Multi-omic Profiling Reveals Dynamics of the Phased Progression of Pluripotency

Abstract

Pluripotency is highly dynamic and progresses through a continuum of pluripotent stem cell states. The two states that bookend the pluripotency continuum, naive and primed, are well characterized, but our understanding of the intermediate states and transitions between them remains incomplete. Here, we dissect the dynamics of pluripotent state transitions underlying pre- to post-implantation epiblast differentiation. Through comprehensive mapping of the proteome, phosphoproteome, transcriptome, and epigenome of embryonic stem cells transitioning from naive to primed pluripotency, we find that rapid, acute, and widespread changes to the phosphoproteome precede ordered changes to the epigenome, transcriptome, and proteome. Reconstruction of the kinase-substrate networks reveals signaling cascades, dynamics, and crosstalk. Distinct waves of global proteomic changes mark discrete phases of pluripotency, with cell-state-specific surface markers tracking pluripotent state transitions. Our data provide new insights into multi-layered control of the phased progression of pluripotency and a foundation for modeling mechanisms regulating pluripotent state transitions (www.stemcellatlas.org).

Keywords: EpiLC; MAPK/ERK; embryonic development; embryonic stem cells; epiblast; formative pluripotency; mTOR; pluripotency; protein phosphorylation; signaling.

Figure 1. High temporal-resolution profiling of the proteome, phosphoproteome, transcriptome, and epigenome during ESC to EpiLC transition.

(A) Developmental events during embryogenesis in mouse embryos. ICM, inner cell mass; ESCs, embryonic stem cells; TE, tropectoderm; PE, primitive endoderm; EpiLC, epiblastlike cells; EpiSCs, epiblast stem cells; PGCs, primordial stem cells. (B) Schematic showing EpiLC induction from ESCs grown in 2i+LIF medium. Proteome, phosphoproteome, transcriptome, and epigenome were profiled at indicated time-points. Phase contrast images correspond to representative ESC colony grown in 2i+LIF medium (left) and cells undergoing morphological changes at 72h post EpiLC induction (right). (C) Schematic of mass spectrometry (MS)-based experimental protocols used for proteome and phosphoproteome profiling. (D) Summary statistics of proteins, phosphosites, transcripts, and epigenetic marks profiled. See also Figure S1 and S2.

Figure 2. Temporal dynamics of the proteome, phosphoproteome, and transcriptome during ESC to EpiLC transition

(A-C) Principal component analysis (PCA) of the transcriptome (A), proteome (B), and phosphoproteome (C) during EpiLC induction. Each circle represents data from a sample collected at a particular time point during ESC to EpiLC transition, with lighter and darker shades of purple denoting earlier and later time points, respectively. Filled green squares represent transcriptomic data from Kalkan et al. (Kalkan et al., 2017). (D) Temporal dynamics of global changes in the proteome, phosphoproteome, transcriptome, and epigenome during ESC to EpiLC transition. Changes in the phosphorylation level of a given phosphosite were normalized to the changes in corresponding protein level. (E) Density plot showing the distribution of magnitude of changes at the protein, mRNA, and phosphosite level. Changes in phosphosite levels were normalized as in (D). (F) Fraction of phosphosites, mRNAs, and proteins dynamically regulated during EpiLC induction, as assessed using ANOVA test. Changes in phosphosite levels were normalized as in (D). (G) Venn diagram showing overlap among genes encoding differentially regulated mRNAs, proteins, and/or phosphosites during ESC to EpiLC transition. Only genes with both protein and mRNA levels quantified were used for this analysis. See also Figure S3 and S4.

Figure 3. Characterization of signaling dynamics during ESC to EpiLC transition and prediction of substrates for key kinases involved.

(A) Clustering of phosphosites based on their temporal dynamics. Four clusters (out of the twelve, see Figure S6A, B) enriched for known substrates of ERK and S6K/RSK (blue), mTOR (green), p38a (orange), or AKT (purple) are shown. Select substrates are highlighted. P-values, Fisher’s exact test. (B) Heatmap representation of the data shown in (A). (C) Gene ontology (GO) analysis of phosphoproteins represented in each of the four clusters in (A). Top five enriched GO categories (biological processes) are shown. Select phosphoproteins within each group are highlighted at the top. (D) Temporal dynamics of relative phosphorylation levels (compared to 0h) of Erk2 (T183/Y185) and Erk1 (T203/Y205, during ESC to EpiLC transition, as quantified using LCMS/MS. (E) Western blot analysis of total and phosphorylated ERK1/2 during ESC to EpiLC transition. Histone H3 is used as loading control. (F) Temporal dynamics of relative protein and mRNA levels (compared to 0h), as quantified using LC-MS/MS and RNA-Seq respectively, of Erk2 and Erk1 during ESC to EpiLC transition. (G)Same as in (F) but showing data for Dusp6, Spry5, and Spred1, all downstream transcriptional targets of ERK1/2 signaling. (H) RT-qPCR analysis of relative expression of genes associated with naïve pluripotent state (right) or post-implantation epiblasts in EpiLCs (48h) compared to naïve ESCs (0h). During the ESC to EpiLC transition (0–48h), cells were left untreated or cultured in the presence of PD0325901 or Rapamycin for indicated time period. Data, normalized to Actin, represents mean of n = 3 biological replicates. Error bars represent SEM. *p < 0.05 (Student’s t-test, two-sided). (I) Violin plots showing the distribution of ensemble prediction scores of all profiled phosphosites indicating the likelihood of they being a substrate of one of the five kinases (S6K/RSK, ERK, mTOR, p38a, AKT); kinases other than this five were grouped into the ‘other’ category. Ensemble score for each kinase-substrate pair was generated using a positive-unlabelled ensemble algorithm (Yang et al., 2016a). Black crosses (‘x’) represent previously known substrates. (J) Sequence motifs enriched within predicted substrates for ERK, S6K/RSK, AKT, mTOR, or p38a kinases. Motifs were identified using IceLogo (Colaert et al., 2009), using precompiled mouse Swiss-Prot sequence composition as the reference set. Y-axis represents the difference in the frequency of an amino acid in the experimental vs the reference set. (K)Temporal profiles of predicted substrates for ERK, S6K/RSK, AKT, mTOR, or p38a kinases. Mean and the standard deviation are shown as line-plot and range, respectively. See also Figure S5.

Figure 4. Comparative analysis of the proteome and transcriptome during ESC to EpiLC transition.

(A) Temporal dynamics of correlation (y-axis) between fold-changes in protein (compared to 0 h data) and fold-changes in mRNA (compared to 0 h) over time (x-axis). See Figure S6B for the actual scatter plots showing correlation at various time-points. (B) Gene ontology (GO) analysis of genes upregulated or downregulated (at both protein and mRNA levels) at 72h vs 0h during ESC to EpiLC transition. Select GO categories (biological processes) enriched among upregulated or downregulated genes are shown. (C) Temporal dynamics of relative protein and mRNA levels (compared to 0h) of select genes. Genes associated with naïve pluripotent state (Esrrb, Tfcp2l1, and Prdm14), primed pluripotent state (Dnmt3a and Otx2), and those whose expression is relatively stable during ESC to EpiLC transition (Jarid2 and Oct4) are shown. (D) Correlation between changes in protein and mRNA levels (y-axis) for individual genes plotted against their relative rank-order (x-axis) in terms of change in gene (mRNA/protein) expression (72h vs. 0h) (see Methods). Genes that were substantially down-regulated at 72h vs. 0h have smaller ranks (positioned to the left) and those that were substantially up-regulated have higher ranks (right). Select transcriptional and chromatin regulators, associated with naïve pluripotent state (ESCs), that are downregulated during EpiLC induction are highlighted as filled red circles; those, associated with primed pluripotent state (EpiLCs), which are upregulated during EpiLC induction are highlighted as filled blue circles. Genes whose expression is relatively stable is during EpiLC differentiation are highlighted as filled yellow circles. Prdm14, whose protein levels are relatively stable but whose mRNA levels are dramatically downregulated, is highlighted as an open circle. See also Figure S6.

Figure 5. Distinct waves of global changes in the proteome mark various phases of pluripotency

(A) Temporal profiles of standardized changes in protein levels (compared to 0h). Top and bottom 20% of the proteins that are the most down- or up-regulated (red and blue, respectively), based on the rank-ordering in Figure 4D, are grouped into clusters based on fuzzy c-means clustering (c = 9). Top six clusters, with the most proteins, are shown. Transcriptional and chromatin regulators, known/implicated to play important roles in ESCs and/or EpiLCs, are highlighted. (B) Genome browser shots of Esrrb and Otx2 showing temporal profiles of gene expression dynamics (RNA-Seq) and ChIP-Seq read density profiles for RNAPII and histone modifications H3K4me1, H3K4me3, H3K27ac, H3K27me3, and H3K9me2. Gene annotation is shown at the bottom, with an arrow representing the direction of transcription from the active transcription start site. Regions containing transcriptionally active promoter and known enhancer are highlighted in yellow and green, respectively. (C) Temporal profiles of gene expression, RNAPII, and histone modification dynamics of genes associated with naïve (ESCs) and primed state (EpiLCs). The top and bottom 20% of the genes that are the most down- or up-regulated, based on the rank-ordering in Fig. 4D, were considered as naïve and primed state genes, respectively. Median and standard deviation are shown as line-plot and range, respectively. ChIP-Seq read density within the promoter region was used for analysis (RNAPII, H3K4me3, and H3K27ac: ±1 Kb of TSS; H3K4me1, H3K27me3, and H3K9me2: ±2.5 Kb of TSS).

Figure 6. Cell-surface markers specific to naïve and formative/primed pluripotent states.

(A) Scatter plot showing expression levels of cell surface proteins in naïve ESCs (x-axis) vs. EpiLCs (y-axis). Data for 49 surface proteins that are differentially expressed at one or more profiled time points during the ESC to EpiLC transition are shown. Based on their distance relative to the diagonal (expressed equally in both cell types), cell surface proteins have been categorized as naïve-specific or primed-specific (darker shades of red and blue, respectively). See Figure S7A for expression dynamics during ESC to EpiLC transition. (B, C) Histograms of flow cytometry analysis using fluorophore-conjugated antibodies showing separation in the fluorescence signal between naïve ESCs (red) and EpiLCs (blue). Data for cell state-specific proteins in naïve ESCs (B) and EpiLCs (C) are shown. (D) Flow cytometry contour plots and dot plots of pairwise antibody combinations in ESCs and EpiLCs (first column) and over the ESC to EpiLC time-course (other columns). (E) Relative gene expression of selected cell surface proteins in mouse and human pluripotent cells based on RNA-Seq data from ESC to EpiLC time-course from this study (0h, 1h, 6h, 12h, 24h, 36h, 48h, and 72h), RNA-Seq data from mouse EpiSCs (Factor et al., 2014), and RNA-Seq data from conventional human ESCs (hESCs) and reset “naïve” hESCs (Takashima et al., 2014). To facilitate direct comparison, all datasets were processed similarly and quantile-normalized. Fold changes relative to expression in mouse EpiSCs are shown. See also Figure S7.

Figure 7. Comparative analysis of mouse and human pluripotent states.

(A) PCA of RNA-Seq data from this study (shades of gray; 0h, 1h, 6h, 12h, 24h, 36h, 48h, and 72h) and previously published studies (in color) (Blakeley et al., 2015; Boroviak et al., 2015; Chan et al., 2013; Chen et al., 2018; Factor et al., 2014; Fiorenzano et al., 2016; Marks et al., 2012; Petropoulos et al., 2016; Takashima et al., 2014; Yan et al., 2013). To facilitate direct comparison, all datasets were processed similarly and quantilenormalized. Each data point represents a biological replicate. mESC, mouse ESC; hESC, human ESC. (B) Heatmap showing unsupervised hierarchical clustering of pair-wise Pearson correlations between the RNA-Seq datasets used in (A). (C) Relative expression of genes associated with naïve pluripotency. Fold changes relative to expression in mouse EpiSCs are shown. (D) Same as in (C) but showing genes associated with formative and/or primed pluripotency.