High-throughput, nonperturbing quantification of lipid droplets with digital holographic microscopy

Abstract

In vitro differentiating adipocytes are sensitive to liquid manipulations and have the tendency to float. Assessing adipocyte differentiation using current microscopy techniques involves cell staining and washing, while using flow cytometry involves cell retrieval in suspension. These methods induce biases, are difficult to reproduce, and involve tedious optimizations. In this study, we present digital holographic microscopy (DHM) as a label-free, nonperturbing means to quantify lipid droplets in differentiating adipocytes in a robust medium- to high-throughput manner. Taking advantage of the high refractive index of lipid droplets, DHM can assess the production of intracellular lipid droplets by differences in phase shift in a quantitative manner. Adipocytic differentiation, combined with other morphological features including cell confluence and cell death, was tracked over 6 days in live OP9 mesenchymal stromal cells. We compared DHM with other currently available methods of lipid droplet quantification and demonstrated its robustness with modulators of adipocytic differentiation in a dose-responsive manner. This study suggests DHM as a novel marker-free nonperturbing method to study lipid droplet accumulation and may be envisioned for drug screens and mechanistic studies on adipocytic differentiation.

Keywords: adipocyte; adipogenesis; label-free.

Fig. 1.

DHM is noninvasive and can be used to assess different features in lipid droplet accumulating cells. A: Schematic representation of DHM image acquisition. An infrared laser beam passes through the sample, while a reference beam passes in parallel without crossing the sample. B: Image reconstruction and analysis. The superimposition of both waves creates a hologram containing the phase shift information, which can then be mathematically reconstructed into a single-plane image. C: Information contained within the hologram. In addition to the phase shift information, DHM records the intensity absorption D: Superimposition of fluorescence lipid staining with DHM image. On the left are a digital phase contrast image (nonquantitative) and LipidTox Deep Red fluorescence images. On the right is the superimposition of LipidTox Deep Red signal with DHM signal. E: Signal intensity histogram of the cross-section seen as a horizontal red line passing through a lipid droplet-accumulating cell on the DHM image in D. F: CellProfiler software image analysis detects single cells, represented by their red outlines. Cells touching the borders are excluded. G: Cell counting using the aforementioned CellProfiler plugin. Increasing numbers of cells were seeded into a 384-well plate and counted from DHM images acquired 24 h after plating. H: Correlation of detection of OP9 dead cells exposed to various toxic compounds from DHM images as compared with fluorescence (Fluo) images from Et-HD and Hoechst 33342 stains. Dead cells were defined in the CellProfiler plugin as small and rounded with a high phase shift signal. Error bars are SD. Scale bar = 100 μm.

Fig. 2.

Six-day time lapse of OP9 cells induced to differentiate toward adipocytes: Comparison of DHM and fluorescence microscopy. A: DHM images from day 1 to 6 postplating and induction. Top, 10× bottom, 20×. B: Fluorescence images taken from day 1 to 6 postplating and induction. Blue is Hoechst 33342, and red is either LipidTox Deep Red (top) or Nile Red (bottom). C: OPD values of OP9 cells induced to differentiate (red) versus noninduced OP9 cells (blue). D, E: LipidTox Deep Red and Nile Red median signal intensities in both induced and noninduced OP9 cells. Note the error during the Nile Red staining procedure on day 4 postinduction of adipocytic differentiation. R² = 0.6 for OPD vs. LipidTox Deep Red and R² = 0.54 for OPD vs. Nile Red. Error bars are SD. Scale bars = 100 μm.

Fig. 3.

A: Total TG and protein content of OP9 cells during the 6 day adipocytic differentiation. Quantification is based on the standard curve from the manufacturer’s control. B: Relative TG content over protein content during the 6 day differentiation. C: The transcriptional profile of relevant genes involved in the adipocytic differentiation of OP9 cells. Values are normalized to noninduced OP9 cells 1 day postplating and the geometric mean of two housekeeping genes: hprt1 and actb. Error bars are SD.

Fig. 4.

OPD signal of FACS-sorted OP9 cells induced to differentiate toward adipocytes during a time period of 6 days. A: Flow cytometric gates were chosen along the SSC signal (representing different stages of lipid accumulation), sorted separately, and subsequently plated into a 384-well plate. B: Average OPD signal of sorted OP9 cells 24 h after plating. Images without cells were removed from analysis. C: Representative DHM images of sorted OP9 cells. Top, 10× bottom, 20×. Error bars are SD. Scale bars = 100 μm.

Fig. 5.

DHM imaging can be used to identify inhibitors and enhancers of adipogenesis in a high-throughput manner. A: Snapshot of a 384-well plate presenting a montage of DHM images of OP9 cells incubated with different PPARγ antagonists and agonists at different concentrations (one image per well). On the left half of the plate, OP9 cells were induced to differentiate to adipocytes while incubated with different PPARγ inhibitors (T0070907, GW9662, and BADGE). On the right half of the plate, OP9 cells were not induced with the standard differentiation cocktail, but incubated with different PPARγ agonists (rosiglitazone, pioglitazone, and indomethacine). The small inset on the top shows dead cells B: Dose-response curves of different tested PPARγ inhibitors. C: Dose-response curves of the different tested PPARγ agonists. Dashed lines represent baseline OPD values for induced (DMI) or noninduced (no DMI) OP9 cells. Error bars are SD.