To the Editor:
In the September 2010 issue of Nature Methods we demonstrated live-cell direct stochastic optical reconstruction microscopy (dSTORM) of histone H2B proteins using a trimethoprim chemical tag (TMP tag) for genetic encoding with photostable standard fluorophores1. The method takes advantage of the fact that cells contain the reducing thiol glutathione—a cysteine-containing tripeptide—at millimolar concentrations, which enables reversible photoswitching of synthetic organic fluorophores2, 3. The generality of the method can be easily understood considering that most Alexa Fluors (Invitrogen) and Atto fluorophores (ATTO-TEC) belong to the class of rhodamine and oxazine derivatives that have similar redox properties, that is, the triplet state of rhodamine and oxazine fluorophores is reduced by thiols such as glutathione3.
We now report that live-cell dSTORM of core histone H2B proteins in different eukaryotic cell lines is possible using commercially available SNAP tags as well. Protein-specific labeling is based on the irreversible reaction of human O6-alkylguanine-DNA alkyltransferase with the rhodamine O6-benzylguanine derivatives SNAP-Cell 505 (rhodamine green) and SNAP-Cell TMR-Star (tetramethylrhodamine) (New England BioLabs)4. We observed cell line–dependent differences in specificity and efficiency of labeling, introduce a very efficient method to minimize nonspecific adsorption of fluorophores and demonstrate how direct excitation of the radical anions at 405 nm enables precise control of the number of fluorophores residing in the 'on' state5.
To test for specific labeling, we transfected cells with plasmids coding for both histone H2B–SNAP-tag fusion proteins and histone H2B fused to enhanced GFP (Supplementary Methods and Supplementary Fig. 1). Unfortunately, most fluorophore conjugates such as SNAP substrates or TMP tags tend to nonspecifically adsorb on glass surfaces, complicating the accomplishment and analysis of super-resolution imaging experiments based on single-molecule localization (Supplementary Fig. 2). In our original work with TMP tags, we detached labeled cells and transferred them into new, freshly cleaned glass chambers to reduce background signals1. However, such a step is incompatible with handling procedures for several cell lines and is time-consuming because the cells have to reattach to the glass for at least 1–3 hours before imaging. We conducted new experiments with different blocking reagents and found that coating with BSA or poly(D-lysine) did not prevent nonspecific adsorption, whereas glycine coating enabled super-resolution imaging in living HeLa cells with two different SNAP substrates and minimal background signal (Supplementary Methods, Figs. 1a–c and Supplementary Fig. 2). Experiments with COS-7 and HEp2 cells revealed similar results (Supplementary Fig. 3). Irradiating the cells at only one wavelength—the readout wavelength for fluorescence excitation at 532 nm in the case of SNAP-Cell TMR-Star—of moderate intensity ensured that the number of fluorophores residing simultaneously in the fluorescent 'on' state was sufficiently low to allow accurate single-molecule localization (Fig. 1d) throughout the whole image stack. However, additional irradiation of the nonfluorescent radical anions at 405 nm (most rhodamine radical anions absorb at ~400 nm)5 with low intensity (<0.05 kW cm−2) recovered the fluorescent 'on' state. Thus, similar to photoactivated localization microscopy6 experiments, the number of fluorophores residing in the fluorescent 'on' state can also be controlled by the irradiation intensities of two independent laser beams. High irradiation intensities at 405 nm have to be avoided to provide useful fluorophore densities e nabling precise single-molecule identification and localization, and to minimize autofluorescence and cell damage (Figs. 1d,e).