**4. Discussion**

The assembly of iron-sulfur clusters is an important cellular process in almost all living organisms, and stress conditions that destroy iron-sulfur clusters increase the need for iron-sulfur cluster assembly. The facultative phototrophic bacterium *R. sphaeroides* harbors a large operon that comprises *isc* and *suf* genes, which are distributed to di fferent operons in *E. coli*. Unlike *E. coli*, *R. sphaeroides* can form photosynthetic complexes that cause an additional demand for the Fe-S cluster and can also cause photooxidative stress that destroys the Fe-S cluster. The synteny of *isc* and *suf* genes is conserved among *Rhodobacteraceae* independently of the ability to perform photosynthesis and is also similar in *Rhizobiales*. This suggests that the combined *isc-suf* operons arose early in evolution, in a common ancestor of these orders.

Previous work revealed iron- and hydrogen peroxide-dependent expression of the *isc-suf* genes of *R. sphaeroides* and the role of OxyR, Fur/Mur, Irr, and IscR regulators in *isc-suf* expression [20,23,30,31]. To verify promoter activities as predicted by dRNA-seq (Figure S2) and to understand the contribution of the individual promoters, we analyzed activities of individual promoters and of promoter combinations and the effect of iron, oxygen, and oxidative stress on the activities by using reporter gene fusions. These data were combined in a model (Figure 7) that visualizes the complex regulatory network controlling *isc-suf* expression in *R. sphaeroides*.

**Figure 7.** A schematic model, combining the influence of other promoters on P2 and the action of different proteins on the activity of the *isc-suf* promoters. + indicates a stimulating effect of P1, P4, and P5 promoter regions (dependent on the length of the upstream region) on P2 activity. The IscR protein binds to IscR boxes (yellow bars) in the promoter regions of P1, P2, and P3, and OxyR binds upstream of the P5 promoter (indicated by solid arrows), and Irr binds to an Irr-box (green bar) in the P3 promoter region. At P2, IscR mediates the response to iron and tBOOH. The iron-dependent activity of P3 is only observed in the absence of IscR or the absence of Irr. Green arrows indicate activation by the protein regulators, and black arrows indicate repression. The small effects of Irr on P1 and P2 and of Fur/Mur on P3 are most likely indirect and do not involve direct binding (indicated by dashed arrows).

Our data supported the view that both P1 and P2 contributed to *isc-suf* expression, and P3 further contributed to transcription of the *suf* genes. P3 was the strongest of the sense promoters, and extending the upstream region of P3 and including P1 and P2 elevated the activity further and conferred iron-dependent expression. An additional promoter for the *suf* genes (P3) might guarantee the high expression of the proteins required for the assembly of iron-sulfur clusters. IscR has a regulatory function and may not be required in a high amount. IscS is a cysteine desulfurase that is required for mobilization of sulfur from L-cysteine. A higher expression of *iscS* might not be required, since, with SufS, another cysteine desulfurase was encoded by the *isc-suf* operon downstream of P3. This arrangemen<sup>t</sup> with two cysteine desulfurases was conserved among *Rhodobacteraceae* and also found in *Rhizobiaceae* (Figure S1).

Two antisense promoters were present downstream of P2. P4 was located within the *iscR* gene, while P5 was located upstream of *iscR* and responsible for transcription of RSP\_0444. The gene product of RSP\_0444 is annotated as a putative hydrolase, but no experimental data on its function is available. The position of this gene on the chromosome was also found in most *Rhodobacteraceae* and *Rhizobiaceae*, although in *Rhizobiaceae, iscR* was not located upstream of IscS (Figure S1). Surprisingly, both antisense promoter regions, P5 and P4, positively affected transcription in sense direction. Transcription of *iscR* from P2 was higher, when, at the same time, P5 was present, and even higher, when, also, the region containing P4 was present. RNA-seq analyses revealed the presence of remarkably high numbers of antisense promoters in many bacterial species, and different effects of antisense transcripts have been reported [50–52]. In many cases, antisense transcription produces non-coding sRNAs that either interfere with translation or influence stability of the sense RNA. We could not confirm the existence of a distinct, small RNA originating at P4 by Northern blot analysis. Alternatively, the formation of an open complex during initiation of transcription might also allow more efficient transcription in the opposite direction. An effect of promoters on superhelicity-dependent processes is well documented [53]. The exact mechanisms by which P5 and the P4 region affect *isc-suf* expression remain elusive, but our data emphasized that it is important to consider the e ffect of such antisense promoters for sense promoter activity.

P2 was the main promoter for transcription of *iscRS*, and binding of IscR to P2 conferred iron-dependent expression to P2 and to the downstream promoter P3 that initiated further *suf* transcripts. Our present study revealed that IscR was also required for the tBOOH-dependent activity of P2. Thus, IscR could function as a sensor for iron availability and organic peroxide stress in *R. sphaeroides*. Furthermore, we could demonstrate the binding of IscR to the other sense promoters—P1 and P3. In contrast to its e ffect on P2 activity, IscR binding to P1 or P3 did not mediate tBOOH-dependent expression. Two types of IscR binding sites—Type 1 and Type 2—were identified in *E. coli*: holo-IscR (containing the Fe-S cluster) shows a higher a ffinity to Type 1 sites than apo-IscR, while both IscR forms bind with similar a ffinity to Type 2 sites [9]. Interestingly, this regulatory mechanism is even conserved in the Gram-positive *Thermincola potens* [54]. Our in vivo data revealed stronger repression of the P2 promoter by holo-IscR, while the repressing e ffect on P3 was similar in iron repletion and iron depletion, and only apo-IscR had a repressing e ffect on P1. This indicated the presence of di fferent types of IscR binding sites also in *R. sphaeroides*.

Despite the presence of an Irr box around the P2 promoter, we detected only a small e ffect of Irr on P2 (1.8-fold) and did not detect binding of Irr to the P2 region in vitro. A small e ffect of Irr on P1 activity (1.5-fold) was observed, but no binding of Irr to the promoter region in vitro was observed. Our results implied that Irr did not make a major contribution to P1 and P2 regulation under the tested conditions and suggested that the influence of Irr might be indirect. There was also an influence of Irr on P3 that was even more pronounced when all upstream promoters were present, and binding of Irr to P3 was demonstrated. Since the P12543 reporter also carries a complete copy of *isc*R, we could not exclude that elevated levels of IscR influence activity of the P3 promoter in this strain. Higher IscR levels could increase the e ffect of iron in the *irr* deletion strain (compare 5C to 5F). The P3 promoter was the only promoter that showed a significant but small e ffect in the mutant lacking Fur/Mur, indicating an activating function of Fur/Mur under iron depletion. Johnston and co-workers suggested that iron-dependent regulation in alpha-proteobacteria mainly occurs by regulators di fferent from Fur [55]. Some Fur homologs in *Rhizobia* and the Fur homolog of *R. sphaeroides* were shown to a ffect the expression of the *sit* operon in an Mn2<sup>+</sup>-dependent manner and were consequently named Mur [23,56–58]. The deletion of the *fur*/*mur* gene resulted in stronger expression of many iron-dependent genes in *R. sphaeroides* [23], suggesting that Fur/Mur has a role in regulating iron metabolism in this bacterium and has a repressing function under iron depletion. Rodionov and co-workers suggested a Fur-box and a slightly di ffering Mur-box as consensus sequences for DNA binding sites in alpha-proteobacteria [43]. Such sequences are not present in the vicinity of P3 of the *isc-suf* operon, while an almost perfect Mur box is located between the –10 and –35 region of the *sitA* promoter. Fur/Mur is required for a strong induction (about 50-fold) of *sitA* expression in response to Mn2<sup>+</sup> limitation [23]. It is likely that the small e ffects of Fur/Mur on many genes of iron metabolism, including P3 of the *isc-suf* operon, does not include direct binding but is rather mediated indirectly.

While none of the sense promoters was influenced by the OxyR protein, antisense promoter P5 was strongly repressed by this regulator, under iron-replete and iron deplete conditions, and binding of OxyR to the P5 upstream region was confirmed. The repressing e ffect of OxyR on the P5 promoter required 112 nt of the upstream region and was not present with only 88 nt of the upstream region (Figure 5E). An e ffect of OxyR on the iron-dependent levels of *isc* and *suf* genes was reported previously [20,41,45]. Since the activity of P12543 was not influenced by OxyR (Figure 5F), it was likely that other mechanisms than binding to P5 were responsible for the OxyR e ffect on the *suf* genes, which might also be indirect. Since P5 is the promoter for RSP\_0444, OxyR could also a ffect the expression of RSP\_0444, but previous microarray studies revealed hydrogen peroxide- and OxyR-independent expression of RSP\_0444 [41,45]. OxyR is mostly known as an activator of gene expression in response to oxidative stress. A repressor function of OxyR under non-stress conditions, as observed for P5, was, however, also described in *R. sphaeroides* [41].
