased U87 apoptosis by ~33% over PDT alone. As shown in Fig. 1B, JQ1 at a much lower concentration (0.3 µM) promoted apoptosis much more substantially, i.e. by ~66%. This recognition was a strong impetus for studying the mechanism of action of JQ1 in the context of PDT.

JQ1 inhibition of iNOS expression -
We showed previously that a PDT oxidative challenge results in prolonged upregulation of pro-survival iNOS in several cancer cell lines, including glioblastoma lines (21)(22)(23)(24)(25)(26). Given that NF-κB is often implicated in iNOS expression (23,36,37) and that Brd4 can serve as a NF-κB co-activator (30,31), we asked whether the observed JQ1 enhancement of PDT cytotoxicity could be explained on this basis. We looked first at the effects of low level JQ1 on U87 iNOS expression after a PDT challenge compared with that in a dark control. As shown by the immunoblot in Fig.  2A, control (ALA-only) cells expressed iNOS protein at a relatively low level, and this was no different from that in untreated cells (not shown). After irradiation, ALA-primed cells exhibited a rapid and prolonged upregulation of iNOS over at least a 24 h post-irradiation period ( Fig. 2A). JQ1 (0.3 µM) strongly inhibited basal iNOS expression, as well as the robust induction of enzyme by photodynamic stress (Fig. 2B). For example, at 6 h after irradiation, the iNOS level for PDT plus JQ1 was only ~40% of the iNOS level for PDT alone. Thus, JQ1 markedly reduced basal as well as stress-induced iNOS, the latter being associated with a strong pro-survival response in stressed cells and a switch to a more aggressive phenotype in surviving cells (21)(22)(23)(24)(25)(26).

Generation of NO and suppression thereof by JQ1 -
We used the fluorophore DAF-FM-DA to probe for NO-derived oxidant levels in photodynamically stressed U87 cells and how NO levels might be altered by JQ1 and 1400W. Fluorescence of the NO-derived triazole product of the probe (DAF-FM-T) (38), was monitored. Cells treated with DAF-FM-DA directly or after addition of the JQ1 vehicle (DMSO) exhibited the same, relatively low probe fluorescence after 5 h and 18 h of pre-incubation (Fig. 3). The intensity of this background fluorescence was significantly reduced by 1400W, implying detection of iNOSderived NO. JQ1 at a small fraction of the 1400W concentration reduced the DAF-FM-T signal more substantially than 1400W, i.e. by ~80% at 18 h relative to control (Fig. 3). A striking 10-to 12fold increase in DAF-FM-T fluorescence was observed at 5 h and 18 h after cells underwent a PDT (ALA/light) challenge (Fig. 3). Whereas JQ1(-), the enantiomer of JQ1 that does not bind BET bromodomains (34), had no effect on the level of photostress-generated NO, active JQ1 reduced NO levels to ~14% of the ALA/light value. A large decrease in NO output was also observed with 1400W, i.e. to ~30% of the ALA/light value. Thus, the effect of 1400W was not as impressive as that of JQ1 at a far lower starting concentration. Therefore, limiting NO by presumed JQ1 inhibition of iNOS expression appeared more effective in increasing PDT cytotoxicity than limiting NO by inhibiting iNOS enzyme activity.
The results in Fig. 3 were confirmed by measuring NO-derived NO 2 expression in U87 cells, but also PDT-upregulated iNOS, suggesting control by NF-κB. As further evidence, we found that PDT (ALA/light) resulted in complete translocation of the p65/RelA subunit of NF-κB from the cytosol to nucleus, and Bay11 prevented p65 translocation (Fig. 4B). On the other hand, JQ1 had little (if any) inhibitory effect on p65 translocation, which is consistent with JQ1 acting in the nucleus (33,34).

Upstream events in photostress induction of iNOS -
We learned previously that human breast cancer COH-BR1 cells underwent a rapid and robust phosphorylation-activation of the prosurvival/progression kinase Akt after an ALA/light challenge and that this subsided during prolonged post-irradiation incubation (23). Akt activation was nullified by Wortmannin, an inhibitor of PI3K, which is required for Akt activation, and this also prevented iNOS induction. These and related findings, e.g. inability of 1400W to inhibit Akt activation, suggested that Akt was an upstream mediator of iNOS induction through phosphorylation-activation of IKK and thence activation and nuclear translocation of NF-κB (23). To assess whether Akt might function similarly in photostressed glioblastoma cells, we monitored its phosphorylation status after ALA/light treatment and how this might be affected by JQ1. As shown in Fig. 5A, Akt was strongly activated in U87 cells (appearance of p-Akt band) over 3-6 h after irradiation, total Akt remaining the same throughout. The PI3K inhibitor LY294002 prevented this activation (Fig.  5C), but JQ1 had no effect on it (Fig. 5B), nor did 1400W (not shown). LY294002 also prevented photostress upregulation of iNOS (Fig. 5D), implying that upstream activation of PI3K and Akt was necessary for iNOS induction, as observed previously for COH-BR1 cells (23). These results ruled out any possible JQ1 impairment of upstream signaling events leading to iNOS upregulation, including Akt activation.
Elevated Brd4 and p65-acK310 levels after photodynamic stress -We discovered that like iNOS, Brd4 was strongly upregulated after subjecting U87 cells to photodynamic action. As shown by the Western blot in Fig. 6B, Brd4 level increased progressively after ALA/light treatment, reaching nearly 3-fold above the dark control basal level 24 h after irradiation. In contrast, there was little, if any, upregulation of another BET family member, Brd2, after the same PDT challenge (Fig.  6A). On the other hand, the level of acetylated lysine 310 (acK310) on the p65 subunit of NF-κB exhibited a substantial increase after PDT, rising to ~3-fold over background after 24 h (Fig. 6C), which was similar to the elevation in Brd4. Total p65 expression was not altered by PDT (results not shown), so the observed acK310 response must have been due to more extensive acetylation at this particular lysine residue. As shown in Fig. 6B, JQ1 had no effect on basal or photostress-induced Brd4 protein level (Fig. 6 B) or on p65-acK310 level (Fig. 6C). This rules out any inhibition of Brd4 expression or extent of p65 lysine 310 acetylation as a possible factor in iNOS/NO suppression by JQ1 (Figs. 2 and 3).

JQ1-inhibitable interaction of Brd4 and NF-κB/p65 in PDT-stressed U87 cells -
Knowing that photostress induction of iNOS depended on nuclear translocation of NF-κB/p65 and that this was accompanied by upregulation of Brd4 and acK310 on p65, we postulated that Brd4 interaction with acK310 is necessary for NF-κB/p65 activation (30,31). To investigate this, we used a pull-down approach in which p65 was immunoprecipitated, collected on Protein A-linked Sepharose beads and, after release, checked for the presence of Brd4 by immunoblotting. As shown in Fig. 7A, the p65 immunoprecipitation revealed a strong Brd4 immunoblot band, the intensity of which was substantially reduced by JQ1. We deduced from this evidence that Brd4 served as a co-activator of NF-κB/p65 in photostressed cells and that JQ1 suppressed iNOS induction by targeting Brd4 and preventing its binding to p65-acK310. We went on to determine whether another BET protein, Brd2, might contribute to coactivation and possibly also be present in the p65 immunoprecipitate. As shown in Fig. 7B, a pulldown immunoblot band for Brd2 was detected, but it was very weak compared with Brd4, yet strong Brd2 and p65 bands were seen in the overall lysate. Although other BET proteins have not been interrogated similarly, we believe, in agreement with others using different cancer cells (30,31), that Brd4 was the predominant (if not sole) coactivator for iNOS expression/overexpression in our system.

JQ1 Suppression of iNOS/NO Anti-PDT Effects
to upregulated iNOS/NO (24)(25)(26), we asked whether this iNOS/NO might stimulate growth and invasion of cells that could withstand a photochallenge and, if so, how JQ1 would affect these responses. Twenty-four hours after ALA/light treatment, surviving U87 cells, along with ALA-only controls, were recovered, re-plated at equal live cell densities, and monitored for proliferation over 48 h in the absence vs. presence of 1400W (25 µM) or JQ1 (0.3 µM). As shown in Fig. 8, proliferation of non-irradiated control cells was slowed somewhat by 1400W and slightly more so by JQ1. On the other hand, surviving ALA/light-treated cells exhibited a sizeable growth spurt (~61% in 24 h) relative to ALA-only controls, and 1400W slowed this spurt much more than it did control cell growth (~50% vs. ~10%). However, JQ1 slowed surviving cell growth to an even greater extent than 1400W, i.e. by ~80%, and JQ1 did this at only ~1% the 1400W concentration in bulk cell system, making JQ1 more impressive than 1400W for pharmacologically arresting iNOS-stimulated proliferation.
In addition to exploiting iNOS/NO for proliferative signaling, glioblastoma cells are known to rely on NO for migratory and invasive potency (26,(42)(43)(44). We compared the effects of 1400W and JQ1 on surviving U87 cell invasiveness after a typical ALA/light challenge. Invasion measurements were started immediately after ALA-primed cells were irradiated, nonstressed controls being analyzed similarly. Fig. 9A shows that control cell invasion rate was only moderately inhibited by JQ1 or Bay11 (~15%), 1400W having a smaller effect and JQ1(-) no significant effect. Cells surviving PDT exhibited a striking 35-40% increase in invasion rate, which was unaffected by JQ1(-). However, JQ1 not only abrogated the more rapid invasion, but brought the remaining invasion rate to ~40% that of the vehicle control (Fig. 9A). Although 1400W also eliminated the more rapid invasion, the residual rate with 1400W was ~80% of the control rate, i.e. much greater than the rate left by JQ1. Thus, JQ1 inhibited invasion to a far greater extent than 1400W at >80-times the JQ1 concentration. As shown in Fig. 9A, Bay11, which strongly reduced iNOS expression via inhibition of NF-κB activation (Fig. 4), suppressed post-PDT invasiveness to nearly the same extent as JQ1. While further supporting the role of iNOS in hyper-invasiveness, this finding raises the issue of specificity because Bay11 can inhibit other progrowth/invasion effectors besides IκB kinase, e.g. protein tyrosine phosphatases (45).
We asked whether NO from an exogenous source might restore invasiveness that JQ1 had strongly suppressed. When photostressed and JQ1treated U87 cells were exposed to the NO donor DETA-NONOate, a concentration-dependent increase in invasion rate from a very low point was observed. This rate maximized at about twice that observed for cells exposed only to ALA, light, and JQ1 (Fig. 9B). Thus, although iNOS, the primary source of endogenous NO, was depleted (along with several other tumor-promoting effectors; see below), these cells were still able to respond to NO by becoming more invasive. The signaling mechanism behind this remarkable response remains to be elucidated.
As shown in Fig. S2, JQ1 also abolished PDTpromoted invasiveness in another human glioblastoma line, U251 cells. Similarly to U87 counterparts ( Fig. 9A), U251 cells exhibited a residual invasion rate after PDT/JQ1 (~35%), which was far below that of control cells or cells treated with 1400W after PDT (~80%). Thus, PDT survivors in at least two different glioblastoma cell lines exhibited greater invasiveness and this could be eliminated by JQ1.
As shown by the immunoblot in Fig. 10A, survivin, a potent inhibitor of apoptosis (46), underwent a time-dependent upregulation during post-PDT incubation, reaching ~2-times the ALAonly control level after 24 h. This control level was no different from that of untreated cells (not shown). A similar response was observed in our previous study (26). The presence of JQ1 not only suppressed basal survivin expression, consistent with others' results (47,48), but also the strong induction of survivin by PDT (Fig. 10A).
Bcl-xL is another NF-κB-regulated antiapoptotic protein that is highly expressed in glioblastoma cells (51). Bcl-xL underwent a gradual upregulation after PDT, reaching approximately twice the control level at 24 h (Fig.  10B). A similar response was observed previously for a breast cancer cell line (52). JQ1 strongly reduced Bcl-xL expression in U87 control cells, consistent with others' results (51), and also attenuated its post-PDT upregulation; e.g. 40% less at 24 h (Fig. 10B).
The responses of p21 were diametrically opposite to those of iNOS, survivin and Bcl-xL. Thus, the p21 level declined progressively over 24 h of dark incubation after PDT (Fig. 10C). A remarkable reversal of this response was observed when JQ1 was present such that p21 reached twice its control level 3 h after PDT and remained there for at least another 21 h, JQ1 alone (without PDT) producing a similar effect (Fig. 10C). Strong induction of p21 by JQ1 has been reported for several other cell lines (47,53,54). Such induction could promote cell death through arrest of cell cycle progression. The striking down-regulation of p21 observed after PDT is consistent with our evidence that surviving cells proliferated more rapidly (Fig. 8). That JQ1 not only inhibited p21 down-regulation, but elicited a long-lasting upregulation of this protein, could explain the strong suppression of this post-PDT growth spurt by JQ1 (Fig. 8). MMP-9, which can catalyze the degradation of extracellular matrices, is associated with the migratory/invasive characteristics of many tumor cells, including glioblastoma cells (55,56).
As shown in Fig. 10D, MMP-9 underwent a slow upregulation after PDT, reaching about 50% greater than its control level by 24 h. JQ1 strongly inhibited this response to photostress, reducing the MMP-9 level by nearly the extent that JQ1 did in control cells (Fig. 10D). In a previous study (26), we found that while U87 MMP-9 expression was only slightly elevated after PDT, MMP-9 activity assessed by in-gel zymography was increased by approximately 80%. This increase was nearly abolished by L-NAME or 1400W, thus implicating iNOS/NO in the MMP-9 activation (26).
The oncogenic protein c-Myc, which is constitutively expressed in a variety of aggressive tumors, including glioblastomas, acts not only as a transcription factor, but as a global regulator of pro-tumor epigenetic modifications (57). We found that c-Myc expression in U87 cells underwent a rapid decline after PDT, beginning immediately after irradiation, reaching a nadir at ~3 h, and then gradually rising so that the c-Myc level at 24 h approximated that of the nonirradiated control (Fig. 10E). JQ1 not only prevented this delayed return to background c-Myc expression after PDT, but by itself (without PDT) nearly abolished all c-Myc expression, in agreement with previous studies involving glioblastoma and other cell lines (51,58,59). The effects of different PDT approaches on c-Myc status have been described previously for other cancer cell types, often with contrasting results. In some cases, expressed c-Myc mRNA or protein steadily increased after PDT (60,61), while in other cases, it decreased (62-64), but no rational explanations were offered. In most of these studies, c-Myc was not monitored for more than 4-6 h after irradiation, whereas we tracked it over 24 h. The striking decline in c-Myc and return to constitutive level after 24 h (Fig. 10E) might reflect a unique stress accommodation response of this effector in preparation for accelerated cell division.
We conclude from these findings that although PDT stress-induced iNOS played a major role in promoting cell resistance and aggressiveness, altered expression of survivin, Bcl-xL, p21, MMP-9, and c-Myc made a significant contribution, which could be significantly counteracted by inhibition of BET bromodomains by JQ1.

DISCUSSION
Malignant gliomas such as glioblastoma multiforme (GBM) are among the most aggressive and lethal of the primary brain tumors. Without treatment, a patient's average survival time after initial diagnosis is typically 4-6 months (65,66). Even with the most advanced surgical techniques or surgery combined with radiotherapy or chemotherapy, survival time remains dismal at 18-24 months (66). Pre-existing or acquired resistance to conventional chemo-and radiotherapy remains a serious impediment to the benefits of these treatments, and this has stimulated development of better alternatives. One such alternative is PDT using Photofrin ® (67-69) or ALA-induced PpIX (69,70) as a photosensitizing agent. In addition to improving average survival time relative to cisplatin-based chemotherapy, for example (69), PDT has the advantage of high tumor-site specificity, i.e. fewer negative off-target effects on normal tissue (12)(13)(14). However, like other therapeutic interventions, PDT can be antagonized by pre-existing or stress-induced factors which could increase cell resistance to photokilling and/or provide surviving cells with a growth and migratory advantage.
One major antagonist of PDT's antitumor effects is iNOS-derived NO, as amply demonstrated in our previous studies on several human cancer cell lines, including glioblastoma lines (22)(23)(24)(25)(26). Three key findings emerged from these studies: (i) cancer cell iNOS undergoes a rapid and prolonged upregulation after a photodynamic (ALA/light) insult originating in mitochondria; (ii) stress signaling by upregulated iNOS/NO increases cell resistance to apoptotic photokilling, pre-existing iNOS typically being much less important in this regard; and (iii) induced iNOS/NO promotes growth and migration/invasion aggressiveness in cells withstanding the photodynamic stress (22)(23)(24)(25)(26). Highly specific inhibitors of iNOS enzymatic activity, viz. 1400W and GW274150, played a key role in our discovery of these anti-PDT responses. For example, 1400W increased the extent of apoptotic cell photokilling, but decreased the hyper-aggressiveness of surviving cells (22)(23)(24)(25)(26). Similar results were obtained with the NO scavenger cPTIO (24)(25)(26). The translational potential of iNOS inhibitors was readily apparent from these findings, viz. their ability to improve PDT outcomes by increasing tumor regression and/or suppressing greater migratory activity. Some of these inhibitors (L-NIL, GW274150) have already been safely tested in clinical trials, although these had no relationship to cancer or PDT (28,71). As an intermediary proof-ofconcept, we recently showed that ALA-PDT suppression of mouse-borne human breast tumor xenografts was substantially augmented by administration of 1400W or GW274150, whereas no significant effect was observed on control tumor growth (52). Consistently, iNOS protein in tumor samples was strongly upregulated after ALA-PDT, and NO-derived NO 2 levels were also elevated relative to control levels -and in 1400Winhibitable fashion (53). However, it was apparent from our previous in vitro and in vivo studies that the modulating effects of iNOS enzyme inhibitors (e.g. increased PDT cytotoxicity or decreased survivor aggressiveness) were far from maximal at relatively high inhibitor concentrations or dosages (24)(25)(26)53). This prompted us to ask whether possible suppression of iNOS expression with a BET bromodomain inhibitor such as JQ1 might be more effective than inhibition of expressed iNOS activity.
JQ1 and other inhibitors of epigenetic (BETcontaining) reader proteins such as Brd2, Brd3 and Brd4 have emerged as highly potent and relatively selective pharmacologic suppressors of cancer cell proliferation, migration, and metastatic dissemination (32)(33)(34)(35). BET bromodomain inhibitors function by binding to ε-N-acetylated lysine (acK) recognition motifs on BET proteins, thereby preventing BET bromodomains from binding to acK sites on histones and transcription factors (32,33). Many BET bromodomain inhibitors are under clinical trial scrutiny for a variety of malignancies, including multiple myeloma, lymphoma, triple negative breast cancer, and other solid tumors (72,73). BETcontaining proteins recognize specific acK residues on histones and transcription factors such as NF-κB (32)(33)(34)(35)72). Recent seminal studies by Chen et al. (30,31), using A549 lung cancer cells, revealed that Brd4 binding to acK310 on the p65 subunit of NF-κB maintains the latter in an active form. Brd4 knockdown by shRNA or inhibition by JQ1 suppressed expression of NF-κB target genes while inducing ubiquitination and degradation of nuclear p65, both in constitutively active and TNF-α-stimulated form (31). Target genes such as E-selectin, A20, and IL-8 were identified in those studies, but there was no indication as to whether iNOS expression was also activated by Brd4, and if so, how JQ1 would affect it. In fact, to our knowledge there is no published prior work on how a BET bromodomain inhibitor such as JQ1 might limit cancer progression by interfering with iNOS expression. However, in macrophagemediated immune responses to bacterial pathogens, analogous interference has been described (74).
In the present study, we found that activation and nuclear translocation of NF-κB in glioblastoma U87 cells played a key role in basal as well as PDT-stimulated iNOS expression. Using JQ1 at a concentration that was minimally toxic to these cells, we showed that this BET bromodomain inhibitor strongly suppressed iNOS expression and NO generation in both control and PDT-challenged U87 cells. Concomitantly, JQ1 caused a striking increase in apoptotic cell death when used in combination with PDT such that an overall synergistic effect was seen ( Fig. 1B; ~50% apoptosis). In contrast, relatively little necrotic cell death occurred (<5%), which is significant because clinical PDT strives to maximize apoptosis and minimize necrosis in order to limit non-specific inflammatory stress from necrosis (12)(13)(14). Based on our previous evidence obtained with breast cancer cells (21-23), we predicted that PDT would activate PI3K-dependent Akt in U87 cells upstream of NF-κB activation and iNOS/NO expression. A robust phosphorylation-activation of Akt did occur after PDT (Fig. 5) and Akt presumably activated NF-κB via phosphorylation of IκB kinase (23). Importantly, JQ1 had no effect on Akt activation, suggesting that JQ1 suppression of iNOS/NO occurred entirely at the site of iNOS expression. While the strong enhancement of U87 photokilling by JQ1 (Fig. 1) has important implications for improving clinical PDT efficacy with this BET bromodomain inhibitor, the observed JQ1 suppression of hyper-aggressiveness in PDT-surviving cells (Figs. 8, 9) has even greater significance in terms of limiting cancer progression. We discovered previously (26), and confirmed here using glioblastoma U87 and U251 cells, that cells not lethally photodamaged can grow and migrate/invade more rapidly, potentially leading to greater metastatic dissemination if occurring in vivo. By suppressing iNOS/NO upregulation as a major contributing factor, JQ1 curbed this enhanced aggressiveness, and much more effectively than an iNOS enzyme inhibitor (1400W) at exceedingly higher concentration. Taken together, our findings suggest that by restraining iNOS expression and NO production, JQ1 could greatly improve clinical PDT outcomes, not only by enhancing tumor regression, but limiting the adverse effects of surviving cells, i.e. more rapid invasion leading to metastasis.
A novel and particularly interesting observation in this study is that Brd4, like iNOS itself, was upregulated several fold in PDTstressed U87 cells, whereas a Brd4 paralog, Brd2, was unaffected. Equally interesting is our observation that the level of acK310 on the p65 subunit of NF-κB also increased several fold in these cells. Like the Brd4 response, p65-acK310 upregulation has not, to our knowledge, been described previously for any type of cancer cell subjected to a therapy-related oxidative challenge. The explanation for the elevated p65-acK310 level is not yet clear. However, one can speculate that stress-induction and/or activation of acetyltransferase p300/CBP was involved, or possibly downregulation of a p65 deacetylase such as Sirt1 (75). These different possibilities will be assessed in our ongoing studies. We postulate that post-PDT Brd4 and acK310 upregulation are cooperative stress responses that promote p65mediated iNOS expression and other prosurvival/expansion genes, leading to a more resistant and aggressive cell phenotype. Although JQ1 did not affect Brd4 or acK310 upregulation, it nearly abolished basal as well as stress-activated iNOS expression, most likely by binding to Brd4 and preventing its access to p65-acK310.
It is clear from the large protective effects of an iNOS enzyme inhibitor (Figs. 3 and 8; Ref. 26) that iNOS played a major role in glioblastoma cell resistance to photokilling as well as greater aggressiveness of surviving cells. However, based on evidence in Fig. 10, other NF-κB-regulated proteins such as survivin, Bcl-xL, and p21 probably contributed to these responses, survivin and Bcl-xL being upregulated by PDT and p21 down-regulated. Each of these responses was strongly affected by JQ1, i.e. inhibited for survivin and Bcl-xL, and reversed for p21. JQ1 may have down-regulated survivin directly by blocking its transcription (47,48). However, an indirect effect due to iNOS down-regulation was also possible, since NO is known to signal for survivin induction (49,50). Similar direct and indirect effects of JQ1 may have occurred in the case of Bcl-xL, given our recent evidence that Bcl-xL upregulation in PDT-stressed breast cancer cells was suppressed by iNOS enzyme inhibitors (52). Whether low level iNOS-derived NO has any influence on p21 expression is not known. Therefore, for at least two of the effectors described in Fig. 10, JQ1 could have acted directly by preventing Brd4 binding at promoter sites as well as indirectly via suppression of iNOS expression.
It is likely that other oxidative stress-based therapeutic modalities will induce prosurvival/expansion adaptations similar to PDT. For example, ionizing radiation has been reported to elicit such responses and overexpressed iNOS/NO has been implicated (76), but the underlying regulation of iNOS expression were not investigated. Of related interest is a recent study demonstrating that JQ1 can function as a radiosensitizer, i.e. act additively or even synergistically with ionizing radiation in dispatching malignant cells (54). Numerous other examples of combining BET bromodomain inhibitors with conventional chemotherapeutic agents have been described recently, e.g. JQ1 with paclitaxel for triple negative breast cancer (77), and OTX015 with temozolomide for glioblastoma (78). A potential clinical advantage of such combined treatments is that lower than normal individual drug dosages can be used, thus causing less off-target toxicity. Another advantage is that different subcellular sites with different negative effects can be targeted. For the ALA-PDT which we describe, mitochondria are the primary targets and nuclear sites are secondary targets, albeit indirect ones through JQ1 binding/inactivation of Brd4. It remains to be seen whether other anticancer therapies will realize similar advantages through use of the conventional modalities combined with BET bromodomain inhibitors in moderate doses.

CONCLUSIONS AND PERSPECTIVES
We describe a PDT-aggravated growth and invasive aggressiveness of glioblastoma cells in which NO from photostress-induced iNOS plays a major driving role. The NF-κB-dependent iNOS response was fostered by upregulation of epigenetic reader Brd4 and acK310 on the p65 subunti of NF-κB (Fig. 11). The Brd4 inhibitor JQ1 suppressed iNOS expression, NO production, and cell hyper-aggressiveness much more powerfully than an inhibitor of iNOS enzymatic activity, suggesting that JQ1 or a related BET bromodomain inhibitor could greatly improve clinical PDT outcomes for glioblastoma and possibly other malignancies. A few examples of combining JQ1 with conventional radio-or chemotherapeutic approaches have been reported (54,77,78), but the present study represents the first time that JQ1 has been combined with PDT, which is recognized as one of the best treatment options for many solid tumors, including glioblastomas (67)(68)(69). Our findings from this in vitro study provide a strong incentive for more advanced work involving JQ1 in a mouse tumor PDT model, which will soon be underway.

Cell culture
Human glioblastoma U87-MG and U251-MG cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). From this point on, these designations are shortened to U87 and U251, respectively. Cells were grown in a humidified incubator at 37 o C/5% CO 2 , using minimal essential medium with Earle's salts (MEM) supplemented with 10% FBS, 1% pyruvate, penicillin (100 units/ml), and streptomycin (100 µg/ml). The cells were switched to fresh supplemented medium every third day, passaged fewer than 6-times for all experiments, and used at no greater than ~65% confluency.

Cell sensitization and irradiation
U87 or U251 cells at 60-65% confluency in 35-mm culture dishes were metabolically sensitized with PpIX by incubating with 1 mM ALA in serum-and phenol red-free MEM for 30 min in the dark at 37 o C. As shown previously (26), most of the PpIX at this point was localized in mitochondria, where it originates via the heme anabolic pathway. Immediately after this step, the medium was removed and cells were overlaid with fresh MEM lacking serum, phenol red, and ALA. Cell dishes were then placed on a translucent plastic platform over a bank of four 40W coolwhite fluorescent lamps and irradiated at room temperature. The light power density (irradiance or fluence rate) at the bottom of each dish was ~1.1 x 10 -3 W/cm 2 . Cells were typically irradiated for 15 min, which corresponds to a delivered light dose or fluence of ~1 J/cm 2 . Immediately thereafter, the cells were overlaid with fresh 10% v/v FBS-supplemented medium, which either lacked or contained the BET bromodomain inhibitor JQ1 or iNOS inhibitor 1400W at a predetermined starting concentration. Immediately before use, stock solutions of JQ1 and 1400W were prepared in DMSO and PBS, respectively. A vehicle control for JQ1 was prepared and examined alongside; for 0.3 µM JQ1 (frequently used), this amounted to 0.001% v/v DMSO. Each inhibitor was maintained at the same starting concentration throughout all subsequent dark incubations. After various post-irradiation incubation times, cell samples were recovered for determination of parameters such as viable fraction, extent of apoptosis, and surviving cell proliferation and invasion rate. Cells treated with ALA alone or light alone were prepared and analyzed alongside as controls. For the in vitro experiments in this study, the term PDT is defined as "photodynamic treatment" and is distinguished from PDT as photodynamic therapy, which should apply only to in vivo situations.

Measurement of viability loss and extent of apoptosis in PDT-treated cells
The effects of photodynamic stress on overall cell viability were determined by Dojindo CCK-8 assay (79), which was typically carried out 24 h after irradiation. Light-only or ALA-only controls were analyzed as well. Early stage apoptosis, as indicated by externalization of plasma membrane phosphatidylserine, was assessed by Annexin V-FITC staining with fluorescence microscopy. A 96-well plate reader system (Biotek Synergy MX) was used with 485 nm excitation and 535 nm emission. Any necrosis due to plasma membrane disruption was assessed by Propidium Iodide (PI) staining. Other details were as described previously (26).

Detection of NO in photodynamically-stressed cells
NO levels in photostressed glioblastoma cells were assessed using the fluorophore DAF-FM-DA (38). Upon entering cells, DAF-FM-DA is hydrolyzed and trapped inside as DAF-FM, which fluoresces weakly. In aerobic systems, NO-derived dinitrogen trioxide (N 2 O 3 ) can nitrosate DAF-FM to give highly fluorescent DAF-FM-triazole. A stock solution of 1 mM DAF-FM-DA in DMSO was prepared immediately before experimental use and shielded from room light. At various postirradiation times, cells in serum-free medium were incubated in the dark for 50 min with 10 µM DAF-FM-DA, then washed and examined for DAF-FM-triazole level by fluorescence microscopy, using a Nikon Eclipse TS100 microscope set at 495 nm excitation and 515 nm emission.
NO generated by photostressed cells was also determined by Griess assay. U87 cells (3 × 10 4 per well) were seeded into a 96-well plate and allowed to attach overnight. After ALA-treatment, irradiation, and washing, cells were overlaid with serum-free medium containing either 0.3 µM JQ1 or DMSO vehicle. After dark incubation for various intervals, the medium was recovered for measurement of NO-derived nitrite (NO 2 -) and nitrate (NO 3 -) by Griess assay, using a protocol recommended by the reagent supplier (Cayman Chemical Co.). The procedure included reduction of any NO 3 in the samples by nitrate reductase. Absorbance of the azo dye product at 540 nm was recorded in a plate reader and quantification of total NO 2 -/NO 3 -(NO x ) was based on a NO 2 standard curve. Standardization was based on determination of total protein in each well.

Western blot procedures
The expression of iNOS, Brd2, Brd4, MMP-9, Survivin, c-Myc, Bcl-xL, and p21 in U87 cells before and after a photodynamic challenge was monitored by Western blot analysis, using commercially available and authenticated primary antibodies (see General materials section). At various dark incubation times after irradiation, beginning immediately (0 h), and extending to 24 h, treated cells along with appropriate controls were recovered by gentle scraping, centrifuged, and washed with ice-cold PBS. Cells were suspended in cold pH 7.4 lysis buffer (10 mM Tris pH 8.0, 1% v/v Triton X-100, 0.1% w/v sodium deoxycholate, 0.1% w/v SDS, 140 mM NaCl, and 1 mM PMSF) containing protease inhibitors (22) and homogenized as described (22). After centrifugation, the supernatant fraction was analyzed for total protein by BCA assay, after which samples of equal protein content (typically ~100 µg) were separated by SDS-PAGE using appropriate acrylamide/bis-acrylamide mixtures. Separated proteins were transferred to a polyvinylidene difluoride membrane, and after blocking using 5% w/v non-fat dry milk in TBST, the membrane was treated overnight at 4 o C with a primary antibody diluted as follows: 1:250 for iNOS and 1:1000 for all other proteins. After washing, the membrane was treated with a peroxidase-conjugated IgG secondary antibody (1:10,000), after which protein bands were analyzed using SuperSignal West Pico chemiluminescence detection (Thermo Scientific, Rockford, IL). Other details were as described previously (22,26).

Detection of NF-κB/p65 in cytoplasmic and nuclear fractions
U87 cells were treated with ALA alone; ALA and light; ALA, Bay11 (5 µM) and light; or ALA, JQ1 (0.3 µM) and light. Irradiated cells, along with controls (ALA-alone) were incubated in the dark for 5 h. The cells were removed by gentle scraping into PBS, pelleted by centrifugation, and recovered. Nuclear and cytoplasmic fractions were then isolated using a NE-PER TM kit supplied by Thermo Scientific. All centrifugation and lysing steps for preparing cytoplasmic and nuclear fractions were according to supplier recommendations. After determination of total protein concentration, samples from each fraction were analyzed by Western blotting, using antibodies against p65, histone H3 as a nuclear marker, and α-tubulin as a cytosolic marker.

Detection of BET protein interaction with NF-κB/p65
The possibility that post-PDT activation of NF-κB/p65 required interaction with a BET protein (Brd2, Brd4) was investigated using an immunoprecipitation approach. After an ALA/light challenge, U87 cells were switched to 10% FBS-containing medium lacking or containing 0.3 µM JQ1, and returned to the incubator. After 8 h of incubation, cells were lysed and total lysate protein was determined. A primary monoclonal antibody against p65 (10 µl from a 1:100 diluted stock solution from Cell Signaling Technologies (#4764S) was added to 250 µg of total cell lysate and incubated for 16 h at 4 o C with mild agitation. After incubation, 100 µl of Protein A-conjugated Sepharose bead slurry was added to each lysate, followed by 6 h of additional incubation at 4 o C with agitation. After centrifugation, the beads with attached proteins were washed 3× with lysis buffer to remove any contaminating proteins, and then treated with 50 µL of 0.2 M glycine buffer (pH 2.6) for 10 min to release bound proteins. After centrifugation, proteins were recovered in supernatant fractions, which were brought to pH 8.0 with 20 mM Tris buffer. Samples were then mixed with SDS sample buffer in preparation for SDS-PAGE, followed by p65, Brd2, and Brd4 immunoblotting, using the antibody dilutions described in the preceding section.

Evaluation of surviving cell proliferation
Twenty four-hours after an ALA/light challenge, followed by a wash to remove detached (dead or dying) cells, remaining (surviving) glioblastoma cells were recovered by gentle scraping along with non-irradiated controls and seeded into a 96-well plate using 10% v/v FBScontaining MEM. Normalization of seeding density was based on prior knowledge of U87 viability losses; e.g. a 25% cell kill was compensated for by plating 25% more cells, so that the initial cell count for each experimental condition was approximately the same. These cells, along with non-stressed controls, were darkincubated in the presence of JQ1 or 1400W at the indicated starting concentrations. At various time points out to 72 h, numbers of viable cells were determined by CCK-8 assay and expressed relative to the 24 h post-irradiation starting point.

Evaluation of surviving cell invasiveness
The invasiveness of U87 or U251 cells that could withstand an ALA/light challenge was examined using a 96-place trans-well device (Model MBA96) from NeuroProbe (Gaithersburg, MD). Immediately after ALA/light exposure, these cells, along with ALA-only or light-only controls, were treated with JQ1, JQ1(-), Bay11, or 1400W at the indicated concentrations in serum-free medium and transferred to the upper wells of the invasion chamber (225 µL per well). Prior to this, 225 µL of 10% FBS-containing medium was added to each lower well of the invasion chamber, the serum serving as a cell attractant. A Matrigelinfused polycarbonate filter with 8 µm pores was fitted over each lower well, after which the unit was pre-warmed at 37 o C. The upper and lower wells were then clamped together and the closed unit placed in a 37 o C incubator. After a given incubation period (typically 24 h), medium in the upper wells was carefully removed and cells remaining on top of the filters were gently wiped off with a cotton swab. Cells that had invaded to the filter bottoms were detached by centrifugation into 10% FBS-containing medium (400 × g; 15 min), allowed to adhere on a 96-well plate, and then either stained and photographed or quantified by CCK-8 assay.

Statistical analyses
Data are presented as means ± SEM of values from at least three replicate experiments. Statistical significance was determined using the Student's t-test in conjunction with PRISM Graphpad Software, P values <0.05 being considered as statistically significant.

Conflicts of interest
The authors declare that they have no conflicts of interest with the contents of this article.      Twenty-four hours after U87 cells were ALA/light-treated (see Fig. 4), any detached cells were carefully removed by aspiration. Remaining live cells were recovered by gentle scraping and seeded into a 96-well plate along with non-irradiated control cells. A pre-determined greater seeding density was used for photostressed cells to account for the portion that succumbed to this treatment. Accordingly, the initial cell count was approximately the same for all post-irradiation conditions studied. PDT-surviving cells, along with non-stressed (ALA-only) controls, in 10% serum-containing medium were dark-incubated in the presence of 0.3 µM JQ1 or 25 µM 1400W (W), which were maintained at these concentrations throughout. At the indicated time points, live cell levels were determined by CCK-8 assay. Plotted numbers are means ± SEM (n=4).