Discovery of Cyclic Peptide Inhibitors Targeting PD-L1 for Cancer Immunotherapy

Abstract

Blockade of the interaction between programmed cell death ligand-1 (PD-L1) and its receptor PD-1 has shown great success in cancer immunotherapy. Peptides possess unique characteristics that give them significant advantages as immune checkpoint inhibitors. However, unfavorable physicochemical properties and proteolytic stability profiles limit the translation of bioactive peptides as therapeutic agents. Studies have revealed that cyclization improves the biological activity and stability of linear peptides. In this study, we report the use of macrocyclization scanning for the discovery of cyclic anti-PD-L1 peptides with improved bioactivity. The cyclic peptides demonstrated up to a 34-fold improvement in the PD-1/PD-L1 blocking activity and significant in vivo anti-tumor activity. Our results demonstrate that macrocyclization scanning is an effective way to improve the serum stability and bioactivity of the anti-PD-L1 linear peptide. This strategy can be employed in the optimization of other bioactive peptides, particularly those for protein-protein interaction modulation.

Graphical Abstract

Introduction

Immune checkpoint inhibitors have been very successful for immunotherapy in the past few years. The most popular agents in this approach are monoclonal antibodies (mAbs), with anti-PD-1 and anti-PD-L1 antibodies being the key players. Currently, a total of seven PD-1/PD-L1 inhibitors have been approved by the United States Food and Drug Administration (FDA). Daurvalumab (Imfinzi), atezolizumab (Tecentriq) and avelumab (Bavencio) are anti-PD-L1 inhibitors, while pembrolizumab (Keytruda), nivolumab (Opdivo), cemiplimab (Libtayo) and dostarlimab (Jemperli) are marketed as anti-PD-1 mAbs. These mAbs are deployed to effectively manage numerous malignancies including cutaneous squamous cell carcinoma, melanoma, non-small cell lung cancer, urothelial carcinoma, and merkel-cell carcinoma among others. Moreover, there are on-going studies to investigate the possibility of using these checkpoint inhibitors either alone or in combination with other agents in the management of other cancer types ^1–3.

mAbs remain very important therapeutic agents largely because of their high stability and specificity to targets ⁴. That said, mAbs as therapeutic agents do have some shortfalls that cannot be overlooked. Poor tumor penetration, high propensity to trigger immune response, high production costs, minimal flexibility in structure modification, and formulation challenges are some limitations of mAbs that create the need for new therapeutic agents like small peptides that are effective without the above-mentioned setbacks ^{5, 6}.

In the past, peptides were not given much attention as potential therapeutic agents partly because of their relatively poor chemical and physical stability towards peptidases in vivo, resulting in their short half-life and rapid elimination from the body ⁷. Also, peptides have high conformational flexibility which often leads to promiscuous interactions with multiple receptors, instigating side effects ⁸. Recent studies have shown that with the right tools and skill set, most of these undesirable characteristics that create an inconvenience in peptide drug development can be surmounted ⁹. Consequently, the last decade has seen a rapid rise in the number of peptides and peptidomimetics approved by regulatory agencies around the world ¹⁰. Peptide drugs are generally more efficacious, more potent, highly selective, and safer than most small-molecule drugs. The relative ease of chemical synthesis and modification make peptides a preferred choice over antibodies.

The development of small-molecule and peptide immune checkpoint inhibitors is not at pace with monoclonal antibodies. Despite the inherent challenges of small-sized checkpoint inhibitors, a number of small molecules, linear peptides, cyclic peptides, and peptidomimetics have been discovered to modulate the PD-1/PD-L1 protein-protein interaction ¹¹. For instance, Sasikumar et al. recently reported the discovery of CA-170, an orally bioavailable small-molecule checkpoint inhibitor. Although CA-170 could not block the formation of the PD-1/PD-L1 complex, it promotes the proliferation and activation of effector T cells in the tumor by inhibiting the PD-L1/PD-L2 and VISTA signaling pathways ^{13, 14}. Several other small-molecule inhibitors of PD-L1 were discovered by modifying the first nonpeptidic chemical inhibitors reported by Bristol-Myers Squibb ^{15, 16}. In another patent, Bristol-Myers Squibb reported several anti-PD-L1 macrocyclic peptides with the peptide-57 and peptide-71 being the most prominent ¹². Very recently, Yin et al. identified the PD-1 mimetic peptide MOPD-1, which exhibits nanomolar binding affinity to PD-L1 and blocks the PD-1/PD-L1 interaction ¹⁷.

We recently discovered a low-molecular-weight linear anti-PD-L1 peptide using phage display. The peptide specifically blocks the PD-L1/PD-1 interaction and shows promising anti-tumor activity in a mouse model of colon cancer ¹⁸. However, like most peptides, a major impediment to the advancement of the peptide is its short serum half-life and relatively low affinity compared to monoclonal antibodies. Various approaches have been developed for improving the stability and efficacy of therapeutic peptides. Alanine scanning remains the mainstay in peptide structure-activity relationship (SAR) studies. Other strategies like glycine and proline scanning, peptide truncation, N-methylation, D-amino acid substitution, and cyclization have also been adopted in peptide structure analysis ^{19, 20}.

In this study, side-chain macrocyclization scanning is utilized to optimize the linear anti-PD-L1 peptide discovered by our group. The two terminal amino acids in the linear peptide were mutated with cysteine, and the position of one or both cysteines was sequentially varied. This was followed by a disulfide bond formation between the two cysteine residues to form the corresponding cyclic peptide. Screening for PD-L1 blocking activity led to the identification of two cyclic peptides with enhanced anti-PD-L1 activity. These macrocycles also demonstrated an improvement in serum stability. The improved bioactivity and enhanced proteolytic stability culminated in a superior therapeutic efficacy of the cyclic peptides even at a low dose of 0.5 mg/kg.

Results

Screening of Anti-PD-L1 Cyclic Peptides Using Blocking Assay

Using macrocyclization scanning, we designed a series of cyclic peptide derivatives (Figure 1A) of the parent linear peptide CLP-2, which was recently discovered by our group.²¹ We first evaluated the blocking efficiency of these peptides against the human PD-L1/PD-1 protein interaction. The blocking efficiency of each peptide at 5 μM was normalized against the parent peptide CLP-2. As shown in Figure 1B, cyclization scanning from the N-terminus (C1-C6 peptides) either maintained or decreased the blocking activity. However, scanning from the C terminal (C7-C13 peptides) resulted in cyclic peptides with enhanced blocking activity. Specifically, peptides C7 and C12 exhibited about a 1.7-fold and a 2-fold increase in the blocking efficiency, respectively. Interestingly, no substantial improvement in inhibitory effect was observed when a cyclization scan was performed from the C-terminal with a simultaneous truncation of the flanking residues (C14-C19 peptides). This suggests that size may play a crucial role in blocking the PD-L1/PD-1 interaction. Finally, a cyclization scan performed on both N and C termini (C20-C22 peptides) did not result in macrocycles with improved activity.

Figure 1. Screening of anti-PD-L1 cyclic peptides.

(A) Sequences of cyclic peptides designed by macrocyclization scanning. (B) Normalized blocking efficiency for anti-PD-L1 cyclic peptides at 5 μM. Data are presented as mean ± SD (n=3). Chemical structures of cyclic peptides C7 (C) and C12 (D).

The addition of a cysteine residue to either the N (Cys-CLP2: CWHRSYYTWNLNT) or C (CLP2-Cys: WHRSYYTWNLNTC) terminus does not significantly change the blocking efficiency of CLP2 (Figure S5A,D,E). To validate our blocking assay, the blocking efficiency of BMS peptide-57 was determined at different concentrations using the same method. Peptide-57 exhibited a dose-dependent blocking activity (Figure S5B). Also, the blocking efficacy of a retro-inverso sequence of the C7 peptide [d(CNLNWTYYSRHC)] was compared with that of C7. Introducing D-amino acids in a reversed C7 sequence resulted in loss of the blocking efficiency (Figure S5C,F). Retro-inverso peptide has a similar side-chain topology to its parent L-peptide with inverted amide peptide bonds; as such, some studies have suggested that retro-inverso peptides have similar activity to parent peptides. However, there have been instances where retro-inverso peptides fail to improve or maintain the bioactivity of parent peptides ²². The result of this study suggests that C7 is very specific, and retro-inverso isomerization leads to loss of blocking activity. Details of peptide synthesis and characterization are available in the Supplementary Information. Table S1 describes the characterization of synthesized cyclic peptides.

Blockade of the PD-1/PD-L1 interaction

We next assessed the half-maximal inhibitory concentration (IC₅₀) of the peptides against human PD-1/PD-L1 and mouse PD-1/PD-L1 interactions using a protein-based blocking assay. Evaluating the IC₅₀ of the peptides is a means to quantify the in vitro potency of the peptides in blocking the PD-1/PD-L1 protein-protein interaction. As shown in Figure 2A–C, the IC₅₀ value of the parent peptide CLP-2 against the human PD1/PD-L1 interaction is 6073 nM, while the IC₅₀ values of the cyclic peptides C7 and C12 are 180 and 440 nM, respectively. This result clearly demonstrates that cyclization improves the blockade effect of the parent anti-PD-L1 peptide. Because in vivo anti-tumor activity study needs to be carried out in a syngeneic mouse cancer model, it is imperative to examine the blocking efficacy of the peptides against the mouse PD-1/PD-L1 interaction. The blockade activities of the peptides C7, C12, and CLP-2 against the mouse PD-1/PD-L1 are similar to that observed in the human proteins (Figure 2D–F).

Figure 2. Blockade efficiency of anti-PD-L1 cyclic peptides against the PD-1/PD-L1 interaction.

Blocking profiles of the peptides C7 (A), C12 (B), and CLP-2 (C) against the human PD-1/PD-L1 interaction. Blocking profiles of the peptides C7 (D), C12 (E), and CLP-2 (F) against the mouse PD-1/PD-L1 interaction. Data are presented as mean ± SD (n = 3).

Competitive Surface Plasmon Resonance (SPR)

We next used competitive SPR to verify the blockade effect of the cyclic anti-PD-L1 peptides. A major advantage of SPR over other analytical methods is that it allows for the measurement of interactions between analytes and ligands without the need of labeling and colorimetric reactions. This is very advantageous because labeling (e.g., biotinylation of proteins in ELISA) may potentially compromise the native functionality of the proteins to be analyzed. In a traditional SPR, a ligand is immobilized on the surface of a sensor chip, and an analyte of interest is injected onto the chip. Interaction of the analyte with the immobilized ligand results in an increase in the refractive index of the medium on the chip surface. This leads to a change in the angle of reflection of reflected light from the light source. The change is captured by a detector and translated as an SPR response. In competitive SPR (Figure 3A), anti-PD-L1 peptides are incubated with PD-L1 proteins in the running buffer and subsequently block the interaction between PD-L1 and immobilized PD-1 proteins. Consequently, the SPR response decreases in the presence of anti-PD-L1 peptides in the running buffer (Figure 3A).

Figure 3. Competitive SPR to investigate the blocking activities of peptides against PD-1/PD-L1 interaction.

(A) Schematic representation of competitive SPR. SPR sensorgrams of the peptides C7 (B), C12 (C), and CLP-2 (D) at different concentrations. Blocking curves of the peptides C7 (E), C12 (F), and CLP-2 (G). Data are presented as mean ± SD (n = 3).

As illustrated in Figures 3B–D, maximum SPR response was observed when only the PD-L1 protein was injected, and the lowest response was observed following an injection of the PD-L1 protein in the presence of the highest concentration of anti-PD-L1 peptides. This suggests that these peptides block the PD-L1/PD-1 interaction in a concentration-dependent manner. The competitive SPR IC₅₀ for C7, C12, and CLP-2 are 119, 143, and 1361 nM, respectively (Figure 3E–G). Both cyclic analogues have better anti-PD-L1 activity than their linear counterpart. This result is consistent with the blocking studies using the blocking assay (Figure 2).

Binding Specificity of Anti-PD-L1 Peptides to PD-L1-Expressing Cells

Binding specificity is one of the most crucial considerations in the design of protein-protein interaction (PPI) modulators. To be an effective PPI modulator, drug candidates must specifically bind to their intended protein target but interact minimally with other molecules in the body. Drug molecules with low specificity usually have indiscriminate binding interactions with unintended targets, resulting in unwanted side effects. Cyclization of small peptides introduces conformational rigidity in linear peptides thereby reducing the entropic cost of binding as cyclized peptides may be pre-organized into a bioactive conformation ²³. Therefore, cyclization improves the binding affinity and selectivity of linear peptides to target biomolecules. In this study, we assessed the binding specificity of the cyclic anti-PD-L1 peptides to PD-L1 proteins expressed on various cancer cells. The peptides were incubated with PD-L1-overexpressing cells (DU145 and MDA-MB-231 cells) and PD-L1-deficient cells (MCF-7 and MDA-MB-231 PD-L1 knockout cells). As illustrated in Figures 4A–B, both the cyclic peptides C7 and C12 exhibit dramatically higher binding to MDA-MB-231 cells than to MDA-MB-231 PD-L1 knockout cells, suggesting high specificity of these peptides to PD-L1 expressed on cancer cells. By contrast, the peptide CLP-2 also shows a higher binding affinity to MDA-MB-231 cells, but the specificity is not as high as the cyclic peptides (Figure 4C). A similar result was observed in DU145 cells and MCF-7 cells (Figure 4D–F). These results demonstrate improved specificity of the cyclic peptides to PD-L1 expressed in cancer cells.

Figure 4. Binding of anti-PD-L1 peptides to human PD-L1-overexpressing cells (MDA-MB-231 and DU145) and PD-L1-deficient cells (MDA-MB-231 PD-L1 knockout and MCF-7).

Binding of Cy5-labeled peptides C7 (A), C12 (B), and CLP-2 (C) to MDA-MB-231 and MDA-MB-231 PD-L1 knockout cells. Binding of Cy5-labeled peptides C7 (D), C12 (E), and CLP-2 (F) to DU145 and MCF-7 cells. Data are presented as mean ± SD (n=3). (* p < 0.05; ** p < 0.01)

Cyclic Anti-PD-L1 Peptides Reinvigorate PBMCs Co-cultured with DU145 Cells

PD-L1 overexpressed on the tumor cell surface interacts with PD-1 on immune cells to provoke an inhibitory effect on the activation and proliferation of immune cells. It also promotes exhaustion, anergy and apoptosis of immune cells ²⁴. In this study, we investigated the ability of the cyclic anti-PD-L1 peptides to reinvigorate immune cells co-cultured with PD-L1-overexpressing cancer cells. The results show that PBMCs cultured alone exhibited a very low degree of apoptosis, but the rate of apoptosis is increased significantly when co-cultured with DU145 cells. However, incubation of the co-cultured cells with the anti-PD-L1 peptides reduces the apoptosis of the immune cells (Figure 5A,B).

Figure 5. Cyclic anti-PD-L1 peptides reinvigorate PBMCs co-cultured with DU145 cell.

(A) Annexin V-FITC/PI flow cytometry analysis of PBMCs co-cultured with DU145 cells in the presence of anti-PD-L1 peptides (10 μM) or antibody (2 μM). (B) Percentage of apoptotic PBMCs co-cultured with DU145 cells in the presence of anti-PD-L1 peptides (10 μM) or antibody (2 μM). Flow cytometry analysis (C) and bar graphs (D) showing percentage of CD8⁺/IFN-γ⁺ T cells from PBMCs co-cultured with DU145. Flow cytometry analysis (E) and bar graphs (F) showing percentage of CD4⁺/IFN-γ⁺ T cells from PBMCs co-cultured with DU145. Data are presented as mean ± SD (n = 3). (* p < 0.05; ** p < 0.01)

Cytokine Assay

Cytokines play an important role in modulating host immune activity. Assessing cytokine levels during immunotherapy is one way to evaluate therapeutic response and the onset or extent of adverse effects [18]. IFN-γ is a pleiotropic cytokine that possesses cytostatic, pro-apoptotic, immunomodulatory, and antiproliferative functions ²⁵. In general, IFN-γ is mostly produced by CD4⁺ and CD8⁺ T cells during an adaptive immune response while in innate immunity, natural killer cells modulate IFN-γ production ²⁶. High levels of IFN-γ may also promote immune escape by upregulating PD-L1 expression ²⁷, serving as a potent immune evasion mechanism in tumor cells ²⁸. Nonetheless, IFN-γ-mediated responses are still associated with positive prognosis in several cancers. Studies have revealed that blocking the PD-1/PD-L1 interaction can increase IFN-γ secretion ^{29, 30} by cytotoxic T-lymphocytes with an associated anti-tumor activity. For instance, in a study by Li et al. ³¹, bacterial surface display was used to identify PD-L1-targeting peptides. These peptides exhibited significant anti-tumor activity in vivo, and the peptide-treated group showed relatively a high level of IFN-γ compared with that of the untreated group, suggesting reactivation of T cells. These results are consistent with our findings that the anti-PD-L1 peptide C7 increases IFN-γ levels in CD8⁺ T cells (Figure 5C,D). It is worthy to note that both the anti-PD-L1 antibody and the peptide C7 increase the percentage of IFN-γ⁺/CD8⁺ T cells from 1.3% of the control group to 1.7%, which is equivalent to 31% upregulation. Cytotoxic T cells produce cytokines like IFN-γ upon activation, leading to elimination of cancer cells. Therefore, the ability of the cyclic peptides to enhance IFN-γ production by CD8⁺ T cells can partly be associated with improved anti-cancer activity.

Cell Viability Assay

Immune checkpoint inhibitors block immune suppressive proteins like PD-L1 expressed on tumor cells from interacting with their receptor PD-1 on T cells. This allows the immune cells to remain in an activated state thus exerting their cancer killing activity ³². Therefore, checkpoint inhibitors do not directly kill tumor cells but rather restore the anti-tumor activity of cytotoxic T cells. Here, we investigated whether the anti-PD-L1 peptides directly induce cytotoxicity of tumor cells. As shown in Figure 6A, these anti-PD-L1 peptides have negligible effect on cell viability. This finding agrees with other studies investigating the cytotoxicity of anti-PD-L1 agents. For example, Ganesan et al. assessed the cytotoxicity of a series of anti-PD-L1 peptides and small molecules in HepG2, CHO, and Jurkat cells using the CellTiter-Glo^® Luminescent Cell Viability Assay. While the small molecules tested elicited significant cytotoxicity, the anti-PD-L1 peptides did not reduce the viability of the cells ³³.

Figure 6. Cytotoxicity and serum stability of cyclic anti-PD-L1 peptides.

(A) Cytotoxicity of peptides C7, C12, and CLP-2 in DU145 cells at various concentrations. Triton X-100 (0.5% v/v) is used as positive control. Percentage of remaining peptides CLP-2 (B), C7 (C), and C12 (D) in human serum. Data are presented as mean ± SD (n = 3).

Serum Stability

Relatively poor serum stability of peptides is a major limitation of peptide-based therapeutics. As a result, serum stability is one of the most critical screening tests in the early stages of peptide drug development. We hypothesize that cyclization of the linear CLP-2 peptide improves its serum stability. In this study, we performed serum stability of the peptides in human serum to evaluate the effect of cyclization on the proteolytic stability of native peptides. As observed in Figure 6B, the linear CLP-2 peptide shows a half-life of 18.8 min in 50% human serum, and the peptide is nearly completed degraded after 3 h. By contrast, the cyclic peptides C7 and C12 exhibit better serum stability with half-lives of 38 and 70 min, respectively. C12 shows the best serum stability, and nearly 25% of the C12 peptide remain intact after 3h.

Anti-Tumor Activity

In vivo therapeutic efficacy of the cyclic anti-PD-L1 peptides was evaluated in a CT26 syngeneic Balb/c mouse model. In our previous study, the parent peptide CLP-2 exhibited significant anti-tumor activity at 2 mg/kg.¹⁸ Compared to the CLP-2 peptide, cyclic peptides C7 and C12 showed better blocking efficiency, higher specificity, and improved serum stability. We, therefore, reduced the dose of peptides to 0.5 mg/kg in this study (Figure 7A). The antibody was also administrated at the dose of 0.5 mg/kg. The peptides and antibody were intraperitoneally injected daily from day 10 to day 18. As shown in Figure 7B, both cyclic peptides and the anti-mouse PD-L1 antibody groups exhibit a significant anti-tumor activity compared with the saline group. By contrast, CLP-2 at the current dose of 0.5 mg/kg does not show significant tumor suppression in comparison with the control. The cyclic peptides demonstrated better tumor growth inhibition (TGI) compared to other groups (Figure 7C). TGI was calculated using the formula TGI (%) = [1 − (relative tumor volume (RTV) of the treated group on day 18)/(RTV of the control group on day 18)] × 100 (%). RTV = (tumor volume on day 18)/(tumor volume on day 10).³⁴

Figure 7. Anti-tumor activity of the anti-PD-L1 peptides.

(A) Balb/c mice bearing CT-26 tumor cells received anti-PD-L1 peptides or antibody (0.5 mg/kg) by intraperitoneal injection daily. (B) Measurement of tumor growth over time. (C) Tumor growth inhibition (TGI) percentage on day 18. (D) Survival curves. Tumor progression curves of individual mice (red-female, black-male) in (E) saline, (F) CLP-2, (G) C12, (H) anti-PD-L1 antibody, and (I) C7 groups, respectively. ( ** p < 0.01)

Survival study showed that all anti-PD-L1 peptides and the antibody improved the survival of the mice when compared with the saline group (Figure 7D–I). However, the CLP-2 peptide did not significantly improve the survival of the mice. Both cyclic peptides C7 and C12 showed significant improvement in survival of the mice. The anti-tumor activity result is consistent with the in vitro affinity, specificity, activity, and serum stability results.

Discussion and Conclusions

We have applied macrocyclization scanning in the optimization of a native linear anti-PD-L1 peptide to obtain cyclic analogues with improved therapeutic efficacy. Using various in vitro assays, we demonstrated that these cyclic derivatives could bind to and block the PD-L1 checkpoint ligand from interacting with its PD-1 receptor. Macrocyclization resulted in modified peptides that markedly inhibit the growth of CT26 colorectal tumor in a syngeneic mouse model and prolong the survival of the mice.

Linear peptides generally possess a high degree of conformational flexibility, allowing them to exhibit a variety of conformations in a solution. This could be disadvantageous as therapeutics as there is a high propensity for off-target binding, leading to low efficacy and increased risk of side effects. On the other hand, cyclization leads to derivatives with a fixed geometry, thereby reducing the entropic cost of binding. Consequently, cyclic peptides can bind more efficiently and selectively to targets ^{8, 35, 36}.

There is currently enormous interest in using constrained peptides as the suitable alternative to mAbs. This is mainly because cyclic peptides, unlike mAbs, are easy to synthesize and chemically modify, safer, and well tolerated. Also, the backbone amides of cyclic peptides undergo extensive intramolecular hydrogen bonding in hydrophobic environment like the lipid bilayer of a cell membrane. This reduces their effective polar surface area and thus affords constrained peptides a high cell and tissue permeability ^{37, 38}.

Macrocyclization scanning is an approach of introducing two cysteine residues in a peptide sequence followed by a sequential change in the position of the cysteines to generate a cyclic peptide library ³⁹. Building on this concept, we designed a series of cyclic peptides using macrocyclization scanning as a strategy to improve the bioactivity and proteolytic stability of the linear anti-PD-L1 peptide CLP-2 previously reported by our group. Employing this approach for lead optimization is beneficial in that it does not involve the use of molecular docking simulations which are often complex and less reliable. Like many scanning techniques, our strategy can be useful when structural characterization of the peptide-receptor complex is lacking. Macrocyclization scanning can be a powerful tool in peptide drug development as it provides an effective means for gaining insight into the relationships between structure, activity, and conformation of therapeutic peptides.

Numerous studies have demonstrated that cyclization confers improvement in the binding affinity, target selectivity, proteolytic stability, and therapeutic efficacy of linear peptides and peptidomimetics ^{37, 40}. For instance, Kumar et al. showed that head-to-tail cyclization of Ac-CIYKYY (a linear peptide), an Src tyrosine kinase inhibitor, improved its inhibitory potency by more than 60 folds ⁴¹. Chen et al. also employed cyclization as a tool to reduce the entropic penalty of linear peptide inhibitors of the nuclear factor-erythroid 2-related factor 2 (Nrf2)/Keap1 (Kelch-like ECH-associated protein 1) interaction. Cyclization resulted in a 1.4-7.5-fold increase in binding affinity of the native peptide ⁴². Cyclization has also been applied to enhance the biological activity and stability of naturally occurring peptides like the antimicrobial peptide gomesin ⁴³. In this study, as revealed by the blocking IC₅₀ assay, cyclization resulted in nearly a 34-fold improvement in PD-L1 blocking activity.

SPR is an optical detection platform that allows for a label-free and real-time analysis of molecular binding interactions ^{44, 45}. In competitive SPR, a mixture of an analyte and its inhibitor is injected to assess the ability of the inhibitor to block the ligand-analyte binding ⁴⁶. This is particularly useful when the inhibitor is a small molecule which has a weak SPR signal on its own. The competitive SPR data suggest that the cyclic peptides exhibit better PD-L1 blocking activity when compared with the parent linear peptide. These results are in congruence with the assertion that cyclization of a linear peptide in the correct conformation reduces the entropic cost of binding to the target of interest thus increasing bioactivity ^{23, 47}.

Therapeutic peptides are generally regarded as safer than most small-molecule drugs. This is attributable to the fact that therapeutic peptides have amino acid building blocks that the body uses in protein synthesis. PD-1/PD-L1 inhibitors are known to promote tumor suppression by blocking the interaction of PD-L1 on tumor cells with its receptor, PD-1 on T cells, mostly without direct cytotoxicity to the cancer cells. This immune checkpoint blockade allows T cells to exert cytotoxic effects against cancer cells. Similarly, the cyclic peptides did not exhibit cytotoxicity in DU145 cancer cells cultured alone. However, when DU145 cells were cocultured with PBMCs, the cyclic peptides inhibited the proliferation of the DU145 cells (Figure S6). Also, these peptides are less likely to exert direct cytotoxicity on normal healthy cells, a dilemma associated with most chemotherapy agents.

A high conformational fluidity allows linear peptides to exist in multiple conformations in a solution. This makes the peptide more likely to exist in a conformation that fits the active site of proteolytic enzymes, ultimately resulting in an accelerated degradation. On the other hand, the rigidity introduced via cyclization reduces the rate of degradation by rendering peptides inaccessible to, or unrecognizable by, the active sites of proteolytic enzymes ³⁶. The linear peptide in this study showed an in vitro serum half-life of approximately 19 min in 50% human serum. Cyclization resulted in nearly a 4-fold improvement in half-life, suggesting that cyclization improves the stability of linear peptides. Our results conform to the findings of other previous studies. For instance, Etayash and colleagues investigated the proteolytic stability of IDR-1018, an immunomodulatory peptide, and its cyclic analogs in trypsin ⁴⁸. The native peptide was degraded in less than 30 min, while the cyclic derivatives demonstrated greater resistance to trypsin degradation for up to 120 min. Elsewhere, Pakkala et al. investigated the effect of cyclization on the stability of human kallikrein-2-specific peptide inhibitors. Backbone cyclization of one of the peptides (KLK2b) conferred a significant improvement in proteolytic stability against trypsin and human plasma ⁴⁹.

Although our previous in vivo anti-tumor activity study using a murine model showed that CLP-2 has good anti-tumor activity at a dose of 2 mg/kg, it did not significantly reduce tumor size or improve the survival at 0.5 mg/kg in this study. By contrast, both cyclic peptides showed superior efficacy compared to the parent peptide CLP-2. The cyclic peptides demonstrated a significant reduction in tumor volume when compared with the saline group. It can be inferred that the improved bioactivity and enhanced proteolytic stability have been translated into a superior therapeutic efficacy of the cyclic peptides even at a low dose of 0.5 mg/kg. Also, the cyclic peptides were able to prolong the survival of CT26 tumor-bearing mice. With these exciting results, we are currently assessing the anti-tumor activity of these peptides in other cancer models either alone or in combination with other anti-cancer agents.

Applying macrocyclization scanning in a unique way, we have designed novel cyclic analogues of the linear anti-PD-L1 peptide CLP-2 with improved proteolytic stability and PD-L1 blocking activity. Our findings suggest that cyclization of the linear anti-PD-L1 peptide in the correct format improves in vivo therapeutic efficacy. This strategy may be applied in the optimization of other bioactive native peptides, particularly those for protein-protein interaction modulation.

Experimental Section

Cell Culture

MDA-MB-231, MCF-7, DU145 and CT26 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). PD-L1 knockout MDA-MB-231 cell line was kindly provided by Dr. Mien-Chie Hung at the MD Anderson Cancer Center. DMEM containing 10% fetal bovine serum (FBS), 100 μg/mL streptomycin, and 100 units/mL penicillin were used in culturing MDA-MB-231 and DU145 cells. Fresh human PBMC (CXPR-15-200M, iXCells biotechnologies, San Diego, CA), CT26, and MCF-7 cells were cultured in RPMI1640 medium with 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. Recombinant human insulin (0.01 mg/mL) was added to the medium for MCF-7 cells. The cells were incubated at 37 °C in a humidified atmosphere containing 5% CO₂.

Blocking Assay and IC₅₀ Determination

The human PD-1 (Biotinylated): PD-L1 Inhibitor Screening ELISA Assay Pair kit (Cat#EP-101, Acro Biosystems, Newark, DE) was used in the blocking assay and IC₅₀ determination. Briefly, 100 ng of recombinant human PD-L1 protein was coated in 96-well plates at 4°C for 16 h with shaking over an orbital shaker. Unbound protein was removed by washing twice with PBS (pH 7.2-7.4), and 2% BSA was then added for 1.5 h at 37°C to block the plate. Specific concentrations of the peptides were dispensed into the appropriate wells and incubated at 37°C for 2 h. All peptides used in this study have a purity of >95 % as determined by HPLC. Biotinylated PD-1 protein was subsequently added and incubated for 1.5 h, followed by the addition of streptavidin-HRP and chromogenic substrate (R&D Systems, Minneapolis, MN). The color reaction was terminated with 2N H₂SO₄, and absorbance was recorded at OD₄₅₀ and referenced to OD₅₄₀. The blocking activity of the peptides was also assessed using mouse PD-1 (Biotinylated, cat# PD1-M82F4, AcroBiosystems) and PD-L1 (Cat# PD1-M5220, Acro Biosystems) following the same procedure as the human proteins.

Competitive Surface Plasmon Resonance

BI-4500A SPR and a CM-5 sensor chip pre-functionalized with streptavidin (Biosensing Instrument Inc., Tempe, AZ) were used for the SPR analysis. Biotinylated human PD-1 protein (Cat# 71109, BPS Bioscience, San Diego, CA) was immobilized on the sensor chip to give a coating of approximately 750 Response Units. Different concentrations of each peptide were incubated with human PD-L1 protein (Cat# EP-101, AcroBiosystems) prior to injection. Human PD-L1 alone was also injected to serve as a control. The ability of the peptides to block the PD-L1/PD-1 interaction was assessed by the change in SPR response between the control and peptide samples. The data were analyzed with SPR Data Analysis Version 3.8.4 (Biosensing Instrument Inc., Tempe, AZ).

Cell Binding Assay

Binding of the peptides to PD-L1-overexpressing cancer cells (DU145 and MDA-MB-231) and cancer cells with low PD-L1 expression (MCF-7 and PD-L1 knockout MDA-MB-231) was evaluated as described previously with some modifications ⁵⁰. Enzyme-free cell dissociation buffer (Cat# 13-151-014, Gibco BRL, Gaithersburg, MD) was used to detach the cells. The detached cells were suspended in PBS at a density of 1×10⁶ cells/mL. Half a milliliter of the suspended cells was incubated with various concentrations of Cy5-labeled peptides for 1 h at 37 °C with low-speed rotation. Following centrifugation at 300g for 10 min, the cells were washed, resuspended in PBS, and analyzed using a FACScalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ).

Cell Viability Assay

In vitro cell viability assay was performed as previously reported with modifications ⁵¹. Briefly, white 96-well plates with a clear bottom were seeded with DU145 cells at a density of 1×10⁴ cells/well and incubated in a humidified atmosphere at 37 °C with a supply of 5% CO₂. After 12 h, peptides with various concentrations were added to the cells and incubated for 48 h. The CellTiter-Glo^® Luminescent Cell Viability Assay kit (Promega, Madison, WI) was employed in the cell viability analysis following the manufacturer’s instructions.

Apoptosis and Cytokine Assay of Co-cultured PBMCs and DU145 Cells

The ability of the peptides to enhance immune cell functional activity was assessed. 7.5×10⁵ fresh human peripheral blood mononuclear cells (PBMCs) (Cat# 10HU-003, iXCells Biotechnologies, San Diego, CA) were cultured either alone or with 1.5×10⁵ DU145 cells in 6-well plates. This was followed by the addition of peptides or antibodies to appropriate wells. After 24h incubation, the supernatant containing PBMCs was collected and stained with APC-labeled anti-CD-3 antibody (Cell Signaling Tech., Danvers, MA). Apoptosis of CD-3⁺ cells was analyzed using a Dead Cell Annexin V-FITC apoptosis kit (Cat# V13241, Invitrogen, Waltham, MA) with a FACScalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ).

To assess the level of CD4⁺/IFN-γ⁺ and CD8⁺/IFN-γ⁺ cells, PBMCs were co-cultured with DU145 cells as described in the apoptosis assay. After incubation for 24 h, the cells were washed with PBS and stained with Fixable Viability Dye eFluor^™ 780 (Cat# 65-0865-14, Life Tech. Corp, Carlsbad, CA) to exclude dead cells. Subsequently, the cells were surface stained with a cocktail of mouse anti-human CD3-PerCP-Cy5.5, CD4-PE, and CD8-APC (Cell Signaling Tech., Danvers, MA) monoclonal antibodies in PBS with 0.5% BSA for 20 min at 4 °C. For intracellular staining, the surfaced-stained cells were fixed with FluoroFix^™ Buffer (Cat# 422101, Biolegend, San Diego, CA) for 20 min at room temperature. The fixed cells were permeabilized with a permeabilization buffer (Cat# 68751S, Cell Signaling Tech., Danvers, MA) for 5 min and stained with mouse anti-human IFN-γ-Alexa Fluor^® 488 monoclonal antibody (Cat# 557718, BD Biosciences, Franklin Lakes, NJ) for 20 min at room temperature. The stained cells were analyzed on a FACScalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ).

Serum Stability

The stability of the peptides in human serum was investigated using LC-MS/MS on an AB Sciex 4000 QTRAP mass spectrometer equipped with a Shimadzu Ultra-Fast Liquid Chromatography (UFLC) system. Solutions of the anti-PD-L1 peptides were incubated with 50% human serum at 37°C, and aliquots were taken at specific time points for analysis.

Anti-Tumor Activity and Survival Study

The animal protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Missouri-Kansas City. A total of 20 male and 20 female Balb/c mice purchased from Charles Rivers Laboratories (Wilmington, Massachusetts) were housed in a temperature and humidity-controlled room on a 12 h light-dark cycle. Approximately 1×10⁶ CT26 cells were subcutaneously implanted into the right flank of the mice. When the tumor size reached 200-250 mm³, the mice were divided into five different groups (4 females and 4 males each group). One group received normal saline administered intraperitoneally as a negative control. Animals in each of the other groups received 0.5 mg/kg of either CLP-2, C7, C12, or anti-mouse PD-L1 antibody (Cat# BE0101, BioXCell, Lebanon, NH) daily. Tumor size was measured with a pair of calipers and the volume was calculated with the formula 1/2 × length×width².

Statistical Analysis

Statistical analyses were performed using the statistical software GraphPad Prism 6.0. All data were presented as means ± SD (or mean ± SEM for anti-tumor activity data). One-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison test was used to compare data with more than two groups, while Student’s t-test was used to compare two independent groups. Statistical analyses of survival curves were carried out with the Gehan-Breslow-Wilcoxon test. p<0.05 is considered statistically significant.

Abbreviations

APC

allophycocyanin

BSA

bovine serum albumin

ELISA

enzyme-linked immunosorbent assay

FACS

fluorescence activated cell sorting

FBS

fetal bovine serum

FDA

Food and Drug Administration

FITC

Fluorescein isothiocyanate

IC₅₀

half maximal inhibitory concentration

LC-MS

liquid chromatography-mass spectrometry

mAbs

monoclonal antibodies

PBMCs

peripheral blood mononuclear cells

PD-1

programmed cell death protein 1

PD-L1

Programmed death-ligand 1

PE

Phycoerythrin

PerCP

Peridinin Chlorophyll Protein Complex

SAR

structure-activity relationship

SPR

Surface Plasmon Resonance

UFLC

ultra-fast liquid chromatography