Graphical Abstract

Introduction

Human pluripotent stem cells (hPSCs) represent a unique tool to study early cell fate decisions as they can be grown indefinitely in vitro while maintaining the capacity to differentiate into the three germ layers: endoderm, mesoderm, and neuroectoderm (Thomson et al., 1998). The role of the cell cycle machinery in this process has recently been explored and various studies have established that specification of the germ layers is regulated by cell cycle regulators (Pauklin and Vallier, 2013, Pauklin et al., 2016, Singh et al., 2013, Singh et al., 2015); however, extensive biochemical and molecular analyses of these interplays have been hindered by the difficulty of successfully synchronizing a large quantity of stem cells in the different phases of the cell cycle.
Of particular interest, the fluorescence ubiquitination cell cycle indicator (FUCCI) system (Sakaue-Sawano et al., 2008) can be used in hPSCs for live imaging and for sorting cells in different phases of their cell cycle for transcriptomic analyses (Pauklin et al., 2016, Singh et al., 2013). Nonetheless, the FUCCI system presents several limitations. Sorting large amounts of cells is challenging, as it compromises viability and decreases efficacy of differentiation, thereby precluding precise biochemical analyses. In addition, cells in S and G2/M phases cannot be separated using the FUCCI system, limiting studies investigating mechanisms occurring specifically in these phases of the cell cycle. Finally, the FUCCI system does not distinguish between cells in early G1 or quiescence cells. These limitations highlight the need for the development of alternative tools and complementary approaches to synchronize the cell cycle in hPSCs.

Traditionally, somatic cells have been successfully synchronized using small molecules inhibiting cell cycle progression. Those include G1 phase inhibitors, such as lovastatin and mimosine. Lovastatin is a 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMG-CoA reductase) inhibitor and results in G1 cell cycle arrest by inducing CDKIs, such as p21 and p27 (Hengst et al., 1994, Keyomarsi et al., 1991, Rao et al., 1999). Mimosine is an iron chelator that blocks initiation and elongation of replication forks (Chung et al., 2012, Kalejta and Hamlin, 1997, Krude, 1999, Vacková et al., 2003), resulting in accumulation of cells in the late G1 phase. Inhibitors of G1/S phase transition are also commonly used, such as aphidicolin and thymidine. Thymidine causes inhibition of DNA replication (Thomas and Lingwood, 1975), while aphidicolin blocks DNA polymerase-α, thereby arresting cells at the G1/S phase boundary (Ikegami et al., 1978, Pedrali-Noy et al., 1980). Furthermore, hydroxyurea results in accumulation of cells in the S phase by inhibiting ribonucleotide reductase and dNTP production (Adams and Lindsay, 1967, Brigitte Maurer-Schultze and Bassukas, 1988). Last, G2/M phase inhibitors include colcemid and nocodazole. Both inhibit microtubule polymerization and were shown to arrest somatic and embryonic stem cells in G2/M (Blajeski et al., 2002, Grandy et al., 2015). Importantly, previous studies have used some of these molecules to synchronize hPSCs (Calder et al., 2013, Gonzales et al., 2015, Grandy et al., 2015, Yang et al., 2016); however, these methods were often associated with cell death and accumulation of genomic anomalies while their impact on pluripotency and self-renewal remains to be comprehensively analyzed. In this study, we optimized and characterized the use of these inhibitors to synchronize the cell cycle of hPSCs. We observed that a low dose of nocodazole successfully enriches for hPSCs in G2/M without affecting pluripotency and genetic stability. In addition, nocodazole-treated hPSCs can successfully differentiate into the three germ layers and can generate functional cell types, including cardiomyocytes, smooth muscle cells, chondrocytes, and hepatocytes. Finally, we used this approach to differentiate hPSCs into endoderm while being synchronized for their cell cycle, thereby creating an approach to study mechanisms occurring during cell cycle progression upon differentiation. Accordingly, we performed single-cell RNA-sequencing (RNA-seq) analysis during definitive endoderm formation using hPSCs synchronized by nocodazole treatment, and showed that cell cycle synchronization does not affect gene expression or efficiency of differentiation. Taken together, our results demonstrate that cell cycle synchronization by nocodazole does not affect the fundamental characteristics of hPSCs while providing a valuable tool to study the interplays between cell cycle and differentiation.

Results

In order to identify small molecules that successfully synchronize human embryonic stem cells (hESCs), we tested a panel of inhibitors commonly used with somatic cell types (Figures 1A and 1B). Conventional doses used in somatic cells resulted in cell death within 6 to 20 hr of treatment (data not shown), indicating that the concentrations of cell cycle inhibitors tolerated by stem cells is different from the threshold tolerated by somatic cells. For this reason, we performed extensive tests to identify the optimal conditions that would block cell cycle progression without toxicity. This screen revealed that only doses up to ten times lower than the ones conventionally used were tolerated by hPSCs. At lower doses, G1 and S phase inhibitors did not affect hPSCs colony morphology with the exception of mimosine, which systematically induced cell death (Figure 1C). Concerning the G2/M inhibitors, most hPSCs were arrested in mitosis and acquired a specific round morphology and increased size (Figure 1C). Having solved the toxicity problem, we then aimed to identify the optimal timing of treatment. For that, we incubated hESCs with each inhibitor for 16 or 24 hr and subsequently performed cell cycle profile analysis using EdU incorporation. Most inhibitors enriched hPSCs in specific cell cycle phases and few differences were observed between the two time-points (Figure S1A). Thus, we decided to apply inhibitors for 16 hr in all subsequent experiments. Concerning the G1 phase inhibitors, lovastatin increased by only 9% the fraction of cells in G1 when compared with DMSO-treated cells. Mimosine treatment resulted in a higher enrichment, with 70% of the cells being in G1 phase; however, most of the cells were dead after 16 hr of treatment (Figure 1C). S phase inhibitors gave different results, with thymidine consistently producing a single population of cells without clear cell cycle phase identity (Figure S1A). This observation can be explained by the fact that cells are blocked at the G1/S transition. Aphidicolin sometimes resulted in the same profile as thymidine, whereas in other cases cells were enriched in the S phase (70%) (Figure S1A), suggesting that the synchronization during G1/S transition is not reliable. Hydroxyurea successfully enriched hPSCs in S phase (70%) while nocodazole treatment successfully enriched hESCs in G2/M (>80%). Colcemid treatment proved less efficient with only around 40% of cells being found in G2/M and was thus excluded from further studies (Figure S1A). Based on these encouraging results, we decided to further refine the optimal dose for each inhibitor. Higher dose systematically improved cell cycle synchronization. However, lovastatin and mimosine treatment still failed to generate homogeneous populations of hESCs blocked in G1 (Figure S1B) and thus were excluded from further studies. Concerning S phase inhibitors, synchronization was very efficient (>70%) (Figure S1B); however, release from these inhibitors systematically resulted in a heterogeneous population. Indeed, removal of thymidine and aphidicolin allowed the cells to progress in S phase (Figures 1D, 1E, S1C, and S1D). However, hESCs became asynchronous 12 hr after release, with 50% of the cells in the G2/M phase upon release from thymidine inhibition, whereas in the case of aphidicolin, cell cycle profile was similar to DMSO-treated cells (Figures 1D, 1E, S1C, and S1D). In the case of hydroxyurea, the percentage of cells in the S phase remained constant throughout the time course after release from inhibition, indicating that the cells remain arrested in the S phase (Figures 1F and S1E). Finally, nocodazole treatment resulted in the most efficient synchronization (>90% of cells in G2/M) while the cells remained synchronous after release and moved homogeneously through the cell cycle for 24 hr. More precisely, the cells progressed into G1 2 hours following removal of nocodazole, with 70% of the cells in G1 at 4 hr and 80% of the cells in S phase after 12 hr (Figures 1G, S1F, 2A, and 2B). Importantly, this synchronization lasted for one cell cycle, after which the cells acquired a heterogeneous cycle profile, thereby suggesting that different hESCs could progress through cell cycle at a different speed (Figures 1G, S1F, 2A and 2B). These observations were confirmed by examining the expression of cyclins D1, D2, and D3, which were specifically enriched in late G1. Accordingly, low levels of cyclin D proteins were observed at time zero after removal of nocodazole when cells were in the G2/M, while their levels steadily increased reaching a peak 4 hr after release when most hESCs are in G1 (Figure 2C). These results demonstrate that nocodazole can be applied to generate a near homogeneous population of hESCs synchronized for their cell cycle without altering cell cycle mechanisms such as periodicity of cell cycle regulators.

Figure 1Nocodazole Is the Most Efficient Small-Molecule Inhibitor to Synchronize the Cell Cycle in hPSCs

To further characterize the effect of cell cycle synchronization, we decided to perform single-cell RNA-seq (scRNA-seq) on nocodazole and DMSO-treated cells before and after differentiation into endoderm. Accordingly, hPSC colonies were treated with DMSO or 100 ng/mL nocodazole for 16 hr and induced to differentiate into definitive endoderm for 3 days. Single cells were subsequently collected in either undifferentiated conditions or after 3 days of endoderm differentiation and then sorted onto 384 well plates for Smart-seq2 processing (Figure 4A). Principal component analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE) analysis showed a clear separation between pluripotent and endoderm cells while cell cycle synchronization has no effect on their transcriptional profile with the vast majority of these cells clustering together regardless of their synchronization condition (Figures 4B, S3A, and S3B). Further PCAs show that the main difference between different cell populations (PC1, 40% of variance explained) is their differentiation stage (pluripotent versus endoderm) irrespective of whether they were treated with DMSO or nocodazole (Figures 4B and S3A–S3C). Accordingly, key pluripotency genes, such as POU5F1 (OCT4), NANOG, and SOX2, were only expressed in pluripotent cells, whereas endoderm genes, such as SOX17, GATA6, and CER1, were expressed in endoderm cells regardless of whether they were treated with DMSO or nocodazole (Figure 4C). These results confirm that nocodazole treatment is compatible with endoderm differentiation.

However, it is important to mention that our analyses also revealed that the endoderm cells could be separated based on their synchronization status (DMSO versus nocodazole). Nonetheless, a more thorough investigation of this dataset showed that this separation is only evident by the third principal component (PC3), which explains less than 6% of the variance among the samples (Figures 4B and S3A–S3C). To confirm this observation, we carried out clustering analysis using a shared nearest neighbor (SNN) modularity optimization algorithm (see Experimental Procedures section). This approach identified five individual clusters that for visualization purposes were presented in a t-SNE plot (Figure 4D). To determine the relationship among these clusters, the average expression for genes in each group was calculated and then used to carry out hierarchical clustering. The top 10 markers of each cluster were selected based on their differential expression when compared with other cells and presented in a heatmap (Figure 4E). This approach revealed that cells coming from the pluripotent cohort were classified in three different clusters. The segregation of clusters 0 and 3 seems to be explained only by biological heterogeneity of the pluripotent population (Figure 4E), whereas cluster 4, based on its proximity to the endoderm cohort, seems to represent a fraction of spontaneously differentiated cells that can be observed in conventional cultures of hPSCs (Figure 4E). Interestingly, these clusters include pluripotent cells both from DMSO and nocodazole conditions, confirming that nocodazole treatment does not affect the fundamental characteristics of pluripotent cells. Concerning endoderm cells, the hierarchical clustering suggests that these two groups are highly similar, although our SNN clustering approach did separate the endoderm cohort based on synchronization status (clusters 1 and 2 for DMSO and nocodazole respectively) (Figure 4E). To further confirm this observation, we carried out differential expression analysis for genes in clusters 1 and 2. Accordingly, we found that key endoderm marker genes, such as SOX17, CXCR4, and GATA6, are not among the differentially expressed between these two clusters, indicating that clusters 1 and 2 are similar in terms of differentiation status. However, this approach unveiled 33 genes significantly upregulated in cluster 1 versus cluster 2, and 38 genes significantly upregulated in cluster 2 versus cluster 1 (Figure 4F). The main difference originates from increased expression in genes involved in lipid and cholesterol metabolism (ACAT2, FDFT1, and MVD in cluster 1 and APOE in cluster 2, Figure 4F). Gene ontology (GO) analyses for the different clusters further confirmed that differences observed in DMSO- versus nocodazole-treated cells relate to metabolic processes, whereas processes common to both clusters involve tissue development (Figure 4G), thereby confirming the endodermal identity of these cells. The suggested change in metabolic activity could be explained by the lower density systematically observed in nocodazole-treated cells since they undergo at least one cell cycle less than their control. In addition, the loss in epithelial morphology occurring during synchronization in G2/M could also change the metabolic requirement in cells treated with nocodazole. In summary, our analyses show that nocodazole synchronization has little effect on the differentiation capacity of the cells into endoderm while it does not affect the cellular identity of undifferentiated pluripotent stem cells or their capacity to differentiate into definitive endoderm.