Synthetic Integrin-Targeting Dextran-Fc Hybrids Efficiently Inhibit Tumor Proliferation In Vitro

Herein, we present the design, synthesis, and biological evaluation of novel integrin-targeting molecular hybrids combining RGD peptides and a potent cytotoxin presented on dextran polysaccharides. Based on an aglycosylated Fc as a centerpiece, endosomal-cleavable cytotoxic agent monomethyl auristatin E (MMAE) and dextran as multimerization site were covalently connected by two bioorthogonal enzyme-mediated reactions site-specifically. Decoration of dextran with cyclic RGD peptides, introduced by copper “click” reaction, resulted in the final constructs with the potential to kill integrin-overexpressing tumor cells. We found that these modifications had little impact on the stability of the Fc scaffold and the RGD-bearing construct showed good binding properties of αvβ3-expressing U87MG cells. Furthermore, the construct showed a remarkable antiproliferative activity. These results demonstrate the general capability of our design to provoke receptor-mediated endocytosis upon binding to the cellular surface, followed by endosomal cleavage of the linkage between Fc-dextran and MMAE and its subsequent release. Our approach opens new avenues to transcribe small molecule binders into tailor-made multimeric molecular hybrids with antitumor potential.


INTRODUCTION
Over the last decades, specific targeting of cancer cells has become an increasingly important issue in the development of novel approaches to antitumor therapy. Numerous strategies have been proposed to reach this long-sought goal, one of them being the combination of unspecific cytotoxic drugs with a targeting moiety (Borsari et al., 2020). In particular, antibody-drug conjugates (ADCs) became popular as vehicles relying on immunoglobins, which bind to overexpressed antigens on cancer cells and therefore facilitate the delivery of cytotoxic agents (Khongorzul et al., 2020). Besides these large and complex architectures, targeting modules based on small molecules and peptides have also been developed (Krall et al., 2013).
Within the range of potential targets for anticancer therapy, integrins have gained special interest due to their importance for tumor progression (Nieberler et al., 2017). Being the key actors in cell-cell and extracellular matrix communication, integrins are important regulators of cellular processes such as adhesion, proliferation, apoptosis, and migration (Humphries et al., 2006;Nieberler et al., 2017). These glycosylated transmembrane proteins are heterodimers; a pool of 18 α-subunits and 8 β-subunits results in at least 24 variants in humans (Yousefi et al., 2021). Expression levels of several integrin subtypes, most prominently αvβ3, have been reported to be elevated in solid tumors, for example, melanoma, breast, pancreatic cancer, and glioblastoma, as well as in tumor blood vessels (Desgrosellier and Cheresh, 2010). Since the discovery of the binding motif arginine-glycine-aspartic acid (RGD) in the natural ligand fibronectin (Pierschbacher and Ruoslahti, 1984), much effort has been directed to the development of improved integrin binders based on this triad sequence. The selectivity for tumor-relevant subtypes and activity was enhanced by cyclization, for example, head-to-tail macrocyclization (Aumailley et al., 1991), and incorporation of diketopiperazine (Marchini et al., 2012). Further progress was achieved by the introduction of D-amino acids (Aumailley et al., 1991) and N-methylation (Dechantsreiter et al., 1999). These improvements culminated in the development of the integrin antagonist cilengitide cyclo[RGDf(NMe)V] (Dechantsreiter et al., 1999) which unfortunately failed in clinical trials (Tucci et al., 2014). Paradoxically, cilengitide as an integrin agonist induced tumor growth and angiogenesis at low nanomolar plasma concentrations which were reached in most clinical trials during the drug administration schedule (Reynolds et al., 2009). This could be attributed to the finding that cyclic RGD antagonists are capable of promoting a major conformational change in the integrin αvβ3 receptor that prompts it to adopt a high-affinity ligand-binding state (Takagi et al., 2002), thus suggesting that a high local concentration of the ligand is required to avoid these unwanted effects.
Alternatively, RGD-containing peptides can be employed as a tumor-homing module in targeted drug delivery to possibly negate the effects of the concentration-dependent ambivalence. The respective strategies have already been realized by either conjugation of RGD-containing integrin binders to cytotoxic drugs like paclitaxel , doxorubicin (Burkhart et al., 2004), cryptophycin (Nahrwold et al., 2013), and monomethyl auristatin E (MMAE) (Lopez Rivas et al., 2019) or genetic fusion to proteins such as tumor necrosis factor-α (TNF-α) (Zarovni et al., 2004). Compared to the solitary ligands, these constructs in general displayed similar binding capabilities to the target integrins and, furthermore, improved pharmacological properties of the conjugated payloads. Besides targeted drug delivery, RGD-containing peptides have also shown their potential for tumor imaging, for example, with radioisotopes as cargo for positron emission tomography (PET) . In order to enhance the affinity of RGD ligands and also the rate of specific internalization (Sancey et al., 2009;Kemker et al., 2020), multimerized RGD constructs were synthesized. Different platforms for the multimeric presentation were used, among them were polymers (Komazawa et al., 1993), peptides (Thumshirn et al., 2003;Sancey et al., 2009), antibodies (Kok et al., 2002), and liposomes (Wu et al., 2020).
Herein, we report the development and evaluation of a novel platform for the multimeric presentation of RGD ligands for targeted drug delivery relying on a dextran-Fc scaffold ( Figure 1A). Dextran is a flexible polysaccharide comprised of α-(1-6)-linked D-glucose units with a low degree of branching (van Witteloostuijn et al., 2016), which can be modified to allow for the covalent attachment of cargo molecules. This polymer has been already used for the RGD-mediated delivery of doxorubicin and bortezomib (Li et al., 2020). Our group has recently reported dextran as a multimerization scaffold for novel ADCs (Schneider et al., 2019a), death receptor 5-targeting peptide (Schneider et al., 2019b), and cell-penetrating peptide L17E (Becker et al., 2021). Analogous to our previous work (Schneider et al., 2019b), a human Fc was chosen as a centerpiece of the construct, which serves as a scaffold for enzyme-mediated site-specific conjugation and, furthermore, enables purification by affinity chromatography. We chose the tubulin inhibitor MMAE as a potent cytotoxic agent (Chen et al., 2017), which was additionally equipped with a lysosomal cleavable and self-immolative valinecitrulline-PAB linker to enable release after endocytosis ( Figure 1B) (Dubowchik et al., 2002).

NMR Spectroscopy
NMR measurements were performed on a 300 MHz Avance II or 300 MHz Avance III spectrometer (Bruker BioSpin GmbH, Rheinstetten, Germany). Samples were dissolved in deuterium oxide or CDCl 3 .

HPLC
Reversed-phase (RP) HPLC measurements were conducted on an Agilent Infinity 1100 device (Agilent, Santa Clara, United States) equipped with an Agilent Eclipse Plus RP column (C18, 3.5 µm, 100 mm × 4.6 mm, 95 Å) at a flow rate of 0.6 ml min −1 . Alternatively, an Agilent Infinity 1260 device equipped with an Interchim (Montluçon, France) Uptisphere Strategy (C18-HQ, 3 μm, 100 mm× 4.6 mm) column was used. The eluent system consisted of eluent A: 0.1% (v/v) aq. TFA, and eluent B: Product purities were estimated upon the percentage of peak area under the curves at 220 nm wavelength using the Agilent OpenLab Software.
Hydrophobic interaction chromatography (HIC) was performed on an Agilent Infinity 1260 device equipped with a Tosoh Bioscience GmbH (Griesheim, Germany) TSK Butyl-NPR column (2.5 µm, 35 mm × 4.6 mm). The eluent system consisted of eluent A: 1.5 M (NH 4 ) 2 SO 4 in 20 mM Tris pH 7.5, and eluent B: 20 mM Tris pH 7.5. Analysis was performed applying a 35 min gradient at a flow rate of 0.9 ml min −1 .

Synthesis of 4-Pentynoic Acid NHS Ester 17
The synthesis was performed according to the literature (Pal and Koner, 2017). Thus, 4-pentynoic acid (200 mg, 2.04 mmol, 1 eq.) and N-hydroxysuccinimide (268 mg, 2.33 mmol, 1.15 eq.) were dissolved in 10 ml anhydrous dichloromethane (DCM). The reaction mixture was cooled with an ice bath, and 1-ethyl-3-(3dimethylaminopropyl)carbodiimide hydrochloride (EDC·HCl) (780 mg, 4.07 mmol, 2 eq.) was added in portions. The reaction mixture was stirred for a further 2 h, filtered, and evaporated under reduced pressure. The residue was dissolved in ethyl acetate, washed thrice with saturated aqueous NaHCO 3 and thrice with brine, and subsequently dried over Na 2 SO 4 . The organic phase was evaporated, and the residue was dissolved in ACN and precipitated by the addition of water.

Synthesis of H-D(OtBu)fK(Boc)R(Pbf)G-OH 11 and H-D(OtBu)fK(Boc)R(Pbf)A-OH 12
Synthesis was performed as described in the literature (Dai et al., 2000) with some modifications. Thus, 2-chlorotrityl chloride resin (2-CTC resin, 0.75 mmol, loading: 1.59 mmol/g) was loaded with 4 eq. Fmoc-Gly-OH and Fmoc-Ala-OH, respectively, in DCM in the presence of 8 eq. DIEA for 2 h. Then the supernatant was drained and the resin was washed thoroughly with DMF. Fluorenylmethoxycarbonyl (Fmoc)-protecting group was removed with 20% (v/v) piperidine in DMF for 5 and 10 min. After thorough washing with DMF, the subsequent building blocks were coupled using 4 eq. Fmoc-protected amino acid, 3.95 eq. HBTU, and 8 eq. DIEA for 1 h in DMF. Fmoc-f-OH was coupled using 3 eq. amino acid, 2.95 eq. HBTU, and 6 eq. DIEA. After final Fmoc deprotection, the peptides were cleaved from resin under mild conditions using the mixture of acetic acid: DCM:methanol (5:4:1, v/v/v) for 2 h. Cleavage cocktail was removed under reduced pressure and the residue was dissolved in ACN, followed by precipitation by the addition of water.

Dextran Functionalization
Dextran functionalization was performed as previously reported (Schneider et al., 2019b). Analytical data of the final dextran derivative cadaverine-dextran-(N 3 ) 15.6 9 are given in Supplementary Material.

Protein Expression and Purification
Expression and purification of aglycosylated (N297) human Fc-LPETGG 1 were realized as previously reported (Schneider et al., 2019b). Briefly, the DNA sequence for the protein was cloned in a standard pEXPR vector for mammalian expression. Forty micrograms of the plasmid was mixed with 120 µg polyethylenimine (PEI) in a serum-free Expi293 TM expression medium (Thermo Fisher Scientific, Waltham, United States). The mixture was added dropwise to 2.5 × 10 6 Expi293F TM cells/ml in 30 ml Expi293 TM expression medium. After 24 h incubation under constant shaking, 0.5% (w/v) tryptone was added and the cells were incubated for further 5 days. Following this, the supernatant was diluted with an equal volume of protein A running buffer (20 mM sodium phosphate, Frontiers in Chemistry | www.frontiersin.org July 2021 | Volume 9 | Article 693097 pH 7) and purified by protein A affinity chromatography (HiTrap Protein A HP column, GE Healthcare, Chicago, United States). Elution of Fc 1 was performed with 100 mM citrate buffer pH 3. Product fractions were pooled, dialyzed against phosphate-buffered saline (PBS) pH 7.4, and concentrated using Amicon Ultra centrifugal filters (Merck Millipore, Burlington, United States). Microbial transglutaminase (mTG, Streptomyces mobaraensis) was produced and activated with trypsin as described in the literature (Deweid et al., 2018).
Sortase A was expressed and purified according to the literature (Chen et al., 2011).

Copper(I)-Catalyzed Azide-Alkyne Cycloaddition on Fc-Dextran
CuAAC of alkyne-modified RGD or RAD to Fc-dextran was performed as described in the literature (Schneider et al., 2019b). Prot A-immobilized Fc-dextran-MMAE 10 or Fc-dextran 22 (1 eq.) in PBS was mixed with either alkyne-RGD 18 or alkyne-RAD 19 (2.5 eq. per N 3 ). Fresh stocks of ascorbic acid (5 eq. per N 3 ) and CuSO 4 ·5H 2 O (2.5 eq. per N 3 ) in water were mixed, incubated for 5 min, and added to the immobilized proteins. The reactions were incubated for 3 h at 30°C, followed by the removal of the solvent and washing of the immobilized proteins with PBS and subsequent elution. The buffer was exchanged to PBS and the products 20, 21, 23, and 24 were concentrated with Amicon Ultra centrifugal filters. Reaction success was confirmed by SDS-PAGE.

Cell Culture
Cells were incubated under standard conditions in a humidified incubator with 5% CO 2 at 37°C. Expi293F TM cells were incubated in serum-free Expi293 TM expression medium in an orbital shaker at 110 rpm and 8% CO 2 . U87MG cells were incubated in minimum essential medium (MEM) supplemented with 10% (v/v) fetal bovine serum (FBS).

Cell Proliferation Assay
U87MG cells were seeded in 96-well plates with a density of 3.5 × 10 3 cells and incubated for 7 h under standard conditions. Serial dilutions of samples (10× concentrated in PBS) were added to a final volume of 100 µl and the cells were incubated for a further 72 h. MTS (AQueous One Solution) was added and cell proliferation was measured using a Tecan Infinite F200 PRO (Männedorf, Switzerland). Samples in reference wells containing untreated cells were set to 100% viability.

Cell Binding Assay
Trypsinized U87MG cells were washed twice with 0.1% (w/v) bovine serum albumin (BSA) in PBS. 2.5 × 10 5 cells were transferred to a 96-well plate and incubated with the respective concentration of constructs in PBS + 0.1% BSA for 35 min on ice. The supernatant was removed and the cells were washed thrice with 0.1% BSA in PBS. Following this, the cells were incubated for 20 min on ice with fluorescently labeled IgG Fc goat anti-human, PE, eBioscience TM (1/100 diluted in PBS + 0.1% BSA). Subsequently, cells were washed once with 0.1% BSA in PBS and analyzed by flow cytometry using a BD influx device (Becton Dickinson, Franklin Lakes, United States).

Thermal Shift Assay
Thermal shift assays were performed as duplicates on a CFX96 device (Bio-Rad, Hercules, United States). Measurements were performed at protein concentrations of 0.1 mg ml −1 in PBS using SYPRO Orange (dilution 1:800) whereby temperature was increased in 0.5°C/30 s increments up to 99°C. Melting temperatures (T m ) were derived from melting curves using the Bio-Rad software.

Statistical Analysis
Statistical analysis was performed with GraphPad Prism version 8.0.1. Results are displayed as the mean ± standard error of the mean and are based on triplicates. Statistical significance was determined via a two-way ANOVA test (Bonferroni t-test). p values ≤ 0.05 were considered to be statistically significant.

Design and Synthesis
To ensure orthogonal and site-specific conjugation of both dextran and cytotoxic agent, two enzyme recognition sites were introduced into Fc 1. For sortase A-mediated conjugation of GGG-PEG 3 -Val-Cit-PAB-MMAE 2, a C-terminal LPETGG extension was incorporated genetically (Macarrón Palacios et al., 2020). Furthermore, Fc was produced as an aglycosylated variant N297A which enables coupling of cadaverine-functionalized dextran to Q295 by microbial transglutaminase (Schneider et al., 2019b). The modification of Fc started with sortase A-catalyzed conjugation of MMAE 2 (Figure 2), and product Frontiers in Chemistry | www.frontiersin.org July 2021 | Volume 9 | Article 693097 3 was isolated using protein A affinity chromatography. Fc-MMAE 3 was analyzed by HIC, whereby the drug-to-protein ratio was 1.6 according to the integrals of the corresponding peaks (Supplementary Figure S28). In the next step, dextran 9 was conjugated to Fc-MMAE 3 using transglutaminase (Figure 2), an enzyme catalyzing the formation of isopeptide bonds between suited glutamine residues and amine counterparts. Therefore, 10 kDa (corresponds to 62 glucose units on average) dextran 4 was equipped with a Boc-protected cadaverine moiety via reductive amination (compound 5) and was further functionalized with 15.6 carboxyethyl groups at C2 position on average (compound 6) ( Figure 3). EEDQ-mediated conjugation of amine-bearing linker 7 to the carboxy groups yielded azide-functionalized dextran 8. Finally, acidic cleavage of the Boc-protecting group at the reducing end resulted in dextran 9. After mTG-mediated conjugation of azide-bearing dextran 9 to Fc-MMAE 3, product 10 was isolated by protein A affinity chromatography and remained immobilized for the subsequent modification via copper "click" reaction. Therefore, alkyne-modified cyclo [RGDfK] 18 and, additionally, cyclo [RADfK] 19 as negative control were required (Figure 4), which were obtained by solid-phase peptide synthesis using the standard Fmoc strategy according to the literature (Dai et al., 2000). After mild cleavage to conserve the side chainprotecting groups, the linear peptides 11 and 12 were head-to-tail cyclized at low concentrations to minimize side reactions.  Figure S15) displayed two product peaks at approximately equal ratios and identical m/z signals. This indicates the formation of stereoisomers by the racemization of the C-terminal FIGURE 2 | Synthetic approach toward the final constructs 20 and 21, starting with sortase A-mediated conjugation of MMAE 2 to Fc 1, followed by transglutaminase-catalyzed conjugation of azide-decorated dextran 9. Construct 10 was further functionalized via CuAAC with either alkyne-decorated RGD 18 or RAD 19 to yield final constructs 20 and 21. Alternatively, Fc 1 was directly functionalized with dextran (construct 22) and further modified via CuAAC to obtain RGD-decorated 23 and RAD-decorated 24. Figure was created with biorender.com.
Frontiers in Chemistry | www.frontiersin.org July 2021 | Volume 9 | Article 693097 amino acid as has been reported for head-to-tail cyclization of peptides (Davies, 2003). This assumption is supported by the fact that this did not take place for the synthesis of cyclo[RGDfK] 15 (Supplementary Figure S12), which is based on a C-terminal glycine. However, it remains unclear why no chromatographic separation of the isomers did occur for the side chain-protected precursor 14. An alkyne moiety was introduced by coupling of 4-pentynoic acid NHS ester 17 to the respective lysine side chains. Conjugation of alkyne-functionalized cyclo [RGDfK] 18 and cyclo [RADfK] 19, respectively, to protein A-immobilized Fc-MMAE-dextran 10 was achieved by copper "click" reaction ( Figure 2) which yielded the final constructs 20 (RGD-decorated) and 21 (RAD-decorated). In addition to the MMAE-functionalized constructs 20 and 21, Fc-dextran-containing RGD (construct 23) and RAD (construct 24), respectively, lacking a cytotoxic warhead were synthesized as controls. To that end, Fc 1 was functionalized with dextran 9 via transglutaminase-mediated transamidation yielding Fc-dextran 22 (Figure 2), which was further decorated with alkyne-bearing cyclo[RGDfK] 18 and cyclo [RADfK] 19 (Figure 2), respectively, to obtain constructs 23 and 24.

Biology
In order to assess, whether the modifications had an impact on the stability of the constructs, the melting properties of construct 20 were compared to parent Fc 1. Thermal shift assay (Supplementary Figures S38, S39) revealed only negligible differences between construct 20 (T m 64.75 ± 0.25°C) and nonfunctionalized Fc 1 (T m 66.00 ± 0.50°C). Next, the binding properties of the constructs 20 and 21 on U87MG cells, which display high levels of αvβ3 on their surface (Benedetto et al., 2006), were investigated. To that end, trypsinized U87MG cells were incubated with either 20 or 21 at different concentrations, followed by labeling of the Fc part with a detection antibody and FIGURE 3 | Synthetic approach toward cadaverine-dextran-(N 3 ) 15.6 9.
Frontiers in Chemistry | www.frontiersin.org July 2021 | Volume 9 | Article 693097 subsequent flow cytometry. RGD-functionalized construct 20 showed higher binding capabilities to U87MG cells compared to RAD-decorated construct 21 ( Figure 5). These findings clearly demonstrate that specific interactions of RGD-containing construct 20 with the cellular surface of U87MG cells play a pivotal role for binding, since RAD-modified 21, which has alanine instead of glycine in the cyclic peptide, shows negligible binding. The binding of RGD-containing construct 20 revealed an apparent dissociation constant K D of 33 nM, whereas no K D value for the RAD-decorated construct 21 could be determined since no sigmoidal binding curve-up to a concentration of 600 nM-was observed (Supplementary Figure S40). Encouraged by these results, we investigated the impact of RGD-and RAD-modified constructs 20 and 21, respectively, as well as Fc-MMAE 3, on the proliferation of U87MG cells. Cells were treated with compounds at different concentrations for 3 days, after which the proliferation was quantified in an MTS assay. RGD-decorated construct 20 displayed higher antiproliferative effects (EC 50 16.15 nM) compared to the RAD-decorated counterpart 21 (EC 50 > 100 nM, Figure 6A). This observation strongly suggests an integrin-dependent uptake of construct 20 which results in higher intracellular concentrations of MMAE and, ultimately, higher cytotoxicity. The antiproliferative activity of construct 21 (EC 50 > 100 nM) was found similar to Fc-MMAE 3, implying nonspecific activity of RAD-decorated 21. This could be caused by either nonspecific uptake or hydrolysis of the linkage between Fc and cytotoxic agent, which is membrane-permeable in free form, within the incubation time. In comparison with a solitary RGD-containing peptide connected to MMAE by a lysosomally cleavable Val-Ala linker, construct 20 displayed higher cytotoxic activity in U87MG cells than its monomeric counterpart (IC 50 39 nM) (Raposo Moreira Dias et al., 2019). To investigate whether multimerized RGD without cytotoxic payload exhibits antiproliferative effect by receptor clustering, cytotoxicity of construct 23 was assessed and compared to RAD-decorated construct 24 ( Figure 6B). Both constructs 23 and 24, respectively, display no cytotoxicity-up to a concentration of 600 nM-highlighting the necessity of a cytotoxic payload for RGD-directed killing of tumor cells.

CONCLUSION AND OUTLOOK
The combination of unspecific cytotoxic agents with targeting moieties offers a promising strategy towards targeted antitumor therapy. Peptides incorporating the RGD-motif have proven their efficacy as integrin binders in numerous publications, whereby their potency was even further enhanced by multimerization. Herein, we present the design, synthesis, and biological evaluation of a novel scaffold bearing multiple RGD peptides in combination with a potent cytotoxin. Based on a nonglycosylated Fc as a centerpiece, endosomal-cleavable MMAE as cytotoxic agent and dextran as multimerization site were introduced by enzymemediated site-specific reactions. Multiple attachments of RGD peptides were achieved by copper "click" reaction resulting in final construct 20. These modifications had little impact on the stability. Indeed, negligible differences in melting temperatures between construct 20 and parent Fc 1 were observed. RGDcontaining construct 20 showed superior binding of αvβ3expressing U87MG cells compared to the negative control, that is, RAD-decorated construct 21. These properties also translated into their antiproliferative activity, and construct 20 displayed significantly higher cytotoxicity than control 21. In these constructs, the binding is not mediated by the antibody part, but by dextran decorated with a high number of cyclic peptidic integrin binders. Thus, this format enables to transcribe small binders into an antibody-like format. The observed potency of 16.15 nM is comparable to a cystine knot peptide genetically fused to Fc which binds integrins containing an αv subunit with high affinity and was equipped with 1.9 MMAF (EC 50 9.2 nM) (Currier et al., 2016). We recently showed that multiple MMAE conjugations to dextran result in higher cytotoxicity compared to single coupling as used in this study (Schneider et al., 2019a). Further work will reveal whether the usage of a dual modified Fc with a variable number of multiple copies of the integrin-binding peptides and of the cytotoxin both installed separately on dextran result in tunable cell-specific cytotoxicity. Already in the present setting, our results demonstrate the general capability of our design to provoke receptor-mediated endocytosis upon binding to the cellular surface, followed by endosomal cleavage of the linkage between Fc-dextran and MMAE and its subsequent release. In conclusion, our approach opens new avenues for the development of tailor-made multimeric molecular hybrids with antitumor potential.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.  Frontiers in Chemistry | www.frontiersin.org July 2021 | Volume 9 | Article 693097