Rhodobacter sphaeroides CarD Negatively Regulates Its Own Promoter

ABSTRACT

Bioinformatic analysis showed previously that a majority of promoters in the photoheterotrophic alphaproteobacterium Rhodobacter sphaeroides lack the thymine at the last position of the −10 element (−7T), a base that is very highly conserved in promoters in bacteria other than alphaproteobacteria. The absence of −7T was correlated with low promoter activity using purified R. sphaeroides RNA polymerase (RNAP), but the transcription factor CarD compensated by activating almost all promoters lacking −7T tested in vitro, including rRNA promoters. Here, we show that a previously uncharacterized R. sphaeroides promoter, the promoter for carD itself, has high basal activity relative to other tested R. sphaeroides promoters despite lacking −7T, and its activity is inhibited rather than activated by CarD. This high basal activity is dependent on a consensus-extended −10 element (TGn) and specific features in the spacer immediately upstream of the extended −10 element. CarD negatively autoregulates its own promoter by producing abortive transcripts, limiting promoter escape, and reducing full-length mRNA synthesis. This mechanism of negative regulation differs from that employed by classical repressors, in which the transcription factor competes with RNA polymerase for binding to the promoter, and with the mechanism of negative regulation used by transcription factors like DksA/ppGpp and TraR that allosterically inhibit the rate of open complex formation.

IMPORTANCE R. sphaeroides CarD activates many promoters by binding directly to RNAP and DNA just upstream of the −10 element. In contrast, we show here that CarD inhibits its own promoter using the same interactions with RNAP and DNA used for activation. Inhibition results from increasing abortive transcript formation, thereby decreasing promoter escape and full-length RNA synthesis. We propose that the combined interactions of RNAP with CarD, with the extended −10 element and with features in the adjacent −10/−35 spacer DNA, stabilize the promoter complex, reducing promoter clearance. These findings support previous predictions that the effects of CarD on transcription can be either positive or negative, depending on the kinetic properties of the specific promoter.

INTRODUCTION

Transcription in bacterial cells is carried out by a multisubunit DNA-dependent RNA polymerase. Bacterial RNA polymerase (RNAP) is comprised of an α subunit dimer, β, β′, and ω subunits and a dissociable σ subunit that directs RNAP to recognize specific elements in promoter DNA. When σ-associated RNAP (holoenzyme) binds promoter DNA, it initially forms a closed complex in which the promoter DNA is duplex (RPc). Through a series of isomerization steps, promoter DNA is then unwound to form the open complex (RPo). The template strand of the unwound DNA is thereby positioned in the active site of RNAP, where nucleoside triphosphate (NTP) addition is catalyzed to make a growing RNA product. As the RNA increases in length, RNAP breaks its contacts with promoter DNA, σ dissociates, and RNAP escapes the promoter (1–7).

Transcription factors can have different effects on different promoters, depending on the kinetic features of the promoter and the binding location of the transcription factor. Some transcription factors (e.g., cyclic AMP receptor protein [CRP], Fis) can activate promoters by interacting with both RNAP and promoter DNA just upstream of the RNAP binding site or can repress transcription from other promoters by binding to DNA sites within the RNAP binding site (8, 9). Other transcription factors (e.g., the MerR family) bind to the spacer DNA between the −10 and −35 hexamers, bending spacer DNA and activating or inhibiting promoter activity depending on whether or not a cofactor is present (10). Still other transcription factors (e.g., DksA) function allosterically by binding only to RNAP and not to DNA, exerting promoter specificity based only on the kinetic features of the promoters themselves (11–13). DksA activates or inhibits hundreds of genes in conjunction with the regulatory nucleotides ppGpp and pppGpp (14).

The transcription factor CarD is a small protein found in several bacterial phyla, although not in Betaproteobacteria and Gammaproteobacteria (15, 16). In contrast to DksA, which binds to the secondary channel and rim helices of the β′ subunit of RNAP (17–19), CarD binds to the lobe of the RNAP β subunit and interacts nonsequence specifically with promoter DNA near the site of interaction of RNAP with the upstream edge of the −10 element (16, 20). In contrast to DksA, which is present primarily in proteobacteria and inhibits rRNA transcription, CarD activates rRNA transcription in mycobacterial species, Thermus thermophilus, Myxococcus xanthus, and Rhodobacter sphaeroides. In several of these species, CarD has been shown to facilitate open complex formation by stabilizing RNAP-promoter complexes (16, 20–27).

In most bacterial species, including Escherichia coli and Bacillus subtilis, a thymine base in the nontemplate strand at the most downstream position in the −10 element (−7T) is more than 90% conserved among promoters (27). However, the majority of promoters in R. sphaeroides and other alphaproteobacteria that have been examined lack −7T (27, 28). The CarD homolog in R. sphaeroides can compensate for the absence of −7T in almost all of these promoters tested in vitro, including those for rRNA and at least some r-proteins in order to activate transcription (27).

carD is an essential gene in mycobacteria, Myxococcus xanthus, and R. sphaeroides (15, 21, 29), and deletion of the carD gene (also known as cdnL) affects the growth properties and rRNA expression of Caulobacter crescentus (30, 31). An early report that depletion of carD increased rRNA transcription in Mycobacterium tuberculosis implied that it functioned as a negative regulator (15), but this conclusion conflicted with the subsequent demonstration that purified CarD increased mycobacterial rRNA transcription in vitro (16). carD depletion or substitutions in the carD gene in M. tuberculosis were also reported to alter the levels of a majority of the cell’s transcripts in vivo, increasing some and decreasing others, leading to the proposal that CarD activates some promoters and inhibits others (24, 25). However, it has not been demonstrated previously that CarD inhibits the activity of any promoter directly, i.e., that the observed increase in transcription in the carD mutant did not result from the loss of activation of rRNA transcription or some other indirect mechanism.

CarD levels are constant during exponential growth in R. sphaeroides and decrease during the stationary phase (27). Even though CarD has been identified as an important transcription factor in various bacterial species, there is little information about how CarD expression itself is regulated. Here, we show that the R. sphaeroides carD promoter lacks −7T, but it nevertheless has high basal activity in vitro that results from the presence of an extended −10 element and other sequences in the spacer region near the extended −10 element. We show that CarD not only does not activate its own promoter, but, rather, it inhibits its own expression directly in vitro. We then determine the mechanism responsible for this autoregulation.

RESULTS

CarD inhibits transcription from its own promoter.

Many of the R. sphaeroides promoters tested previously, like rrnB, were inactive or had very low activity in vitro (27). A majority of R. sphaeroides promoters lack the highly conserved thymine at the last position of the −10 element (−7T; e.g., rrnB) (Fig. 1A). However, in almost all of those promoters tested, CarD compensated for the absence of −7T by activating transcription (27) (unless stated otherwise, CarD and RNAP refer to the R. sphaeroides proteins). Since the promoter for the carD gene also lacked −7T (Fig. 1A), we tested whether it was active in vitro and whether CarD increased its activity.

FIG 1.

CarD inhibits transcription from its own promoter. (A) Sequences of the R. sphaeroides rrnB promoter (−37 to +2), the R. sphaeroides carD promoter (−36 to +1), and the ColE1 plasmid-encoded RNA-1 promoter (−36 to +1). The −35 and −10 elements, transcription start sites (+1), and extended −10 element regions are shaded. Matches to the E. coli consensus sequences are in bold. The last position in the −10 element is indicated as “−7,” even though in the R. sphaeroides rrnB promoter, it is at −8 with respect to the transcription start site. (B) In vitro transcription of the R. sphaeroides rrnB promoter with R. sphaeroides RNAP (20 nM) with or without CarD (1,280 nM) (lanes 1 to 4) or of the carD promoter with or without CarD (wedge indicates CarD concentration range of 10 to 2,560 nM) (lanes 5 to 15). In the experiment pictured, CarD activated transcription from the rrnB promoter ∼13-fold, and it inhibited transcription from the carD promoter ∼4-fold (with an IC₅₀ of 120 nM). The larger band appearing in lanes 5 to 15 likely results from readthrough of the first of the tandem E. coli rrnB terminators in the plasmid downstream of the carD promoter (T1) and termination at the second terminator (T2). T1-terminated transcripts were quantified and are shown in panel C. In the experiment pictured, CarD activated transcription from the rrnB promoter ∼13-fold, and it inhibited transcription from the carD promoter ∼4-fold (with an IC₅₀ of 120 nM). (C) Quantitation of transcription from the R. sphaeroides carD promoter and the plasmid-encoded RNA-1 promoter in the presence of CarD (10 to 2,560 nM), relative to that in the absence of CarD, as in panel B. Previously published data for activation of the rrnB promoter by CarD (27) are shown for comparison with inhibition of the carD promoter by CarD. Best-fit curves were calculated with 0 nM CarD set to a relative transcription value of 1.0. Error bars indicate mean and standard deviation from 5 separate experiments.

Surprisingly, we found that the basal activity of the carD promoter (its activity in the absence of CarD) was high compared to that of the rrnB promoter or most other R. sphaeroides promoters that we had tested previously in the absence of CarD (27) (Fig. 1B, lanes 5 and 6 versus lanes 1 and 2). The high basal activity of the carD promoter could account for the high intracellular levels of CarD in R. sphaeroides cells during log-phase growth (27).

Most notably, in contrast to its activation of rrnB transcription (Fig. 1B, lanes 1 and 2 versus lanes 3 and 4), CarD inhibited its own transcription by as much as 3- to 4-fold (Fig. 1B, lanes 5 and 6 versus lanes 7 to 15) (Fig. 1C). The concentration of CarD required for half-maximal inhibition of the carD promoter (IC₅₀, 120 nM) (Fig. 1C) was similar to the concentration required for half-maximal activation of the rrnB promoter (EC₅₀, 85 nM) (27), implying that CarD might be working by the same mechanism to inhibit and activate transcription. In contrast, there was no effect of CarD on transcription from the RNA-1 promoter present on the plasmid vector into which the R. sphaeroides promoters were inserted for measuring transcription in vitro (Fig. 1). The concentrations of CarD that inhibited transcription in vitro were within the range of protein concentrations present in vivo, ∼1 μM in exponential-phase and <0.1 μM in stationary-phase cells (27).

CarD substitutions that impair activation also impair inhibition.

To examine further whether the mechanisms of activation and inhibition might result from the same interactions with RNAP, we tested the effects of a substitution for a highly conserved tryptophan residue in the DNA interaction domain, W91 in R. sphaeroides CarD. Substitutions for W91 in R. sphaeroides CarD or the corresponding residue in M. tuberculosis or T. thermophilus CarD reduced activation (16, 27). The same substitutions in R. sphaeroides CarD reduced the extent of inhibition of the carD promoter (Fig. 2A, compare lanes 3 and 4 with 5 to 6 or 9 to 10; Fig. 2B) (1.5- or 1.6-fold inhibition by W91L or W91A CarD, respectively, compared to 3.6-fold inhibition by wild-type CarD). A CarD variant with a triple substitution in the RNAP interaction domain (RID) (Q31A, I33A, R53A) that reduced the extent of activation also reduced the extent of inhibition (to only 1.3-fold) (Fig. 2A, lanes 7 and 8; Fig. 2B). Thus, the same CarD variants that caused defects in activation (26, 27) also caused defects in inhibition.

FIG 2.

Inhibition of carD promoter transcription by CarD is reduced by substitutions in CarD. (A) In vitro transcription of the carD promoter with 20 nM RNAP without CarD or with wild-type or W91A, W91L, or Q31A/I33A/R53A triple-mutant CarD (1,280 nM). (B) Transcription from the carD promoter in the presence of wild-type or mutant CarD relative to its transcription in the absence of CarD from experiments like that shown in panel A; error bars indicate range from 2 assays.

carD promoter sequence features important for high activity and inhibition by CarD.

To understand more about the properties of the carD promoter that make it subject to inhibition by CarD, the carD promoter sequence from R. sphaeroides was compared with that of other R. sphaeroides promoters (Fig. 1) (27) and with predicted carD promoters from 106 species in 6 orders of alphaproteobacteria (see Fig. S1 in the supplemental material). For illustrative purposes, Fig. 3 shows a representative subset of the alphaproteobacterial promoters from Fig. S1, as well as carD promoter sequences from a deltaproteobacterium and two actinobacterial species. Promoters were identified by using transcription start site information where available or by searching upstream of the predicted carD genes for likely −10 and −35 elements.

FIG 3.

carD promoter sequences. Alignment of the R. sphaeroides carD promoter sequence with carD promoter sequences from representatives of 6 orders of alphaproteobacteria, 1 deltaproteobacterium, and 2 alphaproteobacteria. The alphaproteobacteria are from the following orders: Rhodobacterales, R. sphaeroides and Gemmobacter aquatilis; Caulobacterales, C. crescentus and Asticcacaulis excentricus; Sphingomonodales, Novosphingobium aromaticivo and Zymomonas mobilis pomaceae; Rhizobiales, Agrobacterium tumefaciens and Sinorhizobium meliloti; Rhodospirillales, Azospirillum humicireducens and Magnetospirillum magneticum; and Sneathiellales, Oceanibacterium hippocampi (56, 57). Promoters were identified using available transcription start site data (TSS) for Novosphingobium aromaticivo (32), Zymomonas mobilis (28), Sinorhizobium meliloti (58), Mycobacterium tuberculosis (59), and R. sphaeroides (32) and for other bacteria by comparison with the R. sphaeroides carD promoter sequence. Promoters are aligned by their −10 elements. The −10 and −35 element positions that match the E. coli consensus −10 element (TATAAT) and E. coli −35 element (TTGACA) are shown in bold. The T at position −18 and the extended −10 element (TG) at positions −15 and −14, which are conserved in most of these promoters, are in bold, red, and underlined.

Like most R. sphaeroides promoters, including the rRNA promoter rrnB, the carD promoter lacks a thymine at the most downstream position in its −10 element (Fig. 1A) (27). The carD promoter has other features in common with most R. sphaeroides promoters, including a TTG sequence at the beginning of the −35 element, a 17-bp spacer between the −10 and −35 elements, and a TA at the beginning of the −10 element, but it also contains an extended −10 element (TGn) (Fig. 1A), a feature only found in about 15% of identified R. sphaeroides promoters (27, 32). Although not highly conserved in R. sphaeroides promoters in general, the extended −10 element was extremely highly conserved in alphaproteobacterial carD promoters, 100/106 of the carD promoter sequences analyzed (Fig. S1 and Fig. 3). In addition, there was also strong conservation of a T at −18 among the alphaproteobacterial carD promoters (78/106 of the alphaproteobacterial promoters analyzed) (although in the order Sphingomonodales, only 1 of 14 promoters contained −18T) (Fig. S1). A thymine at −7 was not well conserved in the set of alphaproteobacterial carD promoters analyzed (Fig. 3 and Fig. S1).

To determine whether the features identified from the alignment of carD promoters contribute to the high basal activity and/or inhibition of the R. sphaeroides carD promoter, we constructed variants with substitutions in the extended −10 element, for the T at −18, in the 6- to 7-bp region upstream of the extended −10 element, including −18, or we introduced a T at −7 (Fig. 4A). Transcription activities of the promoter variants were compared to the wild-type promoter in the absence (basal activity) or presence of a saturating concentration of CarD in vitro (1,280 nM) (Fig. 4B).

FIG 4.

Effects of carD promoter mutations on basal promoter activity and inhibition by CarD. (A) R. sphaeroides carD promoter sequence with key features indicate the −35 and −10 elements and the extended −10 element (positions conserved with the E. coli consensus are in bold), position −18 (conserved among many carD promoters as shown in Fig. 3), and the transcription start site (+1). Mutations introduced into the carD promoter are shown in red below the wild-type sequence. (B) In vitro transcription from the wild-type promoter and promoter variants with RNAP (20 nM), either without CarD (odd-numbered lanes) or with 1,280 nM CarD (even-numbered lanes). (C) Basal activity (i.e., no CarD) for each promoter variant relative to the wild-type carD promoter from experiments like that in panel B; error bars represent standard deviation from 3 to 5 assays. The activity of the carD −15 AC −14 mutant promoter was too low to quantify and is represented here as a bar at the baseline for comparison to the other promoters. P values for the difference in basal activities of the promoter variants relative to wild-type basal activity are 0.008, carD A7T; <0.001, carD −21GTCTGC-16; 0.002, carD −21TAAAGC-16; <0.001, carD −22GATTGGA; <0.001, carD −18G; and 0.002, carD −18A. See Materials and Methods for further details. (D) Inhibition of promoters by CarD, represented as transcription for each promoter in the presence of CarD relative to without CarD (error bars represent standard deviation from 3 to 5 assays). Activity without CarD for each promoter is defined as 1. P values for difference in inhibition of the promoter variants by CarD relative to inhibition of the wild-type promoter by CarD are 0.47 (not significant), carD A7T; 0.019, carD −21GTCTGC-16; <0.001, carD −21TAAAGC-16; 0.003, carD −22GATTGGA; <0.001, carD −18G; and 0.031, carD −18A.

A substitution for the extended −10 element in the wild-type carD promoter (−15, −14AC instead of −15, −14TG) eliminated detectable basal promoter activity in vitro (Fig. 4B, compare lanes 1 and 5; Fig. 4C). Single substitutions for A7 (A7T), or for T18 (T18G or T18A) reduced basal promoter activity 2- to 4-fold (Fig. 4B, lanes 3, 13, and 15, and Fig. 4C). Substitutions for −18T have been reported previously to affect transcription in some promoters in other species (33–36).

Although not quite as conserved in alphaproteobacterial carD promoters as the T at −18 or the extended −10 element, a stretch of T residues, including and just upstream of T18, is highly conserved in the order Rhodobacterales (Fig. S1). To determine whether these or other sequences in the spacer adjacent to the extended −10 element affected promoter activity, variants were created that replaced the sequence from −21 to −16 in the wild-type carD promoter (ATTTCA) with the corresponding sequence from the spacer region of the R. sphaeroides rrnB promoter, a promoter that is activated rather than inhibited by CarD (GTCTGC) (Fig. 1B) (27) or with a sequence from the spacer region of a T7A1 promoter variant with the spacer substitution GATTGGA (−22 to −16) that was reported to reduce the stability of the open complex and to reduce transcription in vitro with Thermus aquaticus RNAP (37). Each of these variants had ∼7-fold lower basal activity than the wild-type promoter (Fig. 4A; Fig. 4B, lanes 7 and 11; Fig. 4C; see also Fig. S2). In contrast, replacing the wild-type sequence from −21 to −16 with TAAAGC reduced basal promoter activity only ∼2.5-fold (Fig. 4A; Fig. 4B, lane 9; Fig. 4C). None of the spacer sequence variants altered the extended −10 element (Fig. 4A).

Taken together, these results indicate that the extended −10 element of the carD promoter (−15 and −14) and the spacer sequence just upstream of the extended −10 element (−22 to −16) are important contributors to the basal activity of the promoter. In contrast to what was observed for many other R. sphaeroides promoters (27), introduction of the consensus −7T did not increase carD transcription, but, rather, it slightly decreased promoter activity (∼1.8-fold) (Fig. 4B, lane 3; Fig. 4C).

The effects of the promoter substitutions on inhibition by CarD are shown in Fig. 4B (compare lanes “−”and “+” CarD for each template) and graphically in Fig. 4D. The A7T and T18A substitutions did not reduce the almost 3-fold inhibition by CarD observed for the wild-type promoter. However, the T18G and the TAAAGC spacer variants were less inhibited by CarD, and inhibition of two of the spacer variants (GTCTGC and GATTGGA) by CarD was almost completely abolished (Fig. 4D). Because we could not detect significant transcription from the extended −10 variant (−15 AC −14) in the absence of CarD (Fig. 2B, lane 5), we could not determine whether CarD reduced transcription further (lane 6). Taken together, the data indicate that the extended −10 sequence and the adjacent spacer region are required for carD promoter activity and for its regulation by CarD. The conservation of these sequences suggests that carD promoters in other alphaproteobacterial species might also be autoregulated.

We note that reduced basal activities of some of the carD promoter variants did not result in a switch from repression to activation by CarD. This reflects the complex contribution of multiple RNAP-promoter interactions to the kinetics of promoter open complex formation and escape of RNAP from the promoter (context effects). Thus, the propensity to be activated or repressed is complex and does not correlate simply with basal promoter activity (25).

Inhibition of the R. sphaeroides carD promoter by CarD leads to accumulation of abortive RNAs.

It was proposed that M. tuberculosis CarD might inhibit transcription from some promoters by slowing the rate of promoter escape by RNAP (24, 25, 38). Abortive RNA synthesis, in which short transcripts are synthesized repeatedly and released without dissociation of RNAP prior to the elongation phase, is characteristic of a defect in promoter escape (39–41). Therefore, we measured abortive and full-length transcript production in vitro from a short DNA fragment containing the carD promoter to test the effect of R. sphaeroides CarD on transcription from its own promoter (Fig. 5A).

FIG 5.

Analysis of abortive transcripts from the carD promoter. (A) carD promoter sequence from −36 to +50 from the 122-nt linear fragment (−72 to +50) used in the abortive transcription reactions. The −35, −10, and extended −10 elements (with positions matching E. coli consensus sequence shown in bold), position T18, and the transcription start site (+1) are all indicated. The runoff (full-length [FL]) transcript is 50 nt. (B) Image of 20% acrylamide-urea gel showing 5′ radiolabeled abortive and productive transcripts produced from the carD promoter fragment with RNAP (120 nM), with or without CarD (4,800 nM) in transcription buffer with 50 mM NaCl. The RNA size markers (lane 5), the positions of the abortive RNAs, and the full-length transcript (FL) are indicated. (C) Amount of full-length (50-nt) transcript without CarD relative to with CarD is plotted. (D) Amount of 6-mer abortive RNA from the carD promoter without CarD relative to with CarD is plotted. (E) Productive yield (FL transcript as a percentage of total product, abortive plus FL) without CarD or with CarD is plotted. The percentage yields are average values, and the error bars represent standard deviations determined from 5 assays. (F) Abortive yield-to-productive yield ratio without CarD (cyan) or with CarD (gray). (G) Abortive probabilities at positions in the ITS of the carD promoter. The abortive probabilities for each position, calculated as described in Materials and Methods and in reference 55, are shown without CarD and with CarD. The 5-mer transcript was not detected in the absence of CarD, so no value is shown. The 5-mer value in the presence of CarD and the 6-mer value in the absence of CarD are the abortive yields for those positions since abortive probabilities cannot be calculated without values for a 4-mer and 5-mer, respectively. Error bars represent mean and standard deviations from 5 assays. P values for the difference in abortive probability in the presence versus the absence of CarD for each of the following ITS positions from R-7 through R-11: R-7 (residue 7), 0.011; R-8, 0.002; R-9, <0.001; R-10, <0.001; and R-11, 0.002.

CarD inhibited full-length (50-nucleotide [nt]) transcript formation from this DNA fragment about ∼2-fold (Fig. 5B, compare lanes 3 and 4 with lanes 1 and 2; Fig. 5C). These reactions also produced abortive RNAs ranging from 5-mer to 11-mer (Fig. 5B). The sizes of the abortive products were determined by performing transcription with a subset of NTPs (see Materials and Methods, Fig. S3, and Fig. S3 legend). The yield of the abortive products was greater in the presence of CarD, correlating with the inhibition of productive RNA transcript formation (Fig. 5B, compare lanes 3 and 4 with lanes 1 and 2). For example, the presence of CarD resulted in a 5-mer abortive product that was not detected in the absence of CarD, and it increased the level of the 6-mer RNA about 2-fold (Fig. 5B and D; see legend). CarD thereby caused an ∼3- to 4-fold decrease in the productive yield (full-length transcript as a percentage of total transcripts) (Fig. 5E). The abortive-to-productive RNA ratio was about 5-fold higher in the presence of CarD (Fig. 5F), i.e., RNAP aborted 9.9 ± 2.2 times per productive cycle in the absence of CarD and 48.9 ± 14 times in the presence of CarD (for explanation of these calculations, see “Abortive transcription” in Materials and Methods).

To determine whether CarD presents a barrier to productive transcription at each position in the initial transcribed sequence (ITS), we expressed the ratios of abortive transcripts as probabilities. CarD increased the probability of RNAP aborting at each position in the ITS from the 7-mer to the 11-mer (Fig. 5G) (probabilities could not be calculated for the 5- and 6-mers since values for the 5-mer product in the absence of CarD were too low for quantitation [see Materials and Methods]). In summary, the carD promoter has a tendency to make abortive products during the early stages of transcription elongation, and CarD increases abortive product formation in the ITS, suggesting it exacerbates the promoter’s difficulty with escape. Consistent with the previously proposed hypothesis that CarD can inhibit promoter escape by increasing the stability of the RNAP-promoter complex (24), R. sphaeroides RNAP-carD promoter complex was stable to challenge with a double-stranded DNA competitor in the presence of CarD but not in the absence of CarD (Fig. S4B).

DISCUSSION

CarD can activate or inhibit transcription, depending on the identity of the promoter.

We previously showed that R. sphaeroides CarD activates many R. sphaeroides promoters, correlating strongly (albeit not absolutely) with the absence of −10 element consensus position −7T in these promoters (27). CarD protein levels varied with growth phase; they were constant during exponential growth and declined more than 10-fold as cells entered the stationary phase, suggesting that transcription from CarD-activated promoters might decrease coordinately as cells enter the stationary phase (27). However, the mechanism(s) responsible for maintenance of constant CarD levels in the exponential phase and for the decrease in CarD protein levels in the transition to stationary phase was not addressed in our previous study.

To begin to address the mechanisms responsible for regulating CarD levels, we first tested whether CarD affects its own promoter. We show here that CarD negatively regulates its own transcription 3- to 4-fold (Fig. 1B), with an IC₅₀ similar to that for activation of the R. sphaeroides rrnB promoter. The same CarD DNA binding residue, W91, is critical for inhibition and activation, suggesting that they work through similar interactions of CarD with RNAP and the promoter. A transcriptome sequencing (RNA-seq) experiment performed previously in M. tuberculosis showed that many transcripts increase in a carD mutant, implying a negative effect of CarD on transcription, but those effects could be direct or indirect (24). We demonstrate here that CarD can directly inhibit R. sphaeroides transcription in vitro.

Mechanism of inhibition by CarD.

It was shown previously that M. tuberculosis CarD wedges open the promoter minor groove just upstream of the −10 hexamer, utilizing a residue in CarD analogous to R. sphaeroides W91 to activate the M. tuberculosis rRNA promoter by stabilizing the intrinsically unstable RNAP-promoter complex (20, 23). Under our experimental conditions, CarD was not able to stabilize the R. sphaeroides rrnB promoter-RNAP complex sufficiently for it to survive a challenge by the competitor heparin, and thus, we were not able to confirm the hypothesis directly (see Fig. S4A in the supplemental material). However, we emphasize that this instability of the R. sphaeroides rrnB-RNAP complex is a function of the rrnB promoter sequence, not a property of R. sphaeroides RNAP, since R. sphaeroides RNAP forms stable open complexes on some promoters even in the absence of CarD (e.g., RNA-1) (Fig. S4A; see also reference 42).

On the basis of the increased activities of large numbers of promoters in vivo in a mycobacterial strain partially defective for CarD activity, it was predicted previously that CarD might directly inhibit some promoters that already formed stable complexes by stabilizing them further and impeding promoter escape (24, 25, 41). Supporting this prediction, the R. sphaeroides carD promoter complex was stable to challenge by a double-stranded DNA competitor in the presence but not in the absence of CarD (Fig. S4B and C). Despite the instability of the carD promoter-RNAP complex to the double-stranded DNA competitor in the absence of CarD, the ratio of abortive to productive transcripts from the carD promoter was high, characteristic of promoters that are limited for escape (38–40), and CarD strongly increased this ratio (Fig. 5F). We note, however, that it is unclear how many of the M. tuberculosis promoters whose activities increased in the carD mutant strain (24) are inhibited directly by CarD rather than by some indirect mechanism.

The mechanism of inhibition by CarD differs from that of classical repressors, which typically function by binding to the promoter and competing with RNAP (8). In contrast, the transcription factors DksA/ppGpp, TraR, and CarD bind to RNAP and can either inhibit or activate transcription, depending on the unique kinetic features of the individual RNAP-promoter complex (13, 24, 25). Although CarD, DksA/ppGpp, or TraR all affect multiple kinetic steps in the pathway, and activation or inhibition depends on the specific steps that are limiting for output from the particular promoter (25), the details of the mechanism by which CarD affects transcription differ from that of DksA/ppGpp or TraR (11–13, 43–45).

DksA/ppGpp and TraR bind to the RNAP secondary channel and alter the conformation of the DNA binding channel region of RNAP allosterically (45), resulting in effects on multiple steps in the mechanism prior to NTP addition that can affect output positively or negatively, depending on the identity of the promoter (7, 10, 11, 43–45). As described at the beginning of this section, CarD also acts on a step prior to NTP addition that can affect output positively if a promoter is limited at this kinetic step (27, 38). However, its effects on the complex can also affect output negatively by inhibiting a step after NTP addition, promoter escape. Thus, its mechanism of inhibition differs from that of both classical repressors and from DksA/ppGpp and TraR (8, 9).

There are also other transcription factors that have been shown to prevent promoter escape, albeit by a mechanism different from CarD. For example, the P4 protein of B. subtilis phage Phi29 can bind to the DNA upstream of the −35 element, interact with the α subunit C-terminal domain of RNAP to stabilize the open complex, and inhibit promoter escape (46). Even strong interactions between the α subunit C-terminal domain of RNAP and UP element DNA sequences upstream of the −35 element in the absence of a transcription factor can inhibit promoter escape in some circumstances (47).

Promoter determinants of inhibition by CarD.

The R. sphaeroides carD promoter displays properties different from promoters that are activated by CarD, accounting for its susceptibility to autoregulation. It has a relatively high basal activity even though it does not contain the thymine at −10 element position −7. In contrast, although only 15% of predicted R. sphaeroides promoters contain an extended −10 element (27), 94% of the 106 alphaprotobacterial carD promoters we analyzed (including the R. sphaeroides carD promoter) contain an extended −10 element, suggesting it plays an important role in carD transcription initiation (Fig. 3 and Fig. S1). Our mutational analysis confirmed that an extended −10 element is essential for the high basal activity of the promoter in the absence of CarD. Furthermore, sequences just upstream of the extended −10 element further affect both basal promoter activity and inhibition by CarD. The extended −10 element and the −10 hexamer have long been known to make large contributions to the stability of a promoter complex (48, 49). We suggest that substitutions in the extended −10 element eliminated basal transcription because they destabilized the complex so much that it decayed before catalysis.

However, several lines of evidence indicate that the presence of an extended −10 element alone is not sufficient to cause inhibition by CarD. First, replacement of the spacer sequence just upstream of the R. sphaeroides carD promoter significantly reduced inhibition by CarD (Fig. 4), indicating that this spacer region plays an important role in the mechanism of inhibition. Second, introduction of an extended −10 element into four CarD-activated R. sphaeroides promoters (including rrnB) increased the basal activities of three of the promoters, but none of the four promoters were inhibited by CarD (27). Third, tests of effects of CarD on six native R. sphaeroides promoters that contain an extended −10 sequence indicated that two of the promoters were not inhibited by CarD, whereas four others were activated by CarD (27). Thus, we suggest that the effects of the extended −10 element on inhibition by CarD are highly context dependent, and the extended −10 element cannot be used by itself to identify negatively regulated promoters (see also reference 50). No other R. sphaeroides promoters that are negatively regulated by CarD have been identified to date.

RNAP interactions with features of the spacer region of the carD promoter located just upstream of the extended −10 sequence contribute to repression of the carD promoter by CarD. However, our information about spacer features is limited by the small number of variants tested, so the precise features responsible have not been identified. Previous studies on other promoters have implicated sequences in the spacer upstream of the −10 element on promoter function (33–37, 51). In one study, this spacer region sequence was termed the “Z element” (37). Although the Z element was proposed to be a base-specific RNAP recognition element in one promoter context, subsequent structural studies on another RNAP-promoter complex suggested that the region of RNAP that interacts with this sequence (the β′ zipper) does not interact with the DNA base specifically, but, rather, it interacts with the DNA backbone (52). In another study, sequences in the spacer interacted with the nonconserved region of σ⁷⁰ (36). Two recent studies (34, 35) implicated position −18 (i.e., a position within this section of the spacer) in promoter function. The DNA backbone at this position interacts with a residue in the linker between σ regions 3 and 2 (R451), and mutational analysis showed that the presence of a G at position −18 correlated with reduced promoter activity. We show here that −18G in the carD promoter not only reduced basal activity but also reduced the extent of negative regulation by CarD. We suggest that interactions of RNAP with this part of the spacer region of the CarD promoter contribute to the stabilization of the complex and that these interactions, together with those of CarD to the upstream edge of the −10 element, reduce the rate of promoter escape.

Recent structural data show that a loop in σ, the σ region 3.2 finger, sterically interferes with extension of the growing RNA chain in the initiation complex beginning at a length of 5 nt (53). Our observation that CarD increased 5-mer abortive product formation is consistent with the model that the mechanism of inhibition of promoter escape by CarD at least in part involves stabilization of σ within the promoter complex. The CarD-dependent appearance of a 5-mer product in our R. sphaeroides abortive initiation experiments is also consistent with observations in reference 38 that M. tuberculosis CarD slowed promoter escape and enhanced a transient fluorescent signal associated with the formation of early transcription complexes (including those producing a 5-mer) in stopped-flow kinetic assays.

Potential role of autoregulation in determining levels of CarD during log-phase growth.

We suggest that autoregulation of the CarD promoter could be a mechanism for maintaining homeostasis during log-phase growth. The IC₅₀ needed for inhibition, the EC₅₀ needed for activation, and the concentration of CarD present during aerobic growth in log phase suggest that the carD promoter-RNAP complex would be substantially but not fully occupied by CarD during log phase. Thus, a change in the CarD concentration would lead to widespread changes in the transcriptome, including many genes involved in ribosome synthesis (27, 29). Feedback control of CarD synthesis might be sufficient to keep CarD levels perched at the concentration required for activation of transcription during log phase, thus contributing to homeostasis (27). Our data, of course, do not rule out roles for regulation of CarD levels at additional steps in gene expression (e.g., mRNA stability, translation) or on regulation of CarD protein stability.

Reduction of CarD levels in stationary phase, and thus, loss of activation of the many promoters it controls, likely plays a major role in shifting the makeup of the transcriptome as cells enter stationary phase in R. sphaeroides, e.g., converting the bacterium’s emphasis on production of the translation apparatus to production of proteins needed for survival in times of low nutrient availability. However, transcriptional feedback inhibition cannot account for the decrease in CarD protein levels observed when cells enter stationary phase since CarD is an inhibitor of CarD expression and CarD levels decrease at this time. Previous work has implicated proteolysis by ClpXP in regulation of C. crescentus and Mycobacterium smegmatis CarD levels (30, 54). Figure S5 shows that the C. crescentus and R. sphaeroides CarD proteins share the C-terminal AA recognition motif for proteolysis by ClpXP. Future studies will be required to address whether the decrease in CarD levels in stationary phase in R. sphaeroides results from increased protein turnover mediated by ClpXP or whether there is an antisense transcript for the carD mRNA as recently reported for Mycobacterium (54).

Finally, we note that there are other regulators of at least some CarD-activated promoters, e.g., ppGpp and its cofactor DksA (42, 55). The amounts of these regulators likely work concurrently with changes in CarD concentrations to regulate expression from particular promoters. For example, an increase in ppGpp concentrations and the decrease in CarD concentrations may work together to reduce rRNA production during the transition to stationary phase. Examining the ways in which different regulators work together to control gene expression in R. sphaeroides (and other bacterial species) remains a challenge for the future.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Bacterial strains are listed in Table S1 in the supplemental material. E. coli strains were grown at 37°C in either LB Lennox medium or on LB agar plates supplemented with kanamycin (30 μg/ml) or chloramphenicol (20 μg/ml) as required.

Purification of R. sphaeroides RNAP holoenzyme and SUMO-tagged wild-type and variant CarD proteins.

R. sphaeroides holoenzyme was purified using Ni-agarose affinity chromatography as described (27). The R. sphaeroides carD gene was cloned into the pETSUMO vector for overexpression and purification from E. coli BL21 using the IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible T7 promoter, and carD variants were constructed by Multisite Lightning Quik Change mutagenesis as described (27). Untagged purified CarD proteins were stored at −20°C in storage buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 2 mM dithiothreitol [DTT], and 50% glycerol). Protein concentrations were determined by the Bradford assay using bovine serum albumin (BSA) as a standard. For in vitro transcription assays, CarD dilutions were prepared in CarD storage buffer.

Construction of plasmids for use in in vitro transcription.

An R. sphaeroides carD promoter fragment (−160 to +50 with respect to the transcription start site, based on transcription start site data (32) and the predicted −10 and −35 elements, was amplified by PCR from R. sphaeroides 2.4.1 chromosomal DNA using oligonucleotides 8006 and 8131 (Table S2). The resulting promoter fragment contained sequences for ligation to linearized pRLG770 (amplified using oligonucleotides 7057 and 7058) by Hi-Fi Assembly (NEB). Assembly reactions were electroporated into E. coli XL1-Blue, and the sequence of a transformed plasmid with the expected insert size on gels was determined with oligonucleotides 4252 and 4253 (p15221). carD promoter variants were created as derivatives of p15221 by Multisite Lightning Quik Change (Agilent) using the mutagenic primers listed in Table S2. Plasmid DNA for use in in vitro transcription reactions was purified from E. coli XL1-Blue by miniprep (Qiagen), phenol extracted, diluted to 200 ng/μl, and stored at −20°C.

In vitro transcription.

Multiple-round transcription reactions contained supercoiled plasmid templates with the promoters listed in Table S1. Twenty-five-microliter reaction mixtures contained transcription buffer (20 mM Tris-HCl [pH 7.9], 10 mM MgCl₂, 1 mM DTT, 0.1 mg/ml BSA, and 170 mM NaCl), NTPs (200 μM ATP, GTP, CTP, 10 μM UTP, and 1.2 μCi [α-³²P]UTP [PerkinElmer]), CarD (or CarD storage buffer), and template DNA (200 nM).

Transcription was initiated by addition of R. sphaeroides RNAP (20 nM) at 30°C for 20 min. The reactions were terminated by addition of 25 μl water-saturated phenol, aqueous and phenol phases were separated by brief centrifugation, and 16 μl of the aqueous phase containing the transcript was transferred to 16 μl 2× loading solution (8 M urea, 10 mM EDTA, 2× Tris-borate-EDTA [TBE], 1% SDS, and 0.02% bromophenol blue). Transcripts were denatured by heating the samples at 95°C for 1 min, resolved on 6% polyacrylamide 7 M urea gels, imaged using a Typhoon phosphorimager, and quantified with ImageQuant 5.2 software.

For the data shown in Fig. 4C, we performed a two-tailed t test (assuming unequal variances) between the wild-type and mutant basal transcription signals. For the data shown in Fig. 4D, we analyzed the significance of the difference between the ±CarD ratio for each promoter variant and the ±CarD ratio for the wild-type promoter using a two-tailed t test (assuming unequal variances). This particular statistical analysis tests whether there is a difference between the two groups, with the assumption that the variances are unequal.

Abortive transcription.

A carD promoter fragment was amplified by PCR from p15221 using oligonucleotides 8660 and 8661 (Table S2) resulting in a 122-nt linear transcription template (−72 to +50). The PCR product was of uniform size on an analytical 2% agarose gel, purified using a Qiagen PCR clean-up kit, and resuspended in water. Abortive transcription reactions were carried out as previously described (1). Ten-microliter reaction mixtures contained template DNA (100 nM); transcription buffer (50 mM NaCl, 10 mM Tris-HCl [pH 8], 10 mM MgCl₂, and 0.1 mM DTT); 500 μM GTP, CTP, and UTP; 25 μM ATP; 8 μCi [γ-³²P]ATP [Perkin Elmer]); and 4,800 nM CarD (or CarD storage buffer). Reactions were initiated by the addition of R. sphaeroides RNAP (120 nM) for 30 min at 30°C and terminated by the addition of 10 μl stop solution (95% formamide, 10 mM EDTA, 0.05% xylene cyanol, and 0.05% bromophenol blue). Transcripts were denatured by heating at 90°C for 30 s and resolved on a denaturing gel (20% bisacrylamide [19:1 acrylamide-bisacrylamide], 7 M urea, and 0.5× TBE) on an electrophoresis apparatus with an anode buffer containing 0.5× TBE and a cathode buffer containing 0.5× TBE and 0.75 M sodium acetate. The gel was wrapped in Saran Wrap and exposed to a phosphorimager screen overnight. Transcripts and abortive products were imaged on a Typhoon phosphorimager and quantified with ImageQuant 5.2 software.

The abortive yield was calculated by dividing the total signal for the abortive RNAs by the total signal for each RNA species in the lane. The productive yield was calculated by dividing the signal for the full-length transcription product by the total signal for each RNA species in the lane. The abortive-to-productive ratio was calculated by dividing the abortive yield by the productive yield. Equations used to calculate abortive probability were derived from reference 40, where X_i is the abortive yield of the ith abortive RNA. The denominator gives the fraction of RNA polymerase-promoter complexes that survives to the ith position. The abortive probability for the ith position, P_i, in percent, is the probability that the transcript will be aborted from the initial transcription complex at the ith position.

$${\mathit{P}}_{\mathit{i}}=\frac{{\mathit{X}}_{\mathit{i}}}{100\%-{\displaystyle {\sum }_2^{i–1}{\mathit{X}}_{\mathit{i}}}}$$

This equation cannot be used to determine the abortive probability for ITS positions where the probability for the previous position is not known. Since the 5-mer product is not detected in the absence of CarD, the values shown in Fig. 5G for the 5-mer and 6-mer in the absence of CarD are the abortive yields (ratio of abortives at that position to total products). In the absence of CarD, the 6-mer is the first abortive product detected, and therefore, the abortive probabilities ± CarD for the 5-mer to 6-mer step cannot be determined. There was no statistical difference (tested by the two-tailed t test for variance) observed for ITS position 6 ± CarD, suggesting that this position does not represent an energy barrier for promoter escape.

Potential products smaller than 5-mer could not be quantified because they would have comigrated on the gel with radiolabeled material present in the [γ-³²P]ATP preparation used to label the product RNAs (Fig. S3C). The size marker ladder was generated from radiolabeling RNA from the Decade markers system (Invitrogen) using the manufacturer’s protocol with the exception that the ladder was not heated before loading onto gel.

Determination of abortive RNA length.

To determine the length of each abortive RNA from the carD promoter, transcription was performed as described above with an NTP mix containing 500 CTP, 25 μM ATP, and 8 μCi [γ-³²P]ATP). The initial transcribed region of the carD promoter contains only A and C for the first 8 nt (Fig. S3A). The NTP mix with only ATP and CTP produced a prominent RNA for the 8-mer RNA, as RNAP could not elongate to the 9th position, which was a G.