Bicyclic Peptides Conjugated to an Albumin-Binding Tag Diffuse Efficiently into Solid Tumors Free

Solid tumors present many barriers to efficient delivery of therapeutics, particularly to larger protein-based agents such as antibodies (1–3). The delivery to cancer cells that are situated distal from functioning blood vessels is challenging. If the activity of cancer therapeutics is based on antagonizing effects, and the therapeutics are unable to access all of the cells within a tumor, their effectiveness will be compromised. Many engineered immunoglobulin fragments of varying size, valence, and pharmacokinetic property have been explored (4). Size and circulation half-life were the factors that most strongly influenced extravasation and tissue penetration (5–10). Large ligands such as antibodies have a long plasma half-life, and thus achieve over time relatively high concentrations in tumors but they tend to distribute heterogeneously and do not access all regions in tumor tissue. In contrast, antibody fragments, small protein scaffolds and peptides show increased vascular permeability, penetrate more rapidly into tumors, and distribute more evenly but they also clear more rapidly from blood due to their lower molecular weight, resulting in lower concentrations in tumors (9–11).

A promising avenue to overcome the tissue delivery limitations of small proteins and peptides may be a “piggyback” strategy in which small polypeptide ligands are tethered noncovalently via an affinity tag to serum albumin. Extending the circulation time allows the polypeptide ligands to diffuse into tissues over time. The ligands can dissociate from the large albumin (66.5 kDa) and may diffuse into spaces such as cavities that are not accessible to large protein-based ligands. This strategy has been applied to enhance the accumulation of antibody fragments such as Fab, scFv, or single-chain diabodies in tumors (12–16). The much prolonged half-life resulted in a greatly improved targeting. In a study of Dennis and colleagues (13) applying a Fab linked to an albumin-binding peptide, a more homogeneous distribution compared with IgG was reported. A limitation of the approaches based on antibody fragments is that, after dissociation from albumin, the ligands are still relatively large (25–50 kDa) and can most likely not diffuse to all regions of tumor tissue. A significant volume of the intercellular space is known to be not accessible to albumin (17). We reasoned that the potential of the “piggyback” strategy may be even better exploited if the ligands were significantly smaller.

Our laboratory is developing bicyclic peptides that bind with high affinity and selectivity to therapeutic targets, including proteases and receptors implicated in tumor growth and/or invasion (18, 19). Bicyclic peptides contain two macrocyclic rings that interact with protein targets much like complementarity determining regions of antibodies, yet they are nearly 100-fold smaller than antibodies (∼2 kDa; ref. 20). The bicyclic peptides have a limited conformational flexibility and are relatively stable in vivo. Bicyclic peptides binding with high affinity and specificity to targets of interest can be generated by phage display (19). In brief, billions of peptides having random sequences are displayed on the surface of filamentous phage, chemically cyclized and panned against immobilized protein targets (21). Because of their small size, bicyclic peptides are rapidly filtered out by the kidney, and delivery of large bicyclic peptide quantities into (tumor) tissue poses a great challenge (22).

In this work, we proposed to exploit albumin-binding peptides to achieve high concentrations of bicyclic peptide in tumor tissue. The small size of bicyclic peptides promises efficient diffusion within tissue and good access to protein targets. In a recent work, we conjugated a bicyclic peptide to the albumin-binding peptide SA21 developed by Dennis and colleagues (K_d for mouse albumin = 24 nmol/L; ref. 23) and showed that its plasma half-life was prolonged from 30 minutes to 24 hours in mice (22), but we did not investigate if the bicyclic peptide diffuses into tissue. Herein, we were interested to learn which concentrations can be reached and if this molecule format would access all regions of a tumor as well as distribute homogeneously. Moreover, we wanted to investigate whether bicyclic peptides in tissue remain functional.

As a model, we used the bicyclic peptide UK18 inhibiting the serine protease urokinase–type plasminogen activator (uPA; K_i = 53 nmol/L; ref. 20), or the recently engineered derivative UK202 being slightly more active and more resistant to proteolysis (K_i = 28 nmol/L; ref. 24). The target of this bicyclic peptide, the protease uPA, was reported to play a role in tumor growth and invasion (25). In a recent study performed in our laboratory, the bicyclic peptide inhibitors UK18 and UK202 did not significantly reduce the growth rate of human MDA-MB-231 xenograft tumors in mice (unpublished observation), and this result raised the question whether the bicyclic peptides reached sufficiently high concentrations in the extracellular space of solid tumors to inhibit uPA. UK18 conjugated to the albumin-binding peptide SA21 (UK18-SA21) inhibits the protease with a K_i of 132 nmol/L in the presence of mouse albumin (22). On the basis of the rather weak affinity and the relatively low expression level of uPA in tumors, we expected that the distribution of the bicyclic peptides UK18 and UK202 in mice is not influenced by their binding specificity and that their tissue distribution properties are representative for other bicyclic peptides. As a tumor model, we used MDA-MB-231 human breast cancer xenografts in mice. This tumor model was applied in the above mentioned therapy study, and we were thus interested to learn how well bicyclic peptides diffuse into this type of tumor. MDA-MB-231 xenograft tumors contain hypoxic regions (26) and this allowed to study diffusion also into weakly vascularized tissue zones.

Fmoc-protected amino acids and Rink Amide MBHA resin were purchased from GL Biotech. (S)-N-Fmoc-propargylglycine (Shanghai Plus Bio-Sci & Tech), 5-azido-pentanoic acid (Bachem), O-Benzotriazole-N,N,N′,N′-tetramethyl-uronium-hexafluoro-phosphate (HBTU, GL Biotech), N,N-Diisopropylethylamine (DIPEA, Iris Biotech GmbH), trifluoroacetic acid (TFA; Sigma-Aldrich), 1,2-ethanedithiol (EDT; Fluka), thioanisole (Fluka), piperidine (Fluka), phenol (Acros Organics), thioanisole (Fluka), piperidine (Fluka), phenol (Acros Organics), and 1,3,5-tris(bromomethyl)benzene (TBMB; Sigma-Aldrich) were used as received without further purification.

Peptides were synthesized on an Advanced ChemTech 348Ω peptide synthesizer (AAPPTec) by standard Fmoc (9-fluorenylmethyloxycarbonyl) solid phase chemistry using Rink Amide AM resin (0.03 mmol scale). The coupling step was carried out twice for all amino acids using 0.12 mmol Fmoc-protected amino acid, 0.12 mmol HBTU, 0.12 mmol HOBt, and 0.18 mmol DIEPA in 1.3 mL DMF. Fmoc groups were removed using a 20% (v/v) solution of piperidine in dimethylformamide (DMF; 2 × 2.5 mL). The peptides were deprotected and cleaved from the resin by treatment with 5 mL of cleavage solution (90% TFA, 2.5% phenol, 2.5% thioanisole, 2.5% EDT, and 2.5% water), for 3 hours at room temperature (RT) under shaking. The resin was removed by centrifugation 1 minute at 1,000 rpm and filtration under vacuum. The peptides were precipitated by addition of ice-cold diethyl ether (50 mL) and incubation for 30 minutes at −20°C and centrifuged 5 minutes at 4,000 rpm. The precipitated peptides were washed twice by adding ice-cold ether (35 and 20 mL, respectively) and centrifugation 5 minutes at 4,000 rpm. Samples were resuspended in water/ACN (ratio 5:2) containing 0.1% TFA and lyophilized.

Linear UK18-alkyne and UK202-alkyne were cyclized with TBMB as follows. To a solution of crude peptide (8 mL, 0.625 mmol/L) in aqueous buffer (20 mmol/L NH₄HCO₃, pH 8.0), TBMB was added (2 mL, 5 mmol/L in ACN). The final concentrations in the reaction mixture were 0.5 mmol/L crude peptide, 1 mmol/L TBMB, 80% v/v aqueous buffer, and 20% v/v ACN. The reaction mixture was incubated at 30°C in a water bath for 1 hour and then lyophilized. TBMB-modified peptide was purified by preparative reversed-phase HPLC (high-performance liquid chromatography) as described below. Fractions containing the desired peptide were lyophilized. Pure peptide (10 mg) was typically obtained from 40-mg crude peptide.

Linear SA21-linker(G)-azide and SA21-linker(KG)-azide were cyclized by disulfide-bridge formation as follows. Air was bubbled for 3 hours through a filter-sterilized oxidizing solution of 95% v/v 20 mmol/L NH₄HCO₃ buffer, pH 8.0, and 5% v/v DMSO at RT. Crude peptide (40 mg) was dissolved in 20 mL of this buffer and incubated for 3 days at room temperature. After lyophilization, disulfide-cyclized peptide was purified by preparative reversed-phase HPLC as described below and the product lyophilized. Pure peptide (10 mg) was typically obtained from 40-mg crude peptide.

Peptides were purified by reversed-phase chromatography on a preparative C18 column (Vydac C18, 218TP column, 22 × 250 mm) using a solvent system of 99.9% H₂O/0.1% TFA and 99.9% ACN/0.1% TFA (20 mL/min, linear gradient over 27 minutes).

The carboxylic group of Alexa 594 (kindly provided by Prof. Kai Johnsson and Dr. Luc Reymond, EPFL, Switzerland) was activated by converting it to an N-hydroxysuccinimide ester as follows. A solution of Alexa 594 (10 mmol/L final concentration in dry DMSO) was mixed with TSTU-NHS (Fluorochem; 12 mmol/L final concentration in dry DMSO) in presence of DIPEA (25 mmol/L final concentration in dry DMSO) and incubated at RT for 30 minutes. The reaction was purified by preparative reversed-phase HPLC as described above. The fraction containing Alexa 594-NHS was isolated and lyophilized.

Alexa 594-NHS was conjugated to disulfide-cyclized SA21-linker(KG)-azide as follows. To a solution of SA21-linker(KG)-azide (2 mmol/L, 1 eq.) and Alexa 594-NHS (12 mmol/L, 6 eq.) in dry DMSO (0.5 mL), 2.5% (v/v) DIPEA was added. The reaction mixture was kept at RT for 2 hours and the product purified by preparative reversed-phase HPLC as described above. Fractions containing the desired peptide were lyophilized. Pure peptide–Alexa conjugate (6 mg) was typically obtained from 10-mg SA21-linker(KG)-azide.

The peptides were linked together by copper-catalyzed azide-alkyne cycloaddition as follows. 1 mmol/L alkyne-functionalized peptide (cyclic UK18-alkyne and UK202-alkyne) and 1 mmol/L azide-functionalized peptide (cyclic SA21-linker(G)-azide and SA21-linker(KG)-Alexa 594-azide) were reacted in the presence of 2 mmol/L CuSO₄·5H₂O, 10 mmol/L sodium ascorbate, 30% (v/v) H₂O mQ, and 70% (v/v) DMSO, for 2 hours at RT. The product was purified by preparative reversed-phase HPLC as described above, and fractions containing the desired peptide were lyophilized. The mass of the conjugates was confirmed by ESI-MS. Pure peptide (3 mg) was typically obtained from 6-mg peptide mixture.

The inhibition assay was performed as previously described (20). In brief, the cleavage of the fluorogenic substrate Z-Gly-Gly-Arg-AMC (50 μmol/L) by uPA (4 nmol/L) in reaction buffer [10 mmol/L Tris-Cl, pH 7.4, 150 mmol/L NaCl, 10 mmol/L MgCl₂, 1 mmol/L CaCl₂, and 0.1% (w/v) casein] was monitored for 15 minutes at 25°C in the presence or absence of inhibitor (2-fold dilutions from 1 μmol/L to 5 nmol/L UK202-SA21^Alexa). Optionally, 1.5 μmol/L mouse albumin was added to the reaction. Fluorescence was excited at 365 nm and emission recorded at 467 nm.

The MDA-MB-231 human breast carcinoma cell line was obtained from the cell bank Cell Lines Service (CLS) less than 6 months before the cell culturing experiments were started. The cell line was authenticated by CLS and tested negative for mycoplasma, bacteria, and fungi. MDA-MB-231 cells were cultured in RPMI-1640 medium (52400041; Lifetech) with 10% FBS (DE14-801F; Lonza), 1% NEAA (BE13-114E; Lonza), 0.5% penicillin/streptomycin (09-757F; Lonza). For 150 cm² flasks, 40-mL medium was used. Cell medium was changed every other day and cells were split every 4 days. Cells were split four times before the cultures were expanded for cell production. When the final flasks reached 95% confluence, cells were detached through addition of 1 mL of trypsin-EDTA (T4049-100ML; Sigma-Aldrich) and collected in 10 mL RPMI complete medium. Cells were centrifuged for 2 minutes at 1,000 rpm and resuspended in 20 mL RPMI complete medium and counted with a hemocytometer. Cells were centrifuged for 2 minutes at 1,000 rpm and resuspended in PBS to obtain 3 × 10⁶ cells per 100 μL.

Animal experimentation was performed according to the Swiss law for animal protection. Female BALB/c^nu/nu mice (4 weeks, between 18 and 22 g) were purchased from Charles River Laboratories (L'Arbresle Cedex). Mice were anesthetized with isoflurane and 3 × 10⁶ MDA-MB-231 cells in 100 μL PBS (14190-094; Life Technology) were injected s.c. into the right flank (day 0). Tumors had a diameter of 8 to 10 mm after around 30 days.

Peptides were iodinated using Pierce Iodination Beads (prod 28665; Thermo Fisher Scientific) according to the manufacturer's instructions. In brief, two beads were washed with PBS and dropped into a reaction tube containing 500 μCi of Na¹²⁵I (around 5 μL of NEZ033A00 in 10⁻⁵ mol/L Na; PerkinElmer) in 200 μL PBS (pH 7.4). The reaction was incubated for 15 minutes at RT to allow oxidation of iodide species into ICl. Each peptide (100 μg; UK18 or UK18-SA21) in around 20 μL H₂O was added to the tube and the reaction was incubated for 5 minutes at RT. H₂O (1.5 vol) containing 0.1% TFA was added and the mixture incubated 10 minutes at RT. Excess iodine species was removed chromatographically (disposable C18 column; Sep-Pak Cartridges WAT054955; Waters). The column was washed with 1-mL ACN and equilibrated with 2 × 1 mL of 5% ACN, 95% PBS (pH 7.4). The sample was added to the column and the flow through kept for radioactivity counting. The column was washed twice with 1 mL of 5% ACN 95%, PBS (pH 7.4), 0.1% TFA, and twice with 1 mL of 5% ACN 95%, PBS (pH 7.4). The sample was finally eluted in 800 μL 50% PBS (pH 7.4), ACN 50%. Acetonitrile was removed by evaporation using a vacuum pump (for UK18) or filtration (3 kDa cutoff Amicon Ultra filter 0.5 mL, UFC500324; Millipore; for UK18-SA21). The radioactivity of the samples and flow-through fractions were measured with a γ-counter (Wallac wizard. 1470; PerkinElmer). The labeling and purification procedures yielded typically 18% of UK18 and 25% of UK18-SA21 as assessed with nonradioactive iodine and analytic HPLC chromatography. The fraction of free iodine in the purified peptide samples was typically 10% for UK18 and 10% for UK18-SA21 as assessed by applying identical labeling procedures without peptide.

Female BALB/c^nu/nu mice (4–6/group) bearing subcutaneous human MDA-MB-231 tumor (8–10-mm diameter) whose thyroids were previously saturated by addition of 3 drops of Lugol solution (62650-100ML-F; Sigma-Aldrich) in water, were injected i.v. with equimolar quantities of radiolabeled peptide UK18 (5 μg in 100 μL PBS pH 7.4, 25 μCi) or UK18-SA21 (10 μg in 100 μL PBS pH 7.4, 30 μCi). After 4 or 24 hours, the animals were anesthetized with 200 μL of 12.5 mg/mL ketamine, 0.05 mg/mL, dorbene (i.p.; Graeub AG), transcardially perfused with 30 mL Ringers lactate buffer (Braun AG) containing 1 g/L procaine, 1 g/L glucose, 10 mmol/L HEPES, 1 mL/L penicillin–streptomycin–amphotericin-b (Gibco, Life Technologies), 5 mL/L heparin (14-gauge blunt-needle; gravity pressure, RT). Tumor and organs were removed, weighted, and the radioactivity measured with a γ-counter (Wallac wizard 1470). The %ID per gram of organ was calculated.

Three female BALB/c^nu/nu mice bearing established MDA-MB-231 tumors (8–10-mm diameter) were injected i.p. either 300 μg of UK202-SA21^Alexa in 200 μL PBS (pH 7.4; mouse 1 and 2) or PBS only (mouse 3). After 24 hours, all mice were transcardially perfused as described above and tumor and organs excised. The tumors and organs from all mice were fixed in 4% PFA (w/v) in PBS (pH 7.4) overnight at 4°C. The next day, they were rinsed twice by transferring to PBS, incubated around 48 hours (or until they sink) in 20% sucrose in PBS (pH 7.4) and embedded in optimal cutting temperature (OCT) compound. Organs were sectioned (8 μm) in a cryostat, mounted on Superfrost glass slides, and dried for at least 30 minutes. DNA was stained by incubation with DAPI (1 mg/mL stock, diluted 1:10,000 in PBS buffer; D9542; Sigma-Aldrich) for 10 minutes at RT. Slides were mounted with a glycerol-based mounting medium using DABCO (D2522; Sigma-Aldrich) as antifading agent. Tissues were imaged using the ×20 UPLSAPO objective on an Olympus slide scanner VS120-L100 microscope. Fluorescence was measured with EX440-500 (for DAPI), BA525-30 (for autofluorescence) and EX607-36 (for Alexa 594) filter sets, respectively. Microscope settings were kept constant for the same type of experiment and corresponding controls. Exposure times are provided in Supplementary Table S1. Sections were also analyzed by confocal microscopy (Zeiss LSM 700 upright confocal microscope). A ×63 oil Plan-Achromat objective was used. DAPI was recorded with a 403 laser and a DAPI filter. Alexa 594 was recorded with a 555 nm laser and a rhodamine filter. Laser intensity and gain were set to obtain optimal contrast and were kept constant for images within the same experiment.

Bicyclic peptide was detached and lost to some extent in the immunofluorescence staining procedure. UK202-SA21^Alexa in tissue sections was, thus, imaged before staining of blood vessels as described above. For immunostaining, sections of tumors and organs were dried for 30 minutes at RT and rehydrated in PBS. The sections were briefly equilibrated with antigen retrieval buffer (10 mmol/L tri-Na citrate, pH 6) and then subjected to heat-induced epitope retrieval for 20 minutes at 95°C using Lab Vision PT Module. The sections were blocked by incubation with 1% BSA in PBS for 30 minutes. Monoclonal rat anti-mouse CD31 antibody (clone SZ31, DIA-310, diluted 1:50 in 1% BSA; Dianova) was then added and incubated overnight at 4°C under constant agitation. Sections were washed three times with PBS (pH 7.4), incubated with biotinylated donkey anti-rat antibody diluted 1:400 in 1% BSA (712-065-153, Jackson Immunoresearch) for 40 minutes at RT under agitation, and washed three times with PBS. Streptavidin–Alexa 488 1:800 in 1% BSA (S-11223; Life Technologies) was added to the slides and incubated for 30 minutes at RT under agitation. Sections were washed three times with PBS and finally counterstained with DAPI for 10 minutes at RT (DAPI 1 mg/mL stock; Sigma-Aldrich; D9542, diluted 1:10,000 in PBS). Slides were mounted with a glycerol-based mounting medium using DABCO (D2522; Sigma-Aldrich) as antifading agent. New images were made with the ×20 UPLSAPO objective on an Olympus slide scanner VS120-L100 microscope as described above (exposure times are indicated in Supplementary Fig. S1).

Peptide concentrations in tumor tissue sections were calculated on the basis of fluorescence intensities measured by confocal microscopy. A standard curve relating peptide concentration and fluorescence intensity was prepared by taking images of serial dilutions of UK202-SA21^Alexa in a solution composed of 50% mounting media and 50% PBS. Pictures of random regions of the tumor sections were taken with the same microscopic parameters and a standard curve generated to relate peptide concentration and fluorescence intensity (Supplementary Fig. S6). Fluorescence intensity was measured in the regions of the sections in which tissue was present. This was achieved by creating a threshold map of the tissue based on the DAPI channel (the threshold was fixed as percentile 0.5 of the maximal fluorescence intensity) and measuring the fluorescence of the red channel on this selection. Nine Z-stacks with a distance of 160 nm were recorded and the one with the most intensive fluorescence signal used for analysis. The number of voxels (volumetric pixel) found for each fluorescence intensity was represented in a graph.

Mice were injected with 300 μg UK202-SA21^Alexa (mouse 4) or PBS (mouse 5), perfused 24 hours later, and tumor and organs removed. Tumor and organs were homogenized with a pestle at 4°C in 2 volumes of buffer containing 100 mmol/L Tris pH 7.4, 150 mmol/L NaCl, 100 μmol/L bestatin, 20 μmol/L E-64, 10 mmol/L EDTA, 100 μmol/L leupeptin, 20 μmol/L pepstatin A, 1 mmol/L phenylmethylsulfonylfluoride and 1.5 μmol/L aprotinin. The lysate was centrifuged 13,000 rpm for 1 minute at 4°C. The extract was aliquoted and 20 μL of tumor extract, 30 μL of liver extract, and 5 μL of kidney extract were analyzed by RP-HPLC (Agilent 1260 HPLC system Agilent) using a C18 column (Agilent ZORBAX 300SB-C18 4.6 × 250 mm 5 μm). A linear gradient with a mobile phase composed of eluent A (94.9% v/v H₂O, 5% v/v ACN and 0.1% v/v TFA) and eluent B (94.9% v/v ACN, 5% v/v H₂O and 0.1% v/v TFA) from 0% to 50% in 30 minutes was applied at a flow rate of 1 mL/min. Fluorescence was monitored with a Shimadzu RF-10AXL detector (Ex., 590 nm and Em., 630 nm).

Radiolabeled bicyclic peptide UK18 was applied i.v. to tumor-bearing mice and the amount of peptide that reached tissue of tumor and organs was quantified. Two peptide formats were compared: bicyclic peptide UK18 and UK18 conjugated to the albumin-binding peptide SA21 (UK18-SA21; Fig. 1A). UK18 was synthesized as described before (20) and UK18-SA21 was prepared by synthesizing UK18 and SA21 individually and conjugating them by azide-alkyne Huisgen cycloaddition (Supplementary Fig. S1). UK18 (5 μg) and UK18-SA21 (10 μg) labeled with around 30 μCi I¹²⁵ at the tyrosine residue were injected i.v. into nude mice bearing a subcutaneous human MDA-MB-231 tumor (8–10-mm diameter, 4–6 mice/group). After 4 and 24 hours, blood samples were taken and the blood circulation of mice perfused to remove the peptide in the blood vessels of organs. The fading of the red color in the liver indicated efficient removal from this organ and suggested that most blood was removed from the vessels in tumor and other organs. UK18 was found in ample quantity in tumor tissue 4 hours after administration (2.5 ± 0.2 %ID/g) but after 24 hours, nearly all was cleared (<0.029 ± 0.004 %ID/g; Fig. 1B and C). In contrast, UK18-SA21 was found in significant quantities at both time points (3.8 ± 1 %ID/g and 1.5 ± 0.3 %ID/g). In tissue of organs, UK18-SA21 was also found in higher quantities than UK18 wherein the difference was again much larger at the 24 hours time point. UK18 was found at levels between 0.3 and 6 %ID/g 4 hours after injection and was entirely cleared after 24 hours (0.01–0.22 %ID/g). The quantity of UK18-SA21 ranged between 1 and 11 %ID/g at the 4 hours time point and remained relatively high after 24 hours (0.4–3.8 %ID/g). The highest quantities of UK18 and UK18-SA21 were found in kidney, indicating renal clearance for both formats.

The spatial distribution of bicyclic peptide in tumors was investigated by microscopic analysis of tissue sections of mice injected with fluorescence-labeled peptide. For this experiment, a bicyclic peptide derivative of UK18 that was newly engineered in the meantime, UK202 (24), was applied. UK202 has a 1.7-fold higher binding affinity for uPA and a 4-fold better stability toward proteases compared with UK18. Alexa 594 was conjugated to the linker connecting the two peptides UK202 and SA21 (Fig. 2A and Supplementary Fig. S2). The conjugate UK202-SA21^Alexa inhibited human uPA with a K_i of 63.1 ± 8 nmol/L in the presence of mouse albumin. UK202-SA21^Alexa (300 μg) was injected into two mice carrying subcutaneous tumor MDA-MB-231 (8–10-mm diameter; mouse 1 and 2). A third mouse with the same tumor was injected with PBS as control (mouse 3). Because of the relatively large dose and the limited solubility, the peptide conjugate was injected i.p. After 24 hours, mice were perfused to clear peptide from the circulation and the organs were removed and fixed. A procedure, including a paraformaldehyde fixation step, was established to immobilize peptides and to prevent their dissociation from the tissue during cryosectioning, DAPI staining, and washing. The entire tumor sections were scanned with a fluorescence microscope at high magnification (×20 objective) to detect UK202-SA21^Alexa. After microscopy, the blood vessels were visualized with anti-CD31 antibody.

The fluorescence micrographs of mice that were injected with peptide and perfused before tumor and organ removal are shown in Fig. 2 and Supplementary Fig. S3, and those of the control experiment in Supplementary Fig. S4 (no peptide injected). The left panels in all figures show the peptide (red) and cell nuclei (blue); the right panels show the anti-CD31 antibody (green) and cell nuclei (blue). An overview of the whole-tumor sections (mouse 1 in Fig. 2B, mouse 2 in Fig. 2C) and three regions with higher magnification are shown (mouse 1 in Fig. 2D–F, mouse 2 in Supplementary Fig. S3). Tissue sections of the mouse not injected with peptide showed almost no fluorescence in the red fluorescence channel (mouse 3; Supplementary Fig. S4). In sections of mice injected with UK202-SA21^Alexa, potential autofluorescence was measured in a green fluorescence channel and identical microscopy settings (Supplementary Fig. S5). These control experiments showed that essentially all signal detected with the red filter was derived from the peptide. The anti-CD31 allowed localization of blood vessels. In the lumen of some small blood vessels, fluorescence was detected that lined the epithelial cells, indicating that the perfusion procedure did not remove all peptide in the vasculature. In larger blood vessels no fluorescence was detected showing that overall, peptide was efficiently cleared from the blood circulation in the perfusion procedure.

Bicyclic peptide was present in all regions of the tumors, including necrotic zones (Fig. 2B and C, left). Particularly high-fluorescence intensities were found in the skin and around vascularized regions. Weaker but clearly detectable fluorescence signal was found in regions distant from blood vessels. In all regions of the tumors, the fluorescent peptide was distributed heterogeneously as shown in figures with higher resolution (mouse 1 in Fig. 2D–F and mouse 2 in Supplementary Fig. S3). Significant fluorescence intensity was found in close distance (<1 cell diameter) of nearly every cell.

Regions of 86 × 86 μm were chosen within the tumor sections and imaged by confocal microscopy with at higher resolution (×63 objective). Peptide was distributed as shown in the representative photomicrographs in Fig. 3A and B. In many regions, the bicyclic peptide appeared to be located in the extracellular space (e.g., Fig. 3A). In some regions, speckles around cell nuclei were observed and indicated that the peptide was located in vesicles inside cells (e.g., Fig. 3B). Z-stacks of tumor sections were recorded and 3D images generated to localize the fluorescent peptide more precisely. Figure 3C and D show two types of peptide distributions that were observed. In the first figure, peptide appears to be bound to filamentous extracellular structures as well as to occupy the cytosolic space of cells (Fig. 3C). In the second figure, peptide appears to be trapped in vesicular structures inside cells (Fig. 3D).

The concentration of peptide in tissue was estimated on the basis of the fluorescence intensity measured by confocal microscopy. Fluorescence intensity was related with peptide concentration by preparing different dilutions of peptide in mounting media and measuring the fluorescence intensity (a standard curve is shown in Supplementary Fig. S6). Three to five random regions (86 × 86 μm) in tumor sections of mouse 1 and 2 (injected with UK202-SA21^Alexa) as well of mouse 3 (negative control not injected with peptide) were recorded and the number of voxels was plotted against the fluorescence intensity (Fig. 3E). Data from the two mice injected with fluorescent peptide showed a similar distribution, peaking at a fluorescence intensity that corresponds to a concentration of around 400 nmol/L (Fig. 3F). The average fluorescence intensity of voxels from the negative control mouse, originating from autofluorescence, was significantly lower (Fig. 3F, mouse 3). On the basis of the fluorescence micrographs of mouse 1 and 2, the average concentration of peptide in tumor tissue was estimated to be in the high nanomolar range. For around 10% of the voxels, the intensity corresponded to a concentration of more than 1 μmol/L.

Kidney and liver were removed from a mouse injected with UK202-SA21^Alexa (mouse 1) and from one not injected with peptide (mouse 3), and sections microscopically analyzed (Fig. 4). In the kidney, the perfusion procedure had efficiently cleared peptide from large and small blood vessels (Fig. 4A–C). Strong fluorescence signal was observed in cells of the proximal tubules. Confocal micrographs showed intense speckles inside cells of the proximal tubules (Fig. 4C). Fluorescence micrographs were recorded at 18-fold shorter exposure times than the sections of tumors described above. Taking the different microscopy setting into account, significantly more peptide was delivered to the kidney than to tumor, as found in the biodistribution experiment. The sections of liver showed high autofluorescence derived from the cytoplasm of hepatocytes (Fig. 4D–F; autofluorescence was recorded in the green fluorescence channel and is shown in green). Peptide was detected in capillaries surrounding the hepatocytes, indicating that perfusion did not clear it completely. In contrast with tumor and kidney, no peptide was detected inside cells.

The integrity of UK202-SA21^Alexa in tissues was assessed by peptide extraction and chromatographic analysis. Two tumor-bearing mice were injected with either 300 μg UK202-SA21^Alexa (mouse 4) or PBS (mouse 5), their circulation was perfused after 24 hours, and their organs and tumors were removed and homogenized. Extracts were analyzed by RP-HPLC using a fluorescence detector. The negative control (mouse 5) was included to detect potential endogenous autofluorescent species. For peptide quantification, different quantities of UK202-SA21^Alexa were run as standards; they eluted at 14 minutes and showed a linear correlation between peptide quantity and fluorescence signal over a wide range (4–140 ng; Fig. 5A). UK202-SA21^Alexa in plasma taken 24 hours after injection (just before perfusion) was almost all intact (Fig. 5B). Tissue probes removed from the mice after perfusion were homogenized in the presence of protease inhibitors to reduce peptide degradation during the extraction process. In the extract from tumor, 74% of the fluorescent species was corresponding to intact UK202-SA21^Alexa (Fig. 5C). The remaining fluorescent species eluted at two earlier time points. The two peaks had symmetrical shapes, suggesting a single degradation product per peak. The small number of degradation products indicated limited proteolysis of the bicyclic peptide conjugate in the tumor. The identity of the degradation products could not be determined by mass spectrometry due to coelution of numerous nonfluorescent molecules derived from tumor tissue (Fig. 5C). Extracts from kidney and liver contained also intact UK202-SA21^Alexa as well as the same two degradation products. Of note, 7% and 58% of the fluorescent species corresponded to intact peptide in the two organs, respectively. The absolute quantity of intact UK202-SA21^Alexa extracted from 1 mg of tumor tissue was 6.7 μg. If this amount of peptide was hypothetically assumed to be homogenously distributed in the 1 mL volume tissue, it had a concentration of 1.2 μmol/L.

We were interested in finding a molecule format that is suitable to target proteins buried in tissues of solid tumors and organs. We tested a “piggyback” strategy in which small bicyclic peptides are noncovalently bound to the long-lived serum protein albumin, preventing fast renal clearance and potentially allowing diffusion into tissues. The strategy allows dissociation of the peptide from the large serum protein, and thus potential diffusion into small cavities that may not be accessible to other molecule formats such as antibodies. A central question in our study was whether a bicyclic peptide with albumin-binding tag would be able to distribute efficiently into tissue, or whether binding to albumin in blood would even hinder delivery to tissues.

The quantification of bicyclic peptide delivered to tumor tissue required differentiation between peptide in tissue and in blood vessels. A technical procedure, in which the circulation of mice was perfused before tumor and organ removal, was applied. We showed that most of the peptide in blood was cleared with this strategy and that predominantly peptide in tissue was subsequently analyzed. Conjugation of bicyclic peptide to an albumin-binding peptide improved much its diffusion into tissue. One day after injection, bicyclic peptide UK18 conjugated to the albumin-binding peptide SA21 was present at a 54-fold higher concentration than free bicyclic peptide UK18. The better delivery of UK18-SA21 to tissue resulted from the significantly longer circulation time (t_1/2 of around 24 hours vs. around 30 minutes). The strong effect was remarkable considering that peptide is most of the time tethered to albumin and that the albumin–peptide complex is >15-fold larger than the peptide alone. Evidently, a long circulation time is much more important than a small size for an efficient delivery into tissue.

In our study, the bicyclic peptide UK18-SA21 did not accumulate more in tumor compared with other tissues. This was expected because the binding affinity for uPA is weak and/or the concentration of uPA in tumor relatively low. This is in contrast with the studies of Dennis and colleagues and Neri and colleagues (13, 16), in which antibody fragments were strongly enriched in tumors. The low affinity for the target protein allowed in our work studying the delivery of bicyclic peptide into tissues of tumor and organs solely based on diffusion as free or albumin-bound peptide. The biodistribution study with radiolabeled UK18-SA21 showed that as much as 1.5% of the injected dose was present per gram of tumor 24 hours after injection, which was remarkably high. On the basis of conjugate solubility in aqueous buffer, up to 0.5 mg UK18-SA21 can be applied i.p. to mice (corresponding to around 20 mg/kg). At such a dose, average concentrations in the high nanomolar range might easily be achieved, assuming a similar biodistribution as observed with the low dose of 10 μg. Bicyclic peptides isolated from naive phage display libraries have typically dissociation constants in the medium-to-low nanomolar range (18), and they can be further affinity matured to bind with picomolar affinity. Antagonists based on the presented molecule format should, thus, be able to block efficiently targets in tumor tissue.

We did not study the mechanisms by which the peptide conjugate travels to the tumor. On the basis of the binding affinity of UK18-SA21 for mouse albumin (K_d = 14 nmol/L; ref. 22), >99.9% of the conjugate is at any time bound to albumin in the blood stream. Most likely, the vast majority of peptide extravasates and diffuses into tissue in its albumin-bound form. Some of the peptide conjugate may also cross the endothelial barrier by transient dissociation from albumin, diffusion, and rebinding to albumin. Given the noncovalent link between peptide and albumin, the bicyclic peptide can diffuse more freely compared with peptide ligands bound covalently to albumin. In our study, we did not compare directly the diffusion of the bicyclic peptide conjugate with that of an antibody or antibody fragment. We neither investigated whether the peptide conjugate diffuses into regions in the extracellular space that are not accessible to larger protein-based ligands. However, on the basis of the affinity for mouse albumin and the binding kinetics, it can be assumed that the peptide conjugate dissociates from albumin frequently. In the short moments being detached from albumin, the peptide might diffuse into cavities that are not accessible to large protein-based ligands such as antibodies.

The microdistribution of bicyclic peptide in tissue was studied by tagging it with a red fluorophore that allowed clear discrimination of peptide from autofluorescent components as shown in control experiments. Fixation of the peptide to tissue was challenging as it is small and contains a limited number of groups that can react with paraformaldehyde. It is likely that a fraction of the peptide not immobilized was washed away after cryosectioning and that the amount of peptide in tissue was underestimated. Control experiments showed that peptide detaching in the process of tissue section preparation did not rebind. Analysis by epifluorescence and confocal microscopy showed that UK202-SA21^Alexa was delivered to all regions of the tumors and revealed a heterogeneous distribution. Peptide was found in the extracellular space as well as inside cells. It is likely that most of the observed peptide is bound to albumin and that its distribution is much influenced by the distribution of albumin in tissue. The concentration of albumin in the interstitial space is lower than in serum but more than half of the body albumin is found in the extravascular compartment (27, 28). Albumin was reported to be enriched in tumors and this effect was exploited to deliver drugs specifically to tumors (29). The delivery of bicyclic peptides with albumin-binding function may be positively affected by albumin levels being elevated in tumor tissue. The confocal micrographs of tumor tissue sections were used to estimate the bicyclic peptide based on fluorescence intensities. Twenty-four hours after injection of 300 μg UK202-SA21^Alexa, the mean concentration of fluorescent peptide was estimated to be around 400 nmol/L. This result was in line with the results of the biodistribution experiment in which administration of 10 μg UK18-SA21 (labeled with I¹²⁵) was estimated to lead to an average concentration of 33 nmol/L 24 hours after injection.

Microscopic analysis of kidney sections from mice injected with UK202-SA21^Alexa showed a high fluorescence intensity in proximal tubular cells. Peptide and protein filtered through the pores of the glomerular membrane are typically reabsorbed in the proximal tubules and degraded. This process was clearly observed for UK202-SA21^Alexa. Reabsorption starts with endocytosis followed by endosomal transport and fusion with lysosomes. The speckles observed in cells of the proximal tubular cells are likely such vesicular compartments. If bicyclic peptides are to be used as ligands for radionuclide therapy or for the targeting of cytotoxic drugs, the concentration in tubular cells should be minimized to limit kidney toxicity. Although delivery to tumor is much better for UK202-SA21 than for the bicyclic peptide alone, also the exposure of kidney is much higher for the albumin-binding bicyclic peptide conjugate. Further studies may be required to assess which format is most suited for such applications. If a bicyclic peptide is used for antagonizing enzymes or receptors or for interfering with a protein–protein interactions, the substantial accumulation in kidney will most likely not be a limitation.

More than 70% of peptide extracted from tumor tissue was fully intact after 24 hours of circulation in a mouse. The peptide conjugate was most likely stabilized through binding to albumin. We found previously that tethering bicyclic peptides to albumin via SA21 increases much its proteolytic stability in blood in vitro and in vivo (22). The large quantity of intact peptide combined with a nanomolar K_i for uPA (63.1 ± 8 nmol/L for albumin-bound UK202-SA21^Alexa and 8.3 ± 3 nmol/L for free UK202-SA21^Alexa) suggests that the activity of the target protein uPA is efficiently blocked.

In summary, we showed that small bicyclic peptides with albumin-binding function reach high nanomolar concentrations in solid tumors. Peptide conjugate was found to distribute to all regions of tumor tissue. Most of the peptide remained fully intact and functional. The noncovalent association of the bicyclic peptide with albumin allows dissociation and further diffusion into small cavities that may not be accessible to large ligands. Given these qualities and the good binding properties for targets of interest, bicyclic peptides offer a promising format to target solid tumors.

C. Heinis has ownership interest (including patents) in Bicycle Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Pollaro, C. Heinis

Development of methodology: L. Pollaro, S. Raghunathan, J. Morales-Sanfrutos, A. Angelini, S. Kontos, C. Heinis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Pollaro, S. Raghunathan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Pollaro, C. Heinis

Writing, review, and/or revision of the manuscript: L. Pollaro, J. Morales-Sanfrutos, A. Angelini, S. Kontos, C. Heinis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Pollaro

Study supervision: C. Heinis

The authors thank Prof. Kai Johnsson and Dr. Luc Reymond for providing Alexa 594, Prof. Jeffrey Hubbell and Prof. Melody Schwartz for support with injection and perfusion procedures, Dr. Shiyu Chen for support in HPLC peptide analysis, Dr. Philippe Diderich, Beatrice Volpe, Dr. Vanessa Baeriswyl and Emmanuela Bovo for help with animal experimentation, Dr. Jessica Sordet-Dessimoz for advice and histologic sample preparation, Dr. Romain Guiet and Oliver Burri for advice and computational image analysis.

The financial contributions from the Swiss National Science Foundation (SNSF Professorship PP00P3_123524/1, SNSF Sinergia project 141945 and SNSF project 146794; to C. Heinis) and the EPFL are gratefully acknowledged.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.