Crystal structure of a transcribing RNA Polymerase II complex reveals a complete transcription bubble

Summary

Notwithstanding numerous published structures of RNA Polymerase II (Pol II), structural details of Pol II engaging a complete nucleic acid scaffold have been lacking. Here, we report the structures of TFIIF stabilized transcribing Pol II complexes, revealing the upstream duplex and full transcription bubble. The upstream duplex lies over a wedge-shaped loop from Rpb2 that engages its minor groove, providing part of the structural framework for DNA tracking during elongation. At the upstream transcription bubble fork, rudder and fork loop-1 residues spatially coordinate strand annealing and the nascent RNA transcript. At the downstream fork, a network of Pol II interactions with the non-template strand forms a rigid domain with the Trigger Loop (TL), allowing visualization of its open state. Overall, our observations suggest that “open/closed” conformational transitions of the TL may be linked to interactions with the non-template strand, possibly in a synchronized ratcheting manner conducive to polymerase translocation.

Introduction

The pre-initiation stage of transcription requires concerted interactions between RNA Polymerase II (Pol II) and the general transcription factors TFIIB, TFIID, TFIIF, TFIIE and TFIIH. During initial promoter melting TFIIH generates an unwound region of 7–9 base pairs. Subsequently, this transcription bubble is unwound to approximately 18–25 bases and a short DNA-RNA hybrid is synthesized. Transcripts of 10 or more nucleotides result in promoter escape and stabilization of a mature bubble (Liu et al., 2011; Luse, 2012; Nechaev and Adelman, 2011). The number of nucleotides unwound in a mature bubble is still a matter of debate since sizes ranging from 8–22 nucleotides have been reported for bacterial, archaeal and eukaryotic polymerases (Fiedler and Timmers, 2001; Naryshkin et al., 2000; Pal et al., 2005). In addition, the size of the bubble might not be fixed but depend on Pol II’s transcriptional stage; evidence of a scrunched state –where template and non-template strand bases are compacted in space relative to relaxed conformations– has been proposed for the early stages of transcription initiation in bacteria (Kapanidis et al., 2006; Revyakin et al., 2006). Similarly, other transcriptional events such as backtracking or interactions with elongation or termination factors might alter the number of bases (and location) in the mature bubble (Fiedler and Timmers, 2001). Notwithstanding numerous Pol II structures published to date, structural details of a complete transcribing complex including upstream and downstream DNA duplexes and a full transcription bubble have yet to be revealed. Here we report the crystal structures of Pol II in complex with a complete nucleic acid scaffold that illustrates the architecture of a Pol II transcribing complex.

Results

Design, Assembly, and Crystallization of Pol II transcribing complexes

Assembly of a Saccharomyces cerevisiae Pol II transcribing complex was achieved by mixing Pol II with pre-assembled nucleic acid scaffolds (see experimental section). The main scaffold used for our experiments (scaffold 1) consisted of two synthetic DNA oligonucleotides (53-nucleotides long), featuring upstream and downstream duplexes, a non-complementary stretch of 15 nucleotides to generate a synthetic transcription bubble, and a 9-mer RNA complementary to the template strand to form a DNA-RNA hybrid (Fig. S1A). The number of non-complementary bases used in the design of the bubble was based on crystal structures of partial Pol II transcribing complexes including PDB:IDs 1Y1W and 2NVZ (Kettenberger et al., 2004; Wang et al., 2006). These structures show at the downstream end, base complementarity at positions i+3 and i+5, respectively (where i+1 indicates the nucleotide addition site and i-1 the first base of the nascent RNA transcript), and at the upstream end, a partial template strand reaches position i-9 below arch residues (comprising rudder (Rpb1^312–319) and fork loop 1 (FL1, Rpb2^470–480)) (Fig. S1B). However, steric clashes with arch residues at this position suggested that at least two additional nucleotides are required to allow template and non-template strand annealing. Collectively, these observations suggested an artificial bubble size with a minimum of 14 nucleotides for in vitro structural studies of a transcribing Pol II.

Initially, crystals of Pol II bound to our scaffolds showed weak electron density for the upstream duplex but none for the non-template strand (Fig. S1C). In search for factors that could contribute to a stabilized transcription bubble, we assembled Pol II or 10-subunit Pol II (lacking Rpb4 and Rpb7 subunits, Δ4/7) transcribing complexes with TFIIF or its 45 kDa β-subunit, Tfg2 (Fig. S1D,E). Two sets of crystals were obtained using PEG 4000 and low salt (Table 1). The first transcribing complex comprises Δ4/7-TFIIF (heretofore referred as Δ4/7-TC for simplicity) and the second comprises Pol II-Tfg2 (heretofore referred as Pol II-TC). Structures were solved by molecular replacement in Phaser (McCoy et al., 2007) using Δ4/7 or Pol II as search models (see experimental procedures). An initial unbiased F_obs-F_calc map revealed the presence of extra density corresponding to upstream dsDNA, the non-template strand (Figs. 1A) and three previously disordered regions of Rpb2 (Fig. S2A,B). Unfortunately, any additional density for Tfg2 or TFIIF was non-interpretable due to either high mobility or partial occupancy in the crystals. Nevertheless, the presence of TFIIF within the crystals (Fig. S1F) was essential to reveal the full structure of the transcription bubble in our complexes. Interestingly, low resolution cryo-EM data of the pre-initiation complex suggests a role of TFIIF, or the Tfg2 winged-helix domain specifically, in stabilizing the upstream duplex, which could not be visualized prior to TFIIF addition within these structures (He et al., 2013; Muhlbacher et al., 2014). While strikingly similar, without clear extra density, we can only speculate that TFIIF or Tfg2 is acting in an analogous fashion within our structures.

Table 1.

Crystallographic data and refinement statistics

	Pol II-Scaffold 1 (i+5)a (APS-GM/CA)	Δ4/7 –TC (i+5) (APS-GM/CA)	Pol II-TC (i+5) (SSRL-11.2)	Pol II - Tfg 2 – Scaffold 2 (i+2) (SSRL-11.2)	Pol II-Scaffold 3 (i+3) (APS-GM/CA)
Data Collection
PDB:ID	5C44	5C4J	5C4X	5C4A	5C3E
Nucleic Acid	Scaffold 1	Scaffold 1	Scaffold 1	Scaffold 2	Scaffold 3
Space Group	C2221	C21	C2221	C2221	C2221
Unit cell (Å)	220.2 391.8 282.3	280.7 223.3 156.4	220.7 393.3 281.6	219.8 396.7 273.6	219 390.9 278
α, β, γ (°)	90 90 90	90 98.1 90	90 90 90	90 90 90	90 90 90
Wavelength (Å)	1.03	0.979	0.979	0.978	1.03
Resolution (Å)	120–3.9	174–4	50–4	200–4.2	178–3.7
Unique Reflections	106,511	80,485	103,153	80,062	126,815
Completeness (%)	96 (93.8)	97(96.1)	97.3(96.4)	99(98.3)	94.88(78)
Redundancy	4.3 (3.2)	5.1(4.3)	3.8(3.4)	4.5(3.7)	3.6(2.8)
<I/σI>	8.2(1.2)	10.3(3.7)	12.5(1.95)	14.2(2.1)	8.5(1.1)
Mosaicity (°)	0.6	0.6	0.35	0.55	0.35
Rmerge (%)	13.8(49)	14.7(61)	9.1(42)	8.5/(47)	12.5(56)
Refinement
No. Atoms	33086	30259	33673	32281	33086
Rcryst/Rfree (%)	22.3/25.4	21/26	21.5/23.2	23.53/27.47	23.2/27.62
Refinement Program	Buster/CNS	Refmac/CNS	Buster/CNS	Refmac/CNS	Refmac/CNS

^aIndicates the position of complementary base pairing in the downstream bubble.

^bNumbers in parentheses correspond to the highest resolution shell

^cResolution limits were extended to include weak intensity data (Karplus and Diederichs, 2012). Using the traditional criterion of I/σI > 2.0, resolution limits are 4.15 Å and 3.9 Å for Pol II-Scaffold 1 and Pol II-Scaffold 3 complexes, respectively.

Figure 1. Architecture of the complete nucleic acid scaffold.

A. Difference F_obs-F_calc electron density map contoured at 2σ. The following color scheme will be used throughout: cyan, template strand; green, non-template strand; red, RNA transcript.

B. Front and side views of the 2F_obs-F_calc map for the final refined map contoured at 1.0σ.

C. Cartoon representation of the 38-nucleotide refined nucleic acid scaffold; downstream and upstream duplexes form an angle of approximately 130° degrees.

D. Surface representation of the overlay between Δ4/7-TC (blue) and Pol II-TC (grey). The two structures overlay remarkably well, minor structural differences occur at the downstream fork (see also Fig. S1).

The molecular replacement models were refined using the programs Buster (Blanc et al., 2004), Refmac (Murshudov et al., 1997) and manual building with B-factor sharpening in Coot, which clarified side chain positioning, thus allowing model refinement (Fig. S1G and Table 1) (Emsley et al., 2010). The nucleic acid scaffold was built into the electron density using the characteristic features of the DNA-RNA hybrid as register (Fig. 1A). The final refined 2F_obs-F_calc map for Δ4/7-TC is illustrated in Figure 1B. The full observable DNA scaffold (38 nucleotides long) spans the length of Pol II and comprises: the downstream duplex, the DNA-RNA hybrid and two previously uncharacterized regions, an upstream duplex and the full transcription bubble including the non-template strand (Fig. 1B,C). Overlay between the DNA scaffolds from Δ4/7-TC and Pol II-TC, show minor differences mainly located at the downstream fork (Fig. 1D). Pol II regions involved in DNA binding include: 1) The previously described Rpb5 “jaw” and Rpb1 “clamp” residues that interact with the downstream duplex (Gnatt et al., 2001); 2) Rpb2 “wedge” residues (Rpb2^862–874) that interact with the upstream duplex; and 3) “arch” residues that interact with the upstream fork (Fig. 2A,B). For clarity, the presentation of our structural findings is based on the 10-subunit Pol II (Δ4/7) transcribing complex comprising TFIIF and scaffold 1, which has been labeled as Δ4/7-TC unless otherwise noted.

Figure 2. Overall structure of Δ4,7-TC.

A. Surface representation (side view, Rpb2 removed) illustrating the position of the scaffold inside Pol II. Wedge, jaw and arch interactions with the scaffold lie on almost a perfect plane, possibly to minimize strain during elongation.

B,C. The scaffold binds asymmetrically inside Pol II’s cleft, more prominently at the downstream end where observed clamp-DNA distances of ≈6 Å vs. lobe DNA distances of ≈14 Å are due to interactions with clamp and jaw residues (see also Fig. S2).

D. Surface electrostatic representation calculated using the APBS (Baker et al., 2001) suite in PyMOL to illustrate how the non-template strand follows a path of positively charged residues inside Pol II’s cleft (lobe and protrusion). The final refined 2F_obs-F_calc map (grey) contoured at 1.0σ is also illustrated to show the continuous density for the DNA scaffold.

Pol II interacts with the minor groove of upstream and downstream DNA duplexes using two domains located 90 Å apart

Our refined models show that Pol II interacts with the upstream duplex, which appears as if propped by a hairpin loop or “wedge” (Rpb2^862–874) –a vertical extension of wall residues (Rpb2^855–861)– that engages the minor groove of the double helix (Fig. 2A). At the tip of the wedge, Met⁸⁶⁸ lies between template and non-template strands, while the amide backbone of Gly⁸⁶⁷ appears to form hydrogen bonds (H-bond) with two contiguous phosphates on the non-template strand (Fig. 3A). No crystal contacts that could potentially stabilize the conformation of wedge residues or the upstream duplex were present in the two different crystal forms (Fig. S3A,B). The framework of the wedge -a hairpin loop within a long concave five-strand β-sheet–is conserved in all multi subunit DNA-directed RNA polymerases (Fig. 3B). Moreover, overlay of the human mitochondrial polymerase elongation complex (mtRNAP) (Schwinghammer et al., 2013) and our structure shows that the upstream duplex and fork adopt similar conformations (Fig. 3C). This is particularly interesting given the scarcity of conserved structural elements between the two structures (Fig S3C). Furthermore, overlay of the archaeal RNAP clamp domain in complex with the heterodimer interface of the Spt4/5 complex (PDB:ID 3QQC, (Martinez-Rucobo et al., 2011)) with the Rpb1 clamp domain of Δ4/7-TC shows that the upstream duplex is situated between the wedge domain and position of the Spt5 NusG domain (Fig. S3E). While speculative, the observed location of the duplex in our structure is consistent with previous biochemical data and Pol II - Spt4/5 elongation models (Klein et al., 2011; Martinez-Rucobo et al., 2011; Sevostyanova and Artsimovitch, 2010).

Figure 3. Rpb2 Wedge residues: structure, conservation, and function.

A. Interaction of wedge residues with the minor groove of the upstream duplex. Sequence conservation of tip residues across species is nearly universal for Gly⁸⁶⁷ whose amide-bond interacts with the phosphate chain of the non-template strand; Met⁸⁶⁸ is conserved in yeast (S. cerevisiae and S. pombe) and is substituted by a bulky hydrophobic residue in other species. A refined 2F_obs-F_calc map contoured at 1.0σ is also illustrated.

B. Conservation of wedge structure: yellow, T. thermophilus (Vassylyev et al., 2007); hot-pink, archaea (S. sulfactarius) (Hirata et al., 2008); sand, S. cerevisiae; blue, S. pombe (Spahr et al., 2009); and purple, 14-subunit Pol I (S. cerevisiae) (Engel et al., 2013).

C. Overlay of mt-RNAP-upstream duplex (PDB:ID 4BOC) with our structure about template strand i+1 and i+2 (see also Fig. S3).

D. In vitro elongation rate of wt Pol II and an rpb2 wedge deletion mutant (K864G/K865G/Δ866–871). Elongation rate determined on nucleic acid scaffolds at a number of NTP concentrations followed by non-linear regression of the rates for determination of maximum elongation rates (bar graph, error bars indicate range of 95% confidence interval).

E. In vivo apparent elongation rates for wt Pol II and the rpb2 wedge deletion mutant at a galactose inducible reporter gene determine by chromatin IP upon glucose shutoff of transcription (schematic of reporter in F). Values are normalized to 0 minutes of glucose and error bars represent standard deviation of the mean for three independent experiments.

F. Steady state occupancy for wt Pol II and the rpb2 wedge deletion mutant at a galactose inducible reporter gene under galactose induction determined by chromatin IP (schematic of reporter with positions of PCR amplicons shown below) (n=3 independent experiments).

G. Steady state RNA levels of reporter used in F for wt Pol II and the rpb2 wedge deletion mutant. Values were normalized to SCR1 levels (a Pol III transcript) and averaged (n=3) with error bars representing standard deviation.

H. Primer extension analysis for ADH1 transcripts in rpb2 wedge alleles (left) with quantification on right showing average change in fraction of ADH1 starts in various positions relative to wild type, with error bars representing standard deviation of the mean (n=3).

To assess a role of Rpb2 wedge residues during in vivo transcription, we constructed a number of alleles and characterized them for growth phenotypes consistent with transcription defects (Fig. S3D, left panel). We observed sensitivity to mycophenolic acid (MPA), which can be indicative of IMD2 transcriptional phenotypes (Kaplan, 2013). Indeed we found that Rpb2 wedge alleles were generally defective for induction of IMD2 gene expression in the presence of MPA (Fig. S3D, right panel). In order to more directly assess the role of the Rpb2 wedge, we examined the K864G/K865G/Δ866–871 rpb2 wedge allele for in vitro or in vivo elongation phenotypes (Fig. 3D,E). We found that this particular rpb2 wedge allele did not confer robust elongation defects in vitro or in vivo, though it did confer very strong defects in steady state Pol II reporter gene occupancy and expression (Fig. 3F,G). The defects observed are consistent with strong defects in initiation, and we observed altered transcription start site selection consistent with altered initiation in vivo (Fig. 3H).

Pol II contacts with the downstream duplex encompass Rpb5 (jaw) and Rpb1 (clamp-head) domains (Fig. 2B,C). Specifically, Rpb5 jaw residue Pro¹¹⁸ (from helix 118–124) is positioned inside the minor groove of the DNA double helix and Thr¹¹⁷ (from loop 112–117) and Ser¹¹⁹ locate within H-bond distance to non-template strand positions i+15 and i+16 (Fig. S2C, also observed in PDB:IDs 1R9T, 2NVQ and 2NVZ). Rpb1 clamp-head residues Lys¹⁰⁰, Lys¹⁰¹, Lys¹⁴³, and Arg¹⁷⁵ locate within H-bond distance of the phosphate chain of non-template strand positions i+8 to i+10 (Fig. S2D). As a result of these interactions, the downstream duplex is asymmetrically positioned inside the cleft (Fig. 2B,C). Interestingly, contacts resembling clamp-head interactions in T. thermophilus polymerase, are located on a helix-loop-helix domain in its β’ subunit (Fig. S2E) forming H-bonds with non-template strand i+8 to i+9. However, interactions with jaw residues are observed only in archaeal and eukaryotic polymerases, since bacterial polymerases lack Rpb5 homologues.

Strand annealing and 3-way coordination of the template strand, non-template strand and RNA at the upstream fork

The transcription bubble in our complexes lies within upstream and downstream duplexes and was enforced by non-complementarity between the template and non-template strands at positions i+5 through i-10 (Fig. S1A, scaffold 1). At the upstream “closing” end of the bubble, rudder and fork-loop 1 (FL1) residues come in close proximity to form an “arch” located 25–30 Å above the bridge helix (Fig. 2A). The arch is situated in between the template and non-template strands, and physically marks the upstream boundary of the bubble (Fig. 4). Arch residues adopt unique conformations that allow simultaneous coordination of the nucleic acid scaffold (Fig. 4A). As the template strand separates from the RNA transcript (i-8) and emerges from Pol II’s active site, non-specific packing interactions with FL1 residues and potential salt bridges between the phosphate chain and arch residues guide the template strand in a straight conformation towards its junction with the non-template strand (Fig. 4A,B and S4A.). Once above the arch, template strand nucleotides anneal with the non-template strand at i-12 (Fig. 4B,C). Stabilization of the nucleic acid scaffold by arch residues include, interactions between template strand i-8, i-9 and i-11 with FL1 residues, and between non-template strand i-11 and i-12 with rudder residue Lys³¹⁷ (Figs. 4 and S4B) (Treutlein et al., 2012). In addition to contacts that arch residues make with the template and non-template strands, rudder residue Arg³²⁰ reaches within H-bond distance of the 2´ hydroxyl of the 8^th RNA base of the nascent transcript (Figs. 4A and S4B, observed also in PDB:ID 1Y1W). This interaction is conserved in bacterial polymerases, where rudder Arg⁵⁹⁸ forms a H-bond with the 2´ hydroxyl group of the 7^th base on the nascent transcript (Fig. S4C, (Kashkina et al., 2007)) and has been corroborated by cross-linking experiments (Korzheva et al., 2000).

Figure 4. Architecture of the upstream fork junction.

A. Stereo-view of the tripartite coordination of the three nucleic acid chains. Arch residues at the back, top and bottom adopt unique conformations –with respect to apo- and elongation structures– that interact with template, non-template and RNA strands respectively. Rpb1 residue Arg³²⁰ forms a H-bond with the 2′-OH of the nascent transcript at position i-8.

B,C. Front view (B) and side view (C) of the upstream (closing) end of the bubble. Rudder (silver) and FL1 (sand) residues reach within 4 Å across the midline to form an arch that provides a scaffold for template and non-template strand annealing at i-12. Contacts include packing interactions between template strand i-8 and Tyr⁴⁵⁹ and potential H-bonds between Rpb2 residues Thr⁴⁶³ with template strand i-8, Glu⁴⁶⁹ with template strand i-10 and Lys⁴⁷¹ with non-template strand i-9. Lys³¹⁷ participates in contacts with non-template strand i-11 and i-12 (see also Fig. S4)

Strand separation at the downstream transcriptional fork

Strand separation in our structure appears to take place at i+5 where base-pair distance begins to increase progressively. However, it was not possible to define its exact location since our artificial bubble enforced non-complementarity precisely at i+5 (see below). However, our structure gives possible insight into the mechanism by which Pol II residues promote and sustain DNA strand separation. The template strand interacts with switch 1 and switch 2 residues driving it towards the active site as originally described (Gnatt et al., 2001), while the non-template strand interacts with two groups of Pol II residues. The first involves a positively charged cleft formed by Rpb1 residues Arg¹³⁸⁶-His¹³⁸⁷-Arg¹³⁹¹ (switch-1), and Lys¹¹⁰²-Lys¹¹⁰⁹-Asn¹¹¹⁰, (located in a structurally conserved U-loop) in the vicinity of non-template strand i+5 and i+6 (Fig. 5A and S5A, (Cheung et al., 2011; Kettenberger et al., 2004)). The second involves interactions with Rpb2 fork loop 2 (FL2) residues 501–510 (Fig. 5A) providing packing contacts with the non-template strand i+3 to i+1, and Arg⁵⁰⁸ reaching within H-bond distance of the non-template phosphate chain (Figs. 5B and S5B). Structural overlay of FL2 residues from published crystal structures suggests they could be grouped in two major states (Fig. 5C). The first one, an “open” state represented by PDB:IDs 3PO2, 1Y1W and 3HOW, where FL2 residues interact with positions i+2 or i+3 on the non-template strand, respectively, allowing access to a non-specific nucleotide-binding pocket (Cheung and Cramer, 2011). The second, a closed state where FL2 residues appear to rotate about Pro⁵⁰¹ and Pro⁵¹⁰ (our structures, PDB:IDs 3FKI and 3K7A) blocking access to the pocket. Positioning of FL2 residues in the latter conformation appear to guide the non-template strand under a 4-strand β-sheet dome (Fig. 5A) towards its junction above the arch. Other interactions include contacts with Rpb2 β-strand 245–255 residues (Fig. S5B) and a patch of positive charges from the Rpb2 protrusion helix (Fig. 2C).

Figure 5. Architecture of the downstream fork junction.

A. Stereo-view of the architecture of the downstream fork. Relevant interactions include U-loop residues with non-template strand i+5 and i+6, FL2 residues with non-template strand i+2 to i+3, and Rpb2 residues 221–282 (forming a 5 strand β-sheet, Dome) with non-template strand i+1 to i-2.

B. Unbiased F_obs-F_calc electron density map contoured at 3σ (FL2 residues 498–512 were not included in map calculation).

C. Structural overlay of published FL2 conformations during different stages of transcription. Δ4/7-TC (sand), PDB:ID 1Y1W (blue, elongation complex), PDB:ID 3HOW (cyan, backtracked complex), and PDB:ID 3PO2 (red, backtracked complex).

D. Representative spectra of 2-AP probes at i+2 or i+3, bound to complementary non-template strand in the absence (open red circles and squares, respectively) and presence (blue circles and squares, respectively) of Pol II. The excitation wavelength was 315 nm and the fluorescence emission (shown in counts per second ×10⁶) was collected from 340–400 nm.

E. Normalized fluorescence values for polymerase bound to ssDNA (primer-template) where 2AP is at the i+2, i+3, i+5 or i+8 position (red) and polymerase bound to dsDNA (primer-template annealed to fully complementary non-template strand) (blue). Error bars are standard deviation of the mean (n=3, see also Fig. S5).

Fluorescence and structural experiments suggest that the downstream fork is dynamic

Given that the precise location of strand separation was not possible to be determined from the structure –and in light of previous structural studies showing a closed bubble at i+2 and i+3, respectively (Cheung et al., 2011; Kettenberger et al., 2004), as well as fluorescence studies where the use of a 35 nucleotide scaffold with a partial (17 nucleotides) non-template strand showed complementarity at i+2 (Kashkina et al., 2007)– we wished to ascertain whether it was possible to detect an open bubble in the presence of a 45 nucleotide scaffold bearing a fully complementary non-template strand (Fig. S5C). 2-aminopurine (2AP) is a fluorescent nucleotide analog that is significantly quenched upon base pairing to either T or C, as well as by stacking interactions with adjacent nucleotides (Liu and Martin, 2001; Stivers, 1998). Therefore, we placed 2-AP in the template strand at positions i+2, i+3, i+5 (at the boundary of strand separation) and downstream at i+8, where the two DNA strands were likely to be paired (Fig. S5C). As expected when the primer-template is annealed to a complementary non-template strand, full quenching is observed consistent with stable base pairing of 2AP (Fig. 5D,E). When polymerase was added to the scaffold with a fully complementary non-template strand, we see a 4.5-fold increase in fluorescence for the i+2 substrate and a 2.8-fold increase in fluorescence for the i+3 (Fig. 5D,E). This increase in fluorescence indicates a disruption in stable 2AP base pairing at both locations and could suggest that Pol II can unwind the non-template strand at i+3. There is 1.2-fold fluorescence enhancement for i+5 and none for i+8 suggesting 2AP is more stably base paired at those positions (Fig. 5E). Moreover, crystal structures of Pol II bound to scaffolds 2 and 3 bearing non-complementary transcription bubbles with strand separation at i+2 and i+3, respectively, both showed base-pairing at i+5, recapitulating the results of 2-AP experiments (Fig. S5D,E).

To observe a full non-template strand it was necessary to utilize a bubble with non-complementary DNA to i+5, since Pol II transcribing complexes bearing non-complementary bubbles at positions i+2 and i+3 showed minimal non-template strand density (scaffolds 2 & 3, Fig. S1A and S5D,E). Thus, we analyzed whether forced non-complementarity at i+5 in the transcribing complex contributed to increased mobility of the non-template strand and its subsequent capture in our scaffold 1 complex. Consistent with this possibility, 2AP experiments where complementary base pairing begins at positions i+3 and i+5 showed an increased fluorescence signal when compared to bases where complementarity begins at the i+1 position (Fig. S5F,G). This might indicate increased motion of the non-template strand at the leading edge of the artificial transcription bubble bearing non-complementary bases. Thus, it is possible that the use of a scaffold with non-complementary base pairing at i+5 resulted in a kinked conformation of the non-template strand in Pol II-TC structure (Fig. S5H). It is also possible, while highly speculative, that the presence of Tfg2 (without Tfg1) in the cleft induced such conformation.

Interactions with the non-template strand are associated with the off (non-catalytic state) of the trigger loop

A highly conserved loop comprising Rpb1 residues 1078–1097, the trigger loop (TL), has been shown to play a fundamental role in nucleotide selection and catalysis, while also proposed to govern translocation (Bar-Nahum et al., 2005; Feig and Burton, 2010; Kaplan, 2013; Kaplan et al., 2012; Kireeva et al., 2012; Larson et al., 2012; Wang et al., 2006). Evidence from single molecule studies on Pol II indicates that mutation of TL residues alters Pol II translocation properties, consistent with these models (Larson et al., 2012). Structurally, the TL locates between Rpb1 helices 1064–1078 (TLα₁) and 1097–1106 (TLα₂), which in turn are part of a universally conserved five-helix bundle, (heretofore known as TL bundle (TLB)), that includes Rpb1 helices 826–846 (bridge helix), 1340–1357 (TLα₄) and 1365–1379 (TLα₅) supported by packing of hydrophobic residues at the bundle core (Fig. S6A). The Pol II TL is intrinsically mobile and has only been detected in X-ray structures when bound to small molecules (such as a matched NTP or α-amanitin) or protein-ligands (TFIIS) that each stabilize a particular conformation (Brueckner and Cramer, 2008; Kaplan et al., 2008; Kettenberger et al., 2003; Wang et al., 2006). Observed conformations define a closed “on” state where TL residues interact with a matched nucleotide in the addition or “A” site (effectively isolating a reaction chamber, and blocking access to additional substrates (Feig and Burton, 2010; Kireeva et al., 2012)); or an open “off state where TL residues move away from the A site. During advanced stages of refinement, electron density for the full TL backbone was clearly discernible (Fig. 6A and S6B) and partial residue placement was feasible with help of map sharpening in Coot. TL backbones were found in “off states that differ from previously reported conformations (Kettenberger et al., 2003; Wang et al., 2009; Wang et al., 2006) such as the Pol II-TFIIS complex, where direct contacts with TFIIS displaced and stabilized TL residues (Fig. S6C). Interestingly, our structures show two distinct off state conformations: The first one present in the Δ4/7-TC closely resembles the TL loop in the “on” conformation (PDB:IDs 2E2H and 2NVZ, Wang et al., 2006) but rotated counter-clockwise approximately 60° away from the addition site (Fig. 6A). The second one, observed in Pol II-TC resembles a hairpin loop and its shape is similar to previously reported off state structures (Fig. S6A–C, (Kettenberger et al., 2004; Wang et al., 2006)). Both conformations induce considerable changes in funnel and neighboring residues (Fig. 6A and S6B). Albeit these differences in the positions of the TL’s off state, a common set of interactions can be observed among these structures and all structures crystallized with a DNA scaffold (exclusive of the on-state structures). These include: 1) interactions between TL hinge regions (TLα1 and TLα2 helices) with neighboring Rpb1 residues, and 2) the burying of Met¹⁰⁷⁹ inside a small hydrophobic pocket at the core of the TLB (Fig. 6B).

Figure 6. Trigger loop (TL) and nucleic acid scaffold interactions during translocation.

On-, off- state residues will be indicated in blue and grey, respectively. A modeled UTP (PDB:ID 2NVZ) is indicated in light grey/orange.

A. Conformational changes observed between off and on (structural overlay with PDB:ID 2NVZ) states of the TL and Rpb1 funnel helices. A 2F_obs-F_calc map rendered at 1.0 σ is contoured around TL and funnel residues.

B. Conformational changes observed between TL off and on states. Red arrows indicate motion. During off/on state conformational changes, most TL stabilizing interactions are disrupted, including: 1) Release of Met¹⁰⁷⁹ from its hydrophobic pocket. 2) Disruption of α2-Bridge Helix H-bonds, resulting in bridge helix displacement. 3) Disruption of Thr¹⁰⁹⁵-Thr¹¹¹³ H-bonds allowing counterclockwise TL motion. 4) Disruption of non-template strand - U-loop bonds, possibly leading to non-template strand release and translocation.

C. Mutations of residues that disrupt Met¹⁰⁷⁹ hydrophobic pocket result in gain of function phenotypes (Kaplan et al., 2012). TLB residues are represented as a solid silver surface. Motion of Met¹⁰⁷⁹ might occur through a defined pathway on the protein surface (orange trace). Mutations of Ala¹⁰⁷⁶ and Gly¹⁰⁹⁷ (red surface) for bulkier residues, can potentially disrupt the vestibule of the hydrophobic pocket (Kireeva et al., 2012).

D. Possible coupling of the global translocation of the scaffold to local motion of the TL. Pol II regions in contact with upstream and downstream duplexes, Rpb2 (sand) and Rpb5 (magenta), respectively, are coupled through TLB residues (dark gray). TL off and on conformations are illustrated in yellow and blue, respectively (see also Fig. S6).

Structural overlay between off and on states suggest substantial rearrangements in TLB helices and TL residues during the nucleotide addition cycle (Fig. 6B). Transition to the on state involves TL residues swinging towards the addition site, where hinge contacts are disrupted and Met¹⁰⁷⁹ moves out of its pocket at the core of the TLB (Fig. 6C). Genetic interactions and gene expression profiling of substitutions of Rpb1 Ala¹⁰⁷⁶, Gly¹⁰⁹⁷ and Leu¹¹⁰¹ with residues that disrupt the Rpb1 Met¹⁰⁷⁹ hydrophobic pocket (Braberg et al., 2013), support a model where the integrity of the pocket is critical for stabilization of the off state in eukaryotic (Kaplan et al., 2012; Kaplan et al., 2008) and archaeal (Fouqueau et al., 2013) polymerases. Such substitutions are highly related to those that hamper off state conformations by destabilization of the C-terminal TL hinge region; therefore, these substitutions are similarly predicted to alter translocation rate and catalysis (Kaplan et al., 2012; Kaplan et al., 2008; Kireeva et al., 2012; Malagon et al., 2006; Wang et al., 2006) suggesting that the TL “off state is specifically required for proper transcription.

Structurally linked DNA-interacting domains: Upstream (Rpb2), downstream (Rpb5) and TLB residues could possibly coordinate translocation

Three regions of Pol II furnish residues that can potentially form H-bond contacts with the nucleic acid scaffold (Fig. S6E). These include (from upstream to downstream): 1) wedge, wall and FL1 (Rpb2) with the upstream duplex, template strand and the non-template strand (respectively); 2) TLB, rudder, switch (1 and 2) and clamp residues (Rpb1), with DNA-RNA hybrid, template strand, downstream fork and downstream duplex (respectively); and 3) jaw residues (Rpb5) with the downstream duplex (Fig. 6D and S6E,F). Remarkably, these regions have substantial interactions among them: TLB helices (TLα₄ and TLα₅) and Rpb1 residues from a seven-helix bundle (Rpb1^846–1064) form a large pocket that buries a two-strand hairpin from Rpb5^193–214, which is in the immediate neighborhood of the jaw motif. Similarly, TLB contacts (via the bridge helix) with Rpb2 are extensive and involve “wall” residues (with the phosphate chain of the template strand) in the immediate neighborhood of Rpb2’s wedge (Fig. 3B and 6D). Since these regions are coupled extensively, it is possible that they could play an important role during translocation.

Discussion

Molecular basis for DNA-tracking

Real time microscopy experiments demonstrated the ability of RNA polymerases to rotate DNA by tracking with high fidelity its right hand helix (Harada et al., 2001). Our structures suggest that it is possible that engagement by wedge (upstream duplex), arch (closing end of the bubble) and jaw and clamp (downstream duplex) residues (Figs. 2, S2 and S6F) could comprise the structural framework that explains such tracking mechanism. The interactions that Rpb1 (head clamp) and Rpb5 (jaw) residues make with the downstream duplex were described in several published crystal structures (Wang et al., 2006; Westover et al., 2004). However, since contacts with the upstream duplex and arch were not previously observed, the correlation between DNA tracking, and its structural underpinnings could not be established. Moreover, given that the crystal structures of Pol II transcribing complex and mtRNAP show interactions with upstream and downstream duplexes, it is possible that tracking mechanisms are conserved in transcription (Figs. 3C and S3).

We assessed multiple rpb2 wedge alleles to ascertain the role of these elements in vivo. The strongest growth effects observed in vivo required removal of the loop (Fig. S3D), which may be considered an extensive perturbation to the Pol II structure. However, our interpretation of the direct or indirect functions of this loop is based on what is known about other Pol II mutants. We found that wedge alleles were MPA sensitive and shifted start sites downstream (Figs. 3H and S3D). This profile is relatively rare for Pol II alleles or known general transcription factor alleles (reviewed in (Kaplan, 2013)), especially for mutants unrelated to the active center (Rpb1 N488D within the Pol II active site has this phenotypic profile (Malagon et al., 2006)). Recent crystal structures of Pol II-TFIIB structures show that the wedge domain interacts with the TFIIB core N-terminal cyclin fold (Sainsbury et al., 2012). Since TFIIB is necessary for start site selection, it seems plausible that changes in Pol II-TFIIB interactions could affect this process. Deletion of the wedge loop confers a strong defect in occupancy of Pol II at a reporter gene, consistent with an initiation defect. However, the wedge mutants do not phenocopy TFIIB (sua7) alleles in relation to MPA-sensitivity or IMD2 expression (Fig. S3D), suggesting that some functions may be independent of TFIIB or relate to an initiation defect not observed in particular sua7 alleles. The wedge domain may possibly have a role during transition from initiation to early elongation by providing interactions with the upstream duplex that could assist TFIIB ejection (Fig. S3F).

Downstream fork flexibility

The downstream fork is a dynamic region and different conformations of the non-template strand could be allowed during different stages of transcription. Structural overlay of the non-template strand from our structures and published structures crystallized in the presence of a nucleic acid scaffold show small positional differences from i+5 to i+7, which are anchored by U-loop and switch 2 residues (Fig. S5A). However, positions i+2 to i+4 differ in all crystal structures and can be found paired or un-paired to the template strand. These observations suggest that there may be a range within which the leading edge of the downstream fork can move; hence, fluidity within this region could be required to promote bubble opening and maintenance. Moreover, non-template strand flexibility correlates with FL2 flexibility (Fig. 5C), which adopts multiple conformations (Cheung and Cramer, 2011; Liu et al., 2010; Meyer et al., 2009; Wang et al., 2006) to allow interactions with the non-template strand during the various transcriptional stages. The size of the bubble itself, which appears to be dynamic, might not be so critical due to the fact that Pol II can accommodate different lengths and possibly different locations of the transcription bubble inside the cleft (Pal et al., 2005).

The crystal structure of the ternary complex between T. thermophilus polymerase (RNAP), σ^A and an open promoter complex (Zhang et al., 2012) shows that FL2 residues fold to form part of a 2′-deoxy-GTP “selectivity-pocket” for non-template strand i+2 during initiation. Also, the crystal structure of the RNAP ternary complex revealed that promoter recognition, melting and bubble loading are carried out by σ^A region-2 residues (Fig. S4C). Thus architecture of the downstream fork in RNAP during initiation is determined by σ^A as well as FL2 residues. The recent crystal structure of E.coli RNAP holoenzyme in complex with a 15-nucleotides bubble at 6 Å resolution revealed the architecture of the RNAP transcription bubble stabilized by σ elements (Zuo and Steitz, 2015). However, overlay of RNAP bubble structure with our structures show different trajectories of the non-template strand and also different position of the upstream duplex. Such differences could be explained structurally since, on the one hand FL1 residues (not conserved in RNAP) would clash with the non-template strand of RNAP; and on the other, the presence of σ, which itself interacts with the non-template strand, in the RNAP cleft shifts dramatically the position of the upstream duplex.

Interestingly, overlay of the co-crystal structures of Pol II in complex with TFIIB (Bushnell et al., 2004; Kostrewa et al., 2009; Liu et al., 2010) with the our structure shows that TFIIB linker domain reaches within 6.5 Å of the fork junction and could assist bubble loading (Liu et al., 2010; Sainsbury et al., 2012) and prevent re-annealing of downstream fork during initiation (Fig. S5I). Thus, σ^A and TFIIB could sustain an open fork (during initiation) using different mechanisms.

TL allosteric effects and DNA translocation

It has been proposed by a number of groups that TL movement contributes or controls translocation (Bar-Nahum et al., 2005; Brueckner and Cramer, 2008; Feig and Burton, 2010; Kaplan et al., 2012; Larson et al., 2012). Our structures show Pol II in a post-translocated state, with a TL in the off state due to interactions with neighboring Rpb1 residues (Figs. 6A,B and S6B). It is possible that a matched NTP at the i+1 position might disrupt these interactions leading to an on state confirmation. The on-state structure (Wang et al., 2006) shows Gln¹⁰⁷⁸ as one of the key residues stabilizing a matched nucleotide through formation of H-bonds with the ribose. Structural overlay of TL residues between off and on conformations shows that Gln¹⁰⁷⁸ moves approximately 3 Å to form H-bonds with the matched nucleotide (Fig. 6B, see also (Cheung et al., 2011)). The position of this residue might be critical since on the one hand, it could constitute part of the nucleotide selection mechanism (Fouqueau et al., 2013; Yuzenkova and Zenkin, 2010). On the other hand, displacement of Gln¹⁰⁷⁸ could trigger extraction of the neighboring Met¹⁰⁷⁹ from its hydrophobic pocket, initiating a cascade of events that would lead into the full on state. Along with our observations, genetic evidence has shown that mutations of Gln¹⁰⁷⁸ have a comparable effect on Pol II elongation activity as mutations on the catalytic His¹⁰⁸⁵ (Kaplan et al., 2012). Importantly, Gln¹⁰⁷⁸ and His¹⁰⁸⁵ are genetically distinguishable suggesting a multistep process in TL function, consistent with initial substrate-Gln¹⁰⁷⁸ interactions and subsequent TL movement or folding. Moreover, Pol II activity is also exquisitely sensitive to substitutions around the Met¹⁰⁷⁹ hydrophobic pocket (Fig. 6C) (Kaplan et al., 2012; Kireeva et al., 2012). These substitutions invariably lead to genetic phenotypes, genetic interaction and gene expression profiling phenotypes consistent with increased Pol II activity, most likely due to destabilization of the TL off state (Braberg et al., 2013).

Critical questions for the Pol II mechanism are how translocation occurs and what are the molecular determinants of its linkage to the nucleotide addition cycle? Our structure reveals that a second pivotal role played by TLB residues could possibly include stabilization of the non-template strand. Crystal structures of Pol II bound to partial or full nucleic acid scaffolds (Cheung et al., 2011; Kettenberger et al., 2004; Wang et al., 2006; Westover et al., 2004) show that TLα₂ and UL residues Lys¹¹⁰², Asn¹¹⁰⁶, Lys¹¹⁰⁹ (amide backbone) and Asn¹¹¹⁰ locate within H-bond distance to the phosphate chain of non-template strand i+5 and i+6 (Fig. S5A). Overlay of the two states shows that the hinged motion of the TL during matched nucleotide binding results in conformational changes leading to increased U-loop – non-template strand distance and hence disruption of potential H-bond contacts (Fig. 6B,C). Although speculative this could suggest that TL “off/on” transitions are allosterically coupled to “latch and release” (respectively) events of the non-template strand phosphate chain by U-loop residues. Furthermore, comparisons between on and off state conformations show that on-state Rpb5 jaw residues (interacting with the downstream duplex) become disordered, hence decreasing the number of effective contacts with the duplex (Fig. S6D). Since TLB, Rpb5 (jaw residues) and Rpb2 (wall, wedge) regions could be coupled, through extensive observed interactions, it is possible that TL conformational changes are transferred allosterically to downstream and upstream duplexes to assist global translocation (Figs. 6D and S6E,F).

Our refined structure sheds light into four fundamental mechanistic aspects of transcription: 1) In addition to hybrid interactions (Gnatt et al., 2001), Pol II has four major contact points with a nucleic acid scaffold: wedge, arch, clamp and jaw domains (Fig. 2A,B). Engagement of the minor groove by Rpb2 wedge (upstream), arch (closing end of the bubble) jaw and clamp (downstream) residues could comprise the structural framework that explains the mechanism for high fidelity DNA “tracking” observed using real-time optical microscopy (Harada et al., 2001). 2) Pol II residues define the architecture of the transcriptional fork. Arch residues coordinate annealing of template and non-template strands at the upstream fork; the downstream fork is a highly dynamic area where FL2 residues accommodate different conformations of the non-template strand. 3) The structure also suggests that Gln¹⁰⁷⁸ is positioned to couple extraction of Met¹⁰⁷⁹ out of the hydrophobic pocket to interactions with an incoming matched NTP, initiating a cascade of events leading to a full on state of the TL, followed by nucleotide incorporation and subsequent DNA/RNA translocation. 4) Finally, our structure shows that Pol II regions in contact with the nucleic acid scaffold are connected as rigid bodies (from Rpb1, Rpb2 and Rpb5) and that TL on/off state conformational changes could possibly be tied to global translocation.

Experimental Procedures

Pol II transcribing complex purification and assembly

Saccharomyces cerevisiae Pol II and Tfg2 were purified as previously described (Chung et al., 2003; Wang et al., 2006). To assemble a Pol II transcribing complex, equimolar concentrations of oligonucleotides containing a single stretch of non-complementary bases and a 9-mer RNA were annealed. The resulting nucleic acid scaffolds were mixed with Pol II (3:1 molar ratio) and excess scaffold was removed using size exclusion chromatography (Superdex200, GE LifeSciences) against Buffer A (25 mM Hepes pH 7.5, 100 mM KCl, 5 mM DTT, 0.5 mM EDTA, 10 µM ZnCl₂). Purification and assembly of Pol II-TC and Δ4/7-TC was achieved as previously described (Pullara et al., 2013). An SDS-PAGE of the final complex is illustrated in Fig. S1E. Ethidium bromide staining confirmed the presence of the nucleic acid scaffold (Fig. S1E). Catalytic activities in the presence of TFIIF or Tfg2 were not determined.

In vitro elongation assay

Pol II enzymes for in vitro assays were purified from yeast strains expressing wild type or the mutant rpb2 gene from a low copy plasmid, as described above. In vitro elongation assays were performed as described in (Kaplan et al., 2012; Kaplan et al., 2008) with minor modifications in the amount of nucleic acids and Pol II used for elongation complexes (extended experimental section).

Chromatin Immunoprecipitation (ChIP) assays

Epitope tagged (RPB3::3XFLAG::kanMX) wild type or mutant strain containing a galactose inducible YLR454w reporter (kanMX::GAL1p::YLR454w) gene were used for ChIP assays. Chromatin immunoprecipitaion experiments for in vivo elongation rate determination were performed as described previously in (Hazelbaker et al., 2013), with slight modifications (extended experimental section).

Northern blotting and primer extension analysis

Northern blotting was performed as previously described (Kaplan, et al., 2012) essentially following the instructions of GeneScreen hybridization membranes (Perkin-Elmer) with minor modifications (extended experimental section).

2 – aminopurine fluorescence spectroscopy

Steady-state fluorescence measurements for Pol II-TCs were performed as previously described (Kashkina et al., 2007; Liu and Martin, 2001) with minor modifications using a Fluoromax-3 (HORIBA Scientific). Excitation wavelengths included both 280 and 315 nm, and fluorescence emission was collected from 340–400 nm. In addition to Pol II-TCs, spectra were collected for buffer, DNA and Pol II alone to correct for background fluorescence as outlined in the extended experimental section. All measurements were performed at 25 °C in triplicate (n=3).

Crystallization and Refinement

Pol II transcribing complexes were concentrated to 8–10 mgs/ml and crystallization trials produced crystals in several conditions (extended experimental section) and were verified by SDS-PAGE analysis of crystals (Fig. S1F). The structures were solved by molecular replacement using 12-subunit Pol II PDB:ID 3FKI (Meyer et al., 2009) for Pol II-scaffolds in the presence or absence of Tfg2 and 10-subunit Pol II PDB:ID 1SFO (Westover et al., 2004) for Δ4/7-TC in Phaser (McCoy et al., 2007). Molecular replacement models were refined using the program Buster (Blanc et al., 2004), Refmac (Murshudov et al., 1997) and CNS (Brunger et al., 1998) followed by several cycles of manual building with B-factor sharpening in Coot (Emsley et al., 2010). Inclusion of weak-intensity, high-resolution data improved refinement behavior and stereochemistry (Karplus and Diederichs, 2012). All figures were rendered using PyMOL (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC).