Among the several available approaches, the biological library (such as phage display library) and OBOC combinatorial library are the two most commonly used methods to identify cancer-targeting peptide ligands [
26]. Yet unlike the biological combinatorial library approach, which is generally limited to
l-amino acids, ligands identified from OBOC libraries can contain unnatural amino acids (e.g.,
d- and β-amino acids), organic moieties, and/or constrained conformation which improve the stability of the ligands against proteolysis. Therefore, it is a rapid, high-throughput, and cost-effective approach to identify integrin-targeting ligands. Using the OBOC approach, we previously discovered three cyclic peptide ligands that target α3β1 in U-87MG cells: LXY1, LXY3, and LXY4. The order of in vitro binding affinity to U-87MG cells is LXY4>LXY3>LXY1. However, LXY3 or LXY4 did not show improvement in tumor targeting over LXY1 in optical imaging studies on a U-87MG glioblastoma xenograft model (data not shown). We hypothesized that further optimization of ligand with higher affinity would be able to improve in vivo tumor-targeting capacity.
In this report, we used a combination of the focused OBOC library and traditional medicinal chemistry approaches to rapidly identify LXY30, an improved and potent ligand against the α3β1 integrin. First, we synthesized three focused OBOC libraries designed for targeting α3β1 integrin. Many
d- and unnatural amino acids were used as building blocks to increase in vivo stability. In order to minimize the time and cost of synthesis and screening of OBOC libraries, all three OBOC libraries L1–L3 contain three sub-libraries that have 0–2 amino acids at the C-terminal of LXY1. In theory, we should have been able to identify LXY30 from the screening of library L1, but we failed to pick it up. The reason for this is that the actual number of screened beads (200-μL beads, ~1.5 × 10
5) is lower than the theoretical permutation (9.18 × 10
5). In order to cover all possible ligands, a tenfold high number of permutation beads would have to be screened which is neither practical nor necessary. Nevertheless, we had obtained very useful SAR information from screening the three OBOC libraries (Fig.
1 and Table
1): No additional amino acid is needed at the N-terminus of LXY1, but a hydrogen donor residue (or a polar residue) at the C-terminus might improve binding since it can provide additional interaction with adjacent binding pocket. In addition, hydroxyproline (Hyp) is more favorable (12 out of 13) than
l-proline (P) for the X
2 position because its hydroxyl group can form an extra hydrogen bond with α3β1 integrin. Based on this SAR information, 28 new analogs of LXY1 were designed with majority having a hydrogen bond-forming (donor or acceptor) amino acid at the C-terminus. Those compounds were quickly synthesized using standard solid-phase peptide synthesis approach and tested using an in vitro cell binding assay. LXY30 was found to be most potent (Table
2). At the X
1 position, the order of ligand binding affinity is Phe(3,5-diF) > Phe(3,4,5-triF) > Phe(3,4-diF) > L, which is consistent with our previous finding indicating the finding was not obtained by chance [
22]. Introduction of
l-arginine, a positively charged amino acid, at the X
4 position of the C-terminus resulted in an increase in binding affinity (LXY30 vs LXY4). In contrast, a negatively charged amino acid at the X
4 position decreased binding (LXY13 vs LXY14 and LXY8 vs LXY1). However, ligands with
d-arginine at the X
4 position have weaker binding than their
l-arginine counterpart (LXY38 vs LXY30, LXY36 vs LXY7), indicating that the conformation of arginine is also important for the binding. Insertion of an oxygen atom in the peptide chain (LXY33 and LXY39) resulted in reduced binding, implying that the size or polarity of the ring should not be changed. Replacement of the amino acid with
N-substituted glycine led to a significant loss of binding (LXY7 vs LXY34), indicating that the side chain appended to the α-carbon is very important for the binding activity. The resulting peptide, LXY30, not only had improved in vitro binding to all three tested glioblastoma cells (Fig.
1) but also demonstrated better in vivo tumor targeting than LXY1 (Fig.
3). We further tested the binding of LXY30 to a panel of lung and breast cancer cell lines and found that LXY30 bound to different tumor cell lines with a wide range of variable affinity (Additional file
1: Table S2 and Figure S3). It can be explained by the heterogeneous expression level of α3β1 integrin on these cell lines verified by a recently published report [
25], in which the trend of α3β1 integrin mRNA expression levels on these cell lines is consistent with their LXY30 binding affinities (Additional file
1: Table S2).
Our study has several translational potentials. First, overexpression of α3β1 integrin has been shown in multiple, aggressive tumor types, including basal-like breast cancer cells and lung cancer cells [
9,
27]. Recent studies suggest that α3β1 integrin-mediated signaling plays a critical role in the initiation and/or growth of mammary tumors [
28]. Overexpression of α3β1 integrin is associated with spontaneous metastasis of breast [
14] and lung cancer cells to the brain [
13] and mediates the resistance of HER2+ breast cancer cells to cancer therapy [
29]. Second, the internalization property of LXY30 on α3β1 integrin-expressing cancer cells could be used to facilitate delivery of conventional chemotherapeutic agents, target-specific agents, siRNAs, and microRNAs into tumor cells, either through direct conjugation or by encapsulation inside LXY30-decorated nanocarriers. Third, there is a largely unmet need to improve drug delivery to refractory, metastatic tumors while sparing the normal cells that have been exposed to accumulative dose of cancer therapeutics. Finally, while malignant gliomas are the most prevalent type of primary brain tumor in adults, central nervous system (CNS) metastases from epithelial tumors are much more common than primary brain tumors [
30]. The incidence of CNS metastases has been rising in recent years due to early detection with imaging and improved systemic therapy (especially oncogene-driven molecularly targeted therapy) for extra-CNS tumors [
31,
32]. CNS metastases may occur in 20–40 % of patients with cancer, of which 60–75 % are symptomatic. In adults, lung cancer is the number one primary cancer that is most likely (>50 %) to metastasize to the brain, followed by breast (15–25 %) and skin (melanoma, 5–20 %) cancer [
30,
33]. Patients with CNS metastases often have short survival and significant mental and physical debilitations that create extra burdens on both the patient and family members. This represents another unmet clinical need to improve the diagnosis and treatment for CNS metastasis. The fact that LXY30-biotin/streptavidin-Cy5.5 complex with over 80,000 Da can target intracranial implanted xenografts (Fig.
4) indicates that LXY30 is an excellent cancer-specific ligand for guiding drug delivery to primary or metastatic tumors in the brain.