Super Resolution Fluorescence Microscopy - Lecture Notes | PHYS 552, Lab Reports of Optics

Download Super Resolution Fluorescence Microscopy - Lecture Notes | PHYS 552 and more Lab Reports Optics in PDF only on Docsity! Seeing the Unseen Before -- Super-resolution Fluorescence Microscopy Juan Guan guan4@uiuc.edu We have all heard of the term “diffraction limit” in Physics 101. If two points are very close, their individual positions cannot be resolved precisely. When it comes to fluorescent microscopy, the light intensity from a point source can be viewed as a point spread function (PSF) having Gaussian width of λ/2(NA). When two fluorophores are separated by less than PSF width, their images will substantially overlap, and thus their individual positions are difficult or impossible to resolve. For visible light λ of ~ 550 nm and a typical NA of ~ 1.4, the width is ~ 200 nm, which determines the image resolution. Meanwhile, cutting-edge research now is trying to understand things happening at the nanometer scale, in biological structures and processes, in nanofabrication industry and so on. Are there ways to overcome the diffraction limit? The answer is yes! Researchers have developed new techniques that have a resolution of as high as ~ 20 nm. This report will cover the basic ideas, the detailed techniques, and some recent successful applications. One active group in this field is Zhuang lab. [1,2,3,4] Interestingly, if you look into the history of how they started, you will find a 2005 paper [5] where they essentially have everything ready for super-resolution microscopy. But they did not know it. They discovered this remarkable phenomenon (a green laser can switch the fluorescence of a molecule on and a red laser switches it off) without really knowing what to do with it. Inspired by the concept of using FRET as a “spectroscopic ruler” [6], they proposed, since this switch is more sensitive to short distances, a “short-range” spectroscopic ruler. There is no way for me to know how the idea of super-resolution microscopy got around, but in less than a year, they had come up with stochastic optical reconstruction microscopy (STORM). And, in a sweeping manner, in the next year or two, they have pushed the technique far along to do multi-color imaging, 3-D imaging, live cell imaging, etc. Another guy called Betzig stepped into the field, more or less naturally, following tens of years of related efforts to go beyond the diffraction limit. He developed near-field scanning optical microscopy (NSOM) in 1992 [7] and in 1995 put forward the concept of control viable [8], essentially the STORM idea and eventually came up with nice applications in studying dynamic processes in living cells. They are not alone in this newly-emerging area. Moerner at Stanford and Marriott at Wisconsin are just a few other examples. Fig 1 (adapted from [2]) here constrasts conventional microscopy image (left) and STORM image (right) of the same area and convincingly shows that the resolution of images can indeed increase many fold. Fig 1 How does STORM beat the diffraction limit, that is, to resolve two points that are very close? The basic idea of STORM is to first light up ONLY one point, precisely determine its position, and then resolve the position of the other point when only that point is lit. Similarly, for multiple points, first, a fraction of molecules is lit up and their Fig 3 Next, I will explain how image analysis is done to resolve individual position of each molecule. A typical frame can look like Fig 3, adapted from [5]. In one frame, fluorescent regions (red dots in Fig 3) are fit to a continuous ellipsoidal Gaussian. Here, (x’, y’) and (x, y) are referring to the same pixel position in different coordinates, since the axis of the ellipsome can tilt from the image edge. Details can be found elsewhere.[1] A is the background fluorescence level, I0 is the amplitude of the peak, a and b are the widths of Gaussian distribution along x’ and y’ directions. A few criteria are used to reject bad data. For example, if the fluorescence in a particular region doesn’t last long enough, or the ellipticity (defined as 2|b-a|/(b+a)) is larger than a certain value, or the total photon numbers collected from one such region (can be roughly estimated from a*b*I0) is smaller than a threshold, this region is rejected. From the image analysis algorithm, one sees that the resolution of the reconstructed final image is determined by how precisely the position of each individual point can be resolved. Now experimentally, the resolution of each point is ~ 20 nm. Consequently, the resolution of the final image is ~ 20 nm, meaning, positions of points separated by 20 nm can be resolved. STORM and related techniques are relatively new, yet they evolve rapidly. It is worthy pointing out a few directions that the techniques are extending into. The striking fact is that these beautiful examples I will cover now are all taken from literatures less than one year old. A. Multi-color super-resolution imaging. Zhuang et al [2] in 2007 successfully incorporated into STORM technique different combinatorial pairings of activators and reporters. Parenthetically, the exact mechanism of how a photo-switchable pair works is not fully understood yet. The process can be viewed roughly as a light pulse activates the activator, which instead of emitting on its own, transfer the energy to the reporter which then fluoresce. The energy transfer process in a way is similar to what happens between a donor-acceptor pair via Forster Resonance Energy Transfer (FRET). The efficiency is more sensitive to distance however. [5] Fig 4 One apparent application for the combinations of one kind of activators and different kinds of reporters is that one can simultaneously track multiple species through different detector channels with one activation wavelength. Conversely, combinations of different kinds of activators and same reporters allow one to use different activation wavelengths to select different species at different time and data can be conveniently collected through one single detector. Fig 4 here shows an image of clathrin coated pits (red) and microtubules (green) when they are labeled by different activators pairing the same reporter, adapted from [2]. Fig 5 B. 3-D super-resolution imaging. Zhuang et al [3] in 2008 successfully acquired 3-D STORM image without scanning the actual sample. They creatively incorporated a cylindrical lens (instead of a spherical lens) to introduce shape anisotropy, that is, different elongation of images in x and y directions, at various z positions. Fig 5 (adapted from [3]) is a scheme of their experimental set-up. The calibration curves (see Fig 5) give widths of point image in x and y directions as a function of z. Referring actual image to the calibration curves gives a good estimation of z and thus 3-D image is reconstructed without scanning the sample. Fig 6 (adapted from [3]) shows that microtubules seemingly overlapping in x-y plane (false color coded in z,