Introduction
Tumors often depend on deregulation of transcription factor activity to maintain the transformed state by growth factor–independent proliferation, suppression of apoptosis, self-renewal and initiation of angiogenesis. Whereas direct activation of oncogenic transcription factors such as c-myc1 and c-myb2 is observed in some cancers, most alterations in transcription factor activity result from events upstream in the cell signaling cascade such as deregulation of tyrosine kinase or G-protein activity3. Thus, transcription factors can act as critical focal points in a number of oncogenic pathways4. Although transcription factors have emerged as important targets for cancer therapy, they are extremely difficult to inhibit by conventional means. Their intranuclear localization precludes the use of therapeutic antibodies. Also, transcription factors (other than steroid hormone receptors, SHRs) lack binding sites for small molecules, which makes development of small-molecule inhibitors difficult. Nonetheless, inhibiting transcription factors has been attempted with varying success; approaches include blocking DNA-binding5, using peptidomimetics6 and G-quartet oligonucleotides7, and inhibiting transcription factor expression by antisense oligonucleotides8. Clinical development of these approaches has remained problematic, however, and non-SHR transcription factors are widely considered undruggable.
Here we report the development of a targeted antisense approach to inhibit the dominant negative transcription factor Id1 in the tumor endothelium of living animals. This test system was chosen for several reasons. Whereas involvement of Id proteins in tumor cell aggressiveness9, 10, 11 and metastatic behavior12, 13 has been suggested from data in cultured cells, their role in tumor angiogenesis in vivo is well established. Vascular defects after Id loss have been observed in a variety of murine models14, 15, 16, 17 and Id upregulation is observed in the endothelium of all solid human tumors examined to date11, 18. Even partial loss of Id1 activity by genetic manipulation in mice has been shown to inhibit tumor angiogenesis, and subsequently the growth of primary tumors and metastases14, 15. We could, therefore, set a reasonable goal of phenocopying Id1+/- mice using our targeted therapy. Id1 is specifically upregulated in tumor endothelial cells11, 15, 16, 17, so targeted inhibition is likely to be nontoxic. Finally, Id1 is downstream of pro-angiogenic factors VEGF-A19, bFGF, IGF-1 (ref. 20) and EGF21 so loss of Id1 activity could short-circuit all of these pathways, an important consideration as tumors can escape mono-directed, antiangiogenic therapy by upregulation of alternate growth factors22, 23.
But hurdles for targeting Id1 are substantial. In addition to the difficulties of targeting transcription factors as outlined above, inhibiting Id1 is complicated even more by the structural similarities of the Ids and their bHLH binding partners. Drugs targeting Id proteins must be selective for the Id-bHLH interaction and not affect bHLH-bHLH-dimerization. We have developed an antisense approach to inhibit Id1. To circumvent poor pharmacokinetic properties of antisense oligonucleotides, we covalently coupled them to an address-peptide that targets tumor endothelial cells. This peptide, fragment F3 of the high mobility group protein (HMG) N2, homes to neo-vessels in xenograft tumors and localizes in nuclei of endothelial cells24. Coupling to F3 enhanced the effectiveness of antisense oligonucleotides by increasing local concentrations in the target cells and facilitating uptake into the correct cellular compartment. Properties of the resulting peptide-conjugated antisense oligonucleotide (Id1-PCAO) and its effects on tumor angiogenesis and tumor growth are reported here.
Results
Coupling of oligonucleotides to peptides
We first identified an antisense oligonucleotide that inhibited expression of both murine and human Id1. Fully phosphorothioated antisense oligonucleotides displayed high toxicity in transfection experiments (Supplementary Fig. 1 online). We therefore developed a gap-mer of the selected antisense oligonucleotide, which is an oligonucleotide consisting of five non-phosphorothioated 2-O'-methyl RNA bases at both the 5'- and 3'- end and a central 13-mer stretch of phosphorothioated DNA (referred to as Id1-AO). Id1-AO but not a reverse complementary control (rcId1-AO) substantially reduced Id1 protein levels after standard lipid transfection in two endothelial cell types derived from human, human umbilical vein endothelial cells (HUVEC), and mouse (MS-1) (Fig. 1a,b). Id3 protein levels were unaffected and served as a control for specificity as its gene sequence shows only four and five mismatches (mId3 and hId3, respectively) with the chosen Id1-AO (Supplementary Fig. 2 online). To couple this antisense oligonucleotide to the F3-peptide, we modified it with a C6-amino-linker at the 5'-end. F3-peptide was coupled by an N-terminal cysteine (Fig. 1c). Conjugation was performed with a hetero-bifunctional linker (GMBS), in a chemo- and regiospecific way. The product was verified by mass spectroscopy and digestion with proteinase K (Fig. 1d).
Figure 1: Characterization of Id1-antisense oligonucleotides and Id1-PCAOs.
(a,b) Western blot analysis of the effect of the selected antisense sequence toward the human (a, HUVEC) and the murine form (b, MS1 cells) of Id1. Transfection with 200 nM Id1-AO or rcId1-AO every 24 h led to substantially reduced levels of Id1, but has no effect on expression levels of the homologous Id3 proteins. (c) Schematic description of the synthesis of Id1-PCAOs. (d) Agarose gel analysis of proteinase K digests of Id1-PCAO. (e) Agarose gel analysis of Id1-PCAO after incubation in mouse plasma at 37 °C for up to 48 h. A nonphosphorothioated DNA oligonucleotide (Id1-AO-DNA) is completely degraded within 4 h. (f) Agarose gel analysis of Id1-PCAO after incubation in buffered saline at 37 °C for up to 28 d.
Full size image (84 KB)Id1-PCAOs showed remarkable stability in plasma relative to unmodified antisense oligonucleotides, similar to the gap-mer alone (Fig. 1e). Indeed, lability of the PCAO is due primarily to degradation of the peptide (Fig. 1e). At 37 °C little degradation was observed after 24 h, suggesting suitability for in vivo experiments. No degradation was observed in buffered saline after 28 d at 37 °C (Fig. 1f) making it possible to administer the drug through subcutaneously implanted pumps (see below).
Uptake of Id1-PCAOs by endothelial cells
To determine if the Id1-PCAO retained the homing specificity of the F3-peptide and could be taken up by cells in the absence of lipid carrier, we supplied fluorescence-labeled Id1-PCAOs to different cell lines at a concentration of 200 nM. Confocal microscopy showed uptake by endothelial cells (HUVECs and murine endothelioma cells, EOMA (Fig. 2a and Supplementary Fig. 3 online)), whereas all other tested tumor cell lines and normal murine embryonic fibroblasts (MEFs) and human dermal fibroblasts were negative. No endothelial cell internalization was observed using non-peptide-conjugated fluorescence-labeled Id1-AOs (see below). Similar results were obtained with fluorescein and PCAOs labeled with tetra-methyl-rhodamine-red. Epifluorescence live imaging on viable HUVECs and HeLa cells yielded similar results thus ruling out fixation artifacts (Supplementary Fig. 4 online).
Figure 2: In vitro homing and activity of Id1-PCAOs.
(a) Standard growth media of exponentially growing cells was supplemented with 200 nM tetra-methyl-rhodamine-labeled Id1-PCAO (Id1-PCAO-TAMRA). Uptake and nuclear localization is only observed in endothelial cells (HUVEC, EOMA). Laser confocal images, scale bars: 20 
m. (b) Western blot analysis of Id1 levels in HUVEC after incubation with different concentrations of Id1-PCAO over 72 h with renewal of Id1-PCAO–supplemented media every 24 h. (c) Western blot analysis of time-dependent Id1 levels in HUVEC after incubation with Id1-PCAO. Treatment with Id1-PCAOs leads to inhibition of Id1 expression in a time and concentration dependent way. (d) Cell proliferation assay. HUVEC were incubated with different concentrations of Id1-PCAO over 5 d. Supplemented medium was renewed every 24 h. (e,f) Prolonged exposure of HUVEC to Id1-PCAOs (72 h, 200 nM Id1-PCAO, exchange of supplemented media every 24 h) inhibits tube formation on Matrigel (e) and migration of HUVEC in a scratch assay (f). Cells were counterstained with calcein AM for tube analysis. Scale bars, 20 m; all errors given in 
s.e.m.
Downregulation of Id1 with PCAOs in vitro
Exponentially growing HUVECs were incubated with standard growth medium (EGM-2) supplemented with Id1-PCAOs in the absence of lipophilic transfection reagents (Fig. 2b,c). This treatment resulted in a dose-dependent downregulation of Id1 expression. Treatment was repeated every 24 h for at least two consecutive days to yield a substantial effect on Id1 levels. Near complete inhibition of Id1 expression was reached after 3 d with a dosage of 200 nM Id1-PCAO, conditions under which unconjugated Id1-AOs showed only minor Id1 inhibition (Fig. 2b). Moreover, Id1-PCAO treatment resulted in upregulation of p16ink4a and downregulation of MMP-2, known Id1 targets25, 26, 27. Id1-PCAO concentrations as high as 1 M for 5 d did not affect HUVEC (Fig. 2d) or tumor cell proliferation (Supplementary Fig. 5 online). However, 200 nM Id1-PCAO over 3 d blocked tube formation of HUVECs on Matrigel and substantially inhibited cell migration (Fig. 2e,f and Table 1). Reverse complement rcId1-PCAO or Id1-AO plus the F-peptide in unconjugated form had no effect.
Mechanism of uptake
Id1-PCAOs colocalized with nucleolin in the nucleus of HUVECs (Fig. 3a) in accordance with published data indicating that nucleolin is the cell surface binding partner for F3 and that F3 is transported with nucleolin into the cytoplasm and subsequently into the nucleus28. To verify that Id1-PCAO is translocated by this mechanism, we performed a blocking experiment using an antinucleolin antibody (ZN004) that recognized nucleolin on the surface of HUVECs (data not shown). ZN004 was able to block uptake of fluorescein-labeled Id1-PCAO into the nucleus of proliferating HUVECs (Fig. 3b). ZN004 did not block binding of PCAOs to cell surface nucleolin (Fig. 3b, rightmost panel) suggesting that F3 binding to nucleolin is not sterically hindered by the antibody, rather internalization of the conjugate is. ZN004, but not an IgG control, efficiently inhibited downregulation of Id1 protein levels by PCAOs in HUVECs (Fig. 3c). It has been reported that nucleolin is translocated to the cell surface of endothelial cells after stimulation with vascular endothelial growth factor (VEGF)29. Consistent with this observation, binding of PCAO to the cell surface and transport into HUVECs is VEGF-A165 dependent (Fig. 3d).
Figure 3: The uptake of Id1-PCAOs is nucleolin dependent and can be stimulated by VEGF-A.
(a) Laser confocal images of HUVEC after incubation with Id1-PCAO or Id1-AO for 4 h. Medium was supplemented with 200 nM FITC-labeled Id1-PCAO (Id1-PCAO-FAM) or Id1-AO-FAM. Subsequently cells were immunofluorescence stained for nucleolin. Id1-PCAO co-localize with nucleolin in the nucleus, Id1-AO is not taken up by HUVEC. (b) Exponentially growing HUVECs were treated with anti-nucleolin AB ZN004 for 2 h before media was supplemented with fluorescein-labeled PCAO (Id1-PCAO-FAM, 200 nM). Antibody ZN004 blocked uptake of the PCAO completely. (c) HUVEC were treated for 72 h with Id1-PCAO together with addition of antinucleolin or control antibodies. ZN004 treatment resulted in inhibition of the Id1-downregulation, with H-250 and IgG-control this effect was not observed. (d) HUVEC were serum- and GF-starved for 24 h and incubated with varying amounts of VEGF-A165 (0, 2, 5 20 ng/ml) for 8 h. Subsequently the cells were treated with 200 nM of Id1-PCAO-FAM plus VEGF-A165 for an additional 2 h, showing that Id1-PCAO uptake is VEGF dependent. Results were similar if the cells were plated on standard culture slides or culture slides coated with fibronectin. All scale bars, 50 m; all errors given in s.e.m.
Homing of Id1-PCAOs to tumor endothelium
The ability of PCAOs to accumulate in tumor endothelium was tested in different murine models. MDA-MB-435S xenografts were first tested because the vasculature of these tumors is efficiently targeted by F3 (ref. 24). Allografts from spontaneous breast tumors arising in Id1- /- MMTV-HER2/neu (YD) mice were also tested15. Fluorescence-labeled Id1-PCAOs were injected systemically into graft-bearing immunodeficient mice. After 4 h, accumulation of fluorescent dye could be observed in the endothelium of dissected tumors as verified by costaining for the endothelial marker CD31 (Fig. 4a, top two panels). Unconjugated antisense oligonucleotides were not delivered into tumor endothelium (Fig. 4a, bottom panel). Id1-PCAOs were not detected in most other organs (brain, heart, colon, liver and spleen, Fig. 4b and Supplementary Fig. 6 online). However, fast uptake of Id1-PCAO into tubular cells of the renal cortex was observed (Fig. 4c, upper left panel). Accumulation in renal cortex was also observed with fluorescence-labeled Id1-AO—that is, the partially phosphorothioated gap-mer (Fig. 4c, lower left panel) and fully phosphorothioated antisense oligonucleotides (Supplementary Fig. 7 online)—and FITC-F3 (Supplementary Fig. 8 online). In addition Id1-AO and fully phosphorothioated antisense oligonucleotides also accumulated in liver (Fig. 4c, lower right panel and Supplementary Fig. 7). This is in accordance with biodistribution studies that showed a preferential accumulation of phosphorothioated antisense oligonucleotides in liver and kidney30. Conjugation to F3-peptide seems to block most of the liver accumulation (Fig. 4c, upper right panel).
Figure 4: In vivo tumor homing and activity of Id1-PCAOs.
(a) Id1-PCAOs accumulate in the vasculature of MMTV-HER2/neu (YD) and MDA-MB-435s graft tumors grown in NudeNCR mice after systemic injection (20 nmol/mouse Id1-PCAO-TAMRA or Id1-AO-TAMRA, dissection of organs 4 h after i.v. injection). Id1-AO alone does not home to the tumor (lowest panel). (b,c) Organ sections of animals bearing MMTV-HER2/neu (YD) allografts 4 h after i.v. injection of 20 nmol Id1-PCAO-TAMRA. The same dose does not lead to accumulation in other organs (b), with the exception of the renal cortex (c). Unconjugated Id1-AO accumulates in the renal cortex and in the liver (lower panel). All images in b and c merged channel confocal images (green: CD31 immunofluorescence staining, red: Id1-PCAO-TAMRA or Id1-AO-TAMRA and DAPI counterstain). (d) IHC staining of tumor sections for Id1 after treatment with Id1-AO or Id1-PCAO. Systemic treatment with Id1-PCAO (15 nmol/d i.v.) over 3 d leads to loss of Id1 expression in endothelial cells (arrows) in spontaneous MMTV-HER2/neu (YD) driven tumors in an Id1+/-
background. All scale bars, 50 m.
Homing studies were also performed in MMTV-HER2/neu (YD) and PTEN+/- animals bearing tumors. Accumulation of Id1-PCAOs in tumor endothelium was observed in these models, showing that the homing properties are maintained in spontaneous tumor models (Supplementary Fig. 6).
To test PCAO activity in vivo we injected tumor-bearing transgenic MMTV-HER2/neu (YD) Id1+/- mice intravenously with Id1-PCAOs. Because repeated application of the drug was necessary to yield significant downregulation in vitro, animals were treated with 15 nmol/d of Id1-PCAO or Id1-AO for three consecutive days. Over 80% of tumor vessels in animals treated with Id1-PCAO were completely negative for Id1 expression by immunohistochemistry (Fig. 4d). The Id1-AO alone showed no detectable downregulation of Id1 in tumor vessels.
Single agent and combination therapy
Short-term treatment of established allograft tumors with Id1-PCAO led to drastically increased hemorrhage and hypoxia and a moderate but substantial growth suppression (Supplementary Fig. 9 online). To determine if longer treatment leads to stronger growth suppression, we implanted animals with osmotic pumps that delivered Id1-PCAO continuously over 21 d. Id1-PCAO was also combined with the Hsp90-inhibitor 17-(allylamino)-17-demethoxygeldanamycin (17-AAG) as genetic Id-loss in combination with 17-AAG is more effective than either alone in reducing tumor burden15, perhaps owing to the requirement of Hsp90 for maintaining Hif1
or HER2/neu stability31. Id1-negative cells from a spontaneous tumor formed in a MMTV-HER2/neu (YD) Id-
/-
animal were used to ensure that observed effects were caused by Id1 inhibition in the microenvironment, presumably the endothelium, and not the tumor cells.
Id1-PCAO was delivered by osmotic pumps at a rate of 7 nmol/d for 21 d. 17-AAG was given by intraperitoneal injection on three consecutive days per week15 after tumor establishment. As controls, saline or the unconjugated components of the Id1-PCAO (F3-peptide and Id1-AO each at 20 nmol/d) were administered. In Id1-PCAO–treated animals the average tumor volume on day 21 was 40% of that observed in untreated tumors (Fig. 5a). The efficacy was comparable to treatment with 17-AAG alone (34% tumor volume).
Figure 5: Combination therapy with 17-AAG affects tumor growth and vascular integrity.
(a) Immunodeficient mice were subcutaneously implanted with osmotic pumps delivering 7 nmol/d Id1-PCAO or 20 nmol/d F3-peptide and Id1AO (red line, working period of pumps). Animals received 17-AAG by i.p. injection on three consecutive days/week (black arrows). (b) Animals receiving 17-AAG with or without Id1-PCAO 14 d after tumor implantation (tumors: dotted outline). (c) Epifluorescence images of whole tumor sections from animals in different treatment arms. Animals were injected with Evans blue 4 h before being killed. The red fluorescent Evans blue-albumin complex indicates vessel leakiness in the Id1-PCAO treated tumors. (d) Tumor sections were evaluated for area percentage positive staining for CD31 and for the density of Id1-positive cells (positive cells/mm2) by IHC. Both values for different treatment arms are normalized to the untreated specimen. Treatment with 17-AAG leads to decreased vascularization, treatment with Id1-PCAOs to decreased Id1 expression in remaining endothelial cells. (e) Tumor sections were stained for Hif1 expression. Treatment with 17-AAG reduces Hif1 levels, whereas Id1-PCAO leads to increased hypoxia. (f) Id1-PCAO–treated tumors stain highly positive for Hif1 (scale bars, 500 m). Hypoxic areas surround necrotic regions showing signs of cystic lesions. Combination with 17-AAG suppresses the hypoxic response while the large necrotic areas remain (black arrow). Error bars, s.e.m.
Combination of both drugs, however, yielded virtually complete inhibition of tumor growth over the treatment period (10% tumor volume, P = 0.0001, P < 0.0001, P = 0.0002 versus control, 17-AAG, and F3 plus Id1-AO plus 17-AAG, respectively). In contrast administration of Id1-AO and peptide in unconjugated form did not inhibit tumor growth and showed no enhancement over the 17-AAG alone effect (Fig. 5a,b). To further control for nonspecific effects, we repeated the experiment with a PCAO with the reverse complementary oligonucleotide sequence (rcId1-PCAO, delivered at the same rate as the Id1-PCAO, 7 nmol/d) and F3-peptide without addition of Id1-AO (20 nmol/d) (Supplementary Fig. 10 online). Id1-PCAO administration in combination with 17-AAG again had a robust effect on growth (P = 0.0079) whereas neither rcId1-PCAO nor F3 enhanced 17-AAG efficacy (P = 0.97 and 1.00, respectively). Injection of Evans blue in selected animals showed massively increased leakage from Id1-PCAO–treated tumor blood vessels (Fig. 5c) probably accounting for hypoxic stress and sensitivity to 17-AAG.
Treatment with Id1-PCAO or 17-AAG alone resulted in a decrease of Id1-positive endothelial cells in the tumor (Fig. 5d and Supplementary Fig. 11 online). Whereas Id1-PCAO downregulated Id1 expression in the cells, 17-AAG led to a decreased vascular density, which resulted in the lower count for Id1-positive cells. Combination of both drugs further diminished Id1-positive cells. Whereas Id1-PCAO administration caused upregulation of Hif1, 17-AAG injections counteracted this response (Fig. 5e,f). The hypoxic regions were characteristically surrounded by necrotic areas that displayed signs of cystification.
Id1-PCAO treatment did not affect animal weight or wound healing (data not shown). Kidneys were examined after the 3-week treatment, and no gross histological signs of toxicity were observed.
Inhibition of metastatic growth
The antitumor properties of Id1-PCAO as a single agent were further examined in a second tumor model. Lewis lung carcinoma (LLC) allografts were chosen because genetic loss of Id1 alone had been shown previously to significantly slow tumor growth and metastasis in this model14. LLC cells form tumors when implanted subcutaneously in nonimmunocompromised mice with a pure or partial C57BL/B6 background14, 25 and formation of metastasis is observed after removal of the primary tumor32.
To follow metastatic spread, we transduced LLC cells before implantation with a retroviral dual-modality reporter vector expressing eGFP and firefly luciferase33, 34. GFP-positive cells, obtained by fluorescence-activated cell sorting (FACS), were injected into the dorsal flank of male C57/B6 mice. When 7 d after injection tumors reached an average size of 20 mm3, osmotic pumps were implanted delivering Id1-PCAO at 20 nmol/d over 14 d. Control animals received saline, rcId1-PCAO (20 nmol/d) or F3 and Id1 AO in nonconjugated form (75 nmol/d). Primary tumors were surgically removed 14 d after injection and animals were monitored for metastatic development by in vivo luciferase imaging (Fig. 6a). Id1-PCAO treatment, although started in progressed, aggressively growing tumors, resulted in a significant reduction of primary tumor growth (P = 0.0079), whereas F3 plus Id1-AO or the reverse complimentary rcId1-PCAO controls had no effect (P = 1.0 and P = 0.4206, Fig. 6b). Id1-PCAO again led to an increase in Hif1-positive cells (data not shown), but baseline Hif1 levels in treatment-naive tumors were strongly elevated relative to the HER2/neu allografts.
Figure 6: Id1-PCAO inhibits primary tumor growth and metastatic spread of Lewis lung carcinoma allografts.
(a) Time-line of the experiment. Male C57/B6 were engrafted with GFP/fluc expressing Lewis lung carcinoma cells. After tumor establishment osmotic pumps were implanted to deliver Id1-PCAO or control substances. (b) Tumor growth was followed for 14 d post implantation. Treatment started on day 8 after the pumps implanted on day 7 started working (gray field: working period of pumps during primary tumor growth). (c) Kaplan-Meier plot of tumor-free survival after primary tumors were surgically removed 14 d after injection. Metastatic growth was monitored by intravital luminescence imaging (gray field, residual working period of pumps after removal of primary tumors). (d) Increased hemorrhage in treated tumors was evaluated by imaging of whole tumor sections stained with H&E. Error bars, s.e.m.
Histological analysis of primary tumors showed a decrease in Id1 levels in the endothelium (data not shown) and enhanced hemorrhage after Id1-PCAO treatment (Fig. 6), indicating a similar therapeutic response as seen in the other models described above. After removal of the primary tumor, all animals developed metastasis to the lung and occasionally to intestines and aggressive invasive local secondary regrowth (Fig. 6c). Median survival, free of secondary tumors, was prolonged by the Id1-PCAO treatment from 3 to 27 d (P = 0.0191), whereas tumor-free survival in the F3 plus Id1-AO and rcID1-PCAO groups was only 12 d, which is not statistically significant (P = 0.483 versus saline-treated animals). Effects of the PCAO on both tumor growth and hemorrhage recapitulated results obtained in Id1+/- animals (compare Fig. 6b,d with Supplementary Fig. 12 online) consistent with partial inhibition of Id1 protein levels revealed by immunohistochemistry.
Discussion
The Id proteins have attractive characteristics as targets for antiangiogenic tumor therapy. They are essential for the mobilization of endothelial progenitors from the bone marrow to the tumor17, 25, are not expressed in normal adult vasculature and lead to severe perturbations in vascular integrity when partially inhibited genetically14, 15, 16, 17. But inhibiting the activity of the Id proteins is difficult because they work by blocking the DNA binding activity of transcription factors by direct physical association (reviewed in refs. 35,36). In an attempt to inhibit expression of Id1 protein in tumor endothelial cells, we have developed an antisense targeting strategy whereby introduction of the antisense moiety into endothelial cells is facilitated by fusion with a peptide (F3) that binds tumor endothelial cells specifically. The resulting PCAO retains its homing specificity and ability to inhibit Id1 protein expression both in vitro and in vivo. The uptake is VEGF-A dependent and can be blocked with antinucleolin antibodies, indicating that the mechanism of uptake is active and in accordance with the mechanism proposed previously for F3 (ref. 28). The dependence on endothelial cell stimulation by VEGF-A and presumably other angiogenic growth factors29 explains the selectivity for the tumor vasculature versus resting blood vessels. Whereas it has been shown that F3 can transport a payload like fluorophores or nanoparticles into the tumor vasculature24, 37, 38, it was anticipated that the homing potential of the highly basic F3-peptide might be affected by conjugation to the anionic oligonucleotide. However, the PCAO seems to show higher selectivity for endothelial cells than F3 itself, which is also taken up by tumor cells in vitro24. This might indicate that the binding affinity of F3 for nucleolin is reduced by the attached anionic antisense moiety and we could, therefore, observe uptake only in cells that have the highest surface concentration of the receptor, thereby increasing the selectivity of the PCAO compared to F3. The vastly improved potency of PCAOs when compared to antisense oligonucleotides might in part be due to enhanced hybridization to the RNA target, as it has been shown previously that conjugation to lysine-rich peptides can accelerate hybridization39. However, the higher local concentration of the oligonucleotide in the endothelium is likely to be more important, as the unconjugated antisense oligonucleotide could not be detected in the tumor vasculature.
In vivo, the action of PCAOs closely recapitulates the effects of Id1 loss observed in genetically manipulated mice, which strongly supports the idea that we have effectively hit the intended target. As a single agent Id1-PCAO was able to reduce the growth rate of experimental breast tumors and the antitumor effect was enhanced when combined with the Hsp90 inhibitor 17-AAG. Also growth of highly aggressive Lewis Lung carcinomas was significantly impeded by Id1-PCAO alone. Moreover, after treatment with Id1-PCAOs and removal of primary tumors, metastatic growth of LLC was substantially delayed as was observed after genetic reduction of Id1 and Id3 levels14.
The tumor vessel phenotype observed after intervention with Id1-PCAOs and after genetic Id1 loss is different from that found after treatment with anti-VEGF agents like bevacizumab (Avastin). Whereas VEGF ablation is reported to lead to a normalization of the tumor vasculature40, Id1 loss causes increased hemorrhage and vascular permeability. That Id1 is a downstream target of VEGF/VEGF-R2 signaling19 suggests that inhibition of different arms of the VEGF-pathway may have different effects on neo-vascularization. The increased vascular leakage observed after Id1-PCAO treatment could be due to effects on endothelial progenitor cells or on non-bone marrow–derived endothelial cells because a requirement for Id1 in both populations has been demonstrated16, 25, 41. Vascular disrupting agents (VDAs) like combretastatins and 5,6-dimethylxantheonone-4-acetic acid show a similar enhancement of tumor vasculature permeability as Id1-PCAO. Treatment with VDAs has been shown to lead to Id1-dependent mobilization of EPCs, which partially rescues the tumor from the therapeutic effect41. Combining VDAs and genetic Id1 loss led to a drastically enhanced antitumor effect, which can now be further tested with a combination of Id1-PCAO and VDAs.
Because we observed increased hemorrhage and vascular permeability in treated tumors, which is in general associated with a higher rate of tumor cell embolization, it is probable that Id1-PCAO interferes with metastasis by blocking angiogenesis in the new distal bed rather than by inhibiting escape of cells into the circulation from the primary tumor. Indeed metastatic cells were observed in the lungs of Id1-PCAO–treated animals but these cells failed to colonize as long as treatment was applied. This is in accordance with earlier findings that genetic Id1 loss prevents the establishment of metastasis in the lung after intravenous injection of LLC cells14. Also, as shown previously, metastatic LLC cells that start to colonize the lungs stay dormant and are unable to induce angiogenesis as long as the primary tumors are not removed32. After elimination of the inhibitory effect emanating from the primary tumor, the dormant micrometastases still fail to stimulate angiogenesis as long as the PCAO treatment is continued.
The exact mechanism by which Id1-PCAOs and 17-AAG cooperate is not clear. Most plausible is a model by which Id1-PCAOs lead to increased tumor hypoxia (Fig. 4e,f) because of vascular leakage and therefore enhanced dependence on Hif1, a protein that is destabilized by 17-AAG42. However, we also observed a decrease in tumor vascularization after treatment with 17-AAG alone. This is likely caused by the inhibition of tumor-derived VEGF as both HER2/neu and Hif1 are upstream effectors of VEGF-A expression43. Therefore, a simple additive effect on tumor endothelial cell viability by combining Id1-PCAOs with 17-AAG is also possible.
As shown in the preclinical models presented, PCAOs have a number of attractive characteristics. First, they are remarkably stable in plasma over prolonged periods thereby bypassing a major impediment that has plagued the development of antisense therapeutics in the past. In addition, the specificity imposed by F3 toward tumor endothelium makes toxicity unlikely and indeed no adverse effects on treated mice have been observed to date over a wide range of drug concentrations. Although clearance of the PCAOs by the kidneys most likely reduced their efficacy, the PCAOs nonetheless were active enough to reduce Id1 protein levels dramatically in the treated animals with no obvious kidney toxicity. Conversely, systemic delivery of nonconjugated Id1-AO did not yield any therapeutic effect or reduction in Id1 levels up to a dosage where hepatic and renal toxicity for phosphorothioated oligonucleotides has been reported44. Thus, although optimization of a number of parameters might improve PCAO activity, the results presented here indicate that they are likely to be active and of low toxicity in a clinical setting.
The PCAO technology should allow rapid validation of other therapeutic targets in the tumor endothelium in a preclinical setting for what has been previously considered undruggable proteins. Selection for an effective antisense molecule can be done rapidly in vitro. In addition, phage display panning methods have already yielded a multitude of peptides with in vivo homing activities to other cellular targets like lymphatics45, tumor cells46, adipose tissue47, urothelium48, synovium49 and hematopoietic cells in the bone marrow50. Whether these peptides will allow efficient delivery of antisense oligonucleotides remains to be determined. If so, the potential already exists to inhibit targets for a variety of diseases in a tissue-specific way. The high local concentrations achievable with PCAOs allows for substantially lower therapeutic doses, thereby decreasing side effects. Thus, directed delivery of antisense molecules or other biologically active molecules using a peptide conjugate may prove to be an important avenue of therapeutic treatment of human cancers and other diseases.
Methods
General.
Chemicals were purchased from Sigma unless otherwise indicated. GMBS was purchased from Fluka. Oligonucleotides were obtained from Operon Biotechnologies. Peptides were synthesized in the microchemistry core-facility at Memorial Sloan–Kettering Cancer Center. 17-AAG and EPL were obtained from the National Cancer Institute. Antibodies for western blot (WB) and immunohistochemistry (IHC) were obtained from Santa Cruz Biotechnology (Id1 (WBs), SC-488; Nucleolin, SC-9893, CD31), Biocheck (Id1 (IHC)), BD Pharmingen (p16ink4a, CD31, CD34), Sigma (
-actin), Chemicon (Hif1). Mass spectrometry was performed by M-Scan Inc.
Cell lines.
Human umbilical vein endothelial cells (HUVEC-2) were obtained from BD Biosciences, maintained inEGM-2 media (Cambrex), and were used at passages four through six in all experiments. Normal human dermal fibroblasts (NHDF) were purchased from Cambrex and maintained in FGM-2 growth medium (Cambrex). Murine embryonic fibroblasts (MEFs) were obtained from animals with a mixed C57B6/129SV background, as described previously16. HeLa, MDA-MB-435S, LLC and EOMA cells were purchased from ATCC and maintained in DME high-glucose medium supplemented with 10% FBS and glutamine. KYSE-520 were obtained from the DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen) and maintained in RPMI 1640 media with 10% FBS.
Animals.
Female nude NCR mice were obtained from Taconic at 4-5 weeks of age. Genetic tumor models used (MMTV-HER2/neu (YD) and PTEN+/- ) were described previously15, 16. Generation of Id1-deficient mice (Id1- /- and Id1+/- ) was described previously14. These animals were bred back into a pure C57BL/6 background and used with wild-type littermates as controls in the experiments. All experiments involving animals were approved by MSKCC's Institutional Animal Care and Use Committee (IACUC).
Coupling of antisense-oligonucleotides to cysteine-modified peptides.
Coupling was done according to Harrison and Balasubramanian51 with some modifications as described below. Fully modified oligonucleotides were obtained directly from Operon Biotechnologies with a C6-amino linker attached to the 5'-end of a gap-mer with the sequence GCACCagctccttgaggcGUGAG (upper case: 2'O-methyl RNA bases; lower-case phosphorothioate-linked DNA bases). The reverse complimentary sequence with the same modification was used as a control oligonucleotide (rcId1-AO and in conjugated form as rcId1-PCAO). For uptake and homing studies, oligos were obtained from the supplier with a 3'-end fluorescein or Rhodamine red label. The cysteine modified F3-peptide sequence is 5'-CKDEPQRRSARLSAKPAPPKPEPKPKKAPAKK-3'.
The oligos were dissolved in 200 mM TrisHCl pH 8.4 to a final concentration of 1 mM and stored at -
20 °C. 2.8 mg GMBS (4-maleimidobutyric acid N-hydroxysuccinimide ester, 10 mol, 100 eq.) in 40 l acetonitrile were added to 100 l (100 nmol) of the oligo solution. The reaction vessel was wrapped in aluminum foil and incubated with shaking at 25 °C for 90 min. The oligonucleotide was precipitated with 1 ml acetonitrile and remaining GMBS was removed by vigorous washing with acetonitrile (9 
1 ml). After drying in vacuo the activated oligo was dissolved in 50 l buffer (100 mM Na-phosphate, 400 mM NaCl at pH 7.0) to which 400 nmol F3-N-Cys (4 eq.) in 50 l buffer (40 mM sodium-phosphate, 20 mM EDTA, pH 7.0) were slowly added. The reaction mixture was incubated with shaking for 24 h at 25 °C. The coupling-product was purified by reversed-phase high-performance liquid chromatography (Äkta Purifier System, GE Healthcare; Column: OligoDNA RP 150 7.8 mm, Tosoh Bioscience). To first eluate the nonconjugated peptide a 0.05% (vol/vol) transcription factor A in water/acetonitrile gradient (5–20% (vol/vol) acetonitrile) was used. To separate the nonconjugated antisense oligonucleotide from the PCAO a second step was done using an ammonium acetate (20 mM, pH 6.8)/acetonitrile gradient (5–25% (vol/vol) acetonitrile). Fractions containing the PCAO were combined, concentrated in a vacuum concentrator system (Eppendorf), and precipitated with ethanol. Yields in a multitude of experiments (>10, scales up to 600 nmol) varied between 40 and 78% as calculated from absorbance measurements at 260 nm. Identity of the conjugate was verified by mass spectrometry. Negative ion electrospray ionization mass spectronomy was performed on a Sciex Q-star/Pulsar instrument (MDS Sciex). Id1-PCAO: calculated mass: 11441.6 Da, found: 11441.8 Da.
Transfection of endothelial cells with antisense oligonucleotides.
Endothelial cells (HUVEC-2 or MS-1) were seeded 16 h before transfection at 105 cells/well in six-well multi-well dishes (MWD) in standard growth medium without antibiotics. Antisense oligonucleotides were reprecipitated with ethanol from sodium acetate (10 mM, ph 4.8) buffer before use. Cells were transfected in 1 ml OptiMEM I medium (Invitrogen) with Lipofectin (Invitrogen) or Cytofectin (Gene Therapy Systems) according to the manufacturers' recommendations. The transfection was repeated at 24 and 48 h.
Plasma stability assay.
Female BALB/c mice were bled by submandibular punctuation using a 4.5 mm lancet (Medipoint). Plasma was separated from heparinized whole blood by centrifugation (18,000g, 5 min at 4 °C). Antisense oligonucleotides and PCAOs were dissolved in plasma at 25 M and 10 l aliquots in microfuge tubes and were incubated at 37 °C for the indicated time. After incubation, 5-l quencher solution (1.6 M NaCl, 100 mM EDTA pH 8.0) were added and the samples were stored at -
80 °C. For analysis samples were diluted to 60 l with agarose sample buffer, incubated 5 min at 95 °C and separated in a 1.5% (wt/vol) low melting point agarose gel. Antisense oligonucleotides/PCAOs were visualized with ethidium bromide under UV-light.
Transfection of cells with Id1-PCAOs.
HUVEC were seeded 24 h before transfection at 105 cells/well in six-well MWDs in standard growth medium (EGM-2) without antibiotics. At the day of transfection the antisense oligonucleotides–conjugates and control oligos were diluted in EGM-2 to the indicated concentration. Medium was replaced with the supplemented EGM-2. The procedure was repeated every 24 h for two more days, samples were drawn at the indicated time points and analyzed by western blot analysis. Other cell lines were treated similarly, with the exception that Id1-PCAO was added to standard growth medium, DME or RPMI 1640, according to cell type.
Uptake studies in different cell lines with fluorescence labeled Id1-PCAO were performed in both the standard growth medium (DME, RPMI 1640 or EGM-2) or in OptiMEM I (supplemented with 2% FBS) to control for media effects.
For uptake studies monitoring dependence on VEGF stimulation, HUVEC were starved for 36 h in OptiMEM I. Serum-free medium was exchanged with different amounts (0, 2, 5, 20 ng/ml) of VEGF-A165 (Peprotech) and cells were incubated for 10 h. Id1-PCAO-FAM was added at 200 nM and cells were incubated for an additional 2 h, fixed and counterstained.
Scratch assay.
HUVEC were plated in fibronectin-coated two-well chamber slides (BD Bioscience) at 2.5 104 cells/well. Growth medium was supplemented with 200 nM Id1-PCAO (or Id1-AO) and renewed every 24 h. Seventy-two hours after plating, a scratch was applied using a 20 l pipette tip. Chambers were washed with medium and supplemented medium was added. Eighteen hours after the scratch was applied, cells were fixed and imaged.
Transduction of LLC cells with an eGFP/Fluc dual-modality reporter.
Ecotropic retrovirus based on the SFG vector34, expressing an Aequorea Victoria eGFP/firefly luciferase (eGFP/FLuc) fusion protein was produced in PhoenixE cells and was used with at least 1 106 infectious particles/ml against NIH3T3cells. The in vitro transduction of early passage LLC cells with the retroviral vector was accomplished by exposing the cell monolayer to a filtered (0.45 m) culture medium obtained from the vector producer cells for 8 h in the presence of 8 g/ml polybrene (Sigma). Stably transduced cells were enriched by FACS using eGFP-expression as an marker for successful transduction.
Tube formation assay.
HUVEC were grown for 72 h in T25 dishes in media containing Id1-PCAO (or rcId1-PCAO or Id1-AO plus F3) at 200 nM. Medium was renewed every 24 h. Treated cells were detached with trypsin/EDTA and seeded at 2.5 104 cells on top of a matrigel layer in 24-well multiwell dishes in the treatment medium. After incubation for 18 h, cells were counterstained with Calcein AM (Invitrogen) and imaged under an inverted fluorescence microscope (Zeiss Axiostar 200). Tube length was evaluated with the MetaMorph software.
Proliferation assay.
Cells were plated at 2 104 cells in the wells of 24 well MWDs. After 18 h medium was exchanged with standard growth medium (EGM-2) supplemented with Id1-PCAO or Id1-AO plus F3 (200 nM or 1 mM). Supplemented medium was renewed every 24 h. Media was removed from sample plates at different time points, and the plates were stored at -
80 °C. All samples were analyzed in parallel using the CyQuant system (Invitrogen) according to the manufacture's instructions.
Nucleolin-antibody blocking studies.
HUVEC were plated in EGM-2 medium at 2.5 104 cells in 4-well chamber slides (BD Bioscience). After 24 h antinucleolin antibodies were added to the medium (tested antibodies: MS-3 (SantaCruz Biotechnologies), H-250 (SantaCruz Biotechnologies), ZN004 (MBL) and 3G4B2 (Millipore). After 2 h cells were fixed with 4% (wt/vol) PFA, probed with fluorescein-labeled secondary antibodies and counterstained with Hoechst 33342. Ability to recognize extracellular epitopes of cell surface nucleolin was assessed by confocal laser microscopy.
To study blocking of PCAO-uptake, Id1-PCAO-FAM was added at 200 nM after the 2 h incubation step with the antibody. Cells were incubated an additional 2h at 37 °C, fixed with 4% (wt/vol) PFA and counterstained.
To study blocking of PCAO-mediated Id1-downregulation, HUVEC were seeded at 4 105 cells in 6-well MWDs, and incubated for 16 h. Antibody (ZN004, 10 g/ml) or IgG-control (mouse-IgG2b, BD Pharmingen) were added for 2 h before Id1-PCAO or rcId1-PCAO (200 nM) was supplemented. After 24 h and 48 h medium was exchanged, containing the same concentration of AB and Id1-PCAO. Seventy-two hours after the first treatment, cells were lysed and lysates were probed by western blot analysis.
Delivery of Id1-PCAOs in vivo.
12 nmol fluorescence-labeled Id1-PCAOs (
6.8 mg/kg body weight (BW)) or Id1-AOs were dissolved in TBS and injected into the tail vein or subcutaneously of tumor-bearing mice. Mice were killed, organs and tumors were dissected, fixed overnight in 4% (wt/vol) PFA and finally immersed in 20% (wt/vol) sucrose for 24 h. After embedding in OCT (Miles Inc.) and sectioning, samples were probed for CD31 using a biotinylated secondary antibody and a streptavidin-Alexa488 conjugate as a tertiary agent.
Allograft model of Her2-overexpressing breast cancer.
Female nude NCR mice (Taconic) were engrafted with 5 106 MMTV-HER2/neu (YD) Id1-
/-
tumor cells in the left flank. The animals were randomly divided into three cohorts of four animals and treatment was started 96 h later when tumors became palpable. The first cohort received 10 nmol Id1-PCAO conjugate in 200 l TBS. The other two cohorts served as negative controls and received either TBS or 10 nmol F3-peptide plus 10 nmol Id1-AO in TBS (5.7 mg/kg BW). Application of the conjugate and control solutions was performed by intravenous injection into the tail vein and was repeated every 24 h for 7 consecutive days. 24 h after the last injection animals were killed. Tumors, kidneys, livers and femurs were collected, fixed with paraformaldehyde and embedded in paraffin.
Alternatively, animals were implanted s.c. with osmotic pumps (Durect Corp.) that delivered 7 nmol/d Id1-PCAO (3.5 mg/kg BW) in TBS over a 14-d period. Controls received 20 nmol/d F3-peptide plus 20 nmol/d Id1-AO (in TBS) or TBS. 2 106 MMTV-HER2/neu (YD) Id1-
/-
tumor cells were injected into the left flank 24 h after implantation of the pumps. After the 14-d treatment period, pumps were replaced using a model with a work period of 7 d. Treatment with 17-AAG (75 mg/kg, i.p. on three consecutive days/week) was started when tumors reached a size of 20 mm3. 17-AAG was dissolved at 50 mg/kg in DMSO and diluted with EPL 1:1 before injection. Control animals received DMSO:EPL 1:1 i.p. at the same schedule. Tumour size was measured using a calliper. Volume was calculated as V = (
/6 longest diameter perpendicular diameter2).
Allograft model of metastatic LLC.
7.5105 Dual reporter labeled LLC cells were implanted in the right dorsal flank of male C57J/B6 mice (Jackson Laboratories). After 7 d, animals were implanted with osmotic pumps (100 l volume, work period 14 d). The pumps were filled with saline solution of either Id1-PCAO (3.5 mM), rcId1-PCAO (3.5 mM) or F3-peptide plus Id1-AO (12.5 mM each). Concentration and release rate of the pumps resulted in a delivery rate of 229 g/d (Id1-PCAO and rcId1PCAO) or 265 g/d and 580 g (F3 and Id1-AO). Fourteen days after tumor implantation animals were anaesthetised and primary tumors were surgically removed. Complete removal of the tumor tissue was checked 3 d post operation by in vivo luciferase imaging and re-growing primary tumors were removed. For in vivo luciferase imaging, 100 l of D-luciferin (Gold Bio Technology, 15 mg/ml potassium-salt in PBS) were injected retro-orbitally to animals anaesthetised by isofluorane inhalation. Photographic and luminescence images were acquired using an IVIS 100 system (Xenogen). Animals were sacrificed when distressed. Tumor burden and metastasis data acquired by in vivo luminescence was confirmed by histology.
Image acquisition and analysis.
Epifluorescence, bright field and phase contrast images were acquired using Zeiss Axiostar 200 microscopes. Leica laser confocal microscopes were used for co-localisation studies. For quantification, large fields of the tissue sections were acquired using an automated image acquisition and montaging system (Zeiss Axiostar 200M microscope with MetaMorph Software, Molecular Devices). For evaluation of single cell staining (Id1, CD31, Hif1) an average of 30 adjacent, single images were acquired from the center of the section using a 20 objective. Images were montaged to yield one large image covering an average area of 0.82 mm2. Three or 4 large field images were used to quantify each section. Stained areas were quantified using MetaMorph or ImageJ software (http://rsb.info.nih.gov/ij/). Images were threshholded and stained area (CD31) was calculated or particles per field were counted (Id1 or Hif1 positive cells). To quantify extent of hemorrhage, H&E whole tumor sections were imaged with a 5x objective and evaluated using the color threshold function of the Metamorph software.
Statistical analysis.
Statistical analysis was performed using the GraphPad Prism software (GraphPad Software). Tumour progression in different treatment arms was compared using the Wilcoxon signed rank sum test. Students t-test was used to analyze results from IHC staining and hemorrhage evaluation experiments. The build in statistical function of the Prism software was also used to evaluate Kaplan-Meyer survival curves. All P-values are two-tailed.
Note: Supplementary information is available on the Nature Biotechnology website.
Author Contributions
All authors contributed significantly to the experimental design and/or execution of the experiments described.