Attempts have been made to develop cell and gene therapies using the adult stem cells from bone marrow referred to as mesenchymal stem cells or marrow stromal cells (MSCs). However, the results have been variable in part because there are no standardized protocols for preparing and characterizing MSCs. In the experiments presented here, we developed a standardized assay by light scattering to measure the content of rapidly self-renewing cells (RS cells) in preparations of MSCs. The assay quickly identifies preparations of MSCs that replicate rapidly in subsequent culture. In addition, the standardized assay enabled us to isolate RS cells that were up to 90% clonogenic and that generated single cell–derived colonies that differentiated into either mineralizing cells or adipocytes with appropriate additions to the medium.
One possible strategy for therapy with stem cells is to use the adult stem cells from bone marrow stroma referred to as mesenchymal stem cells or marrow stromal cells (MSCs) [1–8]. Human MSCs are relatively easy to obtain from a small aspirate of bone marrow under conditions in which they retain the potential to differentiate into multiple cell lineages that include osteoblasts, adipocytes, chondrocytes, myoblasts, and early progenitors of neural cells. MSCs proliferate rapidly and largely retain their multipotentiality for differentiation after expansion in culture [9–11]. However, cultures of expanded cells are heterogeneous in morphology, and they lose multipotentiality as they are replated for six or seven passages [9–12]. The cells are also highly sensitive to plating density, and early progenitors are rapidly lost if the cultures are grown to confluency . Additionally, there is considerable variation in the proportion of early progenitors recovered from different samples of bone marrow, even when the samples are obtained from the same donor at the same time . Currently, there are no surface epitopes that are useful for distinguishing early progenitors from mature cells in the cultures . For these reasons, it is obviously important to devise standardized protocols for isolating and characterizing MSCs.
We previously extended earlier observations  and identified a subfraction of small and rapidly self-renewing cells (RS cells) in early-passage, low-density cultures of human MSCs [9, 10]. The RS cells were characterized by low forward scatter (FSlo) and low side scatter (SSlo) of light. In the experiments described here we demonstrate that FSlo/SSlo MSCs can be isolated by fluorescence-activated cell sorter (FACS) analysis and that the isolated cells are up to 90% clonogenic. Essentially all of the FSlo/SSlo cells differentiate into either osteoblasts or adipocytes. We also demonstrate that a rapid, standardized assay for FS/SS may be useful to identify preparations of MSCs enriched for RS cells that will expand rapidly during subsequent passage in culture. The use of the assay should help to resolve discrepancies in data obtained by different laboratories with apparently similar preparations of MSCs.
Human MSCs were prepared as described previously [10, 11]. In brief, nucleated cells were isolated with a density gradient (Ficoll-Paque; Pharmacia) from 2-ml human bone marrow aspirated from the iliac crests of normal volunteers under a protocol approved by an Institutional Review Board. All of the nucleated cells (30–70 million) were plated in a 145-cm2 dish in 20-ml complete culture medium (CCM) that was prepared with 1 liter of alpha minimum essential media (α-MEM) (GIBCO-BRL, Rockville, MD), 200-ml fetal bovine serum (FBS; lot-selected for rapid growth of MSCs; Atlanta Biologicals, Lawrenceville, GA), 100 units/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine (GIBCO-BRL). After 24 hours at 37°C in 5% CO2, nonadherent cells were discarded, and incubation in fresh medium was continued for 4 days. The cells were lifted with 0.25% trypsin and 1 mM EDTA for 5 minutes at 37°C and then replated at 50 cells per cm2 in an interconnecting system of culture flasks (6320 cm2; Nunc Cell Factory, Roskild, Denmark). Parallel 145-cm2 dishes (Nunc) were plated under the same conditions as pilot samples to observe expansion of the cells. After cells in the pilot samples expanded 500- to 1,000-fold (7–9 days), the cells in the interconnecting flasks were lifted with trypsin/EDTA and frozen at 106 cells/ml in liquid nitrogen as passage 1 cells. Alternatively, some samples were plated at high densities of 5,000 cells per cm2 and incubated for 7–9 days prior to freezing. For most of the experiments here, a frozen vial of 106 passage 1 cells was thawed, plated in 20 ml of CCM in a 145-cm2 dish, and incubated for 1 or 3 days to recover viable passage 2 cells. The passage 2 cells were harvested with trypsin/EDTA and then incubated at 50–100 cells per cm2 for 4–10 days and lifted with trypsin/EDTA to obtain passage 3 cells. The cells were then repeatedly plated without freezing at 50–1,000 cells per cm2 for 4–9 days to obtain passage 6. During each step of expansion, the medium was changed every 3–5 days.
For measurement of FS and SS, a closed stream flow cytometer (Epics XL with ADC; Beckman Coulter Inc., Fullerton, CA) was standardized using microbeads with known diameters (7, 10, 15, and 20 μ; Dynosphere Uniform Microspheres; Bangs Laboratories Inc., Fisher, IN). The gains and voltages on the photomultiplier tubes were adjusted so that the mean value of the FS peak for the 20-μ bead was about 650, and the peak of the SS for the 7-μ bead was about 450. With these settings, the standard deviation for FS of the largest bead was less than ± 0.4 % (n = 3) of the mean, and the slope of FS on a linear scale of 0 to 1,023 was at least 41. Preliminary experiments (not shown) indicated that the values for log(%G/%T) were not substantially changed with samples containing 0.5 or 1 million MSCs per ml when the following parameters were varied: (a ) the flow rate was 250, 500, or 900 cells per second; (b) the FS was assayed with 67 or 122 volts and a gain of 2, or with 353 volts and a gain of 1; (c) the peak for FS of the 20-μ bead was set at 550, 650, or 750; and (d) the peak for SS of the 7-μ bead was set at 350, 450, or 550. For the assays, cells were lifted with trypsin/ EDTA, washed with CCM by centrifugation at 450× g for 10 minutes to neutralize trypsin, counted on a hemocytometer, suspended in PBS at 4°C at a concentration of about 0.5 million cells per ml, and then assayed shortly thereafter. The microbeads could not be used to standardize the FS/SS assays on the open stream FACS instrument (FACSVantage; Becton, Dickinson, Franklin Lakes, NJ) because of the inherent instabilities of the instrument. To isolate distinct fractions on the basis of FS and SS (Fig. 1C), we first divided the Annexin V− events into four quadrants on the basis of FS and SS. We then offset the sort gates from the boundaries.
For the single-cell colony-forming unit (sc-CFU) assay, the accuracy of sorting single cells into each well of a microtiter plate was verified routinely by sorting fluorescent beads (Flowchek; Beckman Coulter) into a test plate and examining the wells with an epifluorescence microscope. The microtiter plates with one MSC per well were incubated in 0.15 ml CCM with change of medium every 4–5 days. After 2 weeks, the medium was removed, and the wells were washed with PBS. Parallel plates were used for assays of sc-CFUs, for differentiation into osteoblasts, and for differentiation into adipocytes. For sc-CFU assays, the samples were incubated with 1 ml crystal violet in methanol for 1 minute and then washed with water, and colonies with diameters greater than 1 mm were counted by microscopy (4× phase contrast). For assay of osteogenic differentiation, the microtiter plates were incubated in 0.15 ml per well of α-MEM supplemented with 10% FBS, 100 units/ml penicillin, 100 μg/ml streptomycin, 1 mM L-glutamine, 1 × 108 M dexamethasone (Sigma, St. Louis, MO), 10 μM β-glycerol phosphate (Sigma), and 50 μM ascorbic acid–2 phosphate (Sigma) for 21 days. The medium was changed every 4–5 days. The plates were fixed in 10% neutral buffered formalin for 20 minutes and stained with 0.5% alizarin red S for 5–10 minutes (pH 4.1; Sigma). The plates were washed at least three times with water, and the mineral deposits were visualized by a dark red stain. For adipogenic differentiation, the microtiter plates were incubated with 0.15 ml per well of α-MEM supplemented with isobutyl methylxanthine (IBMX; Sigma), and 60 μM indomethacin (Sigma) for 21 days. The medium was changed every 4–5 days. The samples were then fixed for 20 minutes in 10% neutral-buffered formalin and stained for 30 minutes with oil red O (stock: 0.5% in isopropanol, mixed 3 parts stock to 2 parts water and filtered through a 0.2-μm filter; Sigma). After thoroughly washing the plates with water, the lipid vacuoles were identified as bright red inclusions within the cells.
Genotyping of single cell–derived colonies was carried out by polymerase chain reaction (PCR) amplification of sequences spanning single-nucleotide polymorphisms (SNPs) in exon 45 of COL1A1 (collagen, type I, alpha 1), exon 25 of COL1A2, and exon 9 of COL2A1 . The reaction volume was 20 μl, and the conditions were 1 cycle at 95°C for 10 min, 34 cycles at 95°C for 25 seconds, and 34 cycles at 72°C for 35 seconds. Heteroduplexes were formed by incubation at 72°C for 4 minutes, 95°C for 4 minutes, and then 68°C for 30 minutes. The heteroduplexes were assayed by conformation-sensitive gel electrophoresis [13, 14].
Staining for senescence-associated β-galactosidase was carried out with one commercial kit (ImaGene Green CFDG lacZ Gene Expression Kit; Molecular Probes Inc., Eugene, OR) and for Annexin V with a second commercial kit (Sigma). Cells treated with the Annexin V-FITC (fluorescein isothiocyanate) were maintained at 4°C to prevent aggregation and death from the presence of calcium and reagent-induced toxicity.
Statistical analyses were performed in GraphPad, PRISM 3.0 (ANOVA, regression, post tests) or Microsoft Excel (%CV, t-test).
In initial experiments, we used late-passage MSCs to further test whether MSCs can be subfractionated by FS and SS (Fig. 1). The uncorrected plot of FS versus SS indicated three distributions of events (Fig. 1A). Staining with Annexin V-FITC demonstrated that the events in the upper left of the plot were cell debris and dead cells (R1 in Fig. 1B). To obtain subfractions of cells, the Annexin V+ events were gated out, and four subpopulations were defined on the basis of FS and SS (Fig. 1C). Cells defined by their light-scattering characteristics—for example, FSlo/SSlo and FShi/SShi (R2 and R3 in Fig. 1E and the four regions in Fig. 1C)—were isolated by FACS. Reassay of the two subfractions from Figure 1E demonstrated that they were distinct in terms of their light-scattering properties and the separations were reproducible (Fig. 1D and 1F). Subsequent experiments demonstrated that the proportion of Annexin V+ events was only about 1% in early-passage, low-density cultures that were harvested by timed digestion with trypsin/EDTA. Therefore, gating to exclude Annexin V+ cells was unnecessary for early-passage, low-density cultures.
To define the properties of the subfractions, cells from each of the separate regions indicated in Figure 1C were sorted into 96-well plates at 1 cell per well (three separate experiments). The sorted cells were then incubated to develop an assay for sc-CFUs (Fig. 2C) and clonal differentiation into adipocytes (not shown) and osteoblasts (Fig. 2B). Cells that were FSlo/SSlo gave a significantly higher value in the sc-CFU assay than cells from any of the other three regions (Fig. 2C). One-way ANOVA followed by Newman-Keuls multiple comparisons was used to test for the significance of the difference in clonogenicity observed for the four regions. The difference was significant with p < .05. The clonogenicity values for FSlo/SSlo cells from different preparations ranged from 80%–90%. The assays of parallel 96-well plates incubated in either osteogenic or adipogenic medium indicated that essentially all the colonies generated from FSlo/SSlo cells differentiated into mineralizing cells (Figs. 2B, C) or adipocytes (not shown). As reported elsewhere , single cell–derived colonies obtained by low-density plating of the cells were previously shown to differentiate into mineralizing cells, adipocytes, and chondro-cytes. Therefore, the FSlo/SSlo cells correspond to the sub-fraction previously referred to as RS cells .
To confirm that the sc-CFU assay accurately measured single cell–derived colonies, genotyping experiments were carried out with SNPs and heteroduplex analyses of PCR products spanning the SNPs [13, 14]. MSCs were prepared from three different donors with distinct genotypes for SNPs in either of the two genes for type I collagen (COL1A1 and COL1A2) or the gene for type II collagen (COL2A1). Pairwise mixtures containing equal numbers of cells from two distinguishable donors were used for the sc-CFU assays, and DNA from the resulting colonies was assayed. In one experiment, colonies generated by mixtures of MSCs from donors 87 and 146 were homogeneous for either the SNP in exon 9 of COL1A1 found in donor 146 or the SNP in exon 25 of COL1 A2 found in donor 87 (Fig. 3, upper panel). In a second experiment, colonies generated by mixtures of MSCs from donors 107 and 146 were homogeneous for the SNP in exon 25 of COL1A2 found in donor 107 (Fig. 3, lower panel). Therefore each of the colonies had arisen from a single donor in 13 out of 13 colonies tested.
The sc-CFU assay was compared with the commonly used CFU assay in which cells are plated in tissue-culture dishes at low densities of about two cells per cm2 (Fig. 4). The coefficient of variation was used as a measure of reproducibility. The coefficient of variation for the sc-CFU assay was about one-third the coefficient of variation for the commonly employed assay in which the cells were plated at low density (14.6 versus 4.52; n = 4 for each experiment). A t-test for the difference in reproducibility assuming unequal variance (Excel program) indicated a p value < .02.
To demonstrate further the differences between the cells based on light-scattering properties, cells that showed the greatest difference in FS and SS were sorted from gates R3 and R5 in Figure 1C or R2 and R3 in Figure 1E, and the mRNAs were assayed with microarrays. The data were first analyzed to select the genes whose signal intensities showed the greatest difference between the two fractions. Of the 13 genes that showed the greatest difference, eight were related to cell proliferation (Table 1). Five genes that are expressed in proliferative cells were expressed at higher levels in FSlo/SSlo cells (Fig. 5A). In contrast, three genes that are expressed in less proliferative cells were expressed at lower levels in the FSlo/SSlo cell fraction (Fig. 5B).
Table Table 1.. Identity of genes shown in Figure 5
In additional attempts to identify subfractions of MSCs, we stained for β-galactosidase, a marker for senescent cells . The stain was not useful because the number of β-galactosidase+ events in late-passage cultures (passage 6) was only twice the 1.5% threshold established for early-passage cultures (passage 3) enriched for RS cells. In another series of experiments, we attempted to identify cells comparable to those defined as side population cells (SP cells) from bone marrow [16, 17]. The dye (Hoechst 33342) used to identify SP cells proved to be toxic to human MSCs. In samples with 106 human MSCs/ml, essentially all the cells were killed after 18 minutes at 37°C in the presence of the recommended concentration of dye (5 μg/ml) [16, 17] and before dye exclusion could be assessed.
To standardize the assay for FS and SS, we used a closed stream flow cytometer (Beckman-Coulter XL with ADC) and calibrated the system with microbeads of known dimensions (Fig. 6). The microbeads, ranging in size from 7–20 μ, generated a linear and reproducible standard curve for FS (Fig. 6B). To standardize the assay, we adjusted the gains and the voltages on the photomultiplier tubes so that the FS peak for the 20-μ bead was 650 and the SS peak for the 7-μ bead was 450. To compare different samples, we compared early-passage, low-density cultures with late-passage, high-density cultures in terms of the distribution of events assayed in different regions, as defined in Figures 6A and 6C. We found that the most sensitive discriminator among the samples was a flow parameter defined as log of % Annexin V− events in region G divided by % events in region T. For example, the flow parameter for low-density cultures of passage 3 cells was 1.71 (Fig. 6A) and 0.44 for high-density cultures of passage 5 cells (Fig. 6C). Therefore, the flow parameter appears to reflect the relative content of RS cells in the preparations [10, 11].
To assess the predictive value of the FS/SS assay, we compared 19 different preparations of human MSCs that had been isolated and expanded with slightly different protocols over a 2-year period and then stored frozen (Fig.7). The thawed vials of cells were plated at high density for 1 day to recover viable cells (P2), and then separate aliquots were assayed for FS/SS and for expansion over 4 days after plating at 100 cells/cm2 (harvested cells are P3). The data indicated that all 10 of the preparations with a flow parameter greater than 0.5 expanded fivefold or greater in 4 days. In contrast, 6 preparations with a flow parameter of less than 0.5 expanded less than fivefold. Three preparations appeared to be outliers in the assay, because they had flow parameters less than 0.5 but expanded more than fivefold. The outliers were not explained by age differences of the donors (20, 27, and 35 years for the outliers versus mean of 25.8 years ± 4.7 S.D. for all 19 donors) or gender (two males and one female for the outliers versus a 2.6 ratio of males to females for the remaining donors). The outliers were also not explained in any obvious manner by differences in the conditions used to prepare or expand the MSCs.
In previous experiments, our optimal preparations of MSCs that were enriched for RS cells were about 60% clonogenic: that is, they generated up to 60 colonies for every 100 cells plated [9, 10, 12]. As demonstrated here, the fraction of RS cells isolated as the FSlo/SSlo events were 80%–90% clonogenic. The higher clonogenicity is probably explained not only by sorting for FSlo/SSlo cells but also by more precise plating of the cells with an automated instrument and expansion of the colonies in individual wells that eliminates the cross-talk among nascent colonies through soluble factors produced by the cells . In addition, only about 60% of single cell–derived colonies that were obtained in previous experiments differentiated into osteoblasts or adipocytes [11, 19]. In contrast, assays of parallel sc-CFU plates here demonstrated that essentially all the colonies that developed from FSlo/SSlo cells could be differentiated into osteoblasts or adipocytes. The increased differentiation may also be explained by growth of the colonies in individual wells that eliminates cross-talk among them. Because of their rapid rate of proliferation, the behavior of RS cells probably is comparable to the transitory amplifying cells of the hematopoietic stem cell system .
We tested over 200 commercially available antibodies (not shown) but did not find any that effectively discriminated between FSlo/SSlo and FShi/SShi cells if adequate allowance was made for the difference in average size of cells from the two regions. Therefore, it appears that no surface epitopes are currently available to distinguish RS cells from more mature cells in different preparations of MSCs. As a result, it has been difficult to adequately assess the high degree of variability encountered with different preparations and to compare data generated by different investigators. The standardized FS/SS assay developed here for cell morphology presents a simple and rapid method of defining some of the variability. We previously found that preparations that do not expand rapidly can be identified by subjective evaluation of cultures for the presence of a few wider, spindle-shaped cells (RS-1B and RS-1C cells) , which apparently have an inhibitory effect on the propagation of RS cells in the same cultures. The flow parameter developed here provides an object measure of the same feature of the cultures. The flow parameter can be assayed in less than 20 minutes and therefore can be used to rapidly eliminate preparations that will have low plating efficiency and may be unsuitable for some experiments. Unexpectedly, however, the assay for FS/SS (Fig. 7) identified a subset of MSC preparations that expanded rapidly even though they contained a significant number of larger cells, as determined by the analysis of FS and SS. This subset will be of interest to examine further. However, it is probably advisable to discard all preparations with low flow parameters in designing experiments that require rapid replication of early progenitor MSCs under reproducible conditions. The light-scattering properties of cells were previously used to identify early progenitors in cultures of periosteal cells from fetal rat . Therefore, the standardized protocol for FS/SS-based morphological analysis may be useful as a rapid method to characterize small stem-like cells from a number of adult tissues.
Jason R. Smith and Radhika Pochampally contributed equally to this work.