Single molecule DNA sequencing in submicrometer channels: state of the art and future prospects

Abstract

We demonstrate a new method for single molecule DNA sequencing which is based upon detection and identification of single fluorescently labeled mononucleotide molecules degraded from DNA-strands in a cone shaped microcapillary with an inner diameter of 0.5 μm. The DNA was attached at an optical fiber via streptavidin/biotin binding and placed ∼50 μm in front of the detection area inside of the microcapillary. The 5′-biotinylated 218-mer model DNA sequence used in the experiments contained 6 fluorescently labeled cytosine and uridine residues, respectively, at well defined positions. The negatively charged mononucleotide molecules were released by addition of exonuclease I and moved towards the detection area by electrokinetic forces. Adsorption of mononucleotide molecules onto the capillary walls as well as the electroosmotic (EOF) flow was prevented by the use of a 3% polyvinyl pyrrolidone (PVP) matrix containing 0.1% Tween 20. For efficient excitation of the labeled mononucleotide molecules a short-pulse diode laser emitting at 638 nm with a repetition rate of 57 MHz was applied. We report on experiments where single-stranded model DNA molecules each containing 6 fluorescently labeled dCTP and dUTP residues were attached at the tip of a fiber, transferred into the microcapillary and degraded by addition of exonuclease I solution. In one experiment, the exonucleolytic cleavage of 5–6 model DNA molecules was observed. 86 photon bursts were detected (43 Cy5-dCMP and 43 MR121-dUMP) during 400 s and identified due to the characteristic fluorescence decay time of the labels of 1.43±0.19 ns (Cy5-dCMP), and 2.35±0.29 ns (MR121-dUMP). The cleavage rate of exonuclease I on single-stranded labeled DNA molecules was determined to 3–24 Hz under the applied experimental conditions. In addition, the observed burst count rate (signals/s) indicates nonprocessive behavior of exonuclease I on single-stranded labeled DNA.

Introduction

Since the first detection of single fluorescent molecules in solution by laser-induced fluorescence detection (Shera et al., 1990), several groups have developed the capability to detect and identify single fluorescent molecules in solution as they flow through a focused laser beam (for review see Eigen and Rigler, 1994, Barnes et al., 1995, Rigler, 1995, Keller et al., 1996, Goodwin et al., 1996, Nie and Zare, 1997, Weiss, 1999). As early as 1989, Keller and coworkers (Jett et al., 1989, Harding and Keller, 1992) proposed a new method for high-speed DNA sequencing based upon fluorescence detection of single molecules. In contrast to current DNA sequencing schemes where only relatively short DNA fragments of up to 1000 base pairs are sequenced, the method would allow to sequence a single fragment of DNA, several tens of kilobases (kb) or more in length, at the rate determined by the cleavage rate of the exonuclease enzyme. As originally proposed, the technique is based on the detection and identification of fluorescently labeled nucleotides in flowing sample streams as they are released sequentially from a DNA strand by an exonuclease enzyme. The DNA to be sequenced should be copied using a biotinylated primer, a DNA polymerase and the four nucleotide triphosphates (dNTPs), each containing a different fluorescent label which exhibits a characteristic laser-induced fluorescence. As a single biotinylated DNA fragment is bound to a microsphere coated with avidin or streptavidin the microsphere is transferred into a flowing sample stream by mechanical micromanipulation or optical trapping. Upon addition of an 3′→5′ exonuclease fluorescent monophosphate molecules (dNMPs) will be cleaved and transported to the detection area down stream. The fluorescence signal of each individual fluorescent dNMP molecule is detected as a fluorescence burst and identified by its characteristic fluorescence decay time using the technique of time-correlated single photon counting (TCSPC) (Wilkerson et al., 1993, Zander et al., 1996, Müller et al., 1996, Sauer et al., 1998) or spectral property. However, besides the problems associated with the complete enzymatic substitution of the native nucleotides by dye-labeled nucleotides and the definite selection of a single DNA strand, the detection and characterization of each released nucleotide with high accuracy is also full of problems. Furthermore, since the DNA sequence is determined by the order in which labeled dNMPs are detected, diffusional misordering of sequentially cleaved dNMPs has to be prevented, i.e. the enzymatic cutting rates, flow velocities and the distance to the detection volume have to be optimized (Pratt and Keller, 1993). To ensure the efficient detection of each dNMP hydrodynamic focusing of the sample stream in a sheath flow cuvette down to <10 μm was applied (Zarrin and Dovichi, 1985). To remove Rayleigh- and Raman scattering pulsed excitation in combination with time-gated detection was used (Shera et al., 1990). Under these conditions, fluorescent molecules released from an optically trapped microsphere have been detected with high efficiencies (Machara et al., 1998). On the other hand, single molecule fluorescence signal-to-background ratios (SBRs) can be drastically improved by using confocal excitation/detection techniques with detection volumes in the femtoliter (fl) range (Rigler et al., 1993, Mets and Rigler, 1994, Nie et al., 1994, Nie et al., 1995). Using confocal fluorescence microscopy individual fluorescent analyte molecules can be detected with SBRs >100. Unfortunately, there is a drawback associated with the use of such small volumes in applications requiring efficient detection of all analyte molecules, e.g. in single molecule DNA sequencing. To constrain all analyte molecules to flow through a volume with linear dimensions of <1 μm the detection volume must be confined by walls, i.e. a detection channel has to be used. While the refractive index differences at the outer walls of such channels can be matched by the use of the appropriate index-matching oil, the refractive index differences at the inner wall and deviations of the beam profile generally result in higher background rates and smaller photon bursts. In addition, the use of channels with such small volume-to-surface ratios results in strong adsorption of analyte molecules on the walls. Due to this dynamic adsorption process burst durations of up to 60 ms have been measured for single rhodamine 6G molecules in aqueous buffer in submicrometer channels (Lyon and Nie, 1997). To circumvent these problems polymethylmethacrylat (PMMA) microchannels with channel diameters of ∼10 μm in the detection area were employed (Dörre et al., 1997, Brinkmeier et al., 1997). The laser beam was shaped by a cylindrical lens and focused over the whole channel. To reduce Raman scattering the volume element was imaged onto a glass fiber bundle where seven fibers were aligned. Each fiber was connected to its own separate detector thus producing seven overlapping smaller volume elements.

However, the sequential counting of single rhodamine 6G molecules dissolved in ethylene glycol in a 1 μm capillary was successfully demonstrated (Zander and Drexhage, 1997). Hence, if the adsorption can be suppressed by addition of detergents or other additives, the use of submicrometer channels should allow a precise control of the movement of single molecules by electrokinetic or electroosmotic forces. Recently (Zander et al., 1998, Becker et al., 1999), we were able to demonstrate the time-resolved identification of individual fluorescent dyes as they flow through a microcapillary with an inner diameter of 500±200 nm applying an electrical tension of a few volts. Addition of a nonionic detergent (Tween 20) suppressed efficiently adsorption of molecules on the glass surface of the capillary and reduced the electroosmotic flow. Using a 3% polyvinyl pyrrolidone (PVP) matrix, single fluorescently labeled nucleotides were detected and identified with comparable efficiency in the microcapillary as they were enzymatically cleaved from DNA strands bound to a fiber (Sauer et al., 1999). Hence, this technique seems to be a valuable alternative for single molecule DNA sequencing. Nevertheless, the ultimate goal of the determination of the sequence of a single dye-labeled DNA fragment by laser-induced fluorescence detection has not been achieved.

In the present communication we demonstrate efficient detection and identification of single fluorescently labeled nucleotide molecules in a submicrometer capillary degraded from synthetic DNA strands by addition of exonuclease I. The 218-mer model sequence which contained six differently labeled cytosine and uridine residues, respectively, at well defined positions was bound via a biotin/streptavidin bond to a cone shaped etched optical fiber placed ∼50 μm in front of the detection area inside of the capillary. Enzymatic degradation was started upon addition of a diluted exonuclease I solution. The flow of the negatively charged cleaved nucleotide molecules (dNMPs) towards the detection area was established by electrokinetic forces.