All animals were handled in strict accordance with good animal practice as defined by University of Texas Health Science Center Institutional Animal Care and Use Committee, and all animal work was approved by the committee (protocol # HSC-AWC-07-099).

Oligonucleotide primers were synthesized by Midland Certified Reagents (Midland, TX). The extracellular domain of recombinant human E-selectin (535 amino acid residues) was purchased from R&D Systems (Minneapolis, MN). Streptavidin-coated magnetic particles were purchased from Pure Biotech (Midlesex, NJ). Human microvascular endothelial cells (HMVECs) were a kind gift from Dr. Rong Shao (Biomedical Research Institute, Baystate Medical Center/University of Massachusetts at Amherst, Springfield, MA, USA). Anti-human CD31 antibody was purchased from BD Pharmingen (San Jose, CA). Anti-E-selectin antibody H18/7 was isolated from hybridoma purchased from ATCC (Manassas, VA) and used as a competitor of TA binding to endothelial cells. Anti-human E-selectin antibodies were purchased from Sigma (St. Louis, MO) and Innovex (Richmond, CA) and used for immunostaining for cultured cells and human carcinoma paraffin sections, respectively. Human carcinoma tissue array was purchased from US Biomax (Rockville, MD).

The synthesis of the DNA thioaptamer (TA) combinatorial library was described previously [24]. Briefly, a single-stranded DNA library (∼10¹⁴ different sequences) with a 30-nucleotide random region (N30) flanked by 23 and 21 nucleotide primer binding regions was chemically synthesized. The library (40 nM) was PCR amplified in a reaction containing Sp dATP(αS), dCTP, dGTP and dTTP (200 µM), MgCl₂ (2 mM), biotinylated forward primer (5′biotin- CAGTGCTCTAGAGGATCCGTG-AC-3′) (300 nM), reverse primer (5′-CGCTCGGATCGATAAGCTTCG-3′) (300 nM) and AmpliTaq DNA polymerase (0.5 U). Biotinylated double-stranded PCR products were incubated with streptavidin-coated magnetic beads for the separation of the ssDNA library.

Screening of TAs that binds to recombinant E-selectin protein was carried out using a solution-based filter binding method as described previously [24]. Briefly, the recombinant human E-selectin protein (240 pmoles) was incubated with TA library (200 pmoles) in selection buffer (PBS with Ca²⁺ and Mg²⁺ and 5 mM MgCl₂) at room temperature for 2 hours. The reaction mixture was filtered through the nitrocellulose membrane and washed 3 times with the selection buffer to remove unbound TAs. The TA/E-selectin complex retained on the filter membrane was eluted with 8 M urea solution. The eluent was used as the template for PCR amplification and the integrity of the TAs was analyzed by 15% polyacrylamide gel electrophoresis. This selection cycle was repeated 10 times and the stringency of the selection was elevated gradually. The TA libraries obtained after rounds 5 and 10 were PCR amplified and subcloned into a plasmid vector for DNA sequencing. The selected sequences were analyzed using the ClustalW program (with DNA identity matrix, gap open penalty of 15, gap extension penalty of 6.66, gap separation penalty of 4 with no end gaps). Cyanine 3 (Cy3)-labeled TAs were produced by PCR amplification of plasmids containing TAs as a template with 5′-terminal Cy3-labeled reverse primer.

Human Microvascular Endothelial cells (HMVEC) were a kind gift from Dr. Rong Shao at the University of Massachusetts. HMVEC were cultured according to the protocol described previously [33]. HMVEC were grown in endothelial basal medium-2 supplemented with 10% (v/v) Tet-approved fetal bovine serum, 100 U/ml penicillin and 10 µg/ml streptomycin, 1 µg/ml epidermal growth factor and 50 µg/ml hydrocortisone. All experiments were performed on 70–80% confluent cultures in 5% CO₂ humid chambers at 37°C. The HMVECs were genetically manipulated to generate Tet-on inducible system for E-selectin expression (ES-Endo). E-selectin expression was induced with doxycycline (2000 ng/ml) for 5 hours unless specified.

To examine TA binding to the ES-Endo, the cells were plated onto a plastic dish and cultured overnight to allow them to attach. After E-selectin induction with doxycycline, the cells were incubated with cy3-labeled TAs at the indicated concentrations (0–200 nM) for 20 minutes at 37 °C. The cells were washed with ice-cold PBS to remove unbound TA and subsequently fixed with 4% paraformaldehyde for 10 minutes. The nuclei were counterstained with 1.0 µg/ml Hoechst 33342 for 10 minutes. The extent of TA binding to the cells was assessed by fluorescence microscopic analysis (TE2000-E, Nikon, final magnification 60x). The relative binding affinity of TAs was determined by the amount of fluorescence detected on the cells based binding assay and the specificity was determined by the extent of doxycycline dose dependent effect seen on TA binding. For competition of TA binding to the cells, the cells were pre-incubated with 10 mg or 25 mg of anti-E-selectin antibody (H18/7) for 2 hours prior to incubation with TA. All images were acquired under the same exposure conditions for the comparison of TA binding.

Human tissues derived from epithelial ovarian cancer patients were collected from surgical cases at The University of Texas M.D. Anderson Cancer Center. Frozen tissue arrays derived from human carcinomas (breast, ovarian, and skin) and their normal counterparts were also used (US Biomax, MD). The tissue sections were fixed with ice-cold acetone, incubated with 50 nM ESTA-1 for 1 hour at RT, and then stained with primary antibody against anti-rat CD31 (1∶1000). E-selectin expression was determined by immunostaining with anti-E-selectin (1∶20). For in vivo experiments a total of 10 µg of chemically synthesized Cy3-labeled ESTA-1 was intravenously injected into mice bearing tumors derived from mouse breast cancer 4T1 cells. The organs and tumors were harvested 3 hours after the injection, and each organ was embedded in OCT. 8 µm frozen section was fixed with acetone and stained with Hoechst 33342 for assessment of ESTA-1 binding to the vasculature.

Equal amounts of ESTA-1 (4.6 pmoles) were incubated with increasing concentrations of the recombinant selectin proteins (0–19 pmoles) in a total volume of 10.5 µl of PBS supplemented with Ca²⁺ and Mg²⁺, 5 mM MgCl_2, and 1% NP40 at room temperature for 45 minutes. The reaction mixtures were loaded onto 6% polyacrylamide tris borate gels and run at 100 V for 90 minutes at 4°C. The gel was stained with SYBR Gold nucleic acid staining dye and visualized using the FluorChem 8800 chemimager (Alpha Innotech). Protein-bound TA and unbound TA were quantified using ImageJ software. The binding curves were generated assuming a single binding site curve fits using the Graph Pad Prism software. For competition experiments, anti-E-selectin antibody (H8/17) (3 µg, 9 µl) or for control experiment human IgG antibody was mixed with recombinant E-selection protein (19 pmoles) and 1% NP40 in PBS (3.7 µl) and incubated for 30 minutes. To this mixture, ESTA-1 (4.6 pmoles) in PBS and 5 mM MgCl₂ was added and incubated for another 10 minutes and loaded onto 6% polyacrylamide gel and ran at 100 V for 80 minutes, stained with SYBR Gold stain and visualized using the chemiimager.

To determine the effect of ESTA-1 on adhesion of sLe^x positive cells to endothelial cells, confluent ES-Endo were incubated with doxycycline for 5 hours followed by ESTA-1 (50 nM and 100 nM) for 20 minutes. HL-60 cells (10⁵ cells) suspended in RPMI containing 1% FBS were added to ES-Endo and incubated at 4°C for 30 minutes with mild agitation. The unbound cells were washed off with RPMI containing 1% FBS. The number of cells that adhered to the ES-Endo was counted on at least 3 random areas using a light microscope (final magnification 100x) and expressed as the mean of triplicate experiments.

ES-Endo were cultured on a 96-well plate at 10,000 cells per well. The cells were incubated with doxycycline for 5 hours and then incubated with ESTA-1 at the indicated concentrations for 48 hours. For the measurement of cell viability, 10 µl MTT (5 mg/ml) were added to each well and incubated for 4 hours. The formazan was dissolved in 150 µl of DMSO and the absorbance at 490 nm was measured.

All experiments were carried out in triplicates and the data were analyzed statistically to provide 80% power for a test at significance level of 0.01. We validated the normality assumption, and proceeded with a parametric test as appropriate. The Student-T test was performed to compare the cell viability among different groups.

We screened a thioaptamer (TA) library to select for those TAs that demonstrated affinity for E-selectin. Each of the 10¹⁴ TAs in the library consisted of a region of random sequence (N₃₀ residues) flanked by two primer regions common to all TAs, and all dA's contained 5′-monothiophosphate substitutions with the exception of the 5′ primer region. A two-step E-selectin TA selection strategy followed. First, a solution-based combinatorial selection method was employed for the identification of thioaptamers that bind the extracellular domain of recombinant human E-selectin protein. The TA library was allowed to interact in solution with glycosylated recombinant E-selectin protein. Then, the E-selectin/TA complexes formed were isolated and PCR amplified to be used in subsequent cycles of selection. After 10 iterative selection cycles, 35 TA sequences were isolated. Sequences were aligned using the ClustalW program based on their primary sequences (Fig S1), and a cladogram tree was generated. Based on the alignment and the cladogram, the 35 sequences were grouped into 14 families (Fig S2). A single sequence was selected from each family based on the lowest predicted free energy of the secondary structure. Several common sequence motifs were identified among the selected 14 sequences (Fig S3). These 14 TA sequences were synthesized by PCR with Cy3-labeled reverse primer for the second step of cell-based selection.

A Tet-on inducible E-selectin endothelial cell line (ES-Endo) was used to identify the TA sequences that specifically bind to E-selectin on the surface of endothelial cells. First, to demonstrate the doxycycline-dependent induction of E-selectin expression, ES-Endo cells were incubated with increasing concentrations of doxycycline (0–2000 ng/ml) for 5 hours, and the E-selectin expression level on the plasma membrane was analyzed by immunofluorescent staining using anti-E-selectin antibody. As a reference for the physiological level of E-selectin expression, the cells were also treated with TNF-α (10 ng/ml) for 5 hours. Elevated expression of E-selectin was detected predominantly on the cell membrane when treated with 500 ng/ml of doxycycline, and its expression level was increased in a doxycycline concentration dependent manner (Fig. 1A). In the absence of doxycycline, the baseline level of E-selectin expression was slightly higher than wild type cells, perhaps due to the leakiness of this inducible system. The level of E-selectin expression with a doxycycline concentration of 2000 ng/ml was equivalent to TNF-α treated cells, and thus, we used doxycycline at this concentration for subsequent experiments, unless otherwise specified. For the second-step selection of TAs that bind specifically to E-selectin, ES-Endo were pre-incubated with doxycycline and then with each of the 14 TAs (100 nM) selected in the first step for 20 minutes at 37°C. Fluorescent intensities associated with the cells were compared using fluorescent microscopy, and the data was evaluated for both binding and specificity. Interestingly, the relative binding affinities and specificities of the TA sequences were found to correlate with the energy of the structures predicted by MFOLD [34] (Table S1). Among the 14 TAs tested, TA-1 exhibited high doxycycline-dependent binding to ES-Endo with minimal binding to ES-Endo in the absence of doxycycline, whereas the rest of the TAs showed weak doxycycline dependent binding with high background (Table S1). The MFOLD prediction exhibited a single secondary structure for TA-1 (with an estimated free-energy of folding of −10.72 kcal/mol) containing two stable hairpin loops (Fig. 2B). In contrast MFOLD predicted four secondary structures for both TA-20 (dG = −7.98 to −7.44 kcal/mol) and TA-31 (dG = −8.64 to −7.94 kcal/mol) with comparable free energy values, and only a single stable hairpin loop was predicted in each of these structures (Fig S4). Interestingly, two highest binders (TA-1 and TA-20) share the ACT(T/C)C(T/A)C(T/C)TCAC sequence motif in the loop region of the hairpin stem-loop (Fig. S3), suggesting that this region might be involved in binding to E-selectin. Presumably the second hairpin loop in TA-1, but not in TA-20 contributes to its increased binding affinity and specificity. Overall, TA-1 showed the highest binding as well as specificity. Based on database search (NCBI BLAST), the TA-1 sequence (Fig. 2A) did not show any homology to existing genes. Together, we performed further studies using E-selectin thioaptamer TA-1 (ESTA-1).

To provide further concrete evidence that ESTA-1 binding to ES-Endo is E-selectin specific, we used three independent approaches. First, the ES-Endo cells were treated with increasing concentrations of doxycycline (up to 2000 ng/ml) for 5 hours to induce different levels of E-selectin expression, and they were then incubated with a fixed concentration of ESTA-1 (100 nM) for 20 minutes at 37°C. ESTA-1 was found to form a speckled binding pattern. The number and brightness of the speckles increased proportionally to the doxycycline concentration up to 2000 ng/ml (Fig. 1A), suggesting E-selectin specific ESTA-1 binding. Treatment of ES-Endo with TNF-α (10 ng/ml) yielded a similar binding pattern and fluorescent intensity (Fig. 1A). In the second approach, ES-Endo expressing E-selectin were first pre-treated with two different concentrations of E-selectin monoclonal antibody for 2 hours and then incubated with 100 nM of ESTA-1. Pre- and co-incubation of the cells with E-selectin monoclonal antibody caused a significant reduction of ESTA-1 binding as evidenced by the disappearance of the speckle pattern (Fig. 1B), suggesting that ESTA-1 and E-selectin antibody share the same epitope on E-selectin. Similar reduction in ESTA-1 binding was also observed by E-selectin antibody for TNF-α treated ES-Endo (data not shown). In contrast, normal IgG pre-treatment did not affect ESTA-1 binding to the cells (Fig. 1B). Thirdly, immunostaining was performed using a monoclonal antibody against E-selectin that does not compete with ESTA-1 binding to doxycycline induced ES-Endo (data not shown). This data demonstrated that ESTA-1 (red fluorescence) partially colocalized with E-selectin (green fluorescence) on the edge of the cells as seen in yellow merge, supporting ESTA-1 binding to E-selectin on the cell surface (Fig. S6). A possible explanation of a partial colocalization of ESTA-1 and E-selectin might be either ESTA-1 binding to other surface molecules on the endothelial cells or intracellular uptake following the initial cell surface binding since E-selectin undergoes internalization [34]. In conclusion, the two-step screening strategy employed here led to the identification of a TA sequence (ESTA-1) that binds specifically to E-selectin expressed on endothelial cells.

Next, we tested ESTA-1 binding to the tumor vasculature using histological sections derived from human carcinomas. First, immunohistochemical analysis was performed to evaluate the level of E-selectin expression on the tumor vasculature using paraffin sections derived from three types of carcinomas including breast, ovarian, and skin. Approximately 70-80% of tumors showed E-selectin expression on the vasculature (Fig. 3A and C). Unlike angiogenic factors such as integrins and vascular endothelial growth factor receptor, E-selectin expression was detected in both the existing mature vessels and the microvessels in the tumor (Fig. 3A). To test ESTA-1 binding to the tumor vasculature, the frozen sections were first incubated with a 50 nM solution of Cy3-labeled ESTA-1 (red fluorescence), then immunostained with CD31 (green fluorescence). Intense ESTA-1 binding was observed on the vessels in ovarian carcinomas as evidenced by the co-localization with CD31 (Fig. 3B). In contrast, ESTA-1 binding was not observed in the vessels in the normal counterpart. Similarly, ESTA-1 bound to the tumor associated vessels in breast (80%) and skin (100%) carcinomas (Fig. 3C). Overall, ESTA-1 binding to the tumor associated vessels was highly correlated with E-selectin expression as indicated by ESTA/E-sel ratios for breast (ratio = 0.89), ovarian (ratio = 1) and skin (ratio = 1) carcinomas (Fig. 3C). As opposed to the binding to the tumor associated vessels, ESTA-1 binding was almost absent in the normal human tissues, including the adrenal, brain, temporal lobe, breast, cervix, heart, kidney, liver, lung, pancreas, placenta, salivary gland, skeletal muscle, small intestine, spleen, stomach, thyroid, and uterus, with the exception of minor binding to the vessels of the skin (data not shown), where E-selectin is constitutively expressed [35].

Next we tested ESTA-1 binding to E-selectin on tumor-associated vasculature in vivo. We used a 4T1 breast tumor mouse model in which high E-selectin expression was observed on the endothelial cells of the tumor-associated vasculature (Supplemental Fig. S5). Intravenous administration of ESTA-1 into mice bearing 4T1 tumor resulted in accumulation of ESTA-1 to the endothelial cells of the tumor vasculature as evidenced by the speckled red pattern on the tumor associated vasculature (Fig. 4). No obvious binding of ESTA-1 to the vasculature of other organs (liver, spleen, kidney, lung, and heart) was detected (data not shown). To confirm E-selectin specific binding of ESTA-1 in vivo, an E-selectin antibody was injected intravenously prior to ESTA-1 injection into mice bearing breast tumor xenograft. E-selectin antibody pre-injection resulted in a significant reduction in ESTA-1 binding to tumor vasculature as compared to untreated mice, whereas pre-injection of control IgG did not cause any changes (Fig. 4), suggesting that ESTA-1 binds to E-selectin on the tumor vasculature.

We next evaluated the binding affinity of ESTA-1 to all selectins using electrophoretic mobility shift assay (EMSA). To determine the binding constant, fixed amounts of ESTA-1 was mixed with increasing amounts of recombinant protein (E-, P-, and L-selectin). Incubation of recombinant E-selectin protein and ESTA-1 resulted in the formation of a DNA/protein complex in equilibrium with unbound states. An increment in ESTA-1/E-selectin complexes was observed with increasing recombinant E-selectin added to the reaction, accompanied by a corresponding decrease in the free (unbound) ESTA-1 (data not shown). As expected, the amounts of ESTA-1/E-selectin complexes reached saturation at a molar ratio of 1∶1, when both of the binding molecules are at a concentration of 500 nM. Based on the densitometric analysis, the binding constant calculated for the ESTA-1 binding to E-selectin was 47 nM (Fig. 5A). The binding of ESTA-1 to P-selectin showed significantly lower affinity (estimated K_D = 13 µM), suggesting a very weak interaction at the concentration range measured (Fig. 5B). ESTA-1 binding to L-selectin was not detectable under the same conditions. To validate ESTA-1 binding to E-selectin, EMSA was performed by pre-incubating the human E-selectin recombinant protein with E-selectin antibody (H18/7) prior to addition of ESTA-1 to the reaction mixture. The addition of E-selectin antibody (H18/7) resulted in the disappearance of the band corresponding to the ESTA-1/E-selectin complex (Fig. 5C). This data implies that ESTA-1 not only binds to E-selectin, but also may share the same binding site with E-selectin monoclonal antibody used for this experiment. To further confirm the binding affinity of ESTA-1 under biological conditions, different concentrations of ESTA-1 (50–200 nM) were incubated with ES-Endo induced with doxycycline. ESTA-1 binding to the cells was detectable at 50 nM and increased in a dose dependent manner (Fig. 5D). Together, these data support a nanomolar affinity of ESTA-1 binding to E-selectin on the endothelial cells.

On the basis of specific binding of ESTA-1 to E-selectin, we next tested ESTA-1-mediated inhibition of leukocyte adhesion to endothelial cells. For this study, we tested a human promyelomonocytic cell line (HL-60) that expresses sLex, a natural ligand for E-selectin. HL-60 cell adhesion to ES-Endo increased by 5-fold when E-selectin expression was induced by doxycycline (Fig. 6A). E-selectin expressing ES-Endo were pre-incubated with indicated concentrations of ESTA-1 for 20 minutes and then the adhesion of these cells to the ES-Endo was compared. Pretreatment of the E-selectin expressing ES-Endo with ESTA-1 inhibited HL-60 adhesion by 80% at 100 nM ESTA-1 (Fig. 6A) (p<0.01). The IC₅₀ for the inhibition of this interaction was approximately 63 nM. These data indicate that the ESTA-1 interaction to E-selectin occurs through the sLe^x-binding site, further highlighting the potential therapeutic utility of ESTA-1 as an antagonist of E-selectin mediated adhesion. Lastly, we tested the cytotoxicity associated with ESTA-1 treatment in ES-Endo. ES-Endo were first incubated with doxycycline for 5 hours and then with increasing concentration of ESTA-1 (up to 200 nM) for 48 hours. MTT assay was performed to test cell viability. Incubation with up to 200 nM of ESTA-1 for 48 hours did not cause any visible morphological changes or a reduction of cell viability in ES-Endo cells (Fig. 6B).