RNA Polymerase II Activity Control of Gene Expression and Involvement in Disease

Abstract

Gene expression is dependent on RNA Polymerase II (Pol II) activity in eukaryotes. In addition to determining the rate of RNA synthesis for all protein coding genes, Pol II serves as a platform for the recruitment of factors and regulation of co-transcriptional events, from RNA processing to chromatin modification and remodeling. The transcriptome can be shaped by changes in Pol II kinetics affecting RNA synthesis itself or because of alterations to co-transcriptional events that are responsive to or coupled with transcription. Genetic, biochemical, and structural approaches to Pol II in model organisms have revealed critical insights into how Pol II works and the types of factors that regulate it. The complexity of Pol II regulation generally increases with organismal complexity. In this review, we describe fundamental aspects of how Pol II activity can shape gene expression, discuss recent advances in how Pol II elongation is regulated on genes, and how altered Pol II function is linked to human disease and aging.

Pol II is a large 12-subunit complex where the active site is mainly formed by the two largest subunits, Rpb1 and Rpb2. Basic functions of the Pol II active site are structurally conserved among highly-related eukaryotic multisubunit RNA polymerases (msRNAPs), including Pols I and III, which are universally present in eukaryotes, and Pols IV and V, which are specialized orthologs that have evolved in plants, and bacterial and archaeal msRNAPs (please see historical reviews^{1, 2, 3, 4, 5, 6}). This high level of conservation means that fundamental insights into catalytic mechanisms can emerge from msRNAPs in different species, with bacterial RNA polymerase studies advancing knowledge or models for eukaryotic msRNAP in many cases. However, within this core framework, there can be differences in precise aspects of eukaryotic Pol mechanism^{7, 8} and regulation is distinct between eukaryotic and bacterial RNAPs. For example, only one transcription elongation factor, Spt5/NusG is conserved across all kingdoms of life. Eukaryotic RNA polymerases have evolved to transcribe different classes of RNA while single prokaryotic RNA polymerases handle all transcription within their respective species. Eukaryotic RNA polymerases have evolved dedicated factors for each of their regulations with Pol II having the greatest number of regulatory factors due to diverse needs for regulation of gene expression. How Pol II and other eukaryotic RNA polymerases originally evolved is an important question for understanding how their mechanisms might be distinct from bacterial or archaeal relatives. Interestingly, recent studies provocatively suggest a complex origin for Pol II and potentially even eukaryotic Pols I and III as being derived from Nucleo-Cytoplasmic Large DNA Virus (NCLDV) RNAP. In one study, it was suggested that Pol II and parts of Pol I might have been captured in the eukaryotic precursor from a giant eukaryotic virus,⁹ while a more recent study proposes a giant virus origin for the eukaryotic nucleus¹⁰ (and with it, eukaryotic RNA polymerases and core aspects of eukaryotic DNA replication machinery).

Basic Pol II mechanisms

Two cycles underlie RNA synthesis by Pol II. One cycle covers the broadly defined phases of transcription: initiation, elongation, and termination, where discrete cofactors and regulatory steps differentiate the phases. The initiation phase can be additionally broken down into the steps of promoter melting, initial transcription where polymerase complexes may be sensitive to abortive initiation, a transition to a stable initial transcribing complex, and finally promoter clearance. The other cycle comprises the iterative functions of the Pol II active site during nucleotide addition. In the nucleotide addition cycle (NAC), DNA template-specified substrates are selected, phosphodiester bond formation is catalyzed, pyrophosphate is released, and finally the enzyme translocates such that Pol II proceeds to the next template position.^{11, 12} The NAC is controlled by conformational changes in Pol II and related msRNAPs that allow fast and accurate substrate selection, catalysis, and translocation.^{13, 14, 15, 16, 17, 18, 19, 20} These conformational changes can be perturbed by Pol II mutations while also potentially being impacted by elongation factors. Furthermore, off-pathway events from the NAC such as pausing, backtracking, and arrest may also occur (discussed below).

Genetic, biochemical, and structural analyses together have revealed critical features of the Pol II active site that underlie the NAC. Upon translocation where a template base is positioned in the Pol II active site, NTP substrates may be sampled for base-pairing potential. Catalysis is then promoted for based-paired substrates with appropriate chemical features by closing of a mobile and flexible loop called the trigger loop (TL). The TL is not essential for catalysis of RNA synthesis as RNAPs across all kingdoms of life show some activity in the absence of the TL or severe mutation of the TL.^{15, 19, 21, 22, 23}This activity, however, is generally 2–3 orders of magnitude below wild type. The TL can obtain a number of different conformations. The major changes during the NAC are moving from an open state that can be folded but may also be flexible or mobile (where substrates are allowed to move in and out of the active site) to a structured, closed state over the NTP (where catalysis is hypothesized to be promoted) (Figure 1).^{14, 15, 24, 25, 26, 27} Additional states, due to restraint or competition with elongation factors, have also been observed. Single molecule Förster Resonance Energy Transfer (FRET) studies on bacterial RNAP link TL closing to the presence of a cognate NTP but not mismatched NTPs.¹⁸ Furthermore, in,¹⁸ Mazumder et al. found the number of observed TL closing events precisely matched the number of template positions specifying a matched substrate, suggesting that efficient TL closing was coupled to presence of a correct NTP, and on average there was exactly one TL opening and closing cycle per catalytic event. Specifically, futile cycles of opening/closing, for example with mismatched substrates, were not observed. Critically, this single molecule FRET assay allowed the rate of TL closing to be estimated on the order of ∼20 ms at 22 °C. Of note, TL opening events were still on the order of milliseconds but generally slower than closing. With Pol II elongation rate in vivo being measured at 1–3 kb/minute (16–50 bp/s),^{28, 29, 30, 31, 32} trigger loop opening and closing along with translocation are likely to be the rate limiting events in the NAC cycle (see below). Such powerful single molecule FRET systems for Pol II have not yet been published.

Figure 1. Multiple states of the Pol II Rpb1 trigger loop support the nucleotide addition cycle.

Nucleic acid scaffold from Sce Pol II PDB 2e2h¹⁴ is shown with RNA in red, template DNA in blue, and non-template DNA in green. The TL from this structure is in the closed form (magenta) and interacting with a matched GTP substrate (orange). Other TL conformations are superimposed. TFIIS and an open TL from Sce PDB 1y1v show how the TL (yellow) can be restricted due to insertion of the elongation factor TFIIS (cyan) into the active site.²⁵ In the absence of substrate, the TL from Sce PDB 5c4j (salmon) can be observed in a folded but open conformation.²⁴ The TL is directly below another highly-conserved structural element, the Bridge Helix (BH, white). This figure was generated using Pymol.²⁷⁴

Gain-of-function (“fast”) and loss-of-function (“slow”) Pol II trigger loop (TL) mutants

Mutations in the TL alter Pol II catalytic speed, with some increasing and others decreasing it.^{13, 20, 33, 34, 35, 36, 37, 38} Our lab refers to these classes of mutation as loss-of-function (LOF) for those that decrease catalysis and gain-of-function (GOF) for those that increase catalysis. GOF mutations have been found in the TL of RNAPs from all kingdoms of life.^{16, 33, 34, 36, 39, 40} Biochemical and biophysical studies are consistent with at least one Pol II GOF mutation altering TL dynamics such that the closed state is promoted and, with it, catalysis.²⁰ Biochemical studies demonstrate that LOF and GOF mutations confer changes to elongation rate and fidelity consistent with LOF mutants slowing catalytic or other rate limiting steps (e.g. translocation) with GOF mutants increasing them.^{13, 20, 34, 36, 41, 42} As a consequence of increased catalytic activity, tested Pol II GOF mutants show decreased fidelity both in vitro^{13, 20} and in vivo.^{43, 44, 45, 46} Single molecule studies also demonstrate clearly that a GOF mutation in yeast Pol II promotes an increase in pause-free elongation velocity^{41, 42} and the effects can be modeled as effects on both catalysis and translocation.⁴² Molecular dynamics (MD) approaches have been used to explore Pol II mechanisms,^{40, 47, 48, 49, 50, 51} with simulations of Pol II mutants showing structural alterations.^{51, 52} However, MD simulations are not yet long enough to fully recapitulate TL movement, Pol II translocation, or easily predict the in vivo phenotypes of mutant enzymes of different classes.⁵²

Catalysis of Pol II transcription promoted by the trigger loop (TL)

Catalysis by Pol II and all msRNAPs has been proposed to use a two-metal mechanism where deprotonation of the 3′-OH of the preceding NTP or RNA chain facilitates nucleophilic attack on the alpha phosphate of the incoming NTP (reviewed in).^{11, 50} This mechanism is based on the proposal by Steitz and colleagues for all nucleic acid polymerases.⁵³ Protonation of the pyrophosphate leaving group then completes the reaction. The reverse reaction is pyrophosphorolysis where pyrophosphate can attack the terminal phosphodiester bond of an RNA to release an NTP. The role of the TL in catalysis was suggested to be through critical substrate contacts of the closed TL with the substrate. These contacts sense multiple substrate features such as hydrophobicity of the substrate base of a base-paired NTP substrate, hydroxyl groups of the substrate ribose moiety, and substrate phosphates.^{14, 15} Furthermore, an ultraconserved histidine residue (Rpb1 H1085 in Saccharomyces cerevisiae (Sce) Pol II) was suggested to function in acid-base catalysis through proton donation to pyrophosphate.¹⁴ Use of a basic residue in the active site for this mechanism has been proposed for major nucleic acid polymerase types.⁵⁴ Work from our group using deep mutational scanning of the Sce TL surprisingly found that Rpb1 H1085L was well tolerated for yeast growth.³⁶ This observation suggested that proton donation by His1085 may not be strongly required for TL function. The Landick lab has biochemically tested Leucine-substituted TLs and has proposed that the TL acts primarily as a positional catalyst and is not the source of proton donation.^{55, 56} One issue, noted by Unarta et al.⁵⁰ in discussion of this model, is the potential for plasticity in the Pol II catalytic mechanism. Here, Unarta et al. suggest that H1085L’s mild catalytic defect might be due to the bypass of direct protonation for this specific TL mutant or by indirect facilitation of donation of a proton from a water. In other words, active site mutations may alter the catalytic mechanism. Interestingly, in a preprint, our lab has found that H1085L perturbs the genetic landscape of other TL mutations differently than the more catalytically defective H1085Y allele.³⁷ In this experiment, SceH1085Y or H1085L mutations were combined with all possible other TL substitutions to examine genetic interactions. H1085L showed much greater epistasis (genetic interactions indicating H1085L effects were dependent on other TL residues) than H1085Y, which showed many more additive genetic interactions (genetic interactions indicating H1085Y was independent of many other TL changes). These additive genetic interactions suggest that when H1085Y was combined with other TL mutants, the individual mutants maintained their characteristics in the context of the double mutant enzymes. In contrast, H1085L’s ability to function was strongly dependent on the rest of the TL being WT. These results suggest that the Pol II active site is distinctly sensitive to H1085L substitution and that H1085L may have some unique properties.

The current highest resolution Pol II structure with a substrate bound is only at 3.0 Å.⁵⁷This structure from the Calero lab is still a major advance because it captures the Pol II TL in a closed conformation and visualizes the two catalytic Mg²⁺ ions confidently for the first time. The fully closed TL has only previously been observed in crystal structures for Sce Pol II in 2006,¹⁴ for Thermus thermophilus (Tth) RNAP in 2007,¹⁵ and at moderate resolution by Cryo-EM for Escherichia coli (Eco) RNAP in 2019.²⁷ This new work required a new crystallization condition to capture the closed TL interacting with a matched NTP substrate, and the use of a free electron laser to capture both Metal A and B Mg²⁺ ions. Intriguingly, a potential third Mg²⁺ was observed in a location similar to that for a third ion in the DNA Polymerase eta active site.^{58, 59} Note, density at this position for Pol II was observed previously by the Wang group and attributed to a water molecule.²³ The role of this ion in catalysis is not clear yet, but residues adjacent to it (Sce Rpb2 E529 and Y769) can alter Pol II activity in both directions when mutated.^{35, 60}

Calero and colleagues also describe the structure⁵⁷ of a Sce Pol II containing the hyperactive GOF Pol II bridge helix (BH) mutant Rpb1 T834P.³⁶ The BH is a structural element that bridges the two lobes of msRNAPs, over which the template DNA must pass during translocation into the active site (Figure 1).^{61, 62, 63, 64, 65} The BH also has close contacts with the TL and an additional helix that supports TL function. The BH has long been suspected as a key element for RNAP translocation and the NAC, potentially through conformational changes.^{39, 63, 64, 66, 67} Consistent with the BH being able to communicate to the active site and tune functions through conformational changes, specific helix-kinking proline substitutions can generate hyperactive enzymes both for an archaeal RNAP³⁹ and Sce Pol II.³⁶ The structure of Sce Pol II Rpb1 T834P shows a number of alterations that may explain its hyperactivity and loss of fidelity, including loss of a BH-TL interaction. Of interest, the T834P structure shows movement of the Rpb1 rim helices, widening the Pol II cleft by a few angstroms, and this might propagate any number of changes to how the TL may function. The rim helices, their connecting loop, and the Pol II Rpb9 subunit are also sites of hyperactivating GOF alleles.^{68, 69, 70, 71, 72}These residues may work in part by restraining the TL in an open conformation with restraint lost in specific mutants.⁶⁹ Small changes propagating across the Pol II structure may represent a paradigm for how elongation factors might allosterically regulate Pol II activity.

Structural states in RNAP elongation and pausing/backtracking

The sensitivity of the Pol II active site to a large number of changes that can increase catalytic rate in addition to those that reduce it suggests that the TL and other flexible domains are delicately balanced. This balance being supported by many surrounding interactions suggests that networks of interactions could converge on the TL or adjacent domains that communicate with it.^{36, 37, 38} TL movement is not the only conformational change implicated in Pol II activity, however some other conformational changes can be connected to TL function as they may control TL folding or TL folding may be incompatible with certain structural states. Off-pathway states are defined as states that do not contribute to the NAC, such as non-obligate pausing or backtracking states (reviewed for bacterial RNAP in ⁷⁴, for Pol II in ⁷⁵). Backtracking is when RNAPs reverse translocate on template DNA due to thermal motion, with unwinding of the RNA 3′ end from the template DNA allowing rewinding of template and non-template DNA in front of the enzyme as it moves backwards, with unwinding of upstream DNA.⁷⁵ This causes extrusion of the nascent RNA 3′ end away from the RNAP active site and such positioning is necessarily incompatible with the NAC. DNA sequence can strongly modulate backtracking as RNA polymerases will be more stable on stronger RNA-DNA hybrid sequences (G-C rich) relative to weaker ones (A-T rich). For example, backtracking from an A-T rich stretch to an immediately preceding G-C rich stretch would be energetically favored in the absence of continued elongation. If Pol II backtracks to an extent where it is highly unlikely to resume elongation on its own, it is said to be arrested. A transcription elongation factor called TFIIS can insert into the Pol II active site and promote cleavage of the backtracked RNA, resetting Pol II with a new 3′ RNA end in its active site.⁷⁴ The clearest example aside from backtracking of a distinct structural state for Pol II is during pausing enforced by negative elongation factors.^{76, 77} Promoter-proximal pausing in higher metazoans occurs in a zone just downstream of Pol II promoters in organisms that have the Negative Elongation Factor complex (NELF).⁷⁸ In Drosophila melanogaster cells, for some promoters the location of the pause is in a nucleosome free zone indicating that it can be factor driven, while at other promoters, the pausing zone overlaps the position of the first downstream nucleosome, suggesting a contribution of chromatin.⁷⁹ In the paused state observed in the Cryo-EM structure of Pol II bound with NELF and DSIF (DRB-sensitivity Inducing Factor comprising higher metazoan homologs of yeast Spt4 and Spt5 proteins), the active site is prevented from NTP addition because of a partially translocated (“half translocated”) state where the RNA is translocated but the DNA has not.⁷⁶ Furthermore, NELF interacts with and restrains the TL. The role of NELF in pausing is complicated because upon NELF depletion, Pol II still accumulates near promoters but now is associated with downstream nucleosomes.^{80, 81} This could be because in the absence of NELF, the Pol II elongation complexes formed may be defective or greatly slowed transiting the first downstream nucleosome.^{80, 81} We note that even acute depletion studies create outcomes where the removal of a factor results in situations that may not ever exist in its presence, so phenotypes should be interpreted cautiously when ascribing functions to the missing factor. NELF conformation and interactions with mammalian Pol II are malleable in that a newly observed “poised” state allows both TL motion and binding of the cleavage factor, TFIIS, which can rescue backtracked Pol II by promoting cleavage of the RNA 3′ end, generating a new 3′ end positioned in the active site for resumption of the NAC.⁷⁷ A number of functions for the poised state can be imagined. For example, in the transitions in or out of paused states, or by allowing input from additional factors, such as action of TFIIS to rectify any backtracking to potentially homogenize complexes for pausing (or exit), or by allowing limited elongation for Pol II to survey local sequence.

It seems likely that distinct structural states will support Pol II translocation while TL movement, or movement of other domains, will be leveraged by the enzyme to support the NAC, or to control the probability/rate of entering and exiting off-pathway states. Structural and biochemical studies on Eco RNAP highlight domains that are mobile or dynamic, those can be tuned by factor binding, and may underlie both on and off NAC pathway states.^{82, 83, 84, 85, 86, 87} Recent structural studies have suggested that a distribution of states is present for “elementally” paused Eco RNAP. The elemental pause reflects a fundamental response of RNAP to specific template sequences where the biophysics of RNAP and DNA sequence interactions increase the probability of entering off-pathway states.^{73, 86} Pol II also can show sequence-dependent pausing in vivo^{88, 89, 90, 91} but much pausing by Pol II in vivo beyond factor-dependent promoter proximal pausing may be due to obstacles in chromatin (e.g. nucleosomes or other bound proteins and DNA damage). Consistent with this, high-resolution mapping of Pol II shows enrichment of Pol II density upstream relative to nucleosomal dyads.^{79, 90} Importantly, an open question is whether the half-translocated RNAP state is obligatory for pausing, and how elongation factors might regulate it. An important recent study has applied single-molecule technology to couple RNAP translocation to the passage of template DNA through a nanopore.⁹² This affords both sub-nucleotide and exquisite temporal resolution for monitoring of translocation during transcription. Here, every RNAP molecule observed to pause at an elemental pause site was also observed to enter the half-translocated state, suggesting that at the pause site assayed, the half-translocated state might be obligatory for pausing. Furthermore, kinetic modeling of enzyme states during the NAC suggested that there is a post-translocated state that is unable to bind an NTP. The physical nature of this state for RNAP and whether it is also present for Pol II remains to be determined. Nanopore experiments with other polymerases and time-resolved Cryo-EM determination of structural ensembles will likely be the way forward to answer these questions.

Pol II and other msRNAPs have a mobile clamp domain that controls how accessible the cleft between the two major lobes of the enzyme is. This flexibility of the clamp was apparent in early structural analysis of Pol II.^{61, 62, 66, 93, 94} There are different ideas and potential controversy around whether clamp opening or closing is required for steps in initiation for bacterial RNAP,^{95, 96} but clamp mobility during elongation by Pol II or RNAP is likely to be lower due to elongation factor binding.^{97, 98, 99} Other motions that can include the clamp, however, are strongly implicated during pausing or backtracking and the reversal of these states, as noted above and discussed below.

Tth RNAP has been described as having ratcheted and non-ratchetable states that are implicated in backtracking or responsiveness to factors that regulate it^{100, 101} (Figure 2A). A swiveled state for Eco RNAP, conceptually similar to the “ratcheting” observed for Tth RNAP has been more recently revealed and studied^{27, 87, 102, 103} (Figure 2B). EcoRNAP swiveling has been described as distinct from Tth RNAP ratcheting by ²⁷, likely due to precise structural details, but they can be considered analogous. Importantly, the terminology of “swiveling” does not imply specifics of mechanism, whereas “ratchet” may convey regulation by a “pawl” or suggest a relationship to translocation that may not be the case. Swiveled and unswiveled states involve the movement of clamp and shelf modules relative to the core, where the unswiveled state is represented by the post-translocated state with a substrate bound and a folded TL. Swiveled states are a continuum of different degrees of rotations of the large subunit shelf/clamp domains and have been observed for both paused and backtracked RNAP complexes, providing a rationale for how these states may be stabilized or regulated. Importantly, some of these states have the RNAP active site occupied by RNA (pre-translocated state or backtracked states) or by the template DNA maintaining interaction with translocated RNA (the partially translocated “half” state, sometimes described as a tilted RNA/DNA hybrid). As noted above, the half-translocated state has also been observed for paused Pol II in complex with NELF and the eukaryotic elongation factor DSIF. Recent data suggest that swiveling is not necessarily obligate for pausing in RNAP, but that different pause sequences and factor binding events can bias RNAP into different amounts of swiveling, and ensembles of structures represent how the rates or extents of pausing and its tunability by factors may be different in bulk for different sequences.^{82, 83, 84} The take home message from Eco RNAP studies is that RNAP conformational states can control the duration of off pathway events and elongation factors can promote or inhibit these states. Each structural state is supported by interaction networks, and factor binding can reinforce specific conformations or alter structural dynamics to bias structural ensembles. Similarly, Pol II interaction networks are attractive targets for how eukaryotic transcription elongation factors might alter Pol II elongation activity. It is also likely that these networks are shaped by evolution and provide the mechanistic basis for distinct behaviors of eukaryotic RNAPs and even for Pol II from different species.

Figure 2. Conformational flexibility in RNAP elongation complexes may underlie regulation by elongation factors.

(A) Tth RNAP complexes showing an “unratcheted” RNAP elongation complex (PDB 4wqs) superimposed on a “ratcheted” Tth RNAP complex (PDB 4wqt) bound to a hybrid elongation factor (Gre-C1), comprising sequences from Tth GreA (analog of eukaryotic TFIIS) and the Gre-homolog Gfh1.¹⁰¹ (B) Eco RNAP complexes showing an “unswiveled” RNAP elongation complex (PDB 6ri9) and a backtracked, “swiveled” RNAP elongation complex (PDB 6rip).²⁷ Complexes in both (A) and (B) were aligned by their entire β subunits. β (dark grey) is the bacterial homolog of the eukaryotic Pol II Rpb2 subunit. β′ (light grey) unratcheted (A), unswiveled (B); light blue ratcheted (A), swiveled (B)) is the bacterial homolog of the Pol II Rpb1 subunit. Bacterial RNAP contains two copies of α, which is the homolog of Pol II Rpb3 (α₁, red) and Rpb11 (α ₂, yellow). ω (slate blue) is the bacterial homolog of the Pol II Rpb6 subunit. This figure was generated using Pymol.²⁷⁴

Structural studies on Pol II complexes and potential for allosteric regulation of Pol II

High-resolution structural studies on Pol II active site mechanisms and enzymology have lagged behind beautiful more moderate resolution X-ray and Cryo-EM studies that increasingly demonstrate organization of Pol II complexes in both initiation and elongation, and even on chromatin.^{76, 77, 97, 98, 99, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139} The recent structural studies we do have show more and more of the preinitiation complex from yeast and humans,^{104, 105, 106, 107, 110, 111, 124} as well as increasingly sophisticated views of Pol II with Mediator,^{108, 109, 112, 124} with TFIID,^{109, 110} in initial transcribing states,^{122, 125} with nucleosomes,^{104, 113, 114, 115, 116, 117, 118, 119, 120, 132} elongation factors,^{76, 77, 97, 98, 99, 117, 127} in transcription coupled repair states,^{133, 134, 135, 136, 137, 138, 139}with termination/RNA processing factor Integrator,¹²⁶ the RNA capping machinery,^{123, 128} and components of RNA splicing machinery.¹²¹ Examples of some of these impressive Pol II-elongation factor complexes are shown in Figure 3. It is likely that time-resolved cryo-EM approaches^{140, 141, 142} will be the path to revealing insights into Pol II elongation and factor mechanisms during the NAC and in pausing and termination. A recent study has applied time-resolved cryo-EM to the early stages of promoter opening for Eco RNAP, enabling a diversity of states to be observed.¹⁴³

Figure 3. Structures of Pol II complexed with diverse elongation factors.

(A) Cryo-EM structure of mammalian Sus scrofa Pol II with human DSIF (comprising homologs of yeast Spt4 and Spt5 SUPT4H1 and SUPT5H, respectively), PAF1C, and SUPT6H, the human homolog of yeast Spt6 (PDB 6ted⁹⁷). (B) Cryo-EM structure of the yeast Komagataella phaffiii Pol II elongation complex with yeast elongation factors Spt4/Spt5, Spt6, Paf1C, and Elf1 (yeast homolog of conserved ELOF1)(PDB 7xn7¹¹⁷). (C) Cryo-EM structure of mammalian Sus scrofa Pol II with human Elongin complex (Elongin A, B, C) and SUPT6H (PDB 8of0¹⁵³). (D) Cryo-EM structure of the mammalian paused elongation complex with Sus scrofa Pol II and human NELF and DSIF (PDB 6mgl⁷⁶). Pol II in all panels is shown in surface representation while elongation factors are in cartoon representation. This figure was generated using Pymol.²⁷⁴

A recent review from Aoi and Shilatifard covers Pol II elongation and its factors deeply⁷⁸; however, we would like to discuss a few key points about factor-regulated Pol II elongation, noting that Chen and Cramer cover this ground more extensively.¹⁴⁴Specifically, what are the precise molecular mechanisms by which elongation factors alter Pol II activity? Elongation factors Spt4/Spt5 (DSIF),^{145, 146} Spt6,^{147, 148, 149} and members of the Paf1 complex^{81, 97, 150, 151, 152} have all been implicated in promoting Pol II elongation in cells or on DNA in vitro. By analogy to bacterial RNAP, a simple model is that elongation factor binding alters or biases conformational states towards on pathway NAC events and away from off-pathway pausing or backtracking. For example, Spt4/Spt5 (DSIF) binds above the Pol II cleft, acting as a clamp for the complex by encircling the upstream DNA.^{99, 127}

A more specific potential mechanism of allosteric control of Pol II by an elongation factor first emerged for the Rtf1 subunit of the conserved Paf1C complex. Specifically, for the Rtf1 homolog within the mammalian complex, a domain inserts close to the active site and contacts the N-terminal side of the BH, providing a tantalizing potential explanation for why addition of RTF1 to Pol II can stimulate Pol II elongation in vitro⁹⁷ (Figure 4). Intriguingly, the location of this Rtf1 “Latch” domain in the Pol II complex is occupied by a sequence insertion in yeast Pol II (compare Figure 4A and B). Sce Pol II is faster than mammalian Pol II in the absence of elongation factors in a simple in vitro assay.¹³ Perhaps mammalian Pol II has evolved to be responsive to the Paf1 complex to differentiate mature elongating Pol II from promoter proximal paused Pol II, while yeast, which lacks promoter proximal pausing, has evolved a “built in” mechanism to stimulate elongation. Consistent with a mechanism to enforce differences between paused and elongating Pol II in higher metazoans, NELF and Paf1C binding are mutually exclusive.^{76, 98} It is important to note that while yeast Rtf1 appears to have sequence similar to the mammalian Rtf1 Latch, density for Rpb2 sequence was still observed in the latch position for the Komagatealla pastoris yeast Pol II elongation complex structure, which includes Rtf1.¹¹⁷ Intriguingly, recent studies show that another Pol II elongation factor, Elongin, also has a Latch domain within its Elongin A subunit and binding of this Latch is mutually exclusive with the Rtf1 Latch (Figure 4C).¹⁵³ Elongin is a heterotrimer originally identified as a positive Pol II elongation factor in vitro^{154, 155, 156, 157} but has been demonstrated to have roles in Pol II ubiquitinylation and degradation upon DNA damage^{158, 159}through function as a Cullin-dependent E3-ubiquitin ligase. Of note, Paf1C was shown to be required for Elongin recruitment to stalled Pol II,¹⁶⁰ though their binding on Pol II should be mutually exclusive, there potentially is a handoff from one factor to the other while maintaining potential Pol II elongation stimulation via a Latch domain. Each of the Rtf1 Latch, Elongin A Latch, and Rpb2 insertion sequence has a similarly positioned arginine side chain that is placed for interactions with an adjacent Rpb2 loop or the Rpb1 BH or both (Figure 4D). In fact, yeast Rpb2 R728 and Elongin R555 appear to have conserved interactions with (yeast numbering) Rpb2 D760 and Rpb1 BH L808 (through its main chain carbonyl). Biochemical studies indicate that truncated variants of mammalian RTF1 or Elongin A that lack their latch domains lack elongation stimulation ability^{97, 153} but site-directed mutagenesis has not been performed as of yet. The next steps will be to rigorously assess requirements of key residues and mutant effects on Pol II biophysical properties.

Figure 4. Potential for allosteric regulation of Pol II by elongation factors.

(A) Sce Pol II crystal structure PDB 2e2h¹⁴ showing position of a non-conserved insertion sequence of Rpb2 (“Rpb2 latch”) adjacent to the Rpb1 bridge helix (BH). Rpb2 Arg728 is positioned to interact with the main carbonyl of BH residue Rpb1 Leu808 and a BH proximal loop of Rpb2 through Rpb2 Asp760. (B) Mammalian Cryo-EM structure of Pol II and elongation factors including PAF1C and RTF1 from PDB 6ted⁹⁷ showing the RTF1 latch domain adjacent to the Rpb1 BH, where RTF1 Arg 596 is positioned to interact with Rpb2 Asp1004. Only a modest repositioning of the RTF1 latch would allow interactions with residues homologous to those that interact with the yeast Sce Rpb2 latch analog Rpb2 Arg728. (C) Human Elongin A latch Arg555 interacts with Sus scrofa Pol II residues Rpb2 Asp792 and Rpb1 Leu831, homologous to targets of the yeast Rpb2 latch Arg728 (PDB 8of0.¹⁵³ (D) Overlay of (A-C). This figure was created using Pymol.²⁷⁴

Sce Pol II does show elongation defects when a subset of Paf1 subunits is acutely depleted in vivo, suggesting additional mechanisms for Paf1C to influence elongation outside of a Latch domain.¹⁵² In vivo, Spt6 promotes Paf1C association with transcribed regions^{161, 162, 163} and interacts physically with the Paf1 component Cdc73 in yeast¹⁶¹ and in the elongation complex structure,¹¹⁷ suggesting some Spt6 mutant effects will relate to Paf1C defects. In vivo studies do indicate that there are differences between these factors¹⁵⁰ and we do not yet know the mechanism by which Spt6 might directly influence Pol II activity beyond its roles in chromatin structure.¹⁶⁴

Pol II activity-mutant effects in vivo

How do we understand how Pol II activity alters transcription and organismal phenotypes in vivo? The primary approach here has been to leverage mutants in the Pol II active site that alter catalytic activity to probe specific effects on the transcriptome and on cotranscriptional processes. Here, many of the mutants used have been studied genetically, biochemically, and biophysically in the budding yeast system while homologous mutants have been ported to metazoans to understand effects in more complicated systems. Gene expression, genetic interaction, and phenotypic profiling clearly indicate that Pol II GOF and LOF mutants are distinct in vivo, consistent with their observed altered elongation properties in vitro.⁶⁸ It is expected that gene expression, genetic interaction, and some growth phenotypes will relate to the sum of direct transcription defects arising from altered Pol II activity and the indirect effects that propagate from changes to transcription across the genome. In the next section, we will discuss the nature of specific phenotypes arising from altered Pol II activity and how we might understand them.

Pol II activity mutants in yeast

As noted above, LOF and GOF Pol II mutants lead to distinct phenotypes in vivo.^{33, 34, 68}These in vivo phenotypes correlate so well with measured biochemical activity in vitrothat they can be used to predict biochemical defects of Pol II mutants.^{36, 37} Our lab has leveraged this correlation to perform deep mutational scanning on the Pol II active site to understand what additional TL residues individual mutants rely on to determine how they alter activity.³⁷ The simple phenotypic profile that can distinguish between LOF and GOF mutants relies on three in vivo phenotypes that represent a range of direct mechanistic connections to Pol II activity.

One phenotype is caused by sensitivity to a drug, mycophenolic acid (MPA), that decreases GTP levels in cells.¹⁶⁵ It was originally thought, and is still promulgated widely, that mutants defective for Pol II elongation should be sensitive to a reduction in substrate levels (e.g. GTP) (discussed extensively in ¹⁶⁶). Therefore, sensitivities of transcription factor or Pol II mutants to MPA or other drugs that alter NTP levels (e.g. 6-azauracil, 6-AU) were discussed as due to elongation defects (see¹⁶⁷). However elegant work from the Reines lab and others strongly argued that nucleotide depleting drug-sensitive transcription mutants were defective for expression of a key drug resistance gene.^{168, 169, 170, 171} The actual mechanism for most factors’ sensitivities to MPA is due to defects resulting in defective initiation at the IMD2 promoter and not bulk elongation defects as indicated by classical studies above. Our lab has demonstrated that Pol II activity mutants are not selectively sensitive to MPA treatment when they are not differentially starved for GTP from WT cells.¹⁶⁶ We discuss this mechanism in detail below to make these points as clear as we can.

Importantly, Pol II GOF mutants with demonstrated in vitro increased elongation activity are sensitive to MPA and similar drugs while Pol II LOF with severe elongation defects in vitro and in vivo are generally resistant^{13, 33, 34, 36} – counterintuitively from the widely made assertion that bulk elongation defects are responsible for sensitivity to GTP depletion. While MPA or related drugs do confer altered Pol II elongation properties in vivo,¹⁷² these altered properties do not appear to be the basis for altered drug sensitivity of Pol II mutants.¹⁶⁶ The basis for altered MPA sensitivity for Pol II active site mutants instead relates to the exquisite sensitivity of the Pol II initiation mechanism in S. cerevisiae to NTP levels. The gene IMD2, which is required for resistance to MPA, has evolved to use this initiation sensitivity to NTP levels in its regulation. IMD2 is one of three IMPDH homologs that perform the rate limiting step in GTP synthesis; the other homologs of this enzymatic activity being encoded by IMD3 and IMD4.¹⁷⁰ Under conditions of GTP limitation, IMD2 transcription is induced. MPA causes GTP depletion because it is an inhibitor of the IMPDH enzymes encoded by IMD3 and IMD4, but the IMPDH encoded by IMD2 is resistant to MPA.^{169, 170} Therefore, wild type yeast survive MPA treatment because they induce IMD2 expression upon limitation of GTP due to inhibition of IMD3 and IMD4, but any mutants defective for IMD2 induction cannot, because they will be starved for GTP.^{168, 169, 171}

This IMD2 induction mechanism is fascinating as it occurs through a transcription start site switch elegantly described by the Brow and Reines groups.^{173, 174} Work examining Pol II active site mutants shows that initiation and not elongation drives Pol II mutant sensitivity to MPA through its effects in IMD2.¹⁶⁶ Transcription start site usage in yeast can be affected by NTP levels because budding yeast utilize a promoter scanning mechanism for identification of transcription start sites.^{174, 175, 176, 177} In this mechanism, there is a kinetic competition between initiation at a given position and a scanning process that moves the PIC in a directional fashion from upstream positions to downstream positions. Pol II mutants alter TSS distributions at promoters by increasing or decreasing initiation efficiency at all TSSs.^{177, 178} GOF mutants that increase Pol II catalytic activity in vitro, also increase Pol II initiation efficiency at all TSSs, which results in a net upstream shift in TSS distributions at nearly all yeast promoters because less promoter DNA is scanned on average before initiation happens.^{177, 178} In contrast, LOF mutants decrease Pol II initiation efficiency at all TSSs, resulting in a net downstream shift in TSS distributions at all yeast promoters because more promoter DNA is scanned on average before initiation happens.

At IMD2 under GTP-replete conditions, IMD2 transcription initiates from upstream TSSs that start with GTP (“G+1” TSSs).^{173, 174} These TSSs are positioned upstream of an attenuator element that prevents Pol II elongation from making transcripts that contain the IMD2 ORF. When GTP is depleted, such as when yeast are treated with MPA, initiation by promoter scanning proceeds past upstream G+1 TSSs, which decrease in usage, to downstream positions. There is a major TSS downstream of the attenuator element that specifies a transcript beginning with ATP (A+1 TSS) and is only reached when upstream TSSs are compromised. This depression of usage of G+1 TSSs upon MPA treatment in fact is genome wide.¹⁷⁷ Pol II LOF mutants that have reduced initiation efficiency at all TSSs allow promoter scanning to continue past the attenuator element because of reduced initiation at upstream G+1 TSSs, allowing use of the downstream +1A TSS in absence or presence of MPA.^{34, 179, 180} Pol II GOF mutants, in contrast, cannot switch to the downstream +1A TSSs when GTP is limiting because they use TSSs that are in between the upstream G+1 TSSs and the functional downstream A+1 TSS.¹⁶⁶ Therefore, they are sensitive to MPA treatment because they have a defect in functional IMPDH expression. Even when Pol II mutants are normalized for IMPDH expression by removing IMD2 and the possibility of differential GTP starvation among WT, Pol II LOF, and Pol II GOF strains, Pol II LOF mutants do not confer additional MPA sensitivity.¹⁶⁶ In other words, transcription elongation may be sensitive to reduced GTP levels, but this sensitivity is not rate limiting for Pol II mutants defective for elongation relative to other GTP-sensitive properties of cells.

The other in vivo growth phenotypes that are predictive along with MPA sensitivity for Pol II biochemical defects are the Spt^- phenotype deriving from altered expression of a specific allele of LYS2, lys2–128∂,^{38, 181, 182, 183, 184} and the suppression of a specific allele of gal10 (gal10Δ56) where altered GAL10 transcription interferes with GAL7initiation.^{38, 163, 185, 186} These have been discussed at length in a previous review of ours,³⁸ but it is likely that the sensitivities of each locus to Pol II activity derives from compound transcription units where there is Pol II elongation and termination over a promoter region.

Cotranscriptional processes are affected by altered Pol II activity in yeast and other organisms

Pol II mutants with altered activity show widespread effects on TSS selection and gene expression in budding yeast. Furthermore, consistent with different biochemical effects in vitro, Pol II mutants also alter termination at non-coding RNAs with GOF mutants exhibiting extended RNA 3′ ends and LOF mutants exhibiting shortened RNA 3′ ends on a subset of tested yeast non-coding RNAs that terminate using the Sen1-Nab-Nrd1 pathway.^{166, 187} Note that these effects could be explained by increased and decreased elongation rates in vivo (as predicted from in vitro studies), respectively, or by decreased and increased probability of termination. Pol II mutants also alter 3′ end formation and polyadenylation site choice for mRNAs in yeast.^{188, 189} Effects at 3′ ends of transcripts are consistent with kinetic models for cotranscriptional events where increased elongation rates favor downstream processing and slowed elongation favors upstream processing.

The employment of Pol II activity mutants as probes of cotranscriptional processes is predicated on assumptions that their in vivo elongation rates reflect those found in in vitro biochemical and biophysical studies. Global Pol II elongation rates are more difficult to measure in yeast than metazoans owing to short genes and lack of pharmacological methods for inhibiting initiation or otherwise synchronizing elongation release from promoters. Where elongation rates have been measured carefully in yeast, both Pol II GOF and LOF mutants confer slow elongation in vivo¹⁶⁶ in a commonly used chromatin IP assay examining bulk Pol II occupancy across a gene upon promoter inhibition.^{28, 172}This result from our group was in contrast to a prior report on a single Pol II GOF mutant using the identical assay.¹⁸⁷ Our studies demonstrated how that prior report could be artifactual, and our result was robust across multiple Pol II GOF alleles.¹⁶⁶ For interpretation of how Pol II GOF mutants confer phenotypes, it will be essential to understand these results more thoroughly. It will be important to employ orthogonal methods to measure elongation rates for Pol II mutants, and at several genes. Careful measurement of Pol II occupancy together with nascent synthesis analyses can estimate elongation rates. However, with likely global changes to RNA stabilities in Pol II activity mutants,¹⁹⁰ mRNA stabilities should also be measured carefully. Assays monitoring single gene transcription can measure elongation rate based on the appearance of fluorescently marked 5′ and 3′ RNA ends but are limited to individually engineered loci.¹⁹¹ Single molecule tracking (SMT) is powerful in that it can measure average duration of Pol II in slow moving states (presumed to be engaged at specific chromatin locations) (for example^{192, 193}) but here the length of transcription unit, required for an estimate of elongation rate, can only be inferred from a weighted average of expressed gene lengths. Additionally, because premature termination events can limit length of chromatin association events, there will be caveats to any elongation rate estimate.

Pol II activity mutants in yeast showed altered RNA splicing when examined by microarray.⁶⁸ It has been reported that Pol II GOF mutants decrease splicing efficiency (increasing intron retention) while Pol II LOF mutants increase splicing efficiency (decreasing rates of intron retention). All current interpretations in the literature are based on assumptions of fast elongation in vivo for Pol II GOF mutants and slow elongation in vivo for Pol II LOF mutants, which may not be the case (as discussed above). Additional looks at potential splicing defects of Pol II activity mutants are warranted using several mutants and technologies that can measure multiple aspects of transcription dynamics. For example, Pol II mutants may have substantial and widespread effects on initiation and RNA degradation rates, potentially complicating interpretations of ratios of intron-containing pre-mRNA to mRNA. Splicing defects for Pol II mutants, shown using microarrays, have been recapitulated using RNA-seq for one Pol II GOF mutant and to some extent for one Pol II LOF mutant.¹⁹⁴ Because of substantial effects across the transcriptome, Pol II mutants’ phenotypes will be both directly and indirectly related to their synthesis defects. These have not been formally assessed by acute perturbations to Pol II activity or acute uncovering of Pol II mutants. Of note, some Paf1C gene deletions in yeast have widespread transcription defects across the genome^{195, 196} but it was not known how direct these changes were to the loss of Paf1C functions. Recent work from the Arndt lab has shown that Paf1C mutants can have elongation rate, elongation processivity, and RNA splicing defects.¹⁵² Intriguingly, they find that elongation defects are immediately apparent upon acute depletion of a subset of Paf1C components, but splicing is unaffected upon acute Paf1C depletion and requires longer term effects of subunit deletion.¹⁵² These experiments argue that splicing can be normal even when elongation is defective and that splicing defects can be an indirect consequence of defective elongation.

Because Pol II elongation could have altered kinetics over different gene regions, disruption of Pol II activity can have consequences on the cotranscriptional events that have evolved to be kinetically matched to Pol II behavior, if not directly coupled to it. One area where this might be important is in chromatin remodeling and modification that occurs over different gene regions. In both yeast and human cells with genetic alteration to Pol II activity, chromatin modifications and other events with specific distributions over genes can be altered.^{197, 198}

Pol II activity and cotranscriptional processing outside of yeast

Transcription elongation rates are not uniform in mammalian cells, and intronic sequences (and therefore long genes) can be more rapidly transcribed.^{199, 200, 201} Elongation rate also increases within the first 10–20 kb downstream from a gene’s promoter in human cells, consistent with a change in Pol II properties for genes longer than 10–20 kb.^{32, 200} Given potential differences in elongation properties in cell types, at different genes, and over different types of sequence, quantitative modeling of elongation at high resolution^{202, 203, 204, 205} will have great value in understanding where elongation properties may be affected in disease.

The connection between potential Pol II elongation and RNA processing are derived from observations much wider than yeast and have been made across organisms for decades (reviewed extensively^{206, 207, 208, 209}). Pol II activity mutants have been introduced into human cell lines where mutants in some cases have been chosen based on yeast studies.^{189, 198, 210, 211} Here, elongation rate in vivo is more easily measured across the genome than in yeast owing to the length of genes. When effects on splicing were examined, results were complex and did not easily fit the simple kinetic “window of opportunity” model in that there were many kinds of defects including a number that were shared between Pol II GOF and LOF mutants.²¹⁰ Kornblihtt and colleagues, in²¹², summarize introns into classes to explain their observed responsiveness or not to Pol II elongation rate and the direction of their responsiveness. A basic model is that splicing decisions could be sensitive to Pol II elongation rate in many ways, and that an optimal elongation rate for individual splicing contexts would be required for proper processing. As with interpretation of yeast experiments, it would be interesting to know which processing defects were directly related to altered Pol II kinetics and which might be indirect. Importantly, Pol II LOF and GOF mutants employed by the Bentley lab do demonstrate altered elongation rates from wild type as predicted.²¹⁰ Pol II mutants also show changes in 3′ end formation consistent with observations in yeast. Furthermore, elegant studies again from the Bentley lab present evidence for cotranscriptional RNA folding differences that are sensitive to Pol II elongation rate, with biological consequences, for example on RNA processing at histone genes and A to I RNA editing.^{211, 213} This result was especially interesting because older models postulated that a process’s sensitivity to Pol II elongation rate could be through altered RNA folding.

In the above sections, we have discussed some of the basic mechanisms for Pol II activity, the nature of factors that regulate it, and how Pol II mutants alter this activity in vitro and in vivo. For the remaining portion of this review, we will discuss how Pol II activity defects can relate to disease.

Pol II defects in aging and disease

Recent findings have illustrated wide-ranging and diverse effects of Pol II activity in aging and disease. Pol II is essential for transcription but makes errors.²¹⁴ These errors can lead to proteostasis defects due to translation errors and protein misfolding.⁴⁵ The accumulation of toxic proteins is common in age-related diseases.²¹⁴ Transcription errors appear to increase in yeast as yeast cells age.⁴⁵ Aging has recently been linked to global changes in Pol II elongation properties, including increased Pol II elongation rate in older individuals from model metazoans to senescent cells from human cell lines,²¹⁵ and counterintuitively, increased Pol II stalling in liver from aged mice.²¹⁶ Debes et al.²¹⁵provocatively showed that senescent human cells showed faster Pol II elongation rate than proliferating, that older C. elegans or D. melanogaster showed faster Pol II elongation rates than younger, and that lifespan-extending alterations slowed Pol II elongation rates. Furthermore, slow and fast Pol II mutants showed shifts in lifespans consistent with slower elongation being lifespan extending and faster elongation being lifespan shortening. Counterintuitively, Gyenis et al.²¹⁶ found that in aging murine livers, the overall number of nascent transcripts was observed to decrease relative to younger livers and there appeared to be a deficit in Pol II completing transcripts based on a skewed distribution of Pol II towards gene 5′ ends. These and additional experiments led to the conclusion that there is an increase in Pol II stalling on genes with age. The increased stalling in aged liver was attributed to increased DNA-damage-induced Pol II blockage, leading to loss of Pol II on genes as transcription proceeds. Interestingly, the apparent increase in total transcribing Pol II on genes was observed and was proposed as a compensatory mechanism for the transcriptional output that is lost due to Pol II stalling. Pol II stalling and loss from genes provides an explanation for a selective decrease in transcripts from longer genes that has been previously observed,²¹⁷ because the longer a transcription unit, the greater the probability of a DNA damaged-induced stall occurring. Even though these two papers find potentially disparate results, the two groups together argue that different assays can detect different alterations to Pol II activity and that global changes to Pol II function are manifold in aging.²¹⁸

Pol II mutations in disease

In addition to global defects in Pol II properties in aging, mutations within Pol II itself can lead to disease. Somatic mutations in the human POLR2A gene (encoding Rpb1 of Pol II) were identified in a subset of benign meningiomas.²¹⁹ Clark et al. discovered somatic mutations in POLR2A in 23 individuals that had no known meningioma drivers.²¹⁹ Of the 23, 19 bore a POLR2A/Rpb1 missense mutation, Gln403Lys, while the other four had a two-amino acid deletion, ΔLeu438His439. These mutations are adjacent to each other in the structure and reside in the Pol II dock domain that interacts with initiation factor TFIIB during initiation.^{220, 221} POLR2A mutant meningiomas do show altered gene expression that differentiates them from other meningiomas but were reported to not have significant RNA processing or transcription fidelity defects.²¹⁹ Additional studies only report these two recurrent variants in meningioma patients,^{222, 223, 224} with an excess in female patients over expected.^{222, 224} The mechanism by which mutations in a globally-required general factor Pol II elicit such a tissue specific phenotype are currently unknown.

Furthermore, it is now clear that mutations in POLR2A can cause a neurodevelopmental disease by unknown mechanisms. In 2019, Haijes et al. identified 16 individuals with neurodevelopmental and neuromuscular issues with very rare and suspected disease-causing dominant de novo mutations in POLR2A.²²⁵ A subsequent report by Hansen et al. documented 11 more individuals with similar phenotypes and additional rare POLR2A mutations,²²⁶ with one individual inheriting their POLR2A variant and the rest being de novo as in the Haijes et al. cohort. Additional individuals have been described in case reports that expand the observed variability in symptoms, including a single patient with a pediatric brain tumor.^{227, 228, 229} POLR2A variants have also been discovered in studies examining individuals with congenital head malformation craniosynostosis^{230, 231, 232} or cerebral palsy-like symptoms.²³³ Each of these studies^{230, 231, 232, 233}identifies a wide range of de novo, dominant causative mutants in genes in addition to POLR2A that have previously been linked to a number of disorders with neurodevelopmental defects. Many of these other genes are related to gene expression, chromatin remodeling, or chromatin modification. The striking feature when considering all identified POLR2A-related cases together is that there are a wide range of phenotypes and severities among them. POLR2A cases can exhibit diverse but commonly neurodevelopmental and neuromuscular issues, including developmental delay, epilepsy, autistic behaviors, aggressive behavior, hypotonia, strabismus, craniofacial defects, and other symptoms. Specific phenotypes are not shared between all affected individuals, but all individuals have neurodevelopmental or neuromuscular phenotypes.

How do these POLR2A mutations result in altered Pol II function and how does altered function result in disease? We will consider how Pol II function might be altered. First, no clear structure function relationship is evident between the locations of mutations in Pol II and the severity or spectrum of symptoms (Figure 5). Considering the mechanism of dominance may help point to a mechanism of action. For example, why are the POLR2A alleles dominant? Many neurodevelopmental disorders can be caused by haploinsufficiency that is likely due to the sensitivity of critical stages of development to gene dosage. It is likely that POLR2A can be haploinsufficient for viability, because inactivating variants (frameshifts, nonsense mutations) are greatly depleted in the general human population.²³⁴ However, some do exist in the general population and have not been attributed as causative of disease. That said, two POLR2A syndrome patients have been found with inactivating mutations (frameshifts unlikely to support any complex assembly),²²⁵ with two more likely inactivating alleles identified in craniosynostosis patients.²³² This observation supports the idea that heterozygosity for a null allele might be sufficient for disease in some patients, but null heterozygosity does not obligately lead to disease in all individuals. Perhaps surprisingly, the two patients from Haijes et al. with inactivating POLR2A mutations have milder phenotypes relative to the entire cohort, and the majority of patients, and all severely affected individuals, have missense mutations.^{225, 226} There are multiple possible interpretations here. It seems likely that stochasticity or environmental conditions during critical developmental windows will affect disease severity. This might explain how inactivating POLR2A mutations are depleted from the human population because in some cases, defects are so severe as to be incompatible with organismal viability, while in other individuals, heterozygosity for inactivating POLR2A mutations leads to no or milder phenotypes. The possibility leads to the prediction that there will be additional individuals identified with inactivating POLR2A mutations covering a range of disease severities.

Figure 5. Disease causing mutations in POLR2A.

(A) POLR2A variation from gnomAD v4.1,²³⁴ updated from Haijes et al., 2019 Figure 2A.²²⁵ The positions of amino acids in the POLR2A coding sequence are mapped from left to right. Positions of missense, in-frame deletions (IF deletions), frameshift, and stop codon mutations present in gnomAD v4.1 are shown in labeled rows. The height of the missense black tick marks indicates the frequency of missense mutations in gnomAD v4.1 at that position (truncated at n = 15). Variants that cause disease are shown as vertical tick marks and labeled, those with X in stop codon row are nonsense alleles. Missense alleles with asterisks have been identified in more than one individual. Color code for severity is from severe (red) to mild (yellow) with dashed lined indicating phenotype not reported individually. POLR2A alleles identified in other studies are colored by key phenotype. Positions of synonymous mutations in gnomAD v4.1 are displayed by gray tick marks. The height of the grey bars also indicates the frequency of mutations at those positions (truncated at n = 10). Regions of POLR2A depleted of missense variation were estimated from the distribution of gnomAD v4.1 missense alleles. AlphaMissense calculations for probability of deleterious mutation was also plotted with a LOWESS smooth.²⁷⁵, ²⁷⁶ Positions that are on average predicted to be deleterious display high peaks and positions predicted to be benign display low peaks. (B) Structure of mammalian Sus scrofa Pol II from PDB 6ted.⁹⁷ All Pol II subunits are shown in surface view. Visible subunits are labeled by Rpb number. Template DNA is in blue, non-template DNA is forest green, and RNA is in red. (C) Structure of mammalian Sus scrofa Pol II from PDB 8b3d¹³⁶ rotated −90° on the y-axis from the same view in (B). Rpb1 (POLR2A) is shown as cartoon with positions of disease mutants shown as spheres with colors corresponding to phenotype and severity as labeled in (A). Non-POLR2A subunits are shown in transparent surface view. Mutant color-coding: yellow, mild POLR2A disease; orange, moderate POLR2A disease; red, severe POLR2A disease; black dashed, no symptom data; purple, alleles identified in meningiomas; deep blue; POLR2A case with Posterior Ependymoma; green, alleles identified in cerebral palsy study; cyan, alleles identified in craniosynostosis patients. (D) Rpb1 (POLR2A) regions depleted for missense alleles as determined in (A) shown in brown cartoon. Rpb1 (POLR2A) is shown as light gray cartoon with other Pol II subunits shown in transparent surface view like in panel (C), but all are dark grey. Regions of Rpb1 (POLR2A) with high mutant desert Z scores are colored in brown. (E) Mutant desert regions of Rpb1 (POLR2A) shown individually by color. This figure was generated using Pymol.²⁷⁴

Second, it is also possible that missense alleles could have stronger effects than just reduction in Pol II dosage because mutant Pol II could interfere with WT Pol II on genes even if partially or mostly functional, thus having dominant negative effects. Haijes et al. noted that POLR2A disease alleles were enriched in regions of POLR2A greatly depleted for missense mutations in the human population (Figure 5A).²²⁵ Very broad surveys of human genetic variation now indicate that genes intolerant of inactivating variants are also enriched in those intolerant of missense mutations.²³⁵ Furthermore, the rate of missense mutations in missense-constrained regions across the genome is 6.6-fold higher for a cohort of individuals with a developmental disorder than controls,²³⁵consistent with POLR2A-constrained regions being enriched for identified disease alleles. The variation-depleted regions of POLR2A are highly conserved, surround the active site, nucleic acid-interacting regions, and Rpb1-Pol II subunit interfaces (Figure 5D–E). These results argue that strong missense loss of function alleles for POLR2A in critical regions for enzyme stability or activity do not support organismal viability. Under this model, it might be predicted that disease alleles represent weaker alleles that are not completely inactivating but might still be predicted to be LOF. Most disease mutations are in residues conserved from yeast to human and both Haijes et al. and Hansen et al. modeled a subset of alleles in yeast.^{225, 226} Results indicated that most alleles did not confer strong growth defects or phenotypes in genetic assays where known Pol II activity alleles give phenotypes. That said, some alleles displayed slow growth, or showed sensitivity to protein denaturants (consistent with reduced stability or protein folding defects) or showed genetic interactions with mutants in Pol II elongation or regulatory factors. Indeed, the majority of disease alleles are buried within the Pol II structure and might be predicted to alter structure or structural dynamics (Figure 5). Notably, there are alleles proximal to the TL and to the TFIIS interaction surface, suggesting effects on catalysis or backtracking in some cases (Figure 5).

Both altered catalysis and loss of TFIIS activity can lead to increased transcription errors in mature mRNA from increased misincorporation and decreased proofreading, respectively. The Vermulst lab has modeled the effects of a subset of disease alleles in yeast and human cell lines to measure transcription fidelity.⁴³ Two of four mutants tested in yeast showed increased levels of transcription errors. One allele, Sce Rpb1 L1101P is in the TL and another substitution in that position (L1101S) is a bona fide GOF allele with increased catalytic activity in vitro. A proline substitution is a bit harder to predict, and this allele was much sicker in our deep mutational scanning dataset³⁷ than reported by Haijes et al.²²⁵ However, our study does indicate that substitutions of H,K,M,N,Q,S,T for L1101 are predicted as GOF.³⁷ The other allele showing increased transcription errors, SceRpb1 N1232S, is in a residue directly adjacent to TFIIS, suggesting it could have compromised TFIIS proofreading. Human genes containing these two alleles were introduced into HeLa cells using α-amanitin resistant transgenes so the WT POLR2A could be degraded due to α-amanitin treatment.⁴³ Only the human version of N1232S (POLR2A N1251S) showed altered transcription fidelity in the human system.⁴³ These experiments were the first to model any aspect of POLR2A disease alleles for transcriptome effects in human cells and show that there will likely be complexity in interpreting modeled alleles in non-human systems.

How do Pol II mutations cause disease?

It is reasonable to speculate that mutations in Pol II cause disease through gene expression defects, as Pol II is essential for the transcription of all protein coding genes and a wide range of non-coding RNAs. The diversity of symptoms for POLR2A syndrome would be consistent with gene expression defects in systems across the body, in any number of genes, but with key defects in development of the neuromuscular system. Adding more complexity, even subtle defects spread across many genes can propagate into indirect effects. Given stochastic and genetic modifier effects during development of individuals, it is possible that even individuals with identical Pol II mutations could have diverse outcomes. There are three cases with individuals sharing POLR2A alleles, and in each where there is information, the described phenotypes were distinct. In Haijes et al., an individual had POLR2A Ile457Thr with severe symptoms including hypotonia and developmental delay, but no epilepsy reported.²²⁵ However, Giacomini et al. observed a patient with this same mutation and shared general hypotonia and developmental delay, but severe epilepsy was reported.²²⁸ Additionally, in Haijes et al. and Hansen et al., three individuals shared POLR2A Asn1251Ser with one patient diagnosed with ASD while the other two were not.^{225, 226} While additional effects of genetic background could be the cause for these phenotypic disparities, it seems more likely that environmental or stochastic effects are a major driver of developmental defect severity in individual patients. Finally, Hansen et al.²²⁶ identified an individual with mild disease with G1418R and this allele was also noted in a craniosynostosis patient.²³² Even if effects on individual genes might be diverse across individuals, there could still be a gene expression signature that is recognizable among patients. This signature might be best determined from patient-derived iPSCs or iPSC-derived organoids or motor neurons. Hope for a recognizable gene expression signature for POLR2A syndrome alleles comes from diverse autism spectrum disorder genes converging on shared gene expression defects.^{236, 237, 238, 239}

What is the basis for the neurodevelopmental effects of POLR2A alleles? Neurodevelopmental disorders, commonly, are due to de novo mutations in genes intolerant of inactivating alleles,²³⁵ consistent with mutations being associated with severe consequences, and neurodevelopment being exquisitely sensitive to gene dosage. These disorders are also commonly associated with genes involved in gene expression from developmental regulators, chromatin remodelers, epigenetic modifiers, and RNA binding or processing factors.^{240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253} The wide range of genes that can perturb neurodevelopment points to sensitivity to defects, and since POLR2A is required for expression of all genes, a combination of diverse, subtle defects might underlie POLR2A pathology.

Long genes in metazoans are sensitive to transcription defects

How do a wide range of gene expression defects lead to neurodevelopmental disorders? There is obviously a huge amount of complexity and there may be countless potential underlying causes. That said, one interesting feature of neurons is that they express genes longer on average than many other tissues.^{254, 255, 256, 257} As noted above regarding Pol II stalling in aging, long genes statistically will be expected to be more difficult to complete in the presence of stochastic DNA damage. Long genes may also have challenges for co-transcriptional processing, whether due to increased numbers of exons, or perhaps more likely, due to processing having to deal with increased intron lengths. Long neuronal genes have distinct requirements for topoisomerase activity²⁵⁸while also being repressed by DNA methylation, disruption of which is important for Rett syndrome.^{255, 259} Long genes in mammalian cell line systems in general are more affected by elongation factor defects, e.g. knockdown or depletion of Spt6,^{148, 150, 162}Paf1C,^{150, 151, 162, 260} U1 snRNP,^{261, 262, 263} CSB,²⁶⁴ CDK12,²⁶⁵ or overexpression of dominant negative TFIIS.^{266, 267} Recent results indicate that loss of the major TFIIS isoform in human cells does not have strong effects on nascent transcription⁹¹ and this adds an important caveat to interpretation of studies employing a dominant negative TFIIS in human cells.^{266, 267} It was previously known from yeast studies that overexpression of dominant negative TFIIS had much stronger phenotypes than deletion²⁶⁸ and therefore it is likely that it promotes additional defects beyond blocking resolution of backtracked Pol II. In Sigurdsson et al.,²⁶⁸ these additional defects were suggested to be a block to intrinsic cleavage by Pol II for rescuing backtracked complexes, but this is difficult to show in vivo and does not preclude additional interfering activities of mutant TFIIS hindering elongation or enhancing or stabilizing backtracking.

It is likely that defects in long gene expression derive from a constellation of different defects that converge on long genes as a diverse class that are enriched for different properties. While long genes can be enriched for stalling due to DNA damage, loss of specific factors that function in regulation of Pol II and DNA repair at stall sites can have different disease outcomes.^{74, 264, 269} These differences appear to relate to kinetics of Pol II removal or repair choice at the stall site. Long gene defects go beyond length, and it is important to understand both sequence and epigenetic state distinctions for long genes relative to short. Sequence composition can affect how well elongation proceeds.^{203, 270}Enrichment of intronic cleavage and polyadenylation sites can prevent full length transcript production while presence of 5′ splice sites can antagonize premature termination,^{271, 272, 273} meaning defects in these processes can have effects on gene expression for genes sensitive to them. Interestingly, the U1 snRNP, which recognizes 5′ splice sites may also function to alter Pol II properties independently of directing splicing and antagonizing premature termination.²⁶³ Most of the gene length for long genes is intronic, and intron and exon sequences can have different makeups. These differences can relate to base compositions, nucleosome positioning, and density of sequence motifs affecting transcribability or directing RNA processing and/or termination. A very interesting recent paper suggests that introns of human genes (but not mouse) are enriched for sequence elements that can promote R-loop formation on which Pol II dynamics are sensitive to activity of the transcription coupled repair factor CSB.²⁶⁴

Conclusions

In this review we have discussed the basics of Pol II mechanisms, focused on the Pol II active site. Pol II activity shapes gene expression in many ways and recent studies link Pol II function to disease. The complexity of transcription elongation in cells relates to the specific demands for each transcription unit where there will be an interplay between Pol II on genes and the processes that take place along with it and in response to its passage. There has been an explosion of structural and genome-level information that inform us about the mechanisms of Pol II function and the consequences for when Pol II and its regulatory factors are disrupted. We have indicated in each section of this review the direction for the next steps in understanding Pol II mechanisms in gene expression and disease. It will be exciting to watch how the field progresses over the coming years.

Highlights.

• RNA Polymerase II (Pol II) is the engine for gene transcription in eukaryotes.

• Pol II mechanism can be studied in model systems using mutants.

• Mutants show that activity is finely balanced in wild type Pol II.

• Activity mutants widely affect the transcriptome in many ways.

• Pol II mutants have recently been linked to human disease.