Viewing dynamic assembly of molecular complexes by multi-wavelength single-molecule fluorescence

Abstract

Complexes of macromolecules that transiently self-assemble, perform a particular function, and then dissociate are a recurring theme in biology. Such systems often have a large number of possible assembly/disassembly intermediates and complex, highly branched reaction pathways. Measuring the single-step kinetic parameters in these reactions would help to identify the functionally significant pathways. We have therefore constructed a novel single-molecule fluorescence microscope capable of efficiently detecting the colocalization of multiple components in a macromolecular complex when each component is labeled using a different color fluorescent dye. In this through-objective excitation, total internal reflection instrument, the dichroic mirror conventionally used to spectrally segregate the excitation and emission pathways was replaced with small broadband mirrors. This design spatially segregates the excitation and emission pathways and thereby permits efficient collection of the spectral range of emitted fluorescence when three or more dyes are used. In a test experiment with surface-immobilized single-stranded DNA molecules, we directly monitored the time course of a hybridization reaction with three different oligonucleotides, each labeled with a different color dye. The experiment reveals which of the possible reaction intermediates were traversed by each immobilized molecule, measures the hybridization rate constants for each oligonucleotide, and characterizes kinetic interdependences of the reaction steps.

FIGURE 1

Optical layouts for TIR fluorescence microscopes. (a) Conventional through-objective TIR. The incoming short-wavelength (λ₁) excitation beam is reflected by a dichroic mirror, passes through the objective lens, and is totally internally reflected at the glass-aqueous interface (inset). The longer-wavelength (λ₂) fluorescence emission is transmitted through the dichroic. (b) Schematic layout (not to scale) of the multi-wavelength single-molecule TIR fluorescence microscope. The design incorporates two small broadband mirrors (M) for directing multi-wavelength laser excitation into the objective (M_in) and deflecting the reflected laser beam that exits the objective away from the image path (M_out). Fluorescence emission is transmitted through the gap between the mirrors. ND, neutral density filter; S, shutter; Q, quarter-wave plate; CL, collimating lens pair; LF, laser filter; D, dichroic; L, lens; I, iris; and NF, long-pass (HQ505LP, Chroma Technology; Rockingham, VT) and 532/633 nm dual notch filter (Barr Associates; Westford, MA) filters; EMCCD, electron multiplying charge-coupled device camera. The dashed box encloses the dual view optics. (c) Photograph showing the physical arrangement of the objective and optics M_in, M_out, I1, and M6 in the multi-wavelength microscope.

FIGURE 2

Transmission of emitted fluorescence through the objective back aperture. Fluorescence from a flow chamber with surface-adsorbed fluorescent beads (F8786, Molecular Probes; Eugene, OR) excited at 532 nm is observed looking at the objective back aperture through a Bertrand lens positioned after M6 (see Fig. 1 b). The green region is the portion of the aperture through which emitted fluorescence is transmitted. The small areas blocked by the input and output mirrors are seen as semicircular outlines at left and right.

FIGURE 3

Removal of objective lens autofluorescence by the laser coupling mirrors. Images show the objective back focal plane as imaged with the EMCCD camera without (a) and with (b) the output mirror M_out (see Fig. 1 b) in place. A sample on a glass coverslip of aqueous buffer without any fluorophore was excited at 532 nm, 0.7 mW incident power. Filters attenuating the 532-nm excitation wavelength by 10¹²-fold were present in the imaging path. Alterations of the attenuation factor caused a far less than proportional change in the observed spot intensity (data not shown), indicating that the spot is autofluorescence rather than scattered or reflected excitation light.

FIGURE 4

Preparation of surface-immobilized DNAs labeled with individual dye molecules that fluoresce at one or more wavelengths. Single-stranded DNA molecules were tethered to a PEG-coated slide surface through biotin-streptavidin (SA) linkages. DNA quantity was limited to produce a surface density ≪1 molecule per diffraction-limited spot. One or more dye-labeled oligonucleotides introduced into the flow cell can then hybridize with this surface-tethered ssDNA.

FIGURE 5

Image of a single Cy3-labeled DNA molecule. (a) Image acquired over 1.5 s. The integrated spot is comprised of ∼30,000 photons. (b and c) The central x row (b) and y column (c) of pixel intensities (points) together with the corresponding values (lines) from the fit of the image to the two-dimensional Gaussian exp.

FIGURE 6

Photobleaching statistics for 137 Cy3-DNA molecules. Images were obtained with 0.7-mW incident 532-nm excitation.

FIGURE 7

Time records of fluorescence from single Cy3-DNA molecules immobilized on fused silica (a) or glass (b) coverslips. Emission at wavelengths <635 nm from a 0.8-μm² (a) or 0.4-μm² (b) area was recorded with 0.2-s time resolution using 0.7-mW 532-nm excitation. After dye photobleaching (arrows), the camera and excitation laser were sequentially shuttered to record the relative amounts of background noise caused by camera noise, stray light, and autofluorescence.

FIGURE 8

Field of DNA molecules hybridized to oligonucleotides labeled with three different dyes (see Fig. 4). (a–e) Fluorescence images (18.3 × 20.2 μm) of the same field excited with a single laser wavelength (Ex) of (a) 633, (b) 532, or (c) 488 nm (close to the excitation maxima of Cy5, Cy3, and Alexa 488, respectively), or (d–e) with all three lasers simultaneously. Emitted fluorescence (Em) images were collected from either the long- (>635 nm; a and e) or short- (<635 nm; b–d) wavelength area of the dual-view optics image. The duplex DNAs were preformed on a fused silica coverslip, and unbound dye-labeled oligonucleotides were flushed from the cell before imaging. Each image is an average of 20 0.1-s duration frames. Typical single-molecule spots consist of ∼3000, ∼3600, and ∼2300 photons in a–c, respectively. (f) Overlaid positions of each kind of dye molecule as detected in (a) (Cy5; red open boxes), (b) (Cy3; green ×), and (c) (Alexa 488; blue open circles). Some DNA molecules do not contain all three dye moieties, due either to incomplete labeling or to photobleaching.

FIGURE 9

Binding of dye-labeled oligonucleotides to a single immobilized DNA molecule. The binding reaction was initiated by introducing a mixture of the three different dye-labeled oligonucleotides labeled with Alexa 488, Cy3, and Cy5 (6.3 nM, 5.0 nM, and 2.0 nM, respectively) into a glass flow chamber with the surface-immobilized target ssDNA. The oligonucleotides were present in solution throughout the duration of the recordings. Fluorescence from the dyes was detected by cyclically exciting the sample at 488 nm (five 200-ms frames), 532 nm (25 frames), and 633 nm (five frames). (a–c, e) Time records of integrated fluorescence from a single spot in the field (0.4 μm²) recorded with the specified excitation and emission wavelengths. Zero emission is taken to be the signal recorded when the camera is shuttered. Presence of Alexa 488, Cy3, and Cy5 (d) were determined from records (a–c), respectively. Record (e) shows signals due to cross excitation of Cy5, FRET to Cy5, and leak-through of Cy3 emission into the >635-nm detection channel.

FIGURE 10

Time courses of dye-labeled oligonucleotide binding to multiple immobilized target DNA molecules in the same experiment shown in Fig. 9. These data were recorded from a single field of view containing ∼260 target DNA molecules. (a) Cumulative number of observed binding events, n, for each oligonucleotide as a function of time, t, after the start of observation (points). One-parameter fits (lines) are to the function n(t) = [n(t_max)/(1 − exp(−t_max / τ))][1 − exp(−t/τ)], where t_max is the longest observation time for each oligonucleotide. The best-fit values of τ are 182, 905, and 2837 s, respectively, for the Cy5, Cy3, and Alexa 488 oligonucleotides. (b) Time until Cy3 oligonucleotide binding and time until Alexa 488 oligonucleotide binding for the 122 individual DNA molecules to which both oligonucleotides were observed to bind (observation times 1598 s for Cy3 and 3233 s for Alexa 488). Clustering of points near the line suggests that binding of the two oligonucleotides is not independent. (Inset) Histograms of Δt = (Alexa 488 oligonucleotide binding time) − (Cy3 oligonucleotide binding time) for the data (top) and for a simulation (bottom) based on the measured second-order rate constants and assuming independent binding of the two oligonucleotides. The large peak in the data histogram that is not seen in the simulation indicates an enhanced rate of Alexa 488 oligonucleotide binding to target DNAs on which the Cy3 oligonucleotide is already bound.