Superresolution optical fluctuation imaging with organic dyes.

Superresolution far-field microscopy has experienced a tremendous growth since the introduction of the first concept in 1994[1], broadening the scope and applications of fluorescence microscopy beyond the diffraction limit. Various powerful and excellent methods, such as stimulated emission depletion[2], saturated structured illumination microscopy[3] and stochastic single-molecule switching methods such as STORM[4], (f)PALM[5, 6], directSTORM (dSTORM)[7, 8] and GSDIM[9], have been developed since. Notably, the key to all superresolution methods lies in the exploitation of a two-state transition (e.g. a fluorescent ‘on’ and a non-fluorescent ‘off’ state) of the molecules. Especially for the stochastic single-molecule switching methods (PALM, STORM etc.) it is necessary to precisely control both the time a fluorophore spends in the fluorescent as well as in the dark state. In the dSTORM concept, this is achieved by irradiation of the sample with one or two wavelengths in the presence of reducing agents. In parallel, the invention of a reducing and oxidizing buffer system (ROXS) made the fluorescence intermittency rates of most organic dyes tuneable to almost any degree[10, 11]. Immediate consequences are increased photo-stability and ‘non-blinking’ fluorescent dyes, which has been recently exploited to increase the performance of STED[12]. Here we demonstrate that reversible photoswitching of organic dyes using similar conditions as in dSTORM can be utilized for another superresolution technique, namely Superresolution Optical Fluctuation Imaging (SOFI)[13]. SOFI makes use of random temporal signal fluctuations of single emitters and uses them to achieve fast, background-free, 3D superresolution microscopy by means of higher order statistics (HOS). In contrast to single-molecule switching methods, many emitters can be ‘on’ at the same time in a diffraction-limited volume and still contribute to enhanced resolution. The SOFI protocol consists of recording a movie of the fluctuating signal on any kind of imaging platform. This movie is then processed by a software-based HOS analysis (that could also be implemented in hardware). So far, SOFI has been demonstrated to work with blinking quantum dots (QDs) and a resolution enhancement of a factor of two could be achieved in generic imaging applications[14]. Since almost all organic dyes exhibit fluorescence intermittency (e.g. due to intersystem crossing to the dark triplet state, which gives rise to a fluctuating fluorescence signal), they could potentially be used for SOFI as well. The use of much smaller and targetable organic dyes (as compared to QDs) could expand the scope of SOFI to a large array of live and fixed cell imaging applications. On the other hand, using dyes for SOFI introduces many challenges: (i) the immanent photo-bleaching of dyes upon illumination limits the acquisition time and potentially compromises the SOFI algorithm (since it requires a temporally constant mean signal); (ii) a lower signal to noise ratio (as compared to QDs) degrades the algorithm’s performance; (iii) typical intersystem crossing rates of conventional organic dyes result in microseconds ‘on/off’ blinking, whereas most movie acquisitions for SOFI are limited to millisecond timescales. Here we apply experimental conditions similar to the ones used in the dSTORM concept to adjust reversible photoswitching of organic dyes to a suitable level for SOFI, using an EMCCD camera acquisition. We chose the dye Alexa647 (which exhibits light-driven microsecond fluorescence intermittency) for these experiments (see Experimental Section)[7]. Note that similar experimental conditions were shown to control blinking of a large selection of organic dyes[8] and therefore our choice of a cyanine dye is almost arbitrary. The experiment was carried out on fixed COS-7 cells whose β-tubulin network was immuno-labelled with Alexa647-conjugated antibodies. We recorded a movie of 1000 frames at a frame rate of 20 Hz (see Supporting Information) and subsequently analyzed it with an extended SOFI algorithm[14]. This enabled us to produce SOFI images that have four-times more pixels than the original acquired image. Fig. 1 clearly shows resolution enhancement, and in addition, a striking reduction in background fluorescence can be observed in the SOFI image. This is due to the fact that the SOFI algorithm inherently eliminates the non-fluctuating background signal. Also, the inherent SOFI’s optical sectioning attribute contributes to a ‘cleaner’ image since out-of-focus light is suppressed even though imaging was done in close-to-TIRF configuration. The resolution enhancement translates in a smaller Rayleigh limit of 169 nm (as compared to 290 nm for the conventional imaging), which was assessed by comparing the same line profile in Fig. 1a and 1b (Fig. 1c). Figure 1 Fluorescence and corresponding SOFI images of β-tubulin network of COS-7 cells. White boxes are magnified regions shown in the upper right corner. An intensity cross-section (white line) is taken to evaluate the resolution enhancement. (A) Original ... In order to circumvent imaging artefacts due to dye photo-bleaching during acquisition (which would manifest as non-resolved features in the SOFI image), the acquired movie was analysed piecewise in blocks of frames. Within each block, the change of the mean signal due to photo-bleaching was negligible (see Experimental Section) and therefore the requirement for successful SOFI imaging was given for each movie block. SOFI images for all movie blocks were summed together before any further analysis. In conclusion, we demonstrated a convenient superresolution imaging method using conventional organic dyes with an unmodified TIRF-microscope. Compared to single-molecule superresolution methods based on switching, SOFI reached superresolution performance in equivalent acquisition times (50 s for the data presented here). It has, however, the potential to run at even faster speeds (topic of current research). Note that even though we applied photoswitching conditions in order to tune the on/off times, the only SOFI prerequisite was to be able to monitor dye fluctuations. Therefore, it is likely that SOFI could be implemented in the future with faster blinking dyes as faster cameras come online[15, 16]. In contrast to PALM and STORM, SOFI does not require long off times or has to maintain a certain on/off time ratio. This obviously points to future experiments, which will be performed in live cells, where tuning capabilities and measurement times are limited and fast acquisition is crucial. Since the signal to noise ratio of dyes is considerably lower in comparison to QDs, SOFI imaging is more demanding, i.e. for QDs an arbitrary frame rate of the camera can be used (leading to an improved signal) whereas for dyes a fast frame rate is desirable since photo-bleaching limits the acquisition time. Photo-bleaching is also the limiting factor for applying higher-order correlations in order to improve resolution more than demonstrated here.