Kinesin moving through the spotlight: single-motor fluorescence microscopy with submillisecond time resolution

Abstract

Kinesin-1 is one of the motor proteins that drive intracellular transport in eukaryotes. This motor makes hundreds of 8-nm steps along a microtubule before releasing. Kinesin-1 can move at velocities of up to approximately 800 nm/s, which means that one turnover on average takes 10 ms. Important details, however, concerning the coordination between the two motor domains have not been determined due to limitations of the techniques used. In this study, we present an approach that allows the observation of fluorescence intensity changes on individual kinesins with a time resolution far better than the duration of a single step. In our approach, the laser focus of a confocal fluorescence microscope is pointed at a microtubule and the photons emitted by fluorescently labeled kinesin motors walking through the spot are detected with submicrosecond accuracy. We show that the autocorrelation of a fluorescence time trace of an individual kinesin motor contains information at time lags down to 0.1 ms. The quality and time resolution of the autocorrelation is primarily determined by the amount of signal photons used. By adding the autocorrelations of several tens of kinesins, fluorescence intensity changes can be observed at a timescale below 100 micros.

FIGURE 1

Schematic representation of the experimental setup and the motility assay. (A) Light paths of excitation (green) and emission (yellow) are indicated. The beam expander is represented by L1 and L2. DM is the dichroic mirror and EM is the emission filter. L3 is a lens located inside the microscope, and L4 and L5 image the confocal spot onto the optical fiber. The dashed line represents the port selection of the microscope. (B) Several kinesins in or near the confocal spot. The microtubule is attached to the glass via a positively charged surface. Casein prevents nonspecific sticking of the kinesin to the glass.

FIGURE 2

Characterization of hK421C's motility on microtubules using single-molecule wide-field fluorescence microscopy assays. (A) Histogram of the velocities of individual motors (N = 657). The solid line is a Gaussian fit to the distribution with its center position at 886 ± 7 nm/s (mean ± SE). (B) Distribution of the probability of being bound after a given run length (N = 204). A single exponential fit (solid line) indicates that the average run length is 1550 ± 30 nm (not corrected for photobleaching). The error bars are calculated from the square root of the absolute number of already detached motors and normalized.

FIGURE 3

Fluorescence intensity traces obtained when the confocal spot was positioned on a locally bleached microtubule. (A) A time trace of 30 s shows several events, most with comparable amplitude. (B) A full event, which is due to a motor landing before the confocal spot, walking in and through it and consequently showing a complete Gaussian profile. The solid line is a Gaussian fit to the trace. The amplitude (A) and width (σ) of the event are indicated for clarity. (C) A vanish event, which is due to a kinesin walking into the confocal spot but abruptly photobleaching or detaching somewhere in the spot and consequently showing only the leading flank of a Gaussian. (D) A landing event, which is due to a kinesin landing on the microtubule somewhere in the spot and consequently showing only the trailing flank of a Gaussian. In (C) and (D) the Gaussian was fit to the points of the leading and the trailing flank, respectively. Each error bar in graphs B, C, and D is the square root of the corresponding intensity.

FIGURE 4

Wide-field and confocal velocity measurements at two different ATP concentrations. Histogram of the widths of full events at 2 mM ATP (A) and 100 μM ATP (B) (N = 213 for 2 mM and N = 55 for 100 μM, events at different excitation powers were pooled). The calculated velocities from these widths are plotted together with the velocities obtained from wide-field measurements for both 2 mM ATP (C) and 100 μM ATP (D). The solid lines represent Gaussian fits to the data.

FIGURE 5

(A) The average amplitude of all full events as a function of excitation power. The solid line represents a linear fit (without offset) with a slope of 3700 ± 80 photons/(s μW). (B) The average width of all full events as a function of excitation power. The solid line represents a fit of a constant value (137 ± 2 ms). (C) Landing ratio (squares) and vanish ratio (dots) (see text) as a function of excitation power. The solid line represents a fit of a constant to the landing ratio data points (0.12 ± 0.01). The dashed curve represents a fit to an exponential decay (y = 1 − a exp(−x/P), where a = 0.97 ± 0.05 and P = 5 ± 0.5 μW).

FIGURE 6

Autocorrelation analysis of the time traces obtained with single-kinesin confocal microscopy. (A) Normalized autocorrelation of a single full event at an excitation power of 9.6 μW. (B) Normalized summed autocorrelation for three different excitation powers of only full events (time binned with 10 μs). Inset shows the exponential decay with their fits at the submillisecond timescale, probably due to triplet blinking of the fluorophore; the amplitudes increase with excitation power (0.049 ± 0.006, 0.079 ± 0.009, and 0.097 ± 0.008, respectively), whereas time constants of the fits are similar (0.10 ± 0.04, 0.056 ± 0.011, and 0.058 ± 0.010, respectively). (C) Normalized summed autocorrelations at three different excitation powers of vanish and full events (time binned with 1 ms). The small shoulders observed around 900 ms are due to small errors in the background correction and the finite time interval of the correlated trace (2000 ms in all graphs shown here). The red lines represent a simulation for the three different powers; see text for details of simulation.

FIGURE 7

(A) Simulated fluorescence intensity traces of kinesin stepping through a Gaussian excitation profile switching between high and low FRET states. The red graph is shifted upward for clarity; its axis is on the right-hand side of the graph. Each error bar is the square root of the intensity. (B) Intensity autocorrelations of the events shown in A. (C) Summing the autocorrelation of 20 separately simulated events improves the signal/noise ratio, in particular on the short timescales. See text for details of the simulations. For all three graphs, the red circles represent a simulation with the low FRET state (emitting 4 photons/ms in the acceptor channel) lasting on average 9.9 ms, and the high FRET state (20 photons/ms) lasting 0.1 ms. Black squares represent a simulation, with the low FRET state lasting 9.5 ms and the high FRET state lasting 0.5 ms.