Although GFP itself has intrinsic photoswitching properties (20), the first effective photoswitchable fluorescent protein was paGFP (15). It is synthesized in a dark “off” state, but is irreversibly converted by irradiation with 405 nm light into a green-emitting, 488 nm excitable “on” state which can emit ∼160 detectable photons before bleaching (21). psCFP2 (22) is another fluorescent protein that emits in the same spectral window (green) as paGFP, but yields more detectable photons (∼250 median; (23)). In its “off” state, it emits blue photons when irradiated with 405 nm light; excitation of this blue form irreversibly converts the protein to a green-emitting “on” state excitable by 488 nm light. Unfortunately neither paGFP nor psCFP2 perform well in densely labeled samples; both have substantial “off” state fluorescence (∼1% of “on” state for paGFP; (15, 17)), and both are also activated to some extent by the 488 nm laser used to image their “on” states.

Several “reversible” green fluorescent proteins have also been developed. The first to be used for SMLM was Dronpa. This FP emits in the green channel, and has its excitation peak at 503 nm (24). Upon excitation, it emits a median of ∼120 detectable photons (17) before entering a stable dark state (with a half life of 840 min; (25)). UV illumination can return it to its “on” state. Dronpa can cycle many times before permanently photobleaching (24). It also appears to have a lower “off” state fluorescence than psCFP2 or paGFP, and so may be more appropriate for high density labeling. Several variants of Dronpa have been created that are slightly brighter or have different spectral properties; some have been used in SMLM experiments (25). One variant, Padron (25), has the opposite photophysical properties to Dronpa; it is switched “on,” and excited by 488 nm light, but switched “off” by 405 nm light. As Padron is not deactivated by excitation light, it may have a high photon yield.

IrisFP (21) is a complex photoactivatable FP with multiple switching behaviors. It initially emits in the green band under 488 nm excitation, but after releasing ∼170 detectable photons, switches to an inactive state. Exposure to 405 nm light drives the protein back into its original green-emitting state, while further UV exposure converts the green FP to an orange-emitting form excitable by a 568 nm light. This later form emits ∼350 detectable photons before photobleaching.

One of the first fluorescent proteins used in SMLM was the orange fluorescent protein EosFP (7). This protein is synthesized as a green emitting FP excitable by 488 nm, but can be—by exposure to UV light—converted to a orange-emitting “on” state excitable by 568 nm (26). The monomeric form of this protein matures poorly at 37°C, and thus seems unsuitable for expression in mammalian cells (26). However, its tandem dimer tdEos is fully functional at 37°C, and has been used extensively in SMLM, emitting a median ∼450 detectable photons before photobleaching (23).

An effort to improve EosFP led to the creation of mEos2, a variant that matures correctly at 37°C when expressed as a single fusion protein (i.e., not as a tandem dimer). It is now widely used in SMLM, and has an excellent photon yield (median ∼360 detected; (23)). Unfortunately, mEos2 has several properties that can cause problems in SMLM experiments. It has a tendency to aggregate in vitro, and thus may cause artifactual clustering of its fusion partner (23). Activated mEos2 molecules have also been shown to enter long-lived dark states before photobleaching (27, 28); thus in a typical SMLM experiment, individual molecules can be localized several times. For quantitative studies of protein clustering, careful control experiments are needed, as bona fide molecular clusters cannot be easily distinguished from single mEos2 molecules that have been localized multiple times. Similar blinking behavior has also been reported for its EosFP parent (26). Dendra2 and mKikGR are both monomeric green-to-orange photocovertable FPs that lack the tendency of mEos2 to aggregate (29, 30); however, neither yields as many photons as mEos2 (e.g., ∼130 median detected for Dendra2; (23, 31)).

Several dark-to-orange photoactivatable FPs have also been developed. PAmCherry1 is reported to yield as many photons as mEosFP (median 413 detected under 568 nm excitation; (32)) and has virtually no “off” state fluorescence. PATagRFP, a second dark-to-red FP, is three times brighter, and roughly twice as photostable as PAmCherry1 (33) in ensemble measurements, and so should also be useful in SMLM experiments. As the fluorescence emission of PATagRFP and PAmCherry1 are spectrally distinct from those of the green fluorescent proteins noted above, they can be used in two-color experiments.

Recently, a new orange-to-red photoswitchable FP, PSmOrange, has been introduced. This psFP is revolutionary as it is the first to possess a red-emitting “on” state. It is initially orange, with peak emission at 565, but, upon exposure to 488 nm light, switches to a red-emitting FP with excitation and emission spectra comparable to those of Cy5 (34). Unfortunately, the photophysical properties of this psFP do not allow it to be localized as accurately as other psFPs (e.g., PAmCherry1). However, its red excitation/emission spectra may make it useful for in vivo experiments.

In addition to fluorescent proteins that have been designed specifically for photoswitching, a number of conventional FPs have been used as photoswitches. GFP, for example, has long been known to enter stable nonfluorescent off states under constant 488 nm excitation; molecules can be returned to their active states by irradiation with UV light (20). EYFP can be similarly operated as a photoswitch, and has been used for super-resolution imaging in bacteria (35). A different strategy has been presented by the Sedat lab who used “oxidative reddening” to operate EGFP as a photoswitch, and demonstrated super-resolution imaging of histone proteins (36).

Although the fluorophore inside a fluorescent protein is somewhat isolated from their environment by the beta barrel surrounding it, their photophysical properties can be easily altered, or destroyed, by exposure to buffers components. Chemical fixation, for example, can inactivate or reduce the activity of psFPs. We and others have found that glutaraldehyde is particularly destructive (18), while formaldehyde has more moderate effects (e.g., formaldehyde does not affect GFP fluorescence; (37)). Other data suggests that fixation can—in some cases—reduce, but not destroy the photon yield of psFPs (32). Mounting media might also have effects on the switching of psFPs. Such media are used to match the refractive index of a sample to that of the coverslip, in order to avoid imaging aberrations. However, most SMLM studies to date have not used mounting media but rather, immersed the samples in innocuous aqueous buffers (e.g., PBS or PHEM; (17, 32, 38)), as a result, little is known about the effects of mounting media on psFP activity. However, one can speculate that the redox environments and reduced oxygen levels of some mounting media might have strong effects on psFP activity. A recent study, for example, has found that reducing agents and oxygen depletion can both influence psFP behavior (28) (the influence of redox reagents is further discussed in Subheadings 5.1 and 5.2.4). However, we do note that recent SMLM studies that did not use special mounting media have achieved excellent results, probably because they observed only molecules found relatively close to the coverslip (using total internal reflection microscopy, which requires refractive index mismatch). As SMLM is applied to molecules found deeper in samples, we imagine that the issue of mounting media compatibility will become increasingly important.

Although psFPs are powerful tools for SMLM, it is sometimes desirable to use fluorescent dyes instead. These fluorophores exhibit photon yields that are considerably higher than those of fluorescent proteins, a fact that directly translates to increased localization accuracy and spatial resolution. Under standard imaging conditions (e.g., phosphate buffer saline), synthetic dyes typically spend most of their time in an “on” state. However, several different strategies have been developed that allow dyes to be switched “off,” most of which involve redox chemical processes. Novel dyes, which can be activated by the light-induced cleavage of a quenching group (39) have also been developed, and thus can be operated similarly to irreversibly activatable FPs.