The theoretical foundation necessary for achieving resolution beneath the diffraction barrier, which is actually composed of a family of physical concepts, was first advanced by Stefan Hell and associates with their introduction of the idea of reversible saturable (or switchable) optical fluorescence transitions (RESOLFT). This scheme focuses on fluorescent probes that can be reversibly photoswitched between a fluorescent "on" state and a dark "off" state (or between any two states A and B). The exact nature of the these states is variable and can be the ground and excited singlet states (S0 and S1) of a fluorophore as will be discussed below for STED microscopy, the excited singlet and dark triplet states utilized in ground state depletion (GSD) and ground state depletion-individual molecule return (GSDIM) microscopy, or the bright and dark states of a reversibly photoswitchable fluorophore (such as Cy5, kindling fluorescent protein, or Dronpa). In contrast, many of the optical highlighter fluorescent proteins, such as Eos, Kaede, Dendra2, and PA-GFP, which are capable of being permanently photoconverted from one emission bandwidth to another through covalent molecular alterations to the polypeptide backbone (and served as the foundation for the first PALM experiments), are not suitable probes for RESOLFT because the changes in fluorescent state are not reversible. The RESOLFT concept also includes switching isomerization states (such as cis-trans) and other optically bistable transitions in fluorophores.
The tutorial initializes with the depletion laser (red curves) and excitation laser (cyan curves) being stationary in the Specimen section of the window. Individual fluorophores are denoted by black boxes and the analog intensity of the labeled specimen outlined in a white curve. The excitation laser intensity drives the fluorophore from a bright (green) to a dark state. Intensity zeros exist between the depletion laser profiles. In order to operate the tutorial, use the Scan Progress slider to transition between fluorophores with the intensity zero or use the Autoscan button to enable automatic operation of the tutorial. As the tutorial progresses, individual fluorophores that are recorded on the detector appear on the CCD Image Plane and a corresponding image is sequentially constructed in the lower portion of the window. The Reset button can be used to re-initialize the tutorial.
For any given fluorophore or class of fluorescent probes, the number of switchable states is strictly limited so it is not surprising that many of the RESOLFT and single-molecule superresolution techniques described below rely on similar fluorophore switching mechanisms. In order to better understand these common properties, a close examination of the basic principles behind incoherently driven optical transitions is useful. When a molecule photoswitches from one state to the other, the probability that the molecule will remain in the first state decreases in an exponential fashion with increasing excitation light intensity. The term saturation intensity is used to define the light intensity at which the switching transition occurs (for example, when 50 percent of the molecules have transitioned from dark to bright) and is inversely proportional to the lifetimes of the two states. In situations where the excitation light intensity exceeds the saturation intensity, it becomes highly probable that one of the incoming photons will initiate the photoswitch. Fluorophores with long lifetimes in the initial and final switching states will afford more latitude in selecting excitation intensities and often exhibit higher fatigue levels (a measure of the ability to repeatedly photoswitch before being destroyed). The various fluorescent probes used for superresolution microscopy have lifetimes that differ significantly, as does the saturation intensity necessary to invoke photoswitching.
Among the differences between the RESOLFT concept (encompassing the related techniques of STED and GSD) and the single-molecule methodology employed by PALM and STORM are the switching mechanisms and the excitation light intensities necessary to photoswitch fluorescent probes. Extremely high intensities are necessary for switching off the excited singlet state with STED (using what is termed a depletion laser), but switching to a metastable triplet or similar dark state (GSD) requires a light intensity between a thousand and a million times lower. Likewise, even lower excitation intensity is capable of switching between the metastable dark and bright states of photoswitchable fluorescent proteins. Thus, the single-molecule localization techniques can accommodate less powerful light sources while providing much larger fields of view. In order to acquire coordinates of the photoswitched molecules, specific areas of the specimen are either targeted by defining scanning regions (STED, GSD, RESOLFT) or the individual fluorophores are allowed to stochastically turn on and off throughout the field of view (PALM and STORM). The different factors involved with these techniques dictate experimental parameters, such as imaging speed, instrument complexity, and the sensitivity of detection.
Presented in Figure 1 are several of the most important concepts surrounding superresolution imaging using RESOLFT techniques. A simplified Jablonski energy diagram (Figure 1(a)) depicts the possible electronic ground and excited states that are associated with on-off reversible optical transitions, such as the cis-trans isomerism associated with photoswitchable fluorescent proteins (as illustrated in Figure 1(c) for Kindling fluorescent protein). Depletion laser profiles 1-5 in Figure 1(b) highlight the spatial region in which the fluorophore exists in state A (or the "on" state) in regions where the standing wave of depletion light exhibits intensities between 10 (profile 1) and 1000 (profile 5) times the saturation intensity (as outlined in Equation (1)), and a zero node at xi. As the depletion laser intensity is increased, the region where the fluorophore is capable of residing in the A state is reduced to generate a subsequent increase in lateral resolution. A three-dimensional wireframe representation (Figure 1(d)) of the depletion and excitation laser profiles illustrates point-spread function modification by RESOLFT techniques.
The targeted photoswitching of RESOLFT (as generalized for STED and GSD) is performed by spatially controlling the depletion laser light intensity distribution so only that a limited number of molecules can be switched on while a majority of the others remain in the off state. Thus, when a specific wavelength having a defined intensity (that is much higher than the saturation intensity) is selected to switch the fluorophore off, applying this light in a spatially modulated manner is used to restrict the fluorophores remaining in the on state to a sharply defined region. The resolution (full width at half maximum; FWHM) for the point-spread function using these techniques is therefore defined by the following equation:
where λ is the wavelength of excitation light and the combined term η • sin(α) is the objective numerical aperture, as described above for the classical Abbe formula (Equation (1)). The variable a is a parameter that takes into consideration the shape of the spatially modulated depletion laser beam, which is often manifested in the form of a line shape or a "doughnut" having a central zero node (see Figure 1(b)). Under the square root, Imax is the peak intensity of the depletion laser and Is is the saturation intensity for the fluorophore being imaged. In cases where Imax equals zero, Equation (1) reduces to the Abbe diffraction limit. Conversely, when Imax is much greater than the fluorophore saturation intensity (in effect, the value of the square root increases), the point-spread function becomes very narrow and superresolution is achieved. For example, when Imax/Is equals 100, the improvement in resolution is about tenfold. Thus, in agreement with theory, sub-diffraction resolution beneath the Abbe limit scales with the square root of the light intensity depleting the ground state. The resolution of all methods that rely on targeted readout of fluorescent probes, including RESOLFT, STED, GSD, and SSIM, is governed by Equation (1) regardless of the fluorophore switching mechanism or the spatial modulation geometrical parameters (doughnut or line) dictated by the microscope configuration.
RESOLFT techniques require scanning the specimen with a zero node in the depletion laser field, but not necessarily using a single beam or a zero region that is geometrically confined to a point. Multiple dark lines or zeros can also be employed in conjunction with a conventional area-array (CCD) digital camera detector, provided the zeros or the dark lines are spaced further apart than the minimal distance required by the diffraction resolution limit. Scanning only with dark lines increases the resolution in a single lateral direction, but repeated scanning after rotating the pattern followed by mathematical deconvolution can provide sub-diffraction resolution across the lateral axis. The basic requirement for scanning the specimen is the reason why RESOLFT transitions (A/B or on/off) must be reversible. Molecules in one state must be able to return to their other state when they are scanned by the zero node. Note that saturated depletion of molecules in the excited state using a zero node focal spot produces a superresolution point-spread function that is not limited by the wavelength but only by the intensity of the depletion laser.
Another attractive feature of the RESOLFT concept is that the simplest mechanism for realizing a saturated optical transition is to excite the fluorophore intensely in what can be considered an "inverse" application of the approach. In this case, the ground state (S0) is depleted and the fluorophore is expected to reside largely in the fluorescent state (S1). The same RESOLFT principles discussed above still apply, except that the reverse state is now fluorescent such that highly defined dark regions are created that are surrounded by brightly fluorescent regions. The result is a "negative" image of the specimen details, which can subsequently be inverted to generate the final image through post-acquisition mathematical processing. The dark regions can either be lines generated by interference patterns or three-dimensional doughnuts. In the case of doughnuts, the result would be dark three-dimensional volumes that are confined by walls of intense fluorescence. Saturated structured illumination (SSIM) and saturated pattern excitation (SPEM) microscopy utilize this approach with line-shaped geometries to achieve superresolution images. The primary limitation of the inverse RESOLFT technique are the mandatory computations and high signal-to-noise ratio that are required in order to obtain the final image.
Presented in Figure 2 are high resolution images of 50-nanometer fluorescent beads along with intensity profiles generated by a uniform layer of fluorescent dye illuminated by a line pattern-modulated Gaussian beam in saturated structured illumination microscopy. In widefield microscopy the fluorescent beads exhibit poor resolution (Figure 2(a)) and appear as a blurred mass. Mathematical filtering of the widefield image (Figure 2(b)) produces scant improvement, but linear structured illumination (Figure 2(c)) yields higher resolution. The most dramatic increase in lateral resolution is achieved with SSIM (Figure 2(d)) using three harmonic orders (Figures 5(e) and (f)) during image processing. Intensity profiles through images of the thin layer of fluorescent dye with a modulation period of 2.5 micrometers are shown in Figure 2(e) at high (red curve) and low (yellow curve) peak energy densities. The lower curve closely follows the sinusoidal illumination pattern because the peak energy density lies beneath the saturation limit. In the top curve, the higher pulse energy induces fluorescence to saturate near the peaks. The Fourier transforms (Figure 2(f)) corresponding to the illumination patterns in Figure 2(e) exhibit several harmonics produced by the non-linearity of the saturating illumination intensity (red curve), whereas only the lowest harmonic is detectable in the lower energy pattern (yellow curve).
The intense excitation necessary for many of the RESOLFT techniques is compromised by the fact that high laser intensities often produce excessive rates of photobleaching so that fluorophores must be carefully chosen. Therefore, although the family of RESOLFT methods is not subject to the traditional diffraction barrier, the dependence of laser intensity on resolution gain installs another barrier that is governed by how much laser power the fluorophore can tolerate. The best remedy for this dilemma is to remove the necessity for strong intensities by implementing molecular transitions that occur with low depletion laser powers. Many bistable fluorescent probes fill this criterion and are able to be optically switched between fluorescent and non-fluorescent states through low-energy mechanisms such as photo-induced cis-trans isomerization (Figure 1(c)). In cases where both of the states are stable, the optical transition back and forth between the states can be completed at arbitrary or otherwise very long time scales, enabling the illumination to be spread out in time and reducing the required intensity (as well as photobleaching artifacts) by many orders of magnitude. Such a strategy has enabled parallelization of RESOLFT methods to allow their use in large area widefield imaging. Note that as described above, one of the principal ideas behind RESOLFT is that superresolution imaging does not necessarily require extremely high light intensities.
Stephen P. Price and Michael W. Davidson - National High Magnetic Field Laboratory, 1800 East Paul Dirac Dr., The Florida State University, Tallahassee, Florida, 32310.