Anal. Chem. 2004, 76, 1618-1626
Focal Volume Confinement by Submicrometer-Sized Fluidic Channels Mathieu Foquet,†,‡ Jonas Korlach,†,‡,§ Warren R. Zipfel,† Watt W. Webb,†,§ and Harold G. Craighead*,†
School of Applied & Engineering Physics, and Field of Biochemistry, Molecular and Cell Biology, Cornell University, Ithaca, New York 14853
Microfluidic channels with two lateral dimensions smaller than 1 µm were fabricated in fused silica for highsensitivity single-molecule detection and fluorescence correlation spectroscopy. The effective observation volumes created by these channels are ∼100 times smaller than observation volumes using conventional confocal optics and thus enable single-fluorophore detection at higher concentrations. Increased signal-to-noise ratios are also attained because the molecules are restricted to diffuse through the central regions of the excitation volume. Depending on the channel geometries, the effective dimensionality of diffusion is reduced, which is taken into account by simple solutions to diffusion models with boundaries. Driven by electrokinetic forces, analytes could be flowed rapidly through the observation volume, drastically increasing the rate of detection events and reducing data acquisition times. The statistical accuracy of single-molecule characterization is improved because all molecules are counted and contribute to the analysis. Velocities as high as 0.1 m/s were reached, corresponding to average molecular residence times in the observation volume as short as 10 µs. Applications of these nanofabricated devices for high-throughput, single-molecule detection in drug screening and genomic analysis are discussed. Fluorescence correlation spectroscopy (FCS) and singlemolecule detection (SMD) yield valuable information about dynamic molecular properties and the nature of biochemical reactions.1-4 Both methods rely on the detection of few, or even one, molecules at a time for the analysis of a given reaction pathway. Valuable information about individual molecular behavior can be inferred, such as the existence and lifetimes of transient intermediate species, conformational fluctuations, memory effects, and distributions in enzyme efficiency and processivity, making these techniques ideal tools to further our understanding of the * To whom correspondence should be addressed. E-mail: mf37@ cornell.edu. Fax: 1-607 255 7658. † School of Applied & Engineering Physics. ‡ These authors contributed equally to this work. § Field of Biochemistry, Molecular and Cell Biology. (1) Rigler, R.; Elson, E. Fluorescence correlation spectroscopy: theory and applications; Springer: Berlin, 2001. (2) Hess, S. T.; Huang, S.; Heikal, A. A.; Webb, W. W. Biochemistry 2002, 41, 697-705. (3) Krichevsky, O.; Bonnet, G. Rep. Prog. Phys. 2002, 65, 251-297. (4) Medina, M. A.; Schwille, P. Bioessays 2002, 24, 758-764.
1618 Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
chemistry of biological systems.5-11 Developments toward analytical applications of SMD and FCS, often using microfabricated devices to allow for high-throughput counting and characterization of single molecules and rare species detection in complex samples, have also been undertaken.12-15 There are many different ways by which single molecules can be investigated.16-22 The experimental strategies attempt to restrict the number of molecules under study in two possible ways. The first is simply to dilute analytes enough so that the probe volume will on average contain a single molecule. The second approach confines the probe volume as much as possible, increasing the likelihood and signal-to-background ratio of single-molecule detection. In practice, a compromise between the two can be found, based on the detection mode, observation time, and concentration range required for the process under study. Fluorescence is a widely used tool for SMD because a single fluorophore can emit many photons in an isolated spectral band. Several different optical methods have been designed for creating small probe volumes using fluorescence, such as confocal optics,23 total internal reflection,24 two-photon excitation,25 near-field optical (5) Adachi, K.; Noji, H.; Kinosita, K., Jr. Methods Enzymol 2003, 361, 211227. (6) Clausen-Schaumann, H.; Seitz, M.; Krautbauer, R.; Gaub, H. E. Curr. Opin. Chem. Biol. 2000, 4, 524-530. (7) Edman, L.; Rigler, R. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 8266-8271. (8) Zhuang, X.; Kim, H.; Pereira, M. J.; Babcock, H. P.; Walter, N. G.; Chu, S. Science 2002, 296, 1473-1476. (9) Xie, X. S.; Lu, H. P. J. Biol. Chem. 1999, 274, 15967-15970. (10) Xie, S. N. Single Mol. 2001, 2, 229-236. (11) Schuler, B.; Lipman, E. A.; Eaton, W. A. Nature 2002, 419, 743-747. (12) Zander, C.; Enderlein, J.; Keller, R. A. Single molecule detection in solution: methods and applications; Wiley-VCH: Berlin, 2002. (13) Castro, A.; Shera, E. B. Appl. Opt. 1995, 34, 3218-3222. (14) Keller, R. A.; Ambrose, W. P.; Arias, A. A.; Cai, H.; Emory, S. R.; Goodwin, P. M.; Jett, J. H. Anal. Chem. 2002, 74, 316A-324A. (15) Vercoutere, W.; Akeson, M. Curr. Opin. Chem. Biol. 2002, 6, 816-822. (16) Barnes, M. D.; Whitten, W. B.; Ramsey, J. M. Anal. Chem. 1995, 67, A418A423. (17) Fan, F. R. F.; Kwak, J.; Bard, A. J. J. Am. Chem. Soc. 1996, 118, 96699675. (18) Lermer, N.; Barnes, M. D.; Kung, C. Y.; Whitten, W. B.; Ramsey, J. M. Anal. Chem. 1997, 69, 2115-2121. (19) Moerner, W. E.; Orrit, M. Science 1999, 283, 1670-1676. (20) Feder, T. J.; BrustMascher, I.; Slattery, J. P.; Baird, B.; Webb, W. W. Biophys. J. 1996, 70, 2767-2773. (21) Ghosh, R. N.; Webb, W. W. Biophys. J. 1994, 66, 1301-1318. (22) Barak, L. S.; Webb, W. W. J. Cell Biol. 1981, 90, 595-604. (23) Koppel, D. E.; Axelrod, D.; Schlessinger, J.; Elson, E. L.; Webb, W. W. Biophys. J. 1976, 16, 1315-1329. (24) Thompson, N. L.; Burghardt, T. P.; Axelrod, D. Biophys. J. 1981, 33, 435454. (25) Mertz, J.; Xu, C.; Webb, W. W. Opt. Lett. 1995, 20, 2532-2534. 10.1021/ac035088o CCC: $27.50
© 2004 American Chemical Society Published on Web 02/05/2004
microscopy,26 and stimulated emission depletion.27 While all these methods have specific advantages, the range of concentrations accessible to SMD is still largely restricted to the pico- or nanomolar range. Bringing the molecule of interest into the interaction volume can present another limitation, especially when diffusion is slow or the molecular species of interest are very rare. In addition, studies of weak binding interactions necessitate relatively high concentrations for sufficient complex formation.28 New techniques are therefore desirable to further expand the range of concentrations amenable to SMD and to combine them with high-throughput analysis systems. Zero-mode waveguide arrays have been described as an approach to utilize nanostructures for the creation of extremely small observation volumes.29 Here we report on another possibility of volume confinement, using nanofabricated channels with a width and depth smaller than the dimensions of the confocal detection volume. The use of microfabricated channels for single-fluorophore detection with high-throughput has been demonstrated previously, and various channel or capillary geometries with dimensions of one to several micrometers diameter have been evaluated.14,30-33 The width of the focal volume was either expanded to span the entire channel width, or only a fraction of the channel was illuminated, and thus, not all molecules flowing through the channel were detected. The upper limit in concentration to avoid simultaneous passage of two or more molecules through the observation volume was therefore similar or worse compared to diffraction-limited confocal optics. Capillaries ending in small circular constrictions of ∼0.5 µm have also been used for singlefluorophore detection,34 but this experimental setup is not easily amenable to integration into more complex “labs on chips”.35-37 The further reduction of channel dimensions is therefore desirable for (i) allowing higher analyte concentrations for SMD, (ii) decreasing background contributions from Raman scattering, and (iii) enabling highly parallel implementations with integrated detection. It has also been shown that new phenomena arise at these small spatial scales, which can be taken advantage of to design new technologies.38-40 (26) Betzig, E.; Chichester, R. J. Science 1993, 262, 1422-1425. (27) Klar, T. A.; Jakobs, S.; Dyba, M.; Egner, A.; Hell, S. W. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 8206-8210. (28) Laurence, T. A.; Weiss, S. Science 2003, 299, 667-668. (29) Levene, M.; Korlach, J.; Turner, S. W.; Foquet, M.; Craighead, H. G.; Webb, W. W. Science 2003, 299, 682-686. (30) Gosch, M.; Blom, H.; Holm, J.; Heino, T.; Rigler, R. Anal. Chem. 2000, 72, 3260-3265. (31) Do ¨rre, K.; Stephan, J.; Eigen, M. Single Mol. 2001, 2, 165-175. (32) Sauer, M.; Angerer, B.; Ankenbauer, W.; Fo ¨ldes-Papp, Z.; Go ¨bel, F.; Han, K. T.; Rigler, R.; Schulz, A.; Wolfrum, J.; Zander, C. J. Biotechnol. 2001, 86, 181-201. (33) Han, J.; Craighead, H. G. Science 2000, 288, 1026-1029. (34) Zander, C.; Drexhage, K. H.; Han, K. T.; Wolfrum, J.; Sauer, M. Chem. Phys. Lett. 1998, 286, 457-465. (35) Craighead, H. G. Science 2000, 290, 1532-1536. (36) Reyes, D. R.; Iossifidis, D.; Auroux, P. A.; Manz, A. Anal. Chem. 2002, 74, 2623-2636. (37) Auroux, P. A.; Iossifidis, D.; Reyes, D. R.; Manz, A. Anal. Chem. 2002, 74, 2637-2652. (38) McKnight, T. E.; Culbertson, C. T.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 2001, 73, 4045-4049. (39) Turner, S. W.; Cabodi, M.; Craighead, H. G. Phys. Rev. Lett. 2002, 88, 128103. (40) Chou, C. F.; Bakajin, O.; Turner, S. W.; Duke, T. A.; Chan, S. S.; Cox, E. C.; Craighead, H. G.; Austin, R. H. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 13762-13765.
As the fluid is confined by such channels to dimensions comparable or smaller than the optical excitation volume, the resulting boundaries restrict diffusion and affect FCS measurements and SMD sensitivities. Their evaluation is therefore essential toward optimizing a nanofabricated device for a given analytical application. It is equally relevant for cellular applications of FCS and SMD for measuring diffusion behavior and extracting diffusion coefficients in subcellular compartments that have similar sizes.41,42 In this article, the use of submicrometer-sized fluidic channels, manufactured by nanofabrication techniques, for FCS and SMD is described. The small size of these channels enables performing FCS at higher, micromolar concentrations and modifies diffusion measurements to two- or pseudo-one-dimensional limits. Precise positioning of the focal volume with respect to the channel has a large effect on the signal-to-noise ratios and shapes of the FCS curves and their deduced parameters. It is also demonstrated that the small volumes created translate to high signal-to-noise ratios for performing SMD. Fast, electrokinetically driven fluid flow of molecules past the observation region for high-throughput singlemolecule analysis is evaluated. MATERIALS AND METHODS Channels used in this study were similar to those described by Foquet et al.,43 with the exception that the walls were formed by thermal oxidation of the sacrificial layer material instead of deposition of doped plasma-enhanced chemical vapor deposition silicon dioxide. The polysilicon sacrificial layer was transformed into silicon dioxide wall material by oxidation using a pyrogenic steam furnace. This has the benefit of reducing the patterned feature sizes to submicrometer dimensions without the need for electron beam lithography or other advanced patterning methods. The channels have a total length of 30 mm, with two central constrictions of 50-µm length and variable widths, shown in Figure 1. The lateral protrusions are a consequence in the wet-etching step of the nanofabrication protocol.43 The smallest channels fabricated have sizes of approximately 350 nm wide by 250 nm tall. Measurements were performed on a setup described previously,43 using a 60 × 0.9 NA objective lens (Olympus, Melville, NY). Alexa Fluor 488-5-dUTP (Molecular Probes, Eugene, OR) was used as the fluorescent dye in 1× TBE (90 mM Tris-borate, pH 8.0, 2 mM EDTA). To suppress dye adhesion to the capillary wall, 0.2% (v/v) of the surfactant Nonidet P-40 (USB Corp., Cleveland, OH) was added. Unless otherwise indicated, a concentration of 1 µM was used for FCS and 2 nM for SMD. Flow in the nanocapillaries was induced by an applied electric field through gold wires in contact with the solution. Field strengths given in the paper refer to the narrowest capillary of the microstructure (∼350 nm wide), calculated from the respective voltage drops in the various sections of the channel. Ten consecutive 1-min measurements were acquired and averaged for the FCS analysis of diffusion confinement. Three consecutive runs of 5 s each were sufficient when flow was applied. Curves were fit with a nonlinear least-squares algorithm, weighted by the (41) Gennerich, A.; Schild, D. Biophys. J. 2000, 79, 3294-3306. (42) Gennerich, A.; Schild, D. Biophys. J. 2002, 83, 510-522. (43) Foquet, M.; Korlach, J.; Zipfel, W. R.; Webb, W. W.; Craighead, H. G. Anal. Chem. 2002, 74, 1415-1422.
Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
1619
Figure 1. (A) Top view micrograph of the central area of an array of nanochannels. The channels are 10 µm wide (light cyan), except in two constricted areas (arrows) of ∼350-nm width. Scale bar, 50 µm. (B) Scheme of detecting single molecules flowing in the nanochannel. The axes define coordinate designations used in this paper.
reciprocal variance measured at each delay time. For singlemolecule experiments, photon arrival times were recorded for 2 min, binned, and analyzed using the custom-made software described in detail elsewhere.43 The diffusion coefficient of Alexa Fluor 488-5-dUTP was measured independently to 2.1 × 10-6 cm2/s by comparison with Rhodamine Green.44 Illumination intensities were kept low at 4-10 µW for FCS to avoid tripletstate population and increased to 1 mW unless stated otherwise for SMD to maximize fluorescence burst sizes. FCS45,46 yields information on the time scale of fluorescence fluctuations originating from diffusion or chemical reactions involving fluorescent molecules in an observation volume. In this study, only diffusional contributions are considered. Neglecting triplet states, the autocorrelation function, assuming a Gaussian illumination profile, can be expressed as,
where wi is the e-2 radius of the excitation volume along axis i, D is the diffusion coefficient, and N is the average number of molecule in the effective observation volume, Veff ) N/C, where C is the concentration. The molecular brightness can be expressed as I/N*, where I is the average fluorescence intensity. As the small size of the nanocapillary introduces additional volume limitations, some factors become unity, as indicated by the brackets in eq 1. Two different cases will be discussed. By comparing the effective dimensions of the focal volume, wi, to that of the fluid sample determined by the channel dimensions, di, the ratio di/wi is of particular relevance. (44) Rigler, R.; Mets, U.; Widengren, J.; Kask, P. Eur. Biophys. J. Biophys. Lett. 1993, 22, 169-175. (45) Magde, D.; Webb, W. W.; Elson, E. Phys. Rev. Lett. 1972, 29, 705-708. (46) Elson, E. L.; Magde, D. Biopolymers 1974, 13, 1-27.
1620
Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
If the focal volume dimensions are small compared to the channel dimensions (di/wi are large), the diffusion volume is unbounded and is treated as an open-volume, three-dimensional diffusion model. However, the nanocapillaries restrict fluid along the axial dimension z over their entire length, defining a plane sheet of fluid with a height of only ∼250 nm, much smaller than the focal volume height, wz. As dz/wz is small for this case, the last term in eq 1 can be neglected, yielding a two-dimensional system. In the narrow channel regions (Figure 1A, arrrows), one more lateral dimension is further restricted to submicrometer sizes. Here, dy/wy is on the order of 1, which requires a modified diffusion model with boundaries in two dimensions,41
Gxy*(τ) )
1 N
( )
dy gy* τ, wy
1
x1 + 4D/wx τ 2
(2)
with the function gy* taking into account the confined diffusion in that direction. In our case, dy/wy e 8, for which a numerical approximation has been formulated.41 For measurements involving scanning the focal volume across the channel, the resulting intensity was fit to the convolution of the Gaussian illumination profile with the channel cross section via,
[(
I(y) ) A erf
) (
y + dy/2
x2wy
- erf
)]
y - dy/2
x2wy
(3)
where A is a scaling factor. In the presence of uniform translational fluid flow in the channels, autocorrelation curves are extended to
(
G(τ) ) G(τ)Diff exp -
τ2 (wx/vx)2(1 + 4D/wx2τ)
)
(4)
where vx is the average velocity of the analyte, containing contributions from both electrophoretic motion and electroosmotic
fluid flow, which depend on surface charges of the channel walls, buffer conditions, and the charge of the fluorescent species. To evaluate the efficiency of single-molecule detection, the average number of photons detected from a molecule passing through the observation volume was estimated by
I E ) 1.5 τ1/2 N*
(5)
where N* ) N/(1 - F) is the average number of molecules, corrected for the fraction F of the molecules in the triplet state, and τ1/2 is the correlation time at which the FCS curves has decayed to 1/2N*. The constant 1.5 accounts for the relationship between τ1/2 and the duration of the bursts, assuming a Gaussian burst shape. RESULTS Nanochannel Fabrication. The nanocapillaries were fabricated using the sacrificial layer method.43,47 This method, adapted from micro- and nano- electromechanical system manufacturing (MEMS and NEMS), allows for the formation of structures with submicrometer dimensions. These features can be obtained using standard 0.5-µm photolithography tools by reducing the size of the sacrificial layer after patterning by a thermal oxidation process. It is therefore a simple and economical method to obtain very small feature sizes. The devices consist of straight tubes of 10-µm width, with two narrow, submicrometer-sized constrictions in their central region (Figure 1A, arrows). The height of the channel was measured to be constant throughout the device at 250 nm. The smallest capillaries fabricated have a cross section estimated at 250 × ∼350 nm. The channels exhibit a straight rectangular profile with little curvature.43 Fluorescence Correlation Spectroscopy. Depending on where the focal volume is positioned with respect to the channel, a different effective observation volume is created. If positioned inside the 10-µm-wide channel region, diffusion is restricted in the axial dimension, whereas diffusion is restricted in both the axial dimensions and one lateral dimension when positioned over the narrow channel region. Here, the excitation intensity is relatively uniform across the channel width and height, facilitating detection of all fluorophores in a sample (Figure 1B). This confinement effect of the nanocapillaries was investigated by comparing diffusion of a 20 nM concentration of the fluorescently labeled nucleotide Alexa Fluor 488-5-dUTP in an open volume to 1 µM inside the nanostructure, using an identical optical setup (Figure 2). The laser intensity was kept low enough to avoid triplet-state formation in these measurements. While the fluorophore concentration was increased by a factor of 50 for the measurements using the nanostructure, correlation amplitudes G(0) were comparable to the open solution measurement, indicating a similar average number of molecules, N, in the restricted effective observation volume. As expected, this effect was more pronounced in the narrow channel due to the additional lateral confinement, resulting in a ∼100-fold reduction in the effective observation volume (Table 1). Signal-to-noise ratios, judged from (47) Turner, S. W.; Perez, A. M.; Lopez, A.; Craighead, H. G. J. Vac. Sci. Technol., B 1998, 16, 3835-3840.
the molecular brightness, were increased ∼2-fold in the channel (Table 1). The correlation curves from diffusion inside the nanostructures were devoid of any long time components, which would indicate fluorophore adsorption to the channel walls.48 Fits to these FCS curves correctly reflected the boundaries to diffusion set by the channel walls (Table 1).41 In all three cases, fits resulted in values for the lateral e-2 focal volume radius of 240-300 nm, consistent with the constant optical setup that was used for these measurements. The FCS curve obtained in an open volume (Figure 2A) was fit with the standard three-dimensional Gaussian approximation (eq 1). When the optical axis dimension was restricted by the nanochannel depth, a better fit was obtained using a two-dimensional model, with the third term in eq 1 accounting for diffusion in the z-direction approaching unity. In the narrow channel, further lateral confinement was manifested by a pronounced long time tail in the FCS curve (Figure 2C), illustrated by a comparison to a fit assuming an incorrect diffusion dimensionality of two (green dashed line). FCS curves from the narrow channels could be fit well using the modified one-dimensional model outlined in the theory section (eq 2). From the fit, the ratio of the lateral channel width to the focal volume radius, dy/wy, yielded an approximation of the channel width of ∼450 nm, in acceptable agreement with a value of ∼350 nm estimated from the fabrication protocol. Further accuracy in channel width determination by FCS fits proved difficult because of the very sensitive positioning effects of the focal volume relative to the channel (see below) and because the ratio dy/wy only has a relatively weak effect on the precise shape of the long-time tail in the FCS curve, compared to the other parameters in eq 2. Precise positioning of the focal volume had a large effect on the FCS results. Figure 3 shows parameters obtained from FCS curve fits when the focal volume was moved along the optical axis (z-direction) and laterally (y-direction) relative to the channel. As expected, the molecular brightness was maximal when the laser focus was centered with respect to the channel interior. This alignment was very sensitive, as displacements of only a few hundred nanometers resulted in decreased signal and signal-tonoise ratio in FCS measurements. The cone of the focused laser light also induced a larger apparent width of the focal volume when the center of illumination was offset along the optical axis (Figure 3A,B). Diffusion times and resulting apparent widths of the focal volume were relatively constant only within 1 µm of the center of illumination. Lateral positioning of the laser focus with respect the narrow channel for maximal fluorescence was even more sensitive (Figure 3C). Fluorescence above the background level was only detected within a 500-nm range as a steep function of position with a maximum fluorescence signal over only ∼200 nm. These beam scanning data could be fit to a simple model based on the convolution of the Gaussian illumination profile with the straight channel (eq 3), yielding a channel width of ∼375 nm, in good agreement with the value estimated from the fabrication protocol. Analyte transport effects were investigated by inducing electrokinetic flow in the channels. Upon application of an increasing electric field along the channel, data acquisition times could be reduced to a few seconds. FCS curves shift to shorter time delays due to the increasing flow speed of the molecules (Figure 4A). (48) Starr, T. E.; Thompson, N. L. Biophys. J. 2001, 80, 1575-1584.
Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
1621
Figure 2. FCS analysis of Alexa Fluor 488-5-dUTP in different nanochannel regions. Correlation curves taken in (A) a conventional open volume, confocal setup, (B) inside the wide channel region, and (C) inside the narrow channel region. Average FCS curves are fit (red curves) according to models assuming diffusion with the corresponding restricted dimensions (eqs 1 and 2). The green dashed curve in (C) is a fit assuming an incorrect dimensionality of two. Dye concentrations were 20 nM for (A) and 1 µM for (B) and (C). Error bars are standard deviations from 10 consecutive measurements (1 min each). Fitting parameters are given in Table 1. Table 1. Parameters Obtained from FCS Measurements on Alexa Fluor 488-5-dUTP in the Nanofabricated Channelsa
open solution wide channel narrow channel
C (µM)
fitting function
I (kHz)
N
Veff (10-18 L)
brightness (kHz/molecule)
wx (nm)
0.02 1 1
3D (eq 1) 2D (eq 1) modified 1D (eq 2)
32.6 91.3 25.8
12.9 ( 0.1 19.9 ( 0.1 6.34 ( 0.2
1070 33 11
2.5 4.6 4.1
250 ( 2 239 ( 1 296 ( 5
wz/wx
dy/wy
10.5 1.65 ( 0.07
a Symbols: C, fluorophore concentration; I, average fluorescence intensity; N, average number of molecules in the effective observation volume; Veff, wx, e-2 focal volume radius; wz/wx, ratio between e-2 focal volume height and width; dy/wy, ratio between channel width to e-2 focal volume width. Parameters in the last six columns were derived from fits to the FCS curves shown in Figure 2.
The long time tail in the FCS curve disappears even at low fields as the molecules are passing the observation volume with more uniform average speed, and the diffusion component accounting for variability in residence times becomes less dominant. Curve fits using the mixed diffusion/transport model (eq 4) yielded the flow speed of the fluorophores (Figure 4B). The flow speed was linearly proportional to the applied voltage and grew inversely proportional to the width of the channel for a fixed channel height, demonstrating that the electrical double layer is 1622 Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
small compared to the size of the capillary under these conditions. At very low electric fields, the diffusion term dominates the correlation curve, translating into a relatively large incertitude in flow rate determinations. Flow speeds could be measured more reliably when the analyte speed exceeded their diffusion rate. This transition from the diffusion-dominated to flow-dominated transport regime is shown in Figure 4C by plotting the characteristic molecular residence time, τ1/2, as a function of the ratio between the characteristic flow time, wx/vx, and diffusion time, wx2/4D. For
Figure 3. Effects of the focus position relative to the nanochannel on FCS results. Plotted in (A) and (B) are the average molecular brightness (left axes) and the apparent width of the focal volume (right axes) for a scan in the z-direction through the wide channel (A, 100-nm step size) and the narrow channel (B, 200-nm step size). (C) shows the intensity dependence of the lateral laser focus position (y-dimension, 100-nm step size) with respect to the narrow channel, fit according to eq 3 (red line).
electric fields higher than 2.75 × 105 V/m, diffusion could be neglected and FCS curves could be fit with a pure translational flow model with errors of less than 10%. At the corresponding flow speeds, the average molecular residency time is essentially a function of the flow speed. At even higher flow speeds, autocorrelation curves became increasingly noisy because the number of detected photons is decreased by the extremely short (500 kV/m), the average time interval between two successive bursts was so short that resolving of individual bursts became increasingly difficult. DISCUSSION Integrated micro- and nanofluidic devices continue to transform the biological and chemical sciences to carry out common analysis tasks faster, more accurately, and with less reagent consumption than current procedures.49-51 To this end, the fabrication of nanocapillaries with submicrometer dimensions has been achieved in several laboratories. The sacrificial layer process utilized here (49) Huang, X. C.; Quesada, M. A.; Mathies, R. A. Anal. Chem. 1992, 64, 21492154. (50) Kameoka, J.; Orth, R.; Ilic, B.; Czaplewski, D.; Wachs, T.; Craighead, H. G. Anal. Chem. 2002, 74, 5897-5901. (51) Ehrfeld, W.; Hartmann, H.-J.; Hessel, V.; Kiesewalter, S.; Lo ¨we, H. Micro Total Analysis Systems 2000; Kluwer Academic: Dordrecht, The Netherlands, 2000; pp 33-40.
Figure 5. Single-fluorophore detection in nanochannels. (A) Time traces of fluorescence intensity (50-µs bin size) from 2 nM Alexa Fluor 488-5-dUTP for four different applied electric fields and (B) corresponding histograms of burst sizes distributions (2-min acquisition), showing the large increase in burst frequency with increasing flow speed.
allows for integration of a large number of microfluidics circuitries on a very small footprint. We have shown that the nanofabricated channels employed here comprise an attractive platform to perform FCS and singlefluorophore detection at relatively high concentrations due to a reduction of the effective observation volume by as large as 2 orders of magnitude. While conventional far-field optics are used, the concentration range accessible to these methods is thereby
extended into the micromolar regime, which is important for the study of many biochemical processes that naturally occur at such concentration levels.28,29 In the absence of flow, this simple fluid confinement can be used to extend analyte residence times in the observation volume, leading to higher statistical accuracy of dynamic processes, such as enzyme turnovers, transient molecular interactions, or photophysical phenomena, without the necessity of analyte immobilization. No artifacts from fluorophore adsorption Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
1625
to the channel walls were observed after inclusion of a low concentration of surfactant. Distinct from a sole effect of size reduction by the small channel dimensions, a second contribution resulting in increased overall signal-to-noise ratio stems from the elimination of out-offocus background fluorescence by guaranteeing that molecules in the channel will pass through the effective center of illumination where photon emission rates are highest. In conventional confocal detection of three-dimensional diffusion, the majority of detected events correspond to molecules traversing the fringe of the probe volume, which carry a high-noise contribution since fewer photons are emitted. Thereby, confinement control of the focal volume allows reducing the measurement time for an autocorrelation analysis. For channel dimensions on the order of or smaller than the excitation volume, the exact position of the focal volume with respect to the channel has significant effects on the apparent molecular brightness, diffusion times, and other parameters extracted from FCS measurements. Therefore, care should be taken in the interpretation of FCS curves where a constant focal volume is conventionally assumed, e.g., for determinations of diffusion coefficients. The effect is more pronounced for the apparent brightness than the molecular diffusion time so that adjusting the channel position to the maximum molecular brightness is a good parameter for optimal alignment. Depending on the desired application, further modifications of the focal volume should be employed to minimize these positioning effects, e.g., by using cylindrical lenses or beam-shaping optical elements for uniform illumination across the channel width and height,52-56 leading to even narrower photon burst distributions, increased single-molecule detection efficiency, and higher resolving power for distinguishing different analyte populations. These studies of diffusion and transport in volumes smaller than the diffraction limit are also relevant for cell biology applications of FCS and SMD. Confinement of analytes in small cellular structures was shown to lead to modified diffusion behavior, resulting in complex expressions for the resulting FCS curves.41,42 Nanofabricated devices may therefore constitute useful model structures toward gaining a better understanding of the complex diffusion dynamics in the cytosol and subcellular compartments. To this end, potential effects of light scattering and refractive index mismatches on the focal volume by the submicrometer compartment should also be characterized. The possibility of applying flow to the channel fluid leads to improved collection statistics, because the molecular detection rate (52) Anazawa, T.; Matsunaga, H.; Yeung, E. S. Anal. Chem. 2002, 74, 50335038. (53) Larson, E. J.; Hakovirta, J. R.; Cai, H.; Jett, J. H.; Burde, S.; Keller, R. A.; Marrone, B. L. Cytometry 2000, 41, 203-208. (54) Zhang, S.; Neil, G.; Shinn, M. Opt. Express 2003, 11, 1942-1948. (55) Hoffnagle, J. A.; Jefferson, C. M. Appl. Opt. 2000, 39, 5488-5499. (56) Roos, P.; Skinner, C. D. Analyst 2003, 128, 527-531.
1626
Analytical Chemistry, Vol. 76, No. 6, March 15, 2004
can be increased substantially. Single molecules are thereby counted with high efficiency, providing intrinsic quantitation. As single-molecule analysis relies on individually characterizing an ensemble of molecules to deduce their average behavior and its underlying distribution, an increase in the rate of detected molecules improves the statistical accuracy of the measurement. Flow speeds as high as 0.1 m/s were achieved in this study, ideal for high-throughput applications such as drug screening. This has to be balanced by the residence time per molecule, which is decreased with faster flow speeds. The analysis of the SMD detection efficiency (Figure 4D) shows that the collected intensity per molecule decreases significantly at very high flow speeds. Thus, there is a tradeoff between signal-to-noise ratio and speed of acquisition, and a balance between intensity and flow speed has to be found to provide optimal detection signal and analyte throughput for a given application. Fluid flow in the nanocapillaries is characterized by its extremely small transported volumes, even at very high flow speeds. For example, at 5 cm/s flow rate, the effective volume carried through a 250 × 350 nm capillary is only ∼4 × 10-6 µL/s. Hence, virtually no volume is consumed by the device, making it an ideal candidate for single-molecule detection where only minute amounts of samples are available. Development of appropriate reagent coupling to and from these nanostructures is essential to exploit this feature in future applications. Molecular biological applications in which accurate counting of the number and type of molecular species is of interest would also benefit from these nanostructures, as very rare events could be detected above a large background. Differences in electrophoretic flow speeds of different analytes could be used advantageously for discrimination of distinct molecular species in a complex mixture. Interactions between the electrical double layer at the channel wall and the analytes can similarly be exploited. The nanostructures are also compatible with surface modification schemes for the study of molecular interactions. Further research using these devices for direct RNA quantitation, cross-correlation spectroscopy, and the monitoring of single-enzyme dynamics is currently being undertaken in our laboratories. ACKNOWLEDGMENT The authors thank S. Kim for critical reading of the manuscript. This publication was made possible by Grant DE-FG02-99ER62809 from the U.S. Department of Energy and Grant P41-2RR04224 from the National Center for Research Resources, National Institutes of Health.
Received for review September 16, 2003. Accepted December 9, 2003. AC035088O
