Design, Synthesis, In Vitro and In Vivo Evaluation of Heterobivalent SiFAlin-Modified Peptidic Radioligands Targeting Both Integrin αvβ3 and the MC1 Receptor—Suitable for the Specific Visualization of Melanomas?

Combining two peptides addressing two different receptors to a heterobivalent peptidic ligand (HBPL) is thought to enable an improved tumor-targeting sensitivity and thus tumor visualization, compared to monovalent peptide ligands. In the case of melanoma, the Melanocortin-1 receptor (MC1R), which is stably overexpressed in the majority of primary malignant melanomas, and integrin αvβ3, which is involved in lymph node metastasis and therefore has an important role in the transition from local to metastatic disease, are important target receptors. Thus, if a radiolabeled HBPL could be developed that was able to bind to both receptor types, the early diagnosis and correct staging of the disease would be significantly increased. Here, we report on the design, synthesis, radiolabeling and in vitro and in vivo testing of different SiFAlin-modified HBPLs (SiFA = silicon fluoride acceptor), consisting of an MC1R-targeting (GG-Nle-c(DHfRWK)) and an integrin αvβ3-affine peptide (c(RGDfK)), being connected by a symmetrically branching framework including linkers of differing length and composition. Kit-like 18F-radiolabeling of the HBPLs 1–6 provided the labeled products [18F]1–[18F]6 in radiochemical yields of 27–50%, radiochemical purities of ≥95% and non-optimized molar activities of 17–51 GBq/μmol within short preparation times of 25 min. Besides the evaluation of radiotracers regarding logD(7.4) and stability in human serum, the receptor affinities of the HBPLs were investigated in vitro on cell lines overexpressing integrin αvβ3 (U87MG cells) or the MC1R (B16F10). Based on these results, the most promising compounds [18F]2, showing the highest affinity to both target receptors (IC50 (B16F10) = 0.99 ± 0.11 nM, IC50 (U87MG) = 1300 ± 288 nM), and [18F]4, exhibiting the highest hydrophilicity (logD(7.4) = −1.39 ± 0.03), were further investigated in vivo and ex vivo in a xenograft mouse model bearing both tumors. For both HBPLs, clear visualization of B16F10, as well as U87MG tumors, was feasible. Blocking studies using the respective monospecific peptides demonstrated both peptide binders of the HBPLs contributing to tumor uptake. Despite the somewhat lower target receptor affinities (IC50 (B16F10) = 6.00 ± 0.47 nM and IC50 (U87MG) = 2034 ± 323 nM) of [18F]4, the tracer showed higher absolute tumor uptakes ([18F]4: 2.58 ± 0.86% ID/g in B16F10 tumors and 3.92 ± 1.31% ID/g in U87MG tumors; [18F]2: 2.32 ± 0.49% ID/g in B16F10 tumors and 2.33 ± 0.46% ID/g in U87MG tumors) as well as higher tumor-to-background ratios than [18F]2. Thus, [18F]4 demonstrates to be a highly potent radiotracer for the sensitive and bispecific imaging of malignant melanoma by PET/CT imaging and impressively illustrates the suitability of the underlying concept to develop heterobivalent integrin αvβ3- and MC1R-bispecific radioligands for the sensitive and specific imaging of malignant melanoma by PET/CT.


Introduction
With a global incidence increasing over the last decades and being among the tumor types with the most increasing prevalence in Europe, malignant melanoma (MM) is the most aggressive type of skin cancer. The probability of developing the disease is increasing for people with a large number of melanocytic nevi, a fair skin type and genetic predisposition [1][2][3]. Repeated exposure to strong UV (ultraviolet) radiation through recurrent intense sun exposure is the most important environmental risk factor [4]. In most cases, an early diagnosis enables a complete surgical removal and thus the patient to be cured. However, early detection is often not possible since the disease has no particular symptoms, and the tumors can rapidly progress from the fully encapsulated stage to infiltrative growth. In the case of basal membrane penetration, the tumor has access to the blood and lymph vessels, and metastases can be formed in organs or lymph nodes [5]. Since a cure is rarely possible when metastasis has already occurred, an early, very sensitive and specific diagnosis of the disease is of the highest importance. Moreover, the correct staging of the disease is critical, as only, in this case, can an appropriate therapy, having the potential to cure the patient, be chosen.
However, primary diagnosis using positron emission tomography (PET), which has the highest sensitivity compared to other whole-body imaging techniques such as computed tomography (CT) or magnetic resonance imaging (MRI), is often not suitable to correctly identify MM lesions. One drawback of the commonly used radiotracer [ 18 F]FDG (2-[ 18 F]fluoro-2-deoxyglucose) is its accumulation in inflamed tissues, giving false-positive results. Furthermore, the detection of slowly growing lesions is often difficult as well, resulting in possible false-negative imaging results [6]. Since the tumor visualization sensitivity and specificity using [ 18 F]FDG can be low, an early and correct diagnosis and staging are often not possible. An alternative to unspecific, metabolically driven imaging is addressing the tumor by a tumor-specific radiotracer. For this purpose, receptors that are overexpressed in the tumor cell surface are especially useful. In the case of MM, the MC1R is best suited, as this receptor type is overexpressed in about 80% of MM primaries [7,8] and thus is a highly important target structure for MM-specific imaging. However, not all lesions express the MC1R, resulting in an incomplete visualization of the tumor load and thus false staging of the disease. In order to improve the diagnostic imaging of MM and enable an adequate, early and sensitive diagnosis and correct staging, a reliable and sensitive imaging method for MM is needed. Therefore, the development of target-specific accumulating agents that are able to address more than just the MC1R is mandatory.
Such agents should be based on radiolabeled peptides being able to bind with high affinity and specificity to surface receptors overexpressed by malignant cells and thus, enable the distinction between benign and malignant tissue. Ideally, radiolabeled peptides exhibit favorable tumor-to-background ratios, due to their tumor-specific accumulation, and thus produce images of high quality. Furthermore, peptides exhibit low toxicity and immunogenicity, are easily synthetically accessible and can be chemically modified at defined sites. Their pharmacokinetics prove to be very advantageous due to rapid tissue penetration, target accumulation and elimination from non-target tissues [9,10]. Therefore, numerous radiolabeled peptides have been developed for both the diagnosis and therapy of malignancies over the last decades [11,12].
Heterobivalent peptidic ligands (HBPLs), consisting of a radionuclide and two different peptides, each addressing its respective target receptor, have the advantage of a higher target avidity compared to monovalent peptide ligands by being able to bind simultaneously or independently to different target receptors on the tumor surface, resulting in stronger binding to the target cell [13]. Furthermore, HBPLs usually exhibit higher metabolic stability than their respective monomers against peptidases, due to their higher molecular weight and introduction of artificial structural elements [14]. The prerequisite for an HBPL with high tumor visualization potential is at least a moderate binding affinity of each of the included peptides to their target receptors. Ideally, both receptor types should be present in high density to achieve a concomitant binding of both peptide binders of the HBPL; however, the presence of only one target receptor is sufficient to achieve a high tumor uptake [9,[15][16][17], resulting in an overall improved imaging sensitivity.
In contrast to HBPLs, monovalent peptides, being able to address only one receptor type and thus only visualize tumors that overexpress this particular receptor, can result in limited tumor visualization sensitivity, as tumor cells can overexpress different receptor types. In such cases of inhomogeneous receptor expression, which can further be caused by tumor dedifferentiation, metastasis or triggered by therapy, the target receptor for the monospecific binder can be absent or present in insufficient density [18][19][20]. This results in an insufficient sensitivity of the peptides' tumor delineation ( Figure 1A). In contrast, HBPLs have the advantage of binding to more than one receptor type and thus exhibit a high tumor visualization efficiency ( Figure 1B) [9,20]. In the case of (A), no binding is possible since the respective target receptor is only expressed to a low extent. In the case of (B), the HBPL can bind since at least one of the target receptors is expressed on the tumor surface.
For the development of HBPLs, some requirements have to be fulfilled. The peptides have to be modified as little as possible in their chemical structure to preserve their binding affinities to their corresponding receptors. In particular, the pharmacophoric site has to remain unchanged. Furthermore, it is important to determine which receptor types are overexpressed in a tumor entity and thus can be addressed by the radioligand to be developed [9,20]. For this purpose, many studies have been performed within recent years regarding the available receptor types on different human malignancies [21]. The results obtained serve as a guideline for the choice of peptidic receptor ligands, yielding potent tumor-targeting HBPLs with highly sensitive visualization properties.
For MM, the MC1R represents one especially useful target structure for the specific imaging of the disease (vide supra). Another receptor type that is of high potential for MM imaging is integrin α v β 3 , as it was shown that this receptor is overexpressed in the blood vessels of many human tumors [22][23][24]. Further studies revealed the involvement of integrin α v β 3 in the progression of the disease and in the change of tumor growth from radial to vertical (thus infiltrative) growth [25][26][27][28][29][30]. Therefore, integrin α v β 3 , although overexpressed in all neo-angiogenetic processes, is also an important marker protein for MM targeting.
Thus, HBPLs based on MC1R-and integrin α v β 3 -affine peptides would be most promising for visualizing MM during all stages of the disease, enabling a highly sensitive and especially correct assessment of the extent of the disease. This is of crucial importance for choosing the optimal therapy approach, adapted to the extent of the disease: an encapsulated tumor can be treated differently than an infiltratively growing or already metastatic tumor. A high sensitivity to tumor imaging, surely identifying all tumor mass, is thus the prerequisite for the choice of the best-suited therapy option.
As no HBPLs based on these peptidic ligands have been described so far, we intended to assess the general feasibility of this concept and to develop different HBPLs, consisting of the mentioned peptidic binders, a SiFAlin-moiety (for efficient radiolabeling of the HBPL with the positron-emitting nuclide 18 F) and a varying molecular design. The molecular scaffold for the HBPLs was based on a symmetrical branching unit exhibiting linkers of different lengths and compositions so as to be able to systematically determine the influence of the used linker type and length on the biological parameters of the resulting HBPLs. The developed agents were labeled with 18 F and evaluated in vitro regarding their lipophilicity, stability in human serum and especially their binding affinity to the respective target receptors. Finally, the most promising 18 F-labeled derivatives were evaluated in vivo, in terms of their tumor visualization potential, in an appropriate preclinical tumor model using PET/CT imaging and ex vivo biodistribution experiments.

General Considerations for the Design of the Heterobivalent SiFAlin-Modified Peptidic Ligands
The molecular design of the target compounds ( Figure 2) included two different peptides, each addressing specifically one of the two target receptors-c(RGDfK) for integrin α v β 3 and GG-Nle-c(DHfRWK) for MC1R binding-and was based on the following conditions: (i) The HBPLs should contain a SiFAlin-moiety exhibiting a permanent positive charge. With this SiFAlin building block, the radionuclide 18 F can be efficiently introduced in one step [35]; (ii) the required molecular building blocks should be connected by a small symmetrically branched framework resulting in homogeneous compounds [9,36]; (iii) a lysine spacer should be introduced between the SiFAlin-moiety and the branched framework to achieve a spatial distance between the SiFAlin and the peptides, preventing interference with the peptide-receptor interaction, and to obtain the products in higher radiochemical yields [9,20,37,38]; (iv) as much as possible, the syntheses should be carried out on a solid support to facilitate the assembly of the rather complex target molecules; (v) different linker structures should be introduced between the peptides and the branching unit to systematically determine the optimal distance between both peptides. An optimal distance between the peptide binders enables the binding of each peptide to the respective receptor while remaining not interfered with by the second peptide and, at the same time, does not result in a high entropy, limiting the benefits of peptide heterodimerization [15,[39][40][41][42][43]. Since the synthetic effort for the SiFAlin-linked framework is higher than that of the peptides, the linkers should be introduced as bis-NHS (N-hydroxysuccinimide) active esters on the peptidic side by derivatizing the N ε -amine of lysine of c(RGDfK) and the N α -amine of glycine of GG-Nle-c(DHfRWK), or the N α -amine of glutamic acid for the EGEGE peptides. Regarding the order of peptide-to-framework conjugations, the smaller c(RGDfK)-based peptides should be reacted first, followed by the bulkier GG-Nle-c(DHfRWK) peptides, to achieve higher product yields. Depiction of the structures of the target HBPLs 1-6 consisting of: a SiFAlin-moiety (blue); a short lysine linker (black); the symmetrically branching framework (pink); linkers Y of different lengths and compositions (PEG 1 , PEG 3 , PEG 5 , PEG 8 , DIG, Ox-EGEGE; PEG = polyethylene glycol; DIG = diglycolic acid; Ox = oxalic acid); the MC1R-and integrin α v β 3 -affine peptides GG-Nle-c(DHfRWK) (green) and c(RGDfK) (orange).

Synthesis of the Heterobivalent SiFAlin-Modified Peptidic Ligands
For the assembly of the SiFAlin-modified HBPLs 1-6, the monomeric peptides were synthesized according to standard Fmoc-based solid-phase peptide synthesis (SPPS) protocols. The c(RGDfK)-peptide was synthesized according to a known procedure [44] and was obtained in an overall yield of 83%. For the synthesis of the peptide GG-Nle-c(DHfRWK) (7) (Scheme 1A), all amino acids were coupled on a rink amide resin. After deprotection of the acid-labile protecting groups (PG)-Mtt and O-2-Ph i Pr-under mildly acidic conditions, the cyclization, deprotection and cleavage from the resin were performed. By optimizing the reaction conditions, peptide 7 was isolated in an overall yield of 42%. For the synthesis of the charged HBPL 6, further amino acids had to be added to the monovalent peptides. For c(RGDfK)-EGEGE (8) (Scheme 1B), c(RGDfK) was first synthesized according to a known procedure [45] and was then modified with the glutamic acids and glycines at the N ε -amine of lysine (still on a solid support) before the deprotection and cleavage from the resin were carried out. Peptide 8 was obtained in an overall yield of 41%. EGEGE-GG-Nle-c(DHfRWK) (9) (Scheme 1C) was prepared in a different way than GG-Nle-c(DHfRWK)-peptide. First, only the first six amino acids were conjugated to the resin. Afterward, the cyclization was performed, and only then the following conjugation of the remaining amino acids followed by the cleavage from the resin was performed. Peptide 9 was obtained in an overall yield of 43% following this route (for analytical data, see Supplementary Materials Figures S1-S3).
After successfully establishing the synthesis of the monovalent peptides c(RGDfK) and 7-9, modification of the peptides with the different linkers (PEG 1 , PEG 3 , PEG 5 , PEG 8 and DIG) was performed as follows.
For the evaluation of the optimal conditions to obtain the peptide-linker conjugates 10-21 (Scheme 2), which can be used for heterodimer synthesis, different solvent/base systems were tested during the reactions of the peptide c(RGDfK) and 7-9 with the respective bis-NHS esters of the linkers introduced to obtain the conjugates 10-21. The best results in terms of isolated yields were found for DIPEA in DMF. Besides the target NHS-PEG n peptides 10-13 and 16-19, small amounts of hydrolyzed compound and homodimer were also isolated. However, in the case of the DIG-and Ox-linker, only the hydrolyzed carboxylic acids 14, 15, 20 and 21 could be isolated, although the reason for this is not obvious. For further reactions of these agents with the framework structure, in order to obtain the target HBPLs, these free carboxylic acid-comprising peptides had to be pre-activated with a suitable activation reagent (for analytical data, see Supplementary Materials Figures S4-S15). To obtain the SiFAlin building block 28 (Scheme 3), the acetal 26 was synthesized following a published procedure [46][47][48] with some modifications. First, the hydroxyl function of the 4-bromobenzyl alcohol was protected with TBDMS-Cl (tert-butyldimethylsilyl chloride) to produce 22, then the SiFA unit was introduced by an in-situ-preceding halogenmetal exchange with subsequent transmetalation to produce 23. After the acidic deprotection of the TBDMS-PG, the resulting alcohol 24 was transferred to the bromide 25 by an Appel reaction. Amination of 25 with 4,4-diethoxy-N,N-dimethylbutan-1-amine led to the desired acetal 26. After the acidic deprotection of 26, the resulting aldehyde 27 was oxidized using KMnO 4 (potassium permanganate) to obtain the desired SiFAlin building block 28. Next, the SiFAlin-comprising symmetrically branching building block 29, being the basis for the following peptide conjugation and thus peptide heterodimer synthesis, was prepared. The branching unit was synthesized on a solid support using the same standard protocols followed for peptide synthesis (Scheme 4).  Finally, the synthesized building blocks 10-21 and 29 were assembled into the heterobivalent target agents 1-6. For this purpose, 29 was first reacted with the c(RGDfK) derivatives 10-15, which produced the monovalent intermediates 30-35. These were further reacted with the GG-Nle-(DHfRWK) derivatives 16-21 into the final products 1-6 (Scheme 5), as this order gave better results (in terms of achievable isolated yields) than did first conjugating the structurally more demanding peptides 16-21 followed by the smaller ones (10-15). The conjugation of the peptide-linker conjugates 10-13 and 16-19 was conducted analogously to the synthesis of the peptide-linker conjugates by directly reacting the starting materials in DMF using DIPEA as a base. For the conjugation of 14, 15, 20 and 21, which were obtained as free acids instead of the respective NHS esters, the linker-modified peptides had to be activated before conjugation using PyBOP as the coupling agent.
The intermediates 30-35 were obtained in yields of 41-67% (for analytical data, see Supplementary Materials Figures S47-S52). The isolated yields of the final products 1-6 varied depending on the reaction pathway. Whereas during the conjugation reactions of the NHS-modified peptides 16-19 to 30-33 relatively high yields of 49-77% could be achieved, the yields during the reactions of 20 and 21 to 34 and 35 were considerably lower at 24% and 21%, respectively. This might be attributable to the additional activation step being required for the free acids 20 and 21 or the shorter linker structure, resulting in a steric hindrance of the conjugation reaction.
For the HBPLs 1-6, 19 F NMR spectra were recorded (for analytical data, see Supplementary Materials Figures S53-S64) along with standard HR mass spectrometry to verify that all agents contained the required fluorine atom in the SiFAlin building block, instead of having formed the hydrolyzed hydroxy-comprising species. All spectra showed a signal with a chemical shift between δ = −175-−177 ppm, which indicates the presence of an intact SiFAlin-moiety [49,50].

18 F-Radiolabeling of 1-6 and Determination of Lipophilicity and Stability of [ 18 F]1-[ 18 F]6 in Human Serum
In the following procedures, the HBPLs 1-6 were radiolabeled with [ 18 F]fluoride as instructed by previously published protocols on other SiFAlin-modified peptides [35,51]. Briefly, [ 18 F]fluoride was dried using the "Munich method" [52] over a QMA carbonate Sep-Pak SPE light cartridge, instead of applying an azeotropic drying, and the activity was eluted from the cartridge using a freshly prepared solution of K 222 (Kryptofix222) and KOH (potassium hydroxide) in acetonitrile (MeCN). After optimizing the reaction and elution conditions, the pH of the obtained solution was adapted with oxalic acid, preventing potential basic hydrolysis of the SiFAlin-moiety. To this mixture, small amounts of the respective precursor molecules 1-6 at 25 nmol were added and incubated at ambient temperature for 10 min. Afterward, the radiolabeled products [ 18  Since high lipophilicity of peptidic radiotracers can lead to a high plasma protein binding, resulting in unspecific organ and high liver uptakes, thus negatively impact tumor visualization [46,53,54], the lipophilicity of the HBPLs was determined to get an approximate estimation of the in vivo biodistribution behavior of the radioligands. Therefore, the log D(7.4) values of the SiFAlin-modified HBPLs [ 18 F]1-[ 18 F]6 were determined via their distribution coefficient between n-octanol and phosphate buffer at pH 7.4. The results are also summarized in Table 1. In these experiments, [ 18 F]4 (log D(7.4) = −1.39 ± 0.03) exhibited the highest hydrophilicity of the PEG n -linker based HBPLs. Furthermore, it is clear that the introduction of negative charges led to the expected substantially increased hydrophilicity of [ 18 F]6 (log D(7.4) = −1.52 ± 0.01). Overall, all 18 F-labeled HBPLs demonstrated hydrophilicity suitable for further in vivo application.
In addition to the investigation of the radiotracers' lipophilicity, their stability in human serum was evaluated in order to determine possible stability issues of the newly developed agents. For this purpose, the respective 18 F-labeled HBPLs were incubated in human serum at 37 • C for 120 min. The results of these experiments are summarized in Table 1, and the corresponding radio-HPLC chromatograms for [ 18 F]2 as representative examples for all compounds studied are depicted in Figure 3A (see Supplementary Materials, Figure S91 for the results obtained for the other radioligands). In Figure 3B, the portions of intact radiotracer over the course of the stability experiments are depicted. As the results indicated the radiolabeled HBPLs [ 18 F]1-[ 18 F]6 to be sufficiently stable in vitro (81-87% intact radioligand after 120 min), all radiotracers were found to be suitable for in vivo imaging via PET/CT.

In Vitro Evaluation of 1-6 Regarding Their Binding Affinities to the Respective Target Receptors
As in vitro receptor affinities represent an important parameter for the in vivo tumor uptake of radiotracers, the binding affinities of 1-6 were determined to both target receptorsintegrin α v β 3 and the MC1R-in competitive displacement assays. During these evaluations on MC1R-positive B16F10 cells [55] and integrin α v β 3 -positive U87MG cells [56], α-MSH   (38) and c(RGDyK) (39), which were used as monospecific reference peptides during the competitive displacement studies.
The resulting binding curves and determined IC 50 values are depicted in Figure 5 and summarized in Table 2.  To ensure that the binding of each peptide binder to its target receptor was unaffected by the other respective peptide of the HBPL, the monomeric peptides c(RGDfK) and GG-Nle-c(DHfRWK), which cannot bind to the receptors MC1R and integrin α v β 3 , respectively, were also examined under the same conditions, showing-as expected-no receptorspecific binding (see Supplementary Materials Figure S92 for details).
Considering the receptor affinity data with respect to the MC1R, none of the developed agents was as potent as NDP (IC 50 of 0.17 ± 0.04 nM), which is however not surprising as NDP is a superpotent synthetic analog of the endogenous ligand α-MSH, thus exhibiting a high potency. However, 3 (IC 50 of 3.44 ± 0.09 nM), 5 (IC 50 of 2.05 ± 0.35 nM), 1 (IC 50 of 1.74 ± 0.25 nM) and 2 (IC 50 of 0.99 ± 0.11 nM) showed considerably higher affinities than the physiological reference α-MSH (IC 50 of 3.75 ± 0.61 nM). In comparison, 4 (IC 50 of 6.00 ± 0.47 nM) showed a decreased affinity and 6, comprising the charged linker (IC 50 of 4.18 ± 0.32 nM), exhibited a fourfold higher IC 50 value compared to its uncharged counterpart, 2 (same distance between both peptide binders but differing linker composition), and thus considerably decreased affinity.
The corresponding experiments on the integrin α v β 3 -positive U87MG cells revealed that neither 5 (IC 50 of 5895 ± 722 nM), 1 (IC 50 of 2881 of 757 nM), 4 (IC 50 of 2034 of 323 nM) nor 3 (IC 50 of 1911 ± 70 nM) were as potent as the highly affine reference peptide c(RGDyK) (IC 50 of 427 ± 37 nM; in accordance with former values obtained on these cells [57]), whereas at least compound 2, showing an IC 50 value of 1300 ± 288 nM, demonstrated a higher integrin affinity than the other reference c(RGDfC) (IC 50 of 1493 ± 210 nM; also in accordance with literature data [44]). For HBPL 6, an IC 50 value towards α v β 3 could not be determined in the same concentration range of the other agents studied but showed a substantially reduced affinity to the target receptor, compared to 1-5. This observed negative influence of anionic charges on the resulting receptor affinities was also described in other studies [58] and could be confirmed here.
From the obtained results, it can be concluded that the introduction of a negatively charged linker impairs binding to the MC1R, as well as to integrin α v β 3 , and thus limits the usefulness of the approach. Within the line of the other linkers used, a similar trend can be observed on both cell lines with regard to the linker length used. In both cell lines and thus for both receptor types, the affinities increased with increasing linker length up to the PEG 3 -unit but then decreased with further increasing linker length, thus giving the best results for the PEG 3 -modified analog 2 on both receptor types.

Evaluation of the In Vivo Pharmacokinetics and Ex Vivo Biodistribution of [ 18 F]2 and [ 18 F]4
For the evaluation of the in vivo pharmacokinetics, the two most promising HBPLs-[ 18 F]2 with the highest affinity to both target receptors and [ 18 F]4 with the highest hydrophilicity and still reasonable binding affinities-were selected. For PET/CT imaging, 6-week-old male nude mice (Balb/cAnNRj-Foxn1 nu/nu ) were subcutaneously injected with 5 × 10 5 B16F10 cells into the right flank and 2.0-2.5 × 10 6 U87MG cells into the left flank to generate the respective receptor-positive tumors. When the tumors reached a sufficient size for imaging, each mouse was administered 4.15 ± 2.28 MBq of [ 18 F]2 or 3.95 ± 2.06 MBq of [ 18 F]4 via the lateral tail vein under isoflurane anesthesia. To determine the receptor specificity of both peptide parts of the labeled HBPLs and their relative contribution to overall tumor uptake, blocking experiments were also performed. For these, the respective radiotracer was coinjected with the corresponding blocking substance-20 µg NDP, 200 µg c(RGDyK) or both for double blocking-via the lateral tail vein. After i.v. injection of the tracers, a dynamic PET scan, followed by a CT scan, was performed. The resulting PET/CT images and time-activity curves (TACs) are depicted in Figures 6-8. After completion of the diagnostic scans, the mice were sacrificed, their organs (blood, spleen, liver, kidney, pancreas, lung, heart, brain, bone, muscle, tail, tumors, stomach, colon and small intestine) were collected and measured in a γ-counter for ex vivo biodistribution (see Supplementary  Materials Table S1 for detailed results).   From the PET/CT scans (Figures 6 and 7 From the PET/CT data depicted in Figure 7, the visual impression obtained is that [ 18 F]4 accumulates only to a low extent in B16F10 tumors. However, ex vivo biodistribution data confirm the data of the TACs and the uptake to be comparatively high as in the case  Figure 7, should be due to the fact that the large tumor already had partially necrotic areas, showing no tracer uptake anymore. This assumption is supported by the literature [59,60]. In the blocking experiments, the receptor specificity of the tracers could be demonstrated, as blocking with NDP and c(RGDfK) resulted in a considerable decrease of the respective tumor uptakes of both tracers in B16F10 and U87MG tumors. Coinjection with NDP substantially reduced the accumulation in the MC1R-positive B16F10 tumors ([ 18 F]2: reduction from 2.32 ± 0.49% to 1.83 ± 0.24% ID/g (change ns, p = 0.19) and [ 18 F]4: reduction from 2.58 ± 0.86% to 1.33 ± 0.27% ID/g (change ns, p = 0.07)). Corresponding results were also found for c(RGDyK) blocking, where U87MG tumor uptakes were reduced from 2.33 ± 0.46% to 1.48 ± 0.12% ID/g for [ 18  In summary, both radiotracers developed were able to clearly visualize both integrin α v β 3 -positive and MC1R-positive tumors, and both parts of the heterobivalent agents contributed to receptor-specific tumor uptakes (see Supplementary Materials Figure S93). However, [ 18 F]4 demonstrated lower non-target organ uptakes and faster clearance than [ 18 F]2, thus resulting in considerably higher tumor-to-background ratios, despite its lower in vitro receptor binding affinities to both target receptor types. Therefore, [ 18 F]4 proved to be the more promising radiotracer for the bispecific imaging of malignant melanoma by PET/CT, having a high potential for clinical translation.

In Vivo PET Imaging
Briefly, 4-5-week-old male nude mice (Balb/cAnNRj-Foxn1 nu/nu ) were obtained from Janvier. A dynamic PET scan over 90 min and a subsequent CT image over 20 min were acquired using a triple-modality Bruker Albira II small-animal PET/CT/SPECT scanner. Three animals were studied per group.

Synthesis of Peptides 7-9
General procedure for peptide synthesis (GP1): Peptides were synthesized on a solid support according to standard Fmoc-based solid-phase peptide synthesis (SPPS) protocols. The resin was first swollen in CH 2 Cl 2 for 60 min and then rinsed with DMF. After deprotection of the Fmoc-protecting group (PG) with piperidine/DMF (1/1, v/v) for 2 + 5 min, the respective amino acid (4.0 equiv.) was pre-activated with DIPEA (4.0 equiv.) and HBTU (3.9 equiv.) in DMF for 2 min and then coupled for 60 min. These steps were repeated until the respective peptide sequence was complete.

Modification of the Peptides with Linker Structures to Obtain 10-21
General procedure for the synthesis of NHS-PEG n peptides (GP2): All steps were carried out under an N 2 atmosphere. A total of 1.0 equiv. of the respective peptide was added to a solution of 1.0 equiv. bis-NHS-ester and 0.5-1.0 equiv. DIPEA in dry DMF. Subsequently, the reaction mixture was stirred for 5-40 min at room temperature, while reaction control was performed by HPLC (analytical, 0-50% MeCN + 0.1% TFA in 5 min). After the removal of the solvent under reduced pressure, the corresponding NHS-PEG npeptide was obtained after purification via semipreparative HPLC. In addition to the respective target product, small amounts of the hydrolyzed compound HO-PEG n -peptide and the dimer peptide-PEG n -peptide were isolated (for analytical data, see Supplementary Materials Figures  1.0 equiv.) were reacted in 6 mL dry DMF for 40 min. After purification by HPLC (semipreparative, 10-30% MeCN + 0.1% TFA in 10 min, t R = 8.51 min), 17 was isolated as a colorless solid (6.9 mg, 4.7 µmol, yield: 49%, purity: >91%). MS (MALDI) m/z calculated for C 68 [M + H]  General procedure for the synthesis of HO-DIG-and HO-Ox-EGEGE peptides (GP3): All steps were carried out under an N 2 atmosphere. In total, 1.0-10.0 equiv. of the respective peptide was added to a solution of 1.0-10.0 equiv. bis-NHS-ester and 1.0 equiv. DIPEA in dry DMF. Subsequently, the reaction mixture was stirred for 5-50 min at room temperature, while the reaction control was performed by HPLC (analytical, 0-50% MeCN + 0.1% TFA in 5 min). After the removal of the solvent under reduced pressure, the corresponding HO-DIG-or HO-Ox-EGEGE peptides were obtained after semipreparative HPLC purification. In addition to the target products, small amounts of the dimers peptide-DIG/Ox-EGEGEpeptide were isolated (for analytical data, see Supplementary Materials Figures S81-S84).

Log D(7.4) Determination of [ 18 F]1-[ 18 F]6
The radiotracer solutions were first diluted with 0.9% NaCl solution to give a final EtOH concentration of <10%. A total of 5 MBq of the respective radiotracer in solution was added to 1.6 mL of a mixture of n-octanol and 0.05 M phosphate buffer (pH = 7.4) (v/v, 1/1), and the solution was vigorously shaken for 5 min. Subsequently, the phases were separated by centrifugation at 13.4 rpm for 2 min, and 100 µL of organic and aqueous phases were collected. The activity of each sample was measured using a gamma counter. The lipophilicity of each compound was determined in triplicate in three independent experiments.

Determination of the Stability of [ 18 F]1-[ 18 F]6 in Human Serum
The radiotracer solutions were first diluted with 0.9% NaCl solution to give a final EtOH concentration of <10%. Then, 6 × 25 µL radiotracer solution was added to 6 × 100 µL human serum and incubated at 37 • C for 120 min. At defined time points (5,15,30,60,90 and 120 min), 125 µL EtOH was added to one of the mixtures and the precipitation of serum proteins was supported by ice-cooling for 2 min. After centrifugation at 13.4 rpm for 2 min, the supernatant was analyzed by analytical radio-HPLC (0-100% MeCN + 0.1% TFA in 5 min). The experiment was performed for each compound trice by three independent experiments.

Cell Culture
All cell lines were cultivated at 37 • C in a humidified incubator at 5% CO 2 . B16F10 cells were cultured in DMEM and the U87MG cells in EMEM, each medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. The medium was exchanged every 2-3 days and cells were split at 70-90% confluence using 0.25/0.02% Trypsin/EDTA (w/v) in PBS. A medium change was performed 24 h before an experiment. For in vivo experiments, the cell resuspension after centrifugation was performed in PBS. The cells were homogenized in PBS to give a concentration of 5 × 10 6 B16F10 cells/mL and 25 × 10 6 U87MG cells/mL and were aliquoted and stored on ice upon use.

Competitive Displacement Studies on B16F10 and U87MG Cells
To determine the binding affinity to the respective receptor, competitive displacement studies were performed on MC1R-expressing B16F10 and on integrin α v β 3 -expressing U87MG cells. Each compound was evaluated at least three times, each experiment being The respective cells were harvested and re-suspended in the binding buffer to give a cell concentration of 2 × 10 6 /mL. After the BSA solution was filtered using the Millipore Multiscreen vacuum manifold, 50 µL of a cell suspension containing 10 5 cells were seeded in each well. Subsequently, 25 µL of the 125 I-labeled competitor solution (0.018 kBq/µL) and 25 µL of the respective compound to be tested were added. The compound to be tested was added in eleven increasing concentrations, whereas the 12th well contained no test compound to ensure the 100% binding of the 125 I-labeled competitor. After incubation of the plate for another hour at 25 • C, the solution was filtrated using the Millipore Multiscreen vacuum manifold, and the cells were washed three times with cold PBS (1 × 200 µL, 2 × 100 µL). Using a Millipore MultiScreen disposable punch and a Millipore MultiScreen punch kit, the filters of the well plate were collected in γ-counter tubes separately and measured by γ-counting. The determination of the half-maximal inhibitory concentration (IC 50 ) values was performed by fitting the obtained data via nonlinear regression using GraphPad Prism (v6.05).

In Vivo PET/CT Imaging and Ex Vivo Biodistribution of [ 18 F]2 and [ 18 F]4
Each male nude mouse (six weeks old) was injected subcutaneously with 5 × 10 5 B16F10 cells into the right flank and 2.0-2.5 × 10 6 U87MG cells into the left flank. The health status and tumor growth of the mice were monitored regularly until the animals could be examined after 15-21 days, depending on the tumor size. For the in vivo experiments, the 18 F-radiolabeled compound was diluted in 0.9% saline to give a final EtOH concentration of <10%. Each mouse was injected with 4.15 ± 2.28 MBq of [ 18 F]2 or 3.95 ± 2.06 MBq of [ 18 F]4 via the lateral tail vein under isoflurane anesthesia. For the blocking studies, the respective radiotracer was coinjected with 20 µg NDP, 200 µg c(RGDyK) or both substances (double blocking). Each 18 F-labeled compound was studied with or without blocking in at least three mice. Mice were measured under isoflurane anesthesia in a small PET/SPECT/CT animal imaging system. First, a dynamic PET scan was performed over 90 min and the scan was framed in 29 timeframes (10 × 1 min, 10 × 2 min, 6 × 5 min, 3 × 10 min). Images were reconstructed using 12 iterations, a maximum likelihood expectation maximization (MLEM) algorithm including corrections for scattered radiation and decay and a voxel size of 0.5 mm. All PET scans were immediately followed by CT acquisition at a voltage of 45 kV and a current of 400 µA. The images were reconstructed using the filtered back projection (FBP) algorithm with a voxel size of 250 µm. The analysis of the data, including the generation of TACs of the kidneys, liver and tumors and the MIPs, was performed via PMOD (v3.8). For ex vivo biodistribution, the mice were sacrificed directly after the PET/CT scan. Organs (blood, spleen, liver, kidneys, pancreas, lungs, heart, brain, bone, muscle, tail, tumors, stomach, colon and small intestine) were collected, weighed and their radioactivity measured in a γ-counter. The percentage injected dose per gram (% ID/g) of each tissue was calculated from the determined values, organ weights, reference values and injected activity.

Conclusions
In the present study, different heterobivalent bispecific 18 F-labeled agents for the sensitive and receptor-specific imaging of malignant melanoma using PET/CT were developed. After establishing the chemical and radio synthesis of the agents, the obtained tracers were studied systematically in vitro, regarding their hydrophilicity, stability in human serum and receptor-binding potential of both target receptor types, integrin α v β 3 and MC1R. It was shown that the distance between the peptide binders strongly influences receptor affinities and that the introduction of negatively charged linkers negatively affects the receptor-binding potential of both receptor types. In vivo, the most potent tracers were studied in direct comparison to PET/CT imaging and ex vivo biodistribution studies. These experiments showed higher absolute tumor uptakes and tumor-to-background ratios and thus more favorable in vivo pharmacokinetics for the agent, demonstrating slightly lower affinities but comprising longer PEG linkers, though not for that agent exhibiting the highest receptor affinities. Heterodimer [ 18 F]4 thus demonstrated an excellent receptor-specific tumor visualization ability and impressively illustrated the suitability of the underlying concept to develop heterobivalent integrin α v β 3 -and MC 1 R-bispecific radioligands for the sensitive and specific imaging of malignant melanoma by PET/CT.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.