Functional display of bioactive peptides on the vGFP scaffold

Grafting bioactive peptides into recipient protein scaffolds can often increase their activities by conferring enhanced stability and cellular longevity. Here, we describe use of vGFP as a novel scaffold to display peptides. vGFP comprises GFP fused to a bound high affinity Enhancer nanobody that potentiates its fluorescence. We show that peptides inserted into the linker region between GFP and the Enhancer are correctly displayed for on-target interaction, both in vitro and in live cells by pull-down, measurement of target inhibition and imaging analyses. This is further confirmed by structural studies highlighting the optimal display of a vGFP-displayed peptide bound to Mdm2, the key negative regulator of p53 that is often overexpressed in cancer. We also demonstrate a potential biosensing application of the vGFP scaffold by showing target-dependent modulation of intrinsic fluorescence. vGFP is relatively thermostable, well-expressed and inherently fluorescent. These properties make it a useful scaffold to add to the existing tool box for displaying peptides that can disrupt clinically relevant protein–protein interactions.

In vitro biosensing of Mdm2 by vGFP variants. p53 and Mdm2 have been used as a model interacting protein pair in the development of several biosensing technologies [20][21][22][23][24][25] . We have previously described a biosensor comprising β-lactamase coupled to an inhibitory protein (BLIP) in cis via a linker containing the M2 peptide sequence 26,27 . Upon binding to Mdm2, the normally disordered M2 peptide adopted an α-helical conformation, effectively shortening linker length and allosterically displacing BLIP and restoring β-lactamase activity. To test this allosteric derepression paradigm in vGFP-M2 we measured its fluorescence in the presence of Mdm2. No  www.nature.com/scientificreports/ significant drop in fluorescence was seen ( Fig. 2A), suggesting that any induced change in M2 conformation upon Mdm2 binding could not propagate sufficiently to displace the Enhancer from sfGFP. Crystallographic analysis of the complex confirmed this to be the case (see below). To improve the likelihood of displacing the Enhancer from sfGFP upon Mmd2 binding we shortened the linker region by removing either one or two residues from the disordered C-terminus of sfGFP (vGFP-M2.1 and vGFP-M2.2 respectively) ( Table 1). However, in the absence of Mdm2 notable reduction in GFP fluorescence was observed when compared to vGFP-M2 (15% for vGFP-M2.1 and 29% for vGFP-M2.2) (Fig. 2B). This suggested that even minimal reduction in linker length significantly reduced the Enhancer-GFP interaction in cis, thereby limiting biosensing utility through measurement of fluorescence changes upon Mdm2 binding. We therefore mutated 4 key residues in the Enhancer domain contributing towards its strong interaction with sfGFP to reduce its affinity 14 . This variant (vGFPE4-M2: S273A, R275A, S299A and F342A) maintained  www.nature.com/scientificreports/ interaction with Mdm2 in a pull-down assay (Fig. 1). In the absence of Mdm2, vGFPE4-M2 fluorescence was the same as for vGFP-M2, indicating that the mutations did not abrogate fluorescence enhancement by the Enhancer. Incubation of vGFPE4-M2 with Mdm2 did not however result in any fluorescence change ( Fig. 2A). We reasoned that differences in the intramolecular affinity of the Enhancer for sfGFP would become more evident at higher temperatures. Thermal melt curve analysis was therefore carried out on vGFP-M2, vGFP-M2C and vGFPE4-M2, using the intrinsic sfGFP fluorescence to report on protein stability and conformation. All three proteins showed a higher temperature melt peak (T m 2, ~ 88 °C) corresponding to denaturation of the GFP chromophore ( Fig. 3, Supplementary Fig. S2). Notably, a clearly distinct lower temperature melt peak (T m 1, ~ 63 °C) was only observed for vGFPE4-M2. This is likely due to thermally induced intramolecular dissociation of the mutated Enhancer from GFP, resulting in decreased fluorescence. The corresponding event in the vGFP-M2/M2C proteins takes place at higher temperatures, represented by less distinct peaks at ~ 73 °C and ~ 78 °C respectively. Co-incubation with Mdm2 resulted in a clear downward shift of T m 2 for vGFP-M2 and both T m 1 and T m 2 for vGFPE4-M2 (Fig. 3B,C, Supplementary Fig. S2). The melt profile of the non Mdm2-binding vGFP-M2C protein showed less difference in the presence of Mdm2 (Fig. 3A, Supplementary Fig. S2). Incubation of vGFPE4-M2 with a non-binding protein control (amidase) did not alter the melt profile, confirming an Mdm2-specific phenotype (Fig. 4A). The Mdm2-dependent shifts were further down-shifted in the presence of DMSO (10% v/v), an www.nature.com/scientificreports/ additive solvent that can modulate protein structure and interactions 28 (Fig. 4B). Binding specificity was further validated using Nutlin 3a, a small molecule Mdm2 inhibitor that competes for binding to the same region of Mdm2 targeted by the scaffolded M2 peptide 29 . In the presence of Nutlin 3a the T m 1 peak drifted towards that of vGFPE4-M2 alone. Intriguingly, the T m 2 peak remained unchanged (Fig. 4C). The control Nutlin 3b enantiomer (with ~ 150-fold reduced binding affinity for Mdm2) did not have any notable effect on either melt peak. Nutlin 3a or 3b alone did not significantly affect the vGFPE4-M2 melt profile ( Supplementary Fig. S3). Whilst these results indicate the potential for biosensing using vGFP, further engineering will be required to improve both sensitivity and robustness. To this end, we next determined the crystal structure of vGFP-M2 bound to the Mdm2 N-terminal domain (residues 6-125). A single binary complex was present in the asymmetric unit, with the M2 peptide linker presented as an α-helix projecting three key residue sidechains (Phe, Trp and Leu) into a hydrophobic groove in Mdm2 (Fig. 5A). Both M2 and Mdm2 showed high similarity to the structure of the parental p53 peptide (from which M2 is derived) bound to Mdm2 (Cα RMSD = 0.43 Å) 19 . A notable difference is a complete one turn extension of the interacting helix in the vGFP-M2 structure comprising the Enhancer residues Val 242, Gln 243, Leu 244 and Val 245 (Fig. 5B). This confirmation is stabilised by the Leu 244 side chain projecting into Mdm2 to more fully occupy its hydrophobic groove.
Comparison with the binary GFP-Enhancer structure 14 shows highly similar conformations of the GFP and the Enhancer components in vGFP-M2-Mdm2 (Cα RMSD = 0.43 Å). A striking deviation is however seen for residues 1-10 of Enhancer in the two structures (Fig. 5C). In the binary complex, these residues comprise β-strand 1 in the highly conserved β sheet framework observed in variable-domain immunoglobulin folds 30 . In the vGFPE4-M2-Mdm2 complex this region is displaced by up to 30 Å, with the first four amino acids forming part of the extended M2 helix as discussed above. The remaining residues act as a linker connecting to the otherwise structurally conserved Enhancer fold. The integrity of the complex, despite this large translocation attests to the high affinity of the Enhancer for GFP (590 pM) 14 that is further improved when the two are connected in cis.
Intracellular targeting using vGFP-scaffolded peptides. Next, the intracellular targeting and activity of vGFP-scaffolded peptides was investigated. Inhibition of the p53-Mdm2 protein-protein interaction was measured by transfection of vGFP constructs into T22 cells. This cell line harbours a stably integrated p53-driven β-galactosidase gene that reports on elevated p53 levels arising from Mdm2 inhibition 31,32 . In addition to vGFP-M2 we also tested a construct with an extended linker sequence added C-terminal to the M2 peptide (vGFP2-M2) ( Table 1), designed to relieve deformation of the Enhancer domain seen in the Mmd2-bound structure of vGFP-M2 that was designed for biosensing (Fig. 5A). The positive control inhibitor (Nutlin 3a) gave the expected increase of p53 activity in the assay (Fig. 6). Expression of both vGFP-M2 and vGFP2-M2 resulted in clear p53 activation above levels observed for the corresponding vGFP-M2C and vGFP2-M2C controls (Fig. 6A). Activity of vGFP2-M2 was higher than that of vGFP-M2 (41% versus 32% of activity seen for Nutlin 3a positive control). The vGFP-M2-Mdm2 structure indicated that the intramolecularly displaced Enhancer residues 242-245 effectively lengthened the α helical Mdm2-binding interface (Fig. 5B). We therefore made a construct incorporating an additional copy of Enhancer residues 242-246 after the M2 peptide sequence in vGFP2-M2 (vGFP3-M2, Table 1) and measured activity in T22 cells. This iteration notably increased activity over the starting vGFP-M2 construct (75% versus 28% activity seen for Nutlin 3a positive control) (Fig. 6B).
We further studied vGFP robustness by using it to scaffold a 14 amino acid peptide (e4pep) that binds to the translational initiation factor eIF4E 34 . Targeting constructs (vGFP-e4pep and vGFP-GSe4pep) were transfected into HEK293 cells and assayed for eIF4E engagement by pull-down from lysate using m7GTP beads. vGFP-GSe4pep comprises an additional Gly-Ser motif in the linker region ( Table 1). The m7GTP beads specifically bind eIF4E via its cap analog interface, enabling its purification along with any bound proteins. After pull-down, bound vGFP constructs were detected by Western blot using anti-GFP antibody. The results indicate clear binding to eIF4E by the vGFP-e4pep/vGFP-GSe4pep proteins, but not the respective vGFP-e4pepC/vGFP-GSe4pepC controls (Table 1) displaying a non-binding variant of e4pep (Fig. 8).

Discussion
We have described use of vGFP as a scaffold to present different peptides targeting Mdm2 and eIF4E. Disruption of the p53-Mdm2 protein-protein interaction by vGFP-M2 was observed, leading to increased p53 activity in cells. Structural analysis of vGFP-M2 bound to Mdm2 show the scaffolded M2 peptide adopting the optimal α-helical binding conformation seen in both linear and chemically scaffolded (i.e. stapled) Mdm2 binding peptides 19,35,36 . Furthermore, the structure revealed an extension of the M2 peptide α-helix that interfaces with Mdm2. This was achieved by co-opting the first 4 residues from the N-terminus of the fused Enhancer domain that normally adopt a β-strand conformation. Incorporation of these residues into a vGFP variant designed to reduce deformation upon Mdm2 binding (vGFP3-M2) enhanced cellular activity ~ twofold. It will be interesting to see if this modification enhances Mdm2-targeting stapled peptides being developed for clinical applications 18,37,38 .
Melt curve analysis of vGFP-M2/M2C highlighted moderate thermostability of the vGFP scaffold. Fluorescence of vGFP-M2C and vGFP-M2 respectively started to diminish at 75 °C and 66 °C, clearly highlighting a destabilizing influence of the hydrophobic 'FWL' signature triad present in M2 (Table 1) and high-affinity Mdm-2 binding peptides retaining this motif. The intrinsic fluorescence of vGFP-M2 is also lower than vGFP-M2C (Fig. 1B), further suggesting a destabilizing feature of this peptide. We previously scaffolded the M2 peptide into a bacterial copper oxidase and noted an inhibition of enzyme activity not seen for the control M2C peptide 1 , further highlighting unusual properties of the hydrophobic 'FWL' triad. We anticipate vGFP thermostability will prevail upon insertion of other bioactive peptides. When coupled with the relatively high expression yields ( ≥ 10 mg per litre in E. coli) and intrinsic high fluorescence, these features make vGFP a strong candidate to add to the scaffolding tool box.    [39][40][41][42][43][44] . However, peptide insertion has been reported to destabilise the scaffold, reducing fluorescence and/or causing solubility issues in some cases. The vGFP scaffold potentially mitigates this liability by not directly interfering with the GFP fold and allowing for improved fluorescence/stability due to intramolecular interaction with the Enhancer. However, as shown for the M2 peptide, indirect destabilization can still occur. An interesting possibility would be to increase valency and sample novel display topology space by co-opting a known permissive insertion site in the GFP component of vGFP.
Our efforts at developing a vGFP-based biosensor based on target-dependent intramolecular dissociation of GFP and the Enhancer were met with limited success. The structure of the vGFP-M2-Mdm2 complex revealed the GFP-Enhancer interaction was concomitant with significant disruption of the conformation of β-strand 1 in the www.nature.com/scientificreports/ Enhancer. Despite introducing four mutations into the Enhancer to reduce affinity for GFP, specific interaction with analyte (Mdm2) could only be discerned in vitro using fluorescence melt curve analysis. Therefore, further rational engineering and/or directed evolution of the GFP-Enhancer binding interface may yield a sensor that works in living cells. Future design iterations could also include use of the companion fluorescence-quenching Minimizer nanobody 14 to improve signal to noise read-outs.

Methods
Constructs. The vGFP cassette from pANT-9L (a kind gift from Dr Swaine Chen) was excised using NdeI and HindIII and ligated into cut pET22b (+) vector. Peptide-displaying vGFP constructs for bacterial expression/purification were subsequently made by inverse-PCR mutagenesis of pET22-vGFP using primers indicated in Supplementary Table S1. vGFP-M2 was made using primer pair 1 and 2. vGFP-M2C was made using primers 3 and 4. vGFP-M2.1 and vGFP-M2.2 were made using primer pairs 5/6 and 5/7 respectively. vGFPE4-M2  They were purified using His-GraviTrap columns (GE Healthcare) following manufacturer's protocol. Mdm2 (amino acids 6-125) was cloned as a GST-fusion protein, expressed and purified using affinity chromatography and Resource S cation exchange column as previously described 35 . G. pallidus RAPc8 amidase was expressed and purified as previously described 45 .

P53-reporter gene assay in T22 cells. T22 reporter cells (stably transfected with pRGCd-Fos-lacZ car-
rying a p53-driven β-galactosidase gene) 31 were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% (v/v) fetal calf serum (FCS) and 1% (v/v) penicillin/streptomycin. Transfection was carried out using Lipofectamine 3000 transfection reagent (Thermo Fisher Scientific) according to manufacturer's instructions. For intracellular targeting activity in T22 reporter cells, luciferase expression plasmid was co-transfected with respective vGFP expression constructs at 1:1 ratio, 1.0 μg total plasmid DNA. eIF4E interaction assay in HEK 293 cells. Twenty-four hours prior to transfection, HEK 293 cells (Thermo Fisher Scientific) were seeded at a cell density of 1 × 10 6 per well of a 6-well plate. Transfections in 6-well plates were performed using Lipofectamine 3000 (Thermo Fisher Scientific) with 0.5 μg of plasmid vectors per well according to the manufacturer's instructions. After 48 h, cells were lysed and m7GTP pull down were performed as described 50 . Briefly, 300 µg of cell lysates were incubated with 20 μL of m7GTP (Jena Bioscience) agarose beads for 3 h at 4 °C on a rotator. Beads were then washed four times with lysis buffer and vGFP-eIF4E complexes eluted by boiling the beads for 5 min at 95 °C in presence of Laemmli buffer. Samples were resolved by SDS/PAGE and transferred onto PVDF membrane using the Trans-Blot Turbo system (Bio-Rad) according to the manufacturer's protocol. Blocked membranes were blotted with 3H9 anti GFP antibody (Chromotek) followed by incubaton with the secondary goat anti-rat for 1 h at room temperature. Chemiluminescence was then detected with a Licor system (Licor Biosciences).
Live cell spinning disk confocal fluorescence microscopy. For live cell imaging, T22 cells grown on 27 mm diameter Nunc glass bottom dish (Thermofisher) were transfected 24 h prior to image acquisition. Images were acquired on Nikon Eclipse Ti inverted microscope, using a 40 × oil immersion objective (NA 1.3). Z-stacks were acquired for all images and sum slices projection was applied to all the stack images using ImageJ Software 51 .

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.