How to follow lipid droplets dynamics during adipocyte metabolism

Abstract

Lipid droplets (LDs) are important cellular organelles due to their ability to accumulate and store lipids. LD dynamics are associated with various cellular and metabolic processes. Accurate monitoring of LD's size and shape is of prime importance as it indicates the metabolic status of the cells. Unintrusive continuous quantification techniques have a clear advantage in analyzing LDs as they measure and monitor the cells' metabolic function and droplets over time. Here, we present a novel machine-learning-based method for LDs analysis by segmentation of phase-contrast images of differentiated adipocytes (in vitro) and adipose tissue (in vivo). We developed a new workflow based on the ImageJ waikato environment for knowledge analysis segmentation plugin, which provides an accurate, label-free, live single-cell, and organelle quantification of LD-related parameters. By applying the new method on differentiating 3T3-L1 cells, the size of LDs was analyzed over time in differentiated adipocytes and their correlation with other morphological parameters. Moreover, we analyzed the LDs dynamics during catabolic changes such as lipolysis and lipophagy and demonstrated its ability to identify different cellular subpopulations based on their structural, numerical, and spatial variability. This analysis was also implemented on unstained ex vivo adipose tissues to measure adipocyte size, an important readout of the tissue's metabolism. The presented approach can be applied in different LD-related metabolic conditions to provide a better understanding of LD biogenesis and function in vivo and in vitro while serving as a new platform that enables rapid and accurate screening of data sets.

Keywords: adipocyte metabolism; adipose tissue; lipid droplets; lipolysis.

Figure 1

Lipid droplet (LD) quantification workflow. (a) Input sample phase‐contrast image of differentiating 3T3‐L1 cells. (b) Segmentation process using waikato environment for knowledge analysis (WEKA) trainable segmentation tool. Three classes were assigned (red—droplet content, green—LD linings, purple—surroundings). The classifier was trained until a satisfactory product was produced. (c) LDs were obtained from the binary segmented image using ImageJ's automatic analyze particles tool. (d) The droplets were analyzed for all available morphologic variables (e) the classifier can be applied to other images to obtain batches of binarized LDs images.

Figure 2

Lipid droplet (LD) analysis through segmentation. (a) Representative stitched images of cultured adipocytes and level of adipogenesis analysis (n = 3) after 14 (gray) and 21 (red) days post adipogenesis induction. (b) Phase‐contrast images and segmentation products of adipocytes after 14 and 21 days post induction (magnification ×200, scale bar = 100 μm). (c) Single LD area distribution after 14 (gray, n = 2847) and 21 days (red, n = 4528). (d) Representative thresholding and segmentation products of differentiating adipocytes (magnification ×200, scale bar= 50 μm). (e) Percentage of detectable LDs in comparison to manual counting (n = 4), by thresholding (blue) and segmentation (green). (f) Single LD area distribution by thresholding (blue, n = 445) and segmentation (green, n = 1264) tools (g) distribution of adipocytes based on their average LD area (n = 40), the cells were divided into three groups, small (red, <10 μm²), large (blue, >35 μm²), and medium (green) (h) histograms of the number of LDs, cell area, and average LD size of the small (red), medium (green), and large (blue) averaged LDs area, significance was calculated using unpaired nonparametric Kruskal–Wallis test, followed by Dunn's posttest. (i) Scatter plot of the cell area, and the number of LDs of the small (red), medium (green), and large (blue) groups, the dots' radius corresponds with the average LDs area, together with a distribution plot of the number of LDs. Unless stated otherwise, significance was calculated using an unpaired nonparametric Mann–Whitney test.

Figure 3

Lipid droplet (LD) analysis during lipolysis. (a) The experimental model, and phase contrast images and segmentation products at 0 and 120 min post lipolysis induction (magnification of ×200, scale bar = 100 μm). (b) Number of LDs and LD area per cell after 0 min (gray), 60 and 120 min (orange), n = 38, significance was calculated using paired one‐way analysis of variance (ANOVA) with Friedman's posttest. (c) Stacked bars of amount of 10− μm LDs (green) and 10+ μm² LDs (gray) through lipolysis and the respective percentage of 10− and 10+ μm² LDs after 120 min (light orange) 60 min (orange) and 0 min (gray). (d) Paired normalized LD area analysis after 0, 60, and 120 min, and normalized relative (to time 0) LD area analysis after 60 and 120 min, normalization was performed to the initial LD number of a cell, n = 38, significance was calculated using paired one‐way ANOVA with Friedman's posttest and Wilcoxon signed rank test, respectively. (e) Illustration of a single cell and organelle LDs dynamics during lipolysis. (f) Paired single organelle LD area analysis after 0, 60, and 120 min, and relative (to time 0) LD area analysis after 60 and 120 min, n = 63, significance was calculated using paired one‐way ANOVA with Friedman's posttest and Wilcoxon signed rank test, respectively.

Figure 4

Stratification of lipolytic adipocytes based on average cellular lipid droplet (LD) size. (a) Distribution of lipolytic adipocytes in 0min based on their average LD area (n = 38), the cells were divided into three groups, small (red, <3.5 μm²), large (blue, >15 μm²), and medium (green), and the averaged LD area after 60 and 120 min of the three groups. (b) Principal component analysis of all morphological parameters of each cell for the small (red), medium(green), large (blue) cells. (c) Number of LDs over time in the small (red), medium(green), large (blue) groups, and the relative number of LD. (d) Normalized relative (to time 0) LD area analysis after 60 and 120 min for the small (red), medium(green), large (blue) groups, normalization was performed to the initial LD number of a cell. Unless stated otherwise, significance was calculated using two‐way analysis of variance with Sidak's posttest.

Figure 5

Lipid droplet (LD) analysis during acidic lipolysis. (a) Lipophagy induction experimental model and representative overlay phase‐contrast images of untreated cells (Control) and metformin‐treated, nutrient‐starved cells (MetF+ NR) after 8 h postinduction, stained with green fluoresce lysotracker dye (magnification ×200, scale bar = 100 μm). (b) Cytoplasmatic lysotracker intensity levels in Control (gray, n = 42) and MetF + NR (pink, n = 37) treated cells after 8 h, significance was calculated using unpaired nonparametric Mann–Whitney test. (c) LD's linings lysotracker intensity analysis in Control (gray, n = 770) and Metformin (pink, n = 770), significance was calculated using unpaired nonparametric Mann–Whitney test. (d) Mean LD area after 0 (gray, n = 571) and 8 h of metformin and nutritional treatment in LysoT^low (pink, n = 394, LysoT intensity>40 AU) and lysoT^high (green, n = 182, LysoT intensity<40 AU), significance was calculated using unpaired one‐way analysis of variance with Friedman's posttest. (e) Representative fluorescent images of MetF + NR LysoT^low and lysoT^high cells after 8 h postinduction, stained with green fluoresce lysotracker dye (magnification × 400, scale bar = 25 μm). (f) A comparison of LD's area and their respective lysotracker intensity in metformin‐treated cells, the correlation was calculated using Spearman's R correlation coefficient. (g) LD area and LD number analysis after 8 h of metformin + NR relative to 0 h (n = 25), the cells divided to LysoT^low (pink) and lysoT^high (green) cells, significance was calculated using Wilcoxon signed‐rank test and unpaired nonparametric Mann–Whitney test, respectively.

Figure 6

In vivo adipocyte quantification workflow. (a) Input of adipose tissue unstained whole‐mount image segmentation process using waikato environment for knowledge analysis (WEKA) trainable segmentation tool. Two classes were assigned (red‐cell content, green‐cell linings. The classifier was trained until a satisfactory product was produced, a probability map was generated and processed using ImageJ's watershed tool. (b) A histogram and a violin plot of adipocyte cell area in whole‐mount sections (n = 152).