Strains and growth conditions

Sulfolobus acidocaldarius was cultured while continuously shaking at 75 °C in Brock medium [22] supplemented with 0.2 % casamino acids, 0.2 % sucrose, 0.02 mg/ml uracil and if mentioned, with 10 mM β-alanine. While for ChIP-seq analysis S. acidocaldarius DSM639 was used, S. acidocaldarius MW001 [23] and MW001ΔbarR [17] were grown for relative gene expression quantification experiments.

ChIP-seq analysis

ChIP-seq analysis was performed for biological duplicates. S. acidocaldarius DSM639 was cultured while continuously shaking at 75 °C in Brock medium supplemented with 0.2 % casamino acids, 0.2 % sucrose, 0.02 mg/ml uracil and if mentioned, with 10 mM β-alanine. Crosslinked and sheared DNA was prepared from 200 ml cultures grown until reaching an optical density (OD_{600 nm}) of 0.3 as described before [4] with the following exception: while in previously established ChIP protocols with hyperthermophilic Sulfolobus spp. cells were cooled down before performing formaldehyde crosslinking at a constant temperature of 37 °C, here crosslinking was performed immediately after the culture was taken out of the incubator at 75 °C [Additional file 1]. After sonication, the sizes of sheared DNA fragments ranged from less than 100 bp to about 800 bp with a major proportion of the fragments between 100 bp and 600 bp.

ChIP was performed with polyclonal anti-BarR rabbit antibodies, which were previously validated for specificity [17], using Dynabeads M-280 sheep anti-rabbit IgG beads (Life Technologies) as described previously [24]. As a control, we also prepared an antibody-free mock sample. Subsequently, all samples were purified with the iPure DNA extraction kit (Diagenode) following 1 x 50 bp sequencing with the Illumina Miseq system (Scilife Lab, Stockholm, Sweden). Sequence reads were mapped to the S. acidocaldarius DSM 639 genome (NC_007181.1) with Burrows-Wheeler Aligner (BWA 0.7.10) [25] with default settings and MACS version 2 (2.1.0) [26] was used for peak calling (q value cutoff = 1.00e⁻⁸), followed by a manual curation. ChIP-seq results were visualized by IGV version 2.3.2 [27]. DNA sequences from ChIP-seq peak regions were extracted by BED Tools [28]. All raw sequence data files are available as supporting data. Binding motifs were identified in ChIP-seq enriched regions with MEME software version 4.10.0 using default parameters [29].

Electrophoretic mobility shift assays

Recombinant His-tagged BarR was overexpressed in Escherichia coli and purified from inclusion bodies as described before [17]. Electrophoretic mobility shift assays (EMSAs) were performed with DNA fragments generated by PCR using oligonucleotides [Additional file 2] of which one is 5’-end labeled with ³²P and using S. acidocaldarius genomic DNA as a template. Probes were designed to harbour the predicted binding motif. EMSAs were performed as described previously [30]. All binding reactions contained an excess non-specific competitor DNA and, when indicated, 1 mM β-alanine.

Reverse transcriptase quantitative PCR

Total RNA was isolated from exponentially growing S. acidocaldarius MW001 and S. acidocaldarius MW001ΔbarR cultures using an RNeasy mini kit (Qiagen). Residual genomic DNA was removed by treatment with Turbo DNase (Ambion) according to manufacturer’s instructions. First-strand cDNA was synthesized from 1 μg RNA with a SuperScript III First-Strand Synthesis SuperMix kit (Invitrogen). Primers [Additional file 2] were designed with Primer3 Plus software [31]. Reverse transcriptase quantitative PCR (RT-qPCR) was carried out in a Bio-Rad iCycler as described before [6]. Biological quadruplicates were assayed and normalization was performed with respect to the expression of reference genes Saci_0691 (encoding RNA polymerase subunit A) and Saci_1336 (encoding TATA binding protein). A paired t-test was performed to validate differential expression.

Genome-scale identification of BarR binding regions

To obtain a genome-wide view of the in vivo DNA interactions of BarR, we performed a ChIP-seq analysis using polyclonal anti-BarR antibodies. Since β-alanine is the specific ligand of BarR, this analysis was done for cells grown in the absence but also in the presence of exogenously added 10 mM β-alanine. Sequencing libraries were constructed of input DNA and immunoprecipitated DNA of cells grown in each of these conditions, which were subjected to next-generation sequencing and mapped to the S. acidocaldarius genome. Upon sequencing the input samples, no obvious bias was observed with the read count evenly distributed across the genome (Fig. 1a). Furthermore, relatively low background signals were observed for a mock IP control.

The largest fraction of the BarR binding loci are exclusively located in intragenic regions and moreover, besides the barR/Saci_2137 intergenic region, only a limited number of binding loci that encompass an intergenic region are located at a reasonably short distance from promoter regions to potentially cause regulation of transcription initiation (i.e., Saci0028, Saci1050, Saci1674, Saci1964 and Saci2166 peaks). Genes located in the direct neighbourhood of the peaks have a variety of functions, including amino acid metabolism (lysine arginine ornithine transport system, gltB) and sugar metabolism (gluconolactonase). Our analysis did not reveal binding in the neighbourhood of a malonate semialdehyde dehydrogenase gene (Saci_1700), which catalyzes the second step in β-alanine degradation and is genomically co-localized with and regulated by the orthologous gene/protein in Sulfolobus tokodaii [17]. Furthermore, genes encoding proteins involved in other aspects of β-alanine or coenzyme A metabolism did not display BarR association in the tested growth conditions. This indicates that there is only BarR-mediated β-alanine-dependent transcriptional regulation of the degradation of β-alanine but not of its biosynthesis.

In vitro validation of BarR binding regions

The interaction between BarR protein and in vivo bound genomic regions was validated in vitro by employing EMSAs for a selection of eight high-enrichment ChIP-seq peaks (Fig. 2). In addition to the Saci2136 and Saci2137 targets, for which stable in vitro binding was demonstrated before [17], six novel targets (Saci0061, Saci0839, Saci1115, Saci1674, Saci1796, Saci2319) were shown to form stable and specific BarR-DNA complexes. BarR-DNA complexes generally displayed relative low migration velocities, indicating a higher stoichiometric nature as is the case for the previously characterized complex with the barR/Saci_2137 intergenic region, in which BarR interacts with multiple regularly spaced binding sites [17]. All other tested targets displayed unstable low-affinity binding, visible by smearing. These results indicate that BarR interacts with its genomic targets in a direct and presumably sequence-specific manner.

Similarly as with the in vivo observations, the effect of β-alanine on the protein-DNA interaction varied for the different targets (Fig. 2). The complexes formed with the high-affinity targets Saci0061, Saci0839 and Saci2319 targets dissociated in response to 1 mM β-alanine, similarly as for the Saci2136 target, while for the other targets binding is only slightly or not at all affected by β-alanine.

Identification of the BarR DNA-binding motif

To predict the putative binding motif in the identified BarR binding loci, ChIP-seq sequences were analyzed by MEME software, a bioinformatic tool that searches for overrepresented sequence motifs in multiple unaligned sequences [29]. This analysis resulted in the identification of a 15 bp semi-palindromic binding motif 5’-TTGGAAAAATTACAA-3’ with an E-value of 8.5e⁻⁴ (Fig. 3), present in 20 of 21 peaks [Additional file 3]. This predicted binding motif is congruent with the sequences of the binding sites that were previously characterized by footprinting of BarR-DNA complexes formed with the barR/Saci_2137 intergenic region [17]. Together, the presence of a conserved recognition sequence and the observed in vitro binding indicates that the ChIP-enriched genomic sites are associated with BarR by direct sequence-specific interactions with the DNA and not through protein-protein interactions, as has been observed for other archaeal Lrp-type regulators [20, 32]. Notably, BarR and Ss-LrpB, an Lrp-type regulator in the related species Sulfolobus solfataricus, share a very similar sequence specificity [21, 33], suggesting that the helix-turn-helix motif-encoding parts of barR and Ss-lrpB share a common ancestral gene [Additional file 4].

In vivo binding at the barR/Saci_2137 genomic region

ChIP-seq analysis confirmed the in vivo association of BarR with the barR-Saci_2137 intergenic region that is responsible for autoregulation and regulation of aminotransferase expression [17] (Fig. 4a). However, in addition to the intergenic region for which binding was characterized previously in in vitro experiments, binding extends into the coding sequence of the BarR target gene Saci_2137, resulting in a complex binding profile with three peak summits. In silico analysis of the Saci_2137 coding sequence indeed identified the presence of two previously uncharacterized BarR binding motifs, which we named site D and site E, in addition to site C that is located upstream of the promoter (Fig. 4b). The three sites presumably mediating Saci_2137 regulation are regularly spaced with a very similar center-to-center distance (277 bp between site C and D and 283 bp between D and E). The locations and regular spacing of the three sites is reflected by the three peak summits in the ChIP-seq profile (Fig. 4a). It can be hypothesized that binding events at the major and auxiliary sites are not taking place independently from each other. Possibly, protein-protein interactions between BarR subunits bound at these different sites, whether or not part of the same pre-existing oligomeric protein molecule, result in the formation of a higher-order nucleoprotein structure in which the intervening DNA is looped out.

Binding of BarR to the intragenic binding sites had not been detected before and might underlay the differences observed previously between in vitro and in vivo detected BarR-DNA interactions with the barR/Saci_2137 intergenic region: while β-alanine causes the disruption of BarR-DNA complexes formed with a DNA fragment encompassing the intergenic region in vitro, this is not the case in vivo [17]. Furthermore, BarR activates Saci_2137 expression in the presence of β-alanine. With regards to binding, the ChIP-seq analysis reveals an opposite effect as compared to the in vitro observations, with higher number of sequence reads when cells were grown in the presence of exogenously added β-alanine (Fig. 4a). Possibly, binding to the newly identified intragenic sites stabilizes the complex upon the conformational changes induced by β-alanine. The presence of auxiliary operator sites in the coding sequence of the regulated gene has also been observed for the Lrp-type transcription factors LysM in S. acidocaldarius [7] and Ss-LrpB in the related Sulfolobus solfataricus [6].

Expression analysis of genes adjacent to BarR binding regions

Next, we determined whether or not BarR is involved in the regulation of the genes adjacent to the identified ChIP-seq binding regions. We performed qRT-PCR to analyze the effect of barR deletion on the expression of the most probable target genes, either located near high-affinity sites (Saci_0061, Saci_0839, Saci_1674, Saci_1797, Saci_2320 and Saci_2321) and/or displaying a binding event in the neighbourhood of their promoter region increasing the probability that BarR binding affects gene expression (Saci_1050, Saci_1797, Saci_2320 and Saci_2321). Gene expression was monitored both in the absence and presence of 10 mM β-alanine.

Only 2 out of the 7 tested genes displayed a significantly different expression in the MW001ΔbarR strain as compared to the isogenic MW001 (Fig. 5). These genes, Saci_2320 and Saci_2321, located in an operon and encoding a glutamate synthase enzyme (gltB), were downregulated 2.7-fold and 2.4-fold respectively in ΔbarR versus WT in the presence of 10 mM β-alanine. Hence, BarR also seems to activate glutamate synthase in response to β-alanine, thereby connecting the regulation of β-alanine to amino acid metabolism.

The 5 other tested genes did not show expression differences in both genetic backgrounds (Fig. 5) indicating that most sites uncovered in the ChIP-seq analysis, including intragenic sites, are non-functional in terms of transcription regulation. Given the presence of genuine BarR binding motifs and the in vitro validation of binding for most of these non-regulatory sites, it is unlikely that they are false positive artefacts of the ChIP-seq analysis. Also for other transcription factors, either in archaeal, bacterial or eukaryotic organisms, proof is accumulating that specific binding on the genome without an apparent regulatory output is a common feature of both global and local regulators [6, 7, 34–44]. Generally, more than half of all binding sites discovered by ChIP-seq for a transcription factor under study are intragenic and not directly linked to transcription regulation [6, 7, 35, 41–44]. The exact function of these sites, sometimes termed “decoy sites” [45] or “transcriptionally dormant sites” [40], is unclear although it is assumed that they optimize regulatory response dosage kinetics and dynamics by causing transcription factor titration and buffering [45–47]. Alternatively, these sites potentially contribute to gene regulation by regulating spurious intragenic transcription initiation events that are undetected, or by establishing long-range regulatory interactions [43]. We therefore hypothesize that most of the newly discovered BarR genomic binding sites in this study also serve an alternative function to direct transcription regulation.

Glutamate synthase is a regulatory junction for Lrp-type regulators in Sulfolobus spp

The highest-enrichment ChIP-seq peak (Saci2319 target) identified in this study (Fig. 6a) is one of the few binding events that results in transcriptional activation (Fig. 5) and can be considered as the only regulatory target besides the β-alanine aminotransferase. Interestingly, the regulated Saci_2320/Saci_2321 operon has previously been identified as a target of other Lrp-like transcription factors in Sulfolobus spp. (Fig. 6b). Indeed, the promoter region of glutamate synthase (gltB) encoded by the Saci_2320/Saci_2321 is a major binding target of the glutamine-responsive non-specific binding protein Sa-Lrp [48]. In the related S. solfataricus, the gltB promoter is associated with the lysine-responsive LysM [7] through direct protein-DNA interactions at a binding site that is conserved in S. acidocaldarius, which harbours a LysM ortholog, directly upstream of the promoter (Fig. 6b). Furthermore, another Lrp protein Ss-LrpB also associates at the S. solfataricus gltB promoter through protein-protein interactions with LysM [6]. Curiously, while S. acidocaldarius does not contain an Ss-LrpB ortholog, the S. acidocaldarius gltB promoter displays at least two BarR binding sites that are very similar to the Ss-LrpB consensus sequence [Additional file 4]. This suggests a common ancestral origin of these regulatory interactions. In contrast to BarR, for none of the above-mentioned Lrp-like regulators, a clear regulatory effect on gltB expression was observed in single deletion strains or when comparing different relevant growth conditions [6, 7, 48]. This observation demonstrates that there is a complex interplay between the different regulators and that regulatory effects are interdependent.

The binding motif that was predicted to be recognized by BarR [Additional file 3] is located quite far upstream of the gltB promoter as it starts at 433 bp upstream of the Saci_2320 translational start (Site A; Fig. 6b). While this binding motif is predicted to be a high-affinity site [Additional file 3] and is bound with high affinity in vitro (Fig. 2), the zoomed ChIP-seq profile displays only low-enrichment precipitation of this region (Fig. 6a). In contrast, high-enrichment precipitation is observed in the region more downstream with respect to this high-affinity site. Indeed, further in silico analysis enables the prediction of another BarR binding motif, not predicted by MEME, located just downstream of the promoter (Site B, Fig. 6b). Similar as for the barR/Saci_2137 target, it appears that BarR binding is more complicated and consists of several operator sites. The presence of other Lrp-like regulators in the control region might explain the discrepancy between the in vivo observed binding profile and the theoretically predicted and in vitro validated binding behaviour. Of note, the predicted LysM binding site is located just upstream of the promoter, a canonical position for transcriptional activators, while the BarR site B, which is presumably bound in vivo and responsible for activation, exerts BarR-mediated activation.

It is interesting to note that the Saci_2320/Saci_2321 operon encoding glutamate synthase is a common target of BarR and several other Lrp-family regulators in Sulfolobus spp. and that its control region can thus be considered as a DNA-binding hotspot for Lrp proteins. This adds to the variety of mechanisms in which archaeal Lrp-like regulators form transcription regulatory networks: i) different Lrp-like regulators bind to adjacent binding sites in the same control region (shown in this work); ii) they interact through protein-protein interactions resulting in genomic co-association [6, 20, 32], iii) paralogs share the same DNA-binding motifs [8], and iv) they regulate each other’s expression [19, 48]. For example, the transcription of barR has been shown to be regulated by Sa-Lrp [48].