Podoplanin interaction with caveolin-1 promotes tumour cell migration and invasion

Podoplanin, a highly O-glycosylated type-1 transmembrane glycoprotein, found in lymphatic endothelial cells, podocytes, alveolar epithelial cells and lymph node fibroblasts is also expressed by tumour cells, and is correlated with more aggressive disease. Despite numerous studies documenting podoplanin expression, the mechanisms underlying its tumour-promoting functions remain unclear. Using a murine melanoma cell line that endogenously expresses podoplanin, we demonstrate interactions with proteins necessary for cytoskeleton reorganization, adhesion and matrix degradation, and endocytosis/receptor recycling but also identify a novel interaction with caveolin-1. We generated a panel of podoplanin and caveolin-1 variants to determine the molecular interactions and functional consequences of these interactions. Complementary in vitro and in vivo systems confirmed the existence of a functional cooperation in which surface expression of both full length, signalling competent podoplanin and caveolin-1 are necessary to induce directional migration and invasion, which is executed via PAK1 and ERK1 pathways. Our findings establish that podoplanin signalling mediates the invasive properties of melanoma cells in a caveolin-1 dependent manner. Summary Statement This manuscript describes a new interaction and functional cooperation between podoplanin and caveolin1 that drives tumour cell invasion into surrounding tissues.

an external stimulus which is not caveolin-driven, however disruption to either is sufficient to prevent a response.

Podoplanin transmembrane domain mediates caveolin-1 interactions
To examine the interactions between podoplanin and caveolin-1 in greater detail, and downstream consequences, further B16.F10 variants were generated in which endogenous podoplanin was knocked down and reconstituted with mutant forms (Fig. 5a and Fig. S3).
We created two variants either lacking the short cytoplasmic domain, removing intracellular signalling capacity, or lacking the signal peptide, preventing podoplanin trafficking to the cell surface. A third variant, in which podoplanin deficient cells were transfected with caveolin-1 was also generated to identify podoplanin-independent but caveolin-1 dependent effects.
Confocal analysis confirmed membrane co-localization of podoplanin with caveolin-1 ( Fig.   5b top panel). Moreover, it confirmed that in the absence of signal peptide (DSP), podoplanin failed to reach the cell membrane, remaining cytoplasmic. Notably, in the absence of surface podoplanin, the majority of caveolin-1 also remained intracellular ( Fig. 5b middle panel). In contrast, without the cytoplasmic domain, podoplanin again reached the cell surface colocalizing with caveolin-1 ( Fig. 5b lower panel). As previously demonstrated, wound closure velocity increased with the level of cell surface podoplanin (Fig. 5c). In DSP cells, closure velocity of B16.F10 cells was significantly impaired, returning to levels comparable with knockdown variants. Similarly, variants expressing surface, but signalling-incompetent podoplanin (DCT) exhibited significantly retarded migration velocities, even with surface caveolin-1. Reconstitution with caveolin-1 in the absence of podoplanin was not capable of restoring migration further confirming that caveolin-1 was not the dominant driver. In fact, velocities were further reduced indicative of a dual function for caveolin-1 which in this case, may be tumour suppressive in the absence of podoplanin. These data point to the requirement of surface expression of signalling competent podoplanin for directional tumour migration, which is dependent on interaction and functional cooperation with caveolin-1. This interaction is likely to occur via the transmembrane domain. In vivo, the type of podoplanin (functional status or cellular location) had no impact on primary tumour growth (Fig. 5d), but the extracellular domain of podoplanin was sufficient for colonization ( Fig. 5e) since both KD+CAV1 OE and DSP exhibited highly inefficient colony formation (Fig. 5e). In contrast, cells lacking the cytoplasmic tail formed lung nodules comparable to signalling competent podoplanin (Fig. 5e). These findings indicate that functional cooperation between podoplanin and caveolin-1 assists egress from the primary tumour through directional migration and invasion rather than a controlling growth or secondary colonization. Here, the formation of emboli through cell-cell or cell-platelet interactions may be sufficient for arrest in lungs and subsequent nodule formation.

Cytoplasmic signalling potential of podoplanin
To address the question as to whether podoplanin transduces signals from outside to inside, we then focused on its small cytoplasmic domain. Previous studies indicated that podoplanin interacts with the cytoskeleton via the ERM (Ezrin, Radixin, Moesin) protein family and it is this path that controls migration (Martin-Villar et al., 2006). However, other reports described that with only 10 amino acids, the cytoplasmic domain may signal via phosphorylation of two serines by PKA and CDK5 (Krishnan et al., 2013;Krishnan et al., 2015). CDK5 (Cyclindependent kinase 5) is a cytoskeleton regulator reported to support melanoma cell motility and invasiveness (Bisht et al., 2015). In our cells, CDK5 is deregulated, increasing with podoplanin ( Fig. 6a), yet intriguingly, siRNA disruption enhanced motility (Fig. 6b). While CDK5 levels increased at the protein level with podoplanin ( Fig. 6a) this was not mirrored at the mRNA level (data not shown). This inverse relation to migration capacity may be explained by recently described functions of CDK5, phosphorylation of S171 of podoplanin inhibited migration (Krishnan et al., 2015). Thus in B16.F10 cells, accumulation beyond basal levels required for podoplanin phosphorylation may have a non-migration related function. To further evaluate the requirement for phosphorylation of podoplanin in execution of directional migration, we created podoplanin phospho-variants, transforming serine 167 into non-phosphorylatable (S167A) or constitutively active phosphomimetic (S167E) forms.
In both phospho variants, serine 171 remained phosphorylatable. Migration in response to physical stimulation was impaired in S167A variants, resembling podoplanin deficient cells (Fig. 6c). Reconstitution with S167E had an even more profound impact on directional migration ( Fig. 6c-d). Together, these data highlight that the cytoplasmic domain can indeed signal directly to the cytoskeleton to modulate migratory responses via CDK5 dependent phosphorylation. Finally, we investigated interactions between podoplanin variants and ERM proteins. As expected, increasing podoplanin levels corresponded to increased interactions with ERM proteins, which didn't happen in DCT variant due to lack of its cytoplasmic domain ( Fig. 6e). Interestingly, no differences were observed with phospho-serine variants indicating that ERM signalling is independent of phosphorylation status, instead relying on KKXXXR motifs.
We have demonstrated that podoplanin is able to influence tumour cell migration via a combination of elements: novel interactions with caveolin-1 or with ERM complexes, which transduce signals from the surface to the cytoskeleton. This is mainly accomplished through a balance of phosphorylation but also via PAK1-ERK1/2-ERM pathways, independent of small Rho GTPases activity.

Discussion
Recent reports have demonstrated that expression of podoplanin expression in tumours correlates with poor prognosis (Martin-Villar et al., 2005;Ochoa-Alvarez et al., 2012;Wicki et al., 2006). Podoplanin induced RhoA activation and epithelial-mesenchymal transition (EMT) in MDCK canine cells (Martin-Villar et al., 2006), yet in MCF7 breast carcinoma cells, it attenuated RhoA activation, and although promoting invasion this was not via EMT (Wicki et al., 2006). Podoplanin is detected in numerous cell types including lymphatic endothelial cells, alveolar epithelial cells, fibroblastic reticular cells, podocytes and osteoblasts. Given the diversity of expression patterns and contradictory data regarding function, it is likely that podoplanin function is cell and context dependent. Moreover, many studies examining podoplanin function to date have been performed in cell lines lacking endogenous podoplanin expression (Martin-Villar et al., 2006;Wicki et al., 2006). Hence we investigated function in the context of behaviours that promote tumour progression using B16.F10 melanoma cells that endogenously express podoplanin. Podoplanin exerted no growth advantage to primary tumours either in vitro or in vivo, yet increasing levels supported significantly higher degrees of lung colonization consistent with previous studies in transfected CHO cells (Kunita et al., 2007) and MCF7 cells, which exhibited enhanced metastasis in the absence of accelerated growth. Here, podoplanin supported tumour lymphangiogenesis and metastasis to lymph nodes (Cueni et al., 2010). Superior capacity to colonize sites such as the lung is now understood to be mediated through platelets via interactions with the C-type lectin CLEC-2 (Kato et al., 2008;Suzuki-Inoue et al., 2007), which convey protective effects through release of growth factors to sustain tumour cell growth, providing physical support for cell immobilization, shielding them from high shear stresses, and limiting anti-tumour cell immune surveillance mechanisms (Labelle and Hynes, 2012;Palumbo et al., 2005). Here, podoplanin has a function in promoting metastasis due to its interaction with the extracellular environment. Our data support the view that it is the physical characteristics occurring as a consequence of surface podoplanin (i.e. platelettumour emboli formation) rather than any active signalling processes that underlie lung colonization since signalling incompetent variants still efficiently formed nodules.
Podoplanin had no influence on proliferation of B16. In agreement with other studies, podoplanin influenced migration and invasion of cells which express it. However, this was only true upon receipt of a physical stimulus, having no impact on random migration of cells in a resting state. This implies that podoplanin has the potential to act as a microenvironment sensor, translating physical cues into an actionable, migration phenotype. To migrate in response to a stimulus cells require cell polarization, the formation of membrane extensions such as lamellipodia and filopodia, translocation of the cell body, and efficient mechanisms of release at the rear of the cell. These processes are tightly regulated by spatial activation of small GTPases Rho, Rac and Cdc42. Podoplanin expression is concentrated in actin-rich microvilli and plasma membrane projections including filopodia, lamellipodia and ruffles indicating a direct role in migration phenotypes (Li et al., 2015;Navarro et al., 2008;Scholl et al., 1999;Wicki et al., 2006) and has been further correlated with cytoskeleton and extracellular matrix remodelling to induce motility of individual cells (Cueni et al., 2010;Scholl et al., 1999). Although podoplanin has been implicated in collective cell migration (Cueni et al., 2010;Wicki et al., 2006) we observed expression to induce a more mesenchymal mode of migration. While podoplanin-induced migration has been proposed to occur via recruitment of ERM proteins and downstream activation of RhoA to drive EMT-like behaviour (Acton et al., 2014;Martin-Villar et al., 2006;Scholl et al., 1999), or via the inactivation or down-regulation of RhoA, Rac1 and Cdc42, to support collective migration (Cueni et al., 2010;Wicki et al., 2006), in endogenously expressing cells the migration phenotype was independent of these small GTPases. Instead, PAK1 emerged as the driver of podoplanin-induced migration in B16.F10 melanoma. Of note, PAK1 has been shown to drive increased invasion potential through cytoplasmic reorganization via the MAPK pathway in melanoma (Ong et al., 2013) and via F-actin rearrangements (Manser et al., 1997;Staser et al., 2013) or regulation of focal adhesion dynamics in other cell types (Arias-Romero and Chernoff, 2008;Slack-Davis et al., 2003).
Together, these data imply that that podoplanin may be sufficient to act as an activator of PAK1 and therefore induce migration in tumour cells via the PAK1-ERM axis, bypassing the need for Rac1 activation.
Beyond colonization of secondary sites via platelet activation and aggregation (Kato et al., 2008;Suzuki-Inoue et al., 2007), it has been reported that pro-tumour functions of podoplanin arise following interactions with other membrane proteins such as CD44, which assisted with tethering tumour cells to ECM components (Martin-Villar et al., 2010;Tsuneki et al., 2013). In melanoma cells, podoplanin-induced migration in response to stimulus relied on its surface localization (Fernandez-Munoz et al., 2011) and physical interactions with caveolin-1 within lipid rafts, unlike in embryonic mouse alveolar epithelial cells (Barth et al., 2010). These opposing observations further highlight the context dependent nature of podoplanin behaviour. The interaction between podoplanin and caveolin-1 was fundamental to the migration capacity and phenotype of cells examined. Since both components have potential roles in formation and regulation of invadopodia, particularly relating to traffic of its constituents such as matrix metalloproteinases (Martin-Villar et al., 2015;Yamaguchi et al., 2009) and invadopodia stability (Martin-Villar et al., 2015), invadopodia turnover may be one mechanism by which the interaction is transduced. Indeed, we observed increased activity of matrix metalloproteinases in podoplanin expressing cells (Table 1 and data not shown).
Although caveolin-1 is key to vesicular transport, endocytosis and signal transduction (Martinez-Outschoorn et al., 2015), more recent pro-tumour activities including invadopodia formation, proliferation, migration and tumour forming capacity (Felicetti et al., 2009) have been reported following its activation (Lobos-Gonzalez et al., 2013;Senetta et al., 2013). As with podoplanin, its expression can be found localized to the leading edge are exposed to mechanical cues, such as stresses associated with adjacent stroma (Felicetti et al., 2009;Yoo et al., 2003). We demonstrated that directional migration responses required both caveolin-1 and podoplanin to be present at the surface since disruption to either significantly impaired the response to stimulation. However, although essential, restoration of caveolin-1 in the absence of podoplanin was not sufficient to rescue migratory phenotypes implying that podoplanin is the dominant partner driving these responses, likely directing caveolin-1 to its correct cell surface locale (Fig 4d) where any pro-tumour functions can occur (Felicetti et al., 2009;Yamaguchi et al., 2009).
Although surface expression of the podoplanin extracellular domain was sufficient for lung colonization, directional migration required both cell surface podoplanin and an intact cytoplasmic tail, indicating a requirement for signal transduction following stimulation. Within the 10 amino acid cytoplasmic domain, two potential serine phosphorylation sites (S167 and S171) exist. Their phosphorylation and subsequent downstream signalling has been described to be mediated through cooperation between PKA and CDK5 (Krishnan et al., 2013;Krishnan et al., 2015) whereby the default resting state of podoplanin is a phosphorylated to suppress migration, and de-phosphorylation results in promotion of a migration phenotype (Krishnan et al., 2013;Krishnan et al., 2015). Hence an increased migration of tumour cells may depend on the balance between the total amounts of podoplanin being expressed, but also the degree of phosphorylation and downstream substrates. In the case of B16.F10, we not only have an increase in podoplanin expression but also an increase in the expression in CDK5; yet we do not see a reduction in motility in response to stimulation as might be expected from previous studies in transfected cells (FIG   6a). While siRNA experiments where blockage of CDK5, substrate for S167, enhanced migration ability consistent with previous studies (Krishnan et al., 2013;Krishnan et al., 2015), mutation to a non-phosphorylatable form surprisingly impaired migration, though not to levels observed with constitutively phosphorylated variants (FIG 6c). Although we were not able to ascertain the phosphorylation status of S171 or PKA levels, the data implies the existence of alternate pro-migratory roles for CDK5 in these cells, so how may the CDK5 function in these cells beyond S167 phosphorylation? CDK5 is a proline-directed serine/threonine-protein kinase that has also been related to motility, invasiveness and metastatic spread in melanoma. It has been suggested that plays important roles in anchorage-independent growth, cell morphology, and in the phosphorylation of proteins required for actin reorganization, endocytosis and exocytosis (Bisht et al., 2015;Krishnan et al., 2015). In B16.F10, CDK5 may also directly modify downstream effectors of cytoskeleton, adhesion and migration including PAK1 (Strock et al., 2006) and ERK1 to modulate migration rather than inactivating podoplanin by serine phosphorylation. Here, we demonstrated that podoplanin may act through PAK1 and CDK5 to promote cell migration whereby CDK5 is sufficient to activate PAK1, independent of the Rho-GTPases, and therefore interact with the cytoskeleton via the ERM protein family. Furthermore, podoplanin- In summary, we have demonstrated that functional cooperation between cell surface podoplanin and caveolin-1 are required for directional migration and invasion of melanoma cells via CDK5, PAK1 and ERK1 mediated signal transduction.

Cloning of podoplanin variants and caveolin-1 plasmid
Insert amplifications (Table S1- Quantitative RT-PCR was performed using RT 2 SYBR Green Mastermix and run on and LC480 cycler (Roche). Post-analysis was carried out using the RT 2 Profiler PCR Data Online Analysis tool (SABiosciences).

Random migration assay
Cells were seeded in chamber slides. After attaching overnight, brightfield live imagining of randomly selected fields (100x total magnification) was performed using a Live Cell imaging microscope (AF7000, Leica Biosystems) for 24 hours. Migration analysis was performed using Volocity Software (Perkin Elmer).

Wound healing scratch assay
Cells were seeded in an 8-well chamber and left to attach overnight. A scratch was then performed with the aid of a pipette tip. Sequential brightfield images of each scratch were taken by a live-cell imaging microscope (AF7000, Leica Biosystems) for 24 hours in Adherent cells were washed with cold PBS twice then scraped directly into cold lysis buffer (20mM Tris-HCl pH 7.4, 0.2% NP-40, 150mM NaCl, 2mM sodium orthovanadate, and protease inhibitors cocktail (Roche)). 1μg of the pull-down antibody (Podoplanin or Caveolin-1) was added to 1.5 mg of total protein incubated for 3 hours at 4ºC. Protein G Plus Agarose beads (Santa Cruz Biotechnology) were added before incubation for a further hour. The precipitated conjugates attached to beads were washed three times with lysis buffer containing 1% Triton X-100. Precipitates were then removed from the beads by incubation with protein loading buffer (50mM Tris-HCl pH6.8, 2% SDS, 10% glycerol, 1% betamercaptoethanol, 12.5mM EDTA, 0.02% bromophenol blue) for 5 minutes and boiling for 10 minutes. Samples were analysed by western blot (Table S3).

Membrane raft extraction
Lipid raft extraction was performed by gradient ultracentrifugation as previously described by George et al. (George et al., 2010).

Immunoblotting
Protein samples were separated by 12% SDS-PAGE and transferred onto a PVDF membrane (Millipore). The membranes were blocked with 5% BSA in TBS containing 0.5% Tween 20 and then incubated with primary antibodies overnight at 4ºC (Table S3). Washed membranes were incubated with the appropriate HRP-conjugated secondary antibody prior to detection with an enhanced chemiluminescence detection kit (Pierce).

In vivo studies
Experiments involving animals were performed in accordance with UK Home Office regulations under HO license PPL 80/2574. Where possible, experimental groups were randomized and blinded. Lung colonization capacity was determined by intravenous injection of 150,000 B16.F10 cells into 7-week old female C57BL/6 mice (Envigo). Animals were sacrificed 21 days later and lungs analysed for nodules and tumour burden. Samples were processed for histology. For primary tumour studies, 250,000 cells were implanted subcutaneously on both shoulders of female C57BL/6 mice and growth was monitored for 9 days or until tumour sizes reached the legal limit of a diameter of 12 mm. Tumour volumes were calculated by: Volume=π/6×small 2 ×longest.

Haematoxylin and eosin stain
Harvested lungs were embedded in OCT medium (TissueTek). 10μm sections were fixed in 4% para-formaldehyde. Hematoxylin and eosin stain was performed using a Leica Autostainer XL. Slides were mounted with a Leica CV5030 and imaged using AxioScan.Z1