Treatment of human liver, pancreatic and kidney tumor cell lines with increasing low relatively concentrations of vorinostat and sorafenib at a fixed dose ratio resulted in a synergistic enhancement in tumor cell killing as measured by median dose effect long-term colony formation assays ( Figure S1
). Re-expression of the VHL protein in 786 renal carcinoma cells did not significantly alter the synergistic interaction of vorinostat and sorafenib (not shown). Additional studies, using a variety of short term cell killing assays, examined the interaction between vorinostat and sorafenib. In HEPG2 and UOK121LN cells, TUNEL, Annexin V – propidium iodide and trypan blue exclusion viability assays generated quantitatively and qualitatively similar data i.e., greater than additive induction of cell killing following low dose vorinostat and sorafenib exposure compared to either agent individually (; Figure S2
). Substratum attachment did not alter the sensitivity of A498 cells to vorinostat and sorafenib exposure (not shown). Based on the findings in , the molecular mechanisms by which vorinostat and sorafenib interacted to kill cells were then investigated.
Sorafenib and Vorinostat interact in a synergistic fashion to kill pancreatic, liver and kidney tumor cells in colony formation assays
In HEPG2 and Mia PaCa 2 cells, pan-inhibition of caspase function using zVAD significantly reduced vorinostat and sorafenib lethality, as did inhibition of cathepsin protease function; in UOK121LN cells the relative role of caspases in the death process appeared to be greater than that of cathepsins (;Figure S3
). Inhibition of caspase 8 function (e.g., by ectopic expression of CrmA) significantly reduced, to a greater extent than inhibition of caspase 9 (e.g. by ectopic expression of dn caspase 9), the lethality of vorinostat and sorafenib in our panel of tumor cells. Similar cell killing data after vorinostat and sorafenib treatment were obtained using the caspase 8 and caspase 9 inhibitors IETD and LEHD, respectively (not shown). Based on data showing that inhibition of caspase 8 and caspase 9 significantly reduced vorinostat and sorafenib lethality, genetically modified transformed mouse embryonic fibroblasts lacking expression of various cell survival modulator genes were employed. Combined loss of BAX and BAK function or loss of BID function significantly reduced vorinostat and sorafenib lethality in transformed fibroblasts (). These findings suggest that lysosomal dysfunction plays a role in the killing process and that caspase 8 signaling and the extrinsic pathway are also involved in vorinostat and sorafenib –induced transformed cell lethality.
Inhibition of caspase 8 and cathepsin function suppresses sorafenib and vorinostat lethality in GI tumor cells
To further define the processes of cell death, immunoblotting analyses in vorinostat- and sorafenib- treated HEPG2 and UOK121LN cells were performed (). In UOK121LN cells, within 6h of combined, but not individual, drug exposure the expression of the pro-forms of BID, pro-caspase 8 and pro-caspase 3 (, section (i)) declined relative to vehicle control treated cells. These observations correlated with decreased expression of BCL-2 and c-FLIP-s (, section (ii)). The cleaved form of BID was poorly visualized by immunoblotting in all studies using this drug combination (not shown). Twenty four hours after drug exposure, the expression of BCL-2, BCL-XL, MCL-1, c-FLIP-s, BID, pro-caspase 8 and pro-caspase 3 had further declined after combined, but not individual, drug exposure (, sections (i) and (ii)).
Sorafenib and vorinostat treatment modulates the expression of c-FLIP-s, BCL-XL and MCL-1 in tumor cells
In HEPG2 cells, little obvious change in the expression of any protein was observed 6h after combined drug exposure with the exception of c-FLIP-s and pro-caspase 3, whereas 24h after combined, but not individual, vorinostat and sorafenib exposure, expression of BID, pro-caspase 8, pro-caspase 3, XIAP, BCL-2, BCL-XL, MCL-1, and c-FLIP-s were all reduced (, sections (i) and (ii)). In both HEPG2 and UOK121LN cells, decreased expression of pro-survival proteins correlated with increased phosphorylation of eIF2α S51; increased phosphorylation of eIF2α S51 is known to correlate with increased activity of this protein and with suppression of translation / initiation in cells (, section (ii)) (21
). Similar immunoblotting data to that obtained in HEPG2 and UOK121LN cells in was obtained in MiaPaca2 cells (Figure S4
). Of note, cell viability data obtained 24h after drug exposure argued that drug-treated tumor cells had just begun to display signs of cell death at this interval (e.g. time course data in , not shown). Thus, the observed reductions in pro-survival protein expression occurred prior
to significant manifestations of cell killing.
The ability of over-expression of proteins whose levels were reduced in to prevent cell killing was then tested. Over-expression of BCL-XL, XIAP or c-FLIP-s significantly reduced vorinostat and sorafenib lethality (). Whether the expression and/or activity of additional pro-apoptotic proteins correlated with increased cell killing was also determined. Treatment of cells with vorinostat and sorafenib increased the expression of BIM, including promotion of BIM dephosphorylation as well as the dephosphorylation of BAD S112 and the activation of BAX and BAK (, upper inset blotting panels). Based on data showing that c-FLIP-s over-expression largely abolished low dose vorinostat and sorafenib lethality, regardless of the fact that in parallel drug exposure also suppressed expression of multiple downstream pro-survival proteins and activated BAX and BAK, the possibility that vorinostat and sorafenib – induced killing was death receptor dependent, specifically CD95 (FAS receptor) –dependent, was explored.
Notably, knock down of FADD or CD95 expression significantly reduced the lethality of low dose combined vorinostat and sorafenib exposure in HEPG2 and UOK121LN cells (). Vorinostat and sorafenib exposure in HEPG2 and UOK121LN cells, in a cell type –dependent fashion, also enhanced expression of FAS-L and/or CD95 proteins as well (Figure S5
). In addition, combined treatment of HEPG2 and UOK121LN cells with vorinostat and sorafenib promoted the rapid association of pro-caspase 8 with CD95 i.e. DISC complex formation (, upper section (i)). Knock down of CD95 abolished drug-induced procaspase 8 and BID cleavage in total cell lysates (not shown). Studies in primary hepatocytes treated with death-inducing natural compounds that act via CD95, such as toxic bile acids, have shown that these agents cause plasma membrane localization and clustering of CD95 as part of the receptor activation / hepatocyte killing process (34
). Treatment of HEPG2 cells with vorinostat and sorafenib caused increased plasma membrane localization of CD95 and the appearance of intense-staining punctate bodies containing CD95, demonstrative of CD95 clustering and its activation (, lower section (ii)). Note that in HEPG2 cells, CD95 activation and DISC formation occurred at time points without alteration in either total protein levels of CD95 or FAS-L / FAS-L cleavage, arguing that CD95 activation was ligand-independent. Over-expression of c-FLIP-s significantly suppressed vorinostat and sorafenib lethality as measured in TUNEL assays and markedly diminished cytochrome c release into the cytosol of HEPG2 cells (Figure S6
Sorafenib and vorinostat interact to kill tumor cells via activation of CD95 and suppression of c-FLIP-s expression
We next defined how low doses of combined exposure to vorinostat and sorafenib could rapidly suppress the expression of multiple pro-survival proteins. High doses of sorafenib,
3 µM as a single agent, have been shown by our laboratories to cause ER stress, translational inhibition, and reduced expression of MCL-1 that correlated with eIF2α phosphorylation (19
); in the present studies, lower doses of sorafenib (~3 µM) do not
enhance eIF2α phosphorylation but did
synergize with vorinostat to cause eIF2α phosphorylation. Expression of dominant negative eIF2α S51A abolished low dose combined sorafenib and vorinostat –induced suppression of c-FLIP-s and MCL-1 expression in HEP3B cells and expression of eIF2α S51A in transformed fibroblasts and in HEP3B cells suppressed the toxic interaction between sorafenib and vorinostat (; Figure S7
). Over-expression of c-FLIP-s suppressed the synergistic lethality of vorinostat and sorafenib in median dose effect colony formation assays reducing the combination index (CI) value from ~0.45 to ~1.10 (). These findings argue that low doses of vorinostat and sorafenib induce activation of the extrinsic apoptotic pathway at the level of the CD95 death receptor, and that the initial rapid loss of c-FLIP-s expression with 24h occurs via activation of eIF2α, and plays a critical role in transmitting death signals from the plasma membrane to the cytosol, resulting in a plieotropic activation of multiple downstream apoptotic processes, including mitochondrial dysfunction.
The synergy of killing by sorafenib and vorinostat is dependent upon loss of c-FLIP-s expression
Sorafenib was originally developed as an inhibitor of Raf
family protein kinases, but subsequently shown to be an inhibitor of several receptor tyrosine kinases and an inducer of ER stress signaling (12
). Vorinostat has been shown to modulate in a concentration- and cell type- dependent manner ERK1/2, NFκB, AKT, JNK1/2 and p38 MAPK pathway signaling (23
). Furthermore, changes in NFκB, ERK1/2, AKT, JNK1/2 and p38 MAPK signaling have been linked to the modulation of CD95 function and that of c-FLIP-s, expression, also in a cell type dependent manner. These considerations prompted further examination of whether significant alterations in signaling pathway function occurred in vorinostat/sorafenib-treated cells and whether these changes could be related to CD95 activation, changes in apoptosis-regulatory protein expression, and overall tumor cell survival.
Twenty four hours after drug exposure, a time at which CD95 activation had occurred, eIF2α phosphorylation had occurred, and the expression of multiple pro-survival proteins such as c-FLIP-s had already declined, no profound change in the basal activities of ERK1/2, AKT, JNK1/2 or p38 MAPK were observed in HEPG2 or HEP3B cells (, section (i), not shown). Over the ensuing 72h, activation status of the JNK1/2 or p38 MAPK pathways did not correlate strongly with cell death induction. In contrast to signaling pathways that promote death, activities of pathways that protect against cell death began to decline in cells treated with vorinostat and sorafenib 24h–96h after drug exposure. Within 48h of exposure, ERK1/2 was inactivated in combined drug exposed cells; within 72h of exposure, AKT became inactivated. Inactivation of the ERK1/2 pathway was not due to the prior activation of CD95 or FADD, in view of the findings that knock down of CD95 or FADD suppressed cell killing but did not maintain ERK1/2 phosphorylation (, section (ii)). Thus, inactivation of ERK1/2 and AKT were relatively late events in cell death induction after sorafenib and vorinostat exposure, but were not causally dependent upon the primary CD95-dependent apoptotic signal.
Delayed inactivation of ERK1/2 and AKT correlates with profound and long-term loss of c-FLIP-s expression: delayed activation of NFκB is a toxic signal following sorafenib and vorinostat exposure
Based on the findings in , attempts were made to determine whether expression of constitutively active MEK1 and/or AKT protected cells from vorinostat and sorafenib exposure. Expression of constitutively active MEK1 maintained ERK1/2 phosphorylation in HEPG2 cells treated with vorinostat and sorafenib, as did expression of constitutively active AKT in maintaining levels of AKT S473 phosphorylation (, upper inset panel to the left). Expression of either activated MEK1 or activated AKT almost abolished the toxicity of the individual drugs and significantly suppressed the toxicity of the drug combination (, lower graphical section). These findings correlated with maintenance of c-FLIP-s expression in tumor cells expressing activated MEK1 and activated AKT, and treated with vorinostat and sorafenib (, upper inset panel to the right). Collectively, these data further argue that maintained c-FLIP-s expression prevents CD95 signaling from activating the caspase 8 – BID pathway to induce mitochondrial dysfunction and death. These findings also argue that primary activation of eIF2α followed by the secondary inhibition of ERK1/2 and AKT represent the likely sequence of events by which low doses of sorafenib and vorinostat suppress c-FLIP-s levels, and subsequently maintain suppression of c-FLIP-s expression, in transformed cells.
Prior studies using vorinostat have shown that this agent activates the transcription factor NFκB and that this can act against the lethal actions of this drug (29
). Whether NFκB function plays any role in cell survival after low dose vorinostat and sorafenib treatment in our cell system was then examined. Treatment of HEPG2 and UOK121LN cells with vorinostat caused a late post-24h exposure –induced activation of NFκB, that was not significantly altered by incubation of cells with low doses of sorafenib (; Figure S8
). As noted above, NFκB activation following vorinostat treatment has been demonstrated as a protective signal in malignant hematologic cells, we determined whether genetic inhibition of NFκB function via expression of the super repressor IκB S32A S36A altered the survival response of drug treated carcinoma cells. Expression of IκB S32A S36A inhibited vorinostat –induced activation of NFκB did not significantly alter the lethality of vorinostat as a single agent (; Figure S8
, data not shown). However, inhibition of NFκB function significantly suppressed the death of cells treated with sorafenib and vorinostat. Our findings argue that hyperactivation of ERK1/2 and AKT can suppress killing that may be due to maintained expression of c-FLIP-s, and that the observed late phase activation of NFκB induced by vorinostat treatment is, surprisingly, a toxic signal.
Finally, we performed in vivo analyses using established ~150 mm3 HEP3B flank tumors to determine whether sorafenib and vorinostat interacted in a toxic manner in vivo. We noted that unselected clones of HEP3B and HEPG2 cells are poorly tumorigenic in the flanks of athymic mice and form tumors that rapidly become necrotic upon growth beyond > 200 mm3, potentially due to a relatively low CD31 staining (data not shown). As such, we chose an in vivo treatment, ex vivo colony formation assay approach to assess tumor cell killing and long-term survival. A 3 day treatment of animals with either vorinostat or sorafenib caused little alteration in the cleavage status of caspase 3 or the TUNEL positivity of flank tumor sections (, sections to the left). Combined exposure of animals/tumors to vorinostat and sorafenib caused a large increase in both the number of apoptotic TUNEL positive cells and a large increase in immunoreactivity for the cleaved form of caspase 3. Combined, but not individual, exposure of animals/tumors to vorinostat and sorafenib caused a large increase in eIF2α phosphorylation and large decreases in the expression of c-FLIP-s and MCL-1 and phosphorylation of AKT (S473). Of particular note, at the in vivo concentrations of sorafenib used as a single agent in our study, we observed near total inhibition of ERK1/2 phosphorylation by the drug; but this inhibition of ERK1/2 phosphorylation did not correlate with the cleavage of caspase 3, enhanced TUNEL positivity or lower c-FLIP-s or MCL-1 levels. In ex vivo colony formation assays using viable cells isolated from treated tumors after cessation of drug treatment, and with cells cultured in the absence of any drug in vitro, combined exposure of animals/tumors to sorafenib and vorinostat caused a greater reduction in cell survival of the explanted tumor cells growing ex vivo than was observed in the cells that had been exposed to either drug individually (, graph to the right). Collectively, these findings argue that our molecular defined markers for sorafenib and vorinostat lethality are observed in vitro and also in drug treated tumors, and that these effects correlate with an increase in both short-term and long-term tumor cell killing.