The metastasis-promoting P1597L mutation in PlexinB1 enhances Ras activity

Abstract

Background

Metastatic prostate cancer is a leading cause of cancer-related morbidity and mortality in men, yet the underlying molecular mechanisms are poorly understood. Plexins are transmembrane receptors for semaphorins with divergent roles in many forms of cancer. We recently found that a single clinically relevant specific amino acid change (Proline1597Leucine, (P1597L)), found in metastatic deposits of prostate cancer patients, converts PlexinB1 from a metastasis suppressor to a gene that drives prostate cancer metastasis in vivo. However, the mechanism by which PlexinB1(P1597L) promotes metastasis is not known.

Methods

Pull down assays using GST-RalGDS or -GSTRaf1-RBD were used to reveal the effect of mutant or wild-type PlexinB1 expression on Rap and Ras activity respectively. Protein–protein interactions were assessed in GST pulldown assays, Akt/ERK phosphorylation by immunoblotting and protein stability by treatment with cycloheximide. Rho/ROCK activity was monitored by measuring MLC2 phosphorylation and actin stress fiber formation. PlexinB1 function was measured using cell-collapse assays.

Results

We show here that the single clinically relevant P1597L amino acid change converts PlexinB1 from a repressor of Ras to a Ras activator. The PlexinB1(P1597L) mutation inhibits the RapGAP activity of PlexinB1, promoting a significant increase in Ras activity. The P1597L mutation also blocks PlexinB1-mediated reduction in Rho/ROCK activity, restraining the decrease in MLC2 phosphorylation and actin stress fiber formation induced by overexpression of wild-type PlexinB1. PlexinB1(P1597L) has little effect on the interaction of PlexinB1 with small GTPases or receptor tyrosine kinases and does not inhibit PlexinB1-stimulated Akt or ERK phosphorylation. These results indicate that the mutation affects Rho signalling via the Rap/Ras pathway. The PlexinB1(P1597L) mutation inhibits morphological cell collapse induced by wild-type PlexinB1 expression, suggesting that the mutation induces a loss of an inhibitory tumour suppressor function.

Conclusion

These results suggest that the clinically relevant P1597L mutation in PlexinB1 may transform PlexinB1 from a suppressor to a driver of metastasis in mouse models of prostate cancer by reducing the RapGAP activity of PlexinB1, leading to Ras activation. These findings highlight the PlexinB1-Rap-Ras pathway for therapeutic intervention in prostate cancer.

Background

Plexins, cell surface receptors for semaphorins, have been implicated in the progression of many cancers [1], including breast [2, 3], ovarian [3, 4], glioma [5], and melanoma [6] as well as prostate cancer [7,8,9]. Nine genes for plexins exist in vertebrates (PlexinA(1–4), PlexinB(1–3), PlexinC(1–2)) [10]. Loss of PlexinB1 expression is a favourable prognostic factor for ErbB2 amplified breast cancer [2] and ovarian cancer [4] but is associated with poor prognosis in melanoma [6] and ER-positive breast cancer [11] suggesting that PlexinB1 can act as an oncogene or a tumour suppressor depending on context. PlexinB1, and its’ ligands Semaphorin4D (Sema4D) and Sema3C are overexpressed [8, 12,13,14,15] in prostate cancer and expression of Sema3C is an independent predictor of biochemical recurrence of prostate cancer [8].

PlexinB1 is also mutated in clinical prostate cancer [7]. The somatic missense mutation, C5060T, changing proline1597 to leucine (P1597L) [7] investigated in this study, was found in two prostate cancer lymph node metastases and three bone metastases, while a different mutation in the same codon, (C5059T) changing proline1597 to serine (P1597S), was found in three lymph node metastases [7].

Functional analysis in vitro shows that the PlexinB1(P1597L) mutation increases cell motility, invasion and anchorage-independent growth of prostate cancer cells [16]. In contrast overexpression of wild-type Plexin-B1 had the opposite effect [16]. Furthermore, prostate epithelial cell-specific expression of PlexinB1(P1597L) in knock-in mice significantly increases invasion and metastasis in two transgenic mouse models of prostate cancer—PbCre + Pten ^fl/fl Kras ^G12V and PbCre + Pten ^fl/fl p53 ^fl/fl. In direct contrast, prostate epithelial cell-specific expression of wild-type PlexinB1 significantly decreases invasion and metastasis in PbCre + Pten ^fl/fl Kras ^G12Vmice [17]. Studies on specifically how this single amino acid change in PlexinB1 has such a dramatic effect on prostate cancer metastasis will increase our understanding of the mechanisms involved in this process.

Activation of B-type plexins results in activation of the receptor tyrosine kinases ErbB2 [18], EGFR [8] and Met [19] and modulates migration and invasion of prostate cancer cells [20, 21] through regulation of several small GTPases including Rho [22] through PZDRhoGEF/LARG [18] and p190RhoGAP [23] activation. B-type plexins also act as GTPase activating proteins (GAPs) for Rap [24] catalysing the conversion of RapGTP to GDP. Active Rap (RapGTP) inactivates p120RasGAP leading to Ras activation [25]. In ovarian cells, Rnd1-mediated activation of PlexinB1 promotes inactivation of RapGTP and Ras, while loss of Rnd1 or PlexinB1 leads to de-repression of Ras and metastasis in mouse models of ovarian cancer [26] showing that Rnd1 and PlexinB1 act as tumour suppressor genes in this context. The P1597L mutation is found in the GAP domain of PlexinB1.

The mechanism by which Plexin-B1(P1597L) promotes metastasis is not known. We investigated how the P1597L mutation affects PlexinB1 signalling and function. A better understanding of the role of mutant Plexin-B1 in metastatic prostate cancer, and the downstream pathways it activates, offers the potential for novel therapeutic approaches to impede the spread of prostate cancer.

Materials and methods

Cell culture and transfection

All cell lines were from ATCC (LGC standards, Middlesex, UK) and were STR typed to confirm their identity. Cos7, HCA-2 [27] or HeLa cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% foetal calf serum (FCS) and penicillin–streptomycin (Invitrogen). Cells were transiently transfected with the indicated plasmids using Lipofectamine 2000 (Invitrogen) as per manufacturer’s instruction. After 16 h cells were either lysed or fixed in paraformaldehyde to undergo immunoprecipitation or immunofluorescence experiments respectively.

Antibodies

The following antibodies were used: Mouse anti-GAPDH (MAB374) antibody, mouse anti-Myc antibody clone 9E10 (13–2500), rabbit polyclonal anti-Myc antibody (Sigma, C3956, 1:4000), rabbit anti-GFP antibody (A-11122), rabbit polyclonal anti-FLAG (M2) (F-7425), mouse anti pSer19 myosin light chain (Sigma-Aldrich); anti-GST (Merck, GE27-4577), rat monoclonal anti-HA antibody clone 3F10 (11,867 423,001) (Roche Applied Science), Rabbit antibody anti-mCherry (Abcam ab183628). Akt Phospho-Akt (Ser473) (D9E) XP® Rabbit mAb (upper) or Akt (pan) (C67E7) Rabbit mAb #4691 ERK Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (D13.14.4E) XP® Rabbit mAb #4370 and p44/42 MAPK (Erk1/2) (3A7) Mouse mAb #9107 (cell signalling), Ras (Upstate, 05–516, 1:3000), Rap1A/1B (cell signalling, 2399), b-catenin (Abcam, ab22656) b-catenin (Abcam, ab22656), mouse monoclonal antibody β-actin (Genetex GTX629630), mouse anti-ubiquitin antibody (Invitrogen 13–1600), rabbit anti-YAP1 antibody (Novus Biologics NB110-58358), rabbit anti- LC3B antibody (Cell Signalling #2775).

Secondary antibodies: horse-radish peroxidase linked (HRP) polyclonal goat anti-mouse IgG (P0447), goat anti-rabbit IgG (P0448), rabbit anti-goat IgG (P0449) (Agilent) or goat anti-rat IgG (sc-2006) (Santa Cruz Biotechnology). Goat anti-rabbit Alexa-Fluor 568 (A11036) or goat-anti-rabbit Alexa-Fluor 546 (A11010), goat anti-mouse Alexa-Fluor 546 (A11003), goat anti-mouse Alexa-Fluor 488 (A11029) and Alexa-Fluor 633 Phalloidin (A22284) were all purchased from ThermoFisher Scientific.

Expression vectors and site-directed mutagenesis

pcDNA3-PlexinB1 full-length-myc (PlexinB1-myc), pcDNA3-PlexinB1 cytoplasmic domain-Myc (cytoPlexinB1-myc), pcDNA3-PlexinB1(R1677A, R1678A, R1984A) [28] full-length-myc (PlexinB1(RA)-myc) and cytoPlexinB1(RA)-myc were kind gifts of Dr I Oinuma (Kyoto, Japan). pGex-GST-cytoPlexinB1 was generated by subcloning from the corresponding full-length Plexin pcDNA3 plasmids using NEBuilder® HiFi DNA Assembly Master Mix (E2621S NEB), using the primers: pGex-GSTcytoPlexinB1: ATCGGATCTGGTTCCGCGTGGCAGGAGGAAGAGCAAGCAG and GTCAGTCACGA TGCGGCCGCTCCTATAGATCTGTGACCTTGTTTTC, [29] pEGF-GFP-Rnd2, Rnd3, ROCK1-727, V38R-Ras, Rac1L61, RhoD, Rap1A, RhoH, MLC2, PDZRhoGEF or Grb2 ROCK(1–420) were described previously [7, 16, 30,31,32,33,34]. GFP-MLC2 was a kind gift from Professor Eric Sahai and Professor Vicky Sanz-Moreno, Grb2 was a kind gift from Dr James Monypenny. The C5060T (P1597L) mutation was introduced into the expression vectors encoding myc-PlexinB1 (cytoplasmic domain and full length) and HA- PlexinB1 (cytoplasmic domain and full length) and pGex-GST-cytoPlexinB1, using the Quikchange system (Stratagene). The nucleotide change was verified by DNA sequencing (Eurofins-MWG, UK).

GST pull-downs

Rosetta™ Competent Cells (Novagen) were transformed with pGex-GST-cytoPlexinB1(wild-type) or pGex-GST-cytoPlexinB1(P1597L) and glutathione-S-transferase (GST) fusion protein expression induced with IPTG, 1 mM for 3 h. Pelleted cells were lysed with lysis buffer (B-PER Bacterial Protein Extraction Reagent, ThermoFisher), lysozyme (100 μg/ml), PMSF (1 mM, Sigma), protein inhibitor cocktail (ThermoFisher), DNAse (5U/ml), at room temperature for 15 min, then on ice for 15 min, then spun for 30 min. The cleared lysate was incubated with Glutathione Sepharose 4B (17,075,601, GE Healthcare) for 2 h at 4 °C then washed. Recombinant GST was used as a control. Cos7 were co-transfected with MCL2-GFP and cytoPlxnB1(WT)-HA or cytoPlxnB1(P1597L)-HA, or with ROCK1(1–420) and cytoPlxnB1(WT)-HA or cytoPlxnB1(P1597L)-HA. Transfected cells were washed and lysed in lysis buffer (1% Triton X-100, 20 mM Tris–HCl [pH 8], 130 mM NaCl, 10 mM NaF, 1% aprotonin, 10 μg/ml leupeptin, 1 mM dithiothreitol ( DTT), 0.1 mM Na₃VO₄ and 1 mM phenylmethylsulfonyl fluoride (PMSF)). Insoluble material was removed by centrifugation, and the cell lysates were combined for 2 h at 4 °C with anti-HA-antibody bound to agarose. Beads were washed extensively with lysis buffer before the proteins were eluted in the Laemmli sample buffer. Bound proteins were analysed by immunoblotting. GFP-Trap agarose beads (gta-20) were supplied by ChromoTek, mouse anti-FLAG agarose (A2220) from Sigma-Aldrich, mouse anti-HA antibody bound to agarose clone HA.7 (A2095) from Sigma-Aldrich.

Immunoblotting

For immunoblotting, proteins were resolved by SDS-PAGE and transferred to nitrocellulose membrane (Schleicher and Schuell). Membranes were blocked with TBS (20 mM Tris–HCl [pH 7.6], 137 mM NaCl) containing 5% non-fat dried milk and 0.05% Tween 20 or 3% BSA for phospho-specific antibodies and protein detected with specific antibodies.

Bound antibodies were visualised with horseradish peroxidase-conjugated goat anti-immunoglobulin G antibodies and enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech). In some cases, blots cut prior to hybridisation with antibodies.

pERK and pAkt analysis

Cos7 cells transfected with cytoPlxnB1(WT)-myc; or cytoPlxnB1(P1597L)-myc; cytoPlxnB1(RA)-myc or vector-myc empty were lysed with ice-cold cell lysis buffer (20 mM Tris–HCl, pH 8, 130 mM NaCl, 1% Triton X-100, 1 mM DTT, 10 mM NaF, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 µg/ml aprotinin, 10 µg/ml leupeptin, 0.1 mM sodium orthovanadate). Lysates were centrifuged for 10 min at 13,000 rpm. Phosphorylated proteins were analysed by SDS-PAGE and immunoblotting with phospho-specific and total antibodies.

Treatment with cycloheximide or proteosomal inhibitor MG132

Cos7 cells were transfected with cytoPlxnB1(WT)-myc; or cytoPlxnB1(P1597L)-myc; cytoPlxnB1(RA)-myc or vector-myc. After 18 h cells were treated with cycloheximide (CHX) at 300 µg/ml or the proteosomal inhibitor MG132 at 10 µM for various time points. Cells were lysed in ice-cold cell lysis buffer (20 mM Tris–HCl, pH 8, 130 mM NaCl, 1% Triton X-100, 1 mM DTT, 10 mM NaF, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 µg/ml aprotinin, 10 µg/ml leupeptin, 0.1 mM sodium orthovanadate). Lysates were centrifuged for 10 min at 13,000 rpm. Proteins were analysed by SDS-PAGE and immunoblotting.

Preparation of GST-RalGDS or Raf1-RBD-GST

Bacterial pellets of Rosetta™ Competent Cells (Novagen) transformed with GST-RalGDS-RBD or Raf1-RBD-GST were lysed in cold STE buffer (10 mM Tris pH 8; 150 mM NaCl; 1 mM EDTA) containing 1 mM PMSF. Lysed bacterial pellets were homogenised with a 19-gauge needle 3–4 times and 100 µg/ml of lysozyme was added, rocked and incubated on ice for 15 min. After 15 min, 5 mM DTT was added followed by 1% Tween 20 and 0.03% SDS. Bacterial lysates were centrifuged at 13,000 rpm for 30 min. During this time, approximately 300 µl of glutathione-sepharose beads were washed in lysis buffer and mixed with supernatant from spun bacterial lysates. Samples were rotated for 2 h at 4 °C. After 2 h, glutathione-sepharose beads were washed in cold STE buffer and resuspended in approximately 200 µl of buffer and used the next day.

Rap activation assay

HeLa cells were transiently transfected with cytoPlxnB1(WT)-myc, or cytoPlxnB1(P1597L)-myc, or cytoPlxnB1(RA)-myc, or vector-myc using Lipofectamine 2000. After 18 h cells were washed with cold PBS and then lysed with 500 µl of Mg²⁺ buffer lysis buffer (125 mM HEPES pH 7.5; 750 mM NaCl; 5% Nonidet 40; 50 mM MgCl₂; 5 mM EDTA and supplemented with 25 mM sodium fluoride; 1 mM sodium vanadate; 1 mM PMSF; 10% glycerol and protease cocktail inhibitor tablets). Lysates were centrifuged for 10 min at 13,000 rpm. The supernatants were then incubated with 30–50 µg GST-RalGDS-RBD for 1 h with rotation at 4 °C. Positive and negative controls were prepared by treating cell lysates by adding 10 mM EDTA followed by 100 mM GTPyS or 1 mM GDP respectively. Lysates were incubated for 30 min at 30 °C with agitation. The reaction was terminated using 60 mM MgCl₂ and then incubated with GST-RalGDS-RBD beads. After 1 h GST-RalGDS-RBD beads were centrifuged at 3000 rpm for 2 min. Supernatant was discarded and beads were washed in lysis buffer three times and 2 × sample buffer containing DTT was added ready for SDS-PAGE analysis. Western blots were incubated with Rap1A/Rap1B antibody to detect changes in Rap1 activity levels.

Ras activity

Levels of RasGTP were assessed by immunoprecipitation using Raf1-RBD-GST fusion proteins bound to glutathione-sepharose beads, which selectively isolate the active form of Ras (RasGTP) in the lysis Buffer: 25 mM HEPES, Ph 7.5, 150 mM NaCl, 1% NP-40, 10 mM MgCl2, 1 mM EDTA, 2% Glycerol, 10 mM NaF, protease inhibitor tablet (Roche), followed by western blot analysis with anti-Ras antibody. Controls for Ras activation assays were made by pre-loading cell lysates with a non-hydrolyzable form of GTP (GTPγS) or GDP, by adding 10 mM EDTA followed by 100 mM GTPyS or 1 mM GDP respectively to the cell lysates. Lysates were incubated for 30 min at 30 °C with agitation. The reaction was terminated using 60 mM MgCl₂ and the pre-loaded lysates were immunoprecipitated with Raf1-RBD agarose beads and immunoblotted with Ras antibody.

Stress fiber analysis

HCA-2 (human dermal fibroblasts) cells were seeded at a density of 5 × 10⁵ per ml on glass coverslips. Post 18 h, cells were transiently transfected with cDNA encoding cytoPlexinB1(WT)-myc, cytoPlexinB1(P1547L)-myc, full length-PlexinB1(WT)-myc, full length-PlexinB1(P1597L)-myc, RhoAV14-myc, or control empty vector. GFP-tagged V14RhoA was used as positive controls to detect actin stress fibers. After 18 h cells were fixed in 3.7% paraformaldehyde and permeabilised in 0.2% Triton for 5 min. Coverslips were treated to blocking buffer consisting of 5% heat inactivated serum in PBS for 30 min. Transfected cells were identified by anti- myc antibody followed by AlexaFluor 488 fluorescent secondary antibodies. Stress fibers were detected using TRITC-phalloidin and nuclei detected by DAPI staining. Images were taken using Eclipse Ti-E inverted confocal microscope (Nikon) with × 60 (1.4 NA) objective. The intensity of actin staining was scored blind using imageJ.

Phosphorylation of pMLC2

The human dermal fibroblast cell line, HCA-2, were seeded onto coverslips, 50,000 cells per coverslip and after 18 h cells were transfected with cytoPlexinB1(WT)-myc, cytoPlexinB1(P1547L)-myc, myc-tagged ROCK 1–420 (used as positive control for MLC2-phosphorylation), or empty vector. After 18 h cells were fixed in 3.7% paraformaldehyde and permeabilised in 0.2% Triton for 5 min. Coverslips were treated to blocking buffer consisting of 5% heat inactivated serum in PBS for 30 min. Cells on coverslips were incubated with rabbit anti- myc antibody (to detect PlexinB1 protein), and mouse anti pSer19 MLC2 antibody for 1 h followed by anti-rabbit AlexaFluor 488, anti-mouse AlexaFluor 567 secondary antibody for a further 1 h. Stress fibers were detected using Phalloidin 647 and nuclei detected by DAPI staining. Images were taken using Eclipse Ti-E inverted confocal microscope (Nikon) with × 60 (1.4 NA). The intensity of phospho-MLC2 staining was scored blind using imageJ.

Collapse assay

Cos7 cells were plated onto glass coverslips, and 24 h later, cells were transiently transfected with plasmids encoding cytoPlxnB1(WT)-myc; or cytoPlxnB1(P1597L)-myc; cytoPlxnB1(RA)-myc or vector-myc empty using Lipofectamine 2000 as per manufacturer’s instruction. After 18 h, cells were fixed for 15 min with 3.7% paraformaldehyde and permeabilised for 5 min with 0.2% Triton X-100. Coverslips were incubated for 1 h with rabbit anti-myc antibody to detect myc-tagged Plexin B1. Mouse anti-β-tubulin (clone DMIA) was used to detect microtubules. AlexFluor 488 antibody conjugated donkey anti-rabbit antibody identified PlexinB1 expressing cells, and AlexFluor 647 antibody conjugated donkey anti-mouse antibody revealed microtubular organisation. After extensive washing, coverslips were mounted on glass microscope slides, and images were generated using Eclipse Ti-E inverted confocal microscope (Nikon) DU-897X-3430 camera (Andor) with PlanFluor × 40/1.3 NA oil objective. ‘Collapsed’ cells were defined as cells with a contracted cell body and with one or more extensions of greater length than the cell body). 102 or more cells counted per condition per experiment (n = 3).

Statistics

Statistical analysis was performed using Microsoft Excel or GraphPad. Error bars on graphs denote the average values ± standard error of the mean (SEM), statistical significance was determined by two tailed paired students’ t-test unless otherwise stated.

Results

The PlexinB1(P1597L) mutation has little effect on protein interaction

GST pull down assays were performed, using GST-PlxnB1(WT), GST-PlxnB1(P1597L) or GST alone as control, to establish if the P1597L mutation in PlexinB1 influences the interaction of PlexinB1 with other proteins. The P1597L mutation had little effect on the interaction of PlexinB1 with: Rnd2, Rnd3, ROCK1-727, V38R-Ras (Fig. 1A), Rac1L61, RhoD, Rap1A, RhoH, MLC2, (Fig. 1B) or PDZRhoGEF or Grb2 (Supplementary Fig. 1A). The interaction of wild-type and mutant PlexinB1 with MLC2 and ROCK(1–420) was confirmed by co-immunoprecipitation studies (Fig. 1C). These results show that the P1597L mutation has negligible effect on the interaction of PlexinB1 with this panel of proteins.

Fig. 1

Effect of P1597L mutation in PlexinB1 on protein–protein interaction. A, B Lysates of Cos7 cells were incubated with GST, GST-PlxnB1(WT) or GST-PlxnB1(P1597L) immobilised to gluthathione-sepharose beads and bound proteins were detected by immunoblotting with indicated antibodies. Stars indicate bands of correct size C MLC2 and ROCK (1–420) co-immunoprecipitate with PlexinB1(WT) and PlexinB1(P1597L) Cos7 were co-transfected with MCL2-GFP and cytoPlxnB1(WT)-HA or cytoPlxnB1(P1597L)-HA; or with ROCK1(1–420)-myc and cytoPlxnB1(WT)-HA or cytoPlxnB1(P1597L)-HA. After 20 h, cell lysates were incubated with anti-HA-antibody bound to agarose. Immunoprecipitated proteins were analysed by immunoblotting with indicated antibodies

The PlexinB1(P1597L) mutation does not inhibit PlexinB1-induced Akt or ERK phosphorylation

Activation of B-type plexins results in PI3K activation and phosphorylation of Akt [8], and in the phosphorylation of ERK, via the receptor tyrosine kinases ErbB2, EGFR and MET [8]. To establish whether the P1597L mutation has an effect on these pathways, Cos7 cells were transfected with various constructs encoding the cytoplasmic domain of PlexinB1. Cells were transfected with: 1). The cytoplasmic domain of wild-type PlexinB1 (cytoPlxnB1(WT)-myc), 2).cytoplasmic domain of PlexinB1 with the P1597L mutation (cytoPlxnB1(P1597L)-myc), 3). A previously characterised mutant version of PlexinB1 which has mutations (R1677A, R1678A, R1984A) in the GAP domain and displays no or very weak GAP activity (cytoPlxnB1(RA)-myc) [28], or 4). empty vector-myc. The effect on Akt and ERK phosphorylation was monitored by immunoblotting, using phospho-specific antibodies. Overexpression of WT PlexinB1 increased phosphorylation of Akt (Fig. 2A) and ERK (Fig. 2B). Transfection of cytoPlxnB1(P1597L)-myc or cytoPlxnB1(RA)-myc had a similar effect on both Akt (Fig. 2A) and ERK phosphorylation (Fig. 2B) to that of PlexinB1(WT). These findings show that the P1597L mutation has little effect on PlexinB1-mediated activation of these signalling pathways.

Fig. 2

A Effect of P1597L mutation on PlexinB1-stimulated AKT phosphorylation. Cos7 cells were transfected with cytoPlxnB1(WT)-myc, cytoPlxnB1(P1597L)-myc, cytoPlxnB1(RA)-myc or vector-myc empty and phosphorylation of Akt identified by immunoblotting. Graphs show average of phosphorylated Akt/total Akt band intensity (n = 4). Error bars denote SEM. B Effect of P1597L mutation of PlexinB1-stimulated ERK phosphorylation. Cos7 cells were transfected with cytoPlxnB1(WT)-myc, cytoPlxnB1(P1597L)-myc, cytoPlxnB1(RA)-myc or vector-myc and phosphorylation of ERK identified by immunoblotting. Graphs show average of phosphorylated ERK/total ERK band intensity (n = 4). Error bars denote SEM. C The P1597L mutation in PlexinB1 has little effect on protein stability. Cos7 cells were transfected with cytoPlxnB1(WT)-myc, cytoPlxnB1(P1597L)-myc or empty vector (control). Cells were treated with cycloheximide (CHX) at 300 µg/ml. Lysates were analysed by SDS-PAGE and immunoblotting with indicated antibodies. Graph shows average band intensities for Ubiquitin (n = 3, error bars = SEM)

The PlxnB1(P1597L) mutation has little effect on protein stability

To determine if the P1597L mutation affects the stability of PlexinB1, Cos7 cells transfected with cytoPlxnB1(WT)-myc or cytoPlxnB1(P1597L)-myc were treated with cycloheximide, an inhibitor of de novo protein synthesis, for 6 h and PlexinB1 protein levels assessed by immunoblotting (Fig. 2C). The levels of PlexinB1 for both constructs were broadly the same showing that the P1597L mutation has little effect on protein stability. Consistent with these results ubiquitin levels were similar in cells transfected with either construct (Fig. 2C). Treatment of transfected cells with the proteosome inhibitor MG132 for various times followed by immunoblotting indicated that the mutation has little effect on transcription of the PlexinB1 gene (Supplementary Fig. 1B).

PlexinB1(P1597L) inhibits the RapGAP activity of PlexinB1

The C5060T (P1597L) mutation occurs in the GAP domain of PlexinB1, and plexins act as GAPs for Rap [24]. To examine whether expression of PlexinB1(P1597L) has any effect on Rap activity, the levels of activated Rap (RapGTP) were measured following transfection of Hela cells with cytoPlxnB1(WT)-myc, cytoPlxnB1(P1597L)-myc, cytoPlxnB1(RA)-myc or vector-myc. Overexpression of cytoPlxnB1(WT) reduced RapGAP activity relative to vector control, as expected (Fig. 3). The RapGTP levels of cells transfected with cytoPlxnB1(P1597L) or with the GAP-defective mutant (cytoPlxnB1(RA)) were similar, and both were above that of cytoPlxnB1(WT). These results are consistent with the hypothesis that cytoPlxnB1(P1597L) has reduced RapGAP activity.

Fig. 3

PlexinB1 regulates levels of GTP-bound Rap A Control for Rap activation assays; cell lysates were pre-loaded with a non-hydrolyzable form of GTP (GTPγS) or GDP, immunoprecipitated with RalGDS RBD-GST agarose beads and immunoblotted with Rap1A/B antibody. B Effect of P1597L mutation on Rap activity. Lysates of Hela cells, transfected with cytoPlxnB1(WT)-myc, cytoPlxnB1(P1597L)-myc, cytoPlxnB1(RA)-myc or vector-myc were immunoprecipitated with RalGDS-RBD agarose beads and the ratio of RapGTP to total Rap1A/B assessed by immunoblotting. C Bar chart of average band intensities, RapGTP/total Rap, (n = 3) (*p =  < 0.05, 2 tailed student T-Test)

PlexinB1(P1597L) increases Ras GTPase activity

We have shown that PlexinB1(P1597L) mutation inhibits the RapGAP activity of PlexinB1. RapGTP inhibits p120RasGAP activity, resulting in Ras activation and RAS-MAPK signalling [25]. Rnd1-mediated activation of PlexinB1 promotes inactivation of RapGTP via the RapGAP activity of PlexinB1. Consequently, loss of Rnd1 and/or PlexinB1 inactivation leads to de-repression of Ras and promotes metastasis in mouse models of ovarian cancer [26].

In order to examine the effect of the PlexinB1(P1597L) mutation on Ras activity we performed Ras activation assays using Hela cells transfected with cytoPlxnB1(WT)-myc, or cytoPlxnB1(P1597L)-myc, cytoPlxnB1(RA)-myc, or vector-myc, using Raf1-RBD beads, which bind GTP-bound Ras specifically.

Cells transfected with cytoPlxnB1(WT)-myc showed a significant reduction in Ras activity, as expected (Fig. 4). Overexpression of PlexinB1 protein with mutations in the GAP domain which lacks GAP activity [35] (cytoPlexinB1(RA)), increased Ras activity relative to cells transfected with WT PlexinB1 or with empty vector, showing that the decrease in Ras activity upon expression of WT PlexinB1 is dependent on the GAP domain of PlexinB1. Overexpression of cytoPlxnB1(P1597L)-myc also had significantly higher levels of Ras activity relative to cells expressing WT PlexinB1 (Fig. 4, Supplementary Fig. 1C). These results show that PlexinB1-mediated Ras repression is reversed by the P1597L mutation.

Fig. 4

PlexinB1 regulates the level of GTP-bound Ras A Control for Ras activation assays; cell lysates were pre-loaded with a non-hydrolyzable form of GTP (GTPγS) or GDP, immunoprecipitated with Raf1-RBD agarose beads and immunoblotted with Ras antibody B. Effect of P1597L mutation on Ras activity. Lysates of Hela cells, transfected with cytoPlxnB1(WT)-myc; or cytoPlxnB1(P1597L)-myc; cytoPlxnB1(RA)-myc or vector-myc were immunoprecipitated with Raf1-RBD agarose beads and the ratio of RasGTP to total Ras assessed by immunoblotting. C Bar chart of average band intensities, RasGTP/total Ras, (n = 3) (*p =  < 0.05, **P =  < 0.001, 2 tailed student T-Test)

Effect of the P1597L mutation on MLC2 phosphorylation and stress fiber formation

B-type plexins regulate Rho via PDZRhoGEF/LARG [22] and p190RhoGAP [23] and act as GAPs for Rap1B and Rap2A [36] which in turn regulate Rho [25]. To establish whether the P1597L mutation in Plexin-B1 might affect Rho activation, phosphorylation of myosin light chain (phospho-MLC2^Ser19—a marker of Rho/ROCK activation [37])   was monitored following transfection of HCA-2 cells with Plexin(WT) or Plexin(P1597L). HCA-2 are human dermal fibroblasts (HDFs); they were chosen because they readily form actin stress fibers upon Rho/ROCK activation. Transfection of ROCK(1–420) was used as a positive control for MLC2 phosphorylation.

Phosphorylation of MLC2 was reduced by overexpression of PlexinB1(WT), implying inactivation of Rho. In contrast, expression of PlexinB1(P1597L) did not reduce MLC2 phosphorylation relative to control levels, indicating a block to PlexinB1-mediated Rho/ROCK inactivation in vitro (Fig. 5).

Fig. 5

PlexinB1-induced repression of MLC2-phosphorylation is blocked by the P1597L mutation. Representative images of HCA-2 cells transfected with cytoPlexinB1(WT)-myc, cytoPlexinB1(P1597L)-myc, ROCK(1–420)-myc or control empty vector, fixed and stained for actin (phalloidin), myc (PlexinB1/ROCK1 (1–420)), pSer19-MLC2, and DAPI. Scale bar represents 20 µm. Graph shows intensity of pSer19-MLC2 staining. The intensity of phospho-MLC2 staining and cell area were scored blind using imageJ. The total number cells scored was: 37 control cells, 13 cytoPlexinB1(WT)-transfected cells, 16 cytoPlexinB1(P1597L)-transfected cells (n = 3). *** p < 0.0005 (ANOVA). Error bars = SEM

Rho/ROCK activation and MLC2 phosphorylation lead to actin stress fiber formation [38]. Consistent with the MLC2 phosphorylation results, overexpression of PlexinB1(WT) (either cytoplasmic domain or full length) reduced stress fiber formation (Fig. 6). The decline in PlexinB1-induced stress fiber formation was inhibited by the P1597L mutation. Transfection of RhoAV14 was used as a positive control for actin stress fiber formation (Fig. 6).

Fig. 6

PlexinB1-induced repression of actin stress fiber formation is blocked by the P1597L mutation. Representative images of HCA-2 cells transfected with cytoPlexinB1(WT)-myc, cytoPlexinB1(P1547L)-myc, full length-PlexinB1(WT)-myc, full length-PlexinB1(P1597L)-myc, RhoAV14-myc or control empty vector, fixed and stained for actin (phalloidin), myc (PlexinB1 or RhoA) and DAPI. Zoomed images of area indicated show actin staining of transfected cells. Scale bar represents 20 µm; arrows indicate stress fibers. Graph shows intensity of actin staining. The intensity of actin staining and cell area were scored blind using imageJ. The total number cells scored was: 54 control cells, 71 cytoPlexinB1(WT)-transfected cells, 56 cytoPlexinB1(P1597L)-transfected cells, 18 RhoAV14 transfected cells (n = 3). ** p < 0.001, (ANOVA) Error bars = SEM

These results suggest that the P1597L mutation reverses the negative effect of wild-type PlexinB1 on Rho/ROCK signalling and actin stress fiber formation.

PlexinB1 (P1597L) blocks PlexinB1-induced cell morphological collapse

Semaphorins were first identified in the nervous system where they mediate repulsive axon guidance via their plexin receptors and induce growth cone collapse of neuronal cells [10]. The RapGAP activity of plexins is required for semaphorin-induced growth cone collapse [36]. Similarly, morphological collapse of non-neuronal cells occurs upon stimulation of plexins with semaphorins or upon ectopic expression of plexins [39] and mutation of the GAP domain of plexins abolishes plexin-induced collapse [28, 40]. We tested the effect of the P1597Lmutation on this well-established test of plexin function– the morphological collapse of Cos7 cells [41].

Overexpression of wild-type cytoplasmic PlexinB1 (cytoPlxnB1(WT)-myc) in Cos7 cells resulted in cell collapse as expected (Fig. 7). In contrast, overexpression cytoPlexinB1(RA) failed to induce cell collapse, showing that the GAP activity of PlexinB1 is required for this phenotype change. The P1597L mutation in PlexinB1 also blocked PlexinB1-mediated Cos7 cell collapse. These results show that the P1597L mutation inhibits the PlexinB1-mediated cell collapse.

Fig. 7

PlexinB1(P1597L) mutation blocks PlexinB1-induced cell collapse A Representative images of Cos7 cells transfected with cytoPlxnB1(WT)-myc, cytoPlxnB1(P1597L)-myc, cytoPlxnB1(RA)-myc or vector-myc empty. Fixed cells were stained with anti-myc antibody to detect PlexinB1 and anti-tubulin antibody to detect microtubules. Cell images were taken using Eclipse Ti-E inverted confocal microscope (Nikon) using × 40 objective. Scale bar represents 20 µm. B,C. Data is expressed as the mean percentage ± standard error mean (SEM) of cells showing a collapsed phenotype (defined as cells with a contracted cell body with one or more extensions of greater length than the cell body). 102 or more cells counted per condition per experiment (n = 3) ****p < 0.0001, ns not significant; ordinary one-way ANOVA (analysis of variance) with Tukey post-hoc test

Discussion

Our results show that overexpression of wild-type PlexinB1 decreases the levels of active Ras, as well as active Rap. The decrease in Ras activity observed upon PlexinB1(WT) expression is likely to result from the RapGAP activity of PlexinB1. RapGTP inhibits p120RasGAP, leading to Ras activation [25]. In vivo, prostate-epithelial cell-specific expression of WT PlexinB1 decreases invasion and metastasis in mouse models of prostate cancer [17]. Our results show that this decrease in metastasis upon PLXNB1(WT) overexpression may in part result from Ras suppression.

We have found that overexpression of PlexinB1(P1597L) blocks the RapGAP activity of PlexinB1 and leads to elevated Ras activity. Loss of the RapGAP activity of PlexinB1 seen for the P1597L mutation may increase Ras activity, leading to the tumour progression and the increase in metastasis observed in transgenic prostate cancer mouse models expressing PLXNB1(P1597L) [17].

Consistent with our findings, depletion of PlexinB1 in ovarian cancer models increases metastasis through loss of Rap-p120RasGAP-mediated Ras repression [26]. In addition, Rap1 activation promotes prostate cancer progression [42].

Our in vitro studies show that overexpression of WT PlexinB1 reduces Rho/ROCK activation, as shown by a reduction in MLC2 phosphorylation and stress fiber formation, while the P1597L mutant form blocks the decrease. In vivo, the increase in metastases induced by the overexpression of PlexinB1(P1597L) is reversed when RhoA/C or PDZRhoGEF is deleted [17], or by treatment with a ROCK inhibitor [17], indicating that Rho/ROCK is required for PLXNB1(P1597L)-induced metastasis. The net effect of plexin activation on Rho activity is determined by the balance between at least three signalling pathways (Supplementary Fig. 2): 1.via PlexinB1-mediated activation of the receptor tyrosine kinase (RTK) ErbB2 leading to Plexin phosphorylation at Y1708/Y1732, PLC binding and PDZRhoGEF activation [18]; 2. PlexinB1-mediated activation of the RTK Met, leading to PlexinB1 phosphorylation at Y1864/Y2094, Grb2 binding and p190RhoGAP activation [23]; 3. Rap inactivation through the GAP activity of PlexinB1 [36] leading to inactivation of Ras [26]; both Rap and Ras regulate Rho [25]. The P1597Lmutation does not affect ErbB2 binding [16], which occurs via the extracellular domains of both proteins [18] and our Akt/ERK phosphorylation studies suggest that the P1597L mutation has little effect on the PlexinB1-receptor tyrosine kinase pathways. The effect of the mutation on Rho activity is therefore most likely to result from de-repression of Rap and/or Ras activity. Loss of the RapGAP activity of mutant PlexinB1 and the consequent increase in Rho via this pathway may account for the dependence on Rho of PLXNB1(P1597L)-induced metastasis in vivo [17] (Supplementary Fig. 2).

We have found that the P1597L mutation inhibits PlexinB1-mediated cell collapse suggesting that the mutation generates loss of a tumour suppressor function. However, the mutation is likely to promote a gain of function in addition as levels of metastases were greater in PbCre + Pten ^fl/fl Kras ^G12V PLXNB1 ^P159L and PbCre + Pten ^fl/fl p53 ^fl/fl PLXNB1 ^P159Lmice relative to their respective parental lines [17]. Furthermore an increase in Ras activity was observed for P1597L and RA PlexinB1 mutants relative to empty vector controls.

PlexinB1 mutations were found in approximately 4% of whole genome/exome sequencing studies of prostate cancer (13 large scale studies combined which each include metastases samples, 112/3092 samples, cBioportal), the rate depending on the depth of sequencing and sample types used. For example, PlexinB1 mutations were found 5% of neuroendocrine prostate tumours. This is comparable to the 7% of mutations in AR detected in the same study [43]. Cancer-specific mutations in PlexinB1 are spread over the whole gene, including the GAP domain, with no obvious hotspots (Supplementary Fig. 1). Our study shows that the P1597L mutation results in the loss of an inhibitory activity. As such, many of the other missense mutations found in the GAP domain may have a similar phenotype.

Previous studies have shown higher levels of PlexinB1 expression in clinical prostate tumour samples [7, 8] and lower levels of PlexinB1 expression is a positive prognostic factor [8]. Wild-type PlexinB1 acts as a tumour suppressor in our in vitro and in vivo models of prostate cancer while mutant PlexinB1 promotes metastasis in this context. It is unclear whether the high levels of PlexinB1 observed in clinical samples is of the wild-type or mutant form of the PlexinB1 protein. Analysis of circulating tumour DNA could be used to establish the presence of PlexinB1 mutations in tumour tissue for individual patients.

Treatment of tumour-bearing mice with an inhibitor of PlexinB1 delays castrate-resistant regrowth of prostate tumours in xenograft models [8], while whole body depletion of PlexinB1 in ErbB2 -expressing mouse models of breast cancer or in PbCre + Pten ^fl/fl Kras ^G12V and PbCre + Pten ^fl/fl p53 ^fl/fl mouse models of prostate cancer reduces metastasis [17] indicating the therapeutic potential of blocking this pathway.

Conclusion

Our results suggest that the clinically relevant P1597L mutation in PlexinB1 promotes prostate cancer metastasis in mouse models through Rap and Ras activation, highlighting this pathway for therapeutic intervention.