Photoinduced reconfiguration to control the protein-binding affinity of azobenzene-cyclized peptides

Abstract

The impact of next-generation biorecognition elements (ligands) will be determined by the ability to remotely control their binding activity for a target biomolecule in complex environments. Compared to conventional mechanisms for regulating binding affinity (pH, ionic strength, or chaotropic agents), light provides higher accuracy and rapidity, and is particularly suited for labile targets. In this study, we demonstrate a general method to develop azobenzene-cyclized peptide ligands with light-controlled affinity for target proteins. Light triggers a cis/trans isomerization of the azobenzene, which results in a major structural rearrangement of the cyclic peptide from a non-binding to a binding configuration. Critical to this goal are the ability to achieve efficient photo-isomerization under low light dosage and the temporal stability of both cis and trans isomers. We demonstrated our method by designing photo-switchable peptides targeting vascular cell adhesion marker 1 (VCAM1), a cell marker implicated in stem cell function. Starting from a known VCAM1-binding linear peptide, an ensemble of azobenzene-cyclized variants with selective light-controlled binding were identified by combining in silico design with experimental characterization via spectroscopy and surface plasmon resonance. Variant cyclo_AZOB[G-VHAKQHRN-K] featured rapid, light-controlled binding of VCAM1 (K_D,trans/K_D,cis ~ 130). Biotin-cyclo_AZOB[G-VHAKQHRN-K] was utilized to label brain microvascular endothelial cells (BMECs), showing co-localization with anti-VCAM1 antibodies in cis configuration and negligible binding in trans configuration.

Introduction

Inducible affinity interactions between ligand and target biomolecules are the underlying functions governing biological systems. The ability to design synthetic ligands whose biorecognition can be activated and controlled using remote and biocompatible stimuli is key to engineer next-generation biomimetic systems. Mainstream engineered ligands, like antibodies and aptamers,^1-4 enable sensitive detection and sorting of biological targets (e.g., proteins, viruses, and cells) in complex media.^5,6 Their biorecognition activity, while strong and selective, is innate and cannot be activated on demand. Adjusting the composition, concentration, and pH of the aqueous environment provides some control over binding strength, but often at the expense of the bioactivity and viability of the target. The ability to design synthetic ligands whose affinity for a target can be activated rapidly and remotely, using external stimuli that do not adversely impact the integrity of the target, is a much sought-after goal in modern ligand engineering.

Peptides represent ideal scaffolds for developing dynamically regulated ligands, owing to their excellent biorecognition activity, modular assembly, and affordable manufacturing at large scale with no batch-to-batch variability.⁷ In particular, peptides with constrained conformation (e.g., cyclic peptides) feature superior target affinity and selectivity – in some instances, on par with antibodies – as well as high biochemical stability.^8-10 Furthermore, cyclic peptides can be engineered to integrate stimuli-responsive moieties that provide remote control over target-binding affinity independently of the physicochemical conditions of the environment. Of particular interest in this regard are photochromic switches, such as azobenzenes, hemithioindigos, and spiropyrans/spirooxazines, which respond to specific light wavelengths and intensities by rearranging their structure. When utilized as peptide cyclization linkers, photochromic switches can reconfigure the peptide structure upon photo-isomerization.^11-18 This mechanism can be harnessed to remotely activate the selective binding of the peptide for a target biomolecule, resulting in adsorption on a substrate or labelling in solution.

In this study, we sought to develop a novel family of azobenzene-cyclized peptide affinity ligands that are selective, sterically inconspicuous, rapidly activated, and thermally stable (note: we adopted azobenzene as photochromic switch owing to its ease of synthesis and commercial availability to ensure the translational potential of this technology). To this end, we have developed a method for converting known linear protein-binding peptides into an azobenzene-cyclized framework (Fig. 1). This comprises a protein-binding peptide segment, the azobenzene cyclization linker, connecting spacers, and a reporter (e.g., a fluorescent dye or biotin). The proposed cyclization geometry provides an optimal balance between structural flexibility and rigidity, which enables rapid and high-yield photo-isomerization under moderate light intensity, and ensures temporal and thermal stability of both trans and cis isomers.

Fig. 1.

Structure of the proposed azobenzene-cyclized peptides.

To demonstrate our method, we chose the Vascular Cell Adhesion Marker 1 (VCAM1) as protein target, and the linear VCAM1-binding peptide VHPKQHR as reference sequence.¹⁹ VCAM1 is a cell surface sialoglycoprotein implicated in directing downstream lineages of human hematopoietic progenitor cells (HPCs).²⁰ VCAM1 negative (VCAM1⁻) HPCs give rise to lymphoid progenitors, while VCAM1⁺ HPCs result in myeloid progenitors, suggesting that VCAM1 expression in HPCs represents a branching point between the lymphoid and myeloid lineages.^21,22 Sorting lineage-committed HPCs into lymphoid (VCAM1^−/low) and myeloid (VCAM1⁺) has the potential to enable stem cell therapies for treating leukemia, lymphoma, cardiac failure, neural disorders, autoimmune diseases, metabolic or genetic disorders. The VCAM1-binding linear peptide has been identified via phage-display screening in apolipoprotein E-deficient mice and has been utilized as a ligand for VCAM1 in numerous applications.^23-26

The design method begins with the in silico analysis of the crystal structure of VCAM1 (PDB IDs: 1VCA) using a “druggability” test to identify binding sites for peptide binding.^27,28 The information on size, structure, and physicochemical properties of the putative binding sites was used to design a set of 25 peptide variants in the form cyclo_AZOB[G-VH(X)KQHR(Z)-K]-GSG (Fig. 1) where (X) and (Z) are interchangeably A, D, N, P, or S. The structures of the azobenzene-cyclized peptides were generated by molecular dynamics (MD)^29,30 and docked in silico^31,32 on VCAM1 to identify leads with conformation-dependent binding. Selected complexes were then refined by MD simulations to select azobenzene-cyclized variants with predicted high binding strength (i.e., either ΔG_B,trans or ΔG_B,cis < −8 kcal mol⁻¹) and loss of binding upon photo-isomerization (i.e., ∣ΔΔG_B∣ = ∣ΔG_B,trans – ΔG_B,cis∣ > 2.5 kcal mol⁻¹).^33,34

Three azobenzene-cyclized peptides selected from in silico screening were characterized to evaluate their (i) VCAM1 binding activity by surface plasmon resonance (SPR), (ii) kinetics of photo-isomerization upon exposure to light by UV/Vis spectroscopy, and (iii) thermal stability of the cis isomers. Variant cyclo_AZOB[G-VHAKQHRN-K] in particular showed efficient photo-isomerization and an ample affinity shift (K_D,trans/K_D,cis ~ 130), which ensured efficient light-controlled binding of VCAM1. Finally, the peptide was fused with biotin and utilized as a light-activated label for brain microvascular endothelial cells (BMECs); VCAM1 expression was induced by a synergistic treatment of Interleukin-4 (IL-4) with lipopolysaccharide (LPS) and confirmed by immunohistochemical staining and RT-qPCR. Cell imaging by fluorescence microscopy confirmed binding of VCAM1 on BMECs by the cis isomer of cyclo_AZOB[G-VHAKQHRN-K], which afforded fluorescent labelling of cells with intensity proportional to the cell surface density of VCAM1, while showing negligible binding in trans. These results demonstrate the effectiveness of our design and selection methods for developing peptide ligands whose target-binding affinity can be rapidly activated via light-controlled structural reconfiguration.

Experimental

Materials

N,N′-Dimethylformamide (DMF), dichloromethane (DCM), ethanol, HPLC-grade acetonitrile, HPLC-grade water, and endogenous biotin blocking kit were from ThermoFisher Scientific (Waltham, MA). Protected amino acids, piperidine, trifluoroacetic acid (TFA), diisopropylethylamine (DIPEA), Rink amide resin (100–200 mesh, functional density ~0.6 mmol g⁻¹), and Hexafluorophosphate Azabenzotriazole Tetramethyl Uronium (HATU) were purchased from ChemImpex Inc. (Wood Dale, IL). Purified cyclo_AZOB-[G-VHAKQHRN-K]-K(Biotin) was sourced from the peptide synthesis facility at UNC Chapel Hill. Bovine serum albumin, Alexa Fluor 488-labeled streptavidin (AF488-streptavidin), DAPI nuclear stain, anti-human VCAM1 (CD106) rabbit monoclonal antibody and polyclonal goat anti-rabbit antibody, azobenzene-4,4′-dicarbonyl dichloride, glycine, glacial acetic acid, diethyl ether, triethylamine (TEA), triisopropylsilane (TIPS), ethane-dithiol (EDT), Tween 20, 1 M aqueous NaOH, phosphate-buffered saline (PBS) pH 7.4, acetic anhydride, Kaiser test kit, and were from Millipore Sigma (St. Louis, MO). Azido-PEG-thiol (N₃-PEG-SH, MW = 600 Da) was from Nanocs Inc. (New York, NY), while amine-PEG-thiol (NH₂-PEG-SH, MW = 2000 Da) and hydroxyl PEG thiol (OH-PEG-SH, MW = 1000 Da) were from Creative PEGWorks (Chapel Hill, NC). Glass sensor chips (12 × 20 × 0.5 mm) sputtered with a 50 nm gold layer were from KSV Instruments OY (Helsinki, Finland). Nitrogen gas was obtained from Airgas National Welders (Raleigh, NC). VCAM1 was obtained from SinoBiologicals (Beijing, China). Brain microvascular endothelial cells (BMECs) were obtained from ATCC (Manassas, VA).

Peptide synthesis and characterization

Synthesis.

The linear peptide precursors VHGKQHRP-K*, G-VHAKQHRN-K*-Prg, G-VHAKQHRP-K*-Prg, G-VHNKQHRP-K*-Prg, G-VHPKQHRS-K*-Prg, G-VHNKQHRS-K*-Prg, G-VHPKQHRP-K*-Prg, and VHPKQHR-GSG-Prg (Prg: propargyl-glycine) were synthesized on Rink amide polystyrene resins via Fmoc/tBu chemistry using standard protecting groups for all amino acids except K* (Fmoc-Lys(Mtt)-OH).^35-37 All amino acid conjugations were conducted using a Biotage Alstra Initiator (Biotage, Uppsala, Sweden) by performing 3 coupling steps with protected amino acid (5 equivalents), HATU (5 eq.), and DIPEA (10 eq.) at 75 °C for 5 min in DMF. Fmoc removal occurred in 20% piperidine in DMF at RT for 20 min. Upon completion of chain elongation, the Fmoc group on the N-terminus was removed, and the resin was copiously rinsed with DCM and vacuum dried. The dry resin was swollen in DCM (dried over molecular sieves), cooled to 0 °C, and anhydrous TEA (1.2 eq.) and azobenzene-4,4′-dicarbonyl dichloride (1.2 eq. in anhydrous DCM) were added dropwise at 0 °C over 10 min. The system was equilibrated at RT and the reaction was allowed to proceed overnight. The resin was copiously rinsed with DCM and the azobenzene conjugation was confirmed by Kaiser’s test.³⁸ The Mtt protecting group on K* was removed by incubating the resin with 2/5/93 (TFA/TIPS/DCM) for 5 min. Completion of K* deprotection was verified by Kaiser’s test. After rinsing with DCM and DMF, the peptide was cyclized through reaction of the carboxyl group on the azobenzene linker with the ε-amino group of K*, by incubating the resin with HATU (5 eq.) and DIPEA (10 eq.) in DMF for 10 min at 75 °C. Completion of the cyclization reaction was verified by Kaiser’s test. The resin was finally rinsed with DMF, DCM, and dried under nitrogen then incubated for 2 h at RT 10 mL g⁻¹ of 95/2.5/2.5 (TFA/TIPS/water). The peptide was precipitated by drop-wise addition into ice-cold diethyl ether. The precipitate was copiously rinsed with diethyl ether, dissolved in 50/50 (acetonitrile/water), and lyophilized. The crude peptide powder was purified by preparative C18 HPLC and lyophilized.

Peptide photo-isomerization

All spectroscopy measurements were performed using Cary 60 spectrometer equipped with a custom cuvette holder on a Peltier stage to maintain the peptide solutions at 37 °C. Orthogonal to the spectrometer beam, the peptide solution was irradiated with a BlueWave 200 lamp (Dymax, Torrington, CT) to induce photo-isomerization. A volume of 700 μL of cyclo_AZOB[VHGKQHRP-K*] in MilliQ water at either at 0.15 mM, 0.38 mM, or 1.5 mM was initially placed in quartz micro cuvette (Thorlabs, Newton, NJ). The cuvette was placed in the spectrometer with a 10 mm pathlength for UV-Vis spectroscopy and a 2 mm pathlength for irradiance. A UV bandpass filter (BP305-390, Thorlabs, Newton, NJ) was used to achieve trans-to-cis isomerization, whereas a 420 nm longpass (LP420, Edmund Optics, Barrington, NJ) filter was used for cis-to-trans isomerization. Spectra of the lamp and filtered outputs are reported in the ESI.† The incident irradiance power was calculated by measuring the output power using an Accu-Cal 50 radiometer (Dymax, Torrington, CT). The kinetics of photo-isomerization was measured by monitoring the absorption at 350 nm. The absorbance vs. time data was fitted against an exponential function to derive the kinetic constant of photo-isomerization κ (s⁻¹). The values of κ were obtained for light intensities between 10 and 50 mW cm⁻². The rate of thermal reverse isomerization (cis-to-trans) was evaluated by monitoring the absorbance at 350 nm of a 0.15 mM solution of cyclo_AZOB[VHGKQHRP-K*], initially conditioned in cis configuration and allowed to relax in the dark over 60 h with absorption measurements taken every 15 min.

Circular dichroism (CD)

A solution of cyclo_AZOB[VHGKQHRP-K*] at 200 μM in PBS was placed in two 1 mL quartz micro fluorescence cuvettes (Thorlabs, Newton, NJ). The peptides in solution were conditioned to either cis or trans configuration as described in the previous section, and further diluted with PBS to reach a final concentration of 80 μM. The CD spectra of the solutions of cis and trans peptide isomer were measured on a J-715 spectropolarimeter (JASCO, Oklahoma City, OK) through the 10 mm path length of the quartz cell. The samples were scanned from 190 to 260 nm, at a resolution of 1.0 nm and a scanning speed of 100 nm min⁻¹. For each sample, five scans were averaged together and a 5-point moving average was applied.

Binding affinity using surface plasmon resonance (SPR)

Surface plasmon resonance SPR sensor chips were prepared, characterized, and utilized for affinity measures as described by Islam et al.^39,40 The sensor chips were initially coated with a PEG-based self-assembled monolayer (SAM), on which the test azobenzene-cyclized peptides and the linear precursor VHPKQHR were conjugated to the surface via azide-alkyne “click chemistry”;⁴¹ sensor chips coated with hydroxyl-PEG-thiol only were utilized as negative controls. Three gold sensors per ligand were prepared and analyzed by variable angle spectroscopic ellipsometry (VASE, J.A. Woollam Co.) to measure ligand density. The measurements of peptide-VCAM1 dissociation constant (K_D) were performed at the UNC Macromolecular Interactions Facility.^39,40 Briefly, sensor chips functionalized with azobenzene-cyclized peptides were initially conditioned to their cis or trans isomers and contacted with a solution of human VCAM1 at concentrations ranging between 0.01 and 1 mg mL⁻¹ in PBS.

In silico design of VCAM1-binding peptides

The crystal structures of the extracellular domain of VCAM1 (PDB: 1VCA)⁴² were subjected to standard protein preparation using Schrödinger’s ProteinPrep Wizard to search for and correct missing atoms and/or entire side chains (using PRIME software), remove extra salts and non-binding ligands, add explicit hydrogens, assign tautomeric states with EPIK, optimize hydrogen-bonding networks, and minimize the protein’s energy with the OPLS3e force field.^43-45 The adjusted structure was then subjected to a “druggability” study using SiteMap to identify putative binding sites capable of accommodating azobenzene-cyclized peptides.^46,47 To this end, the SiteMap (Schrödinger, New York, NY) algorithm identifies binding pockets by placing spherical ‘site points’ across the protein surface. These site points are then clustered together based on (i) their ability to form favorable protein–ligand or protein–protein interactions, (ii) solvent exposure, and (iii) hydrophobic/philic character. Finally, regions that possess a sufficient number of site points and volume are globally scored using an S-score and D-score. The S-score represents the likelihood of the protein’s surface to be a binding pocket, while the D-score provides a measure of the pocket’s ‘druggability.’ S-score and D-score values greater than or equal to 0.7 and 0.9 respectively were considered favorable pockets for ligand binding. Herein, the SiteMap analysis was performed with the default settings and the ‘detect shallow binding sites’ option selected, which adjusts amino acid atomic van der Waals radii to be more accommodating for peptide binding when locating potential binding pockets.

An ensemble of 25 peptide variants of VHPKQHR in the azobenzene-cyclized form cyclo_AZOB[G-VH(X)KQHR(Z)-K]-GSG were designed by varying the positions (X) and (Z) using amino acids A, D, N, P, S, or R. The structures of the linear precursor VHPKQHR and the azobenzene-cyclized variants as both cis and trans isomers were initially designed using the molecular editor Avogadro, and equilibrated by molecular dynamics in the GROMACS simulation package^29,30,48 using the OPLS all-atom force field^49,50 and periodic boundary conditions.^51-53 The force-field parameters for the azobenzene linker were derived from density functional theory using GAUSSIAN98.⁵¹ Every peptide was individually placed in a simulation box with periodic boundary conditions containing 800 water molecules (TIP3P water model). The solvated system was initially minimized by running 10 000 steps of steepest gradient descent, heated to 300 K in an NVT ensemble for 250 ps with 1 fs time steps, and equilibrated to 1 atm by running a 500 ps NPT simulation with 2 fs time steps. The production run for every peptide was performed in the NPT ensemble, at constant T = 300 K using the Nosé–Hoover thermostat^54-56 and constant P = 1 atm using the Parrinello–Rahman barostat.^57,58 The atomic coordinates were saved every 2 ps. The leap-frog algorithm was used to integrate the equations of motion, with integration steps of 2 fs, and all of the covalent bonds were constrained by means of the LINCS algorithm.⁵⁹ The short-range electrostatic and Lennard-Jones interactions were calculated within a cutoff of 1.0 nm and 1.4 nm, respectively, whereas the particle-mesh Ewald method was utilized to treat the long-range electrostatic interactions.^60,61 The non-bonded interaction pair-list was updated every 5 fs, using a cutoff of 1.4 nm.

The peptides were then docked in silico against the putative binding sites on VCAM1 using the docking software HADDOCK (High Ambiguity Driven Protein–Protein Docking, v.2.1).^31,32 Default HADDOCK parameters (e.g., temperatures for heating/cooling steps, and number of molecular dynamics sets per stage) were used. All the residues on each binding site (solvent accessibility of 50% or greater) were defined as “active”, whereas the residues surrounding the binding sites were defined as “passive”. All variable amino acid positions on the peptide ligands were also denoted as “active”, while the GSG tripeptide spacer was defined as not being involved in the interaction to account for the directionality of binding. Docking proceeded through a 3-stage protocol: (i) rigid, (ii) semi-flexible, and (iii) water-refined fully flexible docking. A total of 1000, 200, and 200 structures were calculated at each stage, respectively. Final structures were grouped using a minimum cluster-size of 20 (10% of the total water refined calculated structures) with a Cα RMSD <7.5 Å using ProFit (http://www.bioinf.org.uk/software/profit/). Once the clusters were identified for each peptide-VCAM1 complex pair, FireDock and XScore were used to score the complexes;^33,34 FireDock is an efficient method re-scoring of protein–protein docking solutions, while Xscore computes the dissociation of a protein–ligand complex using an empirical equation that considers energetic factors in a protein–ligand binding process. The selected binding poses of the peptide variants, in both their cis and trans forms, on the putative binding sites of VCAM1 were then refined via 100 ns atomistic molecular dynamics (MD) simulations using the GROMACS simulation package. The peptide-VCAM1 complexes were embedded in a cubic periodic box of 9.7 nm side lengths and solvated with 30 000 TIP3P water molecules. The MD simulations were performed at 300 K and 1 atm using the Amber99SB force field. The MM/GBSA method was used for post-processing of the peptide-VCAM1 complexes derived from MD simulations to estimate the free energy of binding (ΔG_B).^62,63 if complexes had conflicting docking score or MM-GB/SA ranks, then the complex was discarded. If the two rankings agreed, then the complex was saved. The variants that possessed strong binding in either cis or trans configuration (i.e., ΔG_B,trans or ΔG_B,cis < −8 kcal mol⁻¹) and loss of binding upon photo-isomerization (i.e., ∣ΔΔG_B∣ = ∣ΔG_B,trans – ΔG_B,cis∣ > 2.5 kcal mol⁻¹) were selected for experimental analysis.

In vitro evaluation of VCAM1-binding peptides

Cell culture and VCAM1 expression.

Human brain microvascular endothelial cells (BMEC) (cAP-0002; Angioproteomie, Boston, MA, US) were cultured using Microvascular Endothelial Cell Growth Medium 2 (EGM-2 MV) (CC-3202; Lonza, Basel, Switzerland). Human dermal fibroblasts (HDFn) (PCS-201-010; ATCC, Manassas, VA, US) were cultured using Dulbecco’s modified eagle’s medium (DMEM) (10013CV; Corning, Corning, NY, US) supplemented with 10% fetal bovine serum (FBS) (25-514H; Genesee Scientific, Cajon, CA, US) and 1X penicillin-streptomycin (30-002-CI; Corning, Corning, NY, US). Human umbilical vein endothelial cells (HUVEC) (C2519A; Lonza, Basel, Switzerland) were cultured using Endothelial Cell Growth Medium 2 (EGM-2) (CC-3162; Lonza, Basel, Switzerland). Cells were cultured in an incubator maintaining an atmosphere of 5% CO₂ at 37 °C and 100% humidity. A glass bottomed 96 well plate was pretreated with 50 μg of fibronectin (356008; Corning, Corning, NY, US) in deionized water for 1 h. Fibronectin solution was aspirated and 1 × 10⁴ BMECs, HDFns, or HUVECs per well were seeded in 100 μL of EGM-2 MV media, DMEM with 10% FBS, or EGM-2 media, respectively. Cells were allowed to adhere for 24 h. BMECs were treated with three conditions: (i) media only, (ii) 1 μg mL⁻¹ lipopolysaccharide for 24 h, or (iii) 100 ng mL⁻¹ IL-4 for 1 h followed by 23 h of treatment with 1 μg mL⁻¹ lipopolysaccharide. After treatment period, cells were fixed with 4% paraformaldehyde in 1× PBS for 10 min at RT and washed twice with 1× PBS. BMECs were permeabilized with 0.1% Triton X-100 in 1× PBS for 5 min at RT and washed twice with 1× PBS. Cells were then blocked for 1 h with 2% w/v bovine serum albumin (BSA) at RT.

RT-qPCR quantification of VCAM1 expression

Cells were cultured and treated as described in the previous section. To quantify gene expression, cells were lysed in TRK lysis buffer, then RNA from samples were isolated using EZNA isolation kit (Omega Bio-Tek, Norcross, GA). cDNA templates were created from the RNA samples using a Go Script reverse transcriptase kit (Promega, Madison, WI). Primers target genes were determined from PrimerBank from Harvard University and selected for specificity with the NCBI Primer BLAST tool; those primers were: AAGGTGAAGGTCGGAGTCAAC (Forward – GAPDH), GGGGTCATTGATGGCAACAATA (Reverse – GAPDH), GGGAAGATGGTCGTGATCCTT (Forward – VCAM1), and TCTGGGGTGGTCTCGATTTTA (Reverse – VCAM1). cDNA samples were analyzed using SYBR Master Mix (Applied BioSystems, Foster City, CA) with a QuantStudio 3 real time PCR machine (ThermoFisher, Waltham, MA). Relative gene expression of cell-specific genes was calculated by determining the change in VCAM1 expression relative to the housekeeping gene GAPDH, following eqn (1):

$$\Delta \text{CT}=2^{-[{\text{CT}}_{\text{VCAM1}}-{\text{CT}}_{\text{GAPDH}}]}$$ (1)

where CT is the threshold cycle.

Cell-labelling with cyclo_AZOB[G-VHAKQHRN-K*]

Cells were fixed and permeabilized as described above and blocked using the endogenous biotin blocking kit (ThermoFisher, Waltham, MA). First, cells were incubated with 100 μL of the streptavidin “solution A” for 30 min at RT. The cells were then washed thrice with PBS, incubated with 100 μL of biotin solution “solution B” for 30 min RT, washed again, and stored in PBS overnight at 4 °C. The cells were finally stained in three conditions; (i) cyclo_AZOB[G-VHAKQHRN-K*]-K(Biotin) only, (ii) anti-VCAM1 antibody only, or (iii) co-localized with both peptide and antibody.

For peptide-only staining, a solution of cyclo_AZOB[G-VHAKQHRN-K*]-K(Biotin) at 0.1 mg mL⁻¹ in PBS was initially split in two pools: one was conditioned into the cis configuration (VCAM1-binding), while the other was conditioned into the trans configuration (VCAM1-non-binding), as described above. After irradiation, both peptide solutions were diluted with PBS to a final concentration of 50 μg mL⁻¹. A volume of 100 μL of peptide solution was added to the fixed cells and incubated for 2 h at RT in dark. The fixed cells were then washed copiously with PBS, then stained with 100 μL of AF488-streptavidin (2 μg mL⁻¹) in PBS for 35 min at RT, washed with PBS, and stained with 100 μL of 300 nM DAPI in PBS for 5 min at RT.

For antibody-only staining, 100 μL solution of primary anti-VCAM1 rabbit monoclonal antibody (abcam, Cambridge, UK) at 1.75 μg mL⁻¹ in PBS was incubated with the fixed cells overnight at 4 °C. After incubation, the cells were washed with 1× PBS and 100 μL of AF594-labeled anti-rabbit goat polyclonal antibody (Abcam, Cambridge, UK) at 4 μg mL⁻¹ was incubated with the cells for 2 h at RT in the dark. The wells were washed with 1× PBS, and DAPI stained.

For co-localized staining, the fixed cells were stained with cyclo_AZOB[G-VHAKQHRN-K*]-K(Biotin) and subsequently with anti-VCAM1 antibody, as described above.

Cell imaging

Stained cells were imaged using a Zeiss LSM 710 confocal microscope (Carl Zeiss AG, Oberkochen, Germany). The resulting images were processed using ImageJ software. The fluorescence quantification was determined by averaging the raw intensity measurement of 9 images at 20× magnification for each condition. Final values of relative intensity were obtained by normalizing the averaged value to the maximum and minimum measured intensity across all conditions. Error bars are reported as standard error. Statistics were done using a 2-tailed t-test and p values <0.05 (*) were considered statistically significant.

Results and discussion

Azobenzene-cyclized peptide design

Azobenzene linkers have been extensively utilized to endow peptides and small proteins with the ability to reconfigure their structure reversibly upon exposure to light.^64,65 Recently, a novel set of azobenzene linkers have been developed, which feature electron-donating/-withdrawing groups in the ortho and para position to the azo group. These modifications enable photo-isomerization under biocompatible light wavelengths (i.e., red, far-red, and infrared) and tuning of the thermal cis-to-trans isomeric relaxation.^66,67 While the structural and photokinetic properties of azobenzene-constrained peptides and proteins have been extensively studied, both experimentally^68,69 and computationally,^51-53 less effort has been dedicated to the engineering of azobenzene-peptide hybrids with light-controlled biorecognition activity.^70-74 Bellotto et al. have identified streptavidin-targeting light-responsive peptides by screening phage-display libraries;⁷² the selected peptides exhibited a ~3-fold shift in binding strength upon light exposure, with a dissociation constant (K_D) varying between 2.2 μM (trans isomer) and 6.7 ± 2 μM (cis isomer). Using the same peptide cyclization strategy and phage-display library technique, Jafari et al. identified azobenzene-cyclized peptide ligands against streptavidin⁷⁵ with a ~4.5-fold shift in K_D. Using a rational design approach, Woolley et al. modified a leucine zipper DNA-binding protein by crosslinking two cysteine residues using an azobenzene linker;⁷⁴ when the azobenzene is in cis configuration, the protein maintains its α-helical structure and its DNA-binding activity; vice versa, when in trans, the azobenzene disrupts the structure of the protein, resulting in a 20-fold decrease in binding strength.

Following on this work, we sought to develop peptides that feature (i) rapid photo-isomerization kinetics, (ii) a large shift in protein-binding strength between trans and cis isomers, and (iii) high thermal stability in cis configuration. To this end, we have devised an azobenzene-cyclized peptide structure (Fig. 1), constructed by head-to-side-chain cyclization between the peptide N-terminal α-amine group and the ε-amine group on the C-terminal lysine using homobifunctional azobenzene-4-4′-dicarbonyl linker. The length of the protein-binding peptide segment set at 8 amino acids and cyclization through the C-terminal lysine residue were adopted to achieve a balance between chain flexibility, which increases with the number of residues and enables rapid photoisomerization, and chain rigidity, which promotes binding affinity and stability of the isomers. The observations collected in this study indicate that the proposed design indeed affords efficient photo-isomerization under moderate light dose, which translates into a remarkable shift in protein-binding affinity, and a high thermal and temporal stability of the cis isomer.

In silico design of VCAM1-binding peptides

The linear VCAM1-binding peptide VHPKQHR, discovered by phage-display screening,^23-25 was adopted as precursor for the design of azobenzene-cyclized light-responsive variants. This peptide has been demonstrated to target VCAM1 with high affinity and selectivity in numerous works.^25,26 Our method for designing light-responsive variants of VHPKQHR requires the identification of putative binding sites on VCAM1 that are compatible with the size and shape of 8-mer cyclic peptides, and the production of the crystal structure of the trans and cis isomers of the azobenzene-cyclized variants. To this end, the crystal structure of VCAM1 (PDB ID: 1VCA) was evaluated in silico by performing a “druggability” analysis of the protein surface^27,28 using Schrödinger’s SiteMap software.^46,47 Five binding pockets (S1–S5 in Fig. S1, ESI†) with adequate size, shape, physicochemical properties, and solvent exposure were identified as putative binding sites for the azobenzene-cyclized peptides.^76,77 They were ranked according to two scores, namely Site Score and Druggability Score;⁴⁷ the SiteMap outputs for sites S1–S5 are listed in Table S1 (ESI†). All sites are characterized by high pocket exposure (>0.8) and low enclosure scores (<0.4), and are relatively hydrophilic as determined by SiteMap’s balance score. Sites S1–S5, with a S-score > 0.7, were selected as putative binding pockets on VCAM1. The druggability of these sites, assessed with SiteMap’s D-score, featured a D-score > 0.8, supporting their ability to interact with peptide ligand.

We then designed a focused library of 25 variants of the VHPKQHR precursor in the azobenzene-cyclized form cyclo_AZOB[G-VH(X)KQHR(Z)-K]-GSG (Fig. 1) by varying the positions (X) and (Z) with A, D, N, P, or S. These five amino acids were chosen to explore the effect of mild hydrophobicity (A), peptide turn (P), hydrogen bonding (N and S), and incorporation of a negative charge within a positively charged sequence (D). Aromatic amino acids were avoided to ensure water solubility of the selected peptides; sulfur-containing amino acids were also avoided to prevent formation of disulfide bonds and oxidative degradation.

The crystal structures of the linear precursor VHPKQHR and the cis and trans isomers of the azobenzene-cyclized variants were generated by molecular dynamics (MD), and evaluated to determine the values of (i) equivalent hydrodynamic (R_h) radius⁷⁸ and (ii) root-mean-square deviation of the atomic positions (RMSD) between the cis and trans isomers. These values are reported in Fig. 2 together with representative structures of the corresponding peptides. The values of R_h, calculated using HydroPro v.10,^79,80 resulted to be 11.7 Å for the linear VHPKQHR precursor, and varied between ~9 Å and ~18 Å for the azobenzene-cyclized variants, with an average of 14.9 Å for the trans isomers and 12.5 Å for the cis isomers. Most notably, the values of atomic cis-to-trans RMSD fluctuated between ~0.7 Å and ~5.1 Å (average ~3.1 Å), corresponding to a 20–25% size shift upon isomerization. Substantial structural rearrangement is critical towards achieving the needed shift in binding affinity. Finally, several variants (i.e., 5, 12, 13, 16, 21, and 23) featured an energy landscape considerably lower – indicating higher stability – in the trans configuration, whereas variant 25 showed higher stability in the cis configuration. These variants were deemed unable to undergo photo-isomerization and were therefore not considered for the in silico screening against VCAM1.

Fig. 2.

Structure of 25 variants of the VHPKQHR precursor in the azobenzene-cyclized form cyclo_AZOB[G-VH(X)KQHR(Z)-K]-GSG as both cis and trans isomers, and corresponding values of their hydrodynamic radius (R_h) and root mean square deviation (RMSD). Note: the GSG spacer is abstracted for clarity.

Variants 1–4, 6–11, 14, 15, 17–20, 22, and 24 were docked in silico against the putative binding sites S1–S5 on VCAM1 (PDB ID: 1VCA) using HADDOCK (High Ambiguity Driven Protein–Protein Docking, v.2.1)^31,32 following the method employed in prior work.^74,81 In particular, the GSG spacer located on the C-terminus of the peptide was marked as “inactive” to ensure its outward orientation in the peptide-VCAM1 complexes; a short peptide spacer, in fact, is utilized to link the probing moiety (e.g., biotin or a fluorophore) to the peptide and should not be involved in VCAM1 binding. The resulting clusters for every peptide-VCAM1 complex were ranked using the scoring functions FireDock and XScore.^33,34 The top binding complexes were localized mostly on putative binding sites S1, S4, and S5, and were evaluated further by atomistic MD simulations to evaluate the free energy of binding (ΔG_B). Representative examples of modeled peptide-VCAM1 complexes and the corresponding values of binding energy (ΔG_B,trans and ΔG_B,cis) are shown in Fig. 3.

Fig. 3.

Structure of the peptide-VCAM1 complexes formed by docking azobenzene-cyclized peptide variants 1–4, 6–11, 14, 15, 17–20, 22, and 24 (Fig. 2) onto the putative binding sites S1–S5 of VCAM1.

Variants cyclo_AZOB[G-VHPKQHRS-K*], cyclo_AZOB[G-VHAKQHRP-K*], and cyclo_AZOB[G-VHNKQHRP-K*] were predicted to possess strong binding in cis configuration (ΔG_B,cis < −8.5 kcal mol⁻¹, corresponding to K_D < 0.65 μM), while cyclo_AZOB-[G-VHAKQHRN-K*], cyclo_AZOB[G-VHNKQHRS-K*], and cyclo_AZOB-[G-VHPKQHRP-K*] were predicted to possess strong binding in trans configuration (ΔG_B,trans < −7.9 kcal mol⁻¹, corresponding to K_D < 1.5 μM). Importantly, these variants were predicted to lose binding upon photo-isomerization (∣ΔΔG_B∣ > 2.5 kcal mol⁻¹, highlighted in red) and were therefore selected for experimental analysis. All other variants featured insufficient values of either ΔG_B or ∣ΔΔG_B∣ and were therefore abandoned.

Characterization of azobenzene-cyclized peptides

Spectroscopic characterization.

UV spectroscopy was utilized to evaluate the kinetic of photo-isomerization using sequence cyclo_AZOB[VHGKQHRP-K*] as model variant. The comparison of the absorption spectra of the linear precursor VHGKQHRP-K*, cyclo_SUCC[VHGKQHRP-K*] (wherein a succinyl linker is utilized in lieu of the azobenzene linker), and cyclo_AZOB[VHGKQHRP-K*] suggests that: (i) peptide cyclization constrains the histidine residues (His), which carry aromatic imidazole rings, in a spatial orientation that produces an increase in absorbance at 254 nm (solid vs. dot-dash, Fig. 4A); this effect was reported in a published study on the effect of spatial arrangement of histidine molecules upon their spectroscopic behavior;⁸² (ii) cyclization via azobenzene linker magnifies this effect (dot-dash vs. dash spectra), Fig. 4A, suggesting interplay between the imidazole residues and the benzene rings in the azobenzene linker; this results in a red shift in absorbance of the peptide segment and a blue shift in the azobenzene absorbance, which have been observed in other azobenzene-cyclized peptides.⁸³ The histidine-azobenzene interference also attenuates the difference between the spectra of the trans and cis isomers of cyclo_AZOB[VHGKQHRP-K*] compared to those characteristic of the trans and cis isomers of free azobenzene. Despite the subtle difference between the spectroscopic profiles (Fig. 4B), peptide photo-isomerization was detected in the 300–390 nm and 400–420 nm ranges, corresponding to the π–π* and n–π* bands of azobenzene, respectively. We also recorded the temporal evolution of the spectra of cyclo_AZOB[VHGKQHRP-K*] in solution during exposure to UV light (using a bandpass light filter BP305-390), and observed a decrease at 340 nm and an increase at 420 nm, both indicative of trans-to-cis isomerization (Fig. S2A and B, ESI†); upon exposure to blue light (using a longpass light filter LP420), instead, we observed an increase at 340 nm and a decrease at 420 nm, indicative of trans-to-cis isomerization (Fig. S2C and D, ESI†). The difference between the spectra (Δabsorbance) is reported in the inset of Fig. 4B.

Fig. 4.

(A) UV-vis absorption spectra for VHGKQHRP-K* (solid), cyclo_SUCCVHGKQHRP-K*] (dot-dash), and cyclo_AZOB[VHGKQHRP-K*] (dash); (B) Δabsorbance for the cis-induced isomer (red) and trans-induced isomer (blue); (C) trans-to-cis isomerization upon exposure to UV light (305–390 nm) and (d) cis-to-trans isomerization upon exposure to visible light (>420 nm). For (C) and (D) fitted lines are solid (squares), hash (circles), dot-hash (triangles), and error bars mean ± S.D.

Based on the variation in absorbance at 350 nm, the kinetic constants of photo-isomerization were calculated for different values of light intensity and solution concentration (Fig. 4C and D; the data set is reported in the Fig. S3, ESI†). Rate constants vs. exposure intensity were fit, assuming first-order kinetics. The calculated cis-to-trans and trans-to-cis rate constants are within a factor of 2 of each other; further, the broad excitation spectra and the potentially simultaneous excitation of both cis-to-trans and trans-to-cis limit direct attribution of the fundamental rate constants to the individual photoisomerization processes. Accordingly, the reported rate constants are the spectrum-weighted rate constants. The exhibited small difference in the reported rate constants can be attributed to light attenuation by the peptide segment, differences in light attenuation respective to conformation, a difference in photo-isomerization cross sections, or the relaxation back to trans may be the favored process. In this context, we believe that the difference stems primarily from a general attenuation phenomenon, as observed at high concentrations, where the light absorbance by the peptide segment reduces the fraction of incident energy absorbed by the azobenzene linkers and, consequently, the isomerization rate. For dilute solutions (0.15 and 0.38 mM), the rate constants of both trans-to-cis and cis-to-trans isomerization were found to be higher than the constant obtained with a 1.5 mM (Fig. 4C). At the concentration of 1.5 mM, an attenuation of the excitation energy of approximately 40% is observed across the sample in the BP 305–390 range. This effect is illustrated by the larger difference in the rate constants of the trans-to-cis isomerization (λ_ex = 305–390 nm) across the investigated concentration range (Fig. 4C) compared to the difference in the rate constants of the cis-to-trans isomerization (λ_ex = 420–500 nm). The effect of light intensity on the rate constant for the trans-to-cis isomerization ranges from 7.9 (0.15 and 0.38 mM) to 5.9 s⁻¹ W⁻¹ cm² (1.5 mM); for the cis-to-trans isomerization, the difference is significantly smaller (3.6–3.3 s⁻¹ W⁻¹ cm²). This can be attributed to the fact that the absorbance of the peptide significantly decreases beyond 400 nm, which encompasses the LP420-filtered spectra used for the cis-to-trans isomerization. Notably, the isomerization rate constants of cyclo_AZOB[VHGKQHRP-K*] are considerably lower than those of free azobenzene and > 10-fold lower than that of azobenzene self-assembled monolayers.^84,85

The slower kinetics are attributed to the covalent linkage of the azobenzene within the framework of the cyclic peptide structure and, potentially, to the aromatic interaction with the histidine residues in the VCAM1-binding peptide segment. It is also noted that the broad spectra used for excitation have a negligible impact on the observed kinetics, as the filtered spectra (Fig. S4, ESI†) used for excitation feature a narrow peak with the majority of the energy; for the BP305-390 filtered excitation the major peak is at 365 nm, and for LP420 the peak is at 437 nm. The isomerization rate constants were used to calculate the dosage required for a nearly complete (5/κ or 99.3%) photo-isomerization: 630 mJ cm⁻² for trans-to-cis isomerization of diluted peptide solutions (0.15 mM and 0.38 mM) and 840 mJ cm⁻² for the concentrated solution (1.5 mM); for cis-to-trans isomerization, the dosage was calculated to be 1400 mJ cm⁻² for the diluted solutions and 1500 mJ cm⁻² for the concentrated solution. Acknowledging this difference in the trans–cis rate constant at high concentration (i.e., possible deviation from Beer–Lambert Law) is important as a practical consideration for calculating the dosage needed to achieve complete photoisomerization.

We finally evaluated the ability of cyclo_AZOB[VHGKQHRP-K*] to undergo multiple cycles of photo-isomerization and the conformational stability of the cis isomer at 37 °C (“thermal relaxation”). Notably, the photo-isomerization of cyclo_AZOB[VHGKQHRP-K*] is completely reversible and can be cycled repeatedly (Fig. S5, ESI†). The thermally induced cis-to-trans isomerization was evaluated by monitoring the absorbance at 350 nm of a 0.15 mM solution of the cis isomer of the peptide for 60 h at 37 °C (Fig. 5). The resulting half-life of ~44 h proves that the trans isomer of the peptide is thermally stable within the time span required for cell labelling and imaging without significant unbinding of the peptide. The thermal stability of cyclo_AZOB[VHGKQHRP-K*] is coherent with the results of the MD simulations, which showed that the energy landscape of the trans and cis isomers feature minima with similar values, respectively 1063 and 1089 kJ mol⁻¹; these, in turn, are comparable to those of the linear precursor VHPKQHR (958 kJ mol⁻¹) and the cyclic variant cyclo_SUCC-[G-VHAKQHRN-K*] (1035 kJ mol⁻¹), later used in cell labeling.

Fig. 5.

Thermal cis-to-trans isomerization (relaxation) of cyclo_AZOB-[VHGKQHRP-K*]. The red curve represents initial induction of the peptide into the cis conformation and the blue curve represents the recovery of the trans configuration. Isomerization was monitored by recording the absorbance at 350 nm of a 0.15 mM solution of peptide in MilliQ water for 48 h, at 37 °C in the dark.

These results are particular remarkable considering that, unlike other azobenzene-cyclized variants, neither isomer of cyclo_AZOB[G-VHAKQHRN-K] features elements of secondary structure (i.e., α-helix or β-sheet) known to confer stability to peptides;^86-88 both VHPKQHR and cyclo_SUCC[G-VHAKQHRN-K*], in fact, possess a short α-helical segment (Fig. S6C, ESI†), which contributes to a more favorable energetic state. The secondary structures of cyclo_AZOB[G-VHAKQHRN-K], VHPKQHR, and cyclo_SUCC[V-HGKQHRP-K*] predicted in silico were confirmed by circular dichroism (Fig. S6, ESI†).

Lastly, we have conducted a spectroscopic characterization under a broad range of UV light exposure (λ_ex = 305–390 nm and λ_ex = 420–500 nm) to address concerns of potential cytotoxicity. We note that the dose applied for nearly complete photo-isomerization of the proposed peptide has been reported to cause minimal cytotoxic effects, especially in absence of photosensitizers;^89-93 UV dosing in the range of 138–6000 mJ cm⁻² has, in fact, been shown to be viable for multiple cell processing methods.^89-94 The effects of photo-isomerization on cell activity are dependent on several factors, such as the photo-isomerization reaction, wavelength, light power, exposure time, and presence of a photosensitizer. It is also noted that the proposed peptide demonstrates thermal stability, indicating that low-dose rapid pulse excitation can be applied to achieve light-induced activation of binding activity. Future efforts should always optimize the operating conditions for each cell-type or target protein, including photoisomerization parameters, to limit cytotoxicity and off-target effects.

Binding affinity by surface plasmon resonance (SPR)

Efficient photo-controlled activation of biorecognition requires a strong shift in binding energy from the inactive to the binding configuration upon photo-isomerization. The in silico screening provided several sequences that fulfilled both conditions and were therefore tested by SPR to measure both their ΔG_B,trans and ΔG_B,cis. The average ligand density for the SPR chips used was 0.86 molecules nm⁻², as reported in Table S2 (ESI†). The values of VCAM1 mass bound to the peptide sensors were fitted to a Langmuir isotherm (Fig. 6), from which the values of K_D were calculated (Table 1). Variant cyclo_AZOB[G-VHAKQHRN-K*] showed a remarkable variation in binding strength for VCAM1 in solution between the cis and trans conformations.

Fig. 6.

SPR measurements of VCAM1:peptide K_D,trans and K_D,cis for (A) of cyclo_AZOB[G-VHAKQHRN-K*]; (B) cyclo_AZOB[G-VHPKQHRS-K*]; (C) cyclo_AZOB-[G-VHAKQHRD-K*]; and (D) VHPKQHR. The data points were obtained as average of triplicate readings.

Table 1.

Values of ΔG_B,trans, ΔG_B,cis, and ∣ΔΔGB∣ of the interaction between VCAM1 and cyclo_AZOB[G-VHAKQHRP-K*], cyclo_AZOB[G-VHNKQHRP-K*], cyclo_AZOB[G-VHPKQHRS-K*], cyclo_AZOB[G-VHAKQHRN-K*], cyclo_AZOB[G-VHNKQHRS-K*], cyclo_AZOB[G-VHPKQHRP-K*], and VHPKQHR determined by fitting the VCAM1 adsorption data obtained from the SPR sensorgrams to a Langmuir isotherm. The “—” indicates that no accurate SPR reading could be obtained

Peptide	ΔGB,trans (kcal mol−1)	ΔGB,cis (kcal mol−1)	∣ΔΔGB∣
cycloAZOB [G-VHAKQHRP-K*]	—	1.09	—
cycloAZOB[G-VHNKQHRP-K*]	1.29	−0.55	1.84
cycloAZOB[G-VHPKQHRS-K*]	3.20	0.68	2.52
cycloAZOB[G-VHAKQHRN-K*]	2.86	−0.07	2.93
cycloAZOB[G-VHNKQHRS-K*]	2.14	—	—
cycloAZOB[G-VHPKQHRP-K*]	1.58	0.06	1.52
VHPKQHR	−0.93

Cell-labelling with cyclo_AZOB[G-VHAKQHRN-K*]

The light-activated biorecognition activity of variant cyclo_AZOB-[G-VHAKQHRN-K*] was finally demonstrated in vitro via labelling of VCAM1⁺ BMECs. VCAM1 expression was induced by treating the cells with interleukin-4 (IL-4) and lipopolysaccharide (LPS); this pair has been shown to be synergistic in eliciting VCAM1 expression in endothelial cells.^95-97 VCAM1 expression by BMECs was confirmed by RT-qPCR (Fig. S7, ESI†); LPS alone also induced VCAM1 expression in BMECs, though to a lesser extent. Human umbilical vein endothelial cells (HUVECs) and human dermal fibroblasts (HDFn) were also considered, but they did not demonstrate significant changes in VCAM1 expression when treated with either condition.

VCAM1⁺ BMECs were incubated with either the cis or trans isomer of cyclo_AZOB[G-VHAKQHRN-K*]-biotin and stained with AlexaFluor488-Streptavidin. As anticipated, the cells labeled with the cis isomer demonstrated significantly higher fluorescence intensity compared to those labeled with the trans isomer (Fig. 7). The cis conformation of cyclo_AZOB[G-VHAKQHRN-K*] also conferred significantly higher fluorescent intensity to VCAM⁺ cells compared to VCAM1⁻ cells. To confirm selective VCAM1 binding, we also evaluated peptide co-localization with anti-VCAM1 antibodies on VCAM1⁺ BMECs. The images collected using confocal microscopy shows a clear co-localization of the cis isomer of cyclo_AZOB[G-VHAKQHRN-K*] with the VCAM1 antibody (Fig. 8).

Fig. 7.

Cyclo_AZOB[G-VHAKQHRN-K*] in the cis conformation incubated with cells (A) expressing and (B) not expressing VCAM-1, cyclo_AZOB-[G-VHAKQHRN-K*] in the trans conformation incubated with cells; (C) expressing and (D) not expressing VCAM-1, and (E) the relative intensity of biotin-labelled cis/trans cyclo_AZOB[G-VHAKQHRN-K*] incubated with VCAM-1^+/− induced cells. Statistics were done with a 2-tailed t-test with a p value <0.05 (*) considered statistically significant.

Fig. 8.

IL-4/LPS treatment induced VCAM1 expression confirmed with (A) antibody staining (B) cis cyclo_AZOB[G-VHAKQHRN-K*] peptide and (C) the colocalization of the anti-VCAM1 antibody and cis cyclo_AZOB[G-VHAKQHRN-K*] peptide around the nucleus of the cell.

Collectively, these results demonstrate both selectivity and light-controlled VCAM1-binding activity of cyclo_AZOB[G-VHAKQHRN-K*], as well as adequate thermal stability for in vitro studies.

Conclusions

The ability to activate – remotely, efficiently, and rapidly – the binding affinity of ligands to target biomolecules will uniquely enable next-generation biotech applications. As more effective mechanisms for affinity regulation are developed, more advanced applications can be explored. To this end, we have developed and demonstrated a novel approach to design and validate azobenzene-cyclized peptide affinity ligands featuring inducible protein-binding activity via light-controlled structural reconfiguration.

Azobenzene-cyclized variants of a known VCAM1-binding peptide (VHPKQHR) were discovered in silico that promised isomerization-dependent binding of VCAM1. Selected variants were experimentally characterized for thermal stability, isomerization kinetics, and VCAM1 binding affinity. The selected candidate cyclo_AZOB[G-VHAKQHRN-K*] was finally tested in vitro via light-controlled labelling of VCAM1⁺ BMEC. The peptide showed a statistically significant, conformation-dependent binding selectivity, confirming that the in silico and experimental characterizations. This study demonstrates a general method for developing azobenzene-cyclized peptides featuring efficient and robust light-controlled activation of protein-binding affinity. Notably, the light-responsive behavior is imparted to the ligand irrespectively of its protein-binding segment, making our design experimental methods presented herein are applicable to the design of light-responsive labels to virtually any target protein.

These results will enable future studies focusing on the aspects of on-demand, spatio-temporal control, and reversibility of cell labelling. Specifically, future efforts will evaluate (i) the kinetics of binding and unbinding the surface of the cells upon exposure to light at activating and de-activating wavelengths; (ii) the ability to direct the labelling of a subset of cells in solution via spatial control of the incident light; and (iii) the potential downstream effects of the peptide to the cell surface and, potentially, intracellular metabolic pathways upon binding and unbinding.