TPX2 Amplification-Driven Aberrant Mitosis in Culture Adapted Human Embryonic Stem Cells with gain of 20q11.21

Despite highly effective machinery for the maintenance of genome integrity in human embryonic stem cells (hESCs), the frequency of genetic aberrations during in-vitro culture has been a serious issue for future clinical applications. By passaging hESCs over a broad range of timepoints (up to 6 years), the isogenic hESC lines with different passage numbers with distinct cellular characteristics, were established. We found that mitotic aberrations, such as the delay of mitosis, multipolar centrosomes, and chromosome mis-segregation, were increased in parallel with polyploidy compared to early-passaged hESCs (EP-hESCs) with normal copy number. Through high-resolution genome-wide approaches and transcriptome analysis, we found that culture adapted-hESCs with a minimal amplicon in chromosome 20q11.21 highly expressed TPX2, a key protein for governing spindle assembly and cancer malignancy. Consistent with these findings, the inducible expression of TPX2 in EP-hESCs reproduced aberrant mitotic events, such as the delay of mitotic progression, spindle stabilization, misaligned chromosomes, and polyploidy. These studies suggest that the increased transcription of TPX2 in culture adapted hESCs could contribute to an increase in aberrant mitosis due to altered spindle dynamics.


Introduction
While the machinery of genome integrity maintenance is well-developed in human embryonic stem cells (hESCs) [1], genetic abnormalities ranging from full chromosome aneuploidy to single point mutations are relatively common in human pluripotent stem cells (hPSCs) when they are propagated in in vitro culture [2,3]. Hence, genetic aberration in hPSCs during in vitro maintenance has remained a serious hurdle for future clinical applications [4] because of the uncertainty of the biological consequences of genetic aberrations in hPSCs [5]. Indeed, unexpected genetic mutations identified in induced pluripotent stem cells (iPSCs) halted the second human clinical trial of iPSC-based cell therapy [6].
Through massive array comparative genomic hybridization (aCGH), the sub-chromosomal amplification of the 20q11.21 locus has been reported to be the most frequent copy number variant (CNV) with normal a karyotype [4,[7][8][9] derived from in vitro culture because the gain of 20q11.21 does not occur in normal embryos [10]. In particular, BCL2L1 in 20q11.21 has been identified as a driver mutation for survival advantages in cultures [11][12][13], which also results from an independent loss-offunction mutation in p53 [14]. Other than the survival advantage conferred by the amplification of BCL2L1 or the TP53 mutation, the other genetic alterations have not been closely characterized, although various abnormal behaviors, such as tumor formation [15,16] and impaired differentiation [13,17], in their progenies or hPSCs themselves were reported in animal models. Although CNV, including 20q11.21, occurs even in early passage [18,19], the incidence of abnormal karyotypes may occur in a prolonged culture [20].
Trisomy of chromosomes 12 or 17, which is frequent in embryonal carcinomas [21], is commonly found in hPSCs due to the survival benefit of these abnormalities in culture [22]. Recent studies demonstrated that aneuploid hPSCs (with trisomy in chromosome 12) showed increased proliferation and impaired differentiation in vivo [12,17] and also developed tumors when differentiated cells were transplanted [15,16,23]. However, to date, further molecular mechanisms underlying chromosomal instability (CIN) upon in vitro propagation have remained largely unknown, with the exception of a few studies that have revealed that reduced serum response factor (SRF) expression [24] or escape from mitotic cell death upon mitotic errors is responsible for ongoing CIN in aneuploid hPSCs [25]. Additionally, high cell density and consequent acidification of medium also leads to genome instability of hESCs [26,27].
A targeting protein for Xklp2 (TPX2) located in 20q11.21 has been studied with respect to mitosis-regulating microtubule nucleation, spindle assembly, and centrosome separation during mitosis [28]. It has also been shown to play a role in cancer [28] by activating Aurora A kinase [29] through direct interaction [30]. The genetic abrogation of TPX2 in mouse embryos leads to arrest at the morula stage and early embryonic lethality due to the incorrect formation of bipolar spindles [31] and the mis-segregation of chromosomes, implying that critical roles of TPX2 in early embryonic development [32]. On the other hand, the ectopic expression TPX2 is sufficient to induce polyploidy [33] and elevates malignancy in cancers [34]. Considering the strong association of TPX2 expression with CIN in cancer [35], and with poor patient survival in various forms of cancers [28,34], the deregulation of TPX2 is a putative driver of CIN by accelerating aberrant mitotic spindle dynamics and improper chromosome segregation.
Through transcriptome analysis between late-passage hESCs (LP-hESCs), showing typical culture adapted phenotypes (e.g., survival trait and aberrant mitosis) and early passage hESCs (EP-hESCs), we suggested TPX2, one of the genes in 20q11.21, as a putative driver for aberrant mitosis in hESCs, likely due to its roles in the stabilization of mitotic spindles.

RNA Extraction, Quantitative Real-Time PCR, and Genome-Wide Gene Expression Profiling
Real-time PCR was performed using TB green Premix Taq (Takara, #RR820A) on a LightCycler 480 Instrument II (Roche). Primers are listed in the Supplemental Materials. Total RNA was isolated using Easy-blue reagent (Intron, #17,061) according to the manufacturer's instructions. To construct a sequencing library, we used a TruSeq Stranded mRNA Library Prep Kit (Illumina, San Diego, CA). Briefly, the steps for the strand-specific protocol are as follows: first strand cDNA synthesis; second strand synthesis by using dUTPs instead of dTTPs; end repair, A-tailing, and adaptor ligation; and PCR amplification. Each library was then diluted to 8 pM for 76 cycles of paired-read sequencing (2 X 101 bp) with illumine NovaSeq 6000, per the manufacturer's recommended protocol. Quality check and subsequent trimming (removal of adapters and low-quality bases) of raw RNA-seq fastq files were conducted via FastQC and Trim-Galore, respectively (https:// www. bioin forma tics. babra ham. ac. uk/). The reads were mapped to the GRCh38 reference genome using the STAR aligner (v2.7.3a). Transcripts per million (TPM) and expected read counts were calculated using RSEM v.1.3.3. with assembly GRCh38.84. Assessment of differences in gene expression between groups was conducted using the R package "DESeq2." DEGs were selected with |log 2 fold-change|> 1 and a false discovery rate (FDR) of < 0.01. To identify genomic loci containing the most DEGs, an over-representation analysis was performed using a hypergeometric test with the positional gene sets from the Molecular Signatures Database (MSigDB, https:// www. gsea-msigdb. org/ gsea/ msigdb/). After FDRbased multiple testing correction, an enrichment score for each genomic locus was defined as -log 10 Q-value. For GO analysis, GSEA was performed using the GO BP gene sets from MSigDB via the R package "fgsea." Significantly upregulated or downregulated GO terms were selected with an adjusted P-value of < 0.01 and a |normalized enrichment score (NES)| of > 2. The FASTQ files and processed data are available in the Gene Expression Omnibus (GEO: GSE167495).

Determination of Copy Number Variations
Whole-genome genotyping was performed using the Illumina HumanOmni1-Quad Beadchip (Illumina) containing 1,140,419 genetic markers across the human genome. Samples were processed according to the specifications of the Illumina Infinium HD super assay. Briefly, each sample was whole genome amplified, fragmented, precipitated, and re-suspended in an appropriate hybridization buffer. Denatured samples were hybridized on a prepared BeadChip for a minimum of 16 h at 48 °C. Following hybridization, the bead chips were processed for the single-base extension reaction, stained, and imaged on an Illumina iScan system. Normalized bead intensity data for each sample were loaded into the GenomeStudio software package (Illumina). Ratios of signal intensity were calculated using the Log R Ratio (LRR: logged ratio of observed probe intensity to expected intensity; any deviations from zero in this metric are evidence for copy number change) and allelic intensity was determined by the B allele frequency for all samples. Values were exported using Illumina GenomeStudio. Analysis for structural variants was performed using the sliding window approach (window size 10).

Whole-Exome Sequencing
Whole-exome sequencing was performed using the Ion Proton Platform (Life Technologies). Briefly, 100 ng of gDNA was used for AmpliSeq exome amplification, according to the manufacturer's protocol. The final sequencing libraries were inspected and quantified using the Bioanalyzer 2100 Instrument and DNA HS kit (Agilent Technologies). All libraries were diluted to 100 pM working solutions and then pooled as needed to perform the template preparation on an Ion OneTouch 2 according to the manufacturer's protocols. Template was prepared using the Ion PI™ Template OT2-200-Kitv2 on the Ion OneTouch™ 2 System. Templated Ion Sphere Particles were then enriched for positive ISP using the Ion OneTouch ES, and sequencing was performed using the Ion PI™ Chip-Kit v2 and Ion PI™ Sequencing200-Kitv2 on the Ion Proton™ Sequencer. Sequencing data were processed using the Torrent Suite™ Software (ver.4.0.2) on a Torrent Server. Sequences were aligned against the reference genome (GRCh37/hg19) using the Genomebrowser (DNA Nexus).

Spindle Dynamics Assay
To measure dynamics of spindle, microtubule depolymerization assay and microtubule repolymerization assay were performed as previously described [36]. For depolymerization assay, cells were incubated with 1 µg/mL of nocodazole and fixed with ice-cold methanol (MtOH) time dependent-manner before immunofluorescent. For repolymerization assay, cells were incubated with 1 µg/ mL of nocodazole, and wash-off and fix with ice-cold MtOH time dependent-manner before immunofluorescent. Mitotic spindle fluorescence intensity of metaphase cells, determined by IFC with anti-β-tubulin antibody (DSHB, #Q-E7-S) was quantified.

Statistical Analysis
Graphical data are presented as mean ± standard deviation (SD). Statistical significance 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 significance between the two groups was analyzed using unpaired Student's t-tests. Statistical analysis was performed with GraphPad Prism 8 software (https:// www. graph pad. com/ scien tific-softw are/ prism/). The statistical significance was assumed to be ns: not significant, *p < 0.05, **p < 0.01, and ***p < 0.001.
Of note, these cells showed identical short tandem repeat profiles compared to the control H9 hESCs (Fig.  S1A), implying that they were derived from the same embryonic genetic background through long-term in vitro maintenance. It is noteworthy that a clear survival advantage resulting from culture adaptation occurs in hESCs over 200 passages (P3 and P4 hESCs: LP-hESCs) [12]. Considering a recent study that mitotic stress survival [38] is closely associated with gaining aneuploidy [25], we first monitored mitotic progression. Interestingly, the LP-hESCs showed aberrant chromosome segregation during mitosis (Fig. 1B), high incidences of chromatin bridges (Fig. 1C), and cytokinesis failure (Fig. 1D). Aberrant mitosis in LP-hESCs can be observed in real-time images (Movie S1A-D). Notably, mitotic progression was significantly retarded in P4 hESCs (Fig. 1E). Consistent with previous studies that excessive numbers of centrosomes were closely linked to improper chromosome segregation [39], supernumerary centrosomes were more frequently observed in P4 hESCs compared to P1 hESCs (Fig. 1F).
One clear biological effect observed in hESCs after longterm in vitro culture is the growth advantage [11] marked by increased self-renewal, proliferation, and resistance to apoptosis, which are referred to as "features of culture adaptation" [22]. Thus, these LP-hESCs (P3 and P4 hESCs) were hereafter referred as "culture-adapted hESCs (c-hESCs)". Indeed, our established P3 and P4 hESCs were progressively dominant in cultures when mixed with GFP-expressing P1 hESCs ( Fig. 1G and S1B). This growth advantage might result from the acquisition of robust resistance to apoptosis (Fig. S1C) rather than increased self-renewal ( Fig. S1D) or proliferation (Fig. S1E). As shown in Fig. 1H, the polyploid population identified after mitotic arrest using nocodazole (Noc) was significantly enriched in P4 hESCs. In particular, the clear induction of polyploidy was observed in P4 hESCs when mitotic cell death was blocked with z-VAD-FMK, a pan-caspase inhibitor. This data clearly supports the idea that survival from the mitotic stress would contribute to increase of aneuploidy as previously described [25]. Thus, the survival advantage acquired by "culture adaptation" in hPSCs, which leads not only to resistance to cell death but also to escape from mitotic death, contributes to the increase of polyploidy in c-hESCs.

Recurrent Gain of the 20q11.21 Genomic Locus in Culture-Adapted hESCs
Given that 25% of hESC lines with normal karyotypes contain altered CNVs, which are defined as gains or losses of a relatively small region of a genome [9], our established cell models would exhibit this aberrant CNV profile, conferring the distinct cellular characteristics shown in Figs. 1 and S1. To this end, a high-resolution single nucleotide polymorphism (SNP) array was performed in four variants of the H9 hESC lines, and the genome profile of P1 hESCs was used as a reference genome. As shown in Fig. 2A, numerous gains or losses of CNVs were clearly detected in P2, P3, and P4 hESCs (i.e., 5 gains and 19 losses in P2, 6 gains and 4 losses in P3, and 14 gains and 7 losses in P4). The detected CNVs ranged in size from 479 bases to 3.10 Mb, and the average size of CNVs in P2, P3, and P4 hESCs was 55.28, 585.99, and 476.29 kb, respectively. More importantly, large CNVs (over 100 kb) were exclusively observed in P3 and P4 hESCs (Fig. 2B).
The recurrent chromosome abnormalities that most commonly result from in vitro culture adaptation of hPSCs have been identified as the gain of chromosomes 1, 12, 17, 20, or X [22]. Therefore, the repeated CNVs (i.e., gains or losses) in P2, P3, and P4 hESCs would be predicted to exist among the above chromosomes. As reported elsewhere, the 20q11.21 genomic locus was commonly amplified in P2, P3, and P4 hESCs (Fig. 2C), and the gain of the 17q24 genomic locus [3] was also detected in P4 hESCs (Fig. S2A  and B). It is noteworthy that the gain of various loci of 17q has also frequently been associated with chromosome abnormalities [9].
Of note, given that point mutations are mostly observed in alternative allele frequencies (AAFs) at a rate of around 50% [40], the high calling variants in P2, P3, and P4 hESCs found in AAFs ranging from 40 to 60% indicates that their chromosomes are genetically unstable compared to those of P1 hESCs (Fig. 2D). Taken together, these findings indicate that c-hESCs appear to undergo continued genetic alterations, in which the additional gain of 20q11.21 allows the subclones to survive against selective pressures on genomes.

Perturbed Microtubule Dynamics in Culture-Adapted hESCs
To explore the influence of recurrent CNVs on gene expression, RNAseq was performed on the four hESC sublines (P1, P2, P3 and P4 hESCs). We observed distinct overall gene expression patterns between the non-culture adapted hESCs (nc-hESCs: e.g., P1 and P2)-and c-hESCs (P3 and P4) (Figs. 3A and S3A). Differentially expressed genes (DEGs) also increased dramatically in P3 hESCs where passage number exceeded 200 compared to nc-hESCs (Figs. 3B and S3B). To focus on dysregulated genes in c-hESCs, DEGs were selected by comparing nc-hESCs and c-hESCs (Table S1). As expected, based on the observed gain of 20q11.21 in c-hESCs, the top enriched genomic position of upregulated DEGs was 20q11.21, followed by Xq22 and Xp11, while no significant enrichment was detected in downregulated DEGs (Fig. 3C and Table S1). The expression of the 18 upregulated DEGs located on 20q.11.21 increased gradually over the passages (Fig. 3D), supporting a tight relationship between 20q11.21 CNV and the corresponding gene expression.
To examine functional dysregulation in nc-hESCs, gene set enrichment analysis (GSEA) [41] with gene ontology (GO) was carried out. The most upregulated functions were classified into five categories: microtubule organization, protein catabolic processes, regulation of viral processes, mitochondrion organization, and ATP metabolic process ( Fig. 3E and S3C and Table S2). Interestingly, KIF3B and TPX2 in 20q11.21 contributed most to the enrichment signals for microtubule-, spindle-, or mitosis-related GO terms ( Fig.  S3D and Table S2).
Because the most distinct cellular feature of c-hESCs other than the survival advantage [12] was mitotic aberration (Fig. 1), the enrichment of the gene signatures of c-hESCs in the "microtubule organization," "mitotic spindle organization," or "cell cycle G2M transition" categories was intriguing (Fig. 3F). Accordingly, the microtubule organization would be different between c-hESCs and P1 hESCs. To examine this idea, the microtubule stability of P3 hESCs relative to P1 hESCs was determined using the nocodazole treatment protocol as described previously [36]. As predicted, the mitotic spindle in P3 hESCs remained stabilized compared to P1 hESCs' mitotic spindles (Fig. 3G). Importantly, microtubule dynamics (e.g., spatiotemporal polymerization and depolymerization) are critical for proper chromosome segregation. Thus, gain of 20q11.21 and following alteration of gene expressions, would be expected to significantly influence microtubule dynamics, thus providing an explanation for the high incidence of CIN in c-hESCs.

Prediction of TPX2 for Aberrant Mitosis in Culture-Adapted hESCs
Among the genes within 20q11.21 (Fig. 3D), we attempted to identify a gene responsible for aberrant mitosis and microtubule organization or stability, which were both observed in c-hESCs (Figs. 1 and 3G). Chromosome missegregation (Fig. 1B) may result from altered microtubule dynamics, whereas multipolar spindles (Fig. 1E) are possibly due to robust nucleation and the formation of ectopic spindle poles [42]. Interestingly, among the genes at the 20q11.21 locus, TPX2 and KIF3B were consistently associated with GO terms such as "microtubule cytoskeleton organization", "spindle organization", and "mitotic nuclear division" (Fig. S3D). More importantly, TPX2, which encodes a multifunctional protein that has diverse roles in microtubule nucleation and spindle assembly [28] is closely associated with the highest CIN score among multiple genes associated with cancer prognosis [35]. Notably, TPX2 was the gene that contributed most to the enrichment of GO terms of the "spindle pole" and "cell cycle G2M phase transition" categories in GSEA results (Fig. S3D). Therefore, we hypothesized that the upregulation of TPX2 in c-hESCs would be accountable for chromosomal abnormalities resulting from perturbed microtubule dynamics.

Upregulation of TPX2 Stabilizes Mitotic Spindle
To determine the genomic levels of TPX2 in c-hESCs, we analyzed genome expression patterns using a wholegenome SNP array. As expected, the fluorescence intensity of genetic probes bound to TPX2 gradually increased in P3 and P4 (Fig. 4A), and the TPX2 mRNA levels were also clearly upregulated in P3 and P4 hESCs (Fig. 4B). Consistent with the increased TPX2 mRNA levels, protein levels were also markedly upregulated in P3 and P4 hESCs (Figs. 4C and S6A-D). It is noteworthy that TPX2 serves as a positive regulator of Aurora A kinase (encoded by AURKA) [29] through direct interaction [30], activity of which is also critical for not only microtubule dynamics [43] but also the maintenance of the pluripotency of ESCs through its suppression of the p53 response [44]. Accordingly, active phosphorylation of Aurora A (pAURKA) was evident in P3 and P4 hESCs along with high TPX2 expression (Fig. 4C). Because there were no noticeable changes to the genomic and mRNA levels of AURKA (Fig. S4A) as well as a comparable cell cycle profile (Fig. S4B), the high activity of Aurora A in P3 and P4 hESCs could primarily be attributed to the high expression of TPX2 in P3 and P4 hESCs (Fig. 4C).

Establishment of Inducible TPX2 in Early Passage hESCs
To examine the effects of high TPX2 expression on microtubule dynamics and consequent mitotic event (Fig. 1), we first attempted to establish stable knockdown of TPX2 in P4 hESCs to examine whether depletion of TPX2 restores aberrant mitosis. However, stable depletion of TPX2 in P4 was unsuccessful even after multiple trials for unknown reasons (data not shown), which may account for failure of formation of blastocysts in TPX2 null embryo [32]. We also noticed that TPX2 was highly expressed in hESCs compared to human dermal fibroblasts (hDFs) (Fig. S4E). Instead, we next aimed to determine the effect of TPX2 expression in P1 hESCs. Unexpectedly, stable expression of TPX2 in P1 hESCs was also failed after multiple attempts (data not shown). Alternatively, TPX2 expression was conditionally expressed through a doxycycline (Dox) inducible system in P1 hESCs (iTPX2 hESCs) (Fig. 5A). After Dox treatment, the clear induction of mRNA and the production of TPX2 protein were observed (Figs. 5B and S6E-G), along with fluorescence of an enhanced green fluorescent protein (eGFP) tagged to TPX2 (Fig. 5C). The eGFP signal from the mitotic cells after Dox treatment was clearly associated with the mitotic spindle where the mitotic TPX2 is localized (Fig. 5D). We noticed that the prolonged expression of TPX2 by Dox treatment suppressed cell proliferation (Fig. 5E), which would account for the repeated failure to establish TPX2-stable hESCs. Based on the observed increase in the 4N peak in the eGFP-positive population after Dox treatment (Fig. 5F), the growth suppression by stable expression of TPX2 resulted from the induction of G2/M arrest. TPX2 induction (10-15-fold compared to P1 hESCs) by Dox (1 μg/ mL) resulted in G2/M arrest and growth suppression in iTPX2 hESCs (Fig. 5G). Of note, the basal level of TPX2 in iTPX2 hESCs was comparable to that of P4 hESCs; this may have resulted from the leaky expression of the Tet-On system [45] (Fig. 5G). We then aimed to determine the optimal dose of Dox to induce a moderate level of TPX2 that may not interfere with growth. TPX2 induction in iTPX2 hESCs occurred at Dox concentrations as low as 0.05 μg/mL and dramatically increased in a dose-dependent manner (Fig. 5H). Colony formation at various doses of Dox revealed that hESC growth was minimally affected at a dosage of 0.1 μg/mL (Fig. 5I), where moderate levels of TPX2 were achieved (Figs. 5H and S6H-J).

Spindle Stabilization and Abnormal Chromosome Condensation by TPX2 Induction
Next, we examined whether TPX2 expression affects mitosis in hESCs. Using a Dox concentration that exerts only marginal effects on cell growth (Fig. 5I), the duration of mitosis with or without TPX2 expression was assessed. Similar to P4 hESCs compared to P1 hESCs, the mitotic duration of iTPX2 cells was significantly delayed by TPX2 induction (Fig. 6A). It is generally accepted that the disruption of spindle dynamics affects mitotic duration. In line with this, the extended mitotic duration observed as a result of TPX2 expression may occur because of disarrangement of spindle dynamics. To this end, the assembly or disassembly of mitotic spindles was monitored with or without the moderate induction of TPX2 in iTPX2 hESCs. Similar to the high stability of the mitotic spindles of P4 hESCs, where TPX2 expression was higher than P1 hESCs ( Fig. 4F and G), Doxinduced TPX2 expression in iTPX2 hESCs stabilized the mitotic spindle upon depolymerization (Fig. 6B). While the mitotic spindle was completely depolymerized at 9 min after nocodazole treatment, in the iTPX2 where TPX2 expression was induced by Dox, the spindle remained partly polymerized (Fig. 6C). Consistent repolymerization of the mitotic spindle after the removal of nocodazole was significantly enhanced by Dox treatment (Fig. 6D). This data clearly demonstrates that high TPX2 expression stabilizes the mitotic spindle or promotes spindle assembly, leading to prolonged mitotic duration. Next, to closely monitor the mitotic cells expressing high levels of TPX2, individual mitotic cells were examined by immunofluorescence. Surprisingly, we observed widespread mitotic chromosome condensation abnormalities in iTPX2 hESCs that had undergone Dox treatment for 48 h. Unlike P1 hESCs, failure of chromosome alignment in metaphase was dramatically increased after Dox treatment in iTPX2 hESCs (Fig. 6E). It is also noteworthy that failure of chromosome alignment was higher in iTPX2 even without Dox than that of P1 (Fig. 6E), which may result from the relatively high TPX2 in iTPX2 (due to leakage shown in Fig. 5G). In particular, typical spindle fibers depolymerization observed at telophase of P1 was largely disorganized at iTPX2 hESCs by TPX2 induction (Fig. 6F). These evident disorders occurring in mitotic spindles would be strongly associated to drastic event of polyploid in iTPX2 hESCs (Fig. 6G) in TPX2 level dependent manner (Fig. S5). Overall, our findings indicate that the high expression of TPX2 observed in c-hESCs would delay mitotic progression and induce polyploidy by stabilizing the mitotic spindle.

Discussion
In this study, we found that culture-adapted hESCs acquire a functional signature associated with a gain of 20q11.21 function. GSEA revealed these cells to be highly enriched in the biological process for "microtubule dynamics at G2/M." By multiple approaches, we revealed TPX2, located in 20q11.21, as a putative driver for chromosome misalignment and consequent mitotic abnormality, as inducible expression of TPX2 in nc-hESCs affected mitotic progression and spindle dynamics, which were similarly observed in c-hESCs.
The maintenance of genome integrity in human pluripotent stem cells (hPSCs) is exhaustively challenged by simultaneous exposure to multiple artificial selective pressures during in vitro propagation. Given that hPSCs are vulnerable to genomic insults [1], they progressively acquire genetic alterations or mutations upon prolonged culturing, as shown in Fig. 2. Thereafter, those subclones that survive, are able to do so because they have acquired either resistance to apoptosis or rapid proliferation rates [22], thereby becoming predominant in the cultures (Figs. 1G and S1B). These aberrant hPSCs exhibit recurrent CNVs at chromosomes 12, 17, 20, and X [3,21]. The amplification of sub-chromosomal 20q11.21 has emerged as a frequent variant in cultureadapted hPSCs [9], a discovery propelled by advances in single-nucleotide resolution of genome-wide approaches. Considering that 25% of normal karyotype hESC lines, including hPSCs with early passage [18], and our cell models (Figs. 2 and 3) revealed a gain of 20q11.21 [9], it is likely hard to detect this minimal amplicon using conventional cytogenetic testing methods (i.e., spectral karyotyping and G-band karyotyping). This suggests that genetically altered hPSCs possessing malignancy or chemoresistance acquired by gaining this region cannot be readily perceived [11]. Notably, the gain of 20q11.21 has only been detected in cultured hPSCs and not during embryo derivation [10].
Indeed, our results demonstrate that culture-adapted hESCs are resistant to apoptosis (Fig. S1C) and consequently possess survival traits in cultures (Figs. 1F and S1B). More interestingly, chromosome misalignment, aberrant mitosis (Figs. 1B-E), and consequent polyploidy (Fig. 1G) were frequently observed in c-hESCs in our study, which is consistent with previous reports [46,47]. Although the amplification of 20q11.21 has been determined to provide growth advantages mediated by increased levels of anti-apoptotic factor BCL2L1 [48,11], there is no clear evidence that BCL2L1 is involved in the molecular mechanisms underlying the incidence of CIN. Therefore, to uncover the biological cause of CIN in c-hESCs, we performed GO functional analysis for genes commonly amplified in both CNVs and the gene expression profiles of c-hESCs (Fig. 3). Surprisingly, the genes were almost exclusively included in the 20q11.21 region and were highly enriched in spindle/microtubule or G2M transition annotation (Fig. 3F), a process responsible for maintenance of chromosome integrity [49]. Thus, these results indicate that gaining 20q11.21 can confer survival advantages and chromosomal abnormalities in culture-adapted hPSCs.
Given that TPX2 is commonly associated with the "microtubule regulation" GO term and the 20q11.21 region (Fig.  S3D), the gain of 20q11.21-driven TPX2 amplification would likely be responsible for the deregulation of microtubule organization, which can further lead to chromosome misalignment and abnormal mitosis [35]. Indeed, as shown in Fig. 4, both the mRNA and protein levels of TPX2 were significantly upregulated in c-hESCs, which corresponded with high microtubule stability (Fig. 4). TPX2 has frequently been reported to be overexpressed in various tumor types [28], acting as an oncogenic holoenzyme with Aurora kinase A (AURKA) by engaging its enzymatic activity [29] to be a potential target for genomically unstable cancer cells [50]. As expected, TPX2 induction in P1-hESCs activated Aurora-A kinase (Fig. 5B) and markedly affected mitotic spindle dynamics (Fig. 6), which was similar our observations in P4-hESCs ( Fig. 4F-G). Thus, our findings indicate that TPX2 amplification may trigger chromosome misalignment and aberrant mitosis by controlling microtubule kinetics. It is noteworthy that TPX2 overexpression alone failed to result in cellular transformation but not abnormal spindle formation [29].