Human embryonic and induced pluripotent stem cells maintain phenotype but alter their metabolism after exposure to ROCK inhibitor

Human pluripotent stem cells (hPSCs) are adhesion-dependent cells that require cultivation in colonies to maintain growth and pluripotency. Robust differentiation protocols necessitate single cell cultures that are achieved by use of ROCK (Rho kinase) inhibitors. ROCK inhibition enables maintenance of stem cell phenotype; its effects on metabolism are unknown. hPSCs were exposed to 10 μM ROCK inhibitor for varying exposure times. Pluripotency (TRA-1-81, SSEA3, OCT4, NANOG, SOX2) remained unaffected, until after prolonged exposure (96 hrs). Gas chromatography–mass spectrometry metabolomics analysis identified differences between ROCK-treated and untreated cells as early as 12 hrs. Exposure for 48 hours resulted in reduction in glycolysis, glutaminolysis, the citric acid (TCA) cycle as well as the amino acids pools, suggesting the adaptation of the cells to the new culture conditions, which was also reflected by the expression of the metabolic regulators, mTORC1 and tp53 and correlated with cellular proliferation status. While gene expression and protein levels did not reveal any changes in the physiology of the cells, metabolomics revealed the fluctuating state of the metabolism. The above highlight the usefulness of metabolomics in providing accurate and sensitive information on cellular physiological status, which could lead to the development of robust and optimal stem cell bioprocesses.

Single cell cultures in suspension of hPSCs undergo apoptosis despite the use of culture conditions conducive to stem cell maintenance 9,10 . This problem has been addressed with the inhibition of ROCK (Rho-associated protein kinase), a serine-threonine kinase that phosphorylates and activates the myosin II pathway [11][12][13][14][15][16] resulting in the maintenance of the differentiation potential for up to 72 hours, which is considered to be primarily mediated via the inhibition of an E-cadherin-dependent apoptotic pathway [17][18][19][20] . The most effective ROCK inhibitor is Y-27632 21 . hPSCs colonies, both hESCs and hiPSCs, are most commonly treated with 10 μ M of Y-27632 prior to dissociation of the colonies into single cells [22][23][24] . With this protocol, single cell survival is maintained for up to 3 weeks and stem cell phenotype along with differentiation capability into all lineages are sustained 11 . Until now, it has been assumed that ROCK inhibition does not affect physiology since the hPSCs retain their stemness and survive 25 .
Herein, we have undertaken an in-depth assessment of the effect of ROCK inhibitor (Y-27632) on the dynamic metabolism of hPSCs over a 96-hour culture period 5,26,27 . Whereas no differences were observed in the pluripotency and viability of hESCs and hiPSCs on gene and protein expression levels, differences in metabolism were detected. Specifically, metabolomics analysis was able to detect changes on the metabolic physiology of the cells as early as 12 h of ROCK inhibitor treatment. Existing metabolic switches on hESCs and hiPSCs were similar on both cell types up to 48 h of exposure to Y-27632, with 24 h exposure being identified to be critical as a turning point in metabolism. Correlating with these metabolic changes, a differential expression of the metabolic regulators, p53 28,29 and mTORC1 4,30 , was observed. The above indicate a dynamic process of adaptation of the cells to the altered environment, which is mostly projected to the metabolic level.

Results
hPSCs maintain pluripotency phenotype up to 48 h of exposure to ROCK inhibitor. The effects of Y-27632 exposure on both hESCs and hiPSCs were evaluated by assessing gene and protein expression of pluripotency markers. Specifically, TRA-1-81 and SSEA3 expression indicated that the cells retained these pluripotency markers up to 48 h of exposure, whereas there was a reduction at 96 h in both cell types (p < 0.05; Fig. 1b). Intracellular expression of Nanog, Oct4 and Sox2 indicated maintenance of stemness at high levels (over 67% for hESCs and 80% for hiPSCs), which is reduced at 96 h exposure for hESCs (p < 0.05; Fig. 1b). Nanog, Oct4 and Sox2 expression did not differ at any time-point evaluated (data not shown). Consequently, gene and protein expression showed that the pluripotency phenotype remained constant for at least 48 hours after ROCK exposure and was similar to the untreated control. In all cases, cell viability was high (data not shown). Immunostaining for Oct4 and Sox2 confirmed the maintenance of pluripotency, apart from the 96 h time-point in hESCs when the staining appeared to be weaker (Supplementary Figure 1), in agreement with the flow cytometry results (Fig. 1b).

Metabolomics analysis reveals changes in metabolism.
Multivariate statistical analysis revealed that 0 h-untreated cells (grown in colonies and not exposed to Y-27632) had discrete metabolism compared to those exposed to Y-27632 as early as 12 hrs, as highlighted by Hierarchical Clustering (HCL) analyses and heatmaps (Fig. 2b). Grouping of the metabolic profiles from cells exposed 48 h and 96 h to ROCK inhibitor demonstrated that prolonged exposure resulted in adapted cellular physiology not as discrete as that of cells exposed for 12 h and 24 h, as elucidated by the grouping in Principal Components Analysis (PCA) graphs (Fig. 2a). Alas, the distinct grouping of the hESCs and hiPSCs at 12 h exposure indicates a different physiological response of the cells (Fig. 2a).
To further understand how cellular physiology changed with culture time, an in-depth assessment of the metabolic transitions was undertaken: consecutively from 0 h to 12 h, 12 h to 24 h, 24 h to 48 h and 48 h to 96 h by Significance Analysis of Microarrays (SAM) (Fig. 3). At 12 hours of Y-27632 exposure, the metabolic profiles were similar for both hESCs and hiPSCs; glucose concentration increased whereas lactate and alanine production decreased, indicating down-regulation of the glycolytic route. TCA cycle and glutaminolysis were also down-regulated. Some amino acids (glycine, proline, and GABA for both ESCs and iPSCs, and threonine, serine, phenylalanine, tyrosine for hESCs) and urea (for both cell types) were reduced. From 12 h to 24 h, the most important difference was the reduction of metabolites in the serine-glycine-threonine pathway. In addition, aspartate was reduced in both hiPSCs and hESCs. The further reduction at 24 h in the metabolite pools is more intense in hiPSCs. After 48 h of exposure time, metabolism increased as indicated by the activation of glycolysis, glutaminolysis, TCA cycle and amino acid pools (serine, glycine, threonine for both types of cells, ornithine, phenylalanine for hESCs, and aspartate, leucine, valine for hiPSCs). Overall, the metabolic behaviour of hESCs and hiPSCs was similar up to 48 h. In contrast, at 96 h only hESCs increased glycolytic rate and up-regulated the TCA cycle with increased aspartate whereas hiPSCs down-regulated glutaminolysis and aspartate production. Hence, not only was the metabolism of ROCK-exposed hPSCs different from that of the day 0 control, but cells sequentially down-regulated metabolism at 12 and 24 hours, followed by an upregulated metabolic profile at 48 hours and had completely disparate physiology by 96 hours of exposure.

Caspase-3 expression did not differ up to 48 hours of ROCK exposure. ROCK inhibition blocks
caspase-3 apoptotic signalling 31,32 . qRT-PCR analysis confirmed that there was no change in caspase-3 expression at any time-point in the hiPSCs cultures (data not shown). In contrast, there was a significant decrease in caspase-3 expression in hESCs from 48 h to 96 h of ROCK inhibition (Fig. 4a). This alteration of caspase-3 from 48 h to 96 h may reflect on the increased activity of metabolic pathways such as glycolysis, TCA cycle, glycerolipids and phospholipids synthesis in hESCs, which is required for apoptosis 33 , as shown in Fig. 4b, after significant analysis and comparison of the metabolic profiles of the two time-points. Regardless, changes in metabolism observed throughout in the cultures of both hPSC types cannot be explained by changes in caspase-3 expression, which remained stable for at least 48 hours and similar to that of 0 h control unexposed cells. In all cases, the viability of the cells remained high (data not shown).
Scientific RepoRts | 7:42138 | DOI: 10.1038/srep42138 mTORC1 expression correlated with metabolic changes, independent of tp53. Expression of two major metabolic controllers, mTORC1 and tp53, was assessed. In hESC cultures inhibited by ROCK (Fig. 5), no difference in tp53 expression throughout the culture time was observed. In contrast, mTORC1 expression was lower at 24 h and 96 h exposure compared to the control 0 h unexposed cells (p < 0.05; Fig. 5a), correlating with decreased metabolic activity at the same time points (Fig. 5b), as identified by the significant analysis of the metabolic profiles between these time-points. Specifically, at 24 h glycolysis and glutaminolysis, lipid synthesis, most of the amino acid pools and TCA cycle were less active, whereas at 96 h glycolysis and glutaminolysis were reduced compared to the control 0 h unexposed group. In hiPSC cultures inhibited by ROCK (Fig. 6), expression of tp53 was higher only at 96 h exposure compared with that at 24 h exposure, coincident with the increased glycolysis and glutaminolysis, TCA cycle and lipid synthesis at 96 h exposure ( Fig. 6a and b). In contrast, mTORC1 expression was lower following 24 h exposure and then increased again at 96 h exposure (p < 0.05), correlating with similar changes in metabolism. Therefore, in both hESCs and hiPSCs, decreased expression of mTORC1 correlated with reduced metabolism after 24 h of exposure with Y-27632, independent of tp53 and caspase3 expression levels.

ROS/RNS levels and catalase expression suggest a correlation with proliferative metabolism.
The measurement of reactive oxygen and nitrogen species (ROS/RNS) indicated a reduction in ROCK Figure 1. Assessment of pluripotency phenotype by conventional markers following exposure to ROCK inhibitor. (a) Experimental design. hPSCs were pre-treated with 10 μ Μ Y-27236 ROCK inhibitor for 1-2 hours. Then, colonies were dissociated into single cells and exposed to 10 μ Μ Y-27236 ROCK inhibitor. Samples were collected at five time-points: 0 h were untreated with cells remained in colonies whereas single cells were sampled at all other time-points (12 h, 24 h, 48 h and 96 h). (b) hPSC cell phenotype by flow cytometry and qRT-PCR was maintained for at least 48 hours following exposure to ROCK inhibitor, with reduction in stemness phenotype noted by 96 hours. *p < 0.05.
inhibitor-treated hESCs throughout the culture period. In contrast, following an initial reduction of ROS/RNS at 12 h and 48 h, an increase in ROS/RNS was observed at 96 h in hiPSCs (Supplementary Figure 2). Catalase expression was also significantly increased at 96 h in hiPSCs cultures compared to 12 h, 24 h and 48 h. No significant change has been observed in glutamylcysteine synthetase (GCS) expression in both hESC and hiPSC cultures whereas a significant decrease of glutathione peroxidase 1 (GPX-1) expression was observed only at 24 h in hESCs (Supplementary Figure 3).

Discussion
The use of the ROCK inhibitor on hPSCs has been shown not to affect their stemness 18,25 . In contrast, its effect on metabolism has not been investigated. Herein, we confirm that treatment of hiPSCs and hESCs with ROCK inhibitor does not affect the expression of pluripotency markers at gene and protein level, whereas their metabolism is altered and becomes distinctly different. The observed metabolic changes appeared to be irreversible since the cells did not revert back to the metabolic signature of untreated cells. Furthermore, the metabolic transitions, in both cell types, were similar up to 48 h of ROCK inhibitor exposure and correlated with expression levels of mTORC1 at 24 h of ROCK inhibition, but were independent of tp53 or caspase-3 expression. Finally, the detected metabolic differences between hESCs and hiPSCs indicate that although the two hPSC types share many phenotypic similarities, they are not physiologically identical suggesting that their metabolic features should be considered in designing hPSC bioprocesses to deliver robust maintenance and differentiation protocols.
The effects on metabolism could be explained by the single cell culture conditions that result in loss of cell-to-cell contact, changes to nutrient availability in single cell cultures compared to that for cells grown within colonies, and the time-period required for cells to adapt to an environment conducive to single cell culture. hPSCs are highly proliferative cells and share metabolic characteristics with cancer cells 5,26,27 . In tumors, the glycolytic rate is reduced when E-cadherin is unbound 34 similar to what we observed when the 0 h-untreated cells cultured in colonies were compared with treated cells at other time-points. After 12 h and 24 h of exposure to ROCK inhibitor, the metabolism of both hESCs and hiPSCs was down-regulated, less glycolytic and, therefore, less proliferative. According to literature on mammalian signalling, an indirect connection between ROCK and mTORC1 exists. ROCK inhibition has been shown to dramatically suppress phosphorylation and activity of FAK (Focal Adhesion Kinase) 35,36 . FAK activation has also been correlated with increased proliferation via regulation of the Akt/mTORC1 pathway 37 . Recently, this relationship was demonstrated at the metabolic level, where FAK activation increased glycolytic metabolism (aerobic glycolysis), which substituted mitochondrial respiration 38 . Consequently, ROCK inhibition could lead to down regulated FAK activity followed by suppressed mTORC1 signaling and a decreased proliferative (glycolytic) metabolism, which is in agreement with our results, as early as 12 h.
The serine-glycine-threonine pathway is directly connected to one-carbon metabolism, which is the folate and methionine cycle, as independent modules. One-carbon metabolism cycles carbon units, amino acids, to maintain redox balance and promote biosynthesis. Folate cycle activation is a sign of nucleic acid biosynthesis 39 , among others. Studies in cancer metabolism have suggested that pathway activation is correlated with proliferation of cells 40,41 . Serine biosynthesis and glycine metabolism promote tumorigenesis 41 . Our results show that the serine-glycine-threonine pathway is down-regulated following 12 h and 24 h exposure and up-regulated after 48 h exposure for both hESCs and hiPSCs (Fig. 3). Overall, it is clear that the cells lose their highly proliferative physiology after the exposure to ROCK inhibitor and the dissociation into single cells as indicated by their metabolic profiles and the metabolic pathways affected. Their metabolic physiology never returned to the initial condition.
Metabolic transitions following 48 h exposure to ROCK inhibitor were similar for both hESCs and hiPSCs. In contrast, 96 h exposure resulted in divergent metabolism for the two cell types. In hESCs metabolism increased; in contrast, hiPSCs had decreased metabolic activity. hESCs and hiPSCs share similar but not identical metabolism 26 . This fact could explain why the same manipulation (i.e. dissociation of colonies and exposure to Y-27632) could cause different effect(s), especially when the treatment is prolonged, as was observed following 96 h exposure ROCK inhibitor. Both hESCs and hiPSCs reduced their metabolic activity trying to adapt to the new environment, something which is clear from the12 h and 24 h metabolic profiles. After 48 h exposure, the cells appear to finally adapt to the new environment and increase their metabolic activity again, assuming a more proliferative metabolism. The two cell types follow discrete directions afterwards. Moreover, even if gene and protein expression of pluripotency markers remained high following 96 h exposure to ROCK inhibitor, a small but significant decrease in the percentage of pluripotent cells was observed. This could explain the divergent effects on the metabolism between the two types of cells after 96 h of exposure, where hESCs increase their metabolite pools further while hiPSCs show a decrease in their metabolic activity when compared to 48 h exposure. It becomes clear that hESCs respond differentially to the prolonged exposure to ROCK inhibitor and single cultivation than hiPSCs.
Both hESCs and hiPSCs encounter a critical point in culture at 24 h of exposure, as indicated by the metabolic change commensurate with reduction in mTORC1. It has been shown that metabolism plays a significant role in defining human pluripotent stem cell fate, mostly by affecting the epigenome, in an irreversible way following a critical point 5,42,43 . Increased mTORC1 expression favors growth metabolism, i.e. increased glycolysis and glutaminolysis, whereas the opposite effect is expected when mTORC1 is not expressed. Interestingly, tp53 expression is also increased at 96 h compared to 24 h in hiPSCs -this may represent an attempt of cells to control proliferative activity. High mTORC1 expression 44 induces tp53, as the two proteins counter-balance each other 45,46 , with mTORC1 favoring growth when environmental conditions (nutrients, O 2 , etc.) are appropriate 47 , while tp53 suppresses growth and proliferation when necessary 28,29,48 . ROCK inhibition blocks the pathway of myosin II to caspase-induced apoptosis 49 . In hESCs, caspase-3 expression decreased after 96 h exposure but not at 48 h exposure, coincident with activation of the glycolysis pathway at 96 h with respect to production of glycerolipids and phospholipids, consistent with previous work that showed that caspase-driven apoptosis is metabolically demanding 33 .
ROS/RNS and catalase expression analyses indicate a positive correlation between proliferative metabolism and oxidative stress. The decrease in ROS/RNS levels in hESCs is followed by a decrease in mTORC1 expression, while the increase in ROS/RNS levels and catalase expression at 96 h at hiPSCs cultures correlates with a more proliferative metabolism and increased mTORC1 expression. Our results are in agreement with the literature. Specifically, a positive relationship between pluripotency and hPSC maintenance with proliferative metabolism and reduction of ROS 50 has already been established. Similarly, ROCK inhibition has been linked to down-regulation of oxidative stress 51 .
Metabolism represents the cellular function that is most sensitive to even small environmental disturbances; GC-MS metabolomics has already been proven sensitive enough to detect changes in the physiology of cell cultures compared to gene and protein expression 52 , something of great value in the fields of stem cell biology and cell culture engineering in terms of monitoring and bioprocess optimization. Ultimately, metabolomics could be used as a release assay for verifying both product (cells) and bioprocess quality.
In conclusion, exposure to ROCK inhibitor altered cellular metabolism whereas gene and protein expression of pluripotency markers remained unaffected; as early as 12 h exposure to ROCK inhibitor resulted in the metabolism of both hESCs and hiPSCs being different than that of unexposed cells. Generally, hPSC metabolism decreased for the first 24 h of exposure and then increased again at 48 h exposure for both hESCs and hiPSCs, with completely disparate metabolic pathways followed following 96 h exposure. The observed metabolic changes correlated well with similar changes in mTORC1 expression, a principal metabolic regulator, yet were independent of tp53 and caspase-3 expression levels. Our findings indicate that metabolomics is essential in deciphering changes of physiology and is an invaluable tool in the field of stem cell bioprocessing as it can be used to identify optimal physiological transitions and result in robust cultivation processes.  Metabolite profiling. Cells were washed twice with PBS before cold methanol (− 40 °C) was added to culture plates (6-well -Corning, NY, US). At least 10 6 cells were extracted from each replicate-sample of hPSCs using ribitol (1 mg/1 × 10 6 cells) as an internal standard, as previously described 52 . The process of methoximation (50 μ L of 20 mg methoxyamine hydrochloride/mL pyridine) and MSTFA (100 μ L of N-methyl-trimethylsilyl-trifluoroacetamide) derivatization turned the dried polar extracts into their (MeOx) TMS-derivatives 58,59 . Metabolic profiles were obtained using QP2010 Ultra GC-MS and GCMSsolution software Ver.4.11 (Shimadzu Corp., Kyoto, Japan). The raw metabolomics dataset was comprised of 84 peaks of known chemical category metabolites. From each independent sample, three experimental replicates (n = 3) were created. The relative peak areas of all detected peaks (RPAs) were estimated from their normalization with the 103 marker ion peak area of the internal standard ribitol. Data normalization and filtering are applied before incorporation into the final dataset table 58,59 . Cumulative (effective) peak areas were calculated using weight coefficients of the hESCs 0 h-untreated cells. All analyses applied on the acquired metabolic profiles were based on standardized values of the metabolite relative peak areas 52 . Statistical analyses. One-way ANOVA with Bonferroni post hoc test was used for flow cytometry and qRT-PCR data comparison (N = 3, n = 3). The level of statistical significance was set by p < 0.05. The p-values where differences were significant are clearly noted on the figures. Unsupervised HCL and PCA algorithms were used as visualization methods of the differences among samples, based on their metabolic profiles (N = 3, n = 3). The Euclidean distance metric was used in HCL. SAM was used to identify the metabolites, whose concentration was significantly higher or lower in a set of metabolic profiles compared to another, with false discovery rate (FDR) at zero 60 . TM4 MeV v4.9.1 (Dana-Farber Cancer Institute, MA) was used for the multivariate 61 and Origin2016 (OriginLab Corporation, MA) for univariate statistical analyses.
Immunocytochemistry. Sox2 and Oct4 protein expression was assessed using immunofluorescence. Cells were washed with PBS and fixed with 4% (w/v) paraformaldehyde solution in PBS (Sigma-Aldrich, UK) and subsequently blocked with a PBS solution of 10% (v/v) normal donkey serum, 0.1% (w/v) Triton X-100 and 1% (w/v) BSA (Sigma-Aldrich, UK) for 45 min at room temperature in order to avoid non-specific antibody binding. After blocking, the cells were incubated for 3 h with conjugated antibodies at 1:10 dilution (NL557 PE-conjugated Goat Anti-Human SOX2, NL637 Alexa Fluor 647-conjugated Goat Anti-Human Oct-4) (R&D Systems Inc, MN) at room temperature and counterstained with DAPI for 5 min. Images were taken with a BX-51 Olympus microscope (Olympus, UK).

ROS/RNS assay.
Reactive oxygen species (ROS) and reactive nitrogen species (RNS) levels were evaluted with the use of quenched fluorogenic probe dichlorodihydrofluorescin (DCFH-DiOxyQ; OxiSelect, Cell Biolabs, CA). Cells were resuspendend at a cell density of 1 × 10 7 cells/mL in PBS and lysed with 1% (w/v) Triton X-100 solution and spinned at 10000 g for 5 min to remove insoluble particles. They were assayed according to the manufacturer's instructions.