608 Wang et al. / Front Inform Technol Electron Eng 2019 20(5):608-630 Frontiers of Information Technology & Electronic Engineering www.jzus.zju.edu.cn; engineering.cae.cn; www.springerlink.com ISSN 2095-9184 (print); ISSN 2095-9230 (online) E-mail: [email protected] Review: Super-resolution optical microscope: principle, * instrumentation, and application Bao-kai WANG1, Martina BARBIERO1, Qi-ming ZHANG1, Min GU†‡1,2 1Laboratory of Artificial-Intelligence Nanophotonics, School of Science, RMIT University, Melbourne, Victoria 3001, Australia 2Center for Artificial-Intelligence Nanophotonics, School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China †E-mail: [email protected] Received July 27, 2018; Revision accepted Oct. 29, 2018; Crosschecked May 13, 2019 Abstract: Over the past two decades, several fluorescence- and non-fluorescence-based optical microscopes have been developed to break the diffraction limited barrier. In this review, the basic principles implemented in microscopy for super-resolution are described. Furthermore, achievements and instrumentation for super-resolution are presented. In addition to imaging, other ap- plications that use super-resolution optical microscopes are discussed. Key words: Super-resolution; Imaging; Optical microscope https://doi.org/10.1631/FITEE.1800449 CLC number: O439 1 Introduction proach for determining the resolution of a microscope in the frequency domain is by measuring the cut-off The maximum spatial resolution of a conven- frequency of the optical transfer function (OTF). OTF tional optical microscope is limited to half the wave- is the Fourier transform of PSF (Gu, 2000). length of light. The limitation is due to the diffraction In the visible wavelength region, the resolution barrier imposed by the slight bending of a light beam of a conventional optical microscope is around when it encounters an object. According to Abbe 200–300 nm in the lateral direction and 500–700 nm (1873), a light beam of wavelength λ focused by a in the axial direction. lens with numerical aperture nsin α (sin α<1) cannot In 2014, the Nobel Prize in Chemistry was resolve objects closer than distance d=λ/(2nsin α) (Gu, awarded jointly to Eric BETZIG, Stefan W. HELL, 1996, 2000). The diffraction of light causes a sharp and William E. MOERNER for the development of pointed object to blur into a finite-sized image spot super-resolution fluorescence microscopes that allow through the optical microscope. the imaging resolution limitation due to the diffrac- The resolution of microscopes is determined by tion of light to be overcome. Stefan W. HELL was the the size of the point spread function (PSF). The PSF is first scientist to break the diffraction barrier theoret- defined as the three-dimensional (3D) intensity dis- ically (Hell and Wichmann, 1994) and experimentally tribution of the image of a point object. Another ap- (Klar and Hell, 1999) with the concept of stimulated emission depletion (STED). Eric BETZIG (Betzig, ‡ Corresponding author 1995; Betzig et al., 2006) pioneered the method called * Project supported by the Australian Research Council (ARC) through photoactivated localization microscopy (PALM) in the Discovery Project (No. DP170101775) 2006. William E. MOERNER’s contributions center ORCID: Min GU, http://orcid.org/0000-0003-4078-253X © Zhejiang University and Springer-Verlag GmbH Germany, part of on the first optical detection and spectroscopy of a Springer Nature 2019 single molecule in condensed phases (Moerner and Wang et al. / Front Inform Technol Electron Eng 2019 20(5):608-630 609 Kador, 1989) and on the observations of on/off intensity, a fluorophore in the ground state is imme- blinking and switching behavior of green fluorescent diately pumped to the excited state. The fluorescence protein (GFP) mutants at room temperature (Dickson lifetime is determined by the fluorescent emission et al., 1997). rate. When the fluorophore reaches saturation, its In recent years, a number of novel super- fluorescence is not proportional to the intensity of the resolution fluorescence- and non-fluorescence-based excitation light. Under this circumstance, a saturated methods have been developed. The fluorescence-based excitation structured illumination pattern is achieved. methods are further divided into spatial- and time- When the intensity reaches the saturated level, the domain techniques. pattern becomes flat (Fig. 1b). In this review, we summarize the basic principles Higher frequency information is encoded by implemented in fluorescence- and non-fluorescence- mixing saturated excitation structured illumination based optical microscopy. We also review the patterns containing more Fourier components (Fig. 1c). achievements and developments of these optical mi- croscopes, as well as their applications to other fields. Lens Sample 2 Principle 2.1 Fluorescence-based super-resolution 2.1.1 Spatial-domain methods Grating Objective (a) 1. Saturated structured-illumination microscopy Structured-illumination microscopy (SIM) achieves spatial resolution beyond the diffraction limit with a wide-field microscope with spatially structured illumination. A grating is used to generate a structured illu- mination pattern that varies in lateral direction (Gus- tafsson, 2000) (Fig. 1a) or three dimensions (Gus- tafsson et al., 2008). The structured illumination ex- tends resolution beyond the cut-off by moving high- Position frequency information into the observed images in the (b) form of Moiré fringes. The Moiré fringes are gener- ated through spatial frequency mixing (Fig. 1c). The new image is computationally extracted. The method demonstrates imaging with twice the spatial resolu- tion of conventional microscopes. Sample SIM is applied in three dimensions to double the SIM SSIM axial and lateral resolutions. Resolutions of about (c) 100 nm in the lateral direction and about 300 nm in Fig. 1 Principles of structured-illumination microscopy the axial direction were obtained (Schermelleh et al., (SIM) and saturated structured-illumination microscopy (SSIM): (a) structured illumination generated by a 2008; Shao et al., 2011). grating; (b) SIM excitation pattern (blue line) and SSIM In saturated structured-illumination microscopy excitation pattern (red line); (c) imaging the sample with (SSIM), a nonlinear phenomenon, the saturation of SIM and SSIM fluorescence emission, is applied to achieve a theo- References to color refer to the online version of this figure retical unlimited resolution (Gustafsson, 2005). The saturation occurs when a fluorophore is illuminated The resolving power is determined by the signal- by optical intensities in the order of about 106 W/cm2. to-noise ratio, which in turn is limited by photo- Under an excitation beam with high optical bleaching. Experimental results showed that a 610 Wang et al. / Front Inform Technol Electron Eng 2019 20(5):608-630 two-dimensional (2D) point resolution of less than 3. Stimulated emission depletion 50 nm was possible to achieve with bright photostable The concept of STED microscopy was first samples (Gustafsson, 2005) and photoswitchable proposed by Hell and Wichmann (1994), and proteins (Rego et al., 2012; Li et al., 2015). demonstrated experimentally in Klar and Hell (1999). 2. Image scanning microscopy Super-resolution is achieved through the selective Image scanning microscopy (ISM) was previ- deactivation of fluorophores via stimulated emission. ously described by Sheppard (1988) and demon- Under a depletion beam, a fluorophore in the excited strated by Müller and Enderlein (2010). ISM doubles state interacts with a photon that matches the energy the resolution of a scanning confocal image. The difference between the excited state and the ground principle of ISM is to extract the inherent high- state (Fig. 2a). In this case, the fluorescence emission frequency information from a laser scanning confocal of the fluorophore is depleted. microscope (LSCM). The origin of the super- resolution information can be understood in two dif- S1 ferent ways. The first way is to describe the origin based on the idea of overlapping excitation and emission PSFs in an LSCM (Sheppard, 1988; Shep- Excitation Fluorescence Stimulated pard et al., 2013); the second description is the same emission emission as SIM (Müller and Enderlein, 2010). The super- resolution information can be collected simply by reducing the size of the pinhole in an LSCM. How- ever, a smaller pinhole rejects a large amount of light, S0 reducing the signal-to-noise ratio. ISM can recover (a) the lost super-resolution information. The optical setup of ISM involves recording the signal that passes Objective through the pinhole on an array detector and obtaining an image at each scanning position. The simplest way to recover a final super-resolution image is pixel re- Detector Sample assignment (Sheppard et al., 2013). For every scan- ning position, the image obtained is reduced by some factor, and added to a running-total image that is Phase modulation centered at the scanning position of the beam. After a Excitation STED complete picture has been reconstructed, further res- beam beam olution enhancement can be achieved by Fourier (b) reweighting. To achieve faster acquisition, a digital micro- mirror device (DMD) is applied to obtain multifocal + = structured illumination in ISM (York et al., 2012). Effective PSF Following this idea, ISM is also achieved in a con- Excitation PSF focal spinning-disk microscope (Schulz et