The C. crescentus strains used in this study were the wt strain NA1000 and isogenic mutant strains Δfur, ΔoxyR, and ΔiscR; E. coli strains were DH10B, S17-1 and BL21(DE3) (described in Supplementary Table S1). When necessary, the plates were added of the following antibiotics: 1 μg/mL Tetracycline, 5 μg/mL Kanamycin, 20 μg/ml Nalidixic Acid (for C. crescentus) and 12.5 μg/mL Tetracycline, 50 μg/mL Kanamycin (for E. coli). Plasmids used in this work are described in Supplementary Table S2.

Eventually, after RNA sequencing, we realized that the wt NA1000 strain clone used for these experiments had spontaneously lost the genomic island corresponding to a mobile element (Marks et al., 2010) since no reads were obtained between positions 473,069–499,098. This clone was also used to construct the IscR-3xFLAG strain, so no reads were obtained for this region in the ChIP-seq analysis.

Bacterial cultures were grown under aerobic conditions at a temperature of 30°C in PYE rich culture medium (Ely, 1991) in flasks or in 96-well microplates (200 μL cultures). To investigate the effects of iron limitation in cultures, the iron chelator 2′,2-dipyridyl (DP) (Sigma-Aldrich) was added to a concentration of 100 μM during an incubation period of 2 h.

Escherichia coli K12 MG1655 and C. crescentus NA1000 genomes were retrieved from NCBI databases. RBH analysis was conducted with easy-rbh workflow from MMseqs2 (v. 15.6f452) (Steinegger and Söding, 2017) and the results were analyzed by in-house Python scripts. Structural alignment of putative IscR proteins from selected bacteria was made with Expresso from webserver T-Coffee (Notredame et al., 2000). Gene synteny analysis was carried out with the with the SyntTax program (Oberto, 2013).

Protein structures were predicted using AlphaFold2 through the ColabFold platform (Mirdita et al., 2022). For each protein, 5 models were generated and ranked followed by built-in AMBER energy minimization of the top1 structure. Graphics were generated by the open-source software PyMOL.

The in-frame deletion of iscR gene from the C. crescentus NA1000 chromosome was obtained by allelic replacement. The flanking regions of the iscR gene were amplified by PCR with primers 01942IF1/01942IF2 and 01942IF3/01942IF4 (all primers mentioned henceforth are described in Supplementary Table S3). These two fragments were then cloned in tandem into the suicide vector pNPTS138 (M.R.K. Alley, unpublished) using the In-Fusion cloning kit (TaKaRa Bio) and transformed into competent E. coli DH10B cells. The vector was introduced into E. coli S17-1 and subsequently into C. crescentus NA1000 via conjugation. The kanamycin-resistant colonies obtained were grown in PYE for 48 h, followed by plating on PYE 3% sucrose. Kan-sensitive clones were investigated by PCR with primers 42CompF/42CompR, and a clone presenting a single 765-bp amplified fragment was selected as an iscR deletion mutant for further characterization.

Different fragments of the iscR promoter region were amplified by PCR, using the genomic DNA of C. crescentus NA1000 as a template and specific forward primers pLacZFA, pLacZFB, and pLacZFC combined with the reverse primer pLacZ2R. The amplified fragments were cloned into the pRKlacZ290 vector (Gober and Shapiro, 1992), generating vectors placZFragA, placZFragB, and placZFragC. The plasmids were subsequently transferred into C. crescentus strains NA1000, ΔiscR, ΔfuR, and ΔoxyR by conjugation with E. coli S17-1. To evaluate the activity of the promoter regions under study, the activity of the β-galactosidase enzyme was measured, using the method described by (Miller, 1972).

Strains containing these reporter plasmids were grown in PYE/Tet and expression was analyzed at logarithmic phase (OD_600 nm = 0.5) under different experimental conditions: in the presence of 100 μM DP for 2 h, or 60 μM H₂O₂ for 30 min.

The streptonigrin (SNG) sensitivity tests were carried out as described in Nachin et al. (2001) and Justino et al. (2007). The wt and ΔiscR strains were grown at 30°C in PYE, diluted to OD_600nm = 0.1, and aliquoted into new tubes, which received 0.05 μg/mL, 1 μg/mL, or no streptonigrin (Sigma-Aldrich), respectively. The cultures were incubated for 24 h at 30°C, serially diluted, plated on PYE medium and incubated for 48 h at 30°C to determine CFU/ml.

For RT-qPCR assays, cultures were grown in PYE medium up to exponential phase (OD_600 nm = 0.5), 1 mL aliquots were treated with TRIzol^® (Invitrogen Life Technologies), and the RNA was extracted following the manufacturer’s instructions. RNA (2 μg) was treated with one unit of DNAse I (Invitrogen) and a negative control PCR was carried out without reverse transcriptase using primers for the rho gene (Supplementary Table S3). cDNA synthesis was performed using the SuperScript III First-Strand Synthesis Kit for RT-qPCR (Invitrogen). RT-qPCR experiments took place using Power SYBR Green and PCR master Mix (Applied Biosystems), using the CCNA_01991 or rho genes as internal controls for reference. RNA was amplified in duplicate for each biological replicate with primers designed for each gene, and reactions were performed in the StepOnePlus real-time PCR system (Thermo Fisher Scientific). The relative differential expression was calculated using the 2^−ΔΔCt relative expression quantification method (Livak and Schmittgen, 2001).

Three independent biological replicates from strains C. crescentus NA1000 (n = 3) and ΔiscR (n = 3) were grown in PYE medium up to OD_600 nm = 0.5 at 30°C with agitation at 200 rpm. Total RNA was extracted from 10 mL-cultures essentially as previously described (de Araújo et al., 2021). rRNA depletion and cDNA libraries were prepared using the Illumina Stranded Total RNA Ligation Kit with RiboZero Plus (Illumina) and subsequently sequenced using the NextSeq 500/550 Mid Output Kit v2.0 (Illumina) on the Illumina NextSeq 500 (Illumina). RNA sequencing was done in the Core Facility for Scientific Research – University of São Paulo (CEFAP-USP/GENIAL).

A DNA fragment containing the coding region of the iscR gene along with 1 kb upstream and downstream flanking regions was obtained by PCR using the Q5 DNA polymerase (New England Biolabs), genomic DNA from strain NA1000, and primers IscR Fw/IscR Rv. This blunt fragment was cloned using Zero Blunt™ PCR Cloning Kit (Thermo) and the recombinant pCR™-Blunt plasmid was used as template in a PCR reaction with outward primers that hybridized in the 3’end of the iscR coding region, containing 1.5xFLAG coding sequence in each of their 5′ regions. After ligating with T4 DNA ligase (Thermo), this reaction yielded a pCR™-Blunt plasmid containing a 3x FLAG coding sequence inserted in frame with the C-terminal region of IscR. The correct fusion was confirmed by DNA sequencing using the BigDye™ Terminator v3.1 Cycle Sequencing Kit (Thermo). The 1.7 kb-DNA fragment containing the iscR-3xFLAG gene was obtained by digestion with EcoRI (Thermo), cloned into the suicide vector pNPTS138 and inserted in E. coli S17-1 by transformation. The plasmid was transferred to C. crescentus NA1000 by conjugation, and after double recombination, the iscR gene was substituted by the iscR-3xFLAG version.

The protocol was followed essentially as described by (Gebhardt et al., 2020; Gozzi et al., 2022) in a threefold larger scale. Briefly, cultures from the IscR-3xFLAG strain were freshly diluted to OD_600 nm = 0.1 in 180 mL PYE at 30°C at 200 rpm. After reaching an OD_600 nm = 0.3, cultures continued to grow in the normal condition (PYE medium, n = 3) or under iron deficiency condition (PYE with 100 μM DP for 2 h, n = 3). Crosslinking was performed with formaldehyde to a final concentration of 1%, the cells were further incubated for 30 min, and glycine was added to 0.125 M for 15 min at room temperature. Cells were disrupted by sonication long enough to obtain DNA fragments between 250 and 500 nt, followed by centrifugation for 20 min at 12,000 × g at 4°C. 3xFLAG-tagged IscR was immunoprecipitated using 200 μL of Anti-FLAG M2 affinity gel beads (Sigma-Aldrich) at 4°C overnight. After washing the resin 5× at 4°C, the protein was eluted from the resin in 150 μL of elution buffer.

After obtaining the DNA-ChIP fraction, purification was carried out using the Qiaquick PCR purification kit (Qiagen) following the manufacturer’s instructions and eluted using 50 μL of MilliQ H₂O. DNA concentration and fragment size was determined using BioDrop (Biochrom) and Bioanalyzer (Agilent). The purified DNA was used to construct libraries with the NEBNext^® Ultra™ II DNA Library Prep Kit for Illumina^®, and the libraries were sequenced using the NextSeq 500/550 Mid Output Kit v2. 0 (150 cycles) (Illumina) on an Illumina NextSeq 500 (Illumina).

The iscR gene was amplified by PCR with primers pETIscR1/pETIscR2, cloned into vector pET28a in frame with the 6-histidine tag codons at the carboxi terminus, and transformed into E. coli BL21(DE3). The IscR-His protein expression was induced with 300 μM IPTG for 2 h and the bacterial cell pellet obtained from a 1-liter LB culture was resuspended in 40 mL of buffer A (50 mM HEPES, 100 mM NaCl, pH 8). Subsequent sonication was performed on ice at 37% amplitude for 4 min (30 s ON/30 s OFF) followed by clarification via centrifugation at 25,000 × g for 15 min. The clarified extract was loaded onto a pre-equilibrated 5 mL HiTrap Heparin HP column, utilizing an AKTAPrime Plus FPLC system with a flow rate set at 4 mL.min⁻¹. The column was washed with buffer A until the UV signal (measured at 280 nm) reached baseline and a linear gradient was applied, ranging from 0 to 100% of buffer B (50 mM HEPES, 1 M NaCl, pH 7), over a course of 10 column volumes (CV). Fractions collected during this gradient elution were subsequently analyzed via SDS-PAGE.

Reddish fractions containing IscR-His were combined, diluted 10-fold in buffer C (50 mM HEPES, 100 mM NaCl, pH 7), and loaded onto a pre-equilibrated 2 mL HiTrap SP FF column, maintaining a flow rate of 2 mL.min⁻¹. The column was washed with 10 CV and subjected to a linear gradient elution, starting from 0% and ending at 100% of buffer B. Fractions were likewise analyzed via SDS-PAGE and those containing IscR-His were then diluted twice with distilled water, quantified through UV absorption at 280 nm (considering ε_IscR = 11,585 M⁻¹ cm⁻¹ and _MWIscR-His = 21.04 kDa) and stored at 4°C for subsequent use. The protein sample was also loaded into quartz cuvettes and subjected to UV–Vis absorption scanning from 300 to 900 nm (integration time = 0.01 s, interval = 1 nm) with a BioChrom BioDrop Duo spectrophotometer. As a control for the apo state, the protein sample was incubated with EDTA and DP at 100X and 10X the protein concentration, respectively, for 5 min at room temperature.

The electrophoretic mobility shift assay was performed using a 442 bp probe from the iscR promoter region that was obtained by amplification from C. crescentus NA1000 DNA with the primers placZFA and placZ2R. The probes for the other promoters were obtained with specific primer pairs generating DNA fragments of approximately 250 bp. DNA binding assay was performed in 20 μL in 10 mM HEPES, 40 mM KCl, 1 mM MgCl₂, 1 mM MnCl₂, 5% glycerol, salmon sperm DNA (0.1 mg.ml⁻¹), 500 ng of the DNA probe and increasing amounts of purified IscR-His protein (5, 250, and 500 nm). After incubation for 30 min at room temperature, the samples were analyzed by electrophoresis in a 5% polyacrylamide gel in 1X Tris-borate buffer (TBE) for 1 h at 20 mA. The detection of bands was carried out using the Electrophoretic Mobility Shift Assay kit (Thermo Fisher) or ethidium bromide.

RNA and DNA sequencing data were processed using the frtc pipeline¹ (Ten-Caten et al., 2018). We aligned the reads to C. crescentus NA1000 genome using the assembly ASM2200v1.

We used uniquely aligned RNA-Seq reads for gene differential expression analysis as in (de Araújo et al., 2021), using DeSeq2 (Love et al., 2014) and custom R scripts. We analyzed COG category enrichment as in (de Araújo et al., 2021). We removed the genes belonging to the genomic island that is absent only in parental strain (CCNA_00460, CCNA_00464 to CCNA_00482, CCNA_03921, CCNA_03922 and CCNA_03998) from the list of differentially expressed genes (see section “Construction of the ΔiscR strain”).

We used uniquely aligned ChIP-Seq reads as input for peak-calling with MACS2 (v2.2.6) (Zhang et al., 2008). Peak-calling was done individually for each replicate, using the parameters -g 4,047,433 [effective genome size, estimated with the khmer program (v2.1.1)] (Döring et al., 2008; Brown et al., 2015; Irber and Brown, 2016) and-f BAMPE, as the data is from paired-end sequencing. We then used the DiffBind R package (v3.12.0) (Stark and Brown, 2011; Ross-Innes et al., 2012) to (i) find significant MACS2-called peaks among the replicates and (ii) search for differentially bound regions by IscR between treatments. Reads in significant peaks were counted with dba.count command with parameters: bSubControl = F, minOverlap = 2, filter = 1,000, filterFun = mean, summits = 50. We performed the differential binding analysis using the treatment (addition of DP to PYE medium) as the contrast and the DBA_DESEQ2 method for the calculations.

We used the Integrative Genomics Viewer (IGV) (Robinson et al., 2011) to visualize and integrate all data. Comparison of the IscR regulon with previously published Fur regulon used data from da Silva Neto et al. (2013) and Leaden et al. (2018).

All experiments were conducted using a minimum of two independent biological replicates for each experimental condition, and in the qRT-PCR experiment two technical replicates were also employed. Mean and standard deviation were calculated from the data obtained from biological replicates, and to determine the statistical significance between different experimental conditions, the Student’s t-test (unpaired t-test) was used, with a significance level set at p < 0.05.

Caulobacter crescentus and other Alphaproteobacteria have a single operon encoding putative [Fe-S] cluster biosynthesis proteins that contains iscRS followed by orthologous of E. coli suf genes (Figure 1A and Supplementary Figure S1). IscS and SufS consist of cysteine desulfurases responsible for removing the sulfur groups from L-cysteine generating L-alanine and sulfur. The latter is then transferred to a conserved cysteine residue before subsequent transfer to scaffold proteins (Blanc et al., 2015). When comparing C. crescentus and E. coli genomes by Reciprocal Best Hits analysis, we were able to identify the C. crescentus genes CCNA_01941 and CCNA_01936 as putative orthologs for E. coli iscS (34.9% identity, YP_026169.1) and sufS (49.8% identity, NP_416195.1), respectively. To obtain a more robust hypothesis of orthology, we also modeled the tertiary structure of the proteins CCNA_01941 and CCNA_01936 with AlphaFold2. According to Fujishiro et al. (2022), general protein architecture is observable among members of the cysteine desulfurase family, but local structural features can distinguish them. CCNA_01936 presented a shorter catalytic loop and a typical β-hairpin region in the vicinity of the dimer interface. Both features are only observed in SufS-like proteins while longer catalytic loops and the absence of the β-hairpin are characteristic of IscS-like ones, as seen in CCNA_01941 (Supplementary Figure S2). Also, when analyzing the synteny of the iscS gene among several genomes from the Pseudomonadota phylum, we were able to determine that other Alphaproteobacteria present the same gene organization in this operon (Supplementary Figure S1). Moreover, in Betaproteobacteria and Gammaproteobacteria other than the Enterobacteriaceae, iscRS is part of the isc operon, like in E. coli (Figure 1A and Supplementary Figure S1).

The other genes downstream of iscS in C. crescentus are orthologs of the E. coli suf genes, so herein we will refer to this operon as the isc-suf operon for simplicity. The C. crescentus sufB, sufC, sufD and sufS orthologs are in the same order of the E. coli suf operon, except for a small ORF between sufB and sufC annotated as a putative ADP-ribosylglycohydrolase (Figure 1A). Downstream of sufS, CCNA_01935 is annotated as a FeS assembly SUF system protein (KEGG), followed by two ORFs. CCNA_01934 encodes a putative ortholog of sufA (50% similarity) and CCNA_01633 encodes a conserved hypothetical 66 aa-protein containing five cysteines. To verify that the genes were in fact co-transcribed, we designed primer pairs to amplify some of the gene junctions by RT-PCR. All the primer pairs generated an amplified band, confirming that all genes are part of the same operon (Figure 1B).

The protein encoded by CCNA_01942 possesses all the conserved structural features of IscR orthologs, which include the winged HTH motif for DNA recognition and the four residues that form the 2Fe-2S cluster binding (Cys⁹⁵, Cys¹⁰⁵, Cys¹¹², and His¹¹⁵) (Figure 2A). The three-dimensional model of IscR produced in AlphaFold2 shows a dimer with rotational symmetry in which the two Fe/S clusters are arranged on opposite sides, as well as the HTH and wing motifs (Figure 2B). At least 8 residues of each monomer are involved in the maintenance of the dimeric interface through polar interactions (Figure 2C).

To characterize the role of IscR, we generated a strain with an in-frame deletion of the iscR gene (ΔiscR). The mutant growth in PYE medium was slightly slower than the wt, but severely impaired in iron depletion (addition of DP) (Figure 3A). The expression of the iscR gene in trans from the low-copy number plasmid pMR20 directed by its own promoter restored the wt phenotype (Figure 3A). To verify whether the mutant had an imbalance in intracellular iron concentration, we performed a streptonigrin sensitivity test (Figure 3B). Streptonigrin is an antibiotic that requires a redox-active metal to be fully active and has been used as an indirect measurement of intracellular iron concentration (Zik et al., 2022). As seen in Figure 3B, the ΔiscR mutant does not show an increased sensitivity to 0.5 or 1.0 μg/mL streptonigrin when compared to wt, differently from the high sensitivity of the fur mutant (positive control). These results indicate that the lack of iscR does not generate an accumulation of intracellular iron, although cells present a much slower growth rate at low iron levels, probably due to incorrect gene expression for iron starvation response. We have also evaluated the response of the iscR mutant to oxidative stress generated by either H₂O₂ or paraquat (Supplementary Figure S3). Although the mutant seems to be slightly more sensitive to paraquat than the wt after 15 min, at all the other time points the phenotype was the same. These results indicate that the iscR mutant is not more sensitive to oxidative stress than the wt strain.

IscR has been reported to regulate the isc operon in several bacteria (Schwartz et al., 2001; Nesbit et al., 2009; Remes et al., 2015) and in C. crescentus the isc-suf operon was upregulated in low iron condition (da Silva Neto et al., 2013). In C. crescentus iscR is the first gene of the isc-suf operon, so we measured the expression driven by the isc promoter in the ΔiscR strain in conditions of sufficient or scarce iron (Figure 4). For that, three transcriptional fusions to the lacZ gene were constructed, containing successive deletions of the isc regulatory region (Figure 4A), as a tentative to map the sequences involved in gene regulation. The results showed that expression driven by the larger fragment containing 150 bp upstream of the iscR transcription start site (TSS) (Figure 4B, Fragment A) increases about 3-fold in low iron while its basal levels are already high to the same level in the ΔiscR mutant. Moreover, gene expression was further increased when the ΔiscR strain was in low iron conditions. These results indicate that IscR is repressing its own transcription in high iron (as a holo-IscR protein) but there is a second mechanism of regulation occurring at a low-iron condition. This increase can still be observed in construction B, which starts at position-110 bp, but is not seen with construction C that contains only the promoter (starting at position-49) (Figure 4B). These results suggest that an activator might bind upstream of the-35 region. We then tested whether the isc-suf operon upregulation under iron starvation was affected in the fur mutant (Figure 4C). The results showed that transcription driven from all fragments was increased in iron depletion, and that in the fur mutant the levels of expression were similar to wt. These data corroborate previous results (da Silva Neto et al., 2013) showing that the isc-suf operon was not directly regulated by Fur.

To evaluate the transcription levels when cells are exposed to oxidative stress, the wt and ΔiscR strains were incubated with 80 μM H₂O₂ for 30 min. The results showed that H₂O₂ has no effect on iscR expression, and transcription is derepressed in the ΔiscR mutant, as expected (Figure 5A). These results are in accordance with the previously reported C. crescentus oxidative stress stimulon, where iscR was not upregulated after 5 min or 15 min after H₂O₂ addition (Silva et al., 2019). To ascertain whether the oxidative stress regulator OxyR contributed to the isc induction, the same analysis was carried out using the oxyR mutant in different iron concentrations with the three constructs (Figure 5B). The results showed that expression in iron-depleted conditions was higher than in iron sufficiency in all constructs irrespective of the strain (Figure 5B), and the absence of OxyR had no effect on isc transcription. We have also evaluated if the gene responded to superoxide, incubating the cultures with 50 μM paraquat for 2 h prior to the assay, and no induction was observed (Figure 5C). Taken together, these results show that the isc-suf operon is not regulated in response to oxidative stress.

To determine the importance of IscR as a transcriptional regulator in C. crescentus, we carried out a global transcriptional profiling of the ΔiscR mutant vs. wt, grown in PYE at 30°C with agitation, in a condition of sufficient iron. The transcriptomic analysis identified 94 DEGs, with 47 upregulated (log₂FoldChange ≥ 1.0 and adjusted p-value <0.01) and 47 downregulated genes (log₂FoldChange ≤ −1.0 and adjusted p-value <0.01) in the ΔiscR mutant (Supplementary Table S4 and Figure 6A). The DEGs were categorized according to the Cluster of Orthologous Genes (COG) (Figures 6B, 7). The categories Coenzyme Transport and Metabolism, Inorganic Ion Transport and Metabolism and Lipid Transport and Metabolism were overrepresented among the downregulated genes, while Amino acid transport and metabolism, Carbohydrate transport and metabolism, Energy Production and Conversion and Nucleotide Transport and Metabolism were more predominant among the upregulated genes. The expression of some of the DEGs was determined by RT-qPCR and the results validated the RNA-seq expression profile (Figure 6D).

Among the upregulated genes there were operons encoding enzymes for arginine biosynthesis (CCNA_00617–00620), maltose transport (CCNA_02365–02370), and two operons for purine metabolism (CCNA_02687–02690 and CCNA_02699–02701) (Supplementary Table S4). Moreover, operons encoding cytochromes cb and bd from the electron transport chain (CCNA_01468–01471 and CCNA_00801–00802) were also upregulated. While some of these genes encode [Fe-S]- or iron-binding proteins, like CCNA_02690, FixG, CysI, CcoO, CcoP and GltB, others encode proteins binding to different divalent cations such as zinc and molybdenum. Among the most downregulated genes it is of note the operon encoding enzymes for the synthesis of riboflavin (CCNA00929-00932) and the qoxA-D operon (CCNA_01851–01849) that encodes a Heme/copper-type cytochrome/quinol oxidase (Supplementary Table S4). Accordingly, several genes encoding flavin adenine dinucleotide binding enzymes were also downregulated. Several genes involved in lipid transport and metabolism were downregulated, encoding distinct dehydrogenases and NAD-or FAD-binding enzymes (Figure 7).

A very intriguing result was that almost none of the genes belonging to the isc-suf operon were differentially expressed in the ΔiscR mutant, although we observed a threefold increase in expression driven by the isc-suf promoter in the ΔiscR mutant (Figure 4B). The RNA-seq results showed a twofold increase for iscS, but the difference was not considered statistically different by our parameters (adjusted p-value <0.01), and the sufS gene was downregulated in the iscR mutant. To clarify this, we investigated the variation in mRNA levels for several genes of the isc-suf operon by RT-qPCR (Supplementary Figure S5A). We observed that except for iscS, that is highly upregulated in the iscR mutant, the levels of the suf genes do not change between the strains. This result indicates that there must be a second layer of regulation that occurs after iscS, at the mRNA level.

To try to map a little more precisely where this regulation is occurring, we have designed 4 primer pairs for the beginning, middle and end of the iscS gene (Supplementary Figure S5B), and we determined the amount of transcript in each region by RT-qPCR in the ΔiscR mutant and wt. The results showed that the relative amount of mRNA from the beginning of the iscS gene is around 10-fold. However, this relation decreases further toward the end of the gene, suggesting that there must be a rupture of the operon mRNA within this gene. This could be a result either of premature transcription termination or cleavage of the transcript by RNases. The results taken together indicate that the difference in the expression levels through the C. crescentus isc-suf operon could result from distinct rates of mRNA turnover, as described for the E. coli isc operon (Desnoyers et al., 2009).

To determine the IscR binding sites in the whole C. crescentus genome, we constructed a strain encoding a chromosomally encoded 3xFLAG-tagged IscR to use in the ChIP-seq assay. This strain was grown in conditions of sufficient iron (cultures in PYE medium) or depleted of iron (cultures in PYE treated with 100 μM DP for 2 h). The addition of the 3xFLAG-tag did not alter the growth in these conditions (Supplementary Figure S6). The cell extracts were used for chromatin pulldown with an anti-FLAG resin and the DNA fragments isolated were identified by DNA sequencing (Supplementary Table S5).

The IscR regulator has been described to bind to distinct DNA sequences when in the apo-or holo-form, showing a [Fe-S] cluster-dependent DNA sequence discrimination (reviewed in Santos et al., 2015). The ChIP-seq experiment was carried out in two growth conditions: iron-sufficient (PYE medium) and iron-depleted (PYE + DP), which allows the identification of consensus sequences among IscR-binding peaks from each condition (Figure 8). We analyzed the results quantitatively by comparing the number of reads in the peaks between the cultures treated or not with DP, and only peaks with an average of the normalized reads higher than 1,000 were considered. In our analyses, 224 peaks were considered significant among replicates, with 90 differentially bound in each condition (88 more predominantly bound in DP, 2 more predominantly bound in PYE) and 133 with no difference in ligation between conditions. One hundred and twenty-six peaks localized in intergenic regions were considered to have a regulatory role.

While several peaks were present in both treatments, we considered peaks with more reads in PYE as sites predominantly occupied by holo-IscR and peaks with more reads in PYE+ DP as sites predominantly occupied by apo-IscR. However, although a 2 h treatment with DP has been shown to lead to iron depletion (da Silva Neto et al., 2009, 2013), this must be taken carefully since we cannot establish the real [2Fe-2S] occupation of IscR in our samples. We extracted 500 bp DNA sequences around the peak summit (250 bp on each side) and used them to search for conserved sequences with the MEME-ChIP tool (Machanick and Bailey, 2011). As shown in Figure 8B, we obtained two distinct motifs enriched at the peaks found for each of the conditions, which we named Type I (TypeI^Cc) enriched in the iron-replete conditions, and Type II (TypeII^Cc), enriched in the iron-depleted conditions. Only TypeII^Cc motif was found in peaks from both conditions, but with very few exceptions the peaks containing TypeI^Cc and TypeII^Cc were mutually exclusive in each gene. The TypeI^Cc motif (AWWHRAAMMWNYRAHMYMYAVRK TAWRWD) contains a widespread A-rich sequence as the Type 1 motif described for E. coli, which is remarkable given that the C. crescentus genome is 67% GC-rich. The TypeII^Cc motif (RAWDTCCAYRHCHTCBNCRKGGDYWTYS) is similar to the Type 2 motif described in E. coli (AWARCCCYTSnGTTTGMn GKKKTKWA) (Giel et al., 2006; Nesbit et al., 2009), with overall conservation of nucleotides but distinct spacing. We observed a conserved run of A-C in the first half motif followed by a run of G-T in the second half that indicates a symmetric or weak palindromic motif.

To confirm the ChIP-seq results in vitro, we tested IscR binding to some of the predicted DNA regions by electrophoretic mobility shift assay (EMSA) with an IscR-His purified from E. coli BL21(DE3). IscR was purified by heparin and ion exchange chromatography (IEX) since initial trials with nickel affinity showed high protein instability. The fractions from IEX presented a reddish color and when analyzed by UV–Vis spectroscopy demonstrated an absorption peak around 450 nm, suggesting the protein is in its holo state with [2Fe-2S] bound (Supplementary Figure S7).

The EMSA showed that IscR-His bound to several DNA regions that showed peaks in the ChIP-seq experiment (Figure 9 and Supplementary Figures S8, S9). As the expression driven from the iscR promoter is increased in the ΔiscR strain, we carried out a DNA-protein interaction assay using fragment A of the iscR promoter (Figure 4A). A peak for IscR binding was identified in the ChIP-seq overlapping the-35 region, confirming the expression results, and the EMSA results also confirmed the IscR binding (Figure 9). Although a second peak for IscR binding was identified at the end of the sufB gene, it is distant from the promoter of the sRNA R0151 encoded in the opposite strand from sufB, and likely does not regulate it. Moreover, there is still no evidence that it has any role in the expression of the downstream genes.

We identified several relevant genes involved in iron homeostasis and other stress responses as direct targets for IscR binding (Figure 9). The in vitro binding of IscR was confirmed for the aconitase (acn) regulatory region where the peak of IscR binding identified in ChIP-seq was located overlapping the-35 region in the promoter. The gene encoding the bacterioferritin (bfr) has a putative IscR target sequence upstream of its promoter, as well as the sodB gene encoding the iron-manganese superoxide dismutase, CCNA_00028 for the TonB-dependent putative iron transport, the rhlE gene encoding a DEAD-box RNA helicase, the operon for the synthesis of riboflavin and the dnaK gene (Figure 9 and Supplementary Figure S9). Other targets confirmed include the regulatory regions of the genes encoding the Hfq and RpoH (sigma32) protein (Supplementary Figures S8, S9).

A very interesting result was the IscR binding to the fixK-fixT intergenic region (Figure 9). A sequence with good similarity to TypeII^Cc motif was found overlapping the putative promoter and TSS of fixK, indicating that IscR represses fixK transcription. Although the fixK gene was not differentially expressed in the RNAseq most of the genes belonging to the fixK regulon were, as an indirect confirmation that the fixK gene is more expressed in the iscR mutant (Crosson et al., 2005). The patterns of expression of the operons for the respiratory terminal oxidases ccoNOPQ-fixGHI, cydCD, cydAB, and qoxABCD and the fixL gene were consistent with the higher expression of FixK in the cell. Taken together, the results indicate that IscR binds as an apo-enzyme, repressing transcription of fixK at low iron.

The genes differentially expressed in the mutant as determined by RNA-seq can be either directly or indirectly regulated by IscR. Twenty-two differentially expressed genes with an IscR binding site upstream identified by the ChIP-seq were considered as being directly regulated. A comparison between the IscR regulon and the Fur regulon (da Silva Neto et al., 2013; Leaden et al., 2018) showed that 63 genes are shared by both regulators (Figure 6C and Supplementary Table S6). We found several peaks for IscR binding upstream of regulators that could in turn affect the expression of specific sets of genes, as described for FixK above. Among those, it is of notice the stationary phase response regulator SpdR (Da Silva et al., 2016), the response regulator TacA (Marques et al., 1997),and the cell cycle regulator CtrA (Quon et al., 1996), although the relevance of these findings remains to be established.