Citrullination regulates pluripotency and histone H1 binding to chromatin

Pluripotent cells have the capacity to self-renew and differentiate into all somatic and germ cell lineages and, hence, possess immense therapeutic potential for a multitude of medical conditions. Their generation by reprogramming of differentiated somatic cells has been achieved by nuclear transfer, cell fusion and transduction of transcription factors, such as Oct4, Sox2, Klf4 and c-Myc¹². Pluripotent cells have a distinctly open chromatin structure that is essential for unrestricted developmental potential^13,14 and reprogramming involves an almost complete epigenetic resetting of somatic cells¹³. The ability of PADI4-mediated histone citrullination to induce chromatin decondensation in neutrophils¹⁰ prompted us to ask whether it can play a role in pluripotency, where chromatin decondensation is also necessary. To investigate this we first assessed the expression of Padi4 in the embryonic stem cell line ES Oct4-GIP (ES), the neural stem cell line NSO4G (NS) and in induced pluripotent stem (iPS) cells derived NSO4G (see Methods). Padi4 is expressed in pluripotent ES and iPS but not multipotent NS cells (Fig.1a). Culture of ES cells in 2i/LIF medium establishes a ground state of pluripotency¹⁵. This leads to the down-regulation of lineage specific markers and the up-regulation of pluripotency factors, as well as rapid induction of Padi4 (Fig. 1b). The pattern of Padi4 expression follows closely that of Nanog, an essential transcription factor for the transition to ground state pluripotency¹⁶(Fig 1a,b). While other PADIs are expressed in pluripotent cells, PADI4 is the only one whose expression clearly associates with naïve pluripotency (Extended Data Fig. 1a,b). Citrullination of histone H3 (H3Cit), a modification shown previously to be carried out specifically by PADI4¹⁷, is detectable in ES and iPS cells (Extended Data Fig. 1c), indicating that PADI4 is also enzymatically active. H3 and global citrullination are undetectable in NS cells (Extended Data Fig. 1c,d).

To determine the kinetics of PADI4 activation during the establishment of pluripotency, we examined RNA and protein samples collected daily during the course of reprogramming of NSO4G into iPS cells¹⁶. Padi4 is rapidly induced in NS cells after transduction of reprogramming factors but only becomes active to citrullinate H3 after introduction of 2i/LIF, closely following the onset of Nanog expression (Fig. 1c). These observations strongly suggested that PADI4 activity is associated with ground state pluripotency and prompted us to examine whether Padi4 is part of the pluripotency transcriptional network.

First, we asked whether the reprogramming factors regulate Padi4 expression, utilizing the ZHBTc4.1 and 2TS22C cell lines where Oct4 and Sox2, respectively, can be deleted acutely in response to doxycycline treatment (see Methods). Deletion of Oct4, but not Sox2, led to a decrease in Padi4 mRNA levels (Extended Data Fig. 1e). Furthermore, while Oct4 and Klf4 occupy the Padi4 promoter in ES, but not NS cells, Sox2 is bound in both cell types (Extended Data Fig. 1f). To understand the effects of PADI4 on transcriptional regulation in pluripotent cells, we analyzed the transcriptome of ES cells upon PADI4 over-expression and inhibition. Several key pluripotency genes are up-regulated in response to PADI4 over-expression (Fig. 1d, Extended Data Figure 2a and Supplementary Table 2), including Klf2, Tcl1, Tcfap2c, and Kit. Tcl1 was previously identified as the only regulator of self-renewal up-regulated in ground state pluripotency¹⁵ and over-expression of Tcl1 or Tcfap2c positively influence this process¹⁸. Gene Ontology (GO) analysis of this dataset indicates an enrichment of genes involved in stem cell development and maintenance (Fig. 1e). In addition, knockdown of Padi4 in mES cells leads to decreased expression of Tcl1 and Nanog, which is rescued by exogenous expression of RNAi-resistant human PADI4 (Fig. 1f, Extended Data Fig. 2b,c). These genes are under the control of PADI4 enzymatic activity since treatment with the chemical inhibitor Cl-amidine, which disrupts citrullination by PADI4¹⁹, down-regulates their expression (Fig. 1g). Chromatin immunoprecipitation (ChIP) analysis indicated that H3Cit is present on regulatory regions of Tcl1 and Nanog in ES and iPS cells, but not NS cells (Fig. 1h and Extended Data Fig. 3a). Accordingly, exogenously expressed human PADI4 localizes to and is enzymatically active on these regions, as well as regulatory regions of Klf2 and Kit in ES cells (Extended Data Fig. 3b,c). In contrast to PADI4 over-expression, treatment of ES cells with Cl-amidine led to up-regulation of differentiation markers such as Prickle1, EphA1 and Wnt8a and down-regulation of pluripotency markers such as Klf5 (Extended Data Fig. 4a,b, and Supplementary Table 3), in addition to Nanog and Tcl1 (Fig. 1g). GO analysis of this dataset indicated enrichment in genes involved in cell differentiation (Extended Data Fig. 4c). Pou5f1 (Oct4), Klf4, Sox2 and c-Myc, were not affected by PADI4 modulation (Extended Data Fig. 2a and Supplementary Tables 2 and 3). Cumulatively, the above results place PADI4 within the pluripotency transcriptional network, suggesting that it acts downstream of some of the cardinal reprogramming factors to regulate a specific subset of pluripotency genes.

Prompted by the above observations, we investigated whether PADI4 is necessary for pluripotency, as assessed during reprogramming (Extended Data Fig. 5a) and in the early stages of embryo development. NSO4G cells express a GFP reporter under the control of the Oct4 regulatory sequences, which is activated upon acquisition of pluripotency¹⁶, allowing us to trace reprogrammed cells. Knock-down of Padi4 in NSO4G-derived pre-iPS cells impaired the ability of the cells to establish H3Cit upon switch to 2i/LIF medium and led to a clear reduction in reprogramming (Fig. 2a,b, Extended Data Fig. 5b-e and Extended Data time-lapse video). Consistent with this, levels of Tcl1 and Nanog were not elevated upon reprogramming to the same extent as in control cells (Fig. 2b and Extended Data Fig. 5f). Cl-amidine treatment led to a dramatic reduction of reprogramming efficiency and H3Cit (Fig. 2c and Extended Data Fig. 5g-i), suggesting that the catalytic activity of PADI4 is important for the induction of pluripotency.

Padi4 expression and H3Cit are detected in the early embryo^20,21 and Padi4-null mice are born in lower numbers than would be expected by Mendelian inheritance²², suggesting that PADI4 loss affects embryonic development. To assess the role of PADI4 in early development, we cultured mouse embryos in Cl-amidine-containing medium from the 2-cell stage and throughout pre-implantation development (see Methods and Extended Data Fig. 6a-c). Using 200µM Cl-amidine resulted in a complete developmental arrest of the embryos at the 8-cell stage (Extended Data Fig. 6a). We therefore used the maximum dose of Cl-amidine that reduced H3Cit (Extended Data Fig. S6b,c) but did not induce arrest (10µM). This led to a reduced percentage of pluripotent Nanog-positive epiblast cells and an increased percentage of differentiated trophectoderm cells at the blastocyst stage (Fig. 2d,e and Extended Data Fig. 6d,e). Time-course analyses of the cleavage patterns and cell fate decisions in early embryos showed that Cl-amidine increased the number of symmetric cell divisions at the expense of asymmetric divisions at the 8 to 16- and 16 to 32-cell transitions (Fig. 2f,g). This resulted in 16-cell stage embryos with fewer inner cells (destined for pluripotency) and greater numbers of outer cells (destined for differentiation into trophectoderm, reviewed in ²³) (Extended Data Fig. 6f). Treatment with another PADI4 inhibitor, TDFA²⁴, but not the HDAC inhibitor TSA, had similar effects (Extended Data Fig. 7, 8). These results indicate that PADI4 activity also promotes the maintenance of pluripotent cells in the early mouse embryo.

To elucidate the molecular mechanisms by which PADI4 regulates pluripotency, we aimed to identify PADI4 substrates in the chromatin fraction of mES cells using Stable Isotope Labeling of Amino acids in Cell culture (SILAC) (Fig. 3a and Extended Data Fig. 9a,b). Among the identified PADI4 substrates were Atrx, Dnmt3b, Trim28 and variants of linker histone H1 (Fig. 3b-e, Extended Data Fig. S9c-f, S10 and Supplementary Table 4), all of which can impact pluripotency. Histone H1 stabilizes the nucleosome and facilitates chromatin condensation, a state that is less permissive to processes that require access to the DNA, such as transcription¹¹. The identified citrullinated H1 peptides correspond to, and are common between, variants H1.2, H1.3 and H1.4 (Fig. 3d,e), while an additional peptide corresponds to the same residue in H1.5 (Extended Data Fig. 10a,b). In ES cells, H1.2, H1.3 and H1.4 are required for chromatin compaction²⁵, while their depletion leads to increased expression of pluripotency genes such as Nanog and stalls them in a self-renewal state with impaired differentiation capability²⁶. Notably, H1 is more loosely bound to chromatin in ES cells than in differentiated cells¹⁴ and its genomic localization in cancer cells was shown to anti-correlate with that of PADI4⁷.

Mass spectrometric analysis accounted for all arginine residues within H1 but indicated that Arg54 (H1R54) is the only site citrullinated by PADI4 (Fig. 3e and Extended Data Fig. 9c). Indeed, we found that while H1.2 is citrullinated in ES cells (already by endogenous PADI4, and significantly increased upon PADI4 over-expression), it is refractory to modification when Arg54 is mutated (Fig. 3f). Similar results were obtained in in vitro citrullination assays (Extended Data Fig. 11a). H1R54 lies within the globular domain of H1 (Extended Data Fig. 11b), which is highly conserved among the linker histone family and is necessary for interaction with nucleosomal DNA^27,28. To test whether H1R54 is necessary for binding of H1 to nucleosomes, we mutated and assessed it in nucleosome-binding assays. Figure 3g shows that an R54A mutant, which mimics the charge change that accompanies citrullination, is impaired for nucleosome binding. An R54K mutant, which retains the positive charge, is impaired to a lesser extend (Fig. 3g) suggesting that H1R54 is important for electrostatic interactions between H1.2 and the nucleosome.

The above results open up the possibility that PADI4 may affect chromatin compaction in pluripotent cells. To test this hypothesis, we first assessed whether citrullination by ectopic PADI4 can lead to decondensation of differentiated cell chromatin. Recombinant PADI4 was added to permeabilized and stabilized differentiated C2C12 mouse myoblast nuclei (Fig. 4a). This protocol ensures stabilization of the nuclear component while allowing the free diffusion of non-chromatin bound nuclear proteins into the extra-nuclear fraction, and their collection by washing. Incubation with active PADI4 (Extended Data Fig. 12a,b), leads to the eviction of H1 from the chromatin and its diffusion out of the permeabilized nucleus (Fig. 4b). The evicted H1 is citrullinated on R54, as determined by mass spectrometry (Extended Data Figure 12c,d). Consistent with this, PADI4-treated cells showed evidence of decondensed chromatin, as determined by nuclear swelling, diffuse DAPI staining and increased sensitivity to micrococcal nuclease (Fig. 4c,d and Extended Data Figure 12e). Similar results were observed when PADI4 was over-expressed in C2C12 cells (Fig. 4e) or NS cells (data not shown). To monitor if PADI4 can affect H1 binding on pluripotent cell chromatin, we performed ChIP-qPCR analyses of H1.2 on the regulatory regions of Tcl1 and Nanog and found that it is stabilized upon Padi4 knockdown (Fig. 4f). The ability of PADI4 to disrupt the binding of H1 to nucleosomal DNA provides a novel mechanistic example of how citrullination regulates protein function and chromatin condensation.

The work described above identifies citrullination of chromatin components by PADI4 as a feature of pluripotency (Fig. 4g), in addition to its previously described role in the myeloid lineage. One of the reasons for the restricted expression pattern of PADI4 may be the requirement for an open chromatin state in these cell types. The selective expression characteristics and the inducible nature of the catalytic activity of PADI4, suggest that it is under tight spatial and temporal regulation, giving it a unique status among chromatin modifying enzymes. As such, inappropriate PADI4 activity may have deleterious consequences, which may explain its activation in cancers of varying origin during progression to malignancy²⁹. Indeed, citrullination is a common feature of several unrelated diseases, suggesting that strict regulation is likely to be a requirement for the physiological function of all PADIs. During review of this manuscript, Coonrod and colleagues suggested that PADI2, thought to be mainly cytoplasmic, can also citrullinate histones and lead to transcriptional activation³⁰. This opens the intriguing possibility that other PADIs may mediate nuclear events in specific contexts, including in pluripotent cells. Further research into the function and targets of PADIs is likely to shed light into the etiology of several pathologies.