Design and Biophysical Characterization of Second-Generation Cyclic Peptide LAG-3 Inhibitors for Cancer Immunotherapy

Lymphocyte activation gene 3 (LAG-3) is an inhibitory immune checkpoint crucial for suppressing the immune response against can cer. Blocking LAG-3 interactions enables T cells to recover their cytotoxic capabilities and diminishes the immunosuppressive effect s of regulatory T cells. A cyclic peptide (Cys-Val-Pro-Met-Thr-Tyr-Arg-Ala-Cys, disulfide bridge: 1-9) was recently reported as a LAG-3 inhibit or. Based on this peptide, we designed 19 derivatives by substituting tyrosine residue to maximize LAG-3 inhibition. Screening via TR-FRET a ssay identified 8 outperforming derivatives, with cyclic peptides 12 [Tyr6(L-3-CN-Phe)], 13 [Tyr6(L-4-NH 2-Phe)], and 17 [Tyr6(L-3,5-DiF-Phe)] as top candidates. Cyclic peptide 12 exhibited the highest inhibition (IC 50 = 4.45 ± 1.36 µM). MST analysis showed cyclic peptides 12 and 13 bound LAG-3 with KD values of 2.66 ± 2.06 µM and 1.81 ± 1.42 µM, respectively, surpassing the original peptide (9.94 ± 4.13 µM). Docking simulations indicated enhanced binding for cyclic peptide 12, with a docking score of -7.236 kcal/mol compared to -5.236 kcal/m ol for the original peptide.

Lymphocyte-activation gene 3, cyclic peptides, cancer immunotherapy, drug discovery, computational chemistry In 2022, the lymphocyte-activation gene 3 (LAG-3) was approved by the FDA (United States Food and Drug Administration) as the latest immune checkpoint target,1 bringing the total number of FDA-approved immune checkpoint inhibitor drugs to four.2 However, this anti-LAG-3 cancer immunotherapy, as other monoclonal antibodies before, presents some drawbacks compared to other families of cancer therapies, such as small molecules and peptides. Briefly, small molecules and peptides present more efficient tumor penetration properties, less adverse immune responses over time, lower manufacturing costs, and can be optimized to improve their pharmacokinetic properties. 3-5 To date, over twenty different peptides are currently used in cancer therapy .6,7 Peptides are excellent candidates for targeting protein-protein interactions due to their intrinsic properties, which mimic essential features of proteins, providing them with high specificity.8 Consequently , their potential as immune checkpoint inhibitors is being increasingly explored. Chemical modifications are commonly employed to candidate peptides to improve therapeutic properties such as three-dimensional stability, blood circulation time, or shelf-life, among others. One such modification is peptide cyclization. Compared to linear peptides, cyclic peptides offer several advantages in targeting immune checkpoints for cancer immunotherapy. Cyclic peptides are more resistant to enzymatic degradation resulting in extended three- dimensional stability and half-life, thereby improving efficacy. 9 Therefore, cyclic peptides often show enhanced bioavailability and maintenance of therapeutic levels in the system. The cyclic structure also confers peptides with a conformational rigidity, which preserves the peptide's 3D conformation and therapeutic properties, avoiding reduced binding affinity or off-target effects. Overall, these characteristics make cyclic peptides promising candidates for cancer immunotherapy, particularly in targeting immune checkpoints where precise modulation of immune responses is crucial for therapeutic efficacy. Recently, a cyclic peptide targeting LAG-3 immune checkpoint has been discovered using biopanning as the affinity selection technique.10 This peptide successfully inhibited LAG-3 interaction with one of its natural ligands, major histocompatibility complex class II (MHC-II). 10 Based on this peptide’s structure (Cys-Val-Pro-Met-Thr-Tyr-Arg-Ala-Cys, disulfide bridge: 1-9) with submicromolar affinity to LAG-3 protein, we designed 19 different derivatives by incorporating diverse functional groups to the tyrosine amino acid residue (Supplementary Information, Table S1). The conformation of cyclic peptides can be significantly influenced by the chemical composition of the amino acid side chains. Thus, substituting tyrosine with other residues can help stabilize a desired conformation or induce specific structural changes that enhance LAG-3 binding affinity and LAG-3 inhibitory profile. W e first screened the 19 derivatives along with the original cyclic peptide for their ability to inhibit LAG-3/MHC-II interaction using Time Resolved Förster’s Resonance Energy Transfer (TR-FRET) assay. 11 Briefly, both LAG-3 and MHC-II are tagged with donor and acceptor fluorophores, and LAG- 3/MHC-II inhibitors are identified by the reduction in the LAG- 3/MHC-II TR-FRET signal. A primary screening using 20 µM as the tested concentration revealed that eight out of the 19 derivatives outperformed the original peptide, being derivatives 12 [Tyr6(L-3-CN-Phe)],13 [Tyr6(L-4-NH2-Phe)] and 17 [Tyr6(L- 3,5-DiF-Phe)] the three top candidates (Fig. 1A). Subsequent dose-response experiments indicated that cyclic peptide 12 held the highest inhibition capability out of the top candidates, IC 50 = 4.45 ± 1.36 µM, compared to cyclic peptide 13, IC 50 = 131.65 ± 35.30 µM; and cyclic peptide 17, IC 50 = 74.43 µM (Fig. 1B). Regarding cyclic peptide 17, no standard deviation was obtained since two out of the three replicates displayed a wider than accepted IC50’s confidence interval and the nonlinear regression model did not adjust the experimental data. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 6, 2024. ; https://doi.org/10.1101/2024.08.04.606540doi: bioRxiv preprint Fig. 1. Inhibition of LAG-3/MHC-II interaction by cyclic peptide derivatives. (A) Single dosage screening of 19 cyclic peptide derivatives. Tested cyclic peptides at 20 µM (1.5% DMSO) were incubated with Tag1- LAG-3, Tag2-MHC-II, Anti-Tag1 Eu Cryptate reagent, and Anti-Tag2 d2 antibody to study the inhibition ability towards LAG-3/MHC-II interaction. The dashed line indicates the inhibition rate of the original cyclic peptide (0). (B) Dose-response curves of the three cyclic peptides in the LAG-3/MHC-II TR-FRET assay. Inhibition rates were measured in triplicate, with results given as the mean ± standard deviation. W e assessed the binding affinity of the original cyclic peptides as well as the three top candidates (cyclic peptide 12, 13 and 17) to LAG-3 protein using microscale thermophoresis (MST) platform, which we validated for affinity screening using fibrinogen-like protein 1 (FGL-1) as a positive control. The binding affinity analysis revealed that cyclic peptides 12 and 13 bound LAG-3 with an equilibrium dissociation constant ( KD) equal to 2.66 ± 2.06 µM and 1.81 ± 1.42 µM, respectively (Fig. 2B, C). Both results showed a better performance of these two derivatives compared to the original cyclic peptide: 9.94 ± 4.13 µM (Fig. 2A). Although the screening outcome for cyclic peptide 12 revealed an agreement between LAG-3 binding and inhibition profiles, the outcome for cyclic peptide 13 revealed that protein binding affinity for a ligand does not necessarily correlate with its inhibitory potency. 12 On the other hand, no KD value was obtained for cyclic peptide 17. Results regarding cyclic peptide 17 inhibitory capability are in line with LAG-3 binding affinity outcome since no stable IC 50 values together with no binding to the target protein might indicate unstable binding or false positive

in TR-FRET assays (Fig. 1B). Fig. 2. Binding affinity of cyclic peptides 0, 12 and 13 to LAG- 3 pro A range of concentrations (50 µM to 1.53 nM) of the (A) cyclic peptide 0 cyclic peptide 12 and (C) cyclic peptide 13 were incubated with His- lab human LAG-3 protein (50 nM) for 10 minutes to determine the affinity o candidate binder for the His-labeled LAG- 3 protein using MST. The g present the results of three or four independent experiments. independent experiment is colored in different colors. Data points in color represent outliers/excluded data. protein. e 0, (B) labelled ty of the graphs . Each in grey .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 6, 2024. ; https://doi.org/10.1101/2024.08.04.606540doi: bioRxiv preprint Aiming to investigate the impact of the CN substituent in cyclic peptide 12 in improving LAG-3 binding affinity, we utilized Schrödinger's GLIDE software for docking simulations. The cyclic peptide 12 not only binds with higher affinity but also aligns more optimally within the complex's binding pocket compared to the original cyclic peptide (cyclic peptide 0). Specifically, the docking results demonstrated that cyclic peptide 12 achieved an average docking score of -7.236 kcal/mol, an enhancement from -5.236 kcal/mol observed for cyclic peptide 0. The cyclic peptide 12 binds through a two-phase process. Initially, in what we term the tran-conformation (TC), the peptide binds to the pocket but does not completely occupy it, leaving some space unoccupied (Fig. 3A, B). As the peptide relaxes and fully adjusts within the pocket, particularly accommodating the engineered CN group, it transitions to what we refer to as conformation-I (CI). In the TC phase, we observed that the engineered moiety adopts an out-conformation, meaning it protrudes away from the previously unoccupied space, which is now fully occupied due to the peptide's fully relaxed orientation. The detailed docking analysis revealed that the cyclic peptide 12 binds more effectively due to the CN substitution enhancing π -π interactions and hydrogen bonding within the binding pocket. The CN group facilitates a flipping of the peptide's 6-ring towards crucial residues W579 on MHC-II and W106 on LAG-3, optimizing π -stacking interactions essential for stable docking. This molecular alignment allows the cyclic peptide 12 to engage more comprehensively with the protein complex, forming multiple hydrogen bonds and hydrophobic interactions with nearby residues including the critical phenylalanine F443 (Fig. 3A, B). In our simulations, we additionally analyzed the d yna behavior of cyclic peptide 0 an d cyclic peptide 12 w complexed with LAG-3 and MHC-II over a total simulation of 1.0 µs, with each simulation run lasting 0.5 µs. Notably, cy peptide 12 exhibited a progressive decrease in root-mean- sq deviations (RMSD), suggesting that it systematically ex pl stable orientations within the pocket by adjusting latera lly interacting with critical interface residues (Fig. 3D). This stab is corroborated by root-mean- square fluctuation ( RM analyses, which highlighted decreased fle xibility in key b in regions, suggesting a more effective interaction with the ta protein complex (Fig. 3E). Moreover, our observations in dic that cyclic peptide 12 initially bound to LAG-3 and MHC- II manner where it entered the pocket with out fully occup yin However, by the 200 ns mark, we noted the peptide begin nin adjust within the pocket and ultimately adopting a confor ma similar to that of cyclic peptide 0, albeit with a more s nu orientation (Fig. 3D). The analysis of the top conformations from the docking stu emphasized that cyclic peptide 12, despite slight in fluctuations, found a stable orientation within the binding po faster than cyclic peptide 0 (Fig. 3C, D). These fluctuatio ns not indicate i nstability; instead, they represented the p ept process of adjusting within the space to achieve optimal f it. dynamic adjustment process is essential for the pro lon interaction of cyclic peptide 12 with LAG- 3, which w significantly enhance the therapeutic potential of the peptide. Considering the promising results of cyclic peptide 12 in biochemical assays and docking simulations, we next ev alu the ability of the derivative peptide to inhibit tumor gro wt vivo. CT26.WT tumor- bearing mice were treated w it Fig. 3. Molecular docking and dynamics analysis of cyclic peptide 0 and cyclic peptide 12 with LAG-3/MHC-II complex. (A-B) Cartoon representations of the docked complexes of cyclic peptide 0 and cyclic peptide 12, each domain distinguished by different colors. Zoomed- in sections illustrate the superposition of their top docked conformations (cyclic peptide 0 in pink; cyclic p eptide 12 in orange), whit the lower panel illustrating the protein-ligand interaction network, labeling pocket residues. Peptides are rendered in surface representation for clarity. (C) Docking scores of the top five docked complexes highlight the impact of CN substitution, increasing the docking score by approximately 38.2%, demonstrating improved binding affinity of cyclic peptide 12 over cyclic peptide 0. (D) RMSD profiles demonstrate greater stability of cyclic peptide 12 compared to cyclic peptide 0 in the complex. (E) RMSF results for individual residues indicate enhanced stability in binding regions, influencing nearby regions allosterically. ynamic w hen n time , cyclic square plored lly and tability RMSF) inding ta rget dicated II in a ying i t. ning to mation nug f it studies initial pocket ns did eptide's it. This longed would e. in both aluated wth in ith 8 on ns ng he ed lic ns .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 6, 2024. ; https://doi.org/10.1101/2024.08.04.606540doi: bioRxiv preprint mg/kg/day for 21 days, and no differences were observed in either tumor growth or survival rates between the groups (Fig. 4A, B). Since cyclic peptide 0 demonstrated antitumor effects via CD8+ T cells, we aimed to investigate the potential immunomodulatory effect of cyclic peptide 12 in the tumor local environment (Supplementary Information, Fig. S1-2). However, no difference was observed for any of the studied immune populations: helper T cells (total, Th1 and Th17), cytotoxic T cells (total and IFN-γ -expressing cells), regulatory T cells, tumor- specific CD4+ T cells, and tumor-specific CD8+ T cells (Fig. 4C- J). Fig. 4. Effect of cyclic peptide 12 therapeutic treatment on tumor growth and tumor infiltrating immune cells in CT26.WT tumor bearing mice. (A) CT26.WT tumor bearing mice were treated with 8 mg/kg/day (intraperitoneal administration) for three weeks from day 9 after CT26.WT injection. Treatment of mice was initiated after the tumors had been grown until reaching a palpable size of 40-100 mm 3. The graph presents one independent experiment (Vehicle, n=11 and Cyclic peptide 12, n=11). The data are presented as the mean ± standard deviation. (B) Overall mice survival (in days) after the injection of CT26.WT colon carcinoma cell line (500,000 cells/animal). The graph presents one independent experiment (total number of animals at day 0: Vehicle, n=11 and Cyclic peptide 12, n=11). Percentage of (C) helper T cells, (D) Th1 cells, (E) Th17 cells, (F) cytotoxic T cells, (G) IFN-γ -expressing cytotoxic T cells, (H) regulatory T cells, (I) tumor-specific helper T cells, and (J) tumor-specific cytotoxic T cells in the tumor microenvironment. The charts present the results of an independent experiment (Vehicle, n=7 and Cyclic peptide 12, n=8). The data are presented as the means ± standard deviations. Abbreviations: Th: Helper T cell. The absence of immune modulation within the tumor microenvironment led us to speculate whether cyclic peptide 12 might have lost its structural stability upon injection into the animals. To investigate this, we conducted peptide stability assays using both cyclic peptide 0 and cyclic peptide 12 in mouse and human serum over a 24-hour period (Fig. S43). Cyclic peptide 0 remained intact after 24 hours of incubation in mouse serum (Fig. S43A), and similar results were observed in human serum (Fig. S43B). Similarly, the cyclic peptide 12 structure was not affected by mouse or human serum (Fig. S43C, D). The single peak observed for both cyclic peptides in each experimental condition indicates their stability in both types of serum. However, both cyclic peptides appeared to non- specifically bind to mouse and human serum proteins, resulting in a reduced abundance over time, as evidenced by the filtration process prior to HPLC analysis (Fig. S43). Something similar might have been occurring in vivo and so, no effective quantities would have reached LAG-3-expressing immune cells, which could potentially explain the lack of observed effectiveness. Conversely, our results are consistent with previous studies demonstrating a significant presence of PD-1-expressing CD8+ T cells within tumor infiltrating lymphocytes (TILs) in CT26 tumor-bearing mice (Fig. 4J). 13 PD-1, another negative immune checkpoint, is known to suppress anti-tumor immune responses. Consistent with observations in other tumor models, 14-16 our data on LAG-3 monotherapy inefficacy underscore the necessity of combining anti-LAG-3 with anti-PD-1/PD-L1 blockade. Such combination therapies are crucial not only for reducing tumor growth but also for exerting immunomodulatory effects within TILs. These results provide compelling evidence of the effectiveness of molecular modifications in cyclic peptides for therapeutic applications. The introduction of a CN substituent not only optimizes therapeutic interactions but also significantly enhances the binding efficiency and stability of cyclic peptides, offering a robust platform for the development of potent immunotherapeutic agents targeting the LAG-3/MHC-II interaction. This study underscores the critical role of detailed molecular understanding in the design and optimization of peptide-based therapeutics and highlights the importance of dual immunotherapies to achieve therapeutic effects. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data Availability Data will be made available on request. Acknowledgments The authors thank Dr. Natalie Fuchs for technical assistance on processing tumor samples for subsequent flow cytometry analysis. Microscale Thermophoresis (MST) was performed in the Rockefeller University’s Bio-Imaging Resource Center, RRID:SCR_017791. We gratefully acknowledge financial support from the ELSA U. Pardee Foundation (Award ID: 2022- 215). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 6, 2024. ; https://doi.org/10.1101/2024.08.04.606540doi: bioRxiv preprint

and notes 1. Tawbi HA, Schadendorf D, Lipson EJ, et al. Relatlimab and Nivolumab versus Nivolumab in Untreated Advanced Melanoma. N Engl J Med. 2022;386(1): 24-34. 2. Sun Q, Hong Z, Zhang C, Wang L, Han Z, Ma D. Immune checkpoint therapy for solid tumours: clinical dilemmas and future trends. Signal Transduct Target Ther. 2023;8(1): 320. 3. Imai K, Takaoka A. Comparing antibody and small-molecule therapies for cancer. Nat Rev Cancer. 2006;6(9): 714-727. 4. Muttenthaler M, King GF, Adams DJ, Alewood PF. Trends in peptide drug discovery. Nat Rev Drug Discov . 2021; 20(4): 309- 325. 5. Ji X, Nielsen AL, Heinis C. Cyclic Peptides for Drug Development. Angew Chem Int Ed Engl. 2024;63(3): e202308251. 6. Timur SS, Gursoy RN. The role of peptide-based therapeutics in oncotherapy. J Drug Target. 2021;29(10): 1048-1062. 7. Ramadhani D, Maharani R, Gazzali AM, Muchtaridi M. Cyclic Peptides for the Treatment of Cancers: A Review. Molecules. 2022;27(14). 8. Nevola L, Giralt E. Modulating protein-protein interactions: the potential of peptides. Chem Commun (Camb). 2015; 51(16): 3302- 3315. 9. Zorzi A, Deyle K, Heinis C. Cyclic peptide therapeutics: past, present and future. Curr Opin Chem Biol. 2017;38: 24-29. 10. Zhai W, Zhou X, Wang H, et al. A novel cyclic peptide targeting LAG-3 for cancer immunotherapy by activating antigen-specific CD8(+) T cell responses. Acta Pharm Sin B . 2020; 10(6): 1047- 1060. 11. Abdel-Rahman SA, Rehman AU, Gabr MT. Discovery of First-in- Class Small Molecule Inhibitors of Lymphocyte Activation Gene 3 (LAG-3). ACS Med Chem Lett. 2023;14(5): 629-635. 12. Abdel-Rahman SA, Santini BL, Calvo-Barreiro L, Zacharias M, Gabr M. Design of cyclic peptides as novel inhibitors of ICOS/ICOSL interaction. Bioorg Med Chem Lett. 2024; 99: 129599. 13. W augh KA, Leach SM, Moore BL, Bruno TC, Buhrman JD, Slansky JE. Molecular Profile of Tumor-Specific CD8 + T Cell Hypofunction in a Transplantable Murine Cancer Model. J Immunol. 2016;197(4): 1477-1488. 14. Woo SR, Turnis ME, Goldberg MV, et al. Immune inhibitory molecules LAG-3 and PD-1 synergistically regulate T-cell function to promote tumoral immune escape. Cancer Res . 2012;72(4): 917-927. 15. Huang RY, Eppolito C, Lele S, Shrikant P, Matsuzaki J, Odunsi K. LAG3 and PD1 co-inhibitory molecules collaborate to limit CD8+ T cell signaling and dampen antitumor immunity in a murine ovarian cancer model. Oncotarget. 2015; 6(29): 27359- 27377. 16. Goding SR, Wilson KA, Xie Y, et al. Restoring immune function of tumor-specific CD4 + T cells during recurrence of melanoma. J Immunol. 2013;190(9): 4899-4909. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 6, 2024. ; https://doi.org/10.1101/2024.08.04.606540doi: bioRxiv preprint