Design of short peptides to block BTLA/HVEM interactions for promoting anticancer T-cell responses

Antibody based immune-checkpoint blockade therapy is a major breakthrough in oncology, leading to clinical benefit for cancer patients. Among the growing family of inhibitory receptors, the B and T lymphocyte attenuator (BTLA), which interacts with herpes virus entry mediator (HVEM), is a promising target for immunotherapy. Indeed, BTLA inhibits T-cell proliferation and cytokine production. The crystal structure of the BTLA/HVEM complex has shown that the HVEM(26–38) fragment is directly involved in protein binding. We designed and analyzed the capacity of several analogs of this fragment to block the ligation between BTLA and HVEM, using competitive ELISA and cellular assay. We found that the HVEM(23–39) peptide can block BTLA/HVEM ligation. However, the blocking ability was due to the Cys encompassed in this peptide and that even free cysteine targeted the BTLA protein and blocked its interaction with HVEM. These data highlight a Cys-related artefact in vitro, which should be taken in consideration for future development of BTLA/HVEM blocking compounds.


Introduction
T lymphocytes are central players of the adaptive anti-tumor immune response. Activation of T lymphocytes requires two signals: the first one is originating from the T-cell receptor after interaction with an antigen-derived peptide presented by major histocompatibility complex, while the second is delivered upon interactions between co-receptors on T cells and their ligands on antigen-presenting cells (APCs) or target cells [1]. Co-receptors, also called a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 immune checkpoints, can be divided into stimulatory and inhibitory molecules. The CD28 was the first identified co-stimulatory molecule and the interaction between CD28 and B7.1/ B7.2 is one of the dominant pathways required for full activation of naive lymphocytes [2,3]. In contrast, cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and programmed cell death 1 protein (PD-1), when interacting with their ligands B7.1/B7.2 and PDL-1/PDL-2 respectively, restrain the activation of lymphocytes T and are thus considered as co-inhibitory molecules [4][5][6][7].
It is known that manipulation of checkpoint cell-surface signaling molecules can exert potent anticancer effects. Immune checkpoint therapy has recently reached important clinical advances [8,9]. Indeed, it has been reported that cancer patients treated with anti-CTLA-4 (Ipilimumab) or anti-PD-1 (Nivolumab or Pembrolizumab) blocking antibodies showed an increase of the antitumor T-cell response and a higher rate of disease free survival [10][11][12][13]. Additionally, antibodies blocking PD-L1 were shown to be superior to Ipilimumab by having a better safety profile. Therapies combining both Ipilimumab and Nivolumab in patients with metastatic melanoma, have shown significant increase of the progression-free survival rate. However, this combination was also associated with substantial immune related toxicities [12,[14][15][16][17].
Among the growing family of inhibitory receptors, the B and T lymphocyte attenuator (BTLA), which interacts with herpes virus entry mediator (HVEM), inhibits T-cell proliferation and cytokine production [24,25], suggesting that BTLA might be an interesting target in immunotherapy [26]. Naive human CD8 + T cells express high levels of BTLA, which is downregulated upon CD8 differentiation. However, BTLA is not downregulated in melanoma specific CD8 + T cells and remains susceptible to functional inhibition upon HVEM ligation. HVEM is frequently expressed on melanoma cells, suggesting that the BTLA/HVEM pathway might play a role in the inhibition of efficient immune responses against cancer. Preventing/ targeting the BTLA/HVEM interaction can reverse the inhibitory functions of BTLA, thus triggering the immune response against cancer. Additionally, in vitro studies using antibodies to block binding of BTLA to HVEM led to increased melanoma specific CD8 + effector T-cells proliferation and enhanced cytokines production [27][28][29].

Molecular modeling
The crystal structure of BTLA/HVEM complex (A and B chains) served as a starting point for the modeling (Protein Data Bank ID 2AW2) [30]. The atoms of the HVEM B chain were removed, except for residues 23-39. The protonation state of the histidines was set depending on the environment, to optimize hydrogen bonds. All simulations were carried out using GROMACS 4.5 molecular modeling package [32] using the CHARMM22 force field [33]. The structure was located in the center of a TIP3P water box in which the distance between the solute and the edges was 1.4 nm. Na + and Clions were added to neutralize the system. Subsequently the system was subjected to 1000 steps of steepest descent minimization. The structures were then equilibrated according to standard equilibration procedures. The system heating was performed using simulated annealing method by increasing the temperature in four steps: from 0K to 50K, then from 50K to 100K, from100K to 200K and finally from 200K to 300K. Every step took 100ps and at the end of the procedure the system was additionally coupled to the target 300K temperature for 100ps. The force constant of 1000kJ/mol Ã nm was applied to keep the atoms in their positions during the heating procedure. The same force constant was kept also during the first constant volume equilibration simulation during 100ps, followed by another 100ps simulation in the NVT scheme with lowered force constant for position restraint to 300kJ/mol Ã nm. The final step of equilibration consisted in a simulation in the NPT scheme with Nose-Hoover thermostat and Parinello-Rahman barostat with 300K and 1Pa for 500ps and no position restraints imposed on atoms.
The production molecular dynamics simulations were done using the NPT scheme with no restraints on the atoms. For the production molecular dynamics simulation, the Verlet leap frog integrator was used. A 12Å cutoff was applied on non-bonded interactions. The bond length was constraint. A time step of 2fs was used with a length of the simulation of 10ns. For the hydrogen bonds analysis, we used 250 structures extracted every 40ps from the molecular dynamics simulation trajectory.
Molecular graphics images were produced using the UCSF Chimera package from the Computer Graphics Laboratory, University of California, San Francisco (supported by NIH P41 RR-01081) [34].

Peptide synthesis
Peptides were synthesized by solid phase peptide synthesis (SPPS) using semiautomated peptide synthesizer Millipore 9050 Plus PepSynthesizer (Millipore Corporation, Burlington, VT, USA) and general conditions of solid-phase synthesis [35]. Synthesis was performed on a Ten-taGel R RAM resin (0.19 mmol/g), using 9-fluorenylmethoxycarbonyl/tert-butyl (Fmoc/tBu) chemistry with the following side chain protected amino acid derivatives: Fmoc-Pro-OH, Fmoc-Glu(OtBu)-OH, Fmoc-Cys(Trt)-OH, Fmoc-Val-OH, Fmoc-Ala-OH, Fmoc-Gly-OH, Fmoc-Thr(tBu)-OH, Fmoc-Leu-OH, Fmoc-Lys(Boc)-OH, Fmoc-Arg(Pbf)-OH and Fmoc-Tyr (tBu)-OH. Acetylation of the N-terminal amino group was performed using 1-acetylimidazole (1.10 g/1 g of resin at room temperature for 24 h). The peptide was cleaved from the resin for 2 h using a mixture of 88% TFA, 5% fenol, 5% deionized water and 2% triisopropylsilane (10 ml/ 1 g of resin at room temperature for 2 h). After filtration of the exhausted resin, solution was concentrated in vacuum, and the residue was triturated with Et 2 O. The precipitated peptide was centrifuged for 15 min, 4000 rpm, followed by decantation of the ether phase from the crude peptide (process was repeated three times). After evaporation of Et 2 O, the peptide was dissolved in H 2 O and lyophilized.

Peptide purifications
Before the purification the peptides were dissolved in H 2 O, DTT was added (6-fold excess with respect to purified peptide) and the mixture was sonified for 30 min in 60˚C. Purification of the crude peptide was carried out by using RP-HPLC on a semi-preparative Phenomenex Luna C8(2) (250 mm x 20 mm, 5 μm) column. A linear gradient from 5% B to 50% B in A in 150 min was used. The aqueous system (A) consisted of 0.1% (v/v) TFA solution in water, whereas the organic phase (B) was 80% acetonitrile in water, containing 0.08% (v/v) TFA. Purification was monitored by UV absorption at a wavelength of 222 and 254 nm. The purity of the peptide was verified by LC-ESI-IT-TOF/MS (Shimadzu, Shimpol, Warsaw, Poland), and by using RP-HPLC with a Kromasil C8 analytical column (250mm x 4.6 mm, 5 μm), where a gradient of 5% to 100% B in A in 60 min was employed, with A and B as described above.

Disulfide bond formation
The formation of the disulfide bond in peptide HVEM(23-39) was performed after purification. The peptide was dissolved in a mixture of H 2 O and methanol (1:9, v:v), the final volume was 1.5 l and the concentration of the peptide was established at 40 mg/l. The pH was adjusted and kept between 8 and 9 by using ammonia whilst stirring the solution at room temperature for 3 days in the presence of atmospheric oxygen. Reaction progress was monitored by analytical RP-HPLC in a gradient of 5% to 100% B in A in 60 min. When the reaction of oxidation was completed the solvents were evaporated and the peptide was lyophilized.
ELISA 96-well round bottom plates (medium binding; Greiner Bio-One) were coated with 400 ng/ well of recombinant BTLA-Fc protein in PBS (50 μL/well) and incubated overnight at 4˚C. Wells were washed twice with PBS-T (phosphate buffer saline with 0.1% Tween-20, pH 7.4) and blocked with 1% milk in PBS-T for 1 hour at 37˚C. After 2 washes with PBS-T, the peptides were titrated down in triplicates from 10 to 0.1 mg/well in 1% milk/PBS-T and incubated for 2 hours at 37˚C. Maximum binding of HVEM to BTLA was controlled by adding 1% milk/ PBS-T to negative control wells, whereas maximum inhibition was monitored by adding anti-BTLA 7.1 mAb (0.1 μg/well). Next, the plate was washed and incubated with recombinant HVEM-Fc protein at 400 ng/well in 1% milk/PBS-T for 1 hour at 37˚C. Biotinylated anti-HVEM 11.8 antibody was then added (biotinylated preparation diluted at 1/10 000 in 1% milk/PBS-T) and incubated for 1 hour at 37˚C. Streptavidin horseradish peroxidase (Dako) was finally added at a 1/8000 dilution in 1% BSA/PBS-T for 30 min at 37˚C and detected with TMB (Biorad). The substrate reaction was blocked with 0.2 M H 2 SO 4 and the absorbance was read at 450 nm and 630 nm. The percent of inhibition was calculated assuming that the anti-BTLA 7.1 antibody inhibit the complex at 100% and milk/PBS-T in 0%. Statistical analysis of the results was performed using the one-way analysis of variance (one-way ANOVA).

Preparation of microcolumn
The immobilization of BTLA protein in a microcolumn was performed using the NHS-activated acid-coupled Sepharose 4B (Sigma Aldrich). BTLA protein was dissolved in 100 μl of coupling buffer (0.2 M NaHCO 3 , 0.5 M NaCl, pH 8.3) and was added to 0.06 g NHS-activated Sepharose in 100 μl of coupling buffer. The coupling reaction was performed for 1 hour in 25˚C and the mixture was transferred into the microcolumn (MoBiTec, Goettingen, Germany). The microcolumn with immobilized protein was washed alternately with 10 ml of solution blocking buffer (0.1 M amino ethanol, 0.5 M NaCl, pH 8.3) and washing buffer (0.1 M CH 3 COONa, 0.5 M NaCl, pH 4). This step was repeated 4 times. The microcolumn was stored in PBS (pH 7.4).

Affinity tests
Peptide affinity assay using NHS-activated Sepharose microcolumn was already described elsewhere [36,37]. The peptides were added onto BTLA-Sepharose column equilibrated in ammonium hydrogen carbonate (NH 4 HCO 3 , pH 7.4) and incubated for 2 h at 25˚C with gentle shaking. After this time the unbound peptide was removed and the column was washed with 100 ml NH 4 HCO 3 and 20 ml of H 2 O. Dissociation of the protein-peptide complex was performed using 0.1% TFA in H 2 O (2 x 0.5 ml) by 15 minutes. All fractions were analyzed using mass spectrometry (MALDI TOF/TOF™ 5800, AB SCIEX).

Ellman's assay
The presence of free thiols was investigated using Ellman's assay. DTNB (Sigma Aldrich) was first dissolved in reaction buffer (0.1 M sodium phosphate, pH 7.4, containing 1 mM EDTA) at a concentration of 4 mg/ml. Then peptides and recombinant BTLA-Fc, alone or mixed together (ratio 1:1) were added in DTNB reaction buffer. After 15 minutes of incubation, the SH content of protein samples was then assessed by the measurement of absorbance at 412 nm. For each sample, the absorbance value from the blank was subtracted.

BTLA/HVEM interaction-Cell line assay
293T cells expressing BTLA (generous gift from Dr. C. Krummenacher, University of Pennsylvania, USA) were maintained in culture with RPMI 10% FCS supplemented with 0.2 mg/ml hygromycin B. 2x10 5 293T cells expressing BTLA were incubated with rhHVEM-Fc at 8μg/ml in PBS 0.5% BSA for 30min at 4˚C. Cells were then washed and incubated with goat antihuman Fc conjugated with AF647 for 20min at 4˚C (Jackson ImmunoResearch, #109-606-170). Inhibition of HVEM-BTLA binding by different peptides was assessed by pre-incubating the cells with the different peptides at 5 mg/ml for 1h at 4˚C. BTLA expression was controlled by anti-BTLA APC-conjugated antibody (Biolegend, #344510). Maximum inhibition of binding was determined by adding anti-BTLA7.1 antibody for 1h at 4˚C at 2μg/ml before adding rhHVEM-Fc. Sample acquisition was performed on the Gallios Flow-Cytometer (Beckman Coulter) and data were analyzed using the FlowJo Software (TreeStar).

Design of peptides targeting BTLA protein
The crystal structure of BTLA/HVEM complex is stored under ID 2AW2 in the Protein Data Bank [30]. The structure indicates that only the first CRD1 part of human HVEM interacts with BTLA. The domain contains four short β-strands, that are connected by three disulfide bonds, formed between cysteine residues 4-15, 16-29 and 19-37. The fragment of the CRD1 domain of HVEM comprising residues 23-39 forms a β-hairpin structure and the 35-39 fragment builds an anti-parallel inter-molecular β-sheet with G˚strand of BTLA in the center of the interaction interface [30]. Moreover, the HVEM(23-39) fragment comprises two hot-spots of the BTLA/HVEM interaction: Tyr23 and Val36. We sought to determine whether the HVEM(23-39) fragment could bind BTLA. We therefore performed the molecular dynamics simulation of the BTLA structure and the HVEM(23-39) fragment (Ac-YRVKEACGELTGTVCEP-NH 2 ). The simulation starting point was a native-like conformation of HVEM fragment, bound to BTLA as in the BTLA/HVEM complex structure (Fig 1). The complex was stable during the 10ns long molecular dynamics simulation. The averaged RMSD calculated for the backbone atoms of the superimposed over carbons alpha structures was 1.2Å. Besides, the hydrogen Characterization of small peptides inhibiting BTLA/HVEM binding bonds interactions observed in the starting structure were almost all preserved during the simulation (S1 Table). This computational results suggests that the HVEM(23-39) may stably interact with BTLA.

Study of the interactions between BTLA protein and HVEM(23-39) fragment
The designed HVEM(23-39) peptide was synthesized and an affinity test was performed using microcolumn with the immobilized BTLA protein. Three fractions were analyzed: supernatant, last wash and elution by using mass spectrometry techniques. In the supernatant fraction the signal m/z corresponding to the excess of the HVEM(23-39) peptide was observed (Fig  2A). In the last wash fraction, no signal was observed confirming that the excess of peptide was removed (Fig 2B). The presence of m/z signal in elution fraction indicated that the complex between BTLA and HVEM(23-39) was formed ( Fig 2C).
Next, HVEM(23-39) was analyzed in an competitive ELISA test, in order to determine whether this fragment may bind to BTLA and inhibit BTLA/HVEM interaction. Fig 3 shows that the HVEM(23-39) can partially prevent the ligation of HVEM to BTLA, in a dose dependent manner, while scrambled peptide (Ac-ELCAGPVTRKVECTYGE-NH 2 ) has no effect. Moreover, the scrambled peptide (named Ctrl peptide in figures) did not interact with BTLA immobilized in microcolumn ( Table 1).

Assessment of amino acid residues from HVEM(23-39) peptide crucial for inhibition of the BTLA/HVEM complex formation
To assess the importance of each amino acid residues in HVEM(23-39) interaction with BTLA protein, single alanine-substituted HVEM(23-39) variants were synthesized [38,39] and tested in the BTLA/HVEM ELISA. While no or minor diminution in the inhibition of BTLA/HVEM binding was found with the majority of alanine-substituted fragments compared to the native HVEM(23-39), we observed a significant reduction of the inhibition, in a dose depend manner, for the two peptides where cysteine in position 29 (C29A) or 37 (C37A) was replaced by an alanine (Fig 4). This suggests the possible importance of cysteine residues in inhibition BTLA/HVEM complex formation by the HVEM(23-39) peptide.

The role of cysteine residues in targeting BTLA protein
To further determine the role of cysteine residues from HVEM(23-39) peptide in the inhibition of BTLA/VEM interaction, several HVEM(23-39) analogues were designed and synthesized (Table 2). Initially, we aimed to test, if the Cys29 residue could be replaced with aromatic residue that could interact with His127 residue of BTLA to improve the interaction of the peptide. To test this, we synthesized the HVEM(23-39) C29Y peptide. To additionally probe the importance of the cysteine interactions, we replaced the Cys29 residue with the serine residue, which is of similar size and polarity. We assumed that the C29S analog could mimic the cysteine interactions but bears no free sulfhydryl group that could interact non-specifically with BTLA. We were also wondering, if even shorter HVEM-derived peptide could inhibit the interactions between BTLA and HVEM. We synthesized therefore 9 amino acid-long peptides comprising the 1 -strand interface of HVEM, HVEM (31-39) fragment.
Then those analogs were tested in affinity test. We observed that peptides, which could inhibit the BTLA/HVEM interactions in ELISA tests (Fig 5), were also bound by BTLA protein immobilized in microcolumn ( Table 1). The exception was cyclic peptide HVEM(23-39) (C29-C37) which interacted with BTLA in affinity test but in ELISA test its inhibitory effect was not observed. The results suggest that the binding site of cyclic peptides and HVEM(23-39) on the BTLA surface is different.
Overall, our results confirm the prominent role of free cysteine residues in the BTLA/ HVEM interaction blocking capacity of the HVEM(23-39) peptide.
Since the cyclic peptide C29-C37, where cysteine residues were oxidized, does not block the BTLA/HVEM binding, we wondered whether the blocking ability of HVEM(23-39) fragment may be due to the capacity of HVEM-based peptide cysteines to reduce the cysteine disulfide bridges present on the BTLA surface, or possibly make disulfide bridges with these BTLA cysteines instead. Therefore, the effect of free cysteine and methionine amino acids or oxidized dimer of cysteine (cystine) on the BTLA-HVEM binding was analyzed. Results showed that cysteine with free sulfhydryl group targeted BTLA protein and prevented partially the binding of HVEM, whereas for the methionine and cystine, the inhibitory properties were not observed (Fig 6A). An Ellman's assay was then performed on both HVEM(23-39) peptide and recombinant BTLA-Fc in order to quantify free thiol groups. In contrast to BTLA-Fc, we observed free thiol groups in the HVEM(23-39) peptide. However, after 2 hours of incubation, the concentration of free thiol groups in HVEM(23-39) peptide was lower, suggesting that a fraction of the peptide is oxidized. Besides, when the HVEM(23-39) peptide was incubated with BTLA-Fc, we also observed a decrease in the concentration of free thiol groups over time, which was more pronounced than in the condition where the peptide was alone (Fig 6B). These results suggest that BTLA protein may form disulfide bonds with the HVEM (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39) peptide, lowering the amount of free thiol groups in the peptide.
checkpoints such as PD-1 or CTLA-4, have demonstrated major clinical benefit for cancer patients [40][41][42]. It is known that BTLA is involved in the negative regulation of T-cell responses by interacting with HVEM [27,43], highlighting BTLA as a potential target for cancer immunotherapy. Our study focused on the design of blocking peptides of the interaction between BTLA and HVEM, based on the amino acid sequence of HVEM binding fragment. The crystal structure of BTLA/HVEM complex indicated that in the interaction between the two proteins, two fragments of BTLA: (35)(36)(37)(38)(39)(40)(41)(42)(43) and (118-128) bind with HVEM: (26)(27)(28)(29)(30)(31)(32)(33) and (33)(34)(35)(36)(37)(38) respectively. Therefore, HVEM(23-39) fragment was synthesized and its interaction between BTLA and HVEM(23-39) was confirmed by using affinity chromatography. The competitive ELISA tests showed also that the peptide could inhibit the proteins binding. However, we showed using competitive ELISA and cellular assay that this blocking capacity was mainly due to cysteine residues present in the peptide sequence or in solution. Compaan and co-workers showed that the interactions between BTLA and HVEM proteins are primarily stabilized by main chain hydrogen bonds and includes relatively few side chain contacts [30]. The conformation of the HVEM(23-39) peptide is significantly different than the conformation of the same fragment in the native protein due to the fact that it is not stabilized by disulfide bridges and the peptide poses two cysteine residues with free sulfhydryl groups. In the monomeric crystal structure of the extracellular domain of BTLA protein, six cysteine residues form three disulfide bonds between residues 72-79, 34-63, and 58-115. The Cys58-Cys115 disulfide bond is completely buried (solvent accessible surface is 0% for both cysteine) in the hydrophobic core of BTLA [30]. The Cys58-Cys115 (solvent accessible surface is 2,9% and 17,9%) and more Cys34-Cys63 (solvent accessible surface is 45,1% and 33,3%) disulfide bonds are exposed to the solvent and could be sensitive for the other, free or high reactive sulfhydryl group. Therefore, we propose that free sulfhydryl groups of HVEM(23-39) peptide or free cysteine Characterization of small peptides inhibiting BTLA/HVEM binding amino acid compete, in our in vitro assays, with disulfide bridges (Cys58-Cys115 or Cys34-Cys63) of the BTLA protein to form covalent S-S bond between HVEM(23-39) peptide or cysteine amino acid and BTLA protein. This may most likely change the structure of BTLA, thus hindering the binding of HVEM. Yet, we observed that the scrambled peptide, which also contains free thiol groups as confirmed by Ellman's assay (S2 Fig), lacks blocking activity. In order to reduce the disulfide bonds on BTLA, peptides have to pre-dock to the surface of the protein. We hypothesize that scrambled peptide may have a lower affinity to BTLA, compared to HVEM(23-39) peptide. As a consequence of the weaker binding to BTLA the scrambled peptide may not be able to reduce the exposed disulfide bonds of BTLA and form disulfide bonds with the protein.
Finally, the fact that this phenomenon was not observed in the cellular assay might be caused by a spatial arrangement of the BTLA protein on the cell surface making the disulfide bridges on the BTLA surface less accessible for the tested peptides. We hypothesize that the 293T cells expressing human BTLA were incubated with peptides or free amino acids (5mg/mL) prior labeling with rhHVEM-Fc and AF647-conjugated anti-human IgG antibody (at least two experiments in triplicate). The graph shows the Geometric Fluorescence Intensity (GMFI). *: p< 0.05 and ***: p< 0.001 following non-parametric Oneway ANOVA and Dunn's post-test.
https://doi.org/10.1371/journal.pone.0179201.g007 small, compared with peptides, cysteine amino acid could reach the BTLA surface despite the spatial hindrance, hence only activity of cysteine amino acid was observed in the cellular test.
Overall, these data highlight a cysteine-related artefact in vitro, which should be taken in consideration when designing blocking compounds targeting BTLA/HVEM interaction.