Processes that Regulate the Ubiquitination of Chromatin and Chromatin-Associated Proteins

Ubiquitin is a post-translational modification important for many different processes in the cell, including antigen presentation and proteosomal degradation of proteins. It is heavily involved in the regulation of chromatin and the proteins that control chromatin-related processes. In this review, we will focus on ubiquitin-based chromatin regulation involved in four different processes. The first is DNA double strand break (DSB) repair and the role that ubiquitin plays in not just recruiting and stimulating DSB repair, but also the choice of pathway. The second is the PAF1 complex, which is involved in transcriptional elongation and interacts with RNAPII. The third is polycomb repressive complexes, specifically polycomb repressive complex 1, which utilizes ubiquitin to repress constitutively inactive genes. The last role of ubiquitin discussed is ubiquitin as a mitotic bookmark, which serves to provide a record of -active genes as cells transit mitosis. Each of these processes has independent pathways, but each is necessary for proper cellular function and organismal health.


Introduction
Ubiquitin is most clearly associated with the process of targeted protein degradation, but it is involved in many cellular processes such as chromatin regulation, immune response, and antigen processing [1]. Proteosomal degradation is mediated through polyubiquitin chains linked via lysine-48 (K48) on the ubiquitin chain, interacting with the proteasome. Other processes utilize monoubiquitination or polymerization of ubiquitin molecules via another lysine. The ubiquitination system in humans is incredibly complex, with over 1000 known factors and over 10,000 known sites of ubiquitination, enabling its many and diverse roles in cellular biology.
In this review, we focus on four specific roles of ubiquitin in regulating chromatin: DNA repair, transcription elongation, epigenetic silencing via the polycomb repressive complex, and bookmark ubiquitination. In these processes, the ubiquitin moiety interfaces with many other epigenetic marks, such as acetylation, methylation, and histone modification to regulate a given process.
The process of ubiquitination (or ubiquitylation) is the attaching of one ubiquitin protein to a substrate and is performed by a cascade of three enzymes: E1 (activating), E2 (conjugating) and E3 (ligase) [2][3][4][5]. Substrates are proteins, and in the context of chromatin, histones are the most common class of substrate [6]. Histones form octamers containing two each of the four core histones (H2A, H2B, H3, H4), and when a histone octamer is wrapped with two turns of DNA, it is called a nucleosome. Each histone has a tail that extends outside the core of the nucleosome, where it is more accessible to the modifying enzymes. In addition to ubiquitination, other modifications occur, such as acetylation or methylation. The primary role of these marks is governing the localization on the genome of specific epigenetic marks as well as the compaction and decompaction of chromatin, which regulates accessibility of the transcription machinery to chromatin. Histone ubiquitination also serves as signaling molecules for other downstream regulators of transcription, which modulates transcription both directly and indirectly. Part of this concept includes histone cross-talk, where those regulators of transcription integrate signals of multiple distinct histone modifications on the same or nearby histones to generate a phenotype due to the composite signals [7]. Therefore, nucleosome modification has a fundamental function in silencing and activation of transcription. Most histone ubiquitination occurs as a monoubiquitination, but polyubiquitin chains have also been observed. There are a number of small ubiquitin-like modifier (SUMO) proteins that share structural resemblance to ubiquitin and play some similar roles [8]. SUMO proteins come in a variety of isoforms with varying capacity for chain formation and are conjugated to substrates in a similar manner as ubiquitination.

Ubiquitin in DNA damage repair
Survival of organisms and their cells depends on stability and integrity of DNA, but maintaining this integrity is a challenge for cells because they are constantly subjected to DNA damage from a variety of sources [9,10]. DNA damage can cause disease and prevent faithful transfer of genetic information from one generation to the next. DNA double strand breaks (DSB) present a difficult problem to correct since there may be no template to guide error-free repair. In the DNA damage response, cells arrest the cell cycle and activate repair machinery. There are two primary methods eukaryotic cells use to repair DSBs: nonhomologous end joining (NHEJ) and homologous recombination (HR) [11]. The NHEJ pathway occurs throughout the cell cycle (except during mitosis) and is performed more commonly, but it is error-prone since it does not utilize a template [12]. By contrast, HR is only active during S and G2 phases of the cell cycle and because it uses the template in a sister chromatid, the repair has higher fidelity. The faulty repair observed in NHEJ can cause chromosomal rearrangements and mutations, leading to cancer susceptibility. In the following paragraphs, we highlight the roles of a variety of ubiquitin ligases and ubiquitin binding proteins to regulate the DNA damage response.

Initiation of DSB repair
When DSBs occur, ionizing radiation-induced foci form from genome-localized high concentrations of repair machinery with a host of bound factors necessary for DSB repair. To form these ionizing radiation-induced foci, histones near the damage site become modified with K63-linked polyubiquitin chains via the action of the E2 Mms2-Ubc13 and the E3 ligases RNF168 and RNF8 [13]. These K63 chains serve as markers for recruitment of downstream repair proteins and as transcriptional repressors to prevent propagation of problems caused by broken DNA strands.
Once RNF168 has been recruited by RNF8-mediated H1 ubiquitination, it can recognize its own H2A ubiquitination mark due to the UBD on its C-terminus, allowing self-propagation of the DSB repair response [18]. RNF168-mediated H2A is capable of monoubiquitination of H2AK13-15 [19]. Chain elongation by RNF168 is atypical: as mentioned, most DSB-related ubiquitination is comprised of polyubiquitin chains linked by K63, but RNF168, when overexpressed, creates K27 linked chains instead of linkages via K63 [20]. H2AK15ub is important for the recruitment of downstream factors, most importantly 53BP1, which promotes NHEJ [21]. RNF8 and RNF168 recruit more E3 ubiquitin ligases through direct interaction with HERC2 and via their ubiquitin ligase activity, but these other E3 ligases stimulate DSB repair as scaffolds, rather than as ubiquitin ligases [22]. For example, BRCA1 and BARD1 have E3 ligase function, but their role in DSB repair is independent of their ubiquitination activity [23,24].  In NHEJ, DNA broken strand ends are bound by the Ku70-Ku80 heterodimer, which in turns allows recruitment of the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs) [25,26]. Following phosphorylation of DNA-PKcs, the DNA ends are trimmed to make them ready for ligation. This trimming plays a major role in why NHEJ is error-prone and is more damaging to cells than HR. DNA ligase IV and its associated proteins are responsible for ligating these trimmed ends and finishing the NHEJ process.
HR is dependent on a series of posttranslational modifications, including ubiquitination, reviewed in [27]. These modifications control a carefully orchestrated system that recruits and displaces DNA repair factors at multiple different sites during the process of HR. The role of ubiquitin in regulating HR is predicated on the following well-established model of HR. When broken DNA strands are detected, the MRN complex begins end clipping via the endonuclease activity of MRE11, along with other proteins such as EXO1, CtIP, and DNA2, creating extensive ssDNA near the break site. In addition to end clipping, the MRN complex recruits ATM. Resection of these ends at the break site allows binding by single-strand DNA binding protein, replication protein A (RPA), which is then displaced by RAD51, which enables the important difference between HR and NHEJ. RAD51 searches for homology and locates a template with which to use to repair the damaged strand. Because HR depends on this template, it can only occur during the S and G2 phases of the cell cycle, when the damaged DNA has already been replicated and the copied DNA serves as a template strand. Choosing which repair pathway, NHEJ or HR, a cell uses for repair is an important determiner of genome stability and involves complex regulatory processes involving ubiquitin.
Regulation of DSB repair depends on a variety of ubiquitination writers, readers and erasers. The readers and erasers require specific UBDs to recognize their specific conformations of ubiquitination. There are more than 20 unique types UBDs that are found in mammalian proteins and a few of these are enriched in proteins associated with DNA damage repair: ubiquitin-interacting motif (UIM), ubiquitinbinding zinc finger (UBZ) and motif interacting with ubiquitin (MIU) [28]. These proteins possess multiple UBDs that each bind to the target cooperatively to increase specificity and affinity. In several ubiquitin ligases, including RNF168, RNF169, RAD18, and RAP80, specificity is increased further because they possess ligand-binding regions adjacent to their UBDs, allow cooperative binding at higher specificities and affinities than otherwise possible [29].

RNF8 and RNF168 regulation
RNF8 and RNF168 are important regulators of the entire DSB repair response, and require careful regulation themselves. Because these ubiquitin ligases can recognize the same mark they create, their action can cause over-recruitment and overproduction of ionizing radiation-induced foci without control mechanisms. Overproduction of these ionizing radiation-induced foci would cause widespread transcriptional repression across much larger portions of the genome than necessary. One mechanism of limitation is direct ubiquitination of RNF168 by TRIP12 and UBR5, ubiquitin ligases that recognize certain N-terminal domains and direct proteosomal protein degradation on those targets, causing a decrease in the amount of RNF168 in the cell ( Figure 1B) [30].
Another method of RNF8 and RNF168 regulation occurs via deubiquitinating enzymes (DUBs) ( Figure 1C). Ubiquitin-specific protease 3 (USP3) has been shown to increase genomic instability and lead to spontaneous tumors when depleted in mice. This finding was supported when it was shown that UPS3 depletion led to increased levels of H2A ubiquitination, indicating the role of properly regulated H2A ubiquitination in DSB repair [31]. USP3, USP16 and USP44 and their family members also deubiquitinate H2A and thus downregulate the DSB response. One of the significant differences between these DUBs is their affinity for different lysine chains. For example, USP3 is known to target the H2A protein K13 and K15 sites that RNF168 targets as well as the K119 and K120 monoubiquitination sites of PRC1. In contrast, PSMD14 deubiqutinates K63-linked poly-ubiquitin, a different RNF8-RNF168 mediated target [32,33]. Another DUB, USP14, downregulates DSB repair by decreasing RNF168 ubiquitination and RNF168-mediated ubiquitin signals in the setting of inhibited autophagy [34].
RNF8 and RNF168-mediated DSB repair can also be downregulated by phosphorylation. During mitosis, chromatin structure undergoes massive changes and most nuclear processes pause. Phosphorylation of RNF8 and MDC1 (a scaffold protein) prevents DSB repair from occurring during mitosis by blocking their interaction with 53BP1.

Pathway choice
As there are two primary pathways of DSB repair that function through entirely different mechanisms, NHEJ versus HR, cells must decide which pathway to activate. HR requires a perfect homolog to use as a template across the DSB, and for this reason HR should only function following replication of the DNA during S phase or in G2. Cells must have built-in mechanisms to suppress HR during G1, since during this stage of the cell cycle HR would use inappropriate nonhomologous DNAs as template and thus be mutagenic [35]. In addition, during mitosis, NHEJ is repressed by phosphorylation and inactivation of 53BP1 and RNF8 by the cyclin-dependent kinase CDK1 [36]. This inactivation of DNA repair is protective against chromosomal fusions at telomeres that would lead to aneuploidy. The structure of the DSB is also a factor in the decision of cells to engage in which pathway. In general, more complex DSB structures cannot be repaired via NHEJ and require the more time-consuming HR pathway [37,38].
The RNF8-RNF168 ubiquitination pathway plays an important role in determining which DSB pathway will predominate in a cell. BRCA1 stimulates HR but antagonizes NHEJ, and conversely 53BP1 antagonizes HR and promotes NHEJ [39,40]. Ubiquitination via RNF8 and RNF168 leads to retention of 53BP1 and BRCA1 at DSB sites and the balance between these two proteins is the primary decision point between NHEJ and HR. 53BP1 functions to inhibit end resection that is necessary for HR, allowing only NHEJ to be performed. 53BP1 binds H2AK15ub (catalyzed by RNF168) through its own ubiquitin-dependent recruitment motif and also possesses Tudor domains, which recognize H4K20me2 [21,41]. It is proposed that H4K20me2 is the signal that 53BP1 recognizes to promote its recruitment at DSB sites. RNF8-168 ubiquitinates other proteins that impact the pathway selection, including JMJD2A, JMJD3A, and L3MBTL1. When these three factors are ubiquitinated, they are released from H4K20me2. Through the action of JMJD2A, JMJD3A, and L3MBTL1 vacating H4K20me2, 53BP1 can bind freely without competition ( Figure 1D) [42,43]. H4K20me2 is another mechanism supporting the cell cycle dependent decision point between pathways. As S phase continues and more DNA is replicated, H4K20me2 becomes diluted between the two replicated DNA strands, reducing 53BP1 capacity for binding through its Tudor domains, and shifting the balance away from NHEJ to HR [44]. While 53BP1 is bound, RIF1 (and other factors) are recruited to 53BP1 and inhibit resection of the DNA ends. These factors are responsible for replacing BRCA1 at DSB sites, inhibiting HR. BRCA1, in turn, inhibits RIF1 binding at these sites, inhibiting NHEJ.
The antagonist of 53BP1 is BRCA1, which promotes HR over NHEJ by way of supporting RAD51 activity. BRCA1 is a scaffold with activity that also depends on RAP80, which has a ubiquitin binding domain suspected to recognize RNF8-RNF168mediated H2A ubiquitination and is a part of the BRCA1-A complex, which it targets to these sites of RNF8-RNF168 ubiquitination [45,46]. However, RAP80 depletion does not lead to the expected abolishment of HR, but to increased HR activity. To explain this, it has been proposed that RAP80 is functioning to sequester BRCA1 away from DSB sites, so when RAP80 is removed, BRCA1 recruitment to DSB sites is unregulated, leading to the over activity of HR [47]. This would suggest an unknown regulator of BRCA1 recruitment to sites requiring HR activity. BRCA1 also antagonizes 53BP1 by recruiting phosphorylated UHRF1, an E3 ligase that ubiquitinates RIF1, which is bound to 53BP1 at DSB sites. Ubiquitinated RIF1 becomes displaced, reducing 53BP1 mediated repression of DNA end resection [48]. Cockayne syndrome B (CSB) protein has been proposed as fulfilling this role because it seems to antagonize 53BP1 support of NHEJ [49,50]. In addition, when CSB is removed, DNA damage responses have been limited and CSB has been found accumulating at DSB sites.
One mechanism of HR regulation is the proteasome-mediated degradation of factors important to HR, such as CtIP [51]. The decision point depends on the resection of broken DNA ends by factors such as CtIP in HR. During S and G2, when HR is stimulated, CtIP is ubiquitinated by RNF138 to promote CtIP localization to DSB sites [52]. RNF138 also ubiquitinates the NHEJ factor Ku80 during S phase, causing the Ku70/80 heterodimer to dissociate from DSB sites, and thus suppressing NHEJ during S and G2 phases [53].
HR is also inhibited during G1 via ubiquitination of PALB2, a factor involved in HR along with BRCA1 [54]. This ubiquitination by the E3 ligase complex CRL3 is in the BRCA1-binding domain of PALB2 and sterically blocks the two proteins from binding. The ubiquitination is antagonized by USP11, a DUB that is degraded during G1. Thus, while the balance of CRL3 to USP11 is heavily in favor of CRL3 in G1, the BRCA1-binding site on PALB2 is ubiquitinated, and so the PALB2-BRCA1 interaction is blocked, preventing BRCA1 activity, and therefore, HR.

Ubiquitin and transcription elongation
A second key process regulated by ubiquitination of chromatin is transcription. As mRNA is transcribed, one protein complex that associates with the elongating RNA Polymerase II (RNAPII) is the Polymerase Associated Factor 1 Complex (PAF1) (Figure 2A) [55][56][57]. The PAF1 complex regulates RNAPII related transcription elongation and posttranscriptional events and is conserved across many species.
The PAF1 complex in humans is comprised of six protein subunits: PAF1, CDC73, CTR9, LEO1, RTF1, and WDR61 [58]. WDR61 is not present in yeast, although it is present in humans. Cells without PAF1 or CTR9 have a global decrease in protein levels and exhibit growth defects [59, 60]. The complex is found on active genes, at levels directly relating to transcription [61][62][63]. PAF1 binds directly to the carboxy terminal domain (CTD) of RNAPII via the Cdc73 subunit when the RNAPII CTD becomes phosphorylated via CDK9, and via Rtf1 binding along with the elongation factor Spt5 [64][65][66]. The localization and recruitment of PAF1C to specific sites on active genes is dependent upon many factors; in humans, PAF1C recruitment is highest at the transcription start site (TSS) or immediately (~2 nucleosomes) following the TSS [62,67].
The PAF1 complex regulates transcription and the chromatin template to ensure its readiness for transcription. The impact of PAF1C on human chromatin was first established from its role in the ubiquitination of histone H2B at K120 (Figure 2B) [68]. H2Bub is an important epigenetic mark that is associated with both activating and deactivating transcription, though its primary effect on chromatin is disrupting compaction [69]. The ubiquitination at H2BK120 is catalyzed by the E3 ligase complex containing RNF20/40, which interacts directly with the PAF1 complex, and is conjugated by the E2 UBE2A/2B. In addition to H2B ubiquitination at K120, monoubiquitination can also occur at K34 on H2B, a separate mark placed by the heterodimeric E3 ligase, MSL1/2 [70]. Both H2BK120ub and H2BK34ub stimulate histone methylation at H3K79 and H3K4, which has been demonstrated via decreases in both H2BK120ub and H2BK34ub following PAF1 depletion [71]. H2BK120ub is necessary for H3K4 and H3K79 trimethylation, while H2BK34ub functions through trans-tail crosstalk to regulate these methylations. The mechanism of this effect was revealed by experiments showing that depletion of PAF1 caused a decrease in RNF20/40 and MSL1/2 association to chromatin, indicating the role of PAF1C as promoting localization of these E3 ligases, the method by which PAF1C regulates H2Bub [71]. RNF20/40 and MSL1/2 each depend on the specific binding to chromatin by the other ligase and the corresponding histone mark, demonstrating how much interdependence exists between the two co-regulated ligases. This interaction has multiples sources. CDK9 is a kinase that promotes PAF1C association to chromatin, but is itself dependent on both PAF1C-mediated chromatin marks for its chromatin association [71].
Deregulation of this pathway and H2B monoubiquitination is commonly found in cancers. This can occur via multiple mechanisms, such as mutations in CDC73, one of the components of PAF1C, which has been observed in multiple cancers [72]. In addition, silencing of expression by methylation of the RNF20 promoter and RNF20 enhancers has also been observed in many breast cancers [73]. However, it has also been observed that decreased levels of H2Bub have also been shown to be associated with decreased tumor growth, an apparent contradiction to H2Bub as a cancer-causing mutation [73]. Deregulation of a mark can also occur from overactive removal; there are several DUBs responsible for H2Bub deubiquitination, including USP3, USP7, USP12, USP22, USP44, USP46, USP49 [74][75][76][77][78]. Upregulation of these DUBs can cause similar phenotypes as RNF20 depletion. Errors in H2B monoubiquitination lead to errors in chromatin structure on scales larger than the aberrantly ubiquitinated nucleosome [69].
Dysregulation of RNF20 and concomitant H2B ubiquitination has been linked to a wide variety of cancer pathways. One method for H2Bub depletion leading to cancer occurs via H2Bub regulated inflammation and the interaction with NF-κB [79]. Inflammation involves the production of cytokines and chemokines that promote oncogenic activity and NF-κB is a key regulator of the inflammatory system. Reduction in H2Bub has been shown to lead to activated NF-κB, and thus its downstream regulation targets, leading to active inflammation in mice. This was indeed shown to lead to increased colorectal cancer in these animals. Ovarian cancers also display H2Bub dysfunction. One study found that the majority of high grade serous ovarian cancers show global decreases in H2Bub [80]. The most deadly cancer worldwide is lung cancer, and one of the more common forms of lung cancer is lung adenocarcinoma. In human lung adenocarcinomas, H2Bub decreases have been associated with increased cancer burden and a less differentiated carcinoma, a marker of poor prognosis [79].
Mixed lineage leukemia is a classification of cancers that depend on the MLL1 gene, and rearrangements of MLL1 have been shown to be dependent on RNF20 and its role in chromatin regulation [81]. Cells lacking RNF20 showed decreased tumor growth [82]. This role of RNF20 allowing cancer progression is contrary to its role in protecting against the above cancers, but does serve to highlight the fundamental role that H2B ubiquitination plays in maintenance of chromatin.

Polycomb repressive complex
While the preceding section described how histone H2B is ubiquitinated at multiples sites as a part of active transcription process, this section describes the ubiquitination of histone H2A, which has an opposite impact on gene expression. The polycomb repressive complex 1 (PRC1) monoubiquitinates H2A at lysine 119. H2AK119ub is a repressive mark, associated with inactive transcription by condensing chromatin, making it less accessible by transcription factors and associated machinery [83]. This repressive mark is only the most common role of PRC1, as it has also been shown to have diverse effects that have the overall impact of permanently silencing chromatin as part of the differentiation process. The two polycomb group (PcG) complexes, PRC1 and PRC2, modify chromatin to repress transcription and lead to the methylation of the promoter DNA to stably repress transcription at targeted genes. PRC2 contains the methyltransferase EZH2, which methylates histone H3 on lysine 27, H3K27me3. This review will focus on PRC1. PRC2 is necessary for targeted recruitment of PRC1, as experiments have shown that knockdown of PRC2 components also decrease PRC1-mediated H2A ubiquitination. The ubiquitination function of PRC1 is antagonized by the last form of polycomb repressive complex, polycomb repressive deubiquitinase (PR-DUB), which deubiquitinates H3K119 [84,85]. The complimentary actions of PRC1 and PR-DUB to regulate H2AK119ub suggests the fundamental role it plays in repressing transcription.
PRC1 complexes exist in a number of different forms that have the same general structure, with different proteins occupying each position. The core of each complex contains a RING protein and a polycomb group RING finger protein, which bind via  [86]. This core serves as the base for further PRC1 proteins to bind. There are two possible RING proteins, RING1A and RING1B, and six possible polycomb group ring finger (PCGF) proteins, PCGF1-6. All eight of these proteins possess a RAWUL (RING finger and WD40 Ubiquitin-like) domain somewhere in their structure, which bind additional proteins [87,88]. These additional proteins include chromobox and human polyhomeotic homolog (HPH) proteins, which, when included in the PRC1, form what is known as canonical PRC1 [89]. It was previously assumed canonical PRC1 performed the H2Aub function that is associated with PRC1, but it is now known that noncanonical PRC1 also plays an important role in gene regulation [90]. Between the variability of chromobox, HPH, RING, and PCGF proteins, there are well over 100 unique combinations of canonical PRC1 complexes that can form. This diversity plays an important role in the diverse targeting and functions exhibited by PRC1. The PCGF member of the complex binds specifically to a variety of proteins, which are responsible for targeting and regulation of the PRC1 activity [91]. Accordingly, PCGF RAWUL domains exhibit more selective binding than their counterpart RAWUL domains on the RING1A or RING1B protein. The importance of the RING domains is that the RING proteins are E3 ubiquitin ligases, responsible for the primary activity of H2AK119 ubiquitination. However, PCGF-4 (also known as BMI-1) and RING1A both do not directly ubiquitinate H2A, as only RING1B directly ubiquitinates H2A. Instead complexes containing BMI-1 and RING1A serve to promote the RING1B E3 ligase activity [92].

PRC1 function
RING1B monoubiquitinates H2AK119 as part of the PRC1 activity following PRC2 methylation at H3K27 (Figure 3). PcG-regulated genes show aberrant transcriptional levels following removal of PRC1 via RING1B knockdown using shRNA [93]. PRC1-related ubiquitination and subsequent gene silencing is associated with multiple silencing contexts. PcG proteins are known to occupy and thus regulate, developmental genes, X-chromosome inactivation, and parent of origin imprinting. The most widely accepted model of the activity of PRC1-mediated inactivation of target genes is through chromatin compaction. Promoters of active genes become compacted in the setting of PRC1 action, preventing RNA polymerases from accessing the targeted gene, and therefore preventing transcription. This concept has been supported by in vitro experiments and in vivo experiments showing decreased nuclease digestion at genes with PRC1-mediated compaction of chromatin [94]. While the fact that PRC1-mediated ubiquitination of H2A leads to diminished transcription via chromatin compaction is indisputable, the mechanism is currently unclear. It has been shown that PRC1 does not have a role in regulating chromatin accessibility, only to nucleosome spacing and occupancy. Identifying the direct mechanism by which PRC1-mediated H2Aub inhibits transcription needs further elucidation.
Targeting of PcG complexes occurs via Polycomb Response Elements (PREs), which are DNA elements that cannot be recognized by any PcG protein because PcG proteins do not appear to possess any sequence specific DNA binding subunits [95]. While several proteins have been suggested to have a role in recognizing the PRE and enabling recruitment of PcG complexes, none have been confirmed to be sufficient to mediate PcG recruitment alone, suggesting the PcG recruitment is dependent on the interactions of several proteins coordinately creating a stable protein-DNA complex [96]. In addition to protein-DNA interactions to promote PcG recruitment, protein-protein interactions are important. PRC2 has histone methyltransferase function, methylating H3K27. PRC1 can directly recognize the H3K27me3 mark produced by PRC2 [97,98]. Similarly, PRC2 can bind to H2Aub. This complementary interaction can serve to support preservation of PcG silencing across disruptive events to the genome, such as DNA synthesis, when histones are divided between the sister chromosomes [99]. Therefore, the most commonly accepted model of PcG recruitment is that PREs are recognized by adaptor proteins, which recruit PRC2 to promoters, which methylates H3K27. PRC1 recognizes H3K27me3 and is recruited to ubiquitinate H2AK119.
In addition to the repressive effect of PRC1, it has been shown to have activating effects on transcription. PcG proteins mediate their activity by regulating genome architecture [100]. It has been reported in mouse embryonic stem cells that RING1A and RING1B organize genes into three-dimensional interaction networks, which maintains interactions between promoters in the network. When PRC1 was removed, promoter-enhancer interactions were affected, leading to activation of affected promoters and increased transcription. This supports the compactionbased theory of PRC1 transcriptional repression and provides a mechanism for this activity. Deep sequencing of ChIP experiments against selected PRC1 proteins, including both RING1A and RING1B, has shown their enrichment at active transcriptional sites in human fibroblasts [101]. This experiment also showed cell-type specific binding of PRC1. RING1B, the primary ubiquitin ligase involved PRC1mediated H2Aub, has been found associated with Aurora B kinase at active promoters in lymphocytes, while RING1B knockdown decreased transcription at these sites, suggesting an important activating function of RING1B [102]. Cells that have had conditionally-inactivated RING1A and RING1B, and thus inactivated PRC1, exhibit errors in DNA replication [103]. Slow elongation and even stalling of replication forks has been observed in these cells in specific pericentromeric regions. These S phase errors were rescued by monoubiquitination events, suggesting the role of RING1A/B in S phase is dependent on their function as ubiquitin ligases. In breast cancer, RING1B has been found at oncogene promoters, playing an activating role and promoting cancer development and metastasis [104]. All these activating effects of PcG proteins suggest that there is much not known about the diverse array of proteins involved in PRC1 and that understanding PRC1 function may explain many previously unknown chromatin regulation events.

Ubiquitin as a mitotic bookmark
Mitotic bookmarks are a mechanism of dividing cells that maintain the epigenetic and transcriptional state despite the rigors demanded by the mitosis process [105]. Although epigenetic marks persist from mother cell to daughter cell, the compaction of the genome during mitosis requires many epigenetic marks to be temporarily erased. Every mitosis, the epigenome is bookmarked, erased, and reestablished as the cells reenter G1. Cells need a mechanism allowing them to reestablish cell specific chromatin marks after they have been erased during mitosis. The mechanism cells use for "remembering" chromatin architecture is mitotic bookmarking, whereby specific molecules or proteins are found on promoters of genes that enable memory of the chromatin state before mitosis. By definition, bookmarks must be deposited in association with active genes before or at the beginning of mitosis, persist throughout mitosis, and transmit gene expression memory to the cell after mitosis (Figure 4). These mitotic bookmarks involve multiple chromatin changes, including histone modifications and histone variants. Transcription factors also make up a large number of mitotic bookmarks. Many of those transcription factors bookmark specific subsets of genes. One example of a highly selective mitotic bookmark is Brd4, which is found only on the transcription start sites of genes that are expressed at the end of mitosis and beginning of G1 [106]. Mitotic bookmarks can also regulate a specific biological process, as in the case of GATA1, which occupies locations on key hematopoietic genes during mitosis [107]. Ubiquitin has also been found to play a role as a mitotic bookmark, but while many mitotic bookmarks are specific for certain genes or pathways, the mitotic bookmark ubiquitination appears to be generally acting at genes with high transcriptional activity.
This novel role of ubiquitin was first identified through a variant of ChIP-seq experiments that found ubiquitin present on certain sites during mitosis that were previously not described [108]. Those experiments showed that during interphase, ubiquitin was present on the chromatin of transcribed regions of transcriptionally active genes, consistent with the known function of PAF1C. The novel observation was that ubiquitinated chromatin associated proteins were bound to promoters during mitosis, contrasted to interphase, when ubiquitin localized to the promoter was absent. The fundamental difference between interphase and mitosis was a shift of the ubiquitin detected near promoters of the same genes that were previously ubiquitinated on their transcribed regions. For example [109], the GAPDH gene is heavily ubiquitinated over the gene body during G1 (Figure 4A, indicated in blue), while during mitosis that ubiquitination over the gene body is absent but ubiquitination is detected over the promoter (Figure 4A, gold) [108]. The ubiquitinated promoter sites were consistently ubiquitinated at 150 bp upstream of the transcription start sites, suggesting a specific function relating each of these promoters. The fact that this ubiquitin bookmark was identified on promoters of active genes further supported the conclusion that this novel finding of ubiquitin in mitosis was playing a role as a bookmark, not just an incidental observation. This conclusion was also supported by the fact that the promoter-associated, mitotic ubiquitin was found on the same genes as those with PAF1C-associated transcriptional H2B-ubiquitin. The association between these two forms of ubiquitination suggests that the ubiquitin bookmark is dependent upon transcription, as is PAF1C-associated H2B-ubiquitin. However, the mechanisms underlying creation of these ubiquitination marks is different, suggesting that there is no direct relationship between these two transcription-associated marks.
The presence of the ubiquitin bookmark is dependent upon the E3 ligases RING1A and BMI-1, which are both parts of the polycomb repressive complex discussed previously [109]. Surprisingly, RING1B, the primary E3 ligase involved in the PRC1 primary function has no role in bookmark ubiquitination, suggesting that the role of RING1A and BMI-1 in creating the ubiquitin bookmark is independent of their role in the PRC1 complex. However, this remains untested and what factors interact with RING1A and BMI-1 when they are involved in bookmark ubiquitination is an open question. With the discovery of the ligases responsible for the ubiquitin bookmark, it was possible to test experimentally how the process is regulated. RING1A depletion caused a decrease in phosphorylated RNAPII at promoters; the phosphorylated RNAPII was used as a surrogate for transcriptional activity, indicating that the ubiquitin bookmark was necessary for the proper transcription of the bookmarked genes, one of the criteria for mitotic bookmark. So far, what is known about the ubiquitin bookmark is that is present during mitosis, responds to changes in gene expression, and that it impacts transcription. These are the basic requirements to satisfy the definition of a mitotic bookmark.  [108]. The  Beyond the basic outline of a bookmark, there are relatively few facts known about the ubiquitin bookmark. Its localization to promoters of active genes is notable, but only a subset of active genes is found to be bookmarked, suggesting the input of more factors than active transcription and ubiquitination via RNF20/40 with PAF1C. What these factors are, even what kind of signal they are, is still unknown. Only a little is also known about the mechanism by which bookmark ubiquitination affects transcription. H3K4me3 is a histone modification known to associate with sites of active transcription. H3K4me3 has been observed decreasing when RING1A has been depleted, suggesting that the lack of ubiquitin bookmarking has caused a decrease in transcription by decreasing H3K4me3 as a signal for transcription [109]. If this phenomenon is unique to H3K4me3 or if it is common to other histone modifications correlating to active transcription is unknown. Further studies to determine this mechanism will inform the importance of this bookmark and how broadly it affects cellular function and differentiation.
Currently the exact composition of the ubiquitin bookmark is undetermined. The ubiquitin bookmark must have a substrate protein that is directed to the sites identified to have ubiquitin bookmarks, and serves as the connection to chromatin. The signal has been detected via affinity-tagged ubiquitin molecules that do not discriminate between mono-or polyubiquitin, nor between the different lysine residues with which the polyubiquitin chain could be constructed. Ubiquitin is the only known component of the bookmark, but there must be other components yet to be identified. Given that most chromatin-associated proteins dissociate from the genome during mitosis, there are fewer candidates for the substrate than would be in interphase, though the possibility exists that a protein previously unknown to remain during mitosis exhibits that ability as part of the ubiquitin bookmark.
Another aspect for expanding our understanding of the ubiquitin bookmark is expanding the finding of the ubiquitin bookmark to other cell lines. Thus far, all the prior work done on the ubiquitin bookmark has been done in HeLa cells, a common model system. As the ubiquitin bookmark has not been demonstrated in any other cell lines, nor in tissue samples, questions of the ubiquity of the bookmark are raised. It is formally possible that the ubiquitin bookmark is unique to HeLa cells or just cancerous cell lines and is not apparent in tissues in organisms. Obviously, the role of the ubiquitin bookmark is only relevant in an actively dividing cell, although the majority of cells in living tissues are postmitotic. Detecting the presence or lack thereof of the ubiquitin bookmark in other cell lines should be one of the most pressing directions of current research. The significance of the ubiquitin bookmark as a relatively new and poorly understood process suggests a new field in epigenetics, or at least a significant evolution in our understanding of mitotic bookmarks as primarily transcription factors that control limited selections of genes to a much larger, potentially genome-wide scale.

Concluding remarks
Chromatin is dynamically modified as genes are silenced, as genes are expressed, as DNA damage is repaired, and as the genome is prepared for cell division. In this review, we highlighted the diverse roles of ubiquitin in each process. Understanding the complexity of the ubiquitin system is a monumental task of which the scientific community is only scratching the surface. Four important processes were reviewed here, and these processes are paramount to proper cellular functions and deregulation is generally implicated in cancers.