To investigate the feasibility of using oligonucleotides in flow cytometry we describe a model system consisting of human neutrophil elastase (HNE) coated on 3.3 µm beads and a high affinity DNA ligand for HNE isolated by in vitro selection (SELEX). In this system the fluoresceinated DNA ligand was equally effective as an anti-HNE antibody in detecting HNE on beads. The location on and the chemistry of attachment of fluorescein to the DNA ligand is critical for the sensitivity of detection. DNA constructs in which fluorescein was conjugated via an ethylene glycol tether to either the 5′-end or near the 3′-end gave much higher signals than did probes with fluorescein directly conjugated to either end. Second-step staining with strepavidin-conjugated phycoerythrin was accomplished using a biotinylated DNA ligand in the initial staining of HNE beads. These data suggest that instead of, or in addition to, antibodies high affinity oligonucleotide probes can be useful in diagnostic applications based on flow cytometry.
Antibodies have been commonly used in flow cytometry, a technique that is unique in its capability to perform simultaneous multiparameter analysis and to separate (or sort) unique cell types from heterogeneous mixtures. In flow cytometry particles or cells prestained with a target-specific ligand (typically an antibody) conjugated to a fluorophore are passed through a region where they intersect a laser beam. The characteristics of scattered light provide information on the size, shape and granularity of particles, whereas the emitted fluorescence indicates the binding properties of the ligand (1). Flow cytometers are capable of measuring these properties on thousands of cells in a few seconds. Flow cytometry, once used mainly as a research tool, has now made its way into clinical applications. A widely used application is the monitoring of HIV-infected patients to obtain their absolute CD4 counts (number of T-helper cells), a surrogate marker for the progression of the disease (2). In addition, the technique is also being used in immunophenotyping in leukemia and bone marrow screening in patients who undergo transplantation (2).
Systematic Evolution of Ligands by EXponential enrichment (SELEX) is an in vitro selection/amplification technique to enrich specific single-stranded oligonucleotide sequences with desired properties from large random sequence libraries (3,4). The SELEX process has permitted the isolation of specific and high affinity oligonucleotide ligands for a wide range of target molecules, including nucleic acid binding proteins, non-nucleic acid binding proteins and small organic molecules (reviewed in 5). In most cases SELEX-derived ligands have similar affinities and specificities to those of antibodies.
In the present study, using a model system, we describe the behavior of a SELEX-derived high affinity oligonucleotide ligand in flow cytometry. In this model polystyrene beads coated with human neutrophil elastase (HNE) and a DNA sequence that binds HNE with high affinity (6) were used. The performance of the DNA ligand was compared with an anti-HNE antibody in detecting HNE on beads under flow cytometry. By synthesizing various DNA constructs containing one or two fluorescein molecules at different positions of the sequence we found that the site and chemistry of attachment of fluorescein are important for detection. This study demonstrates the feasibility of the use of high affinity oligonucleotide ligands in flow cytometry.
Materials and Methods
A spacer phosphoramidite that introduces an 18 atom spacer arm (six ethylene glycol units) into oligonucleotides, fluorescein-ON phosphoramidite and symmetrical branching 3′-3′ linking CPG were purchased from Clontech (Palo Alto, CA). Mouse anti-HNE antibody was obtained from Dako Corp. (Carpinteria, CA). Rat anti-mouse IgG1, clone X56 (from Becton Dickinson) was labeled with fluorescein isothiocyanate (FITC) to contain approximately one fluorescein per molecule of antibody. Fluoricon™ Polystyrene Assay Particles (3.3 µm) were obtained from IDEXX Laboratories Inc. (Westbrook, ME). HNE was obtained as a salt-free lyophylized solid from Athens Research and Technology (Athens, GA). All other reagents were of analytical grade. Enzymes were purchased from commercial sources.
Oligonucleotides containing fluorescein were chemically synthesized by standard solid phase chemistry using cyanoethyl phosphosphoramidites. Symmetrical branching 3′-3′ linking CPG was used to synthesize the tail-to-tail dimer. After deprotection DNA sequences were purified by denaturing polyacrylamide gel electrophoresis to ensure size homogeneity.
Preparation of HNE beads
Polystyrene beads were successively washed with phosphate-buffered saline (PBS) containing 0.1% SDS, PBS containing 0.01% Tween 20 and then with acetate buffer (50 µM NaOAc, 0.15 M NaCl, pH 5.65). Beads (0.5% solids) in acetate buffer were coated with HNE (0.5 µg/cm2 bead surface area) for 30 min and then washed and suspended (∼5 × 105 beads/µl) in acetate buffer containing 2% bovine serum albumin (BSA). HNE-coated beads were stored at 4°C until use.
Ligand binding to HNE in solution
DNA ligands containing fluorescein either internally or near their 3′-ends were radiolabeled at their 5′-ends with [γ-32P] ATP and T4 polynucleotide kinase. DNAs carrying fluorescein at their 5′-ends were radiolabeled at their 3′-ends with [α-32P]ddATP and terminal deoxynucleotidyl transferase. Radiolabeled DNAs were purified by denaturing polyacrylamide gel electrophoresis. Gel-purified DNAs resuspended at a final concentration of 1 nM in the standard binding buffer (150 mM NaCl, 100 mM Tris-HCl, pH 7.0, 2 mM MgCl2 and 6 mM KCl) were heated to 70°C for 3 min and cooled to room temperature to facilitate secondary structure formation.
Gel-purified radiolabeled DNA (<50 pM) was incubated with varying amounts of HNE in the binding buffer containing 0.02% human serum albumin (HSA) for 10 min at 37° C. DNA/protein mixtures were filtered through pre-wet nitrocellulose filters (0.2 µm) and the filters immediately washed with 5 ml binding buffer. The radioactivity retained on the filters was counted. Retention of DNA on the filters in the absence of HNE was determined and used for background correction. Assuming equimolar binding of DNA to HNE, a non-linear least squares method was used to fit the data using the software package Kaleidagraph (Synergy Software, Reading, PA) to obtain the equilibrium dissociation constant Kd (6).
Staining of HNE-coated beads was achieved by incubating beads (∼3 × 105) with varying concentrations of fuoresceinated oligonucleotides in 50 µl binding buffer (100 mM Tris-HCl, pH 7.0, 150 mM NaCl, 6 mM KCl, 2 mM MgCl2, 0.5% BSA) for 30 min at room temperature. The beads were then washed with 2 ml binding buffer and suspended in 0.5 ml for analysis. Staining with mouse anti-HNE was accomplished by the same method except that bound antibody was detected by a second-step staining with FITC-labeled rat anti-mouse antibody (X56-FITC). To obtain comparative measurements between the antibody and DNA ligand X-56 was stoichiometrically labeled with FITC. Analysis was performed on a FACScan™ flow cytometer (Becton Dickinson). The sensitivity of the flow cytometer was adjusted so that the autofuorescence of the unstained beads was at the very low end of the scale (first decade).
For the determination of Kd values the observed fluorescence signal was plotted against the log concentration of the fuoresceinated ligand.
UV absorption and fluorescence measurements
The number of fluoresceins per oligonucleotide was determined spectrophotometrically. An extinction coefficient of 7200 M−1 cm−1 at 494 nm was used for both free fluorescein in solution and fluorescein tethered to DNA. DNA concentrations were based on the extinction coefficients 13800, 6500, 10500 and 7900 M−1 cm−1 at 260 nm for A, C, G and T respectively. The relative quantum yield for each fluorescein-labeled oligonucleotide was obtained by measuring the fluorescence (excitation at 494 nm and emission at 518 nm) in binding buffer relative to the fluorescence of a solution of FITC. The concentration of each oligonucleotide was adjusted so that the absorbance at 494 nm was approximately the same and small differences from the absorbance of the reference FITC solution were normalized (fluorescence × gain−1 × OD494−1).
A SELEX-derived 45 nt DNA sequence that binds HNE with high affinity (Kd = 17 nM) and specificity has been previously characterized (6). NMR data and comparative sequence analysis of other members of this sequence family supported the folding of the sequence into a G quartet structure with duplex ends. Based on this sequence we made several oligonucleotides containing either one or two fluorescein molecules (shown in Fig. 1). In these constructs fluorescein was attached to either the 5′-end or as the penultimate residue of the 3′-end of the sequence (DNA-5F and DNA-3F respectively). In DNA-LNK-5F and DNA-LNK-3F fluorescein was linked through an ethylene glycol linker (12 units of ethylene glycol). Two dimers of the sequence, each containing two fluorescein molecules, were also synthesized. These dimers were linked by ethylene glycol linkers in either head-to-tail (DNA-DIMER-HT) or tail-to-tail orientation (DNA-DIMER-TT). Complementary regions that base pair in the proposed structure were extended in DNA-EXT and the extension contained two fluoresceins spaced by 7 nt.
To investigate the behavior in flow cytometry of the above oligonucleotides designed to bind HNE we prepared polystyrene beads coated with the target protein. The binding of DNA-3F and DNA-LNK-3F to HNE beads detected by flow cytometric analysis is shown in Figure 2a (panels II and III respectively). The addition of either ligand generated a clear change in the fluorescence intensity compared with the autofluroscence signal of beads alone (panel I). The magnitude of the signal observed with DNA-LNK-3F is similar to that observed with an anti-HNE antibody (panel IV). The fluorescence intensity of DNA-3F is 10-fold less than that produced by DNA-LNK-3F. The only difference between these two sequences is the nature of the fluorescein linkage (see Fig. 1). DNA-LNK-3F, which produced a high fluorescence signal intensity, has fluorescein attached through an ethylene glycol tether.
A comparative analysis of the behavior of different sequences in flow cytometry was carried out by measuring the fluorescence intensity produced by each construct as a function of the ligand concentration (Fig. 2b). Based on these results, DNA sequences can be categorized into three groups; molecules that generate high, moderate and low signals. Sequences that gave high signals included a dimer (DNA-DIMER-TT) containing two fluorescein molecules and a monomer (DNA-LNK-3F) in which a single fluorescein was attached near the 3′-end by an ethylene glycol linker. Monomeric forms in which fluorescein was directly attached either to the 5′-end (DNA-5F) or near the 3′-end (DNA-3F), as well as DNA-EXT, gave very poor signals in flow cytometry. An intermediate signal was produced by DNA-LNK-5F and DNA-DIMER-HT. The data indicate that DNA constructs in which fluorescein was attached through an ethylene glycol linker have higher fluorescence signal intensities than those in which fluorescein was directly coupled to the sequence.
In all cases the signal is saturable, allowing the estimation of the Kd for each construct for its interaction with HNE on beads (Table 1). The unselected random sequence pool (fluoresceinated at the 5′-end via an ethelene glycol linker) did not show detectable binding to HNE beads (data not shown). As a comparision, we also quantified the affinities of these sequences binding to HNE in solution by the nitrocellulose filter binding technique (Table 1). The high affinity binding of all sequences to HNE indicates that tight binding was not affected by the attachment of fluorescein or ethylene glycol linker or both. The Kd values of ligands for their interactions with HNE in solution were generally lower than those obtained with HNE presented on beads. This observation could be due to the adsorption of HNE onto the solid surface, leading to either partial blocking of the DNA binding site, a change in conformation of the protein adsorbed onto the solid surface or both. In a separate binding reaction we used varying concentrations of radiolabeled DNA-3F and a fixed amount of beads to calculate the Kd by measuring the radioactivity retained on the beads. The Kd value obtained by this approach was in agreement with that obtained by flow cytometry, indicating that the binding interaction was independent of the type of signal used to calculate dissociation constants. The two dimers, designed to investigate the effect of dimerization of the sequence, bound HNE better than the monomers, as indicated by their Kd values measured by both techniques.
Flow cytometry was used to investigate whether the DNA ligand (DNA-LNK-3F) and anti-HNE antibody compete for binding to HNE on beads. As shown in Figure 3a, the DNA ligand showed significantly lower binding to HNE beads preincubated with the antibody (squares) as compared with binding to HNE beads in the absence of antibody (circles). This result suggests that the binding sites of the DNA ligand and the antibody overlap. As shown (Fig. 3, triangles), the HNE-specific DNA ligand did not bind to CD3-attached beads, indicating target specificity.
Certain flow cytometric applications utilize a two-step staining procedure in which the primary antibody bound to a target is detected by a secondary antibody. The data for antibody binding (Fig. 3b, squares) was obtained by second-step staining of anti-HNE antibody with FITC-labeled X-56 antibody. The X-56 antibody was labeled with an average of one FITC per molecule for comparison with staining by the oligonucleotide ligand, also containing one fluorescein per molecule. To investigate whether an oligonucleotide ligand could also be used in a two-step staining procedure we designed an experiment in which binding of a biotinylated DNA ligand (biotin was attached to the 3′-end through a linker) to HNE beads was detected by streptavidin conjugated to phycoerythrin (SA-PE). In this experiment the SA-PE conjugate served as a substitute for a secondary antibody and the fluorescence intensity of the conjugate on HNE beads was measured as a function of the biotinylated oligonucleotide. As with the case for antibodies, saturated binding of biotinylated DNA was observed with SA-PE (Fig. 3b; circles), demonstrating the feasibility of second-step staining for oligonucleotide ligands as well.
In the present study we show that the attachment of fluorescein at different locations on a DNA ligand for HNE did not decrease its binding affinity for its protein target. This result indicates that the two structural motifs of DNA proposed to be involved in protein binding (6) were not grossly altered in these constructs. Although the different DNA constructs used in the present study retained their high affinity binding to HNE, fluorescence intensities were quite sensitive to the position and the way in which the fluorescein molecule was attached. Certain sequences gave barely detectable signals in flow cytometry, whereas others generated signals that were comparable with those of an antibody.
Differences in signal intensity observed in flow cytometry of various constructs do not correlate with their binding affinities, nor do they correlate with the number or the position of fluoresceins in each construct. To investigate whether these observations could be due to the intrinsic fluorescence of each sequence we analyzed fluorescence characteristics such as the relative quantum yield and the average number of fluoresceins per molecule (data not shown). These values did not correlate with the observed signal intensities of these ligands, suggesting that the fluorescence behavior of these molecules is complex. The proposed secondary structure of the sequence predicts that the two ends are close to each other and thus the local environment of the two ends is expected to be the same (6). However, fluoresceins attached to either the 3′- or 5′-ends exhibit significantly different signal intensities. This could be due to differences in the way in which the two ends of the oligonucleotide interact with amino acid residues on the surface of HNE, changing the properties of the local environment of the fluorophore and affecting the quantum yield of fluorescence. In general, however, attachment of fluorescein through an ethylene glycol linker improved signal intensity. The generality of the observations on the position of fluorescein on this ligand may not be applicable to a different ligand. However, It is not difficult to try several oligonucleotide constructs for a given ligand to identify the most suitable position for fluoresceination that gives a maximal signal upon binding to the target. Similarly, it is also possible to identify a position for a fluorophore in a ligand that gives maximal quenching upon interaction with a target, a feature that may be useful in other diagnostic applications, such as homogeneous detection.
SELEX-derived oligonucleotide ligands exhibit monovalent interaction with their targets, whereas naturally occuring antibodies are bivalent or multivalent. The dimers used in the study were designed to provide a bivalent target interaction. The two dimers were synthesized with a flexible ethylene glycol linker between the two monomeric units. Affinities of the two dimers were higher than the monomeric forms, probably due to an avidity contribution to the molecular interaction.
For certain flow cytometric applications, where amplification of the signal is necessary, second-step staining of the primary probe is advantageous. Using biotinylated DNA and SA-PE as reagents for primary and secondary staining respectively we demonstrated the feasibility of using oligonucleotides in second-step staining. The affinity of the DNA-HNE interaction measured by second-step staining was lower than that measured directly (compare circles in Fig. 3 a and b). The decrease in affinity could be due to the streptavidin-biotin interaction influencing binding between HNE and DNA. This effect may be overcome by changing the presentation of biotin on the oligo-nucleotide.
In antibodies relatively small target recognition (antigen binding) sites are localized in a defined region in a structurally large protein. This feature allows the attachment of reporter molecules to sites away from those required for target recognition. SELEX-derived ligands, on the other hand, are generally small and contain only the necessary functional groups to interact with their targets and therefore may require additional appendages to bear reporter molecules without loosing their function. The ethylene glycol linkages used in this study appear to meet this requirement. Although antibodies carrying multiple fluoresceins are brighter than probes with a single fluorescein, fluorescence quenching limits the brightness of probes containing a large number of fluoresceins. We have stained cells with a SELEX-derived oligonucleotide containing a single fluorescein. The intensity of oligonucleotide staining is ∼5-fold less than that obtained with a commercially available antibody containing three to five fluoresceins per molecule (unpublished data), indicating that probes with more than one fluorescein molecule are generally better. Oligonucleotide ligands are not resticted in the number of reporter molecules that can be attached to them. Various approaches can be deviced to incorporate more than one fluorescein per oligonucleotide molecule (e.g. complexation of a biotinylated oligonucleotide ligand with streptavidin labeled with several fluorosceins) to enhance the sensitivity of oligonucleo-tide-based probes in flow cytometry (unpublished results).
Nucleic acid-based probes in flow cytometry may offer certain benefits. (i) Unlike antibodies, oligonucleotides derivatized with various molecular probes (fluorophores) at defined positions can be obtained by direct chemical synthesis. Functionalities such as amino and thiol groups that are useful to derivatize oligonucleotides can also be placed in an oligonucleotide chain during synthesis (7). (ii) Although most flow cytometric studies are based on cell surface proteins, detection of intracellular proteins has also been demonstrated (8). The small size of SELEX-derived oligonucleotide probes (<20 kDa) may be advantageous over structurally large antibodies (∼160 kDa) when intracellular proteins are targeted. Nuclease degradation of oligonucleotide-based probes can be overcome by using chemically modified oligonucleotide random libraries for selection (9,10). (iii) The presence of the Fc region on antibodies causes undesirable non-specific binding to Fc receptors on cells, creating difficulties in the interpretation of flow cytometric data (11). The use of oligonucleotides may overcome this problem. (iv) Antibody-based reagents are generally stored at low temperature due to their thermal sensitivity. Nucleic acid ligands are generally thermostable and, further, their native structure can be restored after thermal denaturation, allowing them to be shipped and store at room temperature.
Overall, the data presented here suggests that oligonucleotide-based high affinity ligands may be useful as probes in flow cytometric applications. The development of monoclonal antibodies to specific cell surface proteins has significantly contributed to the diversification of flow cytometric applications in both research and clinical applications. The feasibility of the use of SELEX-derived oligonucleotides in flow cytometry is expected to expand the repertoire of available probes, making the technique even more versatile.
We thank Barry Polisky, Larry Gold and Dan Nieuwlandt for critical reading of the manuscript.