Abstract
Research on human pluripotent stem cells has been hampered by the lack of a standardized, quantitative, scalable assay of pluripotency. We previously described an assay called ScoreCard that used gene expression signatures to quantify differentiation efficiency. Here we report an improved version of the assay based on qPCR that enables faster, more quantitative assessment of functional pluripotency. We provide an in-depth characterization of the revised signature panel (commercially available as the TaqMan hPSC Scorecard Assay) through embryoid body and directed differentiation experiments as well as a detailed comparison to the teratoma assay. We further show that the improved ScoreCard enables a wider range of applications, such as screening of small molecules, genetic perturbations and assessment of culture conditions. Our approach can be extended beyond stem cell applications to characterize and assess the utility of other cell types and lineages.
This is a preview of subscription content, access via your institution
Access options
Subscribe to this journal
Receive 12 print issues and online access
$209.00 per year
only $17.42 per issue
Buy this article
- Purchase on Springer Link
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Accession codes
References
Takahashi, K. et al. Induction of pluripotent stem cells from adult human fibroblasts by defined factors. Cell 131, 861–872 (2007).
Takahashi, K. & Yamanaka, S. Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676 (2006).
Daley, G.Q. The promise and perils of stem cell therapeutics. Cell Stem Cell 10, 740–749 (2012).
Dolgin, E. Putting stem cells to the test. Nat. Med. 16, 1354–1357 (2010).
Müller, F.J., Goldmann, J., Löser, P. & Loring, J.F. A call to standardize teratoma assays used to define human pluripotent cell lines. Cell Stem Cell 6, 412–414 (2010).
Müller, F.J. et al. A bioinformatic assay for pluripotency in human cells. Nat. Methods 8, 315–317 (2011).
Bock, C. et al. Reference Maps of human ES and iPS cell variation enable high-throughput characterization of pluripotent cell lines. Cell 144, 439–452 (2011).
Wang, Z., Oron, E., Nelson, B., Razis, S. & Ivanova, N. Distinct lineage specification roles for NANOG, OCT4, and SOX2 in human embryonic stem cells. Cell Stem Cell 10, 440–454 (2012).
Gifford, C.A. et al. Transcriptional and epigenetic dynamics during specification of human embryonic stem cells. Cell 153, 1149–1163 (2013).
Lipták, T. On the combination of independent tests. Magyar Tud Akad Mat Kutato Int Közl. 3, 171–196 (1958).
Boulting, G.L. et al. A functionally characterized test set of human induced pluripotent stem cells. Nat. Biotechnol. 29, 279–286 (2011).
Stouffer, S., DeVinney, L. & Suchmen, E. The American Soldier: Adjustment During Army Life (Princeton University Press; Princeton, 1949).
Tsankov, A.M. et al. Transcription factor binding dynamics during human ES cell differentiation. Nature 518, 344–349 (2015).
Liao, J. et al. Targeted disruption of DNMT1, DNMT3A and DNMT3B in human embryonic stem cells. Nat. Genet. 47, 469–478 (2015).
Borowiak, M. et al. Small molecules efficiently direct endodermal differentiation of mouse and human embryonic stem cells. Cell Stem Cell 4, 348–358 (2009).
Teo, A.K. et al. Pluripotency factors regulate definitive endoderm specification through eomesodermin. Genes Dev. 25, 238–250 (2011).
Ziller, M.J. et al. Dissecting neural differentiation regulatory networks through epigenetic footprinting. Nature 518, 355–359 (2015).
Lovén, J. et al. Selective inhibition of tumor oncogenes by disruption of super-enhancers. Cell 153, 320–334 (2013).
Pikkarainen, S., Tokola, H., Kerkelä, R. & Ruskoaho, H. GATA transcription factors in the developing and adult heart. Cardiovasc. Res. 63, 196–207 (2004).
Choi, J. et al. A comparison of genetically matched cell lines reveals the equivalence of human iPSCs and ESCs. Nat. Biotechnol. 10.1038/nbt.3388 (2015).
Acknowledgements
We would like to thank all members of the Meissner laboratory as well as D. Tieberg and U. Lakshmipathy for their valuable support and insightful feedback, L. Solomon and L. Gaffney for graphical support, I.S. Kim and B. Bernstein for the JQ1 inhibitor molecule and insight, L. Rubin for the partially reprogrammed cell lines and C. MacGillivray (HSCI Histology core) for teratoma sectioning, fixation and histochemical staining for markers of the three germ layers. A.M.T. is supported by the NIH under Ruth L. Kirschstein National Research Service Award (NRSA) fellowship 5F32DK095537. This work was supported by the National Institute of General Medical Sciences (NIGMS) (P01GM099117), National Human Genome Research Institute (NHGRI) (1P50HG006193), the New York Stem Cell Foundation, and a research grant from Life Technologies. A.M. is a New York Stem Cell Foundation, Robertson Investigator.
Author information
Authors and Affiliations
Contributions
A.M.T. and A.M. conceived and designed the study. A.M.T. optimized all experiments and performed all analysis, V.A. performed cell culture and ran qPCR plates, R.P. performed FACS analysis and ran qPCR plates, C.A.G. trained A.M.T. in cell culture and provided experimental advice, S.C. performed PiPSC embryoid body experiments with advice from L.D., N.M.T. performed all pathological and histological annotations, A.M.T. and A.M. interpreted the data and wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
Harvard University has filed a patent on the original ScoreCard technology and a licensing agreement with Life Technologies (now part of Thermo Fisher). A.M. is an advisor to Life Technologies.
Integrated supplementary information
Supplementary Figure 1 Karyotyping and teratoma assay H&E stainings
a. Summary of karyotyping for the five hESC lines used in Figure 1 of this study (inv(x)=inversion on chromosome x; dup=duplication; t=translocation; [y] denotes y number of cells tested). Parenthesis in “Results” column contains the number of normal/abnormal cells out of the total counted cells.
b. Hematoxylin and eosin (H&E) stained teratoma sections show presence of tissue from the three embryonic germ layers (ectoderm, EC; mesoderm, ME; endoderm, EN; ME*=immature cartilage, EC*=neuroectoderm) for five cell lines discussed in the main text.
Supplementary Figure 2 Validation of pathological annotations using IHC
Immunohistochemical (IHC) stainings for FOXA2 (left, EN), HAND2 (middle, ME), and PAX6 (right, EC/EN) antibodies of sections from five selected teratomas. Stains for markers of the three germ layers were used to further validate the pathologist annotations of different tissue structures. Hematoxylin counterstain was used to show nuclei. Distances for magnification used are displayed in the bottom right. H9 teratoma1 and H1 teratoma1 replicates were fixed while H9 teratoma3 and H1 teratoma3 were unfixed before freezing. IHC can be variable due to amount of fixation as it makes it harder to retrieve the antigen, and antibodies at the same dilution can have different sensitivity and specificity to these antigens. This variability highlights the fact that it is difficult to use IHC for quantifying the germ layer contribution in a teratoma reproducibly.
Supplementary Figure 3 Overview of gene selection for the qPCR expression assay
a. List of gene markers selected for the qPCR gene expression panel. Control (CT) genes were selected for data normalization and flagging sample contamination (only 4 probes of 9 shown). Markers of PL, EN, ME, and EC were selected based on uniqueness of expression for the corresponding cell type (Online Methods) using RNA-seq data for hESC line HUES64 and directed differentiation into endoderm (dEN), mesoderm (dME), and ectoderm (dEC). In addition we included several mesendoderm (MS) markers and other genes of interest. Columns on the right show the mean difference in expression for each selected gene at day 5 and day 12 of embryoid body (EB) formation relative to day 0, averaged across 23 reference cell lines. FPKM, Fragments Per Kilobase of transcript per Million mapped reads; ΔΔCt, difference in threshold cycle (Ct) normalized per experiment to housekeeping genes.
b. Mean expression uniqueness for the germ layer gene sets used in Bock et al. and for the four gene classes used in this study measured using RNA-seq data of the three germ layers. Germ layer expression uniqueness for each gene was measure by calculating the difference in log2(FPKM) between the expression of each marker gene and the mean (left) or max (right) expression of that marker genes in the other three cell types (Online Methods). Error bars show the standard deviation of expression uniqueness for markers within each gene set.
c. Marker expression uniqueness in HUES64 versus mean expression uniqueness averaged across 11 hESC lines based on qPCR expression data presented later in this study for directed differentiation into the three germ layers. Although HUES64 was initially used for germ layer marker selection, we observe high correlation of expression uniqueness calculated using HUES64 data and mean expression uniqueness computed using all other cell line data (R =.83, P < 10-20, Pearson correlation).
Supplementary Figure 4 Reproducibility of EB differentiation and expanded characterization of differentiation potentialEB
a. Pearson correlation coefficient (R) matrices of expression levels between samples with different RNA input (4ug versus 1ug, top) and between plates from three different lots (bottom) both show high technical reproducibility (R >.96, legend on right).
b. Pearson correlation coefficient matrices of day 12 EB gene expression levels from three cell lines maintained for >10 passages (rows/columns). Legend for R is shown on the right.
c. Left/Middle: Pearson correlation coefficient (R) matrices of RNA expression levels from biological replicates for two cell lines (H1 and HUES1) at different time points (days 0, 2, 5, 12) of EB differentiation. Right: Pearson correlation coefficient matrix of RNA expression levels from biological replicates of day 12 EB experiments for four cell lines (iPS27e, H9, H1, and HUES8). Biological replicates correlate at R >.92.
d. Ratio of between to within group variance for germ layer differentiation potential using no data normalization (left), ACTB probes normalization typically used in qPCR analysis (middle), and our control gene normalization (right). Germ layer variance ratios are shown using different colored bars. Variation in differentiation potential between biological replicates of the same cell line (within-group) decreased after gene normalization, when compared to variation in differentiation potential between different cell lines (between-group).
e. Heat map of the unweighted differentiation potentialEB for EC, ME, EN, and PL gene classes at day 0, 2, 5, and 12 of EB formation.
f. Heat map of combined P values corresponding to the unweighted differentiation potentialEB for all three germ layers for day 0, 2, 5, and 12 EB data.
g. Box plots show the distribution of differentiation potentialEB for 23 reference hPSC lines at day 0, 5, and 12 of EB differentiation. Red circles represent all three partially reprogrammed iPS (PiPS) cell lines tested. PiPS differentiation potentialEB is clearly distinct from the reference hPSC lines at day 12 (right) of EB differentiation.
Supplementary Figure 5 Fitting differentiation potentialEB using Lasso regression algorithm
a. Plot of Mean Squared Error (MSE) as a function of decreasing Lambda values when using Lasso regression algorithm fitted to the EC (leftmost), ME, EN, and PL (rightmost) differentiation potential of day 12 EB reference, using all the data (Online Methods). For estimating the variance in MSE of each lambda Lasso fit, five-fold cross validation was used (80% of the data was used for training and 20% for testing). Grey error bars indicate the standard error of the MSE computed using the cross-validation for each lambda fit. The green circles indicate the Lambda with a minimum MSE. The blue circles indicate the largest lambda such that the MSE is within one standard error of the minimum MSE. This lambda makes the sparsest model within one standard error of the minimum MSE, and its fitted coefficients were chosen for all subsequent analysis.
b. Linear regression shows near perfect correlation between EC (leftmost), ME, EN, and PL (rightmost) differentiation potentialEB and the Lasso fitted differentiation potentialEB using the coefficients chosen in part a. The positive y-intercept value indicates that differentiation potentialEB calculated using the weighted Z-method is higher than the Lasso fitted differentiation potentialEB and thus provides more statistical power in detecting differential expression of EB formation when compared to the pluripotent state (differentiation potentialEB = 0).
c. Bar graph displays the Lasso coefficient values (chosen in part a) learned when fitting EC (leftmost), ME, EN, and PL (rightmost) differentiation potentialEB. The number of markers needed to achieve a near perfect fit are shown in parenthesis (EC=12, ME=8, EN=9, PL=5).
d. Heat map of the combined P values corresponding to the differentiation potentialEB for all three germ layers at day 12 of EB formation for the reference set using day 12 EB weights (left) and using the Lasso fit coefficients (right). The Lasso fit combined P values are reduced.
Supplementary Figure 6 Expanded characterization of EB differentiation dynamics
a. Quantification of cell line repression dynamics for pluripotent genes. Heat map shows the ratio of day 5 to day 12 gene expression, averaged across all PL markers.
b. Quantification of cell line dynamics during EB differentiation. Heat map shows the ratio of day 5 to day 12 differentiation potentialEB for the three germ layers (left) and pluripotent markers (right). Ratios near or above 1 indicate rapid germ layer induction/repression, while near or below 0 indicate a delayed response.
c. Linear regression shows a high correlation (R >.95) of EB dynamics in all three germ layers when measured by either mean expression or differentiation potentialEB (day 5 to day 12 ratio).
d. Scatter plot of the day 5 EB weights versus day 2 EB weights. Genes below the diagonal line y=x (e.g. HHEX, PAPLN, SDC2, CABP7, DMBX1) are transiently expressed in day 2. Genes along the diagonal line y=x are rapidly induced (e.g. WNT1, PAX3, NR2F2). Genes above the diagonal are gradually (e.g. POU4F1) or slowly induced (e.g. HAND2, RGS4, COL2A1, ABCA4).
e. Scatter plot of the day 12 EB weights versus day 2 EB weights shows similar trends as the day 5 versus day 2 EB weight scatter plot.
Supplementary Figure 7 Expanded characterization of directed differentiation of hPSCs
a. Schematic of gene selection (left), 2D directed differentiation experiments (middle) and computational technique (right) used for quantifying 2D differentiation potential for pluripotency, endoderm, mesoderm and ectoderm gene classes.
b. Heat maps of the unweighted differentiation potential2D (left) and corresponding combined P values (right) of 14 cell lines for the different gene classes.
c. Heat map of the differentiation potential2D for the four gene classes (EC, ME, EN, PL) for dEC (left), dME, dEN directed differentiation experiments, and hPSC (right) undifferentiated, control samples.
d. Combined P values corresponding to the differentiation potential2D in Supplementary Figure 7c.
e. Left: Box plots illustrate the distribution of differentiation potential2D for the four gene sets (EC, ME, EN, and PL) of the 14 reference dEN experiments. Circles represent HUES64 dEN differentiation potential for unsorted (black) and CD184+ sorted (red) populations. Right: Heat map shows the differentiation potential for unsorted and sorted populations of HUES64 dEN experiments. CD184+ sorted cells have a decreased ME and PL signal, as expected. Legend for differentiation is displayed in the bottom right.
Supplementary Figure 8 Expanded characterization of hPSCs directed differentiation and of JQ1’s effect on lineage bias
a. Linear regression shows a slightly lower correlation between 2D efficiency and differentiation potential2D when using the differentiated state as a reference.
b. Linear regression shows no positive correlation with EN efficiency when using the unweighted Z-transform test to calculate EN differentiation potential2D.
c. Linear regression shows a lower correlation between 2D efficiency and differentiation propensity when averaging t score as done in the ScoreCard method but using our gene signature panel.
d. Linear regression shows low correlation between 2D differentiation potential (top) at day 5 versus day 0 and between EB differentiation potential (bottom) at day 12 versus day 0 for EC (left), ME (middle), and EN (left) gene sets.
e. Box plots show the distribution of differentiation potentialEB for 23 reference hPSC lines at day 12 of EB differentiation. Circles represent the differentiation potentialEB of HUES64 day 12 EBs when treated with different concentrations of JQ1 bromodomain inhibitor (100nM, 200nM, 500nM, and 0nM untreated control) during differentiation. We observe increased EN differentiation potentialEB (P <.01, weighted Z-test) in EBs treated with JQ1 as well as decreased EC and ME differentiation potentialEB relative to the untreated control.
Supplementary Figure 9 Expanded characterization of tissue culture quantification and genomic assays summary
a. Scatter plot of day 12 EB weights versus 2D weights. Labeled genes above versus below the diagonal line y = x are more highly weighted in EB versus 2D differentiation potential calculations.
b. Box plot of the distribution of difference in differentiation potential2D between feeder-free and feeder culture for 11 matching undifferentiated hPSC lines.
c. Table compares PluriTest, the previous ScoreCard, and our improved qPCR ScoreCard methodology against several different criteria (leftmost column). Assay costs estimates as of June 15th, 2015 include the price of Illumina HT12 array (PluriTest), Nanostring Technology codeset (ScoreCard), and Life Technology TaqMan® hPSC Scorecard™ 384-well Kit (qPCR ScoreCard). Buying Life Technology plates and reagents for 8 samples reduces the cost to $150/sample.
Supplementary Figure 10 Fitting differentiation potential2D and 2D efficiency using Lasso regression algorithm
a. Plot of Mean Squared Error (MSE) as a function of decreasing Lambda values when using Lasso regression algorithm fitted to the EC (leftmost), ME, EN, and PL (rightmost) differentiation potential of the 2D reference set using all the data (Online Methods). For estimating the variance in MSE of each lambda Lasso fit, five-fold cross validation was used (80% of the data was used for training and 20% for testing). Grey error bars indicate the standard error of the MSE computed using the cross-validation for each lambda fit. The green circles indicate the Lambda with a minimum MSE. The blue circles indicate the largest lambda such that the MSE is within on standard error of the minimum MSE. This lambda makes the sparsest model within one standard error of the minimum MSE, and its fitted coefficients were chosen for all subsequent analysis.
b. Linear regression shows near perfect correlation between EC (leftmost), ME, EN, and PL (rightmost) differentiation potential2D and the Lasso fitted differentiation potential2D using the coefficients chosen in part a. The small y-intercept value indicates that differentiation potential2D calculated using the learned Lasso coefficients provides similar statistical power to the weighted Z-method in detecting differential expression of 2D differentiation relative to the pluripotent state (differentiation potential2D = 0).
c. Bar graph displays the Lasso coefficient values (chosen in part a) learned when fitting EC (leftmost), ME, EN, and PL (rightmost) differentiation potential2D. The numbers of markers needed to achieve a near perfect fit and similar statistical power are shown in parenthesis (EC=6, ME=8, EN=7, PL=6).
d. Plot of Mean Squared Error (MSE) as a function of decreasing Lambda values when using Lasso regression algorithm fitted to the EC (leftmost), ME, and EN (rightmost) directed differentiation efficiency, as measured using FACS. MSE for each fit was calculated using all the data. For estimating the variance in MSE of each lambda Lasso fit, five-fold cross validation was used. Grey error bars indicate the standard error of the MSE computed using the cross-validation for each lambda fit. The green circles indicate the Lambda with a minimum MSE. The blue circles indicate the largest lambda such that the MSE is within on standard error of the minimum MSE. This lambda makes the sparsest model within one standard error of the minimum MSE, and its fitted coefficients were chosen for all subsequent analysis related to 2D efficiency.
e. Linear regression shows high correlation between EC (leftmost), ME, and EN (rightmost) 2D efficiency measures and the Lasso fit for 2D efficiency using the coefficients chosen in part d.
f. Bar graph displays the Lasso coefficient values (chosen in part d) used to fit EC (leftmost), ME, and EN (rightmost) 2D efficiency measurements. The numbers of markers needed to achieve the fit (R ≥ 0.94) are shown in parenthesis (EC=2, ME=3, EN=2).
Supplementary information
Rights and permissions
About this article
Cite this article
Tsankov, A., Akopian, V., Pop, R. et al. A qPCR ScoreCard quantifies the differentiation potential of human pluripotent stem cells. Nat Biotechnol 33, 1182–1192 (2015). https://doi.org/10.1038/nbt.3387
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/nbt.3387
This article is cited by
-
Effects of low doses of methylmercury (MeHg) exposure on definitive endoderm cell differentiation in human embryonic stem cells
Archives of Toxicology (2023)
-
Single-cell transcriptomics of human iPSC differentiation dynamics reveal a core molecular network of Parkinson’s disease
Communications Biology (2022)
-
Endogenous suppression of WNT signalling in human embryonic stem cells leads to low differentiation propensity towards definitive endoderm
Scientific Reports (2021)
-
Engineering islets from stem cells for advanced therapies of diabetes
Nature Reviews Drug Discovery (2021)
-
Diminished expression of major histocompatibility complex facilitates the use of human induced pluripotent stem cells in monkey
Stem Cell Research & Therapy (2020)
