**3. Results**

#### *3.1. Prediction of Promoters for the isc-suf Operon of R. sphaeroides Based on dRNA-Seq Analysis*

*R. sphaeroides* harbored a cluster of genes, with *isc* and *suf* homologs (Figure 1). RNA-seq data suggested the co-transcription of these genes, and this was also confirmed by further RT-PCR experiments [31]. The first gene of the *R. sphaeroides isc-suf* operon encoded the IscR regulator, which coordinates a Fe-S cluster with a unique Fe-S ligation scheme [31]. The other genes of the operon encoded two cysteine desulfurases (IscS and SufS), the membrane component of an iron-regulated

ABC transporter (SufB), a hypothetical protein (RSP\_0439), the ATPase subunit of an ATP transporter (SufC), an Fe-S cluster assembly protein (SufD), and two proteins of the Yip1 family (RSP\_0433 and RSP\_0432) (Figure 1). This arrangemen<sup>t</sup> of genes was highly conserved in the family of *Rhodobacteraceae* (Figure S1). The synteny of genes that were annotated as *isc* or *suf* genes was highly conserved among the anaerobic anoxygenic phototrophs (AnAPs) and aerobic anoxygenic phototrophs (AAPs). There was some variation in regard to the non-annotated open reading frames. The non-phototrophs *Ruegeria* and *Paracoccu*s also showed a similar synteny of *isc* and *suf* genes, and the gene for the IscR homolog of *Paracoccus denitrificans* was, however, positioned at a different chromosomal locus. A similar synteny of *iscS* and *suf* genes was also found in *Rhizobiales* (Figure S1); however, these bacteria lack IscR homologs [43]. The arrangemen<sup>t</sup> of genes did not provide information on the organization within an operon. In *Rhodobacter capsulatus*, SB1003 RNAseq data indicated co-transcription of *isc-suf* genes (GEO GSE134200) as in *R. sphaeroides*.

**Figure 1.** Schematic overview of the *isc-suf* operon of *R. sphaeroides*. The scheme at the top shows the arrangemen<sup>t</sup> of *isc-suf* genes and the arrows indicate the five transcriptional start sites (TSS) identified by dRNAseq (Figure S2). Fragments used in the reporter assays to examine the promoter activity for individual promoters and promoter combinations are indicated by the arrows below. The designations for the constructs are given on the left, and the number of nucleotides upstream or downstream of the TSS that are present in the reporter plasmids, are indicated on the right. Upstream or downstream regions of different sizes are marked in orange, and the number of the nucleotides (nt) that extend the shorter DNA fragment is added to the name of the shorter construct. According to the genome annotation of *R. sphaeroides*, the *isc-suf* genes are transcribed from the minus strand. We flipped this orientation in all figures to allow direct recognition of promoter sequences or other motifs.

Previous studies addressed the role of proteins with established roles in iron-dependent regulation in *isc-suf* expression in *R. sphaeroides* [23,30,31,36,37]. Different effects of Fur/Mur, Irr, and IscR on individual *iscR* and *suf* genes suggested that not all *isc-suf* genes were under the control of a single promoter and the identical regulatory elements.

Most *R. sphaeroides* promoters share little sequence identity [37], which makes sequence-based promoter identification almost impossible. Therefore, total RNA was isolated from cultures in the exponential growth phase and used for a differential RNA-seq analysis to determine transcriptional start sites (TSS) [37]. This analysis compared RNA, which was treated with terminator exonuclease (TEX) with untreated RNA samples [44]. While TEX degraded RNAs with 5'monophosphate, which were

generated by processing, primary transcripts with 5'triphosphate were protected from degradation. Accumulation of the sequencing reads 5'of a gene in the TEX-treated sample, therefore, strongly supported the presence of a TSS at this position. As shown in Figure S2 (overview in Figure 1), the dRNA-seq read coverage indicated the presence of two promoters (P1 and P2) upstream of *iscR*. Another promoter (P3) was predicted within *iscS* and initiated transcripts spanning the *suf* genes. Furthermore, two promoters on the opposite DNA strand might lead to transcripts that were partially antisense to *isc-suf* transcripts. P4 was represented by a very low number of reads and led to a transcript mostly antisense to the *iscR* mRNA. P5 represented the promoter for transcription of RSP\_0444. In previous studies, RSP\_0444 showed only small differences in expression in the various strains or response to iron-limitation or H2O2 [23,30,31,36,37]. However, the 5'end of this transcript would partially be antisense to transcripts for the *isc-suf* genes initiating at P1 or P2. The dRNA-seq data did not hint to further promoters for *isc-suf* transcription [31]. Figure 1 shows a schematic overview of the promoter arrangement.

Many promoters in *R. sphaeroides* have a TTG around position –35 and an A at position –10/–11 relative to the TSS [37]. For the putative P1 promoter, an A was in position –11 in regard to the TSS, but no TTG around position –35 was present. The putative P2 promoter had an A residue at position –11 and a TTG around position –35 (Figure S3). Rodionov and co-workers predicted the presence of an Irr box and an IscR box upstream of the *iscR* gene [43]. The predicted IscR box spanned the positions –36 to –19 and the predicted Irr box the positions –13 to +1 in relation to the TSS at P2 (Figure S2, TSS indicated in red). Binding of IscR to the upstream region of the *iscR* gene was experimentally verified previously [31]. For the putative promoter within *iscS* (P3), an A residue at position –10 and a TTG around position –34 were present. Putative promoter P4, which seemed to be very weak based on the read number detected in RNA-seq, had a TTG around position –35, and the same was true for the putative promoter P5.

#### *3.2. Activities of the isc-suf Sense and Antisense Promoters Alone or in Combination, as Determined by Transcriptional Reporter Gene Fusions*

In order to confirm the activity of the predicted promoters, we constructed transcriptional fusions to the *lacZ* gene. Fragments containing 60–148 nt upstream and 34–63 nt downstream of the TSS, as determined by dRNA-seq, were cloned in front of the *lacZ* gene of plasmid pBBR1-MCS3-*lacZ* [34] (Table S3). An overview of the cloned fragments is shown in Figure 1, while the exact positions of primers together with predicted TSS are depicted in Figure S3. The plasmids were transferred to *R. sphaeroides* 2.4.1 wild type by conjugation, and the ß-galactosidase activity was determined for exponentially growing cultures. Since the oxygen levels influence the formation of photosynthetic complexes and consequently the demand for Fe-S clusters in *R. sphaeroides*, the cultures were either incubated under aerobic or microaerobic conditions. Under the latter conditions, the formation of photosynthetic complexes was strongly increased. The activity determined for the different promoter fusions showed marked differences (Figure 2). While P1 and P2 exhibited the weak activity of about 20–50 Miller Units (MU), the P3-*lacZ* and P5-88-*lacZ* fusions resulted in about 150–200 MU, the P4-98-*lacZ* fusion in around 400 MU under microaerobic conditions. The high activity of P4 was surprising, considering the fact that only a low number of reads from the P4 promoter was detected by RNA-seq, indicating that the native transcript initiating at P4 might be very unstable in contrast to the P4-*lacZ* fusion. This high activity, however, strongly decreased when the upstream region was reduced from 98 nt to 60 nt (P4–60). Likewise, the activity of P5 was dependent on the length of the upstream region: a shorter upstream region (88 nt) resulted in higher promoter activity than a longer upstream region (112 nt). We describe below that the OxyR protein binds to the P5 upstream region and represses P5 activity. Only for promoter P3, the activity was significantly (increase >1.5-fold and *p* < 0.01) but only slightly higher (slightly more than 1.5-fold) in aerobic conditions compared to microaerobic growth.

**Figure 2.** The activity of individual promoters and promoter combinations, as determined by *lacZ* reporter assays and quantified by measuring the ß-galactosidase activity in Miller Units (MU). Cells were cultured under aerobic or microaerobic conditions, as described in Materials and Methods. The scheme at the top shows the arrangemen<sup>t</sup> of *isc-suf* genes and the arrows indicate the five transcriptional start sites (TSS) identified by dRNAseq (Figure S2).The designations for all constructs are shown in Figure 1. The bars represent the average of technical duplicates from biological triplicates, and the standard deviation is indicated. \*: the difference between the values for iron-replete and iron deplete conditions is >1.5-fold with a *p*-value of <0.01.

Considering the arrangemen<sup>t</sup> of the different promoters on the chromosome, it is conceivable that both P1, P2, as well as P3, contribute to *isc-suf* operon transcription and that the antisense transcripts initiating at P4 and P5 may also influence *isc-suf* operon transcription. To test this hypothesis, we applied the same primers as used for the single promoter fusions to also construct reporter plasmids that harbor combinations of the different promoters (Figure 1). Our results confirmed that the presence of additional promoters could influence the expression of the *lacZ* gene, which was fused to one of the promoters (Figure 2). When P1 was present together with P2, the activity was clearly higher than for the single promoter fusions. Remarkably, also the presence of the P5 promoter (RSP\_0444 promoter) with 88 nt long upstream sequence, which generated transcripts that were antisense to the 28 nt at the 5'end of the *iscR* transcript, increased P2 activity, but not P12 activity. This activation did not occur, when 112 nt of the P5 upstream region were present (P5–112). The additional presence of the strong P4 promoter led to a strong increase of the activity compared to the P25-*lacZ* fusion, when 98 nt long upstream region was present (P254–98) but not with only 60 nt of the upstream region (P254–60).

When we used the 1777 nt fragment harboring all five promoter regions upstream of *lacZ* (P12543), the ß-galactosidase level was lower than for P254 and similar to P12 under microaerobic conditions (Figure 2). While the P12 activity was independent of the oxygen levels, the P12543 activity was slightly (1.8-fold) but significantly increased under aerobic conditions.

#### *3.3. An AntiSense Promoter Stimulates Transcription of the isc-suf Operon*

To study a possible influence of the antisense promoters P4 and P5 (promoter for RSP\_0444) on P2 (main promoter for *iscR*) activity, we used different sized downstream fragments of P2 fused to *lacZ*. As shown in Figure 2, our data indicated that both P4 and P5 might influence P2 activity. However, we also had to consider the possibility that the prolonged DNA fragments fused to *lacZ* might contain other elements, which could affect the *lacZ* transcript level. For determining the influence of P5, the P2 downstream DNA fragment fused to *lacZ* was extended by only 80 nt (P25–88) or by 104 nt (P25–112). For testing a putative additional effect by P4, the P25–112 fragment was extended by 237 nt (P254–60) or by 275 nt (P254–98) (Figure 1). To verify that different activities of the P2-reporter versus the P25and P254- reporters are really due to the promoter activity of P4 and P5, we introduced point mutations into the –35 regions of P4 and P5 promoter regions (see material and methods, TTG changed to AAA) to abolish their activity. Figure 3A shows that the point mutation in the –35 region (mutP5–88) indeed almost abolished the activity of the P5 promoter. While the presence of the wild type P5–88 sequence induced P2 activity, no effect of the mutated P5–88 promoter on the P2 activity was observed, strongly suggesting that the activity of this promoter influenced transcription of the opposite DNA strand (Figure 3A). There were two possibilities, how P5 could influence P2 promoter activity: i) through the production of an antisense transcript, or ii) through changing the local DNA topology by its promoter activity. To discriminate between these possibilities, we introduced a second plasmid (pRK4352-asP2) into the strain harboring the P2-*lacZ* fusion. This plasmid allowed the production of an antisense RNA as produced by P5 from the strong 16S promoter. Real-time RT-PCR proved that in the presence of pRK4352-asP2, the amount of the antisense RNA was about 30-fold higher than in a strain lacking this plasmid (Figure 3B). As shown in Figure 3C, the production of this antisense RNA did not affect the activity of P2. The P2 promoter used in these assays (P2 Gm) was cloned into a different plasmid with gentamicin resistance to allow overexpression of the antisense RNA as P2 from a plasmid with tetracycline resistance.

**Figure 3.** Effects of the antisense promoters P5 and P4 on the activity of P2, as determined by the *lacZ*-reporter assay (ß-galactosidase activity in Miller units). (**A**) The activity of individual promoters P5 or P2 or the combined P25 promoters. The numbers indicate the length of the DNA sequence upstream of the TSS, as shown in Figure 1. "mut" indicates that the TTG at position –35 of promoter P5 was changed to AAA. (**B**) Change of the level of RNA antisense to P2 in a wild type strain harboring the P2 reporter plasmid and plasmid pRK4352-asP2 (overexpression of the antisense RNA) compared to a strain just harboring the P2 promoter. The RNA level was determined by real-time RT-PCR, and the bar represents the average level of technical triplicates from biological triplicates with standard deviation. (**C**) Effect of elevated levels of RNA antisense to P2 on the activity of P5, as determined by the *lacZ*-reporter assay. (Gm) indicates that these reporter constructs carry a gentamycin resistance (all other reporters carry tetracycline resistance) to allow selection of the overexpressed plasmid pRK4352-asP2 (tetracycline resistance). (**D**) The activity of individual promoter P4 or the combined P25 and P254 promoters. The numbers indicate the length of the DNA sequence upstream of the TSS. "mut" indicates that the TTG at position –35 of promoter P4 was changed to AAA.

We applied the same strategy to test the effect of P4 on P2 activity. However, changing the TTG at the putative –35 of P4 only slightly decreased ß-galactosidase activity in the P254–98 construct. Consequently, the resulting construct mutP254–98 would still increase the activity of the P25 fusion (Figure 3D). As already shown in Figure 2, the P4–60 construct had a strongly reduced activity compared to P4–98. The low activity of P4-60 was almost abolished, when the –35 region was mutated. This indicated that the sequences in the –98 to –60 upstream region of the P4 promoter were responsible for the strong activity of P4–98 and the stimulating effect of P4–98 on P25. We were unable to recognize a particular motif that could cause this effect.

#### *3.4. E*ff*ect of Oxidative Stress and Iron Availability on the Activity of the Promoters of the isc-suf Operon*

Previous studies on global gene expression by microarrays or RNA-seq revealed that expression of the *isc-suf* genes is affected by oxidative stress [41,45] and by iron availability [20,23]. In this study, we applied the different reporter fusions to test which promoters are affected by these external factors. We tested hydrogen peroxide and tertiary butyl alcohol (tBOOH) as oxidative stress agents. tBOOH represents organic peroxides, which are generated in the cell, e.g., upon singlet oxygen exposure, and thus represent the photo-oxidative stress that *R. sphaeroides* faces in the presence of pigments, oxygen, and light. None of the promoters P1, P2, P3, P4, P5, or any combination of promoters showed marked changes in activity upon addition of hydrogen peroxide (same values for t0 as shown in Figure 4A and no significant changes at later time points), although hydrogen peroxide was previously shown to strongly induce the *isc-suf* mRNA levels [36]. We had, however, noted before that the response of *lacZ* reporters to H2O2 might be weak, possibly because of a negative effect on the enzymatic activity [46], while tBOOH was shown to induce the activity of certain *lacZ* fusions [47]. The addition of tBOOH resulted in significantly increased activities of the P2 promoter, and, also, the combination P25–88 showed significant tBOOH-dependent expression (all other promoter fusions did not show altered activity in response to tBOOH). The changes in activity were, however, less than 2-fold (Figure 4A). P12543 was also activated by tBOOH, but the increase was only 1.5-fold.

A previous study demonstrated that the *iscR* transcript level increases upon iron depletion [31]. A strong increase (about 5-fold) of P2 promoter activity upon iron depletion was confirmed in this study and was also observed for all other fusions containing P2, including the long fusion extending to P3 (Figure 4B). No significant effect of iron on activities of P1, P3, P4, or P5 was observed in the wild type (Figure 4B).

#### *3.5. Protein Regulators of the isc-suf Operon*

Previous bioinformatic analysis predicted IscR and Irr binding sites upstream of the *iscR* gene [43] (Figure S2), and the binding of IscR to this region was experimentally verified [31]. The role of Irr in the expression of the *suf* genes was supported by microarray analysis [30], but binding of the Irr protein to the *iscR* promoter region of *R. sphaer*oides was not reported. Microarray analyses also revealed a stronger effect of iron depletion on *isc-suf* expression in a mutant lacking the Fur/Mur regulator [23] and an effect of OxyR on *isc-suf* expression [18,20]. We tested the activity of the individual *isc-suf* promoters in different mutant strains to ge<sup>t</sup> more insights into the influence of these regulatory proteins on the *isc-suf* promoters.

Only the mutant strains, lacking the IscR or Irr protein, showed the altered activity of the P1 promoter (Figure 5A). The activity of P1 was almost 3-fold lower in the *irr* mutant than in the wild type under iron repletion and 1.6-fold lower under iron depletion. A sequence (TAGAAGGCATAGTGC) with similarity to the consensus Irr-box (Figure S2) was present directly upstream of the TSS of P1. However, we could not confirm binding of Irr to the P1 promoter region in vitro, while in a parallel control assay binding to the *mbf*A promoter, as described in [30], could be observed and confirmed that the isolated Irr protein was able to bind to one of its known targets (Figure S4A). Activity in the *iscR* mutant was similar to that of the wild type under iron repletion but about 2-fold higher under iron depletion. A sequence with some similarity to the *iscR* box (Figure S2) was present at the

transcriptional start site (TAGACGACCTTGTTGTT, Figure S3). Indeed, gel retardation revealed that IscR showed specific interaction to the P1 promoter region (Figure 6A). Increasing amounts of IscR protein shifted the P1 containing DNA fragment, while the addition of unlabeled, specific competitor released the shift.

**Figure 4.** The activity of individual promoters and promoter combinations, as determined by *lacZ* reporter assays and quantified by measuring the ß-galactosidase activity in Miller Units (MU). (**A**) The ß-galactosidase activity was measured before, 7 min, and 30 min after the addition of tBOOH (100 μM final concentration). (**B**) The ß-galactosidase activity was determined under iron repletion and iron depletion. The designations for all constructs are shown in Figure 1. The bars represent the average of technical duplicates from biological triplicates, and the standard deviation is indicated. \*: the difference is >1.5-fold with a *p*-value of <0.01.

**Figure 5.** The activity of the individual promoters and P12543, as determined by *lacZ* reporter assays and quantified by measuring the ß-galactosidase activity in Miller Units (MU) in the wild type and mutant strains under iron repletion or iron depletion. (**A**) The activity of promoter P1, (**B**) The activity of promoter P2, (**C**) The activity of promoter P3, (**D**) The activity of promoter P4, (**E**) The activity of promoter P5, (**F**) The activity of P12543. The bars represent the average of technical duplicates from biological triplicates, and the standard deviation is indicated. \*: the difference is >1.5-fold with a *p*-value of <0.01.

**Figure 6.** Electrophoretic mobility shift assays showing the interaction of (**A**) IscR to the P1 promoter region (199 bp fragment) and (**B**) IscR to the P3 promoter region (190 bp fragment) and (**C**) Irr to the P3 promoter region (190 bp fragment) and (**D**) oxidized OxyR to the P5 promoter region (147 bp fragment). The star labels the radio-labeled input DNA fragment, and the arrow points to the shifted bands of the DNA protein complexes. The amount of the protein input is given for each lane, as well as the molar ratio of specific, unlabeled competitor DNA (same DNA fragment but without a label).

The activity of P2 was strongly influenced by IscR under iron repletion and iron depletion, as demonstrated previously [31] (Figure 5B). The activity of P2 was also elevated in the presence of tBOOH in iron-replete medium (Figure 4A). This response of P2 to tBOOH was lost in the *iscR* mutant (Figure S5). Furthermore, we observed small but significant effects of OxyR and Irr on P2 activity: lack of OxyR increased P2 activity about 1.5-fold under iron depletion, the lack of Irr decreased P2 activity about 2-fold under iron repletion. Binding of IscR to P2, as well as regulation of P2 by IscR, was demonstrated before [31]. An Irr binding site was predicted for the P2 promoter region [43]. As for P1, we could not confirm binding of Irr to the P2 promoter region, while in a parallel control assay, binding to the *mbf*A promoter, as described in [30], could be observed (Figure S4B).

P3 activity was affected by IscR, both under iron repletion and iron depletion (Figure 5C). Lack of IscR resulted in higher activity (2.8-fold) in iron depletion, as well as in iron repletion (2.5-fold), indicating a repressing effect of IscR under both conditions. In the wild type, P3 showed no significant response to iron, while we observed slightly higher activity (1.5-fold) under iron depletion in the strain lacking IscR. Inspection of the sequence around the predicted TSS of P3 revealed a motif (GACTATTTCTGTCGG) with similarity to the consensus IscR box (Figure S2). Indeed, IscR showed specific binding to a DNA fragment containing the predicted binding site in a gel retardation assay (Figure 6B). Furthermore, the activity of the P3 promoter was significantly reduced in the Fur/Mur mutant under iron depletion (Figure 5C). Only in the strain lacking Irr, we had a significantly increased P3 activity (2-fold increase) in iron depletion compared to iron repletion. This agreed with our previous microarray data: the iron dependency of *suf* transcript levels is more pronounced in a strain lacking Irr than in the wild type. A putative Irr binding site (TTAGAAATATTCTAGA) was present about 50 nt upstream of the TSS of P3, which is a similar distance to the TSS as observed for the confirmed Irr target *ccpA* [30]. A gel retardation assay confirmed the specific binding of Irr to this DNA region (Figure 6C).

Di fferences in activity between wild type and the tested mutants for antisense promoter P4 were small (≤1.5-fold) and/or statistically not significant for the reporter construct with 98 nt upstream of the TSS (Figure 5D) and also for the construct with only 60 nt of the upstream region.

The activity of antisense promoter P5 was strongly a ffected by the OxyR protein when 112 nt upstream of the TSS was present. The activity in the mutant was increased by a factor of 3-4 under both iron repletion and iron depletion (Figure 5E). OxyR did not a ffect P5 activity when only 88 nt of the upstream sequence was present. None of the other tested mutants showed significantly altered P5 activity. Despite the strong di fferences in the activity of P5 in the presence or absence of OxyR, P5 did not show a response to high oxygen levels (Figure 2) or tBOOH (Figure 4A). We tested whether OxyR was able to bind to the P5 promoter by the gel retardation assay. As shown in Figure 6D, the addition of increasing amounts of oxidized OxyR resulted in retardation of a 147 nt DNA fragment harboring the P5 promoter region fragment. The same result was achieved with reduced OxyR protein. The addition of an excess of the same, but un-labeled DNA fragment led to a decrease of retardation due to specific competition with the labeled fragment for binding. This result strongly supported that OxyR was indeed binding to the P5 promoter region.

We also tested the activity of the P12543 fusion in di fferent mutant strains (Figure 5F) to see how the action of the regulators on promoters P1, P2, P4, or P5 would influence P3 expression. Fur/Mur and the RirA proteins did not influence P12543 activity, as this was also the case for P3 alone. Since the P12543 fusion comprises an intact copy of the *iscR* gene and may, therefore, increase IscR levels in the cell, it could not be tested in conditions lacking IscR. The repressing e ffect of Irr under iron depletion was more pronounced when all promoters were present than in the P3-fusion alone (compare Figure 5C–F).

#### *3.6. The RirA Proteins of R. sphaeroides Have No E*ff*ect on isc-suf Expression in R. sphaeroides*

While Fur is the dominant iron regulator in gamma-proteobacteria, other proteins have important roles in iron-dependent regulation in alpha-proteobacteria [43]. Besides IscR, another transcriptional regulator of the Rrf2 protein family, RirA, was identified to have an important role in iron regulation in *Rhizobia* [48,49]. The genes RSP\_2888 and RSP\_3341 of *R. sphaeroides* 2.4.1 have 59%–63% identity to RirA from *Rhizobium leguminosarum* or *Agrobacterium tumefaciens* (*Rhizobium radiobacter*). We constructed knock out strains of *R. sphaeroides* lacking either RSP\_2888 or RSP\_3341 or both genes together. As seen in Figure 5A–F and Figure S6, the promoters of the *isc-suf* operon showed very similar activities in the wild type and the double mutant, and also the e ffect of iron depletion on promoter activity was very similar in both strains. We concluded that the RirA homologs of *R. sphaeroides* had no major role in the iron-dependent regulation of the *isc-suf* operon. The double mutant showed identical growth curves as the wild type in iron repletion and iron depletion (Figure S7), indicating that the RirA proteins have also no major impact on iron regulation in *R. sphaeroides* in general.
