Calcium-binding protein S100A6 interaction with VEGF receptors integrates signaling and trafficking pathways

The mammalian endothelium which lines all blood vessels responds to soluble factors which control vascular development and sprouting. Endothelial cells bind to vascular endothelial growth factor A via two different receptor tyrosine kinases (VEGFR1, VEGFR2) which regulate such cellular responses. The integration of VEGFR signal transduction and membrane trafficking is not well understood. Here, we used a yeast-based membrane protein screen to identify VEGFR-interacting factor(s) which modulate endothelial cell function. By screening a human endothelial cDNA library, we identified a calcium-binding protein, S100A6, which can interact with either VEGFR. We found that S100A6 binds in a calcium-dependent manner to either VEGFR1 or VEGFR2. S100A6 binding was mapped to the VEGFR2 tyrosine kinase domain. Depletion of S100A6 impacts on VEGF-A-regulated signaling through the canonical mitogen-activated protein kinase (MAPK) pathway. Furthermore, S100A6 depletion caused contrasting effects on biosynthetic VEGFR delivery to the plasma membrane. Co-distribution of S100A6 and VEGFRs on tubular profiles suggest the presence of transport carriers that facilitate VEGFR trafficking. We propose a mechanism whereby S100A6 acts as a calcium-regulated switch which facilitates biosynthetic VEGFR trafficking from the TGN-to-plasma membrane. VEGFR-S100A6 interactions thus enable integration of signaling and trafficking pathways in controlling the endothelial response to VEGF-A.


Introduction 41
Receptor tyrosine kinases (RTKs) are integral membrane proteins and enzymes which regulate 42 essential features of cell, organ, tissue and animal function (Lemmon et al., 2016). RTK

50
An archetypal RTK is a Type I membrane protein with a glycosylated, extracellular N-51 terminus which is used to 'sense' exogenous soluble and membrane-bound ligands (Lemmon and

58
Although many RTK phospho-substrates have been identified with specific roles in different 59 aspects of cellular physiology, there is no mechanism to adequately explain how RTK trafficking 60 is regulated in resting and ligand-stimulated conditions to meter RTK bioavailability.  (Stoletov et al., 2001), Shb, Sck, SHP-1, and p66Shc (Simons et al., 2016). VEGFR2 binding to epsin 71 (Rahman et al., 2016) and synectin (Lanahan et al., 2013;Salikhova et al., 2008) suggests that 72 interactions with endocytic regulators facilitates VEGFR2 internalization and delivery to 73 endosomes. However, we were still lacking a mechanism to explain how VEGFR trafficking 74 controls receptor bioavailability for exogenous VEGF-A ligand.

75
In this study, we explored the idea that VEGFR interaction with novel cytosolic factor(s) 76 facilitates integration of signaling and trafficking pathways. We employed a membrane protein-77 based genetic screen to identify VEGFR-interacting cytosolic factors. One such protein was 78 S100A6, a calcium-binding protein which binds both VEGFR1 and VEGFR2. Calcium-dependent 79 S100A6 binding to VEGFR1 and VEGFR2 regulates membrane trafficking and VEGF-A-regulated Results 84 S100A6 identified as a VEGFR2-binding protein using a membrane Y2H screen 85 There has been a lack of genetic screens using a native VEGFR membrane protein to identify new 86 binding partners and potential regulators. To address this, we used the split ubiquitin membrane 87 yeast two-hybrid (Y2H) system (Johnsson and Varshavsky, 1994;Stagljar et al., 1998) which 88 enables the use of native membrane proteins as 'baits' to screen genome-wide libraries. In this 89 membrane Y2H system, a split ubiquitin polypeptide and the LexA-VP16 transactivator were 90 used to control nuclear yeast gene expression ( Figure 1A). The 'bait' and 'prey' are tagged with 91 different halves of the ubiquitin molecule, which when brought together in the cytosol, allows 92 activation of a cytosolic ubiquitin-specific protease which cleaves the LexA-VP16 transactivator 93 from the hybrid prey ( Figure 1A). Cleaved LexA-VP16 can now translocate into the yeast nucleus, 94 and stimulates auxotrophic gene expression to enable cell growth on defined media ( Figure 1A).

95
In this study, we fused the complete human VEGFR2 coding sequence to Cub and LexA-VP16 to 96 form a 'bait' hybrid protein to screen for binding to interacting factors (protein X) fused to the 97 NubG 'prey' hybrid protein ( Figure 1B).

98
We then assessed the expression of VEGFR2 hybrid proteins using either the 'prey' or 'bait' 99 plasmid vectors in transformed yeast cells ( Figure 1C). Yeast expression of either VEGFR2-Cub-

100
LexA-VP16 or VEGFR2-NubG revealed high molecular weight bands ~200-250 kDa 101 corresponding to the predicted size of hybrid proteins ( Figure 1C). Probing yeast cells expressing 102 Alg5-LexA-VP16 or VEGFR2-LexA-VP16 hybrid proteins using anti-VP16 antibodies again 103 detected bands of expected sizes ( Figure 1D). A positive control PLC1-SH2 domain (known to 104 interact with VEGFR2), with an engineered HA tag was fused to NubG, expressed in yeast cells, 105 revealing a hybrid protein of expected size ( Figure 1E). This PLC1-SH2-NubG prey construct 106 could now be used as a positive control in subsequent yeast genetic screens.

137
We constructed and screened a 'prey' cDNA library of human endothelial proteins fused 138 to NubG-LexA-VP16 (see Materials and Methods). From this screen, we identified a calcium-139 binding protein S100A6 as a potential binding partner for VEGFR2. The S100A6 prey plasmid 140 construct was isolated, re-transformed into yeast cells and compared to a range of controls under 141 defined growth conditions (Supplement Table 1). The S100A6 prey showed yeast growth when co-142 expressed with VEGFR2 bait (Supplement Table 1), indicating protein-protein interactions between 143 the two molecules. LacZ activity assay showed that co-expression of VEGFR2 bait and S100A6 144 prey caused a 5-fold rise in LacZ activity (Supplement Table 1), indicating protein-protein 145 interactions between VEGFR2 and S100A6. Biochemical analysis of S100A6-NubG in yeast cells -5-revealed a fusion protein of the expected size that contained an engineered HA tag ( Figure S2A) 147 and cross-reactive with anti-human S100A6 antibodies ( Figure S2B). Probing human endothelial 148 cells with anti-S100A6 antibodies revealed a low molecular weight band of ~10 kDa ( Figure S2C). including using the transferrin receptor as a control ( Figure 2A). As expected, VEGFR2 complexes 156 contained S100A6; surprisingly, VEGFR1 complexes also contained S100A6 (Figure 2A).

157
Immunoisolation of S100A6 complexes from endothelial cells revealed the presence of both 158 VEGFR1 and VEGFR2 (Figure 2A). Another membrane protein, the transferrin receptor, was 159 absent from immunoisolated complexes of VEGFR or S100A6 (Figure 2A), indicating specificity 160 in the interaction between S100A6 and VEGFRs. 161 S100A6 belongs to a family of relatively small (~10 kDa) proteins which undergo calcium-  Figure 2B). We investigated the biochemistry of VEGFR2-S100A6 interactions:

171
VEGFR2 cytoplasmic domain bound to S100A6 in the presence of calcium ions ( Figure 2A).

172
VEGFR2-S100A6 complex formation was blocked in the presence of EGTA with no evidence for 173 VEGFR2 association with immobilized S100A6 ( Figure 2B). We then evaluated requirement for 174 VEGFR2 tyrosine phosphorylation in binding to S100A6: de-phosphorylated VEGFR2 still bound 175 immobilized S100A6 in the presence of calcium ions similar to phosphorylated VEGFR2 ( Figure   176 2B). Such VEGFR2-S100A6 interactions are thus calcium-dependent but do not require VEGFR2 177 phosphorylation.

178
As our data suggested that the VEGFR2 cytoplasmic domain binds to S100A6 (Figure 1), we 179 asked whether the VEGFR1 cytoplasmic domain protein could also bind to S100A6 ( Figure 2B).

180
This VEGFR1 cytoplasmic domain protein bound to immobilized S100A6 in the presence of 181 calcium ions ( Figure 2C). Again, VEGFR1-S100A6 complex formation was inhibited by the 182 addition of EGTA ( Figure 2C). To investigate this further, we used the membrane Y2H system to 183 assess whether VEGFR1 displayed interaction with S100A6 (Supplement Table 1). Yeast cells co-184 expressing the VEGFR1-Cub-LexA-VP16 'bait' and the S100A6-Nub-G 'prey' showed auxotrophic 185 growth and LacZ expression (Supplement Table 1). These findings are consistent with calcium-186 dependent VEGFR1-S100A6 complex formation.

207
Biochemistry of VEGFR2-S100A6 interactions 208 The interaction between two molecules can be described by biochemical parameters. We explored 209 the interactions between S100A6 and VEGFR2 using different assays ( Figure 3). First, surface 210 plasmon resonance (SPR) was used to measure S100A6 binding to the VEGFR2 cytoplasmic 211 domain in the presence of calcium ions ( Figure 3A). SPR data showed that titration of S100A6 212 displayed dose-dependent kinetics of binding to immobilized VEGFR2 in the presence of calcium 213 ions ( Figure 3A). Based on these SPR data, the VEGFR2-S100A6 dissociation constant (Kd) was 214 calculated to be ~0.2 M. S100A6 binding to immobilized VEGFR2 was abolished in the presence 215 of EGTA ( Figure 3A).

250
Depletion of S100A6 caused a significant rise in overall VEGFR2 levels but this did not affect 251 other membrane proteins including VEGFR1 or VEGF co-receptors, the neuropilins (NRP1, 252 NRP2) ( Figure 4A). In contrast, depletion of S100A10 did not significantly affect VEGFR or control 253 membrane protein levels ( Figure 4A). Quantification showed that S100A6 depletion caused ~80% 254 rise in total VEGFR2 levels, with little or no significant effects on membrane-bound or soluble 255 VEGFR1 levels ( Figure 4B). We used cell surface biotinylation to assess plasma membrane VEGFR 256 pools under these conditions ( Figure 4C). Increased mature plasma membrane VEGFR2 levels 257 were detected upon S100A6 knockdown ( Figure 4C). Quantification showed ~2.5-fold increase in 258 mature plasma membrane VEGFR2 levels compared to controls ( Figure 4D). Under basal or 259 resting conditions, S100A6-depleted cells showed no significant change in cell surface membrane-260 bound or soluble VEGFR1 compared to controls ( Figure 4D).

268
Endothelial cells subjected to control, scrambled, S100A6 or S100A10 siRNA treatments followed by cell subjected to control, scrambled or S100A6 siRNA treatments followed by VEGF-A (10 ng/ml) stimulation, 276 followed by cell surface biotinylation, cell lysis and purification of biotinylated proteins before 277 immunoblotting. Color coding indicates non-transfected (red), scrambled siRNA (blue) and S100A6 278 knockdown (purple).

280
A previous study showed that biosynthetic VEGFR1 undergoes VEGF-A-stimulated and 281 calcium-dependent trafficking from the distal Golgi to the plasma membrane (Mittar et al., 2009).

282
One possibility was that S100A6 is involved in this calcium-dependent Golgi-to-plasma 283 membrane trafficking step. To test this idea, we used cell surface biotinylation of S100A6-284 depleted cells to assess VEGFR1 plasma membrane levels in resting or VEGF-A-stimulated cells 285 ( Figure 4E). Upon VEGF-A stimulation, we detected a time-dependent, ~2.5-fold increase in 286 plasma membrane VEGFR1 levels over a 60 min time period ( Figure 4E). However, S100A6 287 knockdown caused a complete block in VEGF-A-stimulated VEGFR trafficking to the plasma 288 membrane ( Figure 4E). Depletion of S100A6, but not S100A10, thus modulates both VEGFR1 and 289 VEGFR2 trafficking and plasma membrane levels.

314
We also carried out confocal microscopy to ascertain subcellular VEGFR and S100A6 315 localization ( Figure 6A, 6B). Steady-state VEGFR2 distribution shows localization to the plasma 316 membrane, endosomes and juxtanuclear Golgi region ( Figure 6A). In contrast, S100A6 is widely 317 distributed in the cytosol and nucleus, with occasional staining of tubular profiles emanating 318 from the Golgi region ( Figure 6A, short arrows). Overlay images suggests co-distribution and 319 close proximity of VEGFR2 and S100A6 ( Figure 6A, boxed). Analysis of the VEGFR1 and S100A6 320 ( Figure 6B) showed also showed co-distribution in the Golgi region ( Figure 6B, long arrows). Co-321 labeling of VEGFR1 and S100A6 elongated tubular profiles emanating from the Golgi region was 322 also detected ( Figure 6B, boxed).

330
We then evaluated the 3-D structures of the S100A6 and VEGFR2 cytoplasmic domain using 331 in silico modelling (Figure 7). Comparison of the structures of free and calcium-bound S100A6 332 shows relatively large movements of helix H3 ( Figure 7A). In silico docking studies using the 333 VEGFR2 tyrosine kinase domain suggests that calcium-bound S100A6 binds close to the cleft of 334 the tyrosine kinase module ( Figure 7B). It is unclear whether VEGFR tyrosine kinase activity and 335 calcium-S100A6 binding are functionally coupled.

337
Discussion 338 How does a cell integrate membrane receptor bioavailability for a specific ligand? In the case of 339 VEGF-A binding to two different membrane receptors, VEGFR1 and VEGFR2, different 340 pathways of signaling, trafficking and turnover need to be integrated to control cellular responses 341 such as cell migration, proliferation and tubulogenesis. Up to now, we lacked molecules that 342 could bridge VEGFR signaling and trafficking. Herein, we now present evidence that a calcium-343 dependent cytosolic protein, S100A6, binds both VEGFR1 and VEGFR2 to integrate signaling and 344 trafficking pathways. Five lines of evidence support this conclusion. Firstly, a genetic screen of 345 human endothelial proteins identified S100A6 as a binding partner for VEGFR2. Second, 346 membrane Y2H assay shows that either VEGFR2 or VEGFR1 can interact with S100A6. This was 347 confirmed by the detection of stable complexes of S100A6 with either VEGFR1 or VEGFR2 in 348 endothelial cells. Third, S100A6 binds the VEGFR2 cytoplasmic domain in vitro, with sub-349 micromolar binding affinity (Kd) and displays calcium-dependence. Furthermore, the VEGFR1 350 cytoplasmic domain also binds to S100A6 in a calcium-dependent manner. S100A6 binding to 351 VEGFR2 maps to the tyrosine kinase module. Fourth, S100A6 regulates VEGFR1 and VEGFR2 352 trafficking with different functional outcomes. Whereas S100A6 depletion causes dysregulated -11-VEGFR2 trafficking and increased plasma membrane levels, loss of S100A6 completely blocks 354 VEGFR1 Golgi-to-plasma membrane trafficking. Finally, S100A6 influences VEGFR2 plasma

373
Our study supports a mechanism where at least two trafficking routes regulate VEGFR

432
Deletion analysis of the VEGFR2 cytoplasmic domain maps calcium-S100A6 binding to the 433 tyrosine kinase module. In this context, de-phosphorylation of the VEGFR2 cytoplasmic domain -13-at Y1175 (within the carboxy-proximal tail region) does not significantly affect VEGFR2-S100A6

443
Our study provides a new mechanism where newly synthesized membrane cargo trafficking 444 to the plasma membrane is dependent on integration with signal transduction pathways. Here, 445 activation of plasma membrane receptors which trigger rise in second messenger levels such as 446 calcium ions causes conformational changes in S100A6 to enable binding to VEGFRs, and cargo 447 selection for TGN-to-plasma membrane trafficking. Importantly, our findings also provide a 448 mechanistic explanation for how trafficking and secretion of newly synthesized membrane cargo 449 is synergized with plasma membrane signaling for replenishment of membrane receptors. In this 450 context, there is increasing evidence that S100 protein family members e.g. S100A10, regulates

464
One likelihood is that calcium-stimulated VEGFR1 arrival at the cell surface (and higher affinity

585
Quantification and statistical analysis

591
Data availability

592
The study did not generate large datasets. All necessary data is included in the figures in this 593 study.