UNDERSTANDING THE BIOLOGY OF FAT METABOLISM is fundamental

Transcription

1 Image-Based High-Throughput Quantification of Cellular Fat Accumulation MIKE DRAGUNOW, 1,3 RACHEL CAMERON, 1 PRITIKA NARAYAN, 1,3 and SIMON O CARROLL 2 A number of biochemical methods are available for measuring fat accumulation in cell culture. The authors report a simple imagebased method for measuring fat accumulation in adipocytes using a combination of high-throughput brightfield microscopy and image analysis, which was validated biochemically using. The quickest and most accurate method of analysis was one based on thresholding brightfield images and determining the area of fat droplets per image. Thus, the authors have developed a simple high-throughput, label-free method for measuring fat accumulation that is applicable to any cell or tissue type where fat droplets are visible under light microscopy. (Journal of Biomolecular Screening 27:999-15) Key words: image analysis, lipids, high-content screening INTRODUCTION UNDERSTANDING THE BIOLOGY OF FAT METABOLISM is fundamental to studies of basic physiology and pathology. A variety of pathophysiological conditions such as obesity, insulin resistance, metabolic syndrome, and hyperlipidemia have been shown to be associated with alterations in the regulation of adipocyte lipid metabolism, which lead to increased levels of plasma free fatty acids. Drugs that alter lipolysis are potential agents for use in the treatment of these diseases. 3T3-L1 cells are a well-characterized model for the study of adipocyte differentiation. They undergo conversion to a lipid-containing adipocyte phenotype when stimulated with a hormone cocktail of dexamethasone, 3-isobutyl-1- methylxanthine (IBMX), and insulin. 1,2 Differentiation involves sequential activation of the transcription factors CCAAT/enhancer binding protein (C/EBP) β and δ, as well as peroxisome proliferatoractivated receptor (PPAR)γ and C/EBPα. 1 PPARγ is a key mediator of the adipogenic process, and forced expression of PPARγ in fibroblasts 3 and myoblasts 4 induces differentiation into adipocytes. A number of different methods have been used to determine the level of adipocyte differentiation, including enzyme assays for Departments of 1 Pharmacology and 2 Anatomy and 3 The National Research Centre for Growth and Development, Faculty of Medical and Health Sciences, University of Auckland, New Zealand. Received Feb 28, 27, and in revised form Jun 23, 27. Accepted for publication Jun 3, 37. Journal of Biomolecular Screening 12(7); 27 DOI: / glycerol-3-phosphate dehydrogenase (GPDH) 5-7 and expression of differentiation markers such as PPARγ and C/EBPα using Northern or Western blotting. 8,9 A simple, reliable, and well-used method of analyzing adipocyte differentiation is to determine the amount of lipid accumulation using the lipid-specific stain Oil- Red-O is specific for triglycerides, the degree of staining is proportional to the amount of cell differentiation, and the dye can be extracted to allow an accurate quantification of differentiation by spectrometry. 14 As staining is specific for triglycerides, it is possible to accurately analyze the degree of lipid accumulation within a culture. We were interested in developing novel assays for analyzing fat accumulation in cells using microscopy and image analysis. We induced differentiation in 3T3-L1 cells and measured their fat accumulation using image analysis techniques on images acquired with brightfield microscopy and to compare the 2 methods. We performed a time course investigating fat accumulation in 3T3-L1 cells in a 24-well plate format (1-5 days) and demonstrated that fat analysis with brightfield microscopy and image analysis accurately and quickly determines the level of lipid accumulation. We also performed experiments in a 96-well format to show that this assay could be used for high-throughput screening (HTS). Cell culture conditions MATERIALS AND METHODS 3T3-L1 preadipocytes were seeded (2 1 4 per well) in 24- well plates and grown to confluence in Dulbecco s modified 27 Society for Biomolecular Sciences 999

2 Dragunow et al. Eagle s medium (DMEM) (5 µl per well) supplemented with 1% fetal bovine serum (FBS), penicillin (1 U/mL), streptomycin (1 µg/ml), and L-glutamine (29.2 mg/ml) at 37 C in humidified 5% CO 2, 95% air. Two days postconfluency, the media were changed to induce hormone-mediated differentiation by the addition of differentiation medium: DMEM, 1% FBS, penicillin (1 U/mL), streptomycin (1 µg/ml), L-glutamine (29.2 mg/ml), and an adipogenic hormone cocktail of 1 µm dexamethasone (Sigma, St. Louis, MO),.5 mm IBMX (Sigma), and 1 µg/ml insulin (Gibco Invitrogen, Carlsbad, CA). After 2 days, media were replaced with 1% FBS DMEM (+PSG and L-glutamine) and changed every 2 days until the end of the experiment. At various times after the addition of differentiation medium (1, 2, 3, 4, or 5 days), cells were fixed with 4% paraformaldehyde (PFA) for 1 h, rinsed, and stored in phosphate-buffered saline (PBS) for image acquisition and staining. Image acquisition Images from differentiated 3T3-L1 cells were acquired using the automated Discovery-1 inverted microscope and imaging platform (Molecular Devices, Sunnyvale, CA). Four sites per well for each well were acquired. Images were acquired using transmitted light (brightfield) microscopy with focusing between wells to obtain consistency in the focal plane optics. The automated focusing algorithm imaged 4 sites per well in 18 s. Biochemical fat assay Biochemical quantification of fat accumulation in differentiated 3T3-L1 adipocytes was assessed with the fat-soluble dye Oil- Red-O. Fixed, differentiated cells were incubated with solution (.3% [Sigma] in 6% isopropanol [BDH], 2 µl per well, 3 min, room temperature). Stained cells were rinsed twice with distilled water, washed for 5 min with 5% isopropanol, and stored in PBS. Then, 4-µL samples of the extracted dye solution were transferred to a 96-well plate (Nunc, Rochester, New York) for absorbance measurements at 51 nm using a SpectraMAX 25 plate reader (Molecular Devices). Image analysis assays To measure fat accumulation in images, we developed a number of assays using MetaMorph image analysis software (v6.2.6, Molecular Devices). The Fat Area Journal/Assay opened each image and then thresholded each image to specifically segment the rim around fat droplets (which were darker) in the images (Fig. 1c). All thresholded objects then underwent image morphometry analysis (IMA). IMA enables the user to filter out any debris/noise that is minute but could contribute to generating false positives. It is an application within MetaMorph that allows each thresholded object to undergo a selection process based on specific parameters that have upper and lower limits. For example, we employed total area and specified that any objects less than 1 µm 2 are excluded from the analysis as they were assumed to be noise/background pixels. Total area was then used to quantify the area of thresholded fat per image. Because the total area measure in MetaMorph measures the total area of each object, including any holes present in the thresholded region, this measure allowed for the accurate determination of the actual area of fat droplets even though the thresholding of each image detected only the rim of the circular fat droplets. The total area (in µm 2 ) of fat per image was then logged in an Excel spreadsheet. The invert-and-count assay inverted each brightfield image (Fig. 1d) and then performed the count nuclei assay, which is available on MetaMorph (but only for fluorescence images) to count the number of lipid droplets per image (Fig. 1e). The count nuclei assay detects and counts objects based on their width and their intensity above local background. The brightfield cell count assay used the FIND SPOTS application in MetaMorph (which detects objects based on their size and intensity) to count fat-positive cells per image (Fig. 1f). The fat droplet cell count assay also used the FIND SPOTS application but at a much smaller spot size for the segmentation of fat droplets (rather than entire cells; Fig. 1g). The top hat and count assay used the top hat filter on MetaMorph to extract the lipid droplets (Fig. 1h) and then performed the count nuclei assay available on MetaMorph (Fig. 1i). For the image analysis assays, the data were automatically logged to Excel spreadsheets. Before final use of the data for statistical analysis, any aberrant readings on spreadsheets were backchecked against the original images to exclude out-of-focus images or images with staining artifacts. Because we used a multiple-site per well acquisition strategy with the Discovery-1 microscope, this meant that there were always sufficient goodquality images for each well for analysis purposes. Finally, to determine whether the image analysis assays would also work on 96-well plates, 3T3-L1 preadipocytes were plated in 96-well plates and grown to confluence in DMEM (1 µl per well) supplemented with 1% FBS, penicillin (1 U/mL), streptomycin (1 µg/ml), and L-glutamine (29.2 mg/ml) at 37 C in humidified 5% CO 2, 95% air. Two days postconfluency, the media were changed to induce hormone-mediated differentiation by the addition of differentiation medium (DMEM), 1% FBS, penicillin (1 U/mL), streptomycin (1 µg/ml), L-glutamine (29.2 mg/ml), and a range of adipogenic hormone cocktails (n = 13, see Fig. 4 legend for details) comprising different concentrations of dexamethasone (.1,.25,.5, and 1 µm), IBMX (.5,.125,.25, and.5 mm), and insulin (1, 2.5, 5, and 1 µg). After 2 days, media were replaced with 1% FBS DMEM (+PSG and 1 Journal of Biomolecular Screening 12(7); 27

3 High-Throughput Quantification of Cellular Fat Accumulation FIG. 1. (a, b) Photomicrographs showing brightfield images of 3T3-L1 cells (a) before and (b) after differentiation with an adipogenic hormone cocktail of 1 µm dexamethasone,.5 mm 3-isobutyl-1-methylxanthine (IBMX), and 1 µg/ml insulin for 5 days (magnification 1). The image in b shows lipid droplets as imaged under brightfield illumination. (c) Photomicrograph showing the brightfield image of differentiated 3T3 cells captured at 1 magnification in b but with orange overlay showing the area captured by thresholding this image for dark objects. This area was then quantified (in µm2) using the fat area assay, written using MetaMorph journaling. (d, e) Photomicrographs showing the brightfield image of differentiated 3T3 cells from b using the Morphology Filters>Invert Image processing application in MetaMorph (v6.2.6). Fat droplets in the inverted image (which produces a pseudo-fluorescence) are then quantified (number of droplets per image, red overlay) using the count nuclei assay in Metamorph (e). (f) Photomicrographs showing the brightfield image of differentiated 3T3 cells from b. Superimposed blue circles show cells detected using the brightfield cell count assay, written using MetaMorph journaling. The image shown in g shows the same original photomicrograph, with blue circles detecting fat droplets using the fat droplet cell count assay, written using MetaMorph journaling. (h, i) Photomicrographs showing the same brightfield image as b, segmented using the top hat filter to extract the dark (lipid droplets, h) features. The segmented image was then quantified (red overlay, i) using the count nuclei assay in MetaMorph. L-glutamine) and changed every 2 days until the end of the experiment (5 days), when cells were fixed with 4% PFA for 1 h, rinsed, and stored in PBS for image acquisition and staining. Images were acquired as described above, and fat content was analyzed with (as described above), the fat area assay (as Journal of Biomolecular Screening 12(7); 27 described above), and blinded manual scoring. The manual scoring was performed by a human observer blinded to the experimental conditions but trained to detect fat droplets in the images. A scoring system from (no fat droplets) to 9 (fat droplets throughout the image) was used to score fat accumulation in images. 11

4 Dragunow et al. a b c Fat Area d Fat Droplets Fat Droplets e Fat Droplets (TopHat Assay) Fat Positive Cells FIG. 2. Graphs showing the correlation between measured fat levels and fat as measured using the (a) fat area assay, (b) invert-and-count assay, (c) brightfield cell count assay, (d) fat droplet cell count assay, and (e) top hat and count assay. All correlations are significant (see Results for details). Each data point on each graph represents the same well measured twice with both methods, and the y-axis shows the average imaging signal of the whole well. Please note that the variation between wells is due to differences in fat levels caused by data derived from adipocytes fixed at different times after differentiation and reflects variations in fat levels, not variability of the assays. RESULTS Differentiation of cells led to an accumulation of fat droplets in 3T3-L1 cells that were easily observed and captured under light microscopy (Fig. 1a,b). The fat area assay measured the total area of fat droplets per image (Fig. 1c). This assay took approximately.19 s per image, and the data derived with this method correlated significantly (R =.8787, p <.1) with the actual fat levels as determined with (Fig. 2a). The invert-and-count assay counted the number of lipid droplets per image (Fig. 1e). This assay required more processing and took approximately 1 s per image, and the data derived with this method correlated significantly (R =.8416, p <.1) with the actual fat levels as determined with (Fig. 2b). The brightfield cell count assay counted fat-positive cells per image (Fig. 1f). This assay took approximately.69 s per image, and the data derived with this method correlated significantly (R =.7618, p <.1) with the actual fat levels as determined with (Fig. 2c). The fat droplet cell count assay counted lipid droplets (rather than entire cells; Fig. 1g). This assay took approximately 17 s per image, and the data derived with this method correlated significantly (R =.8332, p <.1) with the actual fat levels as determined with (Fig. 2d). The top hat and count assay also counted lipid droplets (Fig. 1h,i). This assay took approximately 1 s per image, and the data derived with this method correlated significantly (R =.867, p <.1) with the actual fat levels as determined with (Fig. 2e). The time course of fat accumulation was also used to determine which image analysis assay best represented the Oil-Red- O biochemical assay (Fig. 3a). Although none of the image analysis assays directly matched the results, the fat area assay (Fig. 3b) appeared to be the most accurate and best matched the relative trends observed with. To determine the Z value 15 for the fat area assay, we analyzed a data set comparing images taken from positive (differentiated adipocytes) and negative (undifferentiated adipocytes) controls. The Z statistic gave a value of.48. Finally, we compared the fat area assay with the on adipocytes grown in a 96-well plate format (Fig. 4). Because our preliminary data suggested that the biochemical assay was not accurate on wells from a 96-well plate, due to the volume of extracted dye solution needed for the absorbance readings, we also undertook a blinded fat scoring for comparison purposes. As shown in Figure 4, the various concentrations of treatment cocktails produced (as expected) a broad variation in fat accumulation. When we undertook correlational 12 Journal of Biomolecular Screening 12(7); 27

5 High-Throughput Quantification of Cellular Fat Accumulation a 2 1 b Fat Area Jnl (µm2) 2 1 c Invert&Count Fat-Positive Droplets d e f Brightfield Cell Count Fat-Positive Cells Fat Droplet Cell Count Fat Droplets (TopHat) FIG. 3. Graphs showing the time course of fat accumulation in 3T3-L1 cells measured with (a), (b) the fat area assay, (c) the invert-andcount assay, (d) the brightfield cell count assay, (e) the fat droplet cell count assay, and (f) the top hat and count assay. Solid dark bars show cells treated with vehicle, and checked light bars show cells treated with an adipogenic hormone cocktail of 1 µm dexamethasone,.5 mm 3-isobutyl-1-methylxanthine (IBMX), and 1 µg/ml insulin. studies between the 3 analysis methods, we found that the manual scoring correlated significantly and at a high level with the fat area assay (R =.8896, p <.1) but weakly (R =.57), although still significantly, with the. To investigate this further, we graphed the fat level scores for each of the assays against the different treatments (Fig. 4). We found that the fat area assay detected a pattern of fat accumulation similar to that measured by manual scoring, but the was not able to discriminate clear differences in fat levels between the different treatments. This is particularly obvious in the comparison of treatments 12 and 13 (see Fig. 4). DISCUSSION These results show that image capture and analysis can be applied to analyze cellular fat accumulation. Although all of the image-based methods were accurate, the quickest and most accurate method was based on measuring the area of fat per image with a thresholding method. In general, the results obtained with image-based methods correlated closely with the actual fat content as determined with in the 24-well plate format. However, in the 96-well plate format, the image-based method was much superior to the method because, as depicted in Figure 4, the sensitivity of the assay is not sufficient to detect quite marked differences between levels of fat accumulation in the cells. In contrast, the fat area assay accurately determined the levels of fat accumulation in the 96-well plate format as measured against blind scoring. This result shows that for 96- well plate formats, the image analysis method is superior to biochemical determination of fat levels using a common assay. Also, the method detected basal expression of fat in 3T3-L1 cells at 1 to 4 days in culture in the 24-well plate format, whereas the fat area assay, the invert-and-count assay, and the top hat and count assays did not. This basal fat activity detected with is likely due to the lower limits of this assay and assay noise and indicates that the image-based methods were less likely to produce false-positive results. Our label-free imaging of cellular fat accumulation contrasts with other imaging methods (reviewed in Kuerschner et al. 16 ) that generally require a labeling method. These methods also accurately detect lipids but have a reduced throughput due to the labeling step and the requirement to capture images using fluorescence microscopy compared with our more rapid (and cheaper) brightfield acquisition. However, our label-free methods do not distinguish between types and composition of lipid droplets, and for these applications, labeling methods are still the preferred option. Our label-free high-throughput brightfield image capture and Journal of Biomolecular Screening 12(7);

6 Dragunow et al. a b c Fat AreaJnl (µm2) Treatment Treatment Manual Scoring Treatment d Treatment 12 Treatment 13 e FIG. 4. Graphs showing the measurement of fat content from 96-well plates using the fat area assay,, and blind manual scoring of fat content in images. Photomicrographs show actual images from treatments 12 and 13, where it is clear that treatment 12 generated much more fat accumulation than treatment 13. Both the fat area assay and the manual scoring detected this difference, but the assay did not. Treatments (6 replicates per treatment): 1, control; 2, 1 µm dexamethasone,.5 mm 3-isobutyl-1-methylxanthine (IBMX), and 1 µg insulin; 3, 5 µm dexamethasone,.5 mm IBMX, and 1 µg insulin; 4,.25 µm dexamethasone,.5 mm IBMX, and 1 µg insulin; 5,.1 µm dexamethasone,.5 mm IBMX, and 1 µg insulin; 6, 1 µm dexamethasone,.25 mm IBMX, and 1 µg insulin; 7, 1 µm dexamethasone,.125 mm IBMX, and 1 µg insulin; 8, 1 µm dexamethasone,.5 mm IBMX, and 1 µg insulin; 9, 1 µm dexamethasone,.5 mm IBMX, and 5 µg insulin; 1, 1 µm dexamethasone,.5 mm IBMX, and 2.5 µg insulin; 11, 1 µm dexamethasone,.5 mm IBMX, and 1 µg insulin; 12,.5 µm dexamethasone,.25 mm IBMX, and 5 µg insulin; 13,.25 µm dexamethasone,.125 mm IBMX, and 2.5 µg insulin. analysis method is more suited to an initial screening program where it can be applied to determine the effects of large numbers of compounds on cellular fat accumulation cheaply, accurately, and at high sensitivity in cell culture experiments. ACKNOWLEDGMENTS This work was supported by the National Research Centre for Growth and Development (NRCGD) and the Health Research Council of New Zealand. REFERENCES 1. Rosen ED, Spiegelman BM: Molecular regulation of adipogenesis. Annu Rev Cell Dev Biol 2;16: Green H, Kehinde O: An established preadipose cell line and its differentiation in culture: II. Factors affecting the adipose conversion. Cell 1975; 5: Tontonoz P, Hu E, Spiegelman BM: Stimulation of adipogenesis in fibroblasts by PPAR gamma 2, a lipid-activated transcription factor. Cell 1994;79: Hu E, Tontonoz P, Spiegelman BM: Transdifferentiation of myoblasts by the adipogenic transcription factors PPAR gamma and C/EBP alpha. Proc Natl Acad Sci USA 1995;92: Hirai S, Yamanaka M, Kawachi H, Matsui T, Yano H: Activin A inhibits differentiation of 3T3-L1 preadipocyte. Mol Cell Endocrinol 25;232: Wise LS, Green H: Participation of one isozyme of cytosolic glycerophosphate dehydrogenase in the adipose conversion of 3T3 cells. J Biol Chem 1979;254: Yan H, Kermouni A, Abdel-Hafez M, Lau DC: Role of cyclooxygenases COX-1 and COX-2 in modulating adipogenesis in 3T3-L1 cells. J Lipid Res 23;44: Morrison RF, Farmer SR: Role of PPARgamma in regulating a cascade expression of cyclin-dependent kinase inhibitors, p18(ink4c) and p21(waf1/cip1), during adipogenesis. J Biol Chem 1999;274: Tang QQ, Otto TC, Lane MD: CCAAT/enhancer-binding protein beta is required for mitotic clonal expansion during adipogenesis. Proc Natl Acad Sci USA 23;1: Journal of Biomolecular Screening 12(7); 27

7 High-Throughput Quantification of Cellular Fat Accumulation 1. Green H, Kehinde O: Sublines of mouse 3T3 cells that accumulate lipid. Cell 1974;1: Kliewer SA, Lenhard JM, Willson TM, Patel I, Morris DC, Lehmann JM: A prostaglandin J2 metabolite binds peroxisome proliferator-activated receptor gamma and promotes adipocyte differentiation. Cell 1995;83: Tang QQ, Otto TC, Lane MD: Mitotic clonal expansion: a synchronous process required for adipogenesis. Proc Natl Acad Sci USA 23;1: Sarruf DA, Iankova I, Abella A, Assou S, Miard S, Fajas L: Cyclin D3 promotes adipogenesis through activation of peroxisome proliferator-activated receptor gamma. Mol Cell Biol 25;25: Ramirez-Zacarias JL, Castro-Munozledo F, Kuri-Harcuch W: Quantitation of adipose conversion and triglycerides by staining intracytoplasmic lipids with Oil Red O. Histochemistry 1992;97: Zhang JH, Chung TD, Oldenburg KR: A simple statistical parameter for use in evaluation and validation of high throughput screening assays. J Biomol Screen 1999;4: Kuerschner L, Ejsing CS, Ekroos K, Shevchenko A, Anderson KI, Thiele C: Polyene-lipids: a new tool to image lipids. Nat Methods 25;2: Address correspondence to: Professor Mike Dragunow Department of Pharmacology Faculty of Medical and Health Sciences University of Auckland Auckland, New Zealand m.dragunow@auckland.ac.nz Journal of Biomolecular Screening 12(7);