Abstract
Development is generally viewed as one-way traffic of cell state transition from primitive to developmentally advanced states. However, molecular mechanisms that ensure the unidirectional transition of cell fates remain largely unknown. Through exact transcription start site mapping, we report an evolutionarily conserved BTB domain-containing zinc finger protein, ZBTB12, as a molecular barrier for dedifferentiation of human pluripotent stem cells (hPSCs). Single-cell RNA sequencing reveals that ZBTB12 is essential for three germ layer differentiation by blocking hPSC dedifferentiation. Mechanistically, ZBTB12 fine-tunes the expression of human endogenous retrovirus H (HERVH), a primate-specific retrotransposon, and targets specific transcripts that utilize HERVH as a regulatory element. In particular, the downregulation of HERVH-overlapping long non-coding RNAs (lncRNAs) by ZBTB12 is necessary for a successful exit from a pluripotent state and lineage derivation. Overall, we identify ZBTB12 as a molecular barrier that safeguards the unidirectional transition of metastable stem cell fates toward developmentally advanced states.
Introduction
Stem cell states are recently recognized to be highly metastable1,2. Pluripotent stem cells (PSCs) provide a well-defined model system to understand molecular principles of dynamic stem cell states. PSCs lie in a wide spectrum of pluripotent states, ranging from naïve to primed states. Pre- and post-implantation epiblasts in vivo respectively represent naïve and primed pluripotent states1,2. Recent studies have further shown an intermediate state called formative pluripotency that exists in a developmental continuum between the naïve and primed states and represents a critical period for lineage capacitation3,4,5,6,7,8. Naïve PSCs show resistance to differentiation and the naïve through formative to primed transition that occurs in peri-implantation development is essential for pluripotency exit and three germ layer differentiation6,9,10,11. Furthermore, mouse PSCs cultured with serum exhibit substantial intra-population heterogeneity with a mixture of naïve- and primed-like cells, which represents a dynamic equilibrium of cells transitioning between distinct metastable states12,13. Although human embryonic stem cells (hESCs) are established from inner cell mass of pre-implantation blastocysts, they are more likely to be in a primed pluripotent state1,2. Recently, a number of studies have introduced different combinations of chemical inhibitors to drive hESCs to a naïve-like pluripotent state1. Like serum-maintained mouse PSCs, the presence of naïve-like cells was also reported in a conventional primed hESC population14 although the percentage of naïve-like cells is much lower in hESCs than in mESCs. Given the high degree of flexibility in stem cell states, it is plausible that as cells exit from a primed pluripotent state, hESCs would confront either differentiation into three germ layers or dedifferentiation toward a naïve-like state. The concept of embryonic dedifferentiation is supported by recent studies about neural crest cell development, in which natural dedifferentiation occurs to drive neural crest cell specification from neuroectoderm15. However, it remains unknown how dedifferentiation is blocked once stem cells undergo differentiation.
Endogenous retroviruses (ERVs) are genetic remnants of retroviral infection during evolution and encompass about 8% of the human genome16,17. Increasing evidence suggests that ERVs have significantly contributed to rewiring regulatory networks by creating new transcription factor (TF) binding sites18,19. Given the species-specific ERV integration events and high ERV expression in the early development, it has been proposed that ERVs play key roles in dictating species-specific developmental programs20,21. In particular, human endogenous retrovirus H (HERVH) is a primate-specific ERV family and exhibits high expression in human blastocyst stages22. HERVH-associated long terminal repeat 7 (LTR7) harbors binding sites for several key pluripotency factors such as NANOG, OCT4 (also known as POU5F1), KLF4, and TFCP2L1 (also known as LBP9), and thereby acts as strong promoters and enhancers in hPSCs23. Recently, it was reported that elevated expression of HERVs could be a key feature of naïve-like pluripotent states14,24,25, raising the possibility that primate-specific HERVs could serve as a molecular rheostat for dynamic human pluripotent states. Likewise, retrotransposons including ERVs and LINEs regulate the cell fate transition between two-cell embryos and blastocysts in mice26,27. Beyond embryonic development, it has been increasingly reported that cellular dedifferentiation occurs in various types of cancers with elevated HERV expression28. Therefore, it would be crucial to investigate precise roles and regulation of HERVs in dynamic cell state transition to understand human diseases.
Previous studies have been focusing on TFs that exhibit dynamic expression changes with strong and specific expression in stem cells compared to differentiated cells. Such studies resulted in identification of core pluripotency factors such as OCT4, NANOG, and SOX229,30,31. However, in this study we employ a different approach that is not dependent on expression changes of TFs. Rather, we precisely map transcription start sites (TSSs) of transcripts that show differential expression patterns during stem cell differentiation and then identify potential transcriptional regulators by TF motif analysis on promoter sequences. This approach enables us to discover a new pluripotency regulator with broad expression in multiple cell types and tissues. ZBTB12, a BTB (BR-C, ttk and bab) domain-containing zinc finger protein, is expressed in the nucleus of hESCs and balance self-renewal and differentiation. ZBTB12 depletion skews hESCs toward self-renewal and delays the exit from a pluripotent state. Single cell transcriptomic analysis reveals that upon differentiation ZBTB12-depleted hESCs aberrantly transit backward to a naïve-like state and thus fail to differentiate into three germ layers. This suggests that ZBTB12 serves as a barrier of dedifferentiation. As a downstream mechanism, we find that ZBTB12 transcriptionally regulates HERVH expression. In contrast to KRAB domain-containing ZFPs (KRAB-ZFPs) that have co-evolved with target ERVs17, ZBTB12 is an evolutionarily conserved gene in vertebrates, yet it has acquired a regulatory function on specific transcripts that make use of LTR7/HERVH promoters during primate evolution. These findings provide an interesting example of host cells utilizing an old regulator to deal with newly evolved genetic elements. A growing body of recent evidence suggests that the genomic integration of HERVH with its strong promoter activity substantially contributed to the evolutionary emergence of primate-specific long non-coding RNAs (lncRNAs)32. Consistent with this idea, ZBTB12 suppresses a number of HERVH-overlapping lncRNAs including LINC-ROR and ESRG. ZBTB12-mediated shutdown of the lncRNAs drives the exit from a pluripotent state and safeguards three germ layer differentiation. Overall, we have identified ZBTB12 as a molecular barrier that restricts dynamic stem cell states and safeguards stem cell differentiation.
Results
Identification of ZBTB12 as a key pluripotency regulator by nanoCAGE-seq
To identify TFs that regulate hESC differentiation, we established an hESC-derived neural differentiation model33 and performed nanoCAGE sequencing from cells collected at Day 0, Day 5, Day 15, and Day 23 with two replicates at each time point (Supplementary Fig. 1a, b). The nanoCAGE method captures the 5′ends of transcripts and thereby enables genome-wide mapping of transcription start sites (TSSs) as well as transcriptome profiling34. The reproducibility of the nanoCAGE method was validated by the principal component analysis and the correlation analysis for all the samples (Supplementary Fig. 1c, d). Community detection analysis based on 3712 differentially expressed genes (DEGs) resulted in three modules that represent three sequential transition stages during neural differentiation (Supplementary Fig.1e–g). Given the high correlation between replicates (Supplementary Fig. 1d), we merged the sequencing data of replicates for TSS peaks identification with HOMER (http://homer.ucsd.edu/homer/). With stringent criteria that an individual peak is supported by at least 30 unique reads with high confidence, 11,019–12,503 peaks were identified in each sample (Supplementary Data 1). Overall, about 70% of the peaks were annotated as TSS peaks due to their location near annotated TSSs (<1 kb from nearest TSS). Other peaks were far from annotated TSSs, representing 3′UTRs, intergenic regions, intronic regions, and exonic regions (Supplementary Fig. 1h, i).
Given that our data accurately identified the TSS position of each transcript, we extracted 300 bp promoter sequences (250 bp upstream and 50 bp downstream of TSS) of stage-specific transcripts. TF motif prediction with HOMER uncovered dynamic changes of TFs throughout sequential cell fate transitions during neural differentiation of hESCs (Supplementary Table 1). The number of stage-specific peaks and predicted TFs suggest that widespread transcriptional changes occur during early stages of pluripotency exit (Supplementary Fig. 1j, k). Therefore, we focused on TFs with enriched motifs on Day 0. These TFs include well-known pluripotency factors, such as OCT4, NANOG, SOX2, MYC, and KLF5, and provide confidence in our results (Fig. 1a). Of particular interest was ZBTB12, a zinc finger protein that contains a BTB domain and four zinc finger C2H2 domains, because its functions were unknown. Similar to the core pluripotency factor OCT4, endogenous ZBTB12 proteins were detected in the nuclei of hESCs by immunofluorescence staining (Fig. 1b, Supplementary Fig. 2a). Tagged exogenous ZBTB12 proteins also showed nuclear localization (Supplementary Fig. 2b). To investigate the role of ZBTB12 in hESCs, we downregulated ZBTB12 with two short hairpin RNAs (shRNAs) (Supplementary Fig. 2c, d). ZBTB12 knockdown (KD) in hESCs did not affect pluripotent state maintenance for over ten passages (Supplementary Fig. 2e). Founder cells in hESC colonies have a repopulating capacity when seeded at a clonal density35,36. Strikingly, ZBTB12 KD boosted the colony-initiating capacity with increased number of alkaline phosphatase (AP)-positive colonies (Fig. 1c, Supplementary Fig. 2f). Furthermore, exogenous expression of ZBTB12 reduced the colony-initiating capacity (Fig. 1d, Supplementary Fig. 2g, h). These findings lead to an interesting concept that undifferentiated hESCs express ZBTB12 in order to restrict their self-renewal ability.
