Asparagine deamidation in peptides or in fibronectin fragments containing the asparagine-glycine-arginine sequence generates isoaspartate-glycine-arginine (isoDGR), a new αvβ3 integrin-binding motif. Because αvβ3 is expressed in angiogenic vessels, we hypothesized that isoDGR-containing peptides could be exploited as ligands for targeted delivery of drugs to tumor neovasculature. We found that a cyclic CisoDGRC peptide coupled to fluorescent nanoparticles (quantum dots) could bind αvβ3 integrin and colocalize with anti-CD31, anti-αvβ3, and anti-α5β1 antibodies in human renal cell carcinoma tissue sections, indicating that this peptide could efficiently recognize endothelial cells of angiogenic vessels. Using CisoDGRC fused to tumor necrosis factor α (TNF) we observed that ultralow doses (1–10 pg) of this product (called isoDGR-TNF), but not of TNF or CDGRC-TNF fusion protein, were sufficient to induce antitumor effects when administered alone or in combination with chemotherapy to tumor-bearing mice. The antitumor activity of isoDGR-TNF was efficiently inhibited by coadministration with an excess of free CisoDGRC, as expected for ligand-directed targeting mechanisms. These results suggest that isoDGR is a novel tumor vasculature–targeting motif. Peptides containing isoDGR could be exploited as ligands for targeted delivery of drugs, imaging agents, or other compounds to tumor vasculature. [Cancer Res 2008;68(17):7073–82]
It is well-known that tumors cannot grow beyond 2 to 3 mm without an adequate vascular supply. For this reason, tumors tend to recruit new blood vessels from the preexisting vasculature by neoangiogenesis (1, 2). This concept provides the rationale for developing new strategies for cancer therapy based on administration of drugs that inhibit the formation of new blood vessels (antiangiogenic drugs) or drugs that destroy or alter existing tumor blood vessels (vascular-targeting agents; ref. 3). In the last decade, many investigators have made efforts to identify molecules capable of interacting with receptors expressed in angiogenic vessels, in the attempt to generate new antiangiogenic drugs or to obtain ligands for targeted delivery of other drugs to tumors.
Panning of peptide-phage libraries in tumor-bearing mice has proven useful for selecting peptides able to interact with proteins expressed within tumor-associated vessels and, therefore, to home to neoplastic tissues (4). Among the various ligands identified thus far, peptides containing the asparagine-glycine-arginine (NGR) motif have been exploited for ligand-directed targeted delivery of various drugs and compounds to angiogenic vessels. For instance, we have shown that the therapeutic index of certain cytokines capable of affecting tumor blood vessels, such as tumor necrosis factor α (TNF) and IFNγ, can be improved by fusing their NH2 or COOH termini to peptides containing the NGR motif (5–7). In particular, the CNGRC-TNF fusion protein (called NGR-TNF) has improved vasculature-damaging properties and improved antitumor activity, compared with TNF (5, 8). We have also shown that administration of ultralow doses (picograms) of NGR-TNF, but not of TNF, exert synergistic antitumor effects with various chemotherapeutic drugs, such as doxorubicin, melphalan, cisplatin, paclitaxel, and gemcitabine, by altering drug-penetration barriers (7–11). NGR-TNF, alone and in combination with chemotherapy, is currently tested in phase II clinical trial.
We have recently shown that the NGR motif, which recognizes a CD13 isoform expressed in angiogenic vessels (12–15), can rapidly convert to isoDGR, by asparagine deamidation, and that this motif can interact with αvβ3, an integrin critically involved in angiogenesis (16–18). The transition of NGR to isoDGR can occur also in fibronectin, an extracellular matrix protein that contains four NGR sites, generating new αvβ3 integrin binding sites (16). Considering that αvβ3 integrin is a good marker of angiogenic vessels, natural or synthetic polypeptides containing isoDGR may be exploited, in principle, as ligands for targeted delivery of cytotoxic drugs or nanoparticles to angiogenic vessels in tumors.
To address this hypothesis, we have investigated the tumor vasculature homing properties of a CisoDGRC peptide coupled to fluorescent nanoparticles (quantum dots) or fused to TNF (isoDGR-TNF). We show that indeed CisoDGRC can recognize tumor vessels and that it can be exploited for targeted delivery of TNF to tumors, improving its therapeutic index.
Materials and Methods
Cell lines and reagents. Murine RMA lymphoma, murine WEHI-164 fibrosarcoma, and human EA.hy926 endothelial cell lines were cultured as described previously (6, 19). Human αvβ3 integrin, anti-αvβ3 monoclonal antibody (mAb) LM609, and anti-α5 integrin mAb PB16 were from Immunological Sciences. Murine TNF, NGR-TNF (consisting of murine TNF fused with the COOH terminus of CNGRCG), and DGR-TNF (consisting of TNF fused with the COOH terminus of CDGRCG) were prepared by recombinant DNA technology as described (5, 16).
Preparation and characterization of synthetic peptides. The NH2-terminal acetylated peptides prepared were as follows: CisoDGRCGVRSSSRTPSDKY, CRGDCGVRSSSRTPSDKY, CNGRCGVRSSSRTPSDKY, and CARACGVRSSSRTPSDKY (called CisoDGRC-hTNF1-11, CRGDC-hTNF1-11, CNGRC-hTNF1-11, and CARAC-hTNF1-11, respectively). These peptides consist of CisoDGRCG, CRGDCG, CNGRCG, or CARACG fused to the NH2-terminal sequence of human TNF (underlined, hTNF1-11) followed by a Tyr, to enable detection. These peptides contain only one amino group to enable conjugation to nanoparticles. Each peptide was synthesized and purified as described (20). All peptides were dissolved in sterile. The molecular mass of each peptide was checked by MALDI-TOF analysis.
Preparation of isoDGR-Qdot. Amine-modified Qdot nanoparticles [Qdot605 ITK Amino (PEG) Quantum Dots; Invitrogen] were activated with bis[sulfosuccinimidyl] suberate (BS3; Pierce), a homobifunctional crosslinker, according to the manufacturer's instructions. The N-hydroxysuccinimide-nanoparticles were purified from unreacted crosslinker by gel-filtration chromatography on NAP-5 column (GE Healthcare). Colored fractions were pooled (600 μL) and divided into two separate tubes. CisoDGRC-hTNF1-11 (160 μg in 32 μL of water) was then added to one tube, whereas the other tube was mixed with 32 μL of water alone. Both tubes were then incubated for 2 h at room temperature. The conjugates (called isoDGR-Qdot and Qdot) were then separated from free peptide by ultrafiltration (Ultra-4 Ultracel-100K; AMICON, Millipore), resuspended in 100 mmol/L Tris-HCl (pH 7.4), and stored at 4°C.
Binding assay of isoDGR-Qdot to αvβ3 integrin. Human αvβ3 integrin solution, 0.5 μg/mL in PBS with Ca2+ and Mg2+ (DPBS; Cambrex), was added to 96-wells black ELISA plate (NUNC; Maxisorb; 100 μL/well) and left to incubate overnight at 4°C. All subsequent steps were carried out at room temperature. The plates were washed with DPBS and further incubated with DPBS containing 3% bovine serum albumin (BSA; 200 μL per well; 1 h). The plates were washed with 25 mmol/L Tris-HCl, 150 mmol/L sodium chloride, 1 mmol/L magnesium chloride, 1 mmol/L manganese chloride (pH 7.4; washing solution), and filled with serial dilution of isoDGR-Qdot or Qdot solution (100 μL per well in washing solution containing 1% BSA). After incubation for 2 h, the plates were washed again as described above. The bound fluorescence was determined using a Victor Wallac3 instrument (excitation filter, F355 nm; emission filter, 595/60 nm).
Binding assay of isoDGR-Qdot to cells. Binding of isoDGR-Qdot or Qdot to human EA.hy926 cells and to murine WEHI-164 or RMA cells was carried out as follows: adherent cells were detached with 138 mmol/L sodium chloride, 2.7 mmol/L potassium chloride, 10 mmol/L sodium phosphate (pH 7.3; PBS), containing 5 mmol/L EDTA. After washing with PBS, the cells were resuspended in 25 mmol/L HEPES (pH 7.4), 150 mmol/L sodium chloride, 1 mmol/L magnesium chloride, 1 mmol/L manganese chloride, 1% BSA, 0.02% sodium-azide containing serial dilution of isoDGR-Qdot or Qdot solution (5 × 105 cell/200 μL tube) and left to incubate 2 h at 37°C, 5% CO2. After washing, cells were fixed with 2% formaldehyde in PBS and analyzed by fluorescence-activated cell sorting (FACS).
Binding assay of isoDGR-Qdot to tumor sections. Surgical specimens of human renal cell carcinoma were obtained from the Department of Histopathology, San Raffaele Scientific Institute. The tumor tissues were embedded in ornithine carbamyl transferase medium (Bio-Optica) cryosectioned at −20°C into slice of 6-μm thickness, and adsorbed on polylysine-coated slides. Tumor sections were incubated in 25 mmol/L HEPES (pH 7.4), 150 mmol/L sodium chloride, 1 mmol/L magnesium chloride, 1 mmol/L manganese chloride, and 3% BSA (binding buffer) for 1 h at room temperature. The solution was then removed and replaced with binding buffer containing isoDGR-Qdot or Qdot (1:2,500). After 2 h of incubation at room temperature, tissue sections were rinsed for 10 min with binding buffer and further incubated for 5 min with PBS containing 0.1 μg/mL 4′,6-diamidino-2-phenylindole (DAPI; Sigma). The slides were rinsed again and incubated for 10 min at room temperature with PBS containing 2% paraformaldehyde and 3% sucrose. After extensive washing with PBS, slides were examined under the microscope (Carl Zeiss, Axioscop 40FL; excitation filter, BP 560/40 nm; beam splitter filter, FT 585 nm; emission filter, 630/75 nm). Costaining experiments were performed as follows: tissue sections were incubated with isoDGR-Qdot or Qdot solution containing the mAb anti-CD31, or anti-αvβ3, or anti-α5β1, respectively (4 μg/mL), for 1 h at room temperature. The solution was then removed and replaced with isoDGR-Qdot or Qdot solution containing goat anti-mouse IgG antibody FITC-conjugated FITC (1:100). The slides were incubated for 1 h at room temperature, rinsed again, and processed as described above.
Purification and characterization of isoDGR-TNF by high performance liquid chromatography analysis. IsoDGR-TNF was isolated from NGR-TNF by reverse-phase (RP)–high performance liquid chromatography (HPLC) using a C-3 column (Zorbax 300SB-C3, 5-μm, 300Å, 250×9.4 mm; Agilent), connected in line with 2-μm column prefilter (Supelco), as follows: mobile phase A, 5 mmol/L sodium phosphate buffer (pH 6.8), containing 5% acetonitrile in water; mobile phase B, 5 mmol/L sodium phosphate buffer (pH 6.8), containing 70% acetonitrile in water; 0% B for 10 min, 30% B for 5 min, linear gradient 30% to 65% B in 35 min, 100% B for 10 min, 0% B for 10 min, and flow rate of 4 mL/min. HPLC fractions were partially dried using a Savant Speed-Vac System, to eliminate the organic phase, and stored at −80°C until further analysis. Protein purity and identity were checked by SDS-PAGE and electrospray mass spectrometry. The in vitro cytolytic activity of each HPLC fraction was measured by cytolytic assay with L-M mouse fibroblasts (21) using murine TNF as reference standard (National Institute for Biological Standards and Control).
Protein digestion with trypsin. NGR-TNF (20 μg), diluted in 100 μL of 50 mmol/L phosphate buffer and 150 mmol/L sodium chloride (pH 7.4), was incubated for 24 h at 37°C or at −20°C. Thirty microliters of trypsin-agarose (1:1 suspension; Sigma) were added to each sample. Protein digestion was carried out at room temperature under gentle agitation. After 3 h of digestion, the sample was centrifuged through a Spin-x filter (Corning Incorporated), to remove trypsin-agarose, and stored at −20°C until analysis.
In vivo studies. Studies on animal models were approved by the Ethical Committee of the San Raffaele Scientific Institute, and performed according to the prescribed guidelines. BALB/c (Harlan) or C57BL/6N mice (Charles River Laboratories), weighing 16 to 18 g, were challenged with s.c. injection in the left flank of 106 WEHI-164 or 7 × 104 RMA living cells; 5 to 10 d later, mice were treated i.p. with protein solution (100 μL). All proteins were diluted with 0.9% sodium chloride containing 100 μg/mL endotoxin-free human serum albumin (Farma-Biagini SpA). Tumor growth was monitored daily by measuring tumor volumes with calipers, as previously described (22). Animals were sacrificed before tumors reached 1.0 to 1.5 cm in diameter. Tumor sizes are shown as mean ± SE (5 to 12 animals per group).
IsoDGR-Qdot binds endothelial cells in culture and in tumor vessels. To investigate the tumor vasculature binding properties of CisoDGRC, we have coupled CisoDGRCGVRSSSRTPSDKY, a peptide consisting of CisoDGRC fused to the NH2-terminal sequence of human TNF (CisoDGRC-hTNF1-11), to amine-modified quantum dots. Given that hTNF1-11 residues do not interfere in the binding of CisoDGRC to integrins (16), this sequence was exploited as a spacer. This peptide-Qdot conjugate was called isoDGR-Qdot. In parallel, we prepared similar conjugates with the ARA sequence in place of isoDGR (ARA-Qdot) and quantum dots without peptide (Qdot), to be used as negative controls. To verify that CisoDGRC was functional after conjugation, we analyzed the binding of these conjugates to purified αvβ3 integrin absorbed onto microtiterplates. As expected, isoDGR-Qdot, but not of Qdot, could bind to αvβ3 integrin in a dose-dependent manner (Fig. 1A). Furthermore, no binding of ARA-Qdot was observed, suggesting that the isoDGR sequence was critical for binding. This also confirms that the binding was entirely dependent on the CisoDGRC domain and not the hTNF1-11 sequence added as a spacer.
To assess whether isoDGR-Qdot could also bind membrane-associated integrins, we tested the binding of isoDGR-Qdot to EA.hy926 cells, an endothelial cell line known to express isoDGR receptors (16). Fluorescence microscopy experiments showed binding of isoDGR-Qdot, but not of Qdot, to adherent EA.hy926 cells (Fig. 1B). FACS analysis of EA.hy926 cells after incubation with these conjugates confirmed that isoDGR-Qdot, but not Qdot, could bind these cells in a dose-dependent manner (Fig. 1C,, top). The binding was efficiently inhibited by free CisoDGRC-hTNF1-11 peptide, whereas no inhibition was observed with CARAC-hTNF1-11 (Fig. 1C , bottom). This suggests that the binding was mediated by the isoDGR sequence. Of note, the binding of isoDGR-Qdot was also inhibited by CRGDC-hTNF1-11, although less efficiently. This is in line with the hypothesis that integrin receptors were recognized by isoDGR-Qdot. To assess the role of CD13, the ligand of NGR, we also evaluated the binding of isoDGR-Qdot in the absence and in the presence of mAb WM15 (10 μg/mL), an anti-CD13 mAb able to inhibit the binding of NGR-peptides to CD13 (23). No inhibition of binding was observed (data not shown), suggesting that the interaction with EA.hy926 cells was primarily mediated by integrins.
Interestingly, little or no binding of isoDGR-Qdot was observed to RMA lymphoma cells (Fig. 1D), a cell line that do not express αv subunits (24), and to WEHI-164 fibrosarcoma cells (not shown), suggesting that these cells do not express isoDGR receptors.
Next, we studied the binding of isoDGR-Qdot to tumor vessels using human renal cell carcinoma sections. Fluorescence microscopy analysis of tissue sections with isoDGR-Qdot (red) and with an anti-CD31 antibody (green), a well-known endothelial marker, showed a good colocalization of red and green fluorescence (Fig. 2). In contrast, no binding to vessels was observed with Qdots (Fig. 2). The binding of isoDGR-Qdot to tumor vessels was efficiently inhibited by CisoDGRC-hTNF1-11 peptide, whereas no inhibition was observed with CARAC-hTNF1-11 (data not shown). These results suggest that CisoDGRC is a good ligand of tumor vessels. Interestingly, colocalization was observed also with isoDGR-Qdot and anti-αvβ3 or anti-α5β1 antibodies (Fig. 2). IsoDGR-Qdot could also bind vessels in murine RMA lymphoma tissue sections but not normal vessels in muscle tissue sections (data not shown). These data, together, suggest that the CisoDGRC peptide can bind specific receptors, likely integrins, within tumor vessels.
RP-HPLC of deamidated NGR-TNF separates different fractions (P2 and P3) that contain isoDGR-TNF and NGR-TNF. To assess whether the CisoDGRC motif can be exploited for targeted delivery of drugs to tumors, we prepared CisoDGRC-TNF fusion protein. Because this product cannot be produced by standard DNA technology, we tried to prepare it by deamidation of NGR-TNF and reverse-phase chromatography. RP-HPLC analysis of NGR-TNF revealed two minor peaks, called P1 and P2, and a major peak, called P3 (Fig. 3A,, top). Rechromatography of these peaks, by RP-HPLC, showed single peaks and no change in their retention time (Fig. 3A , bottom), indicating that the corresponding products were stable after separation. The molecular mass of P1, P2, and P3, as measured by electrospray mass spectrometry analysis, were 17,944.8 ± 1.19 Da, (n = 2), 17,844.7 ± 0.75 Da, (n = 5), and 17,843.7 ± 0.62 Da, (n = 4), respectively. Because the expected molecular weight of NGR-TNF is 17,844.2, these results suggest that P3 corresponds to NGR-TNF, whereas P1 and P2 correspond to modified products. Sequence analysis of P1 by Edman degradation showed that the NH2 terminus was modified, as no reaction with phenylisothiocynate could occur. P1, therefore, was discarded. The molecular mass of P2, which is only 1 Da greater than that of P3, suggests that this product corresponds to isoDGR-TNF and/or DGR-TNF, as Asn/isoAsp and Asn/Asp transitions are accompanied by an increase of 1 Da. To verify that indeed P2 and P3 correspond to isoDGR-TNF/DGR-TNF and NGR-TNF, respectively, we performed additional experiments.
First, incubation of NGR-TNF in conditions that favor Asn deamidation [i.e., at 37°C, for 4 hours (pH 8.5)] caused a rapid decrease of P3 and an increase of P2 (Fig. 3A , top). TNF, under the same conditions, was homogeneous and stable (data not shown), suggesting that the different chromatographic behavior of P2 and P3 was indeed related to the NGR domain.
Second, limited trypsin digestion of NGR-TNF, before and after heat treatment for 24 hours, produced two different peptides with monoisototopic masses (MH+) of 894.5 Da and 895.5 Da, respectively, corresponding to disulfide-linked CNGR∼CGLR and CDGR∼CGLR/CisoDGR∼CGLR peptides, cleaved at the RC bond (expected, 894.40 and 895.40 Da, respectively; Fig. 3B).
Third, NH2-terminal sequences of P2 and P3, as measured by the Edman degradation method, were XDGRXG and XNGRXG, respectively, where X is likely Cys, which cannot be determined by Edman degradation.
Fourth, P2 and P3 contained 0.70 ± 0.14 and <0.05 pmol isoAsp/pmol of protein, respectively, as measured by IsoQuant Isoaspartate Detection kit (Promega).
These results strongly suggest that P2 corresponds to a mixture of DGR-TNF and isoDGR-TNF, whereas P3 corresponds to NGR-TNF. Based on these findings, P2 was produced in larger amounts by incubating NGR-TNF for 4 hours, in 0.1 mol/L ammonium bicarbonate buffer (pH 8.5), at 37°C, followed by RP-HPLC. P3 was directly purified from untreated NGR-TNF by RP-HPLC.
Both P2 and P3 are trimeric proteins biologically active in cytolytic assays. Bioactive TNF is a homotrimeric molecule (25). To assess whether P2 and P3 could form bioactive trimers, both products were further characterized by biochemical and biological assays.
Nonreducing SDS-PAGE of both products showed bands corresponding to 17 to 18 kDa subunits (Supplementary Fig. S1A), whereas analytic gel-filtration chromatography showed peaks corresponding to proteins of 45 to 50 kDa, as expected for homotrimeric molecules (Supplementary Fig. S1B). To assess whether the TNF domain was properly folded and functional, we measured the cytolytic activity of P2 and P3 against L-M cell. The specific activities obtained were 6.4 ± 1.98 × 108 U/mg and 2.03 ± 0.18 × 108 U/mg, respectively. Because the specific activity of unmodified TNF is 2 × 108 U/mg, these results suggest that both products were biologically active.
We have previously shown that peptides containing CisoDGRC can bind αvβ3 integrin and promote endothelial EAhy.926 cell adhesion, whereas little or no binding occurs with peptides containing CDGRC or CNGRC (16). To assess whether the CisoDGRC domain of P2 was functional, we analyzed, therefore, its capability to bind αvβ3 and to promote cell adhesion in vitro. The results showed that P2, but not P3, could bind αvβ3 integrin (Fig. 3C) and could induce endothelial cell adhesion (Fig. 3D), as expected. In contrast, DGR-TNF, prepared by recombinant DNA technology, and TNF could neither bind αvβ3 nor promote cell adhesion (data not shown).
These results suggest that both targeting and effector domains of the isoDGR-TNF component of P2 are properly folded and functional.
Antitumor activity of P2 and P3 in the WEHI-164 fibrosarcoma model. The in vivo antitumor properties of P2 and P3 were then investigated using the WEHI-164 fibrosarcoma model. Animal treatment (i.p.) with ultralow doses of both products (1 or 10 pg) were sufficient to induce significant antitumor effects (Fig. 4A and B). Administration of higher doses (30–100 ng) or lower dose (0.2 pg) induced lower antitumor responses (data not shown), suggesting that dose-response curves were bell-shaped in both cases, with maximal activity in the 1 to 10 pg range. Remarkably, no significant effects were obtained with 1 pg of recombinant DGR-TNF or TNF (Fig. 4C). These results suggest that P2 and P3 can induce antitumor effects with similar potency, and that the effect observed with P2 was mainly due to the isoDGR-TNF component. These results also rule out the possibility that the activity of P2 was due to the presence of NGR-TNF contaminants (<1% by RP-HPLC).
No loss of body weight was caused by any tested dose of P2 or P3, suggesting that the antitumor effects described above were induced without causing major toxicity. Furthermore, no loss of body weight was observed when mice were treated with 100 μg/mouse/day of CisoDGRCGVRY peptide for 7 days, suggesting that the isoDGR peptide does not cause major toxicity either when used alone or coupled to TNF (data not shown).
Antitumor activity of P2 and P3 in combination with melphalan in the RMA lymphoma model. In previous work, we showed that administration of picogram doses of NGR-TNF (e.g., 0.1 ng) to RMA tumor–bearing mice increases the penetration of chemotherapeutic drugs in tumors, whereas similar doses of TNF were inactive (7). The antitumor activity of P2 and P3 in combination with melphalan was therefore tested in the RMA lymphoma model. A single administration of melphalan (90 μg) induced little or no effects (Fig. 5A, and B, left). When melphalan was administered to mice in combination with various doses of P2 (5–100 pg), we observed a synergistic effect, as expected (Fig. 5A,, left). Similar effects were also observed when animals were treated with melphalan in combination with various doses of P3 (4–100 pg; Fig. 5B,, left). No increase of melphalan-dependent toxicity (body weight loss) was induced by P2 or P3 (Fig. 5A and B, right), suggesting that both products could increase the antitumor activity of melphalan without increasing its toxicity. Synergy was observed also when melphalan (50 μg) was administered 30 min after P2 (not shown).
Also in this model, no antitumor effects were observed with DGR-TNF/melphalan (20 and 80 pg; Fig. 5C), confirming the hypothesis that the active component of P2 is isoDGR-TNF. Considering that also TNF is inactive in this model when administered at comparable doses (5), these results, together, suggest that fusion of TNF to CisoDGRC improved its antitumor activity, alone and in combination with chemotherapy.
Mechanism of action of P2. To assess whether the antitumor activity of P2 depends on targeted delivery of TNF to isoDGR receptors in tumors, we coadministered P2 (1 pg) with a molar excess of CisoDGRCGVRY (isoDGR-2C) peptide (100 μg) to WEHI-164 tumor–bearing mice. IsoDGR-2C inhibited most of the antitumor effects induced by P2 (Fig. 6A,, left). Similarly, isoDGR-2C could inhibit the activity of isoDGR-TNF in combination with melphalan in the RMA model (Fig. 6B), supporting the hypothesis that the improved activity of P2 was related to an isoDGR-dependent targeting mechanism. Remarkably, the same treatment did not affect the antitumor activity of P3 (Fig. 6A , right). These results are in line with the hypothesis that the receptors of isoDGR and NGR are different, i.e., integrins and CD13, respectively.
We have recently shown that fibronectin fragments containing the isoDGR motif (called isonectins) can bind integrins typically expressed by angiogenic vasculature, such as αvβ3 and α5β1 (16). This notion led us to hypothesize that peptides containing isoDGR could be exploited as ligands for targeted delivery of drugs and nanoparticles to angiogenic vessels.
The results of the current study suggest that isoDGR can indeed be exploited for this purpose. In particular, the results of binding studies obtained with a peptide containing the cyclic CisoDGRC sequence coupled with fluorescent Qdot nanoparticles (isoDGR-Qdot) show that this motif can bind not only purified αvβ3 integrin adsorbed on microtiter plates, as previously shown with isonectins, but also tumor vessels. The good colocalization of isoDGR-Qdot with anti-CD31, anti-αvβ3, and anti-α5β1 antibodies in tumor tissue sections suggests that activated endothelial cells of angiogenic vessels, known to express these markers, are important targets of isoDGR. Accordingly, isoDGR-Qdot could also bind EA.hy926 cells in vitro, an immortalized endothelial cell line that express αvβ3. The binding could be competed by peptides containing CisoDGRC as well as by peptides containing CRGDC, a well-known integrin binding motif (26). This suggests that isoDGR can bind the RGD binding site of integrins on the endothelial cell surface.
The concept that cyclic peptides containing isoDGR can be exploited as ligands for drug delivery to tumors is supported by the results of pharmacologic studies carried out in murine lymphoma and fibrosarcoma models, showing that the antitumor activity of a CisoDGRC-TNF fusion protein (isoDGR-TNF) is greater than that of TNF and of CDGRC-TNF (DGR-TNF), a similar fusion protein unable to interact with αvβ3. Of note, although DGR-TNF and TNF were produced by expression in Escherichia coli cells, isoDGR-TNF could not be prepared by standard recombinant DNA technology, as no codon/anticodon pairs exist in these cells for isoaspartate. Thus, we have developed a method for preparing isoDGR-TNF based on incubation of recombinant NGR-TNF at pH 8.5 for 4 hours (37°C), i.e., in conditions that are known to favor NGR deamidation and consequent formation of isoDGR and DGR in 3:1 ratio (16). Two main components, called P2 and P3, were separated from the product by RP-HPLC. The results of sequence analysis and biochemical characterization studies show that P2 corresponded to a mixture of isoDGR-TNF and DGR-TNF, i.e., the products of NGR deamidation, whereas P3 corresponded to NGR-TNF. P2 and P3 formed bioactive homotrimers, as indicated by the results of SDS-PAGE, mass spectrometry, gel-filtration chromatography, and cytolytic assays. Furthermore, P2, but not P3 (NGR-TNF) and DGR-TNF, could bind to αvβ3 and promote endothelial cell adhesion in vitro. These findings suggest that both targeting and effector domains of isoDGR-TNF (i.e., isoDGR and TNF) present in P2 were properly folded and functional. Moreover, the finding that ultralow doses of P2 (1–10 pg), but not of DGR-TNF, could induce antitumor effects in mice, implies that isoDGR-TNF was indeed the biologically active component of P2. The hypothesis that the improved activity of P2 was related to isoDGR-directed delivery of TNF to tumors is also supported by the observation that coadministration of P2 with an excess of an isoDGR-containing peptide inhibited its antitumor activity.
Interestingly, although the activity of purified isoDGR-TNF (P2) was inhibited by coadministration with an excess of a peptide containing CisoDGRC, the activity of purified NGR-TNF (P3) was not inhibited. This finding supports the hypothesis that isoDGR-TNF and NGR-TNF work via different receptors, likely integrins and CD13 (12, 16). These results, together, suggest that isoDGR is a novel tumor vasculature-targeting motif.
The results of the present work could also have important pharmacologic implications for NGR-drug conjugates previously developed. Various compounds and particles have been coupled to NGR peptides in an attempt to increase their neovasculature-homing attributes, including cytotoxic drugs, cytokines, antiangiogenic compounds, viral particles, fluorescent compounds, contrast agents, DNA complexes, and other biological response modifiers (4–6, 8, 14, 27–39). In principle, isoDGR formation may occur in all these conjugates during preparation and storage, depending on pH and buffer composition. Furthermore, considering that the kinetics of isoDGR formation in cyclic NGR peptides is very fast (half-life, 3–4 hours in DMEM at 37°C; ref. 16), it is possible that transition of isoDGR to NGR occurs also in vivo, after conjugate administration. Although the amount of isoDGR formed in vivo could be low for conjugates with short plasma half-life, such as in the case of NGR-TNF, this reaction could be relevant for drugs with long half-life, such as liposomes (28), or for conjugates that persist in tissues for several hours, such as fluorophore-tagged CNGRC (14). Noteworthy, various conjugates have been prepared using linear and cyclic peptides with different NGR flanking residues (23, 35, 40).
Finally, given that the NGR motif is also present in fibronectin and that generation of isoDGR in this protein may represent a mechanism for regulating its function (16), the finding that peptides containing isoDGR can bind tumor vessels may suggest that also deamidated fibronectin or fibronectin fragments could bind endothelial cells in tumor vessels, with potentially important physiologic and pathologic implications. Of note, damaged vessels can release protein-l-isoAsp-O-methyltransferase, an enzyme that can convert isoaspartate to aspartate (16, 41), pointing to a potential mechanism for regulating its function.
In conclusion, the results obtained with isoDGR-Qdots and isoDGR-TNF suggest that natural or synthetic polypeptides containing the isoDGR motif may be exploited as ligands for targeted delivery of drugs, nanoparticles, imaging compounds, or genes to angiogenic vasculature in tumors or in other angiogenesis-related diseases.
Disclosure of Potential Conflicts of Interest
F. Curnis: Commercial research grant, MolMed SpA. A. Corti: Commercial research grant and consultant, MolMed SpA. The other authors disclosed no potential conflicts of interest.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: Associazione Italiana per la Ricerca sul Cancro.
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.
We thank Angela Cattaneo for mass spectrometry analysis of P2 and P3.