Peter Kner at Histochemistry 2012
Peter delivered the first of the set of five QSTORM talks presented at Histochemistry 2012 at the Woods Hole Marine Biological Laboratory. He provides an introduction to the techniques of super resolution imaging and adaptive optics, and presents the team’s first sets of imaging data, before outlining next steps in the development of the QSTORM technique.
Karine and Carol Lynn teamed up to write this summary of the Research Update Presentation delivered by Peter Kner at the Athens, Georgia Meeting on December 11, 2011.
PETER KNER’S LAB: STORM MICROSCOPY WITH ADAPTIVE OPTICS
Peter’s team includes graduate students Andrew Herrington and Benjamin Thomas. (A highlight of our visit to UGA was touring the STORM imaging lab, and seeing Andrew demonstrate how the STORM system works.)
Peter began with an overview of “localization-based” microscopy – STORM, PALM, FPALM – all these methods are quite new to science, and first published in 2006. The key differences between them have to do with kind of fluorescent probes or markers they utilize, but they all are designed to get around the 250 nanometer (nm) diffraction limit of optical microscopy. Simply put, due to diffraction, we cannot get a focused, direct signal from multiple closely spaced (less than 250 nm) photon-emitting sources. For, how could we visually distinguish between two light sources less than half the width of the very wavelength of light? Furthermore, any point of light smaller in diameter than 250 nm will still appear to be at least 250 nm in diameter. To address this fundamental physical limitation, STORM, PALM, and FSTORM rely on “photo-switchable fluorophores” which are light emitting dye molecules that can be controlled so that only a small scattered fraction of them are “on” at any given time. Each one that is “on” and at least 250 nm from its nearest neighbor can then be accurately localized and plotted without interference. The point source of the light is assumed to be at the very center of the circular diffraction pattern it produces.
The microscopist collects the point source localization data rapidly and repeatedly for each of the signal-emitting fluorophores as they randomly switch on and off. “Rapidly and repeatedly” here means 10-50,000 images every 30 milliseconds. (Contrast this to film, which runs at 30 frames per second, and you understand how STORM might eventually be able to capture essentially video of molecular scale activity within living cells.) After capture, a computer program layers the many multiple images to create one composite image in which all the illuminated structures are rendered in “super” resolution. In this way, STORM tricks nature into yielding up images of intricate structures that- by all the laws of physics – should be secreted in haze too indistinct for our gaze.
The video, below, by Ricardo Henriques, provides a terrific analogy for STORM, applying a similar process of spot detection and image reconstruction to the blinking lights on a familiar structure.
While the microscopist wants only a very small fraction of the fluorophores switched on at any given time; when they are on, he or she wants them to be great and abundant photon-emitters; in other words, super bright. That way the signal of each one will be strong against the background noise of light and diffraction from other sources, particularly in the thicker, live tissue samples the biologists want to probe. The brighter the fluorophore, the higher the signal to noise ratio, and the more precise the point source localization.
To date, STORM microscopists have mostly used organic dyes as their light-emitting agents; the goal of the QSTORM team, however, is to substitute photo-switchable quantum dots (QDs) for the dye molecules. Quantum dots are small (10-20 nm) semiconductor crystals that, depending on their size, absorb particular wavelengths of light, get revved up, and then emit photons in another wavelength. QDs are brighter and longer lasting than dye molecules, and these characteristics should help to produce higher resolution (more precise) STORM images of dynamic molecular-scale structures within cells. But there are several challenges to using QDs for STORM; for instance, some of them are toxic to living cells; and, a reliable method of rapidly switching them on and off has yet to be discovered. That’s a key goal of QSTORM research. Jessica leads the QSTORM QD synthesis effort, so there will be more about progress in this area in her research update report from the 12/11/11 meeting.
The first task Peter’s group tackled was conducting “Drift Tests” with their instruments to determine the stability of their imaging system and particularly of a new stage they had acquired. They tracked the system’s localization of a fluorescent bead over a 40-minute period, finding deviations from 12 to 18 nanometers (nm).
At the meeting, Ge challenged Peter to try to get even better stabilization, since some of the structures the biologists want to image are themselves only a few dozen nanometers in size. At Carnegie Mellon, Ge reports they’ve been able to achieve plus or minus 1 nm stabilization, using a piezo-electric crystal stage on an air table. (See Ge’s 12/11/11 update for more details.) But Carnegie-Mellon’s lab is also built on Pittsburgh granite, so it may be a more stable lab environment overall.
Testing QSTORM with Various QD Species and Photoswitching Control Methods:
Next, Peter’s group tested a few recently published methods of controlling the photo switching of QDs. (Stefan Hell’s group in Germany developed the first two of these methods below.)
– QD “Blueing.” Published by Hoyer, et. al. (Nano Letters 2011). When QDs are exposed to oxygen, they shrink; so the wavelength of light required to activate them – as well as the wavelength they emit – also shrinks, toward the blue end of the spectrum. The microscopist exploits this transition by setting the microscope to detect a particular wavelength of light the QD will have to emit as it passes through its shrinking or “blueing” phase. Since the shrinking of the QDs happens at varying rates and times, the effect is similar to photo-switching – only a small subset of QDs are emitting photons at the targeted wavelength at any given instant. Hoyer diagrammed it as follows:
Peter’s group succeeded in duplicating this blueing technique using a slide with a smear of Invitrogen Cadmium-Selenium QDs on it; a drop of water provided the required oxygen. Below you can see the sequence of recorded point sources as bluing occurs, and below that, a “z-stack” compilation of all the recorded point sources as very faint red dots.
Blue/Red Laser Control. Published by Irvine, et. al. (Angew. Chem. Int. Ed. 14/2008). Manganese-doped ZnSe QDs (also known as Hell’s particles because Stefan Hell first synthesized them). These QDs are activated by a blue laser, and can be quenched with a red laser. Variable switching could be achieved with this technique through careful modulation of the red laser.
Micelle-coated QDs (MQDs) vs. plain QDs. This experiment was not about switching but simply about comparing the signal strength of regular Invitrogen CdSe QDs with micelle-coated versions of small groups of the same QDs. Jessica’s lab synthesized the micelle-coated QDs (MQDs) by building a protective polymer coating around small groups of Invitrogen’s non-doped CdSe QDs. The MQDs emitted significantly stronger signals and kept fluorescing long after the non-coated QDs photobleached out at about 20 minutes. The micelle coating increases QD stability, and the grouping increases the intensity of the light they emit (see data below). However, the increased protection also renders the QDs less susceptible to blueing, so that particular modulation technique did not work well in initial tests.
Testing D-STORM with Conventional Dyes
Peter’s lab also explored STORM imaging using traditional organic fluorescent dye molecules. In the past, the on/off switching necessary for STORM was achieved using two dyes joined together –which were activated by two different wavelengths of light and quenched each other. But recently, some microscopists have started to use a technique called DSTORM, or Direct-STORM, in which just one commercially available (Invitrogen Alexa) dye is excited to the point that some of the dye molecules – rather than emitting photons and then relaxing back down to the initial energy state – instead go into a “triplet” state where they rest awhile without emitting photons, before relaxing back to the initial state. This is another way of producing a situation in which only a small number of the dye molecules are emitting photons at any given moment.
Imaging Fish With Alexa 488
The Brainerd lab sent live zebrafish embryos to Athens with the heads of myosin muscle fiber molecules labeled with Alexa 488. [Check out the set of blog posts from July 19-July 21, 2011, “The fish arrived… Where are they?” wherein the biologists and the physicist/engineers encountered their first culture clash.] The thickness of the fish embryos caused problems for the microscopists. There was a lot of “background noise” from photon emissions from the dye in areas above and below any particular focal plane they chose in the “z” dimension (height). You can see this in the video below, which shows a “z-stack,” a gradual progression of the microscopes plane of focus through the vertical thickness of the fish specimen.
Next, they took a thinner slice of the specimen and were able to begin localizing and recording the centers of all the individual dye molecules as they switched on and off.
By accumulating all the data points in one image frame, they were then able to generate a reconstructed STORM image which begins to reveal the extraordinary level of detail STORM imaging can provide. The tiny dots reveal the anatomical patterns of myosin protein heads in the fixed muscle tissue.
Then Peter’s team applied a “structured illumination” or SIM technique that removes background light. This resulted in a more refined set of z-stacked images, as below:
Imaging Tau Protein in Rat PC-12 Cells
Peter also received sample from Ge Yang’s lab at Carnegie Mellon. Ge, as you will recall, would like to use QSTORM to study the anatomy and dynamics of “cargo transport” in nerve cell axons. Ge sent Peter a thin sample of a rat PC-12 medulla cell that had been induced to grow a “false axon” containing tau proteins – the core subject of his research – which were labeled with Alexa 488. From the initial layered recording of distinct localized points of flourescence,
Peter was able to produce a highly detailed STORM image of the Tau proteins along the false axon of the PC-12 cell.
Can’t wait to try this with QDs!
Peter’s group also did a quick comparison of photon counts per fluorophore, comparing the Alexa 488 dye – used in the rat PC-12 nerve cell and in the fish muscle cells – alone on slides – not within biological samples. Even with less than optimal conditions for the QDs, their photon emissions soared above those of the dye. So everyone on the team is excited about using QDs in place of dye. The biologists however, are still learning how to get the QDs successfully implanted in their samples. (See more on those efforts in Beth and Ge’s 12/11/11 update reports)
Applying Adaptive Optics (AO)
Peter’s team will also attempt to pioneer the application of Adaptive Optics (AO) techniques to imaging biological samples, in order to reduce wavefront distortion produced when light from the fluorophores passes through other layers of the living tissue of the biological sample. Originally developed to help astronomers correct aberrations produced by the Earth’s atmosphere, an AO system senses incoming wave distortions and compensates for them with mirrors that can be precisely deformed in patterns that correct precisely for the pattern of incoming wavefront distortions. To show the difference AO can make, Ben imaged a fluorescent bead through a c.elegans worm, first without using AO and then using AO. The increased focus and clarity is evident in the figure on the right.
Using Astigmatism to Infer Depth in STORM Imaging
If one can correct aberrations due to wavefront distortion, one can also induce them, and in Peter’s lab, Benjamin Thomas will be experimenting with doing just that – in order to increase STORM 3D imaging capability. By inducing astigmatism – in which one of the x-y-z planes is more in focus than another – the glow from a source of light will appear more oval in shape. By comparing the bias, the investigator could then calculate the relative depth and angle of the source. This technique was pioneered by X. Zhuang’s lab at Harvard (B. Huang, et. al., Science 2008). Ben will build on that prior work.
Stability to about 15 nm was achieved with the lab’s new STORM imaging set-up. Tests were conducted with Invitrogen QD’s, Manganese-doped ZnSe QDs, micelle-coated QDs, and two switching methods (bluing and red/blue laser modulation). DSTORM imaging with a single conventional fluorescent dye (Alexa 488) was demonstrated in fixed biological samples, including zebrafish and rat PC-12 cells. Correction of wavefront distortion using adaptive optics in biological tissues was also demonstrated. In the past six months, Peter, Ben, and Andrew have gained valuable experience fine-tuning a new STORM imaging system, testing various light sources and light modulation methods, and working with a variety of biological samples.
Overall engineering goals and methods:
• Demonstrate the utility of quantum dots to STORM. Do a side-by-side comparison of STORM imaging in zebrafish with Alexa dyes vs. QDs. Also test if micelle-coated QDs increase effectiveness.
• Demonstrate the utility of adaptive optics to STORM, showing that high resolution can be obtained in thicker (and eventually living) samples. Do a side-by-side comparison of STORM imaging at a depth of 20-50 microns, with and without AO.
• Obtain better comparative STORM data on bare QDs vs. micelle-coated QDs for both the Mn-doped QDs and the Invitrogen CdSe QDs, and publish a joint paper with the Winter group.
• Conduct further research on “blueing” and the feasibility of its use for QSTORM.
• Try 3D STORM imaging at depths of 20-50 nm, correcting aberration with AO and induced astigmatism. (Probably on rabbit soas muscle from Beth).
• Tackle imaging in thicker samples (like live zebrafish), using AO to demonstrate that high resolution can be achieved in thicker samples. (Approval of protocol from local vet.)
• Compare STORM imaging in fish with Alexa dye vs. QDs vs. MQDs.
• Obtain better data on axonal transport and publish joint paper with Ge.
• Improve mounting medium to discourage photobleaching.