Inhibition of Ras/Raf/MEK/ERK Pathway Signaling by a Stress-Induced Phospho-Regulatory Circuit
SUMMARY
Ras pathway signaling plays a critical role in cell growth control and is often upregulated in human cancer. The Raf kinases selectively interact with GTP-bound Ras and are important effectors of Ras signaling, functioning as the initiating kinases in the ERK cascade. Here, we identify a route for the phospho-inhibition of Ras/Raf/MEK/ERK pathway signaling that is mediated by the stress-activated JNK cascade. We find that key Ras pathway compo- nents, the RasGEF Sos1 and the Rafs, are phosphor- ylated on multiple S/TP sites in response to JNK acti- vation and that the hyperphosphorylation of these sites renders the Rafs and Sos1 unresponsive to upstream signals. This phospho-regulatory circuit is engaged by cancer therapeutics, such as rigosertib and paclitaxel/Taxol, that activate JNK through mitotic and oxidative stress as well as by physiolog- ical regulators of the JNK cascade and may func- tion as a signaling checkpoint to suppress the Ras pathway during conditions of cellular stress.
INTRODUCTION
The three-tiered RAF/MEK/ERK kinase cascade functions as an essential effector cascade required for Ras GTPase signaling in normal and disease states (Aoki et al., 2016; Roberts and Der, 2007). Signal transmission through the cascade begins when members of the Raf family are recruited from the cytosol to the plasma membrane, where they bind directly to active GTP-Ras (Lavoie and Therrien, 2015). Binding to Ras induces conforma- tional changes that promote Raf dimerization, which in turn me- diates kinase activation through an allosteric mechanism that often involves B-Raf/C-Raf heterodimers (Freeman et al., 2013a; Hu et al., 2013). Once activated, Raf initiates the sequen- tial phosphorylation events that ultimately result in ERK activa- tion and the downstream phosphorylation of key substrates required for a specific response. Along with its central function in the forward transmission of signals, active ERK also contrib-utes to inhibitory feedback loops that control the duration and amplitude of Ras pathway signaling (Dougherty et al., 2005; Mendoza et al., 2011).A major advance in understanding ERK cascade regulation has come from the recognition that Raf dimerization is a required step for Raf activation in most cellular contexts (Freeman et al., 2013b). Not only is dimer formation needed for normal Ras- dependent Raf activation, it contributes to the pathogenic activ- ity of many disease-associated Raf proteins, with the exception of select B-Raf mutants, such as V600E-B-Raf, that can function as activated monomers (Yao et al., 2015).
Moreover, Raf dimer- ization and the resulting paradoxical activation of ERK that can occur in cells expressing wild-type Raf alleles has limited the usefulness of ATP-competitive Raf inhibitors as a cancer therapy and often mediates resistance to these drugs (Lito et al., 2013). Therefore, identifying compounds or a means to disrupt Raf dimerization could have significant clinical benefit.Rigosertib (also known as ON01910 and Estybon) is a styrl- benzylsulfone, currently being evaluated in patient trials to treat high-risk myeloid dysplastic syndrome (Garcia-Manero et al., 2016). Rigosertib was originally developed as a non-ATP- competitive, multi-kinase inhibitor of the tyrphostin class, and it is a drug that induces mitotic arrest and apoptosis in a broad spectrum of human cancer cell lines, while having little effect on normal cells (Gumireddy et al., 2005; Lu et al., 2015; Reddy et al., 2011). The precise target(s) of rigosertib action have remained elusive, with early reports describing activity against the PLK1 and PI3K kinases (Gumireddy et al., 2005; Prasadet al., 2009). Although a direct effect on PLK1 was not substan- tiated in subsequent work (Ma€ki-Jouppila et al., 2014; Steegma- ier et al., 2007; Twarog et al., 2016), inhibition of PI3K signalinghas been observed in numerous studies (Chapman et al., 2012; Hyoda et al., 2015; Xu et al., 2014).
Recently, a new mech- anism of rigosertib action has been described in which rigosertib functions as a Ras mimetic, capable of binding to the Ras-bind- ing domains (RBDs) of Ras effectors and blocking the interaction with active Ras (Athuluri-Divakar et al., 2016).Given that Ras binding is typically a pre-requisite for Raf dimerization, if rigosertib can indeed bind the RafRBD and disrupt the Ras/Raf interaction, it should in turn inhibit Raf dimer- ization and downstream ERK cascade signaling. Therefore, this study was conducted to evaluate the effects of rigosertib on Ras- dependent Raf dimerization and signaling. Here, we report thatprolonged rigosertib treatment can inhibit Ras/Raf/MEK/ERK signaling; however, the mechanism is largely, if not completely, indirect, mediated by a stress-induced phospho-regulatory cir- cuit that involves JNK cascade activation. This route for the phospho-inhibition of Ras pathway signaling can be engaged by cancer therapies that activate JNK through mitotic and oxida- tive stress as well as by other stress agents that promote JNK signaling, suggesting a broader function as a stress-induced signaling checkpoint.
RESULTS
To investigate the potential of using rigosertib to disrupt Raf dimerization and activation, we first tested the binding of rigoser- tib to purified RafRBD using Biacore surface plasmon resonance (SPR) analysis. As controls, binding of K-Ras to wild-type (WT)- RafRBD and a mutant RafRBD containing the R89L substitution that disrupts Ras binding was monitored. In this assay, K-Ras loaded with a non-hydrolyzable analog of GTP bound with high affinity to WT-RafRBD, but not R89L-RafRBD, and the interac- tion was greatly reduced when K-Ras was in its GDP-bound state (Figure 1A). As shown in Figure 1B, binding of rigosertib to WT-RafRBD was observed, but only at concentrations>6 mM, making it difficult to determine an accurate binding con- stant and suggesting a low-affinity interaction. In further experi- ments to assess whether rigosertib could disrupt the RafRBD/ K-Ras interaction in vitro, we found that at concentrations up to 50 mM, rigosertib did not inhibit binding between WT-RafRBD and K-Ras-GTP as monitored by SPR analysis (Figures 1C and S1A). Moreover, in proximity-based AlphaScreen assays, rigo- sertib had little effect on RafRBD binding to either GTP-loaded WT-K-Ras or G12V-K-Ras (Figure 1D). Next, to determine if an effect of rigosertib could be detected in cell-based assays, we treated HeLa cells with rigosertib for 2 or 18 hr, following which the cells were evaluated for EGF-induced Ras pathway signaling (Figures 1E and S1B). As expected, EGF treatment rapidly activated ERK cascade signaling as evidenced by the strong induction of B-Raf/C-Raf dimer formation and by the detection of activated pMEK and pERK. EGF also promoted PI3K/AKT signaling, as indicated by the appearance of activated pS473-AKT.
At 2 hr of rigosertib treatment, EGF-induced PI3K/ AKT signaling was disrupted, with only low levels of pS473-AKT observed. In contrast, no change in EGF-induced ERK cascade activation was detected at the 2 hr time point, whereas at the 18 hr time point, the levels of EGF-induced pMEK, pERK, and Raf dimerization were dramatically reduced. Given that both p110-PI3K and the Rafs are RBD-possessing Ras effectors, this difference in the timing of inhibition was unexpected and, together with the in vitro binding data presented above, brings into question the reported function of rigosertib as a specific, high-affinity Ras mimetic. Moreover, these findings raise the possibility that rigosertib may alter Ras pathway signaling through alternative and/or multiple mechanisms. In particular, the rapid inhibition of PI3K/AKT signaling by rigosertib may reflect direct activity on PI3K, as has been reported by Pra- sad et al. (2009), whereas the delayed suppression of ERK cascade signaling could be indirect. Consistent with this model, rigosertib had no effect on Raf kinase activity when added in vitro to purified Raf proteins (Figure S1C). However, prolonged 18 hr treatment with rigosertib caused a shift in the electrophoretic mobility of B-Raf and C-Raf that correlated with their reduced dimerization potential (Figures 1E and S1B), and this shift in mobility could be reversed when the Rafs were treated with l-phosphatase in vitro (Figure S1D).
Comparative mass spectrometry analysis revealed that B-Raf and C-Raf proteins isolated from cells treated for 18 hr with rigo- sertib, but not DMSO-treated cells, were hyperphosphorylated on multiple serine or threonine sites followed by a proline residue (S/TP). As a control, the phosphorylation state of a conserved 14- 3-3 binding site on the Rafs (pS621-C-Raf and pS729-B-Raf) was also monitored and remained unchanged (Figure S1E). The sites of S/TP phosphorylation induced by rigosertib treatment were identified as S29, S244, S296, S301, and S642 for C-Raf and as S151, T401, S419, S750, and T753 for B-Raf (Figures 1F and S1F), many of which are involved in ERK-mediated attenuation of growth factor signaling. It should be noted that A-Raf, the third mammalian Raf family member, was also hyperphosphorylated on S/TP sites and exhibited reduced dimer formation in 18 hr rig- osertib-treated cells (Figures 1E and S1B); however, given the low contribution of A-Raf to overall Raf dimerization/activation, our analysis focused on B-Raf and C-Raf. To determine if rigosertib treatment alters Raf phosphorylation in other cell types, a variety of cancer cell lines were treated with rigosertib for 18 hr, following which the electrophoretic mobility of B-Raf and C-Raf was monitored. As shown in Figure 1G, the mobility of the Rafs was shifted in all lines examined. Moreover, using a phospho-specific antibody that recognizes the C-Raf pS642P site, C-Raf isolated from all of the rigosertib-treated cells showed a dramatic increase in the phosphorylation state of this S/TP site, and in lines expressing mutationally activated Ras proteins (A549 and MOLT-4), this increase in Raf S/TP phos- phorylation correlated with decreased levels of constitutive Ras- induced Raf dimer formation (Figure 1G).
To identify the kinase(s) responsible for the hyperphosphoryla- tion of the Rafs in rigosertib-treated cells, we monitoredpS642P-C-Raf levels, the mobility shift of the Rafs, and the acti- vation state of MEK and ERK over a time course of rigosertib treatment (Figure 2A). Activation of the JNK cascade was also monitored, given that JNK signaling is known to be induced by rigosertib treatment (Chapman et al., 2012; Xu et al., 2014) and because the JNK/MAPK is a proline-directed kinase that shares certain substrates in common with the ERK/MAPK (Weston and Davis, 2007). Within 2 hr of rigosertib treatment, a low level of pS642P-C-Raf could be detected that correlated with a partial shift in Raf mobility and weak activation of MEK and ERK. Similar results were observed at 4 and 6 hr of rigosertib treatment, and given that PI3K/AKT signaling is suppressed at early times of rig- osertib treatment, the weak activation of MEK and ERK seen at these time points may reflect relief from inhibitory crosstalk occurring between the PI3K/AKT and ERK cascades (Mendoza et al., 2011). Following 18 hr of treatment, a more pronounced shift in Raf mobility and a dramatic increase in pS642P-C-Raf levels were observed. Although no further increase in pMEK and pERK levels occurred at the 18 hr time point, activation of the JNK/MAPK was now detected, whereas activation of the p38/MAPK was not.To further differentiate the role of the ERK and JNK kinases in mediating the rigosertib-induced S/TP phosphorylation, we first used the MEK inhibitor U0126 to prevent ERK activation (Fig- ure 2B).
When cells were treated with rigosertib for 2 hr in the presence of U0126, the low level of rigosertib-induced ERK acti- vation was blocked, as were the low levels of pS642P-C-Raf. Although U0126 could still prevent ERK activation at the 18 hr time point, it could not suppress the rigosertib-induced mobility shift of the Rafs or the high levels of pS642P-C-Raf. As expected, the presence of U0126 had no effect on JNK cascade activation at the 18 hr time point, which again correlated with the strong in- duction of S642P-C-Raf phosphorylation and the major shift in Raf mobility. Taken together, these findings indicate that ERK is not the primary kinase mediating the hyperphosphorylation of the Rafs at the 18 hr time point and raise the possibility that JNK is the relevant kinase. In support of this idea, we found that when purified components were used, activated JNK could phosphorylate kinase-inactive C-Raf on the five rigosertib- induced S/TP sites in vitro (Figure S2A).A route for JNK activation in rigosertib-treated cells has been elucidated by the work of Chapman et al. (2012), whereby rigo- sertib induces mitotic arrest and an oxidative stress response associated with mitochondrial depolarization and the generation of reactive oxygen species (ROS), which in turn activates JNK signaling. In our studies (Figures 2C–2E), treatment of HeLa cells with rigosertib caused a linear, time-dependent cell-cycle block, with an increase in G2-arrested cells observed within 2 hr andapproximately 90% of the cells arrested by the 18 hr time point (Figure 2D).
Using MitoSOX Red as a specific sensor for mitochondrial ROS generation, ROS levels were significantly elevated at 18 hr but began to accumulate only at the 12 hr time point (Figures 2D and 2E). When cells were co-treated with rigosertib and the antioxidant N-acetylcysteine (NAC), ROS accumulation was reduced to near background levels, and JNK activation was suppressed, whereas the rigosertib- induced cell-cycle arrest was largely unaffected (Figures 2C– 2E). Together, these findings suggest that the sustained cell- cycle block caused by rigosertib treatment may be the initiating stress that promotes mitochondrial ROS generation and that ROS accumulation mediates JNK cascade activation. Interest- ingly, activated pJNK was still observed at 24 and 30 hr of rigo- sertib treatment, as were the shifts in Raf mobility and S642-C- Raf phosphorylation (Figure S2B), raising the possibility that ROS may also function to sustain JNK activity and preserve the phosphorylation state of the Rafs through the oxidation and inactivation of protein phosphatases (Kamata et al., 2005). Further examination of cells co-treated with rigosertib and NAC revealed that in the presence of NAC, S642P-C-Raf phos- phorylation and the mobility shift of the Rafs were suppressed in a manner that correlated with reduced pJNK levels (Figure 2C). Moreover, when JNK activity was directly blocked using the JNK inhibitors SP600125 and JNK-IN-8, rigosertib-induced pS642P-C-Raf levels were reduced to near background, and lit- tle to no shift in Raf mobility was observed (Figures 2F and S2C).
To further corroborate the involvement of the JNK cascade, we next examined the effects of 18 hr rigosertib treatment in two co- lon cancer lines that express mutant K-Ras proteins, HCT-116 and SW620, of which SW620 is homozygous for a loss-of- function deletion in MKK4 (one of the two upstream kinases that activates JNK). When compared to HeLa cells, similar levels of pS642P-C-Raf and equivalent shifts in Raf mobility were observed in HCT-116 cells, whereas in SW620 cells, pS642P- C-Raf levels and the shift in Raf mobility were significantly reduced (Figure 2G). Analogous results were also observed in the AsPC-1 pancreatic line that, like SW620 cells, are MKK4 defi- cient (Figure S2D). Moreover, rigosertib failed to activate JNK in these MKK4-deficient lines (Figures 2G and S2D). Taken together, these findings strongly support the model that JNK is the principal kinase mediating Raf S/TP phosphorylation in rigo-sertib-treated cells.The above findings identify the Raf kinases as targets of both the ERK and JNK MAPKs, raising the possibility that other Raspathway components known to be phosphorylated by ERK, such as the RasGEF Sos1, might also be a shared substrate of these MAPKs. Therefore, we next examined the phosphorylation state of Sos1 in rigosertib-treated cells. As is shown in Figure 3A, overnight treatment with rigosertib induced a shift in the electro- phoretic mobility of Sos1, and this shifted form of Sos1 was de- tected by antibodies that recognize proteins phosphorylated on S/TP sites.
The shift in Sos1 mobility could be reversed by l phosphatase treatment and was not observed when JNK activity was inhibited (Figures 3B and S2C). Comparative mass spec- trometry analysis further revealed that 18 hr rigosertib treatment had induced the phosphorylation of Sos1 on at least four S/TP sites located in the Sos1 proline-rich (PR) domain: S1082, S1178, S1210, and S1275 (Figure 3C), thus identifying Sos1 as another Ras pathway component targeted by JNK in rigoser- tib-treated cells.Rigosertib-Induced S/TP Phosphorylation Alters Sos1 and Raf FunctionTo further characterize the effect of rigosertib on Ras/Raf/MEK/ ERK signaling, HeLa cells were treated for 18 hr with increasing concentrations of rigosertib, following which they were stimu- lated with EGF and analyzed. As shown in Figure 3D, detection of pS642P-C-Raf was first observed in cells treated with0.25 mM rigosertib and reached a maximum with 0.5 mM treat- ment, correlating with the appearance of slower migrating forms of the Rafs and Sos1 as well as activation of the JNK cascade, as monitored by pMKK4 levels. Moreover, when the hyperphos- phorylation of the Rafs and Sos1 reached peak levels (0.5– 5 mM), EGF-induced MEK activation was suppressed, whereas activation of the EGFR and autophosphorylation on Y1068 (the site that mediates recruitment of the Grb2/Sos1 complex) was largely unaffected.The above results suggest that the shifted Raf and Sos1 pro- teins are no longer responsive to upstream signals, a finding reminiscent of previous work showing that ERK-mediated feed- back phosphorylation of the Rafs and Sos1 inhibits their function in growth factor signaling.
For Sos1, the S/TP sites phosphory- lated in 18 hr rigosertib-treated cells are located in the PRdomain, which mediates binding to the Grb2 adaptor that, in turn, localizes Sos1 to membrane-bound receptor complexes, where it catalyzes Ras GDP/GTP exchange. Phosphorylation of this domain by ERK disrupts binding to RTK complexes, such that Sos1 can no longer promote Ras activation (Corba- lan-Garcia et al., 1996; Porfiri and McCormick, 1996; Rozakis- Adcock et al., 1995). To determine whether the rigosertib- induced phosphorylation of Sos1 might have a similar regulatory function, Sos1 activity was monitored in 18 hr rigosertib-treated cells by measuring the levels of active GTP-Ras generated following EGF stimulation (Figure 3E). Using GST-RafRBD as an affinity reagent to pull down active Ras, we found that EGF rapidly induced Ras GTP-loading in control cells, whereas little to no GTP-Ras could be detected in cells treated with rigosertib. Strikingly, when cells were co-treated with rigosertib and a JNK inhibitor, Ras-GTP loading induced by EGF stimulation was largely restored, and little to no shift in Sos1 mobility was observed.With regard to the Rafs, it has been shown that ERK-mediated feedback phosphorylation of S/TP sites N-terminal to the RBD can inhibit the Ras/Raf interaction, whereas phosphorylation of other S/TP sites in the regulatory and catalytic domains contrib- utes to the disruption of Raf dimers (Dougherty et al., 2005; Ritt et al., 2010).
Therefore, we next examined whether rigosertib treatment alters the ability of the Rafs to bind activated Ras in GST pull-down assays. As shown in Figure 3F, B-Raf and C-Raf isolated from cells treated for 2 or 4 hr with 2 mM rigosertib were fully competent to bind RasV12, whereas the hyperphos- phorylated Raf proteins isolated from cells treated for 18 hr showed a significant impairment in RasV12 binding that could be reversed when JNK activity was inhibited. Further analysis of two C-Raf mutants revealed that a C-Raf protein unable to dimerize because of a mutation in the Raf dimer interface (R401H-C-Raf) was still subject to the regulatory effects of S/TP phosphorylation and exhibited reduced RasV12 binding when isolated from 18 hr rigosertib-treated cells, whereas a C-Raf mutant in which the five serine residues in the SP sites were mutated to alanine (5A-C-Raf) was resistant (Figure 3G), confirming that the Ras/C-Raf interaction is disrupted byphosphorylation of these S/TP sites. Finally, when EGF-induced Raf dimerization and downstream signaling were monitored, only the shifted Raf proteins from cells treated with rigosertib for 18 hr had a reduced potential to dimerize and activate MEK, effects that again were largely reversed when JNK activity was blocked (Figure 3H).Given the inhibitory effects of rigosertib on the Ras/Raf inter- action and Raf dimerization, it would be predicted that rigosertib treatment would alter Raf signaling and MEK activation in cells expressing constitutively active Ras proteins. Indeed, as shown in Figure 3I, although basal pMEK levels were again elevated in cells treated overnight with rigosertib, no further increase in MEK activation was observed when activated G12V-K-Ras was expressed.
Moreover, this block in Raf signaling correlated with the hyperphosphorylation of the Rafs and activation of JNK cascade signaling, as evidenced by the presence of pMKK4. In contrast, prolonged 18 hr rigosertib treatment did not inhibit MEK activation in HeLa cells or melanoma lines expressing the V600E-B-Raf mutant, whose constitutive activity is not depen- dent on Ras binding or Raf dimerization (Figures 3J and 3K). It should be noted that even though V600E-B-Raf is insensitive to the effects of S/TP hyperphosphorylation and can function as a monomer to activate MEK, maintenance of MEK/ERK signaling alone is insufficient to overcome the cumulative pro- apoptotic effects of rigosertib described by Reddy et al. (2011) and Chapman et al. (2012). Thus, like cells expressing activated Ras alleles or those WT for Ras and Raf, cancer lines expressing V600E-B-Raf will undergo mitotic arrest and apoptosis when treated with rigosertib; however, the G2/M block occurs more slowly in the V600E-B-Raf lines (Figure S3).Downregulation of Ras Signaling by a Stress-Induced Phospho-Regulatory CircuitThe above findings suggest that rigosertib can modulate Ras pathway signaling in cancer cells, at least in part, by mediating the phospho-inhibition of key pathway components through an oxidative stress response that is associated with mitotic stress. Due to the potential therapeutic benefit of activating this phos- pho-regulatory circuit, we next examined whether other known inducers of mitotic and/or oxidative stress cause a similar response.
For these studies, we tested the following compounds: Taxol/paclitaxel, which blocks mitosis through microtubule sta- bilization and induces oxidative stress (Horwitz, 1994; Meshkini and Yazdanparast, 2012); KG5, which induces a mitotic block at prometaphase and has been reported to block C-Raf phos- phorylation on activating sites (Mielgo et al., 2011); and, as a negative control, the ATP-competitive B-Raf inhibitor SB- 590885. As expected, 18 hr treatment with either Taxol or KG5 in- hibited mitotic progression and induced mitochondrial ROS gen- eration (Figure 4A). JNK cascade activation was also observed in these cells as was S642P-C-Raf phosphorylation and shifts in the mobility of B-Raf, C-Raf, and Sos1 (Figure 4A), which could be reversed by phosphatase treatment (Figure S4A). In addition, the ability of these agents to induce S642P-C-Raf phosphoryla- tion and the mobility shifts of the Rafs and could be suppressed when JNK activity was inhibited (Figure 4B) or the antioxidant NAC was present (Figure S4B). Subsequent analysis of the micro- tubule disruptors, nocodazole, vincristine, and vinblastine, re-vealed that like Taxol and KG5, 18 hr treatment with these drugs also induced the hyperphosphorylation of the Rafs and Sos1 in a manner that correlated with JNK cascade activation, a block in cell-cycle progression, and mitochondrial ROS generation (Figures 4C, S4C, and S4D).
Interestingly, low levels of ROS induced by hypoxic conditions were not sufficient to activate JNK signaling or engage this phospho-regulatory circuit, even though HIF1a was stabilized (Figures S5A, S5C, and S5D).To investigate whether Ras pathway signaling was altered by Taxol or KG5 treatment, HeLa cells were treated with these drugs for either 2 or 18 hr, following which they were stimulated with EGF and analyzed (Figure 4D). As was observed for rigoser- tib, 2 hr treatment with these drugs caused minimal shifts in pro- tein mobility and had little effect on EGF-induced B-Raf/C-Raf dimerization or MEK activation. However, at the 18 hr time point, when the mobility shifts were maximal, Raf dimerization was significantly reduced in Taxol and KG5-treated cells as were pMEK levels. The shifted Raf proteins in these cells were also defective in their ability to bind active RasV12 (Figure 4F), and Sos1 function was likewise impaired (Figure 4E). Thus, other in- hibitors that activate JNK through a mitotic and oxidative stress mechanism can similarly suppress Ras signaling.Our findings indicate that JNK activation is critical for the stress- induced phospho-regulation of Ras/Raf/MEK/ERK signaling; therefore, we next addressed whether a physiological activator of the JNK cascade, such as TNFa, could induce the hyperphos- phorylation of the Rafs and Sos1. TNFa is known to activate JNK in two phases: an early transient phase that involves the TNF-R1 receptor and a late sustained phase that requires ROS genera- tion and is upregulated when NF-kB-induced expression of the SOD2 antioxidant is suppressed (Kamata et al., 2005; Sakon et al., 2003; Sluss et al., 1994).
When cells were treated with TNFa alone, JNK activation peaked at 20 min but was not observed by the 1 hr time point (Figure 5A). In contrast, treatment with TNFa in the presence of cycloheximide (to prevent transla- tion of SOD2) resulted in the sustained activation of JNK with pJNK levels detected at the 4 hr time point when increased ROS levels were observed (Figures 5A and 5C). During both phases of JNK activation, the S642P-C-Raf phosphorylation and shifts in Raf and Sos1 mobility were observed and correlated with JNK activity (Figures 5A and 5B). In addition, EGF-induced Ras/Raf/MEK/ERK signaling (as indicated by Raf dimerization and MEK activation) could be suppressed by both transient and sustained JNK activation (Figure 5D), suggesting that the hyperphosphorylated Raf and Sos1 proteins were again unre- sponsive to upstream signals. Subsequent analysis of other known activators of JNK signaling revealed that anisomycin addition or UV exposure rapidly induced JNK signaling as well as S642P-C-Raf phosphorylation and the shifts in Raf and Sos1 mobility (Figures S5B–S5D).
DISCUSSION
In this study, we have uncovered a new route for the phospho-in- hibition of Ras/Raf/MEK/ERK pathway signaling that is mediated by the stress-activated JNK cascade. This regulatory circuit was identified while investigating whether rigosertib, a cancer thera- peutic recently proposed to function as a Ras mimetic, could inhibit Ras-dependent Raf dimerization and signaling. In cell- based assays, we found that prolonged 18 hr treatment with rig- osertib could suppress Ras-dependent Raf signaling; however, the effects of rigosertib were largely indirect, mediated by the stress-induced hyperphosphorylation of the Rafs and Sos1 on inhibitory S/TP sites, that renders these key Ras pathway com- ponents unresponsive to upstream signals. Rigosertib has been analyzed in the scientific literature for more than 10 years, and it exhibits broad cytotoxicity in many hu- man cancer cell lines. Rigosertib was first reported to be an in- hibitor of the PLK1 kinase (Gumireddy et al., 2005); however, when no direct effect on PLK1 activity could be substantiated (Steegmaier et al., 2007), other mechanisms of rigosertib action emerged, including the recent description by Athuluri-Divakar et al. (2016) that rigosertib functions as a Ras mimetic, binding with high affinity to RBDs and blocking Ras/effector interactions. Although the findings presented in our study are in agreement with many of the experimental results presented in Athuluri-Diva- kar et al. (2016), the two studies come to different conclusions on how rigosertib mediates its effect.
In SPR analysis, we find that while rigosertib can bind the RafRBD, it is a low-affinity inter- action that only occurs at high concentrations. Interestingly, although this result conflicts with the high-affinity rigosertib/ RafRBD binding constant of 180 pM obtained by Athuluri-Divakar et al. (2016) using microscale thermophoresis, it is consistent with other analyses conducted by this group in which high concentra- tions of rigosertib were needed to demon- strate binding using differential scanning fluorimetry (100 mM) and NMR (250– 1,000 mM). Moreover, through in vitro competition experiments, we obtained no evidence by SPR analysis or prox- imity-based AlphaScreen assays that rig- osertib could directly inhibit the RafRBD/ Ras interaction, an issue not addressed by Athuluri-Divakar et al. (2016). Rather, in cell-based assays, we found that the inhibitory effects of rigosertib treatment on Ras/Raf/MEK/ERK signaling were indirect and dependent upon JNK cascade signaling.In our studies, we found that prolonged rigosertib treatment induced the hyperphosphorylation of the Rafs and Sos1 on mul- tiple S/TP sites and that the hyperphosphorylated forms of these proteins had altered signaling activity. More specifically, the hy- perphosphorylated Sos1 was impaired in its ability to facilitate Ras-GTP loading, whereas the hyperphosphorylated Rafs ex- hibited significantly reduced binding to active, GTP-Ras and were defective in their ability to dimerize and mediate MEK acti- vation. These findings are reminiscent of the altered function of Raf and Sos1 proteins hyperphosphorylated on S/TP sites dur- ing ERK-mediated feedback inhibition of growth factor signaling.
However, unlike growth factor signaling, ERK was not the major kinase mediating the S/TP phosphorylation in rigosertib-treated cells. We found that although the basal activity of ERK was elevated at all times of rigosertib treatment, and a low level of pS642P-C-Raf could be detected at early treatment times, full hyperphosphorylation of the Rafs and Sos1 as well as inhibition of Ras/Raf/MEK/ERK signaling required prolonged 18 hr rigoser- tib treatment and correlated with activation of the JNK cascade. It should be noted that although we did not detect activated p38 in our rigosertib-treated cells over the time course examined, we cannot rule out the possibility that the p38 MAPK may contribute to this phospho-regulatory circuit at later times of rigosertib- treatment or during other conditions of cellular stress. As further evidence that JNK is the primary kinase mediating the phospho-inhibition in rigosertib-treated cells, we found that purified, activated JNK could phosphorylate C-Raf on the iden- tified rigosertib-induced sites in vitro and that disrupting JNK activation or function could block the effects of rigosertib on Ras/Raf/MEK/ERK signaling. In particular, antagonizing the rigo- sertib-induced oxidative stress response by co-treating cells with the antioxidant NAC not only suppressed JNK cascade acti- vation but also inhibited the S/TP hyperphosphorylation of the Rafs.
Hyperphosphorylation of the Rafs was also reduced in can- cer lines lacking MKK4 (one of the two upstream kinases medi- ating JNK activation) and could be blocked when cells were co-treated with rigosertib and a JNK inhibitor. Importantly, inhib- iting JNK activity largely restored Raf and Sos1 function in rigo- sertib-treated cells, a finding inconsistent with the model that rig- osertib inhibits Ras signaling by functioning as a Ras mimetic, given that blocking JNK activity should have no effect on the abil- ity of rigosertib to bind RBDs and disrupt Ras/effector interac- tions. With regard to the study by Athuluri-Divakar et al. (2016), all cell-based assays were performed using cells treated for 18 hr with rigosertib, and because shorter treatment times were not examined, no data were generated showing that rigo- sertib has little to no effect on Ras-dependent ERK cascade signaling prior to the engagement of this phospho-regulatory cir- cuit, even though inhibition of PI3K/AKT signaling could be observed at early time points. Results presented here further indicate that this phospho-reg- ulatory circuit is not unique to rigosertib-treated cells but can also be induced by other agents that cause mitotic and oxidative stress, including KG5 and the microtubule toxins Taxol, nocoda- zole, vincristine, and vinblastine. Eighteen-hour treatment with any of these drugs inhibited the signaling activities of the Rafs and Sos1 in a manner that correlated with JNK activation and
S/TP hyperphosphorylation. Moreover, we found that TNFa, a physiological activator of JNK signaling, also invoked this phos- pho-regulatory circuit, with the hyperphosphorylation of the Rafs and Sos1 in TNFa-treated cells resulting in a reduced response to growth factor stimulation.
These observations, together with the findings that UV exposure as well as anisomycin treatment can induce Raf and Sos1 hyperphosphorylation in a manner that also corresponds with JNK activation, suggest that this regulatory circuit may be engaged by a variety of stress agents and may function as a checkpoint to block Ras/Raf/MEK/ERK signaling during conditions of cell stress (Figure 6).
Taken together, our findings elucidate an unexpected function of the stress-activated JNK cascade that may have important implications for cancer therapy and, in particular, for the use of JNK inhibitors. As with the complex role of the JNK cascade in cell signaling, the effects of this stress-activated phospho-regu- latory circuit may be context dependent and dual in nature. For example, when the stress is irreversible or insurmountable (as is the case for the microtubule toxins), blocking Ras activation and ERK cascade signaling may help promote apoptosis by pre- venting the transmission of survival and growth signals.
Alterna- tively, when the stress or stress damage is transient, the JNK- mediated S/TP phosphorylation could function as a signaling checkpoint, ensuring that either the stress has passed or that the stress-induced damage has been repaired before allowing cells to respond to growth signals transmitted by the Ras- GTPase. Knowledge of this phospho-regulatory circuit may pro- vide insight as to when JNK inhibitors would be beneficial or detrimental in cancer therapy, and engagement of this phos- pho-regulatory circuit may represent an alternative approach to delay or prevent drug resistance involving Ras-dependent Raf dimerization. Antibodies to B-Raf (H-145), C-Raf (C-12), b-Tubulin (H-235), and Sos1 (C-23) were from Santa Cruz Biotechnology; antibodies to pS217/221-MEK, pT202/Y204-ERK, pS473-AKT, pS/TP, pS257/T261-MKK4, pT183/Y185- JNK, pT180/Y182-p38, and pS621-C-Raf sites were from Cell Signaling Tech- nologies; antibodies to pERK were from Sigma; antibodies to Ras were from BD Biosciences. The pS642P-C-Raf antibody has been previously described (Dougherty et al., 2005). Rigosertib, paclitaxel/Taxol, SB-590885 (RAFi), vincristine, vinblastine, SP-600125 (JNKi), and JNK-IN-8 were from Selleck- Chem; KG5 was from Tocris Bioscience; N-acetylcysteine (NAC) was from Sigma-Aldrich. GST-RafRBD was from Millipore; GST-H-RasG12V was from Cell Sciences.
HeLa and A375 cells were grown in DMEM; MCF-7 cells in EMEM with 0.1mg/mL bovine insulin; A549, MOLT-4, SW-620, SK-Mel28, and U937 cells in RPMI-1640; and HCT-116 in McCoy’s 5a medium. All media were supple- mented with 10% fetal bovine serum (FBS) and 2 mM L-glutamine, and cells were cultured at 37◦C under 5% CO2. Drug treatments were conducted using healthy, cycling cells grown to approximately 80%–90% confluency. Growth media were removed and replaced with media containing 2 mM L-glutamine, 0.1% FBS, and the desired drug or inhibitor. For EGF stimulation, HeLa cells were treated with 100 ng/mL EGF for 5 min prior to lysis. Cells were examined for mitochondrial ROS generation by MitoSOX Red staining and were evalu- ated for cell-cycle arrest by fluorescence-activated cell sorting (FACS) anal- ysis, as described in Supplemental Experimental Procedures. Cell Lysis, Coimmunoprecipitation, and GST Pull-Down Assays Cells were lysed in 1% NP-40 buffer (20 mM Tris [pH 8.0], 137 mM NaCl, 10% glycerol, 1% NP-40 alternative, 0.15 U/mL aprotinin, 1mM phenylmethylsul- fonyl fluoride, 0.5 mM sodium vanadate, 20 mM leupeptin), and lysates were clarified by centrifugation. Equivalent amounts of protein lysate were incu- bated either with the appropriate antibody and protein G Sepharose beads (for coimmunoprecipitation assays) or with glutathione resins containing the indicated GST fusion protein (for GST pull-down assays) for 2 hr at 4◦C on a rocking platform. Complexes were washed extensively with 1% NP-40 buffer and examined by immunoblot analysis.
Surface Plasmon Resonance Binding SPR binding experiments were performed on a Biacore S200 instrument (GE). Avi-K-Ras (amino acids 2–188) loaded with the non-hydrolyzable GTP analog, GppNHp, or avi-tagged Raf-RBD (amino acids 55–131) were captured on CM5 sensor chips (GE) containing amine coupled Neutravidin (12,000–15,000 RU). Alternatively, for some experiments, Raf-RBD was coupled directly to the CM5 sensor chips using standard amide coupling chemistry. Samples were injected at 30 ml/min and for a contact time of 60 s. A 2-fold dilution series of rigosertib was prepared (50–0.39 mM) in 20 mM HEPES (pH 7.5), 150 mM NaCl, 0.05% Tween 20, 5 mM MgCl2, and 2.5% DMSO and injected over avi-RafRBD (350 RU). For competition experiments, 5 mM Raf-RBD was mixed with increasing concentrations of rigosertib (0.19–50 mM) and injected over avi-K- Ras-GppNHp (500 RU). RafRBD binding in the presence of rigosertib was compared with a calibration series in which Raf-RBD 0.15–20 mM (2-fold dilu- tions) was injected over the surface. The data were ON-01910 processed by subtracting binding responses on the reference flow cells as well as binding responses when buffer was injected. The samples were also corrected for DMSO mis- matches using a DMSO standard curve.