to BioTechniques free email alert service to receive content updates.
Correlative light microscopy for high-content screening
Benjamin Flottmann*1, Manuel Gunkel*1, Tautvydas Lisauskas*1, Mike Heilemann1,2, Vytaute Starkuviene1, Jürgen Reymann1, and Holger Erfle1
Full Text (PDF)
  • Identify target cells for confocal imaging manually / automatically

  • Image reference markers on confocal setup

  • Perform coordinate transformation of cell positions

  • Load position list and image cells automatically on confocal setup

  • Identify targets for dSTORM acquisition

  • Image reference markers on dSTORM setup

  • Perform coordinate transformation of cell positions

  • Load position list and image cells on dSTORM setup

  • Results and discussion

    To achieve sequential correlative imaging on the same sample in three microscopy modes without additional sample preparation steps in between, we first tested various fluorophores for their suitability. Here, we demanded both high photostability and, in particular, the possibility of inducing photoswitching such that single-molecule super-resolution imaging, following the dSTORM protocol (11), was possible. We identified Alexa Fluor 532 and Alexa Fluor 647 as an optimal pair for dual-color microscopy. Both fluorophores showed excellent photostability and were thus well suited for widefield and confocal imaging, as well as super-resolution imaging.

    We set up a workflow as described in Materials and Methods to target structures of interest within low-resolution wide-field images, determine their spatial coordinates and relocate their positions on multiple other microscope systems. This workflow was applied here in the following way: After widefield imaging, we first transferred the sample to a confocal setup and then to a single-molecule localization microscope and imaged in these systems the same cells of interest identified in the widefield images. In widefield images at 10× magnification, cells of interest that showed the expected phenotype were selected and their positions (coordinates) were determined. The sample was then transferred to a confocal microscope. Using the coordinates, a confocal 3-D stack was acquired for each cell. The precision at which cells of interest could be re-localized was in the range of 10 μm, which was sufficient to position the whole cell of interest within the field of view of ~50 μm. The sample was then passed on to the single-molecule localization microscope where selected cells of interest based on the confocal images were acquired. The re-localization of cells of interest here was slightly worse, since the reference structures could not be fitted within the field of view, and their center positions had to be estimated. However and more importantly, the cells of interest could easily be identified. Images acquired during this workflow are depicted in Figure 1 and illustrate the increase of resolution and information content achieved by this correlative approach. Imaging parameters for the different microscope techniques used are given in Table 1.

    Table 1. 

    Co-localization analysis in confocal and high-resolution imaging

    We then analyzed the capability of our correlative microscopy approach to compare the distribution of the cis- and trans-Golgi markers in NRK cells under diverse conditions (16, 17). Therefore, we imaged the cell samples using the workflow described above.

    Our analysis proceeded as followed: Figure 2 shows the increased resolution of dSTORM (E) compared with confocal imaging (Fig 2A, B). An intensity line profile illustrates this increase in more detail. The spatial distribution was further analyzed to evaluate the co-localization of the cis- and trans-Golgi markers. The increased spatial resolution now enables a more refined analysis of whether two proteins co-localize. To quantify co-localization, we correlated both imaging channels pixel-wise and calculated the Pearson coefficient (intensity correlation analysis, ICA) (18). Positive values report a co-localization of two proteins within a particular pixel. For the confocal image (Figure 2B), ICA (Figure 2C) 11.4% of pixels showed co-localization (corresponding to yellow pixels in Figure 2B). The same analysis on the super-resolution image of the same region (Figure 2E, F) resulted in only 4.5% of pixels showing co-localization. As a consequence, the dimensions of regions in which both proteins co-localize are overestimated by confocal microscopy, as a result of the poorer spatial resolution. Super-resolution microscopy provides a more realistic picture of the co-localization of both proteins. In general, co-localization analysis of confocal images over-estimates both the size of co-localization sites and the size of co-localizing structures, even when analyzing discrete structures.

      1    2    3    4