Identification of Predictive Gene Markers for Multipotent Stromal Cell Proliferation

Abstract

Multipotent stromal cells (MSCs) are known for their distinctive ability to differentiate into different cell lineages, such as adipocytes, chondrocytes, and osteocytes. They can be isolated from numerous tissue sources, including bone marrow, adipose tissue, skeletal muscle, and others. Because of their differentiation potential and secretion of growth factors, MSCs are believed to have an inherent quality of regeneration and immune suppression. Cellular expansion is necessary to obtain sufficient numbers for use; however, MSCs exhibit a reduced capacity for proliferation and differentiation after several rounds of passaging. In this study, gene markers of MSC proliferation were identified and evaluated for their ability to predict proliferative quality. Microarray data of human bone marrow-derived MSCs were correlated with two proliferation assays. A collection of 24 genes were observed to significantly correlate with both proliferation assays (|r| >0.70) for eight MSC lines at multiple passages. These 24 identified genes were then confirmed using an additional set of MSCs from eight new donors using reverse transcription quantitative polymerase chain reaction (RT-qPCR). The proliferative potential of the second set of MSCs was measured for each donor/passage for confluency fraction, fraction of EdU+ cells, and population doubling time. The second set of MSCs exhibited a greater proliferative potential at passage 4 in comparison to passage 8, which was distinguishable by 15 genes; however, only seven of the genes (BIRC5, CCNA2, CDC20, CDK1, PBK, PLK1, and SPC25) demonstrated significant correlation with MSC proliferation regardless of passage. Our analyses revealed that correlation between gene expression and proliferation was consistently reduced with the inclusion of non-MSC cell lines; therefore, this set of seven genes may be more strongly associated with MSC proliferative quality. Our results pave the way to determine the quality of an MSC population for a particular cellular therapy in lieu of an extended in vitro or in vivo assay.

Introduction

Human multipotent stromal cells (MSCs) are highly regarded as having the potential to be used in cell-based therapies. Numerous studies have presented the promising outcomes of MSCs in animal studies, such as osteoarthritis, myocardial infarction, graft-versus-host disease, and cancer [1 –9]. Their potential to have a therapeutic effect stems from their capacity to undergo multilineage cell differentiation, suppress the immune response, and secrete trophic factors [10 –12]. While the scope of this research focuses on bone marrow-derived MSCs, these cell types have also been isolated from other tissue sources, including adipose tissue, muscle, and Wharton's jelly [13 –15].

Efforts have been made to better characterize the identity and therapeutic effect of MSCs in both in vitro and in vivo assays [16 –18]. Furthermore, multidimensional data sets are being produced using microarray and sequencing technologies to evaluate the predictability of identified protein or gene markers to correlate with clinical outcomes [19 –22]. The difficulty in characterizing and identifying potential markers for MSCs is their heterogeneous behavior, which possibly results during tissue isolation and proliferation protocols. The unknown mixture of contaminating cells within a population may negatively impact their therapeutic potential. Similarly, donor variation between MSCs can also impact performance; hence, evidence suggests not all MSCs are equal [20,23 –26].

Our previous work has demonstrated the effect of cellular aging in culture on gene expression and cell proliferation; however, the magnitude of the impact on both parameters is not equal, where MSCs from some donors will have a larger reduction in gene expression or cell proliferation potential with passaging than other MSCs from different donors [27]. Not all MSCs divide at the same rate or an identical number of times every passage, which is a consequence of their heterogeneity. To better characterize MSCs, a more appropriate method would be to correlate MSC markers with an attribute of their quality that is indicative of their therapeutic benefit. MSC markers exhibiting the strongest correlation between these two measurements would be ideal for predicting the therapeutic potential for the specific indication. One such quality attribute is proliferation as cellular expansion is typically needed to generate a sufficient number of cells after tissue isolation for a cell therapy or cell-based product. The objective of this work was to use microarray chips to identify genes from one set of MSCs that exhibit a strong correlation with the outcomes of two different proliferation assays: confluency fraction and fraction of dividing cells. To confirm the predictive power of the top identified genes, another set of MSCs was expanded in different media, evaluated by an alternative method for gene expression reverse transcription quantitative polymerase chain reaction (RT-qPCR) and correlated with an additional measure of cell proliferation. Furthermore, the predictive potential of the identified genes for proliferation was assessed for four different non-MSC populations (osteoblasts, chondrocytes, dermal fibroblasts, and embryonal carcinoma cell line) for comparison.

Methods

Cultivation and expansion of human bone marrow-derived MSCs

Human MSCs were purchased from All Cells (Alameda, CA), Cell Applications (San Diego, CA), and RoosterBio (Frederick, MD) and expanded to passage 8 in marrow stromal cell growth medium (Cell Applications) under a humidified atmosphere of 5% CO₂ at 37°C (Table 1). Other cell types, chondrocytes, osteoblasts, dermal fibroblasts, and NTERA-2 embryonal carcinoma cell line (NT2), were purchased from Cell Applications and ATCC (Manassas, VA) and grown to passage 4. MSCs were detached from the flask using 0.25% trypsin-versene EDTA solution (Lonza, Walkersville, MD) after reaching 80% confluency and replated at a density of 550 cells/cm². The MSCs in the original data set were expanded differently as described in Lo Surdo et al. [28].

Table 1.

Donor Characteristics

No.	Donor	Cell type	Sex	Age	Passage	Company
1	1F3422	MSC	Male	39	P4, P8	All Cells
2	BM2893	MSC	Male	41	P4, P8	All Cells
3	BM3018	MSC	Male	39	P4, P8	All Cells
4	HMSC2483	MSC	Female	Fetal	P4, P8	Cell Applications
5	HMSC2618	MSC	Female	Fetal	P4, P8	Cell Applications
6	HMSC2651	MSC	Male	56	P4, P8	Cell Applications
7	RB14	MSC	Female	20	P4, P8	RoosterBio
8	HMSC1994^a	MSC	Female	Fetal	P4, P8	Cell Applications
9	HC1994^a	Chondrocyte	Female	Fetal	P4	Cell Applications
10	HOb1994^a	Osteoblast	Female	Fetal	P4	Cell Applications
11	NHDF271678	Dermal fibroblast	Female	58	P4	Lonza
12	NT2	Embryonal carcinoma	Male	22		ATCC

Cell types from the same donor.

MSC, multipotent stromal cell.

RT-qPCR

RNA isolation was completed on cell samples at either passage 4 or 8. NT2 cells grow indefinitely; therefore, no passage number was assigned. The Qiagen RNeasy Mini Kit (Valencia, CA) was used to isolate total RNA from each sample per the manufacturer's instructions. The Agilent 2100 bioanalyzer (Santa Clara, CA) was used to assess the RNA quality, where all samples had measured RNA integrity values >9.0. Total RNA concentration was measured using the NanoDrop 1000 spectrophotometer (Wilmington, DE).

RT-qPCR plates containing 30 genes in triplicate with positive/negative controls were purchased from Bio-Rad (Hercules, CA). Reverse transcription was performed per the manufacturer's instruction using the 5× iScript cDNA Synthesis Kit. Quantitative polymerase chain reaction was performed per the manufacturer's instructions using SsoAdvanced Universal SYBR Green Supermix. The reaction protocol was 10 s at 95°C and 30 s at 60°C for 40 cycles. The 96-multiwell plates for three samples, 1F3422 P4, HMSC2483 P4, and HMSC2618 P8, were repeated in triplicate to evaluate the reproducibility of the selected primers.

Cell proliferation assays

Three different cell proliferation assays were performed on all cell types recovered from vials frozen at passages 4 and 8. These cells were considered passages 5 and 9 after thaw and subsequent tissue culture. For the confluency fraction cell proliferation assay, cells were plated at a density of 1,000 cells per well in 24-well plates. Following cell adhesion, multiwell plates were placed into an IncuCyte live-cell imager (Essen BioScience, Ann Arbor, MI). Whole-well images were captured per well every 4 h. Confluency fraction measurements were averaged for each well (n = 4) and each passage of every cell type and plotted as a function of time to assess proliferative capacity.

The Click-iT EdU cell proliferation assay (Life Technologies, Frederick, MD) was used to evaluate cell division per the manufacturer's instructions. Briefly, all cell types at each passage were plated in 24-well plates at a density of 1,000 cells per well and administered the EdU reagent for both 6 and 10 h after adherence. Later, MSCs were fixed and stained with Alexa Fluor 488 for EdU detection and Hoechst 33342 (Life Technologies) to visualize the cell nuclei. Automated fluorescence microscopy using a Nikon Ti-S inverted microscope with a 10× objective was used to capture 40 images per well, and quantitation was performed using CellProfiler as previously described [27].

To determine population doubling time, cells were plated at a density of 1,000 cells per well in four 24-well plates. Plates were fixed and stained with Hoechst 33342 at 24-h intervals (0, 24, 48, and 72 h). Automated fluorescence microscopy captured 40 images per well (n = 4), where cell number was averaged per time point and a population doubling curve was used to estimate the population doubling time.

Data analysis and statistical evaluation

The microarray data sets and subsequent analysis were performed with data files from passages 3, 5, and 7 from 6 different donors as described by Bellayr et al. [27] (Gene Expression Omnibus: GSE74303). Additionally, two-color microarray hybridization was conducted for donors 127756 (at passages 3 and 5) and PCBM1655 (passage 3) in triplicate in a randomized block design as previously described with the other six donors. One technical replicate for sample PCBM1655 (passage 3) was determined an outlier by Mahalanobis and Jackknife distances and subsequently not included in calculating the median expression values for this sample. Donor 127756 did not grow beyond passage 5, and donor PCBM1655 did not grow beyond passage 3. The microarray data were correlated with cell proliferation data for confluency fraction and EdU+ cell fraction after 6 h from our previously published work [27 –29].

For the microarray data, a regression analysis was used to quantify the correlation between two different MSC proliferative potential assays (confluency fraction at 4 days and fraction of EdU+ cells after 6 h) and the 34,555 probes spotted on the microarray corresponding to the MSC donor/passage gene expression. The slope (β₁) and correlation coefficient (r) were calculated for each probe on the microarray chip when correlated with each proliferation assay. A cutoff slope (β_C) was calculated by dividing the technical variability (standard deviation) per probe by the median coefficient of variation from the corresponding proliferation assay. Probes whose calculated slope (β₁) was less than the cutoff slope (β_C) were not included in the statistical analysis as the expression for those probes was considered highly variable. Probes on the microarray were also filtered from the data set if the magnitude of expression for 90% of the samples was not greater than a calculated background cutoff. P values were calculated from a t distribution.

Similarly, for the RT-qPCR data (Gene Expression Omnibus: GSE74203), a regression analysis and statistical analysis were applied between the gene expression data (ΔCT) and the three different proliferation assays: confluency fraction (at 5 and 6 days), fraction of EdU+ cells (6 and 10 h), and population doubling. A paired t-test was also applied to calculate statistical significance between passages 4 and 8 gene expression data and MSC proliferation results.

RT-qPCR results were normalized by eight different methods. For five of the normalization methods, the individual genes ACTB, GAPDH, GUSB, NACAD, and ST13 were used to normalize 96-well plates of RT-qPCR data. Normalization was also performed using the geometric mean of common housekeeping genes (common combined: ACTB, GAPDH, and GUSB) and newly identified potential housekeeping genes (new combined: NACAD and ST13) [30]. The final method to assess the normalization of the RT-qPCR genes was using the geometric mean of all five housekeeping genes (all combined).

For all statistical analyses of genes from both the microarray and RT-qPCR data, an adjusted P value was calculated using the Benjamini and Hochberg method, where significance was determined by P < 0.05 [31]. All statistics and high-throughput analyses were performed using JMP Genomics 7 software.

Results

Correlation of MSC microarray and proliferation data sets

The frequency of the percent coefficient of variation was determined for two previously completed cell proliferation assays: confluency fraction at 4 days and fraction of EdU+ cells after 6-h culture with EdU, where the median percent coefficient of variation is 11.91% and 6.71%, respectively (Supplementary Fig. S1A, C; Supplementary Data are available online at www.liebertpub.com/scd). Likewise, the technical variability of previous microarray data sets was calculated per probe spotted on the microarray, where probes with a lower technical variability were more likely to produce consistent measurements (Supplementary Fig. 1B).

Using the MSC proliferation data for each donor/passage, a change in gene expression per cell proliferation measurement (or slope β₁) could be calculated (Supplementary Table S1). These values for both assays were directly compared to the cutoff slope (β_C) determined by the technical variability for a specific probe divided by the coefficient of variation for a specific assay. A filtering step was used to remove highly variable probes from the data set that would not be good candidates in predicting MSC proliferation. For instances where the absolute value of the slope was observed to be greater than the cutoff slope, the probe was considered a potential candidate of predicting MSC proliferation based on the particular cell proliferation assay (Supplementary Fig. 1D, E). If the absolute value of the calculated slope was less than the cutoff slope, the probe did not pass this filtering step because it was considered too unpredictable. This filtering method resulted in probes that have the greatest amount of correlation and the largest calculated slope for each of the MSC proliferation assays of confluency fraction at 4 days and fraction of EdU+ cells after 6 h (Supplementary Fig. 1F, G).

Unsupervised principal component analysis (PCA) of the microarray data (all 34,555 probes) revealed no distinct clustering of any MSCs by donor or passage when the two MSC donors that did not grow to passage 7 (127756 and PCBM1655) were included in the analysis (Fig. 1A). Using the probes that passed the aforementioned filtering, a statistical analysis with multiplicity adjustment (adjusted P < 0.05) was applied to each of the correlations. When correlating the microarray data with the confluency fraction at 4 days, a total of 643 probes were observed to be significant, whereas 416 probes were found to be significant when correlating the microarray data with the fraction of EdU+ cells after 6 h (Fig. 1B). Of the 342 total significant probes identified in common between both proliferation assays, 24 genes were selected for further analysis that exhibited the greatest correlation and slope (Supplementary Table S2). A hierarchical clustering heat map was used to visualize the changes in gene expression, where the samples are listed from high to low proliferative potential on the x-axis based on the confluency fraction assay (Fig. 1C). It is clear that for a number of genes, expression increases with increases in proliferative potential (positive correlation), while other genes exhibit decreases in expression with increasing proliferative potential (negative correlation). When supervised PCA is performed with the 24 selected genes on the individual MSC donor/passage samples, clustering is observed by proliferation potential instead of by passage (Fig. 1D). A clear distinction is observed between highly proliferative MSC samples, PCBM1641, PCBM1632, 110877, and 167696 at passage 3, compared with the other MSC samples. Furthermore, donor and passage samples clustered much closer to each other compared to the unsupervised PCA, thus exemplifying the minimal technical variability of these genes.

FIG. 1.

PCA of unsupervised microarray data for 21 different samples represented by 62 microarrays (A). Venn diagram of the 342 probes that significantly correlate with both MSC proliferation assays (B). Hierarchical clustering visualizes the gradual increases and decreases of gene expression with the MSC donor/passage samples ordered from highest to lowest based on the proliferative potential of the confluency fraction at 4 days (C). Twenty-four genes with the greatest magnitude of correlation and slope for both MSC proliferation assays are represented by supervised PCA (D). Supervised PCA with 74 probes of interest, including 21 positively correlated proliferative gene markers (blue spheres), 3 negatively correlated proliferative gene markers (red spheres), 3 new housekeeping genes (black spheres), and 15 common housekeeping genes (green spheres) (47 different probes) (E). MSC, multipotent stromal cell; PCA, principal component analysis.

In addition to finding correlative genes with MSC proliferation, similar efforts were made to identify genes consistently expressed by MSCs regardless of proliferative capacity, which could possibly serve as housekeeping genes. Three genes, CYP2B6, NACAD, and ST13, were selected for further analysis as potential housekeeping genes based on their low technical and biological variability and then compared with 15 common housekeeping genes (ACTB, β2M, GAPDH, GUSB, HPRT1, HSP90AB1, PPIA, RPL13A, RPLP0, RPLP1, TBP, TFRC, TUBB, UBC, and YWHAZ) and the 24 potential MSC predictive proliferation genes (Supplementary Table S3). The expression of CYP2B6 was not commonly expressed in all samples through RT-qPCR; therefore, it was not used in further analyses. Through supervised PCA of the 74 probes/genes (Supplementary Table S3), a definitive separation is observed between the positively and negatively correlated genes with the potential new housekeeping genes located between the two groups (Fig. 1E). The 15 common housekeeping gene probes are also distributed in the principal component space between the positively and negatively correlated MSC proliferative genes; however, some are intermingled with the positively correlated genes, indicating that these genes/probes (TUBB) may not be suitable housekeeping genes.

Proliferation data of MSC confirmation set

To confirm the 24 genes identified as potential predictive markers of MSC proliferation, a second set of MSCs from eight donors was expanded to passage 9 (Table 1). Three additional non-MSC cell lines were also evaluated at passage 4/5, where two of the samples were a chondrocyte (HC1994) and osteoblast (HOb1994) cell population isolated from the same donor as HMSC1994; the third cell sample was a human dermal fibroblast cell line. An embryonal carcinoma cell line, NT2, was also evaluated. Each of these cell populations' proliferation potential was characterized quantitatively using confluency fraction, fraction of EdU+ cells, and population doubling time. Most cell samples did not reach 100% confluency after nearly 10 days in culture; however, fetal MSCs HMSC2618 and HMSC1994, and HC1994 (at passage 5) reached 100% confluency by 10 days (Fig. 2A). Based on measuring the confluency fraction at two time points (days 5 and 6), MSCs at the earlier passage demonstrated a significantly greater proliferative potential (P < 0.05) than the later passage MSCs (Fig. 2B, C). MSC cell lines HMSC2618 and HMSC1994 had greater confluency fractions at both passages than all other MSC cell lines, with HMSC2618 having the greatest confluency fraction of all samples. The proliferative potential of non-MSCs was comparable to the MSC donor samples at passage 5, where proliferative potential of HC1994 was only surpassed by HMSC2618. Similar trends were observed for the fraction of EdU+ cells after 6 and 10 h, where passage 5 samples exhibited significantly (P < 0.05) greater cell division than passage 9 samples (Fig. 2D, E]. Cell division after 6 and 10 h with the EdU for the non-MSC samples was comparable to the MSCs at passage 5. Some of the older donor MSCs exhibited little to no cell division at the later passage as measured by the EdU+ fraction.

FIG. 2.

Confluency curves of MSCs at passages 5 and 9, as well as four other non-MSC cell types at passage 5 (A). The differences in confluency fraction at 5 (B) and 6 (C) days are shown. The fraction of EdU+ cells following 6 and 10 hours in culture is shown (D and E, respectively). The population doubling time for the MSCs at different passages and the other cell types is evaluated (F).

Inverse relationships were observed for population doubling time, where doubling time was significantly lower (P < 0.05) at the earlier passage MSCs in comparison to the later passage MSCs (Fig. 2F). By the population doubling time assay, only fetal MSCs, HMSC2618 and HMSC1994, showed substantial proliferation at passage 9 compared with other MSC donors. The population doubling time of the non-MSCs on average was lower than that of the MSCs at passage 5.

Overall, for MSC proliferation assays, proliferation potential was significantly different between early and late passages (Table 2). Comparing early passage to late passage for only adult MSC donors (n = 5), all five MSC proliferation measurements yielded statistically significant results. Nonetheless, significant differences between early and late passages for fetal MSCs (n = 3) were only significant for the measured confluency fraction at 5 days and fraction of EdU+ cells at 6 and 10 h. Comparing the measured differences of MSC proliferation between early and late passages for adult versus fetal MSCs, statistical significance (P < 0.05) was found for fraction of EdU+ cells at both hours (Table 2). This indicated that the differences in MSC proliferation between early and late passages for fetal MSCs are less than those in proliferation observed for MSCs derived from adults.

Table 2.

Proliferation Assay Statistics (P values)

Proliferation assay	P5 versus P9 (all MSC donors)	P5 versus P9 [adult MSC donors (n = 5)]	P5 versus P9 [fetal MSC donors (n = 3)]	Adult versus fetal (difference between P5 and P9)
Confluency fraction at 5 days	0.0003	0.0026	0.0128	NS
Confluency fraction at 6 days	0.0001	0.0012	NS	NS
Fraction of EdU+ cells after 6 hours	0.0027	0.0028	0.0441	0.0067
Fraction of EdU+ cells after 10 hours	0.0014	0.0022	0.0293	0.0077
Population doubling time	0.0012	0.0004	NS	NS

P5, passage 5; P9, passage 9; NS, not significant (P ≥ 0.05).

In addition to analyzing the differences in MSC proliferation between early and late passages, gene expression was also evaluated by multiple normalization techniques using each of the different housekeeping genes. The number of significant genes between passages 4 and 8 per normalization method varied from 16 to 24 significant genes (adjusted P < 0.05) with ranges in fold change from 2 to 32 (Fig. 3A, and Supplementary Table S4). Among all normalization methods, 15 genes were found to be statistically significant in common, where 14 genes were downregulated (P8 < P4) and 1 gene, CDKN1A, was upregulated (P8 > P4) (Fig. 3B). In exploring this gene profile in MSCs, unsupervised PCA of the samples revealed some minor degree of separation between passages 4 and 8 data due to the selection of the 29 genes from the microarray data analysis (Fig. 3C). However, the separation between passages 4 and 8 data is more apparent between the two groups with supervised PCA with the 15 commonly significant genes (Fig. 3E). Similar observations can be made for the unsupervised PCA by genes, where upregulated and downregulated genes are intermixed with the housekeeping genes (Fig. 3D). Once supervised, the 14 downregulated genes are separated from the one remaining upregulated gene by the five housekeeping genes without any intermixing (Fig. 3F).

FIG. 3.

A volcano plot visualizes the number of statistically significant genes between passages 4 and 8 for each of the normalization methods used (A). Hierarchical clustering visualizes the mean differences in gene expression between passages 4 and 8 for the 15 commonly statistically significant genes (B). PCA of unsupervised RT-qPCR data for 16 different MSC samples (P4 and P8) represented by 22 polymerase chain reaction plates (1F3422 P4, HMSC2483 P4, and HMSC2618 P8 were repeated in triplicate) (C). PCA of the unsupervised RT-qPCR data for the 29 different genes is shown (D). Supervised PCAs of the samples (E) and genes (F) are shown when the 15 statistically significant genes common among all normalization methods between passages 4 and 8 are used in the calculation (Supplementary Table S4). RT-qPCR, reverse transcription quantitative polymerase chain reaction.

Correlation of RT-qPCR and MSC proliferation data confirmation set

Filtering of the RT-qPCR data to remove highly variable genes from the data set was performed as previously described with the microarray data. Genes with calculated slopes greater than the cutoff slope were then examined for statistical significance. This aided in the elimination of genes with correlation coefficients and calculated slopes close to zero regardless of the normalization method for the RT-qPCR data or proliferation assay measurement used to complete the correlation (Fig. 4A). For example, a different number of genes were found to have significant correlations with the proliferation data, fraction of EdU+ cells at 10 h, for each of the different normalization methods (Fig. 4B).

FIG. 4.

Using a regression model, a slope and correlation coefficient were calculated per gene for each of the cell proliferation assays for the second set of MSCs. A cutoff slope was determined per gene and per normalization method for all cell proliferation assays using the percent coefficient of variation and the technical variability per gene by RT-qPCR. Genes whose calculated slope from the regression model was greater than the cutoff slope for a given normalization method were considered to have passed this filter and subjected to further statistical analysis. The results presented are from the correlation of RT-qPCR data with the fraction of EdU+ cells after 10 hours (A). A representative Venn diagram indicating that 7 genes were consistently found to be statistically significant of the 8 normalization methods applied when RT-qPCR data were correlated with the fraction of EdU+ cells at 6 hours (B). Supervised PCA of each sample using the 7 common statistically significant genes (C). Supervised PCA of the genes, where each significant gene and housekeeping gene are represented (D). Hierarchical clustering of these 7 significant genes is shown (E). Based on the 5 techniques to observe MSC proliferation, the samples are ranked from left to right in terms of the proliferative quality using the fraction of EdU+ cells after 10 hours, highest to lowest, respectively.

Ultimately, seven genes (BIRC5, CCNA2, CDC20, CDK1, PBK, PLK1, and SPC25) were observed to have common significance among all normalization methods when correlated with fraction of EdU+ cells at 10 h (Supplementary Table S5). These seven genes demonstrated a significant negative correlation between a majority of ΔCT values and other MSC proliferation measurements for all eight normalization methods (Supplementary Table S6). When population doubling time was correlated with gene expression, not all the seven genes were statistically significant for each normalization method (Supplementary Table S5). The median correlation coefficient of all seven genes/normalization methods was highest with the fraction of EdU+ cells after 10 h with a value of −0.789, whereas the confluency fraction at 5 and 6 days had the lowest median correlation value of all the genes/normalization methods, −0.630 and −0.623, respectively.

One gene, IQGAP, was observed to be significantly correlated with confluency fraction after 5 days for all normalizations (median r = −0.648) but not when correlated with other proliferation measurements. Comparing the median of correlation coefficients for the seven genes across different proliferation assays, ACTB and ST13 were the two normalization genes that consistently produced the highest correlation values. The normalization gene NACAD always produced the lowest median correlation coefficient for the seven genes of all the normalization methods.

When supervised PCA was performed by sample for the seven genes, separation was readily apparent between three poorly proliferative (by the EdU assay) MSCs at passage 8, BM2893, BM3018, and HMSC2651, compared with highly proliferative to moderate samples (Fig. 4C). One MSC sample, HMSC2651 at passage 4, is distinctly separated from both clusters in principal component space. Using supervised PCA by gene, the seven negatively correlated genes with proliferation are clearly separated from the housekeeping genes (Fig. 4D). Hierarchical clustering further demonstrates the gene expression correlation as illustrated by ranking the MSCs from the highest proliferative potential (HMSC2618 P4) to the lowest proliferative potential (BM2893 P8) with respect to fraction of EdU+ cells after 10 h (Fig. 4E). The correlation values for the seven genes when correlated with the fraction of EdU+ cells after 10 h are −0.815 (BIRC5), −0.800 (CCNA2), −0.826 (CDC20), −0.787 (CDK1), −0.775 (PBK), −0.826 (PLK1), and −0.779 (SPC25) when ACTB is used for normalization.

To evaluate whether the seven genes can predict the proliferative potential of non-MSCs, four cell lines, HC1994 (chondrocyte), HOb1994 (osteoblast), NHDF271678 (dermal fibroblast), and NT2 (embryonal carcinoma), were evaluated for gene expression and proliferation with the MSC samples. By supervised PCA using the seven identified genes, three of the four non-MSC samples clustered typically with highly proliferative MSC samples at the earlier passage (Fig. 5A). The HOb1994 at passage 4 was distant from the highly proliferative cluster as determined by the fraction of EdU+ cells after 10 h (Fig. 5B). MSC cell lines 1F3422, RB14, and HMSC2483 at passage 8 were slightly separated from the cluster of highly proliferative samples. Even further from the highly proliferative cluster were MSC samples BM2893, BM3018, and HMSC2651 at passage 8, which exhibited the lowest fraction of EdU+ cells after 10 h.

FIG. 5.

Supervised PCA of 7 genes (BIRC5, CCNA2, CDC20, CDK1, PBK, PLK1, and SPC25) for the MSCs and non-MSCs used to evaluate proliferation potential, where the samples are identified by different colors/shapes (A). An identical PCA plot is shown with the color of each sample representing the fraction of EdU+ cells at 10 hours (B). Hierarchical clustering is used to visualize the potentially predictive genes of cell proliferation for both MSCs and non-MSCs (C).

Hierarchical clustering using MSCs and non-MSCs visualizes the gene expression patterning from the highest (HMSC2618 P4) to the lowest (BM2893 P8) ranking proliferative potential by the fraction of dividing cells after 10 h (Fig. 5C). The correlation values for the seven genes when correlated with the fraction of dividing cells after 10 h using all cell types are −0.743 (BIRC5), −0.566 (CCNA2), −0.759 (CDC20), −0.619 (CDK1), −0.792 (PBK), −0.803 (PLK1), and −0.782 (SPC25) when ACTB is used for normalization. Of the seven genes, only two genes, PBK and SPC25, exhibited slightly increased strength in correlation values when non-MSCs were included in the correlation analyses. However, a reduction in the strength of the correlation (based on the absolute value of the correlation coefficient) between gene expression and proliferation assays for the eight different normalization methods was observed for a majority of the values with the inclusion of non-MSCs in the analyses (Supplementary Table S6).

Discussion

The regenerative and therapeutic qualities of bone marrow-derived human MSCs may vary based on the donor, tissue type, isolation, and manufacturing method. Identifying markers that could predict therapeutic qualities of different MSC preparations represents a significant challenge to the field of regenerative medicine. Identifying markers that can predict the proliferation potential of MSCs may help address this challenge. Using gene expression data and proliferation assays, we attempted to identify gene markers that can predict the proliferation potential of MSCs. Our findings indicate that seven genes (BIRC5, CCNA2, CDC20, CDK1, PBK, PLK1, and SPC25) significantly correlate with MSC proliferation potential. These conclusions were drawn based on the experimental design used: (1) two platforms for measuring gene expression, (2) two different sets of MSCs with each set expanded in different media to numerous passages, and (3) multiple assays to measure MSC proliferation potential. Despite differences in normalization methods for the two gene expression platforms, seven genes showed consistent correlations with proliferation. Our data further suggest that the expression of these genes may be used for predicting the proliferation potential of MSC preparations regardless of culture conditions.

In our previous study to examine the differences in gene expression as a result of cellular aging, we found that significant differences observed between passages for both gene expression and cell division were not of consistent magnitude [27]. These observations in combination with respective differences in MSC differentiation, cell size, protein expression, immunosuppression, and epigenetic profiles complicate the process of identifying predictive markers of MSC quality [23,24,27 –29,32 –34]. Instead of evaluating these changes solely by passage, functional changes should be correlated with markers of interest as MSCs may reach senescence at different rates from different donors; thus, all MSC passages are not equal.

In this study, MSC microarray data from eight donors at multiple passages were correlated with two sets of proliferation data: confluency fraction and fraction of dividing cells. Following rigorous filtering and statistical analysis, a collection of 342 markers were identified as candidates that may be predictive of MSC proliferation. Similar efforts have been made by other investigators to evaluate aspects of MSC growth kinetics and senescence [20,35 –41]. In Ren et al., attempts were made to correlate MSC senescence, using β-galactosidase staining, with microarray data to determine the percentage of the maximum life span of clinical bone marrow-derived MSC products [20]. These researchers initially identified 155 age-correlated genes and further narrowed down this set to 24 genes that were used in a prediction model for the age of MSC samples at different passages. In another study by Wagner et al., researchers investigated the effect of long-term culture on replicative senescence of MSCs [40]. In addition to observing increases in cell size and a propensity for osteogenesis with later passages, differences in the expression of 36 genes were found to correlate with the onset of senescence following long-term culture.

In our study, from the 342-gene list identified for MSC proliferative potential using the microarray data, the top 24 genes exhibiting the strongest correlation with MSC proliferation and the greatest magnitude of change were selected for further evaluation. A new set of MSCs consisting of eight different donors expanded from passages 4 to 8 in different media was evaluated for their proliferative potential and gene expression by RT-qPCR. The goal of these experiments was to determine whether the 24 selected genes using the microarray data would maintain similar correlation trends using a different technological platform. One limitation in conducting this analysis was the difference in normalization methods between the microarray and RT-qPCR data. To resolve this issue, three new housekeeping genes that demonstrated consistent expression across all donors and passages in the microarray data were chosen for normalization in comparison to three common housekeeping genes (ACTB, GAPDH, and GUSB). For one of the three new housekeeping genes, no expression was observed and thus was not used in subsequent analyses.

For the majority of applied normalization methods, seven genes were consistently observed to be significantly and positively correlated with each of the MSC proliferation potential assays. These genes, BIRC5, CCNA2, CDC20, CDK1, PBK, PLK1, and SPC25, have previously been found to impact cell cycle and cell division [42 –48]. CCNA2 belongs to the cyclin family and is known to exhibit periodic protein expression throughout the cell cycle, which in turn promotes transitioning between the cell cycle phases, specifically during the late S/G2 phase when bound in a complex to CDK1 [49,50]. Likewise, four additional genes exhibiting high correlation with cell proliferation potential, CDC20, PBK, PLK1, and SPC25, are involved in the mitotic phase of the cell cycle for regulating chromosome alignment and chromosome separation during cell division [51 –54].

The BIRC5 gene, which is a member of the inhibitor of apoptosis gene family associated with binding and inhibiting caspases, also exhibited correlation with MSC proliferation potential [55]. It is typically expressed at low to absent levels in most normal adult tissue; however, a clear expression pattern is present in adult and fetal MSCs based on these results [56,57]. The expression of BIRC5 by MSCs may be due to the culture conditions, which can impact gene expression and MSC behavior. While a number of investigations have identified some of these genes as markers or targets for tumors, our results demonstrate that they may also be beneficial in identifying MSC population with acceptable cell growth [47,52,56,58].

Other genes tested, such as CDKN2A, CENPF, and TPX2, showed promise with their expression correlating with the different proliferation assays for some of the normalization methods, although not as consistently significant as the aforementioned seven genes. Comparing differences in MSC gene expression between passages 4 and 8, 15 genes are observed to be statistically significant (adjusted P < 0.05). Eight of the genes (CDKN1A, CENPF, HAS2, IQGAP3, KIF20B, KIF23, SGOL2, and TPX2) that can distinguish MSCs between early and late passages ultimately are not beneficial in predicting proliferative potential despite their involvement in the cell cycle and cell division.

When the expression of the seven most highly correlated genes is visualized by supervised PCA and hierarchical clustering, a discernable degree of separation is present between highly and lowly proliferative MSCs. Furthermore, individual MSC gene profiles of the seven genes had relatively comparable levels of expression (Fig. 4E). One MSC sample, HMSC2651 P4, exhibited dissimilar levels of expression across the set of seven genes, where lower expression of BIRC5, CDC20, and PLK1 is seen in comparison to the others and was distant from the cluster of highly proliferative MSC samples in the supervised PCA (Fig. 4C). Based on the expression profile of this MSC donor, it appears as an outlier; therefore, it most likely would not be a suitable candidate for further expansion.

Cell-based therapies are not just limited to bone marrow-derived MSCs; instead, researchers have explored the use of stem cells or tissue-specific cells for a given application [59 –65]. To evaluate the ability of the seven identified genes to predict cell proliferation potential, four non-MSC cell types, osteoblast, chondrocyte, dermal fibroblast, and embryonal carcinoma cell line, were included in an additional analysis with the MSC samples. The non-MSCs were characterized at an early passage and exhibited proliferation potentials tantamount or better to early passage MSCs; however, for the majority of correlations between different proliferation assays and normalization methods and genes, the strength of the correlation between gene expression and proliferation potential was reduced (Supplementary Table S6).

Using PCA, non-MSC cell lines typically clustered with MSC samples that exhibited high proliferation potential (Fig. 5A, B); however, one non-MSC cell line, HOb1994, was noticeably separated from the highly proliferative cluster. In contrast to most of the MSC expression profiles of the seven genes, the expression of three genes (SPC25, PLK1, and PBK) was drastically higher than that of the other four genes. Similarly, the expression profile of the NT2 sample, which also had a relatively high proliferative potential, had five genes (CCNA2, CDC20, CDK1, PLK1, and SPC25) expressed much higher than two other genes. Based on these data and limited sample size, we cannot confirm that these seven genes are good candidates for predicting proliferation potential beyond bone marrow-derived MSCs.

The work presented demonstrates that seven genes may be ideal candidates for predicting the proliferative potential of bone marrow-derived MSCs. This is important as many cell-based therapies require cellular expansion to generate sufficient cell numbers. Knowing whether a freshly harvested population of MSC has a high or low proliferative potential can greatly reduce costs associated with their manufacturing processes [66]. This gene expression analysis may be beneficial during the manufacturing process to expand and assess MSC proliferative potential. If the expression of a particular MSC population crosses a cutoff threshold that correlates with minimum lot release proliferation capacity criteria, then its expansion would be terminated. Assays that are used during cell manufacturing processes may be advantageous over ones applied after a freeze–thaw cycle. While the proliferation potential of a population of MSCs is significant, it is not the sole determining factor of its quality. The correlative methods used to identify markers of MSC proliferative potential are applicable to future work needed to identify potential markers that correlate with other important aspects of clinical applications, such as cellular engraftment, survival, motility, differentiation, and immunosuppression. Ultimately, identifying markers as performed using these methods can considerably aid in distinguishing an optimal population of MSCs for a desired therapeutic indication.

Footnotes

Acknowledgments

This project is supported by the Medical Countermeasures Initiative and the FDA Modernizing Science Initiative. We would like to thank Drs. Heba Degheidy and Bharat Joshi for reviewing the article. We would like to acknowledge the members of the FDA Modernizing Science Consortium for their discussion of the presented work.

Author Disclosure Statement

No competing financial interests exist.

References

Baboolal

, Mastbergen

, Jones

, Calder

, Lafeber

and McGonagle

. (2015). Synovial fluid hyaluronan mediates MSC attachment to cartilage, a potential novel mechanism contributing to cartilage repair in osteoarthritis using knee joint distraction. Ann Rheum Dis, 0:8.

Harada

, Nakasa

, Mahmoud

, Kamei

, Adachi

, Deie

and Ochi

. (2015). Combination therapy with intra-articular injection of mesenchymal stem cells and articulated joint distraction for repair of a chronic osteochondral defect in the rabbit. J Orthop Res, 33:1466–1473.

Gao

, Yao

, Zhang

, Lu

, Yu

, Deng

, Yin

, Xia

, Sun

and Wang

. (2010). Human mesenchymal stem cells overexpressing pigment epithelium-derived factor inhibit hepatocellular carcinoma in nude mice. Oncogene, 29:2784–2794.

Kai

, Wang

, Prabhakaran

, Zhang

, Tan

and Ramakrishna

. (2014). Stem cell-loaded nanofibrous patch promotes the regeneration of infarcted myocardium with functional improvement in rat model. Acta Biomater, 10:2727–2738.

Khakoo

, Pati

, Anderson

, Reid

, Elshal

, Rovira

, Nguyen

, Malide

, Combs

, et al. (2006). Human mesenchymal stem cells exert potent antitumorigenic effects in a model of Kaposi's sarcoma. J Exp Med, 203:1235–1247.

, Wang

, Lu

, Pan

and Xu

. (2014). Effects of bone marrow mesenchymal stem cells on hematopoietic recovery and acute graft-versus-host disease in murine allogeneic umbilical cord blood transplantation model. Cell Biochem Biophys, 70:115–122.

Nakamizo

, Marini

, Amano

, Khan

, Studeny

, Gumin

, Chen

, Hentschel

, Vecil

, et al. (2005). Human bone marrow-derived mesenchymal stem cells in the treatment of gliomas. Cancer Res, 65:3307–3318.

Robles

, Liu

, Cao

, Xiang

, Cai

, Manio

, Tang

and Chan

. (2015). Immunosuppressive mechanisms of human bone marrow derived mesenchymal stromal cells in BALB/c host graft versus host disease murine models. Exp Hematol Oncol, 4:13.

Tao

, Cui

, Wang

, Ma

, Wu

, Yi

, Qin

, Liu

, Wang

, et al. (2015). Percutaneous intramyocardial delivery of mesenchymal stem cells induces superior improvement in regional left ventricular function compared with bone marrow mononuclear cells in porcine myocardial infarcted heart. Theranostics, 5:196–205.

10.

Abumaree

, Al Jumah

, Pace

and Kalionis

. (2012). Immunosuppressive properties of mesenchymal stem cells. Stem Cell Rev, 8:375–392.

11.

Baraniak

and McDevitt

. (2010). Stem cell paracrine actions and tissue regeneration. Regen Med, 5:121–143.

12.

Muraglia

, Cancedda

and Quarto

. (2000). Clonal mesenchymal progenitors from human bone marrow differentiate in vitro according to a hierarchical model. J Cell Sci, 113 (Pt 7):1161–1166.

13.

Jin

, Bae

, Kim

, Kwon

, Jeon

, Choi

, Kim

, Yang

, Oh

and Chang

. (2013). Comparative analysis of human mesenchymal stem cells from bone marrow, adipose tissue, and umbilical cord blood as sources of cell therapy. Int J Mol Sci, 14:17986–18001.

14.

Okumachi

, Lee

, Niikura

, Iwakura

, Dogaki

, Waki

, Takahara

, Ueha

, Sakai

, Kuroda

and Kurosaka

. (2015). Comparative analysis of rat mesenchymal stem cells derived from slow and fast skeletal muscle in vitro. Int Orthop, 39:569–576.

15.

Ryu

, Kang

, Park

, Kim

, Sung

, Woo

, Kim

and Kweon

. (2012). Comparison of mesenchymal stem cells derived from fat, bone marrow, Wharton's jelly, and umbilical cord blood for treating spinal cord injuries in dogs. J Vet Med Sci, 74:1617–1630.

16.

Mendicino

, Bailey

, Wonnacott

, Puri

and Bauer

. (2014). MSC-based product characterization for clinical trials: an FDA perspective. Cell Stem Cell, 14:141–145.

17.

Mentink

, Hulsman

, Groen

, Licht

, Dechering

, van der Stok

, Alves

, Dhert

, van Someren

, et al. (2013). Predicting the therapeutic efficacy of MSC in bone tissue engineering using the molecular marker CADM1. Biomaterials, 34:4592–4601.

18.

Platt

, Wilder

, Wells

, Griffith

and Lauffenburger

. (2009). Multipathway kinase signatures of multipotent stromal cells are predictive for osteogenic differentiation: tissue-specific stem cells. Stem Cells, 27:2804–2814.

19.

Williams

, Schuldt

and Muller

. (2011). A guide to stem cell identification: progress and challenges in system-wide predictive testing with complex biomarkers. Bioessays, 33:880–890.

20.

Ren

, Stroncek

, Zhao

, Jin

, Castiello

, Civini

, Wang

, Feng

, Tran

, et al. (2013). Intra-subject variability in human bone marrow stromal cell (BMSC) replicative senescence: molecular changes associated with BMSC senescence. Stem Cell Res, 11:1060–1073.

21.

Gruvberger-Saal

, Cunliffe

, Carr

and Hedenfalk

. (2006). Microarrays in breast cancer research and clinical practice—the future lies ahead. Endocr Relat Cancer, 13:1017–1031.

22.

Gruvberger-Saal

, Eden

, Ringner

, Baldetorp

, Chebil

, Borg

, Ferno

, Peterson

and Meltzer

. (2004). Predicting continuous values of prognostic markers in breast cancer from microarray gene expression profiles. Mol Cancer Ther, 3:161–168.

23.

Lo Surdo

, Millis

and Bauer

. (2013). Automated microscopy as a quantitative method to measure differences in adipogenic differentiation in preparations of human mesenchymal stromal cells. Cytotherapy, 15:1527–1540.

24.

Mindaye

, Ra

, Lo Surdo

, Bauer

and Alterman

. (2013). Global proteomic signature of undifferentiated human bone marrow stromal cells: evidence for donor-to-donor proteome heterogeneity. Stem Cell Res, 11:793–805.

25.

Shao

, Koch

, Gupta

, Lin

, Lenz

, Laufs

, Denecke

, Schmidt

, Linke

, et al. (2013). Induced pluripotent mesenchymal stromal cell clones retain donor-derived differences in DNA methylation profiles. Mol Ther, 21:240–250.

26.

Walter

, Kohli

, Khan

, Major

, Fuller

, Wright

, Kuiper

and Johnson

. (2015). Human mesenchymal stem cells stimulate EaHy926 endothelial cell migration: combined proteomic and in vitro analysis of the influence of donor-donor variability. J Stem Cells Regen Med, 11:18–24.

27.

Bellayr

, Catalano

, Lababidi

, Yang

, Lo Surdo

, Bauer

and Puri

. (2014). Gene markers of cellular aging in human multipotent stromal cells in culture. Stem Cell Res Ther, 5:59.

28.

Lo Surdo

and Bauer

. (2012). Quantitative approaches to detect donor and passage differences in adipogenic potential and clonogenicity in human bone marrow-derived mesenchymal stem cells. Tissue Eng Part C Methods, 18:877–889.

29.

Mindaye

, Ra

, Lo Surdo

, Bauer

and Alterman

. (2013). Improved proteomic profiling of the cell surface of culture-expanded human bone marrow multipotent stromal cells. J Proteomics, 78:1–14.

30.

Vandesompele

, De Preter

, Pattyn

, Poppe

, Van Roy

, De Paepe

and Speleman

. (2002). Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol, 3:12.

31.

Benjamini

and Hochberg

. (1995). Controlling the false discovery rate—a practical and powerful approach to multiple testing. J R Stat Soc Series B-Methodol, 57:289–300.

32.

Lynch

, Thompson

, McGinnis

, Rovira Gonzalez

, Lo Surdo

, Bauer

and Hursh

. (2015). Chromatin changes at the PPAR-gamma2 promoter during bone marrow-derived multipotent stromal cell culture correlate with loss of gene activation potential. Stem Cells, 33:2169–2181.

33.

Nazarov

, Lo Surdo

, Bauer

and Wei

. (2013). Assessment of immunosuppressive activity of human mesenchymal stem cells using murine antigen specific CD4 and CD8 T cells in vitro. Stem Cell Res Ther, 4:128.

34.

Marklein

, Lo Surdo

, Bellayr

, Godil

, Puri

and Bauer

. (2016). High content imaging of early morphological signatures predicts long term mineralization capacity of human mesenchymal stem cells upon osteogenic induction. Stem Cells, 34:13.

35.

Bernardo

, Zaffaroni

, Novara

, Cometa

, Avanzini

, Moretta

, Montagna

, Maccario

, Villa

, et al. (2007). Human bone marrow derived mesenchymal stem cells do not undergo transformation after long-term in vitro culture and do not exhibit telomere maintenance mechanisms. Cancer Res, 67:9142–9149.

36.

Eom

, Oh

, Lee

, Baik

, Rhee

, Shin

, Kim

, Ahn

, Kong

, Kim

and Shim

. (2014). The role of growth factors in maintenance of stemness in bone marrow-derived mesenchymal stem cells. Biochem Biophys Res Commun, 445:16–22.

37.

Hwang

, Park

, Choi

, Lee

, Oh

, Yeo

, Kim

and Sun

. (2013). Age-related characteristics of multipotent human nasal inferior turbinate-derived mesenchymal stem cells. PLoS One, 8:e74330.

38.

Izadpanah

, Kaushal

, Kriedt

, Tsien

, Patel

, Dufour

and Bunnell

. (2008). Long-term in vitro expansion alters the biology of adult mesenchymal stem cells. Cancer Res, 68:4229–4238.

39.

, Xu

, Wang

, Liu

, Li

, Wu

, Jing

, Li

and Wang

. (2015). Senescent mesenchymal stem cells promote colorectal cancer cells growth via galectin-3 expression. Cell Biosci, 5:21.

40.

Wagner

, Horn

, Castoldi

, Diehlmann

, Bork

, Saffrich

, Benes

, Blake

, Pfister

, Eckstein

and Ho

. (2008). Replicative senescence of mesenchymal stem cells: a continuous and organized process. PLoS One, 3:e2213.

41.

, Park

, Kim

, Jin

, Jung

, Ryu

and Jo

. (2014). Characterization of long-term in vitro culture-related alterations of human tonsil-derived mesenchymal stem cells: role for CCN1 in replicative senescence-associated increase in osteogenic differentiation. J Anat, 225:510–518.

42.

Alushin

, Musinipally

, Matson

, Tooley

, Stukenberg

and Nogales

. (2012). Multimodal microtubule binding by the Ndc80 kinetochore complex. Nat Struct Mol Biol, 19:1161–1167.

43.

Ayeni

, Varadarajan

, Mukherjee

, Stuart

, Sprenger

, Srayko

and Campbell

. (2014). Dual phosphorylation of cdk1 coordinates cell proliferation with key developmental processes in Drosophila. Genetics, 196:197–210.

44.

Bendris

, Loukil

, Cheung

, Arsic

, Rebouissou

, Hipskind

, Peter

, Lemmers

and Blanchard

. (2012). Cyclin A2: a genuine cell cycle regulator?. Biomol Concepts, 3:535–543.

45.

Chen

, Yu

, Fu

, Li

, Niu

, Liao

and Lu

. (2010). Functional roles of PC-PLC and Cdc20 in the cell cycle, proliferation, and apoptosis. Cell Biochem Funct, 28:249–257.

46.

Dou

, Wei

, Sun

, Shao

, Childress

, Yang

and Lin

. (2015). PBK/TOPK mediates geranylgeranylation signaling for breast cancer cell proliferation. Cancer Cell Int, 15:27.

47.

Ito

, Yoshida

, Matsuzuka

, Matsuura

, Nakamura

, Nakamine

, Kakudo

, Kuma

and Miyauchi

. (2004). Polo-like kinase 1 (PLK1) expression is associated with cell proliferative activity and cdc2 expression in malignant lymphoma of the thyroid. Anticancer Res, 24:259–263.

48.

Wang

, Zheng

, Shen

and Shi

. (2012). MicroRNA-203 suppresses cell proliferation and migration by targeting BIRC5 and LASP1 in human triple-negative breast cancer cells. J Exp Clin Cancer Res, 31:58.

49.

Brown

, Korolchuk

, Martin

, Stanley

, Moukhametzianov

, Noble

and Endicott

. (2015). CDK1 structures reveal conserved and unique features of the essential cell cycle CDK. Nat Commun, 6:6769.

50.

Jernman

, Valimaki

, Hagstrom

, Louhimo

, Haapasalo

, Arola

and Haglund

. (2014). Cyclin a predicts metastatic potential of rectal neuroendocrine tumors. Hum Pathol, 45:1605–1609.

51.

Baumann

, Korner

, Hofmann

and Nigg

. (2007). PICH, a centromere-associated SNF2 family ATPase, is regulated by Plk1 and required for the spindle checkpoint. Cell, 128:101–114.

52.

Mao

, Gujar

, Mahlokozera

, Chen

, Pan

, Luo

, Brost

, Thompson

, Turski

, et al. (2015). A CDC20-APC/SOX2 signaling axis regulates human glioblastoma stem-like cells. Cell Rep, 11:1809–1821.

53.

McCleland

, Kallio

, Barrett-Wilt

, Kestner

, Shabanowitz

, Hunt

, Gorbsky

and Stukenberg

. (2004). The vertebrate Ndc80 complex contains Spc24 and Spc25 homologs, which are required to establish and maintain kinetochore-microtubule attachment. Curr Biol, 14:131–137.

54.

Shinde

, Gangula

, Kavela

, Pandey

and Maddika

. (2013). TOPK and PTEN participate in CHFR mediated mitotic checkpoint. Cell Signal, 25:2511–2517.

55.

Yano

, Otsuka

, Hirano

, Uenami

, Satomi

, Kuroyama

, Niinaka

, Yoneda

, Kimura

, et al. (2015). Nuclear survivin expression in small cell lung cancer. Anticancer Res, 35:2935–2939.

56.

Danilewicz

, Stasikowska-Kanicka

and Wagrowska-Danilewicz

. (2015). Augmented immunoexpression of survivin correlates with parameters of aggressiveness in prostate cancer. Pol J Pathol, 66:44–48.

57.

Schimmer

. (2004). Inhibitor of apoptosis proteins: translating basic knowledge into clinical practice. Cancer Res, 64:7183–7190.

58.

Ayllon

and O'Connor

. (2007). PBK/TOPK promotes tumour cell proliferation through p38 MAPK activity and regulation of the DNA damage response. Oncogene, 26:3451–3461.

59.

Hoben

, Yan

, Iyer

, Newton

, Hunter

, Moore

, Sakiyama-Elbert

, Wood

and Mackinnon

. (2015). Comparison of acellular nerve allograft modification with Schwann cells or VEGF. Hand (N Y), 10:396–402.

60.

Kern

, Loeckermann

, Fritsch

, Hausser

, Roth

, Magin

, Mack

, Muller

, Paul

, Ruther

and Bruckner-Tuderman

. (2009). Mechanisms of fibroblast cell therapy for dystrophic epidermolysis bullosa: high stability of collagen VII favors long-term skin integrity. Mol Ther, 17:1605–1615.

61.

Kuramoto

, Takagi

, Ishitani

, Shimizu

, Okano

and Matsui

. (2015). Preventive effect of oral mucosal epithelial cell sheets on intrauterine adhesions. Hum Reprod, 30:406–416.

62.

Mehnert

, Kisch

and Brandenburger

. (2014). Co-culture systems of human sweat gland derived stem cells and peripheral nerve cells: an in vitro approach for peripheral nerve regeneration. Cell Physiol Biochem, 34:1027–1037.

63.

Okuizumi

, Tohyama

, Kondo

and Yasuda

. (2004). The effect of cell-based therapy with autologous synovial fibroblasts activated by exogenous TGF-beta1 on the in situ frozen-thawed anterior cruciate ligament. J Orthop Sci, 9:488–494.

64.

Saito

, Tamaki

, Hirata

, Hashimoto

, Nakazato

, Nakajima

, Kazuno

, Sakai

, Iida

and Okami

. (2015). Reconstruction of multiple facial nerve branches using skeletal muscle-derived multipotent stem cell sheet-pellet transplantation. PLoS One, 10:e0138371.

65.

Schinhan

, Gruber

, Dorotka

, Pilz

, Stelzeneder

, Chiari

, Rossler

, Windhager

and Nehrer

. (2013). Matrix-associated autologous chondrocyte transplantation in a compartmentalized early stage of osteoarthritis. Osteoarthritis Cartilage, 21:217–225.

66.

Carmen

, Burger

, McCaman

and Rowley

. (2012). Developing assays to address identity, potency, purity and safety: cell characterization in cell therapy process development. Regen Med, 7:85–100.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.21 MB

0.02 MB

0.08 MB

0.02 MB

0.03 MB

0.16 MB