How to Set Up a 3D-SIM Microscope (User Manual)
How 3D-SIM Super-Resolution Can Be Implemented in a Facility
- How to Set Up a 3D-SIM Microscope (User Manual) - How 3D-SIM Super-Resolution Can Be Implemented in a Facility
- Fig. 1: Influence of the oil. HeLa cells were stained for microtubles (α-Tubulin, red), and F-actin (Phalloidin, green). On the top row, the Phalloidin staining is sharper when the 1.514 oil is used, whereas on the bottom row, the α-Tubulin staining presents the best result with the 1.516 oil. Bar: 5µm.
- Fig. 2: Influence of the coverslip. HeLa cells were prepared as in figure 1. The side view of the stack taken for the 2 types of coverslips is shown. The VWR coverslip stack (a) presents more bleaching that the Zeiss coverslip one (b). Bar: 5 µm.
- Fig. 3: Influence of the structure´s shape. Sections from zebrafish retina were stained for double cones (zpr1, red) and glutamate receptors (mGluR6, green) and were imaged using the wide-field or 3D-SIM modes of the OMX. The 3D-SIM image (bottom) shows artefact that are absent in the deconvolved wide-field image (top). Bar: 5 µm.
- Fig. 4: Influence of the algorithm parameters. HeLa cells were stained for DAPI, and the raw data from the 3D-SIM was reconstructed with the Wiener parameter 0.0010 (a) or 0.0060 (b). Zebrafish retina was stained as in figure 3. The data was recontructed with the Wiener parameter 0.0010 (c) or 0.0070 (d). In both cases, a higher Wiener constant improves the image quality. Bar: 5 µm.
- Fig. 5: 3D-SIM live cell imaging. HeLa cells were stained with MitoTracker Red and imaged in 3D-SIM every 3 minutes for 6 z-sections. After 6 timepoints, the stainnig is nearly completely bleached. Bar: 5 µm.
In the past two decades, light microscopy has seen a tremendous improvement with super-resolution techniques. Many of the super-resolution microscopes (3D-SIM, STED, and PALM/STORM) are now available in a commercial solution, and are entering labs and facilities worldwide. This offers an important step forward in the field of research in biology, but when not used in optimal conditions, those powerful techniques can give rise to artefacts. Here we focus on parameters that can deeply influence the image quality for 3D-SIM and the next challenges for this technique.
A new generation of highly sophisticated light microscopes allows various deeper insights into living organisms. In recent years, super-resolution microscopes have been used to examine details which were previously only accessible to the electron microscope. Using three-dimensional structured illumination microscopy (3D-SIM), the resolution of light microscopy can be doubled (up to 100 nm), which means that even the smallest cell structures can be seen „live" with the light microscope [1, 2].
At the University of Basel, we recently installed an OMX imaging system, a state-of-the-art 3D-SIM microscope for super-resolution (Applied Precision/GE). This microscope instrument can be used for doubling resolution in both fixed and live cell imaging. Due to the specificity of the system, we have been testing the influence of different parameters, such as the coverslip thickness precision, the temperature, the oil used for imaging, and the influence of noise/background. We also tested our system for different applications, ranging from cell biology, cancer biology to developmental biology. We had various success depending on the structure observed, and invested some time to resolve the issues.
We will name here the elements to take into account when setting-up an OMX in a facility, the parameters that can deeply influence the quality of image acquisition, and how to solve the issues that can be encountered. Finally, we will discuss the perspectives that the OMX system can bring in the near future.
As for any light microscopy technique, the rules of basic optics have to be followed.
But in super-resolution techniques, even minor changes to those best practice rules will be of major effect. Spherical aberrations, and/or bleaching due to less stable fluorophores will result in bad quality images or artefactual results. As the system is based on a wide-field platform, the material to image needs to be at most at 16 µm from the coverslip. Otherwise, the spherical aberrations are so immense that the results are full of artefacts, and a standard confocal (with or without deconvolution) will produce better results. The sample itself, in its preparation, needs full attention. Apart from the optimal 0.17 mm coverslip that should always be used in light microscopy, users need to optimize their fixation to have a high signal to noise ratio (SNR). To further increase the SNR, primary antibodies giving strong signal to the structure of interest and photostable secondary antibodies must be used. Without a good SNR, again, the resulting images will be of bad quality. In the design of the experiment for multiple stainings, the fluorophores must be with close wavelengths to avoid artefacts. Better results should be obtained with 488 and 568 fluorophores, than with 488 and 647 for example.
Influence of the Oil, Temperature, Fluorophores and Coverslip
We tested different conditions for the acquisition of OMX images. We first examined the influence of the immersion oil on the imaging. HeLa cells were fixed in 3.7% PFA and stained for microtubules and for F-actin. The sample was mounted with Vectashield H-1000 to best match the refractive index of the immersion oil. The temperature of the room was measured at 23°C. Oils ranging from 1.512 to 1.518 were used (fig. 1). As soon as the oil is not matching, "ghosts" of the structures stained appear, as it can be observed in panel d (Top row, fig. 1). The stainings can also appear fuzzy or more diffuse, as shown in panels a and e for Phalloidin or α-Tubulin, respectively (fig. 1). We also tested different temperatures. Performing the same procedure as described previously, but at 25°C, we determined that the 1.516 is the oil of choice for the 488 channel, and the 1.518 is best suited for the 568 channel (data not shown). As a consequence, the room temperature should not vary by more than one degree. With a variation of +/- two degrees, the oil must be changed for imaging.
With samples having more than one fluorophore, oils can be mixed to obtain the best match. At 23°C for a combination 488 and 568, the 1.515 oil could be used. From our different tests, it seems that the 488 channel is more susceptible to show artefacts compared to the 568 or the 647 channels, so in that same example, the 1.514 oil could also give good results for the 568 channel, even if not fully optimal (see figure 1 panel f). Finally, to get the best OMX-SIM reconstruction, it is recommended to image three channels maximum and the oil has to be chosen on the channel of most interest.
Therefore, we show that the immersion oil must match the room temperature and has to be adapted with the fluorophores used in the sample.
We also performed a comparison of standard VWR #1.5 coverslips and Zeiss #1.5 Precision coverslips (fig. 2). Both coverslips give good results in terms of XY resolution. With both types we observed a similar intensity over the z sections for stacks under 3 μm. However, in stacks over 3 μm, we could observe more photobleaching with the VWR coverslip than with the Zeiss Precision coverslip samples (fig. 2). We therefore recommend for bleaching prone samples to use the Zeiss #1.5 Precision coverslips.
Influence of the Structure's Shape
From the design of its algorithm, the OMX works well for the imaging of fibers and small defined objects [3, 4]. However, for more diffuse stainings, the reconstruction step can create artefacts. For example, sections from zebrafish retina were stained for double cones and glutamate receptors, and imaged (fig. 3). The double cone signal is showing an artefactual honeycomb structure (bottom row) that is not present in the deconvolved wide-field image (top row). In this particular case, using different oils did not resolve the artefact problem. Even for cell biology samples, DAPI staining can present "halo" artefacts, albeit the staining has a high signal to noise ratio. In this particular case, the algorithm can be altered, by changing the value of the Wiener filter constant for example. The Wiener filter is intended to filter out the noise of an image, and the use of Wiener 0.060 instead of 0.010 clearly improve the quality of the DAPI image (fig. 4 a and b). The same algorithm alteration can be applied for the retina sample mentioned previously, and also improve the image quality (fig. 4 c and d). However, homogenous stainings will never give as good results as small defined objects. We also noticed that the region selected for the reconstruction can change the result of the OMX-SIM image. In the particular case of Salmonella infected tissue imaging, it is tempting to acquire larger zones of tissue. However, the algorithm will compute the image differently depending on the dynamic range of the image. With different intensities in different zones, the result is altered to take into account the differences. As this can also create artefacts, we advise to restrict the zone of interest to the minimal zone required (unpublished results). It is therefore critical to optimize the reconstruction algorithm and to carefully select the region to be imaged.
Next Challenges for the OMX: Live Imaging, 3D-SIM-CLEM and PALM
As early as 2010, fast 3D-SIM live imaging was performed with "customized" versions of the OMX [5-7]. Now, the commercial version OMX Blaze is fast enough to perform live experiments (frame rate of 100 fps in 3D-SIM) and offers the possibility for live imaging to a wider community. However, the need of at least 6 z-sections and the 15 images per z-section increases the photobleaching and the difficulty to reach a good SNR. When using HeLa cells stained with MitoTracker Red, we could observe a nearly complete bleaching after 6 timepoints (fig. 5). We are still testing the best conditions to avoid bleaching and artefacts. Results could be improved with anti-oxydant in the media of the cells (Ascorbic acid or Trolox) and media that preserve the fluorescence (such as DMEM GFP media) (personal communication from the OMX GE specialists).
More projects go towards combining 3D-SIM and EM. A first paper of 3D-SIM-CLEM was published recently in plants . This new application of 3D-SIM-CLEM will offer amazing new opportunities. However, the setup of the experiment is still challenging for the sample preparation and imaging. It is usually from a transmitted light image that the user can find the zone to analyze under the EM microscope. For this application, the absence of oculars is very unpractical; it is indeed rather hard to set up a proper Köhler illumination, which would help to get a good transmitted light image and relocalization of the sample zone in EM.
The OMX-Blaze will soon benefit from a new module implementation, called MONET (pointillism). This new feature will transform the OMX system into a multi-function platform, where 3D-SIM, TIRF, and pointillism techniques will be possible on the same system. This new feature has not been released yet, and we are looking forward to this implementation in the near future.
A year after the installation of our new OMX Blaze system, we are very positive about the results it produced. We have already two papers published that include super-resolution work, and at least another 10 ongoing projects. In terms of resources in a facility, we find it more productive to have specific facility staff in charge of the OMX and doing all the imaging for the users, rather than having all the users autonomous on the system. As described in this article, there are many different aspects to be aware of, and we believe that a strong experience on the system helps for a better use of it. We are enthusiastic about the future directions that the OMX Blaze will allow us to follow, and are looking forward to testing them.
The acquisition of the OMX was supported by the University of Basel and the Schweizerische Nationalfonds (SNF). We thank M. Haug and S. Neuhauss (IMLS, University of Zurich) for providing us with zebrafish samples. We also thank the GE team for their ongoing support.
 Gustafsson M.: Journal of Microscopy 198, 82 (2000)
 Schermelleh L. et al.: Science (New York, N.Y.), 320, 1332 (2008)
 Schermelleh L. et al.: Science 320, 1332 (2008)
 Gustafsson M. et al.: Biophysical Journal 94, 4957 (2008)
 Carlton P. et al.: Proceedings of the National Academy of Sciences of the United States of America 107, 16016 (2010)
 Shao L. et al.: Nature Methods 8, 1044 (2011)
 Fiolka R. et al.: Proceedings of the National Academy of Sciences of the United States of America 109, 5311 (2012)
 Bell K. et al.: Plant Physiology 161, 1595 (2013)
Dr. Alexia Ferrand (corresponding author via e-mail request button below)
Dr. Oliver Biehlmaier
Imaging Core Facility
University of Basel