TPX2 Induces Mitotic Survival via BCL2L1 Induction Through YAP1 Protein Stabilization in Human Embryonic Stem Cells


 Genetic alterations have been reported in most human embryonic stem cells (hESCs) for decades. ‘Survival advantage,’ a typical trait acquired during long-term in vitro culture, results from induction of BCL2L1 upon frequent copy number variation (CNV) at locus 20q11.21 and is one of the strongest candidates associated with genetic alteration via escape from mitotic stress. However, the underlying mechanisms for BCL2L1 induction remain undefined. Furthermore, abnormal mitosis and ‘survival advantage’ frequently occurring in the late passage are respectively associated with the expression of TPX2 and BCL2L1, which are located in locus 20q11.21. In this study, we observed that 20q11.21 CNV was not sufficient for BCL2L1 induction and consequent survival traits in pairs of hESCs and human induced pluripotent stem cells (iPSCs) with normal and 20q11.21 CNVs. Inducible expression of TPX2 and basal TPX2 expression due to leakage of the inducible system in hESCs with normal copy number was sufficient to promote BCL2L1 expression and promoted high tolerance to mitotic stress. High Aurora A kinase activity by TPX2 stabilized YAP1 protein to promote YAP1 dependent BCL2L1 expression. Thus, a chemical inhibitor of Aurora A kinase and knockdown of YAP/TAZ significantly abrogated the aforementioned high tolerance to mitotic stress through BCL2L1 suppression. These results suggest that the collective expression of TPX2 and BCL2L1 from CNV at loci 20q11.21 and a consequent increase in YAP1 signaling would promote genome instability during long-term in vitro hESC culture.


Introduction
Due to their pluripotency, human embryonic stem cells (hESCs) have great potential in stem cell-based cell therapy; however, their frequent genetic aberrations during in vitro maintenance are considered a major hurdle that compromises the safety of stem cell-based therapy [1,2]. However, despite these concerns [3], few studies have explored the biological consequences, risks, and biomarkers of hESC genetic alterations [4]. As demonstrated by several massive genomic studies, copy number variations (CNVs) occur most frequently during the sub-chromosomal ampli cation of locus 20q11.21 and trisomy of chromosome 12 or 17 [5] [6, 7].
One of the most characteristic phenotypic changes of genetically aberrant hESCs is 'culture adaptation' [5], which is an acquired survival trait (also referred to as 'survival advantage') under various stress conditions such as culture, genotoxic stress, and single-cell dissociation, among others [8,9]. Such acquired survival trait results from induction of BCL2L1 due to ampli cation of 20q11.21 [10] (encoding BCL-xL), a typical anti-apoptotic gene located at 20q11.21 [8,11] and/or dominant mutation in p53 [12], as the mitochondria are primed to cell death upon genotoxic stress through p53 translocation in hESCs [13][14][15][16]. A recent study revealed that escape from mitotic cell death during mitotic errors due to this acquired survival trait (either induction of BCL2L1, or dominant mutation of NOXA, a pro-apoptotic factor) leads to aneuploidy [9]. In this context, a targeting protein for Xklp2 (TPX2) also located in 20q11.21 was suggested to be the putative driver for abnormal mitosis by disrupting spindle dynamics [17]. However, despite the signi cance of 'survival advantage' in the phenotypic changes of culture-adapted hESCs and even aneuploidy, the molecular mechanisms underlying this phenomenon are not fully understood except for BCL2L1 induction. Moreover, a reduction in serum response factor (SRF) expression [18] or high oxygen concentrations during in vitro culture [19] have been associated with hESC genetic alterations.
To ensure genome integrity, hESCs become highly sensitive to genotoxic stress via mitochondrial priming to apoptosis [16]. However, the contemporaneous expression of anti-apoptotic factors maintains a ne balance between survival and apoptosis [20]. It is also worth noting that the deletion of Yap1 leads to embryonic lethality [21], thus suggesting that Yap1 is required for self-renewal and differentiation of mESCs [22]. A recent study demonstrated that Yap1 in mouse embryonic cell lines (ESCs) serves as a safeguard to attenuate mitochondrial apoptosis by upregulating typical anti-apoptotic factors including BCL2L1 [23]. Similarly, Rho-dependent activation of YAP1 promotes long-term survival of hESCs [24]. In addition to ESCs, YAP1 activation in cancer cells elevates survival through the induction of BIRC5 and BCL2L1, both of which are important for hESC survival [25].
In this study, using isogenic hESC with different passage numbers and CNV status, we demonstrated that a 20q11.21 CNV was insu cient not only for BCL2L1 induction but also to induce 'acquired survival.' TPX2, located at locus 20q11.21, was highly induced in culture-adapted hESCs and conferred resistance under mitotic stress through BCL2L1 induction. Our ndings also revealed that YAP1 protein stabilization by Aurora A kinase activated by TPX2 induction was responsible for BCL2L1 induction, which resulted in mitotic stress escape, suggesting that additional signaling (e.g., YAP1) would be required for determining culture adapted phenotypic changes in hESCs other than CNVs.
Mo Jung-Soon at Ajou University.

Transfection (Plasmid DNA and siRNA)
Cells were prepared about 1 x 106 per 100uL in Opti-MEM. Each cell was separated into cuvettes (EC-002) with plasmid DNA 2µg (or siRNA 100pmole) added per 100uL. Electroporation was done by NEPA21 super Electroporator. After electroporation, cells were seeded to Matrigel coated culture plate with 1 µM of Y27632, and cultured for 24hr~72hr before harvest. After 24 hours, the media was changed to fresh media without Y27632.
RNA extraction and Quantitative real-time PCR Total RNA was extracted using easy-BLUE TM Total RNA Extraction Kit (#17061, iNtRON), and dissolved in RNase-free DEPC-treated water. cDNA was made using extracted RNA and RT master mix (#RR036, TAKARA). Synthesized cDNA, TB green (#RR420, TAKARA), and primers were used as a template for examining the real-time PCR. (Primer list was shown in Table 1.) TAKARA's 2 step protocol was used by Light Cycler 480 Instrument II from Roche.

Immunoblotting
Cell lysates were extracted with RIPA buffer supplemented with a 1% protease inhibitor cocktail and 0.1% sodium orthovanadate. After 1-hour incubation on ice, total protein was extracted by centrifuge. The concentration of total protein was quanti ed by the BCA protein assay kit (#23225, Thermo Scienti cTM). Approximately 25µg of total protein were separated on various (7.5%, 10%, 15%) concentrations of SDS-PAGE. Separated protein in the gel was transferred to the PVDF membrane.
Membrane with protein was blocked with 5% skim milk in TBS-T (Tris-buffered saline with 0.1% Tween-20) for 1 hour and then washed by TBS-T for every 5 minutes three times. The membrane was incubated with the primary antibody in TBS-T (1:1000) with 0.1% sodium azide overnight, 4°C. The incubated membrane was washed for 5 minutes three times with TBS-T. The membrane was incubated with HRPconjugated secondary antibody (Jackson Immunoresearch Laboratories) in TBS-T (1:10000) for 1 hour, room temperature. The incubated membrane was washed for 15 minutes three times with TBS-T. Immunoreactivity was detected by Chemi-Doc using WEST-QueenTM (#16026, iNtRON Biotechnology) kit. The band intensity was measured using Fusion FX software and normalized with the loading control.
Dual-luciferase assay Cells were transfected with a speci c promoter-luciferase vector and pRL vector using the above description. Cell lysates were extracted with 1X passive lysis buffer. After 1-hour incubation on ice, the total lysate was extracted by centrifuge. The supernatant was used for reaction with LARII and Stop & Glo reagent. The reporter assay was performed according to the Dual-Luciferase Reporter Assay System (#E1980, Promega).

Cell Death Assay
Cell death was analyzed by ow-cytometry. Regarding Annexin V/7-AAD staining, cells at 24 h after treatment of each avonoid were washed twice with PBS and stained with FITC conjugated Annexin V antibody (BD Bioscience, Franklin Lakes, NJ, USA, #556419) and 7-AAD (BD Bioscience, #559925) or Propidium iodide(PI) for an additional 45-60 min at room temperature in the dark. Cells stained with Annexin V/7-AAD were analyzed by FACS Calibur (BD Bioscience). Concerning all of the bright eld images captured, a Light channel optical microscope (Olympus, Tokyo, Japan, CKX-41) or JULI-stage (NanoEntek, Seoul, Korea) was used in accordance with the manufacturer's protocol.

Statistical analysis
Graphical data were presented as mean ±S.E.M. Statistical signi cance for more than three groups was determined using one-way or two-way analysis of variance (ANOVA) following a Tukey multiple comparison post-test. Statistical signi cance between the two groups was analyzed using unpaired Student's t-tests. Statistical analysis was performed with GraphPad Prism 8 software (https://www.graphpad.com/scienti c-software/prism/). Signi cance was assumed for p<0.05(*), p<0.01(**), p<0.001(***).

Results
The survival advantage of culture-adapted hESCs under mitotic stress Culture-adapted hESCs become highly resistant to various stressors such as DNA damage [11], single-cell dissociation, and YM155[8], a stemtoxic agent that selectively induces death in undifferentiated pluripotent stem cells [13] through SLC35F2 [26,27]. In addition to the increased survival of hESCs under genotoxic stress, escape from cell death during mitosis has been suggested to increase aneuploidy during long-term culture [9]. Thus, we examined whether P4 hESCs exhibiting the survival trait under YM155 treatment (Fig. 1A) due to BCL-xL induction (encoded by BCL2L1, Fig. 1B) were more resistant to mitotic stress compared to P1 hESCs. P1 or P4 hESCs were treated with typical mitotic drugs targeting mitotic spindles such as nocodazole (Noc), spindle destabilizer, and taxol (Tax), as well as a spindle stabilizer, after which cell death was evaluated. Similar to YM155, P4 hESCs were highly resistant to mitotic stress inducers (Figs. 1C and D).
20q11.21 CNV was not su cient to induce the survival trait in hESCs Except for the role of BCL2L1 induction followed by CNV at locus 20q11.21 [10] or p53 dominant mutations [12], little is known regarding the mechanisms by which the survival trait is acquired during hESC culture adaptation. Further, no previous studies have determined how BCL2L1 transcription becomes upregulated, thus resulting in the survival advantage trait in culture-adapted hESCs. We observed that the survival trait became manifested in P3 (over 200 passages) and P4 (over 300 passages), but not P2 (over 100 passages) (Fig. S1A)  Similarly, hFmiPS1 (passage 20, normal iPSCs: hFmiPSC1-D1) and mutant cells (passage 30 with 20q11.21 CNV: hFmiPS2-DCB1) [28] underwent similar cell death by YM155 treatment (Fig. S1B) with comparable BCL2L1 expression [28]. These results imply that the acquisition of the survival trait via BCL2L1 induction may require additional events other than 20q11.21 CNV.

BCL2L1 induced survival associated with TPX2
As previously demonstrated, ampli cation of 20q11.21 was the most frequently occurring genome aberration and CNV of ID1, BCL21L, and HM13 at locus 2011q.21, suggesting that this was a marker of genome aberration [5]. However, intriguingly, not all genes at locus 20q11.21 were transcriptionally active in hESCs having 20q11.21 CNV (e.g., P2, P3, and P4 hESCs) ( Fig. 2A). Among the genes in the 20q11.21 locus, TPX2, which was previously shown to induce aberrant mitosis in culture-adapted hESCs [17], was concurrently induced with BCL2L1 expression. Similarly, BCL-xL protein level was closely associated with TPX2 protein level in P3 and P4 hESCs where the survival trait was evident (Fig. 2C). Concretely, P4 hESCs with high TPX2 expression also expressed higher levels of BIRC5 (encoding Survivin), an antiapoptotic factor highly expressed in undifferentiated hESCs [13,29] along with BCL2L1 (Fig. 2D). Considering the roles of TPX2 in cancer malignancy [30] and survival/chemoresistance [31], it is readily presumed that TPX2 expression may be associated with BCL2L1 induction and consequent survival traits. As predicted, depletion of TPX2 in P4 hESCs signi cantly attenuated BCL2L1 expression (Fig. 2E), suggesting that TPX2 induction somehow regulates BCL2L1 expression and confers the survival trait. The doxycycline (Dox) inducible TPX2 cell line established from P1 hESCs (iTPX2-hESCs) was used to explore this possibility (Fig. S2A). TPX2 mRNA (Fig. S2A) and protein (Fig. S2B) induction occurred in a dose-dependent manner. The signal from green uorescence protein (GFP) conjugated with TPX2 was evident in the mitotic spindle where TPX2 is located during mitosis [32] (Fig. S2C). Using the iTPX2-hESCs, we examined whether BCL2L1 transcription was promoted by TPX2 induction with Dox. Surprisingly, TPX2 induction was su cient to increase BCL2L1 mRNA (Fig. 2F) in a dose-dependent manner (Fig. 2G), as well as protein (Fig. 2H). It is important to note that iTPX2-hESCs were established from P1 hESCs with normal copy numbers and therefore the copy number of iTPX2-hESCs remained normal regardless of Dox treatment unlike P4 hESCs (Fig. 2I).

TPX2 induction rescues mitotic cell death of normal hESCs
We have previously demonstrated that TPX2, located at locus 20q11.21 along with BCL2L1, would be a putative driver for abnormal mitosis [17]. Using iTPX2-hESCs, cell death under mitotic stress (e.g., Noc) was determined after TPX2 induction by doxycycline (Dox). We noted that only a portion of cells expressed GFP-TPX2 even after Dox treatment for unknown reasons. Thus, a GFP negative population was used as an internal control and the cell death of the GFP positive population (expressing TPX2) of iTPX2-hESCs was monitored after mitotic stress (Fig. 3A). As predicted, GFP positive cells (expressing TPX2) were more resistant to mitotic stress induced by Noc (Fig. 3A) and Tax (Fig. S3A) than GFP negative cells. Similar results were obtained in a dose-dependent manner (Fig. 3B). The resistance to the mitotic stress by TPX2 induction was reproduced in other iTPX2-hESCs derived from hCHA3 (Fig. S3B). To rule out the potential pro-survival effect of GFP expression (rather than TPX2), P1 expressing enhanced green uorescent protein (EGFP-P1) was co-cultured with P4 or iTPX2 hESCs (Fig. 3C), as basal TPX2 level in P4 and iTPX2 (due to leakage of Tet-O system [33]) was comparably higher than that of P1 hESCs (Fig. S3C). Consistently, the hESCs with high TPX2 expression (e.g., P4 and iTPX2) were more resistant to Noc-induced mitotic stress regardless of GFP expression (Figs. 3D and E). Higher TPX2 in P4 hESCs than P1 hESCs occurred constantly regardless of cell cycle phase [17]. To further assess whether high TPX2 expression was responsible for the resistance to mitotic stress, TPX2 was transiently depleted by siRNA#3 (Fig. S3D). Depletion of TPX2 in P4 hESCs at a similar level of TPX2 in P1 hESCs (Fig. S3D), sensitized the cells to Noc induced mitotic stress (Fig. 3F). It is also worth noting that stable knockdown or expression of TPX2 in hESCs was unsuccessful even after multiple attempts, suggesting that the basal level of TPX2 is critical for self-renewal of hESCs.
Aurora-A stabilizes the YAP1 protein Previously, Aurora A, a mitotic kinase strongly associated with TPX2 for spindle assembly, was shown to phosphorylate and stabilize YAP1 in cancer [9,36]. Constant activation of Aurora-A, which corresponded with TPX2 induction regardless of cell cycle phase, was a distinct cellular phenotype of P3 and P4 hESCs [17]. P4 hESCs with high Aurora-A (determined by phospho-Aurora A) and TPX2 activities exhibited high protein levels of BCL-xL and YAP1 (Fig. 5A). As transient depletion of TPX2 also signi cantly lowered the protein level of BCL-xL (encoded by BCL2L1) and YAP1, high TPX2 induction in P4 hESCs would be closely associated to YAP dependent BCL2-xL expression (Fig. 5B). Next, to con rm the increased protein level of YAP1 results from protein stability, YAP1 protein was monitored after cycloheximide (CHX) treatment to inhibit protein translation in P1 and P4 hESCs (Fig. 5C). Similarly, the increased protein level of YAP1 was evident in Dox-dependent TPX2 induction (Fig. 5D), whereas the YAP1 mRNA level remained unaffected (Fig. 5E). Aurora-A active phosphorylation was clearly induced by TPX2 induction in iTPX2-hESCs, which was concurrent with the increase of YAP1 protein expression (Fig. 5F). Similar to P4 hESCs, YAP1 protein and not TEAD4 (Fig. S4B) was highly stabilized after Dox treatment in iTPX2-hESCs (Fig.  5G). These data suggest that high YAP1 protein levels in culture-adapted hESCs result from high Aurora-A activity by TPX2 induction.

Inhibition of Aurora-A abrogated the resistance to mitotic stress by YAP1 destabilization
Given that the activity of Aurora-A in hESCs with high TPX2 expression appeared to stabilize YAP1, we next tested whether chemical inhibition of Aurora-A may destabilize YAP1 and sensitize hESCs with high TPX2 expression to mitotic stress. To this end, we rst determined the concentration of Aurora-A inhibitor (MLN8237: MLN) for inhibition of Aurora-A in hESCs with high TPX2 expression (P4 and iTPX2-hESCs with Dox). Intriguingly, Aurora-A activity determed by its active phosphorylation was signi cantly reduced by 50nM of MLN treatment in P1 hESCs (Fig. S5A), where it remained active up to 100nM in P4 hESCs (Fig. S5B). Clear attenuation of active phosphorylation in P4 hESCs was distinct from 0.5mM of MLN treatment (Fig. 6A). Therefore, all downstream experiments for survival test in P4 were performed using 0.5mM of MLN. In pararllel with Aurora-A inhibition by MLN (Fig. 6A), BCL2L1 transcription was reduced along with CTGF and SERPINE1 [34,37] (Fig. 6B). Mitotic resistance in P4 hESCs, as determined in the sub G1 and 3N populations, was also signi cantly attenuated by additional MLN treatment with Noc (Fig.  6C). The same result was reproduced by ow cytometry to determine the live cell population (Fig. 6D).

Discussion
Despite the widespread concerns regarding the frequent genetic aberration and consequently acquired survival trait of hESCs [3], the molecular mechanisms underlying such drastic phenotypic changes have not been fully determined, except for the known role of BCL2L1 induction followed by 20q11.21 CNV [8, 10,11]. Here, we demonstrated that BCL2L1 transcription was induced by stabilized YAP1 due to high Aurora-A activity via TPX2 induction at locus 20q11.21. Further, BCL2L1 induction (Fig. 2) followed by YAP1 protein stabilization (Fig. 5) was readily achieved by inducible TPX2 induction without 20q11.21 CNV (Fig. 2I), resulting in survival under mitotic stress (Fig. 3). Therefore, 20q11.21 CNV itself would not be su cient to lead to the survival trait. This result is consistent with a recent study that human iPSCs with 20q11.21 CNV only show the lesser commitment of the ectodermal lineage in teratoma without expression change in 20q11.21 loci [28]. Similarly, no drastic phenotypic changes (e.g., BCL2L1 expression and survival trait) were observed in P2 hESCs with CNV at 20q11.21 (passage number of approximately 100), unlike P3 hESCs (passage number of approximately 200), which exhibited a clear BCL2L1 expression and survival trait[8] along with high TPX2 expression and abnormal mitosis [17]. Therefore, we propose that other cues to activate BCL2L1 promoter activity such as YAP1 activation would be required to confer the cells with the survival advantage trait (e.g., YAP1 stabilization by Aurora-A activity due to TPX2 induction). A recent study demonstrated that the Hedgehog signal is responsible for TPX2 induction through the FOXM1 transcription factor in cancer cells [30]. Considering the complexity of the YAP1 activation mechanism, we could not rule out the role of other events (e.g., Factin stabilization [24]) in YAP1 activation for culture adaptation of hESCs other than the TPX2-Aurora-A axis. TPX2, of which basal expression was higher even in normal hESCs than differentiated cells [17], was critical for self-renewal of hESCs as stable depletion of TPX2 failed to be achieved multiple times (data not shown). Thus, only transient depletion of TPX2 with siRNA was performed. These results further support that TPX2 knockout leads to failure of early embryogenesis even prior to blastocyst formation [38]. As TPX2 induction stabilizes the mitotic spindle and leads to abnormal mitosis [17], TPX2 induction occurring in multiple culture-adapted cell lines [17] would serve as a potential driver not only for 'acquired survival' but also further 'genetic alteration' by saving hESCs from cell death after abnormal mitosis.
Unlike other somatic cells, hESCs are highly sensitive to genotoxic stress, which is believed to be a major safeguard system to ensure genome integrity [39]. This extremely high susceptibility to DNA damage is mostly triggered by high mitochondrial priming to apoptosis [16] through p53 mitochondrial translocation [13]. Thus, p53 dominant-negative mutation [12], NOXA mutation [9], and induction of BCL2L1 [11] abrogate the unique genome safeguard mechanism of high mitochondrial priming, which occurs in culture-adapted hESCs. Therefore, the incidence of random mutations including CNV at 20q11.21 (where TPX2 and BCL2L1 are located) during prolonged culture to favor survival through TPX2 induction to activate YAP1 dependent BCL2L1 induction in hESCs with 20q11.21 CNV would be su cient for dominant selection during multiple bottleneck events [40]. Thus, continuous loss of high mitochondrial priming and building resistance to mitotic stress may collectively enhance chromosome instability and even induce aneuploidy during long-term hESC culture.

Conclusion
CNV at locus 20q11.21 was not su cient to lead to culture adaptation. Further, TPX2 induction and subsequent activation of Aurora-A stabilizes YAP1 to promote BCL2L1 transcription from an ampli ed 20q11.21 locus, thus promoting the survival trait in hESCs.