Single-Molecule Measurement of Protein Folding Kinetics

See allHide authors and affiliations

Science  29 Aug 2003:
Vol. 301, Issue 5637, pp. 1233-1235
DOI: 10.1126/science.1085399


In order to investigate the behavior of single molecules under conditions far from equilibrium, we have coupled a microfabricated laminar-flow mixer to a confocal optical system. This combination enables time-resolved measurement of Förster resonance energy transfer after an abrupt change in solution conditions. Observations of a small protein show the evolution of the intramolecular distance distribution as folding progresses. This technique can expose subpopulations, such as unfolded protein under conditions favoring the native structure, that would be obscured in equilibrium experiments.

Fluorescence detection of Förster resonance energy transfer (FRET), in which the probability of excitation energy transfer between two chromophores is used to determine their separation, is a powerful tool for investigating the behavior of single biological macromolecules (1, 2). Two variants of the technique are used. In the first, a dye-labeled molecule is immobilized on a surface, and its fluorescence emission is recorded continuously until one of the dyes has bleached. This allows uninterrupted observation and has been applied to both RNA and protein folding (36). Alternatively, fluorescence can be observed as molecules diffuse freely through the focus of a laser beam (7). This approach has the advantage of preventing interactions between the sample and a surface.

Free-diffusion experiments have yielded direct enumeration of thermodynamic states and information about the sizes and dynamical properties of unfolded molecules when both folded and unfolded molecules are present, information that is difficult or impossible to obtain from bulk samples (810). A limitation of the free-diffusion experiments performed to date is that they could only be used to examine states that are substantially populated at equilibrium. In order to study the properties of single molecules under nonequilibrium conditions, we have coupled a microfabricated laminar-flow mixer to a confocal optical system. This has allowed us to measure the FRET efficiencies of individual dye-labeled protein molecules as they flow through the illuminated volume at selected times subsequent to the triggering of folding by an abrupt change in denaturant concentration.

The microfabricated device and the sample flow in the mixing region are illustrated in Fig. 1. The mixer (11) is similar to that designed by Kauffmann et al. (12) for ensemble infrared measurements. The Reynolds numbers (about 10-2) in these experiments are small enough to guarantee laminar flow, and the microscopic dimensions of the mixer channels ensure that rapid mixing can occur by diffusion alone. Under these conditions, displacements due to diffusion and flow are comparable over a period of a few milliseconds, and the solutions are mixed by the time (about 50 ms) they have reached a point 50 μm beyond the center inlet.


Microfluidic mixer. (A) Channel pattern and photograph of mixing region as seen through the microscope objective and bonded coverslip. Solutions containing protein, denaturant, and buffer were driven through the channels with compressedair (11). Arrows indicate the direction of flow. (B) View of the mixing region. Computed denaturant concentration (11) is indicated by color. The laser beam (light blue) and collected fluorescence (yellow) are shown 100 μm from the center inlet. (C) Cross section of the mixing region. The 1/e2 intensity contour of the laser beam is illustrated along with the cone of fluorescence emission collected by the microscope objective. Actual measurements were made at distances ≥100 μm from the center inlet channel. Because the flow velocity at the vertical center of the observation channel is about 1 μm ms-1, this corresponds to times ≥0.1 s.

A 488-nm laser was used to excite sample molecules in the mixer (11), and the resulting fluorescence was collected with an optical system identical to that used by Schuler et al. (9). The beam was positioned downstream of the mixing region at distances chosen to correspond, via the flow rate, to the desired delays. Counts from the detectors were collected in 1-ms intervals as individual molecules passed through the focal volume.

The molecule studied here is the cold shock protein (Csp) from the bacterium Thermotoga maritima (11), a small β-barrel protein that exhibits two-state thermodynamics and kinetics (13, 14). Csp was labeled with a green fluorescent donor dye (Alexa Fluor 488, Molecular Probes, Eugene, Oregon) and a red fluorescent acceptor (Alexa Fluor 594) via amino- and carboxyl-terminal cysteine residues (9, 11).

Representative FRET efficiency histograms measured at various times after dilution of denaturant to trigger folding are shown in Fig. 2A. The maximum at high transfer efficiency corresponds to folded molecules. For this subpopulation, excitation of the donor dye results in rapid energy transfer to the acceptor because the termini of the protein are separated by only 1 nm in the native structure (15). Most of the fluorescence photons are therefore emitted by the acceptor. A second maximum occurs at a transfer efficiency of about 0.5. This is produced by unfolded protein, for which the average distance between the dyes is greater. Consequently, the energy transfer rate is decreased, and the fraction of photons emitted by the acceptor is reduced.


(A) Histograms of measured FRET efficiency, Em, during folding. (Top) The Em distribution at equilibrium before mixing. The vertical redline indicates the mean value for Em in the unfolded state after mixing. Measurements of Em were taken (typically for 30 minutes) at various distances from the mixing region and thus at various times after the change in GdmCl concentration. Data collection and analysis were carried out as previously described(9, 11). (B) Comparison of single-molecule and ensemble folding kinetics (11). The decrease in donor fluorescence from stopped-flow experiments is shown in green. Single-molecule data are represented by filled circles, with a corresponding single exponential fit (black). The fit to the single-molecule data has a rate constant of 6.6 ± 0.8 s-1, in agreement with the result from ensemble measurements (5.7 ± 0.4 s-1). Measurements of Csp unfolding are shown in fig. S2.

After mixing, a redistribution of the populations is observed. As the number of unfolded molecules decreases, there is a corresponding increase in the number of folded molecules (Fig. 2A), and eventually the distribution converges to its equilibrium form at the final concentration of denaturant. The positions of the peaks, however, remain constant, indicating that the average end-to-end distances of the molecules in the subpopulations do not change significantly during the course of the reaction. This is the behavior expected for a two-state system, as inferred for this protein from ensemble methods (13, 14). The fidelity of the technique for measuring accurate kinetics is demonstrated in Fig. 2B, which shows that nearly identical folding rates are obtained from single-molecule data and ensemble stopped-flow measurements.

Whereas the peak positions are constant during the folding reaction, the mean FRET efficiency of unfolded molecules is shifted from 0.51 in the presence of 4 M guanidinium chloride (GdmCl) to 0.64 after mixing is complete and solution conditions are conducive to folding (Fig. 2A). The lower denaturant concentration after dilution results in more compact unfolded molecules, which exhibit higher FRET efficiency. For a Gaussian chain (16), this shift would correspond to a decrease in the radius of gyration of about 20% (9), which implies a halving of occupied volume. This reconfiguration is obviously complete by the time of the first observation at about 100 ms, consistent with ensemble measurements on other proteins that show collapse of unfolded molecules to be a sub-millisecond process (1721). It is noteworthy that qualitatively different signal changes result from collapse and folding. Collapse of the unfolded state causes a shift of the corresponding peak to higher efficiency, whereas folding increases the folded state population (represented by peak height) and depopulates the unfolded state. This permits separate observations of the two processes. In an ensemble FRET experiment, both collapse and folding would result in an overall increase of the transfer efficiency, and the respective contributions could be identified only indirectly by kinetic modeling.

The power of single-molecule measurements is in the ability to resolve and analyze the properties of subpopulations. The kinetic experiment presented here extends the scope of the technique so that such subpopulations can be studied under conditions far from equilibrium, where they exist only briefly. Experiments in the absence of flow have shown that unfolded Csp must reconfigure quickly relative to the observation time of about 1 ms (9), in that no broadening of the measured energy transfer distributions was observed beyond that seen for rigid polypro-line peptides. These equilibrium measurements were limited to solution conditions that resulted in sufficiently high concentrations of unfolded molecules. By transiently populating the denatured state with the microfluidic mixer, we can now extend the observations to denaturant concentrations that strongly favor the native structure and are therefore closer to physiologically relevant conditions. The present results (Fig. 3) show that the unfolded peak width is constant to GdmCl concentrations as low as 0.5 M, indicating that, even though there is significant compaction of the unfolded state between 6.0 M and 0.5 M GdmCl, there is no increase in the reconfiguration time beyond the previously calculated limit of 0.2 ms (22).


Dependence of (A) the mean values Embedded Image and (B) peak widths (σ) of Em histograms for unfolded Csp as a function of GdmCl concentration. Peak widths are standard deviations of Gaussian fits. Data were obtained from single-molecule measurements in the microfluidic mixer. The dotted lines show a third-degree polynomial fit to the data (A) or indicate their mean value (B). The colored region indicates the range of denaturant concentrations where reliable data cannot be obtained from corresponding equilibrium experiments.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

References and Notes

View Abstract

Navigate This Article