New roles for elongation factors in RNA polymerase II ubiquitylation and degradation

Abstract

RNA polymerase II (RNAPII) encounters numerous impediments on its way to completing mRNA synthesis across a gene. Paused and arrested RNAPII are reactivated or rescued by elongation factors that travel with polymerase as it transcribes DNA. However, when RNAPII fails to resume transcription, such as when it encounters an unrepairable bulky DNA lesion, it is removed by the targeting of its largest subunit, Rpb1, for degradation by the ubiquitin-proteasome system (UPS). We are starting to understand this process better and how the UPS marks Rbp1 for degradation. This review will focus on the latest developments and describe new functions for elongation factors that were once thought to only promote elongation in unstressed conditions in the removal and degradation of RNAPII. I propose that in addition to changes in RNAPII structure, the composition and modification of elongation factors in the elongation complex determine whether to rescue or degrade RNAPII.

RNA polymerase II (RNAPII) undergoes cycles of productive elongation, pausing, arrest, and reactivation during transcription. An extensive collection of elongation factors (EFs) and chromatin-modifying enzymes travel with RNAPII facilitating its passage across genes[1–3]. Some pauses and arrests are quickly resolved by EFs, however prolonged or terminal arrest prevents transcription and cell death. Examples of conditions that lead to severe RNAPII arrest include exposure to transcription inhibitors, physical barriers, replication fork conflicts and DNA damage [4–9]. Arrested RNAPII can be rescued by elongation and DNA repair factors or as a “last resort” or be removed by the ubiquitin-proteasome system[10]. The pathways to degrade RNAPII are largely conserved among eukaryotes, but some differences exist. This review will provide an overview of our understanding of the processes in yeast and human cells and describe recent advances over the last 4 years.

RNAPII destruction in eukaryotes

RNAPII that encounters bulky DNA lesions, such as that induced by ultraviolet radiation (UV), becomes arrested until repair occurs. Years of work in the yeast model system revealed that the damage is repaired by transcription-coupled repair (TCR), which depends on Rad26 that exposes the lesion to repair factors (for review see [11]). TCR rapidly repairs the transcribed strand. Global genomic repair (GGR) is an alternative pathway that corrects the non-transcribed stand and unexpressed regions of the genome[12]. If the TCR repair pathway fails, RNAPII is removed by the ubiquitylation and proteasome degradation of its largest subunit, Rpb1[5, 8, 10].

What distinguishes RNAPII arrested over lesions versus stochastic pausing is still not clear. Yeast RNAPII terminally arrested over a lesion is marked for degradation by the HECT domain E3 ligase Rsp5 (human Nedd4) and the E2 ubiquitin-conjugating enzyme Ubc4/5 to build lysine-63-linked ubiquitin chains onto Rpb1[13, 14]. The recruitment of Rsp5 to RNAPII is through the carboxyl-terminal domain (CTD) of Rpb1[14, 15] and, as described below in detail, by the Ccr4-Not complex[16]. This K63 chain is trimmed and remodeled by the deubiquitinase Ubp2[14]. The monoubiquitylated Rpb1 then recruits Degradation Factor 1, Def1, to the elongation complex (EC)[17]. Def1 binds Rpb1 and then recruits a second E3 ubiquitin ligase, the Elongin-Cullin complex, containing Elc1-Ela1-Cul3, by interacting with a ubiquitin-like domain in the Ela1 subunit[17]. Elc1-Ela1-Cul3 then adds lysine-48 linked ubiquitin chains onto K1246 of Rpb1[18]. The ubiquitylated Rpb1 is then extracted from DNA by a Cdc48-containing complex and degraded by the proteasome[19, 20].

Many components of the Rpb1 ubiquitylation machinery in yeast are conserved through to humans, with identifiable homologs in each organism, but there are differences in the mechanism of action. One significant difference is that Rpb1 degradation is not necessary for TCR in yeast, but it is in mammals[6, 8]. Rpb1 ubiquitylation regulates the recruitment of the TCR machinery in human cells. The major players in resolving RNAPII blocking lesions are CSB, CSA, UVSSA and TFIIH[6, 21–26]. Cockayne syndrome protein B (CSB) is a member of the SWI/SNF family of ATP-dependent translocases that binds to upstream DNA of arrested RNAPII and is the homolog of yeast Rad26. CSB then recruits a CSA-containing E3 ubiquitin ligase complex (also known as CRL4^CSA) that ubiquitylates Rpb1 and other proteins in arrested RNAPII complexes, such as CSB. Recruitment of USP7 protects CSB from ubiquitin dependent degradation [27]. Recent Cryo-EM structures revealed that CSB displaces Spt4/5, bends DNA and pushes polymerase forward, in a similar manner as its yeast homolog Rad26 [28, 29]. If CSB fails to push polymerase past the lesion, CSA ubiquitylates K1268 in human polymerase[4, 11, 29–31]. K1268 ubiquitylation promotes the recruitment of the UV-stimulated scaffold protein A (UVSSA), which in turn recruits TFIIH to initiate TCR[24]. After repair, CSB is required to restart polymerase, at least in part by maintaining a sufficient pool of RNAPII [18, 32]

If DNA is not repaired or the arrested polymerase is not re-activated, human Rpb1 is marked for degradation as in yeast. It begins with the Rsp5 homolog NEDD4 monoubiquitylating Rpb1, followed by polyubiquitylation by ElonginABC-CUL5 (ElonginABC-CUL5). The ubiquitin ligases used in yeast are defined, but humans have multiple E3 enzymes that have been proposed to carry out this function [26]. For example, the CLR4^CSA complex, and a CUL3 complex, have been implicated in this process. In yeast, Def1 bridged monouniquitinated Rpb1 and the elongin E3 complex. Until recently, a Def1 homolog has remained elusive[33]. Once Rpb1 is ubiquitylated, VCP/p97 (CDC48) extracts RNAPII from chromatin, and the proteasome degrades Rpb1. It has been hypothesized that Rad26/CSB plays a role in distinguishing DNA damage arrested from stalled or paused RNAPII [5]. However, Rpb1 is ubiquitylated and destroyed under transcriptional stress conditions that do not induce bulky lesions and is CSB/Rad26-independent. I will propose below that the composition of the ECs may also signal a terminally arrested polymerase from one that can resume transcription.

Human UBAP2/UBAP2L is the missing link in RNA polymerase II degradation

Degradation factor 1 (Def1) was identified as a protein co-purifying with Rad26, and it plays a critical role in the ubiquitylation and degradation of RNAPII, as well as other cellular functions[17, 34–37]. As described above, it acts as an adaptor to Rpb1 by binding to monoubiquitylated polymerase and recruits the Elongin-Cullin E3 ubiquitin ligase, which in turn polyubiquitylates Rpb1[17]. Def1 has a ubiquitin binding domain (CUE) that recognizes a ubiquitin-like domain(UBL) in the Ela1 subunit of the Elongin-Cullin complex[17]. In addition to the CUE domain, Def1 has a long polyglutamine containing C-terminus that sequesters it in the cytoplasm. DNA damage leads to the proteolytic processing of Def1 and its movement into the nucleus[17]. Protein homology searches have failed to identify a human Def1 homolog, which led to the speculation that Def1 is a fungal-specific protein.

The Svejstrup lab performed a thorough “data mining” exercise using known features of yeast Def1 to narrow down the functional homolog in humans[33]. They identified human proteins with a ubiquitin-binding domain that associated with both RNAPII and CSB (Rad26), and the candidates were further narrowed down by expecting that the protein is a substrate for phosphorylation by ATM/ATR kinase. This strategy identified UBAP2 and UBAP2L as potential homologs, which appear to act redundantly. Cells depleted of both UBAP2 and UBAP2L, from here on referred to as the double knock-out cells (DKO) cells, displayed reduced Rpb1 ubiquitylation and turnover and reduced recruitment of Elongin-Cul5 to DNA damage sites in cells [33]. Furthermore, expressing UBAP2 in yeast cells suppressed the slowed-growth phenotype of a def1Δ mutation, also providing evidence that it is the functional homolog. However, they did not examine if UBAP2 can rescue the DNA damage sensitivity phenotype or the Rpb1 ubiquitylation defects of the def1Δ mutant. Thus, it remains to be seen if the suppression is limited to correcting the loss of other functions of Def1 (for review see [34]).

An interesting side story emerged from this study that warns about using KO cells. The authors found significant adaption of the DKO cells over time that reduced the cell’s dependence on UBAP2/UBAP2L for DNA damage responses. The DKO cells had a very modest Rpb1 ubiquitylation defect and an unnoticeable Rpb1 degradation defect. However, when they used a Dox-inducible version of UBAP2 that maintained expression until doxycycline withdrawal, they observed a dramatic reduction in Rpb1 ubiquitylation and impaired Rpb1 degradation. This result suggested that the DKO cells undergone adaption that greatly underestimated the role of UBAP2/2L in Rpb1 ubiquitylation and degradation. One must wonder if some of the differences reported in the literature about the relative contributions of different E3 ligases on Rpb1 ubiquitylation have been obscured by cellular adaption during gene KO construction or by long knockdown experimental strategies.

DSIF (Spt4/5) protects RNA polymerase II from degradation

DSIF (Spt4/5) is among the most conserved elongation factors and is well known for its transcription-stimulating activities [38–40]. It does so by forming a tight complex with RNAPII, interacting with the emerging transcript, the non-template strand in the transcription bubble, and upstream DNA behind RNAPII[41–43]. It is also critical for setting up promoter-proximal pausing by recruiting the negative elongation factor NELF to polymerase[1, 3]. It was unclear why promoter-proximal paused polymerase is not targeted for ubiquitin-mediated degradation since it is well-documented that arrested RNAPII can be degraded without exogenous DNA damage[8, 10]. Models to explain this phenomenon invoked the “recycling” of paused polymerases or that other factors masked the paused polymerase from the degradation machinery. Two recent papers from the Shilatifard and Chen laboratories shed light on this mystery[44, 45].

Using a combination of rapid protein destruction (AID-degron and dTAG) and shRNA depletion, both studies report that depleting Spt5 led to the rapid degradation of Rpb1. These results led to the hypothesis that Spt5 protects or hides RNAPII complexes from degradation. Curiously, this function was unique to the Spt5 subunit of DSIF, because depleting Spt4 did not induce Rpb1 degradation [44]. We have found in biochemical assays that Spt5 denuded of Spt4 can associate tightly with yeast RNAPII elongation complexes (Crickard and Reese, unpublished). Perhaps our result explains why Spt4 is dispensable for protecting Rpb1 from the degradation machinery. Furthermore, Spt5 protection was prevalent at the beginning of genes, over promoter-proximal regions, but much less so in gene bodies. Although the degradation of promoter-proximal polymerase was induced by Spt5 depletion, depleting other pausing regulators, such as PAF1, BRD4, and NELF did not cause degradation. Moreover, the stabilization of Pol II by Spt5 does not require phosphorylation at S666 (KOW domain), a modification necessary for pause release or phosphorylation of its C-terminal repeat region (CTR) [45]. These results suggest that the protective function of Spt5 is not necessarily tied to its role in pausing. The lack of a strict requirement on pausing in human cells is consistent with the observation yeast Spt5 protects RNAPII from degradation. Stripping yeast Spt5 from polymerase using the anchor-away method initiated Rpb1 degradation[44]. Yeast lacks NELF and does not have promoter-proximal pausing.

It was known that promoter-bound RNAPII was subject to UV-induced degradation, but this was independent of the canonical CSB-CRL4^CSA pathway that operates on elongating RNAPII[46]. Interestingly, the Shilatifard lab provided convincing evidence that the degradation of Rpb1 caused by the removal of Spt5 is independent of the NEDD4 HECT domain and CUL5 RING domain ligases used in the DNA damage-induced removal of polymerase, but instead relied on CUL3[44]. Thus, a different set of factors are required for the ubiquitylation of Rpb1 caused by the depletion of Spt5. However, the removal of both types of “damaged” RNAPII complexes was dependent on the VCP/p97 separase and the proteasome, suggesting that only the ubiquitylation machinery differs. Recently, it was shown that TCR is largely independent of Rad26 and Spt4/5 exchange near the +1 nucleosome, but the Spt4/5 suppression of Rad26 is prevalent downstream in gene bodies[47]. Collectively, these observations suggest that the Rpb1 degradation pathways are not redundant in the classic sense but are rather highly specialized in resolving RNAPII arrested by different conditions and that distinct pathways are utilized across a gene.

A novel function for the elongation factor ELOF1/ELF1 in Rpb1 ubiquitylation

ELF1 was first identified in a synthetic lethality screen with yeast elongation factors [48]. It has subsequently been verified to be an elongation factor that travels with RNA polymerase II and functions with Spt4/5 in promoting elongation through transcription blocks[42, 49–54]. Cryo-EM structures of Elf1-containing ECs indicate that it binds across the central cleft of RNAPII, facing downstream DNA and towards the nucleosome, where it and Spt4/5 promote polymerase progression into the nucleosome [49].

ELOF1 is the human homolog of yeast Elf1. Its roles in DNA damage responses and transcription-coupled repair were identified in three different CRISPR screens [55–57]. These works also confirmed ELOF1’s role in transcription elongation and also demonstrated synthetic lethality between ELOF1 KO and KO of SUPT4H1 (DSIF),TCEA1 (TFIIS), SUPT6H (SPT6) and the Paf1C complex subunits CTR9 and LEO1 [55, 57], establishing it as the functional homolog of yeast Elf1. The more exciting outcome of these works was that they identified ELOF1 as a novel TCR factor and that it helps resolve transcription-replication conflicts. Detailed molecular analysis revealed that ELOF1 is required to recruit CRL4^CSA E3 ligase to ECs, which leads to the ubiquitylation of K1268 on Rpb1. Rpb1 ubiquitylation causes the recruitment of UVSSA and TFIIH and, ultimately, DNA repair. Using the Komagataella pastoris EC structure containing Elf1 as a model, the authors proposed that ELOF1 positions CRL4^CSA near the ubiquitylation site on Rpb1. Furthermore, ELOF1-depleted cells displayed greatly reduced restart of transcription following DNA damage. Mutants of ELOF1 that prevented its binding to RNAPII could not rescue the KO phenotypes, firmly establishing its function to regulating RNAPII. ELOF1-depleted cells are UV- and DNA damage sensitive[55–57], but yeast elf1Δ cells are not[56]. Yeast cells defective for TCR are mildly sensitive to UV, such as rad26Δ cells, but display synthetically enhanced UV sensitivity with mutants in the global genome repair pathway, GG-NER[37, 58]. Consistent with Elf1’s involvement in TCR, the authors found that deleting ELF1 sensitized cells defective in GG-NER to UV [56]. It is not known if yeast ELF1 is required for Rpb1 ubiquitylation and degradation, and this should be examined soon.

Paf1C regulates RNAPII degradation and the rescue of transcription after DNA damage

The Paf1C complex orchestrates multiple transcriptional processes, including promoting elongation, regulating histone modifications, 3’ end processing, and promoter-proximal pause release[59]. Paf1C recruitment to the elongation complex depends on the phosphorylation of the carboxy-terminal repeat (CTR) of the Spt5 subunit of DSIF (Spt4/5 complex), and the CTD of Rpb1 [3, 60–63]. About 14 years ago, Aguilera and colleagues used epistasis analysis and lesion repair assays to make a compelling case that Paf1C is required for DNA damage resistance and transcription-coupled DNA repair [64]. This work was shortly followed by another paper suggesting that nucleotide excision repair, both TCR and GGR, required Paf1C [65]. These papers made the connection between Paf1C and DNA repair, but the molecular mechanism or if it directly regulates repair was unknown. It is a concern that reduced repair could be caused by altered transcription in Paf1C mutants. Two recent developments, one in yeast and the other in human cells, have provided important mechanistic insights into the role of Paf1c in DNA repair and the degradation of Rpb1. These molecular insights may explain the complex genetic interactions between Spt5 and Rad26 in the TCR repair pathway.

Using a combination of genetic analysis and elegant reconstitution biochemistry, the Long laboratory identified a critical role for Paf1C subunits in the DNA damage-induced degradation of Rpb1[66]. Deleting PAF1, LEO1, CTR9 LEO1, but not RTF1, blocked Rpb1 degradation in cells. Interestingly, deleting RTF1 actually accelerated Rpb1 degradation, which will be addressed below. They found that Paf1C, specifically the Paf1-Leo1 dimer, stimulated Elongin-Cullin-dependent ubiquitylation of Rpb1 in vitro, which was explained by the binding to the Ela1 subunit of the E3 ligase complex to the Paf1-Leo1 complex. These observations provided a molecular explanation for how Paf1C mediates Rpb1 destruction. As noted above, deleting the Rtf1 subunit accelerated Rpb1 degradation, unlike the other mutants in the Paf1C complex. Judith Jaehning’s lab reported many years ago that deleting RTF1 suppressed the hydroxyurea and stress sensitivity phenotypes of a paf1Δ mutant, further hinting at opposing effects of deleting RTF1 versus other subunits of the complex[67]. Rtf1 binds to the phosphorylated form of the CTR in Spt5 through its PLUS-3 domain [60]. The Long lab demonstrated that accelerated destruction of Rpb1 can also be caused by inactivating the Spt5-CTR kinase Bur2, mutating S/T residues in the CTR or deleting the PLUS3 domain of Rtf1; thus, suggesting that the phosphorylation status of the CTR of Spt5 is an important determinant in RNAPII. The genetic analysis clearly implicated the Rtf1-Spt5 CTR interaction in the pathway. Accelerated destruction of Rpb1 in response to UV was first reported in a rad26Δ mutant [37]; thus, altering the association of Paf1C with Spt5 phenocopied the rad26Δ mutation. This argues that weakening the association of Spt5 with Paf1C is the equivalent to inhibiting Rad26 activity. Structural studies of arrested elongation complexes with Spt4/5, Paf1C, and Rad26 suggested a model where the binding of Spt4/5 and Rad26 are mutually exclusive[28, 29]. The Long group proposed a model where the phosphorylation of the CTR and reduced association of Rtf1 “orchestrates” the exchange of Spt4/5 and Rad26 to regulate the balance between Rpb1 destruction and TCR. However, their results also would be consistent with the possibility that disrupting CTR phosphorylation recruits an inhibitor of Rad26 to the stalled elongation complex or that another component of the ubiquitin ligase cascade binds preferentially to the unphosphorylated CTR. Finally, deleting RFT1 could accelerate Rpb1 degradation by exposing the CTD of Rpb1 to de-phosphorylation by Ssu72. Removal of Ser5 phosphorylation is required for Rpb1 ubiquitylation and destruction[68].

Rad26 is the yeast homolog of CSB, Cockayne-Syndrome group B, ATP-dependent helicase. The connection between Paf1C and Rad26 extends to human cells, with some different twists. The Luijsterburg lab found that Paf1C binds to CSB/Rad26 in a UV-damage-dependent manner[32]. The association was dependent on the Leo1 subunit of the complex, like the yeast Paf1c-Rad26 interaction. They found that while the binding of Paf1C to RNAPII depends on p-TEFb-dependent phosphorylation of RNAPII CTD, the Paf1C -CSB interaction did not require pTEF-B activity. This suggests that phosphorylation of the CTD of Rpb1, Spt5 or NELF does not mediate Paf1C binding to CSB. Interestingly, AID-degron induced depletion of PAF1 did not block Rpb1 ubiquitylation or impair transcription-coupled DNA repair. Thus, Paf1C performs a different function in humans than in yeast. However, the release of polymerase from arrest was impaired in PAF1-depleted cells. Thus, human Paf1C was essential for the recovery of transcription following UV-induced arrest, a novel function of the Paf1C complex. It remains to be seen how Paf1C restores transcription after arrest.

Both studies described above defined novel functions for the Paf1C complex in the DNA damage response and suggest an important role for the Paf1C-CSB/Rad26 interaction in allowing cells to cope with transcription-blocking lesions. However, there are some unresolved questions. Firstly, there were differences in the effects of PAF1 depletion in the two systems. Both TCR and Rpb1 degradation requires yeast Paf1C. In contrast, human Paf1C is dispensable for these functions. What accounts for the differences between the two organisms? Second, how does Paf1C restart polymerase? Paf1C promotes elongation in vitro by allosterically effecting RNAPII translocation[43] and it may activate stalled polymerase after DNA repair by directly affecting polymerase activity. In addition, it could contribute to the process indirectly by regulating histone modifications. Two Paf1C-dependent modifications, H3K79 and H3K4 methylation, have been implicated in transcription restart following DNA damage in human cells and in DNA repair in yeast [65, 69, 70].

Yeast Ccr4-Not promotes Rpb1 ubiquitylation

The Ccr4–Not complex has been identified in all eukaryotes, but its composition varies among species [71, 72]. First identified as a regulator of promoter utilization by TFIID, it is now appreciated for its many roles in gene expression control, including RNA polymerase II elongation, mRNA transport, mRNA decay, protein ubiquitylation, and translation quality control [71–77]. Biochemical and genetic studies in yeast revealed that the complex could be separated into two functional “modules” linked via the Not1 protein: the Ccr4–Caf1 deadenylase and the Not-module [71]. One subunit of the “Not” group is Not4/Mot2 (cNOT4 in human). Not4 is tightly associated with the yeast complex, but its status in metazoan Ccr4-Not is less clear because it fails to copurify with other subunits [71, 78]. Not4 is a RING domain-containing protein that utilizes the E2 enzymes Ubc4/5 to ubiquitylate a few known proteins in the cytoplasm and nucleus [73, 75, 79–81].

Several years ago, we demonstrated that Ccr4-Not associates with elongating RNAPII via the Rpb4/7 module and the transcript, where it promotes transcription of reconstituted elongation complexes [79, 81–84]. Ccr4-Not, the Not5 subunit specifically, was implicated in transcription-coupled repair [64]. Cells deleted of Ccr4-Not subunit genes are sensitive to mutagens, show genetic interactions with DNA repair factors and had reduced TCR capacity [16, 64, 72, 85, 86]. It was not known how it regulates DNA repair or if it does so directly or indirectly. Since Ccr4-Not travels with RNAPII, contains an E3-ligase subunit, and is required for repair, we hypothesized it regulated RNAPII destruction after DNA damage by ubiquitylating Rpb1. While the overall hypothesis was correct, the mechanism differed from what we expected.

Deleting the NOT4 gene or mutating the RING domain greatly impaired DNA damage-induced Rpb1 ubiquitylation and degradation.[16]. In contrast, deleting the deadenylase module subunits Ccr4 or Caf1 did not affect Rpb1 degradation, indicating that the RNAPII destruction phenotype was not a consequence of altered mRNA turnover. A screen of viable Ccr4-Not subunit mutants revealed that deleting NOT2 and NOT5 led to the same phenotype, but these mutations also led to the depletion of Not4 protein from cells. The Rpb1 degrading activity was dependent on the RING domain. Since NOT4 RING mutant was incorporated in the Ccr4-Not complex, the preponderance of evidence indicated that Rpb1 destruction required Ccr4-Not’s ubiquitylation activity.

Although Not4 can ubiquitylate other proteins, biochemical reconstitution assays indicated that it does not directly ubiquitylate Rpb1. The HECT domain E3 ligase Rsp5 and the Elogin-cullin E3 ligases act sequentially to ubiquitylate Rpb1[10, 14]. It was surprising that these two ubiquitin ligases could not modify Rpb1 in the not4Δ mutant unless Ccr4-Not acts upstream of the known ubiquitylation pathway. Consistent with this idea, Ccr4-Not formed a ternary complex with Rsp5 and RNAPII ECs and enhanced Rsp5-dependent ubiquitylation of Rpb1. Biochemical analysis of a Ccr4-Not complex lacking the RING domain indicated that another function of Not4 is to recruit the Ubc4/5 E2 enzyme to the RNAPII-Rsp5 complex. Thus, Ccr4-Not sits atop the Rpb1 destruction cascade where it recruits the ubiquitylation machinery. Therefore, Ccr4-Not may help identify RNAPII that needs removal by the UPS.

We proposed that Ccr4-Not acts as a “hub” or coordinator of the cellular response to arrested polymerase [16]. Does Ccr4-Not help restart arrested polymerase after repair and if rescue fails help initiate Rpb1 destruction? It could restart arrested polymerases by recruiting TFIIS to resolve backtracking or use its anti-arrest activity [83, 84]. Does Ccr4-Not ubiquitylate other components of the EC? Mechanistically, Ccr4-Not functions similarly to Paf1c. Yeast Paf1c recruits the Elongin-Cullin complex to RNAPII and human Paf1c reactivates arrested polymerase (see above). Do Ccr4-Not and Paf1c work together? Ccr4-Not can recruit Rsp5 to monoubiquitylate Rpb1 and then Paf1C can assist Def1 in recruiting the Elogin-cullin complex. Does Ccr4-Not modify Paf1C? Our initial attempts to ubiquitylate purified yeast Paf1C with Ccr4-Not were unsuccessful (Reese, unpublished). However, we have not thoroughly tested this model and it would be worth attempting this on reconstituted ECs. Finally, is this function of Ccr4-Not conserved in metazoans? Ccr4-Not is conserved, but the binding of cNOT4 to the human complex is not certain, or it does so through another mechanism: Yeast Not4 binds to Not1, whereas cNOT4 binds to cNOT9/CAF40 in human Ccr4-Not and is not stably associated with the complex[71, 78].

INO80 extracts RNAPII from chromatin to resolve replication-transcription conflict

The INO80 complex regulates DNA transcription, replication, and repair[87]. INO80 resolves replication-transcription conflicts and regulates transcription by positioning and remodeling the +1 nucleosome [87–92]. One function relevant to this review is its ability to extract RNAPII from chromatin during DNA damage in S-phase [93, 94].

Two papers revealed the role of INO80 in the extraction of RNAPII during S-phase, each showing different aspects of the process. The first paper, described the co-purification of INO80 with the CDC48 complex[93], a factor known to extract ubiquitylated Rpb1 from chromatin [20], and the other report, identified synthetic genetic interactions between subunits of INO80 and an S-phase-specific allele of MEC1, mec1-100 [94]. The ubiquitylated form of Rpb1 accumulates on chromatin in ino80Δ and in an ATPase mutant (K73AA) in cells treated with hydroxyurea (HU) and methyl methanesulfonate (MMS) in S-phase. The failure to remove ubiquitylated Rpb1 also impaired Rpb1 degradation [93]. HU-induced destruction of Rpb1 was not observed in G1 arrested cells, indicating that replication is required and implicating replication-transcription conflicts as the trigger of Rpb1 ubiquitylation[94]. HU-induced Rpb1 destruction was not observed in asynchronous cells either, likely due to only a minority of cells undergoing replication. Significantly, UV-induced degradation of Rpb1 in S-phase was independent of INO80, and epistasis analysis between mutants of INO80 subunits and the rad26Δ allele suggests specificity for HU and MMS damage, two agents that cause replication fork collapse in S-phase. Consistent with this, INO80 is dispensable for the repair of UV lesions [88]. Collectively, the data suggest that INO80 is required to remove and degrade Rpb1 during replication-transcription conflict.

CDC48 extracts RNAPII from chromatin throughout all stages of the cell cycle, but the requirement for INO80 was specific to S-phase. INO80 and the CDC48 complex (and its adaptors) bind together and form a ternary complex with RNAPII. The authors concluded that INO80 and CDC48 function as a complex to remove RNAPII from chromatin. These reports identified a novel pathway in RNAPII degradation, but also raised some interesting questions. What are the relative contributions of the two complexes in Rpb1 extraction and degradation? A plausible model is that INO80 exposes Ub-Rpb1 “entrapped” in chromatin using its nucleosome-remodeling activity, which the CDC48 complex can then act upon to promote polymerase extraction from DNA. INO80 could act on chromatin at the nucleosomal- or at a higher order chromatin-level of organization. It is not clear why the INO80-CDC48 complex is dispensable for Rpb1 degradation outside of S-phase or during UV-induced damage in S-phase. One untested possibility is that the binding of CDC48 to INO80 occurs predominantly in S-phase and is triggered by the activation of the replication checkpoint. Phosphoproteomic analysis indicates that INO80 subunits are phosphorylated in a Mec1-dependent manner during replicative stress, and this could regulate the INO80-CDC48 interaction[94].

Updated models for Rpb1 ubiquitylation in repair and degradation.

Integrating prior models with the new developments described above leads to a revised model for Rpb1 ubiquitylation in DNA repair (Figure 1). The arrest of RNAPII over a lesion causes the recruitment of CSB and its exchange with Spt4/5. What initiates the exchange is unclear, but perhaps CSB translocating along DNA displaces Spt4/5 by breaking Spt5-nucleic acid contacts. The CLA-containing cullin complex (cullin, Figure 1) is recruited, and the pulling on DNA by CSB, together with ELOF1, positions CLA over K1268, initiating ubiquitylation. K1268-Ub recruits UVSSA-USP7, and USP7 protects CSB from removal. CLA ubiquitylates UVSSA at K414, recruiting TFIIH to initiate downstream repair events. Following the recruitment of TFIIH, CSB is polyubiquitylated by CSA. This may be initiated by the release of UVSSA-USP7 or a change in the elongation complex after repair that repositions CSA towards the back side of polymerase and CSB. The destruction of CSB opens the upstream side of polymerase for the reengagement of Spt4/5 to begin the reactivation process. Finally, Spt4/5, TFIIS and Paf1C restore RNAPII to the elongation competent form.

Figure 1. A revised model for DNA damage induced Rpb1 ubiquitylation in DNA repair and transcription reactivation in human cells.

This figure was constructed using Biorender software (agreement# FX25FVZYX5).

New developments in understanding RNAPII destruction have been gleaned from recent studies in the yeast model system (Figure 2). Placing Ccr4-Not atop the cascade identified an early factor in marking Rpb1 for ubiquitylation. Ccr4-Not recruits the first ubiquitin ligase, Rsp5, and the E2 Ubc4/5 to the arrested EC. The E2 is then handed off to Rsp5, which adds K63-linked chains to Rpb1. Ubp2 trims the K63 chain down to the monoubiquitylated form (or shorter chains) and Def1 is recruited through contacts with RNAPII and Ccr4-Not. Next, the Ela1-Cul3-Elc1 cullin complex is brought in by Def1 and Paf1c, which extends poly K48 ubiquitin chains. The CDC48 separase complex, containing ubiquitin-binding adaptor proteins, extracts Rpb1 from chromatin for proteasome degradation. After Rpb1 removal, the GGR repair pathway repairs the damage and restores the template for subsequent rounds of transcription.

Figure 2. An updated model for RNAPII destruction in the yeast model system.

This figure was constructed using Biorender software (agreement# GW252W4MP0).

Do the composition and integrity of the elongation complex signal Rpb1 destruction?

The recruitment of CSB/Rad26 to RNAPII, which is DNA damage dependent, is a distinguishing feature of polymerase arrested by bulky lesions. The futile action of pushing RNAPII through the lesion leads to structural changes in the elongation complex (EC) and the ubiquitylation of Rpb1. But what about RNAPII arrested by other types of transcriptional stress or those in the paused state? The cell must discriminate temporarily paused and salvageable polymerases from terminally arrested forms needing destruction. The identification of new roles for elongation factors (EFs) in the re-activation of RNAPII after repair and in its degradation by the UPS should this fail, suggests that the composition and integrity of the EC may trigger the recruitment of ubiquitin ligases and Rpb1 destruction. The changes can include the presence or absence of a factor, the covalent modifications of the components of the EC, and/or changes in the structure of EFs. In some cases, EFs can hide the rescuable polymerase from the destruction machinery, like how Spt5 protects promoter proximal paused RNAPII from degradation. Does the collision between the trailing and the arrested RNAPII downstream initiate degradation, as proposed by others [95]? One model is that the pulling on DNA by the upstream polymerase mimics the action of RAD26/CSB, changing the structure of the downstream arrested polymerase. An interesting possibility is that this action ejects EFs such as Spt4/5 from the terminally arrested polymerase located downstream, identifying it for degradation (Figure 3). Once the leading arrested polymerase is degraded, the trailing EC is reactivated by the collection of EFs associated with RNAPII and TFIIS is recruited resolve backtracking. A non-mutually exclusive possibility is that elongation factors play the role of judge and jury on the decision to rescue or degrade polymerase and reactivate transcription.

Figure 3. A model for the roles of elongation factors in mediating RNAPII removal and reactivation of upstream polymerases.

This figure was constructed using Biorender software (agreement# FS252WEUWO).

Over the last four years, gaps in knowledge have been filled by identifying novel functions of known elongation factors in RNAPII ubiquitylation. We have learned that more factors are involved in Rpb1 ubiquitylation than once appreciated. More are likely to be discovered, but the experimental strategy utilized to do so may be important. For example, the protective role of Spt5 in preventing Rpb1 degradation was only uncovered using rapid, induced depletion systems, which are less prone to cellular adaptation. Many prior studies using Spt5 knockdown strategies did not reveal this function. Likewise, constructing double knockout UBAP2/L cells underestimated these gene’s importance in Rpb1 ubiquitylation due to changes in the cells. Finally, as described in different parts of this review, there are conserved elements in the mechanism of Rpb1 ubiquitylation between the simple eukaryote yeast and humans, but also differences. It will be critical going forward to distinguish organism-specific functions from the fundamental elements of the pathways.