Automated Holotomographic Microscopy
problematic [12–14]. Nuclear probes oſten interfere with DNA, which perturbs fundamental processes like replication and transcription [15] of the genetic material. Moreover, while being excited, nuclear compounds generate a massive photo- toxic stress [16] where it hurts the most: in the genome [17–19]. Most of today’s image-based investigations require the
segmentation of thousands of entire cells [16,20]. Chemical compounds like CellTrace VioletTM
, cholera toxin B, or other
similar “cell outliners” [21] can provide a cytosolic fluorescent stain that provides a signal with enough intensity to be differ- entiated from the background. However, due to the heteroge- neous nature of the material they stain (the cell’s dry mass), their signal may show spatial and temporal variations mak- ing simple thresholding a challenge. In addition, the complex curvature of cell edges makes it difficult to distinguish the boundaries of touching cells, resulting in under-segmentation problems such as two or more cells being recognized as one, which presents a more complex problem, similar to the chal- lenge posed by touching nuclei [22]. Tis explains why, in the case of entire cells, an object
segmentation strategy, like the one applied for a simple signal generated by a stained nucleus, fails. Te solution for segment- ing an entire fluorescent cell is to use primary nuclei detec- tion to anchor (or seed) a secondary step for the detection of a cell [9]. Such a secondary step relies on the propagation of the segmentation of the primary object until the propagated object touches another propagated object, or until a local drop in signal signifies the limit of the object [23]. When coupled to local thresholding methods, this approach can segment very fine details, making propagation strategies the current gold standard for full cell detection [23]. However, segmentation finesse comes at a cost. Firstly, it
requires more work from the user when using cell segmenta- tion soſtware [24]. Secondly, and more importantly, staining cells generates major perturbations of the living sample [25] and increases the risk of chemically perturbing the observed processes while generating both primary and secondary pho- totoxicity [17,18]. To circumvent these problems, we have developed a new
segmentation technique that relies solely on the RI signal of cells, using proprietary AI for signal preparation together with advanced thresholding methods. EA can detect the fine details of cellular objects that display complex subcellular details without using stains. 3T3-derived pre-adipocytes were imaged for 2 hours at a
frequency of 1 image every 4 minutes using HTM and epifluo- rescence simultaneously. Figure 4 shows the first timepoint of this time-lapse experiment. HTM allows the acquisition of an image of the cell’s RI distribution in 3D. We subsequently pro- jected it along the z-axis, and this image was segmented using the EA segmentation tool (Figure 4a). EA segmentation captures very fine details of the cellular boundaries, as shown in the Fig- ure 4a insets, and is especially good at capturing live membrane protrusions such as filopodia, lamellipodia, and dendrites. Te cells were also stained using Draq5 and CellTrace . Draq5 is a modified anthraquinone that easily
VioletTM
permeates the cell and interacts with double-stranded DNA through weak stacking and hydrogen bonding [26]. Draq5 is a reference for live fluorescent imaging of cell nuclei, as it is
28
reportedly less toxic for genetic processes than Hoechst, for example [12,16,27], and it generates less phototoxicity due to its excitation at long wavelengths. CellTrace VioletTM
is an
Figure 4: Segmentation of live cells using refractive index or low fluorescence signal. (a) Segmentation outlines (green) of 3T3-derived pre-adipocytes (preA) overlaid on the refractive index (RI) signal. Cell segmentations were computed with EVE Analytics (EA) using the RI signal. (b) Segmentation outlines (red) of preA overlaid on the far-red fluorescent signal emitted by Draq5 and CellTrace VioletTM
(low D+C). Cell segmentations were computed with CellProfiler4 (CP4) using low D+C signal. (c) Segmentation outlines of (b) overlayed on RI signal show that large parts of cells are missed using the low D+C signal for live cell segmentation.
www.microscopy-today.com • 2021 September
after low-power excitation (laser intensity: 10%, exposure time: 200 ms)
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80
