Skip to main content
Tissue Engineering. Part A logoLink to Tissue Engineering. Part A
. 2012 Sep 5;19(1-2):40–48. doi: 10.1089/ten.tea.2012.0127

Live-Cell, Temporal Gene Expression Analysis of Osteogenic Differentiation in Adipose-Derived Stem Cells

Hetal V Desai 1, Indu S Voruganti 1, Chathuraka Jayasuriya 2, Qian Chen 2, Eric M Darling 1,3,4,5,
PMCID: PMC3530940  PMID: 22840182

Abstract

Adipose-derived stem cells (ASCs) are a widely investigated type of mesenchymal stem cell with great potential for musculoskeletal regeneration. However, use of ASCs is complicated by their cellular heterogeneity, which exists at both the population and single-cell levels. This study demonstrates a live-cell assay to investigate gene expression in ASCs undergoing osteogenesis using fluorescently tagged DNA hybridization probes called molecular beacons. Three molecular beacons were designed to target mRNA sequences for alkaline phosphatase, type I collagen, and osteocalcin (ALPL, COL1A1, and BGLAP), genes characteristically expressed during osteogenesis. The percentage of cells expressing these genes in a population was monitored daily to quantify the uniformity of the differentiation process. Differentiating ASC populations were repeatedly measured in a nondestructive fashion over a 21-day period to obtain temporal gene expression data. Results showed consistent expression patterns for the investigated osteogenic genes in response to induction medium. Peak expression was observed at days 3–4 for ALPL, day 14 for COL1A1, and day 21 for BGLAP. Additionally, the differentiation response of sample populations became more uniform after 2 weeks in osteogenic induction medium, suggesting a syncing of ASCs occurs over time. These findings are consistent with previous studies of osteogenic differentiation and suggest that molecular beacons are a viable means to monitor the spatiotemporal gene expression of live, differentiating ASCs.

Introduction

Stem cell-based therapies hold immense promise for treating myriad diseases, and researchers across multiple fields have dedicated themselves to exploring this potential.1 Tissue engineers frequently use stem cell populations induced with biochemical and/or biomechanical stimuli to generate tissues of interest, such as muscle or bone. Adipose-derived stem cells (ASCs) are especially attractive because of their relative abundance and nonimmunogenicity and have shown good potential for use in musculoskeletal regeneration.24 However, experiments using ASCs are often confounded by heterogeneity, which can negatively affect cellular differentiation and matrix production. Single-cell and subpopulation effects are often obscured by the whole-population assays typically used by researchers.5 A new method capable of nondestructively assessing stem cell differentiation and heterogeneity in populations over time would alleviate many of the issues currently faced by researchers in the field.

Mesenchymal stem cell heterogeneity exists at multiple levels. First, cell populations harvested from stromal tissues are nonuniform, containing a mixture of differentiated and undifferentiated cell types that can respond to environmental conditions in dramatically different fashions.6 Second, the stem/progenitor cells themselves possess disparate differentiation capabilities (unipotency, bipotency, multipotency, and pluripotency).79 This heterogeneity is problematic for both basic science experiments and translational applications because individual cells can only respond according to their capabilities. Understanding the degree of uniformity in differentiating populations is critical for identifying important subpopulations that hold the key to regenerating tissues and treating diseases.

This study establishes a live-cell analysis approach using fluorescently tagged DNA hybridization probes called molecular beacons to determine gene expression patterns in osteogenically differentiating ASCs. Molecular beacons are hairpin-shaped nucleic acid probes functionalized with a fluorophore and a quencher on opposing ends.10 The loop region of the probe is complimentary to a nucleic acid sequence of interest. In the absence of the target sequence, the probe retains its stem-loop structure and fluorescence is quenched. When the target sequence is bound by the loop region, the stem unfolds, affording fluorescence. Molecular beacons have been used in many capacities, including single-nucleotide polymorphism detection, real-time polymerase chain reaction (PCR) applications, and many live cell imaging applications.1116

For this study, molecular beacons were designed to target mRNA molecules coding for alkaline phosphatase, type I collagen, and osteocalcin (ALPL, COL1A1, and BGLAP), genes characteristically expressed during osteogenesis.17,18 By delivering novel beacons for genes expressed at progressive stages of osteogenic differentiation, we were able to investigate the uniformity of this process in living populations and identify spatiotemporal patterns of expression, which would not be possible using standard, destructive techniques.

The goal of this study was to analyze temporal gene expression patterns in living cells during osteogenic differentiation at the population and subpopulation levels. Human ASC populations were chemically induced along the osteogenic lineage over a 21-day period, and gene expression was quantified using custom-designed molecular beacons. This experimental approach marks the first time that an investigation of gene expression patterns in living mesenchymal stem cells has been performed in a repeatable, nondestructive fashion.

Materials and Methods

Cell culture

All cells were maintained in a humidified incubator at 37°C and 5% CO2. MG-63 and HEK-293 cell lines (ATCC, Manassas, VA) were cultured in growth medium containing phenol red-free MEM (CellGro, Manassas, VA) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, 100 mM Glutamax, and 100 mM sodium pyruvate (ThermoFisher Scientific, Waltham, MA). Cells were passaged at 80% confluence using 0.25% trypsin-EDTA (ThermoFisher Scientific). For molecular beacon experiments, cells were seeded into 96-well plates at a density of ∼25,000 cells/cm2.

ASCs derived from subcutaneous adipose tissue, originally harvested from seven, healthy, nondiabetic donors between the ages of 18 and 60 years old, were purchased from Zen-Bio, Inc. (superlot #36; Research Triangle Park, NC). Cells were grown in expansion medium containing DMEM/F-12 (ThermoFisher Scientific), 10% FBS (Zen-Bio), 1% penicillin/streptomycin, 0.25 ng/mL transforming growth factor-β1, 5 ng/mL epidermal growth factor, and 1 ng/mL fibroblast growth factor (R&D Systems, Minneapolis, MN).19 All ASCs used for experiments were at passage 4.

Beacon development and design

Three custom-designed beacons were developed corresponding to alkaline phosphatase, type I collagen, and osteocalcin mRNA (ALPL, COL1A1, and BGLAP, respectively), which are common markers of osteogenesis (Table 1). Each beacon was functionalized with a 6-FAM (excitation [Ex]: 492 nm/emission [Em]: 517 nm) fluorophore on the 3′ end and a Black Hole Quencher-1 on the 5′ end. A nucleic acid folding program, mfold, was used to model the secondary structures of each mRNA molecule based on thermodynamic stability.20,21 The five structures with the lowest Gibbs' free energy were analyzed for regions of largely unpaired or looped secondary structure. A 20–30-base sequence was chosen and assessed using NCBI BLAST to ensure uniqueness.22,23 ALPL, COL1A1, and BGLAP beacons were highly specific to their target sequences (e-values 105, 102, and 106 times smaller than the next sequence match, respectively). The stem region of each beacon was designed to give the probe an optimal melting temperature of 60°C–80°C.24 The folding of the beacon sequence was also assessed to ensure that a hairpin structure existed. All beacons were manufactured and HPLC purified via commercial sources (MWG Operon, Huntsville, AL).

Table 1.

Molecular Beacon Sequences for Osteogenic Genes

Gene Beacon sequence 5′→3
GAPDH28 CGACGGAGTCCTTCCACGATACCACGTCG
ALPL CGCTCCAGAGTGTCTTCCGAGGAGGTCAAGGAGCG
COL1A1 CGTCCCAAAAAAAAAAAAAAAAAGAAAAATATCAGGGACG
BGLAP TCCGCCGGAAAGAAGGGTGCCTGGAGAGGAGCGGCGGA

Stem regions are underlined. Remainder of oligonucleotide forms the loop region, which is complementary to the gene of interest.

Molecular beacon hybridization assay

Validation of hybridization efficiency was done by measuring the fluorescence of fixed concentrations of beacon hybridization to varying concentrations of target sequence (Supplementary Fig. S1; Supplementary Data are available online at www.liebertonline.com/tea). ALPL molecular beacon in pH 7.4 1× Tris-EDTA buffer (ThermoFisher Scientific; 100 μM solution) was added to wells in an opaque 96-well plate at a final beacon concentration of 5 μM/well. Stepwise concentrations of ALPL target sequence (0.5–5.0 μM) were then added to the wells. Controls included wells containing only beacon and Tris buffer, only target and Tris buffer, and only Tris buffer. Sample plates were incubated at 37°C for 10 min, and fluorescence was read with a spectrofluorometer (Spectramax Plus 384; Molecular Devices, Sunnyvale, CA; Ex: 492 nm, Em: 517 nm) every 10 min for a total of 270 min.25

Beacon validation and testing

MG-63 cells, which highly express osteogenic genes,26 and HEK-293 cells were seeded at a density of 50,000–60,000 cells per well in a 24-well plate. Two nanograms of ALPL, COL1A1, and BGLAP molecular beacon (2 μL of 100 μM solution in Tris-EDTA buffer, pH 7.4) was each encapsulated in 4 μL xtremeGENE HP reagent (1:2 ratio beacon:reagent; Roche Biotech, Pleasanton, CA) and suspended in 100 μL base medium (MEM) according to product instructions. The complexes were delivered to wells at a concentration of 0.5 μM to ensure that the molecular beacon would be in great excess of the mRNA transcripts (∼6×1012 −10×1012 beacons/well).27 A previously published molecular beacon for the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was encapsulated and delivered in the same way for use as a positive control.28 Following beacon treatment, cells were allowed an uptake period of at least 2 h before being imaged on a Nikon Eclipse Ti-U epifluorescent microscope (Nikon Instruments, Inc., Melville, NY). Images were captured with a scope-mounted QICAM 12-bit digital camera (Qimaging, Surrey, BC, Canada). Signal intensity for presented figures was uniformly thresholded to minimize background levels.

Real-time quantitative polymerase chain reaction gene expression verification

mRNA was isolated from MG-63 cells and HEK-293 cells using the RNAqueous Kit (Ambion, Austin, TX) according to manufacturer's instructions. Real-time quantitative polymerase chain reaction (RT-qPCR) was conducted using the DNA Engine Opticon 2 (Bio-Rad, Hercules, CA) using the QuantiTect SYBR Green PCR kit (Qiagen, Valencia, CA). For RT-qPCR, 0.5 μg of RNA was reverse-transcribed using iScript cDNA Synthesis Kit (Bio-Rad) according to the manufacturer's instructions. Species-specific primer pairs were used to compare COL1A1 and BGLAP mRNA transcript levels in MG-63 and HEK-293 cells. Calculations were done using the delta delta Ct (ΔΔCt) method, normalized to rRNA 18S expression.

Osteogenic differentiation of ASCs

ASCs were chemically induced for osteogenesis following established protocols.29 Control medium contained DMEM/F-12 (ThermoFisher Scientific), 10% FBS (Zen-Bio), and 1% antibiotic/antimycotic. Osteogenic induction medium included the addition of 1 nM dexamethasone, 21.6 mg/mL β-glycerophosphate, 50 μg/mL ascorbate-2-phosphate, and 10 μg/mL vitamin D3 (Sigma-Aldrich, St. Louis, MO). Two separate, 96-well plates were seeded with 30,000 cells/well in 24 wells per plate using control medium. All cells were treated with 1 ng/mL Hoechst dye to visualize nuclei (Sigma-Aldrich). After 24 h, 12 wells were given 180 μL osteogenic medium while the remaining 12 wells were given 180 μL control medium. Ninety percent of the medium was changed every other day for 21 days.

Beacon treatment and imaging of differentiating ASCs

mRNA-specific beacons were introduced to cells during the 21-day differentiation process. From days 2 to 10, four osteogenic wells and four control wells were treated with ALPL beacon as described previously and imaged daily. For days 8–16, four separate osteogenic and control wells were treated with COL1A1 beacon and imaged daily. For days 17–21, the remaining four osteogenic and control wells were treated with BGLAP beacon and imaged daily. Four fields of view in each well were taken of Hoechst-stained nuclei, fluorescent beacon signals, and bright field images of cells at 10× magnification (16 fields of view total for osteogenic/control conditions). Cells were treated with the appropriate beacons on days 2, 5, 7, 10, 14, and 17 to maintain saturating intracellular concentrations. This re-treatment schedule was chosen based on a beacon persistence assay in living cells that indicated that the signal was diminished by day 4 (Supplementary Fig. S2).

Image processing and analysis

CellProfiler image analysis software was used to generate a MATLAB-based algorithm that relates “child” fluorescent signals to “parent” Hoechst-stained nuclei.30,31 This relation of fluorescence signal to parent nuclei is valid since no extracellular fluorescence was observed in any of the images. The software sets thresholding parameters for each image by first identifying the maximum and minimum values of pixel intensity in each image, then defining all pixel intensities in the lowest 20% as background. The program analyzed each set of images by first counting Hoescht-stained nuclei, which were recognized as ellipsoidal objects with a major axis between 5 and 20 microns (10 and 40 pixels). An area encompassing the nuclear/perinuclear region (25×10 μm ellipsoid) was defined when monitoring fluorescent signal in each cell. Individual fluorescent events were identified by pixel and grouped with the nearest nucleus, thus defining a cell as displaying positive signal or not (Fig. 1). The number of total fluorescing cells was divided by the total number of cells per image, giving a percentage of positively signaling cells for each sample well. Possible sources of error for this method included classifying weak signals as background, attributing signals to the incorrect cell due to overlapping nuclei, and discarding signals that were outside the analysis area. Despite these sources, however, the error rate from this method was only±8%, which is comparable to human error for similar samples. This analysis relied heavily on a preexisting modification to the CellProfiler program offered by the Broad Institute; the program and its modifications can be viewed at http://www.cellprofiler.org, and http://www.cellprofiler.org/CPmanual/RelateObjects.html.

FIG. 1.

FIG. 1.

The percentage of cells with positive signals for specific genes of interest was calculated using image analysis algorithms. A cartoon representation of the analysis is shown illustrating the basic concept (A). Fluorescence signals were assigned to the nearest Hoescht-stained nuclei, establishing parent–child relationships between the two images (B, merged for illustration purposes). “Low/no-signal” cells exhibited no fluorescence events in the region of interest above a minimum threshold value (A1, B1). “Positive” signals varied in type and intensity and included point signals (A2, B2), punctate, compressed speckling (A3, B3), and widespread fluorescence throughout the perinuclear region (A4, B4). All categories were included when calculating percentage of positive cells. Color images available online at www.liebertpub.com/tea

Verification of osteogenesis

Alkaline phosphatase activity in differentiating ASCs was determined according to instructions for the BioVision alkaline phosphatase assay kit (Mountain View, CA). Briefly, four induced and four control wells per plate were either treated with ALPL molecular beacon or left untreated. After 7 days, these cells were lysed in 200 μL lysis buffer. Lysate was stored at −80°C until testing. For analysis, lysates were thawed and centrifuged at 13,000 rpm for 5 min. About 50 μL of the resulting solutions was transferred into individual wells of a 96-well plate and brought to volume with 110 μL of lysis buffer. The remaining 50 μL from each sample was transferred into separate wells on the same plate, brought to volume, and then treated with 20 μL stop solution to act as background controls. A standard curve using 0–0.5 mM alkaline phosphatase was made for quantification of samples. All wells were treated with 10 μL of 5 mM methylumbelliferone-4-phosphate solution for detection. The wells were covered and incubated at room temperature for 30 min, after which stop solution was added to all wells. A spectrofluorometer (Spectramax Plus 384; Molecular Devices) determined the fluorescence of each well at 360 nm/440 nm.

Alizarin Red-S (ARS; Sigma-Aldrich) staining was done for both control and osteogenically induced wells to examine calcified matrix production. ARS (2% in distilled water) was pH-adjusted (4.1–4.3) and filtered through a 0.2-μm-pore filter prior to use. On day 21, beacon-treated wells were fixed with 3.7% paraformaldehyde in PBS (ThermoFisher Scientific). The fixed cell monolayers were washed in distilled water for 5 min, stained with 2% ARS for 20 min, and then thoroughly rinsed. After staining was complete, wells were imaged at 20× magnification using bright field microscopy and a scope-mounted digital camera (Labomed TCM 400; Labomed, Culver City, CA). ARS dye was then eluted with 10% cetylpyridinium chloride (ThermoFisher Scientific) overnight at 4°C, and the optical densities were measured at 540 nm with a spectrofluorometer (Spectramax Plus 384; Molecular Devices).

Statistical analysis

Gene expression patterns were determined using percent expression in differentiating ASC populations (n=4 for ALPL, COL1A1, and BGLAP). Data were analyzed using two-factor analysis of variance (treatment, time; α=0.05) with Fisher's Least Significant Difference post hoc analysis. Osteogenic protein depositions were analyzed using a Student's t-test for control (n=4) and induced (n=4) samples.

Results

Molecular beacon hybridization assay

A 5 μM concentration of ALPL molecular beacon was used to assess hybridization to set concentrations of target sequence. At lower concentrations, the fluorescence intensity increased rapidly, while at the highest concentration the binding was saturated (R2=0.98). Average fluorescence values increased threefold over a target concentration range from 0.5 to 5 μM (Supplementary Fig. S1). A signal-to-noise ratio of 24:1 was calculated based on these measurements, which is consistent with previous studies using molecular beacons.32

Beacon validation and testing

After administration of GAPDH beacon to MG-63 and HEK-293 cells, 97%–99% of cells in each sample population displayed positive signal. MG-63 and HEK-293 cells were treated with molecular beacons corresponding to ALPL, COL1A1, and BGLAP. About 95%–99% of treated MG-63 cells showed robust positive signal for all three beacons (Fig. 2). In treated HEK-293 cells, only 1%–2% of cells showed positive signal. All cells displayed typical morphology and remained spread over a 72-h observation period. To validate the expression levels of osteogenic genes detected by molecular beacons, we performed RT-qPCR to quantify the mRNA levels of COL1A1 and BGLAP in beacon-treated MG-63 and HEK-293 cells. Both mRNA levels were greatly increased in MG-63 cells versus HEK-293 cells.

FIG. 2.

FIG. 2.

Molecular beacons for ALPL (A), COL1A1 (B), and BGLAP (C) were tested in MG-63 (positive control) and HEK-293 (negative control, insets) cells to ensure functionality and specificity. Fluorescence signals were observed throughout the MG-63 cells, while no signals were observed in HEK-293 cells. Color enhanced for presentation purposes. Scale bars: 100 μm. Color images available online at www.liebertpub.com/tea

Molecular beacon signaling in differentiating ASCs

ASCs undergoing osteogenic differentiation were treated with molecular beacons for ALPL, COL1A1, and BGLAP in a stepwise fashion over 21 days. Percentage of positively signaling cells were calculated to monitor temporal gene expression patterns (Fig. 3). ALPL beacon-treated ASCs showed consistent positive signals during the first week of differentiation, starting at 86% on day 2, peaking at 92%–95% on days 3 and 4 (p<0.0001), and decreasing to 63% on day 10. COL1A1 beacon-treated ASCs showed 6% positively signaling cells at day 9, rising to a peak of 91% on day 14 (p<0.0001), and dropping off to 64% by day 16. Less than 1% of the ASC population showed positive BGLAP beacon signal from days 16 to 18, but the percentage of cells increased to 52% by day 19 and reached a peak percentage of 86% on day 21 (p<0.0001), the final day of testing. Control ASCs that were not induced for osteogenesis showed less than 2% fluorescence in all instances.

FIG. 3.

FIG. 3.

Percentage of cells expressing ALPL, COL1A1, and BGLAP (induced, large insets; control, small insets. Scale bar: 50μm) measured daily from days 2 to 21. Expression percentages reflect the expected gene expression profiles for a differentiating population of adipose-derived stem cells (ASCs) (filled symbols). Peak percentages represent the point at which most cells in the induced population displayed positive signal for the gene of interest. These values can also be used as a measure of stem cell purity since only positively differentiating cells should express all three osteogenic genes. Control populations (open symbols) have signal levels close to zero, indicating a lack of osteogenic gene expression. Color images available online at www.liebertpub.com/tea

Verification of osteogenesis

On day 7 of osteogenesis, both beacon-treated and untreated induced ASC samples had ∼300 units of alkaline phosphatase activity per cell, while all control ASCs had ∼5–10 units of activity per cell (Fig. 4; p<0.0001). On day 21, the optical density of Alizarin Red dye eluted from osteogenically induced cells was three times higher than in control samples (Fig. 5A; p<0.0001). Imaging of stained cells revealed a bright crimson color in osteogenically induced samples, while control samples retained little of the dye (Fig. 5B, C).

FIG. 4.

FIG. 4.

Alkaline phosphatase activity in control and induced ASC populations was significantly different, regardless of beacon presence (*p<0.0001). Activity in control samples was nearly zero, while activity in induced samples was two orders of magnitude higher. There were no significant differences in activity between beacon-treated and untreated samples undergoing osteogenesis, serving as both a signal of successful differentiation and evidence for uninterrupted protein synthesis in the presence of molecular beacons. Color images available online at www.liebertpub.com/tea

FIG. 5.

FIG. 5.

Chemically induced ASCs successfully underwent osteogenesis over 21 days. Absorbance values for eluted Alizarin Red-S dye indicated that induced samples deposited three times more calcified matrix than control samples (A, **p<0.0001). Induced (B) and control (C) ASCs stained with Alizarin Red-S for 30 min showed clear, qualitative differences in calcified matrix deposition. Scale bars: 100 μm. Color images available online at www.liebertpub.com/tea

Discussion

This study marks the first time that temporal gene expression patterns in living ASCs undergoing osteogenesis have been measured and quantified. Results suggest that the differentiation process in individual populations becomes more uniform over time. ASCs, like other mesenchymal stem cells, display considerable heterogeneity in their ability to differentiate in a uniform fashion, and few methods exist to investigate this behavior.9 The goal of this study was to assess gene expression patterns during osteogenic differentiation at the subpopulation and population levels, thereby establishing a technique for investigating heterogeneity in differentiating ASC cultures. Differentiating ASC populations displayed clear, dynamic expression patterns of osteogenic genes, assessed by measuring the percentage of positively signaling cells over a 3-week period.

These experiments establish a technique to elucidate gene expression patterns and degree of uniformity during osteogenesis. By testing our system in established cell lines—MG-63 and HEK-293, and using a housekeeping gene, GAPDH—we were able to verify specificity and uptake of the beacon into >97% of the cells in a population. Subsequent measurement of the percentage of fluorescent, and therefore expressing, cells in the differentiating ASC populations showed a distinct pattern of upregulation followed by a slow decline in the number of expressing cells. Initial expression levels began lower for all genes (<10% for COL1A1 and BGLAP) and then slowly rose to a peak of >90% in all cases. Thus, we were able to identify the peak expression times of characteristic genes in osteogenesis. An important point to note about this approach is that although the same samples were imaged each day, cellular proliferation and migration made it difficult to assess expression levels in the exact same cells over time. Hence, the “peak” in gene expression seen with our experiments reflects the time when the most cells have begun to express the gene of interest.

Interestingly, the percentage of positive signal cells increases in larger jumps during the later phases of osteogenesis. The ERK/Akt pathways have been implicated in osteogenic lineage commitment and have previously been induced by culture in type I collagen–coated flasks, implying that the presence of type I collagen in the extracellular environment may assist in committing a cell to the osteogenic lineage.33 The data for ALPL gene expression show data points very close together in their slow rise and fall from the peak percentage point, with a difference of only a few percentage points between days 3 and 4. But as we map patterns for COL1A1 and BGLAP, we see that the percentages jump from around 50% to 75% and from 60% to about 90%, respectively, over a single day. From these patterns, we could infer that the cells lock into osteogenesis after an initial induction period during which type I collagen is upregulated. While differentiation begins slowly in the first week, it quickly picks up speed and momentum, with more and more cells joining the differentiation process and settling into a more uniform gene expression pattern in later weeks.

Notably, each cell appears to be on its own track during differentiation. Some cells express the genes of interest very early on; for example, around 85% of cells express ALPL on day 2, though the peak percentage of cells expressing ALPL occurs on day 4. Cells that express genes early may also stop expressing genes sooner, which would explain the gradual decrease from the peak percentage as opposed to a steep drop. Additionally, when expression percentages are analyzed on a per-well basis, the changes in expression patterns from well to well are most apparent for ALPL, begin to even out for COL1A1, and are almost nonexistent for BGLAP (Supplementary Fig. S3). These expression patterns serve as additional evidence for the heterogeneity of stem cell populations, given the apparent differences in variation in gene expression from well to well during osteogenesis. Additionally, it is important to note the possibility that even when nearly every cell from a population is differentiating, not all of them so do in the same timeframe.

Use of molecular beacons can be complicated by a number of possible artifacts. One common concern is false signal due to nonspecific binding, probe instability, probe background noise, and probe degradation.14,20,3235 Nonspecific binding, or binding of probes to unintended targets, was not apparent in negative control HEK-293 cells. Likewise, noninduced ASCs treated with beacons showed almost no fluorescence, further indicating the absence of false-positive signals. We feel that these experiments are convincing evidence that nonspecific binding did not occur in our systems. Artifacts due to probe instability, background noise, and degradation were also not observed in the negative control cells. If the probes were unstable or degrading, a signal would be noticeable, either as a point source or by elevated background levels. Positive signal intensities were much higher in the positive control samples (MG-63 cells) than in any of the negative controls.26 Independent of live-cell experiments, the beacon hybridization assay measured the background signal of probe alone, and data were normalized to this value, which was negligible compared with positive signals obtained in the same assay.

To confirm that osteogenic differentiation was successful, protein activity and deposition tests were performed on experimental samples. The alkaline phosphatase activity assay revealed protein activity two orders of magnitude higher in osteogenic samples than for controls, in both beacon-treated and untreated samples. There were no significant differences in activity between treated and untreated, induced samples, indicating that beacon hybridization to mRNA did not interfere with protein synthesis. Additionally, optical densities of eluted AR dye showed three times higher levels in osteogenic samples compared with controls, and imaging of alizarin-red-stained cells revealed a scarlet coloring in osteogenic samples while control samples remained uncolored. Alizarin red staining for calcified matrix is a commonly used and well-established method for assessing osteogenic differentiation. Staining with alizarin red and testing for activity with the alkaline phosphatase assay were meant to ensure that differentiation had indeed occurred successfully, which would not have been possible had significant gene knockdown occurred.

While these methods are commonly used to assess success rates of osteogenic differentiation, they possess pertinent limitations that our molecular beacon-based analyses overcome. Both of these assays collect data from the differentiating population as a whole, meaning that whether an individual cell has successfully differentiated or not, it is lumped into the analysis. These nondifferentiated cells obscure contributions from differentiating cells and can yield unclear data. Additionally, quantification of the data is oversimplified by reporting each parameter on a per-cell basis; this again makes the incorrect assumption that all cells are contributing equally to the assay. The use of molecular beacons to quantify the rate of differentiation allows us to make accurate subpopulation- and population-based observations about a study by accounting for the heterogeneity present in the sample.

A practical limitation of this technique is its inability to quantify gene copy numbers within individual cells, akin to “gold standard,” real-time PCR analyses. Theoretically, signal intensity from a beacon corresponding to an upregulated gene could be compared to the fluorescence intensity of a beacon for a housekeeping gene, such as 18S rRNA or GAPDH, but intensities of different fluorophores vary and are difficult to compare from cell to cell. Other limitations include the slight false-positive rate, indicated by the 1%–2% positive signal seen in HEK-293 cells, and the false-negative rate, indicated by the 1%–2% negative signal seen in MG-63 cells. These may be delivery method dependent, and thus could be solved for more sensitive assays. However, false-positive and -negative rates of 1%–2% were sufficiently small for the current study and had a minimal impact on the interpretation of the results.

A comparison between RT-qPCR and beacon data would provide interesting insight into the strength and weaknesses of both techniques. As such, we conducted validation tests on two cell lines with well-defined mRNA expression: HEK-293 and MG-63 cells. MG-63 (osteosarcoma) cells constitutively express our osteogenic genes, whereas HEK-293 (human embryonic kidney) cells do not. RT-qPCR data were collected for COL1A1 and BGLAP to verify that high mRNA levels corresponded to extensive positive beacon signals in the same samples (Supplementary Fig. S4). As expected, MG-63 cells had much higher expression levels for osteogenic genes than HEK-293 cells, corroborating visual evidence of beacon signal differences. Additionally, many studies have been performed that assess the RT-qPCR–based gene expression patterns of osteogenically differentiating ASCs.3436 One of the strengths of RT-qPCR is that it allows for quantitative comparisons between groups, whereas molecular beacons only provide semi-quantitative comparisons associated with the number of cells expressing a given gene. In these validation experiments, we can quantify the difference in osteogenic gene copies using RT-qPCR between MG-63 and HEK-293 cells, but we cannot describe how that gene expression is distributed across the sample. Molecular beacons, while not providing hard mRNA copy numbers, can describe which portions of a population are expressing the genes of interest. Additionally, beacons can be used to repeatedly assess the same cell population over time. This is not possible for RT-qPCR, which requires lysing of the sample to obtain mRNA for reverse transcription. Well-to-well variations in RT-qPCR data could also be significant for heterogeneous samples, resulting in disparate differentiation behaviors for separate sample groups over time. To illustrate the value of having repeated measurement capabilities, we included a Supplementary figure showing the variation in the percentage of cells displaying positive signals for each mRNA target on a well-by-well basis (Supplementary Fig. S3). While not large, there are clear differences among the wells (e.g., ALPL, days 5 and 7). These differences lessen as differentiation proceeds (e.g., BGLAP shows uniform expression on almost all days). Molecular beacons provide opportunities to analyze data in this way, whereas conventional methods do not.

Increases in signal were attributed to upregulation of the genes of interest, while decreases were attributed to a cessation or downregulation of gene expression for the target sequence. However, other possibilities do exist. Beacon degradation was discussed previously as one but was shown to be only a possible contributor based on the low false-positive rates observed in HEK-293 cells. A second possibility involves cellular proliferation; as cells divide, the cytoplasmic concentration of beacons decreases, resulting in less fluorescence per cell. To counter this, cells were treated with beacon at multiple time points, not just initially, to ensure sufficient beacon concentration in all cells. Interestingly, other groups have reported the persistence of nondegraded, chemically modified molecular beacons up to 21 days after treatment, indicating their resiliency.37

The uniformity of differentiation apparent from the results indicates that the cell population used in this study contained very little heterogeneity. Freshly isolated ASCs have shown more heterogeneity and lower differentiation capability than a serially passaged population.38 As a result, the peak percentages of cells expressing genes of interest during differentiation are very high, between 85%–95%. In freshly isolated, unpurified populations of ASCs, heterogeneity is expected to be more prevalent and could yield peak percentages lower than those seen here.

The current study revealed dynamic gene expression patterns for three characteristic genes in osteogenesis and established a method by which gene expression can be assessed in live cells over extended periods, findings that are useful for both continued studies of differentiation as well as elucidation of uniformity in differentiating populations. Molecular beacons can be developed for genes characteristic of other lineages and used to generate similar gene expression timelines. Additionally, stem cell populations from diverse sources could be monitored in this way to compare their heterogeneity and differentiation capability. Likewise, these beacons could also be used to determine the effectiveness of other differentiation media or stimulation techniques in live-cell, experimental designs.

Conclusions

Stem cell heterogeneity continues to be a research obstacle for both basic science experiments and clinical applications. Tissue-engineered constructs that rely on stem cells face this problem regularly, though it is largely undefined. This study proposes a novel approach to investigate gene expression in stem cell populations, which can be applied to assessments of uniformity in these populations. Live, osteogenically differentiating ASCs were assessed repeatedly over a 21-day period to obtain temporal gene expression data that help elucidate up- and downregulation of key, osteogenic genes in response to induction medium. Additionally, the percentage of actively differentiating cells in a population was quantified, providing a novel method to measure and define heterogeneity in stem cell populations. The demonstrated molecular beacon technology allows for acquisition of live-cell, gene expression data to clarify aspects of heterogeneity in these potentially transformative cell populations.

Author Contributions

E.M.D. and H.V.D. designed the study, analyzed all molecular beacon and protein deposition data, and wrote the article. H.V.D. conducted all live-cell molecular beacon and protein deposition experiments. I.S.V. conducted and analyzed the in vitro hybridization assay. Q.C. and C.J. designed and analyzed all RT-qPCR experiments.

Supplementary Material

Supplemental data
Supp_Fig1.pdf (36.8KB, pdf)
Supplemental data
Supp_Fig2.pdf (43.8KB, pdf)
Supplemental data
Supp_Fig3.pdf (60KB, pdf)
Supplemental data
Supp_Fig4.pdf (36.8KB, pdf)

Acknowledgments

The authors thank Kian Adabi, Olivia S. Beane, Vera C. Fonseca, Manisha Kanthilal, and Nicholas R. Labriola for helpful discussions and contributions. This work was supported in part by National Institutes of Health grants AR054673 and GM104937.

Disclosure Statement

No competing financial interests exist.

References

  • 1.Donizetti-Oliveira C. Semedo P. Burgos-Silva M. Cenedeze M.A. Malheiros D.M. Reis M.A., et al. Adipose tissue-derived stem cell treatment prevents renal disease progression. Cell Transplant. 2012 doi: 10.3727/096368911x623-925. Epub ahead of print. [DOI] [PubMed] [Google Scholar]
  • 2.Gimble J.M. Grayson W. Guilak F. Lopez M.J. Vunjak-Novakovic G. Adipose tissue as a stem cell source for musculoskeletal regeneration. Front Biosci (Schol Ed) 2011;3:69. doi: 10.2741/s133. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Guilak F. Lott K.E. Awad H.A. Cao Q. Hicok K.C. Fermor B., et al. Clonal analysis of the differentiation potential of human adipose-derived adult stem cells. J Cell Physiol. 2006;206:229. doi: 10.1002/jcp.20463. [DOI] [PubMed] [Google Scholar]
  • 4.Guilak F. Estes B.T. Diekman B.O. Moutos F.T. Gimble J.M. 2010 Nicolas Andry Award: Multipotent adult stem cells from adipose tissue for musculoskeletal tissue engineering. Clin Orthop Relat Res. 2010;468:2530. doi: 10.1007/s11999-010-1410-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Locke M. Feisst V. Dunbar P.R. Concise review: human adipose-derived stem cells: separating promise from clinical need. Stem Cells. 2011;29:404. doi: 10.1002/stem.593. [DOI] [PubMed] [Google Scholar]
  • 6.Yoshimura K. Shigeura T. Matsumoto D. Sato T. Takaki Y. Aiba-Kojima E., et al. Characterization of freshly isolated and cultured cells derived from the fatty and fluid portions of liposuction aspirates. J Cell Physiol. 2006;208:64. doi: 10.1002/jcp.20636. [DOI] [PubMed] [Google Scholar]
  • 7.Gimble J.M. Guilak F. Nuttall M.E. Sathishkumar S. Vidal M. Bunnell B.A. In vitro differentiation potential of mesenchymal stem cells. Transfus Med Hemother. 2008;35:228. doi: 10.1159/000124281. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Zuk P.A. Zhu M. Mizuno H. Huang J. Futrell J.W. Katz A.J., et al. Multilineage cells from human adipose tissue: implications for cell-based therapies. Tissue Eng. 2001;7:211. doi: 10.1089/107632701300062859. [DOI] [PubMed] [Google Scholar]
  • 9.Russell K.C. Phinney D.G. Lacey M.R. Barrilleaux B.L. Meyertholen K.E. O'Connor K.C. In vitro high-capacity assay to quantify the clonal heterogeneity in trilineage potential of mesenchymal stem cells reveals a complex hierarchy of lineage commitment. Stem Cells. 2010;28:788. doi: 10.1002/stem.312. [DOI] [PubMed] [Google Scholar]
  • 10.Tyagi S. Kramer F.R. Molecular beacons: probes that fluoresce upon hybridization. Nat Biotechnol. 1996;14:303. doi: 10.1038/nbt0396-303. [DOI] [PubMed] [Google Scholar]
  • 11.Manganelli R. Tyagi S. Smith I. Real time PCR using molecular beacons: a new tool to identify point mutations and to analyze gene expression in mycobacterium tuberculosis. Methods Mol Med. 2001;54:295. doi: 10.1385/1-59259-147-7:295. [DOI] [PubMed] [Google Scholar]
  • 12.Mhlanga M.M. Vargas D.Y. Fung C.W. Kramer F.R. Tyagi S. tRNA-linked molecular beacons for imaging mRNAs in the cytoplasm of living cells. Nucleic Acids Res. 2005;33:1902. doi: 10.1093/nar/gki302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Rhee W.J. Bao G. Simultaneous detection of mRNA and protein stem cell markers in live cells. BMC Biotechnol. 2009;9:30. doi: 10.1186/1472-6750-9-30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Santangelo P.J. Nix B. Tsourkas A. Bao G. Dual FRET molecular beacons for mRNA detection in living cells. Nucleic Acids Res. 2004;32:e57. doi: 10.1093/nar/gnh062. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Tsourkas A. Bao G. Shedding light on health and disease using molecular beacons. Brief Funct Genomic Proteomic. 2003;1:372. doi: 10.1093/bfgp/1.4.372. [DOI] [PubMed] [Google Scholar]
  • 16.Baker M.B. Bao G. Searles C.D. In vitro quantification of specific microRNA using molecular beacons. Nucleic Acids Res. 2012;40:e13. doi: 10.1093/nar/gkr1016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Kim E.K. Lim S. Park J.M. Seo J.K. Kim J.H. Kim K.T., et al. Human mesenchymal stem cell differentiation to the osteogenic or adipogenic lineage is regulated by AMP-activated protein kinase. J Cell Physiol. 2012;227:1680. doi: 10.1002/jcp.22892. [DOI] [PubMed] [Google Scholar]
  • 18.Abrahamsson C.K. Yang F. Park H. Brunger J.M. Valonen P.K. Langer R., et al. Chondrogenesis and mineralization during in vitro culture of human mesenchymal stem cells on three-dimensional woven scaffolds. Tissue Eng Part A. 2010;16:3709. doi: 10.1089/ten.tea.2010.0190. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Estes B.T. Diekman B.O. Guilak F. Monolayer cell expansion conditions affect the chondrogenic potential of adipose-derived stem cells. Biotechnol Bioeng. 2008;99:986. doi: 10.1002/bit.21662. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Mathews D.H. Turner D.H. Zuker M. RNA secondary structure prediction. In: Beaucage S.L., et al., editors. Current Protocols in Nucleic Acid Chemistry. Hoboken, NJ: John Wiley & Sons; 2007. Chapter 11, Unit 11.2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Zuker M. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res. 2003;31:3406. doi: 10.1093/nar/gkg595. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Mount D.W. Using the Basic Local Alignment Search Tool (BLAST) CSH Protoc. 2007;2007 doi: 10.1101/pdb.top17. pdb top17. [DOI] [PubMed] [Google Scholar]
  • 23.Altschul S.F. Gish W. Miller W. Myers E.W. Lipman D.J. Basic local alignment search tool. J Mol Biol. 1990;215:403. doi: 10.1016/S0022-2836(05)80360-2. [DOI] [PubMed] [Google Scholar]
  • 24.Bao G. Rhee W.J. Tsourkas A. Fluorescent probes for live-cell RNA detection. Annu Rev Biomed Eng. 2009;11:25. doi: 10.1146/annurev-bioeng-061008-124920. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Bratu D.P. Catrina I.E. Marras S.A. Tiny molecular beacons for in vivo mRNA detection. Methods Mol Biol. 2011;714:141. doi: 10.1007/978-1-61779-005-8_9. [DOI] [PubMed] [Google Scholar]
  • 26.Rothem D.E. Rothem L. Dahan A. Eliakim R. Soudry M. Nicotinic modulation of gene expression in osteoblast cells, MG-63. Bone. 2011;48:903. doi: 10.1016/j.bone.2010.12.007. [DOI] [PubMed] [Google Scholar]
  • 27.Mueller S.M. Glowacki J. Age-related decline in the osteogenic potential of human bone marrow cells cultured in three-dimensional collagen sponges. J Cell Biochem. 2001;82:583. doi: 10.1002/jcb.1174. [DOI] [PubMed] [Google Scholar]
  • 28.Santangelo P.J. Nitin N. Bao G. Direct visualization of mRNA colocalization with mitochondria in living cells using molecular beacons. J Biomed Opt. 2005;10:44025. doi: 10.1117/1.2011402. [DOI] [PubMed] [Google Scholar]
  • 29.Bunnell B.A. Estes B.T. Guilak F. Gimble J.M. Differentiation of adipose stem cells. Methods Mol Biol. 2008;456:155. doi: 10.1007/978-1-59745-245-8_12. [DOI] [PubMed] [Google Scholar]
  • 30.Lamprecht M.R. Sabatini D.M. Carpenter A.E. CellProfiler: free, versatile software for automated biological image analysis. BioTechniques. 2007;42:71. doi: 10.2144/000112257. [DOI] [PubMed] [Google Scholar]
  • 31.Carpenter A.E. Jones T.R. Lamprecht M.R. Clarke C. Kang I.H. Friman O., et al. CellProfiler: image analysis software for identifying, quantifying cell phenotypes. Genome Biol. 2006;7:R100. doi: 10.1186/gb-2006-7-10-r100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Vet J.A. Marras S.A. Design and optimization of molecular beacon real-time polymerase chain reaction assays. Methods Mol Biol. 2005;288:273. doi: 10.1385/1-59259-823-4:273. [DOI] [PubMed] [Google Scholar]
  • 33.Tsai K.S. Kao S.Y. Wang C.Y. Wang Y.J. Wang J.P. Hung S.C. Type I collagen promotes proliferation and osteogenesis of human mesenchymal stem cells via activation of ERK and Akt pathways. J Biomed Mater Res Part A. 2010;94:673. doi: 10.1002/jbm.a.32693. [DOI] [PubMed] [Google Scholar]
  • 34.Frank O. Heim M. Jakob M. Barbero A. Schafer D. Bendik I., et al. Real-time quantitative RT-PCR analysis of human bone marrow stromal cells during osteogenic differentiation in vitro. J Cell Biochem. 2002;85:737. doi: 10.1002/jcb.10174. [DOI] [PubMed] [Google Scholar]
  • 35.Pittenger M.F. Mackay A.M. Beck S.C. Jaiswal R.K. Douglas R. Mosca J.D., et al. Multilineage potential of adult human mesenchymal stem cells. Science. 1999;284:143. doi: 10.1126/science.284.5411.143. [DOI] [PubMed] [Google Scholar]
  • 36.Zuk P.A. Zhu M. Ashjian P. De Ugarte D.A. Huang J.I. Mizuno H., et al. Human adipose tissue is a source of multipotent stem cells. Mol Biol Cell. 2002;13:4279. doi: 10.1091/mbc.E02-02-0105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.King F.W. Liszewski W. Ritner C. Bernstein H.S. High-throughput tracking of pluripotent human embryonic stem cells with dual fluorescence resonance energy transfer molecular beacons. Stem Cells Dev. 2011;20:475. doi: 10.1089/scd.2010.0219. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Estes B.T. Wu A.W. Storms R.W. Guilak F. Extended passaging, but not aldehyde dehydrogenase activity, increases the chondrogenic potential of human adipose-derived adult stem cells. J Cell Physiol. 2006;209:987. doi: 10.1002/jcp.20808. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental data
Supp_Fig1.pdf (36.8KB, pdf)
Supplemental data
Supp_Fig2.pdf (43.8KB, pdf)
Supplemental data
Supp_Fig3.pdf (60KB, pdf)
Supplemental data
Supp_Fig4.pdf (36.8KB, pdf)

Articles from Tissue Engineering. Part A are provided here courtesy of Mary Ann Liebert, Inc.

RESOURCES