Anti-pausing activity of region 4 of the RNA polymerase σ subunit and its regulation by σ-remodeling factors

The basal transcription factors of cellular RNA polymerases (RNAPs) stimulate the initial RNA synthesis via poorly understood mechanisms. Here, we explored the mechanism employed by the bacterial factor σ in promoter-independent initial transcription. We found that the RNAP holoenzyme lacking the promoter-binding domain σ4 is ineffective in de novo transcription initiation and displays high propensity to pausing upon extension of RNAs 3 to 7 nucleotides in length. The σ4 domain stabilizes short RNA:DNA hybrids and suppresses pausing by stimulating RNAP active-center translocation. The anti-pausing activity of σ4 is modulated by its interaction with the β subunit flap domain and by the σ remodeling factors AsiA and RbpA. Our results suggest that the presence of σ4 within the RNA exit channel compensates for the intrinsic instability of short RNA:DNA hybrids by increasing RNAP processivity, thus favoring productive transcription initiation. This “RNAP boosting” activity of the initiation factor is shaped by the the thermodynamics of RNA:DNA interactions and thus, should be relevant for any factor-dependent RNAP.

The initial transcription pause occurring after the synthesis of 6-nt RNA functions as a checkpoint on the branched pathway between productive and non-productive transcription (Dulin et al, 2018).
During RNA synthesis, RNAP performs a stepwise extension of the RNA chain by one nucleotide that is called nucleotide addition cycle (NAC) (reviewed by (Belogurov & Artsimovitch, 2019). During the NAC, the fist initiating nucleoside triphosphate (NTP) or the 3' end of RNA occupies the product site (i-site) (pre-translocated state), while the incoming NTP enters in the substrate site (i+1-site). After formation of the phosphodiester bond, the RNA 3' end moves from the i+1-site to the i-site (post-translocated state). The concerted translocation of RNA and DNA to the active site is controlled by the b' subunit trigger loop that folds into the trigger helix upon transition from the pre-translocated to the post-translocated state (Toulokhonov et al, 2007b) (Zhang et al, 2010).
Specifically, the s subunit region 3.2 hairpin loop (s3.2-finger) contacts the template ssDNA strand at positions -4/-5 and controls its positioning in the active site (Pupov et al, 2010) (Tupin et al, 2010). The s3.2finger can indirectly modulate the priming of de novo RNA synthesis at promoters (Kulbachinskiy & Mustaev, 2006)  and promoter-like DNA templates, such as the M13 phage origin (Zenkin & Severinov, 2004). The s subunit may also exert an inhibitory effect on initial transcription. Indeed, the unstructured linker connecting domains s3 and s4 (formed by the s regions 3.2 and 4.1) is located in the RNA-exit channel and represents a barrier for growing RNA chains. This linker is displaced by RNA upon promoter escape (Li et al, 2020). A clash between the s3.2-finger and >4-nt RNA chains hinders RNA extension and may cause the formation of abortive RNAs, thus contributing to pausing during initial transcription (Murakami et al, 2002) (Basu et al, 2014) (Duchi et al, 2016) (Zhang et al, 2012) Binding of the regions s3.2 and s4.1 within the RNA exit channel takes place during assembly of the RNAP holoenzyme when the s subunit undergoes the transition from the "closed" to the "open" conformation (Callaci et al, 1999). Recent single-molecule fluorescence resonance energy transfer studies demonstrated that in Mycobacterium tuberculosis, this transition is regulated by the activator protein RbpA (Vishwakarma et al, 2018) that interacts with the s2 and s3.2 domains (Boyaci et al, 2018). Whether RbpA can influence initial transcription has never been explored.
Several mechanisms to explain the s3.2-finger stimulatory activity during initial transcription have been proposed: (1) stabilization of the template ssDNA in the RNAP active site (Zhang et al, 2012); (2) decreased K m for 3'-initiating NTP binding in the substrate i+1-site (Kulbachinskiy & Mustaev, 2006); and (3) stabilization of short RNAs in the active site (Campbell et al, 2002) (Zenkin & Severinov, 2004) (Zenkin et al, 2006). The last mechanism was also suggested for the B-reader domain of TFIIB, which is the structural homologue of the s3.2-finger (Bushnell et al, 2004) (Chen & Hampsey, 2004). As the s subunit occludes the RNA path and contacts all principal regulatory domains of RNAP (b'-clamp, b-lobe, b-Flap), it may affect RNA synthesis in several ways: through ssDNA template positioning, RNA binding, or direct modulation of the RNAP domain motions.
Here, to discriminate among these different scenarios, we investigated how the s subunit and RNA:DNA hybrid length affect branching between productive and non-productive RNA synthesis during initial transcription by two RNAPs from phylogenetically distant bacterial lineages: Escherichia coli (EcoRNAP) and M. tuberculosis (MtbRNAP). Compared with EcoRNAP, MtbRNAP presents several structure-specific features, particularly the lack of Eco-specific TL-insertion and the presence of the ~90 amino acid-long Actinobacteria-specific insertion in the b' subunit (b'-SI) (Lane & Darst, 2010) (Lin et al, 2017). To analyze directly the effects of the s subunit on the RNAP catalytic site activity, we used promoterless DNA scaffold templates (Tupin et al, 2010) (Zenkin & Severinov, 2004). DNA scaffolds have been widely used in structural studies on ITCs of bacterial and eukaryotic RNAPs (Cheung et al, 2011) (Liu et al, 2011). The scaffold model allows bypassing the complexity of promoter-dependent initiation that is strongly influenced by the promoter configuration and by the interactions of s with promoter elements. When complexed with RNAP, scaffold DNA templates harbor a "relaxed" conformation lacking the topological stress observed in the transcription bubble due to DNA scrunching during initial transcription at promoters. Moreover, as DNA scaffolds lack non-template strand ssDNA, transcription initiation should be less affected by interaction with the core recognition element (CRE) (Vvedenskaya et al, 2014). We found that the promoter-binding domain s4 (i.e. the structural homologue of the eukaryotic TFIIB B-ribbon), located ~60 Å away from the active site, strongly stimulates RNAP translocation and stabilizes short RNA:DNA hybrids in the RNAP active site. The combination of these activities provides the basis for the initiation-to-elongation transition regulation by the auxiliary transcriptional factors that binds to the s subunit.

The s 70 subunit is required for initial transcription from the promoter-less scaffold DNA template
To explore the role of the s subunit in initial transcription we used two types of minimal DNA scaffold templates (Fig. 1A): a Short Duplex Template (SDT), which included the 9-bp downstream DNA (dwDNA) duplex, and a Long Duplex Template (LDT), which comprised the 18 bp dwDNA duplex. The dwDNA duplex of the LDT scaffold forms additional contacts with the b' subunit residues 202-247 that stabilize the RNAP-scaffold complex (Kulbachinskiy et al, 2002) (Vassylyev et al, 2007a). Previously, we demonstrated that extension of the 3-nt RNA primer on SDT DNA is strongly stimulated by the s 70 subunit (Tupin et al, 2010). Here, we found that the RNAP core from E. coli (EcoRNAP) was inactive in de novo transcription initiation at SDT and LDT templates performed in the presence of [a-32 P]-UTP, CTP and GTP (Fig. 1B).
Conversely, the s 70 -EcoRNAP holoenzyme synthesized a single 3-nt RNA (pppC[a-32 P]UpG) starting 8 nt downstream of the 3' end of the template DNA (designated as "+1") ( Fig. 1B). This start site assignment was validated by using the antibiotic rifampicin that inhibits the synthesis of the second phosphodiester bond.

The promoter-binding domain s4 is essential for de novo initiation of RNA synthesis
To identify the s 70 regions that influence RNA synthesis initiation on minimal scaffolds, we generated a panel of s 70 mutants (Fig. 1D, E). The s D3-4 fragment (residues 1-448) in which the regions 3 and 4 were deleted, is inactive in promoter-dependent transcription initiation (Zenkin et al, 2007). The s D4 fragment (residues 1-529) lacked part of region 3.2 and the entire s4 domain. In the s D4.2 fragment (residues 1-553), only region 4.2 was deleted, but not the 4.1 a-helix, which binds inside the RNA exit channel (Fig. 1E). In agreement with previous studies (Kumar et al, 1993) (Kumar et al, 1994), the EcoRNAP holoenzymes harboring the s D4 or s D4.2 fragments were inactive in abortive transcription assays with the -10/-35 type lacUV5 promoter, and displayed reduced transcriptional activity with the "extended -10" type galP1cons promoter (Fig. 1F). The s D3.2 subunit, in which residues 513-519 in the s3.2-finger were deleted, was active in transcription initiation with both promoters. The EcoRNAP holoenzymes assembled with the mutant s 70 subunits were inactive in de novo transcription initiation on the SDT scaffold (Fig. 1B). We detected no synthesis of dinucleotide RNA products by the EcoRNAP core and by the holoenzyme, differently from what reported for initial transcription on the M13 minus-strand origin (Zenkin & Severinov, 2004). This difference might be explained by the low NTP concentration (22 mM) used in our experiments. Conversely, on the LDT scaffold, the activity of s D3.2 corresponded to 42% of the activity of full length s 70 . Thus, strengthening the interaction between RNAP and the dwDNA duplex beyond position +10 can compensate for the lack of interaction between the s3.2-finger and template-strand ssDNA. This result suggests that the s3.2finger/DNA interaction contributes to, but is not essential for initial transcription. caused by deletions in domain s4, which does not interact with scaffold DNA, cannot be compensated by the dwDNA interactions, suggesting that s4.2 is essential for initial transcription and exerts its activity through interaction with RNAP. Conversely, it has been suggested that s4 is dispensable for initial transcription on the M13 phage minus-strand origin (Zenkin & Severinov, 2004). This discrepancy might be caused by differences in the DNA template architecture. dwDNA duplex and RNA primers suppress the translocation defect caused by deletions in the s subunit.
In promoter-dependent transcription initiation, short ( 3-nt) RNA primers (pRNAs) can rescue the defects linked to deletions in the s3.2 and s4 regions (Campbell et al, 2002) (Zenkin & Severinov, 2004) (Kulbachinskiy & Mustaev, 2006. To determine whether they have the same effect also when using minimal scaffold templates, we carried out transcription in the presence of a 2-nt pRNA (GpC, pRNA2) the 3' end of which was complementary to the third position upstream of the DNA duplex (designated as position "+1") ( Fig. 2A). We assumed that the first catalytic step, addition of [a-32 P]-UTP to pRNA2 (synthesis of GpC[a-32 P]U), does not requires translocation of the RNAP active center because the 3' end of pRNA2 binds to the "product-site" (i-site, facing the position +1 of the DNA template), thus leaving the "substrate site" (i+1, facing position +2 of the DNA template) available for incoming NTP (Fig. 2B). This hypothesis is supported by the structures of the RNAP initiation complexes with synthetic scaffolds observed in posttranslocated states (Cheung et al, 2011) (Zhang et al, 2012). The next catalytic steps (synthesis of the GpC[a-32 P]UpG and GpC[a-32 P]UpGpA products) require the translocation of the RNA 3' end from the i+1 site to the i-site ( Fig. 2A, B). As only the first incorporated NTP (U) was labeled, the fraction of the longest reaction product (RNA[N+2]) reflected the overall "efficiency of RNAP translocation" from register +2 to + 4.
In the presence of GpC and three nucleotides ([a-32 P]-UTP, GTP, and ATP), the EcoRNAP core synthesized two [ 32 P]-labeled RNA products (3-nt and 5-nt RNAs), with a bias toward the shorter one (overall translocation efficiency: ~ 40%, Fig. 2C). In the same conditions, s 70 strongly stimulated [a-32 P]-UTP incorporation and translocation. Consequently, the 5-nt RNA was the major reaction product synthesized by the EcoRNAP holoenzyme (95% translocation efficiency). The observed low efficiency of the [a-32 P]-UTP-addition reaction by the EcoRNAP core might reflect its low affinity for pRNA2. Indeed, increasing pRNA2 concentration increased [a-32 P]-UTP incorporation, but did not stimulate translocation ( Fig. S1). Even the "minimal" s D3-4 fragment strongly stimulated [a-32 P]-UTP incorporation, compared with the EcoRNAP core ( Fig. 2D, Fig. S2). Therefore, binding of domain s2 to the EcoRNAP core may strengthen its interaction with the scaffold DNA/RNA and/or directs it to the active site cleft.
Like with promoter DNA templates, the initiation defect on DNA scaffolds upon s3.2-finger deletion was rescued by pRNA2 (~80% translocation efficiency, patterns similar to those of the EcoRNAP core. As deletion of the region 4.2 and deletion of the entire s3 and s4 domains led to the same effect, we conclude that s4.2 is a principal determinant for the efficient synthesis of the 5-nt RNA product. It is unlikely that the translocation defect conferred by deletion of the region s4.2 is due to a decreased affinity of the 3-nt RNA product (GpC[a-32 P]U) for the RNAP active site because the overall [a-32 P]-UTP incorporation was similar with full length s 70 and s D4.2 (Fig. 2C). Our results suggest that RNAP translocation from register +2 to register +3 is a rate-limiting step for 5-nt RNA synthesis, and is slower than [a-32 P]-UTP addition and the 3-nt RNA product dissociation. The s4 deletions had similar effects on translocation also when using the SDT scaffold and the 3-nt pRNA (CpGpC, pRNA3) ( Fig. S2).
However, the 3-nt pRNA suppressed the transcription defects caused by the s4 deletion on the LDT but not the SDT scaffold, indicating the RNAP interaction with dwDNA stimulates translocation. To explore the effect of the DNA duplex on RNAP translocation we used modified versions of the SDT scaffold (SDT2 and SDT2+1), with more stable dwDNA duplexes (Fig. 2D). Moreover, in the SDT2+1 scaffold, the 5' end upstream edge of the DNA duplex was extended by one base pair (G:C). Thus, translocation from register +2 to register +3 on SDT2+1 requires the unpairing of 1 bp of dwDNA followed by the formation of the contact between G at position +3 of the non-template DNA and the CRE pocket of the RNAP core that is known to counteract transcriptional pausing (Vvedenskaya et al, 2014). Translocation was more efficient on the SDT2+1 scaffold compared with the SDT1 scaffold ( Fig. 2E), suggesting that dwDNA duplex melting is not a barrier for translocation and that the interaction with CRE stimulates translocation. We conclude that EcoRNAP pauses after the addition of the first NTP to the RNA primer and that the interaction with the downstream DNA and RNA promotes forward translocation. The region s4.2 may act on translocation by strengthening this interaction.

The s subunit regions 3.2 and 4.2 stabilize  4-nt RNAs in the RNAP active site
To determine whether the s subunit can stabilize short RNAs in the RNAP active site, we immobilized ITCs on Ni 2+ -agarose beads and tested their ability to retain RNAs by washing the complexes with transcription buffer. We used 2 to 6 nt-long pRNAs the 3' end of which was aligned to the same position of the template, designated as position "+1" (Fig. 3A) Control experiments in which complexes were formed by the core EcoRNAP on SDT and LDT scaffolds showed that after washing with the "high salt" buffer containing 1M NaCl (Fig. S3), 5-, 6-, 7-, 13-nt pRNAs were stably bound in ITCs. However, reduced retention of 5-and 6nt pRNAs, was observed with the SDT scaffold. This indicates that the dwDNA duplex contributes to the overall stabilization of the complex. Consequently, the LDT scaffold was used in the next experiments. To measure the retention efficiency, ITCs containing 2-to 6-nt RNAs were either washed with the transcription buffer containing 250 mM NaCl and then labeled with [a-32 P]-UTP, or directly labeled without washing step ( Fig. 3B). This experiment demonstrated that 2-to 4-nt pRNAs were weakly bound to ITCs compared with 5-to 6-nt pRNAs (Fig. 3B-D). Therefore, the 4-bp RNA:DNA hybrid is a conversion point between stable and unstable ITC states. Moreover, we observed a clear difference in the capacity to hold 4-nt pRNA by ITCs containing the full length s subunit (75% retention efficiency) and s mutants (s D3.2 and s D4 ; <50% retention efficiency) (Fig. 4E). The defect was stronger for s D4 -EcoRNAP than s D3.2 -EcoRNAP. We obtained similar result with the SDT scaffold (Fig. S4). To determine whether the slight difference in 4-nt pRNA retention observed between s D3.2 and s D4 (Fig. 4E) was significant, we performed several washing steps on ITCs formed by s D3.2 -EcoRNAP and s D4.2 -EcoRNAP (Fig. 4F,G). After the third washing step, almost no bound RNA was left in s D4.2 -ITC (~10% retention relative to full length s 70 ), while RNA retention was higher for s D3.2 -ITC (~40% retention relative to full length s 70 ). RNA binding remained stable with wild type s 70 -ITC (60% retention relative to the 'no washing' condition). We conclude that the s 70 subunit stabilizes 4-5-nt-long RNAs in the RNAP active site and that the region s4.2 is a major determinant of this activity.

The s region 4.2 promotes extension of  7 nt-long RNAs
To explore the relationship between translocation efficiency and RNA:DNA hybrid stability, we performed 2-min primer extension reactions with pRNAs of various lengths in the presence of [a-32 P]-UTP, GTP and ATP (Fig. 4A). In these experiments, to facilitate the detection of the initial pause, we used the SDT scaffold that displayed stronger s-dependence in translocation. The EcoRNAP core efficiently extended  8-nt pRNAs (>90% efficiency) (Fig. 4B,C), and its translocation efficiency decreased gradually with the RNA length shortening. This dependence on RNA length might be explained by the intrinsic instability of RNA:DNA hybrids and/or by the disengagement of the RNA 3' end from the active site. The translocation efficiency was independent from the RNA length when the wild type EcoRNAP holoenzyme was used ( Fig.   4B,C). The s D3.2 -EcoRNAP holoenzyme displayed strong translocation defects with 3-nt pRNA (~25% efficiency), moderate defects with 4-5-nt pRNAs (~80% efficiency), and no defect with 6-nt pRNA (>90% efficiency). On the basis of the RNAP-promoter complex structure data, the 5' end of  5-nt-long RNAs clashes with the s3.2-finger (Fig. 4D). Thus, the s3.2-finger should hinder the extension of RNAs longer than 5 nt and favor abortive initiation (Murakami et al, 2002) (Li et al, 2020). A model of the s3.2-finger deletion on the EcoRNAP structure ( Fig. 4E) showed that 5-to 7-nt RNAs can be accommodated in the active site cleft. Yet, the remaining segment of the region s3.2 can still contact the template DNA strand at positions -6/-7. Strikingly, in our experiments, 'abortive' RNAs accumulated when using the EcoRNAP core and the s D3.2 -EcoRNAP holoenzyme, but not with the wild type EcoRNAP holoenzyme. This suggests that the s3.2-finger is not the primary cause of abortive RNA formation and that abortive RNA synthesis is not an obligatory event in initiation (Duchi et al. 2016).
Unlike s D3.2 -EcoRNAP, translocation stimulation was defective with the s D4.2 -EcoRNAP holoenzyme and pRNAs shorter than 8 nt. The properties of the s D4.2 -EcoRNAP holoenzyme were identical to those of the EcoRNAP core except with 5-nt-long RNA that displayed increased translocation efficiency with the EcoRNAP core. We did not investigate the reason of this unusual behavior. The results of the "RNA- there is no correlation between RNA:DNA hybrid stability and translocation efficiency. Indeed, 5-and 6-nt RNAs were stably bound to ITC, but still displayed s-dependence for translocation. Thus, we conclude that the low efficiency in nascent RNA extension (any lengths) by the s D4.2 -EcoRNAP holoenzyme is due to RNAP pausing after the first NTP addition, and that the stimulation of RNAP translocation by s is unlikely to occur through RNA:DNA hybrid stabilization.

The RNA 3' end nucleotide identity determines the initial-transcription pause duration
To explore the impact of s4.2 on pausing, we studied the kinetics of 5-nt and 7-nt RNA synthesis initiated with pRNA3 (ITC3, unstable RNA:DNA hybrids) and pRNA5 (ITC5, stable RNA:DNA hybrids), respectively (Fig. 5A). The overall [a-32 P]-UTP addition rate was similar for ITC3 and ITC5 in the presence of the s D4.2 , and was highest in the presence of the full length s 70 (Fig. 5B,C). Translocation from register +4 to +5 was at least 100-fold faster (t 1/2 ~ 1.7 s) in the presence of full length s 70 than of the s D4.2 mutant (t 1/2 2 00 s) (Fig. 5B,D). Reactions were completed in 120 s, without any further incorporation of [a-32 P]-UTP. Therefore, the labeled 5-nt RNA product remained bound to RNAP. Extension of the pRNA 5' end by 2 nucleotides (pRNA5) accelerated the forward translocation only by 2-fold. Therefore, in agreement with the conclusion drawn in the previous section, the overall RNA:DNA hybrid stabilization has little effect on pausing.
In our assay, the RNA chain elongation starts with addition of U that forms an unstable U:dA pair with the DNA template (Huang et al, 2009). Therefore, the 3' end nucleotide might disengage from the active site, and induce pausing. If this hypothesis is correct, the substitution of the U:dA pair by the more stable C:dG pair should suppress pausing, and favor forward translocation. To test this assumption, we used a scaffold (SDT-G) harboring G instead of A at position +2 (Fig. 5A), and initiated the primer extension with [a-32 P]-CTP. Unlike UTP, the CTP addition rate with the s D4.2 -EcoRNAP holoenzyme was close to that observed with the wild type EcoRNAP holoenzyme (compare Fig. 5C and 5F). Thus, without s4, EcoRNAP senses the difference between UTP and CTP, while CTP suppresses the effect of the s4 deletion.
Irrespectively of the RNA length (4-nt or 6-nt), translocation of the mutant s D4.2 -EcoRNAP from the register +2 to the register +3 was accelerated by ~10-fold on SDT-G DNA compared with SDT-A DNA. This suggests that the stability of base pairing at the 3' end, but not the RNA:DNA hybrid length is crucial for forward translocation. As s D4.2 -EcoRNAP translocation rate was significantly reduced even when the RNA 3' end was stabilized, compared with wild type EcoRNAP, we conclude that the region s4.2 may affect the active site cycling or the clamp opening-closing dynamics that control RNAP translocation.
The s4 remodeling co-activator AsiA stimulates pausing Region s4.2 binds to the flap-tip-helix (FTH) of the RNAP b subunit (Kuznedelov et al, 2002) (Geszvain et al, 2004) that is implicated in the regulation of pausing (Kang et al, 2018). To test whether s4.2 exerts its effect on RNA synthesis through interaction with b-FTH, we used the T4 phage co-activator protein AsiA. AsiA remodels exactly the same region in the s 70 subunit (residues 528 -613) that was deleted in the s D4 mutant, and disrupts the interaction between s4.2 and b-FTH (Hinton & Vuthoori, 2000) (Shi et al, 2019) (model in Fig. 6A). If the s4.2-b-FTH contact were essential for RNA synthesis, AsiA should fully inhibit initial transcription. To test this hypothesis, we performed de novo and primed transcription by the s 70 -EcoRNAP and sD4.2-EcoRNAP holoenzymes, with and without AsiA, on the SDT2 template (Fig. 6B). As control, we used an abortive transcription assay on the lacUV5 promoter. AsiA inhibited lacUV5-dependent transcription initiation by 85% (Fig. 6B, lanes 1,2 and Fig. 6C). Conversely, de novo initiation from the scaffold was much less sensitive to AsiA. Indeed, 3-nt RNA synthesis was inhibited only by 50%, which coincided with the accumulation of the short 2-nt RNA product (Fig. 6B, lanes 7,8 and Fig. 6C). AsiA also influenced transcription initiated with the GpC primer (Fig. 6B, lanes 3,4 and Fig. 6C). The amount of 3-nt RNA increased simultaneously with the increase in total [a-32 P]-UTP incorporation. Such effect was consistent with the AsiA-induced destabilization of short RNAs, leading to accumulation of "abortive" transcripts. The finding that AsiA did not affect [a-32 P]-UTP incorporation with the sD4-EcoRNAP holoenzyme (Fig. 6B, lanes 5,6 and Fig. 6C) indicates that AsiA modulates RNA synthesis through s4.
However, the weak impact of AsiA on initial transcription was in striking contrast with the strong inhibitory effect of the s4 deletion. The only possible explanation for this discrepancy can be that the s4 physical presence in the RNA exit channel is essential for initial transcription. In the presence of AsiA, the s4 domain remains bound inside the RNA exit channel (Shi et al, 2019), and therefore AsiA exerts only a weak effect on scaffold-dependent transcription. We conclude that the interaction of s4.2 with b-FTH modulates the catalytic site activity, but is not essential for initial transcription.

RbpA from M. tuberculosis stimulates translocation through the s4.2/b-FTH interaction
To determine whether the s subunit anti-pausing activity can be observed with RNAP from other bacteria, we studied initial transcription by MtbRNAP. We used the M. tuberculosis s B subunit that requires the activator protein RbpA to stabilize its active conformation in the MtbRNAP holoenzyme (Vishwakarma et al, 2018). As RbpA N-terminus binds within the RNA-exit channel, it could modulate s4.2 anti-pausing activity (model in Fig. 6D). First, we compared the kinetics of 5-nt pRNA extension by MtbRNAP and EcoRNAP on SDT2 DNA in the presence of [a-32 P]-UTP and GTP ( Figure S5). As observed with EcoRNAP, MtbRNAP paused at the register +6 during initiation from pRNA5, and its translocation was stimulated by the s B -RbpA complex. To better understand the role of s4.2-b-FTH interaction in initial pausing, we constructed a MtbRNAP mutant in which the b subunit residues 811-825 were deleted (MtbRNAP DFTH ), and then assessed how the translocation activity of the mutant and wild type enzymes were influenced by the pRNA length ( Fig. 6E,F). ITCs were assembled with 2, 3, 5 and 6-nt pRNAs (ITC2 to 6) in the presence of s B and RbpA, or without transcription factors, and supplemented with [a-32 P]-UTP and GTP. As observed with EcoRNAP, MtbRNAP translocation efficiency increased gradually with the RNA length, and reached 80% with the 6-nt pRNA. Like for s 70 , the s B -RbpA complex stimulated the forward translocation with short pRNAs (3-6-nt in length). However, the s B subunit alone did not stimulate translocation, in agreement with fact that its conformation in the MtbRNAP holoenzyme differs from that of s 70 in the EcoRNAP holoenzyme. The deletion of b-FTH abolished s B anti-pausing activity with ITC2 and ITC3 (2-and 3-nt pRNAs, respectively). Furthermore, the RNA amount produced by unstable ITC2/ ITC3 formed by the MtbRNAP DFTH mutant increased by ~4-fold, compared with the amount produced by the stable ITC6. (Fig. S6). This "abortive-like" behavior was observed only in the presence of the s B subunit, and might be caused by a clash between the inappropriately positioned region 3.2 and RNA. Addition of RbpA only partially restored s B capacity to stimulate MtbRNAP DFTH translocation ( Fig. 6F and Fig. S5). In agreement with the conclusion drawn from the experiments with AsiA, this result suggests that s4.2 interaction with b-FTH regulates, but is not essential for the anti-pausing activity of s B . The b-FTH deletion should dramatically destabilize s4 positioning/interaction within RNA channel and consequently enhance pausing and abortive transcription.
RbpA compensates for the lack of b-FTH probably by facilitating s4 positioning within the RNA-exit channel.

DISCUSSION
Our study demonstrates that RNAPs from different bacterial species, are inefficient in initial transcription and prone to pause upon extension of short RNAs (3 to 7-nt in length). The s subunit region 4.2, which was implicated only in promoter binding, counteracts the initial transcription pausing and thus plays an essential role in organizing of the RNAP active center for efficient initiation of de novo RNA synthesis. The s4.2 region displays two distinct activities: RNA:DNA hybrid-stabilizing activity, and translocation-stimulating activity. Modulation of these activities by the s-remodeling factors may be a general mechanism to tune initial transcription.

Initial transcription pausing on the pathway to abortive transcription
Our results suggest that at each nucleotide addition step, ITCs harboring 3-8 nt RNAs can enter into a paused state in which the RNA 3' end is disengaged from the active site. The paused ITC (PITC) bifurcates in two pathways: abortive pathway in which nascent RNA dissociates from RNAP, and productive pathway in which nascent RNA remains bound to RNAP and slowly translocates to the next register (Figure 7)). PITC conversion to productive ITC is accelerated by (1) (Liu et al, 2011) (Cheung et al, 2011), the RNA 3' end may be displaced from the active site. In addition, the 3'-U forms unstable base pairs (U:dA) that in the absence of s4, may favor the formation of the frayed state, leading to backtracking and pausing (Artsimovitch & Landick, 2000) (Toulokhonov et al, 2007a). Consequently, the PITC remains blocked in one of the inactive states (halftranslocated, hyper-translocated, or backtracked) that slowly isomerize to an active post-translocated state.
The 3'-C that forms more stable (C:dG) base pairs remains in the active site, thus promoting forward translocation. In support to this model, the PITC half-life time was more strongly biased by the RNA 3' end nucleotide identity than by the RNA:DNA hybrid length. Indeed, the RNA 3' end nucleotide also modulated the translocation bias in stable elongation complexes with 8-9-bp RNA:DNA hybrids (Hein et. al, 2011).
The s subunit may restrain RNA and DNA motions by strengthening the RNAP core interactions with RNA and DNA, thus allowing the correct alignment of the template to the active site and promoting translocation .
The s-mediated stabilization of short RNAs in the active site and stimulation of the forward translocation should drastically reduce the probability of abortive transcription and shift the equilibrium toward promoter escape. As the half-life time of the initial pause strongly depends on the RNA 3' end nucleotide identity, it should be a promoter-specific event determined by the initial transcribing DNA sequence.

Two steps in the maturation of RNA:DNA hybrids and DNA scrunching
Our results underline two phases in initial transcription: (1) transition from unstable to stable RNA:DNA hybrids when RNA length reaches 5 nt; and (2) transition from ITC prone to pause to productive ITC/EC when RNA length exceeds 8 nt (Figure 7B). Before the first phase, abortive transcription is predominant and the RNA:DNA hybrid stability depends on s. Before the second phase, pausing is predominant and translocation depends on s. These phases perfectly fit with the three ITC types observed at natural promoter templates: unstable ITCs with RNAs < 4 nt, intermediate stability ITCs with RNAs of 4-8 nt, and stable productive ITC with RNAs > 8 nt (Metzger et al, 1993) (Carpousis & Gralla, 1980). Our results on RNA:DNA hybrid stability are in agreement with structural studies on eukaryotic RNAPII showing that ITCs with 4-5-nt RNAs form distorted or mismatched RNA:DNA hybrids, while the 6-8 bp hybrids harbor identical canonical structures (Liu et al, 2011). Unlike ITCs formed on promoters, ITCs formed on our DNA scaffolds lack the upstream part of the transcription bubble and non-template ssDNA. Therefore, the initial transcription on scaffold is not affected by the stress arising due to DNA scrunching, which is a major cause of abortive transcription (Kapanidis et al, 2006). The lack of stress in DNA templates may explain the quantitative retention of 4-7-nt RNAs in s-containing ITCs formed on scaffolds, compared with the a small fraction of 5 to 7-nt RNAs retained in ITCs formed on promoter DNA templates (Brodolin et al, 2004) (Duchi et al, 2016). hybrids, and selectively stimulates UTP addition to the RNA 3' end. These activities were previously attributed to the s3.2-finger (Kulbachinskiy & Mustaev, 2006) (Zenkin & Severinov, 2004) that stabilizes template ssDNA in the active site (Zhang et al, 2012) (Tupin et al, 2010). Altogether, our results show that s4.2 activity during initial transcription can be clearly differentiated from that of the s3.2finger. Indeed, s3.2-finger deletion abolished RNA retention in ITCs, but had only a moderate effect on translocation. Conversely, s4 deletion strongly affected RNA retention and also translocation. Furthermore, the s3.2-finger was dispensable for de novo initiation on scaffold template harboring a long dwDNA duplex, while s4.2 was essential. Therefore, we conclude that s4 interaction with core RNAP is a major determinant of the s subunit transcription-stimulatory activities. Our results also suggest that the s3.2 and s4 regions are functionally connected, and changes in s3.2 structure may affect s4 function. Therefore, the s3.2-finger deletion might have an impact on promoter-dependent transcription via allosteric changes in s4 conformation or/and its positioning in the RNA exit channel. Finally, our results provide a rational explanation to the finding that defects caused by s3.2-finger deletion are promoter-sequence dependent (Kulbachinskiy & Mustaev, 2006) .

The s subunit domain 4 as allosteric regulator of NAC
The domain s4 does not make any contact with the scaffold DNA or RNA, and therefore exerts its antipausing activity through interaction with the RNAP core. In the RNAP holoenzyme, the region 4.1 of s4 is located in the RNA exit channel and occupies the place of RNA, 13-16 nucleotides from the RNA 3' end.  (Weixlbaumer et al, 2013). The PITC complexes formed in our assay likely resemble the crystallized elemental paused elongation complexes (ePEC) formed with scaffolds the architecture of which was almost identical to ours (lacking non-template strand ssDNA) and that were trapped in a partially open clamp conformation (Weixlbaumer et al, 2013). We hypothesize that in the absence of 9-bp RNA:DNA hybrids and s4, RNAP We demonstrated that the anti-pausing activity can be observed with structurally distinct s subunits and phylogenetically distant RNAPs (E. coli and M. tuberculosis). Considering that the s4 and b-flap interaction is invariant between all classes of s subunits, we propose that s anti-pausing activity is an universal feature of initial transcription in bacteria. We hypothesize that the function mechanism of archaeal TFB and eukaryotic TFIIB, which are implicated in RNA synthesis priming, might be similar to that of the s subunit.

Scaffold-based transcription assays
Transcription reactions were performed in 5 ml of TB. 240 nM RNAP core was mixed with 1 mM of full length s or 2mM of s mutants, and incubated at 37°C for 5 min. 1mM RbpA was added when indicated. were resolved on 26% PAGE (acrylamide : bis-acrylamide ratio 10:1) with 7M urea and 1x TBE.

RNA retention assay
EcoRNAP-scaffold DNA-RNA complexes were assembled as described above except that 100 mM of 2-3-nt pRNA and 10 mM of 4-5-nt pRNA were used. pRNA7 and pRNA13 were mixed with scaffold DNA before annealing. Complexes formed in 5ml TB in Axygen® 1.7 ml MaxyClear Microtubes were incubated at 18°C for 5 min. Then, 5ml of Ni-NTA agarose beads slurry (Qiagen) in TB was added, and tubes shaken using an Eppendorf ThermoMixer ® at 18°C for 5 min. To separate the Ni-bound RNAP fraction, 0.5 ml of TB/250mM NaCl was added. Samples were briefly stirred, pelleted by centrifugation at 1000g for 1 min, and supernatants were discarded. A second washing step was performed with 50 ml of TB as before. Supernatants were removed to leave a sample volume of 10ml. [a-32 P]-UTP (0.4 mM final concentration) was added to all samples that were then incubated at 22°C for 3 min. Reactions were quenched and analyzed as above.

Promoter-based transcription assays
Transcription on the lacUV5 and galP1cons promoters was performed with 200 nM EcoRNAP, 500 nM full length s 70 or 2mM s 70 mutants, and 300 nM DNA template in TB. Samples were incubation at 37° for 10 min, and supplemented with 100 mM ApA (lacUV5 assay) or CpA (galP1cons assay) and 0.4 mM [a-32 P]-UTP. Transcription reactions were performed at 37°C for 10 min.

AsiA inhibition assay
500 nM s 70 was first mixed with 1 mM AsiA and then with 100 nM EcoRNAP core. Samples were incubated at 30°C for 10 min. Next, DNA templates were added and samples were incubated at 30°C (with lacUV5 promoter) and at 22°C (with SDT2 scaffold) for 5 min. Transcription from the lacUV5 promoter (50 nM) was initiated by adding 100 mM ApA and 0.4 mM [a-32 P]-UTP at 37°C for 5 min. Transcription from scaffold DNA (0.8 mM) was performed in the presence of 0.4 mM [a-32 P]-UTP and 25 mM NTPs or 50 mM GpC and carried out at 22°C for 3 min.

Calculation of the pause half-life times
The pause half-life times (t 1/2 = ln2/k) were calculated by fitting the fractions of RNA in pause [P i PS ]/ [P i total ] in function of the time t using the following single-exponential equation:     Models were built with COOT (Emsley & Cowtan, 2004) and UCSF Chimera (Pettersen et al, 2004).