PET Diagnostic Molecules Utilizing Multimeric Cyclic RGD Peptide Analogs for Imaging Integrin αvβ3 Receptors

Multimeric ligands consisting of multiple pharmacophores connected to a single backbone have been widely investigated for diagnostic and therapeutic applications. In this review, we summarize recent developments regarding multimeric radioligands targeting integrin αvβ3 receptors on cancer cells for molecular imaging and diagnostic applications using positron emission tomography (PET). Integrin αvβ3 receptors are glycoproteins expressed on the cell surface, which have a significant role in tumor angiogenesis. They act as receptors for several extracellular matrix proteins exposing the tripeptide sequence arginine-glycine-aspartic (RGD). Cyclic RDG peptidic ligands c(RGD) have been developed for integrin αvβ3 tumor-targeting positron emission tomography (PET) diagnosis. Several c(RGD) pharmacophores, connected with the linker and conjugated to a chelator or precursor for radiolabeling with different PET radionuclides (18F, 64Cu, and 68Ga), have resulted in multimeric ligands superior to c(RGD) monomers. The binding avidity, pharmacodynamic, and PET imaging properties of these multimeric c(RGD) radioligands, in relation to their structural characteristics are analyzed and discussed. Furthermore, specific examples from preclinical studies and clinical investigations are included.


Multimeric Radioligands
The idea of multimeric radioligand, which consists of multiple structural motifs connected to a single backbone, started from the need to achieve higher binding avidity through cooperative multivalent interactions. In a way, this mimics the behavior of antibodies, which have multiple individual weak interactions with the target. These multiple weak interactions combine in a cooperative way and increase binding avidity [1,2]. Several binding modes have been proposed as an explanation for this cumulative effect, such as the simultaneous binding of the ligand to multiple receptors on the cell surface (Figure 1a), or the improved statistical effect, where the ligand binds to one receptor, but the ligand's apparent local concentration is increased (Figure 1b) [1].
Multimeric radioligands resemble a typical radioligand by retaining some of its structural features-e.g., a chelator group for radiometals-and differ from it, as they include multiple pharmacophores. Radioligands targeting specific receptors on tumor cell surfaces consist of a receptor binding section (pharmacophore or ligand), and a radiolabeled domain. The latter is either a radiometal chelator, which coordinates with a radiometal, or a Binding models for multimers on the cell surface: (a) The binding of a radioligand to a cell surface receptors and the multimeric approach resulting in simultaneous binding of two pharmacophores connected via a long linker with two receptors, (b) improved binding efficiency of a ligand, due to the increased apparent local concentration of the pharmacophore (statistical effect) in the micro-environment of the receptor; (c) basic principles for the design of monomeric and multimeric radioligands (where n = number of pharmacophores).
The linker properties are very important since they influence the orientation of the ligand, when it approaches the extracellular receptors binding domain, enabling or preventing simultaneous multi-receptor binding of the multimers [6]. Long, flexible linkers increase the distance between the binding domains, permitting simultaneous binding of the ligand to multiple receptors, while rigid linkers have less entropy penalty from binding [1].
For multimeric radioligands, a wide spectrum of pharmacophores has been used like small molecules, peptides, or fractions of antibodies [1,[6][7][8][9]. A limited number (n = 2-8) of pharmacophores e.g., c(RGD) n has been covalently connected using low molecular weight linkers e.g., polyethylene glycols (PEGs), while higher numbers of pharmacophores have been tethered on the surface of large molecules, such as liposomes or nanoparticles. The nanoparticle approach is not examined in this review.

18 F Labeled Cyclic RGD Multimers
The majority of PET imaging preclinical and clinical studies refer to 18 F labeled c(RGD) n multimers, while following them are the 68 Ga and 64 Cu labeled ones. 18 F is a PET radioisotope combining favorable physical and chemical characteristics (half-life = 109 min, β + 96.7%, 0.64 MeV), which can be easily produced as aqueous fluoride by means of proton irradiation of 18 O-enriched water with a biomedical cyclotron [31]. 18 F-radiolabeling is achieved using reactions of nucleophilic substitution on suitable precursors, which carry appropriate leaving groups e.g., halides, sulfonates, and ammonium cations, or aromatic rings with activating residues like electron-withdrawing groups in appropriate positions [32]. Poethko [22,33]. Biodistribution studies for c(RGDfE) derivatives 1-3 were performed in mice bearing human melanoma M21 xenographs, overexpressing α v β 3 integrin and M21-L, weakly expressing α v β 3 integrin (negative control). Tumor uptake was higher for the dimeric structure 2 (60 min post-injection [p.i.]: 2.48 ± 0.15 percent of injected dose per gram [% ID/g]; 120 min p.i.: 1.63 ± 0.13% ID/g) compared to the monomer 1 (60 min p.i.: 1.56 ± 0.15% ID/g; 120 min: 1.49 ± 0.10% ID/g). Tumor uptake (M21 tumor) was higher by a factor of 5-6 than the negative control (M21-L), and specific, as it was proved by blocking studies. In another complementary study between ligands 1-3, the tetrameric ligand 3 showed similar tumor uptake with 2 (120 min p.i.: 1.65 ± 0.08% ID/g), while its tumor-to-normal organ contrast ratios in all cases were better than both, 1 and 2. Regarding normal organ biodistribution, 2 showed rapid renal excretion and low activity levels in blood, liver, and intestines. Renal excretion rates were initially comparable for both 1 and 2, with monomer 1 showing higher excretion rates with time, while 2 remained in the kidneys. The tetrameric ligand, 3, showed intermediate clearance kinetics, predominant renal excretion, and lower unspecific tissue distribution, while its liver uptake was lower than that of 1 and 2. However, 3 showed significant metabolization, especially in liver, kidney, and tumor, which according to the authors was due to the effect of lysine linkers, which are present between PEG and Dpr groups, cleaving the ligand's bridging to monomeric and dimeric units.

PEG Linkers on 18 F Labeled Cyclic RGD Multimers
Following the example of Poethko et al. [22,33] who introduced a long PEG 6 group the Lys linker with the c(RGD) pharmacophores, shorter PEGs were introduced in between the linker and the labeling domain. The introduction of PEG spacer groups between the pharmacophores and the labeling domain improved the overall radiolabeling yield and the radioligands pharmacokinetics. Specifically, PEGs were introduced between the amino acid linker (E or K) and the radiolabeling domain and/or between the linker and the pharmacophores (Figure 3, 6). For example, when a PEG 3 was introduced in the structure of 4, resulting in compound [ 18 F]FB-PEG 3 -E(cRGDyK) 2 or [ 18 F]FB-PRGD 2 7 it increased the radiochemical yield (Yield > 60% and purity > 99%) [36]. In addition, 7 showed an improved pharmacological profile compared to 4, since it showed higher receptor binding avidity 7 (IC 50 = 40.6 ± 4.6 nM), 4 (IC 50 = 55.1 ± 6.5 nM), higher metabolic stability, similar integrin α v β 3 -specific tumor uptake (U87MG glioma xenograft model, 30 min p.i.: 4.9-5% ID/g) and reduced renal uptake 7 (60 min: 2.0 ± 0.2% ID/g), 4 (60 min: 3.0 ± 0.2% ID/g) [36]. The reproducibility of [ 18 F]FP-PRGD 2 7 as an integrin α v β 3 -targeting PET probe was also verified using a small animal PET and mouse tumor xenograft (human HCT116 colon cancer) model. During those studies, 7 showed reproducible results with relatively low variability [37].
[ 18 F]Alfatide, 10 were studied as a predictor of anti-angiogenic response in alveolar adenocarcinoma (A549, high angiogenesis) and prostate cancer (PC-3, low angiogenesis) animal models, at baseline and after treatment with an anti-angiogenic drug (apatinib, bevacizumab) or without (control). The evaluation index for the inhibition of tumor growth in the individuals in the treated groups was represented by the treatment/control (T/C) ratio (%). Anti-angiogenic responses were denoted by the changes in uptake of 10 in the same animal. Uptake of 10 in the A549 models was significantly higher than in the PC-3 models (SUV mean A549: 0.64 ± 0.07 and PC-3: 0.25 ± 0.02) and the same was true for T/N ratios (A549: 2.76 ± 0.62 and PC-3: 0.82 ± 0.11), before treatment. After anti-angiogenic treatment tumor growth was suppressed resulting in lower tumor/control values for the A549 xenografted tumors than the PC-3 tumors (apatinib, A549: 47 ± 11.46 and PC-3: 69 ± 26.74%; bevacizumab, A549: 57.80 ± 13.82 and PC-3: 90.27 ± 13.09%, respectively). In addition to the decreased [ 18 F]Alfatide 10 uptake in tumors after treatment, the degree of tumor response was also associated with the tumor uptake prior to treatment, indicating that [ 18 F]Alfatide PET may be a useful molecular imaging tool for individual patient selection prior to anti-angiogenic drug therapy [40].
In another study, 10 was stereotactically compared with 18 F-fluorodeoxyglucose ([ 18 F]FDG) and immunohistochemistry (IHC) staining in Lewis lung carcinoma (LLC) tumor-bearing C57BL/6 mouse model. Tumor-to-brain, tumor-to-lung, and tumor-to-heart ratios of 10 were significantly higher than those of [ 18 F]FDG (p < 0.001). The spatial heterogeneity of the tumors was detected, and the tracer accumulation enhanced from the outer layer to the inner layer consistently. The different SUVs in the different tumor areas represent different levels of angiogenesis and probably deserve different radiation doses for tumor control. Correlations were found between: 1) SUV RGD and the α v β 3 expression in tumors (R = 0.595, p = 0.019) and 2) SUV FDG and SUV RGD (R = 0.917, p < 0.001). The latter corresponded to the IHC findings for the expression levels of glucose transporter proteins and α v β 3 , which were also correlated (R = 0.637, p = 0.011) [41].
The introduction of PEG 4 (15-amino-4,7,10,13-tetraoxapentadecanoic acid) in the linker was another very popular choice for the synthesis of c(RGD) multimers. A PEG 4 group was introduced either in between the alpha-amino group of the Glu linker (ligands 11-13) and/or connecting with amide bonds each of the pharmacophore groups (ligands 14-16) ( Figure 4). Among those radioligands, the most studied is [ 18  protein-bound particles through a dual tracer imaging approach. The parameters fitted with compartmental modeling from the dual-tracer dynamic imaging were consistent with those from single-tracer imaging, substantiating the feasibility of this methodology. Moreover, dual-tracer imaging was able to measure 14 binding potential value and [ 18 F]FDG influx simultaneously to evaluate tumor angiogenesis and metabolism. Such changes are known to precede anatomic changes, offering the promise of early prediction of therapy response with this method [42]. 14 could also monitor early treatment response and predict the therapeutic efficacy of the multi-targeted, anti-angiogenic drug sunitinib in U87MG tumors [43]. Finally, 14 was also evaluated on muscular inflammation [44].
A comparison between the dimeric peptide [ 18 [45] showed a higher radiolabeling yield for 16 (55 ± 12%) than 11 (25 ± 11%), because of the additional PEG 4 spacers, which decreased the steric hindrance during the addition of the prosthetic group. In vitro testing (U87MG cells) showed slightly improved avidity for 16 (IC 50 = 35.8 ± 4.3 nM) in comparison to 11 (IC 50 = 47.4 ± 5.8 nM). In vivo comparison of 16 and 11 in U87MG human glioblastoma and MDA-MB-435 human melanoma tumor xerographs in mice, also favored 16, which presented higher U87MG-tumor uptake at all-time points studied (e.g., 16: 5.32 ± 0.36% ID/g vs. 11: 3.02 ± 0.44% ID/g, 30 min) and tumor-to-non-tumor background ratios. MicroPET imaging for 16 corresponded to the results of the biodistribution. In addition, micro-PET imaging in the 4T1 murine breast cancer model, which expresses integrin only on the vasculature, confirmed explicit binding of 16 in the tumor vasculature [45].  [46]. The study proved the superiority of 14 over the other two. All three multimeric radioligands remained intact after 120 min incubation in mouse serum. Comparison of the cell-binding avidities in U87MG cells with the unlabeled dimer E[c(RGDfK)] 2 also favored 14 compared to 13 and 6 (E[c(RGDfK)] 2 > 14 > 13 > 6). While all radioligands showed a rapid and relatively high uptake in U87MG tumors with satisfactory tumor-to-background ratios, 14 had the highest tumor uptake and the lowest accumulation in the liver [46].
The effects of a symmetric beta-Glu linker in combination with a PEG 2 (3-(2-(2aminoethoxy)ethoxy)propanoic acid) connecting the β-glutamate's amino group with the radiolabeling domain were studied with the 18 F-labeled c(RGD) homodimeric peptide 2.68% ID/g) in A549 mice xenografts, while PET imaging in PC-3 and A549 tumor xenographs showed slightly higher values for PC-3 than A549 (60 min p.i.: 3.38 ± 0.44% ID/g vs. 2.85 ± 0.35% ID/g, respectively). Radioligand 17 was rapidly cleared from the blood by predominately renal excretion and had good stability in vitro and in vivo [47]. However, a direct pharmacokinetic comparison with similar radioligands with asymmetric Glu linkers e.g., 11 was not attempted.

Sugar Amino Acid Linkers on 18 F Labeled c(RGD) Multimers
Other linker variations like galactose-based sugar amino acids have been also tested without any significant improvement [48]. In particular, the introduction of a galactosebased sugar amino acid spacer (SAA = 7-amino-L-glyero-L-galacto-2,6-anhydro-7deoxyheptanamide) in [ 18 Figure 6) resulted in superior pharmacokinetics than their monomeric analogs, but when 18-19 were compared to [ 18 F]FP-PEG 3 -E[c(RGDfK)] 2 8, which has a PEG 3 linker, they all showed a similar pharmacokinetic profile [48]. However, the fact that the radiolabeling yield of 8 was substantially higher (80%) than the yield of 18 (52%) with its laborious four-step synthesis; makes the PEG 3 linkers a more useful and easily available choice.

The Effect of Linkers on the Stability and Production of 18 F Labeled Cyclic RGD Multimers
While [ 18 F]Alfatide 10 was already in clinical studies, Lang et al. investigated the stability problems presented during its production [49]. The by-products observed during the production process were due to the neighboring of the glutamic acid to the α-amine and thiurea groups, both participating in an intermolecular Lewis acid-catalyzed hydrolysis reaction under acidic conditions. Specifically, during the preparation of [ 68 Ga]Ga-NOTA-E(c(RGDyK)] 2 21 from 20 under acidic conditions and elevated temperature the nucleophilic addition of thiocarbonyl group of thiourea moiety at the neighboring carbonyl of the glutamic acid leads to the formation of a thiazolidinone ring in 20 and the release of c(RGDfK) (Figure 7) [49]. The introduction of a PEG 3 spacer in [ 18 F]Alfatide I 10 increased the distance between the Glu linker and the thiourea group, reducing this phenomenon. However, when the NOTA chelator was linked to the dimer with a carboxamide bond, it minimized the oxidation of the thiourea motif resulting in a more stable compound: In [ 18 F]Alfatide II, the tyrosine of the cyclic RGD part was replaced by a D-phenylalanine, which is not oxidized by heating [50], while the two PEG 4 groups introduced between each of the c(RGD) motifs and the Glu linker further increased the distance between the pharmacophores, enabling their simultaneous binding to the receptors [50]. It should be mentioned that 14 has also be radiolabeled using the kit formulation method [51][52][53][54].

64 Cu Labeled c(RGD) Multimers
The introduction of chelator groups i.e., NOTA, DOTA (1,4,7,10-tetraazadodecane-N,N,N,N-tetraacetic acid), facilitates the radiolabeling of c(RGD) multimers with various radionuclides; among them the positron-emitting PET radionuclide 64 Cu (half-life = 762 min, β + 17.9%, 0.64 MeV), which can be produced from a biomedical cyclotron with various methods [31]. Radiolabeling is achieved by complexation of 64 Figure 8) were used in biodistribution, micro-PET imaging, and whole-body autoradiography studies in athymic female nude mice with MDA-MB-435 breast carcinoma xenografts. Radiotracers 23 and 25 showed specific α v β 3 integrin tumor accumulation (i.e., 60 min p.i.: 3-4% ID/g), with higher retention than previously tested monomeric c(RGD) radioligands. However, activity accumulation of 25 in tumors was significantly higher compared to the D-Phe analog 23. Liver uptake of the D-Tyr derivative 25 was lower than the D-Phe 23 derivative at early time points, but the difference became marginal with time. Generally, 25 yielded better PET images than 23. The authors attributed this to the increased hydrophilicity of D-Tyr in 25 compared to D-Phe in 23 [55].  Figure 8) [56]. The c(RGD) tetramer 26 showed higher integrin-binding avidity (IC 50 = 16.6 ± 1.3 nM) in comparison to the corresponding dimeric analogue 23 (IC 50 = 48.4 ± 2.8 nM). Biodistribution and noninvasive microPET studies of tetramer 26 showed rapid, high and specific tumor uptake (U87MG) (30 min p.i.: 9.93 ± 1.05% ID/g; 24h p.i.: 4.56 ± 0.51% ID/g). Ligand 26 showed rapid blood clearance and predominantly renal excretion. The initial high tumor uptake and prolonged tumor retention of the tetramer were attributed to the high integrin avidity and the long blood circulation time, respectively. The latter was due to the increased molecular size. However, the enhanced tumor uptake of the tetramer 26 compared with the dimer 23 was accompanied by a similar increase in renal uptake, while tumor-to-kidney ratios did not increase significantly [56]. Consequently, the therapeutic/diagnostic advantage of a tetramer 26 over a dimer 23 might be modest.  (Figure 8). For 27, the distance between two distant c(RGD) pharmacophores is about 30 bond lengths, which is considered sufficient for simultaneous binding to adjacent integrin a v β 3 receptors, while for the octamer 28 the distance is increased to 40 bond lengths, enabling simultaneous binding with two or more receptors. The octamer 28 showed higher tumor uptake and longer tumor retention compared to the tetramer 27 in both tumor models tested i.e., U87MG, 30 min p.i. 28: 11.7 ± 0.7% ID/g, 27: 10.3 ± 1.6% ID/g; c-neu onco-mice 60 min p.i. 28: 8.9 ± 2.1% ID/g, 27: 4.4 ± 0.9% ID/g), while the integrin a v β 3 specificity of both tracers was confirmed by successful receptor-blocking experiments. However, a higher uptake and slow clearance in the kidneys was noted for 27, which was attributed to the integrin positivity of the kidneys, and to its larger molecular size [30]. The above studies indicate that the advantages observed for tetramers and octamers regarding receptor avidity and tumor uptake are counterweighted by their slow renal clearance, which eventually decreases their potential as diagnostic or therapeutic agents.
Hedhli  [57]. Ligands 29 and 30 are similar to other previously mentioned multimeric c(RGD) peptides bearing PEG groups in the linker region, but they differed in the Tz group, which was formed using CuAAC instead of forming a peptide bond. The binding kinetics against a v β 3 receptor were investigated using surface plasmon resonance. The association K on and dissociation K off constants of 29 and 30 and the commercially available monomeric c(RGDyK) were investigated in immobilized α V β 3 receptors. The unlabeled dimeric peptide NOTA-PEG 4 -E[PEG 2 -Tz-c(RGDfK)] 2 corresponding to 27, showed a binding avidity (K d = 0.19 pM), which was approximately 50-fold higher than the binding affinity of the monomeric NOTA-c(RGDfK) (K d = 9.6 pM), while 64 Cu labeled 29 and FITC labeled 30 showed marginally reduced binding avidity with K d values 1.5 pM and 8.6 pM, respectively. According to the authors, the receptor-bound dimeric c(RGD) peptides 29 and 30 dissociated from α V β 3 at a much slower rate (k off = 2.1 × 10 −6 s −1 ) compared to the typical adhesive proteins, e.g., the fibrinogen (k off = 9.  2 and also used a COCH 2 instead of the C(=S)NHC 6 H 4 CH 2 moiety connecting the chelator with the spacer-pharmacophore part of the radioligand (Figure 9). [58] The structural alteration marginally increased receptor avidity for 32 (IC 50 = 62 ± 6 nM), compared to 31 (IC 50 = 74 ± 3 nM). Dimers 31 and 32 were 64 Cu-radiolabeled with high yields and specific activity being >50 Ci/mmol. Biodistribution studies showed a very similar kinetic profile for the two radiotracers regarding U87MG tumor uptake and clearance with the PEG 4 -based dimer 31 showing slightly faster blood clearance and lower kidney values [58]. Liu 64 Cu under mild conditions, in combination with the linker AmBa (AmBa = 4-(Aminomethyl)benzoic acid).
Ligands were formed by reacting 4bromomethylbenzoic acid with one 33 or both 34 amine groups of Sar protruding the cage cavity ( Figure 10). After radiolabeling, the dimers were further evaluated in vitro and in vivo [59,60]. Radioligands 33 and 34 were proved very stable (intact tracer > 95% during HPLC analysis, 60 min after injection), both in vitro and in vivo and this was attributed to the cross bridged and cage-like configuration of the Sar chelator. Dimeric radioligand 33 showed higher tumor uptake than its respective monomeric, (20 h p.i. 33: 1.76 ± 0.38% ID/g, [ 64 Cu]Cu-AmBaSar-c(RGD): 0.65 ± 0.05% ID/g) and generally more favorable pharmacokinetics, due to the polyvalency effect [59]. However, the bi-functionalized Sar dimeric ligand 34 was proved superior to 33 both in vitro, by displaying higher avidity, i.e., 34: IC 50 = 6.0 ± 0.9 nM, 33: IC 50 = 10.0 ± 0.5 nM, and in vivo, by showing higher tumor uptake i.e., 1h p.i 33: 3.04 ± 0.25% ID/g, 34: 6.16 ± 0.88% ID/g. The difference according to the authors was due to the distance between the two pharmacophores, which is 5 bonds (Glu linker) in the case of 33, while it is 22 covalent bonds in 34 due to the intervention of the AmBaBaSar group ( Figure 10). Thus, in 33 the simultaneous binding on two integrin receptors was less likely compared to 34, where the increased distance and flexibility permit such binding interactions [59,60].

68 Ga-Labeled RGD Multimers
An interesting alternative for the 18 F and 64 Cu cyclotron-produced PET radionuclides is the generator-produced 68 Ga, which can be eluted from an inhouse 68 Ge/ 68 Ga generator ( 68 Ge, T 1/2 = 270.8 days) and has optimal physical characteristics (β + 89%, 1.92 MeV) for PET imaging. Furthermore, its half-life of 68 min, is compatible with the pharmacokinetics of many peptides [31]. The following section refers to the 68 Ga radiolabeled c(RGD) multimers in comparison to their 18 F and 64 Cu structural analogs, as well as to some 68 Ga radiolabeled c(RGD) multimers not previously mentioned. Siitonen (Figure 8), which was then used for PET imaging of Shank-associated RH domain-interacting protein (SHARPIN)-Regulated Integrin Activity in mice. [61] SHARPIN is a cytosolic protein that plays a key role in the activation of nuclear factor κ-light-chain enhancer of activated B cells and regulation of inflammation. Furthermore, SHARPIN controls integrin-dependent cell adhesion and migration in several normal and malignant cell types. Loss of SHARPIN correlates with increased integrin activity in mice. Increased integrin activity due to loss of SHARPIN protein would affect the uptake of the α v β 3 -selective 23, both in several tissue types and in the tumor microenvironment. PET imaging in vivo was evaluated in wildtype (wt) and SHARPIN-deficient mice (Sharpin cpdm , where cpdm 5 designates chronic proliferative dermatitis in mice) with and without melanoma tumor allografts. Sharpin cpdm mice with a spontaneous null mutation in the Sharpin gene and their wt littermates with or without B16-F10-luc melanoma tumors were studied using in vivo PET/CT imaging and ex vivo measurements with 23. The ex vivo uptake of 23 in the mouse skin and tumor was significantly higher in Sharpin cpdm mice than in wt mice, while B16-F10-luc tumors were detected 4 d after inoculation, without differences in volume or blood flow between the mouse strains. PET imaging even after 10 days of inoculation revealed significantly higher uptake in the tumors transplanted into Sharpin cpdm mice than in wt mice, while tumor vascularization was also increased in the Sharpin cpdm mice, indicating that SHARPIN may also have important regulatory roles in controlling the tumor microenvironment.  [62]. These values are in the same range as their DOTA analogs (32 IC 50 = 62 ± 6 nM, 31 IC 50 = 74 ± 3 nM), [58] indicating that the chelator did not significantly affect avidity. Instead, binding avidity was seriously affected by the applied spacer group. The benefits of using a PEG 4 or G 3 in the spacer were also observed in the biodistribution experiments, where it was shown that both radiotracers had higher (U87MG and MDA-MB-435) tumor uptake than the reference NOTA-E[cRGDfK)] 2 [63], while again, small differences were observed in comparison to their DOTA counterparts [58].
In a later study, Kaeopookum et [23]. All 68 Ga-labeled c(RGDfK) peptides displayed fair hydrophilicity (logD = −2.96 to −3.80), low protein binding, and were stable in phosphate buffered-saline and serum up to 2 h. In vitro receptor binding avidity and internalization assays in M21 cells showed specific uptake of all derivatives, which increased with the number of c(RGD) motifs i.e., 45 > 44 > 43. However, these in vitro avidity values did not exactly correspond with the in vivo (U87MG xenographs in mice) results, the monomer had the lowest tumor, but the dimer showed higher tumor uptake compared to the trimer i.e., 90 min p.i., 44: 8.19 ± 0.41% ID/g > 45: 3.98 ± 0.64% ID/g > 43: 2.73 ± 0.28% ID/g, while tumor uptake for the reference trimeric [ 68 Ga]Ga-FSC-(suc-c(RGDfK)) 3 was 4.95 ± 1.10% ID/g. The dimeric 44 also showed the best tumor-to-background ratios. All radiolabeled compounds showed fast blood clearance and high accumulation in kidneys. The authors correlated the high tumor uptake of dimeric 44 compared to the trimeric 45 to the lower density of α v β 3 -integrins on U87MG cells in comparison with the human melanoma M21 cells (preventing simultaneous binding) and this explanation was partly confirmed by the in vitro binding results [23].
The studies so far have shown that there is no significant advantage in using radiolabeled tetramers: E{E[c(RGD-X-K)] 2 } 2 (X = f or y) over their dimeric analogs: E[c(RGD-X-K)] 2 (X = f and y), regarding tumor to background (T/B) ratios or normal organ uptake. It seems that there is a limit on the benefits provided by the increasing pharmacophore multiplicity since over two c(RGD) pharmacophores increased the uptake in normal organs (kidneys, liver, lungs, and spleen). In addition, multiplicity increases production complexity and costs; two factors, which act prohibitively for the future development of multimers c(RGD) n with n > 4, as integrin α v β 3 -targeting radiotracers [20].

Clinically Applied RGD Multimers
A number of multimeric integrin-targeting c(RGD) radioligands have been applied since 2014 for prospective human studies, mainly focusing on oncological diseases (Table 1) [24,25,71]. Ligand [ 18 F]FP-PRGD 2 8 was the first dimeric c(RGD) tracer that was clinically applied in healthy volunteers. Mitra et al. showed that this tracer had good tolerance as well as favorable biodistribution and dosimetric characteristics [72], leading to its FDA approval as an exploratory investigative new drug (IND 104150) in human subjects.
Iagaru et al. reported that in a pilot evaluation of radioligand 8, eight women with newly diagnosed or recurrent breast cancer (BCa) underwent PET/CT with 8 and the commonly used radiopharmaceutical in PET imaging [ 18 F]FDG [73]. The radioligand 8 showed high and specific uptake in primary cancer, as well as in the metastatic lesions, with no safety issues reported or measured. The biodistribution of 8 in cancer patients was extensively investigated in a later study by Minamimoto et al. reporting high tracer accumulation in the bladder and kidneys, due to the tracer's predominant renal clearance, followed by the choroid plexus, spleen, salivary glands, thyroid, liver, pancreas, and bowel [74]. The above results in addition to the observed good tumor-to-background ratios suggested the ligand's 8 suitability for further clinical use. Furthermore, the dimeric c(RGD) 8 was found clinically superior to the monomers [ 18 [24,75], which was in accordance with the preclinical animal studies [48]. Additionally, the lack of significant correlation between tumor uptake for 8 and [ 18 F]FDG confirmed that the two PET tracers provide different molecular information [74]. In a clinical study investigating 10's feasibility for lung cancer detection, 26 patients with suspected lung cancer underwent PET/CT with this tracer before surgery and puncture biopsy. Standardized uptake values (SUVs) and tumor-to-blood ratios were measured, and diagnoses were also pathologically confirmed. Results showed that 10 was able to clearly identify all primary lesions with desirable image contrast (sensitivity = 100%, specificity = 44%, accuracy = 81%, positive predictive value [PPV] = 77% and negative predictive value [NPV] = 100%). The SUV for malignant lesions was significantly higher than that for hamartomas. However, it was difficult to clearly differentiate inflammatory or inflammatory pseudotumors from malignant lesions [76].
In another study of 10, 13 patients with non-small cell lung cancer (NSCLC) underwent PET/CT before surgery [77]. All malignant lymph nodes (LNs) were successfully visualized with a sensitivity of 100%, a specificity of 95%, and an accuracy of 95%. SUVmax, SUVmean, and SUV ratios in malignant LNs were significantly higher than in benign LNs. Similar results were observed in patients with adenocarcinoma and squamous cell carcinoma. Tracer 10 showed high sensitivity (83.9-100%), specificity (78.6-96.7%), and accuracy (81.7-96.9%) according to thresholds calculated from receiver operating characteristic curves.
Hitherto, the majority of clinical studies regarding c(RGD) multimers have been conducted for tracer [ 18 F]Alfatide II 14 [52][53][54]. The radiotracer was initially (2015) investigated in five healthy volunteers and nine patients with brain metastases (identified by MRI and/or CT) originating from various primary tumors i.e., lung, ovarian, gastric [54]. 14 was well tolerated without any serious tracer-related adverse events. The tracer showed rapid clearance from the blood pool and kidneys, while the organs with the highest absorbed dose were the kidneys and the spleen. Further, the detection rate of 14 was compared to other imaging modalities, in particular, CT and [ 18 F]FDG PET/CT. The comparison revealed that all 20 brain lesions were visualized by 14, while 13/20 lesions were visualized by CT and only 10/20 by [ 18 F]FDG PET/CT [54]. Of note, however, is the fact that [ 18 F]FDG is not an optimal imaging biomarker for brain tumors; instead, amino acid PET tracers, such as L-  [80,81]. 14 was also investigated for the diagnosis of bone cancer metastasis (2015) in 11 patients (n = 7 lung cancer, n = 2 cancer of unknown primary site, n = 1 gastric cancer, n = 1 urinary bladder cancer associated with gastric cancer) who underwent PET/CT with [ 18 F]FDG and 14 [53]. The final diagnosis of bone lesions was established based on the comprehensive assessment of all available data and clinical follow-up. Bone metastases were divided into four groups: osteolytic, osteoblastic, mixed, and bone marrow. PET/CT imaging using 14 detected the bone metastatic lesions with good contrast and higher sensitivity (positive rate of 92%) than [ 18 F]FDG (77%), especially in detecting osteoblastic (70% vs. 53%) and bone marrow metastatic lesions (98% vs. 77%). PET/CT sensitivity of 14 in osteolytic metastasis was 100%, while for [ 18 F]FDG was 90% [53]. Moreover, 14 was compared to [ 18 F]FDG for detecting BCa, in a cohort of 44 female patients [52]. PET/CT image analysis was based on visual and semi-quantitative analysis (SUV max , SUV mean ). In total, 42 BCa lesions and 11 benign breast lesions were confirmed by histopathology.  [51].
The dimeric structure 12 (Figure 3) was investigated in comparison to [ 18 F]FDG in 91 LCa patients [78]. Tracer 12 was well tolerated, while it rapidly cleared from the blood pool, mainly through the urinary system. The SUVs for proven malignancies were significantly higher than benign lesions with 12 showing a sensitivity of 84%, specificity of 91%, and accuracy of 86%, exhibiting a diagnostic value comparable to [ 18 F]FDG for LCa detection. Moreover, 12 was more specific than [ 18 F]FDG PET/CT in assessing LN metastasis, with PPV of 90% (30% for [ 18 F]FDG) and NPV and 94% (91% for [ 18 F]FDG), respectively [78].
Although being structurally different, all the clinically investigated RGD peptides, including monomers and dimers, depict very similar in vivo pharmacokinetic properties [24]. Regarding 8, and 14, the two most clinically studied dimers, they were able to detect integrin-positive tumors with good imaging contrast, while showing comparable imaging properties and pharmacokinetics, and while exhibiting high sensitivity (for primary lesions 83.3-100% and for metastatic lesions 70-100%) and specificity for tumor detection and staging. Besides the urogenital system, moderate to prominent off-target uptake was observed for both in the liver, which can be an issue for detecting hepatic tumors or metastases [43,52,82].
Until recently, only a small number of clinical investigations of dimeric RGD peptides have been reported, and the sensitivity/specificity between dimers and monomers has not been compared in the same patients. Thus, additional evaluation with large cohorts is needed to determine if the multimeric strategy provides higher sensitivity and specificity for tumor detection and staging than the monomeric RGD compounds and the golden standard [ 18 F]FDG [24].

Discussion
Multimeric PET radioligands consist of identical binding motifs (pharmacophores) connected to a single backbone (linker) attached to a group, which can be labeled with a positron-emitting radionuclide suitable for PET molecular imaging (radiolabeled domain). Among the various PET multimeric radioligands investigated for targets like integrin α v β 3 , PSMA, GRPr, VEGFR, and EGFR-TKI, the ones targeting integrin α v β 3 are the most studied and the only category which has reached the clinical stage of development (Table 1). Multimeric c(RGD) analogs are a fine example, providing proof that multimerization can improve ligands characteristics like receptor avidity and tumor uptake.
Several factors regarding the design of c(RGD) multimeric radioligands should be taken under consideration. One of the most important factors is the length and flexibility of the linker (1) connecting the chelator group with the multimeric scaffold and (2) connecting the various pharmacophores. Regarding the first case of linker (1), several examples of 18 F-labeled compounds, 7-10 ( Figure 3) have shown that the introduction of PEG group in-between the pharmacophores and the labeling site not only improves the overall radiolabeling yield but also reduces the renal uptake and increases tumor-targeting efficacy [36]. Specifically, the introduction of PEG 3 minimized the instability factors observed in the acidic and high-temperature radiolabeling conditions of 10, due to the thiourea linkage of the labeling site with the α-amine of the Glu linker [50]. Considering the second case (2), the length of the linker defines the distance between the two pharmacophores. Cyclic RGD dimeric peptides, where the Glu linker was connected with each c(RGD) using additional groups like PEG 4 or G 3 e.g., 14-16 ( Figure 4) or 29-32 ( Figure 9) have shown better results, with respect to tumor uptake, than other ligands with shorter linkers [20]. In particular, for c(RGD) dimers, it seems that this distance has to be in the range of 30-38 bonds like in 26, 27 [30] or the SAR conjugate 33 [57,58] for achieving bivalency, and eventually leading to higher integrin α v β 3 binding avidity [20].
Several pharmacodynamic models have been proposed as an explanation for the observed improvements in binding avidity, reduced receptor off-rate, which eventually result in high tumor uptake. One of the models suggests simultaneous binding of the ligand with two receptors on the cell surface; this can be accomplished with the utilization of extremely long spacers, which cover the distance between two receptors on the cell surface. However, extremely long spacers do not always prove to be advantageous for the pharmacodynamic or pharmacokinetic ligand characteristics, because they may prevent other actions such as the internalization of the ligand or may worsen its pharmacokinetic properties, resulting in reduced tumor uptake. Another model describing the improved effects observed for multimeric ligands is based on the improved statistical effect, in this case, the ligand binds to one receptor, but its apparent local concentration of the ligand in proximity to the receptors is increased. This seems to be the most likely explanation for ligands with short linkers [1].
Another factor regarding the design of c(RGD) multimers is the number of c(RGD) pharmacophores included. Increasing the number of pharmacophores had a positive effect on ligand avidity and cell binding in vitro. Nevertheless, this effect did not always correspond with similar advantages in vivo, since increasing peptide multiplicity, in many cases resulted in a parallel increase of ligand uptake in normal organs. Thus, the benefits of multiplicity seem to have limits. Among the multimers summarized in this article, the dimers seem the most successful cases and that is also the reason they have further advanced in the clinic, over the other tracers.
Several examples of c(RGD) multimers for PET imaging of integrin α v β 3 have been studied so far, which have been labeled with various PET radionuclides i.e., 18 F, 68 Ga, and 64 Cu (for a list of compounds included in the article along with the α v β 3 expressing cells/tumor models tested refer to Table 2). The 18 F analogs and specifically dimeric [ 18 F]Fc(RGD) 2 ligands have been the most successful and have already reached the stage of clinical development for various applications utilizing integrin α v β 3 imaging like cancer (ligands 8, 10, and 14). However, the NOTA analogs, which are labeled using [ 18 F]AlF, a much easier, faster, and high yielding production procedure (40 min and 20 min for kit radiolabeling, yield 42%, radiochemical purity > 95%), have significant advantages for their future clinical application, especially after the introduction of kit formulations. Among the two compounds developed so far: [ 18 F]Alfatide I 10 and [ 18 F]Alfatide II 14; the second is considered more stable regarding a possible intermolecular Lewis acid-catalyzed hydrolysis during its production. Besides 18 F, another popular positron-emitting radioisotope is 68 Ga, which can be produced by a 68 Ge/ 68 Ga generator. So far one clinical study has been published for each of the 68 Ga labeled dimeric structures, 12 and 15. Tracer 12 showed similar pharmacokinetics with 10, which can also be labeled with 68 Ga, and is of alike chemical structure to 14. Consequently, additional results of clinical studies regarding [ 68 Ga]Ga-c(RGD) 2 dimers i.e., 12 or 15 are expected to be published in the near future.         Finally, regarding the advantages of multimeric c(RGD) analogs in clinical studies, according to the limited data generated so far, the dimeric structures seem to have an advantage over their monomeric analogs. Even so, additional evaluation is needed with large cohorts of patients in order to determine if the multimeric strategy provides higher sensitivity and specificity for tumor detection and staging over the monomeric RGD compounds and the golden standard [ 18 F]FDG.

Concluding Remarks
This article describes multimeric c(RGD) ligands targeting integrin α v β 3 receptors for PET molecular imaging of tumors. Multimerization is generally advantageous and multimeric c(RGD) radioligands have increased binding avidity against α v β 3 receptors and are more effective in tumor targeting compared to monomeric c(RDG) radioligands. However, when the number of c(RGD) pharmacophores was increased above 2, it did not enhance the pharmacokinetic properties of the ligand in vivo, despite the benefits of multimerization observed in vitro. The length and flexibility of the linker connecting the c(RGD) pharmacophores and the linker connecting the multimerization scaffold with the chelator group have a significant role in the biological activity of the multimeric c(RGD) ligands. Clinical studies are expected to bring forward valuable information regarding the application benefits of multimeric c(RGD) ligands.