Regulation of BRAF protein stability by a negative feedback loop involving the MEK-ERK pathway but not the FBXW7 tumour suppressor.

The (V600E)BRAF oncogenic mutation is detected in a wide range of human cancers and induces hyperactivation of the downstream MEK-ERK signalling cascade. Although output of the BRAF-MEK-ERK pathway is regulated by feed-forward RAF activity, feedback control also plays an important role. One such feedback pathway has been identified in Caenorhabditis elegans and involves ERK-mediated phosphorylation of BRAF within a CDC4 phosphodegron (CPD), targeting BRAF for degradation via CDC4 (also known as FBXW7), a component of the SKP1/CUL1/F-box (SCF) E3 ubiquitin ligase complex. Here we investigate this pathway in mammalian cells. Short-term expression of autochthonous (V600E)BRAF in mouse embryonic fibroblasts (MEFs) leads to down-regulation of BRAF protein levels in a proteasome-dependent manner and (V600E)BRAF has a reduced half-life compared to (WT)BRAF in HEK293(T) cells. These effects were reversed by treatment with the MEK inhibitor PD184352. We have identified the equivalent CPD at residues 400-405 in human BRAF and have found that mutation of ERK phosphorylation sites at residues T401 and S405 in (V600E)BRAF increases the half-life of the protein. While BRAF and FBXW7 co-immunoprecipitated, the overexpression of FBXW7 did not influence the half-life of either (WT)BRAF or (V600E)BRAF. Furthermore, disruption of the substrate-binding site of mouse FBXW7 using the R482Q mutation did not affect the interaction with BRAF and the expression levels of (WT)BRAF and (V600E)BRAF were not altered in MEFs derived from mice with the homozygous knockin (R482Q)FBXW7 mutation. Overall these data confirm the existence of a negative feedback pathway by which BRAF protein stability is regulated by ERK. However, unlike the situation in C. elegans, FBXW7 does not play a unique role in mediating subsequent BRAF degradation.

Introduction

The BRAF protein kinase is a key component of the RAF–MEK–ERK signalling cascade. This pathway plays a major role in controlling numerous cellular events including cell proliferation, survival, differentiation and migration [1], [2]. Tight regulation of the pathway is vital for normal cell, tissue and whole body homeostasis. This is evidenced by the fact that hyperactive, oncogenic forms of BRAF are prevalent in several human cancers [3]. The mutation is the most common BRAF mutation detected in human cancers and is thought to mediate its transforming effects by deregulation of the MEK–ERK pathway [4], [5], [6]. Loss of function mutations in components of the pathway such as BRAF also give rise to embryonic lethality with a failure to thrive associated with placental failure [7], [8].

The mechanisms of regulation of the ERK pathway have been subjected to extensive investigation with evidence showing that the pathway can be controlled by feed-forward and feedback loops [9], [10], [11]. Regulatory loops fall into immediate and late temporal domains. The immediate responses include protein–protein interactions such as dimerization [12], [13], [14] and covalent protein modifications particularly cycles of phosphorylation and dephosphorylation [15], [16], [17], both of which influence RAF activation/deactivation and subsequent downstream signalling. Late events involve newly synthesised proteins, for example the induction of expression of adapters SPROUTYs and SPREDS [18], [19], [20] as well as phosphatases [21], [22], [23] that have been shown to suppress ERK pathway activity through feedback mechanisms. The activity of feedback control pathways is important in cancer development as evidenced by the fact that disabled feedback inhibitory pathways are detected in V600EBRAF transformed cells [24], while clinical and experimental evidence is consistent with most negative regulators of the ERK pathway being tumour suppressors [25].

Regulation of protein stability is an important factor in controlling signalling pathway output. An example of this is for the EGF receptor whereby ligand-induced autophosphorylation allows recruitment of the CBL E3 ubiquitin ligase, which controls EGFR internalisation and degradation [26], [27]. With regard to RAF, the correct folding and stabilisation of the proteins is dependent on the molecular chaperone HSP90 complex [28] and pharmacological inhibition of HSP90 leads to their ubiquitin-mediated degradation, particularly for V600EBRAF which shows a greater dependence on HSP90 than WTBRAF, CRAF or ARAF [29], [30]. A study using siRNA has shown a requirement for the E3 ubiquitin ligase Cullin-5 in mediating V600EBRAF degradation following HSP90 inhibition [31]. The Ring finger protein 149 (RNF149) has also been proposed to be an E3 ubiquitin ligase active operating on the kinase domain of WTBRAF [32] while a further study in Caenorhabditis elegans identified that the BRAF homologue, LIN-45, is a substrate for the multiprotein E3 ubiquitin ligase Skp1/Cul1/F-box (SCF) complex [33]. The F-box containing substrate receptor SEL-10 (FBXW7 in mammals) was shown to target LIN-45 through a conserved Cdc4 phosphodegron (CPD) and, additionally, the ERK homologue MPK-1 was found to be required for controlling LIN-45 degradation through the CPD in a negative feedback loop [33].

Here, we have investigated BRAF protein turnover in mammalian cells. Using MEFs derived from mice bearing a conditional knockin allele for V600EBRAF [34] we show that expression of V600EBRAF leads to downregulation of BRAF protein expression. This downregulation is not associated with alterations in Braf mRNA levels, but can be rescued by proteasome and MEK inhibition. Ectopically expressed V600EBRAF also has a shorter half-life by ~ 3 h than WTBRAF and this can be rescued by MEK inhibition. A conserved CPD at residues 400–405 in human BRAF was identified and we show that the feedback regulation can in part be explained by ERK-mediated phosphorylation of T401 and S405 since mutation of these residues increases the half-life of V600EBRAF. However, although BRAF has the capability to bind to FBXW7, FBXW7 over-expression or loss of function does not alter either WTBRAF or V600EBRAF protein stability, suggesting that this E3 ubiquitin ligase component is not uniquely involved in regulating mammalian BRAF turnover. We also demonstrate an association between this novel feedback pathway and the oncogene-induced senescence phenotype.

Materials and methods

Animal work

Braf+/ mice have been described previously [34] and the induction of lung tumour development by intercrossing with the CreER™ strain [35] has been reported in [36]. The conditional knockin Fbxw7+/ mice have been previously reported [37]. All animal experiments were performed according to local ethical and UK Home Office guidelines, under regulatory approval. Haematoxylin and Eosin (H&E) staining of paraffin-embedded sections was performed as described [34]. PCR genotyping of BRAFV600E floxed and Cre-recombined alleles was performed using methods and primers described in [34] and detection of the Cre-recombined allele for the Fbxw7 alleles using PCR is described in Davis et al. 2011 [37]. MEFs were derived by timed matings between relevant mouse strains and embryos were harvested at embryonic day 14.5 as described [38].

Cell culture and treatments

MEFs were cultured in DMEM with 10% (v/v) Foetal Calf Serum (FCS) and 1% (v/v) penicillin and streptomycin (P-S) at 10% CO2. They were plated at a density of 2x105 cells/well of a 6-well dish and infected with 4 × 107 PFU of Adenoviral Cre (AdCre) for 2 h in 2 ml of media lacking FCS and P-S. Media was replaced and cells were cultured for up to 96 h. For MEK inhibition, 1 μM PD184352 in DMSO was added 48 or 72 h after AdCre infection. In both cases, cells were harvested 24 h after PD184352 addition. For proteasome inhibition, 30 μM MG132 in DMSO or 0.5 μM Epoxomicin in DMSO were added for the last 5 h before harvesting. HEK293T cells were cultured in MEF media and transfected with 5 μg plasmid using Lipofectamine 2000 in accordance with the manufacturer's instructions (Invitrogen). Cells were incubated for 48 h before harvesting. NIH3T3 cells were cultured as for MEFs and transfected with expression vectors using Lipofectamine 2000. Human cancer cell lines were grown in MEF media and treated with or without 1 μM PD184352 in DMSO for 6–48 h before harvesting.

Western blot analysis

Cells and lung tissue were lysed with gold lysis buffer (GLB) as described previously [39]. The soluble fraction (SF) was obtained by taking the supernatant following centrifugation at 13,000 rpm for 10 min at 4 °C. The insoluble fraction (IF) was obtained by treating the pellets with 1 × Sample Buffer (0.05 mM Tris pH 6.8, 2% (v/v) SDS, 0.1% Glycerol), vortexing for 1 min and subsequently boiling for 5 min at 95 °C. Protein concentrations were measured by the Bradford assay (SF) or BCA protein assay kit (IF), both obtained from Pierce and following the manufacturer's guidelines. The following antibodies were used for analysis: BRAF (Santa Cruz 5284), ERK2 (Santa Cruz SC154), Phospho-ERK1/2 (Cell Signalling 9101), Phospho-MEK1/2 (Cell Signalling 9154S), GAPDH (EMD Millipore MAB374), β-ACTIN (Sigma S-A2103), MYC-TAG (Santa Cruz SC40), GFP (ABcam AB6556) and FLAG (Sigma F3165).

Plasmids and site-directed mutagenesis

Plasmids used for half-life experiments were either the pEF Myc-tagged BRAFWT or pEF Myc-tagged BRAFV600E expression vectors as previously described [3]. Site directed mutagenesis was performed using the Gene Tailor site-directed mutagenesis system (Invitrogen). For co-immunoprecipitations, vectors expressing GFP-BRAFWT, GFP-BRAFV600E or empty GFP vector were used together with vectors expressing WTFBXW7 or R482QFBXW7.

Immunoprecipitations

HEK293T cells were transfected with the appropriate plasmid vectors and GLB soluble protein lysates were generated. For GFP vectors, 200 μg protein lysate was incubated with 20 μl GFP-TRAP beads (Chromotek) in a volume of 500 μl with GLB lysis buffer rotating at 4 °C overnight. For Myc-tagged vectors, 200 μg protein lysate was incubated with 15 μl of the 9E10 antibody (200 μg/ml; Santa Cruz SC40) together with 25 μl of Dynabeads protein G (Life Technologies) in a volume of 500 μl with GLB lysis buffer rotating at 4 °C overnight. Immunoprecipitated proteins were washed three times in GLB lysis buffer before resuspension in Laemmli buffer and boiling at 95 °C for 5 min.

Pulse–chase experiments

HEK293T cells were transiently transfected with relevant vectors as above in triplicate. 48 h after transfection cells were washed twice with sterile PBS and cultured in 3 ml of Met/Cys-free Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 2 mM l-glutamine (Sigma) and 5% (v/v) dialysed FCS (Invitrogen) for 1 h at 37 °C. The cells were labelled with 50 μl of EXPRE35S35S protein labelling mixture (100 μCi/dish; PerkinElmer Life Sciences) for 3 h at 37 °C. After extensive washing, chase was initiated by adding 3 ml of MEF media to each dish and cells were lysed at appropriate time points in 500 μl of cold radioimmune precipitation assay buffer (20 mM Tris pH 7.5, 150 mM NaCl, 2 mM EDTA, 10% (v/v) glycerol, 1% (v/v) NP40, 1 mM phenylmethylsulfonyl fluoride, 1 mM NaV3O4, and 1 mM NaF). The samples were left on ice for 30 min to allow complete lysis, and then centrifuged at 13,000 rpm at 4 °C for 10 min. Supernatants were collected and protein levels quantitated using the Bradford assay (Pierce). Samples were subjected to immunoprecipitation as described above, followed by SDS-PAGE. Gels were fixed in 50% (v/v) methanol and 10% (v/v) acetic acid for 30 min, washed with water several times, dried under vacuum and then exposed to a phosphoimager screen overnight. For the MEK inhibitor experiments, 1 μM PD184352 or DMSO carrier control was added at the beginning of the chase and retained in the culture media throughout the time-course.

Gel and blot quantitations

For Western blots, Image J software was used to quantitate protein levels. The background on the blots was first subtracted from the pixel counts for each band and final values were divided by the values obtained for loading controls. The control samples were set at 1.0 and all other samples were calibrated accordingly into Arbitrary Units (AU). For half-life determination, the intensity of each individual band was quantitated using the phosphoimager Quant programme (GE HealthCare). The half-life was determined to be the point at which 50% of the protein remained relative to that at t = 0.

Quantitative RT-PCR

Total RNA was extracted from MEFs using the RNeasy Mini Kit (Qiagen) and 0.5 μg RNA was reverse-transcribed using Superscript III (Invitrogen) according to the manufacturer's instructions. Quantitative RT-PCR was performed as described [40] using SYBRGreen (BioRad) in a LightCycler® 480 Real-Time PCR (Roche). Primers used were: Gapdh For: 5′-AGG TCG GTG TGA ACG GAT TTG-3′ and Rev: 5′-TGT AGA CCA TGT AGT TGA GGT CA-3′; Braf For: 5′-GAA TGT GAC AGC ACC CAC AC-3′ and Rev: 5′-ATA AGC TGG AGC CCT CAC - 3′.

Statistical analysis

Comparison between any two groups was performed by the student's t test. P values of < 0.05 were considered statistically significant.

Results

Expression of V600EBRAF down-regulates BRAF protein levels

We have previously generated conditional knockin Braf+/ mice that allow expression of the oncogenic V600EBRAF protein from one allele of the endogenous Braf gene following Cre-mediated recombination [34]. MEFs were derived from these mice and AdCre delivered to the cells ex vivo. Following AdCre treatment, recombination of the LSL allele was induced, generating the Lox-V600E allele with virtual 100% recombination at 96 h post-Cre treatment (Fig. 1A). The cells were morphologically transformed following V600EBRAF expression (Fig. 1A).

Fig. 1

Downregulation of BRAF protein levels following V600EBRAF expression in MEFs. (A) Detection of LSL and Lox alleles by PCR. Primary MEFs with the Braf+/LSL-V600E genotype were infected with AdCre for up to 96 h, DNA was isolated and PCR assays undertaken to identify LSL and Lox alleles. The photographs show the morphology of Braf+/ MEFs with and without AdCre for 96 h. (B) Drop in BRAF protein levels following BRAFV600E expression. Soluble protein lysates were generated from Braf+/+ and Braf+/ MEFs at the time points indicated following AdCre treatment. Protein lysates were analysed by immunoblot for BRAF and phosphoERK with GAPDH as a loading control. Representative immunoblots for four independent experiments are shown. (C) BRAF protein quantitation. BRAF protein levels were quantitated using Image J analysis of Western blot signals. The control samples were set at 1.0 and all other samples were calibrated accordingly into Arbitrary Units (AU). Data represent mean ± SD of four experiments of four independent MEFs for Braf+/ MEFs and three experiments of three independent MEFs for Braf+/+ MEFs. (D) BRAF expression level is not affected by cell density. Soluble protein lysates were generated from Braf+/+ and Braf+/ that had been treated with AdCre for 96 h and plated at high (H) or low (L) density. Protein lysates were analysed by immunoblot for BRAF and GAPDH was used as a loading control. (E) Analysis of phosphoMEK and phosphoERK. Soluble protein lysates were generated from Braf+/ MEFs at the time points indicated following AdCre treatment. Protein lysates were analysed by immunoblot for BRAF, phosphoMEK, phosphoERK with GAPDH as a loading control.

The levels of expression of the BRAF protein were monitored by immunoblot analysis (Fig. 1B) and quantitation (Fig. 1C) in both Braf+/ and Braf+/+ MEFs following AdCre treatment. Due to the presence of polyadenylation sequences within the LSL cassette, the Braf allele is known to generate a hypomorphic Braf allele that expresses the BRAF protein at < 10% the level of the wild-type Braf allele (Fig. 1B). Thus, following removal of the LSL cassette, BRAF protein levels within the Braf+/ cells are increased at 24 h following AdCre delivery (Fig. 1B and C). However, at subsequent time points (72–96 h), there was a significant decrease in BRAF protein levels (Fig. 1B and C). Such alterations in the levels of BRAF protein were not observed in Braf+/+ cells following a time course of AdCre treatment, suggesting they occur as a consequence of V600EBRAF expression (Fig. 1B and C). This effect was not related to cell density as indicated by the observation that BRAF levels do not change in high or low-density cultures of Braf+/+ and Braf+/ cells (Fig. 1D). The fact that BRAF levels drops by more than 50% in the Braf+/ cells would suggest that both WTBRAF and V600EBRAF proteins are affected as a consequence of expression of the oncogene.

The initial expression of V600EBRAF was accompanied by induction of phosphorylated MEK/ERK at 24–48 h post-AdCre treatment (Fig. 1B and E). However, at later time points, when BRAF levels were observed to drop, there was a noticeable decrease in phosphorylated MEK/ERK (Fig. 1B and E). This result suggests that the drop in BRAF expression leads to downregulation in signalling through the MEK/ERK pathway.

To investigate whether similar responses occur in vivo, we assessed BRAF protein levels in lung tissue derived from Braf+/;CreER™ mice that develop lung adenomas at 3–10 weeks post-partum [36] (Fig. 2A). As with the MEFs, BRAF protein levels were found to progressively decrease in the V600EBRAF-expressing lung and this was accompanied by a decrease in phosphorylated MEK and ERK levels at the 10-week time points (Fig. 2B).

Fig. 2

Downregulation of BRAF protein levels following V600EBRAF expression in the lung. (A) Histological staining of lung tissue with H&E. Lung sections from Braf+/; CreER™ mice at 10 weeks of age and Braf+/; CreER™ mice at 3, 6 and 10 weeks of age were generated and representative H&E stained images are shown. Scale bars, 100 μm. These mice have been previously reported in [36]. (B) Drop in BRAF protein levels following BRAFV600E expression. Soluble protein lysates were generated from the lungs of Braf+/+; CreER™ (WT) and Braf+/; CreER™ (VE) mice at the ages shown using GLB lysis. Representative immunoblot analysis of protein lysates for BRAF, phosphoMEK and phosphoERK are shown. ERK2 was used as a loading control.

The decrease in BRAF protein level is associated with alterations in protein stability

We quantitated Braf mRNA across the time course using qRT-PCR. Braf mRNA levels were decreased in both Braf+/+ and Braf+/ cells at 24–72 h post-AdCre treatment, suggesting this is a response to the presence of Cre (Fig. 3A). However, at the 96 h time point, when BRAF protein levels are most markedly decreased in the Braf+/ cells (Fig. 1B, C and E), there was no significant difference between the expression of Braf mRNA in the Braf+/+ and Braf+/ MEFs (Fig. 3A). This would suggest that the difference in protein expression is not underpinned by transcriptional alterations in the Braf gene or mRNA stability.

Fig. 3

Differences in BRAF protein levels are attributable to alterations in protein stability. (A) Assessment of Braf mRNA levels. Quantitative RT-PCR (qRT-PCR) analysis was used to investigate Braf mRNA levels in Braf+/ and Braf+/+ MEFs following a time course of AdCre treatment. The graph shows the fold increase or decrease in Braf mRNA levels relative to the levels in the Braf+/+ samples without AdCre (t = 0). Data represent mean ± SD of three independent experiments of three independent MEFs of each genotype. (B) Rescue of BRAF protein levels by proteasome inhibitors. Representative BRAF immunoblot analysis of protein lysates derived from Braf+/ and Braf+/+ MEFs following treatment with proteasomal inhibitors are shown. Braf+/+ and Braf+/ MEFs were infected with AdCre for 96 h and treated either with DMSO, 30 μM MG132 in DMSO or 0.5 μM Epoxomicin in DMSO for 5 h before harvesting soluble (SF) and insoluble (IF) fractions. GAPDH and β-ACTIN were used as loading controls for the SF and IF respectively. Data are representative of three experiments. (C) Half-life determination of V600EBRAF and WTBRAF in HEK293T cells. Data were obtained in triplicate for three independent experiments and the graph shows mean ± SD at each time point. The average half-life of each sample is indicated.

To investigate protein stability, the Braf+/+ and Braf+/ MEFs at 96 h post-AdCre treatment were exposed to proteasomal inhibitors MG132 and epoxomicin. For these experiments, soluble and insoluble protein lysates were generated and analysed since proteasomal inhibition is known to lead to the accumulation of ubiquitylated proteins predominantly in detergent-insoluble fractions. Treatment with both inhibitors led to an accumulation of BRAF protein levels in soluble and insoluble fractions in both cell types (Fig. 3B). Indeed this treatment raised BRAF protein levels in the AdCre-treated Braf+/ cells to those similar to levels in AdCre-treated Braf+/+ cells without inhibitor (Fig. 3B).

We also investigated the half-lives of WTBRAF and V600EBRAF by transfecting vectors expressing MYC-tagged versions of either of the proteins into HEK293T cells and using pulse–chase analysis. As shown in Fig. 3C, ectopic V600EBRAF demonstrated a significantly shorter half-life than ectopic WTBRAF by ~ 3 h.

The decrease in BRAF protein level is associated with MEK activity

The fact that BRAF protein levels drop in the AdCre-treated Braf+/ MEFs (Fig. 1C) suggests that both WTBRAF and V600EBRAF proteins are affected by the presence of the oncogene and therefore that a downstream event is involved in this regulation. To examine a role of the MEK/ERK pathway, we treated Braf+/ MEFs with the MEK inhibitor PD184352 and found that this increased BRAF protein levels in both insoluble and soluble protein fractions compared to controls (Fig. 4A). Furthermore, treatment with PD184352 rescued the half-life of V600EBRAF to levels similar to WTBRAF in HEK293T cells (Fig. 4B). Thus, the increased instability of the BRAF protein following V600EBRAF expression can be explained by ERK feedback regulation.

Fig. 4

BRAF protein levels are increased by MEK inhibition. (A) MEK inhibition in MEFs. Representative BRAF immunoblot analysis of protein lysates derived from SF and IF of Braf+/ MEFs following treatment with MEK inhibitor 1 μM PD184352 or carrier control are shown. ERK2 and β-ACTIN were used as loading controls for the SF and IF respectively. Data are representative of two experiments. (B) Half-life determination of V600EBRAF in HEK293T cells following treatment with 1 μM PD184352 (+ PD) or carrier control (− PD). Data were obtained in triplicate for two independent experiments and the graph shows mean ± SD at each time points. The average half-life of each treatment is indicated. (C) No alteration in BRAF levels in human cancer cell lines following MEK inhibition. Colo205, A375 and HT29 were treated with 1 μM PD184352 (+ PD) or carrier control (− PD) for 6–48 h. Protein lysates were immunoblotted for BRAF and phosphoERK. GAPDH was used as a loading control. Of note, the human cancer cell lines used for this analysis do not carry mutations in FBXW7 or known F-box containing genes (http://cancer.sanger.ac.uk/cosmic), as documented to date.

Association of feedback pathway with Oncogene-induced Senescence (OIS)

V600EBRAF expression in primary mouse and human cells is known to be associated with OIS [41], [42], [43], [44], [45]. To investigate whether the feedback pathway identified here is associated with OIS, we analysed the V600EBRAF-expressing MEFs and HEK293T cells by immunoblot with a range of senescence markers (Fig. S1). Although the HEK293T cells only demonstrated weak OIS marker expression, the MEFs demonstrated induction of p21CIP1 and p19ARF expression at the late time points. Similarly, we have previously demonstrated expression of senescence markers (p21CIP1 and γH2AX) in the lung tissue analysed in Fig. 2 at the 10 week time point [36] when BRAF levels are observed to drop (Fig. 2B).

To further confirm an association with OIS, we investigated whether the feedback pathway is operational in advanced, highly proliferative human cancer samples with long-term BRAF mutation acquisition. To this end, three human cancer cell lines Colo205, HT29 (both from colorectal cancers) and A375 (from a melanoma) bearing the BRAF mutation were treated with PD184352 for 6–48 h and BRAF protein levels were assessed. Although this treatment resulted in a noticeably faster migrating form of BRAF, reflecting inhibition of ERK phosphorylation of BRAF [15], there were no alterations in BRAF expression levels in any of the cell lines (Fig. 4C).

BRAF stability is regulated by ERK phosphorylation sites within the CPD

Although a number of pathways controlling BRAF protein stability have been identified and reported in the literature, given the role of the ERK pathway (Fig. 4), we were particularly interested in the reported control of LIN-45 stability by MPK1 and SEL-10 (FBXW7, FBW7 or CDC4 in mammals) in C. elegans [33]. FBXW7 is a subunit of the SCF complex [46], [47]. Substrates for SCF have a high affinity-binding site for FBXW7 within a CPD [48], [49], [50] and LIN-45 contains an ERK docking site D domain adjacent to the CPD with MPK1 capable of phosphorylating two residues (T432 and S436) within this CPD, allowing binding of FBXW7. A conserved CPD is located at residues 400–405 in human BRAF with ERK phosphorylation sites identified at residues T401 and S405 (Fig. 5A).

Fig. 5

BRAF turnover is regulated by ERK phosphorylation in the CPD. (A) Schematic of human BRAF to indicate potential conserved CPD at residues 400–405 and putative ERK phosphorylation sites at T401 and S405. T401 has previously been validated as a bona-fide ERK phosphorylation site using experimentation [15]. (B) Half-life determination of V600EBRAF, WTBRAF and V600EBRAF with CPD mutations in HEK293T cells. Data were obtained in triplicate for three independent experiments and the graph shows mean ± SD at each time point. TS;V600EBRAF represents V600EBRAF with non-mutated CPD, AA;V600EBRAF represents V600EBRAF with the T401A;S405A double mutation in the CPD, AS;V600EBRAF represents V600EBRAF with the T401A single mutation in the CPD and TA;V600EBRAF represents V600EBRAF with the S405A single mutation in the CPD. The average half-life of each mutant is indicated. The data for non-CPD mutated WTBRAF and V600EBRAF are the same as that shown in Fig. 3C. (C) Immunoblot analysis of phosphoERK and MYC in protein lysates derived from HEK293T cells transfected with empty vector (EV), the MYC-tagged WTBRAF expression vector and the MYC-tagged V600EBRAF expression vector with or without CPD mutations for 48 h. GAPDH was used as a loading control. (D) Photographs of NIH3T3 cells that were transfected with empty vector (EV) or the MYC-tagged V600EBRAF vector with or without CPD mutations for 48 h.

To investigate the role of this putative CPD in mammalian BRAF turnover, we mutated T401 and S405 to non-phosphorylatable alanine residues within Myc-tagged human V600EBRAF as single and double mutations and tested their effect on protein stability (Fig. 5B). Both of the single mutations (T401A or S405A) led to a substantial increase in the half-life of V600EBRAF compared to the non-mutated versions and the double mutation did not increase the half-life any further (Fig. 5B). This suggests that phosphorylation of both T401 and S405 is important in regulating the turnover of V600EBRAF. The half-lives of all CPD phosphorylation mutants in V600EBRAF were also increased by 4–5 h above that for WTBRAF, suggesting that disruption of the CPD may interfere with other pathways normally involved in the regulation of BRAF turnover.

To confirm that mutation of these residues within the CPD itself does not interfere with downstream signalling of V600EBRAF, we were able to show that the mutants induced ERK phosphorylation to the same levels as the non-mutated version of V600EBRAF (Fig. 5C) and they were also able to induce morphological transformation of NIH3T3 cells (Fig. 5D). This result is consistent with previous data showing that the T401A mutation does not have a strong effect on BRAF transforming activity [15].

The role of FBXW7 in BRAF turnover

To investigate a role of FBXW7 in mammalian BRAF protein turnover, we first examined whether the two proteins are able to interact. HEK293T cells were co-transfected with vectors expressing GFP alone, GFP-tagged WTBRAF or GFP-tagged V600EBRAF with FLAG-tagged FBXW7. Soluble protein lysates were generated, immunoprecipitated using GFP trap beads and analysed by immunoblot for FLAG (Fig. 6A). The data show co-immunoprecipitation of FBXW7 with WTBRAF and V600EBRAF but not with GFP alone (Fig. 6A).

Fig. 6

FBXW7 interacts with BRAF in HEK293T but does not promote BRAF turnover. (A) Co-immunoprecipitation of BRAF with wild-type and mutant FBXW7. HEK293T cells were co-transfected with vectors expressing GFP-tagged WTBRAF (WT), V600EBRAF (VE) or GFP alone (EV) together with either a vector expressing FLAG-tagged WTFBXW7 (top panels) or R482QFBXW7 (bottom panels). After 48 h, soluble protein cell lysates (WCL) were generated and these were subjected to immunoblot for GFP or FLAG. GFP-expressing proteins were immunoprecipitated using GFP-trap beads and these were subjected to immunoblot for FLAG (right panels) to examine co-immunoprecipitation of FBXW7 with BRAF. (B) Half-life determination of V600EBRAF and WTBRAF in the absence or presence of over-expressed FBXW7 in HEK293T cells. HEK293T cells were transfected with vectors expressing MYC-tagged WTBRAF or V600EBRAF in the presence or absence of a vector expressing FBXW7. Data were obtained in triplicate for three independent experiments and the graph shows mean ± SD at each time point. The average half-life of each condition is indicated. The data for WTBRAF and V600EBRAF are the same as that shown in Fig. 3C.

FBXW7 is involved in the binding of a number of substrates through a C terminal interacting domain made up of WD40 repeats, and arginine residues within these repeats are important for substrate recognition [48], [51]. Mutation of arginine residues at 465 and 479 within the WD40 repeats are hotspots for mutations in human cancers [50], [52] and mutation of the equivalent R479 mutation in mice (R482) generates increased expression of FBXW7 target substrates KLF5 and Tgifl in the mouse lung [37]. To investigate if R482 of mouse FBXW7 is involved in the binding of BRAF we repeated the above coimmunoprecipitation experiment with R482QFBXW7. This mutated form of FBXW7 was able to interact with BRAF in a similar manner to WTFBXW7, suggesting the interaction with BRAF is not mediated by the substrate recognition domain of FBXW7 (Fig. 6A).

We then performed over expression and loss of function studies to address the role of FBXW7 in the turnover of both WTBRAF and V600EBRAF. Using the HEK293T system the half-lives of WTBRAF and V600EBRAF in the presence and absence of overexpressed FBXW7 were compared. As shown in Fig. 6B, the half-lives of both WTBRAF and V600EBRAF were not shortened by the co-overexpression of FBXW7, counteracting the view that FBXW7 is involved in promoting BRAF turnover. In fact, the half-lives of both proteins was observed to increase in the presence of over-expressed FBXW7 (although not to statistically significant levels), potentially suggesting that over-expression of FBXW7 interferes with the binding of other proteins involved in regulating BRAF turnover.

To further investigate a role of FBXW7 we first attempted to undertake Fbxw7 siRNA knockdown experiments in MEFs. However, these proved not to be successful due to the lack of specific antibodies for endogenous mouse FBXW7 (data not shown). As an alternative, we derived MEFs from mice homozygous for a floxed allele for the conditional knockin Fbxw7 mutation [37] and examined BRAF protein expression following Cre-mediated recombination of the floxed allele. As shown in Fig. 7A, the expression level of WTBRAF was not altered following R482QFBXW7 expression.

Fig. 7

Mutation of FBXW7 does not affect WTBRAF or V600EBRAF expression in MEFs. (A) Expression of WTBRAF in MEFs with the homozygous Fbxw7 mutation. MEFs homozygous for the Fbxw7 allele were treated with or without AdCre for 96 h. Protein lysates were generated and immunoblotted for BRAF. GAPDH was used as a loading control. DNA was also generated and subjected to PCR for the Cre-deleted Fbxw7 floxed allele. Quantitation of BRAF protein levels is shown in the bar graph on the right. Data represent mean ± SD of three independent experiments of three independent MEFs. (B) Expression of V600EBRAF in MEFs with or without the homozygous Fbxw7 mutation. MEFs were generated from mice that contained the LSL-V600EBraf genetic modification with (Mut) or without (WT) the Fbxw7 allele. MEFs were treated with AdCre for 96 h. Protein lysates were generated and immunoblotted for BRAF. GAPDH was used as a loading control. DNA was also generated and subjected to PCR for the Cre-recombined LSL-Braf allele and the Cre-deleted Fbxw7 floxed allele. Quantitation of BRAF protein levels is shown in the bar graph on the right. Data represent mean ± SD of three independent experiments of three independent MEFs. (C) Morphology of Braf+/ MEFs with or without the homozygous Fbxw7 mutation treated with AdCre for 96 h.

A similar experiment was performed with V600EBRAF in MEFs. MEFs were derived from mouse embryos containing the Braf allele with or without the conditional homozygous knockin Fbxw7 mutation. MEFs were treated with AdCre for 96 h, and the expression of BRAF was compared. On the Fbxw7 mutant background, the expression level of V600EBRAF was not significantly altered compared to controls (Fig. 7B). Additionally, there were no alterations in the morphological transformation of the MEFs (Fig. 7C). These data are consistent with the view that FBXW7 does not play a unique role in regulating the turnover of V600EBRAF.

Discussion

Exquisite control of the RAF–MEK–ERK signalling pathway is of critical importance in the maintenance of tissue and body homeostasis and deregulation of the pathway is associated with several human pathologies including cancer, RASopathies, some neurological disorders and diabetes [53]. An understanding of the mechanisms that regulate the pathway in normal and diseased cells is imperative for the design of optimal treatments as well as for understanding drug resistance. Here we have confirmed the existence of an ERK-mediated feedback pathway controlling BRAF protein turnover in mammalian cells that was previously identified in C. elegans [33]. However, despite conservation of the feedback loop via ERK, the subsequent BRAF degradation is different in C. elegans and mammalian cells in that mammalian cells do not uniquely rely on the FBXW7/SEL-10 component of the SCF E3 ubiquitin ligase complex.

A variety of different methods for fine-tuning the RAF–MEK–ERK pathway have been identified in recent years, with a central point of control existing at the level of RAF, which can be controlled by both feedback and feed-forward mechanisms [9]. Regulation of RAF activity involves cycles of phosphorylation and de-phosphorylation and has become increasingly more complex with the discovery that RAF isoforms can form RAF homo/hetero dimers with different levels of kinase activity [12]. RAF is also part of a multiprotein complex, components of which can influence protein folding, stability and consequently activity [28], [29], [30]. With regard to the control of RAF protein stability, a requirement for CRAF autophosphorylation of residue S621 has been reported [40]. However, this mode of control was found not to be conserved for BRAF [40] but, instead, we have confirmed an involvement of MEK/ERK activity in the regulation of BRAF protein stability.

This regulation of BRAF protein stability by ERK extends the repertoire of feedback mechanisms by which ERK controls output of the RAF/MEK/ERK pathway. It has been previously documented that ERK1/2 can down-regulate MEK1 activity by phosphorylation of T292 and T386 [54], [55], although the control of cellular MEK1 levels by ERK-dependent transcriptional methods has also been identified recently [56]. ERK control of upstream regulators has also been documented including the phosphorylation of SOS to regulate its interaction with GRB2 [57] and disruption of the CRAF–RAS interaction by phosphorylation of CRAF on multiple sites [16]. With regard to BRAF, several target ERK phosphorylation sites are known to exist at S151, T401, S750 and T753 and phosphorylation of all of the sites has been shown to contribute to disruption of BRAF/CRAF heterodimerisation [12], [15]. We now show that ERK phosphorylation of T401 and S405 has an additional function in controlling BRAF protein stability.

We were first drawn to investigating BRAF protein turnover by the observation that BRAF protein levels are significantly decreased in mouse cells and tissue expressing autochthonous BRAF following short term V600EBRAF expression (Fig. 1, Fig. 2). This finding was supported by observations in the HEK293T over-expression systems, demonstrating a shorter half-life of V600EBRAF than WTBRAF (Fig. 3, Fig. 4, Fig. 5). The physiological role of this feedback mechanism is presently not clear, although our data (Fig. S1) suggest a potential link with the OIS phenotype. Short-term expression of V600EBRAF is known to be associated with the induction of senescence in the mouse lung model [36], [42] as well as V600EBRAF intestinal [43] and melanoma mouse models [44] and the RAF/MEK/ERK pathway is known to mediate growth inhibitory signalling as well as cell proliferation [58]. Low levels of phospho-ERK are associated with this phenotype in the mouse lung models (Fig. 2 and [59]) and, furthermore, human melanocytic naevi bearing the V600EBRAF oncogene demonstrate hallmarks of senescence [45] but < 25% show detectable phospho-ERK immunohistochemical staining [60]. On this basis it has been suggested that feedback loops may be a requirement for maintenance of V600EBRAF-induced senescence [60] as has been documented for oncogenic RAS-induced senescence [61]. Further investigation of this will require manipulation of the various feedback loops associated with the RAF/MEK/ERK pathway including autochthonous mutation of ERK phosphorylation sites within the CPD and correlation with physiological responses. The fact that the ERK-mediated control of BRAF protein stability is not detected in human cancer cell lines (Fig. 4C), is consistent with a hypothesis that advanced cancers have evolved mechanisms to overcome this feedback regulation in order to bypass OIS.

FBXW7 is a member of the F-box protein family, which contain seven tandem WD40 repeats besides the F box, and is a substrate receptor component of the SCF ubiquitin ligase complex. Substrate phosphorylation within a CPD motif instigates FBXW7 binding leading to degradation of the substrate by the SCF complex [50]. Indeed, FBXW7 mutations are associated with many human cancers and FBXW7 has been implicated as a tumour suppressor by targeting the degradation of oncoprotein substrates including Cyclin E, Notch 1 and C-Myc. Although we found that FBXW7 is capable of binding BRAF, neither the binding nor stability of BRAF was affected by the R482QFBXW7 mutation within its WD40 substrate recognition domain (Fig. 6, Fig. 7), the equivalent mutation of which (R479Q) is detected in many human cancers. Taken together, these results suggest that other E3 ubiquitin ligase complexes are involved in regulating BRAF turnover, the identity of which has yet to be revealed. Our data currently do not rule out a role of other components of the SCF complex or F-box proteins.

In summary, our data show that BRAF protein stability is controlled by a negative feedback loop involving ERK phosphorylation of T401 and S405 within a conserved CPD. Although this feedback pathway is activated upon short-term expression of V600EBRAF, its functional role in mediating proliferative/senescence responses are not clear and will require further investigation.

The following are the supplementary data related to this article.

Fig. S1

Expression of senescent markers following V600EBRAF expression in HEK293T and MEFs. A. HEK293T cells were transfected with either empty vector pEF EV) or pEF vectors expressing WTBRAF (WT) or V600EBRAF (VE). 48 hours after transfection, soluble protein lysates were generated and analysed with the antibodies indicated by immunoblotting. B. Braf+/LSL‐V600E MEFs were treated with AdCre and protein lysates harvested at 0–96 h following AdCre treatment. Lysates were analysed with the antibodies indicated by immunoblotting. For both panels, protein lysates from a control cell line (C) expressing high levels of senescent markers was utilised as a positive control. For the HEK293T cells, senescent marker induction was not detectable following V600EBRAF expression although a small increase in γH2AX levels was observed. In the MEFs, expression of p21CIP1 and p19ARF was observed at the later time points, providing some evidence for senescence induction by V600EBRAF in these cells. Protocols used for this analysis were the same as those described in the Materials and Methods section of the main text. Antibodies used were: γH2AX (Cell Signalling 9718S), p21CIP1 (Santa Cruz SC471), p16INK4a (AbCam AB-54210) and p19ARF (AbCam AB80).

Disclosure statement

The authors declare that there are no conflicts of interests.