MEK guards proteome stability and inhibits tumor-suppressive amyloidogenesis via HSF1

SUMMARY

Signaling through RAS/MAP kinase pathway is central to biology. ERK has long been perceived as the only substrate for MEK. Herein we report that HSF1, the master regulator of the proteotoxic stress response, is a new MEK substrate. Beyond mediating cell-environment interactions, the MEK-HSF1 regulation impacts malignancy. In tumor cells, MEK blockade inactivates HSF1 and thereby provokes proteomic chaos, presented as protein destabilization, aggregation, and, strikingly, amyloidogenesis. Unlike their non-transformed counterparts, tumor cells are particularly susceptible to proteomic perturbation and amyloid induction. Amyloidogenesis is tumor-suppressive, reducing in vivo melanoma growth and contributing to the potent anti-neoplastic effects of proteotoxic stressors. Our findings unveil a key biological function of the oncogenic RAS-MEK signaling in guarding proteostasis and suppressing amyloidogenesis. Thus, proteomic instability is an intrinsic feature of malignant state and, disrupting the fragile tumor proteostasis to promote amyloidogenesis may be a feasible therapeutic strategy.

INTRODUCTION

Following environmental challenges, cells stimulate production of heat-shock proteins (HSPs). This HSP induction is the hallmark of the heat-shock, or proteotoxic stress, response (PSR) (Lindquist, 1986). As molecular chaperones, HSPs facilitate folding, transportation, and degradation of other proteins (Morimoto, 2008). In guarding the proteome against misfolding and aggregation, the PSR preserves proteostasis (Balch et al., 2008).

In vertebrates heat shock transcription factors (HSFs) govern the PSR. Among them is HSF1, the master regulator of this response (Morimoto, 2008; Xiao et al., 1999). As a multi-step process, HSF1 activation entails trimerization, nuclear translocation, posttranslational modifications, and DNA binding (Morimoto, 2008). Yet, our understanding of this process remains incomplete.

The HSF1-mediated PSR antagonizes many pathological conditions, including hyperthermia, heavy-metal toxification, ischemia and reperfusion, and oxidative damage, and impacts aging and neurodegeneration (Dai et al., 2012a). HSF1, not surprisingly, acts as a longevity factor (Hsu et al., 2003). In contrast, our and others’ work has revealed a pro-oncogenic role of HSF1 (Dai et al., 2007; Dai et al., 2012b; Jin et al., 2011; Meng et al., 2010; Min et al., 2007). Despite its dispensability under non-stress conditions, HSF1 is crucial for tumor cells’ growth and survival (Dai et al., 2007). Nonetheless, the mechanisms underlying its activation in malignancy remain unclear.

Herein we report that RAS-MEK-ERK signaling critically regulates the PSR. It is MEK that phosphorylates and activates HSF1. MEK inhibition destabilizes the proteome, provoking protein aggregation and amyloidogenesis. Combinatorial proteasome blockade potently augments this tumor-suppressive amyloidogenic effect. Hence, our findings not only suggest HSF1 as a new MEK substrate but also uncover a biological function of RAS-MEK-ERK signaling in governing proteostasis. Beyond shifting the canonical view of RAS-MEK-ERK signaling, our proof-of-concept experiments suggest that intrinsic proteomic instability associated with malignant state may be exploited to combat cancer.

RESULTS

MEK and ERK inversely regulate the PSR

Phosphorylation notably impacts HSF1 activation (Guettouche et al., 2005), suggesting a key role of signaling pathways. To illuminate how such pathways regulate the PSR, we first examined their responses to stress, focusing on RAS-MEK-ERK signaling. To inflict proteotoxic stress, we applied stressors with diverse mechanisms of action, including heat shock (HS), proteasome inhibitor MG132, histone deacetylase 6 inhibitor tubastatin, amino-acid analog azetidine, and HSP inhibitors (17-DMAG for HSP90 and VER155008 for HSP70) (Kawaguchi et al., 2003; Massey et al., 2010; Morimoto, 2008; Neckers and Workman, 2012). Transient exposure to stressors did not impair cell viability (Figure S1A), but elevated phosphorylation of MEK and ERK (Figure 1A), two key components of this pathway. MEK Ser218/222 and ERK Thr202/Tyr204 phosphorylation signify their active state (Dhillon et al., 2007; Roux and Blenis, 2004). Congruently, all stressors activated ELK1 (Figure 1B), a transcription factor downstream of ERK (Roux and Blenis, 2004).

Figure 1. MEK and ERK oppositely regulate the PSR.

(A) NIH3T3 cells were treated with HS at 43°C for 30min, 10μM tubastatin A for 5hr, 40μM VER155008 for 1hr, 500nM MG132 for 1hr, 200nM 17-DMAG for 1hr, and 2.5mM azetidine for 15min.

(B) The dual ELK1 reporter system, comprising a serum response element (SRE)-driven secreted embryonic alkaline phosphatase (SEAP) plasmid and a CMV-driven Gaussia luciferase (GLuc) plasmid, was transfected into HEK293T cells. After 24 hr, cells were treatments as in (A) and recovered overnight before measuring reporter activities (mean±SD, n=6, ANOVA).

(C and D) NIH3T3 cells were treated with 20μM U0126 or 20nM AZD6244 for 3 hr followed by HS and 4-hr recovery. mRNA levels were quantitated by qRT-PCR (mean±SD, n=3, Student's t-test).

(E) Immediately after HS, nuclear proteins of NIH3T3 cells treated as in (C) were extracted to measure HSF1-DNA binding by an ELISA-based assay (mean±SD, n=3, ANOVA).

(F) HEK293T cells were transfected with dual HSF1 reporter plasmids, a heat-shock element (HSE)-driven SEAP plasmid and a CMV-GLuc plasmid. After 24 hr, cells were treated with 20μM U0126, 20nM AZD6244, 1μM FR180204, or 100nM Sch772984 for 3 hr followed by HS for 30min and overnight recovery (mean±SD, n=6, ANOVA).

(G) HEK293T cells were treated with different inhibitors overnight.

(H)-(K) A LacZ or MEK isoform plasmid was co-transfected with dual HSF1 reporter plasmids into HEK293T cells transduced with lentiviral shRNAs. After 24 hr, cells were heat shocked at 43°C for 30min followed by overnight recovery (mean±SD, n=3, ANOVA).

(L)-(O) HEK293T cells were transfected with siRNAs for 48 hr followed by transfection with dual HSF1 reporter plasmids for 24 hr before HS (mean±SD, n=6, ANOVA).

See also Figure S1.

To determine whether MEK-ERK signaling regulates the PSR, we employed U0126 and AZD6244, two specific MEK1/2 inhibitors (Favata et al., 1998; Yeh et al., 2007). Both inhibitors impeded the HS-induced transcription of Hsp genes, and impaired the DNA-binding capacity and transcriptional activation of HSF1 (Figure 1C-1F and S1B), suggesting that MEK activates the HSF1-mediated PSR. ERK, phosphorylated by MEK (Ahn et al., 1991), is widely recognized as the master effector of this pathway (Dhillon et al., 2007; Roux and Blenis, 2004). Surprisingly, ERK inhibitors, FR180204 and Sch772984 (Ohori et al., 2005; Morris et al., 2013), activated HSF1 (Figure 1F and S1C). Both MEK and ERK inhibitors impaired two ERK-mediated events—MSK1 phosphorylation and ELK1 activation (Figure 1G and S1D; Roux and Blenis, 2004). While MEK inhibitors reduced ERK phosphorylation, two ERK inhibitors showed distinct effects (Figure 1G). Sch772984 suppressed ERK phosphorylation, likely due to ERK conformational changes that block MEK-mediated phosphorylation (Morris et al., 2013); conversely, FR180204 promoted ERK phosphorylation (Figure 1G), suggesting feedback MEK activation.

The impacts of MEK and ERK inhibitors on HSF1 were validated via genetic depletions of MEK and ERK (Figure 1I-1O). While depletion of one ERK isoform diminished the other isoform at the protein level (Figure 1L and 1N), mRNA levels of the isoform not targeted were elevated (Figure S1E-S1H), suggesting posttranscriptional mechanisms underlying reduced proteins. These results not only pinpoint RAS-MEK-ERK signaling as a key regulator of the PSR, but also reveal divergent impacts of MEK and ERK on HSF1.

MEK physically interacts with HSF1

To determine whether MEK directly activates HSF1, we examined endogenous MEK-HSF1 interactions by co-immunoprecipitation (co-IP). While no evident MEK1/2 proteins were precipitated with HSF1 without HS, HS caused a marked co-IP (Figure 2A and 2B), showing a stress-inducible MEK-HSF1 interaction. The mobility shift of HSF1 marks HS-induced phosphorylation (Figure 2A). These MEK-HSF1 interactions were verified via expression of recombinant proteins (Figure S2A and S2B). To determine whether MEK and HSF1 are in direct contact, we employed the Proximity Ligation Assay (PLA) technique (Clausson et al., 2011). Antibody specificities were validated by immunostaining (Figure S2C and S2D). In MEK-proficient cells, PLA signals were marginally visible without HS and HS intensified these signals (Figure 2C). In MEK-deficient cells, only faint signals were detected even after HS (Figure 2C), confirming the specificity of PLA. Of note, PLA signals were more manifest in the nucleus than in the cytoplasm (Figure 2C), revealing a prominently nuclear localization of interactions. These results strongly suggest a direct MEK-HSF1 association.

Figure 2. ERK, MEK, and HSF1 form a stress-inducible protein complex.

(A and B) After HS at 43°C for 30min, endogenous HSF1 proteins were precipitated from HEK293T cells. WCL: whole cell lysate; HC: heavy chain.

(C) Endogenous MEK1-HSF1 interactions were detected by PLA in HeLa cells using a rabbit anti-MEK1 antibody and a mouse anti-HSF1 antibody. Scale bars: 50μm for LM, 10μm for HM.

(D and E) Endogenous MEK-HSF1 interactions were detected by IP in HEK293T cells stably expressing shRNAs.

(F) Endogenous MEK-HSF1 interactions were detected in HEK293T cells transfected with siRNAs.

(G) Endogenous HSF1-MEK and HSF1-GFP-ERK1 interactions were detected in HEK293T cells transfected with indicated plasmids.

(H) Schematic depictions of three possible scenarios. P: phosphorylation.

(I) Immediately after HS, HSF1-ERK interactions were detected by co-IP.

(J) Endogenous ERK-HSF1 interactions were detected in HEK293T cells stably expressing shRNAs.

(K and L) Endogenous ERK-MEK and ERK-HSF1 interactions were detected by PLA in HeLa cells. Scale bar: 50μm for LM, 10μm HM.

See also Figure S2.

MEK1 and MEK2 form either homo- or heterodimers in vivo (Catalanotti et al., 2009). To address which type of dimer binds HSF1, we examined MEK1-HSF1 interactions in the deficiency of MEK2. Under HS more MEK1 proteins were precipitated with HSF1 in MEK2-deficient cells (Figure 2D). Similarly, MEK1 deficiency heightened MEK2-HSF1 interactions (Figure 2E), revealing a competition between the two MEK isoforms for HSF1 binding and suggesting that MEK homodimers can interact with HSF1.

ERK suppresses MEK-HSF1 interactions to inactivate HSF1

To elucidate how ERK inactivates HSF1, we first examined the impact of ERK on MEK-mediated HSF1 activation. Whereas ERK1 depletion promoted MEK-HSF1 interactions (Figure 2F), ERK1 overexpression mitigated these interactions and suppressed HSF1 (Figure 2G and S2E). Thus, we contemplated three possible scenarios (Figure 2H): 1) both MEK substrates, ERK and HSF1, compete for MEK interaction; 2) ERK, like MEK, binds HSF1 and thereby competes for HSF1 interaction; and 3) ERK inhibits MEK kinase activity towards HSF1. Each of the first two scenarios predicts competition between two protein complexes; in contrast, the third scenario predicts that ERK assembles with MEK and HSF1 into a single protein complex. Interestingly, under HS HSF1 precipitated both MEK and ERK (Figure 2I), and ERK precipitated both MEK and HSF1 (Figure 2J). Although these results do not exclude the existence of independent MEK-ERK and MEK-HSF1 complexes, they argue against the two complexes being stable and prevalent, as depicted in the first scenario. To test the second scenario, we detected ERK-HSF1 interactions by PLA, as this scenario predicts HSF1 as a substrate for both ERK and MEK. The specificity of ERK antibodies was validated in ERK-depleted cells (Figure S2F). In contrast to evident MEK1-ERK interactions (Figure 2K), no discernible PLA signals denoting ERK-HSF1 interactions were detected (Figure 2L), suggesting no direct contact between these two proteins. Moreover, while ERK1 overexpression mitigated MEK-HSF1 interactions, less ERK1 proteins were precipitated with HSF1 (Figure 2G), conflicting with heightened ERK1-HSF1 interactions predicted by the second scenario. Thus, these results not only refute the second scenario but also suggest that ERK complexes with HSF1 via MEK, in line with the third scenario. Importantly, MEK depletion markedly diminished ERK-HSF1 co-IP (Figure 2J). These results suggest existence of a protein complex comprising ERK, MEK, and HSF1, wherein ERK suppresses HSF1 indirectly, via inhibition of MEK.

MEK phosphorylates Ser326 to activate HSF1

Under HS HSF1 undergoes a series of phosphorylating events, among which Ser326 phosphorylation stimulates its activation (Guettouche et al., 2005). Yet, the identity of the kinase remains elusive. To determine whether MEK phosphorylates Ser326, we examined the effect of MEK blockade on this modification using a phosphospecific antibody that recognized HSF1^WT, but not HSF1^S326A, proteins (Figure S3A). Either MEK knockdown or U0126 treatment impaired Ser326 phosphorylation (Figure 3A and S3B). Conversely, a constitutively active mutant, MEK1^DD (S218D/S222D) (Brunet et al., 1994a), induced Ser326 phosphorylation and activation of HSF1 without HS (Figure 3B and 3C). ERK inhibition enhanced Ser326 phosphorylation; and, MEK depletion abolished this effect (Figure 3D and S3C), indicating MEK-dependent regulation. HSF1^S326A mutants displayed impaired transcription activities (Figure 3E), congruent with their defective nuclear translocation and DNA-binding capacity (Figure S3D and 3F). Moreover, HSF1 proteins were reduced in MEK-deficient cells (Figure 3A and S3E). To determine whether MEK impacts HSF1 stability, we performed cycloheximide chase experiments. MEK depletion shortened the half-life of HSF1^WT protein; importantly, while HSF1^S326A mutants were less stable in MEK-proficient cells, their stability was not evidently affected by MEK deficiency (Figure 3G), suggesting that MEK stabilizes HSF1 largely via Ser326 phosphorylation. These results indicate that MEK controls in vivo HSF1 Ser326 phosphorylation, a modification critical for its activation and stability.

Figure 3. MEK phosphorylates Ser326 to activate HSF1.

(A and B) HSF1 Ser326 phosphorylation was measured by immunoblotting in HEK293T cells stably expressing shRNAs or transfected with MEK1^DD plasmid.

(C) GFP or MEK1^DD plasmids were co-transfected with dual HSF1 reporter plasmids into HEK293T cells (mean±SD, n=5, Student's t-test).

(D) Control or ERK-targeting siRNAs, A (siERK1_1 and siERK2_1) and B (siERK1_3 and siERK2_2), were transfected into HEK293T cells stably expressing shRNAs.

(E) GFP or FLAG-HSF1 plasmids were co-transfected with dual HSF1 reporter plasmids into HEK293T cells stably expressing shRNAs (mean±SD, n=6, ANOVA).

(F) FLAG-HSF1 plasmids were transfected into HEK293T cells stably expressing HSF1-targeting shRNAs. HSF1-DNA binding was measured after HS as described in Figure 1E using anti-FLAG antibodies. The results were normalized against nuclear FLAG-HSF1 levels (mean±SD, n=3, Student's t-test).

(G) FLAG-HSF1 proteins were detected in HEK293T cells treated with 20μg/ml cycloheximide. Co-expressed GFP proteins served as internal controls.

(H) 100ng purified GST-MEK1 proteins were incubated with U0126 at RT for 20min followed by incubation with 400ng purified His-HSF1 proteins at RT for 30min. HSF1 phosphorylation was detected by immunoblotting.

(I) ERK complexes precipitated from HEK293T cells were treated with U0126 or FR180204, followed by incubation with 400ng His-HSF1, 400ng GST-ERK1, or 1000ng MBP proteins.

(J) Inactive GST-ERK1 proteins were incubated with 100ng GST-MEK1 and 400ng His-HSF1 proteins at RT for 30min.

(K) LacZ or GFP-ERK1 plasmid was co-transfected with MEK1^WT or MEK1^T292A,T386A plasmid into HEK293T cells stably expressing MEK-targeting shRNAs.

(L) HSF1 Ser326 phosphorylation was detected in HEK293T cells transfected with indicated plasmids.

(M) HSF1 activities were measured by the dual reporter system in HEK293T cells transfected with indicated plasmids (mean±SD, n=6, ANOVA).

(N) WM115 cells were treated with 20nM AZD6244 or 20μM U0126 overnight.

(O) HSF1 ChIP assays were performed using WM115 cells treated with DMSO or 20nM AZD6244 overnight. The results were normalized against the values of IgG controls (mean±SD, n=3, ANOVA).

See also Figure S3.

In vitro, recombinant MEK1 proteins directly phosphorylated HSF1 at Ser326; and U0126 blocked this event (Figure 3H). ERK was reported to phosphorylate HSF1 at Ser307 (Chu et al., 1998), implying a direct ERK-HSF1 interaction. To exclude direct Ser326 phosphorylation by ERK, we performed in vitro HSF1 phosphorylation using immunoprecipitated endogenous ERK complexes that would comprise ERK associated with or without MEK. Although precipitated complexes phosphorylated Ser326, this event was blocked by U0126, but not by FR180204 (Figure 3I). U0126, but not FR180204, blocked phosphorylation of recombinant ERK1 proteins by the same precipitates (Figure 3I). In contrast, FR180204, but not U0126, blocked phosphorylation of myelin basic protein (MBP), a known ERK substrate (Ahn et al., 1991), by the same precipitates (Figure 3I), showing ERK blockade by FR180204. These results strongly suggest that MEK, rather than ERK, directly phosphorylates Ser326. Moreover, recombinant ERK1 proteins impeded in vitro HSF1 Ser326 phosphorylation by recombinant MEK1 proteins (Figure 3J), consistent with the suppression of Ser326 phosphorylation by ERK in vivo. In contrast, ERK promotes HSF1 Ser307 phosphorylation. ERK depletion diminished Ser307 phosphorylation; however, this effect was largely abolished in MEK-deficient cells (Figure S3F), again indicating MEK dependence. MEK inhibition impaired Ser326 phosphorylation but enhanced Ser307 phosphorylation of HSF1^WT proteins (Figure S3G). Interestingly, phosphomimetic mutant HSF1^S326D proteins displayed reduced basal Ser307 phosphorylation and resisted induction of this phosphorylation by MEK inhibition (Figure S3G). These results support that Ser326 phosphorylation by MEK represses Ser307 phosphorylation, and that ERK impacts HSF1 Ser326 and Ser307 phosphorylation via MEK inhibition. Activated ERK phosphorylates Thr292/386 to inhibit MEK1 (Brunet et al., 1994b). MEK1^T292A,T386A mutants both heightened basal HSF1 Ser326 phosphorylation and blocked ERK-mediated suppression of this phosphorylation in cells depleted of endogenous MEK (Figure 3K), indicating that ERK suppresses Ser326 phosphorylation via feedback phosphorylation of MEK.

Interestingly, two MEK1 mutations identified in human melanomas, P124S and E203K (Nikolaev et al., 2012), caused constitutive HSF1 phosphorylation and activation (Figure 3L and 3M). Conversely, in human melanoma cells MEK inhibitors impaired constitutive HSF1 phosphorylation and binding to HSP promoters (Figure 3N and 3O). These results indicate that MEK controls both inducible HSF1 activation in stressed cells and constitutive HSF1 activation in malignant cells.

MEK preserves cellular proteostasis

HSF1 could maintain cellular proteostasis via HSPs. To examine the impacts of HSF1 on protein folding, we employed the glucocorticoid receptor (GR) as a model. Proper GR folding depends on HSP90 and misfolded proteins are cleared by the ubiquitin-proteasome system (Taipale et al., 2010). HSF1 knockdown induced GR-GFP ubiquitination and depletion (Figure S4A), indicating protein destabilization by HSF1 deficiency. This resulted from diminished cellular chaperoning capacity, as lysates of HSF1-deficient cells were less efficient in reactivating denatured luciferase (Figure S4B). Similarly to HSF1 deficiency, MEK blockade depleted GR-GFP; and this depletion is not due to GFP instability or general expression changes, since co-expressed GFP was not affected (Figure 4A). Instead, MEK blockade ubiquitinated GR-GFP (Figure 4B). This is not due to impaired proteasomal function, as proteasome inhibition by Bortezomib caused GR-GFP accumulation and MEK inhibitors did not affect proteasomal activities (Figure S4C-S4E). In fact, AZD6244 and MEK knockdown both depleted chaperoning capacity (Figure 4C and S4F), revealing modulation of protein folding and stability by MEK.

Figure 4. MEK preserves proteostasis.

(A) GFP and GFP-GR plasmids were co-transfected into HEK293T cells followed by treatments with 20nM AZD6244, 20μM U0126, or 200nM 17-DMAG for 4 hr.

(B) GFP-GR plasmids were co-transfected into HEK293T cells with HA-Ub-K48 plasmids, which encode a mutant ubiquitin that can be conjugated to protein substrates only via lysine 48. Following treatments with 20nM AZD6244 or 200nM 17-DMAG for 4 hr, GFP-GR proteins were precipitated and ubiquitination was detected using anti-HA antibodies.

(C) Denatured firefly luciferases were incubated with lysates of A2058 cells treated with DMSO or 20nM AZD6244 (mean±SD, n=4, ANOVA).

(D) A2058 cells were treated with 20nM AZD6244, and ubiquitinated proteins were detected in both detergent-soluble and –insoluble fractions using Lys48-specific ubiquitin antibodies.

(E) A2058 cells stably expressing LacZ or HSF1^S326D were treated with 20nM AZD6244 for 8 hr.

(F) C57BL/6J mice were i.p. injected with DMSO or AZD6244 three times a week for 2 weeks. S: spleen; K: kidney; L: liver.

(G) Experimental procedures of MS-based quantitation of ubiquitinated peptides, two technical replicates per treatment.

(H) Scatter plot of relative changes in peptide abundance between treated and control conditions. The green and red lines indicate 2.5-fold cutoffs.

(I) The classification of the 68 proteins was performed using the PANTHER gene list analysis tool (www.pantherdb.org).

(J) The Gene Ontology (GO) biological process enrichment analysis was performed using the web-based Enrichr software application.

(K) Interaction network of the 68 proteins. Known and predicted protein interactions were derived from the STRING database (www.string-db.org), and the network was visualized using Cytoscape software.

(L and M) V5-TOR1AIP2 or V5-RPL3 plasmids were co-transfected with HA-Ub-K48 plasmids into HEK293T cells. Following 20nM AZD6244 treatment for 8 hr, proteins were precipitated with anti-V5 antibodies.

(N) Following AZD6244 treatment, endogenous c-MYC proteins were precipitated from A2058 cells and immunoblotted with anti-ubiquitin antibodies.

(O and P) V5-RPL15 and V5-RPL3 plasmids were co-transfected with HA-Ub-K48 plasmids into HEK293T cells stably expressing shRNAs. Cells were treated with 500nM MG132 alone or co-treated with 20nM AZD6244 for 8 hr.

(Q) Endogenous RPL15 and RPL3 proteins were detected in A2058 cells treated with 20nM AZD6244 alone or co-treated with 500nM MG132.

(R) Endogenous RPL15 and RPL3 proteins were detected in A2058 cells stably expressing LacZ or HSF1^S326D with AZD6244 treatment.

See also Figure S4 and Table S1.

In line with a key role of HSF1 in governing cellular proteome, HSF1 depletion induced protein Lys48-specific ubiquitination, a modification marking proteins for proteasomal degradation (Pickart and Eddins, 2004), in both detergent-soluble and -insoluble fractions (Figure S4G). This change suggests global protein destabilization. Consistent with HSF1 inactivation, AZD6244 diminished Ser326 phosphorylation, reduced HSPs, and induced overall ubiquitination (Figure 4D). Overnight AZD6244 treatment also destabilized HSF1 (Figure 4D and S4H). MEK knockdown induced global ubiquitination as well (Figure S4I). Importantly, AZD6244 failed to deplete HSF1 and provoke ubiquitination in cells stably overexpressing HSF1^S326D (Figure 4E and S4J), indicating a causative role of HSF1 inactivation in protein instability due to MEK inhibition. In vivo MEK inhibition also depleted HSPs and HSF1, and provoked ubiquitination in primary tissues (Figure 4F).

To investigate ubiquitomic changes due to MEK inhibition, we conducted mass spectrometry (MS)-based analyses of ubiquitinated peptides enriched by a novel ubiquitin branch motif antibody (Figure 4G; Kim et al., 2011). We compared the ubiquitomes of A2058 cells treated with and without AZD6244 for 8 hours. In total, 3,425 non-redundant ubiquitinated peptides, assigned to 1,715 distinctive proteins, were profiled (Figure 4H, S4K, and Table S1). AZD6244 both increased and decreased peptide ubiquitination (Figure 4H). When a 2.5-fold cutoff was defined as the significant change, a collection of 76 non-redundant peptides were distinguished. These peptides represent 68 unique proteins that perform diverse molecular functions and engage in a wide array of biological processes (Figure 4I and 4J). Intriguingly, the most enriched pathway was translation elongation (Figure 4J). Analyses of these 68 proteins revealed a functional association network encompassing three subnetworks (Figure 4K). Of particular interest is the “translation” subnetwork that consists of 7 ribosomal subunit proteins, highlighting a prominent impact of MEK on ribosome machinery. Moreover, embedded within the network are several oncogenes and tumor suppressors, including c-MYC, Cyclin D1, HIF1A, TP53, and NF1 (Figure 4K). Although the mechanisms whereby MEK regulates these key players in oncogenesis are likely multifaceted, accumulating evidence has implicated HSPs in modulating their stabilities (Isaacs et al., 2002; MÜller et al., 2004). Thus, MEK could impact these proteins at least in part via HSF1.

To validate our MS findings, we elected several target proteins. Torsin-1A interacting protein 2 (TOR1AIP2) and ribosomal protein L3 (RPL3) exhibited 61.0- and 13.7-fold increases, respectively, in ubiquitination (Table S1). To facilitate detection, we expressed V5-tagged TOR1AIP2 and RPL3 proteins via a constitutive promoter. AZD6244 treatment for 8 hours did not alter levels of both V5-tagged proteins but increased their ubiquitination (Figure 4L and 4M). Our MS results also revealed decreased ubiquitination of proteins including c-MYC, RPL15, RPL24, and RPS20 (Table S1). We confirmed reductions in both ubiquitination and total levels of endogenous c-MYC proteins (Figure 4N). Similar results were also observed for V5-tagged RPL15, RPL24, and RPS20 (Figure S4L-S4N), suggesting shortened protein half-life. Indeed, proteasome blockade by MG132 prevented V5-RPL15 depletion by AZD6244 and revealed its elevated ubiquitination (Figure 4O). Consistent with a critical role of HSF1 inactivation, both V5-tagged RPL15 and RPL3 proteins were highly ubiquitinated following HSF1 knockdown, and AZD6244 subtly affected this ubiquitination (Figure 4O and 4P). Importantly, MG132 prevented depletions of endogenous RPL15 and RPL3 by AZD6244 and MEK knockdown (Figure 4Q and S4O), confirming destabilization of ribosomal proteins by MEK deficiency. While HSF1 knockdown diminished endogenous RPL15 and RPL3, HSF1^S326D expression elevated their basal levels and protected them from AZD6244-induced depletions (Figure S4P and 4R). These findings together indicate that MEK inhibition inactivates HSF1 to deplete cellular chaperoning capacity. In consequence, protein destabilization and ubiquitomic imbalance ensue.

MEK inhibition provokes aggregation and amyloidogenesis in malignant cells

Increased ubiquitination in detergent-insoluble fractions suggests protein aggregation (Figure 4D). To demonstrate this, we performed ubiquitin immunostaining. Melanoma cells were treated with Bortezomib and AZD6244 for 24 hours to provoke evident aggregation. As expected, bright fluorescent punctate foci emerged in Bortezomib-treated cells (Figure 5A), demarcating ubiquitin-containing aggregates. AZD6244 depleted HSF1 and, albeit to a lesser extent, induced punctate foci (Figure S5A and 5A). We theorized that blockade of proteasomal degradation of AZD6244-induced misfolded proteins would aggravate aggregation. Indeed, Bortezomib co-treatment augmented AZD6244-induced punctate foci (Figure 5A). We further confirmed the impact of MEK on aggregation using an expanded polyglutamine tract protein (polyQ79) (Sanchez et al., 2003). Proteins with expanded polyQ fragments are aggregation-prone and causally related to neurodegenerative disorders (Orr and Zoghbi, 2007). PolyQ79-expressing cells, as expected, contained large aggregates (Figure S5B). HSF1 knockdown and AZD6244 both enlarged polyQ aggregates; importantly, HSF1^S326D expression antagonized AZD6244-induced aggregate enlargement (Figure S5C and 5B). Bortezomib also enhanced aggregation, and combined treatment produced the largest aggregates (Figure 5B). Thus, both MEK and the proteasome suppress protein aggregation.

Figure 5. MEK and proteasome inhibition provoke protein aggregation and amyloidogenesis.

(A) WM115 cells treated with 20nM AZD6244, 100nM Bortezomib, or both for 24 hr were stained with Lys48-specific ubiquitin antibodies. Arrowheads mark ubiquitin-positive aggregates. Scale bar: 10μm. Amounts of aggregates per cell were quantitated using ImageJ (median, n≥100, ANOVA).

(B and C) Following transfection with polyQ79 plasmids alone or with both polyQ79 and HSF1^S326D plasmids for one day, HEK293T cells were treated with inhibitors as described in (A). Cells were either analyzed for aggregate size or stained with 10μM ThT.

(D) Treated tumor cell lines were stained with 10μM ThT. Geometric means were used to calculate fold changes in ThT fluorescence intensity and the log₂(FC) values were presented as a heat map.

(E) HEK293T cells were transfected with LacZ or polyQ79 plasmid. Following treatments, AOs were quantitated by ELISA using A11 antibodies (mean±SD, n=3, ANOVA).

(F) Intrinsic AOs were detected in human tumor cell lines (mean±SD, n=3, ANOVA).

(G and H) A2058 cells stably expressing LacZ or HSF1^S326D were treated for 24 hr. Amyloids were quantitated by ELISA (mean±SD, n=3, Student's t-test).

(I) 20μM synthetic Aβ1-42 peptides were incubated at RT with gentle shaking with 20μg lysates of A2058 cells treated with inhibitors. AF formation was monitored by ThT binding (mean±SD, n=3, ANOVA).

(J) For TEM studies (left panel 80,000X, right panel 200,000X), 20μM synthetic Aβ1-42 peptides were incubated with A2058 cell lysates in PBS at 37°C with gentle shaking for 2 days. Scale bars: 100nm.

(K) HEK293T cells stably expressing different shRNAs were stained with 10μM ThT.

(L and M) After pre-incubation with 10μM ThT for 6hr, A2058 cells were treated for 24 hr. Amyloids were quantitated (mean±SD, n=3, Student's t-test).

(N and O) After pre-incubation with 10μM ThT for 6 hr or transfection with 100ng A11 antibodies using JBS-Proteoducin for 16 hr, A2058 cells were treated for 24 hr. Viable cells were quantitated by CellTiter® blue (mean±SD, n=6, Student's t-test).

(P) A2058 cells stably expressing LacZ or HSF1^S326D were treated with DMSO or 20nM AZD6244. Viable cells were quantitated (mean±SD, n=6, ANOVA). Relative changes in viable cells after treatment were calculated by normalizing the values of AZD6244-treated cells against the values of DMSO-treated cells at each time point.

(Q)-(S) Following treatments with 20nM AZD6244, 100nM Bortezomib, or both for 24 hr, AOs were quantitated in primary MEFs and human cells (mean±SD, n=3, ANOVA).

(T and U) Cells were treated with 50μM Q-VD-OPh overnight and AOs were quantitated (mean±SD, n=3, Student's t-test).

See also Figure S5.

Aggregation-prone proteins can form amyloid fibrils (AFs) enriched for β-sheet structures (Eisenberg and Jucker, 2012). To assess whether HSF1 and MEK impact amyloid formation, we stained polyQ79-expressing cells with Thioflavin T (ThT) and Congo red (CR), two fluorescent dyes widely used to diagnose amyloids (Chiti and Dobson, 2006). PolyQ79 expression enhanced ThT and CR staining (Figure S5D and S5E), as expected. AZD6244, Bortezomib, and combined treatment further intensified this staining; and, HSF1^S326D expression antagonized the effect of AZD6244 (Figure 5C). Treatments also enhanced ThT and CR staining of human tumor cell lines (Figure 5D and S5F), suggesting emergence of endogenous amyloid-like structures. The presence of soluble amyloid oligomers (AOs) confirmed amyloidogenesis. AOs are believed to constitute a key toxic species in neurodegenerative disorders and can be detected by the conformation-dependent antibody A11 (Chiti and Dobson, 2006; Glabe, 2008; Kayed et al., 2003). Treatments not only exaggerated AO induction by polyQ79, but also provoked genesis of endogenous AOs in human tumor cell lines (Figure 5E and 5F). Similarly to AZD6244, HSF1 depletion induced endogenous AOs and AFs (Figure S5G and S5H). A previously characterized antibody, OC, was used to detect AFs (Kayed et al., 2007). Again, HSF1^S326D expression suppressed AZD6244-induced amyloidogenesis (Figure 5G and 5H).

A unique feature of amyloids is their ability to seed AFs (Chiti and Dobson, 2006). In amyloid seeding experiments, lysates of HSF1-depleted cells accelerated formation of Aβ AFs (Figure S5I). Similarly, lysates of cells treated with AZD6244, Bortezomib, and combination all exhibited augmented seeding efficacy (Figure 5I), which was confirmed using OC antibodies (Figure S5J). HSF1^S326D expression abolished the effect of AZD6244 (Figure 5I). Furthermore, transmission electron microscopy revealed that while seeding with DMSO-treated cell lysates resulted into disordered rod-like protofibrils, a dense lattice of fiber-like structures emerged following seeding with AZD6244-treated lysates; in comparison, seeding with Bortezomib-treated lysates produced mature fibrils, and a compacted network of fibrils were assembled after seeding with lysates from combined treatment (Figure 5J).

The amyloidogenic effects of AZD6244 and Bortezomib were validated genetically. Depletion of the β5 subunit (PSMB5) of the 26S proteasome, a primary target of Bortezomib (Oerlemans et al., 2008), caused accumulation of ubiquitinated proteins (Figure S5K). Mimicking pharmacological inhibitors, genetic depletions of MEK, PSMB5, or both all provoked amyloidogenesis (Figure S5L-S5Q and 5K).

To determine whether amyloids contribute to inhibitor-induced toxicities, we blocked amyloidogenesis with ThT, which impedes amyloid fibrillization via physical binding (Alavez et al., 2011). In melanoma cells, ThT suppressed amyloid induction by inhibitors, and improved cellular growth and survival by 50% (Figure 5L, 5M, and 5N). CR treatment and neutralization of AOs with A11 antibodies exerted similar protection (Figure S5R and 5O). Congruent with mitigated amyloidogenesis, HSF1^S326D expression not only stimulated the growth of melanoma cells but also rendered them refractory to MEK inhibition (Figure 5P).

Surprisingly, AZD6244 did not induce AOs in primary mouse embryonic fibroblasts (MEFs) and tissues (Figure 5Q and S5S). This is not due to inability to detect murine amyloids, as severe stress did induce AOs in murine cells (Figure S5T). These results suggest that non-transformed cells may be more refractory to amyloidogenesis than malignant cells. To assess this, we compared AO levels in primary human mammary epithelial cells (PHMC), immortalized human mammary epithelial (MCF10A) cells, and tumorigenic mammary epithelial (MCF7) cells treated with AZD6244, Bortezomib, and the combination. Each of these three treatments caused marked AO induction in MCF7 cells, slight induction in MCF10A cells, and no induction in PHMC (Figure 5R). A similar pattern was observed in primary human Schwann cells (PHSC) and their malignant counterparts, 90-8TL and S462 cells (Figure 5S). Immortalized and transformed cells, except S462, also showed elevated basal levels of AOs (Figure 5R and 5S). Intriguingly, AO levels positively correlated with malignant states (Figure 5T), supporting proteomic imbalance as an intrinsic feature of malignancy. We theorized that the lack of elevated basal AOs in S462 cells might be due to amyloid-associated toxicity. Indeed, blockade of cell death by a pan-caspase inhibitor elevated AO levels in immortalized and transformed cells, revealing heightened amyloidogenesis; in contrast, it did not elevate AOs in primary cells (Figure 5T and 5U), supporting an absence of amyloidogenesis. These results indicate that malignant cells are distinctively susceptible to amyloidogenesis.

Combined proteasome and MEK inhibition disrupts tumor proteostasis and suppresses malignancy

MEK and proteasome inhibition, individually, disturbed proteostasis in tumor cells to certain degrees; however, the combination of both augmented this effect and, accordingly, markedly impaired the growth and survival of human tumor cell lines (Figure 6A). Of note, this combination did not impact primary cells (Figure 6A).

Figure 6. Combined MEK and proteasome inhibition exerts potent tumor-suppressive effects.

(A) After treatments with 20nM AZD6244, 100nM Bortezomib, or both for 24 hr, viable cells were quantitated by CellTiter® Blue (mean±SD, n=6, ANOVA).

(B and C) 1×10⁶ A2058 cells were s.c. injected into NOD/SCID mice. After 7 days, mice were treated with DMSO, 5mg/Kg AZD6244, 0.5mg/Kg Bortezomib, or the combination via i.p. injection three times a week. Tumor volumes were measured using a caliper weekly (mean±SEM, ANOVA). Tumor growth curves were fitted to exponential growth models to derive tumor-doubling time (DT). Kaplan-Meier survival curve was plotted for each group (Log-rank test).

(D) Proteins were detected by immunoblotting, 3 tumors per group.

(E) Tumor lysates were used to quantitate AOs, 5 tumors per group (mean±SD, n=3, ANOVA).

(F) Tumor lysates were used to seed Aβ1-42 peptides, 5 tumors per group (mean±SD, n=3, ANOVA).

(G) Tumor sections were stained with CR, 5 tumors per group. Ten random fields were taken for each section. Scale bar: 50μm. Total CR fluorescence in each field was quantitated using ImageJ and normalized against total nuclei (median, n=50, ANOVA).

(H) Following CR staining, tumor sections were visualized under polarized light microscopy. Scale bar: 50μm.

(I) Following staining with AF-specific antibodies (OC), sections of tumors receiving combined treatment were further stained with CR. Scale bar: 50μm.

(J) 1×10⁶ A2058 cells stably expressing firefly luciferase transgene were i.v. injected into NOD/SCID mice. Treatments were initiated one day after as described in (B) for 6 weeks. Body weights were monitored weekly (mean±SD, n=10, ANOVA).

(K) Detection of metastases by in vivo bioluminescence imaging.

(L) Representative micrographs illustrate metastatic melanomas in the lung, skeletal muscle, pelvic adipose tissue, and ovary. T: tumors; L: lung; M: muscle; B: bone; A: adipose tissue; OF: ovarian follicle. Scale bar: 500μm.

(M) Combined MEK and proteasome inhibition prevents melanoma metastasis (Barnard's exact test).

See also Figure S6 and Table S2.

In vivo, whereas low doses of AZD6244 or Bortezomib alone exhibited no significant impacts on xenografted melanomas, the combination potently retarded their growth (Figure 6B and S6A). All mice receiving the combined treatment remained alive and their body weights remained constant; in contrast, all mice in the other groups died and lost about 25% of body weight (Figure 6C and S6B). AZD6244 or Bortezomib alone slightly elevated ubiquitination in tumors; however, the combination markedly aggravated this effect (Figure 6D and S6C). While Bortezomib induced HSF1 Ser326 phosphorylation and HSP expression, AZD6244 co-treatment suppressed this stress response and induced caspase 3 cleavage (Figure 6D and S6C). Accordingly, AOs were evidently elevated in tumors receiving combined treatment (Figure 6E). Of particular interest is an inverse correlation between amounts of AOs and tumor masses (Figure S6D), supporting an adverse impact of AOs on malignant growth. Congruent with amyloidogenesis, tumors receiving combined treatment displayed potent seeding capacities and enhancement of CR staining (Figure 6F and 6G). Intratumoral AFs were further demonstrated by the hallmark birefringence of CR staining (Figure 6H), ThT labeling (Figure S6E), and immunostaining with OC antibodies (Figure 6I). In contrast, the combined treatment did not induce AOs and apoptosis in primary tissues of the same tumor-bearing mice, despite elevated ubiquitination (Figure S6F and S6G).

To investigate whether the combined treatment impedes metastasis, we intravenously injected melanoma cells expressing a luciferase transgene into NOD/SCID mice. During a 6-week period, only mice receiving combined treatment gained body weight (Figure 6J), suggesting improved health. In vivo imaging detected luminescence in 40% of mice treated with DMSO, AZD6244, or Bortezomib alone (Figure 6K). Histological examination confirmed melanoma metastases to the lung, skeletal muscle, adipose tissue, and ovary (Figure 6L and Table S2). In contrast, none of the mice receiving combined treatment displayed discernible luminescence or metastases (Figure 6K and 6M). Together, these results demonstrate that combined MEK and proteasome inhibition provokes proteotoxic stress and amyloidogenesis within tumors, and exerts robust anti-neoplastic effects.

Amyloidogenesis is tumor-suppressive

Evident apoptosis in tumor regions showing intense CR staining suggests a causative role of amyloidogenesis in treatment-induced toxicity (Figure 7A). To confirm this, we blocked intratumoral amyloid induction via in vivo CR administration. CR not only accelerated melanoma growth but also potently antagonized the tumor suppression imposed by combined MEK and proteasome inhibition (Figure 7B). Penetration of CR into tumor tissues was indicated by intense light absorption of tumor lysates at 498nm (Figure 7C), a characteristic of this amyloid stain (Sladewski et al., 2006). Congruent with enhanced malignancy, CR treatment enlarged tumor masses, deteriorated body conditions, and shortened animal survival (Figure S7A, S7B, and 7D).

Figure 7. Amyloidogenesis suppresses tumor growth.

(A) Sections of melanomas receiving combined treatment were stained with cleaved caspase 3 antibodies followed by CR staining. Arrowheads and arrows indicate condensed and fragmented nuclei, respectively. Scale bar: 50μm.

(B) 1×10⁶ A2058 cells were s.c. injected into NOD/SCID mice. After 7 days, mice were treated with 1mg/30g CR via i.p. injection one day prior to combined treatment. Tumor volumes were measured weekly (mean±SD, ANOVA).

(C) Lysates of CR-treated tumors exhibited absorbance at 498nm, 3 tumors per group (mean±SD, n=3). Lysis buffer containing CR served as a positive control.

(D) Kaplan-Meier survival curves were compared (Log-rank test).

(E and F) Both detergent-soluble and -insoluble fractions of tumor lysates were used to quantitate amyloids, 3 tumors per group (mean±SD, n=3, Student's t-test).

(G) Proteins were detected by immunoblotting, 3 tumors per group.

(H) Schematic depiction of the interplay among MEK, ERK, and HSF1, and its role in regulating proteome stability. Balanced proteostasis suppresses toxic protein aggregation and amyloidogenesis, thereby facilitating tumorigenesis.

See also Figure S7.

While CR reduced amyloids in tumor tissues, it did not diminish ubiquitination (Figure 7E, 7F, and 7G). These results indicate no interference of CR with MEK and proteasome inhibitors, and further support a specific action of CR in blocking amyloid genesis. In accordance with accelerated growth, CR-treated tumors displayed reduced caspase 3 cleavage (Figure 7G). Collectively, these results strongly suggest that amyloidogenesis is tumor-suppressive and evidently contributes to the anti-neoplastic effects of combined MEK and proteasome inhibition.

DISCUSSION

HSF1 is a new MEK substrate

Unexpectedly, our results reveal HSF1 as a physiological substrate for MEK, challenging the prevailing paradigm wherein ERK exclusively instigates the effects of RAS-RAF-MEK signaling. Our results further show that MEK activates but ERK inactivates HSF1. Importantly, our findings integrate these two seemingly contradictory actions and support the assembly of a ternary ERK-MEK-HSF1 protein complex. In aggregate, our findings propose a bifurcated, rather than a linear, RAS-RAF-MEK cascade. MEK, as a central nexus, both conveys upstream stimuli and governs two discrete but interconnected downstream effector pathways, of which one is mediated by ERK and the other by HSF1 (Figure 7H). In a negative feedback fashion, ERK finely attunes HSF1 activation via inhibitory phosphorylation of MEK (Figure 7H). While our studies focused on MEK-mediated Ser326 phosphorylation, other kinases can also regulate HSF1.

Guarding of proteostasis by RAS-RAF-MEK signaling

Our findings uncover a new function of RAS-RAF-MEK signaling in regulating proteostasis. Diverse proteotoxic stressors commonly activate MEK (Figure 1A). Through HSF1 activation, RAS-RAF-MEK signaling heightens cellular chaperoning capacity to guard proteomic integrity.

MEK-HSF1 regulation could have key physiological implications. Mitogens stimulate RAS/MAPK signaling and downstream mTORC1 (Laplante and Sabatini, 2012). However, heightened protein synthesis driven by mTORC1 encumbers cellular protein quality-control machinery. It thus appears necessary for mitogens, via MEK, to concurrently mobilize the HSF1-controled chaperone system to ensure productive protein synthesis and, thereby, avert proteomic imbalance. Interestingly, MEK also governs translation capacity via HSF1 (Figure 4K). Thus, RAS-RAF-MEK signaling synchronizes protein quantity- and quality-control machineries to support cellular growth.

It is also tempting to speculate that RAS-RAF-MEK signaling may antagonize protein-misfolding diseases, such as amyloidosis, via guarding proteostasis.

Proteomic instability of cancer

Our findings pinpoint a pro-amyloidogenic nature of malignant state. The susceptibility of malignant cells to amyloid genesis likely originates from their debilitated proteostatic state, which is particularly vulnerable to perturbations. Unlike non-transformed cells, malignant cells constantly endure proteomic imbalance, evidenced by elevated basal levels of amyloids (Figure 5T and 5U). Accordingly, HSF1, otherwise latent in primary cells, is constitutively mobilized in tumor cells to retain the fragile proteomic equilibrium (Figure 7H). Proteomic chaos inevitably ensues following either HSF1 inactivation or even mild proteotoxic insults. Hence, tumors cells rely on HSF1 to sustain their malignant phenotypes (Dai et al., 2007). In contrast, the lack of intrinsic proteotoxic stress empowers primary cells to effectively buffer intense proteomic fluctuations and thereby avert deleterious consequences—aggregation and amyloidogenesis.

Excitingly, the distinct susceptibilities to proteomic perturbation between primary and malignant cells may be exploited to combat malignancy. Our findings support important roles for proteotoxic stress and amyloidogenesis in the toxicity of MEK inhibition in malignancy. Through protein destabilization, MEK inhibitors act as a proteotoxic stressor, mechanistically distinct from proteasome inhibitors. When applied as single agent, a MEK or proteasome inhibitor is incompetent to distress tumor proteostasis. However, combinatorial application exerts a profound impact, eliciting amyloidogenesis. Importantly, our findings strongly suggest a tumor-suppressive nature of amyloidogenesis (Figure 5L-5O and 7B). These findings imply that amyloidogenesis, indicative of grave proteomic imbalance, may be of prognostic value in monitoring tumor progression and evaluating therapeutic responses. Conceptually, our findings suggest that proteomic instability is an intrinsic characteristic associated with malignant state and that, therefore, disruption of fragile tumor proteostasis may be a feasible therapeutic strategy.

EXPERIMENTAL PROCEDURES

Proximity ligation assay

Cells were fixed with 4% formaldehyde in PBS for 15 min at RT. After blocking with 5% goat serum in PBS with 0.3% Triton X-100, cells were incubated with a pair of rabbit and mouse primary antibodies 1:200 diluted in the blocking buffer overnight at 4°C. Following incubation with Duolink^® PLA^® anti-rabbit Plus and anti-mouse Minus probes (OLINK Bioscience) at 37°C for 1hr, ligation, rolling circle amplification, and detection were performed using Duolink^® In Situ Detection Reagents Red (OLINK Bioscience). Nuclei were stained with Hoechst 33342. Signals were visualized using a Leica TCS SP5 confocal microscope.

CR and ThT staining of tumor sections

Following deparaffinization and rehydration, tumor sections were stained with 0.5% CR in PBS at RT for 20min followed by differentiation in alkaline solutions (0.01% NaOH, 50% alcohol). Nuclei were stained with either Hoechst 33342 or hematoxylin. Fluorescence was visualized using a Leica TCS SP5 confocal microscope and the birefringence visualized using a Leica DM5000B upright microscope equipped with polarized light filters. For ThT staining, sections were stained with 0.2% ThT in PBS at RT for 10min, rinsed in 1% acetic acid for 2min, and washed with ddH₂O for 3 times. Nuclei were stained with SYTO® 62 (Life Technologies)

Melanoma xenograft models

A2058 cells were s.c. injected into the left flanks of 9-week-old female NOD.CB17-Prkdc<scid>/J (NOD/SCID) mice (The Jackson Laboratory). For CR treatment, mice were i.p. injected with PBS or CR one day prior to combined AZD6244 and Bortezomib treatments. Tumor volumes were calculated following the formula 4/3πR³. For experimental metastasis, engineered A2058 cells were transplanted into 10-week-old female NOD/SCID mice via tail vein injections. All mouse experiments were performed under a protocol approved by The Jackson Laboratory Animal Care and Use Committee.

Statistical methods

All statistical analyses were performed using Prism 6.0 (GraphPad software). Statistical significance: *p< 0.05; **p<0.01; ***p<0.001.