RNA polymerase II- mediated transcription can be subdivided in three phases: initiation, elongation and termination. These phases are associated with distinct histone modifications as well as specific phosphorylation patterns of the heptad repeats of the C-terminal domain (CTD) of the largest subunit of RNA polymerase II. Initiation is associated with H3K4 di- and trimethylation and CTD serine 5 phosphorylation (s5pCTD), while H3K36 trimethylation and CTD serine 2 phosphorylation (s2pCTD) correlate with elongation 38. In vitro and in vivo approaches were undertaken to address if uH2A affects transcription initiation or elongation.

First, uH2A engagement in the initiation step of transcription is supported by in vitro studies by Nakagawa and colleagues 25. GAL4-VP16 driven transcription was assayed on reconstituted chromatin templates using Drosophila nuclear extracts as a source of RNA polymerase II. As previously mentioned, uH2A inhibited transcription if the chromatin template was assembled before addition of the RNA polymerase. On similarly reconstituted chromatin templates, the authors showed that uH2A prevents H3K4 di- and trimethylation by the methyltransferase MLL3, and that USP21-mediated deubiquitination could relieve this inhibition (Figure 1B). Importantly, chromatin assembled with a specific mutant of histone H3 (H3K4R), allowed initiation despite the presence of uH2A. Under these conditions, elongation occurred normally, suggesting that uH2A does not inhibit transcription per se, but that it might rather act by preventing H3K4 methylation, and thereby transcription initiation (Figure 2A).

In vivo support for this conclusion came from the study of Zhu et al. on 2A-DUB 27. Expression of the PSA gene is induced upon stimulation of the AR. The authors examined chromatin changes at the PSA promoter, upon AR activation, through detailed ChIP experiments. AR activation resulted in a decrease of uH2A accompanied by increased levels of S5pCTD at the promoter and, therefore, increased transcription initiation. All these events were 2A-DUB dependent. Although the levels of H3K4Me were not examined, this study suggests that uH2A limits initiation in vivo as well.

Second, the link between uH2A and transcription elongation was examined in two recent papers on H2A E3 ligases, Ring1A/B and 2A-HUB 3940.

Stock and colleagues examined chromatin changes at genes that are derepressed upon conditional deletion of Ring1A and B and global loss of H2A monoubiquitination in mouse ES cells 39. The promoters of these genes have been shown to associate with histone marks characteristic of both active (H3K4Me) and inactive (H3K27Me, uH2A) chromatin and are therefore referred to as bivalent 394142, reviewed in 43. Stock et al. showed that in wildtype cells, low-level transcription of the 5' region of the coding region of these genes was detectable. This result indicates that transcription was initiated, despite the presence of Ring1A/B and uH2A (Figure 2B). Indeed, at these promoters, S5pCTD RNA polymerase was present at levels comparable to actively transcribed genes. Given that RNA polymerase II was shown to associate with regions downstream of the promoter upon Ring1A/B deletion, the data support a model in which Ring1B-dependent uH2A hinders transcription at the stage of elongation (Figure 2C).

The involvement of uH2A in transcriptional events downstream of initiation is further supported by the characterization of a novel RING-type E3 ligase for H2A, 2A-HUB 40 (Table 2). Knockdown of 2A-HUB stimulates elongation but not initiation of transcription of one of its target genes, RANTES. uH2A is mainly present at the promoter of RANTES, as opposed to its distribution over both promoter and exonic regions at bivalent genes. As a consequence, in the case of RANTES, uH2A might regulate the transition of initiation to elongation, a process referred to as promoter escape.

In conclusion, it is still unclear at which stage uH2A affects the transcription cycle. It is of note that current data have been obtained in different in vivo and in vitro systems. For example, uH2A inhibits H3K4 methylation in vitro, whereas these marks, by definition, coexist at bivalent promoters in ES cells in vivo. This suggest that the strict inhibition of H3K4 methylation by uH2A observed in vitro is somehow circumvented at bivalent promoters in vivo, allowing initiation. If this interpretation holds true, it would predict that there might be different categories of target genes, regulated by dedicated mechanisms. In addition, we can envisage that, in vivo, uH2A may act as a "landing platform" for recruitment of, yet to be isolated, regulatory proteins containing ubiquitin binding domains (Figure 2C). The identification of such proteins will be pivotal to the understanding of uH2A-mediated regulation. Finally, it cannot be excluded that effects of the DUBs and E3 ligases targeting H2A on transcription phases are partially independent of their enzymatic function.

Work on the DUBs USP21 and 2A-DUB, suggest that uH2A may affect post-translational modifications on other histones, including H3K4 methylation (as discussed before) and H1 phosphorylation 2527 (Figure 1B and 1C). Such "trans-histone" cross-talk between uH2B and H3 methylation has been well characterized 10. We have previously discussed how uH2A negatively affects H3K4 methylation. Here we will report on the uH2A-H1 interplay. Also, we will present findings pointing to a distinct mechanism, involving the histone chaperone FACT.

The C-terminal tail of H2A, containing the K119 ubiquitination site, can reach the linker histone H1 in nucleosomes 4. Immunoprecipitation of nucleosomes containing wt or its ubiquitination mutant, K119R, revealed that H1 preferentially associated with wt H2A, suggesting that ubiquitination of H2A facilitates interaction with H1 40. In agreement, this has been shown in vitro 44. H1 can be phosphorylated: a modification linked to enhanced H1 dynamics and chromatin dissociation 45. Zhou and colleagues reported that a global increase in uH2A, in 2A-DUB knock down cells, was associated with a decrease in phosphorylated H1 40. Knockdown of Ring1B had the opposite effect. Altogether, these findings suggest that uH2A hinders H1 eviction from chromatin, which is generally associated with an open chromatin conformation favorable to transcription (Figure 1C).

Finally, a functional link between uH2A and the histone chaperone FACT (

FA

C

T

In summary, the discussed studies highlight a positive role of H2A DUBs in gene expression. In addition, albeit with some redundancy 272829, it seems that Ubp-M, USP21, USP22 and 2A-DUB participate in the regulation of specific transcriptional programs (Table 1). To understand how DUBs regulate transcription and how specific their activities are, important questions are i) where, along the chromosomes, do these enzymes bind and ii) which genes do they regulate. Gene expression profiles and genome-wide identification of the in vivo DNA binding sites of the DUBs (by ChIP on ChIP, Chip-Seq or DamID techniques) will be needed and will require extensive effort in the next years. Because of the difficulty in developing highly specific antibodies to uH2A, an additional challenge will be to map the chromosomal regions containing uH2A. Finally, integration of these data with available genome-wide chromatin association data on histone modifications and crucial chromatin modifiers, including histone E3 ligases, methyltransferases, demethylases and remodeling enzymes, will likely lead to further insight into the intricate cross talk between histone modifications.

Four independent studies recently identified RNF8 as the E3 ligase responsible for H2A/H2AX ubiquitination in the response to DSBs 18195455. RNF8 rapidly accumulates at DSBs upon IR, concomitantly with early IRIF markers, namely γH2AX, ATM, the MRN complex, and the mediator protein MDC1 19. The presence of a phosphothreonyl-binding FHA domain together with a RING finger domain enable RNF8 to link phosphorylation with ubiquitination signaling at IRIFs (Figure 1D). The data presented by the four laboratories are consistent with a signaling cascade starting from phosphorylation of H2AX by ATM (Figure 3). This step is well known and allows direct recruitment of MDC1 and its subsequent phosphorylation by ATM 59. Through its FHA domain, RNF8 can in turn bind to phospho-MDC1 at DSBs, where it catalyzes polyubiquitination events, among which H2A and H2AX ubiquitination (Figure 3A). At IRIF, RNF8 likely acts in concert with the ubiquitin conjugating enzyme Ubc13, a previously reported interactor of RNF8 1853545563. At last, RNF8-Ubc13-dependent ubiquitination is required for recruitment and retention of BRCA1 and 53BP1 at DSBs 18195455 (Figure 3B). Besides histone ubiquitination, ubiquitin signaling at IRIF likely comprises a complex variety of ubiquitination events, as also suggested by the partial requirement of a second E3 ligase activity, BRCA1, for efficient ubiquitin accumulation 6062. These may include amplification of the ubiquitination signal on histones or other substrates, as well as autoubiquitination of the E3 enzymes as autoregulatory mechanism 6465. In agreement, both RNF8 and BRCA1 can promote auto-ubiquitination and ubiquitination of histones in vitro 19636667.

These studies revealed an exciting link between ATM signaling and ubiquitination of histone H2A and H2AX at double strand breaks. If uH2A is directly involved in recruitment of DDR factors at IRIF remains to be established.

The ubiquitin-interacting-motif (UIM)-containing protein Rap80 is a good candidate for fulfilling a ubiquitin receptor function at IRIF. Rap80 was initially shown to recruit the Rap80-ccd98/abraxas-BRCA1 complex to ubiquitin-conjugates at DSBs 616869 (Figure 3B). More recently, it was demonstrated that accumulation of Rap80 at IRIF is dependent on functional RNF8-Ubc13 proteins 18195455. In vitro and in vivo Rap80 displays preferential binding to lysine63-isopeptide-linked tetraubiquitin polymers, and to a lesser extent to K6-linked chains 61. Importantly, Rap80 ubiquitin binding properties fit well with the findings that i) chain formation through lysine 63 has been previously shown to regulate, among others, DNA repair processes 60, ii) ubiquitin foci at IRIF form most efficiently through K63-linked polyubiquitination 61 and iii) Ubc13 is the only E2 able to catalyze polyubiquitination through K63 70, and it is an E2 partner for RNF8 63. Although these evidences are consistent with Rap80 linking RNF8-Ubc13 ubiquitylation at IRIF to the DDR mediator BRCA1, studies addressing the potential direct binding of Rap80 to ubiquitinated H2A/H2AX are still needed.

It is important to note that other mechanisms may exist through which H2A ubiquitination could affect DDR. One plausible possibility is that uH2A/H2AX may facilitate a more global alteration of the chromatin, known to occur at DSBs 5771, allowing exposure of other histone marks. This mechanism might be relevant for 53BP1 relocalization to IRIF, as suggested by the findings that 53BP1 binds to methylated histones 7273 and that its recruitment is independent from Rap80 18195568. Interestingly, Ikura and colleagues reported enhanced mobility of H2AX at sites of microirradiation and a rapid release of polyubiquitinated H2AX from the chromatin 20. H2AX ubiquitination and histone release was dependent on TIP60 and on its acetylation target, H2AX K5 (Figure 1D). Drosophila Tip60 can acetylate H2Av (the Drosophila H2AX) and promotes its exchange with unphosphorylated H2Av 74. This work suggests a mechanism by which ubiquitination might promote histone dynamics/removal.

The mechanisms by which the chromatin post-translational modifications are cleared, the protein assemblies are disassembled, and the checkpoint is turned off after completion of repair is an intriguing open question. Work form our group recently showed that USP3, previously identified as a DUB 75, displays deubiquitination activity both towards uH2A as well as uH2B in vivo 17. USP3 appears to be engaged in DDR in two majors ways: in the response to IR and during normal S-phase progression (as discussed below). Upon IR, IRIF containing uH2A, ubiquitin, and γH2AX persist in USP3 knockdown cells 17. Previously published results demonstrated that dephosphorylation and removal of H2AX is needed for efficient repair and resumption of the cell cycle 767778. In agreement, USP3 knock down cells are significantly delayed in G2/M transition. Together, these data suggest that removal of ubiquitination marks at the chromatin surrounding DSBs is necessary to coordinate IRIF disassembly, γH2AX dephosphorylation and cell cycle recovery (Figure 3C). Unlike in S. cerevisiae 7778, dephosphorylation of H2AX by the phosphatase PP2A is thought to occur on the chromatin, given that this enzyme localizes to IRIF 76. If and how USP3 crosstalks with PP2A, for example by affecting recruitment or activity of the latter, is currently unknown. Although USP3 seems to favor deubiquitination of substrates (among which uH2A) at IRIF, we could not detect local accumulation of USP3 at those foci (unpublished results). This may be a consequence of the fact that USP3 rapidly releases its chromatin substrate upon catalysis as determined by Fluorescence recovery after photobleaching (FRAP) and co-immunoprecipitation 17. Also, it is possible that USP3 DUB activity may be required 'off the chromatin" or may be indirect. Isolation of USP3 interacting partners and potential additional substrates, as well as addressing if and how USP3 is regulated, is needed to gain a better understanding of the molecular mechanism of USP3-mediated DDR.

A JAMM domain containing DUB, BRCC36, may also play a role at DSBs: it is part of the BRCA1 A complex (together with RAP80 and Abraxas), localizes to IRIF and positively regulates BRCA1/BARD1 E3 ligase 55617980. However, BRCC36 role at IRIF is not clear and its activity towards ubiquitinated histones has not been investigated.

In summary, the discovery of novel E3 ligase (RNF8) and DUB activities (USP3) for H2A provides us with valuable tools to address ubiquitin-mediated signaling at the chromatin. We can envision several experimental approaches towards the elucidation of the molecular mechanism(s) of uH2A/uH2AX-mediated DDR. These include: i) biochemical characterization of the E3 (RNF8, BRCA1) and DUB activities (USP3, others?) on nucleosomes; ii) identification of the IR-induced ubiquitination site(s) on H2A/H2AX; iii) isolation of uH2A/uH2AX binding proteins; iv) definition of the extent of the ubiquitin mark around the DSB. Also, a proteomic approach towards the characterization of the complex mixture of ubiquitinated proteins and ubiquitin receptors at the chromatin will potentially reveal novel players in DDR.

Among the DUBs targeting H2A, USP22, Ubp-M and USP3 have been connected to cell cycle progression. Consistently, knockdown of these enzymes resulted in growth impairment in different human cell lines, albeit in different ways: i) upon shRNA-mediated knock-down of USP22, both p53 deficient H1299 cells and normal human fibroblast accumulated in G1 28; ii) USP3-depleted U2OS cells showed a strong delay in S-phase progression and a low mitotic index 17; iii) stable Ubp-M knockdown in HeLa cells resulted in a decrease in the proportion of cells in G2/M 24. The potential impact of the two other H2A DUBs, 2A-DUB and USP21, on cell cycle progression has not been investigated yet 2527.

As discussed before, uH2A is functionally linked to transcription regulation and DNA damage signaling pathways. Does uH2A affect cell proliferation through transcriptional regulation of cell cycle genes or DDR activation, or through direct mechanism(s)? The data suggest that both indirect and direct mechanisms may be relevant.

First, USP22 has been characterized as a transcriptional co-activator in the context of the SAGA complex (see also the previous paragraph on transcription) 2829. Consequently, it is possible that the USP22 inhibitory effect on G1/S transition is dependent on its transcriptional activity. In addition, USP22 is required for Myc-driven transcription and transformation. Given that USP22, as well its Drosophila homolog Nonstop, appear to regulate a large subset of genes, gene expression analysis and genetic studies will be required to identify the key USP22-target genes involved in cell cycle progression 282935.

Second, data by our group point to a distinct mechanism by which USP3 may affect cell cycle. S-phase delay in USP3-KD cells is accompanied by spontaneous γH2AX foci, accumulation of DNA breaks and full activation of an ATM- and ATR-mediated checkpoint response 17. Specific phenotypes were observed upon USP3 KD, including i) reduced ability to incorporate BrdU, witnessing impaired DNA synthesis, ii) the presence of single-stranded DNA coated by replication protein A (RPA), iii) activation of the ATR-Chk1 pathway. These features are consistent with USP3 KD cells undergoing replication stress and suggest that S-phase delay and G2/M arrest are intimately linked to DDR activation in these cells. Wheter USP3 acts directly affecting the DNA replication process/machinery, or rather through uH2A-mediated DDR signaling, is presently the subject of investigation. Persistence of γH2AX and uH2A DNA damage foci in USP3 KD cells upon thymidine-induced replication stress, as well as upon exogenous DNA damage (as discussed before), supports the participation of USP3 in ubiquitin (uH2A)-mediated signaling (F. Nicassio and E. Citterio unpublished results). Finally, the observation that γH2AX foci, in USP3-depleted cells, coincide with the onset of S-phase suggests that DNA replication is required for DDR activation (F. Nicassio and E. Citterio unpublished results). Investigating how USP3 influences replication dynamics will possibly contribute to the understanding of USP3/uH2A role at replication forks and its functional interplay with the S-phase checkpoint.

Third, H2A is globally de-ubiquitinated in the G2/M transition, which correlates with phosphorylation of H3 at S10, a hallmark of mitosis 24. Interestingly, Joo et al. showed that Ubp-M is required for mitotic H2A de-ubiquitination and that knockdown of UbpM results in G2/M delay. In addition, in vitro experiments using reconstituted nucleosomes reveal that uH2A inhibits H3 phosphorylation on serine 10 by AuroraB, at least in part by preventing the binding of the kinase to nucleosomes (Figure 1E). De-ubiquitination of H2A by UbpM relieves the inhibition of AuroraB. Since H3S10 phosphorylation by AuroraB is needed for chromosome condensation, these in vitro results put forward the exciting possibility that Ubp-M-mediated H2A deubiquitination may affect G2/M transition through a direct mechanism. Further investigation will reveal if this is true in vivo.

Intriguingly, despite targeting the same substrate, H2A DUBs display specificity in their roles in cell cycle regulation. It is unknown whether this specificity is due to additional protein targets, besides H2A, or to other mechanisms. Therefore, it will be interesting to assess if different H2A DUBs can rescue each other defects in cell cycle progression.

Finally, the H2A E3 ligases Ring1B and RNF8 are clearly implicated in cell proliferation. Work by different groups support a functional role for RNF8 in mitosis, showing its requirement for mitotic exit 8485. Whether the effects of RNF8 on mitosis are dependent on ubiquitination of H2A remains to be investigated. As to Ring1B, its deletion in mice is associated with gastrulation arrest and embryonic lethality 86. Given that Ring1B deletion strongly impact on expression of its target genes, it is difficult to distinguish between direct and indirect effects of Ring1B on cell cycle progression 8687. Knockdown of Ring1B results in cell cycle arrest in U2OS cells 11. It will be interesting to analyze this phenotype in more detail to assess which phase of the cell cycle is affected.