A non-native C-terminal extension of the β’ subunit compromises RNA polymerase and Rho functions

SUMMARY

Escherichia coli RfaH abrogates Rho-mediated polarity in lipopolysaccharide core biosynthesis operons, and ΔrfaH cells are hypersensitive to antibiotics, bile salts, and detergents. Selection for rfaH suppressors that restore growth on SDS identified a temperature-sensitive mutant in which 46 C-terminal residues of the RNA polymerase (RNAP) β’ subunit are replaced with 23 residues carrying a net positive charge. Based on similarity to rpoC397, which confers a temperature-sensitive phenotype and resistance to bacteriophages, we named this mutant rpoC397*. We show that SDS resistance depends on a single nonpolar residue within the C397* tail, whereas basic residues are dispensable. In line with its mimicry of RfaH, C397* RNAP is resistant to Rho but responds to pause signals, NusA, and NusG in vitro similarly to the wild-type enzyme and binds to Rho and Nus factors in vivo. Strikingly, the deletion of rpoZ, which encodes the ω “chaperone” subunit, restores rpoC397* growth at 42 °C but has no effect on SDS sensitivity. Our results suggest that the C397* tail traps the ω subunit in an inhibitory state through direct contacts and hinders Rho-dependent termination through long-range interactions. We propose that the dynamic and hypervariable β’•ω module controls RNA synthesis in response to niche-specific signals.

Graphical Abstract

The largest subunits of RNA polymerase, β’ in bacteria, have dynamic C-terminal tails. We show that a frameshift mutation that adds a short non-native tail to the Escherichia coli β’ confers temperature-sensitive growth and impaired Rho-mediated polarity. The loss of the ω subunit suppresses the ts defect but does not restore termination. We hypothesize that alternative, context-dependent tail conformers contact distinct sites on RNA polymerase and accessory factors to modulate RNA synthesis.

INTRODUCTION

In all living cells, RNA synthesis is carried out by multi-subunit RNA polymerases (RNAPs) that share an evolutionary origin, a crab-claw 3D architecture, and several structural domains that mediate RNAP interactions with the nucleic acids and catalysis (Werner, 2012). The evolutionary relationship among the largest RNAP subunits that jointly form the active site (β’ and β in bacteria, Rpb1 and Rpb2 in eukaryotes) is reflected in 15 conserved regions (A–I in β and A–H in β’) implicated in key functions common to all RNAPs (Lane & Darst, 2010). Beyond these signature motifs, RNAP sequences, and even subunit composition, have diverged considerably. While RNAPs from Escherichia coli and many other bacteria have five “core” subunits (α₂ββ’ω), the Bacillus subtilis enzyme has seven (α₂ββ’ωδε) (Miller et al., 2021). Furthermore, core subunits contain many phylum-specific sequence insertions (SIs) (Lane & Darst, 2010), which have been shown to modulate every step of the transcription cycle. For example, in E. coli, β SI2 makes functional interactions with the termination factor Rho (Hao et al., 2021, Said et al., 2021), whereas β’ SI3 modulates RNAP responses to pause and termination signals, the initiation factor DksA, and the elongation factor RfaH (Ray-Soni et al., 2017, Furman et al., 2013, Svetlov et al., 2007).

Among these variable regions none is more renowned than the C-terminal domain (CTD) of Rpb1, the largest subunit of RNAP II in eukaryotes, which is connected to region H via a linker region (Fig. 1). The CTD is dispensable for basic catalysis but is required for transcription on chromatin templates, RNA splicing and processing, binding of numerous accessory factors, etc. (Hsin & Manley, 2012). The CTD is composed of several heptad repeats, as few as five or as many 52, that vary in sequence among different organisms, and may have arisen independently during evolution (Chapman et al., 2008). In bacteria, the β’ C-termini are also divergent in length and sequence (Fig. 1B), but do they have any function? Observations that C-terminal fusions to the E. coli β’, such as small purification tags or large proteins, do not alter transcription patterns or viability (Bakshi et al., 2012, Mooney & Landick, 2003, Kashlev et al., 1993), suggest that the β’ C-terminal regions are dispensable, at least under favorable conditions.

Fig. 1. The C-terminal tail of the largest RNAP subunit.

A. Sequence logo of Proteobacteria is mapped to the E. coli wild type (WT) sequence (NP_418415.1). The WT and frameshift alleles are shown below.

B. Tail lengths vary significantly within domains. H, region H; CTD, C-terminal domain of RNAP II. NCBI ID: H. sapiens, NP_000928.1; M. bicuspidate, XP_018711502.1; S. solfataricus, WP_009990475.1; S. marinus, WP_011839096.1; R. bellii, WP_011477811.1; E. coli, NP_418415.1; S. heliotrinireducens, WP_012799105.1.

C. The cryo-EM structure of the E. coli pre-termination complex bound to NusA, NusG, and Rho; PDB ID: 6Z9P. In this and other figures, ω is colored in magenta and other RNAP subunits - in shades of gray; the β’ C-tail is shown in cyan and NusA – in slate.

We became interested in the β’ tail when we isolated a frameshift variant of E. coli rpoC, a gene that encodes the β’ subunit, using a genetic selection of suppressors of ΔrfaH (Hu & Artsimovitch, 2017). RfaH, a non-essential paralog of the housekeeping NusG, inhibits Rho-dependent termination during transcription of long horizontally acquired operons that encode capsules, F-pilus, and cell wall biosynthesis functions (Tomar & Artsimovitch, 2013). A deletion of rfaH or a polar insertion in the lipopolysaccharide (LPS) core biosynthesis waa operon (Hu & Artsimovitch, 2017, Moller et al., 2003) permits premature transcription termination and prevents growth in the presence of sodium dodecyl sulfate (SDS). Among rfaH suppressors that grew on SDS, most mapped to rho and to hns (Hu & Artsimovitch, 2017), a gene that encodes histone-like nucleoid-structuring (H-NS) protein. Since both H-NS and Rho silence the expression of xenogenes (Lawson & Berger, 2019, Navarre, 2016), loss-of-function suppressors (partial in the case of rho, which is essential in E. coli) could be expected. We also identified an SDS-resistant allele of rpoC, a frameshift that replaces 46 C-terminal amino acids with a 23-residue tail enriched in positively charged residues (Fig. 1A), suggesting that the altered β’ tail may confer resistance to Rho.

A similar frameshift has been isolated as a survivor of λ cI857, a temperature-sensitive (ts) prophage, at non-permissive temperature (Isaksson et al., 1977). This variant, named rpoC397 (Christie et al., 1996), encodes an identical 23-residue extension but removes additional six native amino acids (Fig. 1A). The rpoC397 strain is defective in RNA synthesis at 43.5 °C, explaining its resistance to λ (Christie et al., 1996), and is also resistant to bacteriophages N4 and P2 (Christie et al., 1996, Zivin et al., 1981). While resistance to N4 is due to the loss of direct interactions with the phage transcription factor (Miller et al., 1997), defects in P2 have been hypothesized to result from changes in the cell envelope structure that affect P2-mediated lysis indirectly (Markov et al., 2004).

A termination defect would readily explain the SDS-resistant phenotype of the frameshift mutant (Hu & Artsimovitch, 2017). In structures of E. coli pre-termination complexes (Fig. 1C), the C-terminus of β’ is located far from Rho and from NusA and NusG, which are required for Rho-mediated termination (Hao et al., 2021, Said et al., 2021). However, the β’ end proximity to the ω subunit, which modulates termination in vitro (Said et al., 2021) and RNAP distribution genome-wide in vivo (Yamamoto et al., 2018), suggests that β’ extensions may alter the conformation of ω or its interactions with other RNAP subunits.

Here, we characterized the new frameshift variant, which we named rpoC397* to preserve the historical context. We show that the 23-residue extension, rather than the β’ truncation, is required for growth on SDS. Mutagenesis of the tail region revealed that a single hydrophobic residue is critical for SDS resistance whereas all eye-catching basic residues are dispensable. We also show that the β’ C397* extension confers resistance to Rho in vivo and in vitro but does not compromise transcriptional pausing or binding to NusA and NusG.

Our results reveal complex genetic interactions between rpoC397* and rpoZ. On the one hand, in contrast to the well-documented ability of ω to correct the ts defects of other rpoC alleles (Minakhin et al., 2001), deletion of the rpoZ gene (that encodes ω) restored the rpoC397* growth at 42 °C. On the other hand, the ability to grow on SDS was independent of the ω subunit. We propose that distinct conformations of the non-native β’ C-terminal tail underlie its genetically separable effects on RNAP. We further speculate that hypervariable β’ tails control RNA synthesis in response to signals that are encountered by different bacterial species in their natural habitats. Solute- and protein-induced changes in the β’ tail conformation can be sensed by distal RNAP elements and by the adjacent ω subunit. The dominant-negative effects of ω reported by us and others (Sarkar et al., 2013) support a notion that ω, widely regarded as a resident RNAP chaperone, can directly control RNA synthesis.

RESULTS

The evolutionarily conserved region H (Allison et al., 1985) comprises residues 1308–1362 in the E. coli β’ subunit, and the GTG motif at positions 1360–1362 is invariant from bacteria to eukaryotes (Fig. 1A). Many substitutions conferring resistance to antibiotics, defects in termination, and stringent response have been mapped to β’H (Mukhopadhyay et al., 2008, Nedea et al., 1999, Rutherford et al., 2009, Weilbaecher et al., 1994). The G1360D substitution was the first RNAP ts allele identified in E. coli (rpoC1) (Khesin et al., 1968) and in yeast (rpb1–1) (Scafe et al., 1990). By contrast, the region that follows β’H (residues 1363–1407 in E. coli) is variable in sequence and length (Fig. 1B) and, to our knowledge, no mutations that map uniquely to this region have been identified.

In structures of E. coli RNAP obtained by cryogenic electron microscopy (cryo-EM) and X-ray crystallography (Murakami, 2013, Hao et al., 2021, Kang et al., 2018, Said et al., 2021), β’ residues 1361–1374 form an α-helix, and the remainder of the β’ tail is disordered. In both rpoC397 and rpoC397*, this helix and the following region are replaced with an identical 23-residue C-tail enriched in basic residues (pI 12.8). In rpoC397, six additional residues are deleted, including Gly1360. The highly unusual residue composition of the C397 extension, noted at the time of its sequencing (Christie et al., 1996), suggested that its charge may be responsible for the observed phenotypes. The tail emerges near a negatively charged pocket on the RNAP surface (Fig. S1A), raising a possibility that fortuitous electrostatic interactions with nearby residues may be responsible for the observed phenotypes. The lack of Gly1360 could contribute to the rpoC1-like ts phenotype of rpoC397 but not to that of rpoC397*, so other tail features must be, at least in part, responsible.

Genetic interactions between rpoZ and rpoC397*

Similarly to rpoC397 (Christie et al., 1996), the rpoC397* strain fails to grow at 42 °C (Fig. 2A); these ts phenotypes are observed in rfaH+ and ΔrfaH backgrounds and the latter is used here to evaluate defects in Rho-dependent termination. However, the rpoC397* strain has relatively mild phenotypes under permissive conditions, showing only a modest increase in doubling times at 32 °C (Fig. 2B) and essentially unchanged sensitivity to antibiotics (Fig. S2). Both frameshift alleles are dominant: they confer resistance to SDS when expressed from a plasmid in the rpoC⁺ strain (Fig. 3A).

Fig. 2. Growth phenotypes of MG1655 ΔrfaH strains carrying indicated mutations.

A. Deletion of rpoZ suppresses the ts phenotype of rpoC397* and promotes growth in the absence of SDS.

B. Doubling times of MG1655 ΔrfaH strains carrying indicated mutations represent the average of four-five independent biological replicates ± standard deviation.

C. Deletion of rpoZ does not prevent growth of rpoC397* on SDS at the permissive temperature.

Fig. 3. Determinants of SDS resistance.

A. Growth on SDS depends on the C397* tail. Plasmids with inserts shown on the left (or an empty vector control) were transformed into ΔrfaH MG1655 strain carrying the WT rpoC allele. Following IPTG induction, serial dilutions of exponential phase cultures were plated on LB-Carb-IPTG (left) or LB-Carb-IPTG supplemented with 1.5 % SDS (right) and incubated for 24 h at 32 °C.

B. Growth phenotypes of plasmid-encoded rpoC alleles transformed into MG1655 ΔrfaH on SDS. Colors/symbols indicate the level of resistance. Pink (Resistant), robust growth at 1.5 % SDS; light pink (weakly Resistant), very small colonies; light green (Sensitive), no growth on SDS.

C. Effects of residues at position 16 on growth in the presence of SDS.

The proximity of the β’ tail to the ω subunit prompted us to investigate a possible interplay between C397* and ω. The ω subunit and its eukaryotic homolog RPB6 promote RNAP assembly and, when overexpressed, suppress the ts phenotypes of rpoC1 and rpb1–1 but, intriguingly, not that of rpoC397 (Minakhin et al., 2001). Unlike β’•ω contacts observed in the Thermus RNAP structure (Vassylyev et al., 2002), the β’ C-terminus does not interact with ω in the E. coli RNAP structure (Murakami, 2013). However, ω is located near the disordered β’ tail (Fig. 1D) and its interactions with the β’ C-terminus have been proposed based on biochemical studies (Ghosh et al., 2001).

We wondered if the C397* tail could interfere with the proper ω function by altering its conformation. For example, a dominant-negative mutant of rpoZ that alters the ω structure has been characterized (Sarkar et al., 2013). If rpoC397* phenotypes were due to trapping of ω in an inactive state, deletion of the (non-essential) rpoZ gene that encodes ω would be expected to suppress these phenotypes. To test this, we moved the rpoZ::kan allele into the rpoC397* and rpoC⁺ strains using generalized P1 transduction. We found that the deletion of rpoZ restored ΔrfaH rpoC397* growth at 42 °C (Fig. 2A) and promoted growth at 32 °C, reducing the doubling time from 62 to 47 min (Fig. 2A, B). By contrast, ΔrpoZ had little effect on growth of the rpoC⁺ strain. We also tested the effect of Rho depletion using an allele with a IS2 insertion in the leader region of rho, which reduces Rho levels ~2.5-fold (Hu & Artsimovitch, 2017). We found that a strain containing ΩIS2 rho and rpoC397* alleles had the doubling time of 78 min, a value close to 75 min predicted if the growth defects of both alleles were independent (Fig. 2B), suggesting the lack of epistasis.

Surprisingly, however, deletion of rpoZ did not restore the sensitivity of the ΔrfaH rpoC397* strain to SDS (Fig. 2C), demonstrating that the ts and SDS^R phenotypes of rpoC397* are genetically separable. These results support a model in which RpoC397* makes unproductive interactions with the ω subunit that inhibit RNA synthesis but shed no light on the determinants of SDS resistance.

The C397* tail, but not its charge, is required for growth on SDS

Is the presence of the non-native C-tail, or simply the absence of the wild-type (WT) C-terminus, responsible for the growth on SDS? To test this, we expressed rpoC alleles from an IPTG-inducible P_trc promoter on a plasmid, an approach commonly used to test complementation (Weilbaecher et al., 1994). As expected, the expression of the WT rpoC did not support the ΔrfaH strain growth on SDS, whereas the expression of rpoC397* or rfaH did (Fig. 3A). We also found that the P_trc-rpoC397 plasmid supported growth on SDS, albeit less well than the P_trc-rpoC397* variant. By contrast, the truncation of the last 46 β’ residues did not suppress the ΔrfaH SDS sensitivity (Fig. 3A), prompting us to conclude that the C-terminal extension is required. Next, we carried out mutagenesis of the tail region to identify the features required for the growth on SDS. We report the results of this analysis as a table, numbering the C-tail residues starting at the first Pro at position 1362 in C397* (Fig. 3B).

To confirm that a specific sequence, rather than a random extension, is required for SDS resistance, we replaced the C397* tail with two different tails of equal length. The first tail, composed of GGS repeats, was designed to be flexible, whereas the second tail contains a neutral α-helix from E. coli DksA (Perederina et al., 2004). We found that neither variant was able to restore growth on SDS in the ΔrfaH background (Fig. 3B). During mutational analysis, we isolated a variant of the C397* tail in which Arg at position 15 was substituted with Lys, with a concomitant acquisition of a silent restriction site that did not have any effect on the observed SDS resistance but was instrumental during mutagenesis.

Five out of the last eight residues of the C397* tail are basic (positively charged at physiological pH), and contacts of the tail with the nucleic acids or with nearby acidic patches (Fig. S1) could alter RNA synthesis. To test this idea, we replaced two C-terminal Arg residues with Ala or Glu. To our surprise, these double substitutions did not compromise growth on SDS. Furthermore, even simultaneous substitutions of five basic residues at positions 17, 19, 21, 22, and 23 (replacing KLRLHRR with DLEGEE) did not confer SDS sensitivity (Fig. 3B).

In the absence of experimental data on the location of the C397* tail in the transcription complex, we used a combination of deletion analysis and random mutagenesis to identify residues required for SDS resistance. To shorten the C397* tail, we introduced stop codons at positions 21 (Δ3), 19 (Δ5), and 16 (Δ8). We found that while the deletions of three and five residues supported growth on SDS, albeit less efficiently in the case of the larger deletion, removal of the last eight residues completely abolished resistance (Fig. 3B). As expected, both the identity of residues 16–23 and their proper location are essential: replacing these residues with a flexible tail (GSGSGSGS) or adding them to the end of the WT α-helix (shifted tail) did not confer resistance to SDS (Fig. 3B).

The latter observation prompted us to test if any pattern emerges from random mutagenesis of the last eight positions of the tail. We replaced the corresponding region of the SDS-sensitive Gly-Ser tail (pIA1456) with an oligonucleotide cassette in which each codon was represented by NNS nucleotides specifying all 20 amino acids, albeit at unequal frequencies determined both by the commercial supplier and the codon space (Kawai et al., 2018). We expected that this approach may work with the C397* tail, as only a small number of residues are sufficient for resistance – otherwise, the “resistant” variants would be lost in the astronomically large number of residue combinations specified by the degenerate cassette. Indeed, we were able to identify SDS-resistant variants and sequenced 14 of these (Fig. 3B); two were obtained as duplicates and are not shown. This limited analysis suggested that SDS resistance may correlate with a hydrophobic residue at position 16 (Val in the C397* tail), with contributions from nearby residues.

Next, we repeated random mutagenesis for the 16^th position only. Sequencing of large colonies selected on 1.5 % SDS identified, in addition to the WT Val, Ile, Leu, Met, Phe, and Tyr residues (Fig. 3C). Colonies that appeared on SDS plates after prolonged incubation had Cys, Thr and Trp at position 16. Sequencing of SDS-sensitive variants revealed Arg, Gly, Pro, and Ser. We conclude that the presence of a non-polar residue at the 16^th position of the frameshift tail is necessary for robust growth on SDS in the absence of RfaH. Since Pro at the 16^th position did not confer SDS resistance, local backbone flexibility may additionally be important. In contrast, the most striking feature of the C397* tail, its positive charge, is non-essential and could be even reversed.

C397* is resistant to Rho

RfaH inhibits Rho-dependent termination in the cell wall biosynthesis operons, and selection for ΔrfaH suppressors identified many loss-of-function rho and hns alleles (Hu & Artsimovitch, 2017). Since both Rho and H-NS are known to trigger premature transcription termination (Shen & Landick, 2019), rpoC397* could also reduce termination. In support of this idea, a termination-defective rho15 allele with an IS1 insertion that replaces nine C-terminal Rho residues with seven non-native residues and the rpoC397 frameshift inhibited phage P2 similarly (Markov et al., 2004), but the effects of rpoC397 on termination have not been reported.

Many RNAP elements, as well as NusG and NusA, have been implicated in regulation of Rho-dependent termination, either through direct contacts to Rho or through conformational changes in the transcription elongation complex (EC) that modulate Rho function (Burns et al., 1998, Hao et al., 2021, Jin et al., 1992, Lawson et al., 2018, Said et al., 2021). Among these, only the ω subunit is located near the β’ C-terminus, and contacts between the C397* tail and ω are possible. We have previously shown that the ω-less RNAP is more sensitive to Rho in vitro (Said et al., 2021), and strengthening ω contacts with other RNAP subunits or Nus factors could be a mechanism by which C397* inhibits termination. However, our observation that the rpoZ deletion has no effect on SDS resistance (Fig. 2C) is inconsistent with this model.

To test the effects of rpoC397* and ΔrpoZ on Rho-dependent termination in vivo, we used a strain with a chromosomal β-galactosidase reporter (Chalissery et al., 2007). In this strain, a λRS45 lysogen carries a Rho-dependent T_R1 terminator derived from the nutR-cro region of lambdoid phage H-19B between the P_lac promoter and the lacZYA operon (Fig. 4A). In agreement with published data, a G146D substitution in nusG, which reduces the efficiency of Rho-dependent termination (Harinarayanan & Gowrishankar, 2003), and an IS2 insertion that reduces Rho levels (Hu & Artsimovitch, 2017) increased lacZ expression 2- and 3.2-fold, respectively (Fig. 4A). The rpoC397* frameshift also inhibited Rho-dependent termination, as evidenced by a 2.2-fold increase in β-galactosidase activity, similarly to the nusG G146D allele. By contrast, the deletion of rpoZ did not have a significant effect on lacZ expression.

Fig 4. C397* enzyme is defective in Rho-dependent termination.

A. Top: A chromosomal lacZ reporter with the lambdoid phage H19-B T_R1 terminator inserted in the leader region of the lac operon. Bottom: β-galactosidase activity measured from exponentially growing stains carrying alleles shown below the graph. The data represent the average of four independent experiments ± standard deviation.

B. Top: An in vitro transcription template with the phage λ Rho-dependent T_R1 termination region under control of the λ P_R promoter. Bottom: Single-round termination assays were carried out on a linear DNA template with Rho and NusG (where indicated) added to the indicated RNAP during chase (see Experimental Procedures). The gel section containing C397* reactions was overexposed to correct for the very low activity of this mutant enzyme. The zone of termination and run-off (RO) regions are shown on the right. The calculated fraction of the run-off RNA represents the average of three independent experiments ± standard deviation.

Next, we wanted to investigate the effect of the C397* tail on RNA synthesis in vitro. Transcription assays in defined systems provide direct means to assess RNAP elongation properties and its response to accessory factors. However, the C397 enzyme could not be purified in an active form (Christie et al., 1996), impeding this analysis. We cloned the rpoC397* allele into a polycistronic expression vector which has been shown to suppress some RNAP assembly defects (Artsimovitch et al., 2003); in this vector, the rpoABCZ genes are co-expressed from the phage T7 promoter with a purification tag on the target subunit (Svetlov & Artsimovitch, 2015). However, although we used similar vectors to successfully purify tens of RNAP variants, C397* purification was challenging even after introducing several modifications into the purification protocol (see Experimental Procedures); the C397* enzyme was unstable during purification and the final yield was low, as reported previously for C397 RNAP (Christie et al., 1996). We were concerned that the His-tagged C397* RNAP could be contaminated with a chromosomally-encoded RNAP that weakly binds to metal affinity resins. To eliminate this possibility, we constructed a derivative of the E. coli BL21 strain in which the chromosomal rpoC gene was replaced with the rpoC397* allele. Using this strain and an affinity tag on the N-terminus of the β subunit, we were able to purify the mutant enzyme for in vitro analysis; we note that C397* RNAP specific activity is <10 % of the WT RNAP activity (from multiple purification trials) and that it is defective in binding to nucleic acids. A contrast between in vitro defects and relatively mild in vivo phenotypes suggests that some factors/solutes may aid the C397* enzyme in the cell.

We compared the WT and C397* enzymes in a standard in vitro termination assay on a linear DNA template encoding bacteriophage λ tR1, an archetypical Rho-dependent terminator (Fig. 4B). On this template, transcription initiated from the λP_R promoter in the absence of CTP is stalled after the addition of the A residue at position 26 (Artsimovitch & Landick, 2002). Upon addition of CTP, RNAP resumes transcription but only 15 % of WT RNAP molecules reach the end of the linear template, generating run-off transcripts, in the presence of Rho (Fig. 4B). NusG promotes Rho-dependent termination, shifting the termination window to promoter-proximal sites and reducing the fraction of run-off RNA to 9 %. By comparison, when the most Rho-resistant RNAP among a panel of enzymes we recently characterized (Said et al., 2021), the βV550A enzyme, was used, 31 % of RNAPs reached the end of the template in the absence of NusG, and 16 % - in its presence (Fig. 4B). We found that the C397* enzyme was even more resistant to Rho-dependent termination, with 42 % full-length RNA observed in the presence of Rho alone, but that it was still responsive to NusG, which decreased run-off RNA to 24 % (Fig. 4B).

The majority of Rho-resistant RNAP mutants, including the βV550A enzyme, are also defective in recognition of pause signals and appear to be “fast” (Jin et al., 1992, Malinen et al., 2014). However, we found that C397* elongation rate and pausing patterns were similar to those of the WT enzyme, as was its response to NusA (Fig. S3). We conclude that the C397* tail strongly interferes with Rho-mediated termination in a purified in vitro system. This effect is unlikely to be due to increased propensity to pause (Fig. S3). The EC assembled with the C397* RNAP appeared to form a stable complex with Rho and NusA, as assessed by size exclusion chromatography (Fig. S4), so the loss of binding to either factor is unlikely to explain the termination defects. Interestingly, NusG was apparently loosely bound to the C397* RNAP (Fig. S4B). Observations that NusG potentiates Rho termination in vitro (Fig. 4B) suggested that this apparent binding defect may not be critical, particularly because our in vitro conditions are very different from those in the cell, where DNA supercoiling, crowding, and many DNA- and RNAP-binding proteins would be expected to modulate RNA synthesis. Therefore, we evaluated C397* RNAP interactions in the cellular context.

Crosslinking reveals changes in the C397* enzyme in vivo

We noted that, in multiple trials, the ω subunit was sub-stoichiometric in the purified C397* core RNAP (data not shown). By itself, this observation was not particularly striking because ω is only loosely bound to core RNAP and we have previously observed its loss in some RNAP preparations. However, our observations that the presence of ω confers growth defects in the rpoC397* strain (Fig. 2A) suggest that ω makes inhibitory contacts in the mutant enzyme. This inhibition could be relieved upon ω dissociation in vivo.

To evaluate this possibility, we used in vivo cross-linking in E. coli BL21 strains containing the WT or rpoC397* chromosomal alleles transformed with the matching polycistronic rpoZ-rpoA-rpoB-rpoC plasmids; the β subunit carries an N-terminal His₁₁ tag that enables capture of proteins that are associated with RNAP (or the β subunit alone). Following induction of core RNAP expression with IPTG (see Experimental Procedures and Fig. S5), we treated the cultures with formaldehyde, isolated the β subunit and associated proteins by metal affinity chromatography (Fig. S5), and analyzed the samples by mass spectrometry.

As expected, the α subunit-derived peptides were present in similar quantities (corrected for the β subunit signals) in samples from the WT and rpoC397* strains. The mutant β’ subunit was sub-stoichiometric (~25 % reduction; Fig. 5A), likely reflecting its dissociation from the α₂β sub-assembly, as inferred previously for the C397 enzyme (Christie et al., 1996, Nedea et al., 1999). Strikingly, and in support of the selective loss of ω, we observed ~3-fold fewer ω-derived peptides in the rpoC397* sample (Fig. 5A). Since ω binds to the α₂ββ’ complex, these data show that ~60 % fewer C397* enzymes contain ω. We next evaluated the abundance of general transcription factors that directly bind to RNAP. We found that crosslinking was somewhat more efficient with the rpoC397* strain, a trend that may be related to the altered cell wall structure reported for rpoC397 (Markov et al., 2004). The C397*/WT ratios of peptides derived from transcription elongation factors GreA, GreB, Mfd, NusA, NusG, and RapA ranged from 1 to 1.4, whereas Rho was an outlier, with ~1.9-fold enrichment in the C397* sample; σ⁷⁰ was slightly less abundant in the mutant (Fig. 5A). The results were similar between replicate experiments with the exception of σ^S (Fig. 5A); an apparent increase in σ^S levels (observed in only one sample) is consistent with preferential binding of alternative σ factors to ω-less RNAP; reviewed in (Kurkela et al., 2021). These results indicate that defects in Rho-dependent termination cannot be explained by the reduced affinity of C397* RNAP to Rho and Nus factors.

Fig. 5. In vivo crosslinking analysis.

A. Quantification of RNAP subunits and general transcription factors pulled down with the His₁₁-tagged β subunit; see Figure S5. For each experiment, the peptide counts were normalized to total peptides derived from the β subunit. Results of two experiments are presented as ratios of peptide counts for each protein shown in samples obtained from the rpoC397* and WT strains; see Supporting Information, Dataset S1.

B. Peptides in the β’ and ω subunits that show more than 20 % change relative to WT in two experiments are mapped to PDB ID 6C6U (left) and listed in a table (right). Red, increased peptides in rpoC397*. Blue, decreased peptides in rpoC397*. See Supporting Information, Dataset S1 and Table S3. Color coding in this and Figure S6: β, black; β’ white; β’ clamp, pink; β’ C-tail, cyan; α subunits, different shades of grey; ω, magenta; template DNA, brown; non-template DNA, beige; RNA, gold.

An aberrant conformation of C397* RNAP may confer ts growth and termination defects, as well as an apparent increase in association of Rho. An altered conformation could be reflected in altered patterns of recovered tryptic peptides: upon formaldehyde crosslinking, the target peptide would become “invisible”, reducing its relative abundance. We analyzed only changes in the β’ and ω subunits because the β subunit and the α₂β subassemblies are present in excess and we can’t distinguish changes induced by the presence of the C397* tail from those due to the absence of β’. Our analysis (see Supporting Information Dataset 1) revealed changes in several regions (Fig. 5B). The loss of β’ peptide encompassing residues 1356–1369 is expected due to its absence in the mutant protein. We also observed changes in abundance of peptides derived from the α2 helix of ω, which makes polar contacts to β’D in structures of E. coli, Thermus thermophilus, and Mycobacterium tuberculosis enzymes (Kurkela et al., 2021), and in the ω α3 helix. These changes could result from either direct interactions between ω and the C397* tail or from changes in ω that are induced by these contacts, but are relatively modest. It is possible that more drastic changes would trigger ω dissociation from the C397* enzyme. We also observed differences in several β’ regions implicated in regulation of pausing and termination (the catalytic bridge helix and trigger loop, the jaw and SI3 domains, the zinc-binding domain (ZBD), and the lid and rudder loops; Fig. 5B). For example, the β’ 997–1005 peptide in the SI3 domain was overrepresented in the C397* sample, perhaps reflecting a loss of contacts present in the WT complex. A recent study identified this region as part of a pocket that captures a key Phe on the β’ jaw domain to bias SI3 motions (Bao & Landick, 2021), which in turn modulate every step of transcription (Bao & Landick, 2021, Artsimovitch et al., 2003, Ray-Soni et al., 2017). Notably, the regions whose tryptic peptides were altered in the C397* vs the WT enzyme surround the affected β’ tail (Fig. 5B). Thus, the observed differences could be a consequence of allosteric changes imparted by the C397* β’ tail or reflect differential direct contacts of the WT and C397* β’ tails. Our structural modeling suggests that the C397* tail may fold into various stable structures and disordered loops, reaching quite far from the last β’ residue visible in the RNAP structures (Fig. S1B); the longer WT tail may also extend to different regions.

DISCUSSION

In this work, we investigated the mechanism by which C397*, the deletion-substitution at the C-terminus of the β’ subunit, alters RNA synthesis. The rpoC397* allele was isolated during a selection for SDS-resistant mutants that overcome Rho-mediated polarity in the waa operon, alongside with mutants in rho (Hu & Artsimovitch, 2017). In this work, we show that the rpoC397* mutation reduces Rho-dependent termination in vivo and in vitro but does not compromise Rho binding to RNAP. Instead, and consistent with a revised Rho mechanism (Hao et al., 2021, Said et al., 2021), our results suggest that the non-native β’ tail induces changes in other regions of RNAP. The rpoC397* strain has a second phenotype: similar to the rpoC397 variant, in which the same 23-residue extension enriched in basic residues replaces a slightly longer segment of the native β’ tail, it is unable to grow at 42 °C. The smallest RNAP subunit (ω/RBP6), which is conserved across all life, can suppress ts defects in the β’/RPB1 subunit (Minakhin et al., 2001). Here we show that, strikingly, ω has an opposite effect on rpoC397* - the rpoZ deletion restores growth at 42 °C and promotes faster growth at 32 °C but does not confer sensitivity to SDS. These observations lead us to propose that the C397* tail exists in (at least two) alternative states, whose distribution depends on temperature and other variables. In one state, the β’ tail triggers conformational changes that interfere with Rho termination. In another state, the β’ tail may form non-productive interactions with ω or other parts of RNAP. Our findings, together with recent structural advances, suggest that β’ and ω engage in dynamic interactions that modulate properties of RNAP and its response to accessory factors.

C397* and Rho-dependent termination

Several mechanisms could explain defects in Rho-dependent termination. First, many RNAP variants that are resistant to Rho are also insensitive to pause signals, which coincide with Rho release sites (Epshtein et al., 2010, Jin et al., 1992); recent data argue that conformational changes in some RNAP elements triggered by Rho and pause signals, rather than the speed at which RNAP makes RNA, explain this correlation (Said et al., 2021). Second, NusA and NusG have been implicated in termination (Burns et al., 1998, Said et al., 2021) and interact with Rho in structures of pre-termination complexes (Hao et al., 2021, Said et al., 2021); altered interactions with either factor could compromise termination. Third, in pre-termination complexes, Rho also interacts with many RNAP elements, including the flap, lobe and SI2 domains of the β subunit, the β’ ZBD, and the α C-terminal domains (Hao et al., 2021, Said et al., 2021). Perturbation of these contacts has been shown to alter termination, either directly or allosterically (Hao et al., 2021, Said et al., 2021). Finally, the RNAP-dependent Rho termination pathway seems to proceed along several intermediates, involving changes in the contacts of Rho to RNAP subunits, factors, and nucleic acids (Said et al., 2021); selective stabilization or destabilization of any of these dynamic contacts may alter termination efficiency or kinetics.

Since, to our knowledge, no transcription-altering substitutions have been characterized in the region removed by the rpoC397* frameshift, we considered a model in which changes in elongation properties of the mutant enzyme underlie its termination defects unlikely. Consistently, our in vitro experiments on standard templates used to study elongation phenotypes of RNAP variants did not reveal any defects in pausing; in fact, C397* was somewhat “slower” than the WT enzyme (Fig. S3), a phenotype typically associated with increased termination.

Structures of E. coli transcription complexes bound to NusA, NusG and Rho (Guo et al., 2018, Kang et al., 2018, Hao et al., 2021, Said et al., 2021) place the β’ C-terminus far from their binding sites, arguing against the direct impact of C397* on binding of either factor to RNAP. Indeed, our in vitro biochemical analysis (Fig. S4) and in vivo crosslinking (Fig. 5A) suggest that WT and C397* RNAPs bind to Rho and NusA similarly (Fig. S4), although minor differences cannot be ruled out; e.g., the mutant enzyme may bind to Rho more stably (Fig. 5A). Even though NusG appears to dissociate more readily from transcription complexes assembled with the C397* enzyme (Fig. S4), in vivo crosslinking did not reveal differences in NusG recovery (Fig. 5A).

Our analysis of the β’ tail variants (Fig. 3) argues that specific effects elicited by the C397* tail, rather than (merely) the abrogation of functional interactions of the WT tail, underpin the observed SDS^R phenotype, and thus RNAP resistance to Rho. Having eliminated changes in elongation properties or binding to NusA, NusG, and Rho as likely explanations, we are left with a model in which the C397* tail changes RNAP conformation to resist Rho. Since after multiple attempts we could not obtain a structure of the C397* RNAP, alone or on a nucleic acid scaffold, molecular details of this hypothetical change remain elusive. Throughout its stepwise inactivation of TEC, Rho contacts all RNAP subunits and induces conformational changes in many enzyme elements, culminating in the clamp opening and broken contacts to the nucleic acids (Hao et al., 2021, Said et al., 2021). Our results imply that the WT and C397* RNAPs may differ in conformations of the ω subunit and many β’ regions (Fig. 5B). While changes in ω could be expected based on its proximity to the altered tail, the lack of ω effect on SDS resistance (Fig. 3C) is inconsistent with the C397*•ω contacts being a sole reason for termination defects. By contrast, almost every β’ element that showed differential crosslinking (Fig. 5B) has been implicated in regulation of RNA chain elongation and termination; for example, we reported that mutations in E. coli ZBD compromise Rho termination in vivo (Said et al., 2021). Additional experimental evidence would be required to determine the exact mechanism by which C397* RNAPs resists Rho.

While C397* is a mutant with a drastically altered, non-native β’ tail, its analysis provided important new fundamental insights into transcription regulation by a hitherto underappreciated region of β’. Our results strongly suggest that the C397* tail can assume multiple conformations, each able to exert different effects on RNAP. The native β’ C-termini are disordered in RNAP structures, implying that they exist in a dynamic continuum of conformations. Binding of a regulatory protein or a small molecule could lock the tail in a given state, altering β’ interactions with nearby or distant targets, mimicking the effects of the tail replacement. In particular, our results emphasize that changes in the β’ contacts with ω may have profound effects on RNA synthesis.

ω as a modulator of RNAP activity

In all domains of life, ω-like proteins are thought to function as RNAP chaperones (Minakhin et al., 2001). Genes encoding ω homologs are present in all cellular genomes and are essential in Archaea, Eukarya, and some bacterial species, such as Mycobacterium tuberculosis; in the latter, rpoZ essentiality has been linked to RNAP assembly (Mao et al., 2018). Consistent with its chaperone function, ω restores assembly defects of some rpoC mutants (Minakhin et al., 2001) and can be functionally replaced by GroEL (Mukherjee et al., 1999). However, unlike GroEL and other general protein chaperones, ω is a live-in chaperone that has diverse and presently enigmatic effects on RNA synthesis.

Bacterial strains lacking rpoZ have pleiotropic phenotypes, many of which are thought to result from preferential recruitment of alternative σ factors; reviewed in (Kurkela et al., 2021, Patel et al., 2020). The molecular mechanism by which ω, which binds far away from σ, perturbs σ competition remains unknown. Furthermore, many other examples of (likely σ-independent) ω effects on RNA synthesis have been reported. First, ω prevents RNAP association with cryptic prophages (Yamamoto et al., 2018). Second, ω interacts with the stringent response effector ppGpp (Zuo et al., 2013) and is necessary for ppGpp-mediated repression of transcription initiation in the absence of DksA (Vrentas et al., 2005). Third, ω is required for the antibacterial activity of antibiotic MRL-436 (Walker et al., 2017). Fourth, missense or silent amino acid substitutions that stabilize the ω secondary structure are dominant lethal in vivo and destabilize RNAP interactions with nucleic acids in vitro (Sarkar et al., 2013). Fifth, ω inhibits termination by E. coli Rho in vitro (Said et al., 2021) and is ejected to enable B. subtilis RNAP recycling by HelD in vivo (Pei et al., 2020); both regulatory events require opening of the RNAP clamp. Finally, we show that ω prevents the rpoC397* strain growth at non-permissive temperature (Fig. 2A).

In structures of transcription complexes from different bacteria, ω binds at a similar location but makes different contacts to the β and β’ subunits (Fig. S6) that, together with species-specific variations in ω, recently reviewed in (Kurkela et al., 2021), could explain diversity in ω effects on gene expression. However, all ω-like proteins interact with the universally conserved β’ region D, which houses the catalytic Asp triad (E. coli β’ residues 460, 462, 464). Based on their finding that a substitution of the adjacent β’ Tyr457 residue suppresses the lethality of a N60D substitution in ω, Chatterji and colleagues proposed a model in which changes in ω are communicated directly to the RNAP active site (Sarkar et al., 2013). It is possible that the non-native β’ tail may promote a similar switch of ω into an RNAP inhibitor. In addition to hypothetical crosstalk with the active site, allosteric communications with other distal regions of RNAP may explain ω effects on σ factor selection.

Unrestrained motions of ω may be required for proper RNAP function. Stabilizing the ω structure by substitutions (Patel et al., 2020) or, as we hypothesize, restricting its mobility by the C397* tail, inhibits RNA synthesis. Effectors that bind to ω and/or to the C-terminus of β’ could alter the structural dynamics of ω, possibly modulating the RNAP activity. The C-terminal helix of E. coli ω are disordered in most structures but refolds upon interactions with Stringent starvation protein A (Wang et al., 2021), which regulates transcription initiation of stress response and virulence genes, and general elongation factor NusA (Guo et al., 2018). Both SspA and NusA make multi-dentate contacts to RNAP and their interactions with ω are likely modulatory. Interestingly, NusA-mediated ω restructuring was observed only in the hairpin-stabilized paused elongation complex (Guo et al., 2018), but the loss of ω potentiated NusA effects on pre-termination complexes which lack stable NusA/ω contacts (Said et al., 2021). Experiments to investigate ω interactions with other core RNAP subunits and its effects on transcriptional regulation are ongoing in our laboratories.

The β’•ω interface as a regulatory target

From their similar locations on RNAPs, ω subunits make contacts with the C-termini of their cognate β’ subunits, but these contacts differ among species (Fig. S6). These differences are reflected in significant sequence divergence of ω (Kurkela et al., 2021) and β’ tails (Fig. 6), which could be taken as an indication of lesser importance. Yet phylum-specific differences are necessary to achieve a balance between the two essential roles of RNAP. On the one hand, each RNAP has to faithfully process genetic information, guided by fundamental rules shared by homologous enzymes and utilizing conserved (although not necessarily at the primary sequence level) enzyme elements. On the other hand, RNAP has to adjust its output to cellular demands based on instructions it receives from accessory proteins. A recent global survey showed that more than 90 % of prokaryotic genes are habitat specific (Coelho et al., 2021) and even ubiquitous transcription factors, such as NusG, play different regulatory roles (Wang & Artsimovitch, 2020).

Fig. 6. The C-terminal tails of the β’ subunits.

Weak phylum-specific sequence motifs in Firmicutes and Proteobacteria are adjacent to β’H (ending at the GTG motif). Divergent sequence signatures are apparent in selected orders of Proteobacteria; additional representatives are shown in Figure S7, and sequences used for analysis are compiled in Supporting Information, Dataset S2. The sequence logos were generated by WebLogo (version 3.7.8).

Thus, the use of divergent RNAP elements to communicate with lineage-specific effectors is anticipated, and bacterial enzymes display significant sequence variations, particularly in surface-exposed regions, including many lineage-specific insertions (Lane & Darst, 2010). The β’ C-tails differ significantly between evolutionary distant phyla, such as Proteobacteria and Firmicutes (Fig. 6), which use divergent strategies to control gene expression. E. coli and B. subtilis RNAP, representative enzymes from these groups, differ both in the core structure of RNAPs (Miller et al., 2021) and in their dependence on accessory transcription factors. For example, in contrast to E. coli, in which Rho and NusG jointly mediate polarity control of un-translated RNAs and are essential (Lawson & Berger, 2019), neither rho nor nusG is essential B. subtilis, and transcription is uncoupled from translation (Johnson et al., 2020). B. subtilis lacks the dispensable SI2 insertion in the β flap domain which interacts with Rho in E. coli RNAP (Hao et al., 2021, Said et al., 2021) as well as two other large insertions in the β and β’ subunits that modulate RNA synthesis in E. coli. Variable regions on RNAP are frequently exploited by phage proteins during the host takeover (Tabib-Salazar et al., 2019), and the E. coli β’ tail is thought to be a target for the phage N4 SSB protein (Miller et al., 1997).

The presence of regions with limited sequence conservation within each phylum prompted us to examine sequence conservation at the order level (Fig. 6). Our analysis revealed different patterns of sequence conservation: while the β’ tails are highly divergent in Burkholderiales, except for the segment adjacent to region H, they are more conserved in other orders, including Enterobacterales and Rhodobacterales (Fig. 6). The C-termini of β’ are particularly striking in Sphingomonadales, where they are nearly twice as long and conserved throughout the entire length (Fig. S6). It is logical to assume that these differences reflect divergent regulatory needs. The dynamic and hypervariable ω •β’ interface, far away from the business center yet capable of allosteric signaling, is an attractive spot for phylum-specific regulators, such as SspA (Wang et al., 2021). Recent advances in structural analyses and mass spectrometry could facilitate identification of proteins targeting this and other sites on transcription complexes.

EXPERIMENTAL PROCEDURES

Reagents, strains and plasmids

General reagents were obtained from Millipore Sigma and Fisher. Strains and plasmids used in this study are listed in Supporting Information Tables S1 and S2, respectively. Strains were constructed using standard generalized transduction with phage P1 and verified by PCR and sequencing of the target alleles. All cells were grown in LB at temperatures indicated in each figure. For plates, LB was supplemented with 1.6 % (wt/vol) agar. Carbenicillin (100 μg/mL), kanamycin (40 μg/mL), IPTG, and SDS were added to LB media when indicated. Plasmids were constructed using standard molecular cloning protocols with restriction and modification enzymes, including Gibson assembly mixes, obtained from New England Biolabs. DNA oligonucleotides were obtained from Millipore Sigma, gBlock DNA cassettes – from Integrated DNA Technologies; sequences will be made available upon request. All plasmids were sequenced at the Ohio State University Comprehensive Cancer Center Genomics Shared Resource facility.

Plating efficiency assays

MG1655 ΔrfaH cells transformed with plasmids carrying inserts of interest under the control of P_trc promoter were grown overnight in LB supplemented with carbenicillin. Cells were diluted 1/50 into fresh LB medium supplemented with carbenicillin and grown for 3 hours at 30 °C. Plasmid encoded rpoC variants were expressed upon the addition of isopropyl-β-ᴅ-thiogalactopyranoside (IPTG) to a final concentration of 1 mM for 1 hour. Serial dilutions were spotted onto LB plates containing carbenicillin and IPTG, with or without SDS.

Growth assays

Single colonies were inoculated overnight into 2 mL of EZ Rich Defined Medium (Teknova) supplemented with 0.2 % glucose and grown at 30 °C. The overnight cultures were diluted 67x into 100 μL of the same medium in a 96-well microplate (CELLSTAR). Using a microplate reader (BioTek Epoch 2), the plate was incubated at 32 °C for 18 h with continuous double orbital shaking. OD₆₀₀ readings taken at 15 min increments were used to determine the doubling times during the exponential phase of growth.

β-Galactosidase assays

Single colonies were inoculated overnight in 5 mL of LB medium at 30 °C. Cultures were diluted 1:50 in 5 mL of EZ Rich Defined Medium (Teknova) supplemented with 0.2 % lactose and grown at 32 °C to mid-log phase (OD₆₀₀ of 0.4–0.6). Cells were pelleted, resuspended in 1 mL of Z-Buffer (60 mM Na₂HPO₄•7H₂0, 40 mM NaH₂PO4•H₂0, 10 mM KCl, 1 mM MgSO₄•7H₂O, 50 mM β-mercaptoethanol) and permeabilized with the addition of 100 μL CHCl₃ and 50 μL of 0.1 % SDS and vortexed for 10 s. In a 96-well microplate (Costar), 50 μL of permeabilized cells was added to 100 μL of Z-buffer. 100 μL of 5 mg/mL o-nitrophenyl-β-ᴅ-galactopyranoside (ONPG) was added and β-galactosidase activity was determined from the rate of increase of the o-nitrophenyl concentration measured every 10 s for 10 min at 420 nm in xMark spectrophotometer (Bio-Rad).

Protein expression and purification

Purification of the WT and β V550A RNAP, NusG, NusA, and Rho were carried out as described previously (Said et al., 2021, Svetlov & Artsimovitch, 2015, Lawson et al., 2018) using E. coli BL21 XJb λDE3 strain (Zymo Research). The standard RNAP purification protocol (Svetlov & Artsimovitch, 2015) was modified for the RpoC397* enzyme. pNS18 was transformed into strain IA523, a derivative of XJb λDE3 that carries a chromosomal rpoc397* allele; all incubations were done at 30°C. Colonies were selected on LB plates supplemented with kanamycin and carbenicillin and grown overnight. Cells were diluted 1/100 into fresh LB medium supplemented with carbenicillin and kanamycin and grown to OD₆₀₀ = 0.5. After induction with 0.4 mM IPTG, cells were grown for 5 hours, supplemented with 0.1 % arabinose for the last hour of growth, collected, and stored at −80 °C. The cell pellet was thawed and resuspended in His A (50 mM Tris-HCl pH 6.9, 500 mM NaCl, 10 μM ZnCl_2, 20 % glycerol) buffer supplemented with cOmplete protease inhibitor (Roche) and 1 mg/mL lysozyme. Cells were incubated for 1 hour, lysed with sonication, and cleared through centrifugation. The cleared lysate was loaded onto a HisTrap HP column (Cytiva) and washed with 20 column volumes of 2 % His B (His A + 1 M Imidazole). RNAP C397* was eluted with a shallow gradient from 2 % to 100 % B. Fractions containing the enzyme were verified by SDS-PAGE, pooled, and diluted to ~75 mM NaCl with buffer Hep A (50 mM Tris-HCl pH 6.9, 10 μM ZnCl_2, 20 % glycerol). The diluted solution was then applied to a Heparin HP column (Cytiva), washed with 5 % Hep B (HepA + 1.5 M NaCl), and eluted with a shallow gradient (5–100 %) of Hep B. The RNAP-containing peak fractions were pooled, diluted four times in Hep A buffer and applied to a Mono Q column (Cytiva). RNAP was eluted using a linear NaCl gradient (0.15 to 0.5 M) in Hep B buffer. Fractions containing the pure enzyme were verified by SDS-PAGE, pooled, and dialyzed overnight at 4 °C into storage buffer (10 mM Tris-HCL pH 7.5, 50 % glycerol, 100 mM NaCl, 10 μM ZnCl₂, 0.1 mM DTT).

Formaldehyde crosslinking

The wild-type BL21 of IA523, its derivative carrying a chromosomal rpoC397* allele, were transformed with the plasmids (Table S1) encoding the WT (pRM843) or mutant (pNS18) polycistronic rpoZ-rpoA-His₁₁rpoB-rpoC transcript under control of the T7 RNAP promoter, respectively. Cells were cultured at 32 °C in Terrific Broth medium with appropriate antibiotics. Kanamycin was used at 50 μg/mL and carbenicillin at 100 μg/mL. When OD₆₀₀ reached ~0.5, IPTG was added at 0.4 mM for 3 h at 32 °C. After induction, formaldehyde (Millipore Sigma) was added to a final concentration of 0.062 %, followed by a 10-min incubation at 32 °C. To stop the crosslinking reaction, glycine was added to 50 mM and the cells were pelleted at 8,000 x g at 4 °C for 6 min. Cells were disrupted by sonication in lysis buffer (50 mM HEPES, pH 7.5, 500 mM NaCl, 0.5 mM MgCl₂, 0.5X Protease Inhibitor Cocktail (Gold Biotechnology, cat# GB-376–5) in the presence of Baseline-ZERO DNase (Fisher Scientific, cat# NC1424104) and Micrococcal nuclease (New England Biolabs, cat# M0247S). The cell lysate was cleared by centrifugation at 20,000 x g at 4 °C for 30 min and loaded onto Histrap FF column (Cytiva, cat# 17531901). The column was washed with 80 column volumes (CVs) of 10 mM imidazole in Ni-buffer (50 mM HEPES, pH 7.5, 500 mM NaCl). Proteins were eluted with a linear gradient from 10 mM to 250 mM imidazole in Ni-buffer over 60 CVs. Peak fractions were assessed by SDS–PAGE and Coomassie staining (Fig. S5). RNAP fractions eluted with above 100 mM imidazole were pooled and concentrated with 100 kDa cutoff Centrifugal Filter Unit (Sigma, cat# UFC910024). The concentrated samples were used for nano-liquid chromatography-nanospray tandem mass spectrometry (Nano-LC/MS/MS) analysis, as described in Supporting Information. Data in sf3 format can be downloaded from “https://github.com/BingWangK/rpoC397-“.

In vitro Rho termination assays

RNAP holoenzymes were assembled by mixing the core RNAP (WT or mutationally altered) with a three-molar excess of σ⁷⁰ initiation factor, followed by incubation at 30°C for 20 min. Phage λ T_R1 DNA template was generated by PCR amplification of pIA267 and purified by PCR cleanup kit (Qiagen). Halted A26 ECs were formed at 37 °C for 12 min by mixing 40 nM RNAP holoenzyme with 20 nM λ tR1 DNA template, 100 μM ApU, 10 μM ATP and UTP, 2 μM GTP, and 5 Ci/mmol [α³²P]-GTP in Rho termination buffer (40 mM Tris-Cl, 50 mM KCl, 5 mM MgCl₂, 0.1 mM DTT, 3 % (v/v) glycerol, pH 7.9). Rho (and 500 nM NusG, when present) was added to 20 nM and the reactions were incubated for an additional 3 min at 37 °C. Transcription was restarted by addition of a pre-warmed (at 37 °C) mixture containing 200 μM NTPs and 50 μg/mL rifapentin and incubated at 37 °C for 6 min. Reactions were quenched by addition of an equal volume of stop buffer (45 mM Tris-borate, 10 M urea, 20 mM EDTA, 0.2 % xylene cyanol, 0.2 % bromophenol blue, pH 8.3) and separated by denaturing 5 % PAGE (7 M urea, 0.5X TBE). Gels were dried and products were visualized using a Typhoon FLA 9000 PhosphorImaging system (GE Healthcare). Readthrough and termination RNA products were quantitated with ImageQuant and Microsoft Excel software.

Sequence alignment of RpoC

To get an overall view of the tail length, representatives from different domains of life were collected. For Eukarya, two rounds of jackhmmer (HmmerWeb version 2.41.2) (Potter et al., 2018) (https://www.ebi.ac.uk/Tools/hmmer/search/jackhmmer) were done by using NP_000928.1 from H. sapiens as the query. 285 sequences are included in the final alignment; For Archaea, one round of jackhmmer was done by using WP_010189657.1 from Candidatus Nitrosarchaeum limnium as the query. 101 sequences are included in the final alignment; to build the Bacteria dataset, the representatives from reference (Johnson et al., 2020) were used, and the RpoC-TIGR model (TIGR02386) was applied to this dataset. A total of 1526 sequences are included in the final alignment. All sequences were fetched from NCBI. Sequence alignment was done by MAFFT (version 7) with the E-INS-I method (Katoh et al., 2017).

To analyze the tail conservation of Proteobacteria and Firmicutes, representatives from the Bacteria dataset were collected, and all sequences were reduced at 95 % identity using cd-hit (version 4.8.1)(Huang et al., 2010). For building the dataset (Supporting Information Dataset S2) to analyze tail conservation inside the Proteobacteria, we selected 1–3 representatives from each genus from the latest version (release202) of GTDB (Parks et al., 2021). Sequence alignment was done by MAFFT with the default setting of Jalview (version 2.11.1.4) (Procter et al., 2021). Columns with > 20 % gaps were removed from the alignment. The sequence logos were generated by WebLogo (version 3.7.8) (Crooks et al., 2004).

Data availability

Additional raw data that support the findings of this study are openly available in github at https://github.com/BingWangK/rpoC397-.