|Home | About | Journals | Submit | Contact Us | Français|
We compared human mesenchymal stem cells (hMSCs), expanded long term with and without fibroblast growth factor (FGF) supplementation, with respect to proliferation, and the ability to undergo chondrogenesis in vitro. hMSCs expanded in FGF-supplemented medium proliferated more rapidly than the control cells. Aggregates of FGF-treated cells exhibited chondrogenic differentiation at all passages tested although, in some preparations, differentiation was diminished after seventh passage. Aggregates made with control cells differentiated along the chondrogenic lineage after first passage but exhibited only marginal differentiation after fourth and failed to form cartilage after seventh passage. Microarray analysis of gene expression identified 334 transcripts differentially expressed in fourth passage control cells that had reduced chondrogenic potential, compared with the fourth passage FGF-treated cells that retained this capacity, and 243 transcripts that were differentially expressed when comparing them to the first passage control cells that were also capable of differentiating into chondrocytes. The intersection of these analyses yielded 49 transcripts differentially expressed in cells that exhibited chondrogenic differentiation in vitro compared with the cells that did not. Among these, angiopoietin 1, secreted frizzled-related protein 1, and six transmembrane epithelial antigen of the prostate 1 appear to be of higher relevance. These preliminary data must now be validated to verify whether different gene expression profiles translate into functional differences. In summary, these findings suggest that the chondrogenic potential of hMSCs is vulnerable to cell expansion and that care should be exercised when expanding these cells for cartilage tissue engineering applications. Supplementation with FGF-2 allows reaching target cell numbers more rapidly and extends the level of expansion within which these cells are useful for tissue-engineered cartilage repair.
Tissue engineering is the process of creating functional living replicas of tissues through the use of cells, scaffolds, and biomolecules. Tissue engineering, and its application to regenerative medicine, stands to benefit greatly from progress in the isolation, expansion, and characterization of multipotent stem cells from adult tissues.
Stem cells are defined as cells with the ability to renew themselves through cell division and, under certain conditions, differentiate and acquire specialized phenotypes.1,2 Mesenchymal stem cells (MSCs) are derived from the stroma of bone marrow. Human MSC (hMSC) preparations have significant, although highly variable and not unlimited, proliferative potential.3 In some cases, MSCs can be expanded through 30–40 population doublings (109–1012:1 expansion), whereas other preparations stop proliferating after 4–5 population doublings (50:1 expansion).4–6 This different behavior may be because of multiple factors, such as the protocol used to obtain the bone marrow,4–7 the low frequency of MSCs in marrow,8 and the age and condition of the donor.5,9
Interestingly, the loss of differentiation potential is not generalized. It has been reported that hMSCs maintain their osteogenic potential through extensive subcultivation.4,5 Further, in some cases, late-passage hMSCs spontaneously differentiate into osteoblasts.10,11 However, as cells approach senescence they lose their ability to differentiate into adipocytes.5 The ability to differentiate into chondrocytes is lost in earlier passages.12
Self-renewal and pluripotentiality, the two key properties that characterize a stem cell, are progressively lost in MSC preparations as cultured MSCs senesce in vitro.5,10–12 By and large, cell-based tissue engineering therapies require large numbers of cells that make extensive in vitro subcultivation an absolute requirement. This loss of proliferation and differentiation potential are critical hurdles that must be overcome if MSC-based tissue engineering and regenerative medicine are going to make the transition from the laboratory bench to the clinic.
Recent reports have demonstrated that hMSCs cultured in the presence of fibroblast growth factor 2 (FGF-2) can undergo more population doublings than control cells before they lose their proliferative capacity.13–17 Our long-term goal is to identify the cellular mechanisms involved in the regulation of the chondrogenic potential of hMSCs. We have demonstrated that gene expression analysis by microarray can aid in the identification of the molecular mechanisms18 underlying the aspects of these phenomena, through generation of association networks that link differentially expressed genes with the cellular processes in which they participate. The fact that FGF-2 treatment has the potential to delay the loss of chondrogenic potential can be exploited to help us identify the mechanisms that regulate this potential.
Six hMSC preparations were used in this study. These cells were derived from bone marrow aspirated from the iliac crest of normal healthy human donors at the Hematopoietic Stem Cell Core Facility at Case Western Reserve University after informed consent was obtained under the terms of an Internal Review Board–approved protocol.
The procedures for establishing human bone-marrow-derived MSC cultures followed previously published methods.19,20 Briefly, bone marrow aspirates were washed with control medium consisting of low-glucose Dulbecco's modified Eagle's medium (DMEM-LG) (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS) from a selected lot (lot # ACQ 23291; Hyclone, Logan, UT)20 (DMEM-LG+10% FBS) and subjected to a preformed Percoll (Sigma Chemical, St. Louis, MO) density gradient to isolate mononucleated cells. Serum lot selection is a standard procedure performed before purchasing a new shipment of serum; all these experiments were conducted with serum from a single lot. The mononucleated cells were washed with control medium and seeded at a density of 1.8×105 cells/cm2 in control medium to establish primary cultures of human bone-marrow-derived MSCs. All cell culture was done at 37°C in a humidified atmosphere of 95% air and 5% CO2.
From the first medium change (day 4), some of the plates were assigned to receive control medium consisting of DMEM-LG+10% FBS, and the rest of the plates received DMEM-LG+10% FBS supplemented with 10ng/mL of FGF-2 (Biological Resources Branch of the National Cancer Institute, Bethesda, MD). This dose was chosen based on the previous studies.18 Cultures were fed twice a week.
To keep their growth at an exponential rate and prevent spontaneous differentiation or loss of differentiation potential, hMSCs had to be subcultured before the cells became confluent.19,21 Typically, they were passaged when the cultures were 80–90% confluent. Primary cultures were usually subcultured around 14±3 days. Subsequently, the cells were subcultured approximately every 7±2 days. Plates assigned to the different study groups (control or FGF treated) were subcultured at the same time, which resulted in different levels of confluence in the treatment groups as a result of the previously reported differences in cell proliferation and cell size.18 Cells were subcultured by trypsinization, counted, and reseeded at a density of 4.5×103 cells/cm2.
Trypsinized cell suspensions were counted manually in triplicate using a Neubauer hemocytometer. Population doublings were calculated as the base 2 logarithm of the number of cells obtained at the end of a given passage divided by the number of cells seeded:
For primary cultures we used the number of colonies formed in primary culture as the denominator, assuming that one colony is derived from one MSC:
Population doubling times for each passage and treatment group were calculated by dividing the number of population doublings in that passage by the duration of the passage time in days.
At the end of first, fourth, and seventh passages, the cells were resuspended in chondrogenic medium at 1.25×106 cells/mL. Chondrogenic medium is high-glucose DMEM (Invitrogen) supplemented with 1% ITS+Premix (BD Biosciences, Franklin Lakes, NJ), 100μM ascorbate-2-phosphate (Wako, Richmond, VA), 10−7 M dexamethasone (Sigma Chemical), and 10ng/mL transforming growth factor-β1 (Peprotech, Rocky Hill, NJ). This is the same chondrogenic differentiation medium formulation used in all our18,22–25 and other's12,26 previous studies. This medium is serum free and contains no FGF. Two hundred microliter aliquots containing 2.5×105 cells were placed in polypropylene 96-well plates (Phenix, Hayward, CA), centrifuged at 500 g, and placed in the incubator at 37°C in a humidified atmosphere of 95% air and 5% CO2.24,27–29 After 12–16h, the cells coalesced and formed a free-floating aggregate. The chondrogenic medium was changed three times per week. On days 7, 14, and 21, cell aggregates were harvested. Replicate aggregates from each group were processed for histologic and immunohistologic evaluation, or for proteoglycan and DNA quantification.
For the chondrogenic aggregate cultures, cell numbers were determined indirectly by measuring the DNA content. Aggregates were digested with papain (Sigma Chemical).30 The papain-digested extract was combined with 0.1N NaOH, incubated at room temperature for 20min, and neutralized with 0.1N HCl in 5 M NaCl, 100mM NaH2PO4, and this mixture combined with 1mL of 0.7μg/mL Hoechst 33258 (Sigma Chemical) in water. Fluorescence was read in a GENios Pro microplate reader (Tecan, Durham, NC; λEx=340nm, λEm=465nm) and compared with that of a certified Calf Thymus DNA standard (Amersham, Piscataway, NJ).
Aggregates were papain digested as described above. A nitrocellulose membrane (0.45μM pore size) was placed into a dot-blot apparatus. A 250-μL aliquot of 0.02% Safranin O (Sigma Chemical) in 50mM sodium acetate (pH 4.8) was pipetted into each well, and 25-μL aliquots of the papain-digested extracts were then pipetted into the wells. Vacuum was applied until the samples filtered through. The wells were rinsed with distilled water, and the filter was removed from the apparatus and allowed to air dry. The individual dots were cut out, transferred to microcentrifuge tubes, and eluted for 20min in 10% cetylpyridinium chloride (Sigma Chemical) at 37°C. The absorbance of these extracts was read at 536nm and compared with that of the chondroitin sulfate standards (Seikagaku America, East Falmouth, MA). Proteoglycan content was normalized to DNA content of the chondrogenic cultures.31
To further evaluate chondrogenesis in the aggregates, adjacent sections were stained for collagen types I, II, and X, using anti-Collagen Type I (Clone Col-1; Sigma Chemical, Cat# C2456), anti-Collagen Type II (Developmental Studies Hybridoma Bank, Iowa City, IA; Cat# II-II6B3), and anti-Collagen Type X (from Dr. Gary J. Gibson, Breech Research Laboratory, Bone & Joint Center, Henry Ford Hospital & Medical Centers, Detroit, MI) antibodies, respectively. Sections were deparaffinized in xylene, digested with pronase for 15min, and then blocked with 5% bovine serum albumin (BSA) in PBS for 30min. Either primary antibodies or control immunoglobulin G were applied in 1% BSA in PBS for 1h. Secondary antibody was fluorescein isothiocyanate–conjugated goat anti-mouse immunoglobulin G (Chemicon, Temecula, CA). In some cases, 4',6-diamidino-2-phenylindole (DAPI) was used as a nuclear counterstain. Wet mounts were immediately documented with a 10×objective, using a fluorescence microscope and a SPOT-RT digital camera. The sections were evaluated qualitatively for the presence (or absence) and location of the immunoreactivity.
To attempt to determine the molecular basis of the effect of FGF-2 on hMSCs, total RNA was Trizol extracted from three independent hMSC preparations at the end of first, fourth, and seventh passages. DNA microarray was performed by the Case Western Reserve University (CWRU) Gene Expression Array Core Facility, using methods that have been described previously.32–37 Briefly, total RNA was cleaned-up using Qiagen (Valencia, CA) spin columns (only RNA samples with 260/280nm absorbance ratios between 1.9 and 2.1 were carried further). Five to eight micrograms of total RNA was converted into double-stranded cDNA using Superscript II reverse transcriptase (Gibco BRL, Rockville, MD) and an oligo-dT primer with a T7 RNA polymerase promoter linked to its 5′ end. cDNA was cleaned and used in an in vitro transcription (IVT) reaction to generate adequate amplified, biotin-labeled cRNA (50–80μg). cRNA was then fragmented and 15μg added to 300μL hybridization cocktail (100mM 2-(N-morpholino)ethanesulfonic acid (MES), 1M NaCl, 20mM ethylenediaminetetraacetic acid, 0.01% Tween 20; 0.1mg/mL Herring Sperm DNA, 0.5mg/mL acetylated BSA, and Affymetrix IVT controls, as per manufacturer). Four IVT products of bacterial genes were added at final concentrations ranging from 1.5 to 100pM. Spiked transcripts served as hybridization controls and were always monitored as part of the quality control regimen. Fragmented biotin-labeled cRNA was added, and the final volume was adjusted to 300μL with molecular grade water. Samples were hybridized onto the Affymetrix (Santa Clara, CA) H-U133A 2.0 human microarrays, which represent ~18,400 full-length genes and expressed sequence tag clusters from UniGene build 133 for 16h at 45°C. The manufacturer's standard posthybridization wash, double-stain, and scanning protocols use an Affymetrix GeneChip Fluidics Station 400 and a Hewlett Packard Gene Array scanner.
Raw data from microarray scans were initially normalized and analyzed with Affymetrix Microarray Suite v.5.0. Pair-wise comparisons were made between samples from the different culture conditions. Transcripts were defined as differentially regulated if they met the following criteria: (i) consistent increase/decrease call in all replicates of a specific condition, based on the Wilcoxon's signed rank test (p<0.05) and (ii) absolute value of the average fold difference ≥2.0 (Tables 2 and and33).
Affymetrix-provided transcript annotations were replaced with official gene nomenclature using National Center for Biotechnology Information (NCBI) databases, and gene functions were assigned based on gene ontology and other data in NCBI LocusLink (www.ncbi.nlm.nih.gov/LocusLink/).
The statistical significance of the differences in population doubling times was determined by paired t-test. Analyses of the biochemical measurements were performed by two-factor repeated measurements analysis of variance and paired t-tests; differences were considered significant for p-values<0.05.
FGF-treated hMSCs exhibited higher proliferation rates than those expanded in control conditions. The population doubling time for FGF-treated hMSCs was always shorter than that of the cells expanded under control conditions at any passage (p<0.01). This is shown in Figure 1 where the slope of the growth curves approximate the proliferation rates of the treatment groups. Plates assigned to the different study groups (control or FGF treated) were subcultured at the same time, which resulted in different levels of confluence in the treatment groups as a result of the previously reported differences in cell proliferation and cell size.18 In no case were control cultures more confluent than their FGF-treated counterparts at the time of subcultivation.
The chondrogenic potential of hMSCs was affected by both time in culture and culture conditions (Table 1). FGF-treated hMSCs always formed aggregates that were larger than the ones made with the matching cells expanded in control conditions after 7, 14, and 21 days of chondrogenic culture.
At the end of the first passage, histology (Fig. 2) revealed a more developed and more homogenous cartilaginous matrix in aggregates made with FGF-treated hMSCs (Fig. 2E–H) compared with those made with cells expanded in control conditions (Fig. 2A–D).
Aggregates made with fourth passage FGF-treated hMSCs (Fig. 3E–H) presented an overall appearance very similar to that of the aggregates made with first-passage cells although they were, in most cases, slightly smaller. Aggregates made with fourth passage cells expanded in control conditions (Fig. 3A–D) demonstrated little or no chondrogenic differentiation.
In all of the cell preparations tested, aggregates made with seventh passage FGF-treated cells were smaller than those made with first or fourth passage FGF-treated cells. In some of the preparations, the aggregates exhibited robust chondrogenic differentiation (identified as “++” in Table 1), whereas some other preparations (identified as “+” or “±” in Table 1) exhibited weaker or limited chondrogenic differentiation (Fig. 4E–H). None of cells expanded under control conditions exhibited signs of chondrogenic differentiation at the end of the seventh passage (identified as “–” in Table 1) (Fig. 4A–D).
Analysis of the DNA content of the aggregates demonstrated no differences in the amounts of DNA per aggregate between the treatment groups (p=0.39) but DNA content decreased as a function of time in culture (p<0.01). In contrast, analysis of the glycosaminoglycan (GAG) content revealed that, within any given passage, aggregates made with FGF-treated cells contained more GAG, both per aggregate and normalized to DNA content (p<0.05) (Fig. 5). GAG content diminished as a function of time in culture in both treatment groups but much more rapidly in the cells expanded in control conditions. In aggregates made with control cells, the GAG content per DNA dropped significantly (p<0.01) between first and fourth passage. It remained at the same low level in fourth and seventh passage (p=0.26), while in aggregates made with FGF-treated cells, a similar decrease did not take place until passage 7 (first to fourth passage p=0.11; fourth to seventh passage p<0.01). It should be noted that these passage numbers correspond to different population doublings; in all cases the control group had a lower number of population doublings. The fact that, despite this, the differences are in favor of the FGF-treated cells strengthens the case for the use of this supplement.
As previously reported,18 even successful chondrogenic aggregates made with cells expanded under control conditions (Fig. 2) presented an outer layer of flattened cells embedded in a fibrous extracellular matrix (ECM) that did not contain type II collagen (Fig. 2C) but did contain type I collagen (Fig. 2B), as determined by immunostaining at 3 weeks. In contrast, aggregates made with FGF-treated cells did not exhibit this fibrous outer layer, and type II collagen immunoreactivity was present throughout the aggregate (Fig. 2G). Type X collagen immunoreactivity was present throughout the ECM of aggregates made with FGF-treated cells except for a peripheral area of approximately six cell layers (Fig. 2H). Type X collagen immunoreactivity was also present throughout the matrix of aggregates made with cells expanded under control conditions (Fig. 2D).
Pair-wise comparisons between cells from different passages within the same treatment group (Table 2) or different treatment groups within the same passage (Table 3) were made for each cell preparation to determine the effect of time in culture and culture conditions on gene expression, respectively. In both treatment groups, the highest number of differentially expressed genes was detected in the comparison of first to seventh passage. In control cells, the comparison of fourth to seventh passage identified six transcripts that were differentially expressed in all three preparations, whereas in the FGF-treated cells, the fourth passage cells had fewer differentially expressed transcripts when compared with the first passage cells (Table 2). When comparing cells within the same passage, the number of differentially expressed transcripts as a result of the different conditions was similar for all the three passages tested (Table 3).
We further analyzed the transcripts identified as differentially expressed in the comparison of control cells at the first and fourth passages and the comparison of the control and FGF-treated fourth passage cells. These are comparisons between chondrogenic and nonchondrogenic cells (Table 2). The intersection of these two groups of transcripts yielded 46 genes differentially expressed in chondrogenic versus nonchondrogenic cells (Table 4). Almost 20% of these differentially expressed genes are involved in intracellular signaling; approximately 15% relate to immune responses and 12% to epidermis development. Half of the signaling-related genes were upregulated and the other half downregulated in cells with chondrogenic potential, while all of those related to immune responses were upregulated and all of the genes related to epidermis development were downregulated.
Cell-based therapies require large numbers of cells with the necessary characteristics for successful implantation, engraftment, and function. For cartilage tissue engineering, for example, approximately 8–10×107 cells/mL of tissue are required as starting material.38 Table 5 shows the projected cell requirements for a variety of repair diameters and thicknesses, assuming disc-shaped defects. The current state-of-the-art in hMSC technology does not consistently achieve this degree of proliferation without partial or total loss of chondrogenic potential. This loss3,5,10,12 limits the applicability of hMSC-based therapies.
As we have reported previously, exposure of hMSCs to FGF-2 during expansion increases the cell yield and enhances their chondrogenic potential. By first passage, the FGF-treated cells average three population doublings more than the control cells: The FGF-treated cells are smaller than their control counterparts, allowing more cells per cm2 at the same level of confluence.18 In addition, the FGF-treated hMSCs exhibit enhanced chondrogenic potential. After four passages, the FGF-treated cells have doubled eight times more than the control (a 256-fold increase in yield). The degree of differentiation, reflective of the interindividual variability among cell preparations, was variable. Nonetheless, they retain a chondrogenic potential which is very diminished or lost entirely in the control cells. After seven passages, no control cells from any of the cell preparations tested retained any chondrogenic potential, whereas some of the FGF-treated cells did. This is consistent with our previous findings of stimulatory effects of FGF-2 on the proliferation and differentiation of bone-marrow-derived progenitors18 and are confirmed by other groups.13,15–17 However, in these reports, bone-marrow-derived progenitor cells exhibited differentiation potential only when expanded in the presence of FGF-2 or cultured at low density.16 In our hands, early-passage hMSCs expanded under control conditions do exhibit chondrogenic potential, which is enhanced in the FGF-treated hMSCs. These differences might be explained by our rigorous screening and testing of serum lots to select those that support the growth of hMSCs and maintain their multipotentiality. This is an important and documented element of our technology,20 as the use of suboptimal serum results in the rapid loss of multipotentiality and slow expansion of hMSCs. Subtle differences in the protocols (seeding density of primary and subsequent cultures, or formulation of the base media) may also play a role in the lack of differentiation potential reported by others for cells grown without FGF.13,15,16 Analysis of DNA and GAG content suggests that the larger size of the aggregates from FGF-treated cells is because of the increased matrix production, and not because of the presence of more cells in the aggregates. Histologic analyses of the pellets revealed qualitative differences between the treatment groups beyond the difference in the size of the pellets.
Clearly, if hMSCs are to be of value to clinical tissue engineering, it will be critical to understand how their proliferation and differentiation potentials are regulated or by what mechanism FGF-2 supplementation enhances their proliferative and chondrogenic potentials. In this study, we take a first step toward this goal and present gene expression signatures of hMSCs with enhanced chondrogenic potential. As in most microarray studies, our dataset presents some sample-to-sample variability39; the genes identified here as differentially expressed met very stringent selection criteria, increasing our confidence in the data. The analysis revealed differences due to both time in culture and culture conditions. The effects of time in culture in control cells were rapid, with most of the expression changes occurring during the initial passages, and the cells then settling into a fairly stable phenotype with only very few changes occurring later on. Many of these changes might be part of the adaptation of the cells to the monolayer culture conditions. But, in conjunction with the rapid sequential loss of differentiation potentials, this suggests that, under ordinary culture conditions, chondrogenic potential and, perhaps, true “stemness” is a very transient property of hMSCs.
Unsurprisingly, given the mitogenic effects of the treatment, cell signaling, metabolism, and cell-cycle-related genes figure prominently in the list of genes differentially expressed in the FGF-treated cells compared with their control counterparts. We also observed robust changes in the ECM-related gene expression, which trend toward reduced expression of differentiation-related ECM molecules and enhanced expression of ECM catabolism-related genes. The pervasiveness of this pattern suggests that it is more than merely related to the enhanced cell division, but rather represents a fundamental shift in the phenotype of the cells. This is supported by the fact that merely delivering a mitogenic stimulus via platelet-derived growth factor-beta polypeptide (PDGF-BB) had a similar effect on cell division, but failed to improve the maintenance of chondrogenic potential (data not shown). In future experiments, we plan to use PDGF-BB-stimulated cells in the expression array experiments to tighten the focus on nonmitogenic FGF-2 effects.
Our previous gene expression studies in first passage cells18 suggested that FGF-2 might be mediated through mitogen-activated protein kinase (MAPK) and Wnt signaling pathways.40–44 Here, we focused on the differences between hMSC preparations that can successfully differentiate into chondrocytes and those that cannot. Some differences were consistent with those reported previously between the first passage control and FGF-treated cells.18 We found differences in Wnt signaling between chondrogenic and nonchondrogenic cells with upregulation of secreted frizzled-related protein 1 and downregulation of pregnancy-specific beta-1-glycoprotein 1 in chondrogenic cells. The effects on MAPK signaling were not through elements of the feedback loop as previously reported,18 but rather through the upstream regulators of MAPK signaling such as angiopoietin 1 and midkine, which are upregulated in cells with chondrogenic potential. These transcripts may have a role not only in the maintenance of the chondrogenic potential but also in the overall physiology of hMSCs; angiopoietin 1 has been implicated in the paracrine tissue repair activity displayed by MSCs45–47 and as an antiapoptotic factor48; secreted frizzled-related protein 1 has also been reported to play a role in the tissue repair activity of MSCs49,50 and in controlling osteogenic differentiation.51–53 Additionally, a recent report indicates that six transmembrane epithelial antigen of the prostate 1, also upregulated in cells with chondrogenic potential, may be a marker for MSCs.54 Our data here must now be validated at the level of protein expression and signaling pathway activation to verify whether the different gene expression profiles detected translate into functional differences.
In summary, the chondrogenic potential of MSCs is vulnerable to cell expansion; the expanded cells may exhibit weak or no chondrogenic potential, rendering them suboptimal or useless for the intended application. Supplementation of the growth medium with FGF-2 not only allows investigators to reach target cell numbers quicker but also extends the level of expansion within which these cells are still useful for tissue-engineered cartilage repair. The FGF-treated cells remain chondrogenic after 30 population doublings, whereas control cells are no longer chondrogenic after approximately 20 population doublings, that is over 1000-fold difference in the number of cells. Taken together, our results indicate that we must be extremely careful when these cells are extensively expanded to achieve a target cell number for cartilage tissue engineering applications in particular, but perhaps also when used in other applications.
This research was supported by a grant from the National Institutes of Health (R01 AR50208; PI Welter) and the Hematopoietic Stem Cell Core Facility and the Gene Expression and Genotyping Facility of the Case Comprehensive Cancer Center (P30 CA43703). Anti-collagen type X antibody was a gift from Dr. Gary J. Gibson, Breech Research Laboratory, Bone & Joint Center, Henry Ford Hospital & Medical Centers, and the human FGF-2 was provided by the Biological Resources Branch of the National Cancer Institute.
No competing financial interests exist.