Single-molecule high-resolution imaging with photobleaching

Abstract

Conventional light microscopy is limited in its resolving power by the Rayleigh limit to length scales on the order of 200 nm. On the other hand, spectroscopic techniques such as fluorescence resonance energy transfer cannot be used to measure distances >10 nm, leaving a "gap" in the ability of optical techniques to measure distances on the 10- to 100-nm scale. We have previously demonstrated the ability to localize single dye molecules to a precision of 1.5 nm with subsecond time resolution. Here we locate the position of two dyes and determine their separation with 5-nm precision, using the quantal photobleaching behavior of single fluorescent dye molecules. By fitting images both before and after photobleaching of one of the dyes, we may localize both dyes simultaneously and compute their separation. Hence, we have circumvented the Rayleigh limit and achieved nanometer-scale resolution. Specifically, we demonstrate the technique by measuring the distance between single fluorophores separated by 10-20 nm via attachment to the ends of double-stranded DNA molecules immobilized on a surface. In addition to bridging the gap in optical resolution, this technique may be useful for biophysical or genomic applications, including the generation of super-high-density maps of single-nucleotide polymorphisms.

Fig. 1.

(a) A plot of total integrated intensity versus time for two closely spaced Cy3 molecules, showing a two-step photobleaching behavior. (b) I_pre′ as calculated from I_pre and I_post.

Fig. 2.

Three examples of resolved, overlapping molecules with different separations. The black lines indicate the computed center-to-center separations. The errors are computed from the standard error of the mean (σ_μ) of the fit. (a) Separation = 132.9 ± 0.93 nm. (b) 72.1 ± 3.5 nm. (c) 8.7 ± 1.4 nm.

Fig. 3.

Histograms of measured end-to-end separations of 30-, 40-, and 51-bp DNA oligos, with fits (black lines). The estimated separations are 17.7 ± 0.7 nm, 13.0 ± 0.5 nm, and 10.7 ± 1.0 nm.

Fig. 4.

Centroid values of the histograms in Fig. 3, with the best fit for the line y = ax drawn through them. The error bars are derived from the standard deviation of the mean of the histograms in Fig. 3. The parameter a is estimated by the fit to be 3.4 Å per base pair.