Expression of focal adhesion kinase (FAK) in endothelial cells (EC) is essential for angiogenesis, but how FAK phosphorylation at tyrosine-(Y)397 and Y861 regulate tumor angiogenesis in vivo is unknown. Here, we show that tumor growth and angiogenesis are constitutively reduced in inducible, ECCre+;FAKY397F/Y397F–mutant mice. Conversely, ECCre+;FAKY861F/Y861F mice exhibit normal tumor growth with an initial reduction in angiogenesis that recovered in end-stage tumors. Mechanistically, FAK-Y397F ECs exhibit increased Tie2 expression, reduced Vegfr2 expression, decreased β1 integrin activation, and disrupted downstream FAK/Src/PI3K(p55)/Akt signaling. In contrast, FAK-Y861F ECs showed decreased Vegfr2 and Tie2 expression with an enhancement in β1 integrin activation. This corresponds with a decrease in Vegfa–stimulated response, but an increase in Vegfa+Ang2- or conditioned medium from tumor cell–stimulated cellular/angiogenic responses, mimicking responses in end-stage tumors with elevated Ang2 levels. Mechanistically, FAK-Y861F, but not FAK-Y397F ECs showed enhanced p190RhoGEF/P130Cas-dependent signaling that is required for the elevated responses to Vegfa+Ang2. This study establishes the differential requirements of EC-FAK-Y397 and EC-FAK-Y861 phosphorylation in the regulation of EC signaling and tumor angiogenesis in vivo.
Distinct motifs of the focal adhesion kinase differentially regulate tumor blood vessel formation and remodeling.
Deciphering the molecular mechanisms that control tumor angiogenesis is fundamental to understanding its regulation (1, 2). Vegfa is essential in triggering angiogenesis, while angiopoietin 2 (Ang2) controls vessel maintenance and is upregulated in response to Vegfa in both endothelial (EC) and tumor cells (2–4), during tumor progression (5–9). Despite this, how essential signaling mediators, such as focal adhesion kinase (FAK), control these regulatory responses is poorly understood.
FAK is a ubiquitously expressed nonreceptor tyrosine kinase that regulates several EC processes important for angiogenesis, including motility and proliferation (10). FAK is composed of an N-terminal FERM domain, a kinase domain, preceded by a tyrosine (Y)397 residue, and a C-terminal FAT (focal adhesion–targeting) domain, preceded by Y861 residue. The C-terminal FAT domain indirectly links FAK to integrins via other focal adhesion signaling complex proteins, such as p190RhoGEF, a RhoA-specific guanine exchange factor (GEF; refs. 10–13). FAK kinase activity results in Y397 phosphorylation, which, in turn, allows proteins containing the Src-homology (SH2) domain to bind to FAK, for example, Src and PI3K. FAK-Src complexing potentiates further FAK phosphorylation at other FAK domains including FAK-Y861 (10, 14). In addition, other FAK-binding partners, such as p130Cas and phospholipase C-gamma (Plcγ), enable FAK to function as a transducer of integrin and growth factor stimulation, leading to phosphorylation of Erk and Akt (15–17).
Previous work indicates that constitutive FAK genetic ablation or pharmacologic inhibition of FAK kinase activity results in embryonically lethal vascular defects (18–21). In addition, deletion of exon 5, which encodes FAK-Y397, also leads to an embryonic lethal phenotype with vascular permeability defects (20). In adult mice, however, data resulting from EC loss of FAK are not consistent and depend on the model used: FAK deletion in adult Tie2-Cre–positive ECs reduced human glioma burden, with increased blood vessel density and decreased vascular permeability (22); in contrast, tumor growth is inhibited in Pdgfb–CreERT;FAKfl/fl mice, with reduced blood vessel density but with no changes in vascular maturation (23). Notwithstanding these data, the in vivo requirement for phosphorylation of FAK-Y397 and Y861 in controlling tumor angiogenesis has been unknown until this report.
Here, using inducible EC-specific FAKY397F/Y397F and FAKY861F/Y861F mice (24), we have identified the differential in vivo requirements for these motifs in tumor angiogenesis.
Materials and Methods
To study the effect of inducible EC-specific FAK mutations in vivo, a Cre-inducible system for mutant FAK expression was generated by designing mutant chicken-FAK constructs (preceded by a STOP sequence flanked by loxP sites) that were targeted to the ubiquitous Rosa26 locus (24). R26FAKKI/KI mice were bred with EC-type–specific Pdgfb-Creert;FAKfl/fl mice to generate mice with tamoxifen-inducible mutant FAK-knockin and endogenous FAK-knockout in endothelial cells (24), Pdgfb-iCreert;FAKfl/fl;R26FAKY397F/Y397F (ECCre; FAKY397F/Y397F) and Pdgfb-iCreert;FAKfl/fl;R26FAKY861F/Y8617F (ECCre; FAKY861F/Y861F) mice, in which tyrosines 397 and 861 were mutated to phenylalanine, respectively. Pdgfb-iCreert;FAKfl/fl;R26FAKWT/WT mice expressed wild-type FAK, as controls.
All animals were used in accord with United Kingdom Home Office regulations (Home Office license number 70/7449). The in-house Ethics Committee at Queen Mary University of London (London, United Kingdom) approved all experiments, using mice under the project license.
B16F0 and CMT19T cell culture
B16F0 melanoma cells (ATCC CRL-6322) and CMT19T lung carcinoma cells (CR-UK Cell Production) were grown in DMEM (Invitrogen), supplemented with 10% FCS (PAA). Cells were tested for Mycoplasma by PCR in 2015. Cells were cultured on uncoated T175 flasks at 37°C 5% CO2 and passaged 2–5 times until used in mouse injections. Tumor cells were 70%–80% confluent when trypsinized, washed three times in 50 mL PBS, passed through a 70-μm cell strainer (BD Falcon) and counted using the automated cell counter (Scepter 2.0, Millipore). The cell suspension concentration was adjusted to 107 cells/mL in PBS prior to 100 μL tumor cell suspension subcutaneous injection.
Subcutaneous tumor growth
Twelve- to 16-week-old mice were given tamoxifen (150 μL of 10 mg/mL solution; Sigma) via intraperitoneal (i.p.) injection on two consecutive days; mice were fed with tamoxifen-containing diet (Tamoxifen400; Harlan Laboratories) starting from the day of the second tamoxifen injection until the end of the experiment. Five to 7 days after initial tamoxifen injection, mice were anesthetized with isoflurane (Abbott) and subcutaneously injected into the flank with 100 μL of 1 × 106 B16F0 mouse melanoma cells or CMT19T lung carcinoma cells. Once tumors became palpable (usually at day 5–7 for B16F0 or at day 8–10 for CMT19T), the tumor size was measured with digital calipers every other day until the experimental endpoint.
When required, ante-mortem procedures were performed as described in the next section, and dissected tumors were snap-frozen in liquid nitrogen or fixed in 10% formalin.
Blood vessel density analysis
Snap frozen sections were fixed with cold acetone for 10 minutes, blocked with 5% BSA in PBS, incubated with a rat anti-endomucin antibody (dilution 1/400 clone V.7c7, Santa Cruz Biotechnology) overnight at 4°C, washed in PBS 0.02% Triton X-100, and then incubated for 60 minutes at room temperature with an Alexa Fluor–labeled anti-rat antibody (dilution 1/300 Invitrogen). Tumor blood vessels were counted across entire midline sections, and the numbers expressed as vessels per mm2 of tumor section.
Pimonidazole-HCl (Hypoxyprobe) at 60 mg/kg body weight was intraperitoneally injected 1 hour before culling mice. Snap-frozen tumor sections were prepared as described previously and incubated overnight at 4°C with a FITC-conjugated mouse anti-pimonidazole antibody (Hypoxyprobe HPI Burlington; MA01803), and then washed and mounted. The hypoxic index was calculated as the area of pimonidazole staining/area of tumor section.
Subcutaneous sponge assay
Two sterile polyether sponges (approximately 1 × 0.8 × 0.8 cm; Caligen Foam) were inserted subcutaneously in the flanks of adult mice, previously injected with tamoxifen as described for subcutaneous tumor experiments. The sponges were injected every other day with 100 μL of 10 ng/mL Vegf (PeproTech, 450-32), 10 ng/mL of platelet-derived growth factor (PlGF; R&D Systems, 465-PL- 010), 5 ng/mL of basic fibroblast growth factor (bFGF; PeproTech, 450-33), 10 ng/mL of Ang2 (R&D Systems, 7186-AN-025), 10 ng/mL of both Vegfa and Ang2, or 100 μL of PBS as a negative control.
After 15 days, mice were culled and sponges removed and fixed in 10% formalin overnight. The next day, sponges were transferred to 70% ethanol, paraffin-embedded, and 5-μm sections were cut for endomucin staining.
Blood vessel IHC and quantitation
Paraffin-embedded sections of sponges and Matrigel plugs were dewaxed, rehydrated, and fixed with Methanol+H2O2. Sections were processed as follows: antigen-retrieval using 10 mmol/L trisodium citrate (pH 6) in the microwave; slides washed in distilled water; incubated with blocking buffer containing 2% goat serum/1% BSA/0.1% TritonX100/PBS; incubated with anti-endomucin (Endomucin clone v.7C7; Santa Cruz Biotechnology; 1/200 in blocking buffer) overnight at 4°C; washed three times for 5 minutes at room temperature in 1× PBS; incubated with Elite ABC reagent (Vector kit PK6100 or 6102) according to the manufacturer's instructions for 30 minutes at room temperature; washed twice in PBS; incubated with DAB substrate (Vector kit SK4100) until the color of the reaction was evident (∼5 minutes); slides counterstained with hematoxylin; dehydrated, cleared, and mounted with DPX (Sigma).
Slides were scanned using the Pannoramic 250 scanner. For sponges, infiltrated microvessels were quantified across the width of the sponge at ×30 magnification. The number of microvessels was quantified for the fields that presented with infiltrated cells and the average number of microvessels per field was determined for each sponge. For Matrigel plugs, infiltrated microvessels were counted across each individual plug and divided by the area of the plug (No. × 10−5/μm2).
Primary nonimmortalized and immortalized endothelial cells were starved overnight in Opti-MEM and then treated with Vegfa (30 ng/mL) and/or Ang2 (30 ng/mL) for the indicated times.
For generation of mouse lung endothelial cell lines (MLEC) and immunofluorescence of cells, see Supplementary Materials and Methods.
A total of 1,000 MLECs were incubated overnight on a hanging drop to form a spheroid and transferred to a collagen type I gel (1 mg/mL). After polymerization, 1% FBS EBM2 medium with the indicated growth factors at the same concentration as cell treatment were added on top of the gels and sprout formation was quantified 24 hours later.
Embedded spheroids were fixed in 4% PFA for 30 minutes, washed in PBS, permeabilized in 0.3% Triton X-100 PBS for 1 hour, and stained overnight with rhodamine phalloidin (Invitrogen). The next day, spheroids were washed for 30 minutes three times with 0.1% Triton X-100 PBS, and mounted with Fluoromont-G (Thermo Fisher Scientific). Superimposed brightfield and fluorescent images were acquired using a confocal spinning disk microscope with a 20× objective and sCMOS confocal camera (Nikon).
For details on preparation of Dunn Chambers, see Supplementary Methods. A total of 45,000 mouse endothelial cells were plated on coated 22-mm glass coverslips and cultured overnight. The following day, cells were washed with PBS and serum-starved for 5 hours in serum-free medium (OPTI-MEM I + GlutaMAX, Gibco, Invitrogen). Coverslips were inverted onto the Dunn Chambers leaving a gap in the outer well and sealed on three sides with hot wax mixture (Vaseline:paraffin:beeswax—1:1:1). The medium was removed from the outer well by capillary action and was rinsed with Opti-MEM before filling with Opti-MEM containing 30 ng/mL of Vegf (PeproTech, 450-32) or Ang2 (R&D Systems, 7186-AN-025) or both together. The chamber was then sealed with wax and mounted on an Olympus IX71 inverted microscope. Images were acquired by phase contrast imaging using a 10× phase contrast objective (NA 0.25). Cell images were collected using a Retiga R6 CCD camera (Photometrics/Qimaging), taking a frame every 10 minutes for 16 hours using Micro Manager acquisition software (NIH, open source). Subsequently, all the acquired time-lapse sequences were displayed as an AVI file and cells from the time-lapse sequence were tracked using ImageJ. Tracking resulted in the generation of a sequence of position coordinates relating to each cell in each frame, motion analysis was then performed on these sequences using Wolfram Mathematica 7 software. Data are depicted as pooled example tracks of cells or rose plots that show the proportion of cells with migratory direction lying within each 20° interval. The red arrow represents the mean direction of migration and the green segment represents the 95% confidence interval determined by Rayleigh test.
ECs were lysed in ice-cold RIPA Buffer supplemented with 50 mmol/L NaF, 60 mmol/L β-glycerol phosphate, 10 mmol/L sodium pyrophosphate, 2 mmol/L Na3VO4, and a dilution 1/100 of protease inhibitor cocktail III (Roche) in distilled water. Lysates were then centrifuged at 10,000 × g for 10 minutes. The clear supernatants were transferred into a clean tube and frozen at −80°C until analysis. Before electrophoresis, lysates were thawed quickly at 37°C and transferred to wet ice. Protein concentrations were assessed using DC Bio-Rad assay and adjusted between lysates using RIPA Buffer. Samples were prepared for Western blotting by supplementing with NuPage 4× loading buffer and Beta-Mercaptoethanol 1/100 volume and heat-treating for 10 minutes at 70°C.
Total protein lysates were separated by reducing SDS-PAGE and transferred onto nitrocellulose membranes. The membranes were blocked for 30 minutes at room temperature in 5% milk in Tris buffered saline + 0.1% Tween-20 (TBST) then incubated overnight at 4°C in 4% BSA in TBST with the following antibodies: myc-tag clone 9E10 (#600-302) from Novus Biologicals, pY397-FAK (#3283), phospho-FAK (Tyr925; 3284S), pY410-p130Cas (#4011), pY416-Src (#2101), p-Akt (#4058), p-Erk (#9102), PLCγ1 (2822), pY783-PLCγ1(2821), total Vegf-r2 (#2479), and phospho-PI3K p85 (Tyr458)/p55 (Tyr199; #4228) from Cell Signaling Technology; pY861-FAK (#PS1008), GAPDH (#MAB374), and total β1 Integrin (#MAB1997) from Millipore; Tie2 (AF762) from R&D Technologies, and Ang2 (#PA5-27297), phospho-paxillin (44722g) from Invitrogen, pY576-FAK (44652G) from Biosource/Thermo Fisher Scientific and with HRP-conjugated sheep anti-mouse or goat anti-rabbit, anti-rat antibodies (Jackson Immunoresearch). Membranes were stripped and reprobed for total protein levels: FAK (#610088) from BD Transduction Laboratories, p130Cas (#05-469, Millipore), Src (#2110), Akt (#2966), and Erk (#9102) from Cell Signaling Technology and paxillin (610052) from BD Biosciences. p190RhoGEF was a kind gift from David Schlaepfer (Moores Cancer Center, University of California, San Diego, La Jolla, CA; 8E8E5 mAb, mouse IgG1, 82% purity, protein G–purified antibody, directed against aa 1502–1519 of the Rgnef protein).
Stress fiber alignment analysis
Endothelial cells were prepared and stained as described (see IF of endothelial cell lines). Phalloidin Z-stack images were acquired using a confocal spinning disc microscope with a 40× objective and sCMOS camera (Nikon). For further details of analysis, see Supplementary Materials and Methods.
Endothelial cells were lysed in ice-cold RIPA buffer supplemented with 50 mmol/L NaF, 60 mmol/L β-glycerol phosphate, 10 mmol/L sodium pyrophosphate, 2 mmol/L Na3VO4 and a dilution 1/100 of protease inhibitor cocktail III (Roche) in distilled water. One microgram of anti-myc tag antibody (Novus Biologicals, #600-302) was bound to 50 μL of protein G magnetic Dynabeads (Life Technologies) prior to overnight incubation with 300 μg of lysate at 4°C. Beads were subsequently washed prior to the addition of sample buffer containing 0.1 mol/L DTT. Following this, immunoprecipitates were analyzed on SDS-PAGE gels as outlined above.
RhoA activation G-LISA
A total of 65,000 endothelial cell lines were plated in precoated 6-well plates and allowed to attach overnight, washed twice in PBS, serum-starved for 3 days in serum-free medium (Opti-MEM I + GlutaMAX, Gibco, Invitrogen); media were removed and replaced with Opti-MEM containing 30 ng/mL Vegfa (PeproTech, 450-32) or Ang2 (R&D Systems, 7186-AN-025) or the combination, for 10 and 40 minutes. G-LISA was then performed according to the manufacturer's instructions (BK 124 from Cytoskeleton, Inc.).
Small interfering RNA transfection
ECs were plated the day before transfection and left in Opti-MEM overnight. Lipofectamine transfection reagent was used according to the manufacturer's instructions (Lipofectamine RNAi/MAX, 13778, Invitrogen) with 20 μmol/L of either scrambled (Scr; ON-TARGETplus Non-targeting Pool, D-001810-10-20, Dharmacon) or p190RhoGEF siRNA (Smartpool ON-TARGETplus Arhgef28 siRNA, L-040493-01-0005, Dharmacon) or Ang2 (Smartpool ON-TARGETplus Angpt2 siRNA, L-040375-01-0005, Dharmacon) siRNAs. Cells were left in Opti-MEM for 5 hours posttransfection and then supplemented with FBS (Sigma). The day after transfection, cells were detached and plated for the essays (proliferation and spheroids). For conditioned medium experiments, after transfection, B16F0s cells were washed and medium changed for Opti-MEM and left for 48 hours to produce conditioned medium that was then used in stimulation essays on ECs.
Analysis of statistical significance
Prism software was used to statistically analyze data. Student t test and Mann–Whitney test were employed when two individual groups were compared. One-way ANOVA was used when comparing with a control group (or selected columns) and two-way ANOVA was used when comparing between different conditions or groups. Selected columns were compared with Sidak post test or Dunnett post test used when comparing with a control. Bonferroni multiple comparison test was used when multiple comparisons were performed. Comparison of variance was performed, confirming no difference in variance between two samples nor interaction between mean and variance.
Characterization of ECCre+;FAKY397F/Y397F and ECCre+;FAKY861F/Y861F mice
To investigate the in vivo effects of EC-specific homozygous nonphosphorylatable FAK-Y397F and FAK-Y861F, we developed Pdgfb-iCreert;FAKfl/fl;R26FAKWT/WT, Pdgfb-iCreert;FAKfl/fl;R26FAKY397F/Y397F, or Pdgfb-iCreert;FAKfl/fl;R26FAKY861F/Y861F mice (Supplementary Fig. S1A and Supplementary Methods). Tamoxifen treatment of Pdgfb-iCreert;FAKfl/fl;R26FAKWT/WT, Pdgfb-iCreert;FAKfl/fl;R26FAKY397F/Y397F, or Pdgfb-iCreert;FAKfl/fl;R26FAKY861F/Y861F mice (24), generated ECCre+;FAKWT/WT, ECCre+;FAKY397F/Y397F, and ECCre+;FAKY861F/Y861F mice, respectively (Supplementary Fig. S1B–S1D). DNA sequencing confirmed mutation specificity in mouse genomic DNA (Supplementary Fig. S2A). Tamoxifen treatment induced endogenous EC mouse-FAK deletion and simultaneous myc-tagged knockin chicken-FAK RNA expression (Supplementary Fig. S2B) and myc-tagged FAK-knockin protein expression in Cre-positive but not Cre-negative mouse lung ECs (Supplementary Fig. S2C and S2D).
Tumor growth is inhibited in ECCre+;FAKY397F/Y397F but not ECCre+;FAKY861F/Y861Fmice
ECCre+;FAKY397F/Y397F and ECCre+;FAKY861F/Y861F and respective ECCre− control mice were injected subcutaneously with melanoma (B16F0) or lung carcinoma cells (CMT19T). B16F0 and CMT19T tumor growth were reduced in ECCre+;FAKY397F/Y397F mice (Fig. 1A) but not ECCre+;FAKY861F/Y861F mice (Fig. 1B). Knockin efficiency was confirmed in freshly isolated tumor ECs that showed significant downregulation of mouse FAK and expression of mutant chicken FAK (Supplementary Fig. S2E and S2F) and myc-Tag expression in ECCre+ but not ECCre− tumor blood vessels in vivo (Supplementary Fig. S2G). Endpoint tumor hypoxia was elevated significantly in tumors grown in ECCre+;FAKY397F/Y397F mice (Fig. 1C), but not affected in ECCre−;FAKY861F/Y861F mice (Fig. 1D; Supplementary Fig. S3A; ref. 33).
Tumor angiogenesis is significantly decreased in ECCre+;FAKY397F/Y397F mice but only transiently in ECCre+;FAKY861F/Y861Fmice
Given that tumor angiogenesis involves distinct molecular processes at initiation and maintenance, we examined angiogenesis in early- and end-stage tumors. Blood vessel density was significantly reduced in early-stage tumors from ECCre+;FAKY397F/Y397F (Fig. 1E) and ECCre+;FAKY861F/Y861F (Fig. 1F; Supplementary Fig. S3B). In contrast, although reduced blood vessel density was still apparent in end-stage B16F0 and CMT19T tumors grown in ECCre+;FAKY397F/Y397F mice (Fig. 1G), no significant differences in blood vessel density was observed in ECCre+;FAKY861F/Y861F mice when compared with respective controls (Fig. 1H).
Furthermore, in both early- and end-stage tumors in ECCre+;FAKY397F/Y397F or ECCre+;FAKY861F/Y861F mice, no significant differences in blood vessel perfusion, size, relative blood vessel leakage, or NG2-positive pericyte association were observed in these mutant mice (Supplementary Fig. S4A–S4D).
Overall, the loss of phosphorylation of either FAK-Y397 or FAK-Y861 reduces the initiation of tumor angiogenesis and appears to recover in ECCre+;FAKY861F/Y861F mice only, suggesting that phosphorylation of FAK-Y861 or FAK-Y397 are involved in distinct molecular mechanisms in the control of sprouting angiogenesis.
Differential responses of ECCre+;FAKY397F/Y397F and ECCre+;FAKY397F/Y397F mice to Vegfa + Ang2
Tumor angiogenesis is controlled by several different growth factors including Vegfa, Ang2, bFGF, and PlGF (2). Angiogenic responses to Ang2 involve essential costimulation with Vegfa (3, 25–29). We hypothesized that the restoration of angiogenesis observed in end-stage tumors in ECCre+;FAKY861F/Y861F, but not in ECCre+;FAKY397F/Y397F, mice could be a consequence of differential responses to tumor angiogenic growth factors. To test this, synthetic sponges were implanted subcutaneously into ECCre+;FAKY397F/Y397F, ECCre+;FAKY861F/Y861F, and respective ECCre− control mice, and subsequently impregnated with growth factors to induce angiogenesis. Vegfa, PlGF, bFGF, and Vegfa + Ang2–stimulated angiogenesis were all reduced in ECCre+;FAKY397F/Y397F when compared with ECCre−;FAKY397F/Y397F mice (Fig. 2A and B; Supplementary Fig. S5A). In contrast, Vegfa was reduced and PlGF- or bFGF-stimulated angiogenesis was not affected in ECCre+;FAKY861F/Y861F mice. However, the combination of Vegfa + Ang2 significantly increased angiogenesis in ECCre+;FAKY861F/Y861F mice, relative to control mice (Fig. 2C and D; Supplementary Fig. S5B).
In vitro, we confirmed that Vegfa and Vegfa + Ang2 treatment stimulated FAK-WT and FAK-Y861F but not FAK-Y397F EC sprout formation. Stimulation with Vegfa alone reduced sprouting response in FAK-Y861F EC spheroids compared with FAK-WT spheroids. Importantly, Vegfa + Ang2 stimulation enhanced FAK-Y861F spheroid sprouting compared with stimulation with Vegfa alone, but no significant difference in sprouting was observed in WT spheroids stimulated with Vegfa versus Vegfa + Ang2 (Fig. 2E).
Ang2–stimulated angiogenic responses have been described to be autocrine in ECs (3, 25–29). However, Ang2 expression levels were similar between FAK-WT, FAK-Y397F, and FAK-Y861F ECs (Supplementary Fig. S5C), suggesting that EC-FAK mutations did not regulate autocrine EC Ang2 levels. In contrast, B16F0 cells secrete high levels of Ang2 compared with Vegfa (Fig. 2F). Importantly, a significant increase in Ang2 transcript levels was detected in end- versus early-stage tumors (Fig. 2G).
Together, endothelial-specific mutations in FAK-Y397F and FAK-Y861F trigger distinct angiogenic responses depending on the angiogenic stimuli, particularly to combined Vegfa + Ang2.
Nonphosphorylatable mutations in FAK-Y397 and FAK-Y861 differentially affect cell proliferation and migration responses to Vegfa, Ang2, and combined Vegfa + Ang2
To further dissect the differential angiogenic responses that we observed, particularly to Vegfa and combined Vegfa+Ang2, FAK-WT, FAK-Y397F, and FAK-Y861F–mutant EC lines were used in in vitro assays of proliferation and migration (30). FAK-WT ECs showed increased cell proliferation upon Vegfa stimulation, but not Ang2 alone or Vegfa + Ang2. FAK-Y397F ECs did not respond to Vegfa, Ang2, or Vegfa + Ang2. In contrast, FAK-Y861F ECs, while not responding to Vegfa or Ang2 alone, exhibited increased proliferation upon combination Vegfa + Ang2 stimulation (Fig. 3A).
We next examined whether phosphorylation of FAK-Y397 or FAK-Y861 played a role in EC chemotactic migration toward Vegfa, Ang2, or combined Vegfa + Ang2 using the Dunn Chamber assay. Vegfa–induced directional migration was apparent in WT ECs, but not in FAK-Y397F or FAK-Y861F ECs, whereas Ang2 or Vegfa + Ang2 combination treatment did not promote directional migration in any of the cell lines tested (Fig. 3B–D). Migration speed was decreased in FAK-WT and FAK-Y397F cells stimulated with Ang2 or Vegfa + Ang2 compared with Vegfa alone treatment (Fig. 3E). In contrast, FAK-Y861F–mutant ECs, despite exhibiting reduced migration speed in Vegfa, Ang2, or combined Vegfa + Ang2 compared with both FAK-WT and FAK-Y397F cells, increased their migratory speed in Vegfa + Ang2 compared with Vegfa alone (Fig. 3E).
We next tested the effect of Ang2 depletion in B16F0 cells on angiogenic responses in vitro and in vivo. FAK-Y861F versus FAK-WT EC spheroids cultures were exposed to conditioned media (CM) from Ang2-siRNA and Scr-siRNA transfected B16F0s in vitro. Ang2 expression levels were reduced partially in Ang2 siRNA–depleted B16F0 cells 48 hours after transfection when CM was collected (Supplementary Fig. S6A). FAK-Y861F endothelial sprouting was enhanced compared with FAK-WT endothelial cell sprouting when exposed to B16F0-scr siRNA CM (Fig. 3F and G). In contrast, CM from Ang2 siRNA–depleted B16F0s inhibited the enhanced sprouting of FAK-Y861F ECs, but had no significant on FAK-WT EC sprouting (Fig. 3F and G).
In an in vivo Matrigel plug angiogenic assay, Matrigel containing either Ang2-siRNA–depleted B16F0 cell CM, Scr-siRNA–transfected B16F0 CM or PBS was injected into ECCre−;FAKY861F/Y861F and ECCre+;FAKY861F/Y861F mice and blood vessel infiltration assessed 6 days postinoculation. Matrigel plugs were immunostained to detect endomucin-positive vessel infiltration. CM from Scr-siRNA–transfected B16F0 cells resulted in enhanced microvessel infiltration in Matrigel plugs from ECCre+;FAKY861F/Y861F mice compared with ECCre−;FAKY861F/Y861F mice. In contrast, CM from Ang2-siRNA–transfected B16F0 cells reduced the enhanced microvessel infiltration in ECCre+;FAKY861F/Y861F mice down to that observed in PBS controls (Supplementary Fig. S6B and S6C).
EC FAK-Y397F and FAK-Y861F mutations differentially regulate Vegfr2 and Tie2 expression and β1 integrin activation
Ang2 has dual roles during angiogenesis: in the absence of Vegfa, Ang2 binds Tie2 promoting EC death and vessel regression (31); while in the presence of Vegfa, Tie2 is downregulated and the proangiogenic function of Ang2 is mediated mainly through β1 integrin signaling (25, 26, 28, 29). Thus, the balance between these receptors can influence the effect of Ang2 on angiogenesis. We therefore examined the levels of Vegfa and Ang2 receptors (Vegfr2 and Tie2, respectively) in FAK-Y397F and FAK-Y861F ECs. Vegfr2 expression levels were significantly decreased in FAK-Y397F, and more modestly decreased in FAK-Y861F ECs. In contrast, Tie2 levels were increased in FAK-Y397F, but decreased in FAK-Y861F ECs, relative to FAK-WT EC controls (Fig. 4A). Although total β1 integrin levels appeared unchanged (Supplementary Fig. S7A and S7B), immunofluorescence analysis indicated that the localization of activated β1 integrin to focal contact sites at the end of stress fibers, was severely disrupted in FAK-Y397F ECs when cells were either unstimulated, or stimulated with Vegfa or a combination with Vegfa and Ang2. In contrast, elevated levels of activated β1 integrin were observed in focal contact sites of FAK-Y861F ECs, in both unstimulated and combined stimulation with Vegfa + Ang2 (Fig. 4B and C). Notably, in FAK-WT ECs, only Vegfa induced increased activation of β1 integrin compared with unstimulated levels, whereas in FAK-Y861F ECs, only combined Vegfa + Ang2 induced an increase-activated β1 integrin (Fig. 4B and C).This observation correlated with a significant disruption in F-actin organization and fiber alignment, assessed by image analysis of phalloidin staining, in FAK-Y397F ECs, but not FAK-Y861F ECs, under all the conditions tested (Fig. 4C and D).
The in vivo relevance of these observations was then investigated. End-stage B16F0 tumor sections from ECCre− controls, ECCre+;FAKY397F/Y397F and ECCre+;FAKY861F/Y861F mice were double immunostained for the endothelial marker endomucin together with either Vegfr2 or Tie2. Corroborating our in vitro data, Vegfr2 expression levels were reduced significantly in both ECCre+;FAKY397F/Y397F and ECCre+;FAKY861F/Y861F murine tumor blood vessels (Fig. 4E). In contrast, Tie2 expression was significantly enhanced in ECCre+;FAKY397F/Y397F mice, but significantly reduced in ECCre+;FAKY861F/Y861F–mutant mice tumor blood vessels (Fig. 4F).
This indicates that nonphosphorylatable mutations in FAK-Y397 or FAK-Y861 control an inside-out signal that regulates the expression of Vegfr2, Tie2, and β1 integrin activation, and that these changes correlate with distinct angiogenic responses to Vegfa + Ang2 or tumor cell–derived conditioned medium.
Nonphosphorylatable mutations in FAK-Y397 and FAK-Y861 residues trigger distinct downstream signals
We next sought to investigate how FAK-Y397F and FAK-Y861F mutations differentially affect downstream signaling and compare these responses with our previously published FAK-kinase dead (FAK-KD) mutant ECs (32). Western blot analysis of FAK-Y397F and FAK-KD ECs showed reduced FAK phosphorylation at Y397, Y577, Y861, and Y925. In contrast, the Y861F mutants showed reduced Y861 and Y925 phosphorylation, but no appreciable change in upstream Y397 or Y577 phosphorylation when compared with controls (Fig. 5A). Moreover, double immunofluorescence analysis detection of the knock-in-FAK (FAK-Myc-tag) and p-Src showed reduced colocalization in FAK-Y397F and FAK-KD ECs, but not in FAK-Y861F ECs (Fig. 5B and C). Accordingly, immunoprecipitation of Myc-tagged FAK followed by Western blotting revealed markedly less coprecipitation of FAK with Src, or p-Src, in FAK-Y397F and FAK-KD ECs compared with FAK-WT ECs. In contrast, FAK-Y861F–mutant ECs, showed evidence of coprecipitating well with p-Src and Src (Fig. 5D). Src phosphorylation dynamics were then analyzed in FAK-mutant primary ECs in response to Vegfa + Ang2. Stimulation with Vegfa + Ang2 tended to induce an increase in p-Src/Src levels in FAK-WT ECs. In contrast, Src phosphorylation kinetics showed no significant increase over time in FAK-Y397F or FAK-Y861F, possibly due to constitutively high basal p-Src levels and/or other FAK-independent mechanisms (Supplementary Fig. S8A). In addition, p-Src association with paxillin (focal adhesion marker) was disrupted in FAK-Y397F ECs but not in FAK-Y861F ECs (Supplementary Fig. S8B) despite no changes in paxillin phosphorylation across the mutants in whole-cell lysates (Supplementary Fig. S8C).
To further investigate other downstream signaling pathways that could be affected by Y397F and Y861F FAK mutations in ECs, we performed Western blot analysis using both immortalized and primary ECs. Phospho-p130Cas was reduced in FAK-Y397F ECs and increased in FAK-Y861F ECs (Fig. 5E; Supplementary Fig. S8D). Both phosphorylated-Plcγ, and total Plcγ were elevated in FAK-Y397F ECs when compared with WT and FAK-Y861F ECs (Fig. 5E). These data suggest a possible compensation by Plcγ. PI3K p85-phosphorylation was also increased in the FAK-Y397F ECs, and not changed in FAK-Y861F while p55 phosphorylation was decreased in both FAK-Y397F and FAK-Y861F mutants when compared with WT (Fig. 5E; Supplementary Fig. S8D); Akt-phosphorylation was significantly decreased in FAK-Y397F ECs, but not in FAK-Y861F ECs; and Erk phosphorylation remained unchanged in FAK-Y397F and FAK-Y861F ECs (Fig. 5E). RPPA analysis, designed for FAK signaling, additionally identified other known FAK targets/signaling partners, such as PKC, GSK-3, AMPK, Smad1/5, and IRS that were downregulated in FAK-Y397F ECs to a greater extent than in FAK-Y861F ECs (Supplementary Fig. S8E).
Together, these data suggest that the loss of Y397 phosphorylation, observed also in KD ECs, constitutively disrupts FAK-p-Src coassociation and causes downregulation of other FAK-dependent downstream signaling pathways. In contrast, nonphosphorylatable Y861F mutation does not disrupt FAK–p-Src interaction but Vegfa + Ang2 stimulation is not sufficient to induce an increase in p-Src kinetic response above the constitutively elevated levels of p-Src, suggesting that the increased angiogenesis observed in FAK-Y861F ECs is independent of FAK/Src signaling.
Loss of EC FAK-Y397 and -Y861 phosphorylation affects RhoA activation and p190RhoGEF binding
To further examine how Vegfa and Ang2 affect EC migration and endothelial sprouting differently, and because FAK has been implicated in the regulation of cell migration through RhoA activity (33–35), we investigated possible changes in RhoA activity. FAK-WT, FAK-Y397F, and FAK-Y861F ECs were serum starved, and either unstimulated or stimulated with Vegfa + Ang2 for 10 or 40 minutes, and RhoA activity measured using an ELISA-based assay. RhoA activity was increased in FAK-WT cells following 10 minutes of Vegfa + Ang2 stimulation, which reduced back to unstimulated levels at 40 minutes. This was in line with previous work demonstrating that growth factor–stimulated migration is associated with transient RhoA activity (36). FAK-Y397F ECs showed no change in RhoA activity at either 10 or 40 minutes of Vegfa + Ang2 stimulation. In contrast, FAK-Y861F ECs exhibited a time-dependent increase in active RhoA following Vegfa + Ang2 stimulation, and levels were significantly higher at 40 minutes than in FAK-WT or FAK-Y397F cells (Fig. 6A).
Moreover, because p190RhoGEF is a known RhoA activator and a specific C-terminal FAK-binding partner (13, 37), we investigated p190RhoGEF binding to FAK. Immunoprecipitation of myc-tagged-FAK followed by Western blotting for p190RhoGEF from protein lysates isolated from FAK-WT, FAK-Y397F, and FAK-Y861F ECs showed that FAK-Y861F ECs had higher levels of p190RhoGEF/FAK association when compared with FAK-Y397F and WT ECs (Fig. 6B). This was further validated by proximity ligation assays where FAK/p190RhoGEF association was enhanced in FAK-Y861F, but not in FAK-Y397F compared with FAK-WT ECs (Fig. 6C). The functional relevance of the enhanced FAK-Y861F/p190RhoGEF interaction was tested by siRNA depletion of p190RhoGEF in FAK-Y861F endothelial cells. In FAK-Y861F ECs, depletion of p190RhoGEF reversed the increase in phosphorylation of p130Cas, compared with WT ECs (Fig. 6D). Furthermore, knockdown of p190RhoGEF also reverted Vegfa + Ang2–stimulated enhancement in cell proliferation (Fig. 6E) and sprouting in endothelial spheroid analysis (Fig. 6F) in FAK-Y861F but not WT ECs.
Together, these data suggest that nonphosphorylatable FAK-Y861 mutation in ECs causes constitutively increased binding to p190RhoGEF, sustained RhoA activation, and downstream p130Cas phosphorylation that is required for the enhanced endothelial cell proliferation and sprouting after Vegfa + Ang2 stimulation.
Our data demonstrate that phosphorylation of EC-FAK-Y397 and Y861 act as critical regulators of tumor angiogenesis through distinct molecular mechanisms where FAK lies upstream of growth factor receptor regulation, signaling, and responses (Fig. 7).
Our results show that in ECCre+;FAKY397F/Y397F mice, angiogenesis is constitutively disrupted, corresponding with reduced tumor growth. In contrast, although angiogenesis is reduced in early-stage ECCre+;FAKY861F/Y861F mice tumors, this defect is resolved in end-stage tumors, resulting in no overall effect on tumor growth. Reduced levels of tumor angiogenesis have been previously shown in EC-FAKKO mice (23). EC-FAK-kinase–dead mice (ECCre+;FAKKD/KD; ref. 32) have a similar phenotype to ECFAKKO mice in inhibiting tumor angiogenesis, enhancing tumor hypoxia and reducing tumor growth, all effects that we also observe in ECCre+;FAKY397F/Y397F mice. However, unlike the ECCre+;FAKKD/KD mice, which presented a vascular leakage defect (32), the ECCre+;FAKY397F/Y397F mutants did not. This could be explained by the fact that the kinase domain in ECCre+;FAKY397F/Y397F mice is intact and thus may affect other downstream responses apart from the phosphorylation of FAK-Y397 in the control of vessel leakage (10). Importantly, although previous studies have begun to address the effect of FAK-Y397 deletion (exon 15) or FAK kinase inactivation (20, 32, 38), ours is the first study to investigate the requirements of phosphorylation of ECFAK-Y397 and -Y861 in tumor angiogenesis in vivo.
We hypothesized that differential phosphorylation of FAK could regulate an inside-out signaling cascade affecting sensitivity to angiogenic growth factors. In fact, because the balance of tumor-derived angiogenic growth factor expression is known to temporally regulate tumor angiogenesis (2), we explored the effect of different proangiogenic stimuli that represent the angiogenic growth factor profile of early and later stage tumors. For example, we and others have shown that Ang2 is upregulated in advanced stage tumors (5, 8, 9) providing a rationale to investigate how angiogenesis is apparently rescued in later stage tumors of ECCre+;FAKY861F/Y861F mice.
Previous work has shown that in a Vegfa/Ang2–rich environment, Vegfa causes downregulation of Tie2, the canonical Ang2 receptor. This leaves Ang2 free to bind and activate β1 integrins and in turn activate downstream phosphorylation of FAK-Y397 to induce sprouting (26, 29). Our data indicate that loss of phosphorylation of FAK-Y397 is associated with a constitutive increase in Tie2 expression levels but decreased Vegfr2 levels with disrupted activation and clustering of β1 integrins (27). This alteration in the profile of these surface receptors correlates with a constitutive loss of angiogenic responses in FAK-Y397F ECs in vitro and in vivo, which is in line with previous in vitro reports showing that deletion of FAK or FAK-Y397 leads to loss of integrin- and growth factor–stimulated downstream signaling, proliferation, migration, and RhoA activation (20, 39–41). FAK-Y397F ECs have drastically impaired FAK/Src interactions. In these cells, Src is predominantly observed in intracellular puncta, potentially reflecting an increase in Src degradation, as reported previously in FAK-deficient cells (42). Although loss of FAK-Y397 phosphorylation is responsible for reducing FAK–Src complex formation, levels of p-Src are not reduced, possibly due to compensatory mechanisms independent of FAK (18). Indeed, phosphorylation of the p85 subunit of PI3 kinase, which has been shown to be Src dependent and FAK-independent is also constitutively increased in FAK-Y397F (43). However, as expected, other SH2-binding proteins that directly bind to p-Y397 domain, such as PI3K p55 and subsequent downstream signaling via Akt are reduced (43). The loss of FAK–Src complex formation with the decrease in p-p130Cas may be responsible for the disorganized F-actin bundling observed in FAK-Y397F ECs (44).
In contrast, FAK-Y861F ECs showed constitutively reduced Vegfr2, which may account for the reduced tumor angiogenesis in early-stage tumors where tumor-derived Ang2 levels are lower. In addition, FAK-Y861F ECs have constitutively decreased Tie2 expression and enhanced activation of β1 integrin. The role of β1 integrins in tumor angiogenesis has been extensively studied (45) and a better understanding of the cooperative roles of these integrins with other growth factor receptors are emerging (46). In fact, the reduction in Tie2 is relevant in end-stage tumors where enhanced Ang2–β1-integrin signaling leads to increased angiogenesis, in line with previous findings (26, 29). It has been shown that FAK binds to p190RhoGEF via its C-terminal FAT domain (13), and this interaction is critical for the GEF-dependent RhoA activation (33). Moreover, p190RhoGEF has also been shown to function as a scaffold, via a GEF-independent mechanism, to enhance FAK activation upon integrin clustering (47). Indeed our data suggest that upon Vegf + Ang2 stimulation while FAK-Y861F–mutant cells depend on p190RhoGEF for increased sprouting, FAK-WT cells do not. This suggests the increased complexing of p190RhoGEF with nonphosphorylatable FAK-Y861F is required for the enhanced responses to Vegfa + Ang2, consistent with published data showing that p190RhoGEF links FAK to integrins (13). Furthermore, we speculate that the GEF-independent function of p190RhoGEF, acting upstream of FAK (37, 47) results in constitutively increased activation of β1 integrin in FAK-Y861F cells, making them more responsive to Vegfa + Ang2 or tumor cell–derived conditioned medium. In FAK-Y861F ECs, we also observed constitutively increased phosphorylation of p130Cas and sustained RhoA activation after Vegfa + Ang2 stimulation. Therefore, we postulate that, upon Vegfa + Ang2 stimulation, mimicking the growth factor environment in end-stage tumors, FAK-Y861F binding to p190RhoGEF enhances p130Cas phosphorylation and sustains RhoA activation in maintaining F-actin binding. Previous studies have shown that p190RhoGEF–FAK complexing enhances β1 integrin–mediated downstream signaling (47). Together, these molecular pathways lead to recovery of angiogenesis in an Ang2–enriched tumor microenvironment.
This differential expression of surface receptors between FAK-Y397F and FAK-Y861F ECs indicate that the altered phosphorylation status of these mutants is sufficient to differentially control inside-out signaling. In the same way that FAK activity is required for integrin activation, regulating its recycling, degradation, and surface expression (27, 48–50), it could be speculated that it also regulates Vegfa and Ang2 receptors. On the other hand, it has been demonstrated that adhesion mediated through integrins and FAK can also influence the expression of surface proteins (51), so the effect on Tie2 and Vegfr2 may also be a consequence of constitutively altered β1 integrin activation.
Although beyond the scope of this study, it is of interest to consider that inhibitors of FAK kinase activity, which lead to a reduction in FAK-Y397-phosphorylation, are currently in clinical trials for the treatment of cancer (52). These inhibitors are reported to have effects on targeting both tumor cells and stromal cells such as ECs (53), but as long-term resistance to kinases in malignant cells is likely, the effects of inhibitors in affecting the tumor stroma become important. Allosteric FAK inhibitors that bind to distinct domain sites are also being developed, but are not specific only to FAK (54, 55). Although inhibitors and genetic manipulation may have different biological effects, our work predicts that inhibitors that target FAK-Y397 phosphorylation are likely to be effective antiangiogenic drugs.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: A.-R. Pedrosa, N. Bodrug, K.M. Hodivala-Dilke
Development of methodology: A.-R. Pedrosa, N. Bodrug, J. Gomez-Escudero, D.M. Lees, V. Kostourou, A.N. Alexopoulou, S. Batista, B. Tavora
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.-R. Pedrosa, N. Bodrug, J. Gomez-Escudero, E.P. Carter, P.N. Georgiou, I. Fernandez, V. Kostourou, B. Serrels, M. Parsons
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.-R. Pedrosa, J. Gomez-Escudero, E.P. Carter, L.E. Reynolds, P.N. Georgiou, B. Serrels, M. Parsons, T. Iskratsch
Writing, review, and/or revision of the manuscript: A.-R. Pedrosa, N. Bodrug, J. Gomez-Escudero, E.P. Carter, L.E. Reynolds, D.M. Lees, M. Parsons, T. Iskratsch, K.M. Hodivala-Dilke
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.-R. Pedrosa, D.M. Lees
Study supervision: A.-R. Pedrosa, K.M. Hodivala-Dilke
We would like to thank Julie Holdsworth, Bruce Williams, Gabriela D’Amico for their technical help, George Elia from Histopathology and Linda Hammond from Microscopy Barts-CRUK-Centre core facilities, and David Schlaepfer (Moores UCSD Cancer Center in La Jolla, CA) for providing the p190RhoGEF antibody. The host laboratory's work was sponsored by Cancer Research UK (C8218/A21453 and C8218/A18673) and Worldwide Cancer Research (16-0390 2016). Work in M. Parsons’ lab was funded by the MRC (MR/K015664/1 and MR/M018512/1). Work in T. Iskratsch's lab was funded by a British Heart Foundation Intermediate Basic Science Research Fellowship (FS/14/30/30917) and a BBSRC New Investigator Award (BB/S001123/1).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.