Open-source Single-particle Analysis for Super-resolution Microscopy with VirusMapper

The overall goal of this procedure is to produce high precision molecular models of viruses or macro-molecular complexes by applying single-particle analysis to super-resolution images of structures with fluorescently labeled components. This method can answer key questions in the field of virology, including the protein architecture of complex viruses, and how this changes during the course of infection. The main advantage of this technique is that by collecting multiple images of the same structure it becomes possible to generate a high-precision map of its components.

Though this method can provide insight into the structure of viruses, it can also be applied to other systems such as other pathogens and macromolecular complexes within mammalian cells. First, image the sample with super-resolution fluorescence microscopy. Acquire images of several fields of view containing hundreds to thousands of well-separated particles with no unwanted fluorescent structures.

Once the images have been obtained and processed, import and concatenate the images into a stack with intercalated channels. If necessary, convert the concatenated image from a hyper-stack to a stack. Next, select Extract Viral Structures"in the VirusMapper sub-menu and set the file path for the extracted particles.

Fill in the number of fluorescence channels imaged. Set the reference channel to the fluorescence channel in which the particles have the most consistent appearance. If those particles lack a central maximum, apply pre-detection Gaussian blurring to induce the appearance of one.

Next, estimate the diameter in pixels of the largest particles. Set the ROI radius to slightly more than half of that value. Estimate the number of ROIs needed per frame.

Use no more than one hundred ROIs initially. Set the maximum ROI overlap based on the particle separation. Preview the ROIs for that frame.

Adjust the radius, the number of ROIs, and the maximum overlap so each particle in the frame is enclosed in a single ROI. Ensure that the ROIs are at least a few pixels wider than the largest particles. Then, click OK"to run the segmentation.

Close the sample image in the ROI manager after extraction. Do not rename the extracted particle sets. Select Generate Seeds"and open the folder containing the extracted particle data.

Set the reference channel and select all channels for which a seed should be generated. If needed to match previous models, rotate the seeds by ninety degrees. For channels in which the particles lack a central maximum, increase the pre-alignment Gaussian blur value as before.

If the channels are not closely aligned, enable shift correction for the non-reference channels. Search the particle sequence for a consistently appearing structure. Identify a representative particle and enter the corresponding frame number in the Frames to use"field.

Review the resultant seeds. Seed selection is a critical stage of this procedure. View the raw data carefully to identify one or multiple structures to be modeled.

Model quality is highly dependent on choosing seeds which accurately reflect these structures. Adjust the reference channel, the Gaussian blur radius, and the shift correction as needed to optimize identification of additional frames with similar seeds. Continue adding frames and adjusting the generation parameters until the average structure best represents an observed structure in the data.

Enter the folder name and file prefix for the seeds. Click OK"to save the seed images for later modeling. Select Generate Models Based On Seeds"and open the extracted particle folder.

Load the seed averages for each channel. If doing reference-based structure discovery, select a reference channel for alignment. Known particle structure in one channel can be used as a reference to align a second unknown channel.

This allows for unbiased mapping of the unknown structure. Keep in mind that chromatic shift between channels must be corrected before-hand. If the analysis is looking for small differences or subtle features in the model, select Square image intensity during template matching.

Set the minimum similarity to between sixty and eighty percent, and the number of iterations to one. Select models and particles to be shown during the calculation, and generate the preview models. Inspect the preview models.

Increase the minimum similarity to include only particles with the desired morphology. Optimize other model calculation parameters as needed and select additional elements at the model generation process to be shown if desired. Once the preview models are satisfactory, increase the number of iterations to ten.

Set the folder name and file prefix. Click OK"to save the model evolution stacks containing all iterations of the final model. A recombinant Vaccinia virus with two proteins tagged with green and red fluorescent proteins was imaged by structured illumination microscopy and modeled with the VirusMapper plug-in.

Seeds were generated separately for the frontal and sagittal orientations, each averaged from five representative particles. Depending on the orientation of the virus, one or two lateral bodies can be distinguished. Therefore, separate models can be generated for the two orientations.

While attempting this procedure, it's important to remember that the quality of the models largely depends on the quality of the raw data acquired. While single particle analysis can increase precision, it cannot compensate for low quality images. Following this procedure, further quantification or model fitting can be done on these models.

This can help answer additional questions specifically around for example nano scale changes on the viral architecture during infection. After watching this video, you should have a good understanding of how to use single-particle analysis software VirusMapper to generate models of molecular architecture from super-resolution images.