Anal. Chem. 1997, 69, 2115-2121
High-Efficiency Molecular Counting in Solution: Single-Molecule Detection in Electrodynamically Focused Microdroplet Streams N. Lermer, M. D. Barnes,* C.-Y. Kung, W. B. Whitten, and J. M. Ramsey
Chemical and Analytical Sciences Division, Oak Ridge National Laboratory, Mail Stop 6142, P.O. Box 2008, Oak Ridge, Tennessee 37831
We report fluorescence detection of individual rhodamine 6G molecules using a linear quadrupole to focus streams of microdroplets through the waist of a counterpropagating cw Ar+ laser. Since the terminal velocity scales as the square of the droplet diameter, the droplet-laser interaction time was “tunable” between 5 and 200 ms by using water samples spiked with a small, variable (2-5% v/v) amount of glycerol. Fluorescence bursts from droplets containing single molecules were clearly distinguished from the blanks in real time with an average signal-tonoise ratio of about 10, limited primarily by photobleaching and droplet size fluctuations (99% confidence) were obtained for 100 and 15 fM rhodamine 6G solutions, in good agreement with detailed theoretical calculations and statistical limitations. Since the first reports of single-molecule detection in liquids several years ago1-4 several exciting new developments and applications of single-molecule fluorescence detection techniques have been described.5-18 Perhaps most notable has been the development of far-field microscopy techniques19-21 as a probe of individual solvated molecules employing low-power continuous (1) Hirschfeld, T. Appl. Opt. 1976, 15, 2965-2966. (2) Nguyen, D. C.; Jett, J. H.; Keller, R. A.; Martin, J. C. Anal. Chem. 1987, 59, 2158-2161. (3) Peck, K.; Stryer, L.; Glazer, A. N.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1989, 86, 4087-4091. (4) Shera, E. B.; Seitzinger, N. K.; Davis, L. M.; Keller, R. A.; Soper, S. A. Chem. Phys. Lett. 1990, 174, 553-557. (5) Soper, S. A.; Davis, L. M.; Shera, E. B. J. Opt. Soc. Am. B 1992, 9, 17611769. (6) Li, L. Q.; Davis, L. M. Rev. Sci. Instrum. 1993, 64, 1524-1529. (7) Castro, A.; Fairfield, F. R.; Shera, E. B. Anal. Chem. 1993, 65, 849-852. (8) Soper, S. A.; Mattingly, Q. L.; Vegunta, P. Anal. Chem. 1993, 65, 740-747. (9) Wilkerson, C. W.; Goodwin, P. M.; Ambrose, W. P.; Martin, J. C.; Keller, R. A. Appl. Phys. Lett. 1993, 62, 2030-203. (10) Lee, Y. H.; Maus, R. G.; Smith, B. W.; Winefordner, J. D. Anal. Chem. 1994, 66, 4142-4149. (11) Ishikawa, M.; Hirano, K.; Hayakawa, T.; Hosoi, S.; Brenner, S. Jpn. J. Appl. Phys. 1994, 33, 1571-1576. (12) Castro, A.; Shera, E. B. Anal. Chem. 1995, 67, 3181-3186. Appl. Opt. 1995, 34, 3218-3222. (13) Haab, B. B.; Mathies, R. A. Anal. Chem. 1995, 67, 3253-3260. (14) Li, L. Q.; Davis, L. M. Appl. Opt. 1995, 34, 3208-3217. (15) Mertz, J.; Xu, C.; Webb, W. W. Opt. Lett. 1995, 20, 2532-2534. (16) Dickson, R. M.; Norris, D. J.; Tzeng, Y.-L.; Moerner, W. E. Science 1996, 274, 966-969. (17) Xue, Q.; Yeung, E. S. Nature 1995, 373, 681-683. S0003-2700(97)00093-0 CCC: $14.00
© 1997 American Chemical Society
wave (cw) laser excitation and extremely small (e1 fL) illumination volumes. A principal drawback, however, common to all single-molecule detection techniques where the excitation volume is defined by the laser, is the difficulty in controlling, or arbitrarily extending, the laser-molecule interaction time. We have shown previously that this problem can be overcome by using microdroplet levitation techniques, where the laser-molecule interaction time can be easily extended to exceed the photochemical lifetime of the molecule.22,23 However, such a technique has obvious disadvantages in terms of sample throughput. It is clearly desirable to develop an approach to single-molecule detection in solution which offers both experimental control over the lasermolecule interaction time and a reasonable (5-10 pL/s) sample throughput rate. In this paper, we report observation of single-molecule fluorescence in electrodynamically focused microdroplet streams (positional stability to within ∼1 µm in the radial dimension) using a linear quadrupole. These experiments employ a cw Ar+ laser as an excitation source and relatively simple gated photon-counting techniques to detect the fluorescence. Two important advantages are realized by this technique. First, because the target molecules are confined to droplets, every molecule can be brought to the excitation region and probed. Second, since the droplets fall with a terminal velocity which scales as the square of the diameter, the interaction time can be “tuned” (5-200 ms) through small changes in the droplet size. In the experimental configuration described here, single-molecule fluorescence signals were limited primarily by photobleaching, with signal-to-noise ratios comparable to those obtained in confocal fluorescence microscopy techniques without limitations or complications due to fluorophore diffusion. In many potential applications of single-molecule detection technology, a primary issue is the optimization of the molecular detection efficiency (MDE), or, the fraction of molecules in an ultradilute sample flowing through the apparatus which are actually detected. Factors which contribute to the MDE can be (18) See also the recent review: Keller, R. A.; Ambrose, W. P.; Goodwin, P. M.; Jett, J. H.; Martin, J. C.; Wu, M. Appl. Spectrosc. 1996, 50, A12-A32 and references cited therein. (19) Mets, U.; Rigler, R. J. Fluorescence 1994, 4, 259. Eigen, M.; Rigler, R. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 5740-5747. Rigler, R. J. Biotechnol. 1995, 41, 177-186. (20) Nie, S.; Chiu, D. T.; Zare, R. N. Science 1994, 266, 1018-1021; Anal. Chem 1995, 67, 2849-2857. (21) Sauer, M.; Drexhage, K. H.; Zander, C.; Wolfrum, J. Chem. Phys. Lett. 1996, 254, 223-228. (22) Ng, K. C.; Whitten, W. B.; Arnold, S.; Ramsey, J. M. Anal. Chem. 1992, 64, 2914-2919. (23) Barnes, M. D.; Ng, K. C.; Whitten, W. B.; Ramsey, J. M. Anal. Chem. 1993, 65, 2360-2365.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997 2115
divided roughly into experimental and photophysical categories; minimizing experimental limitations has been the subject of a great deal of effort by several groups, with methods including capillary confinement of the sample stream,10 sample delivery into a flowing stream through a submicrometer diameter orifice,14 and molecular confinement using microdroplets23 and, recently, in poly(acrylamide) gels.16 However, for single-chromophore molecules such as rhodamine 6G (R6G), the nonzero overlap between the exponential photocount distribution24 associated with the molecule of interest and the background photocount distribution precludes unit MDE (for any nonzero detection threshold), even if every molecule in the sample is probed.25 It is straightforward to show that the MDE (for single-chromophore molecules) is approximately 1 - ∫t0 exp(-x/〈n〉) dx, where x is the number of photocounts, the parameter t is an arbitrary detection threshold, and 〈n〉 is the average number of photocounts measured per molecule.26 Implicit in the above (approximate) expression is that a molecular detection efficiency is not precisely defined without a confidence limit (threshold value) which specifies the error probability associated with each measurement. If the background and (single-molecule) signal photocount statistics are characterized by Gaussian and exponential distributions, respectively, with an average signal-to-noise ratio of 10, a 3σ threshold (99.9% confidence) results in an MDE of about 80%. Realization of an MDE g95% with the same threshold criterion requires an average signal-to-noise ratio of 55. The limitation imposed on the MDE by photocount statistics is fundamental representing an approximate upper bound on the MDE in any experiment. We have previously shown that microdroplets are an interesting sample medium for single-molecule detection in solution, both as a potential analytical tool22,23 and as a vehicle for exploring single-molecule quantum optics.27 Because the molecules being probed are confined to droplets which, in turn, define the illumination volume, effects of molecular diffusion on singlemolecule signals are small. By levitating droplets in an electrodynamic trap,22 the laser-molecule interaction time can be easily extended to exceed the photochemical lifetime, which permits observation of photobleaching of individual molecules as discrete steps in the fluorescence signal versus time. This enhances the ability to discriminate single-molecule fluorescence events from (blank) background or multiple-molecule events, since the discrete, or “digital”, nature of the fluorescence signal versus time is a unique signature of single-molecule events. In addition, since the background (primarily Raman scattering from the droplet) signals were relatively small and very close to shot-noise limited, discrimination of single-molecule events from blanks was insensitive to droplet size (between 8 and 15 µm diameter).23 Our earlier results demonstrated the feasibility of “digital molecular analysis” of ultradilute solutions in which a sample would be segmented into many hundreds or thousands of “pieces” (droplets) to be analyzed individually. In an extremely dilute concentration regime, where the probability of multiple-molecule (24) Whitten, W. B.; Ramsey, J. M. Appl. Spectrosc. 1992, 46, 1587-1589. (25) Note that this is not the case with multiple-chromophore molecules such as β-phycoerythrin, where the overlap between signal and background photocount distributions can be orders of magnitude smaller (see refs 24 and 22). (26) Barnes, M. D.; Whitten, W. B.; Ramsey, J. M. Anal. Chem. 1995, 67, A418A423. (27) Barnes, M. D.; Kung, C.-Y.; Whitten, W. B.; Ramsey, J. M.; Arnold, S.; Holler, S. Phys. Rev. Lett. 1996, 76, 3931-3934.
2116
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
events is negligible, measurements on each droplet represent separate, discrete experiments whose result would be a “yes” or “no”, corresponding to the presence or absence of a molecule, respectively. In principle, a significant advantage in terms of concentration detection limits is possible with such an approach, since the fraction of molecules in the sample which are detected is limited only by the average signal-to-noise ratio for each individual “shot”, or, droplet. However, experimental realization of this concept requires a much higher analysis rate than is possible with a levitated microdroplet approach. EXPERIMENTAL SECTION While extension of a levitated droplet approach to singlemolecule detection in falling droplet streams is, perhaps, simple in concept, experimental execution is nontrivial for several reasons. By imposing, as a design constraint, laser-molecule interaction times of the order of several milliseconds, the ability to follow the photobleaching (in time) of individual molecules for a large number of droplets becomes impractical. In general, the result of each droplet measurement will be a single number, the sum of fluorescence and background photocounts, integrated over the laser beam transit time, implying a much higher sensitivity to average droplet size and size fluctuations. Thus, considerably greater care is required in the droplet production stage of the experiment. The crucial parameters for droplet production in this experiment are small size (3σ) from the background distribution, corresponding to the presence of individual R6G molecules. At present, we have not accumulated enough data to comment definitively on the distribution of photocounts from single R6G molecules. The absence of “clustering” of photocounts about a central value and the resemblance of a decaying exponential in the histogram of the 15 fM data seem to indicate complete photobleaching. However, results of detailed Monte Carlo calculations (to be presented elsewhere) indicate that the distribution of single-molecule photocounts in microdroplets is not predicted to be a pure exponential (even if all the molecules undergo photobleaching), due to a nonuniform spatial dependence in the light collection efficiency within the droplet.36 Table 1 shows a summary of 100 and 15 fM fluorescence and blank data shown in Figures 4 and 5. The average number of molecules per droplet for each concentration was determined from the concentration and droplet size. The molecular detection efficiencies were determined by taking the fraction of events which exceed 3σ, subtracting the corresponding fraction produced by the blank, and dividing by the average number of molecules per droplet (obtained from the concentration and droplet size). The measured MDEs, using a 3σ threshold criterion, for both runs agree to within experimental error (76% and 80% for the 100 and 15 fM runs, respectively. These numbers also agree well with statistical limitations (∼80%), assuming an exponential distribution (36) Hill, S. C.; Barnes, M. D.; Whitten, W. B.; Ramsey, J. M. Appl. Opt. 1996, 35, 6278-6285; Appl. Opt., in press.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
2119
Table 1. Summary of fluorescence data from 100 and 15 fM R6G in 6.5 µm Diameter Droplets (900 and 1000 Droplets for Each Set, Respectively)
average no. of molecules per droplet (∼17 µm initial diameter) P(0)a P(1)a P(2)a P(3)a obsd fraction of blank events > 3σ obsd fraction of “signal” events > 3σ molecular detection efficiencyb
100 fM R6G
15 fM R6G
0.2
0.025
0.819 0.163 0.016 0.001 0.007 0.158 0.76 ( 0.21
0.975 0.0244 0.0003 2.5 × 10-6 0.009 0.029 0.80 ( 0.26
a P(n) represents the Poisson probability of 0, 1, 2, etc. molecules occupying each droplet. b The uncertainty figure represents an upper and lower bound from estimated error in droplet diameter ((0.5 µm) and estimated relative error in concentration of 0.05 and 0.10 for 100 and 15 fM solutions, respectively.
for single-molecule fluorescence counts, and results of detailed calculations37 (∼75%) for a 3σ threshold and an average signalto-noise ratio of 10. To complement the direct analysis of single-molecule event frequencies by simple peak counting, the waiting “time” (or the number of droplets in between single-molecule events) distributions for events exceeding 3σ for the 100 fM and 15 fM runs are shown in Figure 7a and b, respectively. In the context of singlemolecule detection,2 the waiting time analysis provides a means of examining the burst statistics to see if the bursts represent a set of uncorrelated, random events. For systems which are not discretely sampled, such as in flow cells or freely diffusing molecules in solution, molecular re-entry into the probe volume may bias this distribution toward shorter times, causing biexponential decay behavior in the waiting time distribution.38,39 For a system such as ours, which is discretely sampled, this distribution should follow single-exponential behavior, since the molecules are probed only once. For both sets of data, the histogram of the number of droplets giving a signal exceeding 3σ is very well described by a single exponential. The exponential decay constant, τ, gives an additional measure of the detection efficiency (independent of the number of single-molecule fluorescence bursts), since 1/τ is just the average number of molecules per droplet. The decay constants associated with exponential fits to the waiting time histograms for the 100 fM and 15 fM runs were 6.2 and 46, respectively, which correspond to measured numbers of molecules (which give a signal >3σ) per droplet of 0.16 and 0.022. Dividing these numbers by the average number of molecules per droplet obtained from the concentration and droplet size, the waiting time analysis gives molecular detection efficiencies of 0.85 and 0.87 for the 100 and 15 fM runs, respectively. This analysis obviously does not take into account the slightly non-Gaussian nature of the blank photocount distribution, and, as a result, MDE values obtained in this way are somewhat higher than the efficiencies obtained from peak counting. (37) Hill, S. C.; Barnes, M. D.; Whitten, W. B.; Ramsey, J. M. to be submitted to Anal. Chem. These calculations model single-molecule fluorescence signals using Monte Carlo techniques, taking into account molecular diffusion and spatial anisotropies in both the excitation intensity and fluorescence collection efficiency within the droplet. (38) Chiu, D. T.; Zare, R. N. J. Am. Chem. Soc. 1996, 118, 6512-6513. (39) Fister, J. C.; Jacobson, S. C.; Davis, L. M.; Ramsey, J. M., to be submitted to Anal. Chem.
2120 Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
Figure 7. Waiting time distributions and exponential fits for (a) 100 and (b) 15 fM data. The points represent the histogram of the “waiting time”, or number of droplets between successive events which exceed 3σ. The decay constants determined from the fits to the histograms were (a) 6.2 and (b) 46 droplets, corresponding to average numbers of molecules per droplet of 0.16 and 0.021.
SUMMARY AND CONCLUSIONS We have demonstrated single-molecule detection (average signal-to-noise ratio of ∼10) in electrodynamically focused microdroplet streams using a counterpropagating cw Ar+ laser beam and relatively simple gated photon-counting techniques. The passage of droplets through the laser beam containing single molecules was clearly distinguished from the background in real time without the need for any statistical or postprocessing analysis. Total molecular detection efficiencies were determined to be ∼80% (>99% confidence), in very good agreement with results of detailed theoretical calculations and statistical limitations. Currently, improvements are being made in analysis rate, light collection efficiency, and droplet size stability, which will allow detailed exploration of single-molecule photocount statistics, subfemtomolar concentration detection limits of single-chromophore molecules in solution, and microcavity effects associated with single molecules in microspheres. ACKNOWLEDGMENT This research was sponsored by the Office of Research and Development, U.S. Department of Energy, under Contract DEAC05-96OR22464, managed by Lockheed Martin Energy Research Corp. N.L. acknowledges an NSERC Postdoctoral fellowship and partial support from the ORNL Postdoctoral Research Associates
Program. C.-Y.K. also acknowledges support from the ORNL Postdoctoral Research Associates Program. The authors also thank Steven C. Hill for several stimulating conversations regarding preliminary results of Monte Carlo modeling of single-molecule fluorescence in microdroplets.
Received for review January 24, 1997. Accepted March 19, 1997.X AC970093B X
Abstract published in Advance ACS Abstracts, May 1, 1997.
Analytical Chemistry, Vol. 69, No. 11, June 1, 1997
2121
