In the current study, we show that the mRNA levels of CENP-A in human PSCs and day 7 differentiated progeny are significantly higher than fully differentiated cycling somatic fibroblast cells. This elevated level of CENP-A mRNA is found regardless of whether hPSCs were derived by culture-induced conversion of the inner cell mass, or by induced reprogramming of human fibroblasts. However, despite this higher mRNA load, protein analysis indicates that the amount of CENP-A in hPSC chromatin is identical to fibroblasts. A second novel finding is that depletion of CENP-A mRNA and protein in hPSCs under self-renewing conditions causes no significant phenotype, whereas inducing differentiation of the CENP-A depleted cells causes accumulation of cells in G2/M, and a significant increase in p53-dependent apoptosis. Together these results suggest that the amount of CENP-A protein required to define the centromere may be cell type, and context specific in mammalian cells, with undifferentiated hPSCs requiring less CENP-A protein to faithfully retain the centromeric epigenetic mark relative to fibroblasts (Fig. ). Finally, given that the threshold for significantly depleting CENP-A protein in hPSCs was lower than fibroblasts, we induced DNA damage in CENP-A-depleted self-renewing hPSCs, and determined that CENP-A-depleted hPSCs undergo significantly more apoptosis 24 h after irradiation, which is 22 h after the dynamic remodeling of CENP-A foci that occur in response to DNA damage.
Figure 6. Summary of differences between CENP-A expression and functional dynamics in fibroblasts and hPSCs. Our data demonstrate that the relative levels of CENP-A mRNA are significantly higher in undifferentiated hPSCs and hPSCs treated with retinoic acid (RA) (more ...)
Previous reports using microarray analysis determined that CENP-A
mRNA levels were higher in oocytes and hESCs relative to somatic cells (27
). Our study confirms these findings; however, we further show that despite the elevated CENP-A
mRNA in undifferentiated hPSCs, protein levels of CENP-A are equivalent to fibroblasts. This uncoupling of relative RNA to protein levels could be explained by the unique RNA translational controls recently identified in murine PSCs (41
). For example, in murine ESCs, ‘parsimonious translation’ is hypothesized to define the pluripotent state, with undifferentiated ESCs containing 78% lower ribosome loading of RNA transcripts relative to their differentiated progeny, which establishes an RNA pool of specific transcripts that do not undergo productive protein synthesis (41
). Translational regulation is also a critical feature of oocyte growth and the transition to an embryonic developmental program after fertilization. Translational regulation in the oocyte is a complex process involving multi-component RNA-binding complexes, compartmentalization of maternal RNAs and polyadenylation-induced translation (reviewed in 43
). Whether the elevated levels of CENP-A
mRNA are a product of poor polyribosome activity and content in hPSCs and/or the presence of a hPSC-specific CENP-A RNA-binding protein that regulate translation remain to be determined.
The depletion experiments in the current study suggest that under circumstances where CENP-A
mRNA and protein levels are reduced, the threshold of CENP-A required to maintain a functional centromeric mark must be lower in hPSCs relative to fibroblasts because proliferation and self-renewal were unaffected. Depletion of CENP-A
was achieved using shRNAs which are known to mostly act through RNA degradation (45
). This would suggest that the high mRNA reserves of CENP-A
in hPSCs are required to sustain wild-type levels of CENP-A protein. The paradox in this analysis is that reducing CENP-A mRNA and protein levels in undifferentiated hPSCs using shRNAs did not affect proliferation or self-renewal, whereas reducing CENP-A by the same technique in fibroblasts was lethal. One possible explanation is the unique chromatin configuration of PSCs compared with fibroblasts. For example, hPSCs have what is termed an ‘open’ chromatin configuration largely devoid of heterochromatin (37
). In support of this, we show that the heterochromatic mark H3K27me3 is low in the undifferentiated hPSCs used in the current study, and that this epigenetic mark is significantly increased upon hPSC differentiation. We also show that as H3K27me3 levels increase during differentiation so do the levels of CENP-A in chromatin. The purpose of an open chromatin state is hypothesized to create the plasticity for mounting a rapid response to differentiation signals (reviewed in 47
). Using human artificial chromosomes, it has been shown that heterochromatin can act as one of the major determinants of minimal centromeric length (48
). Therefore, if reduced heterochromatin is also found at the centromeres of undifferentiated hPSCs, this unique chromatin architecture may create the opportunity for sustaining a functional centromere under circumstances where CENP-A protein levels become significantly reduced. This hypothesis would only be plausible if heterochromatin were not necessary to sustain a functional centromere, and indeed this was recently shown using neocentromeres as a model in lymphoblast and fibroblast BBB lines (49
Our data indicate that once CENP-A-depleted hPSCs exit the self-renewing state, lineage-committed CENP-A-depleted cells undergo P53-dependent apoptosis. In support of this, a recent study using human primary fibroblasts revealed that entry into senescence in CENP-A-depleted fibroblasts is also P53 dependent (50
). Therefore, our data support the hypothesis proposed by Maehara et al
) that p53 is a major surveillance factor for centromeric defects, and our study extends these original findings to indicate that reduced levels of CENP-A during early lineage commitment in the embryo would also act to prevent aneupoidy as a consequence of compromised centromere function.
A non-centromeric role for CENP-A
in DNA damage repair has previously been proposed using Xenopus sperm as well as mouse and human somatic cells (38
). In order to evaluate the dynamics of CENP-A in hPSCs following DNA damage, we used irradiation to induce double-strand breaks, and in agreement with previous studies, we identified an increase in the number of CENP-A
foci shortly after inducing damage (38
). However, our data stand in contrast to previous reports, as we did not observe an increase in foci size, or direct correlation with γH2AX staining. These differences may be explained by species differences, or an inherent difference between diploid hPSCs and haploid sperm. Given that in our study, CENP-A foci did not correlate with γH2AX after DNA damage, this would indicate that CENP-A is not localizing stably to induced break points in hPSCs (38
). Furthermore, at 60 min after DNA damage, CENP-A foci also do not co-localize with CENP-C. This would indicate that the centromere is no longer functional during this period and therefore, before the cells can resume mitosis after DNA damage and repair, functional centromeres will need to be rebuilt. Inducing DNA damage of CENP-A-depleted hPSCs revealed that 24 h after DNA damage, CENP-A depleted cells are undergoing significantly higher levels of apoptosis relative to control. Combined with the result that CENP-C and CENP-A foci do not correlate following DNA damage, our interpretation is that increased apoptosis at 24 h is not due to defects in DNA repair associated with γH2AX, because CENP-A and γH2AX show no major overlap. Instead, we favor the hypothesis that increased apoptosis in CENP-A-depleted cells following DNA damage is due to reduced availability of new CENP-A protein to rebuild the CENP-A/CENP-C uncoupled centromeres following DNA damage.
In conclusion, cultured hPSCs are unique cell types that closely resemble the inner cell mass cells in blastocysts that normally persist for only a few days before changing fate to create every cell type in the body including the germline. In the first week post-fertilization, the human embryo is dividing on a rigid time scale governed by female reproductive physiology that creates a small window for successful implantation. If cell division occurs too slowly, then the embryo is lost because the window for receptive implantation is over. Our results suggest that hPSCs create both a reserve of CENP-A mRNA as well as the flexibility to reduce the amount of CENP-A necessary to create a functional centromere. This mechanism functions in favor of cell division, while minimizing the potential for post-fertilization aneuploidies due to defective centromeres. The fact that hIPS cells maintain this same phenomenon indicates that establishing a dynamic range for CENP-A is not merely a relic from the pre-implantation embryo, but an integral aspect of hPSC self-renewal. Finally, our data indicate that hPSCs provide a new and unique tool with which to evaluate the biology of CENP-A (Fig. ) and the pathways by which centromeric nucleosomes in primary human cells of the same genotype are dynamically remodeled under conditions of self-renewal, differentiation and DNA damage.