Human serum albumin (HSA) (Baxter); human natural αvβ3 integrin (Immunological Science); anti-polyethyleneglycol (PEG) monoclonal antibody (mAb), clone 26A04 (Abcam); bovine serum albumin (BSA) fraction-V (Sigma); gold nanoparticles (25 nm, A_{520 nm} ∼1 unit/ml) (Aurion); MAL-dPEG₁₁-lipoamide (Mal-PEG₁₁-LPA) and MAL-dPEG₃-lipoamide (Mal-PEG₃-LPA) (Quanta Biodesign). Murine TNF was prepared as described previously (Curnis et al., 2013).

Linear peptide CGisoDGRG was assembled using the Fmoc-based solid-phase method (Fields and Noble, 1990) on a 2-CTC resin. Upon synthesis, the peptide was cleaved from the resin, using 1% trifluoroacetic acid (TFA) in dichloromethane, and dried. The resulting material was dissolved in N,N′-dimethylformamide (DMF, 50 mM) and treated with HBTU/DIEA (1 eq.) to obtain the fully protected cyclo-CGisoDGRG peptide. DMF was removed under vacuum and the resulting crude material was directly treated with TFA-based cleavage mixture (TFA 95%, TIS 2.5%, Thianisole 2.5%) to obtain the unprotected peptide, which was recovered by precipitation in cold diethyl ether. The peptide was purified by reverse-phase (RP)-HPLC, using a Shimpack GWS C18 column (10 µm, 21.2 mm x 250 mm, Shimadzu), lyophilized, and stored at −80°C. Aliquots of the product were dissolved in water and stored at −80°C until use. The identity and purity of the product, called iso1 (see Figure 1), were confirmed by electrospray ionization mass spectrometry [expected monoisotopic mass (MH⁺): 546.22 Da; found: 546.27 Da] and RP-HPLC analysis (purity > 95%).

Iso1-PEG₁₁-LPA was prepared by coupling iso1 to Mal-PEG₁₁-LPA using a 1:1.2 molar ratio, as follows: 4 mg of iso1 in 100 µl of 0.4 M sodium phosphate buffer, pH 7.8, were mixed with 50 µl of acetonitrile (final pH: ∼7). The resulting solution was chilled on ice and mixed with 7.7 mg of MAL-PEG₁₁-LPA in 100 µl of 50% acetonitrile and left to react under gentle shaking at room temperature. The coupling reaction was monitored by RP-HPLC and stopped when complete reagent conversion occurred. In parallel, a control conjugate was prepared using cysteine in place of iso1. Coupling reactions were monitored by RP-HPLC using a Shimpack GWS C18 column (5 µm, 4.6 mm x 150 mm, Shimadzu) connected to Shimadzu Prominence HPLC (mobile phase A, 0.1% trifluoroacetic acid in water; mobile phase B, 70% acetonitrile, 0.1% trifluoroacetic; linear gradient 10–100% B; in 20 min; flow rate, 1 ml/min). The final products were purified by RP-HPLC using a Jupiter C18 column (10 µm, 21.2 mm x 250 mm, Phenomenex), and the same mobile phases indicated above (liner gradient, 0–90% B, 45 min; flow rate, 14 ml/min). The products were lyophilized, resuspended in water, and stored at −20°C. Furthermore, iso1 and Cys were also coupled to MAL-dPEG₃-LPA, i.e., a compound with a shorter PEG chain, using the same procedures. The identity of each conjugate [iso1-PEG₁₁-LPA (compound 1), Cys-PEG₁₁-LPA (compound 2), iso1-PEG₃-LPA (compound 3), and Cys-PEG₃LPA (compound 4)] was checked by mass spectrometry analysis using an LTQ-XL Orbitrap mass spectrometer (Thermo Fischer) (Supplementary Table S1 and Supplementary Figure S1).

To functionalize gold nanoparticles with iso1 we mixed aliquots of colloidal gold (1 ml, adjusted to pH ∼7.5 with sodium hydroxide) with 4–32 µg of compound 1 in 5 mM sodium phosphate buffer, pH 7.4 (100 µl). The mixture (final pH ∼7.5) was incubated at room temperature for 2 h. To saturate gold nanoparticles, we then added 0.5% HSA (in 25 µl aliquots every 2 min, for four times) and left to incubate for 10 min at room temperature. The product was then centrifuged at 13,000x g for 15 min. The pellets were resuspended in 5 mM sodium phosphate buffer, pH 7.33, containing 0.05% HSA (buffer A). The centrifugation/washing steps were repeated twice. The final product (called 1-Au) was resuspended with 1 ml of buffer A and stored at 4°C.

Bifunctional gold NPs bearing compound 1 and TNF (called 1-Au/TNF) were prepared as follows: a solution containing 0.540 mg of compound 1 and 2.88 mg of TNF in 18 ml of 5 mM sodium phosphate, pH 7.4, was slowly added (1 ml/min) to 180 ml of 25 nm-nanogold, with pH adjusted to ∼7.5 with sodium hydroxide, and left to incubate for 40 min at room temperature under shaking. The product was mixed with 18 ml of 0.5% HSA in water (slowly added) and left to incubate for an additional 15 min, to saturate gold NPs. The mixture was centrifuged (13,000x g for 30 min) and resuspended in buffer A (three times). The pellet was then resuspended with 1.8 ml of buffer A, filtered (0.22 µm pore size, Millex-GV Filter), and stored at −80°C. Control nanoparticles bearing compound 2 and TNF (called 2-Au/TNF) were prepared following the above procedure, except that in this case, 0.360 mg of compound 2 was used.

Absorption spectra of coated- or uncoated-NPs were recorded using an UltroSpec 2100 spectrophotometer (Amersham Biosciences), a 1 cm path-length quartz cuvette, and buffer A or 5 mM sodium citrate buffer, pH 6.0, respectively, as blanks. The concentration of coated-NPs was calculated by interpolating the absorbance values at 530 nm on a calibration curve obtained using uncoated nanogold (stock solution: 3.3 × 10¹¹ NPs/ml, A₅₃₀ nm: 0.96 U/ml).

Transmission electron microscopy (TEM) analysis was performed using a TALOS L120C microscope (ThermoScientific) as described previously (Gasparri et al., 2019). Morphometric analysis of nanoparticle’s shape and diameter was performed using the ImageJ software, essentially as previously reported (Rice et al., 2013).

Human MSR3 melanoma cells were grown in Iscove’s modified Dulbecco’s medium (IMDM) (Lonza) supplemented with heat-inactivated 10% fetal bovine serum (FBS), 2 mM L-glutamine, and 1% penicillin/streptomycin. Cells were seeded into a 96-well plate (20,000 cells/well) and incubated for 24 h in complete cell culture medium. After medium removal, the cells were incubated with nanogold functionalized with compound 1 or 2 in cell culture medium (1.7 × 10¹¹ NPs/ml, 100 µl/well, 1.5 h at 37°C, 5% CO₂). After washing, the cells were further incubated for 4 h at 37°C, 5% CO₂, and then photographed using a bright field microscopy.

The bifunctional properties of 1-Au/TNF were checked using two sandwich assays based on αvβ3-and anti-PEG antibody-coated plates followed by an anti-TNF polyclonal antibody. The αvβ3/anti-TNF polyclonal antibody sandwich assay was performed essentially as described (Gasparri et al., 2019). Briefly, various amounts of nanoparticles in 25 mm Tris-HCl, pH 7.4, containing 150 mM sodium chloride, 1 mM magnesium chloride, 1 mM manganese chloride, 1% w/v BSA (binding buffer), were added to microtiter plates coated with or without αvβ3 (0.5–1 µg/ml). The binding of nanoparticles was then detected using a rabbit anti-TNF polyclonal antiserum (1:1,000), followed by a polyclonal goat anti-rabbit HRP-conjugate. Bound peroxidase was detected by adding the o-phenylenediamine chromogenic substrate. The anti-PEG antibody/anti-TNF polyclonal antibody sandwich assay was carried out essentially as described above using a microtiter plate coated without or with an anti-PEG mAb (5 µg/ml) in the capture step, and the anti-TNF polyclonal antibody in the detection step.

The amount of bioactive TNF bound to gold NPs was determined using an in vitro bioassay based on TNF-induced cytolysis of murine L-M fibroblasts, as described (Curnis et al., 2013), except that cell viability was quantified using the PrestoBlue Cell Viability (Invitrogen). International murine TNF reference standard (NIBSC, ID: 88/532) was used to calibrate the assay.

All procedures on mice were approved by the San Raffaele Institutional Animal Care and Use Committee, according to institutional guidelines, and in compliance with national (D.L. N.26, 04/03/2014) and international law and policies (new directive 2010/63/EU). Murine WEHI-164 fibrosarcoma cells (ATCC, cat. CRL-1751) were cultured in DMEM supplemented with 10% fetal bovine serum, 2 mM glutamine, 50 μg/ml streptomycin, 100 U/ml penicillin, and 0.25 μg/ml amphotericin-B. WEHI-164 cells were tested for mycoplasma contamination using the Mycoplasmacheck testing service provided by Eurofins Genomics before their use in vivo. BALB/c mice (6–8 weeks old, Charles River Laboratories), weighing 18–20 g, were challenged with subcutaneous injection in the left flank with 1.5 × 10⁶ WEHI-164 cells. Six days later, mice were injected (i.v.) with nanodrugs in 0.9% sodium chloride. Tumor growth was monitored by measuring the tumor size with a caliper. Tumor volumes were estimated by calculating r1 × r2 × r3 × 4/3π, where r1 and r2 are the longitudinal and lateral radii, and r3 is the thickness of the tumor protruding from the surface of normal skin. Animals were sacrificed before tumors reached a volume of 1 cm³. Tumor sizes are shown as mean ± SE.

To have at hand an isoDGR peptide that can be directly attached to gold nanoparticles, we have coupled cyclo-CGisoDGRG (iso1) to MAL-PEG₁₁-lipoamide and MAL-PEG₃-lipoamide, two cross-linking reagents with different polyethylene glycol chains (Figure 1). Both reagents contain a maleimide group (MAL), which can form a thioether bond with the sulfhydryl group of iso1, and a lipoamide group that can react with nanogold to form dative bonds. In parallel, we prepared control ligands consisting of cysteine in place of iso1. These conjugates, called iso1-PEG₁₁-LAP (compound 1), Cys-PEG₁₁-LAP (compound 2), iso1-PEG₃-LAP (compound 3), and Cys-PEG₃-LAP (compound 4), were purified by RP-HPLC. The identity of each product was confirmed by mass spectrometry analysis (Supplementary Table S1 and Supplementary Figure S1).

To functionalize nanogold with isoDGR, we then mixed 25 nm-gold nanoparticles with various concentrations of compound 1 or 3 and left them to incubate for 2 h. We observed nanogold aggregation with compound 3 at concentrations > 4 µg/ml, as indicated by a marked change of nanogold color, but not with compound 1 (Figures 2A,B, upper panels). Although compound 3 did not cause nanogold aggregation at concentrations < 4 µg/ml, this compound was unable to inhibit the aggregation induced by the addition of 5% (w/v) NaCl (Figure 2A, lower panels). This indicates that compound 3 loaded on gold nanoparticles was not sufficient to inhibit salt-induced aggregation. In contrast, all the tested doses of compound 1 (4–32 µg/ml) did not cause nanogold aggregation by itself and could efficiently protect nanoparticles from salt-induced aggregation (Figure 2B, lower panels). Based on this finding, nanogold functionalized with compound 1 and its relevant control (compound 2, lacking the iso1) were selected for further studies, whereas nanogold functionalized with compounds 3 and 4 were not further investigated. Nanoparticles prepared with compound 1 were further stabilized with human serum albumin and characterized by UV-Vis spectrophotometric analysis. The results showed that these nanoparticles were homogeneous, not aggregated, and resistant to salt-induced aggregation (Figures 2C,D).

To assess whether nanogold functionalized with compound 1 (called 1-Au) could bind membrane-associated αvβ3, we tested the binding of this product to MSR3 cells, a human melanoma cell line that expresses high levels of αvβ3 and previously used for the characterization of iso1 (Nardelli et al., 2018). In parallel, control nanogold functionalized with compound 2 (2-Au) was also tested. Phase-contrast microscopy experiments showed granular dot-like staining in MRS cells incubated with 1-Au, but little, or not at all, with 2-Au (Figure 3), suggesting that the binding of 1-Au was mediated by iso1. These data are in line with the hypothesis that 1-Au can recognize the αvβ3 integrin expressed on the cell surface.

To demonstrate that compound 1 can be exploited for delivering nanogold to αvβ3-positive tumors, we coated gold NPs with compound 1 and murine TNF, a cytokine endowed with potent anti-tumor activity. Optimal nanogold loading with cytokine and ligand was achieved with 16 µg/ml of TNF and 3–4 µg/ml of compound 1 or 2–4 µg/ml of compound 2 (Supplementary Figures S2–S4 and see also Supplementary Material). Based on these results, we prepared a larger batch of bifunctional nanoparticles functionalized with isoDGR and TNF using 16 µg/ml of TNF and 3 µg/ml of compound 1 (1-Au/TNF), a larger batch of control nanoparticles lacking isoDGR using 16 µg/ml of TNF and 2 µg/ml of compound 2 (2-Au/TNF).

UV-Vis spectrophotometric analysis of both products showed single absorption peaks at 530 nm with a similar width, indicating that both products contained low (undetectable) amounts of aggregates (Figure 4A and Table 1). Transmission electron microscopy showed that both nanodrugs had a maximal diameter of about 28 nm and were composed by spherical nanoparticles (Figure 4B).

To assess the presence of compound 1 or 2 and TNF on the two nanodrugs, we measured the capability of each product to form molecular sandwiches with an anti-PEG monoclonal antibody (mAb) and an anti-TNF polyclonal antibody (pAb), or with αvβ3 and anti-TNF pAb. To this aim, we performed assays based on the use of microtiter plates coated with anti-PEG mAb or αvβ3 in the capture step and anti-TNF pAb in the detection step (see Figure 4C for a schematic representation of the assays). 1-Au/TNF and 2-Au/TNF showed comparable binding properties when anti-PEG mAb-coated plates were used (Figure 4C, left panel), suggesting that these nanodrugs were loaded with similar amounts of PEG-containing ligands and TNF. In contrast, only 1-Au/TNF was detected when αvβ3 was used in place of anti-PEG mAb in the capture step, as expected (Figure 4C, right panel). Of note NPs functionalized with compound 3 (a ligand characterized by a shorter PEG length) and stabilized with TNF to prevent aggregation (3-Au/TNF) bound αvβ3 less efficiently than 1-Au/TNF (Supplementary Figure S5), suggesting that the PEG₁₁ chain of 1-Au/TNF is important for the nanodrug/αvβ3 interaction.

To quantify the amount of bioactive TNF loaded onto both nanodrugs, we then tested their cytotoxic effects against murine L-M fibroblasts, using TNF as a reference standard. Based on the biological effects observed, we estimate that the potency of each nanoparticle of 1 -Au/TNF and 2 -Au/TNF was equivalent to ∼11 and ∼14 TNF molecules, respectively (Table 1), suggesting that both conjugates were loaded with similar amounts of bioactive TNF. Overall, these and the above results suggest that the functionalization of nanogold with lipoamide-isoDGR peptide and TNF is feasible.