N-terminus-independent activation of c-Src via binding to a tetraspan(in) TM4SF5 in hepatocellular carcinoma is abolished by the TM4SF5 C-terminal peptide application.

Active c-Src non-receptor tyrosine kinase localizes to the plasma membrane via N-terminal lipid modification. Membranous c-Src causes cancer initiation and progression. Even though transmembrane 4 L six family member 5 (TM4SF5), a tetraspan(in), can be involved in this mechanism, the molecular and structural influence of TM4SF5 on c-Src remains unknown.
Methods: Here, we investigated molecular and structural details by which TM4SF5 regulated c-Src devoid of its N-terminus and how cell-penetrating peptides were able to interrupt c-Src activation via interference of c-Src-TM4SF5 interaction in hepatocellular carcinoma models.
Results: The TM4SF5 C-terminus efficiently bound the c-Src SH1 kinase domain, efficiently to the inactively-closed form. The complex involved protein tyrosine phosphatase 1B able to dephosphorylate Tyr530. The c-Src SH1 domain alone, even in a closed form, bound TM4SF5 to cause c-Src Tyr419 and FAK Y861 phosphorylation. Homology modeling and molecular dynamics simulation studies predicted the directly interfacing residues, which were further validated by mutational studies. Cell penetration of TM4SF5 C-terminal peptides blocked the interaction of TM4SF5 with c-Src and prevented c-Src-dependent tumor initiation and progression in vivo. Conclusions: Collectively, these data demonstrate that binding of the TM4SF5 C-terminus to the kinase domain of inactive c-Src leads to its activation. Because this binding can be abolished by cell-penetrating peptides containing the TM4SF5 C-terminus, targeting this direct interaction may be an effective strategy for developing therapeutics that block the development and progression of hepatocellular carcinoma. © The author(s).

Introduction

The c-Src proto-oncogene, non-receptor tyrosine kinase is overexpressed and activated in a diversity of cancer types 1. c-Src plays key roles in the regulation of cellular growth, division, adhesion, and motility 2. Thus, exploring how and which signaling molecules can affect c-Src activity may lead to the development of therapeutic agents against c-Src-mediated cellular functions 3. c-Src consists of the N-terminal myristoyl group attached to the Src homology (SH) 4 domain, SH3 domain, SH2 domain, proline-rich SH2 kinase linker region, and the SH1 or kinase domain 4. Regulation of c-Src occurs through inhibitory intramolecular interactions between the SH3 domain and the polyproline II helix formed by the SH2-kinase linker, as well as between the SH2 domain and Tyr530 in the C-terminal regulatory tail 2. Activation of c-Src is achieved when the SH3 domain is displaced from the SH2 linker by interacting with other proteins 5. Dephosphorylation of Tyr530 opens the inhibitory, or closed, conformation of c-Src, thereby activating the tyrosine kinase activity 6. Phosphorylation of Tyr419 in the activation loop essentially locks the kinase into its catalysis-ready conformation 7. Once activated further, c-Src translocates from the perinuclear regions to the plasma membrane via RhoA family-mediated actin organization 8. N-terminal myristoylation of c-Src can regulate trafficking to the cellular periphery, which determines its activity 9. Although many studies have focused on c-Src activation to reveal the mechanism of how the inhibitory intramolecular interactions occur, it remains unclear how diverse molecules, such as upstream effectors, regulate the conformation and activity of c-Src.

Previous studies have identified a few binding partners for c-Src, including focal adhesion kinase (FAK) 10 and platelet-derived growth factor receptor (PDGFR) 11, which bind SH2; negative regulatory factor (Nef) and sex-lethal interactor (Sin), which bind SH3 5; and p130Cas (Crk-associated substrate), which binds SH2 and SH3 12. These interactions with the SH2 and SH3 domains have been established to lead to c-Src activation. However, the protein that binds to pY530 in c-Src to positively regulate its activity regulation has not been known. Meanwhile, there are c-Src binders that inactivate c-Src, such as DOC-2/DAB2 (differentially expressed in ovarian carcinoma-2/disabled-2) 13, C-terminal Src kinase (CSK) 14, E3 ubiquitin‑ligase Cbl 15, 16, and Cullin‑5 17. In addition, c-Src activity was negatively regulated by Csk-binding protein (Cbp) in NIH3T3 cells 18, by cooperative roles of Cbp and Caveolin-1 19, and by Connexin43 in C6 glioma cells 20. The transactivator of transcription (TAT) cell-penetrating sequence fused to Connexin43 (TAT-Cx43266-283) has been shown to inhibit c-Src activity 21. The SH3 domain in c-Src interacts with the first proline-rich domain (amino acid 619- 627) of DOC-2/DAB2, leading to inhibition of c-Src Tyr416 phosphorylation. The edge of the active site of CSK interacts and phosphorylates Tyr530 on the C-terminal tails of c-Src. A binding partner for the c-Src SH1 domain and the C-terminal regulatory segment with Tyr530 may indeed affect c-Src conformation and activity.

Transmembrane 4 L six family member 5 (TM4SF5) is involved in tumorigenesis and tumor progression 22. TM4SF5 is a membrane glycoprotein with four transmembrane domains, two extracellular loops (SEL for short and LEL for long extracellular loop), an intracellular loop (ICL), as well as N-terminal and C-terminal tails that are located in the cytosol 23. Whereas the ICL of TM4SF5 binds the F1 lobe of the four-point-one, ezrin, radixin, moesin (FERM) domain in FAK for cell adhesion-dependent directional migration 24, the C-terminus of TM4SF5 supports c-Src-dependent invasive protrusions 25. However, it remains largely unknown how TM4SF5 regulates c-Src at the molecular and structural levels.

Here, we examined the mechanistic and biological significance of the interaction between the membrane protein TM4SF5 and c-Src. We found that the C-terminus of TM4SF5 bound to the SH1 domain of closed, inactive c-Src more strongly than open, active c-Src. Protein tyrosine phosphatase 1B (PTP1B) was additionally recruited for dephosphorylation at c-Src Tyr530. Thus, the TM4SF5 C-terminus activated the N-terminus-deficient c-Src by binding to its SH1 kinase domain. Computational simulation was performed to further explore the fundamental principles of how their structural organization and movements affect their binding to each other and ultimately govern their functions. Further, a unique interaction between the c-Src SH1 domain and TM4SF5 C-terminus was targeted by cell-penetrating peptides (CPPs) containing the TM4SF5 C-terminus during carcinogenesis and progression.

Methods

Cells: Human hepatocellular carcinoma cells expressing TM4SF5-null control (SNU398, SNU449, SNU761 parental, SNU449Cp [a pooled control clone], and SNU761-mock), TM4SF5WT (NM-003963, 197 amino acids), SNU449Tp or SNU449T7 [a pooled or single cell-driven clone with ectopic TM4SF5 expression, respectively], SNU761-TM4SF5WT or mutant SNU761-TM4SF5ΔICL19 [a mutant in which the intracellular domain from amino acids 71 to 89 is deleted], and TM4SF5ΔC [a mutant in which the C-terminus from amino acids 187 to 197 is deleted]) have been described previously 25. PLC/PRF/5 hepatoma cell line endogenously expressing TM4SF5 was purchased from Korea Cell Bank (Seoul National University, Seoul, Korea). HEK293FT cells and endogenous TM4SF5-expressing Huh7 and HepG2 cells were maintained in DMEM-H (WelGene, Daegu, Republic of Korea) containing 10% FBS and 1% penicillin/streptomycin (GenDEPOT, Barker, TX, USA) at 37°C in 5% CO2. Cells were routinely monitored for mycoplasma contamination.

Peptides and DNA mutagenesis: CPPs for TAT-Cscram (TCsr, 2703.1 Da), TAT-Cter (TC, 2703.1 Da), TAT-Caax-Cter (TcxC, 3149.8 Da), Antp-Caax-Cscram (3960.4 Da), or Antp-Caax-Cter (3960.4 Da) were synthesized and purchased (Anygen, Gwangju, Republic of Korea). The sequences of TAT and Antennapedia (Antp) are RKKRRQRRRP and RQIKIWFQNRRMKWKK, respectively, and the sequence of the TM4SF5 C-terminus is 187GDCRKKQDTPH197. The sequence of the Caax motif is CVIM, and it was conjugated between the TAT and TM4SF5 C-terminal sequences to avoid structural interference of the short TM4SF5 C-terminus, presumably without any secondary structure, although the Caax sequence is conjugated to the end of K-Ras for the cell-penetrating effects 26. Peptides were maintained at 1 mM for stock solutions in PBS and then diluted into culture media containing 10% FBS prior to treating the cells at different concentrations or conditions. c-Src point mutations (Y419F, Y530F, Y419F/Y530F, K298M, K298M/Y530F, K391, Q531E/P532E/G533I, or Y419F/Q531E/P532E/G533I, D454A, R460A, D454A/G459P) were engineered with pfu polymerase (Agilent-Stratagene, Santa Clara, CA, USA) and confirmed by direct sequence analyses. TM4SF5 mutants were also generated and confirmed by direct sequence analysis for C118A/C145A, D188A, C189A, and R190A. The ΔICL19 TM4SF5 mutation was described previously 23.

Transfection of plasmids: SNU449 or SNU761 parental or stable cells transfected with mock or TM4SF5-expressing plasmids were further transiently transfected using Lipofectamine PlusTM (Thermo Fisher Scientific, Waltham, MA, USA) or electroporated with different plasmids for 48 h before whole cell lysates were collected. The plasmids included mock, STrEP-tagged TM4SF5WT, STrEP-tagged TM4SF5C189A mutant, STrEP-tagged TM4SF5C118A/C145A mutant, STrEP-tagged TM4SF5D188A mutant, STrEP-tagged TM4SF5R190A mutant, FLAG-tagged TM4SF5, c-Src WT, c-Src-SH432, c-Src-SH321, c-Src-SH1, c-Src1-397, c-Src-SH1Y419F, c-Src-SH1Y530F, c-Src-SH1Y419F/Y530F, c-Src-SH1K298M-HA, c-Src-SH1K298M/Y530F-HA, c-SrcD454A-HA, c-SrcR460A-HA, c-SrcD454A/G459P-HA, c-Src250-450, c-Src250-450-Y419F, c-Src-SH321Y419F, c-Src-SH321Y419F/Y530F, c-Src-SH321Q531E/P532E/G533I, c-Src-SH321Y419F/Q531E/P532E/G533I, or PTP1B. siRNA for a control sequence (siControl) or PTP1B (siPTP1B) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were transiently transfected for 48 h.

Cell lysate preparation and western blots: Whole cell lysates were harvested from subconfluent cells cultured in normal media containing 10% FBS (Thermo Fisher Scientific) or transiently transfected with plasmids for 48 h using a lysis buffer containing 1% Brij58, 150 mM NaCl, 20 mM HEPES (pH 7.4), 2mM MgCl2, 2 mM CaCl2, and protease and phosphatase inhibitors. Then the lysates were normalized and immunoblotted using primary antibodies against pY397FAK, c-Src (ab39546), pS10p27Kip1 (ab62364) [Abcam, Cambridge, UK], pY577FAK (sc-16665-R), pY861FAK (sc-16663-R), pY925FAK (sc-11766-R), pY118paxillin (sc-101774), paxillin (sc-5574), PTP1B (sc-1718), c-Src (sc-8056, clone H-12), α-tubulin (sc-5286) [Santa Cruz Biotechnology], α-smooth muscle actin (A2547, Sigma-Aldrich, St. Louis, MO, USA), pY419c-Src (#2101), pY530c-Src (#2105) [Cell Signaling Technology, Danvers, MA, USA], FAK (clone 4.47, Cat #: 610088), pY397FAK (Cat #: 611723), p27Kip1(Cat #: 610242) [BD Transduction Laboratories, Bedford, MA, USA], or TM4SF5 27. Commercial antibodies were used at 1:1000 dilution in 1% BSA-containing PBS, and anti-TM4SF5 antibody was used at 1:10,000 ~ 20,000 dilution. Quantification of band intensity was performed from three independent blots using NIH Image J software and was normalized to loading controls.

Co-immunoprecipitation: Whole cell extracts were precipitated with anti-FLAG antibody (Cell Signaling Technology) or biotin-precoated beads (Sigma-Aldrich) in a cold room overnight prior to immunoblotting with the indicated antibodies. For co-immunoprecipitation of endogenous proteins, whole cell lysates of endogenous TM4SF5-expressing HepG2 cells were precipitated with either anti-TM4SF5 27 or c-Src (sc-8056, clone H-12) antibody prior to immunoblotting for FAK, c-Src, and TM4SF5. Immunoprecipitated proteins were boiled in 2× SDS-PAGE sample buffer before undergoing western blotting using standard protocols. Quantification of band intensity was performed as described above.

Transwell migration or invasion assay: Stable (TM4SF5-expressing) SNU449Tp cells were infected with adenoviruses expressing the control HA vector (Ad-TA) or various forms of FAK, including WT, an N-terminal deletion mutant (FAKΔ [1-100]), a kinase dead K454R mutant (KAK-KD), or the FERM domain (FAK-FERM), in the absence or presence of 10 μM DMSO, PP2, PP3 (AG Scientifics, San Diego, CA, USA), or CPPs for 24 h. SNU761 cells stably expressing TM4SF5WT were transiently transfected with control siRNA against a scrambled sequence (siControl) or PTP1B (siPTP1B) for 48 h before undergoing the Transwell migration assay for 4 h. Cells were treated with the CPPs before being loaded into the upper chambers. The cells were then analyzed for migration or invasion using Transwell chambers with 8-μm pores (Corning Inc., Corning, NY, USA), where the bottom chamber was filled with 1% BSA alone (in serum-free media) or normal culture media containing 10% FBS as explained previously 28. After incubation, the number of migrated cells from randomly saved images (n = 10) per condition was counted. The data are graphed as the mean ± standard deviation (SD) for each experimental condition.

Invasive extracellular matrix (ECM) degradation analysis: Cellular invasive degradation of the ECM was analyzed using Oregon Green® 488-conjugated-gelatin (10 μg/mL, Thermo Fisher Scientific) as described previously 28. Briefly, cells expressing FLAG (mock) or FLAG-TM4SF5 in the absence or presence of CPP treatment (10 μM) were placed on Oregon Green® 488-conjugated gelatin for 4 h prior to fixation with 3.7% formaldehyde in PBS for 10 min at room temperature [RT]); permeabilization with 5% Triton X-100 for 5 min at RT; blocking with 2% BSA in PBS for 30 min at RT; and then staining with phalloidin for 30 min at RT. Then the samples were examined for black spots, which indicate degradation of fluorescent gelatin, using a fluorescent microscope (BX51, Olympus, Tokyo, Japan). Relative ECM degradation by percent (%) was graphed as the mean ± SD (n=10) per condition.

Indirect immunofluorescence: Cells maintained under normal culture conditions on glass coverslips were treated with peptides for 1 day in the absence or presence of 10 μM TSAHC, a specific TM4SF5 inhibitor 29. The cells were incubated with either PBS (non-permeabilized) or 0.5% Triton X-100 in PBS (permeabilized) for 5 min at RT and then stained using an antibody against TAT (1:100 dilution, Cell Applications, Inc., San Diego, CA, USA). Cells treated with the CPPs were also fixed with 4% formaldehyde for 10 min at RT, incubated with 30 mM glycine for 30 min, permeabilized as described above, and then stained with antibodies against pY397FAK (Cat #: 611723, BD Transduction Laboratories, 1:500 dilution) or pY861FAK (sc-16663-R, Santa Cruz Biotechnology, 1:500 dilution). Phalloidin was also used to stain for F-actin. Immunofluorescent images were acquired on a confocal (Nikon eclipse Ti; Nikon, Melville, NY, USA) or a fluorescent microscope (BX51, Olympus, Japan).

Mouse tumor xenograft: Four-week-old female or male BALB/c-nu/nu mice were purchased from Orient Co. Ltd (Seungnam, Korea). All animal procedures were performed in accordance with the Seoul National University Laboratory Animal Maintenance Manual and with Institutional Review Board approval (SNU-130311-6-2, SNU-140423-11-7). Mice were housed in a specific pathogen-free room under controlled temperature and humidity. Viable SNU449T7 cells (5 × 106) stably expressing TM4SF5 or PLC/PRF/5 cells (107) endogenously expressing TM4SF5 were resuspended in sterile PBS and injected subcutaneously into 5-week-old mice (n = 5 or 6, respectively). Peptide treatment (0.074 or 0.222 nmol/g for TCsr and 0.064 or 0.190 nmol/g for TcxC for 7~8 days; 0.111 nmol/g for TCsr and 0.095 nmol/g for TcxC for 15 days) was initiated when the average tumor volume was approximately 100 mm3. Peptides or PP2/PP3 (3 mg/kg) were administered by intraperitoneal or subcutaneous injection, respectively, every day for 7~8 or 15 days. Body weight and tumor volume, which was determined as described previously 28, were measured daily.

SNU449T7 cells (5 × 106 cells/100 μL sterile PBS) were injected in the lateral tail vein of female BALB/c-nu/nu mice (n = 4). Two weeks later, peptides (0.037 or 0.185 nmol/g for TCsr and 0.032 or 0.158 nmol/g for TcxC) were intraperitoneally injected every other day for 2 weeks. All mice were sacrificed for lung analysis 2 weeks later. Metastatic colonies were counted macroscopically on the lung surface after staining with Bouin's solution (Sigma-Aldrich). In an orthotopic model, SNU449T7 cells stably transfected with tGFP-shScram or tGFP-shTM4SF5 (Origene Technologies, Inc) were orthotopically injected once into the livers of 6-week-old male BALB/c nude mice (5 × 105 cells/mouse, n = 5). Four weeks later, the livers and lungs were analyzed for tumor formation.

Statistical methods: Data are presented as the mean ± SD and analyzed by ANOVA with Tukey's range-test or two-tailed unpaired Student's t-test to determine the statistical significance of differences observed between groups (GraphPad Prism, version 7, San Diego, CA, USA). p-values less than 0.05 were considered statistically significant.

Homology modeling of TM4SF5: The entire human TM4SF5 sequence was retrieved from the UniProtKB Database (UniProt id: O14894). An HHpred search 30 of this sequence suggested the X-ray crystal structure of CD81 (PDB id: 5TCX) 31 as a template; therefore, the sequence alignment deduced by HHpred was utilized as the initial sequence alignment. The sequence alignment was refined manually by comparing the secondary structure predicted by PSIPRED 4.0 32, transmembrane region predicted by TMHMM 33 and UniProt, and the disordered region predicted by DISOPRED 34. The homology model was then built using the RosettaCM modeling protocol 35. Among the 100 generated models, the best model was selected based on its scores, energy values, and appropriateness of the transmembrane (TM) region. The intracellular loop (ICL) was refined using RCD+, a fast loop modeling server 36. The highest scoring loop model was selected and merged to construct a whole sequence model.

Protein-protein docking of the TM4SF5 C-terminus and the c-Src kinase domain: For the direct docking of two different proteins, the fourth TM helix followed by the C-terminal tail of TM4SF5 was modeled using Modeller software 37 with the IVLCGIQLVNATIGVFCGDCRKKQDTPH sequence. (The underlined residues belong to the disordered C-terminal tail, and the remaining resides constitute the fourth TM helix). The resulting final helix model was docked to the X-ray crystal structure of inactive c-Src (PDB id: 2SRC) 38 using ClusPro protein-protein docking software 39. We manually investigated all final reported complexes (100 clusters) and selected the best model based on the following biological criteria: i) the helix binds to the SH1 domain of c-Src, ii) no steric hindrance to the possible membrane location, iii) no steric hindrance during the c-Src conformational change, and iv) the helix binds a region that is structurally different between the inactive and active forms of c-Src kinase. Then, the homology model of the whole TM4SF5 sequence was aligned to the final helix model of the docked complex structure.

Simulations of molecular dynamics (MD) between TM4SF5 and the c-Src complex: The TM regions of TM4SF5 were projected along the z-axis using the 5TCX.pdb (our template) entry in the Orientations of Proteins in Membranes (OPM) database. The input system of the TM4SF5 and c-Src complex structure was prepared using the Input Generator 40 and Membrane Builder 41 modules of CHARM-GUI (http://www.charmm-gui.org/). The TM regions of TM4SF5 were inserted into the 1-palmytoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) lipid bilayer, and the whole complex structure was solvated and neutralized with 0.15 M NaCl. The entire system, which was approximately 92.1×91.8×151.7 Å3 in size, contained one TM4SF5 chain, one c-Src chain, one cholesterol bound in TM4SF5, one ATP, and two Mg2+ ions bound to c-Src, 268 lipids, 92 sodium ions, 94 chloride ions, and 33,999 TIP3 water molecules. Then, the MD simulations were performed using Gromacs v.2018.4 42 with CHARMM36 force fields 43. After equilibration for 10 ns with an NVT ensemble, an additional equilibration step was performed with the NPT ensemble for 100 ns. The pressure was sustained at 1 bar using the Parrinello-Rahman method while the temperature was maintained at 300 K using the Nose-Hoover method. Non-bonded interactions had a 1.2 Å cutoff and 1.0 Å switching distance, and electrostatic interactions were calculated using the particle mesh Ewald (PME) method with a 1.2 Å cutoff. The bonds connected to hydrogen atoms were constrained using LINCS. Three independent production runs were performed for 1,000 ns with a time step of 2 fs. The trajectory analysis was performed using GROMACS v.2018.4 and VMD v.1.9.4.

Sequence conservation analysis: The sequences were retrieved from UniProtKB using the protein family name “L6 tetraspanin family” as a keyword. 'Fragment' sequences and those longer than 600 amino acids in length were removed. Multiple sequence alignment was conducted by MUSCLE (MUltiple Sequence Comparison by Log-Expectation) 44 and analyzed using Jalview 45. The sequence logo was created with WebLogo 46.

All molecular graphic figures were generated using PyMOL v.2.4.0 software (http://www.pymol.org), and the movies were prepared in VMD v.1.9.4 (http://www.ks.uiuc.edu/Research/vmd). Other graphs for the computational studies were generated with Grace-5.1.22/QtGrace v.0.2.6 (https://sourceforge.net/projects/qtgrace).