A Comprehensive Evaluation of the Activity and Selectivity Profile of Ligands for RGD-binding Integrins

Integrins, a diverse class of heterodimeric cell surface receptors, are key regulators of cell structure and behaviour, affecting cell morphology, proliferation, survival and differentiation. Consequently, mutations in specific integrins, or their deregulated expression, are associated with a variety of diseases. In the last decades, many integrin-specific ligands have been developed and used for modulation of integrin function in medical as well as biophysical studies. The IC50-values reported for these ligands strongly vary and are measured using different cell-based and cell-free systems. A systematic comparison of these values is of high importance for selecting the optimal ligands for given applications. In this study, we evaluate a wide range of ligands for their binding affinity towards the RGD-binding integrins αvβ3, αvβ5, αvβ6, αvβ8, α5β1, αIIbβ3, using homogenous ELISA-like solid phase binding assay.

-values for the integrin ligands investigated for the subtypes αvβ3, αvβ5, αvβ6, αvβ8, α5β1, and αIIbβ3. Specificity or subtype with the lowest IC 50 -value are highlighted. All values were referenced as given in the description of the assay and in the SI. * (S-Enantiomer). cell binding on different surfaces 39,40 . These molecules were developed by ligand oriented molecular design and later refined on the basis of X-ray structures of the integrins α vβ 3 and α IIbβ 3 32,33,41 and the homology model for α 5β 1 [42][43][44] . It is interesting to note that the later published crystal structure of α 5β 1 structurally confirmed the selectivity described for these ligands 45 .
Both, the α vβ 3-and the α 5β 1-selective ligands bind the integrins via the kinked RGD motif, but differences are found in the binding mode of the arginine side chain. As illustrated in Fig. 1, which depicts a linear RGD-ligand in the binding pocket, the guanidine group of arginine is binding in a side-on manner to the Asp218 of the α -subunit of α vβ 3, forming a bidendate salt bridge. In addition to this side-on interaction (Asp227 in α 5), an end-on interaction of guanidine and Gln221 can be observed in the crystal structure of α 5β 1 45 . Keeping this difference in mind, the selectivity of the ligands can be explained as follows: the amino pyridine in sn243 interacts side-on with the α v-pocket but does not allow an end-on binding due to sterical hindrance, thus reducing strongly the α 5β 1 affinity. On the other hand, 44b has a full guanidine function allowing side-on and end-on interactions. Recently, we described a way how this small difference between the α v-and the α 5-binding pocket can be utilized to design selective peptidic integrin ligands. By alkylation of the N ω of the guanidine group of arginine, the α 5-specific end-on interaction is blocked, leading to a shift in selectivity for α v-integrins 46 .
Apart from targeting integrins α vβ 3, α 5β 1, or α IIbβ 3, other clinically relevant integrin subtypes have been explored 16 . For instance, several linear peptides, containing a helical DXLLX motif, were shown to selectively bind α vβ 6 and α vβ 8, and display low affinity towards all other subtypes. The biological role of α vβ 6 and α vβ 8 is quite similar, as they are both participating in the activation of transforming growth factor-β (TGF-β ) by interacting with the same endogenous ligands TGF-β 1 and TGF-β 3 47 . Finally, α vβ 3 and α vβ 5 share very similar biological roles (stimulate angiogenesis), but they perform this task via different mechanisms 48 . Nevertheless, their close structural similarity hampers the development of selective ligands to these integrins 49,50 .
The aims of this work. Since the discovery and first application of integrin-binding RGD peptides in the 1980s, and based on their great impact in medicine, biology, and biophysical sciences, the design and use of synthetic integrin ligands attracts much attention. Most of the current research is focused on the discovery of new integrin-selective ligands and their use for drug delivery, diagnosis, and tumor imaging, which are crucial for developing effective personalized medical platforms.
However, unequivocally ascribing a specific biological role to one integrin receptor remains problematic. Even when high binding affinity towards one distinct integrin subtype is achieved, it often remains unclear whether the observed biological effect is not based on a residual effect on another subtype. This may be attributed to the fact that most studies so far have only focused on the selectivity between a reduced subset of integrins, e.g. α vβ 3 vs. α IIbβ 3, or α vβ 3 vs. α 5β 1, but have totally neglected the influence of other closely related integrins of the RGD-binding family. This is particularly problematic as integrin expression strongly depends on cell and tissue type, crosstalk within distinct integrin subtypes, time point of study, and biological environment (e.g. tissue type). Last but not least, the activities reported for integrin ligands are usually evaluated using different experimental protocols and are, thus, highly variable.
Consequently, no reliable comprehensive comparison of the IC 50 -values of biologically prominent integrin ligands has been made, so far. Newly designed integrin ligands have seldom been evaluated for their selectivity against a full panel of RGD-binding integrin subtypes, mainly because there were no reliable testing systems established. In this work, we have evaluated a large number of well-known and widely used integrin-targeting molecules using the same standardized competitive ELISA-based test system, by measuring the inhibition (i.e. IC 50 values) of integrin binding to immobilized natural ECM ligands. In order to facilitate a direct comparison, we show here (Tables 1 and 2) the affinity values determined with our test system, which always contained reference compounds to standardize the biological data. For some ligands, IC 50 values were already reported in the literature and the inhibitory activities might slightly deviate from our present data. However, we consider that a direct comparison under identical conditions is very important and thus only represent the data determined during this work. This study includes RGD-based linear and cyclic peptides, peptide-mimetics as well as commonly used reference compounds. Furthermore, in this study we demonstrate, with carefully selected molecules, how functionalization of integrin ligands (e.g. with chelators, anchoring groups) can affect their binding affinity and selectivity. The investigated integrin subtypes studied here include: α vβ 3 and α vβ 5 (both binding to vitronectin), α vβ 6 and α vβ 8 (binding to LAP), α 5β 1 (binding to fibronectin) and α IIbβ 3 (binding to fibrinogen), which are all RGD-binding (the only missing RGD-binding integrins are α vβ 1 and α 8β 1, which could not be screened in our test system) and have relevant clinical implications. Since none of the presented ligands have previously been evaluated against such an exhaustive panel of integrin subtypes, the results of this study will provide unprecedented insights into the binding and selectivity profiles of synthetic integrin ligands, thus being of great value for the further development of integrin inhibitors for medical applications. In general, these binding activities correlate very strongly with the inhibition of signal transduction and with the binding affinity of the biochemically highly complex focal adhesions to ECM proteins.
An overview of the main approaches for testing integrin subtype-specific ligands. For the development and optimization of biologically active integrin ligands it is of utmost importance to use a reliable and reproducible test system, which yields values for biological activity with low statistical variance and high precision. A careful revision of previously published integrin-binding affinity data for well-known integrin ligands seems to indicate clear differences between the methods used. In general, cell-based methods are strongly dependent on the experimental condition of the study. Thus, the affinity data obtained for the same compound and the same integrin may greatly vary in different cell-based studies. In contrast, non-cellular systems, which are based on the use of isolated integrin receptors, tend to exhibit better biochemical precision and reproducibility. The major drawback of these methods, however, is the fact that they represent a simplified and artificial system, which does not fully mimic the intricate nature of integrin-ligand interactions and the subsequent response of the adhesome-associated signaling. Thus, an efficient combination of both systems is highly recommended for an optimal and efficient development of integrin-targeting drugs. In the following section, several cellular and non-cellular tests are briefly described.
In vitro cellular tests have been widely used to obtain integrin affinity data. A well-established and commonly used method is based on the concentration-dependent inhibition of cell attachment to a surface that is usually coated with the native cell adhesive proteins 5   by transmitted light microscopy, by fluorescence microscopy (in cases when the cells are tagged), or upon use of other functional imaging approaches. Such tests can be performed with a variety of cell types (e.g. NRK-epithelial cells, MG-63, MDA-MB-435 51 or REF52 52 ), as well as with platelets isolated from platelet-rich plasma. Another popular cell-based technique is the calculation of integrin binding affinities based on ligand binding assays 53 . Suspended or adherent cells are incubated with increasing concentrations of an integrin ligand, and afterwards with a radiolabeled ligand that also shows integrin-binding affinity, such as 125 I-Echistatin or 125 I-c(RGDyK). The radioligand is therefore competing with the compound for binding to the integrin receptors on the cell surface and serves as an internal standard reference for the binding affinity. As described for the cell adhesion assay, a great choice of integrin expressing tumor cell lines as well as epithelial cells are available, such as 293-b β 3 54 or U87MG glioblastoma 55 .
Cell-based tests hold great potential to evaluate not only ligand binding but also its capacity to trigger biological responses relevant for the physiological context. This holds true at least for cases in which the "reporter" cell type is physiologically relevant. It is noteworthy that cellular tests possess some serious intrinsic limitations that are of particular importance during drug development. The major drawback is the limited control over homogeneity in the levels of integrin expression present in the different cell lines used, or even in individual cells in the tested population. Typically, there is one highly overexpressed integrin subtype presented on the surface, but there may also be other minor populated integrin subtypes to which the tested compounds may bind. It is even more problematic that these subtype expression levels may change over time. In addition, the total surface receptor density can strongly vary depending on specific conditions (number of passages in culture, presence of other integrins, cell culture conditions, state of cell cycle etc.). For example, a phenomenon known as "integrin crosstalk" has been proven for α vβ 3 and α 5β 1. Specifically, α 5β 1 integrin was shown to modulate (up or down) the expression of another subtype and thus significantly alters the expression pattern 56 . Consequently, the comparison of affinity data measured for different cell lines is highly challenging.
Tests carried out in cell-free systems, use isolated extracellular domains of integrins in conjunction with ECM proteins, either extracted from human tissue or produced by recombinant methods. Most of these tests are based on competitive solid phase binding assays, in which one component is bound to the multi-well plate and in a subsequent step, soluble ligands are added, testing their capacity to block the binding to that component. Mainly two procedures are described in the literature, differing in the molecule used to coat the surfaces: the isolated integrin extracellular domains or the ECM protein. In the first case, the integrin is coated on the surface, followed by an incubation with a mixture of the native ECM protein and increasing concentrations of the ligand of interest 49 .
In an alternative test system, the natural ECM protein (vitronectin for α vβ 3 and α vβ 5, LAP (TGFβ ) for α vβ 6 and α vβ 8, fibronectin for α 5β 1 and fibrinogen for α IIbβ 3) is immobilized onto the surface, and the soluble integrin, together with a serial dilution of the inhibitory ligand, is added afterwards 57 . The read-out in both procedures is usually done in an ELISA-like manner by using conjugated antibodies recognizing the integrin head groups. A detailed schematic illustration of the different steps of the integrin binding assay is presented in Fig. 2. In comparison to many other test systems, this system allows the accurate (SD~10%) and reproducible determination of IC 50 values for almost all RGD-binding integrin subtypes. For a detailed description of the reference to the SI. As the quality of the integrins strongly depend on the batch and providers a reference compound always have to be used for each test plate as internal standard (we used: Cilengitide, c(RGDf(NMe)V) (α vβ 3-0.54 nM, α vβ 5-8 nM, α 5β 1-15.4 nM), linear peptide RTDLDSLRT (α vβ 6-33 nM; α vβ 8-100 nM) and tirofiban (α IIbβ 3-1.2 nM).
Preliminary studies in our laboratories comparing the two methods (surface-bound integrin vs. soluble integrin) revealed significant differences in the antagonistic activity of control ligands in regards to the integrin subtype used. Whereas very similar IC 50 values were found regardless of the method used for integrins α vβ 3 or α vβ 5, the activity towards α 5β 1 seemed to be highly dependent on the experimental protocol. The drug Cilengitide exhibited very high affinity (in the low nanomolar range) towards soluble α 5β 1 when fibronectin (i.e. the natural ECM ligand) was immobilized on surfaces. However, coating of the integrin and subsequent incubation of Cilengitide with fibronectin generally resulted in low antagonistic activities and poor reproducibility within assays (unpublished data). This was already observed in our stem peptide c(RGDfV) 26 , and may be attributed to the possible denaturation of α 5β 1 integrin.

Results and Discussion
The gold standard integrin inhibitor, Echistatin. The disintegrin Echistatin was first isolated in 1988 from snake venom as an effective inhibitor of platelet-fibrinogen interaction, as well as of platelet aggregation 58 . This small folded protein contains an RGD-sequence in a well-exposed loop which was described to bind to α IIbβ 3, α vβ 3, α vβ 5, and α 5β 1 with very high affinity 59 . Since the Tyr-31 residue of Echistatin can be labeled with 125 I using a standard procedure has turned this compound into a commonly used positive control for many competitive cellular integrin binding assays 60,61 . We therefore included Echistatin in our study to determine its selectivity profile and to compare its binding activities to those previously obtained from cellular assays. Echistatin showed a very broad affinity pattern. It binds to the whole panel of investigated integrins with IC 50 -values in the low nano-molar range ( Table 1). As already reported, it shows particularly low IC 50 -values for α vβ 3 (0.46 nM), α 5β 1 (0.57 nM), and α IIbβ 3 (0.9 nM). Interestingly, Echistatin exhibited the lowest IC 50 -values for these integrins compared to all other compounds investigated within this study. Thus, it also represents an ideal candidate to be included as a positive control in cell-free integrin binding tests.
Linear RGD integrin inhibitory peptides. The RGD-sequence was originally discovered as the minimal binding epitope of fibronectin and has extensively been investigated over the last decades. Moreover, it has been shown that the presence and chemical nature of flanking residues have a strong influence on its activity 5 . Thus, early studies with the RGD-motif were conducted with linear tri-to heptapeptides, based on the sequence found in fibronectin. In our study, the following linear peptides were included: RGD, RGDS, GRGD, GRGDS, GRGDSP, and GRGDSPK. In original reports, these peptides were described to bind to α vβ 3 and α vβ 5, but also showed relatively good IC 50 -values for the integrin subtypes α 5β 1 and α vβ 6, as well as low affinity for α IIbβ 3 62 . Apart from the compounds mentioned above, GRGDNP 63 and GRGDTP 64 peptides were also included in our test system, as they are frequently used in biological studies. Especially, GRGDNP has been described to prefer binding to α 5β 1. In our evaluation, all linear peptides showed the lowest IC 50 -values for the integrin subtype α vβ 3 (12-89 nM), with IC 50 -values for α vβ 5 ranging from 167 to 580 nM and for α 5β 1 from 34 to 335 nM. These peptides generally displayed high IC 50 -values towards α vβ 6 and α vβ 8. More surprisingly, none of the linear peptides exhibited an IC 50 -value below 10 μ M on α IIbβ 3. These results demonstrate that the linear RGD peptides are active on integrins α vβ 3, α vβ 5, and α 5β 1, and selective against α vβ 6, α vβ 8 and α IIbβ 3. As presented in Table 1, the residues flanking the RGD-motif essentially contribute to the binding affinity for α vβ 3 (and to a lower extent for α 5β 1 and α vβ 5). In particular, the IC 50 -value to α vβ 3 increases 7-fold from the linear tripeptide fragment RGD (89 nM) to the heptapeptide GRGDSPK (12.2 nM).
Cyclic RGD peptides. A major disadvantage of linear peptides is their low stability regarding enzymatic degradation, limiting their applicability for in vivo studies 65 . This can be significantly improved by cyclization and incorporation of a d-amino acid residue, as illustrated by cyclic pentapeptides of the formula c(RGDxX) and cyclic hexapeptides 66 . Moreover, reduction of the conformational space by cyclization can improve the biological potency of linear peptides when the bioactive conformation is matched 67 . One of the first cyclic compounds that was developed and used in cellular studies was c(RGDfV) which showed outstanding affinity for α vβ 3, while retaining total selectivity against α IIbβ 3. This core structure was later modified to develop a series of new ligands with improved activity and selectivity profiles 68,69 . Extensive Structure-Activity Relationship (SAR) studies on the model sequence c(RGDxX) showed, that the presence of an aromatic amino acid in the d-configuration (i.e. d-Phe, d-Tyr, d-Trp) at the position 4 (residue x) was essential for the α vβ 3-binding affinity, whereas the amino acid at position 5 (residue X) had little effect on the biological activity 68,69 . Based on the stem peptide c(RGDfV), Cilengitide, c(RGDf(NMe)V), the most active cyclic pentapeptide described to date, was developed via a systematic N-methylation scan. Cilengitide has a half-life in man of about four hours and is not metabolized systemically 70 . It became a drug candidate in phase II and III clinical studies for the treatment of different tumors [71][72][73][74] , however failed in phase III as drug against glioblastoma. Despite its extraordinarily low IC 50 -value for α vβ 3 and α vβ 5, the higher, but still significant value for α 5β 1 subtype is often neglected 75 . In this regard, our study clearly shows that this compound has also a remarkably low IC 50 -value for α 5β 1 (14.9 nM). Noteworthy, the IC 50 for α vβ 3 (0.61 nM) and α vβ 5 (8.4 nM) were the highest obtained among all synthetic peptides developed and studied. As previously indicated, the valine residue can be substituted by almost any other amino acid 66 . Hence, for biophysical or medical applications where a functionalization or ligation of integrin ligands is needed, the derivative c(RGDfK) 76 is often used. Lys has been found to be a good anchoring point for the attachment of functional units not only because it does not affect the binding affinity of the stem peptide significantly, but also because it easily allows the linkage of other chemical groups via the free amine. A pilot study by us used acrylate-functionalized derivatives of c(RGDfK) to mediate the adhesion of osteoblasts on a polymethylmetacrylate (PMMA) surface 77 . Since that time, c(RGDfK) and also c(RGDfE) were functionalized for a large number of biological applications 78,79 . Other cyclic penta-peptides used for functionalization are c(RGDyK) 80 as well as c(RGDfC) 81 , are both included in our study. Just like the linear peptides, all of the cyclic pentapeptides of the type c(RGDxX) tested showed moderate to low IC 50 -values for α vβ 3, α vβ 5, and α 5β 1, and no binding to α IIbβ 3, which is of major importance for in vivo applications. Noteworthy, all the cyclic peptides displayed lower α vβ 3 IC 50 -values (i.e. in the range of 1.5 to 6 nM) compared to the linear derivatives, and followed the order c(RGDfV) < c(RGDfK) < c(RGDyK) < c(RGDfC). Therefore, cyclic RGD-peptides should be the preferred choice when high α vβ 3-binding activities are required. The IC 50 values for α vβ 5 and α 5β 1 varied from 250 to 503 nM and from 141 to 236 nM, respectively. Interestingly, all the cyclic compounds showed relatively low IC 50 -values for α vβ 6 (49-75 nM), which has not been discussed in any of the references so far.
Ruoslahti et al. discovered the RGD-containing double cysteine-bridged (1-4, 2-3) peptide RGD-4C 82 (ACDCRGDCFCG) by phage display and reported a low IC 50 -value towards the subtypes α vβ 3 and α vβ 5, and specificity over α 5β 1. This peptide represents one of the most commonly used molecules in cellular tests and in in vivo studies and has been conjugated to target α vβ 3-overexpressing cells 83 . To reduce the synthetic complexity that arises from two disulfide bridges, Hölig et al. developed the single cysteine-bridged peptide RGD10 (GARYCRGDCFDGR) 84 , which has the same IC 50 value and selectivity properties as the original RGD-4C, and has been functionalized for different applications as well, (e.g. for surfaces coating or targeting liposomes). In our studies, the RGD-4C peptide exhibited an IC 50 -value for α vβ 3 of 8.3 nM, which is comparable to that observed for the best linear RGD-sequences but considerably lower than that of cyclic RGD-containing penta-peptides. Furthermore, it also shows a good value for α vβ 5 (46 nM), a high IC 50 -value for α vβ 6 and α 5β 1, and no affinity for α vβ 8 and the platelet integrin α IIbβ 3. As reported in other studies, RGD-10 exhibits a similar pattern of bioactivity, though with a trend towards increased α vβ 3/α vβ 5 selectivity: the IC 50 for α vβ 3 remains almost the same, whereas the α vβ 5 affinity drops to 102 nM. Based on the RGD-4C peptide, Indrevoll et al. developed a PEGylated bicyclic, mono cysteine-bridged peptide with the sequence KCRGDCFC (NC100717) 85 targeting the α vβ 3 and α vβ 5 integrin subtype. This scaffold is the targeting unit of functionalized compounds (chelators and dyes), e.g. the 18 F-labeled compound 18 F-AH111858 (Fluciclatide 86 , GE), described in detail in the section of functionalized molecules. NC100717 showed a low nanomolar IC 50 -value for α vβ 3 (1.1 nM) and α vβ 5 (41 nM) in our test system. Nonetheless, for application of disulfide-bridged cyclic peptides it may be of interest to consider the stability of disulfide bridges in vivo 87 . Finally, another integrin binding motif in fibronectin is the inverse sequence isoDGR which is based on the NGR motif after in situ rearranging from asparagine to iso apartate 88,89 . The integrin subtype selectivity strongly varies depending on the flanking residues of the sequence. Recently, the Scientific RepoRts | 7:39805 | DOI: 10.1038/srep39805 compound c(phgisoDGRk) 50 , which is bi-selective for α vβ 6 and α 5β 1, was identified and used for cellular studies. In our test system, it could be shown here that the compound moreover binds to α vβ 8 as well.
Peptidomimetics and other ligands. The non-RGD linear pentapeptide Ac-PHSCN-NH 2 was derived from the synergy domain of fibronectin and is clinically developed under the trade name ATN161 90 for the treatment of several solid tumors as it is highly active for the α 5β 1 subtype, with some affinity for α vβ 3 and α vβ 5 16 . Interestingly, ATN161 showed a clear selectivity for α 5β 1 (4.2 nM) in our testing system, being essentially inactive for all other integrins investigated.
The non-peptidic compound JSM6427 36 was designed by Stragies et al. and later developed in clinical phase for the treatment of age-related macular degeneration as it strongly inhibits neovascularization in the eye. It is described to be a relatively selective α 5β 1 antagonist, although values for α vβ 6 and α vβ 8 were not published. Indeed, we found JSM6427 to be tri-selective for α 5β 1 (2.5 nM), α vβ 6 (23 nM), and α vβ 8 (8.2 nM).
Recently, we reported the synthesis and binding affinity of the highly active α vβ 3-selective sn243 37 and the α 5β 1-selective 44b 91 peptidomimetic ligands. In a proof-of-concept study, these compounds were functionalized and used for molecular imaging 92 as well as for biophysical studies 39,40 , showing their potential to discriminate the two integrin subtypes α vβ 3 and α 5β 1 both, in vitro and in vivo. Here, we also included them to evaluate their full pattern of integrin selectivity, and found, besides the expected high activities for the corresponding subtype (sn243: 0.65 nM α vβ 3; 44b: 2.3 nM α 5β 1), a low nanomolar IC 50 -value for α vβ 8 for 44b (37 nM).
Integrin αvβ6-binding ligands. The α -helical binding motif DLXXL, an α vβ 6 subtype-specific binding motif, was initially discovered by phage display and later shown to be also present in the natural α vβ 6 ligand latency associated peptide (LAP). Two compounds containing this motif are included in this study as they have extensively been used for addressing selectively α vβ 6-expressing cells in vivo, e.g. as targeting unit for molecular imaging. The RTD containing 9-mer peptide RTDLDSLRT 93 as well as the 20-mer peptide A20FMDV2 94 , which is derived from a foot and mouth disease virus peptide (sequence: NAVPNLRGDLQVLAQKVART), were proven to be subtype-selective in our study and exhibited an IC 50 -value of 29.5 and 0.93 nM for α vβ 6, respectively. The peptidomimetic compound Mol11 49 , which was described as the first α vβ 6 selective small molecule ligand, was resynthesized for this study as an enantiomerically pure (S-enantiomer) compound and indeed showed a very low IC 50 -value for α vβ 6 (1.3 nM), but also good values in the lower nanomolar range for α vβ 3 (13.2 nM), and α vβ 8 (18.5 nM). Recently, we were able to develop the cyclic peptide c(FRGDLAFp(NMe)K) 38 , which mimics the binding epitope of the helical DLXXL-motif. The IC 50 of this peptide was determined to be 0.28 nM for α vβ 6, the highest affinity among the compounds investigated.
Integrin αIIbβ3-binding ligands. The α IIbβ 3 integrin receptor, also known as glycoprotein receptor (GP)-IIb/IIIa, is expressed uniquely on the surface of platelets and megakaryocytes, a type of platelet-producing cells in the bone marrow. By binding to its natural ligand fibrinogen, α IIbβ 3 is involved in primary hemostasis during platelet formation. Thus, application of α IIbβ 3 ligands was earlier explored as a clinical approach for anti-thrombotic therapy 15 . Today, there are two FDA-approved drugs targeting selectively the α IIbβ 3 receptor, Intrifiban (Eptifibatide) 95 and Tirofiban (Aggrastat) 96 , both clinically used for patients with acute coronary syndromes undergoing percutaneous coronary intervention. Eptifibatide is a cysteine-bridged cyclic RGD-containing hexapeptide. It was first described in 1993 as a potent subtype-selective integrin antagonist and is, as well as the small molecule α IIbβ 3 inhibitor Tirofiban, used both clinically and preclinically. Another α IIbβ 3 selective compound included in this comparison of integrin ligands is GR144053 97 . It is described as an orally available, highly potent subtype-selective fibrinogen inhibitor and is used in many preclinical studies. In our test system, these three compounds have been confirmed to be selective α IIbβ 3 inhibitors. The lowest α IIbβ 3 IC 50 -value was determined for Tirofiban (1.3 nM), followed by Eptifibatide (2.8 nM) and GR144053 (18 nM).
Functionalized compounds. Many of the integrin ligands that are used for biophysical or medical experiments require a functional unit (e.g. for a strong covalent binding to the surface) or a chelator for molecular imaging. For that reason, the bioactive moiety is linked to this functionality via a spacer that separates the two entities of the molecule. Ideally, by using the right anchoring point in the molecule, the loss of activity upon functionalization is low. Nevertheless, the size, lipophilicity, and other parameters like the rigidity of the functional as well as the spacing unit can influence the binding of the bioactive moiety to its target 98,99 . To estimate the influence on the IC 50 -values after modification of the ligand, five compounds were chosen as examples ( Table 2). Among three of them, c(RGDfK), which is functionalized via its lysine side chain, represents the bioactive targeting unit. The IC 50 -value of the unmodified ligand was determined to be 2.3 nM for the α vβ 3 integrin subtype (Table 1). After modification to F-Galacto-c(RGDfK) 100 (for molecular imaging) and c(RGDfK)-Peg-MPA) (MPA = mercapto propionic acid), the IC 50 moderately decreased to 8 and 15 nM, respectively. Exactly the opposite, namely a better α vβ 3 IC 50 was observed for the (Ga)NOPO-c(RGDfK) (1.1 nM) 101 . This phenomenon could already be observed for previously published compounds and can be explained, inter alia, by charge effects and/ or altered van-der-Waals interactions. For example, large substituents like chelators possess a high surface area and can randomly interact via unspecific van-der-Waals interactions with parts of the protein. This weak interaction decreases the k off -rate and thus can lead to IC 50 -values compared to the original targeting peptide alone. Concerning the selectivity profile of the modified compounds, no changes are observable. This is also found for the α vβ 6/α 5β 1 bi-selective peptide c(phgisoDGRk), where the selectivity is not affected but the IC 50 -value for the two integrins doubles (2-fold) after functionalization to c(phgisoDGRk)-Peg-MPA) 52 . Fluciclatide, an imaging agent developed by GE Healthcare, consists of NC100717 as targeting unit 84 . For this example, the IC 50 -value for α vβ 3 decreased 3-fold and no change in selectivity was observable. To sum up, the functionalization of a bioactive molecule can alter its integrin binding affinity. This means that in principle, every functionalized compound for Scientific RepoRts | 7:39805 | DOI: 10.1038/srep39805 both in vitro and in vivo applications should be tested individually to obtain comparable results. It is important to mention here that e.g. the introduction of a PEG spacer does not guarantee better ligands properties (solubility, affinity). Peg is not always extended in aqueous solution and thus the distance between the biomolecule and the functional unit is not defined 99 . Changes in affinity induced by a spacer and a functional unit are hard to predict. Additionally, functionalization of bioactive compounds can also strongly alter the pharmacodynamics, e.g. the total uptake and distribution of the compound in the organs after its intake. Mostly, this is observable because of a big change in lipophilicity due to the different modifications. Especially for molecular imaging, where only defined structures in the body should be visualized, it is important to evaluate and optimize the effect of every modification (e.g. different spacer, chelator, coordinating metal) in this regard. Important points to be addressed regarding the applicability of the probes in standard procedures of diagnosis are the simplicity of production and the flexibility in the type of tracer that can be introduced. In the case of functionalization for surface coating, the strength and stability of the binding to the surface by the anchoring unit and the length and chemical structure of the spacer for a defined purpose (e.g. cell adhesion) has to be taken into account.
Molecules used as negative control in the determination of binding activities. For cellular and in vitro experiments, molecules with comparable steric properties and lipophilicity but without biological activity often serve as control compounds. Targeted substitution of any of the three amino acids in the RGD-sequence leads to inactivation of the ligand. A substitution of glycine by alanine leads to steric repulsions on the binding groove between the α -and β -subunit. Moreover, any elongation of the ligand (e.g. glycine to β -alanine and aspartate to glutamate substitution) leads to complete loss of binding activity. The molecules can be functionalized in the same way as their active biologically active counterparts (e.g. via lysine side chain). For this study, linear and cyclic control molecules have been synthesized and evaluated.

Conclusion
After the initial discovery of the RGD sequence in fibronectin, a large number of integrin ligands, binding to RGD-recognizing integrins, were developed by many groups around the world and became highly important for medical applications and for biophysical studies. For the development of each of these compounds, different evaluation techniques (e.g. various cel-based and cell free methods) have been used, allowing a good comparison and selection process of the compounds within a study. However, comparing the values determined by different groups for the very same compound shows very high deviations. For this reason we evaluated for the first time the most frequently used compounds in a homogenous solid phase binding assay for their binding affinity to six RGD-binding integrins (α vβ 3, α vβ 5, α vβ 6, α vβ 8, α 5β 1, α IIbβ 3). This gives the possibility to choose the ligand with the ideal affinity and selectivity pattern for a given application and opens new doors for the application of those ligands. However, the complex mechanism including several steps of conformational transitions in which the initial ligand binding to the resting state of the integrin and the stronger binding in the focal adhesion, might have consequences for the IC50-value given here with data under different environmental condition in vivo102.

Methods Section
Integrin Binding Assay. The activity and selectivity of integrin ligands were determined by a solid-phase binding assay according to the previously reported protocol 103  Flat-bottom 96-well ELISA plates (BRAND, Wertheim, Germany) were coated overnight at 4 °C with the ECM-protein (1) (100 μ L per well) in carbonate buffer (15 mM Na 2 CO 3 , 35 mM NaHCO 3 , pH 9.6). Each well was then washed with PBS-T-buffer (phosphate-buffered saline/Tween20, 137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , 0.01% Tween20, pH 7.4; 3 × 200 μ L) and blocked for 1 h at room temperature with TS-B-buffer (Tris-saline/BSA buffer; 150 μ L/well; 20 mM Tris-HCl, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 1 mM MnCl 2 , pH 7.5, 1% BSA). In the meantime, a dilution series of the compound and internal standard is prepared in an extra plate, starting from 20 μ M to 6.4 nM in 1:5 dilution steps. After washing the assay plate three times with PBS-T (200 μ L), 50 ul of the dilution series were transfered to each well from B-G. Well A was filled with 100 ul TSB-solution (blank) and well H was filled with 50 ul TS-B-buffer. 50 ul of a solution of human integrin (2) in TS-B-buffer was transfered to wells H-B and incubated for 1 h at rt. The plate was washed three times with PBS-T buffer, and then primary antibody (3) (100 μ L per well) was added to the plate. After incubation for 1 h at rt, the plate was washed three times with PBS-T. Then, secondary peroxidase-labeled antibody (4) (100 μ L/well) was added to the plate and incubated for 1 h at rt. After washing the plate three times with PBS-T, the plate was developed by quick addition of SeramunBlau (50 μ L per well, Seramun Diagnostic GmbH, Heidesee, Germany) and incubated for 5 min at rt in the dark. The reaction was stopped with 3 M H 2 SO 4 (50 μ L/well), and the absorbance was measured at 450 nm with a plate reader (POLARstar Galaxy, BMG Labtechnologies). The IC50 of each compound was tested in duplicate, and the resulting inhibition curves were analyzed using OriginPro 7.5G software. The inflection point describes the IC50 value. All determined IC50 were referenced to the activity of the internal standard. αvβ3 (1) 1.0 μ g/mL human vitronectin; Millipore.