Plexin-B1 Mutation Drives Metastasis in Prostate Cancer Mouse Models

Metastasis is the primary cause of morbidity and mortality in prostate cancer, yet despite increasing interest in treating oligometastatic disease with curative intent (1), few effective treatments specifically developed to counter the metastatic process are currently available in the clinic. It is also not clear what drives prostate cancer progression from locally advanced to invasive/disseminated disease but it is likely to result from the activation of a complex combination of multiple signaling pathways (2). Studies using in vivo models of prostate cancer of different genetic backgrounds allow the role of individual pathways contributing to the process to be defined and potential therapeutic targets identified.

The early stages of metastasis involve the migration and invasion of tumor cells out of the primary tumor through the basement membrane and beyond. One set of genes implicated in this process is plexins, cell surface receptors for semaphorins (3). Vertebrates possess nine plexin genes, classified into four subfamilies [class A(1–4), B(1–3), C1, and D1] (3). Plexin stimulation delivers directional cues for cell migration and axon guidance through the regulation of several small GTPases and semaphorin-plexin signaling can either be attractive or repulsive depending on the particular plexin coreceptor expressed (4, 5), while nonpolarized stimulation of cells with semaphorins results in cell collapse (6). B-class plexins can interact with the GTPases Rnd1–3 (7, 8), Rac (9), RhoD (10), R-Ras (11), and M-Ras (12) and regulate Rho via PDZRhoGEF/LARG (13) and p190RhoGAP (14). In addition, the plexin cytoplasmic tail contains a GTPase-activating protein (GAP) domain (15) divided into two regions by a Rho binding domain (16) and plexins act as GAPs for Rap1B and Rap2A (17). Plexin-B1 interacts with the receptor tyrosine kinase receptors ErbB2 (18) and c-Met (19) and regulates the translocation of the hormone receptors androgen receptor (AR) or glucocorticoid receptor to the nucleus where they are active (20, 21). Loss of Plexin-B1 expression is associated with poor prognosis in melanoma (22) and estrogen receptor–positive breast cancer (23) but is a favorable prognostic factor for ErbB2-amplified breast cancer (24) and ovarian cancer (25) suggesting that Plexin-B1 can act as a tumor suppressor or an oncogene depending on context.

Somatic missense mutations in Plexin-B1 have been detected in patient samples of prostate cancer metastases (26). Recent studies have shown a PLXNB1 mutation rate of around 5% [ranging from 3.3% (27), 3.4% (28), 4.9% (29), to 7.8% (30)] (cBioportal), compared with 9.8%, 6.8%, 2.4%, and 13.7% in PTEN in the same studies. Functional analysis in vitro of three such mutations (T1697A, T1795A, L1815P), demonstrated that these sequence changes inhibit the interaction of Rnd1, Rac, and R-Ras with Plexin-B1 and block the ability of Plexin-B1 to mediate R-Ras inactivation (26, 31). Overexpression of either of these three mutations, or of another GAP domain mutation (P1597L; Supplementary Fig. S1A), increased cell motility, invasion, and anchorage-independent growth of prostate cancer cells in vitro. In contrast, overexpression of WT Plexin-B1 reduced cell motility and invasion (26). The P1597L mutation, investigated in this study, was found in three prostate cancer bone metastases and two lymph node metastases, while a different mutation, at the same site (P1597S) was found in three lymph node metastases (26). 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 in this setting.

To gain in vivo insights into the cellular mechanisms involved in metastasis and to investigate the role of Plexin-B1 in this process, we genetically manipulated the expression of WT and mutant (P1597L) Plexin-B1 in two models of prostate cancer and studied the effect on prostate cancer progression in vivo. The two mouse models were chosen to recapitulate common defects found in prostate tumors (32), namely: loss of PTEN or dysregulation of PI3K/Akt pathway, Ras/MAPK activation and/or p53 loss or mutation. We find that Plexin-B1 plays a significant role in the spread of prostate cancer, with high expression of the clinically relevant P1597L-mutant enhancing metastasis but expression of the WT protein suppressing it, likely through activation of the Rho/ROCK (Rho-associated protein kinase) pathway. This pathway has potential as a therapeutic target in locally advanced or oligometastatic prostate cancer, particularly when mutant Plexin-B1 is present.

This study was approved by the Cardiff University Animal Welfare and Ethical Review Body and carried out under the authority of appropriate Home Office Personal and Project Licences and with reference to ARRIVE guidelines. Animals were monitored regularly and predefined humane endpoints were strictly adhered to (33). Research was hypothesis and objective driven to minimize the number of animals used but power calculations ensured sufficient animals were included in cohorts to gain statistically significant results. Randomization was not appropriate as animals had to be assigned to cohorts according to their genotype. Pb-Cre⁺ (Pb-Cre4, ARR2PB) mice were obtained from the Mouse Models of Human Cancers Consortium (NCI, Frederick, MD). The PB-Cre transgene was incorporated into cohorts using male mice, as PB-Cre⁺ female mice have been shown to recombine in the ovaries (34, 35). Littermate controls lacking the Pb-Cre transgene were used in all experiments. Mice homozygous for floxed Pten exons 4 and 5 (Pten^fl/fl; ref. 36), mice carrying inducible endogenous Kras^G12V oncogene (37), mice homozygous for floxed p53 exons 2–10 (p53^fl/fl; ref. 38) and constitutive Plexin-B1–knockout (PlxnB1^−/−) mice lacking exons 13–16 (39) have been described previously. Mice were maintained on an outbred background. RhoA^flox/flox, RhoC^Δ2–3, and PDZRhoGEF^−/− mice were obtained from Yi Zheng lab.

Mice carrying a conditionally activated knock-in construct of human Plexin-B1 containing WT or a C5060T mutation causing a substitution of proline 1597 to leucine (loxP-STOP-loxP-PLXNB1^WT or loxP-STOP-loxP-PLXNB1^P1597L mice; hereafter PLXNB1^WT and PLXNB1^P1597L) were commercially generated by genOway. The cDNA construct P1597L-PLXNB1 had been made and sequenced previously (26). A transgenic cassette expressing the WT or mutated P1597L-PLXNB1 cDNA – hGH polyA under the control of the CAG promoter and with a neomycin STOP sequence flanked by two loxP sites was generated. GenOway's validated Rosa26 “Quick Knock-in” approach was then used to introduce a single copy of the cassette into the Rosa26 locus on chromosome 6 through homologous recombination in embryonic stem (ES) cells (Supplementary Fig. S1). The linearized construct was transfected into mouse 129Sv ES cells according to standard electroporation procedures (i.e., 5 × 10⁶ ES cells in presence of 40 μg of linearized plasmid, 260 V, 500 μF). Positive selection was started 48 hours after electroporation by addition of 200 μg/mL of G418. G418-resistant clones were screened for the correct homologous recombination event at the Rosa26 locus by PCR and southern blotting. Three correctly recombined ES clones were used for injection into C57BL/6J blastocysts. Injected blastocysts were reimplanted into OF1 pseudopregnant females and allowed to develop to term. After approximately 3 weeks, a total of 3 male chimeric mice were produced per construct with a chimerism rate above 50%. These animals were mated with WT C57BL/6J females to generate heterozygous mice carrying the Rosa26 floxed allele in their germline.

To assess whether the ES cells contributed to the germ layer of the chimeras, mouse coat color markers were used. The coat color marker of the 129Sv ES cells (agouti) is dominant over the black coat color of the C57BL/6 mice. Agouti F1 progeny were screened for the Rosa26 knock-in allele by PCR and southern blotting, using genomic DNA isolated from tail biopsies. A total of 9 animals out of 30 carried the allele and 6 animals were verified further by southern blotting (Supplementary Fig. S1).

PLXNB1^WT and PLXNB1^P1597L knock-in animals were bred to Pb-Cre⁺ mice to obtain animals with prostate epithelial–specific expression of WT or mutant Plexin-B1 protein. Mice were genotyped for the PLXNB1^WT or PLXNB1^P1597L-inducible allele (heterozygous or homozygous) by PCR either according to a genOway protocol (forward AAGACGAAAAGGGCAAGCATCTTCC, reverse GCAGTGAGAAGAGT ACCACCATGAGTCC, 94°C for 2 minutes and 35 cycles of 94°C for 30 seconds, 65°C for 30 seconds, 68°C for 5 minutes, giving a 1,870 bp product in the inducible PLXNB1^P1597L mouse) or according to our simplified protocol identical to quantitative RT-PCR analysis of PlxnB1 expression. To distinguish between heterozygous and homozygous PLXNB1^P1597L we followed a protocol developed by genOway (forward CAATACCTTTCTGGGAGTTCTCTGC, reverse CTGCATAAAACCCCAGATGACTACC, 94°C for 2 minutes and 35 cycles of 94°C for 30 seconds, 55°C for 30 seconds, 72°C for 30 seconds, giving a 304 bp product in WT or heterozygous PLXNB1^P1597L mouse and no product in homozygous PLXNB1^P1597L mouse).

Primary tumors were excised and weighed (Supplementary Table S1). Prostate tissue was dissected in 1× PBS and fixed in ice-cold 10% neutral buffered formalin for no longer than 24 hours before being processed into paraffin blocks according to standard procedures. For IHC, 5 μm sections were dewaxed in xylene, rehydrated in ethanol, and antigen retrieval was performed by heating in either citrate (pH 6.0) or Ethylenediaminetetraacetic acid (EDTA) buffer (pH 8.0) in a pressure cooker for 15 minutes after reaching full pressure. Sections were cooled for 15 minutes, blocked in 0.5% hydrogen peroxide for 5 minutes at room temperature, and then blocked with 20% normal rabbit or goat serum (DAKO, Agilent) for 20 minutes and incubated with the primary antibody overnight at +4°C. After washing in TBS/0.05% Tween, sections were incubated in secondary antibody for 30 minutes (EnVision+ System- HRP Labeled Polymer; Dako) and the staining was visualized with DAB (EnVision+ System). Details for all the primary antibodies used in study are listed in Supplementary Table S2. Tissue sections were assessed by a pathologist with a special interest in urological pathology and on the specialist register who was masked to the genotype. Prostate epithelial cells were visualized by AR and pan-cytokeratin staining.

Ki67, phospho-MLC2^Ser19, and AR staining were each quantified in prostates of 100-day-old mice (for each stain n = 3 per cohort, 5 fields for each sample). Ki67 staining was scored “blind” as positive or negative and compared with the total number of nuclei in a field. For semiquantitative analysis of phospho-MLC2^Ser19 staining, we used a Histo-score (H-score) formula: 3 × percentage of strongly staining cells + 2 × percentage of moderately staining cells + percentage of weakly staining cells, giving a range of 0 to 300, scored “blind.”

For quantitation of invasion, 5–10 random images of tumor sections, from each of three biological replicates, were scored blind for invasion into the stroma, at 40× magnification. Invasion was scored by counting pan-cytokeratin positive cells breaking the basement membrane or located inside the stromal compartment, divided by total number of pan-cytokeratin positive cells.

Prostate lobes were dissected in ice-cold PBS. Tissues were homogenized in TRIzol Reagent (Invitrogen), extracted using standard phenol–chloroform protocol and RNA was purified further using an RNA extraction kit (Norgen). RNA from dorsolateral lobes of age-matched 100-day-old mice (n = 2 of each genotype) were used for qRT-PCR. Reverse transcription was performed using the SuperScript III reverse transcriptase kit and random hexamers (Invitrogen) according to the manufacturer's instructions. SYBR Select Master Mix (Applied Biosystems, Thermo Fisher Scientific) was added to cDNA samples and primers. Samples were run using QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems). Reverse transcriptase negative controls were included in all analyses. Plexin-B1 primers were used for detecting both endogenous PlxnB1 and PLXNB1^MUTtranscripts (forward-TGTCACTATCAGGGGCTCCA, reverse-CTCCCCGCTGGCTCCAGTGAT, 94°C for 2 minutes and 35 cycles of 94°C for 30 seconds, 55°C for 30 seconds, 72°C for 30 seconds, giving 145 bp products for both WT and inducible PLXNB1^MUT). β-Actin and GAPDH were used as reference genes.

Data available on request.

To understand the contribution of the clinically relevant Plexin-B1(P1597L) mutation to prostate cancer progression, we established two lines of mice carrying a targeted insertion of either PLXNB1^WT or PLXNB1^P1597L cDNA preceded by a flox-STOP-flox cassette into the Rosa26 locus. Targeting was confirmed by Southern blotting (Supplementary Fig. S1). Activation/overexpression of these conditional alleles in the prostate was achieved by crossing with a line in which CRE recombinase was expressed under the Probasin promoter, which is specifically expressed in the prostate epithelium only (see Materials and Methods). Expression of Plexin-B1 in prostates of PLXNB1^WT and PLXNB1^P1597L mice was compared with unmanipulated WT mice and mice with a germline deletion of Plexin-B1 (PlxnB1^−/−; ref. 39; Supplementary Fig. S2). Plexin-B1 protein was expressed in the epithelial cells of all lobes of WT mouse prostates, localizing to the cell membrane, cytoplasm, and nucleus of prostate epithelial cells, but was absent from prostate stroma (Supplementary Fig. S2) and was absent in PlxnB1^−/− mice. Similar levels of high expression of Plexin-B1 were observed in prostate epithelial cells of PLXNB1^WT and PLXNB1^P1597L mice (Supplementary Fig. S2). No obvious structural or histologic changes in the prostate were found between WT and PLXNB1^WT or PLXNB1^P1597L mice. Mice from all lines were fertile and survived over 500 days.

Prostate tumors are characterized by loss of PTEN and dysregulation of the PI3K/AKT pathway (32, 40), p53 deletion or mutation (41), and activation of the Ras/Raf pathway (40). To determine whether Plexin-B1 contributes to prostate cancer progression in vivo, we tested the effect on tumor growth and metastasis of manipulating Plexin-B1 expression in two different transgenic mouse models of prostate cancer which recapitulate these common genetic alterations and which also metastasize: PbCre⁺Pten^fl/flKras^G12V (ref. 42; hereafter abbreviated to Pten^fl/flKras^G12V) and PbCre⁺Pten^fl/flp53^fl/fl (ref. 43; hereafter abbreviated to Pten^fl/flp53^fl/fl). Both the Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl models have moderate/low metastatic ability that enables the analysis of additional alleles which may accelerate tumor formation/progression; prostate tumors in both models can metastasize to lumbar lymph nodes with limited ability to form distant metastases (42, 43). Pten^fl/flKras^G12V differs from the previously described highly metastatic PbCre⁺Pten^fl/flKras^G12D model (42) in using a less aggressive Kras^G12V (37). The Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl lines were crossed with PLXNB1^P1597L mice. In addition, the Pten^fl/flKras^G12V line was crossed with PLXNB1^WT mice to establish the effect of overexpression of WT Plexin-B1 on tumor progression.

Two cohorts were established for each cross, one for euthanasia at a fixed timepoint of 100 days and one for euthanasia when required for welfare reasons due to tumor morbidity. Primary prostate tumors, local lymph nodes, and visceral organs were all processed for histologic analysis to assess primary tumor morphology, extent of local invasion, and the presence of local or distant metastases.

Plexin-B1 was expressed at moderate levels in the epithelial cells of Pten^fl/flKras^G12V prostate tumor cells (Supplementary Fig. S2E, antibody specificity shown in Supplementary Fig. S2F), and expressed at high levels in Pten^fl/flKras^G12VPLXNB1^WT (Supplementary Fig. S2G and S2H) and Pten^fl/flKras^G12VPLXNB1^P1597L primary tumors and metastases (Supplementary Fig. S2I–S2K).

As described previously (40), prostates from Pten^fl/flKras^G12V-based cohorts displayed distorted glandular structure with focal areas of microinvasion adjacent to reactive stroma and regions of sarcomatoid metaplasia (Fig. 1A–D). There were no overt differences in histology of primary tumors between the Pten^fl/flKras^G12V, Pten^fl/flKras^G12VPLXNB1^WT, and Pten^fl/flKras^G12VPLXNB1^P1597L models (Fig. 1B–I).

PLXNB1^WT overexpression made no significant difference to survival of the Pten^fl/flKras^G12V line (Fig. 1J). Unexpectedly, however, expression of PLXNB1^P1597L significantly increased the survival of Pten^fl/flKras^G12V mice (Fig. 1J). Pten^fl/flKras^G12VPLXNB1^P1597L mice survived for a median of 226.5 days, compared with 182 and 176 days for Pten^fl/flKras^G12V and Pten^fl/flKras^G12VPLXNB1^WT mice, respectively (P = 0.0148 vs. Pten^fl/flKras^G12V; Fig. 1J). Prostate tumor growth in Pten^fl/flKras^G12VPLXNB1^P1597L animals was heterogeneous. Cell proliferation rates in Pten^fl/flKras^G12VPLXNB1^WT tumors were also not statistically different from those of Pten^fl/flKras^G12V cohorts at 100 days, as demonstrated by Ki67 staining of prostate epithelial cells (Fig. 1K). However, there was a wide variation in tumor cell Ki67 staining between different Pten^fl/flKras^G12VPLXNB1^P1597L mice at 100 days, suggesting that this cohort developed a heterogeneous mix of slowly and rapidly growing primary tumors which overall resulted in an increase in median survival.

Expression levels of Plexin-B1 were moderate in Pten^fl/flp53^fl/fl tumor cells (Supplementary Fig. S2L, antibody specificity shown in Supplementary Fig. S2M) and high in Pten^fl/flp53^fl/flPLXNB1^P1597L primary tumors and metastases (Supplementary Fig. S2N–S2P).

Prostate tumors from all Pten^fl/flp53^fl/fl-based cohorts had a marked increase in mesenchymal phenotype with little epithelial component compared with that of Pten^fl/flKras^G12V-based cohorts (Fig. 2A–I) and these sarcomatoid tumors were the cause of morbidity in accordance with earlier data (43). Prostate tumors in Pten^fl/flp53^fl/fl and Pten^fl/flp53^fl/flPLXNB1^P1597L mice showed similar progression from adenocarcinoma at day 100 toward sarcomatoid metaplasia at 6 months (Fig. 2B–I).

Expression of PLXNB1^P1597Lsignificantly increased the survival of Pten^fl/flp53^fl/fl mice (Fig. 2J). Median survival of Pten^fl/flp53^fl/flPLXNB1^P1597L mice was 211 days, compared with 177 days in Pten^fl/flp53^fl/fl mice (P < 0.001;Fig. 2J). Consistent with these findings, Ki67 staining of prostates of 100-day Pten^fl/flp53^fl/flPLXNB1^P1597L mice showing a 1.32-fold decrease in cell proliferation compared with the Pten^fl/flp53^fl/f cohort (P < 0.01; Fig. 2K).

Prostate cancer mouse models typically need to be euthanized as a result of local complications associated with primary tumor bulk. The suppression of primary tumor proliferation by PLXNB1^P1597L overexpression may contribute to the extended survival in these models.

Next, we quantified the metastatic lesions in the different tumor cohorts by histologic examination. The prostate origin of metastatic lesions was confirmed by staining for AR, a prostate epithelial cell marker (Supplementary Fig. S3–S5).

Importantly, expression of mutant PLXNB1^P1597Lin Pten^fl/flKras^G12V significantly increased the percentage of mice with metastases, compared with the parental line (P = 0.0452) and to Pten^fl/flKras^G12VPLXNB1^WT mice (P < 0.0010; Fig. 3A–C; Table 1). This increase in metastases was particularly evident in sites distant from the prostate—the number of mice with lung metastases increased from 5% to 21.43% upon expression of PLXNB1^P1597L.

The metastatic deposit composition varied between adenocarcinoma, sarcomatoid (low pan-cytokeratin staining), and squamous areas (high pan-cytokeratin staining; Supplementary Fig. S3–S5). Adenocarcinoma was the predominant tissue type in Pten^fl/flKras^G12V cohort metastases (Fig. 3D–F; Supplementary Fig. S3) with one animal with a lymph node metastatic deposit of predominantly squamous differentiation at day 200 (Supplementary Fig. S3C) and one animal with sarcomatoid deposits in nodes, peritoneum and lung at day 253 (Fig. 3D–F; Supplementary Fig. S3F).

Metastases in Pten^fl/flKras^G12VPLXNB1^P1597L mice showed a greater heterogeneity (Fig. 3G–M; Supplementary Fig. S4); 6 of 10 Pten^fl/flKras^G12VPLXNB1^P1597L mice that were taken by day 200 developed predominantly sarcomatoid metastases (Fig. 3H, J, L; Supplementary Fig. S4A–F and S4H) and Pten^fl/flKras^G12VPLXNB1^P1597L mice taken after day 200 displayed mixed adenocarcinoma/squamous differentiation (Fig. 3; Supplementary Fig. S4). There was no correlation between the increased age of Pten^fl/flKras^G12VPLXNB1^P1597L mice and the presence of local or distant metastases (Supplementary Fig. S4). Mice without metastasis lived on average longer than those with distant metastasis, showing that the presence of metastases cannot be explained by an increase in survival.

In contrast, overexpression of PLXNB1^WT significantly suppressed the metastatic spread of tumors in Pten^fl/flKras^G12V mice (P = 0.0121 and P < 0.001 compared with Pten^fl/flKras^G12V and Pten^fl/flKras^G12VPLXNB1^P1597L mice, respectively; Fig. 3B and C; Table 1), with only a single local metastasis observed in Pten^fl/flKras^G12VPLXNB1^WT mice (4.5%; Fig. 3N; Supplementary Fig. S5) of adenocarcinoma histology.

Expression of mutant PLXNB1^P1597Lin Pten^fl/flp53^fl/fl mice significantly increased the percentage of mice with metastases, in comparison with the parental Pten^fl/flp53^fl/fl line (P = 0.0154; Fig. 4A–C; Table 1). Metastatic deposits in Pten^fl/flp53^fl/fl mice were all represented by the primary sarcomatoid tumors encroaching the lumbar lymph nodes and invading further into the peritoneum (Fig. 4D–F; Supplementary Fig. S6). Pten^fl/flp53^fl/flPLXNB1^P1597L mice showed a marked increase in locally invasive tumors demonstrating invasion of the primary tumor into local structures or organs including peritoneum, pelvic or bladder muscle, vas deferens, and lymph node (Fig. 4G–K; Supplementary Fig. S7). The relationship between the presence of local metastasis and survival in Pten^fl/flp53^fl/flPLXNB1^P1597L mice is shown in Supplementary Fig. S7M.

Our results show that prostate epithelial cell–specific overexpression of mutant (P1597L) Plexin-B1 increases metastasis. Cancer metastasis is a multistep process which begins with tumor cell invasion from prostate acini through the basal membrane and into the prostate stroma. To establish the stage at which stage Plexin-B1 expression affects metastasis, we investigated the effect of expression of the different forms of Plexin-B1 on invasion of primary tumor cells into the prostate stroma.

Primary tumors from all cohorts of Pten^fl/flKras^G12V or Pten^fl/flp53^fl/fl mice were immunostained for the epithelial marker pan-cytokeratin at the early timepoint of 100 days (before the onset of metastasis) and the percentage of tumors breaking into the stroma was scored (Fig. 5). Overexpression of PLXNB1^P1597Lin prostate epithelial cells increased invasion in both Pten^fl/flKras^G12V (Fig. 5A–D) and Pten^fl/flp53^fl/fl (Fig. 5E–G) models (2.6%–8%, P < 0.05 and 1.6%–2.3%, P < 0.05, respectively). In contrast, overexpression of WT Plexin-B1 in prostate epithelial cells of Pten^fl/flKras^G12V mice suppressed tumor cell invasion into the stroma (P < 0.05). These results show that mutant Plexin-B1 enhances metastasis and WT Plexin-B1 inhibits metastasis at an early stage in the metastatic process.

Plexin-B1 activates RhoA and RhoC through PDZRhoGEF/LARG (13), which bind to the C-terminus of Plexin-B1, and inactivates RhoA and RhoC through p190RhoGAP activation. Plexin-B1–mediated activation of RhoA/C is a key pathway promoting metastasis in ErbB2-mouse models of breast cancer (24). To establish whether Plexin-B1 might signal via RhoA/C to promote metastasis in the mouse models of prostate cancer, phosphorylation of myosin light chain (phospho-MLC2^Ser19—a marker of ROCK activation (44)) was evaluated in tumors of Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl models (Fig. 5).

Overexpression of PLXNB1^P1597L significantly increased MLC2 phosphorylation H-score in both Pten^fl/flKras^G12V (Fig. 5J) and Pten^fl/flp53^fl/fl (Fig. 5M) models (4.8-fold and 3.3-fold increase in the percentage of cells with strong staining for Pten^fl/flKras^G12VPLXNB1^P1597L and Pten^fl/flp53^fl/flPLXNB1^P1597L, respectively, P = <0.05; Fig. 5).

In contrast, overexpression of WT Plexin-B1 in the Pten^fl/flKras^G12V model decreased MLC2 phosphorylation in tumors (P = <0.05; Fig. 5I). These results suggest that Plexin-B1 promotes metastasis at least in part, via the Rho-ROCK pathway.

We next investigated whether the increase in metastasis observed upon expression of PLXNB1^P1597L in Pten^fl/flKras^G12V mice was dependent on RhoA/C expression. Pten^fl/flKras^G12V mice were crossed with mice containing a conditional inactivation of RhoA (RhoA^fl/fl) combined with constitutive deletion of RhoC (RhoC^−/−;Fig. 6A). Pten^fl/flKras^G12VPLXNB1^P1597L RhoA^fl/flRhoC^−/− mice had a marked reduction in life span (Fig. 6B) largely attributed to enhanced skin wart formation around the penis area. Importantly, deletion of RhoA/C in Pten^fl/flKras^G12VPLXNB1^P1597L mice resulted in complete elimination of metastases (Fig. 6C and D).

Plexin-B1 activates RhoA/C by activation of PDZRhoGEF, which binds to the C-terminal of Plexin-B1 (13). To assess the contribution of PDZRhoGEF /Rho/ROCK signaling to metastatic tumor progression in Pten^fl/flKras^G12VPLXNB1^P1597L mice, we crossed these mice with mice that harbor deletion of PDZ RhoGEF (PDZRhoGEF^−/−;Fig. 6A). Survival was not affected in the PDZRhoGEF^−/− cohort (Fig. 6B). However, deletion of PDZRhoGEF completely inhibited metastasis (Fig. 6C and D).

Consistent with these results, treatment of Pten^fl/flKras^G12VPLXNB1^P1597L mice with ROCK inhibitor GSK269962 showed similar survival data to the untreated group (Fig. 6B) but had a significant decrease in metastases (χ² test, P = 0.033; Fig. 6C and D).

These results indicate that the increase in metastasis observed upon PLXNB1^P1597L expression in Pten^fl/flKras^G12V mice is dependent on RhoA/C signaling

We have shown that prostate-specific overexpression of mutant (P1597L) Plexin-B1 promotes metastasis, while prostate-specific WT Plexin-B1 had the opposite effect. Endogenous Plexin-B1 is expressed by a variety of cell types including endothelial cells and Plexin-B1 activation promotes angiogenesis (45). To model the effect of deletion of Plexin-B1 in all cells, we next crossed both Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl lines with mice containing a germline deletion of Plexin-B1 (Fig. 7A). As before, one cohort was established for euthanasia at a fixed timepoint of 100 days and one for euthanasia when required for welfare reasons. Plexin-B1 expression was absent from all tissues in PlxnB1^−/− mice, as expected (Supplementary Fig. S2B) and no overt differences in histology of primary tumors were observed, compared with the respective parental lines (Fig. 7B–F).

PlxnB1 ablation made no significant difference to the survival of either Pten^fl/flKras^G12V or Pten^fl/flp53^fl/fl mice (Fig. 7G and H). Median survival of Pten^fl/flKras^G12VPlxnB1^−/− mice was 187.5 days, compared with 182 days for Pten^fl/flKras^G12V mice; median survival of Pten^fl/flp53^fl/fl mice was 185 days compared with 177 days for the Pten^fl/flp53^fl/fl parental line. Consistent with these findings, germline deletion of Plexin-B1 had a negligible effect on cell proliferation in either model at 100 days, as demonstrated by Ki67 staining of prostate epithelial cells (Supplementary Fig. S9A–C).

Germline deletion of Plexin-B1 reduced metastasis substantially in both Pten^fl/flKras^G12V (Fig. 7I and J) and Pten^fl/flp53^fl/fl (Fig. 7K and L) models (P = 0.0411 for Pten^fl/flKras^G12V mice; Table 1). Deletion of Plexin-B1 resulted in a 3-fold reduction in the number of mice with metastases in the Pten^fl/flKras^G12V model: 35% of Pten^fl/flKras^G12V mice had node metastases, including one mouse with an additional lung metastasis; in contrast, 10.7% of Pten^fl/flKras^G12VPlxnB1^−/− mice had node metastases and no distant metastases were found (Fig. 7I and J; Supplementary Fig. S8). Deletion of PlxnB1 in Pten^fl/flp53^fl/fl mice completely blocked local invasion and metastases (Fig. 7K and L).

Consistent with these results, deletion of Plexin-B1 significantly reduced invasion into the stroma in both Pten^fl/flKras^G12V (2.6%–0.7%, P < 0.05; Fig. 7M) and Pten^fl/flp53^fl/fl (1.6%–0.4%, P < 0.05; Fig. 7N) backgrounds, showing that PlexinBI is required for the earlier stages of metastasis in these models.

Deletion of Plexin-B1 reduced semiquantitative MLC2 phosphorylation H-score, with a 2-fold reduction in cells with “strong” staining in both models upon Plexin-B1 ablation in 100-day-old mice (P = <0.05; Supplementary Fig. S9D and S9E), suggesting that deletion of Plexin-B1 reduced Rho/ROCK signaling in these tumors.

Together, these results indicate that systemic inhibition of Plexin-B1 has potential as a treatment for prostate cancer.

Our results show that Plexin-B1 status has a major effect on prostate cancer metastasis (see Supplementary Fig. S10 for summary). Overexpression of WT Plexin-B1 targeted specifically to the prostate epithelial cells of Pten^fl/flKras^G12V mice, decreased invasion and metastasis in comparison with Pten^fl/flKras^G12V mice expressing normal levels of Plexin-B1. Sema4D, the ligand for Plexin-B1, is expressed by cells in the prostate stroma and tumor-associated macrophages (46) and Sema4D secreted from the tumor microenvironment may act as a repellent cue to inhibit migration and invasion of tumor cells expressing WT Plexin-B1, confining the tumor cells to the primary tumor mass. Sema4D produced by tumor cells may also act as an autocrine or paracrine signal to suppress migration; nonpolarized activation of Plexin-B1 over the whole cell results in cell collapse in vitro (47). Alternatively, ligand-independent Plexin-B1 signaling due to WT Plexin-B1 overexpression and receptor clustering (19), may repress migration and invasion. B-type plexins have been shown to act in a ligand-independent manner to suppress the rate of cell division through the detection of mechanical forces, in embryonic skin development (48).

In direct contrast to overexpression of WT Plexin-B1, similar levels of mutant (P1597L) Plexin-B1 in the mouse models significantly increased metastasis.

These findings reflect our previous results where overexpression of WT Plexin-B1 decreased migration and invasive capacity of HEK293 cells and overexpression of the P1597L mutant form of Plexin-B1 increased motility and invasion (26). The increase in invasion and metastasis of prostate cancers observed in the Pten^fl/flKras^G12VPLXNB1^P1597L and Pten^fl/flp53^fl/flPLXNB1^P1597L mice may result from a change in response of tumor cells to semaphorins produced by the stroma—a switch from repulsion to attraction (4). Sema4D has been shown to promote or suppress migration and invasion depending on cellular context and the Plexin-B1 coreceptors expressed by the responding cell (5). The contrasting results from similar levels of overexpression of the WT and mutant proteins argue against the findings being an artefact of overexpression per se.

The signaling mechanism by which the single clinically relevant Pro1597Leu amino-acid change converts Plexin-B1 from a metastasis-suppressor to a metastasis-promoter in a Pten^fl/flKras^G12V background, in which the Ras-MAPK pathway is overactivated, is unclear.

Expression of PLXNB1^WT or PLXNB1^P1597L in Pten^fl/flKras^G12V decreased or increased MLC2 phosphorylation, respectively, while deletion of RhoA/C or PDZRhoGEF in Pten^fl/flKras^G12VPLXNB1^P1597L mice inhibited metastasis. Deletion of RhoA/C in Pten^fl/flKras^G12VPLXNB1^P1597L mice shortened their survival. This reduction in survival may have contributed to the reduction in metastasis observed in these mice. However, deletion of PDZRhoGEF, a guanine nucleotide exchange factor for RhoA, had little effect on survival, yet it significantly reduced metastasis, showing that the PDZRhoGEF-Rho-ROCK pathway is required for metastasis in Pten^fl/flKras^G12VPLXNB1^P1597L mice. Together, these results implicate Rho signaling in PLXNB1^P1597L-induced metastasis. RhoA/C are upregulated in many cancer types including prostate cancer and promote metastasis in mouse models (49).

The P1597L mutation is in the GAP domain of Plexin-B1 and so may disrupt the RapGAP activity of Plexin-B1. Rap has diverse functions in tumor progression (50) and Rap1 activation promotes prostate cancer metastasis (51). Activation of the RapGAP activity of Plexin-B1 (WT) through Rnd1 binding has been shown to inhibit Rap and Ras—Rap1 inhibition leads to derepression of p120 Ras-GAP resulting in Ras inactivation (52). Consequently, WT Plexin-B1 activation inhibits tumor progression in mouse models of metastasis (52). Inactivation of the GAP activity of Plexin-B1 by the P1597L mutation is therefore expected to result in Rap1 and Ras activation, promoting metastasis (Supplementary Fig. S10).

The mice models expressing PLXNB1^P1597L showed suppression of primary tumor proliferation and consequently extended survival, yet a considerable increase in metastasis. These findings support a model in which expression of mutant Plexin-B1 switched prostate tumor cells from a proliferative to an invasive phenotype. This led to longer survival of the mice, as there were fewer local complications due to primary tumor mass, but increased tumor cell motility, escape from the primary tumor site and metastasis to lymph nodes and organs. It is important to note that although suffering from morbidity due to local advancement if not treated, men with prostate cancer rarely, if ever, die of complications of the primary tumor (unlike with mice), but rather of the metastatic burden.

Survival times can bias the detectable rates of metastasis. However, there was no correlation between the age of Pten^fl/flKras^G12VPLXNB1^P1597L mice and metastases even though survival was increased in this cohort. Furthermore, deletion of PDZRhoGEF in Pten^fl/flKras^G12VPLXNB1^P1597L mice did not significantly affect their survival but did reduce metastasis, showing that the increase in metastasis in Pten^fl/flKras^G12VPLXNB1^P1597L mice cannot be explained by an increase in survival. Expression of PLXNB1^WT in Pten^fl/flKras^G12V mice or knockout of PlxnB1 in both models did not affect survival but did significantly reduce metastasis.

While overexpression of PLXNB1^P1597L increased invasion and metastasis in both Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl models, a marked difference in histology and means of spread were seen between tumors from the different backgrounds. Expression of PLXNB1^P1597L in Pten^fl/flKras^G12V mice resulted in tumors with an epithelial phenotype and a significant increase in metastasis to lymph nodes and lung, metastasis in this background occurring via the lymphatic and/or via blood vessel route. In contrast, expression of mutant Plexin-B1 in Pten^fl/flp53^fl/fl mice resulted in an increase in tumors with a mesenchymal phenotype which were predominantly locally invasive into surrounding tissues such as the peritoneum, pelvic or bladder muscle, or vas deferens. Consistent with these results, Sema3C drives epithelial–mesenchymal transition in prostate cells (53) promoting a spindle-like morphology. These results demonstrate the context dependence of specific semaphorin/plexin-mediated signaling pathways controlling metastasis.

Most human metastatic prostate cancers exhibit AR-dependent epithelial histology. One limitation of the models used in this study is the finding that Pten^fl/flKras^G12VPLXNB1^P1597L mice and especially Pten^fl/flp53^fl/flPLXNB1^P1597 mice develop metastases with predominantly sarcomatoid and squamous histology. This may have implications on the clinical significance of these models.

While the Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl models used in this study recapitulate alterations in the P13K, Ras/Raf/MEK/ERK, and p53 signaling pathways (32), the effect of mutation of PLXNB1 in a background of other commonly found changes in human prostate cancer, such as AR mutation or amplification, was not assessed here.

Plexin-B1 is overexpressed in some cancers and appears to act as a tumor suppressor gene in others (54). Indeed, high levels of Plexin-B1 expression predict longer overall survival in bladder carcinoma, head and neck squamous cell carcinoma, and kidney papillary renal cell carcinoma but shorter overall survival in thymoma and kidney renal clear-cell carcinoma (data from Kaplan-Meier Plotter Pan-Cancer Tool; https://kmplot.com/analysis/index.php?p=back ground; ref. 55). Data on the prognostic significance of Plexin-B1 in prostate cancer are complicated by the use of different baseline comparators. A large-scale gene expression comparison between prostate cancer and normal tissue (32) showed Plexin-B1 expression was altered in 30% of patients with prostate cancer (z-score = ±2) and Plexin-B1 expression downregulation was three times more common (22.67%) in prostate cancer than its increase (7.33%). However, other large-scale genomics projects using a different baseline (diploid tumor samples instead of normal prostate tissue) have suggested that Plexin-B1 upregulation is more common than its decrease [cBioportal (56, 57) summarized in Supplementary Table S3]. Notably, cBioportal data also show that Plexin-B1 expression is inversely correlated with PTEN (Spearman correlation −0.374, P = 1.34E-05, q-value = 2.56E-05). It is interesting to speculate that elevated levels of WT Plexin-B1 may suppress metastasis in a PTEN-deleted tumor, but if the Plexin-B1 acquires a mutation it may switch to a driver of aggressive metastatic disease.

Germline deletion of Plexin-B1 in Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl mice significantly decreased metastasis in comparison with Pten^fl/flKras^G12V and Pten^fl/flp53^fl/fl mice with normal levels and patterns of Plexin-B1 expression, demonstrating that Plexin-B1 is required for metastasis in these mouse models. Consistent with these results, whole body knockout of Plexin-B1 inhibited metastasis in ErbB2-expressing models of breast cancer (24). Plexin-B1 is expressed by endothelial cells (45, 46) in the tumor microenvironment in addition to tumor epithelial cells. Activation of Plexin-B1 on endothelial cells by Sema4D which is expressed by tumor cells (58), promotes angiogenesis (45, 46, 58); therefore, whole-body knockout of Plexin-B1 in our mouse models would inhibit Sema4D-induced angiogenesis and this may contribute to the decrease in metastasis observed.

Normal levels of WT Plexin-B1 in tumor cells may promote metastasis while overexpression of PLXNB1^WT reduces metastasis. This hypothesis is consistent with our in vitro findings: knockdown of Plexin-B1 reduces migration and invasion in prostate cancer cells expressing ErbB2, while activation of endogenous Plexin-B1 with Sema4D promotes migration and invasion. In contrast, forced overexpression of PLXNB1^WT decreases invasion and migration, while overexpression of PLXNB1^P1597L has the opposite effect in transfected cells (59). Overexpression of PLXNB1^WT may alter the balance of Rho activation (via PDZRhoGEF/LARG) and Rap/Ras Rho inhibition (via the RapGAP activity of Plexin-B1; Supplementary Fig. S11). PLXNB1^WT overexpression may also result in the activation of negative feedback loops.

Human prostate tumors are highly heterogenous and consist of a complex mixture of clones of different genetic make-up which complicates analysis (60). The use of mouse models of a defined genetic background allows the effect of the many different genetic changes found in human tumors to be analyzed separately. Current treatments for metastatic prostate cancer are effective only in the short term, highlighting the need for new therapies for late-stage disease. To test such therapies, preclinical models in which metastasis is driven by clinically relevant mutations, such as those we have developed here, are a key requirement. Together, our results have demonstrated that Plexin-B1 has a complex yet significant role in metastasis and is a potential therapeutic target to block the lethal spread of prostate cancer.

T. Worzfeld reports a patent to WO2020165094A1 issued. No disclosures were reported by the other authors.

B. Shorning: Formal analysis, investigation, methodology, writing-original draft. N. Trent: Investigation. D.F. Griffiths: Formal analysis, investigation, visualization. T. Worzfeld: Resources. S. Offermanns: Resources. M.J. Smalley: Supervision, project administration, writing-review and editing. M. Williamson: Conceptualization, funding acquisition, writing-original draft, writing-review and editing.

This work was funded by Prostate Cancer Research (registered charity number, 1156027), grant no. 6929. We also thank Elaine Taylor for assistance with mouse husbandry and Derek Scarborough and Marc Isaacs for their assistance in histology. RhoA^flox/flox, RhoC^Δ2–3 and PDZRhoGEF^−/− mice were obtained from Yi Zheng lab. This article is dedicated to the memory of the late Professor Alan Clarke.