The development of super-resolution fluorescence microscopy has allowed the diffraction-limited resolution to be surpassed1–2
. This advancement has been achieved either by spatially modulating the fluorescence emission with patterned illumination as in the cases of stimulated emission depletion microscopy (STED or RESOLFT)1,3
and saturated structured illumination microscopy (SSIM)4–5
, or by stochastically switching individual molecules on at different times as in the case of stochastic optical reconstruction microscopy (STORM) or (fluorescence) photoactivation localization microscopy ((F)PALM)6–8
. The latter approach further requires high-precision localization of single molecules9–10
and photoswitchable probes. These techniques have allowed biological structures to be imaged with resolution as high as ~20 nm. Recent demonstrations of super-resolution imaging in living cells, as exemplified by the video-rate STED imaging of synaptic vesicles in live neurons11
, have further enabled the characterization of cellular dynamics with sub-diffraction-limit resolution.
However, due to the intrinsic trade-off between spatial and temporal resolutions, the image resolutions achieved in live cells are substantially lower than that for fixed samples where the imaging speed is not a concern. The spatial resolution reported for the video-rate live-cell STED is ~60 nm in the lateral dimensions, 3 fold larger than what has been achieved on fixed cells11
. A recent live-cell STED study reports ~150 nm axial resolution when imaging samples in the xz
. Live-cell SIM has achieved ~10 Hz imaging speed in a wide field with a spatial resolution of ~100 nm in the lateral dimensions13
. For the single-molecule-based imaging methods, such as (F)PALM or STORM, this trade-off arises from the requirement that a sufficiently large number of localizations need to be accumulated for each snapshot in order to define a structure with a desired spatial resolution. This requirement is best characterized by the Nyquist criterion which equates the image resolution to 2/(localization density)1/D
, where D = 1, 2, or 3 for one- (1D), two- (2D) or three-dimensional (3D) imaging, respectively14
. Therefore, although photoactivation-facilitated high-density particle tracking has proven powerful for probing molecular motions in living cells15–17
, the Nyquist criterion has so far limited the spatial resolution of the single-molecule-based imaging methods to 40 – 70 nm in 2D with a 30 – 60 sec time resolution14,18
, when imaging photoactivatable fluorescent proteins in live cells. 2D super-resolution imaging has also been performed in living cells with photoswitchable dyes19–22
, but the localization density has not been characterized in these cases and thus the image resolution achieved is unclear. Because more localizations are inherently required to define a structure in 3D, it is expected that extending super-resolution imaging to 3D will further deteriorate the time resolution. Indeed, 3D (F)PALM or STORM has not yet been achieved for live cells. These limitations have significantly hindered the application of super-resolution fluorescence microscopy to the ultrastructural characterization of living cells.
Here, we report 2D and 3D super-resolution imaging of live cells with high spatial and temporal resolutions using photoswitchable dyes. We achieved a Nyquist resolution of ~20 nm with a time resolution as high as 0.5 sec for 2D STORM imaging. Moreover, 3D volumetric super-resolution imaging of live cells was achieved with an overall resolution of 30 nm in xy and 50 nm in z at time resolutions down to 1 – 2 sec, albeit with a relatively low number of independent snapshots. We demonstrated these imaging capabilities not only for exogenously added molecules, but also for intracellular proteins by delivering bright, fast switching cyanine dyes into living cells and by using genetic fusion strategies to specifically label proteins with these probes. In addition, we compared the 3D image resolutions achieved with six photoswitchable probes in live cells, including the photoswitchable cyanine dye, Alexa Fluor 647 (Alexa647), three cell permeable photoswitchable dyes, Atto655, TMR and Oregon Green, and two photoactivatable Eos fluorescent protein derivatives, mEos2 and tdEos.