SEARCH

SEARCH BY CITATION

Keywords:

  • adipocyte;
  • preadipocyte;
  • SGBS;
  • differentiation;
  • DNA damage;
  • laser scanning cytometry;
  • high content analysis

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information

Understanding adipocyte biology and its homeostasis is in the focus of current obesity research. We aimed to introduce a high-content analysis procedure for directly visualizing and quantifying adipogenesis and adipoapoptosis by laser scanning cytometry (LSC) in a large population of cell. Slide-based image cytometry and image processing algorithms were used and optimized for high-throughput analysis of differentiating cells and apoptotic processes in cell culture at high confluence. Both preadipocytes and adipocytes were simultaneously scrutinized for lipid accumulation, texture properties, nuclear condensation, and DNA fragmentation. Adipocyte commitment was found after incubation in adipogenic medium for 3 days identified by lipid droplet formation and increased light absorption, while terminal differentiation of adipocytes occurred throughout day 9–14 with characteristic nuclear shrinkage, eccentric nuclei localization, chromatin condensation, and massive lipid deposition. Preadipocytes were shown to be more prone to tumor necrosis factor alpha (TNFα)-induced apoptosis compared to mature adipocytes. Importantly, spontaneous DNA fragmentation was observed at early stage when adipocyte commitment occurs. This DNA damage was independent from either spontaneous or induced apoptosis and probably was part of the differentiation program. © 2013 International Society for Advancement of Cytometry

Obesity is major risk factor for a number of chronic diseases, including diabetes, cardiovascular diseases, and cancer. In physiological context, body fat mass is driven via the balance between the adipocyte's growth and removal. Adipose tissue can grow by either hyperplasia or hypertrophy or both, whereas adipocyte removal is facilitated by apoptosis. Adipogenesis, when adipocytes develop from mesenchymal stem cells, is a fairly well known process ([1-3]). In current studies, multiparametric model systems have been developed, where adipocyte differentiation, its nuclear, cytoskeletal, and suborganellar morphological changes can be monitored together with different drug applications in an automated manner. Among these works, Lee et al. ([4]) had introduced a noticeable flow cytometric assay to assess differentiation of preadipocytes, based on cytoplasmic granularity changes corresponding to lipid droplet accumulation. However, application of flow cytometry, which requires cell suspensions, strongly violates physiological optima of solid tissues. This technique also omits the possibility of in situ visualization and reanalysis of samples attached to culture dish. These latter obstacles were solved by applying image cytometry in solid tissue quantifications, as in Lin's work, where laser scanning cytometry was used to study adipocytes and measure capacity of drugs to induce adipocyte apoptosis ([5, 6]). Yet, cell recognition of attached cells was not solved or at least not clearly explained whether single cell contouring or phantom contouring was used in their study. When attached cells get more confluent during their culturing, cell-by-cell segmentation becomes challenging and requires sophisticated algorithms and tools. There have been a few attempts to enable cell-by-cell analyses, for example, McDonough et al. ([7, 8]) introduced an algorithm to recognize cells in confluent cultures and a later study applied this algorithm ([9]). However, their assumptions that preadipocytes or adipocytes cover all the surface of culture disk, and nuclei are located in center of cells, seem inapplicable for human fat cells. Currently, there is no satisfactory quantitative study about human adipogenesis that focuses on single-cell imaging cytometry, and none of them would be efficiently adaptable for high throughput applications.

Our aim was to quantitate objectively the morphological transition from preadipocytes to adipocytes and to measure responses of differentiating fat cells to apoptosis induction on cell-by-cell basis with statistical relevance. Preadipocytes derived from a patient with Simpson-Golabi-Behmel syndrome (SGBS), a recently developed model for studying human adipocytes with subcutaneous origin, was used ([10, 11]). SGBS cells have high capacity for adipose differentiation and display a gene expression pattern similar to mature human fat cells ([12]). The results showed that our approach could inspect accurately adipocyte differentiation in physiological conditions at different time points, could quantitate single cells in high-content manner, and could conveniently study both adipocytes and preadipocytes simultaneously in a slide-based platform. Cellular responses to apoptotic induction, chromatin aggregation, and uniquely, DNA fragmentation in single cells were also monitored at each differentiation stage. Spontaneous apoptosis was found to exist in both preadipocytes and adipocytes during differentiation and there was higher sensitivity of preadipocytes to the death stimuli compared to that of mature adipocytes. Furthermore, we have shown for the first time on single cell level that human maturing fat cells accumulate DNA damage along adipogenesis. This novel imaging approach can potentially be expanded to evolve into a more advanced high-throughput application.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information

Cell Culture and Differentiation Induction

SGBS preadipocytes were seeded in ibidi eight-chamber slides at a density of 1.5 × 104 cells/cm2 in DMEM/F12 (Dulbecco's Modified Eagle's Medium/Nutrient Mixture F-12 Ham 1:1 mixture; Sigma-Aldrich, Budapest, Hungary) medium containing 100 U/ml penicillin/streptomycin, 33 µM biotin, 17 µM pantothenate (serum-free, basal medium), and 10% FBS at 37°C in 5% CO2 for 24 hours. Differentiation was induced for 4 days using serum-free basal medium supplemented with 2 µM rosiglitazone (Caymen Chemicals, Ann Arbor, MI, USA), 25 nM dexamethasone (Sigma), 500 µM 3-isobutyl-1-methylxantine (Sigma), and 10 µg/ml human apo-transferrin (Sigma), 20 nM human insulin (Sigma), 100 nM cortisol (Sigma), and 200 pM triiodothyronine (Sigma). From the fifth day, the complemented serum-free basal medium without rosiglitazone, dexamethasone, and isobutylmethylxantine was used ([10]).

Apoptosis Induction

Every third day of the differentiation period, apoptosis induction was conducted, in which SGBS cells were treated with a mixture of 10 nM human TNFα (PeproTech, London, UK) and 10 ng/ml of cycloheximide (Sigma) for 12 hours.

Experimental Design

Imaging cytometry measurements were performed along two parallel courses of study. In sequential study, a similarly induced group of fresh SGBS cells was divided into subgroups, which were then placed on separate slides; each subgroup was allowed to grow in an incubator until its scheduled daily measurement. In longitudinal follow-ups, a particular sample was studied in a cohort pattern where cellular changes were tracked on cell-by-cell basis.

Analyses of same cells in cohort type protocols were also carried out by scanning the sample before and after apoptotic induction, and after the lysis of halo assay. Subsequently, “before” and “after” image layers were merged, that is, computationally stacked over each other, to allow analyses of correlated information from exactly the same cells at different experimental phases.

Cell Staining

On the day of measurement, cells were stained with Hoechst 33342 (50 µg/ml) for 60 minutes, Nile Red (25 µg/ml) and Nile Blue (750 µg/ml) for 20 minutes. Cell death was detected by propidium iodide and fluorescein isothiocyanate (FITC)-conjugated Annexin V with concentration indicated in Apoptosis Detection Kit (MBL) manuscript. Samples were washed once with phosphate buffered saline (PBS) and then kept in fresh medium.

Cell Lysis for Halo Assay

Upon accomplishing the examination of live cells, medium was discarded from the samples, and each well of ibidi 8-chamber slide was covered immediately by 150 μl volume of melted-39°C 1% low melting point agarose (Boehringer Mannheim, Indianapolis, IN). The agarose was solidified by placing the slides on ice for 2 minutes. Cells were lysed in ice-cold alkaline lysis buffer [1% lauryl-sarcosine, 2.5 M NaCl, 10 mM Tris, 100 mM ethylenediaminetetraacetic acid (EDTA), 10% DMSO, 1% Triton-X-100 (pH 10)] for 10 minutes twice. Slides were neutralized and washed in cold 1× PBS with 5 μM EDTA buffer (pH 7.4) for 4× 4 minutes and fixed in cold 1% formaldehyde for 4 minutes. The samples were stained with SYBR Gold (Molecular Probes, Eugene, OR) diluted 1:10,000 in TE [10 mM Tris (pH 8.0) and 2 mM EDTA] for 15 minutes.

Image Acquisition

Images were obtained by using iCys Research Imaging Cytometer (iCys, CompuCyte Corporation, Westwood, MA). Sample slides were mounted on a computer-controlled stepper-motor driven stage. Area with optimal confluence was determined in low-resolution scout scan with ×10 magnification objective (NA 0.30) and 10-µm scanning step. High-resolution images were consequently obtained by using ×40 objective (NA 0.75) and 0.25-µm step. Size of a pixel was set to 0.25 µm × 0.245 µm at ×40 magnification.

Laser lines were separately operated, namely 405-nm Violet diode laser was used to excite Hoechst 33342, 488-nm Argon-ion laser was used for FITC, Nile Red, and propidium iodide, 633-nm HeNe gas laser for Nile Blue. Emission was collected by four photomultiplier tubes; Hoechst was detected at 463 ± 20 nm, FITC at 530 ± 15 nm, Nile Red at 580 ± 15 nm, propidium iodide and Nile Blue at 675 ± 25 nm. Transmitted laser light was captured by diode photodetectors in which light loss and shaded relief signals were measured to gain information about light absorption and light scattering of the objects.

Automated Recognition of Cellular Objects

Images were processed and analyzed by our high through-put automatic cell recognition protocol using iCys companion software (iNovator Application Development Toolkit, CompuCyte Corporation, Westwood, MA), Image J (National Institute of Health, MD) ([13]), and CellProfiler (The Broad Institute of MIT, MA) ([14, 15]).

Hoechst-stained nuclei from both cell types were first identified and marked as primary objects using adaptive Otsu's method ([16]). Based on parent nuclei, the secondary objects, for example, a whole cell, were later recognized. First, adipocytes were segmented if the sum of fluorescence signal of lipid droplet specific Nile Red and absorption signal was higher than the background. Image regions occupied by adipocytes were then virtually excluded from further search for preadipocytes. Preadipocytes were subsequently segmented based on intensity of phospholipid specific Nile Blue fluorescence (Fig. 1 and Supporting Information Fig. S1).

image

Figure 1. Segmentation used for autodetection of cellular events in adipocytes. Signals from each channel are shown in images: (a) Hoechst 33342 signal monitored nuclei (in cyan), (b) Nile Red staining showed lipid droplets (red), (c) light absorption signified lipid droplets, (d) Nile Blue staining showed phospholipid (blue). (e) Merged image of Hoechst, Nile Red, and Nile Blue channels showed preadipocytes (arrow) and adipocytes (asterisk) with discerned properties. (f) Cell recognition was processed by the automated segmentation module of CellProfiler software. First, nuclei were identified as primary objects based on Hoechst signal, shown with the inner (green) contours. Subsequently, whole cells as secondary objects with the outer (red) contours were detected using Nile Red and Nile Blue signals in close association with predefined primary objects. Images were taken from day 12 of sequential measurements. [Color figure can be viewed in the online issue which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Texture Analysis

Texture analysis per identified objects was done with built-in modules in CellProfiler. Parameter entropy measured the randomness of intensity distribution; sum entropy (ΣE) roughly informed about the number of lipid droplets. Parameter variance measured the difference between intensity of the central pixel and its neighborhood; sum variance (ΣV) roughly depicted the size of lipid droplets.

Assessment of Lipid Accumulation

The combination of the Nile Red and texture signal was used to quantitate lipid accumulation, and cells that contained lipid above a preset threshold value were considered differentiated adipocyte. The ratio of number of adipocytes over the total count of nuclei gave rise to differentiation ratio. This ratio was calculated in a region of at least 100 field images where each field covered an area of 1024 × 768 pixels. Consequently, data of at least 1000 gated cells were collected and analyzed in cell-by-cell as well as cell-population bases using statistical softwares (GraphPad Prism and Microsoft Excel). All steps of gating were confirmed by relocation and visual identification of cellular events.

Quantification of Halo Structures

Cellular event before and after lysis was computationally traced object-by-object to associate DNA halo with its original parent cell. Multiple thresholds were applied to analyze SYBR Gold signal, nucleoid matrix was distinguished from migrated DNA in halo ring (Figs. 5c–5f). Consequently, Tau—a parameter monitoring DNA damage—was defined as the fraction of DNA migrated to the periphery of the original nucleus (“halo”) over the total DNA integral of a cell.

  • display math

where FIHalo Ring is the integrated fluorescence in the peripheral halo (Fig. 5f), FITotal Halo is the integrated DNA fluorescence of the matching nucleus (Fig. 5d). Tau was thus a quantitative entity, linearly proportional to DNA damage (modification of previous study ([17])).

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information

Enhanced Inspection of Adipocyte Differentiation Using Organelle-Specific Dyes and Texture Analysis

Our principal ambition was to assess quantification of adipocyte differentiation in vitro on the single cell level. The designed high-content analysis with enhanced algorithms efficiently identifies cellular objects in highly confluent adipocyte cultures, a major challenge that was well recognized in other studies ([14, 15]). Throughout cell autodetection, nuclei of cells were first identified and marked as primary objects, based on which the secondary objects were later recognized, for example, associated cell membrane, organelles, or a whole cell (Fig. 1 and Supporting Information Fig. S1). This two-step recognition process minimized the existence of noncellular events in data evaluation.

Cellular morphology, lipid accumulation, cell and nuclear size alterations, DNA content and cell phases were elucidated at each time point during the progression of the differentiation. To enhance the assessment of stages of differentiation, information of Nile Red-staining together with texture properties of light loss and nuclear condensation were analyzed (Fig. 2). Along the differentiation pathway, characteristic texture and granularity was identified in undifferentiated cells (Fig. 2a, cell #1), committed differentiating preadipocytes (Fig. 2a, cell #2), and terminally differentiated adipocytes (Fig. 2a, cell #3). Strong correlation between texture properties and fluorescence from lipid accumulation was found (Figs. 2b–2e). Significant increase in texture sum entropy in parallel with lipid droplet accumulation was typified during the differentiation from preadipocytes (Fig. 2b) to adipocytes (Fig. 2c). Preadipocytes showed lower sum entropy and sum variance of texture compared to adipocytes (Figs. 2e and 2f). Benefiting from these two texture parameters, discrimination between preadipocyte and adipocyte populations was more robust than using only Nile Red fluorescence.

image

Figure 2. Correlation between texture parameters and lipid specific staining of adipocytes. Texture analysis of light loss signals facilitated cell deciphering. (a) Characteristic texture patterns of different cell types: object #1 was identified as preadipocyte, while objects #2 and #3 were adipocytes. (b, c) Parameter entropy measured the randomness of the light scatter intensity distribution. The sum of entropy (ΣE) roughly informed about the number of lipid droplets, which was positively correlated with Nile Red staining and both increased during adipocyte differentiation. (d, e) Parameter variance measured the difference between intensity of the central pixel and its neighborhood. The sum of variance (ΣV) roughly depicted the size of lipid droplets and as the cells accumulated more lipid, sum variance increased similarly to sum entropy. [Color figure can be viewed in the online issue which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

The study of cell differentiation was conducted for 15 days on both cellular (Fig. 3a) and histocytometric aspects (Fig. 3b). The preadipocyte commitment was found to occur on day 2–3 with compacting cytoplasm revealed by increasing light-loss signal (Fig. 3a). At this point, forming lipid structures still did not stain with Nile Red (triglyceride) but with Nile Blue (phospholipid). Shortly after that transient monosignal stage, minor but measurable Nile Red signal appeared (Fig. 3a, Day 4–5). Relatively few small separated lipid droplets and enhanced nuclear condensation were shown in committed preadipocytes on day 2–5. This was in contrast to undifferentiated cells, which had more cytoplasm, round giant, and faintly stained nuclei with no Nile Red fluorescence (Fig. 3a, Day 0–1). From day 6 to day 12, most of the adipocytes reflected typical morphological signs of in vitro-differentiated adipose cells, such as round distending shape, a cytoplasm filled with lipid droplets as well as strong Nile Red fluorescence and light-loss signals. Shrinking and more brightly stained nuclei, signs of nuclear condensation, became even more pronounced (Fig. 3a, Day 8–9, 10–14). At day 9–12, when ratio of fully differentiated cells reached 40% at the region of optimal confluence (two-dimensional in vitro confluence), the differentiation curve was considered saturated (Fig. 3b). Indeed, from day 12 until the completion of the experimental regime (18 days), no further formation of lipid droplets was seen, rather the lipid droplets fused and became distorted (Fig. 3a, Day 14–15).

image

Figure 3. Inspection of adipocyte differentiation. (a) Representative SGBS cells at different cell ages. Upper panels show phospholipid (blue: Nile Blue), neutral lipid droplet (red: Nile Red) and nuclear staining (cyan: Hoechst 33342), while their lower counterparts are merged images of light loss and nuclear signals. During the transformation from preadipocytes to adipocytes, the cells underwent morphological changes including shrinkage of nuclei, increase of chromatin compactness, and gradual formation of lipid droplets. (b) Kinetics of ratio of adipocytes matured over total number of cells. Data were taken from sequential measurements (mean ± SD, n = 7).

Download figure to PowerPoint

Detection and Quantification of Apoptosis in Preadipocytes and Adipocytes

Individual information of each particular cell on how it progressed through the apoptotic induction was gained by merging images taken before and after TNFα stimulation. Annexin V-labeled cells in the former image layer represented spontaneous apoptosis. AV+ cells in the image layer recorded after the induction (Figs. 4a–4f) indicated the inducing effect. A stable spontaneous apoptotic rate was seen in both adipocytes and their precursors (Figs. 4e and 4f, dark data points), ranging from 7.05% to 17.91%. After apoptotic induction, significant increase of Annexin V and propidium iodide positivity were seen in both cell types at every time point of differentiation (Figs. 4e and 4f, light data points). Particularly, preadipocytes showed more sensitivity to the apoptotic induction, from 26.06% to 50.09% of preadipocytes, with a rising tendency, showed Annexin V positivity (Fig. 4e) compared to a plateau at 30% average in the adipocyte population (Fig. 4f).

image

Figure 4. Preadipocytes are more sensitive to TNFα-induced apoptosis than adipocytes. Multifluorescent signals exhibited different cellular components after subjecting cells to apoptotic induction: (a) nuclei with Hoechst 33342 staining (cyan), (b) lipid droplets stained with Nile Red (red), (c) apoptotic membrane with Annexin V labeling (white), (d) merged images of Hoechst, Nile Red, and Nile Blue (phospholipid stain, shown in blue). (e, f) Response to apoptotic induction of preadipocytes and adipocytes was compared at each time point of differentiation. Apoptotic cells were recognized by their Annexin V labeling (mean ± SD, n = 4). [Color figure can be viewed in the online issue which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

image

Figure 5. Characteristic changes of nuclear morphology, DNA condensation, and DNA fragmentation on population level during adipocyte differentiation. Nuclei of adipocytes (a) were found to have decreased size and increased chromatin compactness (increased intensity of Hoechst signal) compared to that of preadipocytes (b). (c–f) To quantify DNA fragmentation, alkaline halo assay was performed, in which DNA was stained with SYBR Gold. Fluorescent signals of total integrated DNA (contoured in (d)), within nucleoid matrix (e) and within halo ring (region between two grey lines in (f)) were, respectively, enumerated. (g) DNA fragmentation is started to accumulate early in differentiating adipocytes. Tau, the ratio of migrated DNA in halo over total nuclear DNA (halo + matrix), is linearly proportional with degree of DNA fragmentation in alkaline halo assay. Tau was calculated from nonapoptotic Annexin V-negative population. From day 0 and 3, data were collected from preadipocyte population, while data from day 6 to 15 was collected from maturating adipocytes. Graph shows median, 25th and 75th percentiles, and whiskers indicate the range from minimal to maximal values of populations. At least 150 cells were included into calculation of Tau at one time point. [Color figure can be viewed in the online issue which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Additionally, the apoptotic induction brought on apoptotic bodies, which showed highly fragmented DNA in halo assay as well as strong fluorescence of Hoechst intensity. Typical membrane blebs in close association with condensed nuclei were observed in the majority of apoptotic cells (Fig. 4c). There was remarkable cell loss, which was proportional to the percentage of cells responding to the apoptotic induction: 24.12% ± 10.5% of cells was lost in induced samples compared to 10.67% ± 4.52% in controls (from data comparing nuclei counts before and after treatment in day-12 samples).

Spontaneous DNA Fragmentation During Human Adipocyte Differentiation

Signs of nuclear condensation previously seen in microscopic images (Fig. 3a) were also manifested at population level. In preadipocytes, large nuclear size was found to correlate with fainter Hoechst stained DNA (Fig. 5a), while adipocytes showed more compact nuclei and brighter Hoechst intensity (Fig. 5b). Max-pixel fluorescence was used to identify degree of chromatin condensation, particularly when it increased in mitotic or apoptotic cells ([18]). To clarify whether nuclear condensation that occurs alongside the differentiation was an indication of DNA damage, alkaline halo assay was performed at each time point of differentiation (Figs. 5c–5g).

In healthy cells, DNA fluorescence of a halo was confined to the center of nuclei since unfragmented high molecular weight DNA diffuse out at proximal distance (Fig. 5c, left structure). In cells that had accrued more damage to the DNA, smaller fragments migrated further from the central nucleoid matrix (Fig. 5c, right structure). There was a strong correlation between high fragmentation of apoptotic DNA and Annexin V positivity, as previously suggested by Bacso et al. ([19]). Of note, various degrees of spontaneous DNA fragmentation were revealed in Annexin V-negative cells during their differentiation to adipocytes. Our observation was enumerated by using a quantitative parameter Tau (see Material and Methods section). As cell differentiation progressed, the Tau value rose from 0.16 to 0.27 on day 3, and at day 5 its value saturated at level of 0.4. Increase of Tau revealed a continuous accumulation of DNA damage during the differentiation process (Fig. 5g). As an original observation on single cell level, this finding suggests that spontaneous DNA fragmentation occurs during adipocyte differentiation in the absence of apoptosis.

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information

Application of LSC to Fat Cell Analysis

Among vast variety of methods to investigate adipocyte biology, genetic approaches, biochemical assays, flow cytometry, and conventional microscopy are commonly used. Although biochemical techniques have largely contributed in understanding adipogenesis and adipoapotosis, it is incomplete to resolve potential heterogeneity of cell populations following a bulk analysis. While common microscopic methods are able to study single cells, their subjective application is user dependent and thus lacks of unbiased quantifications. Flow cytometry has the advantage of being quantitative on the single cell level, but it too has inherent shortcomings stemming from its design: handling samples in tubes and lifting cells in suspension. Attached adipocytes obviously do not tolerate the hazard of trypsinization as they deteriorated and lost ability to reattach, as reported by Lin et al. ([6]).

Automated microscopy-based high-content screening has gained significant momentum recently due to its ability to study many features simultaneously in complex biological systems ([20, 21]). In light of this contemporary trend, our work has been developed based on a laser-scanning cytometry platform. Although LSC has a flow cytometry heritage, it is not limited to analyzing cells in fluids ([22, 23]). It allows automated analysis of solid-phase samples and adherent cultured cells while preserving the sample structure. The precise localization and time of each cellular image measurement, multicolor fluorescence, bright field laser-scatter, and transilluminated images can be simultaneously recorded by multiplexed light sensors ([17]). In our case, living adipocytes were cultured in chamber slides during the entire experiment and measurement; this was especially useful for follow-up studies of time-resolved differentiation and apoptotic destruction processes. Morphological properties of individual cells as well as stoichiometric data of large populations were inspected and interrogated with minimized perturbation. Plotted data points were also relocated and examined by visualization and thus improved time-lapse experiments with live subjects.

In a slide-based histocytometric system, advanced image processing is required to enable cell-by-cell analysis in highly confluent samples. When facing this challenge, McDonough et al. ([7, 8]) have introduced a tessellation algorithm to estimate the boundaries between the image partitions that assumed to contain adipocytes. Still, this procedure would only be valuable when lipid droplets in adipocytes are closely associated with a centric nucleus. This is seldom satisfied, as mature adipocytes have typical eccentric nuclei ([24, 25]). On the other hand, tessellation generates pseudo-cell boundaries, that is, instead of selectively propagating from foreground pixels, the whole image is segmented and thus collected data would unavoidably include background pixels (Supporting Information Fig. S1h and Table, ([26])). In our case, we used CellProfiler 2.0 with customizable adaptive Otsu's segmentation and a declumping algorithm that constituted of two-step object identification (primary and secondary). This protocol has improved the accuracy of cell recognition and yielded more precise quantitative data acquisition at high cell density of adipocytes (Supporting Information Figs. S1e, S1f and Table). Hence, this work introduced a reproducible and information-rich protocol to follow the differentiation process of preadipocytes to adipocytes from the same SGBS cell population.

Using Texture Properties Significantly Improves Adipocyte Inspection

Beside fluorescent intensity profiles as in conventional imaging microscopy, texture properties of cellular objects were also analyzed in this work and were demonstrated to be an improved approach to investigate morphology of a mixed population of undifferentiated cells, preadipocytes and adipocytes.

In flow cytometry, light scattering is generally used for event recognition, although forward and side scatter signals (SSC) are also valuable parameters to resolve size and granularity of cellular objects. In LSC technology, there is no direct SSC signal, since it would need to be measured in a perpendicular plane, 90o to the projection of forward light beam; however, transmitted light was captured by diode photodetectors in which light loss and shaded relief signals were measured to gain information about light absorption and scattering characteristics of the objects, respectively ([27]). These LSC signals allowed defining several computed texture parameters of image data; therefore, it was still possible to gain fully that information given by flow cytometric SSC. The computational freedom of texture properties of cells proved to be even more beneficial and gave richer information compared to SSC.

Among the most common parameters of texture analysis, “sum entropy” and “sum variance” were featured in our data presentation, which informed us about the number and size of formed lipid droplets, respectively. These parameters were rather more sensitive for interpreting adipocyte differentiation or at least give similar results compared to signals of conventional fluorescent or chromatic lipid staining, such as Nile Red in our instance. In agreement with our results, Lee et al. ([4]) found that side scatter granularity signal in flow cytometry efficiently recognized lipid droplet formation of mature adipocytes differentiated from 3T3-L1 cells. As shown in our slide-based cytometric approach, texture parameters provided even better granularity information and thus could be improved indicators for early identification of lipid droplet formation.

Adipogenesis Undergoes Two Stages

The observation that most adipocytes completed terminal differentiation on day 10–12 coincides well with the generally accepted time point when adipocytes were considered fully differentiated. Our results are also in favor of the concept of two stages in adipogenesis: the initial commitment of mesenchymal stem cells to a preadipocyte fate and then terminal differentiation ([1]). Adipogenic stimuli induce terminal differentiation in committed preadipocytes through the activation of peroxisome proliferator-activated receptor-γ ([1]). Cell shape and extracellular matrix remodeling have been found to regulate preadipocyte commitment and competency by modulating WNT and RHO-family GTPase signaling cascades.

There Is Uniform Level of Spontaneous Apoptosis During SGBS Cell Differentiation

Our LSC technique allowed us to detect apoptosis in two ways at once: ([1]) detect phosphatidylserine externalization at early apoptosis using Annexin V labeling and ([2]) detect DNA fragmentation at late apoptosis by alkaline halo assay ([17]). Alkaline halo assay robustly indicates apoptotic DNA fragmentation, which is the irreversible step at the end of the apoptotic execution. This assay also introduced a modified quantitative parameter Tau to signify and compare different levels of DNA damage, ranking from smaller Tau of less fragmented DNA to high value of Tau in apoptosis [modified from previous work ([17])].

An equal level of spontaneous apoptosis was found in both preadipocytes and adipocytes during each checkpoint of the differentiation. A plausible explanation for the prevalence of spontaneous apoptosis is that, during the development process not all cells would evolve terminal differentiation; instead, a certain fraction of cells that was unable to adapt to any stages of the differentiation program, was thus led to self-destruction. Besides, it cannot be excluded that a nonideal environment may contribute to this phenomenon and set a certain level of spontaneous apoptosis.

SGBS Preadipocytes Are More Sensitive to Induced Apoptosis Than Mature Forms

TNFα stimulation induced apoptosis in both SGBS preadipocytes and adipocytes. Of note, the inducing effect was more pronounced in the preadipocyte population compared with their differentiated counterparts. In agreement with our results, different susceptibility of fat cell subpopulations to TNFα stimuli was also demonstrated in other studies, for example, an intrinsic depot-specific susceptibility enhancement was described in human omental preadipocytes to programmed cell death compared with preadipocytes in the subcutaneous region ([28]). Preadipocytes' prominent vulnerability to TNFα might be the result of the combined apoptotic and lipolytic effect of the TNFα ([29]).

During Preadipocyte–Adipocyte Transition SGBS Cell Nuclei Become Shrunken and Their Chromatin Become More Condensed

Verstraeten et al. ([30]) recently showed that level of lamin A and B proteins decreased at peripheral regions of nuclei during adipocyte differentiation, while vimentin reorganized into cage-like structures near lipid droplets. Here in our work, we observed that the transformation of preadipocytes to adipocytes involved significant nuclei size shrinkage along with intensifying staining of DNA (Figs. 3a, 5a, and 5b). These observations might relate to each other, as in mitosis or apoptosis, where chromatin condensation is affiliated with the disintegration of nuclear lamina ([31, 32]).

Accumulating DNA Damage Was Seen During SGBS Adipocyte Differentiation

After evaluating spontaneous and induced apoptosis of maturing SGBS cells, we asked whether noticeable DNA damage was associated to nonapoptotic cells. Programmed cell death was identified by Annexin V labeling and Tau was calculated following halo assay in gated nonapoptotic cells at every time point of differentiation. Interestingly, we have observed that Tau, which indicates the level of DNA damage, was continuously increasing from day 3, coinciding with the initiation of textural changes or lipid droplet formation.

In murine 3T3-T proadipocytes, it was previously shown that DNA repair capacity was reduced during the differentiation program ([33, 34]). However, conflicting data have recently risen regarding the repair of double strand breaks in adipocytes, in which a specific DNA repair mechanism for double strand breaks was shown to be increased ([35]). Nevertheless, in all of these works DNA damage was monitored in bulk samples, where it was unable to resolve changes on single cell level. Our study introduced, for the first time, the detection of direct DNA damage in single cells during the adipogenesis. Conflicting data mentioned above might originate from heterogeneous populations with double strand breaks developing in adipogenesis. These potentially heterogeneous subpopulations might carry different capacity towards various DNA repair pathways. This issue might be resolved by our method if direct single and double strand breaks would be measured specifically in single cells.

Accumulation of DNA damage or decrease of capacity of complete replication of genome seems to be a sign of differentiation of somatic cells in multicellular organisms ([36-38]). In human, for example, red blood cells loose completely their nuclei, platelets are cytoplasmic fragments of megakaryocytes, T and B cells go through DNA editing, muscle cells form syncytia, and there are many other signs of decreased or specified functionality of the chromatin along the differentiation program. We believe the increased level of DNA damage observed in our alkaline halo assay is a part of this general phenomenon, or there could be certain heterogeneity of DNA damage that was generated during the maturation of nonapoptotic SGBS cells. We are continuing our work in this field to resolve this possibility.

Conclusion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information

Presented techniques showed strong capacity to satisfy both reliability and high-throughput performance assessing lipid accumulation and viability in a population of fat cells based on analysis of individual cells. We have proved that our slide-based cytometry, when extended by halo assay, could make valuable observations related to nuclear changes and DNA damage. This novel approach can be potentially expanded to evolve into a more complex and powerful research modality combining cytometric data with gene expression analysis, immunohistochemistry, and detection of organelle specific markers.

Literature Cited

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information
  • 1
    Cristancho AG, Lazar MA. Forming functional fat: A growing understanding of adipocyte differentiation. Nat Rev Mol Cell Biol 2011;12:722734.
  • 2
    Lowe CE, O'Rahilly S, Rochford JJ. Adipogenesis at a glance. J Cell Sci 2011;124:26812686.
  • 3
    Ahima RS. Adipose tissue as an endocrine organ. Obesity (Silver Spring) 2006;14 (Suppl 5):242S249S.
  • 4
    Lee YH, Chen SY, Wiesner RJ, Huang YF. Simple flow cytometric method used to assess lipid accumulation in fat cells. J Lipid Res 2004;45:11621167.
  • 5
    Lin J, Della-Fera MA, Baile CA. Green tea polyphenol epigallocatechin gallate inhibits adipogenesis and induces apoptosis in 3T3-L1 adipocytes. Obes Res 2005;13:982990.
  • 6
    Lin J, Page KA, Della-Fera MA, Baile CA. Evaluation of adipocyte apoptosis by laser scanning cytometry. Int J Obes Relat Metab Disord 2004;28:15351540.
  • 7
    McDonough PM, Ingermanson RS, Loy PA, Koon ED, Whittaker R, Laris CA, Hilton JM, Nicoll JB, Buehrer BM, Price JH. Quantification of hormone sensitive lipase phosphorylation and colocalization with lipid droplets in murine 3T3L1 and human subcutaneous adipocytes via automated digital microscopy and high-content analysis. Assay Drug Dev Technol 2011;9:262280.
  • 8
    McDonough PM, Agustin RM, Ingermanson RS, Loy PA, Buehrer BM, Nicoll JB, Prigozhina NL, Mikic I, Price JH. Quantification of lipid droplets and associated proteins in cellular models of obesity via high-content/high-throughput microscopy and automated image analysis. Assay Drug Dev Technol 2009;7:440460.
  • 9
    Linehan C, Gupta S, Samali A, O'Connor L. Bisphenol A-mediated suppression of LPL gene expression inhibits triglyceride accumulation during adipogenic differentiation of human adult stem cells. PLoS One 2012;7:e36109.
  • 10
    Fischer-Posovszky P, Newell FS, Wabitsch M, Tornqvist HE. Human SGBS cells—A unique tool for studies of human fat cell biology. Obes Facts 2008;1:184189.
  • 11
    Keuper M, Dzyakanchuk A, Amrein KE, Wabitsch M, Fischer-Posovszky P. THP-1 Macrophages and SGBS adipocytes—A new human in vitro model system of inflamed adipose tissue. Front Endocrinol (Lausanne) 2011;2:89.
  • 12
    Wabitsch M, Brenner RE, Melzner I, Braun M, Moller P, Heinze E, Debatin KM, Hauner H. Characterization of a human preadipocyte cell strain with high capacity for adipose differentiation. Int J Obes Relat Metab Disord 2001;25:815.
  • 13
    Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nat Methods 2012;9:671675.
  • 14
    Carpenter AE, Jones TR, Lamprecht MR, Clarke C, Kang IH, Friman O, Guertin DA, Chang JH, Lindquist RA, Moffat J, et al. CellProfiler: Image analysis software for identifying and quantifying cell phenotypes. Genome Biol 2006;7:R100.
  • 15
    Kamentsky L, Jones TR, Fraser A, Bray MA, Logan DJ, Madden KL, Ljosa V, Rueden C, Eliceiri KW, Carpenter AE. Improved structure, function and compatibility for CellProfiler: Modular high-throughput image analysis software. Bioinformatics 2011;27:11791180.
  • 16
    Otsu N. A threshold selection method from gray-level histograms. IEEE Trans Syst, Man Cybernet 1979;9:6266.
  • 17
    Bacso Z, Eliason JF. Measurement of DNA damage associated with apoptosis by laser scanning cytometry. Cytometry 2001;45:180186.
  • 18
    Darzynkiewicz Z, Halicka HD, Zhao H. Analysis of cellular DNA content by flow and laser scanning cytometry. Adv Exp Med Biol 2010;676:137147.
  • 19
    Bacso Z, Everson RB, Eliason JF. The DNA of annexin V-binding apoptotic cells is highly fragmented. Cancer Res 2000;60:46234628.
  • 20
    Buchser W, Collins M, Garyantes T, Guha R, Haney S, Lemmon V, Li Z, Trask OJ. Assay Development Guidelines for image-based high content screening, high content analysis and high content imaging, 2012 Oct 01. In: Sittampalam GS, Gal-Edd N, Arkin M, Auld D, Austin C, Bejcek B, Glicksman M, Inglese J, Lemmon V, Li Z, et al., editors. Assay Guidance Manual. Bethesda, MD: Eli Lilly & Company and the National Center for Advancing Translational Sciences; 2004. Available from: http://www.ncbi.nlm.nih.gov/books/NBK100913/.
  • 21
    Ljosa V, Carpenter AE. Introduction to the quantitative analysis of two-dimensional fluorescence microscopy images for cell-based screening. PLoS Comput Biol 2009;5:e1000603.
  • 22
    Darzynkiewicz Z, Bedner E, Li X, Gorczyca W, Melamed MR. Laser-scanning cytometry: A new instrumentation with many applications. Exp Cell Res 1999;249:112.
  • 23
    Henriksen M, Miller B, Newmark J, Al-Kofahi Y, Holden E. Laser scanning cytometry and its applications: A pioneering technology in the field of quantitative imaging cytometry. Methods Cell Biol 2011;102:161205.
  • 24
    Atanassova P. Histochemical and ultrastructural criteria for early differentiation of human subcutaneous adipocytes. Folia Med (Plovdiv) 1998;40:2427.
  • 25
    Granneman JG, Li P, Lu Y, Tilak J. Seeing the trees in the forest: Selective electroporation of adipocytes within adipose tissue. Am J Physiol Endocrinol Metab 2004;287:E574E582.
  • 26
    Arbelaez P, Maire M, Fowlkes C, Malik J. Contour detection and hierarchical image segmentation. IEEE Trans Pattern Anal Mach Intell 2011;33:898916.
  • 27
    Luther EG, William. Laser scatter in clinical applications. Proc. SPIE 6864, Biomedical Applications of Light Scattering II, 686412 (February 15, 2008); doi:10.1117/12.763996.
  • 28
    Niesler CU, Siddle K, Prins JB. Human preadipocytes display a depot-specific susceptibility to apoptosis. Diabetes 1998;47:13651368.
  • 29
    Petruschke T, Hauner H. Tumor necrosis factor-alpha prevents the differentiation of human adipocyte precursor cells and causes delipidation of newly developed fat cells. J Clin Endocrinol Metab 1993;76:742747.
  • 30
    Verstraeten VL, Renes J, Ramaekers FC, Kamps M, Kuijpers HJ, Verheyen F, Wabitsch M, Steijlen PM, van Steensel MA, Broers JL. Reorganization of the nuclear lamina and cytoskeleton in adipogenesis. Histochem Cell Biol 2011;135:251261.
  • 31
    Kihlmark M, Imreh G, Hallberg E. Sequential degradation of proteins from the nuclear envelope during apoptosis. J Cell Sci 2001;114:36433653.
  • 32
    Oberhammer FA, Hochegger K, Froschl G, Tiefenbacher R, Pavelka M. Chromatin condensation during apoptosis is accompanied by degradation of lamin A+B, without enhanced activation of cdc2 kinase. J Cell Biol 1994;126:827837.
  • 33
    Tofilon PJ, Meyn RE. Reduction in DNA repair capacity following differentiation of murine proadipocytes. Exp Cell Res 1988;174:502510.
  • 34
    Bill CA, Grochan BM, Vrdoljak E, Mendoza EA, Tofilon PJ. Decreased repair of radiation-induced DNA double-strand breaks with cellular differentiation. Radiat Res 1992;132:254258.
  • 35
    Meulle A, Salles B, Daviaud D, Valet P, Muller C. Positive regulation of DNA double strand break repair activity during differentiation of long life span cells: The example of adipogenesis. PLoS One 2008;3:e3345.
  • 36
    Nouspikel T. DNA repair in differentiated cells: Some new answers to old questions. Neuroscience 2007;145:12131221.
  • 37
    Nouspikel T, Hanawalt PC. DNA repair in terminally differentiated cells. DNA Repair (Amst) 2002;1:5975.
  • 38
    Sjakste N, Sjakste T. Possible involvement of DNA strand breaks in regulation of cell differentiation. Eur J Histochem 2007;51:8194.

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Conclusion
  7. Literature Cited
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
cytoa22333-sup-0001-suppinfo.doc41KSupporting Information.
cytoa22333-sup-0002-suppfig1.tif8176KFigure S Comparison of cell recognition algorithms on a representative image. (a) Cell nuclei stained by Hoechst. (b) Adipocytes stained by Nile Red. (c) Preadipocytes stained by Nile Blue. (d-f) Object identification method presented in the article: nuclei were automatically identified as primary objects, contoured with green lines in (d). Secondary objects as adipocytes and preadipocytes were contoured with red lines in (e) and (f), respectively. (g) A merged image of a, b and c. Arrows show typical preadipocytes, asterisks indicate adipocytes. (h) The result of the segmentation method developed by McDonough (7,8). (i) Ground-truth segmentation by human manual contouring. Table summarizes identified objects enumerated from the two automated methods and manual segmentation. Area covered by identified objects is calculated as percentage of the total image. Accuracy of each algorithm is evaluated as the proportion of the true prediction achieved by the segmentation. F-measure, which is a more widely used statistical parameter to evaluate imaging algorithms, was also calculated (26). Both parameters ranked our method higher.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.