ABSTRACT
The CarD-CarG complex controls various cellular processes in the bacterium Myxococcus xanthus including fruiting body development and light-induced carotenogenesis. The CarD N-terminal domain, which defines the large CarD_CdnL_TRCF protein family, binds to CarG, a zinc-associated protein that does not bind DNA. The CarD C-terminal domain resembles eukaryotic high-mobility-group A (HMGA) proteins, and its DNA binding AT hooks specifically recognize the minor groove of appropriately spaced AT-rich tracts. Here, we investigate the determinants of the only known CarD binding site, the one crucial in CarD-CarG regulation of the promoter of the carQRS operon (PQRS), a light-inducible promoter dependent on the extracytoplasmic function (ECF) σ factor CarQ. In vitro, mutating either of the 3-bp AT tracts of this CarD recognition site (TTTCCAGAGCTTT) impaired DNA binding, shifting the AT tracts relative to PQRS had no effect or marginally lowered DNA binding, and replacing the native site by the HMGA1a binding one at the human beta interferon promoter (with longer AT tracts) markedly enhanced DNA binding. In vivo, however, all of these changes deterred PQRS activation in wild-type M. xanthus, as well as in a strain with the CarD-CarG pair replaced by the Anaeromyxobacter dehalogenans CarD-CarG (CarDAd-CarGAd). CarDAd-CarGAd is functionally equivalent to CarD-CarG despite the lower DNA binding affinity in vitro of CarDAd, whose C-terminal domain resembles histone H1 rather than HMGA. We show that CarD physically associates with RNA polymerase (RNAP) specifically via interactions with the RNAP β subunit. Our findings suggest that CarD regulates a light-inducible, ECF σ-dependent promoter by coupling RNAP recruitment and binding to a specific DNA site optimized for affinity and position.
INTRODUCTION
A global regulatory complex formed by two proteins, CarD and CarG, acts in regulating multicellular fruiting body development, the transcriptional response to light culminating in carotenoid synthesis, and a number of other processes in the Gram-negative soil bacterium Myxococcus xanthus (1–4). CarD and CarG are found only in a number of other myxobacteria. One striking feature of CarD and its orthologs in the closely related myxobacteria Stigmatella aurantiaca, Myxococcus fulvus, and Corallococcus coralloides is that these bacterial proteins have a DNA binding domain resembling eukaryotic high-mobility-group A (HMGA) proteins (5, 6). The latter are small, relatively abundant, minor-groove DNA binding factors that remodel chromatin in the assembly of specific nucleoprotein complexes essential in transcription, replication, recombination, and repair (7, 8). HMGA proteins have multiple copies of the RGRP or AT hook DNA binding motif embedded among less conserved basic and proline residues and a flanking highly acidic segment (9). Intrinsically unstructured, they bind in a defined conformation to the minor groove of a DNA site made up of at least two appropriately spaced AT-rich tracts, with each tract being ≥4 bp (10–12). M. xanthus CarD has a C-terminal domain with four AT hooks that resembles human HMGA1a (with three AT hooks) in its physical, structural, and DNA binding properties (Fig. 1A) (13–15). In contrast to mammalian HMGA, CarD also has a structurally well-defined N-terminal domain of ∼180 residues which is essential for function and defines the widely distributed CarD_CdnL_TRCF family of bacterial proteins or protein modules that have been implicated in interactions with RNA polymerase (RNAP) (5, 13–17). Through this domain CarD forms a stable complex with CarG, which is indispensable in every known CarD-dependent process (13, 15). CarG has an H/C-rich zinc binding motif similar to archaemetzincin metalloproteases (15, 18) but lacks protease activity and does not bind DNA directly (13, 15). CarG thus appears to function as a transcriptional adaptor. Interestingly, the CarD-CarG regulatory complex, composed of an HMGA-like DNA binding factor and a factor that does not bind DNA, is reminiscent of HMGA1a and the p300/CBP adaptor in a eukaryotic enhanceosome, a large nucleoprotein assembly that also includes the basal transcriptional machinery and gene-specific regulators (3, 15, 19).
CarD, human HMGA1a, and their respective binding sites at the carQRS and human beta interferon (IFN-β) promoter regions. (A) Amino acid sequences of CarD (316 residues; NCBI accession number CAA91224) and human HMGA1a (107 residues; NCBI accession number P17096), with the DNA binding AT hooks highlighted by the gray box and the flanking acidic region underlined. The ∼180-residue CarD N-terminal region absent in human HMGA1a that is implicated in interactions with RNA polymerase and CarG is in italics. (B) DNA sequence of the carQRS promoter region showing the TTT tracts of the CarD binding site (boldface), as well as the −10 and −35 promoter regions, the transcription start site (+1), and the initiator ATG codon (boldface and underlined). The 169-bp and 222-bp fragments used as DNA probes in EMSA and DNase I footprinting assays correspond to positions −117 to +52 and −170 to +52, respectively. (C) DNA sequence of the human IFN-β promoter, with the PRDII/NRDI and PRDIV control elements (gray boxes), the TATA element (white box), and the transcriptional start site (+1) indicated. The AT-rich tracts are in boldface. The dotted line with the arrowheads facing in opposite directions spans the interferon response element (IRE).
We have found that the CarD C-terminal HMGA-like domain can be replaced with no loss of function not only by human HMGA1a but also by human histone H1 or its C-terminal region (H1-CTR) or by the domain resembling H1-CTR in the CarD ortholog found in the myxobacterium Anaeromyxobacter dehalogenans (CarDAd) (13). CarDAd forms a stable complex with CarGAd (a CarG ortholog in A. dehalogenans) but not with M. xanthus CarG, and, despite the lower DNA binding affinity of CarDAd relative to that of CarD in vitro, the CarDAd-CarGAd pair could functionally replace the CarD-CarG pair in M. xanthus (13). HMGA and H1-CTR are both intrinsically disordered domains, and this has been invoked to explain why the two are equally versatile in directing CarD function (13). Even so, the remarkable functional plasticity of the DNA binding domain then raises the question as to what are the determinants for site-specific DNA recognition by CarD in carrying out its function. We addressed this in the present study by carrying out a detailed analysis of a site upstream of the light-inducible promoter of the carQRS operon (PQRS), (Fig. 1B), the only one that has thus far been experimentally demonstrated to be important for both CarD binding and function (6, 13–15). Expression from PQRS is driven by the RNAP holoenzyme containing CarQ, an alternative σ factor of the extracytoplasmic function (ECF) family (20). It also absolutely requires the CarD-CarG complex and a site that spans position −65 to −77 relative to the PQRS transcription start site that is important for CarD function in vivo and to which CarD binds in vitro (6, 13–15). This CarD binding site contains two optimally phased 3-bp-long AT tracts (Fig. 1B). It is also specifically recognized by human HMGA1a in vitro (13–15). This is noteworthy since it has been reported that mammalian HMGA binding and function require each AT tract to be at least 4 bp long (11, 12). This is the case, for example, with HMGA1a binding to the interferon (IFN) regulatory element (IRE) (Fig. 1C), an event that is crucial in the assembly of a transcriptionally competent enhanceosome at the human beta interferon (IFN-β) gene. Here, a molecule of HMGA1a binds to the PRDIV segment with two AT tracts, one 5 bp long and the other 4 bp, and another HMGA1a molecule binds to a second pair of AT tracts (again, 5 and 4 bp long) at the PRDII/NRDI region (Fig. 1C) (21).
To probe the determinants of CarD binding at PQRS, we examined the effects on DNA binding in vitro of the following: (i) mutating one or both AT tracts of the CarD recognition site; (ii) shifting the position of the AT tracts relative to PQRS; and (iii) replacing the entire CarD binding site at PQRS by the segment corresponding to the IRE PRDII/NRDI region, whose AT tracts are longer. The consequences of these changes in the CarD binding site on PQRS expression in vivo were then investigated in wild-type (WT) M. xanthus, as well as in a strain where CarD and CarG are replaced by CarDAd and CarGAd, respectively. Our results indicate that the natural CarD binding site is highly specific and appears to have evolved for optimal rather than maximal affinity and that its position relative to PQRS is key for CarD regulation to be operative in vivo. We discuss the implications of these findings in the context of the association that we show occurs between CarD and RNAP.
MATERIALS AND METHODS
Strains, plasmids, and growth conditions.M. xanthus strains and plasmids used in this study are listed in Tables 1 and 2, respectively. M. xanthus vegetative growth was carried out at 33°C in the rich Casitone-Tris (CTT) medium as described previously (13); when required, the antibiotic kanamycin (Km) was added to a final concentration of 40 μg/ml. Escherichia coli DH5α was used for plasmid constructions, and strain BL21(DE3) was used for protein overexpression. These were grown in Luria broth at 37°C or, for protein overexpression, at 18 or 25°C overnight.
Myxococcus xanthus strains
Plasmids
Plasmid and strain construction.Plasmids were constructed using standard protocols. M. xanthus genomic DNA was isolated using a Promega Wizard kit. Transcriptional fusions of a reporter lacZ gene to PQRS bearing the wild-type (WT) or a variant CarD binding site were constructed in pMR3183, which bears a multicloning site, a Kmr-positive selection marker, and a 1.38-kb M. xanthus DNA fragment for chromosomal integration at a heterologous site by homologous recombination. With M. xanthus genomic DNA as the template and appropriate primers, the 240-bp fragment spanning position −170 to +70 relative to the transcription start site of PQRS was PCR amplified with a 5′ PstI site and a 3′ XbaI site. The product was purified and cloned into the PstI and XbaI sites of pMR3183 to yield pMR3236, which bears the PQRS-lacZ transcriptional fusion. The PCR overlap extension method was employed for site-directed mutagenesis (25) with pMR3236 as the template and appropriate primers to generate the changes in the CarD binding site mentioned in the text (except for Mut5b and Mut10b, which were synthesized [Genscript] with the required displacement of the CarD binding site in the 240-bp fragment) and then cloned into the PstI and XbaI sites of pMR3183. Every construct was verified by DNA sequencing. The plasmid construct with the required PQRS-lacZ (containing the wild-type or the given CarD binding site change) was introduced into the indicated M. xanthus strain (Table 1), and transformants that had the plasmid integrated into the chromosome by homologous recombination were selected using Kmr selection.
β-Galactosidase activity.Reporter lacZ expression was qualitatively assessed by the blue appearance of spots on CTT plates containing the required antibiotic, when necessary, and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal; 40 μg/ml), and by β-galactosidase-specific activity measurements (in nmol of o-nitrophenyl β-d-galactoside hydrolyzed/min/mg protein) in a SpectraMax 340 microtiter plate reader (Molecular Devices), as described elsewhere (13). Briefly, cell cultures were grown in CTT medium to early exponential phase in the dark, divided into two, and grown for a further 14 h, one half of the cultures in the dark and the other half in the light; β-galactosidase-specific activities were then estimated. All β-galactosidase activities reported in this study are expressed as a percentage of the value for the wild-type probe (WT) in the light (148 ± 16 nmol of o-nitrophenyl β-d-galactoside hydrolyzed/min/mg protein), under which PQRS expression is induced. The value obtained in the dark for all the reporter probes used was 5.3 ± 2.4 nmol of o-nitrophenyl β-d-galactoside hydrolyzed/min/mg protein. The standard error from at least three independent measurements is shown scaled relative to an activity of 100% for WT in the light.
Protein purification.CarD was overexpressed as a fusion to intein, from which it was subsequently released by intramolecular cleavage. Native human HMGA1a was expressed from a construct in the pET11b overexpression vector. Expression, purification, and concentration determination of CarD or HMGA1a employed procedures described previously (13, 14).
DNA binding assays.DNA binding in vitro was examined using 32P-5′-end-labeled probes spanning position −117 or −170 to +52 relative to the transcription start site of the carQRS promoter region (Fig. 1B). All probes were generated by PCR (except for Mut5b and Mut10b, which were synthesized by GenScript USA, Inc.), and 32P 5′ end labeled at the coding strand as described elsewhere (5). Electrophoretic mobility shift assays (EMSA) were carried out in 20-μl total reaction volumes of binding buffer (50 mM NaCl, 15 mM HEPES, 4 mM Tris, pH 7.9, 1 mM DTT, 10% glycerol, 1 mg/ml bovine serum albumin [BSA], and 0.1% Nonidet P-40) containing 1 to 3 pM of end-labeled probe (∼13,000 cpm), the amount of protein indicated in the corresponding figures and legends, and 1 μg of double-stranded poly(dG-dC) as nonspecific competitor DNA. After a 30-min equilibration period at 4°C, DNA binding was analyzed by electrophoresis for 30 min in nondenaturing 6% polyacrylamide minigels (37.5:1, acrylamide/bis-acrylamide), prerun at 300 V at 4°C for 30 min in 0.5×TBE buffer (45 mM Tris base, 45 mM boric acid, 1 mM EDTA) using a Mini-Protean II system (Bio-Rad Laboratories Inc.). Gels were dried and analyzed by autoradiography. Identical conditions were used in DNase I footprinting assays except that 10 mM MgCl2 was added to the incubation buffer, and the reaction mixture was treated with 0.7 U of DNase I for 2 min at 4°C, followed by quenching with EDTA. After phenol extraction and ethanol precipitation, samples were run in 6% 8 M urea polyacrylamide gels.
Two-hybrid analysis.The E. coli two-hybrid system employed is based on functional complementation of the T25 and T18 fragments of the Bordetella pertussis adenylate cyclase catalytic domain (CyaA) when two test proteins interact (23). pKT25-carD and pUT18-carG have been described before (13, 15, 16). Coding sequences and flanking regions for each M. xanthus RNAP subunit were obtained from genome data (NCBI accession numbers YP_631525, YP_631280, YP_631281, and YP_633047, respectively, for the α, β, β′, and ω subunits of RNAP). These coding sequences as well as carQ were PCR amplified from genomic DNA with 5′ XbaI and 3′ BamHI (carQ, α, β, and ω) or 5′ XbaI and 3′ EcoRI (β′) restriction sites and cloned into pKT25, pKNT25, pUT18, and pUT18C. A given pair of pUT18/pUT18C and pKT25/pKNT25 hybrid constructs was introduced into E. coli strain BTH101 (cya mutant) by electroporation. As negative controls, pairs in which one construct corresponded to the empty vector were used. Interaction was assessed qualitatively from the intensity of the blue that developed on LB-Amp-Km plates with 40 μg/ml of X-Gal due to reporter lacZ expression.
RESULTS
CarD requires both AT tracts for PQRS activation in vivo and DNA binding in vitro.CarD, as mentioned earlier, exhibits minor-groove binding to a DNA stretch that contains two 3-bp AT-rich tracts separated by 7 bp, located in the region of −65 to −77 relative to the PQRS transcription start site (13, 14). We first examined how modifying the AT tracts of this site affects PQRS expression in vivo and DNA binding in vitro. Figure 2A shows a comparison of the native CarD binding site (WT) with the sites altered in the left (Mut-L), the right (Mut-R), or both (Mut-LR) AT tracts, which were used in probes for EMSA and DNase I footprint analysis of DNA binding in vitro and fused to a lacZ reporter gene for analyzing PQRS expression in vivo. As can be seen in Fig. 2B, light-induced activation of the PQRS-lacZ reporter in vivo was impaired with the Mut-L, Mut-R, or Mut-LR probe. In EMSA, the single retarded complex that occurs in vitro from specific binding of CarD to the WT probe was not detected with the Mut-L, Mut-R, or Mut-LR probe (Fig. 2C). With the WT probe, increasing concentrations of CarD generate a DNase I footprint that spans positions −78 to −60, where the two AT-rich tracts and the DNA in between are specifically protected (with the exception of position −68, immediately upstream of the AT tract that lies closer to the transcriptional start site) (Fig. 2A and D). Consistent with the EMSA results, no DNase I footprint was observed with the Mut-L, Mut-R (Fig. 2D), or Mut-LR probe (not shown). Thus, mutating either one of the two AT tracts of the CarD binding site is sufficient to abolish DNA binding in vitro and to block light activation of PQRS in vivo.
Mutating the TTT tracts impairs PQRS activity in vivo and DNA binding in vitro. (A) Schematic representation of the PQRS-lacZ reporter probe showing the sequence of the wild-type (WT) CarD binding site or its mutated versions (Mut-L, Mut-R, and Mut-LR); the numbers above the WT sequence indicate the positions relative to the transcription start site (+1). The stretch protected against DNase I by CarD is indicated by the line below the WT sequence (thicker for the TTT tracts). The numbers correspond to the first and last nucleotides in the protected stretch, and the asterisk indicates a site that is not protected from DNase I digestion, based on the data in panel D. (B) Light-induced expression of the PQRS-lacZ reporter probe containing the WT or the indicated mutant CarD binding site. Values are shown relative to the estimated value with the WT, as described in Materials and Methods. (C) EMSA using the 169-bp DNA probe with the WT CarD binding site or its mutant versions indicated at the bottom. −, no CarD; +, 1.5 μM CarD. (D) DNase I footprint of CarD binding to WT or mutant probes as indicated (bottom), with the coding strand 32P 5′ end labeled. CarD concentrations increase in the order 0, 3.75, 7.5, and 11.25 μM. The line on the left (thicker for the TTT tracts) indicates the region that is protected against DNase I in the WT as the concentration of CarD increases. The numbers demarcate the first and last nucleotides in the protected stretch relative to the transcription start site (+1), and the asterisk marks position −68.
DNA distortion and enhanced affinity for CarD accompany substitution of the native CarD binding site at PQRS by the HMGA1a binding site at the IRE.The CarD DNA binding domain, as mentioned in the introduction, resembles human HMGA1a in its physical, structural, and DNA binding properties (14), and we have shown that human HMGA1a can replace the CarD C-terminal DNA binding domain in M. xanthus with no loss of function (13). We therefore asked how this functional equivalence translates at the level of the DNA site involved, a question pertinent both in the context of a detailed analysis of the natural CarD site and in the broader one of HMGA DNA binding. Several previous studies have shown that HMGA1a binds to the minor groove of AT tracts with various sequences due to the recognition of a specific aspect of B DNA conformation (11, 12). A single AT tract at least 4 bp long has been reported to be sufficient for low-affinity binding, while high-affinity HMGA1a sites typically correspond to two or more appropriately spaced AT tracts, each at least 4 bp in length, with spacing of 6 to 8 bp correlating with maximum affinity (11). It is therefore striking that both AT-rich tracts of the CarD binding site at PQRS are only 3 bp long. Hence, we studied the consequences of having the CarD binding site at PQRS replaced by the HMGA1a IRE site at the human IFN-β enhancer, whose AT tracts are longer.
We first examined how CarD binding in vitro was affected when the 15-bp stretch from position −77 to −63, which includes the AT tracts and the 7-bp spacer at PQRS, was replaced by the 15-bp stretch encompassing the two longer AT tracts and a 6-bp spacer at the IRE of the human IFN-β enhancer (Fig. 1C and 3A). Thus, in the new construct (referred to as the IRE) the upstream AT tract starts at position −77, as in the native CarD binding site. Within the constraints imposed by the length of the AT tracts and the spacing between them in the IRE, this DNA swap largely conserves the center of the site relative to the PQRS transcription start site (position −71 for the native CarD site and −70 for the IRE) and the flanking sequences. Figure 3B shows that CarD generates a clear DNase I footprint in the region that contains the two AT-rich tracts of the IRE at a concentration about 3-fold lower than that required with the WT, indicating that CarD binds to the IRE with a greater affinity than to the WT. The DNase I footprint at the IRE resembles that observed with the wild-type probe in that the two AT tracts are mostly protected from digestion and are in the span of the footprint, whose downstream edge appears to be about 3 bp closer to the transcription start site than the native site (footprint from positions −77 to −57 for IRE versus −78 to −60 for the wild-type probe) (Fig. 3). As with the WT probe, the position immediately upstream of the AT tract that lies closer to the transcriptional start site (−67) is not protected; rather, it appears hypersensitive. The DNase I digestion pattern with the IRE is additionally characterized by two other hypersensitive sites, at positions −66 and −77, in the presence or absence of CarD. Generally, DNase I-hypersensitive sites are indicative of a local bending toward the major groove to produce a wider and more DNase I-susceptible minor groove (26). In the context of the PQRS region, the presence of the DNase I-hypersensitive sites within the footprint at the IRE, even in the absence of protein, suggests that this segment is intrinsically more distorted than the WT. This is also consistent with the intrinsic DNA bend reported for the IRE in its natural context in the human beta interferon promoter (27). It is also interesting that human HMGA1a yields a DNase I footprint with the IRE probe that matches very well the one observed for CarD (Fig. 3C), consistent with the two proteins sharing similar DNA binding properties (13, 14). Thus, these analyses indicate that CarD binds to the IRE with a greater affinity than to the WT and that, in the context of the PQRS promoter region, the IRE is intrinsically distorted compared to the WT.
Binding of CarD in vitro to the PQRS promoter region with the IRE replacing the native CarD binding site. (A) Schematic showing the PQRS promoter region with the native CarD binding site (WT) and that with it replaced by the IRE, the HMGA1a binding site. The line (thicker at the AT tracts) below each sequence spans the stretch protected against DNase I by CarD. The numbers correspond to the first and last nucleotides in the protected stretch, and the asterisks indicate the DNase I-hypersensitive sites from the data in panel B. (B) DNase I footprint analysis comparing CarD binding at the indicated concentrations to the 169-bp probe with the WT and IRE sites (note the significantly lower CarD concentrations required for binding to the IRE site). In both probes, the coding strand was 32P 5′ end labeled. Regions protected by CarD against DNase I are indicated by the lines (thicker at the AT tracts) on the left (for WT) and on the right (for IRE), with the numbers above and below being the positions relative to the transcription start that demarcate the protected region. (C) DNase I footprint of CarD (3.75 μM) or human HMGA1a (0.75 μM) binding to the 169-bp probe with the IRE site. In panels B and C, asterisks indicate positions in the binding site that are not protected against DNase I or that are hypersensitive to digestion.
Replacing the native CarD binding site by the IRE blocks light-induced expression of PQRS in vivo.Given that human HMGA1a can replace the CarD C-terminal DNA binding domain in M. xanthus with no loss of function (13) and that CarD binds to the human HMGA1a site, the IRE, in vitro, we next asked if a human HMGA1a site can serve as a functional CarD site in M. xanthus. For this, we checked how light-induced expression of PQRS in vivo is affected if the native site at PQRS is replaced by the human IRE site, to which CarD binds with a higher affinity in vitro. Interestingly, the presence of the IRE at PQRS abolished activation of the PQRS-lacZ reporter in the light (Fig. 4B), suggesting that the tighter binding of CarD to this intrinsically distorted site is not conducive to proper functioning in vivo. Previous studies have shown that mutations within either AT tract of the IRE decrease the binding affinity of HMGA1a in vitro, and while the IRE behaved as one multivalent, high-affinity HMGA1a binding site, the single-AT tract IRE variants generated by mutating either tract constituted a univalent, low-affinity HMGA1a binding site (11). If the inability of the IRE to function at PQRS was a consequence of tighter binding by CarD, then decreasing it by disrupting one of the AT tracts might be able to restore activity. Hence, we checked how IRE variants with only one AT tract mutated (Fig. 4A, IRE-L or IRE-R) and, as a negative control, one with both AT tracts mutated (Fig. 4A, IRE-LR), affected PQRS expression in vivo and DNA binding in vitro. In vivo, neither of the IRE variants with one AT tract activated PQRS, just as previously observed with the IRE (Fig. 4B). In EMSA (Fig. 4C), the specific binding of CarD to the IRE probe was not detected with the IRE-L, IRE-R, or IRE-LR probe. Moreover, DNase I footprinting assays yielded consistent results, with no change provoked by the presence of CarD in the DNase I digestion pattern obtained with the IRE-L, IRE-R, or IRE-LR probe, even at CarD concentrations that exhibit a clear footprint with the IRE (Fig. 4D). These DNA binding analyses in vitro therefore indicate that both AT tracts in the IRE are necessary for the tight binding by CarD. However, neither the IRE, which binds CarD tightly, nor its variants with significantly reduced affinities for CarD could functionally replace the native CarD binding site in vivo. Thus, whereas human HMGA1a can functionally replace the CarD DNA binding domain in M. xanthus (13), the IRE HMGA1a binding site cannot do so for the native CarD binding site at PQRS.
The IRE at the PQRS promoter region enhances specific DNA binding in vitro but impairs promoter activation in vivo. (A) Schematic showing the PQRS-lacZ reporter probe bearing the WT, IRE, and the IRE with one (IRE-L or IRE-R) or both (IRE-LR) of its AT-rich tracts mutated. Numbers above the WT sequence indicate positions relative to the transcription start site (+1). The lines (thicker at the AT tracts) below the WT and IRE sequences span the stretch protected against DNase I by CarD. The numbers correspond to the first and last nucleotides in the protected stretch, and the asterisks indicate the DNase I-hypersensitive sites, based on the data in panel D. (B) Light-induced expression of each of the lacZ reporter probes in panel A relative to the β-galactosidase-specific activity determined with the WT, as described in Materials and Methods. (C) EMSA using the 169-bp DNA probe with the IRE or the IRE variant sites shown in panel A. −, no CarD; +, 1.5 μM CarD. (D) DNase I footprint of the probes indicated at the bottom in the presence of CarD with the coding strand labeled. CarD concentrations used increase in the order 0, 0.9, 1.8, and 3.75 μM for each probe. The segment protected by CarD against DNase I in the IRE probe is indicated by the line (thicker at the AT tracts) on the left, with the numbers above and below being the positions relative to the transcription start site that demarcate this protected region. Asterisks indicate DNase I-hypersensitive positions.
Shifting the position of the native CarD binding site impairs light-induced expression of PQRS in vivo.The above results suggest that DNA sequence and conformation of the native WT site, and the affinity of this site for CarD, are optimized for light-activated expression from PQRS in vivo. The importance of the location of the CarD binding site relative to the PQRS promoter region still remained to be addressed. We checked this by generating variants of the CarD binding site which moved the site by 5 bp or 10 bp upstream of the natural site at PQRS, corresponding to displacements of approximately one-half or one helical turn, respectively (28–30). Figure 5A depicts the specific site displacement variants that were examined. In Mut5a and Mut10a the segment spanning the two TTT tracts and the intervening sequence was displaced upstream by 5 or 10 bp, respectively. In these, the regions immediately upstream and downstream of the displaced segment are not conserved relative to the WT. Since the 10-bp displacement in Mut10a would also lead to an A residue next to each TTT tract, the two As were replaced with G residues in order to retain the 3-bp AT-rich tract of the native CarD binding site (Fig. 5A). With shorter AT stretches, the flanking DNA sequences have been reported to influence binding by HMGA by affecting the DNA conformation required (12). Since this could also occur with CarD, we tested two additional variants in which 5 bp or 10 bp of a random, GC-rich stretch was inserted immediately upstream of position −51 relative to the transcription start site (Fig. 5A). The two variants thus generated involve shifting upstream, by 5 bp (Mut5b) or 10-bp (Mut10b), a 37-bp segment composed of the two TTT tracts with intervening sequence and native 11-bp flanking regions, which matches the oligonucleotide DNA probe used in earlier studies of CarD binding in vitro (13, 14).
The position of the CarD binding site from PQRS is critical for promoter activation in vivo. (A) Scheme of the PQRS-lacZ reporter probe with the following: (i) the WT CarD binding site; (ii) the region including the two TTT tracts and intervening sequence (in boldface) shifted upstream relative to the position in the WT by 5 bp (Mut5a) or 10 bp (Mut10a) (the Gs shaded gray are changes from As in the WT sequence introduced to avoid a 4-bp AT tract that would otherwise result from this 10-bp shift); (iii) insertion of the 5-bp (Mut5b) or the 10-bp (Mut10b) sequence shaded gray. Numbers above the sequences indicate the positions relative to the transcription start site (+1). The oligonucleotide probe to which CarD binds is indicated by the underlined stretch in the WT and shown only partially in Mut5b and Mut10b. (B) EMSA using the 222-bp WT, Mut5a, or Mut10a probe. CarD concentrations increase in the order 0, 0.625, and 1.25 μM. (C) EMSA using the 222-bp WT, Mut5b, or Mut10b probe. CarD concentrations increase in the order 0, 0.625, and 1.25 μM. (D) Light-induced expression of each of the indicated mutant lacZ reporter probes relative to the β-galactosidase activity estimated with the WT as described in Materials and Methods.
In EMSA (Fig. 5B and C), CarD exhibited specific binding to DNA probes corresponding to each of these site displacements. CarD appeared to bind to the Mut5a and Mut10a probes with a somewhat lower affinity than to the WT one since the gel-retarded band corresponding to the CarD-DNA complex obtained with the WT probe was more intense and defined and was detectable at lower CarD concentrations (Fig. 5B). On the other hand, the similar EMSA patterns with the WT, Mut5b, and Mut10b probes suggest that CarD binds to these two variants with affinities comparable to that observed for the WT (Fig. 5C). These conclusions are supported by DNase I footprinting analyses, which, moreover, mapped the region protected by CarD binding to Mut5a, Mut10a, Mut5b, or Mut10b to the same segment as in the WT (data not shown). Overall, these data indicate that while the two TTT tracts with the intervening spacer determine CarD binding specificity, regions flanking either side of this segment may have some influence on the affinity in vitro, as has been reported for HMGA (12).
The functional consequences of the above displacements of the CarD binding site were examined next. For this, light-induced expression of the PQRS-lacZ reporter from probes bearing Mut5a, Mut10a, Mut5b, or Mut10b was compared to that of the WT in vivo. As seen in Fig. 5D, light-activated lacZ expression was dramatically reduced with all four variants relative to the WT level, irrespective of whether CarD binds to these variants in vitro with affinities that are similar to or lower than those observed with the WT. Hence, we conclude from these data that maintaining the normal position of the CarD binding site relative to PQRS is critical for function of this promoter in vivo.
The CarDAd-CarGAd complex, functionally equivalent to CarD-CarG, also requires the intact native CarD binding site for PQRS expression in M. xanthus.We mentioned earlier that CarDAd-CarGAd is functionally equivalent to CarD-CarG in M. xanthus and can activate PQRS expression in the light, despite CarDAd having a histone H1-like DNA binding domain rather than an HMGA-like one and a lower affinity for the site at PQRS than CarD in vitro (13). Histone H1 and HMGA DNA binding domains, although distinct, are both intrinsically disordered basic regions that preferentially recognize minor-groove AT-rich tracts, with HMGA typically having a greater affinity (13, 31). Given their functional equivalence in M. xanthus, despite different DNA binding domains and affinities for the CarD site at PQRS, it would be worthwhile to test if the CarD binding site determinants observed for the CarD-CarG pair also carry over to the CarDAd-CarGAd one. We therefore tested how each one of the different variants of the CarD binding site described above affected light-induced PQRS expression in an M. xanthus strain in which the endogenous CarD-CarG pair was replaced by CarDAd-CarGAd. In this strain, in contrast to the behavior of the PQRS-lacZ reporter probe bearing the WT CarD binding site, little or no effect of light on lacZ expression was observed from the reporter probe with any of the CarD binding site changes described above (Fig. 6A). When tested in vitro using EMSA, CarDAd was found to bind the IRE probe with a somewhat higher affinity, with retarded bands being detected at lower protein concentrations with the IRE than with the WT (Fig. 6B), whereas the binding of CarDAd to Mut5a, Mut10a, Mut5b, or Mut10b was comparable to that of the WT (Fig. 6C and D). Moreover, CarDAd concentrations required for observable binding to each of these DNA probes (Fig. 6B to D) were always significantly greater than for the same probe with CarD (Fig. 2C, 4C, and 5C). CarDAd, with a histone H1-like DNA binding domain, thus binds to the native CarD binding site in vitro with a lower affinity than CarD, whose DNA binding domain is HMGA-like. Yet despite this difference in affinities between CarDAd and CarD in vitro, the sequence and position determinants of the DNA binding site for PQRS activation in vivo are conserved for both proteins.
The native CarD binding site is also essential for PQRS expression in M. xanthus with carDAd-carGAd. (A) Light-induced expression of PQRS-lacZ reporter probes with the CarD binding site corresponding to the WT, IRE, or their variants as indicated (Fig. 2A, 4A, and 5A). Values are shown relative to the estimated WT value, as described in Materials and Methods. (B) EMSA comparing binding of CarDAd to the WT 169-bp DNA probe and to its variant with the IRE. CarDAd concentrations increase in the order 0, 0.47, 0.94, 1.875, and 3.75 μM. (C) EMSA using the 222-bp DNA probe with the Mut5a or Mut10a variant. CarDAd concentrations increase in the order 0, 1.875, and 3.75 μM. (D) EMSA using the 222-bp DNA probe with the Mut5b or Mut10b variant. CarDAd concentrations increase in the order 0, 1.875, and 3.75 μM.
CarD interacts with RNAP β subunit but not with other RNAP subunits or CarQ.We reasoned that because the position of the CarD binding site is crucial for light-triggered activation of PQRS in vivo, the binding of CarD to this site might be involved in ensuring the correct recognition and docking of RNAP onto the PQRS promoter elements. If this were so, stable or transient interactions could occur between RNAP and CarD. Two-hybrid analysis had actually indicated that the CarD N-terminal domain can interact with a fragment from the N-terminal region of the M. xanthus RNAP β subunit (16). However, it remained to be validated if this physical interaction persisted for the full-length forms of CarD and RNAP β. It was also unknown if CarD could simultaneously participate in interactions with any of the other RNAP core subunits or with CarQ, the specific ECF σ factor implicated in PQRS activation. Furthermore, CarG, which associates physically with CarD to form a stable complex, has been found to be absolutely required in every known CarD-dependent process in vivo, suggesting that CarD and CarG act as a single, functional unit (13, 15). It was therefore relevant to examine if CarG, which has no DNA binding ability, can also interact with RNAP. We probed these different interactions using two-hybrid analysis. Both N- and C-terminal hybrid fusions of each RNAP subunit (α, β, β′, and ω) or CarQ were individually probed for interaction against the CarD or CarG hybrid. Interaction between CarD and CarG demonstrated previously using CarG-T18 and T25-CarD fusion proteins served as the positive control (13, 15), while cells expressing only one of the two fusion proteins were used as negative controls. This analysis did not detect interactions of CarD or of CarG with the RNAP core subunits α, β′, or ω or with CarQ (Fig. 7). The only interaction observed by the two-hybrid analysis was that of CarD with the RNAP β subunit, concordant with our earlier data using specific domains of either protein (16). Thus, CarD physically associates with RNAP, and this appears to be achieved via the β subunit of RNAP, without involving the other RNAP core subunits or CarQ.
Two-hybrid analysis in E. coli of the interaction of CarD or CarG with RNAP core subunit α, β, β′, or ω or with CarQ. Cells expressing a given protein pair, as indicated, were spotted on LB-Amp-Km plates containing X-Gal. Interaction, if any, for a given protein pair leads to reporter lacZ expression and is manifested by the blue pigment of the spot. CarD/CarG is the positive control, expressing T25-CarD and CarG-T18 fusion proteins; CarD/− and −/CarG are the negative controls corresponding, respectively, to cells expressing only the CarD and only the CarG fusion. The plates on the left show the interaction assay of T25-CarD against T18-α, -β, -β′, -ω, or -CarQ (cell drops aligned vertically on the left) or against α-, β-, β′-, ω-, or CarQ-T18 (cell drops aligned vertically on the right). The plates on the right show the interaction assay of CarG-T18 against T25-α, -β, -β′, -ω, or -CarQ (cell drops aligned vertically on the left) or against α-, β-, β′-, ω-, or CarQ-T25 (cell drops aligned vertically on the right).
DISCUSSION
Light triggers release of the ECF σ factor, CarQ, from its membrane-bound anti-σ factor, CarR. RNAP holoenzyme assembled with the liberated CarQ then drives transcription from the M. xanthus PQRS promoter (1, 2, 20, 32). Thus, availability of CarQ for its association with the core RNAP and the selective recognition of PQRS by RNAP-CarQ holoenzyme are fundamental in activating transcription from PQRS. Previous studies have established two more proteins, CarD and CarG, as indispensable for PQRS activation in vivo (5, 6, 13–16). Both factors are coexpressed from the same operon and act as a single, physically linked functional unit, but only CarD recognizes and directly binds to a DNA site upstream of PQRS. CarD exhibits minor-groove binding to this site via an intrinsically unfolded domain akin to eukaryotic HMGA that can also be like the C-terminal domain of histone H1, indicating that the basic CarD DNA binding domain can be quite plastic as long as it is intrinsically disordered (13, 14). Hence, as shown in this study, it is remarkable that the CarD binding site identified upstream of the light-inducible PQRS promoter is so strictly optimized for affinity and position.
The CarD binding site at PQRS is, as noted in the introduction, the only one that has thus far been experimentally established as essential in the function of the CarD-CarG complex (6, 13–15) even though this global regulator acts in numerous cellular processes (1–4). Thus, elucidating the CarD binding site determinants is not only important in understanding PQRS regulation by the CarD-CarG complex but also could provide valuable insights into the as yet unknown DNA elements recognized by this complex in controlling the other processes where it is involved. Two AT tracts, each 3 bp long and with a 7-bp spacer, constitute the CarD binding site at PQRS. Data provided here demonstrate that each of these AT tracts is critical since mutating just one of them suffices to abolish specific DNA binding in vitro and PQRS activity in vivo. By comparison, in the human HMGA1a recognition sites, such as the IRE at the IFN-β promoter (Fig. 1C), each AT tract is at least 4 bp in length, and the spacer is 6 to 8 bp, a design that confers high-affinity binding (11). We find that although CarD also binds more tightly in vitro to the IRE than to its natural site, replacing the latter with the IRE abolishes PQRS activation in vivo. Thus, the tighter recognition site for CarD correlates with lower promoter activity. Various factors may account for why the IRE is unsuitable for activating transcription from PQRS. One may be the intrinsic differences in DNA conformation that we observe between the IRE and the natural CarD binding site. Tighter CarD binding to the IRE may also be detrimental to a key step in activating transcription: promoter clearance/escape by RNAP, where the RNAP relinquishes its hold of PQRS so that transcription can proceed. Stronger promoter-RNAP interactions usually correlate with a poorer ability of RNAP to escape the promoter (33). The finding that RNAP can associate physically with CarD suggests that tighter binding of CarD to its site could impose a greater constraint on the associated RNAP to exit from PQRS. A third possibility, given that the CarD footprint with the IRE is 3 bp closer to the PQRS promoter than with the natural site, is that IRE-bound CarD interferes with RNAP binding and prevents PQRS activation. DNase I footprinting and in vitro PQRS transcription analyses with M. xanthus RNAP-CarQ holoenzyme (so far unavailable) could help resolve the various possibilities. Besides differences in DNA conformation, binding, and/or promoter clearance/escape, other factors could contribute to why replacing the native site by the IRE leads to loss of PQRS activity in vivo, and additional aspects of the CarD binding site identified as crucial for function are discussed next.
We find that PQRS activation in vivo requires the CarD binding site to be at a fixed position relative to the promoter. Shifting the site upstream by approximately one-half or one helical repeat prevented transcription from PQRS in vivo, as did deleting 6 bp from the region between the CarD binding site and the −35 promoter element reported in another study (34). What could be the basis for this positional specificity? The answer may lie in the multiple interactions involving the CarD-CarG complex. It is noteworthy that IFN-β enhanceosome assembly also requires HMGA1a binding to sites positioned precisely and in phase on the DNA helix to enable allosteric changes in the DNA and facilitate numerous protein-protein interactions for promoter activation (21, 35, 36). While several discrete activator proteins are involved with HMGA1a in communicating with the eukaryotic general transcription apparatus, CarD employs its N-terminal domain, a module whose equivalent is absent in HMGA1a, to mediate interactions with CarG and RNAP.
Bacteria employ promoter-centric or RNAP-centric mechanisms (where factors interact with the promoter or with RNAP, respectively) to initiate transcription activation (37). Often transcription activators must bind to a fixed face of the DNA helix relative to the promoter in order to maintain the appropriate register and facilitate necessary interactions, including those with RNAP, and the activator may bind to the target promoter region prior to interaction with RNAP, or first bind to free RNAP and then to promoter DNA (37, 38). The ability of CarD to interact with RNAP β alone in the heterologous E. coli two-hybrid system suggests that CarD can bind to free RNAP even when PQRS and the associated CarD site are unavailable. Binding to its site would anchor CarD at a fixed position upstream of the promoter region, and this must necessarily be compatible with binding of the associated RNAP to PQRS for transcription initiation. That these two binding requirements of CarD, one to a specific DNA site and the other to RNAP, must be simultaneously met could thus explain why transcription activation requires the strict positioning of the CarD binding site relative to PQRS. The CarD binding site determinants observed in wild-type M. xanthus also apply in strains in which the CarD-CarG pair was replaced by the functionally equivalent CarDAd-CarGAd, the natural pair in A. dehalogenans. CarDAd must therefore share with CarD similar DNA site recognition and interactions with RNAP in vivo even though in vitro CarDAd (with an H1-like DNA binding domain) has a lower affinity for the site upstream of PQRS than the HMGA-like CarD.
By directly associating with RNAP and binding to a fixed site relative to the promoter, CarD could effectively recruit RNAP and ensure productive RNAP-PQRS binding. Such a mechanism for transcriptional activation by synergistically coupling RNAP recruitment and its binding to promoter is fairly prevalent in bacteria (37, 39, 40). Any of the RNAP subunits can provide potential interacting surfaces. RNAP α and σ are frequently favored interacting partners, given their accessibility to factors binding upstream of a promoter, and numerous examples of activators contacting these RNAP components have been reported (37, 39). But there are also some examples known of factors that contact RNAP β (E. coli DnaA and MerR [41, 42] and Rhizobium meliloti DCTD [43]) or RNAP β′ (E. coli GreB, RfaH, and BglG [44–47], phage N4SSB [48], and Mu protein C [49]). Our data implicate CarD in a specific physical interaction with RNAP β. The inability in our two-hybrid analysis to detect CarD (or CarG) interactions with other RNAP subunits including the ECF σ factor CarQ does not, however, rule out the possibility that these may occur once CarD, CarG, and RNAP have assembled at PQRS. This possibility may, we speculate, be the basis for the critical role of CarG that has been shown to be essential in every process dependent on CarD but whose molecular mechanism of action has remained elusive. The data from this and previous studies suggest that CarD acts at PQRS by recruiting RNAP as well as binding to a specific DNA site optimized for affinity and position to foster transcriptional activation. Reconstituting PQRS transcription in vitro using purified CarD, CarG, and CarQ-associated M. xanthus RNAP holoenzyme could provide valuable molecular insights into the workings of this bacterial enhanceosome and is the objective of our current efforts.
ACKNOWLEDGMENTS
This work was supported by grants BFU2009-12445-C02-01 (to M.E.-A.) and BFU2009-12445-C02-02 (to S.P.) from the Ministerio de Ciencia e Innovación (MICINN)-Spain cofinanced by the European Union (FEDER) and by Fundación Séneca-Murcia (08748/PI/08 to F.J.M.). F.G.-H was supported by MICINN grant BFU2009-12445-C02-01 and J.A.-R. was supported by an FPI fellowship from MICINN.
We thank J. A. Madrid for technical assistance.
FOOTNOTES
- Received 18 September 2012.
- Accepted 6 November 2012.
- Accepted manuscript posted online 9 November 2012.
- Copyright © 2013, American Society for Microbiology. All Rights Reserved.
