Synthesis and characterization of a high-affinity α_vβ₆-specific ligand for in vitro and in vivo applications

Peptides that recognize cancer cells with high affinity and specificity have promise as reagents for tumor-specific delivery of chemotherapeutics and molecular imaging agents. Phage display has proven a powerful method for the discovery of peptides that bind to cancer cells but not their nonmalignant counterparts. Purified tumor-associated biomarkers, cells in culture, and tumors in animals have been used as bait to identify cancer-binding peptides. However, when these isolated peptides are removed from the phage scaffold and used as free monomeric peptides, their binding affinities are often too weak to be clinically useful as delivery or diagnostic reagents.

We previously identified several 20-amino-acid peptides from a phage-displayed peptide library that bind to different human non–small cell lung cancer cell lines (1). One peptide, named H2009.1, binds to the integrin α_vβ₆ and does not bind other more widely expressed RGD-binding integrins (2). This integrin is overexpressed in many epithelial-derived carcinomas (3–10) but is not found in normal primate tissues (11). The expression of this integrin is correlated with cellular behaviors that are typical of more aggressive tumors (12, 13). Furthermore, α_vβ₆ expression increases during the epithelial-mesenchymal transition, the process in which cells lose their epithelial phenotypes to become motile, implicating a role in metastasis (10)_. Integrin α_vβ₆ is a biomarker for poor prognosis in several human cancers, including lung, colorectal, and cervical cancers (2, 9, 10). We have shown that the integrin is widely expressed in early-stage non–small cell lung cancer and its expression increases in a stepwise manner with grade (2). In sum, this integrin is emerging as an important target for anticancer therapies. As such, further development of this α_vβ₆-targeting ligand is warranted.

Branched lysine cores provide scaffolds for the synthesis of multimeric peptide dendrimers, often referred to as multiple antigen peptides (14, 15). Tetramerization of the H2009.1 peptide on this trilysine dendrimeric core results in higher affinity for its cellular target than the corresponding monomer. Previous syntheses of the tetrameric H2009.1 peptide were achieved by stepwise solid-phase Fmoc synthesis on a trilysine core (1, 16). Due to the macromolecular structure and complexity, unambiguous synthesis of tetrameric peptides on a trilysine core is difficult. Separation of the desired molecule from the mélange of closely related truncated and deletion products that arise during the synthesis is challenging (15). In addition, the purified products, despite the appearance of being homogeneous by high-performance liquid chromatography (HPLC) analysis, are often heterogeneous, being contaminated with numerous coeluting sequences. Segment condensation has been used to generate tetrameric peptides but current methods have limitations in their use for tumor targeting (15). For these reasons, the utility of these branched peptides is limited in the clinical setting.

To facilitate the use of this α_vβ₆-binding peptide, we report here a facile convergent strategy for multimeric peptide synthesis based on the chemoselective reaction of a thiol to an activated double bond of a maleimido group. We have optimized the H2009.1 peptide on the trilysine scaffold, reducing its size by half while improving the affinity 10-fold. The affinity of the optimized H2009.1 peptide for its target cell rivals the affinity of more commonly used antibody-targeting reagents.

Monomeric peptide synthesis was performed on a Symphony Synthesizer (Rainin Instruments, Protein Technologies, Inc.) by Fmoc solid-phase peptide synthesis on a Rink Amide AM resin (substitution level 0.71 mmol/g). Details are available in the Supplementary Data.³

The maleimido tetrameric cores were synthesized on Fmoc₄-Lys₂-Lys-β-Ala-CLEAR Acid Resin, Fmoc₄-Lys₂-Lys-Lys(Biotin-PEG)-β-Ala-CLEAR Acid Resin, and Fmoc₄-Lys₂-Lys-Cys(Acm)-β-Ala-CLEAR Acid Resin, respectively (substitution level 0.21 mmol/g, Peptides International). The maleimido trimeric core was synthesized on Fmoc-β-Ala-CLEAR Acid Resin (substitution level 0.52 mmol/g). Fmoc-Lys(Mtt)-OH was coupled at a 5-fold excess using HBTU, HOBt, and NMM coupling (45 min). Piperidine in DMF (20%) was used to remove NH₂-terminal Fmoc protecting groups. Fmoc-Lys(Fmoc)-OH was coupled in the same manner. The 4-methytrityl (Mtt) protecting group was removed by treatment with 1% trifluoroacetic acid in dichloromethane for 2 min, five times (17). Maleimidopropionic acid was coupled to all lysine cores in 5-fold excess using HBTU, HOBt, and NMM (24 h, room temperature). The activated core was cleaved from the resin using a trifluoroacetic acid/triisopropylsilane/H₂O cocktail, precipitated in cold diethyl ether, and purified by reverse-phase HPLC. Characterization is reported in the Supplementary Data.³

The peptide (8 μmol) and appropriate maleimide-activated core (1 μmol) were dissolved in 1.5 mL of Ar-purged PBS (pH 7.4) containing 0.01 mol/L EDTA. The reaction mixture was stirred at room temperature for 2 h. Details are included in the Supplementary Data.³

The ability of the peptides to block their cognate phage was determined as described (1, 18).

The acetamidomethyl group on the cysteine was removed according to published procedures (19). A solution of 0.015 μmol H2009.1 tetrameric-SH peptide was prepared in 500 μL of Ar-purged 1× PBS/0.01 mol/L EDTA. A solution of 0.15 μmol Alexa Fluor 488 C₅-maleimide dye (C₃₀H₂₅N₄NaO₁₂S₂, Invitrogen Corporation) in 50 μL PBS was added. The mixture was stirred at room temperature for 12 h before adding 0.75 μmol mercaptoethanol to consume excess thiol-reactive dye. The products were purified by HPLC and analyzed by matrix-assisted laser desorption/ionization mass spectrometry (Supplementary Data).³

The biotin-labeled tetrameric peptide (4,000 nmol/L) was incubated with Streptavidin-Qdot 605 made in-house (SAQdot605, 200 nmol/L) for 2 h at room temperature. The peptide-SAQdot605 conjugate solution were diluted with PBS⁺/0.1% bovine serum albumin to a final Qdot concentration of 20 nmol/L before overlaying onto cells as described (18). Similarly, H2009 cells were incubated for 10 min at 37°C with 1 μmol/L peptide-Alexa Fluor 488 conjugate. β₆ was observed using anti-β₆ antibodies (2). Images were captured on a Leica TCS SP5 confocal microscope at 400× magnification with 7× optical zoom.

Cells were grown to 80% to 90% confluency in a 12-well plate (3.87 cm²). Cells were treated with serum-free RPMI (2 h, 37°C) to clear the cellular receptors. Peptide-dye solutions in PBS⁺ supplemented with chloroquine and protease inhibitor cocktail were placed on the cells (1 mL/well, 1 h at 37°C). Cells were washed four times with PBS⁺, released from the wells by incubation in Enzyme-Free Cell Dissociation Buffer (10 min, room temperature), and analyzed on a Quanta Flow Cytometer (Beckman Coulter).

One million cells were inoculated s.c. on the flanks of immunodeficient mice. Tumors were grown for 19 d. Mice were perfused with PBS followed by the formalin fixative. Tumors were processed for paraffin block sectioning. One slide from each tumor was stained with H&E whereas serial sections were stained with peptide-labeled Qdots(605 nm) or unlabeled Qdots. For Qdot staining, slides were treated with EZ-DeWax (BioGenex) per manufacturer's protocol. Slides were incubated in 300 mL of 10 mmol/L sodium citrate-citric acid buffer (pH 6.0), 90°C, for 10 min, then allowed to cool to room temperature. Slides were rinsed with TBS [100 mmol/L Tris-HCl (pH 7.4), 138 mmol/L NaCl, 17 mmol/L KCl] and blocked in TBS supplemented with 1% goat serum and 1% bovine serum albumin. H2009.1-Qdots were prepared as detailed above. Tumor sections were incubated with peptide-labeled or control Qdots suspended in PBS + 0.1% bovine serum albumin for 30 min at room temperature. Slides were washed four times in TBS + 0.1% bovine serum albumin. Vectashield containing Hoechst dye was applied to the slide before imaging.

The maleimide precursor (4,10-Bis-carboxymethyl-7-{[2-(2,5-dioxo-2,5-dihydro-pyrrol-1-yl)-ethylcarbamoyl]-methyl}-1,4,7,10tetraaza-cyclododec-1-yl)-acetic acid; Macrocyclics) was reacted with the thiol of the cysteine (molar ratio: 5:1, 2 h, room temperature) in 10 mmol/L PBS containing 10 mmol/L EDTA. The 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA)–conjugated peptides were separated by Waters XTera C₁₈ HPLC system (Waters) and confirmed through mass spectrometry. ¹¹¹In was loaded by a standard protocol. Details are included in the Supplementary Data.³

To 100 μL of rat serum, 10 μL of ¹¹¹In-labeled peptide solutions were added. The solution was incubated at 37°C. After 1 and 4 h postaddition, an aliquot of 30 μL was mixed with 100 μL of ethanol to precipitate serum proteins. The mixture was centrifuged at 14,000 rpm for 10 min. The pellets were resuspended with 80% ethanol, followed by centrifugation. The pellets and the pooled supernatants were counted by a γ-counter. The ¹¹¹In-labeled peptides in the supernatant were identified by radio-HPLC. The serum stability of ¹¹¹In-labeled peptides is presented as the product of the HPLC-determined radiochemical purity of the labeled peptide in the supernatant and the percentage of the pooled supernatant γ counts of the total counts.

Tumors were established by s.c. injection of a cell suspension (1 × 10⁶ cell) as described and grown for 16 d, resulting in tumors of 60 to 200 mg. Each mouse received 5 μCi of the ¹¹¹In-labeled peptides i.v. (n = 4). Mice were sacrificed 4 h postinjection. The percentage of injected dose per gram was determined for the tumor, blood, and muscle. The animal protocols were approved by the University of Texas Southwestern Medical Center's Institutional Animal Care and Use Committee.

As an alternative to the reported synthesis of tetrameric peptides on trilysine cores, we used a synthetic route that takes advantage of the chemoselective reaction of a cysteine with a maleimide (Fig. 1). Although this approach has been suggested (15), its use has not been reported for the synthesis of these dendrimers. The activated N-maleimido trilysine core was synthesized by the reaction of 3-maleimidopropionic acid with Lys₂-Lys-β-Ala on a CLEAR Acid Resin solid support. The fully deprotected H2009.1 peptide containing a unique cysteine (NH₂-RGDLATLRQLAQEDGVVGVR-PEG₁₁C-CONH₂) was synthesized in 63% yield. A polyethylene glycol (PEG) linker was incorporated to increase solubility of the peptides in aqueous solution and prevent possible compaction of the peptides on the trilysine core. The reaction of the cysteinyl peptide with the activated trilysine core on the solid support was unsuccessful under a variety of conditions, including a 2.5-fold molar excess of peptide, reaction times up to 48 hours, and use of organic cosolvents. However, the removal of the N-alkyl maleimide-activated trilysine from the solid support followed by a solution phase coupling of the peptide in aqueous buffer for 2 hours at room temperature resulted in production of four-branch peptide in 85% yield. The tetrameric peptide was purified from the small amounts of contaminating one-, two-, and three-branched peptides by reverse-phase HPLC. The resultant tetrameric peptide was >95% pure based on analytic HPLC analysis.

There are structural differences in the linker of the peptides obtained by linear synthesis and the convergent synthesis. Although distant from the cell binding peptide, this may affect cell binding. As this peptide was initially selected as a bacteriophage clone through biopanning on the human lung adenocarcinoma cell line H2009 (1), the ability of this peptide to block binding of its cognate phage on H2009 cells was assessed. This assay has the advantage in that it can be used in cases in which purified membrane proteins are unavailable or when the cellular target of the peptide ligand is unknown. Additionally, it screens for binding in a more relevant cellular context and can be used when the peptides mediate cellular uptake in addition to cell surface binding. By this assay, the peptides perform identically over a 10,000-fold concentration range and display half-maximal binding in the 0.6 to 0.8 nmol/L range (Fig. 2A).

The ease of synthesis allows for optimization of the peptides by chemical approaches as opposed to repanning mutagenic phage-displayed peptide libraries. This has the benefit in that optimization occurs on the format in which the peptides will later be used. The H2009.1 20-mer peptide was selected from a random library of 20-amino-acid peptides. Alanine scanning was done to address which residues are critical for binding of the peptide to its receptor. A set of four monomeric peptides was synthesized in which amino acids 1 to 5, 6 to 10, 11 to 15, or 16 to 20 were replaced by alanines. Changing amino acids 1 to 5 or 6 to 10 to alanine inactivates the ability of the peptide to block phage uptake (Supplementary Data).³ The last 10 amino acids can be changed to alanine with no loss of binding affinity of the peptide for H2009 cells. Consistent with this observation, the NH₂ terminus contains a known binding motif for α_vβ₆ (2, 20). As confirmation, the H2009.1 10-mer peptide possessing the NH₂-terminal 10 residues of parent peptide (sequence: RGDLATLRQL-CONH₂) was synthesized. As expected, the 10-mer peptide behaves the same as the parental 20-mer peptide (Fig. 2B). Of note, the carboxy-terminal amide is important for binding as the corresponding peptide, RGDLATLRQL-COOH, shows significantly reduced affinity (data not shown). Amino acid deletions were made from the carboxylterminus of the 10-mer peptide. The peptide retains full activity as a 7-mer. Truncations past this point result in loss of peptide binding.

We anticipated that tetramerization of the shorter H2009.1 peptides would increase its affinity for H2009 cells in the same manner as observed for the 20-mer peptide. The H2009.1 10-mer and 7-mer tetrameric peptides were synthesized in 91% and 89% yield, respectively. The H2009.1 10-mer and 7-mer tetrameric peptides inhibit phage uptake by >97% at concentrations above 100 nmol/L on H2009 cell (Fig. 2C). Similar results are observed for H2009.1 20-mer tetrameric peptide at the same concentration. Surprisingly, at 1 nmol/L, the H2009.1 10-mer and 7-mer tetrameric peptides inhibit phage uptake 10.8- and 1.4-fold more than H2009.1 20-mer tetrameric peptide, respectively; the monomeric peptides have uniform potency at this concentration. Impressively, the half-maximal binding is in the 40 to 60 pmol/L for the 10-mer tetrameric peptide and 80 to 100 pmol/L for the 7-mer. These results indicate that the carboxy-terminal 10-amino-acid residues contribute negatively to the binding affinity in the context of the four-branched peptides. The reason for this is unknown. It is possible that the orientation of the peptide branches on the 10-mer is better suited for receptor binding on the cell surface. Alternatively, the last 10 amino acids do not contribute to the enthalpy of binding but their entropy may decrease in the process of binding to the cellular receptor, resulting in a decreased binding affinity (21). Nonetheless, our results show that the tetrameric 10-mer has enhanced cellular binding. Furthermore, the peptidic targeting reagent is surpassing the affinities observed with more commonly used antibody-targeting reagents (22).

In attempt to reduce the peptide size, the minimal required valency was explored. The H2009.1 10-mer dimeric and trimeric peptides were synthesized by the convergent strategy. The dimeric lysine core was made on Rink Amide AM resin using Fmoc solid phase peptide synthesis strategy with a yield of 93%. The maleimido trimeric core was synthesized using a chemically orthogonal–protected lysine (Mtt-protected C-amino group of the first lysine). Coupling of the peptides to these scaffolds was done in the same manner as that used for the tetrameric cores.

The dimeric and trimeric peptides display higher phage blocking ability than the corresponding monomeric peptide, corroborating the superiority of peptide multimerization in increasing the affinity of the peptide for its cellular target (Fig. 2D). However, the tetramer construct obstructs phage uptake at lower concentrations when compared with the dimeric and trimer peptide, suggesting that the optimal scaffold is the tetramer for this group of peptides. Higher valencies were not tested.

The sulfhydryl group of a cysteine is frequently used for attachment of peptides to drugs, drug carriers, or imaging agents (16, 23). Thus, the synthetic strategy used here would have greater utility if it is compatible with the protection/deprotection of other chemically orthogonal thiol-protecting groups. To determine if the four-branch peptide could be synthesized and subsequently deprotected to reveal a free thiol, the H2009.1 10-mer tetrameric peptide was coupled to a maleimide trilysine core containing an acetamidomethyl-protected cysteine before the branch point. The acetamidomethyl group was removed by treatment with Ag(I) (98% yield; ref. 19). No loss of the peptide branches was observed (Supplementary Data).³ The peptide can be labeled by reaction of Alexa Fluor 488 C₅-maleimide dye, confirming the integrity of the revealed thiol moiety. Thus, the thiol ether bond formed between the peptide and the lysine core is chemically compatible with removal of the acetamidomethyl group from a strategically placed cysteine.

Another useful chemical moiety is biotin as many streptavidin reagents exist for biological applications. Therefore, we synthesized an activated trilysine core that contained a biotinylated lysine before the branch point using a Fmoc₄-Lys₂-Lys-Lys(Biotin-PEG)-β-Ala-CLEAR Acid Resin. The biotin was fully compatible with the downstream chemistry and was functional in binding streptavidin.

For the smaller peptide to be of utility as a drug delivery reagent, it must mediate cellular uptake specifically in α_vβ₆-positive cells. To address this, the Alexa Fluor 488–labeled H2009.1 10-mer tetrameric peptide was incubated with H2009 cells. The cellular localization of the peptide was determined by confocal microscopy. The peptide-dye conjugate is internalized into the cells (Fig. 3A; Supplementary Data).³ As confirmation, the biotinylated peptide was conjugated to a streptavidin-coated fluorescent quantum dot (SAQdot605) and incubated with H2009 cells. Punctate staining is observed in the cytoplasm by fluorescent microscopy, indicative of cellular uptake of the nanoparticle through an endocytosis process (Fig. 3A). A Z-stacked series obtained by confocal microscopy confirms the internalization of the particle (Supplementary Data).³ Additionally, the peptide mediates internalization of the β₆ integrin. The peptide Qdot complex colocalizes with α_vβ₆ within the H2009 cells, indicative that the H2009.1 10-mer tetrameric peptide is mediating uptake through α_vβ₆-mediated endocytosis (Fig. 3B). No uptake is observed for unconjugated Qdots or when a scrambled H2009.1 peptide (DALRLQGTLR) is conjugated to Qdots. Additionally, no uptake is seen when another peptide with no affinity for α_vβ₆ is used (H1299.2, sequence: YAAWPASGAWTGTAPCSAGT) (Fig. 3; Supplementary Data).³ When H2009 cells are treated with the H1299.2-Qdot conjugate, β₆ remains localized on the cell surface, demonstrating that the internalization of β₆ is triggered by binding of the H2009.1 10-mer tetrameric peptide. The peptide retains its cell specificity as well; no uptake of the H2009.1 10-mer tetrameric peptide is observed in a cell line with insignificant expression of α_vβ₆ (Supplementary Data).³

Peptide-Qdot uptake is observed in cells that are in varying stages of the cell cycle. As seen in Fig. 3C, the peptide-Qdot conjugate is observed in cells in prophase, metaphase, early anaphase, and late telophase/cytokinesis. As the incubation time is only 10 minutes, this suggests that cellular uptake of the peptide by α_vβ₆ can occur at any point along the cell cycle. Thus, cells do not need to be synchronized to use this reagent in vitro, and in vivo targeting will be cell cycle independent. Furthermore, it appears that the peptide-Qdots are being retained during cell division. This is consistent with other reports that show the utility of Qdots to track cells through multiple cell divisions (24).

Antibodies are routinely used to obtain molecular information about tumors. Clinical samples are most commonly presented as formalin-fixed paraffin samples, not fresh tumors. The utility of the H2009.1 peptide for clinical diagnosis would be expanded if it is bound to fixed tumor samples. To address this, H2009 tumors grown s.c. in a severe combined immunodeficiency mouse were fixed in paraformaldehyde and embedded in paraffin. As seen in Fig. 4, the peptide-Qdot conjugate can bind a fixed non–small cell lung cancer α_vβ₆-positive tumor sample. Less binding is observed when Qdots alone are used. Thus, the peptide can serve as a complementary reagent to α_vβ₆ antibodies. As peptide-Qdot conjugates can be multiplexed by using different peptide-Qdot pairs, this opens the possibility of assaying for multiple biomarkers on a single sample.

Additionally, flow cytometry is used to determine the expression levels of biomarkers as an aid for diagnosis (25). Although flow cytometry has been primarily used for nonadherent cells, there is increasing interest in using it for the detection of circulating cancer cells that have been shed from the solid tumor. As α_vβ₆ is overexpressed on many epithelial-derived cancers, this peptide may be of utility to detect α_vβ₆-positive cells in the bloodstream. The Alexa Fluor 488–labeled H2009.1 10-mer tetrameric peptide was incubated with H2009 cells. As shown in Fig. 4B, the peptide is functional as a flow cytometry reagent to recognize α_vβ₆-expressing cells, such as H2009. Cells treated with labeled H2009.1 10-mer tetrameric peptide have mean fluorescence intensity that is 2-fold greater than when the scrambled peptide is used. H460 cells that do not express significant amounts of α_vβ₆ (2) exhibit minimal binding to the H2009.1 10-mer tetrameric peptide (Fig. 4B). The results suggest that this peptide may have utility as a detection or capture reagent of α_vβ₆-expressing cells.

To be of utility in vivo, the peptide must remain intact in serum long enough to home to its cellular target. Historically, serum stability has often been cited as a drawback of peptide therapeutics, although PEGylation and glycosylation can increase the peptide half-life in vivo (26). Peptides synthesized on trilysine cores are more serum stable than their monomeric counterparts (27, 28). However, we have introduced a thiol ether bond into the peptide that might have different serum stability than the native amide.

To evaluate the serum stability of peptide conjugates in vitro, we radiolabeled the monomeric H2009.1 10-mer, H2009.1 10-mer tetrameric peptide synthesized by convergent strategy, and the corresponding tetramer synthesized by linear Fmoc SPPS. The peptides were modified by placing a DOTA chelator towards the COOH terminus of the peptide and subsequently loaded with ¹¹¹In (Fig. 5). The¹¹¹In-labeled peptides were incubated with rat serum at 37°C. After 1 and 4 h, an aliquot was removed and the serum proteins were precipitated in ethanol. The supernatants were counted by a γ-counter and the amount of full-length ¹¹¹In-labeled peptides were identified by radio-HPLC. The results are shown in Table 1. As expected from previous reports, the tetrameric peptides are more stable than the monomeric peptide (27, 28). At 4 hours, a significant fraction of the peptide remains intact for all three peptides. At this time point, similar amounts of tetrameric peptides remain intact regardless of the synthetic route, indicating the thioether bond to be suitable for in vivo applications. This suggests that the peptides will have the serum stability necessary for in vivo applications.

This method measures total radiochemical stability, such that both proteolysis and loss of the radio metal from the DOTA moiety will contribute to the overall loss of the HPLC signal. However, this method is more sensitive and quantitative than standard HPLC methods. Additionally, it measures the total stability of a peptide conjugate that can be used for single-photon emission computed tomography or positron emission tomography imaging. The resultant peptides after treatment in sera were not characterized. Additionally, these studies do not address the clearance rate of the peptides in an animal.

To determine if the peptide could reach its target in vivo and might possibly serve as a molecular imaging reagent, the ¹¹¹In-labeled peptides were injected i.v. into tumor-bearing severe combined immunodeficiency mice. One and 4 hours after injection, the animals were sacrificed and the tumors were collected, weighed, and counted by a gamma counter. The percentage of injected dose per gram was determined for the tumor, blood, and muscle. At 1 hour, the H2009 tumor retains 3.15 ± 0.47 of the total dose per gram (%ID/g) of tissue and the tumor to blood ratio is 1.37 ± 0.14. By 4 hours, the tumor accumulates 2.25 ± 0.47% D/g of tissue and the tumor to blood ratio is 2.43 ± 0.14. By comparison, the corresponding scrambled sequence peptide had only 0.59 ± 0.30% of the total dose per gram of tissue retained in the tumor at 4 hours (tumor to blood ratio of 0.44 ± 0.22). Substantially less accumulation of peptide in the tumor is observed when a non–α_vβ₆-expressing tumor is implanted in the animal (α_vβ₆-negative H460 lung cancer cells, at 1 hour: tumor accumulation of 0.96 ± 0.16%ID/g, tumor to blood ratio of 0.81 ± 0.14; at 4 hours: tumor accumulation of 0.84 ± 0.11%ID/g, tumor to blood ratio of 2.0). The tumor to muscle ratio for the H2009 tumors is 3.92 ± 0.35%ID/g and 4.43 ± 0.03%ID/g at 1 and 4 hours, respectively. By comparison, the tumor to muscle ratio for the H460 tumors is 0.80 ± 0.15%ID/g (1 hour) and 1.54 ± 0.19%ID/g (4 hour). In sum, these data show the H2009.1 peptides have the stability, specificity, and affinity to target α_vβ₆-positive tumors in animal models. The tumor to blood ratio is favorable for the development of the peptide as a single-photon emission computed tomography and positron emission tomography imaging agent. A full pharmacokinetic profile must be done to fully understand the fate of the peptide in other organs and the α_vβ₆ expression profile in a mouse (11, 29).

The integrin α_vβ₆ is increasingly being considered as a candidate for targeted therapies for epithelial-derived cancers. As such, high-affinity and high-specificity ligands for this integrin target are needed. Toward this goal, we have identified a peptide that binds specifically to α_vβ₆ within the context of the cell membrane. Our advancements to the overall synthetic route and optimal presentation expand the possible clinical uses of this targeting ligand. Furthermore, our peptide is functional in multiple applications.

The ease of the synthetic route allowed us to determine the optimal length and valency of a specific peptide, H2009.1. Through systematic mutations and truncations, the peptide was reduced from 20 amino acids to 7. Consistent with our results, the α_vβ₆ is an RGD binding integrin and native ligands for α_vβ₆ typically have an RGDLXXL/I motif (20, 30). The decrease in the size of the peptide speeds the synthesis and reduces the cost of the targeting moiety. It also has the added benefit of reducing the potential of untoward proteolysis and immune responses to the peptide. Experiments to determine the optimal amino acids in positions 5 and 6 of the peptide are under way in our laboratory. Although we could reduce the size of the peptide chain, the tetrameric peptide has a higher affinity than the corresponding dimer or trimer. However, for in vivo delivery, there is a fine balance between affinity, stability, nonspecific uptake, and clearance. As such, the lower valency peptides may be superior for some applications.

The emergence of α_vβ₆ as an important therapeutic target for epithelial-derived cancers has led to the design of ligands for this cell surface receptor. Small-molecule inhibitors (31) and function blocking antibodies are being developed against this target (32, 33). Although promising, peptides combine the high specificity and affinity of antibodies with the small size and synthetic availability of small molecules. DiCara and colleagues (34) have recently characterized three 20-mer peptides for α_vβ₆ binding. The peptides are derived from two of the α_vβ₆ naturally occurring ligands, foot and mouth disease virus, and the latency-associated peptide of transforming growth factor-β (35, 36). The peptides A20FMDV1 (sequence: YTASARGDLAHLTTTHARHL), A20FMDV2 (sequence: NAVPNLRGDLQCLAQKVART), and A20LAP (GFTTGRRGDLATIHGMNPF) contain the RGDLXXL/I motif and bind α_vβ₆ with affinities in the low nanomolar range. Interestingly, the affinity of the peptides is correlated with the α helicity of the peptides. The conserved leucine or isoleucine at positions Asp+1 and Asp+4 may make important contacts with the integrin at the binding interface, and placing these residues on the same face of the helix increases the affinity. The A20FMDV2 peptide has recently been used as a positron emission imaging reagent for α_vβ₆-positive tumors in a mouse model (37). In contrast to our work, shorter peptides (7–12 amino acids) derived from these sequences had reduced affinity compared with the longer 20-mer peptides.

Of the H2009.1 monomeric peptides, the predicted α-helical propensity of the peptides is 20-mer >10-mer >>7-mer using Agadir secondary structure prediction software (Supplementary Data).³ Yet, we observe no difference in the binding affinity of these three peptides. Perhaps the H2009.1 peptides may be unstructured as in the case of the A20FMDV1 peptide. Alternatively, the flanking sequences to the RGDLXXL/I motif may play a role in binding that is independent of the secondary structure. Further structural analysis is needed to better understand the structure function relationship of these peptides.

Peptides are particularly appealing ligands for cell targeting. They are easier to chemically manipulate and less costly to produce than currently used antibody reagents. Small peptides can avoid uptake of the reticuloendothelial system, which can be problematic with antibodies. Importantly, peptide libraries can be screened for cell binding. Both one-bead/one-compound and phage-displayed peptide libraries have been used a source for cell binding ligands, and both are amenable to using intact cells (38, 39). This opens up the route to identifying cell-specific peptide ligands without a priori knowledge of the cellular receptor. Due to the need for tumor-specific reagents for chemotherapeutic targeting, cancer cell lines have been used as the bait in many of these selections (40).

Retaining the affinity and activity of peptides selected from phage-displayed libraries has been a challenge in the field. To overcome this, many have used a cyclic peptide library whose conformation is constrained by a disulfide bond. Although useful, peptides with disulfides may present challenges in downstream in vivo applications or in the attachment of these peptides to drug carriers. The trilysine scaffold mimics the presentation of the peptides on the pIII protein of the phage in both the valency and the orientation of the displayed peptides. We have found the tetrameric trilysine framework to be a general platform for cell-targeting peptides selected from bacteriophage-displayed peptide libraries (18, 41–43). Peptides on this scaffold retain their cell specificity as well as their ability to mediate cellular internalization when removed from the phage particle.

Bracci et al. (44, 45) first reported the utility of trilysine scaffolds for placing targeting peptides. Surprisingly, multimerization did not increase the affinity of the peptide for its target but increased the serum stability, leading to an increase in efficacy in vivo. Recently, they have placed a six-amino-acid neurotensin fragment on a tetrameric trilysine dendrimer; the peptide can deliver an active chemotherapeutic to tumors overexpressing the neurotensin receptor (28). Of note, the peptides used in these experiments were not selected from a phage-displayed library. Additionally, the peptides were optimized on purified protein targets not whole cells. This could account for similar affinities of the monomeric and tetrameric peptides.

Despite the utility of the trilysine scaffold, the synthesis of these multimeric peptides is difficult, especially for longer peptides. The possibility for incomplete coupling for every additional residue along any of the branches results in complicated mixture of compounds. To overcome the problems involved in linear synthesis, segment condensation has been used. However, the coupling rates in segment condensation are slow and the solubility of protected peptide segments is unpredictable (46). Other methods based on the site-specific ligation of unprotected peptide segments through formation of thioether (47), oxime (48), hydrazone (49), or thiazolidine (50) linkages have been developed. These approaches have met with success; however, the oxime- and hydrazone-derived dendrimers may not be appropriate for tumor targeting due to their potential instability in the acidic tumor environment. Additionally, formation of thiazolidine linkage occurs through an NH₂-terminal cysteine. This orientation is opposite of how the peptides are displayed as a fusion on the phage particle and may not be appropriate for the synthesis of multimeric peptides that require a free amino terminus for activity. Last, the formation of thioethers through reaction of a sulfhydryl with a haloacetyl often requires basic conditions, resulting in the formation of peptide dimer side products as a result of disulfide formation. By comparison, our method allows for tetrameric peptides to be produced in 85% to 95% yield and >95% purity. Of note, the overall synthetic yield of the H2009.1 10-mer tetrameric peptide is 68% after purification, taking into account the yield of the monomeric peptide synthesis and the conjugation yield. By comparison, the same peptide synthesized by the Fmoc linear method has an overall yield of 8% after purification.

This chemistry is not restricted to laboratories with expertise in synthetic peptide chemistry. As such, this approach is useful to all performing phage display selections. We have recently selected a peptide from the commercially available NEB PhD 12-mer phage-displayed peptide library that binds to a squamous cell lung carcinoma. Tetramerization of this peptide by this synthetic route increases its affinity for its target cell type by 200-fold and has a half-maximal binding affinity of 30 pmol/L.⁴

This simple synthesis of tetrameric peptides widely expands the use of ligands isolated from phage-displayed peptide libraries. This method has produced a peptide that binds specifically to the integrin α_vβ₆ with an affinity that rivals that of antibodies. The increased affinity and stability of the tetrameric peptide opens the door for this peptide to be used as an in vivo reagent for targeted therapies of the treatment of aggressive α_vβ₆-positive tumors and molecular imaging.

No potential conflicts of interest were disclosed.

We thank Shaghayegh Rezaie, Sapna George, Jonathan Lee, and Naveen Pattisapu for technical assistance, and Bethany P. Gray for critical reading of the manuscript. This article represents publication no. CSCN036 from the Cell Stress and Cancer Nanomedicine Program of the Simmons Comprehensive Cancer Center.