Subtle sequence differences between two interacting σ⁵⁴‐dependent regulators lead to different activation mechanisms

Abstract

In the strictly anaerobic nitrate reducing bacterium Aromatoleum anaerobium, degradation of 1,3‐dihydroxybenzene (1,3‐DHB, resorcinol) is controlled by two bacterial enhancer‐binding proteins (bEBPs), RedR1 and RedR2, which regulate the transcription of three σ⁵⁴‐dependent promoters controlling expression of the pathway. RedR1 and RedR2 are identical over their length except for their N‐terminal tail which differ in sequence and length (six and eight residues, respectively), a single change in their N‐terminal domain (NTD), and nine non‐identical residues in their C‐terminal domain (CTD). Their NTD is composed of a GAF and a PAS domain connected by a linker helix. We show that each regulator is controlled by a different mechanism: whilst RedR1 responds to the classical NTD‐mediated negative regulation that is released by the presence of its effector, RedR2 activity is constitutive and controlled through interaction with BtdS, an integral membrane subunit of hydroxyhydroquinone dehydrogenase carrying out the second step in 1,3‐DHB degradation. BtdS sequesters the RedR2 regulator to the membrane through its NTD, where a four‐Ile track in the PAS domain, interrupted by a Thr in RedR1, and the N‐terminal tail are involved. The presence of 1,3‐DHB, which is metabolized to hydroxybenzoquinone, releases RedR2 from the membrane. Most bEBPs assemble into homohexamers to activate transcription; we show that hetero‐oligomer formation between RedR1 and RedR2 is favoured over homo‐oligomers. However, either an NTD‐truncated version of RedR1 or a full‐length RedR2 are capable of promoter activation on their own, suggesting they should assemble into homohexamers in vivo. We show that promoter DNA behaves as an allosteric effector through binding the CTD to control ΔNTD‐RedR1 multimerization and activity. Overall, the regulation of the 1,3‐DHB anaerobic degradation pathway can be described as a novel mode of bEBP activation and assembly.

RedR1 and RedR2 are bacterial enhancer‐binding proteins than only differ in 17 residues (labelled orange and lilac) but are controlled through opposite mechanisms: RedR2 is constitutive and kept inactive by binding the membrane protein BtdS. RedR1 is negatively controlled by its N‐terminal domain. In the presence of 1,3‐DHB, metabolized to hydroxybenzoquinone (HBQ) by BtdSL, active RedR2 is released from BtdS and RedR1 is activated by HBQ. Both proteins assemble into heterodimers, which oligomerize to heterohexamers and initiate transcription.

Abbreviations

1,3‐DHB

1,3‐dihydroxybenzene (resorcinol)

BACTH

bacterial adenylate cyclase two‐hybrid

bEBP

bacterial enhancer‐binding protein

CTD

C‐terminal domain

HBQ

hydroxybenzoquinone

HHQ

hydroxyhydroquinone

HTH

helix‐turn‐helix

IPTG

isopropyl β‐d‐1‐thiogalactopyranoside

MBP

maltose‐binding protein

NTD

N‐terminal domain

RNAP

RNA polymerase

Introduction

Aromatoleum anaerobium (formerly Azoarcus anaerobius strain LuFRes1 [1]) is a strictly anaerobic nitrate reducing betaproteobacterium capable of using 1,3‐dihydroxybenzene (1,3‐DHB, resorcinol) as the only carbon source under anoxic conditions. The pathway involves two oxygen‐independent oxidative reactions as the initial steps in the degradation, differing from the classical anaerobic aromatic degradation pathways which run through initial reductive reactions [2]. First 1,3‐DHB is oxidized to hydroxyhydroquinone (HHQ) by the activity of resorcinol hydroxylase, encoded by rehLS genes, followed by HHQ oxidation by HHQ dehydrogenase to the non‐aromatic hydroxybenzoquinone (HBQ), encoded by btdLS genes. The ring is finally cleaved by HBQ dehydrogenase, encoded by bqdLMS genes (Fig. 1A). The genes responsible for 1,3‐DHB degradation are encoded in a 25 kb cluster, organized in five transcriptional units (Fig. 1B). Interestingly, the cluster includes two genes for almost identical bacterial enhancer‐binding proteins (bEBPs), redR1 and redR2 [3, 4]. The first three enzymatic steps in the pathway are organized in three operons that are induced when the substrate 1,3‐DHB is transformed into HBQ. Transcription of these operons depends on three σ⁵⁴‐dependent promoters, P _rehL1 , P _bqdL and P _orf14 (Fig. 1B), which show conserved upstream activation sequences (UASs) for bEBP binding [4] (Fig. 1C).

Fig. 1.

Aromatoleum anaerobium 3,5‐DHB anaerobic degradation pathway, gene cluster and promoter alignment. (A) Anaerobic 1,3‐DHB degradation pathway. Each reaction is represented by a black arrow. The proteins involved in each step are indicated below each arrow. The doted arrow represents the series of reactions from HBQ to the final products. (B) Anaerobic 1,3‐DHB degradation gene cluster. The genes of known functions (coding for the sequences shown in A) are represented as grey arrows; the genes for the bEBPs RedR1 and RedR2 are coloured in red; the genes of unknown function are shown in white. The small square arrows indicate the presence of a promoter. P _rehL1 , P _bqdL and P _orf14 are the three σ⁵⁴‐dependent promoters. P _rehL2 is a weak constitutive promoter located downstream P _rehL1 . (C) Alignment of the three σ⁵⁴‐dependent promoters controlling cluster operon expression. The sequences extracted from EF078692.1 were defined in [4] and aligned using clustlw2. The transcription initiation site is indicated with +1. The upstream sequence limit of each sequence is indicated with the corresponding negative number (−128, −136, −130). The conserved Eσ⁵⁴ binding elements are shown in red, the IHF‐binding site is schematized in green with the consensus sequence indicated above the region, and the sequence of the UASs are coloured in dark blue (conserved in the three promoters) and light blue (conserved in 2 promoters). The −35/−10 elements of P _rehL2 σ⁷⁰‐dependent promoter are underlined. The sequences between the mentioned regulatory elements are contracted as N _x , were x is the number of bases present between the two shown sequence segments. Modified from Ref. [4].

The promoters transcribed by σ⁵⁴‐dependent RNA polymerase (RNAP) resemble eukaryotic promoters in that they are activated at a distance by the bEBP hexameric complex bound at its target upstream activating sequences (UASs) in the promoter DNA. Sigma‐54‐dependent promoters specifically require specialized AAA+ transcription activator proteins, the bEBPs, to provide the energy required for transcription initiation through ATP hydrolysis [5]. The current knowledge on the activation mechanism of bEBPs has been established from extensive research in the past three decades, and especially through deep structural analysis [6, 7, 8, 9, 10, 11, 12, amongst others]. The bEBPs are module proteins typically composed of three domains [6]. The regulatory N‐terminal domain (NTD) senses the environmental signal and modulates the activity of the regulator. The NTD is not conserved within the members of the family and consists of different types of sensory domains, which can even be absent [5]. Three main modes of NTD signal sensing have been described: phosphorylation by an associated sensor kinase, direct ligand binding, and protein–protein interaction. Examples of bEBPs of each class, bearing a number of sensory domain types, have been described [5]. The central AAA+ domain, responsible for ATP binding and hydrolysis, protein oligomerization and interaction with σ⁵⁴, is essential to change σ⁵⁴‐dependent RNAP conformation to allow open complex formation. ATP hydrolysis promotes contact with σ⁵⁴, which triggers transcription initiation [9, 10]. The C‐terminal DNA‐binding domain (CTD) bears a Fis‐type HTH for specific promoter recognition at the target UASs. Its structure is composed of four helixes: a positioning helix A and a three‐helix bundle that constitutes the HTH core domain, where helix D is the recognition helix contacting the DNA major groove [13]. This structure directs the regulator to its UASs, guaranteeing specific promoter recognition [13, 14].

The NTD regulates the activity of bEBPs by controlling ATPase activity in the central domain in response to the specific environmental signal. In solution, most bEBPs exist as dimers, inactive in transcription activation. Activation of the NTD through any of the possible sensing modes results in transition from inactive dimers to active hexameric rings, where self‐interactions are generally established through the central domains that stimulate ATP hydrolysis and transcription activation [15]. This control over the central domain can be exerted either positively or negatively [7]. The majority of bEBPs described so far are subjected to negative control. In the bEBPs of this group, such as DctD, DmpR, XylR and NtrC1, the NTD represses ATPase activity in the absence of the sensed signal [11, 16, 17, 18]. The changes in the NTD following the reception of the signal promote monomer re‐arrangement, allowing self‐association into hexameric rings capable of ATPase activity. In these proteins, deletion of the NTD results in a constitutively active regulator (reviewed in Ref. [5]). In contrast, in positively controlled bEBPs, essentially represented by Klebsiella pneumonia NtrC, the NTD is not an ATPase repressive element. Once phosphorylated, the presence of the NTD is essential to stabilize the assembly of the monomers into hexamers, where each NTD contacts the central domain of the adjacent monomer, promoting ATPase activity [19, 20]. In this case, a truncated regulator lacking its NTD is constitutively inactive.

The two bEBPs controlling expression of 1,3‐DHB anaerobic degradation pathway in A. anaerobium, RedR1 and RedR2, show 97% sequence identity. The two regulators present the typical domain organization of bEBPs: their NTD has a distinctive signature, composed of a GAF and a PAS domain connected by a long α‐helix (Fig. 2A,C); the sequence of the central domain is identical in both regulators and bears the typical GAFTGA, Walker A and Walker B motifs that characterize the AAA+ family; the CTD encompasses the Fis‐like HTH DNA binding motif. Only 17 residues, distributed between the NTD and the CTD, are different between the two proteins. In the NTD, the N‐terminal tail of both proteins is different both in sequence and length (six and eight residues in RedR1 and RedR2, respectively), and Thr233 in RedR1 PAS domain is substituted by Ile235 in RedR2. In the CTD, the α‐helix preceding the recognition helix D in the DNA‐binding HTH, and the C‐terminal tail, include nine non‐identical residues (Fig. 2A). Induction of the pathway requires the transformation of the substrate 1,3‐DHB into the intermediate HBQ, the actual effector of the pathway [4]. We have previously shown that despite the high sequence identity between the two regulators, these are not redundant and both are required for full transcription activation, although RedR2 can by itself partially activate transcription from the three promoters [4]. RedR2 is the master regulator of the pathway that also controls RedR1 expression by activating the bqd operon that includes the redR1 gene (Fig. 1B). In contrast, RedR1 is inactive in the absence of RedR2. Preliminary assays indicated that the two proteins interact with each other, which suggested that the two almost identical bEBPs were working as heterohexamers to activate transcription of the three pathway promoters and rendering maximum expression levels [4].

Fig. 2.

Homologies and differences between RedR1 and RedR2. (A) Schematic representation of the domains of the regulatory proteins RedR1 and RedR2: the N‐terminal receiver domain (NTD) that includes a GAF (orange) and a PAS (blue) domain; the central domain (AAA+, yellow); and the DNA binding domain (red) including an HTH motif (pink), are shown. The two helices shaping the HTH, designated as in [14], are shaded in grey in the sequence, and the final C‐terminal tail is indicated. The amino‐acid sequences where the 17 differences between the two proteins are found are enlarged and the two sequences aligned, with the amino‐acid residues that are different between the two regulators highlighted in red. The remaining sequence, which is not shown, is identical in the two proteins. (B) Schematic presentation of the two regulators after deletion of their NTD. (C) Model structure of the NTD of RedR1 using DhaR as the template. The structure was built with phyre2 [60] (see the Materials and methods section). The amino‐acid residues at the N‐ and C‐terminal ends of the domain are indicated. The GAF and PAS domains are coloured as in (A). The IITI stretch is highlighted in red except for the Thr which is shown in green, and the residues are presented as sticks. Modified from Ref. [4].

The aim of this study was to analyse the proposed interaction between the two bEBPs and to elucidate the control mechanism operating for each regulator. Our results demonstrate hetero‐oligomer formation, and unexpectedly identify two totally different modes of control of the activity of the regulators, despite their high sequence identity; furthermore, we describe a control mechanism based on interaction with an integral membrane protein that sequesters one of the regulators to the membrane. Overall, the regulation of 1,3‐DHB anaerobic degradation pathway can be described as a novel mode of bEBP activation and assembly.

Results

Only RedR1‐RedR2 hetero‐oligomers are stable in vitro

Previous bacterial adenylate cyclase two‐hybrid (BACTH) assays suggested that neither RedR1 nor RedR2 were capable of self‐interaction, and only interactions between RedR1 and RedR2 were observed [4]. Consistent results with different truncated versions of the two regulators showed that hetero‐interaction occurred through the central domain, and that the NTDs were not involved. Data also suggested that self‐interaction of each regulator was impeded by their CTD [4]. Because the central domain of the two proteins is identical, the absence of self‐interaction was unexpected. To analyse physical interactions between these proteins, we carried out column‐binding experiments with differently tagged RedR1 and RedR2 proteins (Fig. 3A). An N‐terminal 6His fusion to RedR2 (6His‐RedR2) was purified and then bound onto a Ni²⁺‐sepharose column. Then a C‐terminal fusion of hemagglutinin epitope (HA) to RedR1 (RedR1‐HA) was overexpressed in Escherichia coli and the crude extract was loaded onto the column where 6His‐RedR2 was previously bound. After thorough washing to eliminate the unbound material, the 6His‐RedR2 protein was finally eluted with 250 mm imidazole. Since the two proteins have a similar molecular weight (72.5 and 72.8 kDa) they were detected with specific anti‐tag antibodies. Figure 3B shows that RedR1 (detected with anti‐HA antibody) was initially retained in the RedR2‐bound column, and co‐eluted with this regulator (detected with an anti‐His antibody) when detached with imidazole. If the reverse experiment was performed, i.e. a 6His‐RedR1 fusion protein was first bound to the column, followed by the load of a crude extract containing overexpressed HA‐tagged RedR2 (RedR2‐HA), this last protein was retained in the column and co‐eluted with 6His‐RedR1 when the imidazole gradient was applied (not shown). This result is a strong indication of the direct interaction between the two regulators, consistent with the preliminary two‐hybrid analysis evidence. It is worth noting that RedR1‐HA did not bind the column in the absence of pre‐bound 6His‐RedR2 (Fig. 3C).

Fig. 3.

RedR1 and RedR2 interaction detected by column‐binding assays. (A) Scheme of the experimental strategy. A solution of purified 6His‐RedR2 (6His‐R2) protein was loaded onto a HisTrap HP (GE Healthcare, Uppsala, Sweden) column and thoroughly washed with buffer B as described in the Materials and methods section. Then a crude extract containing overproduced RedR1‐HA (R1‐HA) protein was loaded, the column was washed with 10 bed volumes of loading buffer, then imidazole buffer was applied and the eluted fractions analysed in western blots. (B) Unbound material (unbound) and washing fractions (W) were collected; then the bound proteins were eluted using buffer B supplemented with 250 mm imidazole. The collected fractions (bound proteins) were labelled as IE1 and IE2. The proteins in the indicated samples were fractionated in SDS/PAGE, transferred to polyvinylidene difluoride (PVDF) membranes, and tested using either anti‐His or anti‐HA antibodies as indicated. The figure shows a representative result of two replicates. (C) A control experiment to show that the HA tagged regulator did not bind the HisTrap HP column in itself. A solution of purified RedR1‐HA protein was loaded onto a HisTrap HP (GE Healthcare) column, thoroughly washed with buffer B before the final elution with 250 mm imidazole in buffer B. The protein fractions were fractionated in SDS/PAGE and processed as above. The figure shows a representative result of two replicates. (D) A similar experiment as in B except that the crude extract contained overproduced RedR2‐HA. The figure shows a representative result of two replicates. (E) A similar experiment as in B except that the initially loaded purified protein was 6His‐RedR1. The double bands of 6His‐RedR proteins was due to stop codon readthrough of the redR genes cloned in pET28b+, which rendered a second 6His‐RedR protein with 12 extra residues at the C‐terminal end (see the Materials and methods section). The figure shows a representative result of two replicates.

To explore the occurrence of self‐interactions between the two proteins, we repeated the experiment described above except that we changed either the HA‐tagged protein in the crude extract or the His‐tagged protein bound to the column. Figure 3D,E show that neither RedR1‐HA nor RedR2‐HA were retained in the column when the same regulator (except for a 6His‐tag substituting the HA tag) was pre‐loaded in the column, but they were directly eluted with the unbound material and further washing. These data show that in vitro full‐length RedR1 and RedR2 can only form hetero‐oligomers, which probably constitute the competent conformation in the promoter activation mechanism. It is worth noting that in vivo, full‐length RedR2 can activate transcription on its own to a certain extent (see below), which suggests that in certain conditions this protein should be capable of self‐interaction.

RedR1 and RedR2 activity are controlled through different mechanisms

The great majority of bEBPs are negatively controlled by their NTD, which blocks the protein dimers in a face‐to‐face orientation that impedes the oligomerization to hexamers unless the triggering signal produces an appropriate conformational change [20]. As mentioned, in these proteins deletion of the NTD renders regulators no longer to be repressed for hexamer assembly and capable of transcription activation in the absence of the triggering signal [18, 19]. In contrast, in positively regulated bEBPs the activated NTD is required for the assembly of the hexameric ring [7], and renders an inactive regulator when deleted. To identify the mechanism controlling RedR1 and RedR2 activity, we generated NTD‐truncated versions of the proteins (ΔNTD‐RedR1 and ΔNTD‐RedR2) by deleting the 319 and 321 N‐terminal residues, respectively, leaving intact the linker helix (Fig. 2B). As target promoter to measure the activity of both the wild‐type and truncated versions of the bEBPs, we selected the P _orf14 promoter, to avoid potential artefacts due to the divergent location of P _rehL1 and P _bqdL promoters and the presence of the weak constitutive P _rehL2 promoter (Fig. 1). We used a P _orf14 ::lacZ fusion in p220Porf14 [4] in the heterologous host E. coli MC4100 in the absence of inducer (E. coli cannot transform the substrate 1,3‐DHB into the effector HBQ). Interestingly, whilst RedR1 was inactive in E. coli, its NTD‐truncated version ΔNTD‐RedR1 produced high transcription levels from P _orf14 (Table 1). On the contrary, whilst the truncated ΔNTD‐RedR2 was totally inactive, the wild‐type RedR2 regulator was constitutively active in E. coli, although the transcription levels obtained were lower. To confirm that the measured activities depended on the σ⁵⁴‐dependent promoter, assays of ΔNTD‐RedR1 and RedR2 were performed in E. coli BW25113 (wild‐type) and its rpoN and ihfA mutants. Table 1 shows that activation of transcription from P _orf14 by the two active versions of the regulators was strictly dependent on the RNAP with σ⁵⁴ and required IHF, a specific feature of most bEBP‐regulated promoters that require this factor to bend the DNA and bring into proximity the regulator bound at a distance and the RNAP [21].

Table 1.

P _orf14 activity with different versions of RedR1 and RedR2 proteins in different Escherichia coli backgrounds.

Strain	bEBP a	β‐Galatosidase b (MU)
MC4100	None	6.4 (±0.2)
	RedR1	9.9 (±3.5)
	RedR2	53.7 (±5.5)
	∆RedR1	166.6 (±36.5)
	∆RedR2	9.1 (±2.3)
BW25113 (wt)	∆RedR1	156.2 (±16.2)
rpoN c	∆RedR1	10.1 (±1.2)
ihfA d	∆RedR1	6.4 (±1.0)
BW25113 (wt)	RedR2	49.9 (±4.0)
rpoN	RedR2	6.4 (±8.9)
ihfA	RedR2	8.9 (±4.6)

^a Cloned in pMMB67EH [50].

^b Measured from P _orf14 promoter::lacZ fusion in pMP220 (p220Porf14) [4].

^c Strain JW3169‐1.

^d Strain JW1702‐1.

These results clearly show that the two bEBPs, despite their remarkable sequence identity, are subjected to different control modes. RedR1 behaves as a typical negatively regulated bEBP such as NtrC1, DctD or XylR [11, 22, 23], and is probably controlled through activation by the previously identified effector HBQ [4]. In contrast, RedR2 is capable of significant levels of constitutive transcription activation in itself. This is rather unique, since positively regulated bEBPs require stimulation of their NTD (e.g. through phosphorylation) to become active [5, 19]; but in the experimental conditions shown in Table 1 with E. coli as host, no effector was present.

The inactive form of the regulators impairs promoter activation

The results above suggest that activated RedR1 and RedR2 can interact to promote transcription, and that the active forms of the regulators are different. To test this prediction, we analysed transcription activation by several combinations of the different forms of the two regulators (active and inactive/full‐length and truncated). To that end, we used the different variants of the regulators cloned in compatible plasmids, and assayed effector‐independent activation of the P _orf14 ::lacZ fusion in E. coli. Figure 4A,B shows that the co‐expression of the two truncated regulators (expressed simultaneously from different plasmid constructs), i.e., ΔNTD‐RedR1 (constitutively active) and ΔNTD‐RedR2 (inactive) resulted in no promoter activation, whilst ΔNTD‐RedR1 alone, either expressed from a single construct or simultaneously from two different plasmids, resulted in maximum activation of P _orf14 . Similarly, the co‐expression of full‐length RedR1 (inactive) and RedR2 (constitutively active) resulted in no promoter activation, although RedR2 alone was capable of promoter activation. In contrast, the co‐expression of ΔNTD‐RedR1 and full‐length RedR2, the two constitutively active forms of the regulator, resulted in maximum promoter activation, at similar levels to those obtained with ΔNTD‐RedR1 alone. These results suggest that the presence of an inactive form of either regulator inhibited the function of the active regulator on P _orf14 , confirming the proposed hetero‐interaction. We concluded that the formation of hetero‐oligomers, which occurs through their identical central domains, was always favoured over self‐association, and only rendered an active protein complex if the two components were active.

Fig. 4.

Expression from P _orf14 with different combinations of RedR versions. Betagalactosidase activity was determined from Escherichia coli BL21 carrying P _orf14 promoter cloned in (A) pMP220 (p220P _orf14 , IncP‐1 incompatibility group) or (B) pMP190 (p190P _orf14 , IncQ incompatibility group). These reporter strains were transformed with one or two plasmids carrying the different versions of RedR proteins. When assayed independently (red bars), the truncated regulators were expressed from pET24b (ColE1 incompatibility group) and the full‐length proteins from pMMB67EH (RSF1010 incompatibility group) (RedR1) or pJB3Ap (IncP‐1 incompatibility group) (RedR2). In the combination of two regulators (blue bars), the first regulator was expressed from pET24b, and the second regulator from pMMB67EH, except RedR2, which was expressed from pJB3Ap (see Table S1). The plasmids were maintained by the presence of the corresponding antibiotic in the medium. Controls (grey bars) included BL21 only bearing the promoter in the indicated vector, without any additional plasmid [Porf14 (no regulator)], and the indicated constructions with empty pMP220 or pMP190. It is worth noting that the results when only one protein was tested were similar in the presence or absence of the empty vector used to express the second protein (not shown). Promoter activity was determined 2 h after induction of the proteins expressed from pET24b with 1 mm IPTG. The assays were performed minimally in triplicate; error bars show the standard deviation of the mean.

Promoter activity is effector‐independent (constitutive) in a btsS background

Because RedR2 was constitutively active in the heterologous host E. coli, where no other specific regulatory elements of A. anaerobium were present, we hypothesized that in its natural host an unknown mechanism was maintaining this regulator inactive in the absence of an inducer. Amongst the possible mechanisms involved in the process, we considered protein–protein interaction with a protein factor encoded in the 1,3‐DHB degradation cluster as a possible repressing mechanism of the RedR2 activity. To assay the possible role of different genes of the cluster in transcription regulation, we selected different mutants in the pathway. Because A. anaerobium is a strict anaerobe that cannot grow on solid agar plates, all the experiments requiring the modification of the genetic background were carried out in the facultative anaerobe Azoarcus sp. strain CIB [24] bearing the pR⁺ cosmid, which encompasses the entire 1,3‐DHB degradation cluster of A. anaerobium [3] and confers upon Azoarcus sp. strain CIB the capacity to grow anaerobically on 1,3‐DHB, thus reproducing the promoter expression pattern of the original A. anaerobium host [4]. We screened several available pathway gene mutants [3] for constitutive promoter activation. Some of them (rehL, btdL and bqdS, coding for different sub‐units of the enzymes involved in the three first steps of the pathway and responsible for the transformation of the substrate into the effector) had been analysed in a previous study and showed that the activity in these mutants was in all cases effector‐dependent [4]. In contrast, primer extension analysis of P _orf14 promoter in a knock‐out mutant of the btdS gene, coding for the small subunit of HHQ dehydrogenase composed of BtdL and BtdS (second step of the pathway), showed that expression was similar in the presence and absence of the pathway substrate, unlike the wild‐type strain where expression was only observed in the presence of 1,3‐DHB (Fig. 5). A similar behaviour was observed with the remaining σ⁵⁴‐dependent promoters P _rehL1 and P _bqdL , which required the presence of the substrate in a wild‐type background but showed constitutive expression in a btdS mutant.

Fig. 5.

Primer extension analysis of P _rehL1 , P _bqdL and P _orf14 in a btdS background. Primer extension of total RNA of wild‐type Azoarcus sp. CIB (pR⁺) and its btdS mutant [Azoarcus sp. CIB (pR⁺ btdS::Gm)] growing on glutarate (−) or glutarate plus resorcinol (+) with a ³²P‐labelled primer complementary to rehL, bqdL and orf14. The cDNA extension band corresponding to each promoter is indicated with an arrow. Replicate experiments were performed with each primer. A representative gel is shown in each case.

To confirm constitutive expression of the promoter in an Azoarcus sp. CIB btdS background, we introduced the P _orf14 ::lacZ fusion (p190Porf14) in the btdS mutant and determined β‐galactosidase activity in the presence and absence of the substrate 1,3‐DHB. The results confirmed constitutive activity of P _orf14 in the btdS background (Table 2): whilst in the wild‐type strain P _orf14 activity was negligible in the absence of 1,3‐DHB and was induced more than 10‐fold in the presence of the substrate, the activity in the btdS background was similar both in the presence and absence of 1,3‐DHB and reached approximately 55% of the activity in the wild‐type strain in the presence of the effector.

Table 2.

P _orf14 activity in wild‐type Azoarcus sp. CIB (pR⁺) and different single and double mutants.

Strain	β‐Galatosidase a (MU)
Azoarcus sp. CIB (pR+)	Glutarate	Glutarate + 1,3‐DHB
wt	16.65 (±6.70)	197.47 (±1.18)
btdS::Gm	105.59 (±15.18)	107.98 (±13.62)
btdS::Gm, redR1::Km	99.75 (±15.18)	106.56 (±16.32)
btdS::Gm, redR2::Km	6.13 (±2.32)	7.34 (±8.97)

^a Measured from P _orf14 ::lacZ fusion in pMP190 (p190Porf14) [4].

To determine whether the observed constitutive activity depended on the regulatory proteins, we constructed double mutants of btdS and either of the two regulator genes (redR1 and redR2). Table 2 shows that the constitutive activity was unaffected in a redR1 btdS double mutant. In contrast, the activity was negligible in a redR2 btdS double mutant, both in the presence and absence of the inducer, indicating that the constitutive P _orf14 expression in the btdS background was regulated, and RedR2 was the only regulator responsible for this control. Thus, we can conclude that btdS is the regulatory element repressing RedR2 activity in Azoarcus sp. CIB (pR⁺) when no substrate is present. Furthermore, the presence of 1,3‐DHB in the btdS redR1 double mutant did not further increase activity with respect to the values in the absence of the inducer, which suggests that once de‐repressed by the absence of BtdS, no additional signal is sensed by RedR2, which when working as the only regulator could not reach the maximum activation levels. Maximum activity of the promoter depends on the co‐ordinated activation by RedR1 and RedR2 in the presence of the substrate [4].

RedR2 but not RedR1 interacts with the integral membrane protein BtdS

To confirm direct protein–protein interaction between RedR2 and BtdS, we carried out column binding experiments. Because BtdS is an integral membrane protein (Fig. 6A,B) which is insoluble when overexpressed (not shown), we generated a protein fusion of BtdS to the maltose‐binding protein (MBP) in pMAL‐pV, which includes an internal 6‐His tag [25]. The resulting chimera (MBP‐6His‐BtdS) was purified and loaded onto a Ni²⁺‐sepharose column, and an E. coli crude extract containing overproduced RedR2‐HA was applied to the column and subsequently washed to eliminate unbound material. Figure 6C shows that RedR2 was retained by the BtdS protein bound to the column and co‐eluted with it when imidazole was applied. MBP‐6His‐BtdS eluted as a 52‐kDa protein. On the contrary, when extracts containing RedR1‐HA instead of RedR2‐HA were applied to the column, RedR1 was not retained by BtdS and eluted with the unbound material (Fig. 6D). These results confirm the above genetic indications that RedR2 was repressed by BtdS through direct protein–protein interaction but not RedR1. It is worth noting that in a control assay with the MBP lacking the BtdS moiety bound to the column, RedR2 was not retained (Fig. 6E).

Fig. 6.

Interaction of BtdS with RedR proteins detected by column‐binding assays. (A, B) BtdS structure prediction. (A) Scheme of BtdS membrane topology indicating the amino acid residues encompassing the four helices. Rainbow colours go from the N‐terminal end (blue) to the C‐terminal end (red). The sequence of the exposed loop between helices 2 and 3 is indicated. (B) Cartoon of the computational model of the BtdS three‐dimensional structure built with I‐TASSER server, after model assessment by the SwissModel Structure Assessment Tool, represented as protein surface to allow the visualization of the predicted protein surface exposed to the cytoplasm. The optimal membrane location (represented as dot surfaces) was based on OpenStructure estimations available in Swiss‐Model repository. The inner face of the membrane is only visible as its profile. (C–E) RedR2 and BtdS interaction detected by column‐binding assays: (C) A solution of purified MBP‐6His‐BtdS (6His‐BtdS) was loaded onto a HisTrap HP (GE Healthcare) column and thoroughly washed with buffer B as described in Experimental procedures. Then a crude extract containing overproduced RedR2‐HA (R2‐HA) was loaded and the column was washed with 10 bed volumes of loading buffer. Unbound material (unbound) and washing fractions (W) were collected. Finally the bound proteins were eluted using buffer B supplemented with 250 mm imidazole. The collected fractions (bound proteins) were labelled as IE1 and IE2. The proteins in the indicated samples were fractionated in SDS/PAGE, transferred to PVDF membranes, and probed using either anti‐His or anti‐HA antibodies as indicated. The figure shows a representative result of three replicates. (D) A similar experiment as in C except that the crude extract contained overproduced RedR1‐HA. The figure shows a representative result of two replicates. (E) A similar experiment as in D except that the protein initially bound to the column was purified MBP. The figure shows a representative result of two replicates. (F) Intracellular distribution of RedR1 and RedR2 proteins in Azoarcus sp. CIB (pR⁺) wild‐type and btdS mutant. Western blot analysis of membrane (mb) and cytosol (cyt) fractions of Azoarcus sp. CIB (pR⁺) and its btdS mutant Azoarcus sp. CIB (pR⁺ btdS::Gm) bearing a plasmid expressing either HA‐tagged RedR2 protein (pMMRedR2HA) (upper row) or HA‐tagged RedR1 protein (pMMRedR1HA) (lower row), grown in glutarate and glutarate plus 1,3‐DHB. Lysates of the different cultures with whole cells removed were subjected to centrifugation at 12 000  g for 1 h to separate membrane and cytosol fractions. The resulting pellet (membrane fraction) was solubilized in the same volume of Laemmli buffer, and the proteins from the supernatant (cytosol fraction) were precipitated with 10% (w/v) trichloroacetic acid and resuspended in the same volume of Laemmli buffer. Equal volumes of samples were loaded. Western blot analyses were performed with anti‐HA antibodies as described in the Materials and methods section. The figure shows a representative result of four replicates.

We used bacterial two hybrid assays to identify the RedR2 domain involved in interaction with BtdS. We designed a series of BACTH constructs in pUT18C and pKT25, bearing the adenylate cyclase T18 and T25 fragments, respectively [26], of the different RedR2 domains and tested them against BtdS. BACTH assays are especially suitable for BtdS analysis since they were designed to be applicable to detect protein interactions involving integral membrane proteins [26]. Figure 7A clearly shows that RedR2 but not RedR1 showed binding interaction with BtdS, thus confirming the above results with column binding assays. Furthermore, the binding determinant could be located to the NTD of RedR2, which gave strong positive binding values with BtdS. As expected, RedR1 NTD was negative in interaction with BtdS.

Fig. 7.

Analysis of protein–protein interaction through BACTH assays. (A) Binding of BtdS to RedR proteins, truncated derivatives and mutants. RedR1 and RedR2 proteins and derivatives were fused to the N‐terminus of T25 and BtdS was fused to the C‐terminus of T18 in pKT25 and pUT18C BACTH vectors, respectively [26]. In each assay pair, separated by a hyphen, the first construct was based on pKT25 and the second on pUT18C. R1 and R2 are the full‐length proteins, R1NTD and R2NTD are the NTD of either protein, I/T anf T/I indicate the mutation Ile235 to Thr in RedR2 and Thr233 to Ile in RedR1 in their PAS domain, and ∆6 and ∆8 denote the deletion of the indicated number of N‐terminal end residues of the N‐terminal tail of either the full‐length or NTD proteins. pUT18C and pKT25 are the two‐hybrid vectors devoid of insert. Results with RedR1 derivatives are coloured in red whereas those with RedR2 derivatives are in blue. Controls are assays with one of the plasmids devoid of insert (grey). Assays were minimally performed in triplicate; error bars show the standard deviation of the mean. (B) Control experiments to confirm stable mutant protein expression. RedR1 and RedR2 proteins and their mutant derivatives were fused to the N‐terminus of T25 or to the C‐terminus of T18 in pKT25 and pUT18 BACTH vectors, respectively [26]. In each assay pair, separated by a hyphen, the first construct was based on pKT25 and the second on pUT18. The constructs nomenclature is as in A. pUT18 is the two‐hybrid vectors devoid of insert. Positive RedR1/RedR2 hetero‐interactions are depicted in dark colours; RedR2 or RedR1 self‐interactions are shown in light colours. Controls with one of the plasmids devoid of insert are shown in grey. Interactions between wild‐type proteins are drawn in orange, Ile‐stretch mutants in green and N‐tail deletion mutants in blue. Assays were minimally performed in triplicate; error bars show the standard deviation of the mean. (C) ΔNTD‐RedR1 homodimer stabilization in the presence of promoter DNA. RedR1, RedR2 and ∆RedR1 proteins were fused to the N‐terminus of T25 or the C‐terminus of T18 in pKT25 and pUT18 BACTH vectors, respectively [26]. In each assay pair, separated by a hyphen, the first construct was based on pKT25 and the second on pUT18. R1 and R2 are the full‐length proteins, ∆RedR1 is RedR1 depleted of its NTD. Assays were performed in the absence (−) or presence of P _orf14 promoter DNA (a 285 bp fragment spanning from positions −221 to +58 with respect to the transcription start site) cloned in pBBR‐MCS5 (Porf14) or of the empty pBBR‐MCS5 vector (pBB). ∆RedR1‐∆RedR1 self‐interactions are depicted in red, RedR2‐RedR2 self‐interactions in blue, RedR2‐RedR1 hetero‐interactions in green, ∆RedR1‐RedR2 in orange and full‐length RedR1‐RedR1 in lilac. The presence of P _orf14 DNA is indicated by a darker colour. The assays were minimally performed in triplicate; error bars show the standard deviation of the mean.

The NTD domains of the two proteins are identical except for their N‐terminal tail, where a sequence of 6 and 8 non‐conserved residues (in RedR1 and RedR2, respectively) precede the conserved sequence, and for a single change in the Ile stretch in RedR2‐NTD PAS domain (I233‐I234‐I235‐I236), which is interrupted by a Thr in RedR1 (I231‐I232‐T233‐I234) (Fig. 2). Therefore, the different behaviour of the two proteins in the binding assays must lie in either or both sequence differences. We constructed pKT25 fusions to mutant RedR2 NTD with the substitution Ile235→Thr (R2‐NTD(I/T)) and its reciprocal mutant RedR1 NTD with the substitution Thr233→Ile (R1‐NTD(T/I)), and tested them against the pUT18C fusion to BtdS in BACTH assays. As shown in Fig. 7A, the R2‐NTD(I/T) mutant had lost the capacity of binding interaction with BtdS. This was further confirmed in interaction assays of BtdS with a full‐length mutant RedR2 protein bearing an Ile235→Thr substitution (RedR2(I/T)), where interaction was also lost (Fig. 7A). To verify that the mutant protein was stably expressed, it was tested for hetero‐interaction with RedR1, which should be unaltered in this mutant [4]; effectively, the RedR2(I/T)‐RedR1(T/I) interaction was the same as for the wild‐type proteins, confirming that the mutant RedR2(I/T) protein fused to the adenylate cyclase T25 fragment in pKT25 was a stable protein (Fig. 7B). These results suggest that the Ile stretch in the PAS domain of RedR2 is involved in RedR2 interaction with BtdS, and has an essential role in the control of the NTD through BtdS binding. Interestingly, the reciprocal mutant R1‐NTD(T/I) did not acquire the capacity to interact with BtdS (Fig. 7A), suggesting that the presence of the Ile stretch in the PAS domain did not suffice for interaction with BtdS and that the presence of RedR1 N‐tail probably interfered with this binding. To assess the role of these N‐tails in binding to BtdS, we constructed pKT25 fusions to the NTDs devoid of their N‐terminal tail (six or eight residues, according to the specific protein) and tested them against the pUT18C fusion to BtdS. As shown in Fig. 7A, the RedR2 NTD devoid of its N‐terminal tail had also lost its capacity of binding interaction with BtdS. To rule out any protein misfolding due to the mutations in the NTD tail, the stability of the fusion proteins was examined by testing their capacity to establish hetero‐interactions with the corresponding RedR partner as above. Figure 7B shows that N‐tail depleted regulators established hetero‐interactions similar to the wild‐type, confirming the two mutant protein fusions had active stable conformations. Thus, the two minor differences between RedR1 and RedR2 NTD sequences are responsible for the different behaviour of the two regulators in BtdS binding.

BtdS controls RedR2 activity through sequestration of the regulator to the membrane

The closest relative of BtdS in the databases (85% identity) was a DoxX/D‐like family protein of the Betaproteobacterium Candidimonas bauzanensis [27]. DoxX/D is a poorly characterized family of small integral membrane proteins of unknown function (PFAM04173). The members of this family are similar in size (<200 residues, BtdS has 135 residues) and are characterized by four predicted transmembrane helices (Fig. 6A,B). Thus, as BtdS is located in the membrane, the observed lack of activity of RedR2 in the absence of any effector would result from BtdS‐mediated sequestration of this bEBP to the cytoplasmic membrane. To test this hypothesis, we examined the intracellular distribution of RedR2 in different growth conditions, and the role of BtdS in the localization of the regulator in Azoarcus sp. CIB bearing the entire 1,3‐DHB degradation pathway. Azoarcus sp. CIB (pR⁺) and its btdS mutant [4] were provided with an HA‐tagged version of RedR2 regulator expressed from the broad‐host‐range vector pMMB67EH. The two strains were grown anoxically with glutarate as carbon source or with glutarate plus 1,3‐DHB for pathway induction, and the cellular localization of HA‐RedR2 was determined in each case. Figure 6F shows that under basal conditions (growing on glutarate in the absence of the substrate 1,3‐DHB), RedR2 was entirely localized in the membrane fraction of the cell. However, when the BtdS protein was absent, the amount of RedR2 in the membrane fraction was negligible, and the protein appeared in the cytoplasmic fraction. In contrast, under inducing conditions (in the presence of 1,3‐DHB), no RedR2 protein was found in the membrane fraction, and the protein was only detected in the cytoplasmic fraction irrespective of the presence of BtdS. This strongly suggests that BtdS is located in the membrane, as predicted by its sequence, and through protein–protein interaction sequesters RedR2 to the membrane. RedR2 would otherwise be a cytoplasmic protein, as shown by its cytoplasmic location in the btdS mutant. When the experiment was repeated with HA‐tagged RedR1 protein, the regulator appeared in the cytoplasmic fraction in all growth conditions and genetic backgrounds (Fig. 6F), confirming that this regulator does not interact with BtdS.

The predicted structure of BtdS places an Arg residue of the protein N‐terminal end and a short 5‐residue loop (LFGIY) exposed to the cytoplasm, which would be a candidate for protein–protein interaction.

ΔNTD‐RedR1 homodimers are stabilized by promoter DNA

The column‐binding assays shown above ruled out self‐interactions of the two regulators in vitro. This is especially intriguing since the hetero‐interactions were shown to occur through the central domains of RedR1 and RedR2 [4], and both proteins have identical central domains (Fig. 2A). This suggests that in vitro some determinant(s) in either or both the NTD and CTD are impeding the self‐interaction of each protein through their central domain. In the case of RedR1, even the NTD‐truncated ΔNTD‐RedR1 version of the protein, which is transcriptionally active (Table 1) and should thus be capable of homo‐hexamer formation in vivo, was also incapable of detectable binding‐interaction with itself in two‐hybrid assays, pointing to the CTD being the domain restricting self‐interaction in our assays [4]. Because in vivo ΔNTD‐RedR1 activity is always measured in the presence of its target DNA, which was absent from the BACTH assays, we analysed the possible involvement of the UASs DNA as a stabilizing factor for dimer self‐assembly. To that end, we tested ΔNTD‐RedR1 dimerization in BACTH assays in the presence of promoter DNA. Figure 7C clearly shows that the presence of the P _orf14 promoter DNA (a 285 bp fragment spanning from positions −221 to +58 with respect to the transcription start site) in a plasmid provided in trans produced a strong increase in β‐galactosidase activity, indicative of a stable interaction between the two ΔNTD‐RedR1 monomers. The UASs for regulator binding in Porf14 were identified at position −136 to −120 with respect to the transcription start site (D. Pacheco‐Sánchez, unpublished). When we introduced a P _orf14 promoter with mutated UASs (incapable of promoting transcription), no interaction (increase in β‐galactosidase activity) was observed, confirming that the UASs in the promoter DNA were responsible for the stabilization of ΔNTD‐RedR1 dimers. The presence of an empty plasmid (pBB in Fig. 7C) had no effect.

The active form of RedR2 is the full‐length protein, which can by itself activate transcription from P _orf14 to 50% of the maximum values obtained when both proteins are present (Tables 1 and 2 [4]). As for ΔNTD‐RedR1, self‐interaction of the full‐length protein was neither detected in two‐hybrid assays, nor in the pull‐down assays (Fig. 3D,E). Unexpectedly, the presence of the P _orf14 promoter DNA failed to stimulate self‐interactions of RedR2 (Fig. 7C), as failed the DNA of the two other pathway promoters (P _rehL1 and P _bqdL ) controlled by these bEBPs (not shown), suggesting that possibly additional unidentified factor(s) were required for RedR2 protein self‐interaction and activity.

Finally, the presence of the promoter DNA did not further stimulate binding interaction between RedR1 and RedR2 and between ΔNTD‐RedR1 and RedR2 above the levels observed previously (Fig. 7C). It is worth noting that our attempts to measure in vitro transcription activation by either active form of the two regulators were unsuccessful (not shown).

Different predicted structures despite high sequence identity

The significant differences in the activation mode of the regulators resulting from the few changes in their protein sequence implied significant differences in the structure of the two proteins. Our attempts to prepare concentrated solutions of the NTD domains for crystallization to obtain the corresponding protein structures resulted in protein aggregation. As an alternative, we built models of the NTD domain of the two regulators based on the DhaR protein crystal structure [28], and compared the two highest quality models obtained (see the Materials and methods section). We found that the differences between the two protein sequences had relevant consequences on their structure: (a) the beta sheet encompassing the Ile stretch in RedR2 PAS domain was exposed to the surface, forming an acute angle with the linker helix and creating a hydrophobic cavity (Fig. 8A,B). In RedR1, the hydrophobicity of this patch was disrupted by Thr233 and resulted in a different orientation of the PAS domain. (b) When we compared the predicted local flexibility of the two protein structures, determined as deformation energy [29], we observed that local flexibility in the linker helix between GAF and PAS domains was higher in RedR2‐NTD than in RedR1 (Fig. 8C,D). Modelling of the two tails, which are absent in the DhaR template, showed low reliability. Overall, these differences point to RedR2‐NTD being a more flexible protein close to the Ile patch involved in protein binding, substantiating why RedR2 NTD would be capable of binding interaction with BtdS (Fig. 7A), which would not be possible in the more rigid RedR1 protein.

Fig. 8.

Structural model of the PAS domain in the NTD of RedR1 and RedR2. (A) Partial ribbon presentation of the superimposed structures of RedR1 (blue) and RedR2 (orange) NTDs showing the final segment of the linker helix connecting the GAF and PAS domains, and the PAS domain. The structures were built with I‐Tasser [61] (see the Materials and methods section). The 4xIle stretch in RedR2 and the corresponding Ile‐Ile‐Thr‐Ile sequence in RedR1 are coloured in green. (B) Surface of the PAS domain of RedR1 and RedR2 NTDs showing the hydrophobicity surface in the Kyte‐Doolittle scale [66]. Colours range from blue for the most hydrophilic, to white at 0.0, to orangey‐red for the most hydrophobic. The surface was drawn upon the structures in (A). Visible in the centre of the PAS domain, the hydrophobic 4xIle patch in RedR2 interrupted in RedR1 by Thr231 are coloured in green. (C, D) Differences in the deformation energy of RedR1 and RedR2 NTDs. Putty tube representation of the amount of local flexibility along the sequence of the RedR1 and RedR2 NTD predicted structures, calculated as the deformation energy in the Dynamut server using the normal mode analysis and selecting the C‐alpha force field [29]. The magnitude of the deformation is represented by the tube width and colour, where thick and red tubes indicate higher deformation energies, thin and blue tubes indicate low deformation energies, and white‐coloured tubes moderate energy values. The N‐tail, GAF and PAS domain and the linker helix are labelled.

Discussion

The anaerobic 1,3‐DHB degradation pathway of A. anaerobium is unique not only because it represents a different aromatic degradation strategy in anaerobes [2], but also because it constitutes a novel and complex mode of bEBP‐mediated transcriptional regulation, as evidenced in this study. The presence of two genes coding for almost identical bEBPs in the cluster initially raised the possibility of their function being redundant for pathway activation. Our previous genetic analysis established that both regulators were required for full strength pathway induction in a manner that required them to form hetero‐oligomers [4]. The results presented in this study have helped to clarify important points in the activation steps involved in pathway regulation.

The first relevant finding of this study is that despite the high sequence identity between the two regulators, they operated using completely different activation mechanisms. RedR1 behaved as a typical negatively regulated bEBP, as the protein depleted of its NTD (ΔNTD‐RedR1) was constitutively active (Table 1). In contrast, the active form of RedR2 was the full‐length protein, thus resembling positively regulated bEBPs. However, RedR2 was active in the absence of any external activating signal, which contrasts with the strict need of NTD activation to promote transcription in the known positively controlled bEBPs [5, 7]. It is remarkable that subtle changes in the regulatory domain of the activators suffice to totally change the way of sensing the inducing signal and controlling their activity.

In the natural host, the activating signal is the presence of the substrate 1,3‐DHB, which needs to be metabolized to HHQ by the resorcinol hydroxylase, and further to HBQ by the activity of BtdSL (Fig. 1). This latter activity promotes RedR2 release from the membrane to the cytoplasm, and produces HBQ, the effector of RedR1. Thus the active forms of the regulator are RedR1 activated by HBQ, which can be mimicked by truncated ΔNTD‐RedR1, and RedR2 when freely available in the cytoplasm. Both are capable of interacting, probably forming hetero‐hexamers, the most efficient conformation in promoting transcription from P _orf14 . The dual effect of 1,3‐DHB metabolic processing on the activity of the regulators connects a pathway component with its regulatory system, to avoid futile induction by potential non‐substrate analogues.

A second piece of evidence from this study is RedR1‐RedR2 hetero‐interactions (Fig. 3), and the absence of self‐interactions of the two regulators in vitro and in BACTH [4]. Although column binding assays cannot differentiate hetero‐dimer formation from heterohexamers, the protein–protein interactions detected in BACTH assays indicate that the regulators do form heterodimers, which most probably further associate into hetero‐hexamers to promote transcription. Regulator co‐expression assays in E. coli confirmed that hetero‐interactions were favoured over self‐interactions, even if the protein combinations resulted in inactive regulators. Previous data suggested that self‐association of the regulators was impeded by their CTD in vitro and in BATCH assays in the absence of promoter DNA [4]. However, in the truncated form ΔNTD‐RedR1, which is constitutively active and thus is expected to self‐interact to associate into active hexamers, homodimer formation was allowed only if the CTDs, which encompass the DNA‐binding domain, were bound to their UASs in the target promoter DNA (Fig. 7C). In the absence of a stimulatory signal, bEBPs are normally bound to their UASs as two dimers [14] and assemble into a hexamer when activation of the NTD reorients this domain, eliciting the conformational change in each sub‐unit required for oligomer assembly on the DNA [30]. A case of extreme dependence on DNA binding for oligomerization is the NorR protein of E. coli, which is capable of DNA binding and hexamer formation in the absence of stimulatory signals, provided that its three target UASs are present [31]. Our data indicate that in ΔNTD‐RedR1 the CTD is also involved in ΔNTD‐RedR1 self‐association by binding its target UASs at the target promoter. This binding produces the conformational change that stabilizes the otherwise weak interactions between ΔNTD‐RedR1 monomers. In other words, the DNA of the UASs would behave as an allosteric effector that determines ΔNTD‐RedR1 conformation by controlling multimerization, and hence the activity of the regulator. We were not able to identify similar DNA‐mediated oligomer stabilization for RedR2 full‐length protein, which is able on its own to activate P _orf14 in the absence of any external signal, as shown in the heterologous host E. coli (Table 1) and in a btdS redR1 double mutant of Azoarcus sp. CIB (pR⁺) (Table 2).

A third key finding in this study is the identification of the unusual control mechanism of RedR2, mediated by its NTD binding the integral membrane protein BtdS, which sequesters the regulator to the membrane and restricts its availability to activate its target promoters (Fig. 6E). This implies that rather than being a positively activated bEBP, RedR2 would be a constitutive bEBP which, if made available in the cell, would readily be capable of binding its operator and undergo the series of events leading to transcription activation. BtdS is a transmembrane protein belonging to the poorly described DoxX/D family. The best characterized member of the family is DoxD from the archaeon Acidianus ambivalens, the small sub‐unit of a thiosulfate:acceptor oxidoreductase encoded by doxDA and involved in sulfur oxidation [32]. However, a divergent cluster of DoxX/D family members, annotated as DoxX, lacks any accompanying partner doxA gene; a different function is suggested for this group, to which BtdS would belong [33]. Analysis of the DoxX genome contexts points to these proteins playing a role in the delivery of quinone derivatives to the cell surface, as part of the respiratory electron transport at the membrane [34]. In the 1,3‐DHB degradation pathway in A. anaerobium, BtdS and BtdL are the small and large sub‐units, respectively, of HHQ dehydrogenase. Interestingly, the electron acceptor of DoxD in A. ambivalens is a quinone [33], as a quinone (HBQ) is also the product of the reaction carried out by BtdLS [3]. BtdS would thus play a triple role in the anaerobic degradation of 1,3‐DHB in A. anaerobium: (a) BtdS would control pathway expression by sequestering its main regulator RedR2 to the membrane; (b) in the presence of the pathway substrate 1,3‐DHB, transformed into HHQ, BtdS subunit would be involved in HHQ oxidation to HBQ as the second step in the degradation pathway carried out by HHQ dehydrogenase (simultaneously releasing RedR2 into the cytoplasm); and (c) the production of HBQ by BtdSL would activate RedR1. BtdL, the large sub‐unit of HHQ dehydrogenase, has no function in RedR2 sequestration [4]; this role is exclusive to BtdS. RedR2 can thus be included amongst the regulators controlled by the so called trigger enzymes, i.e. enzymes with dual functions: besides their specific enzymatic activity, they play an active role in the control of gene expression [35].

From a structural point of view, it is striking that RedR1 was not repressed by BtdS. RedR2 NTD sequence, which is responsible for its binding to BtdS (Fig. 7A), shows two differences with respect to RedR1 NTD: the N‐terminal tail preceding the conserved sequence between the two proteins is different in sequence and length; additionally, Ile235 in RedR2 PAS domain Ile stretch is substituted by a Thr (Thr233) in RedR1 (Fig. 2A). RedR2 NTD lacking its N‐terminal tail completely lost its capacity of binding interaction with BtdS, as did the single point mutation Ile235 to Thr (Fig. 7A). The same was true for the equivalent full‐length protein mutants. Our results suggest that both the N‐terminal 8‐residue tail and the Ile stretch in the PAS domain are essential in RedR2 control through the interaction with BtdS. Interestingly, the Thr233 to Ile mutation in RedR1 NTD did not confer this regulator or its NTD the capacity to bind the BtdS, confirming that both elements are involved in the interaction. Protein modelling showed that the Ile stretch in RedR2 was exposed to the surface, forming a hydrophobic cavity in the junction between the PAS domain and the GAF‐PAS linker helix, which was interrupted by the polar Thr233 residue in RedR1 (Fig. 8). Additionally, the RedR2 N‐terminal tail (MSRETSLA) has a more polar character than the RedR1 tail (VDGFLS). The BtdS structure model suggests the presence of three potential contact regions facing the cytoplasm: Thr15‐Arg16 in the N‐terminal end, a short hydrophobic loop (LFGIY) between helices 2 and 3, where Phe76, Ile78 and Tyr79 are especially exposed, and Gln134‐Pro135 in the C‐terminal end. This conferred a hydrophobic character, with polar spots, to the protruding region, placing these structures as candidates for protein–protein interaction.

Several bEBPs are regulated by protein–protein interactions, some of which also respond to changes in the membrane redox state. The best‐characterized examples are FleQ, NifA, PspF and DhaR, the regulation of which shows interesting similarities with RedR2 regulation. DhaR is a special case of bEBP, since it does not control a σ⁵⁴‐dependent but rather a σ⁷⁰‐dependent promoter, and its central domain lacks the σ⁵⁴‐binding motif [36]. However, the structure of its NTD with GAF and PAS‐like tandem domains resembles the RedR proteins NTD and was successfully used to model their structure (Fig. 2C). DhaR regulates the expression of the two sub‐units of the dihydroxyacetone kinase in E. coli, DhaK and DhaL, which are also the regulatory elements that control DhaR activity: DhaR is activated by interaction with DhaL whilst it is repressed by DhaK in a manner that is dependent on the ATP versus the ADP load of DhaL [28]. The NifA protein, which includes a GAF domain in its NTD, is controlled through protein–protein interaction with its anti‐activator NifL, to prevent the hexamer assembly [37]. Interestingly, NifL of K. pneumoniae, containing tandem PAS domains in its NTD, is membrane‐associated (sequestered) when the FAD of one of its PAS domains is reduced by the reduced quinone pool generated during anaerobic respiration. In the presence of oxygen, NifL dissociates into the cytoplasm and represses NifA through direct protein–protein interaction, where its N‐terminal tail seems to be essential [38, 39]. Although ΔNTD‐RedR1and RedR2 are also capable of promoter activation under aerobic conditions in E. coli (Table 1), the 1,3‐DHB degradation pathway in the natural host is strictly anoxic, and activation of the regulators through 1,3‐DHB transformation into HBQ by the pathway enzymes only occurs in the absence of oxygen. In fact, HBQ is highly unstable in the presence of oxygen. Thus, it is plausible that this mode of pathway control through an enzymatic pathway involved in quinone production was selected to guarantee that the environmental conditions (i.e., oxygen concentration) are appropriate before the pathway is fully induced. This is especially important in nitrate reducing bacteria, which are generally facultative anaerobes and normally thrive in fluctuating oxic‐anoxic environments [40].

PspF, the phage shock protein regulator, is another example of a bEBP being controlled through membrane related proteins capable of sensing the redox state of the quinone pool [41]. This bEBP lacks an NTD and is constitutively active, being repressed through protein–protein interaction by PspA, a peripheral membrane protein that inhibits its ATPase activity [42]. Finally, control of FleQ regulator varies depending on the host. In Pseudomonas aeruginosa, FleQ ATPase activity is repressed through direct protein–protein interaction with FleN, a P loop ATPase [43, 44].

We have shown that under natural conditions in Azoarcus sp. strain CIB bearing pR⁺, maximum promoter activity is only obtained when both RedR1 and RedR2 are present, and the activating signal (the presence of the 1,3‐DHB substrate metabolized to HBQ) is provided [4] (Table 2). The RedR2‐dependent constitutive transcription level in the absence of RedR1 was 50% lower than the induced levels in the wild‐type, where RedR1 and RedR2 associate to promote transcription. Because native RedR1 by itself is incapable of promoter activation (see negligible activity values in the redR2 mutant, Table 2), the activity in the wild‐type strain is not the sum of the activities corresponding to each regulator, but rather the activity stemming from the novel active conformation of the regulators composed of RedR1 and RedR2 dimers presumably assembled into heterohexamers. Our data strongly suggest that in their native conformation, the affinity of the RedR proteins for their counterpart is higher than their affinity for themselves, favouring the formation of heterodimers/oligomers. Two examples of bEBP heterodimer formation have been described: Pseudomonas syringae HrpS‐HrpR and Rhodobacter sphaeroides FleQ‐FleT systems. However, both of them significantly differ from the RedR1‐RedR2 system analysed here. Although FleQ is believed to form hetero‐oligomers with FleT, this latter bEBP lacks both the NTD and HTH motifs. Interestingly, FleQ is capable of transcription activation on its own at P _fleT promoter, but both FleQ and FleT are required to activate the remaining target promoters [45]. This resembles the situation with RedR2, which can on its own partially activate its target promoters, but strictly requires RedR1 for full activation [4]. HrpS and HrpR bEBPs are two interesting paralogues that lack an NTD. They form transcriptionally active heterohexamers, and are negatively controlled through protein–protein interaction between HprS and HrpV, which fulfils a similar role to that of PspA with PspF. These two bEBPs are believed to have evolved from a single ancestral gene duplication event, after which sequence degeneration has led to two paralogues with a specialized role in bEBP activity control [46]. It is highly probable that redR1 and redR2 also originated from a gene duplication event, and they further evolved to sense the same ultimate signal (the presence of a metabolizable substrate) through different mechanisms. The mechanism of bEBP control of σ⁵⁴‐dependent promoters shares certain features with eukaryotic promoters, such as their binding at a distance from the RNAP binding site. The RedR system of A. anaerobium, which integrates different levels of signals through the formation of hetero‐hexamers of differently regulated protein factors to produce a single transcriptional output, would mimic another relevant feature characteristic of eukaryotic transcription. The final benefits of this dual mode of regulation in the control of A. anaerobium 1,3‐DHB degradation pathway are not obvious, but it probably allows a finer co‐ordination of the expression levels of the five pathway operons coding for 20 genes, which certainly involves a significant energy cost to the cell that needs to be tightly controlled, especially under anoxic conditions.

Materials and methods

Standard procedures

The plasmids were isolated with a Qiapreps spin plasmid kit from Qiagen (Hilden, Germany). PCRs and cDNA were purified with Qiaquick gel extraction kit from Qiagen. The oligonucleotides were synthesized by Sigma‐Aldrich (St. Louis, MI, USA). DNA sequencing was performed by the DNA Sequencing Service at the López‐Neyra Parasitology and Biomedicine Institute (IPBLN; Granada, Spain). Transformation, PCR amplification with the Expand high‐fidelity system (Roche, Mannheim, Germany), and protein analysis were performed according to standard protocols [47].

Bacterial strains, media, and culture conditions

The bacterial strains and plasmids used in this study are summarized in Table S1. Azoarcus sp. strain CIB [48] was used for heterologous expression of the 1,3‐DHB degradation pathway encompassed in the pR⁺ cosmid [3]. Azoarcus sp. strain CIB bearing the pR⁺ plasmid and plasmid mutant derivatives was cultured anaerobically at 30 °C without shaking in 50 or 100 mL infusion bottles containing non‐reduced Widdel mineral medium (WMM) under nitrogen gas [49]. The medium was buffered with 30 mm 3‐(N‐morpholino) propane sulfonic acid (MOPS) instead of bicarbonate and supplemented with 8 mm nitrate. 1,3‐DHB (2 mm), glutarate (5 mm) or both (1 and 3 mm, respectively) were added from 0.5 m stock solutions prepared under nitrogen gas. For genetic manipulations, Azoarcus sp. strain CIB (pR⁺) and derivatives were grown aerobically in WMM supplemented with 5 mm glutarate as the carbon source. Solid media for the Azoarcus sp. strain CIB derivatives were prepared with 1.6% twice‐washed Difco agar. When necessary, antibiotics were used at the following concentrations: tetracycline, 5 μg·mL⁻¹; kanamycin, 25 μg·mL⁻¹; ampicillin, 50 μg·mL⁻¹ and chloramphenicol, 15 μg·mL⁻¹. E. coli strains were cultivated in Luria‐Bertani (LB) medium (10 g L⁻¹ tryptone, 5 g L⁻¹ yeast extract, 5 g L⁻¹ NaCl) or 2xYT medium (16 g L⁻¹ tryptone, 10 g L⁻¹ yeast extract, 5 g L⁻¹ NaCl) in Erlenmeyer flasks at 37 °C and 200 r.p.m. on a rotary shaker. When required, the cultures were supplemented with 100 μg·mL⁻¹ ampicillin, 50 μg·mL⁻¹ kanamycin, 50 μg·mL⁻¹ streptomycin, 0.8 mm 5‐bromo‐4‐chloro‐3‐indolyl d‐galactopyranoside (X‐Gal) and 0.5 mm isopropyl‐d‐1‐thiogalactopyranoside (IPTG).

Plasmid construction

Truncated versions of redR1 and redR2 genes lacking the NTD were amplified from isolated pR⁺ cosmid DNA by PCR using the primers indicated in Table S2. The resulting products were purified and cloned in pGEM‐T (Promega, Madison, WI, USA) and confirmed by sequence analysis. The pGEM‐T derivatives were digested with NdeI/HindIII, and the fragments containing the ∆redR1 and ∆redR2 genes were ligated and fused in frame to pET28b+ (Novagen, Madison, WI, USA) to create pET28b‐∆redR1 and pET28b‐∆redR2. These were cut with XbaI and HindIII and the resulting fragment containing the protein gene was cloned in pMMB67EH [50], to render plasmids pMMBΔRedR1 and pMMBΔRedR2 used for the β‐gal assays. For the full‐length version of RedR1, the same approach was used except that plasmid pET28bRedR1 [4] was cut with XbaI and HindIII and the resulting fragment was cloned into pMMB67EH to obtain pMMBRedR1. To construct pJBRedR2, a HindIII fragment from pR⁺ encompassing the redR2 gene and short flanking regions was first cloned in pBlueScript II SK (Stratagene, San Diego, CA, USA) and then sub‐cloned as an EcoRI‐KpnI fragment into pJB3Km [51]. To construct pMALBtdS, the btdS gen was amplified using the primers indicated in Table S2. The resulting fragment was purified and cloned in pGEM‐T (Promega) and confirmed by sequence analysis. The resulting plasmid was digested with SalI and HindIII and the fragment was cloned into pMAL‐pV [18] to obtain pMALBtdS.

Site‐specific homologous inactivation of genes in the pR⁺ cosmid

The facultative anaerobe Azoarcus sp. strain CIB [24] bearing the pR⁺ cosmid was used for the physiological characterization of the mutants in the 1,3‐DHB degradation pathway. All the plasmids used for allelic replacement were based on the pKNG101 suicide vector [52] and are listed in Table S1. The mutants in pR⁺ were generated by double homologous recombination using pKNG101 plasmid derivatives as described previously [53]. To maximize recombination efficiency and counter selection, mutagenesis was performed with the recipient pR⁺ plasmid or its derivatives in E. coli strains ET8000 [54] and DH5α [55]. The btdS mutant was constructed by two rounds of overlapping PCR. In a first round, the upstream btdS flanking region (647 bp) and a gentamicin resistance cassette from pBBR1‐MCS5 [56] were amplified with the appropriated primers (Table S2) to create overlapping ends, and used as template for a second round of PCR with the outer set of primers of the first reactions. The procedure was repeated combining the downstream btdS flanking region (678 bp), and then the two PCR products containing the same gentamicin resistance gene were used as template for a final PCR with the outer set of btdS primers used in the first reactions. The final fragment was cloned in pGEM‐T (Promega) and sequenced to exclude the presence of point mutations in the sequences flanking btdS. The fragment was cut and cloned between the ApaI and SpeI sites of pKNG101 to obtain pKNGBtdS::Gm. This plasmid was used to deliver the mutation to pR⁺ cosmid in E. coli ET8000 by RP4‐mediated mobilization, as follows: the E. coli donor strain CC118λpir (pKNGBtdS::Gm), recipient strain ET8000 (pR⁺) or mutant derivatives, and HB101(pRK600) helper strain were grown overnight on LB medium in the presence of the appropriated antibiotics. Samples of 0.5 mL of each culture were harvested by centrifugation, the cells were washed with 1 mL LB, the resulting pellets were combined in 100 μL LB and dropped onto a sterile 47 mm diameter, 0.22 μm‐pore‐size filter (Schleicher and Schuell, Dassel, Germany) placed on an LB plate, and incubated overnight at 30 °C. The filters were transferred to 3 mL of M9 salt solution [57], and cells were washed off by vigorous vortexing. E. coli ET8000 (pR⁺) transconjugants were selected for their resistance to nalidixic acid, tetracycline and gentamicin in LB solid medium. Amongst them, those where double crossovers had occurred were checked for sucrose‐resistance/streptomycin‐sensitivity. The btdS::Gm mutation was also delivered in the redR1 and redR2 mutants and in the redR1 redR2 double mutant of pR⁺ directly in Azoarcus sp. CIB bearing the corresponding mutants of (pR⁺). To that end, we first selected a transconjugant bearing a co‐integrate of pKNG101BtdS into the pR⁺ derivative on WMM with 5 mm glutarate as the sole carbon source and 50 μg·mL⁻¹ streptomycin in addition to tetracycline. Streptomycin‐resistant transconjugants were analysed by PCR with primers flanking the btdS gene to confirm the presence of a wild‐type and a mutated copy of the btdS gene, and a correct clone was selected and cultured in liquid LB medium with only tetracycline during 12–16 h, to promote the second crossover event and the allelic exchange to occur. To select double recombinants, colonies were plated on LB with 10% (w/v) sucrose. Streptomycin‐sensitive/sucrose‐resistant colonies were analysed by PCR to confirm the double cross‐over. The correct insertion of all mutations was confirmed by PCR analysis and Southern blotting.

Triparental conjugation of cosmids and plasmids into the Azoarcus sp. strain CIB

Wild‐type and mutant pR⁺ cosmid derivatives, as well as pMP190 derivatives were transferred to the Azoarcus sp. strain CIB by RP4‐mediated mobilization. The Azoarcus sp. strain CIB was grown aerobically to saturation in WMM supplemented with glutarate (5 mm), and E. coli donor strains bearing either pR⁺ or its mutant derivatives, or pMP190 derivatives, and the HB101 (pRK600) helper strain were grown overnight on LB medium in the presence of the required antibiotics. Samples of the E. coli cultures (1 mL each) and 15 mL of the Azoarcus sp. strain CIB culture were harvested by centrifugation and the cell pellets were washed once with 1 mL of WMM and combined in 100 μL WMM. The resulting cell suspension was dropped onto a sterile 47‐mm‐diameter, 0.22‐m‐pore‐size filter (Schleicher and Schuell) placed on an LB plate, and incubated overnight at 30 °C. The filters were then transferred into 1 mL of WMM, and the cells were washed off by vigorous vortexing. Azoarcus sp. strain CIB transconjugants where selected aerobically for their resistance to tetracycline (pR⁺ derivatives) and chloramphenicol (pMP190 derivatives) on WMM supplemented with glutarate (5 mm) to counter‐select against E. coli donor and helper strains.

Bacterial adenylate cyclase two‐hybrid assays

BtdS was amplified by PCR using the primers indicated in Table S2. The resulting amplicon was purified and cloned in pGEM‐T (Promega) and verified by sequencing. The pGEM‐T derivative was digested with SalI and HindIII and sub‐cloned into pUT18C to create pUTC18btdS. The constructions of the complete and truncated versions of the RedR1 and RedR2 proteins fused to pKT25 were already available [4]. Nucleotide substitutions Ile235 to Thr in RedR2 and Thr233 to Ile in RedR1 in their PAS domain were generated by overlapping PCR on both the full‐length proteins and their cloned NTD with two pairs of complementary mutagenic primers (Table S2) using pGEMT‐NTD‐RedR1, pGEMT‐NTD‐RedR2, pGEMTRedR2 and pGEMTRedR1 plasmids [4] as templates (Table S1). After the mutagenesis, RedR1(T/I) and RedR2 (I/T) were subcloned in pKT25 and pUT18, respectively. Plasmid pBBporf14 was obtained by cloning a 341 bp DNA sequence encompassing P _orf14 promoter between positions +58 and −227 as XbaI/BamHI fragment into pBBR1MCS‐5 [56]. To create pBBporf14‐UAS, encompassing P _orf14 promoter with mutated UASs, nucleotide substitutions in the P _orf14 promoter was carried out as above using p220Porf14 [4] as template and the primer pairs indicated in Table S2. The resulting mutant promoter was cut with XbaI/BamHI and subcloned into pBBR1MCS‐5 [56]. The BACTH assays were performed as previously described [26]. The pKT25, pUT18 and pUT18C plasmids carrying redR1, redR2 and btdS genes or their derivatives, and pBBporf14 and pBBporf14‐UAS when required, were transformed in different combinations into E. coli BTH101 cells, and interaction efficiency between the plasmid‐encoded proteins was quantified by measuring β‐galactosidase activity of the cultures following overnight induction with 0.5 mm IPTG. Assays were minimally performed in triplicate and standard errors of the mean were calculated.

Betagalactosidase assays

For assays in E. coli strains BTH101, MC4100, BW25113, JW1702‐1, JW3169‐1 and BL21, 3 mL LB supplemented with the required antibiotics were inoculated with a single colony of the E. coli strain bearing the appropriate combination of plasmids, and incubated overnight at 30 °C. Fresh 3 mL LB was then inoculated with 1/100 of these overnight cultures and incubated at 30 °C with shaking. When the cultures reached the exponential phase, or after overnight growth, β‐galactosidase activity was determined as described previously [58]. In the case of assays in BL21 strain, 1 mm IPTG was added in exponential phase and the cultures were incubated two more hours before enzyme assays. For assays in Azoarcus sp. strain CIB, 3 mL WMM supplemented with 5 mm glutarate as the carbon source and the required antibiotics were inoculated with a single colony of each strain bearing the appropriate combination of plasmids and incubated aerobically (100 r.p.m.) for 3 days at 30 °C. The pre‐grown culture was then inoculated into 45 mL of WMM supplemented with the indicated carbon source to an initial OD of 0.05 and incubated under anoxic conditions at 30 °C without shaking. At the indicated times, β‐galactosidase activity was determined in permeabilised whole cells (0.1 mL) as above.

Construction of human influenza haemaglutinin epitope‐tagged RedR1 and RedR2

To construct the plasmids for the production of C‐terminal HA‐tagged RedR1 and RedR2 proteins, the genes were amplified using the primers indicated in Table S2 that included the sequence for the HA tag. The resulting fragment was purified and cloned in pGEM‐T (Promega) and confirmed by sequence analysis. The pGEM‐T derivatives were digested with NdeI and HindIII and the fragments containing the tagged redR1 and redR2 genes were fused to pET24b+ (Novagen) to create pET24bRedR1HA and pET24bRedR2HA. E. coli BL21(DE3) [59] transformed with plasmids pET24bRedR1HA or pET24bRedR2HA were used to overproduce RedR1HA and RedR2HA with a C‐terminal HA‐tag. To construct the plasmids pMMRedR1HA and pMMRedR2HA used in RedR proteins localization, the pET24bRedR1HA and pET24bRedR2HA plasmids were digested with XbaI/HindIII and the fragments coding for RedR1HA and RedR2HA were cloned in pMMB67EH [50] digested with the same enzymes.

Overproduction and purification of 6His‐RedR1, 6His‐RedR2 and MBP‐6His‐BtdS by affinity chromatography

The RedR1 and RedR2 proteins were expressed from pET28b as N‐terminal His‐fusion proteins keeping the stop codon of their own sequence, located 36 bp upstream the pET28b polylinker stop codon. Stop codon readthrough originated two proteins: the cloned 6His‐RedR protein and a 6His‐RedR protein with 12 additional residues at the C‐terminal end, rendering a double band in western assays. E. coli strains BL21 (pET28bRedR1), BL21 (pET28bRedR2) [4] and BL21 (pMALBtdS) cells were grown in 1 litre of 2xYT medium supplemented with 25 μg·mL⁻¹ kanamycin in an Erlenmeyer flask on a rotary shaker at 37 °C. When the cells had reached an A ₆₆₀ of 0.6, protein expression was induced with 0.1 mm IPTG and incubation was continued for 20 h at 18 °C. The cells were then harvested and stored at −20 °C until use. The frozen cells were suspended in 20 mL of lysis buffer A (20 mm Tris–HCl [pH 8.0], 500 mm NaCl, 10% glycerol, 10 mm Triton X‐100, 0.1 mm EDTA, 2.5 mm β‐mercaptoethanol, and 10 mm imidazole) and lysed by two passages through a French pressure cell. The crude extract was centrifuged at 13 000  g for 1 h at 4 °C. The resulting supernatant containing 6His‐RedR1, 6His‐RedR2 or MBP‐6His‐BtdS protein was loaded onto a HisTrap HP affinity column (GE Healthcare) prepared as indicated by the manufacturer and equilibrated with the lysis buffer A in an Äkta fast protein liquid chromatography system (GE Healthcare). After column washing with 10 bed volumes, the protein was eluted with 250 mm imidazole. The purity of the 6His‐RedR1, 6His‐RedR2 and MBP‐6His‐BtdS proteins was verified by sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS/PAGE). After dialysis against buffer B (20 mm Tris–HCl [pH 8.0], 300 mm NaCl, 10% glycerol, 0.1 mm EDTA, and 1 mm DTT), the protein concentration was determined with the Biorad protein assay kit using bovine serum albumin as the standard.

Column binding assay

Before starting these assays, the 6His‐RedR1, 6His‐RedR2, or MBP‐6His‐BtdS proteins were purified as described in the previous section. After dialysis against buffer B without imidazol, the assay started by binding the purified protein to a HisTrap HP affinity column (GE Healthcare). Then a crude extract containing the overproduced protein of interest (RedR1‐HA or RedR2‐HA) was obtained by 3 passages of 1 L culture of either BL21 (pET24bRedR1HA) or BL21 (pET24bRedR2HA) concentrated in 30 mL of buffer A through a French pressure cell. The crude extract was centrifuged at 13 000  g for 1 h at 4 °C. The resulting supernatant containing the HA‐tagged protein was loaded onto the column where the 6His‐RedR1, 6His‐RedR2, or MBP‐6His‐BtdS protein had been previously loaded, then the column was washed with up to 10 column volumes of buffer A. The bound proteins were then eluted in buffer B supplemented with 250 mm imidazole. Fractions (2 mL) were collected throughout the process and analysed by western blot with either anti‐His or anti‐HA antibodies to identify each protein.

Preparation of cytosol and membrane fractions for western blot

Azoarcus sp. strain CIB carrying the wild‐type pR⁺ plasmid or its btdS mutant (pR⁺ btds::Gm), together with pMMRedR1HA or pMMRedR2HA, was grown anaerobically at 30 °C without shaking in 50 mL WMM in infusion bottles supplemented with glutarate (5 mm) and glutarate (3 mm) plus 1,3 DHB (1 mm). When the cultures reached an OD₆₆₀ of 0.3–0.4, 15 mL of culture were harvested by centrifugation and the cell pellets were resuspended in 1 mL sonication buffer (20 mm Tris·HCl pH8, 5 mm EDTA), snap frozen in liquid nitrogen and thawed on ice. After 5 min incubation at 40 °C and transfer to ice, the sediments were sonicated on ice in five pulses of 20 s. The extract was then centrifuged at 3000  g for 5 min. The resulting supernatant was centrifuged at 12 000  g during 45–60 min at 4 °C. The pellets, which constituted the membrane fraction, were solubilized in the same volume of Laemmli buffer; the proteins from the supernatant, which constituted the cytosol fraction, were precipitated with 10% (w/v) trichloroacetic acid, dried with acetone, and suspended in the same volume of Laemmli buffer. The proteins from cell lysates, membrane and cytosol fractions were analysed by SDS/PAGE and western blot as indicated in the next section.

SDS/PAGE and western blot

Proteins from column binding assays and cytosol and membrane fractions (25 μL) were fractionated by SDS/PAGE in 10% polyacrylamide gels and electro‐transferred to nitrocellulose membranes. The membranes were then blocked with 5% dry milk in PBS. Immunodetection was performed using anti‐His antibody [Penta·His Antibody (1 : 5000), BSA‐free from QiagenR] or a monoclonal antibody directed against the influenza hemagglutinin epitope (1 : 2000) (HA.11; Covance, Burlington, NC, USA) by incubating the membranes for 1 h at room temperature. Then the membranes were washed three times with PBS. The second antibody, polyclonal rabbit anti‐mouse immunoglobulins/HRH (Dako, Glostrup, Denmark) or the horseradish peroxidase‐conjugated rabbit anti‐mouse (Dako) (1 : 1000) was added and incubated for 1 h, the membranes were washed three times with PBS, and the signal was detected using the SuperSignal® West Femto Chemiluminescent Substrate (Thermo Scientific, Waltham, MA, USA). The blots were scanned and analysed using the quantity one version 4.6.7 (Bio‐Rad, Hercules, CA, USA).

Bioinformatics tools for sequence analysis

Nucleotide and amino‐acid sequences were analysed using the tools provided by the NCBI (http://blast.ncbi.nlm.nih.gov/Blast.cgi) and the ExPASy molecular biology server (http://www.expasy.org). The promoter sequences were aligned with clustalw2 at EMBL‐EBI (http://ebi.ac.uk). Computational models of the NTD three‐dimensional structure were built with phyre2 (http://www.sbg.bio.ic.ac.uk/phyre2/) [60] and I‐Tasser [61]. To improve the reliability of protein models, reiterative models were obtained for each protein NTD using different modelling servers and approaches. phyre2 models were built in quadruplicate, whilst a single model per protein was allowed from the I‐Tasser server. The homology models were assessed using the Swiss‐Model Structure Assessment Tool [62, 63] and for each protein the model with the lowest number of residues in the Ramachandran plot disallowed regions was selected for comparison and further analysis. The deformation energy of the predicted models was estimated using the normal mode analysis approach (NMA) as provided in the Dynamut server, selecting the C‐alpha force field option [29]. BtdS secondary structure model was built with C‐I‐Tasser and assessed in Swiss‐Model. Porter 4.0, PaleAle 4.0 secondary structure and relative solvent accessibility prediction server (http://distillf.ucd.ie/porterpaleale/) [64] was used to predict BtdS transmembrane regions. The protein structure models were visualized with ucsf chimera [65].

Conflict of interest

The authors declare no conflict of interest.

Author contributions

DP‐S planned and performed experiments and analysed data; PM and AM‐F performed experiments; SM planned the experiments and wrote the manuscript.

Peer review

The peer review history for this article is available at https://publons.com/publon/10.1111/febs.16576.

Supporting information

Table S1. Strains, plasmids and cosmids used in this study.

Table S2. Oligonuclkeotide primers used in this study.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.