Illustration by Don Bishop
The development of the polymerase chain reaction (PCR) technique, which amplifies genetic materials to detectable levels, mitigates problems in genetic analysis caused by limited sample amounts. PCR can also significantly decrease the time from onset to diagnosis of certain diseases. On the down side, the technique is time-consuming, not quantitative, has stringent conditions, and can still fail to produce any amplification.
Single-molecule DNA screening is a powerful complement to PCR. First, researchers need sensitivity beyond what is now available to establish molecular profiles for species that are present at low concentrations in tissues or a single cell. Second, it is likely that biochemical changes can be recognized well before physical changes occur in the progress of a disease, so the technique may help with diagnosis. Profiling individual cells and probing single DNA molecules reveals and quantifies the diversity of a population, which may tell what fraction of cells need to be infected (mutated) before a disease becomes unchecked. In the specific example of HIV detection, the most sensitive PCR tests require 20 to 50 copies per mL. Recent case studies, however, have documented a need to detect levels as low as one copy per mL. high-throughput detection
Single-molecule detection (SMD) methods based on fluorescence are limited by the presence of background noise. Hence, researchers often use single-channel detectors. Optical pinholes placed in strategic locations eliminate out-of-focus emission and further improve the signal-to-noise ratio (SNR). This type of setup requires extremely low concentrations to prevent two molecules from being present in the detection volume simultaneously.1 While such tactics enable the undeniable detection of individual fluorescent molecules, in concert they squelch throughput.
Using off-the-shelf components, our group has developed a method that addresses the issue of high throughput in single-molecule analysis. SMD requires more than just getting a sufficient signal; we also need high specificity to pinpoint a target, for example through electrophoretic mobility provide such selectivity. At the single-molecule level, researchers have demonstrated electrophoretic mobility in a micrometer-sized flow stream by correlating the photon bursts created by two axially separated laser beams. With this method, only one DNA molecule at a time can be probed, and the measurement time is inversely related to the distance of separation between the two laser beams. Researchers have also used the relative fluorescence intensities of double-stranded DNA stained with intercalating dyes to size DNA, but the precision is quite poor.
The experimental configuration we developed for single-molecule electrophoresis imaging is a miniaturized version of a typical capillary electrophoresis setup. Our system uses an argon ion laser emitting at 488 nm, scientific-grade microscopes, and a scientific-grade CCD camera featuring a 512 x 512 detector with 25 µm2 pixels.2 The system can simultaneously illuminate and image many distinct molecules and record their time-dependent motion in free solution for subsequent analysis.
Figure 1. An SMD system uses a flattened laser beam to irradiate a sample of DNA in solution flowing through the capillary. A dye attached to the DNA fluoresces when excited by the laser light, and the image is captured by a CCD camera (not shown).
A DC electric field (-80 V/cm) drives the dye-labeled DNA through the capillary. Two cylindrical lenses shape the laser beam into a planar sheet. The beam irradiates the capillary window when we open the laser shutter. In experiments, the total laser power at the capillary is 14 mW (see figure 1). The laser produces a Gaussian power distribution along the x- and z-axes. The irradiated region in the capillary measures about 75-µm wide (x-axis), 75-µm long (y-axis), and about 10-µm thick (z-axis). The axis of the laser beam is perpendicular to one side of the square capillary.
We captured images of the irradiated region through a 20X, 0.75-NA Plan Apochromat microscope objective with the CCD camera cooled to -35°C. A 488-nm holographic notch filter with an optical density of better than six was placed between the objective and the CCD camera, which was synchronized with the laser shutter. The digitization rate of the CCD camera was 1 MHz (16 bits). demonstrating advantages
The DNA assay described can be adapted to detect deletions, for example in the mitochondrial DNA (mtDNA) of a human, which have been linked to various diseases.3 The percentage of mutant and normal mtDNA is often indicative of the degree of severity or progression of a disease. Single-molecule screening eliminates the need for PCR amplification, yielding more quantitative information in a significantly shorter time.
In our study, we mixed human mtDNA and the corresponding fragment resulting from a deletion and drove the sample through the observation channel. We determined the mobility of each molecule by looking at consecutive single-molecule images. We were able to achieve unambiguous identification of the normal versus deleted DNA.
Figure 2. The system first captures the raw image of fluorescing DNA molecules (left), then calculates the position of each molecule according to intensity maxima. Correlating results from several consecutive images allows the system to plot the distribution of mobilities identifying target molecules. The relative standard deviation of mobilities in this figure is 6.7%, while the difference in the mobilities of 6.1 kilobase vs. 16.5 kilobase mitochondrial DNA is 25%.
We can also automate the data analysis. Software can estimate the mobility of a given molecule in terms of the number of pixels N it will move between frames. From the first real image in a series, the program calculates a new virtual image in which each molecular spot is translated N pixels. The program multiplies the calculated virtual image by the next real image in the data set. The resulting correlated image picks up those molecular spots that fit the predicted electrophoretic motion (see figure 2).
We found a narrow distribution in the numbers of molecules identified as a function of the prescribed shift N. This implies that the method provides good discrimination against other molecules not traveling at N pixels per frame. Such a calculation takes only a few seconds on a desktop computer. measuring concentration
If we can detect single molecules of a target DNA, then we should, in theory, be able to detect them at an infinitely low concentration. However, this does not mean that we are able to quantify the concentration itself. To achieve this task, we need to count the number of target DNA molecules in a unit volume of the sample solution with 100% accuracy.
The two strands of DNA are held together in a double helix shape by the bonds between two base compoundseither adenine and thymine or guanine and cytosine. Base pairs (bp) are a convenient measurement for the length of a DNA molecule.
The first problem in measuring concentration was labeling DNA with YOYO-1, a dye that fluoresces when bonded to DNA at an appropriate ratio, for example, one dye molecule per five base pairs for DNA at high concentrations (10-10 M). For our system, we diluted the labeled DNA to about 10-13 M, a concentration low enough for single-molecule detection. In quantitative analysis, DNA must be efficiently labeled with YOYO-1 at a constant concentration because the concentration of DNA is unknown and low. At the same time, the concentration of YOYO-1 must be minimized so that the background intensity from free YOYO-1 is as low as possible.
We found by experiment that YOYO-1 at concentrations of better than 10-8 M is necessary to adequately label β-actin DNA (838 bp) at less than 10-12 M. Almost all of the YOYO-1 dye molecules thus remain free after the reaction reaches equilibrium. We expect that the β-actin DNA is uniformly and sufficiently labeled at the concentrations we tested (10-16 to 10-12 M). This means that the number of labeled β-actin DNA molecules in a unit volume of any of the sample solutions should be linearly proportional to the concentration of β-actin DNA.
The second problem is the presence of fluorescent particles other than DNA. Although YOYO-1 is supposed to be essentially non-fluorescent in the absence of nucleic acids, we observed many fluorescent particles when we applied the SMD system to buffer solutions containing YOYO-1 but not DNA. The positively charged YOYO-1 probably becomes associated with some impurities that carry a negative charge, such as dust and educts in the buffer solution.
Such impurities present a sizable background for SMD and limit the sensitivity of such measurements. We were unable to remove them by filtering the buffer solution and the YOYO-1 stock solution through 0.22-µm filters. However, we serendipitously discovered that adding poly(ethylene oxide) (PEO) to the solution greatly reduced the number of labeled impurity particles, while the quantity and the intensity of the labeled β-actin DNA remained unchanged.
We hypothesize that the affinity between YOYO-1 and the PEO polymer is stronger than that between YOYO-1 and the impurities but weaker than that between YOYO-1 and DNA. Molecules of YOYO-1 dye associated with the PEO polymer are not detected because they are uniformly dispersed throughout the solution, rather than concentrated as they are with each DNA molecule. Furthermore, we found that the difference in mobility during electrophoresis in the 0.3% PEO polymer solution allowed us to discriminate the labeled impurities from the labeled β-actin DNA. This allowed us to count only the β-actin DNA molecules.
The fluorescence intensity of the 73 molecules of β-actin DNA in the images was 615 counts (arbitrary unit) on the average, while the background intensity was 550 ± 4.8 counts. The SNR for β-actin DNA molecules is thus 14. Each molecule of β-actin DNA (838 bp) is expected to carry about 168 YOYO-1 molecules when the domain of the YOYO-1 binding site on the DNA is 5 bp/dye molecule. The limit on the detectable size for DNA molecules with this system is estimated to be 120 bp (with an SNR of 2).
The number of β-actin DNA molecules and the number of impurities constitute the signal and noise of the quantitative assay, respectively. Discrimination based on electrophoretic mobility improved the SNR. We estimate the limit of quantifiable concentration of β-actin DNA to be 3 x 10-16 M (SNR = 2). This corresponds to a limit on the quantifiable amount of 3 x 10-21 M, or 1800 molecules of β-actin DNA for the 10-µL sample volume used in this study. This sensitivity is the highest of any method for the quantitative assay of nucleic acids in which amplification is not used.
SMD has matured to the point that it is possible to pick out one mutant DNA molecule in a sea of normal molecules and measure the concentration of such molecules down to the attomolar range. As we begin to correlate genetic information with disease states, SMD can play an important role in early clinical diagnosis. oe
The Ames Laboratory is operated for the U.S. Department of Energy by Iowa State University under contract No. W-7405-Eng-82. This work was supported by the Director of Science, Office of Biological and Environmental Research, Division of Chemical Sciences, and by the National Institutes of Health.
1. S. Weiss, Science 283, p. 1676 (1999).
2. E. S. Yeung, The Chemical Record, Vol 1; M. Kobayashi, Ed. The Japan Chemical Journal Forum and John Wiley & Sons: Tokyo, Japan, p. 123 (2001).
3. Y. Ma et al., Electrophoresis, 22, p. 421 (2001).
4. T. Anazawa, H. Matsunaga, et al., Anal. Chem., 74, p. 5033 (2002).
Edward Yeung is a distinguished professor in Liberal Arts and Sciences at Iowa State University, Ames, IA.