In this new series "From Image to Results", explore various case studies explaining how to reach results from your demanding samples and acquired images in an efficient way. For each case study, we highlight different samples, imaging systems, and research questions.
We start the series with an analysis of vesicle trafficking in Cos7 cells. The goal is to generate a high-resolution longitudinal study of vesicle trafficking within a single cell to explore some basic observations concerning vesicle transport. We then showcase image analysis tools to segment vesicles and generate vesicle tracks to derive meaningful data to confirm our observations.
|Case Study Overview|
Cos7 cells transiently transfected with mEmerald-Rab5a and Golgi7-tdTomato.
Track both types of vesicles, volume & distances over time, colocalization.
Graphic visualization of tracks and cells in a movie and resulting data as time line along track, scatter plots.
ZEISS Lattice Lightsheet 7
For this case study, we selected a simple, yet analytically demanding Cos7 cell model that has been tagged with two different markers for vesicular transport: tdTomato-Golgi-7 and mEmerald-Rab5a-7.
The tdTomato-Golgi-7 fluorescent protein encodes the signal sequence of the protein Beta-1,4-Galactosyltransferase 1 (B4GALT1), which functions in the glycosylation of trans-membrane proteins and localizes mainly to the Golgi Apparatus and the cell membrane (Figure A).
The mEmerald-Rab5a-7 fluorescent protein encodes parts of the Ras GTPase RAB5A. This protein is involved in vesicle trafficking and localizes mainly to endosomes and to the cytosol (Figure B).
The raw data for this use case was captured with ZEISS Lattice Lightsheet 7. This system is specifically tailored for fast, volumetric live cell imaging with near-isotropic resolution, making it the perfect instrument for imaging subcellular, dynamic structures such as vesicles in 3D over time.
Using lattice light sheet microscopy, the sample is imaged at an angle, due to the geometry inherent to this technology. As a consequence, the raw data needs to be processed (deskewed) before it can be further analyzed. For this use case, 200 time points were acquired in 2 channels, one volume (85x80x20 µm³) every 3.2s, 205 planes per volume and a total of 41.000 frames in just under 11 minutes.
After acquisition, the raw data was deconvolved, deskewed and cover glass transformed using ZEN (blue edition), resulting in a processed .czi file with 0.145 µm pixel size in x, y and z. The processed .czi file could then be directly imported into arivis Vision4D and converted into arivis' .sis file format, a format specifically designed for fast handling of large data sets, for further analysis.
In order to define a suitable image analysis strategy, it is important to look at the raw images (Figure C).
In this single time-point 3D representation, some imaging artifacts are visible as stripes in the green channel. As a result, some image pre-processing is necessary to remove these artifacts. Vision4D has two tools; “Morphology filter” and “Denoising”, which can remove these artifacts by enhancing sphere-like structures of an approximate diameter (here: ~1.5 µm) and suppressing any other signals.
Also, the images show that Golgi7-labelled vesicles strongly co-localize with a fraction of the Rab5a-labelled vesicles. In fact, careful inspection shows that there are no vesicles in the data set that are Golgi7-positive and Rab5a-negative. Hence, the dataset consists of two vesicle species: “endosomal” vesicles that are Rab5a-positive/Golgi7-negative and “Golgi-associated” vesicles that are Rab5a-positive/Golgi7-positive. The image analysis strategy needs to account for these two species by applying differential segmentation.
Figure C: Single time-point 3D view of data set. Single channels and merge image are shown. Note the (complete) overlap of Golgi7-labelled vesicles with a fraction of the Rab5a-labelled vesicles. Also note “striping artefacts” in the green channel.
In order to reduce the data size of the original raw dataset (2 channels; 150 slices; 200 time points; 16-bit; ~ 40 GB), the images were cropped to include only the relevant cellular areas. These cropped images were then used to generate a stack subset containing only every second slice and the data were then transformed to 8-bit. The final data set had 2 channels, 22 slices and 200 time points, it was 8-bit and was ~ 1.2 GB. in size.
The image analysis in Vision4D consisted of three steps. Firstly, image processing to denoise the images was performed, with special emphasis on preserving spherical structures and removing striping artefacts. This was done using the image processing functions “Morphology Filter” and “Denoising” independently for both channels.
Secondly, vesicles in both channels were independently segmented using Watershed segmentation with minimum vesicle volume of 0.3 µm³. The Rab5a-positive vesicles were then filtered based on the “minimal” distance to the next Golgi7-positive vesicle and assigned to the group of Rab5a-positive/Golgi7-negative “endosomal” vesicles (if distance > 0). Rab5a-positive/Golgi-positive objects were assigned to “Golgi-associated” vesicles.
Finally, to derive the movement of vesicles over time, object tracking was performed with both vesicle types.
The sketch summarizes the image analysis procedure. In addition, the Vision4D pipeline to perform these operations is available for download at the bottom of this page as part of the case study data package ("Green-Magenta vesicle detection_Tracking_Rev4.xml”).
For validation of tracking, selected single vesicle tracks were examined to ensure they correctly represented the vesicle movement. In the lower panel, three tracks are shown as video sequences.
The overall distribution of vesicles helps in deducing the overall cellular architecture of the sample (Figure D). The area of high-density Golgi-associated vesicles presumably represents the main Golgi apparatus region of the cell. Next to it is a large “vesicle-free” region that represents the nuclear area. The three-dimensional contours of the nucleus can also be deduced by the spherical volume void of vesicle tracks in a 3D representation (Figure F).
A further validation is the option to visualize the cell as a “scatter plot” based on the objects’ x-y coordinates (Figure E). This shows again an accumulation of vesicles (and tracks) at the Golgi and an exclusion of vesicles from the nuclear area. This plot type is also a valuable tool for visualization of certain observations.
All other cellular regions may be assigned to “Periphery”. Assigning these areas aids a more precise description for the subsequent vesicle and track observations.
Figure F: 3D movie with visualized vesicle tracks; note the nuclear sphere in the side view.
Having defined “endosomal” and “Golgi-associated” compartments during image analysis it is now possible to make observations and prove them with measurements and suitable visualizations.
Vesicular compartments mainly differ in size and cellular localization. The ~ 26000 identified endosomal vesicles have an average volume of 0.63 µm³, while the ~ 22000 identified Golgi-associated vesicles have an average volume of 1.062 µm³. The volume distributions are also shown in Figure G. Endosomal vesicles are devoid in the Golgi area and evenly distributed over the cell periphery (Figure H). Golgi-associated vesicles accumulate more in the central “Golgi area” of the cell (Figure I). By color-coding the plot with vesicle volumes, it becomes apparent that large vesicles are primarily located in the Golgi area. This is further illustrated by Figure J, that compares vesicle sizes within the central Golgi region and in the cell periphery.
The next goal was to determine the density of vesicles throughout the cell. Vesicle density takes into account both vesicle volumes and the accumulation of vesicles in a certain region of a cell, and is hence related to these two features.
There is no direct way of plotting object densities in Vision4D. Vesicle positions and volumes were therefore exported via spread sheet and calculated density maps weighted for vesicle volumes were generated using Python (Figure K-M).
The results confirm the observations made for vesicle sizes and distribution: The Golgi area is the vesicle center of the cell. The vesicle density here is more than 10 times higher than in the cell periphery (Figure K). This enhanced density is attributed almost exclusively to Golgi-associated (Golgi7-positive) vesicles (compared Figure L and M).
Vesicle trafficking is a process of directed vesicle transport along cytoskeletal filaments (e.g. microtubules). This process generates cellular zones of strong vesicle motion. To analyze vesicle motion in this data set, the tracks from endosome vesicles (1408 tracks) and Golgi-associated vesicles (551 tracks) were used to determine the track speed, which is the average movement of vesicles from one frame to another.
Endosomal vesicles displayed an average speed of 0.206 nm/s and therefore were significantly faster than Golgi-associated vesicles with 0.125 nm/s (Figure N). When analyzing the distribution in the cell, vesicle motion was found to be fast in the periphery, more specifically in a zone around the nucleus and the cell’s protrusions (Figures O-Q). In contrast, vesicle motion in the central Golgi area was 2-3 times slower. The fast vesicle movement in the periphery was mainly attributable to endosomal vesicles.
Vesicle track speed, as determined in the previous section, does not take into account the overall direction of vesicle movement. Hence, a vesicle that ends up at its original starting point, might still have a high track speed. In order to consider directed motion, the Mean Square Displacement (MSD) of the vesicle tracks was determined, which measured the total distance from a track’s starting point.
Endosomal vesicles and Golgi-associated vesicles displayed an average MSD of 1.994 µm² and 0.721 µm², respectively (Figure R). As for track speed, vesicles with high MSDs were located in the periphery, more specifically, in a zone around the nucleus and towards the cell’s protrusions (Figures S-U). Again, endosomal vesicle tracks were mainly responsible for high MSDs.
Both parameters track speed and MSD allows evaluation of vesicle velocity and directed motion. However, these evaluations do not display active transport, as they cannot take into account the track directions. Visual inspection of the dataset suggests, that such active transport takes place in the zones of fast vesicle motions around the nucleus and towards the cell protrusions. To illustrate that vesicles indeed have a preferred motion direction it is possible to use track displacement vectors of vesicles and tracks.
Vision4D offers limited options for visualizing vectors. As a result, the relevant vesicle positions were exported together with their parent tracks and data science tools in Python were used to obtain different visualizations of track directions (Figure W-Y).
All visualizations show that there are main motion directions in different parts of the cell. Vesicles move across a ring zone around the nucleus as well as towards the cellular protrusions at the upper end of the cell.
This case study has showcased how to approach the analysis of a complex imaging data set employing the Vision4D software. Vision4D provided tailor-made image processing and segmentation to derive and track the biological objects that represent the full complexity of the sample. After definition of the objects, visualization or plotting important object features for both vesicles and tracks was possible from within Vision4D. For features not accounted for by Vision4D, the necessary data were extracted and analysis was continued using customized analysis with Python.
The biological results obtained for the two vesicle compartments fit nicely with what is expected from their biological background. Golgi-associated vesicles are packed densely in a zone next to the nucleus that appears to form the Golgi apparatus of the cell. These are large and move relatively slowly. On the other hand, endosomal vesicles, responsible for directed transport throughout the cell, appear to be more distributed throughout the cell periphery and are devoid from the Golgi zone. They are smaller and move faster and in a more directed way than Golgi-associated vesicles, especially in a specific peripheral zone around the nucleus.
Of note, the intention of this use case was not to perform a thorough analysis, but to provide an overview of the analysis strategy. Other important observations may be buried within the data set, e.g., the measurement of “active vesicle transport” by means of analyzing the MSD over time, or vesicle fusion and budding by means of a more complex tracking strategy. Vision4D has the tools to find answers to these questions also. This dataset is available for you to try these analysis approaches for yourself in arivis Vision4D so you can find out more about these observations.