Synthetic integrin-binding immune stimulators target cancer cells and prevent tumor formation

Immuno-oncology approaches mainly utilize monoclonal antibodies or protein-based scaffolds that bind with high affinity to cancer cells and can generate an immune response. Peptides can also bind with high affinity to cancer cells and are intermediate in size between antibodies and small molecules. They are also synthetically accessible and therefore easily modified to optimize their stability, binding affinity and selectivity. Here we describe the design of immune system engagers (ISErs), a novel class of synthetic peptide-based compounds that bind specifically to cancer cells and stimulate the immune system. A prototype, Y9, targets integrin α3, which is overexpressed on several cancer cells, and activates the immune system via a formyl methionine-containing effector peptide. Injection of Y9 leads to immune cell infiltration into tissue and prevents tumor formation in a guinea pig model. The anti-tumor activity and synthetic accessibility of Y9 illustrate that ISErs could be applied to a wide variety of targets and diseases.

per PEG 27 25 ) was chosen to cover similar distances as the two paratopes in an antibody (Fig. 1A). By combining two binder peptides per ISEr, we aimed to avoid the fast dissociation and low retention times that can reduce the efficacy of even high-affinity monovalent binders under non-equilibrium physiological conditions 26 .
We selected an immune stimulatory effector peptide based on an N-formyl methionine containing peptide employed to activate the innate immune system via granulocytes, monocytes and macrophages (denoted as F1 or F2, Fig. 1). Such peptides are typically of bacterial origin and are well-known to elicit an innate immune response by interaction with various immune cell receptors such as three members of the N-formyl-peptide receptor family (FPR1-3) [27][28][29] . Previous experiments in which a formyl-methionyl-leucyl-phenylalanine peptide (fMLF) was covalently linked to an IgG antibody induced monocyte chemotaxis 30 and a two-fold increase in macrophage infiltration of hepatomas and a decrease in tumor weight in guinea pigs 31 . At the same time no relevant toxicity in a human phase I clinical study was observed 32 .
The selected binder peptide (denoted as B9, Fig. 1) binds with high affinity to the integrin α3 chain and was identified by one-bead-one-compound (OBOC) screening 33 . The eight-residue binder consists of several non-proteinogenic amino acids and has been evolved as a peptidomimetic, cyclized via two terminal cysteine residues, using the one-bead-one-compound approach 34 .
Here, we describe the synthesis and activity of a prototypical ISEr abbreviated Y9 (Y denotes the trimeric functionality of the molecule) that targets specific cell surface structures (here the integrin α 3 chain) as well as engage the innate immune system (with an N-formyl effector). In a proof-of-concept study, this ISEr effectively prevented tumor formation in an allogenic adenocarcinoma model in guinea pigs.

Results
Effector activation of the immune system. Separate testing of the two components comprising the ISEr Y9, the effector (F2) and the binder moiety (B9), were necessary to select suitable building blocks as well as to learn what changes of the individual properties occur upon incorporation into an ISEr (Figs 1A and 2A). We initially started with testing the well-known N-formyl peptide effector fMLF (denoted as F1, Figure S1) at concentrations between 1-10 nM. fMLF was selected based on prior reports of its use for inducing chemotaxis and immune reactions when linked to antibodies 30,32 . Our experiments confirmed that this short, N-formyl peptide prompts chemotaxis of human leukocytes and superoxide production in human granulocytes (dihydro-rhodamine (DHR) oxidation assay, Figure S2A,B) 35 . However, incorporation of F1 into a suitable scaffold to attach two binder moieties, consisting of a short glycine-glycine linker and two additional lysine residues linked to PEG 27 via their ε-amino groups (F1-2PEG, Figure S1), reduced the potency to induce chemotaxis and neutrophil activation by two orders of magnitude ( Figure S2A,B). This behavior is best explained by the much slower diffusion of the larger molecule and potential occlusion of F1 by the PEG chains. Additional assays indicated that F1 and F1-2PEG are less potent immune activators in mice and guinea pigs ( Figure S2C,D) 36 . In order to incorporate an effector suitable for activation of innate immune responses in humans, mice and guinea pigs, we selected fMIFL (F2), a sequence previously identified from Staphylococcus aureus [37][38][39] . F2 induces chemotaxis and superoxide production of human neutrophils starting at 10 pM concentration ( Fig. 2A,B). Here, we also observe a loss of potency by 1-2 orders of magnitude when linked to the PEG chains (F2-2PEG, Figs 2A,B, S3). However, sufficiently high concentrations of effector to induce an effective immune response could still be reached, as demonstrated by subcutaneous application of F2-2PEG in mice and guinea pigs. Analysis of tissue sections of mouse skin samples 24 h after subcutaneous injection of 100 nmol F2-2PEG clearly showed infiltration by immune cells at the injection site in comparison to control injected animals (Fig. 3C). We could demonstrate that 6 h incubation of PBMCs with 100 nM F2-2PEG increased the levels of IL-1β secreted from human monocytes ( Fig. 3D) similarly to bacterial lipopolysaccharides (LPS). Other cytokine levels such as IL-6, IL-8 and TNFα were not significantly increased by F2-2PEG while the positive control used here (LPS), induced increases of these cytokines (Fig. 3D).
Specific binding to cancer cells. Our selected binder B9 was prepared by Fmoc-based SPPS and carefully analyzed with respect to binding to a panel of human, mouse and guinea pig tumor and blood cells (Tables 1  and S1). For the detection by streptavidin in flow cytometry, B9 was C-terminally labeled with biotin (Fig. 4A). Concentration-dependent flow cytometry measurements using this B9-Biotin demonstrated high affinity binding to A431 tumor cells expressing integrin α 3 β 1 (also known as VLA-3) on their surface (K D ~90 nM, Fig. 3 and S3). B9 competes with a commercially available anti-integrin α3 antibody (anti-CD49c) for the same binding site, thereby unequivocally establishing the target of this peptide binder (Fig. 4C). Binding affinity depends on the addition of MnCl 2 (at 2 mM) during incubation as Mn 2+ induces the high-affinity state of the integrin (Fig. 4B 40,41 ). In the absence of Mn 2+ the apparent K D drops to 176 nM. An overlay of the 1 H-13 C HSQC NMR spectra of reduced (grey contours) and oxidized (black contours) Y9 is shown with Cys Cβ peaks labeled. Additional information on chemical shifts can be found in Table S2.
Additional flow cytometry analyses with a variety of primary cells such as macrophages, fibroblasts and endothelial cells (HUVECs) from humans, mice and guinea pigs were performed to exclude unspecific binding of B9 (Tables 1 and S1). No binding at concentrations below 1 µM was detected on most of the primary cell types, except for a weak, heterogeneous binding signal observed on human macrophages and HUVEC. Expression of integrin α 3 on human, rabbit and guinea pig cells was verified with a mouse anti-human CD49c antibody and correlated with binding of B9 to these cell lines ( Table 1). The CD49c specific antibody did not recognize mouse integrin α 3 , however binding of B9 to most murine tumor cell lines could be demonstrated (Tables 1 and S1).
Synthesis and characterization of ISEr Y9. The ISEr Y9 combines F2 and B9, and was synthesized by Fmoc SPPS. N-terminal formylation of MIFLGGKK on resin was quantitatively achieved with p-nitrophenylformate 42 and followed by Mtt removal from both lysine side chains onto which two PEG 27 linkers were coupled prior to generating identical binder peptides, also via stepwise SPPS. The complete molecule, denoted as Y9, was obtained in 20% yield based on the synthesis scale of 0.2 mmol and purified product (207 mg). Purification via RP-HPLC of reduced Y9 (Fig. 2B) was followed by oxidative cyclization of both binder moieties to give active Y9 with more than 98% purity (Fig. 2C).
To characterize Y9 structurally, 2D NMR data were recorded in aqueous solution at pH 3.0 and chemical shifts were assigned for the reduced form of Y9 (Table S2). Most of the secondary Hα shifts of both the binder and effector peptides in Y9 were smaller than 0.1 ppm, showing that both peptides are in a predominantly random coil conformation ( Figure S5 43 ). NMR data of the oxidized form of Y9 were compared with that of reduced Y9, confirming (in agreement with MS data) the presence of the disulfide bond. Most notably, the Cβ resonance of the cysteine residues in B9 shifted from ~28 ppm in reduced to ~41 ppm in oxidized Y9 (Fig. 2C). in chemotaxis and oxidative burst assays ( Fig. 5A-C). Before using Y9 in any subsequent experiments we excluded endotoxin contamination using a commercially available test system based on the limulus amebocyte lysate (LAL) assay ( Figure S6) 44 . In vivo experiments in mice and guinea pigs showed that local immune cell infiltration occurs upon subcutaneous injections (Figs 5C, S7). Immune cell infiltration into the injection site was most pronounced after the application of 100 nmol Y9 in guinea pigs. Immune infiltrate was also found (but was less intense) after the subcutaneous injection of 100 nmol Y9 into guinea pigs that were immune suppressed by four weekly injections of cyclophosphamide (CPA) as used in a subsequent efficacy study with the allogenic GPC-16 tumor model in guinea pigs ( Figure S7A,B). To explore the possible use of the large variety of xenogeneic tumor models available in mice we also injected Y9 into immune-deficient Balb/c nu/nu mice. Here, dose-dependent increases in immune infiltrates between 200 nmol to 500 nmol could be observed (Fig. 5C). This finding indicates that higher dosages of Y9 are required for inducing an immune response in immune-deficient Balb/c nu/nu mice than for immune-suppressed guinea pigs. Immunohistochemical staining of the skin sections with antibodies against the myeloid cell marker myeloperoxidase (MPO) and the macrophage marker F4/80 demonstrated that the immune infiltrate consisted, at least in part, of granulocytes, monocytes and macrophages ( Figure S7C-F).
Specific integrin binding of ISEr Y9. Binding to integrin α 3 was shown in vitro via surface plasmon resonance (SPR) measurement using immobilized integrin α 3 . Here, a K D value of 60 nM for Y9 was determined (Fig. 5D). This value is in the same range as found for B9-Biotin in previous flow cytometry measurements of the binder moiety alone and agrees with measurements of biotinylated Y9 on cells. To determine binding constants on integrin α3β1 expressing cells, we used biotinylated Y9 (biotin attached to the N-terminus of the effector) and determined an apparent K D of 244 nM for binding to A431 cells (Fig. 5E). This three-fold decrease in affinity when compared to B9-Biotin could be due to an occluding effect of the PEG spacers. To test this hypothesis we used our scaffold (fMIFLGGKK with two PEG 27 chains) just carrying one B9 binder peptide. B9 competed with 200 nM B9-Biotin at an IC 50 of 318 nM whereas the scaffold with only one B9 binder achieved the same effect only at 1533 nM. This confirmed our hypothesis that the PEG spacers have a negative impact on binding affinity. In competition experiments Y9 displaced B9-Biotin at a four-fold lower concentration than the monovalent binder (380 nM), which also demonstrated the positive effect of bivalency in Y9 ( Figure S8). Further analysis of binding of B9 and Y9 under physiological conditions (37 °C) revealed that Y9 remains present on the surface of various cancer cell lines longer than B9 ( Figure S9). B9-Biotin bound to PC-3 or A431 cells disappeared from cells quickly when incubated at 37 °C and no signal above background was detected by streptavidin staining after 5 min. However, biotinylated Y9 remained on cell membranes with t 1/2 of at least 15 minutes ( Figure S9). To exclude contributions by the effector F2 to binding of Y9 to A431 cells, staining with a formyl peptide receptor 1 (FPR-1)-specific antibody confirmed that no FPR-1 was present on these cells ( Figure S10). Stability and availability of ISEr Y9. The amount of effector available to attract and stimulate immune cells upon administering Y9 is not only affected by internalization and dissociation but also by its degradation. To this end, we have tested stability of Y9 in mouse serum using extraction and HPLC-based quantification ( Figure S11). Y9 remains largely intact for extended periods of time (>24 h) by combining the protective properties of the PEG chains against proteolysis as well as by the presence of several non-proteinogenic amino acids in the binder peptides and their cyclic structure ( Fig. 2A). After 48 hours only 20% of Y9 were degraded. Bioavailability of Y9 in vivo was tested in sera taken from mice after subcutaneous application of Y9, using a sensitive dot-blot method and a custom-made antibody against B9 (detection limit ~1 nM, Figure S12)  Figure S13). To further confirm that Y9 can be safely applied in animals, we performed initial dose escalation studies testing acute toxicity after repeated application of Y9 in immune-competent Balb/c mice. Five animals per group were subcutaneously injected with 200, 400 or 800 nmol Y9 every two days for 14 days. A control group of ten mice received 0.9% NaCl. Four days after the last application of Y9, blood and serum parameters as well as relative organ weights were measured and tissue samples of organs were analyzed by histopathology. As expected Y9-treated animals exhibited local effects such as increased neutrophil and monocyte infiltrates at the injection site and, correlating with this observation, a mild but significant dose-dependent immune infiltration and lymphoid hyperplasia in the draining lymph nodes was found. No histopathological effects were detected in any of the organs tested (liver, kidney, pancreas, spleen, heart, lung, ovaries, brain, stomach, intestine, mesentery) after repeated 200 and 400 nmol Y9 injections ( Figure S14). Only with the highest dose of 800 nmol moderate degenerative changes of the kidney, including distended tubules and protein casts, were detected. Measuring relative organ weights, a slightly enlarged liver weight was observed. However, no histopathological abnormalities or steatosis was detected in the liver ( Figure S15). The analysis of serum and blood samples after the application of the highest dose of 800 nm Y9 revealed slightly but significant increased lipase and creatinine (CREA) levels and ). An isotype control with mlgG1-PE was used to determine background signal.
significantly increased amylase, triglycerides, cholesterol and blood urea nitrogen (BUN) levels in serum ( Figure S15). Hematological analysis including a differential hemogram of the white blood cells showed no significant changes. Overall, these results do not indicate any significant systemic toxic effect at these effective dosages.

Discussion
Conventional antibodies, fragments thereof and other protein-based binders address a molecular weight range above 25 kDa, in contrast to small molecule drugs that are in the range of 300 to 1000 Da, which leaves the intermediate mass range open for new concepts such as the ISEr here. The trivalent ISEr Y9 provides a purely synthetic alternative to conventional antibodies and related biological techniques. It offers a versatile synthesis route to medium sized drug molecules (~5 kDa) that combine bivalent, specific target binding with activation of the innate immune system. Target specificity is achieved by attaching two binders (B9) to a generic scaffold carrying fMIFL (F2) as an effector for activating formyl peptide receptors (FPRs). This effector induces chemotaxis and activation of human and guinea pig leukocytes in vitro and in vivo without undesired side effects. Formyl-peptides can also induce the expression of IL-1β, TNFα and IL-8 48 , but we did not observe a significant induction of the latter two cytokines by F2. Our data thus indicates that treatment of tumors with F2-2PEG does not contribute to the expression of these potentially tumor-promoting cytokines 49 . Furthermore, the flexible synthesis route allows for the adaption of effectors to specific requirements with respect to immune stimulation and species.
The design of Y9 allows access to more than one binding site on a cellular target such as in dimeric or clustered receptors, thereby improving binding to the targeted cells beyond simply increasing the local concentration of a receptor ligand towards more complex avidity effects. Such effects will help to accumulate Y9 at a concentration in tumor tissue sufficient to activate macrophages. Bispecific or multispecific molecules targeting two or more different cell-surface receptors can also be easily envisioned based on this design and synthesis scheme. Such molecules will combine the ease of chemical synthesis with well-known advantages of bi-or multispecific binders, as demonstrated for approved bispecific antibodies such as Catumaxomab 50 . The PEG chains reduced the affinity of Y9 for integrin α 3 -expressing A431 cells. Nevertheless, efficacy could still be demonstrated in vivo, possibly due to the role of PEG in protecting against biodegradation, improving solubility, and promoting the formation of monodisperse molecules in aqueous solution 25 .
Y9 reduced the establishment of a GPC-16 based tumor in guinea pigs by 67% upon a single dose treatment during injection of tumor cells. As no direct toxic effects of Y9 on cancer cells were observed in vitro, we conclude that activation of immune cells interfered with tumor establishment. Clinical use of such a treatment can be envisioned in preventing tumor relapse or development of metastases after first line treatments such as surgery, chemotherapy and/or radiation. As the simultaneous injection of tumor cells and Y9 mimics a local administration of the drug, as for instance anticipated during chemoembolization or peri-surgical application, other therapeutic applications are possible as well. Considering that Y9 was still effective in cyclophosphamide treated guinea pigs, it might provide an effective treatment in combination therapies. No severe toxic effects other than the expected moderate local reactions to Y9 were observed in guinea pigs and mice. Only repetitive doses four-fold higher than used in the effective treatment led to mild pathological changes in kidneys and some serum and blood parameters. The moderate degenerative changes of the kidney including distended tubules and protein casts in the highest dosage group indicate renal clearance of Y9. In general, Y9 can be safely used within a dose range of 100 to 400 nmol (0.5-2 mg). Overall, Y9 is an alternative to conventional antibody-based therapeutics and related protein-based approaches due to its entirely flexible and robust chemical synthesis. ISErs can be generated quickly, providing access to customized drugs targeting specific receptors not only relevant for cancer therapy but also for other disease areas requiring modulation of innate immune responses at distinct target sites such as inflammatory and autoimmune diseases.

Online Methods
Synthesis of effector and binder peptides. Peptides were obtained via standard Fmoc SPPS. In brief, the first amino acid was DIC-activated and subsequently coupled to Wang resin (100-200 mesh, 0.9 mmol/g substitution level) in the presence of DMAP as a catalyst. All following couplings were performed using HBTU as activator. To check for complete coupling, a Ninhydrin test was conducted. Fmoc was removed using 20% piperidine in DMF. Formylation was achieved using 3 eq p-nitrophenylformate in DMF for 3 h. The peptide was cleaved using TFA/TIS/H 2 O/DMS 92.5:2.5:2.5:2.5 (v/v) for 2 h, following peptide precipitation with cold diethyl ether and re-solubilization in ACN/H 2 O 1:1 (v/v) with 0.1% TFA. After freeze drying, peptides were dissolved in 6 M GndHCl, pH 4.7 and purified via RP-HPLC.

Synthesis of Y9.
Removal of the Mtt protecting group was achieved by flow washing the peptidyl-resin with DCM/TFA/TIS 98:1:1 (v/v) until the solution turned colorless. Each lysine side chain was PEGylated overnight using 1.12 equivalents of Fmoc-NH-(PEG) 27 -COOH and HATU as activator. Subsequently, the synthesis was continued using standard SPPS methods as described above.
Surface plasmon resonance (SPR). Single cycle kinetic experiments were conducted on a Biacore 3000 system using a CM5 sensor chip at 25 and 37 °C with a flow rate of 10 µl/min. After coupling of the VLA-3 receptor (R&D Systems) via standard EDC/NHS immobilization chemistry (7 min activation with EDC/NHS followed by a 7 min flush with VLA-3) in 10 mM NaOAc buffer at pH 6 containing 1 mM MnCl 2 , a 1 mM solution of ethanolamine was injected for 7 min to block all amine reactive sites. Subsequently the analyte was flushed over the chip surface at a flow rate of 10 µl/min in HEPES buffered saline supplemented with 1 mM MnCl 2 . As no method was found to regenerate the surface from Y9 without damaging the VLA-3, Y9 was injected in increasing concentrations, taking care not to reach saturated binding (single cycle kinetics). Human cells. Primary cells were obtained from healthy volunteers that served as donors at our blood bank, after written informed consent and approval by the local ethics committee of the Uniklinik RWTH Aachen, Germany.

Animals. All experiments
For more information on experimental details about NMR measurements, binding and stability assays as well as on animal testing, please see Supplementary information.