Nuclear-capture of endosomes drives depletion of nuclear G-actin to promote SRF/MRTF gene expression and cancer cell invasiveness


 Signals are relayed from receptor tyrosine kinases (RTKs) at the cell surface to effector systems in the cytoplasm and nucleus, and coordination of this process is important for the execution of migratory phenotypes, such as cell scattering and invasion. The endosomal system influences how RTK signalling is coded, but the ways in which it transmits these signals to the nucleus to influence gene expression are not yet clear. Here we show that an RTK, cMET promotes Rab17-dependent endocytosis of EphA2, another RTK, followed by centripetal transport of EphA2-positive endosomes. EphA2 then mediates physical capture of endosomes on the outer surface of the nucleus; a process involving interaction between the nuclear import machinery and a nuclear localisation sequence in EphA2’s cytodomain. Nuclear capture of EphA2 promotes RhoG-dependent phosphorylation of the actin-binding protein, cofilin to oppose nuclear import of G-actin. The resulting depletion of nuclear G-actin drives transcription of Myocardin-related transcription factor (MRTF)/serum-response factor (SRF)-target genes to implement cell scattering and the invasive behaviour of cancer cells.

Many plasma membrane receptors, including RTKs, are endocytosed following ligand engagement and trafficked through the endosomal system, and the way in which this occurs is thought to influence signalling outcomes 1 . An extensively studied example is the signal termination after endocytosis where ubiquitinated receptors are sorted for lysosomal degradation 2 . More recently, an increasingly complex picture is emerging in which certain endosomal compartments and membrane subdomains, including late and recycling endosomes, constitute platforms which influence downstream signalling coding 3 . Many of these endosomal compartments are positioned very close to the nucleus, and this is thought to facilitate the communication of signals to the transcriptional machinery. For example, the Rab5 effector, APPL translocates into the nucleus to drive transcription and cell proliferation, and delivery of cMET to perinuclear endosomes is necessary to promote nuclear accumulation of the STAT3 transcription factor 4,5 . There are numerous reports of RTKs, and the intracellular domains of RTKs, being imported into the nucleus to perform signalling roles 6 . However, mechanisms accounting for this have been unclear until, more recently, EGFR1 was shown to be delivered to the nucleoplasm via docking and fusion of endosomes with the nuclear membrane 7,8 . Indeed, these investigators provide evidence for physical association of endosomes with the nucleus as a prelude to nuclear import of an RTK and thus provide a rationale for how the endosomal system may directly interface with gene expression. These studies prompted us to investigate whether endosomes associate with the nucleus following activation of HGF signalling with a view to determining whether this may contribute to transcriptional landscapes which favour HGF-driven cell migratory behaviour. Here we report that activation of an RTK, cMET by HGF promotes endocytosis of EphA2, another RTK followed by capture of endosomes upon the nuclear surface, in an interaction mediated by a nuclear-localisation signal in the cytoplasmic domain of EphA2. This event promotes both actin polymerisation in the juxtanuclear region and phosphorylation of the actin monomer-binding protein, cofilin. We have then proceeded to use a combination of cell biological and mathematical modelling approaches to determine that it is the ability of nuclear-captured EphA2 to phosphorylate cofilin which leads to depletion of nuclear G-actin which, in turn, drives changes in gene expression to enable cell migration and invasion.

EphA2-mediates capture of endosomes at the nuclear surface
To study the influence of cMET signalling on the intracellular destination of endosomes, we surface-labelled H1299 cells with a cell-impermeant biotinylation reagent, incubated the cells in the present and absence of HGF, and studied the distribution of biotinylated internalised material using fluorescence microscopy. This revealed a population of endosomes transporting plasma membrane-derived material into very close proximity of the nuclear membrane, and this population was significantly increased by HGF addition (Fig. 1a).
Moreover, transmission electron microscopy (TEM) of nuclei purified from HGF-treated cells indicated endosomes containing plasma membrane-derived cargo docked with the nuclear membrane in an interaction which appears to be direct and not to rely on intervening structures, such as ER of Golgi membranes ( Supplementary Fig. 1a). These observations prompted us to develop a stable isotope labelling (SILAC) proteomic approach to identify cargoes that traffic from the plasma membrane to endosomes that are physically associated with the nucleus. H1299 cells were SILAC-labelled with medium (M) or light (L) amino acids, surface-biotinylated and incubated in the presence or absence of HGF. Following this, nuclei were purified (together with their associated endosomes), biotinylated proteins isolated from these using streptavidin beads and the proteome of isolates analysed by mass spectrometry (Fig. 1b). This indicated that HGF addition promoted translocation of a number of receptors from the plasma membrane to the nucleus, and another RTK, EphA2 was prominent amongst these ( Fig. 1b; supplementary spreadsheet 1). Surface biotinylation/capture-ELISA-based approaches confirmed that surface-labelled EphA2 relocated to a compartment which copurified with the nucleus within a few minutes of HGF addition, and super-resolution Airyscan microscopy confirmed that these EphA2-positive vesicles were attached to the nuclear surface ( Fig. 1c). Furthermore, siRNA of EphA2 opposed recruitment of plasma membrane-derived vesicles to the nuclear surface indicating that EphA2 is not a passive cargo of this pathway, but that it is necessary for their delivery to and/or attachment to the nucleus (Fig. 1d). To determine how HGF drives packaging of EphA2 into endosomes, we tested the involvement of a battery of Rab GTPases and other endosomal regulators known to be involved in cell adhesion and migration (including Rabs 4,5,7,11,17,21 and 25,CD63 and LAMP1/2) in EphA2 internalisation and trafficking. This highlighted Rab17 as being required for HGF-driven internalisation ( Fig. 1e) and nuclear-capture (Fig. 1f) of EphA2, and live cell fluorescence imaging indicated that Rab17 accompanied EphA2 on its centripetal journey from the plasma membrane to vesicles captured at the nuclear surface ( Fig. 1f, g, Supplementary Fig. 1c, Nuclear location signals (NLS) are found in proteins that translocate to the nucleus and are recognised by α-importins which recruit β-importins to promote association with the nuclear pore 9 . A bioinformatic analysis 10 predicted a putative NLS in the juxtamembrane region of EphA2 and we generated two mutants of this (NLS1 and NLS2) designed to disrupt key basic residues responsible for α-importin binding (Fig. 1h). Fusion of EphA2's NLS, but not NLS1 or NLS2, with GFP increased nuclear delivery of GFP, demonstrating that this sequence can mediate functional interaction with the nuclear import machinery ( Supplementary Fig. 1d).
Consistently, following HGF addition, EphA2 coimmunoprecipitated with importins -α5 and -β1 (Fig. 1i), and fluorescence time-lapse microscopy indicating that EphA2-positive vesicles associated with these components of the nuclear import machinery near the nuclear surface ( Supplementary Fig. 1e). Importantly, mutation of EphA2's NLS opposed coimmunoprecipitation of EphA2 with importins ( Fig. 1i) and reduced nuclear capture of EphA2 vesicles (Fig. 1j), without compromising canonical signalling downstream of EphA2 ( Supplementary Fig. 1f). Taken together, these data indicate that addition of HGF promotes Rab17-dependent internalisation and centripetal trafficking of a population of endosomes which are then captured on the nuclear surface by an NLS in the cytotail of EphA2.

Nuclear-capture of endosome is required for HGF-driven cell scattering and cancer cell invasion
In the 'KPC' autochthonous mouse model of pancreatic ductal adenocarcinoma (PDAC), knockout of EphA2 opposes invasive migration and metastasis 11 . Invasive migration of EphA2 knockout PDAC cells towards a gradient of HGF was restored by re-expression of wild-type EphA2 and, interestingly, EphA2s with mutated NLSs are ineffective in this regard (Fig. 2a).
Consistently, knockout or siRNA of EphA2 opposed scattering of H1299 and PDAC cells respectively and this was restored by re-expression of wild-type, but not NLS mutants of, EphA2 ( Fig. 2b-d). Moreover, siRNA of Rab17 (to reduce internalisation of EphA2) opposed HGF-driven cell scattering (Fig. 2d) indicating that internalisation, centripetal transport and NLS-mediated nuclear capture of EphA2-positive endosomes is required for cells to mount a scattering response following addition of HGF.

EphA2-mediated nuclear-capture influences expression of MRTF/SRF target genes
We proposed that interaction of endosomes with the nuclear may influence invasive behaviour by controlling gene expression. We used RNAseq to profile gene expression signatures driven by HGF and to determine whether components of this are dependent on EphA2. This indicated that a number of HGF-responsive mRNAs were sensitive to siRNA of EphA2 (  3b) and HGF-driven cell scattering ( Supplementary Fig. 3c). We next used DNase I in combination with fluorescence microscopy and flow cytometry to visualise G-actin, and found that addition of HGF significantly decreased nuclear, but not cytoplasmic, G-actin (Fig. 3c,dc).. Importantly, this was opposed by siRNA of EphA2 and rescued by wild-type, but not an NLS mutant of, EphA2 (Fig. 3c). These data indicate that nuclear capture of endosomes allows HGF signalling to drive depletion of nuclear (but not cytosolic) G-actin and to promote actin polymerisation in the juxtanuclear cytosol.
G-actin influences nucleocytoplasmic shuttling 16 and transcriptional activity 17 of MRTF/SRF, so we used high content imaging to measure nuclear shuttling of MRTF and the role that EphA2, RhoG and G-actin play in this. MRTF translocated to the nucleus following HGF addition. However, this was not opposed by knockdown of either EphA2 or RhoG (Fig. 3d).
Moreover, by deploying non-polymerisable actins -actin R62D and the nuclear targeted, NLSactin R62D actin 18 -we found that nuclear shuttling of MRTF was only minimally affected by levels of actin monomer in the cytoplasm and nucleus respectively (Fig. 3d). Despite this, expression of actin R62D profoundly inhibited MRTF-target gene expression ( Supplementary   Fig. 3b) and HGF-driven cell scattering (Fig. 3e). These data indicate that, although G-actin may make a contribution to nuclear translocation of MRTF, it is the levels of G-actin in the nucleus that ultimately determine its transcriptional activity in response to HGF. Consistently, when we elevated nuclear G-actin levels by knocking-down the exportin-6 transporter (XPO6) ( Supplementary Fig. 3e), which mediates efflux of G-actin from the nucleus 19 , HGF-driven transcription of MRTF-target genes and cell scattering were inhibited ( Supplementary Fig. 3f, g).

Nuclear-capture promotes depletion of nuclear G-actin via phosphorylation of cofilin
The rate of juxta-nuclear actin polymerisation (denoted by E c µ1 in Fig. 4a), which we have found to be nuclear-capture and RhoG-dependent, is likely to affect availability of G-actin for nuclear import. Indeed, a recently-described pathway linking mechanical force to gene expression in the skin involves tension-induced F-actin polymerisation in the immediate vicinity of the nuclear membrane, and this restricts availability of G-actin in the nucleus to oppose transcription in a chromatin-dependent manner, but may also influence SRF/MRTF target gene expression 20 . Another event that might influence nucleo-cytoplasmic actin dynamics is phosphorylation of cofilin at Ser 3 (by LIMK) (denoted by E c µ2 in Fig. 4a) 21 , as this opposes formation of the cofilin-actin complex which is the species transported into the nucleus via importin-9 22 . Addition of HGF increased levels of phospho-cofilin, and this was opposed by siRNA of EphA2 or RhoG and rescued by expression of wild-type (but not NLS mutants of) EphA2, indicating that cofilin phosphorylation is controlled by nuclear-capture and thus may influence communication with the transcriptional machinery. We designed a computation approach for modelling nucleo-cytoplasmic actin dynamics, and thus test the contribution of signalling events occurring downstream of EphA2 and RhoG to depletion of nuclear G-actin. This model clearly predicted that when juxta-nuclear actin polymerisation (E c µ1) and cofilin phosphorylation (E c µ2) (Fig. 4a) are both active this drives rapid depletion of nuclear G-actin and modestly increased cytosolic G-actin, thus recapitulating the actin dynamics observed experimentally following HGF addition (Fig. 4c). We then tested the consequences of independently reducing E c µ1 and E c µ2 to zeroi.e. mimicking selective inhibition of cytosolic actin polymerisation and cofilin phosphorylation respectively. This indicated that cofilin phosphorylation (E c µ2) was likely to exert a greater influence over the depletion of nuclear G-actin depletion than actin polymerisation (E c µ1) (Fig. 4c). This prediction was confirmed experimentally, as treatment with the Arp2/3 inhibitor, CK-666 23 completely opposed juxta-nuclear actin polymerisation without opposing expression of MRTFtarget genes ( Supplementary Fig. 4a, b). Conversely, addition of the LIMK inhibitor, BMS-5 24 to oppose cofilin phosphorylation, completely opposed the ability of HGF to drive both expression of MRTF target genes and cell scattering (Fig. 4d, e).
To conclude, this study describes a new paradigm for transmission of signals between the plasma membrane and the nucleus in which one RTK (cMET) promotes endocytosis of another (EphA2) whose function is to capture endosomes at the nuclear surface. This nuclear captured EphA2 leads to both local actin polymerization and phosphorylation of cofilin.
However, experimental observations and mathematical modelling have allowed us to conclude that it is the phosphorylation of cofilin leadingto depletion of nuclear G-actin which provides the main impetus to SRF/MRTF-dependent gene expression leading, in turn to cell scattering and invasive responses (Fig. 4e). This mechanism of signal transduction provides opportunities for coordination and integration of inputs from other pathways and therefore, we anticipate that nuclear-captured endocytic compartments will function as a key signalling nexus for a range of cellular processes in a variety of cells types.

Cell culture, transfection, constructs and siRNA
H1299 cells were obtained from ATCC. The genetic identity of all these cell lines has been confirmed at the CRUK Beatson Institute for Cancer Research. Cell lines were cultured at 37 °C and 10% CO2 in a humidified incubator. H1299 cells and in-house PDAC cells were cultured in DMEM. All media were supplemented with 10% foetal calf serum, 2 mM Lglutamine, 100 IU/ml penicillin, 100 μg/ml streptomycin and 250 μg ml −1 fungizone. For expression vectors, cells were transfected using lipofectamine 2000 (Thermo Fisher), and for siRNAs transfection was performed using the Nucleofector system (kit V; Lonza).

Antibodies and immunoprecipitation
For Western blotting and immunofluorescence, antibodies were from the following sources:

qRT-PCR
Trizol (Ambion) was used to isolate total RNA from the relevant cell lines following the manufacturer protocol. The cDNA was obtained by using the Quantitect reverse transcription kit (Qiagen). qRT-PCR reactions were prepared using the SYBR Green kit (QuantaBio). The amplified products were obtained and analysed by a CFX96 qPCR System (BioRad). ΔΔC(t) was determined using GAPDH as a reference. Control transfected transcript levels were assigned the arbitrary value of 1. FosB, JunB, ATF3 and GAPDH human Quantitect primers were purchased from Qiagen and the following primers manufactured by Invitrogen (Thermo Unbound material was removed by extensive washing with PBS-T and wells were incubated with streptavidin-conjugated horseradish peroxidase (Vector Laboratories) in PBS-T containing 1 % BSA for 1 hr at 4ºC. Following further washing, biotinylated integrins were detected by chromogenic reaction with ortho-phenylenediamine as described previously 25 .
Nuclear-capture: Cells were collected in PBS, centrifuged for 10 s in a table top centrifuge and the supernatant discarded. Cells were resuspended in PBS containing 0.1% NP-40 (PBS-N) and passed through a 26-gauge needle 5 times, centrifuged for 10s and resuspended in PBS-N. This procedure was then repeated. The pellet was finally resuspended in lysis buffer containing 1% Triton X-100 and 1% NP-40, sonicated for three rounds of 20 s/round. Finally, samples were centrifuged at 8000 rpm in a table top centrifuge for 30 s and protein content was measured by using Optiblot (Thermo Fisher) following the manufacturer's protocol. The levels of nuclear-captured biotinylated-EphA2 were determined by capture-ELISA as described above using ELISA plates coated with either mouse anti-EphA2 or mouse anti-GFP (Abcam) antibodies. To image purified nuclei by immunofluorescence, nuclei were seeded onto glass bottom dishes previously coated with poly-D-lysine, fixed with 4% paraformaldehyde and stained with various antibodies.
To image internalisation of biotinylated proteins, the same cell-surface biotinylation and MesNa reduction procedure as described above was used followed by fixation in 4% paraformaldehyde. Biotinylated proteins were visualised with Alexa488-conjugated streptavidin (Vector). Images were acquired using a Zeiss LSM 880 Airyscan confocal microscope, and processed using Zen Black Zeiss software. To quantify distances between vesicles and the nucleus in ImageJ software a mask for the DAPI staining was created and a distance map obtained from the boundaries of the nucleus. Internalised particles were selected and the distance to the nucleus was analysed by overlaying them with the distance map.

Cell scattering and invasion
Cell scattering: H1299 cells were seeded onto six-well plates for 48 h, during which time the cells formed small colonies. Cells were then visualized using a Nikon time lapse microscope in the presence and absence of HGF (10 ng ml −1) , with or without CK-666 (XXX) or BMS-5 (XXX). Images were collected every 5 min from six different regions per well. To track scattering, ImageJ manual tracking and chemotaxis plugins were used.
Invasion: Inverted Matrigel assays were performed as described previously 26 . Briefly, Matrigel was allowed to polymerize in Transwell inserts (Corning) for 1 h at 37°C. Inserts were then inverted and cells seeded directly onto the upper face of the filter. Cells were placed in a chemotactic gradient of medium supplemented with 10% FCS and 10 ng/ml HGF and, 48 h after seeding, migrating cells were stained with Calcein-AM and visualized by confocal microscopy with serial optical sections being captured at 15-μm intervals.

High-content image analysis
Cell scattering: PDAC cells were seeded onto glass six-well plates. After 24 h cells were fixed and stained with DAPI (Sigma) and Cell Mask (1:10,000, Invitrogen). Cells were imaged at 10X magnification using the Opera Phenix High-Content Screening System (PerkinElmer) and cell distribution was quantified using Columbus Image Data Storage and Analysis System (PerkinElmer). A cell cluster was defined as five or more nuclei <1μm from each other with an area >600μm 2 .
Nucleocytoplasmic distribution of MRTF-GFP: H1299 cells stably expressing an MRTF-GFP construct (gift from Drs. Christian Baarlink and Robert Grosse, University of Marburg) were plated in 6-well dishes and treated in the presence or absence of HGF. Cells were processed and imaged as described before. MRTF localisation was quantified by using DAPI as a nuclear mask and DAPI-negative Cell Mask staining for the cytoplasmic mask.

RNAseq
RNA was extracted as described in the qRT-PCR methods. Quality of the purified RNA was tested on a 2200 Tapestation using RNA screentape (Agilent). Libraries for cluster generation and DNA sequencing were prepared using the TruSeq Stranded mRNA LT Kit (Illuma). Quality and quantity of the DNA libraries was assessed on the 2200 Tapestation (D1000 screentape) and Qubit (Thermo Fisher Scientific) respectively. The libraries were run on the Next Seq 500 Light-labelled cells were warmed to 37°C for 30 min to allow internalisation of labelled protein in the presence of HGF (10 ng/ml), whilst the heavy labelled cells were warmed to 37 ºC in the absence of HGF. Biotin remaining at the cell surface was removed by surface reduction as above. Cells were washed in PBS and nuclei were purified as described above. The supernatants from the heavy and light SILAC labelled cells were then mixed, and biotinylated proteins isolated using streptavidin-conjugated agarose beads as described above.

Transmission electron microscopy
Purified nuclei were fixed for 2 hours in and ice-cold buffer containing 2.5% glutaraldehyde, 4% paraformaldehyde and 0.1M sodium cacodylate pH7.4. After fixation, nuclei were rinsed with a buffer containing 0.1M sodium cacodylate and 2% sucrose. To visualise internalised biotin, streptavidin-conjugated nanogold beads (10nm InnovaCoat® GOLD 10OD Streptavidin Conjugate -Expedeon) were used to stain the samples. Resin embedding, ultra-thin sectioning and staining for transmission electron microscopy were performed by the Electron Microscopy service (University of Glasgow). Images were obtained using a JEOL 1200 TEM -80kv electron microscope.

Quantification of nuclear G-actin
Purified nuclei were resuspended in PBS, fixed in 4% paraformaldehyde for 20min and permeabilised with 0.2% Triton X-100 in PBS for 10 min at 37 ºC. Nuclei were stained with Alexa594-conjugated DNAseI (0.3μM) in combination with DAPI for 1h 30min at 37°C under continuous shaking, followed by resuspension in PBS containing 1mM EDTA and 1mM MgCl 2 .
Data were acquired in LSRFortessa flow cytometer and analysed with the FlowJo software.
Geometrical mean value corresponding to the DAPI positive population was calculated, and data were expressed as the ratio of the geometrical mean value of HGF treated cells versus vehicle treated cells.

Computational model to predict nuclear G-actin dynamics
We model the cell as two static circular compartments, Ω and Ω , which represent the cytoplasm and the nucleus, respectively, for some points 1 and 2 , and some radii 1 and 2 . We also let Γ and Γ denote the external boundary of the cytoplasm and the nucleus:cytoplasm boundary respectively, A graphical illustration of the simulation domain is provided ( Supplementary Fig. 5).
It is assumed that the cytoplasm contains a population of G-actin, ( , ), which diffuses and polymerises to form F-actin, ( , ), at a rate of . In the presence of a catalytic growth factor, ( , ), ( , ) polymerises to form ( , ) at a rate of (1 + 1 ). ( , ) depolymerises at a rate of to form ( , ). We also assume that the cytoplasm contains a population of cofilin, ( , ), which associates with ( , ) at a rate of Ξ to form cofil-actin, Ξ ( , ), and dissociates from ( , ) at a rate of Ξ . Additionally, ( , ) is assumed to phosphorylate at a rate of Θ (1 + 2 ), giving phospho-cofilin, Θ ( , ), which dephosphorylates at a rate of Θ . Finally, we assume that the cytoplasm contains a population of profilin, ( , ), which associates with ( , ) at a rate of Υ to form profil-actin, Υ ( , ), and dissociates from Υ ( , ) at a rate of Υ . The governing equations in the cytoplasm are therefore for ∈ , where [⋅] denotes the diffusion coefficient of species [⋅].
In the nucleus, we assume that there exists a population of nuclear G-actin, ( , ), and a population of cofilin, ( , ), which associate with each other at a rate of Ξ to form nuclear cofil-actin, Ξ ( , ). Nuclear cofil-actin dissociates at a rate of Ξ to give ( , ) and ( , ).
Finally, we assume that the nucleus contains a population of profilin, ( , ), which associates with ( , ) at a rate of Υ to form nuclear profil-actin, Υ ( , ), and dissociates from Υ ( , ) at a rate of Υ . The governing equations in the nucleus are therefore for ∈ .
At the external boundary of the cytoplasm, each species obeys the no-flux condition where denotes the outwards-facing unit normal vector. At the internal boundary between the cytoplasm and the nucleus, we impose the boundary conditions for ∈ , where [⋅] denotes the transfer rate between the cytoplasm and the nucleus for It was assumed that F-actin and the catalytic growth factor are both non-diffusing, such that = = 0. The remaining diffusion coefficients were chosen such that [⋅] ≫ [⋅] to ensure that the concentration profile of each species remains reasonably 'flat' within the nucleus (relative to the cytoplasm). We also made the assumption that, once introduced, the total amount catalytic growth factor is constant over time, such that = 0. In the model, the primary purpose of cofilin is to transport G-actin from the cytoplasm to the nucleus; we therefore set Ξ > (i.e. cofilin and G-actin associate at a faster rate in the cytoplasm than in the nucleus) and Ξ < (i.e. cofilin-actin disassociates at a faster rate in the nucleus than in the cytoplasm), so that cofilin-actin quickly dissociates upon entering the nucleus. Since the profilin species is used to transport G-actin from the nucleus to the cytoplasm, we set Υ > (i.e. profilin and G-actin associate at a faster rate in the nucleus than in the cytoplasm) and Υ < (i.e. profil-actin disassociates at a faster rate in the cytoplasm than in the nucleus), so that profilin-actin quickly dissociates upon entering the cytoplasm. To ensure that the equilibrium concentration of cofilin is approximately the same on either side of the nuclear boundary, we set = ; similarly, we set = so that the equilibrium concentration of profilin is approximately the same on either side of the nuclear boundary. Wet lab experiments have revealed that, in the absence of a catalytic growth factor, the concentration of G-actin in the nucleus is typically higher than the concentration of G-actin in the cytoplasm; given sufficiently large values for and , this implies that > . Finally, the first catalytic rate parameter, 1 , was chosen to ensure that G-actin is polymerised at a faster rate in the presence of a catalytic growth factor (as observed experimentally), whilst the second catalytic rate parameter, 2 , was chosen to ensure that cofilin is rapidly phosphorylated in the presence of a catalytic growth factor (which should have the effect of reducing the amount of cofilin that is available to transport G-actin from the cytoplasm to the nucleus).

Rate of dissociation for profilin-actin in the nucleus
Nuclear import rate of profilin 1 Nuclear export rate of profilin 1 Nuclear export rate of profilin-actin 1 Nuclear import rate of cofilin 1 Nuclear export rate of cofilin 1 Nuclear import rate of cofilin-actin To determine appropriate initial conditions, the model system was simulated using the default parameters outlined above (with ( , ) = 0, ∀ ∈ ) until it had approached equilibrium, and then the average concentration of each species (across its associated domain) was used as a spatially constant initial value in subsequent simulations. The computed initial value for each species is as follows:

Symbol Value
Initial concentration of G-actin in the cytoplasm Note that the equilibrium concentration profile for each species is not necessarily flat, so our simulations (which use a spatially constant value as the initial condition for each species) do not start at steady state. Finally, the initial condition for the catalytic growth factor was chosen to be were not fractionated and these were used to determine the surface proteome and proteins which are present in the surface proteome and which have, therefore, moved from the plasma membrane to the nuclear fraction are highlighted by red dots.

Figure legends
c, Surface-labelled H1299 cells were warmed to 37°C in the absence or presence of HGF for the indicated times to allow internalisation. Nuclei were purified and the presence of biotinylated EphA2 in these purified nuclei was determined using a capture-ELISA (left graph) and by immunoflourescence followed by super-resolution microscopy (right panels, Bar 5 μm).     f, cMET drives Rab17-dependent endocytosis of EphA2 [1]. EphA2 endosomes are then transported centripetally under control of Rab17 [2] and become physically attached to, or 'captured' by, the nucleus by an interaction formed between the nuclear import machinery and a nuclear localisation sequence located in EphA2's cytodomain [3]. This nuclear-capture event, in turn, drives actin polymerisation which is restricted to the juxtanuclear region [4] and LIMK-driven phosphorylation of cofilin [5], and both events are dependent on the RhoG GTPase. Phosphorylation of cofilin opposes nuclear import of cofilin-actin, leading to depletion of G-actin from the nucleus [6] which, in turn, activates transcription of MRTF/SRF-target genes [7] to implement cell scattering and invasion.