The AadR-FixK hierarchy coordinates iron-responsive metabolism via Fur-family regulators in Rhodopseudomonas palustris TIE-1

Abstract

Rhodopseudomonas palustris TIE-1 (TIE-1) is a metabolically versatile environmental bacterium that flourishes across gradients of iron, oxygen, and light. This versatility necessitates extensive regulatory control, exemplified by the aerobic-anaerobic metabolic shift controlled by the hierarchy of CRP/FNR-family regulators AadR and FixK. Many anaerobic metabolic pathways demand expression of iron cofactor-intensive proteins, and TIE-1 in particular can generate energy through phototrophic iron oxidation via the PioABC system. However, TIE-1 lacks canonical iron-sensing regulators: IscR, ancestral Fe(II)-sensing Fur, and Fe(II)-sensing RirA of Rhizobiaceae, leaving it unclear how TIE-1 coordinates expression of these iron-requiring metabolisms with bioavailable iron levels. Here, we demonstrate that the AadR-FixK hierarchy plays a previously underappreciated role in iron regulation in TIE-1 by comparing growth and transcription in wild-type and regulatory mutants across wetland-inspired naturomimetic conditions. ΔaadR and ΔfixK showed defects in iron-dependent growth and Fe(II) oxidation, and the ΔaadRΔfixK double mutant was synthetically lethal under anaerobiosis. The regulatory hierarchy of FixK and AadR influences expression of Fur-family regulators: the two irr paralogs were oppositely regulated in the presence of AadR, and absence of AadR perturbed iron-responsive expression of mur. Furthermore, the AadR regulon was significantly enriched for iron-related and iron-containing proteins. Despite initial predictions that AadR directly regulates pioABC, we found no conclusive evidence for direct AadR activity at the pioABC promoter, refining the search for pio regulators. Together, these findings establish AadR as a central integrator of oxygen and iron signals to coordinate iron-requiring anaerobic metabolism in TIE-1.

3. Introduction

Rhodopseudomonas palustris, a ubiquitous and free-living phototrophic purple non-sulfur bacterium, is renowned for its metabolic versatility across varying environments (1). R. palustris inhabits wetland sediments and water columns (2, 3), microbial mats (4), and wastewaters (5), where it is exposed to a variety of environmental gradients. As a “metabolic opportunist” (1), it uses various sources of carbon, electrons, and energy. It grows aerobically or anaerobically, and along the microaerobic transition, and it is capable of metabolisms including photoheterotrophy, photoautotrophy, chemoheterotrophy, nitrogen fixation, and degradation of aromatic organics such as monolignols (1, 6, 7). One metabolism of interest for biotechnological applications is photoferrotrophy: a form of photoautotrophy involving fixation of carbon dioxide, energy from light, and electrons for reducing power derived from the oxidation of reduced iron (Fe(II)) (8–10). This process is facilitated by the PioABC protein complex for the phototrophic oxidation of iron (4, 9–11). While many pioABC-encoding strains can oxidize Fe(II), the ability to grow photoferrotrophically is not equally widespread (9, 10, 12).

Deploying anaerobic metabolic pathways requires expression of iron-requiring enzymes: nitrogenase and anaerobic benzoyl-CoA reductase, for example, each require more than ten iron atoms per holoenzyme (13). Cytochromes and ferredoxins used in the photosystems and electron transport chain are also expensive to the cell’s iron stores (14), and PioABC includes both a high-potential iron-sulfur protein and a decaheme cytochrome (10, 11, 15, 16). Expression of these various “iron-requiring” pathways must be controlled in coordination with differing levels of bioavailable iron and other metabolic resources to maintain homeostasis and growth. Many narrow, pathway-level regulators have been characterized (17, 18), but how R. palustris coordinates its response to Fe(II) more broadly remains poorly understood.

One key dynamic that broadly affects the sophisticated regulatory landscape of R. palustris is the aerobic-to-anaerobic regulatory hierarchy centered on the CRP/FNR-family transcriptional regulators FixK and AadR (19). The FixL-J two-component sensor activates expression of fixK as oxygen decreases, at which point FixK activates expression at numerous other loci (19). One gene activated by FixK is its own paralog aadR, which represses further fixK transcription to complete the shift to the anaerobic regulatory program (19). Precise regulons for the two regulators were not successfully determined, primarily owing to their similarity, but Rey and Harwood (2010) identified dozens of potential sites regulated by the pair. They also observed that AadR’s sphere of influence extends well beyond its originally characterized role in anaerobic aromatic degradation (19–21). Compared with a ΔaadR strain, wild-type R. palustris CGA009 repressed >20 ferric iron transport genes, activated a putative Fur-family iron-response regulator, and activated a locus later identified as the fez operon, encoding the iron-storage ferrosome (19, 22). Work in R. palustris TIE-1 (TIE-1) showed that FixK directly binds the promoter of pioABC, and that pio expression is activated by FixK only during growth with Fe(II). Together, these findings imply an important role for the AadR-FixK hierarchy in broader iron-responsive regulation, prompting us to study them in this context.

Iron-responsive gene regulation in α-proteobacteria has undergone substantial lineage-specific remodeling (15, 23–28). In taxonomic order Rhizobiales, the ancestral Fur (ferric uptake regulator) has functionally diverged to become the manganese uptake regulator Mur, while IscR (Iron-sulfur cluster regulator) has been lost entirely (15, 28–31). In the Rhizobiaceae family, the acquisition of the regulator RirA (Rhizobiaceae iron regulator) has filled the resulting regulatory gap as a direct Fe(II)–binding regulator (28, 29). The sister family Bradyrhizobiaceae, including R. palustris, lacks RirA entirely (29). They are thought instead to manage their iron-responsive regulation exclusively through Irr, which has predicted binding sites at many loci that the missing RirA regulates (15, 24, 27). Irr activates genes involved with Fe³⁺ uptake when iron is limited, while binding heme iron deactivates the Irr protein (30, 32). Irr has undergone at least one discrete duplication event in this lineage, leading most Bradyrhizobiaceae to possess two paralogs (Fig. S1) (15, 27). R. palustris IrrA2 retains the conserved residues for heme sensing, while IrrA1 lacks the motif, leaving its role less certain (27). The gene irrA1 shares an upstream intergenic region with the fez operon, has predicted tandem FNR sites in its promoter (33), and appears to be strongly activated by both FixK and AadR under anaerobic conditions (9, 19, 22, 27, 29). Without a dedicated Fe(II)–sensing Fur or RirA, and with no IscR to regulate [Fe-S] cluster-dependent pathways, how R. palustris coordinates iron-responsive gene expression in response to Fe(II) remains an open question.

The pioABC operon provides a compelling case study for this question. The homologous mtrCAB in Shewanella oneidensis is regulated by Fur (34, 35), but no Fur-family regulators have yet been implicated in pioABC regulation. The pioABC promoter (Ppio) instead contains two confirmed transcription factor binding sites with palindromic AT-rich motifs of the type bound by CRP/FNR-family regulators, henceforth termed Binding Site 1 (BS1) and 2 (BS2) (9). Ppio-BS1 lies at −44.5 relative to the transcription start site and is bound by FixK protein in vitro, while Ppio-BS2 lies upstream of it (at −137.5 relative to TSS) and is not bound by FixK (Fig. S2) (9). Deletion of BS2 reduced expression by 60–80% and abolished the iron-responsiveness of regulation by the promoter, suggesting that the unknown BS2-binding regulator is either iron-responsive or under the control of an iron-responsive regulator (9). AadR and FixK are paralogs with near-indistinguishable binding motifs (Table ST1) (19, 33, 36) which hierarchically regulate many shared loci, making AadR a rational candidate for a possible pio regulator. In this work, we compared growth and transcriptional regulation in wild-type (WT) TIE-1 and in regulator deletion (Δ) mutants. We performed experiments under “naturomimetic” (nature-mimicking) growth conditions: the lighting, carbon sources, and iron availability were carefully tuned to better reflect the conditions experienced by R. palustris in the field. We show that AadR coordinates iron-responsive gene regulation as an additional hierarchical level atop the Fur-family regulators, and that the list of pathways differentially regulated by its absence is significantly enriched for genes encoding both iron-requiring and iron-related proteins. The effects of AadR on pioABC expression are indirect— narrowing the search for the unknown Ppio-BS2-binding factor.

4. Results

4.1. Phenotypes during naturomimetic growth reveal oxygen- and iron-dependent defects in ΔaadR and ΔfixK strains, including anaerobic lethality in the ΔaadRΔfixK double mutant

For a real-world understanding of microbial growth under the diversity of natural conditions, we analyzed how the variables light, oxygen, and iron could best represent nature in what we term naturomimetic conditions (SI 1). The attenuated and yellow-shifted light available in a wetland water column or sediment is better mimicked by fluorescent lighting than it is by the incandescent lamps used for phototrophic growth of R. palustris in the laboratory (SI 1.1, Fig. S3) (37, 38).

No significant growth defect was observed for any mutant strain under aerobic conditions (YPS, Fig. S4). While both ΔaadR and ΔfixK were capable of anaerobic growth, the ΔaadRΔfixK double mutant was synthetically lethal under all tested anaerobic conditions (Figs. S5–S8). During photoheterotrophy with basal iron levels (−Fe), ΔfixK showed a growth rate defect as expected based on past work (Table ST2, Fig. S9) (9). The TIE-1 ΔaadR strain showed the expected defect on benzoate but, surprisingly, also showed a significant defect on succinate photoheterotrophy, contrary to what was previously reported in CGA009 (Figs. S6–S7) (19, 20). Knocking out either fixK or aadR significantly reduced expression of light-harvesting genes under our conditions: for example, two pufB paralogs were expressed 2.8- and 5.6-fold higher in wild-type than ΔaadR (q<10⁻⁵) in freshwater acetate. This suggests that the growth defects caused by these mutations may be exacerbated by the reduced photon flux of naturomimetic lighting compared with the conventional high-intensity incandescent illumination used in past work (SI 1.1, Fig. S9) (9, 19).

Iron-supplemented (“+Fe”) conditions contained 5 mM FeCl₂ with equimolar nitrilotriacetic acid (NTA), emulating the millimolar-scale Fe(II) found in groundwater-fed wetland sediments (39–41). Basal iron (“−Fe”) conditions contained ~4 μM Fe (as FeSO₄) added as part of the trace element solution to the media (SI 1.2) (42). Two of the growth conditions we categorize as +Fe mixotrophy (i.e., acetate or malate as the organic carbon source and Fe(II) supplemented to +Fe levels). R. palustris typically consumes most carbon sources (e.g., acetate (43, 44)) before it begins performing measurable iron oxidation, allowing us to harvest RNA during the photoheterotrophic phase of mixotrophic growth (0.5 < OD₆₆₀ < 1.0 at harvest) (SI 1.3). As this was before iron oxidation commenced, the transcriptomic differences reported for mixotrophy in this study directly reflect the response of TIE-1 to supplemental Fe(II), rather than a more general response to autotrophic metabolism, thereby avoiding the confounding limitation inherent in previous methods that directly compared transcriptomes of photoheterotrophy and photoferrotrophy.

+Fe mixotrophy revealed substantially more severe phenotypes than were displayed under −Fe growth. ΔaadR showed a significant growth defect in +Fe mixotrophy on both acetate +Fe and malate +Fe, despite growing comparably to wild-type on both carbon sources with basal iron levels (−Fe) (Fig. 1A, B, C). ΔaadR also oxidized significantly less Fe(II) than wild-type, with less than 1 mM oxidized over the course of the experiment during growth on either carbon source, compared to ~2.5 mM oxidized by WT on acetate and ~4.5 mM on malate (Fig. 1D). ΔfixK failed to grow at all under +Fe conditions (Fig. 1E, S5, S8), contrasting the delayed but successful photoferrotrophy it exhibited under high-intensity incandescent light (9). Neither single mutant grew on photoferrotrophy (FW+Fe) (Fig. 1E, S8). The iron-dependent growth defects, particularly the conditional ΔaadR defect (absent under −Fe but pronounced under +Fe, Fig. 1C), motivated our transcriptomic investigation of AadR’s iron-related regulatory role.

Figure 1: AadR and FixK are important for growth of TIE-1 under iron-supplemented conditions.

Growth curves of wild-type and ΔaadR TIE-1 on FWAcetate (A.) or FWMalate (B.). ΔaadR showed a slight defect in lag time, but no significant defect in doubling time relative to wild-type under photoheterotrophy with basal iron (−Fe) on either carbon source. Growth curves for each strain on each carbon source are also reported for supplemental iron (+Fe, +5 mM) conditions. (C.) Doubling time for each strain on each carbon source under either −Fe or +Fe concentrations. CLD Letters represent the results of a two-way ANOVA (Tukey-adjusted) on calculated doubling time. (D.) Oxidation curves for the +Fe growth experiments, measured via the ferrozine assay (see Methods) and normalized to the abiotic control. Fe oxidation curves were calculated as asymmetric sigmoidal 5-parameter lines fit to the data (Table ST3). Iron oxidation rate and final amount oxidized were both significantly decreased in ΔaadR relative to wild-type. For wild-type, significantly less iron was oxidized by the end of growth with acetate compared to with malate, though the rate at which it was oxidized did not significantly differ. (E.) Growth curve for FW+Fe(II) photoferrotrophy, with a FW-Hydrogen photoautotrophic pre-growth. The ΔaadRΔfixK double mutant did not grow on the pre-growth (“NG” = no growth), so its performance on photoferrotrophy was not determined (“ND” = not determined). While ΔaadR and ΔfixK grew in the pre-growth, neither began iron-oxidizing photoferrotrophic growth.

4.2. Loss of AadR disrupts iron-responsive expression of Fur-family transcriptional regulators

In wild-type TIE-1, transcription of fixK was significantly repressed with supplemental Fe(II) (log₂FC = −0.73, q = 0.002) while aadR was significantly activated (log₂FC = +1.40, q< 10⁻¹⁵) (Fig. 2A). This pattern held true for both of the +Fe conditions (FWAc+Fe and FW+Fe), demonstrating that it is a response to iron rather than an effect of autotrophic metabolism. Repression of fixK by iron was also replicated in the independent Guzman et al. (2019) dataset (Fig. 2A) (45). Critically, this iron-mediated repression of fixK was abolished in ΔaadR (log₂FC = +0.20, not significant) (Fig. 2B), implicating AadR’s involvement in iron-responsive modulation of fixK expression, and supporting the possibility that this repression occurs through the established mechanism of AadR’s direct repression of fixK (19).

Figure 2. Iron levels in the growth medium mediate the transcriptional regulation of aadR— and by extension, putative and confirmed AadR-FixK hierarchy regulatory targets.

(A.) Wild-type TIE-1 on photoferrotrophy vs photoheterotrophy. Photoferrotrophy induces repression of fixK and activation of aadR. Log₂-adjusted fold change (log₂FC) in each transcript between +Fe and −Fe conditions for selected transcriptional regulators. *Alongside this study’s RNA-seq data, expression data from the same conditions released by Guzman et al. (2019) are presented (hatched bars) (45). (B.) Wild-type TIE-1 and ΔaadR, during photoheterotrophy (+Fe vs −Fe). Fur-family and FNR-family regulators show differential regulation between +Fe vs −Fe in WT and ΔaadR. Log₂FC in transcript between FWAc +Fe and −Fe, in wild-type (solid bars) and ΔaadR (hatched bars). Results of a two-way ANOVA comparing the Fe-driven change in a transcript between WT and ΔaadR are shown above each pair. (C.) Wild-type TIE-1 vs ΔaadR, across several growth conditions. Deletion of aadR leads to opposite expression patterns in the two irrA paralogs encoded in TIE-1’s genome, labeled here as irrA1 and irrA2, under anaerobic growth. (D.) Wild-type TIE-1 during photoheterotrophy (supplemental iron (+Fe) vs basal iron (−Fe)). The trend between fixK and aadR expression in (A) is replicated here, and fur-family regulators also show strong differential regulation under +Fe vs −Fe in WT TIE-1. Note the inverted direction of the change in mur expression in response to iron.

Three of the four Fur-family transcription factors were differentially expressed in ΔaadR relative to wild-type under anaerobic conditions (Fig. 2B). Expression of irrA1 was significantly greater in wild-type than ΔaadR, while irrA2 showed the opposite pattern, and this relationship held true across all anaerobic growth conditions we tested (Fig. 2C). In wild-type, irrA2 showed a striking 28-fold increase in expression under photoferrotrophy relative to photoheterotrophy (log₂FC = +4.58, q = 3.9×10⁻¹⁷⁰), far outstripping its response to +Fe during heterotrophic (mixotrophic) growth (Fig. 2B, 2D). This suggests that irrA2 responds to the specific metabolic context of photoferrotrophy, rather than to iron concentration alone. Contrasting the lack of an AadR-mediated iron response in irr, iron-responsive differential expression of mur was AadR-dependent. In wild-type, mur was activated under FWAc+Fe mixotrophy (log₂FC = +0.76, q = 3.2×10⁻⁵) (Fig. 2B), but in ΔaadR this response was inverted: mur was significantly repressed in ΔaadR under the same condition (log₂FC = −0.64, q = 0.017) (Fig. 2B). Photoferrotrophically, mur in wild-type (Fig. 2A) showed the same trend as in ΔaadR under FWAc+Fe (Fig. 2B). This finding establishes that mur is repressed specifically under photoferrotrophic conditions, or in absence of AadR, further implicating AadR as important for the activation of mur transcription under mixotrophic +Fe growth. Activation of zur by +Fe occurred in wild-type, but this was not significantly attenuated in ΔaadR (log₂FC = +0.60 vs. +1.42, ANOVA = 0.6299, p_adj = 0.1386). Whether direct or mediated by intermediaries such as FixK, these findings establish AadR’s position as upstream regulator of at least three of the Fur-family transcription factors in a condition-dependent regulatory cascade, though its role for each differs in importance.

4.3. The global transcriptome is more severely disrupted by AadR loss under growth conditions where greater control is needed over iron-related or iron-requiring pathways.

Principal component analysis of mutant and wild-type transcriptomes revealed that photoheterotrophy-related gene expression (changing expression in response to light, oxygen, and carbon source) was the primary driver of the differences observed between whole transcriptomes. PC1 explained 31% of variance (organic carbon r = −0.74, O₂ |r| = 0.59) (Fig. 3A, B). Iron was nearly as strong a driver, explaining a quarter of the variation observed (PC2, 25%, [Fe] r = +0.89), and together these two components explain over half of the total variation across conditions. The displacement of ΔaadR from wild-type was largest with benzoate growth (59 units), consistent with AadR's role regulating aromatic degradation (Fig. 3C) (20). The strain’s second largest displacement was on FWAc+Fe (42 units), predominantly along the PC2 iron axis, and more than double the displacement shown on basal iron (18 units) (Fig. 3C). PC3 (14.3%) and PC4 (8%) both correlated most strongly with (anaerobic) aromatic carbon source metabolism (Fig. 3A, D), emphasizing the importance of AadR’s role regulating these pathways yet falling short of iron in importance. The disproportionate iron-related axis displacement suggests a substantial expansion from AadR’s characterized roles as a regulator of aromatic or anaerobic metabolism.

Figure 3. Principal component analysis (PCA) biplots and displacement plot reveal the hierarchy of importance between tested environmental stimulus factors on wild-type and ΔaadR TIE-1.

(A.) Correlation matrix, mapping metadata indices to principal components. PC1 correlates positively with oxygen, and negatively with light and amount of organic carbon provided. PC2 correlates very strongly (r=+0.89, p<0.001) with iron concentration, positively with light, and negatively with oxygen. (B.) PCA biplot of PC1 vs PC2. Metadata vectors and biological replicate points for VST-adjusted whole transcriptomes are shown. PC1 explains 31% of the variation, while PC2 explains 25%. (C.) PC2 vs PC1 displacement plot of how the ΔaadR knockout mutation shifted the transcriptome relative to that of wild-type on each growth condition (centroids of replicate point averages are shown). (D.) As B, but in this panel comparing PC3 vs PC1. PC3 explains 14.3% and PC4 explains 8% of the variation between samples.

To test whether AadR's regulon is significantly enriched for iron metabolism-related and/or iron-cofactor-containing proteins— or whether this apparent enrichment merely reflects the inherent iron requirements of anaerobic metabolism enzymes (46, 47)— we annotated the TIE-1 protein-coding genome for “iron-related” and “iron-requiring” gene sets to test if these were significantly over-represented in the differentially-expressed gene sets between wild-type and ΔaadR (Fig. 4). We classified 438 genes as iron-related based on an analysis of the TIE-1 genome using FeGenie (i.e., involved in iron uptake, oxidation/reduction, etc.) (48). The 174 genes classified as iron-requiring were ones whose products coordinate Fe-S clusters, heme, or other iron cofactors, and ones that are subunits of apoenzymes which do contain iron (Fig. 4A), identified using the tools HMMER and MetalPredator (49, 50) (SI Dataset S03).

Figure 4. Iron-related and iron-requiring genes are significantly enriched in the set perturbed by ΔaadR.

(A.) “Iron-requiring” genes were defined as ones whose products incorporate iron cofactors. Most iron-requiring genes encode proteins with four iron atoms (often a [4Fe-4S]), though some have substantially more— among the highest are PioA (10 hemes) and nitrogenase (with its Fe-Mo cofactor) (SI Dataset S03). Genes with zero Fe weight here are ones whose protein is a subunit of an enzyme that contains an iron cofactor. (B.) Fisher’s exact test on whole transcriptome differential expression between iron-requiring vs. non-iron-requiring genes (unweighted). The ΔaadR strain differentially regulates iron-related genes on FWAcetate, and iron-requiring genes on FWAc+Fe, FWB, and YPS. Stars mark significance compared to the background of all other genes differentially expressed in wild-type between aerobic and anaerobic conditions (log₂FC > 1, q < 0.05). Bubbles are shaded by significance (−log₁₀ q value (Benjamini-Hochberg FDR)) and sized by overlap (number of genes of each category in the differentially expressed set). The analysis was repeated for a set with badDEFG and badAB excluded (for comparison with the ΔbadM mutant, hatched bars). (C.) Weighting the iron-requiring genes by the number of iron atoms needed to express each one further strengthens the observed trend, which is significant with or without exclusion of badDEFG (12 Fe) and badAB (4 Fe). Binary tests (Fisher and Mann-Whitney U) and weighted tests (PERMANOVA, Spearman’s rank correlation, and weighted average of iron cost) are displayed. (D.) As C, but comparing transcriptomes of ΔbadM to wild-type, to serve as an analytical control. Weighted and unweighted tests found the ΔbadM set significantly enriched for iron-requiring genes only before filtering out its established regulatory associations badDEFG and badAB.

The set of genes differentially expressed between wild-type and ΔaadR was significantly enriched for those encoding iron-requiring proteins (pooled Fisher OR = 2.34, q = 2.6×10⁻⁶) (Fig. 4B). Significance of this enrichment persisted even after restricting genes that are already differentially expressed between aerobic and anaerobic conditions from the analysis. This confirms that the iron-related signal is not an artificial byproduct of regulation of anaerobic metabolism, which inherently demands iron cofactors to catalyze anoxic biochemical reactions (OR = 2.92, q = 2.0×10⁻⁵). After weighting these genes by the number of iron atoms expended to express them (Fig. 4A), iron-requiring genes continued to be over-represented (Mann-Whitney q = 1.5×10⁻¹⁰) (Fig. 4C). Targets under the regulatory scope of AadR collectively contained an average of 45% more Fe atoms than random chance would predict.

As an analysis specificity control, we used a ΔbadM knockout mutant strain (Fig. 4D). Consistent with CGA009 (6, 18, 20), deletion of badM in TIE-1 constitutively derepressed the iron-requiring badDEFG operon, encoding benzoyl-CoA reductase (SI 2, Fig. S10–S13). No other loci showed differential regulation approaching the magnitude or degree of significance as canonical targets in the bad operon. Across growth conditions, the increase in badDEFG transcription usually exceeded 300-fold relative to wild-type (Fig. S13). The degree of overexpression also seemed to allow for probable read-through into badAB to occur, which are not considered members of the operon but follow badDEFG on the same strand (Fig. S10). The set of differentially expressed genes between wild-type and ΔbadM was not significantly enriched for iron-related genes, but it was for iron-requiring genes (q = 3×10⁻⁴). This enrichment became insignificant after excluding the bad genes (q = 0.16) (Fig. 4D). The ΔbadM PCA displacement contrasted with ΔaadR's: it was stronger under −Fe (28 units on FWAc) than +Fe, consistent with the burden of overexpressing a single iron-requiring operon being greatest when iron is less available (SI 2, Fig. S13).

4.4. Characterizing the roles of AadR, the Ppio-BS2 binding site, and broader regulation of pioABC

Ppio-BS1 binds FixK and is nearly essential for the promoter’s function, shown to be reduced to ~1% of wild-type expression levels in a Ppio-ΔBS1 reporter strain. Ppio-BS2 has an unknown cognate regulator. A Ppio-ΔBS2 reporter showed constitutively reduced expression (~4-fold average reduction), and the iron-responsive activation of the promoter was abolished (9). Paradoxically, reporter activity increased in ΔfixK under non-photoferrotrophic conditions, despite the established role of FixK as an activator of PpioABC (9). Our finding that supplemental Fe(II) decreases fixK transcription implies that another factor appears to be involved in the iron-responsive activation of pioABC, as the established mechanism of pioABC activation by FixK is insufficient to explain the observed pattern.

Wild-type TIE-1 showed modestly greater pioA transcription than ΔaadR when grown on FWAc (log₂FC = +1.48, q = 2.9×10⁻⁴), but it showed no significant difference on FWAc+Fe (Fig. 5A). On FWB, only pioC expression was significantly affected (log₂FC = −1.59, q = 1.1×10⁻⁸). These modest, inconsistent effects contrast sharply with the ~4-fold changes in reporter expression in the Ppio-ΔBS2 reporter strain or the coordinated 6–18-fold expression changes in ΔfixK (9). Differences in pioABC regulation in ΔaadR are instead most parsimoniously explained by altered fixK dosage in response to the missing AadR. Transcription of fixK was 3- to 30-fold derepressed in ΔaadR across anaerobic growth conditions (Fig. 5A), which would perturb pioABC regulation by FixK via Ppio-BS1 without requiring direct AadR binding at either site in the pio promoter. Supporting this interpretation, iron supplementation during growth on acetate significantly induced pioA and pioB in ΔaadR (log₂FC = +1.59, +1.39; q < 10⁻⁴) but not in wild-type (Fig. 5B). This potentially reflects the loss of the iron-driven repression of fixK that we observed in ΔaadR (Fig. 2A,D), and these data support a lack of direct pioABC regulation by AadR at either Ppio binding site.

Figure 5. AadR is not a primary, direct regulator of pioABC iron-responsive expression.

(A.) Log₂ fold change in transcript between WT and ΔaadR. pioA, pioB, pioC, and fixK transcription was compared across four growth conditions: YPS (aerobic), FWAcetate with or without +Fe, and FWBenzoate. Positive values indicate higher expression in wild-type. (B.) Log2FC in transcript abundance between +Fe and −Fe photoheterotrophic conditions (FWAc+Fe / FWAc (−Fe)) for pioABC transcripts, in wild-type (solid bars) and ΔaadR (slashed bars). (C.) Growth curves of wild-type, single mutants, and Ppio-ΔBS deletion strains on photoferrotrophy (FW+Fe). Only WT and (after a lag) Ppio-ΔBS2 grew. (D.) Ferrozine assay Fe(II) oxidation curves of wild-type, single mutants, and Ppio-ΔBS deletion strains on photoferrotrophy (FW+Fe). Only WT and (after a lag) Ppio-ΔBS2 oxidized Fe(II).

To directly assess Ppio-BS2's functional role in photoferrotrophy, we grew Ppio-ΔBS1 and Ppio-ΔBS2 strains on FW+Fe alongside wild-type and regulator mutants (Fig. 5C). ΔBS1 entirely failed to grow, copying the phenotype of ΔfixK and confirming BS1 as the essential FixK-dependent activation site. ΔBS2 grew and measurably oxidized Fe(II), but with substantially delayed onset and less complete oxidation of the provided iron (~1.75 mM Fe(II) remaining vs. ~0.25 mM) (Fig. 5D). The defective but not abolished photoferrotrophic capability of ΔBS2 supports the previous finding that BS2 contributes to full pio activation under photoferrotrophy without being essential to growth under this condition (9).

The unknown Ppio-BS2 factor must satisfy two constraints: it must bind at or near BS2 to regulate pioABC expression, and its regulatory effect (or regulation of it) must be iron-responsive. We found that the AadR-FixK hierarchy plays a significant role in regulation of mur and both irr paralogs in response to iron. IrrA2 shows a 28-fold induction under photoferrotrophy (Fig. 2C), retains conserved residues indicating heme-sensing capability, and belongs to a family of regulators known for binding AT-rich palindromic sequences. We observed a statistically significant but imperfect Irr-binding motif overlaps Ppio-BS2 in the pioABC promoter (FIMO, MEME-suite, p = 4.73 ×10⁻⁷, q = 0.032) when scanning against the complete TIE-1 genome using a matrix built from the 122 Irr-regulated sites in Rhizobiales, supporting the hypothesis that a Fur-family regulator may bind there (Fig. S14) (33, 51, 52).

5. Discussion

5.1. AadR as a high-level iron-allocation orchestrator

Our data establish an important role for AadR as a central coordinator of iron-responsive metabolism in TIE-1 (Fig. 6). Three orthogonal lines of evidence support this conclusion. First, transcription of fixK is repressed and aadR is activated by iron supplementation (Fig. 2A), and this iron-mediated repression of fixK does not occur in the ΔaadR strain (Fig. 2D). This shows that the AadR-FixK hierarchy is modulated not just by O₂ levels but also by Fe(II) levels, even when not actively oxidizing it.

Figure 6. Model of the role played by AadR as a central integrator of environmental signal response.

AadR and FixK form a regulatory hierarchy (16), and the ratio of AadR to FixK increases in response to [Fe(II)]. This leads to the regulation of loci including the genes encoding the Fur family regulators, the protein products of which then go on to regulate their own sets of loci. AadR’s regulon is significantly enriched for iron-requiring and iron-related pathways, and it regulates them differently based on iron concentrations. Under basal iron, conversely, FixK remains more involved than AadR, regulating its set of microaerobic and anaerobic metabolic pathways. Both regulators appear to play an important role in expression of photosynthetic machinery. Except where clearly specified otherwise in the text, regulation shown in this model (even by solid arrows) is not implied to be direct.

Second, AadR mediates the differential regulation of Fur-family transcription factors in response to Fe(II) (Fig. 2B–D). Expression of irrA1 was significantly greater in wild-type than in ΔaadR, while that of irrA2 showed the inverse pattern (Fig. S5). The biological purpose of this relationship could be a regulatory shift from an aerobic, iron-responsive program primarily concerned with intracellular heme iron levels to an anaerobic iron-regulation program, potentially one more concerned with Fe(II) levels. Wild-type cells grown photoheterotrophically showed significantly increased mur expression when supplemental Fe(II) is added, but autotrophic growth via photoferrotrophy negates this increase (Fig. 2C). ΔaadR shows the same pattern: supplementing photoferrotrophic growth with Fe(II) leads to reduced mur transcript levels instead of the increase seen in wild-type (Fig. 2D). This observation demonstrates AadR’s pivotal role in shaping the important regulatory difference between growth in the presence of Fe(II) and metabolism of Fe(II). Mur is the closest homolog to ancestral Fur in R. palustris and it is predicted to directly regulate irrA2 (Fig. 6) (29, 33). While Bradyrhizobium Mur is definitively a Mn-responsive regulator with only minor or indirect roles in iron regulation (31), past work performed in TIE-1 has shown some genetic redundancy in transport pathways for manganese and iron, and close homologs of Mur have been shown to become functionally active with Fe²⁺ replacing Mn²⁺ as the effector (23, 53). Ultimately, the change of mur transcript in ΔaadR in response to iron makes a compelling case for its involvement in the AadR-mediated iron-responsive regulation in TIE-1.

Third, the AadR regulon is significantly enriched for iron-cofactor-containing proteins even after controlling for the inherent iron-intensity of the anaerobic transcriptome (Fig. 3; SI Dataset S03). This enrichment is AadR-specific, as the ΔbadM transcriptome’s iron-requiring signal vanishes to insignificance when its canonical regulatory targets (bad) are excluded (Fig. 4D), while that of ΔaadR remains significant. Whether AadR itself senses iron remains an open question. Canonical FNR regulators sense oxygen through a [4Fe-4S] cluster coordinated by a set of conserved cysteine residues, the presence of which is retained in AadR, though spacing between them is different than that in O₂-sensing FNR-family proteins (Fig. S15, SI 3) (20, 30). Thus, it is a possibility that the cysteine motif in AadR may also directly sense Fe²⁺ ions, as does the [4Fe-4S]-coordinating domain in Rhizobium RirA (54, 55). Whether direct (in this method) or indirect (e.g., via a Fur-family regulator), the consequence of the modulation of the AadR-FixK hierarchy in response to Fe(II) is the same: AadR integrates cellular oxygen and iron signals to coordinate global expression of iron-related anaerobic metabolism.

5.2. Regulation of pioABC via the second regulator binding site

AadR does not appear to regulate pioABC via Ppio-BS2. The modest changes to pioABC expression observed in ΔaadR are instead most parsimoniously explained by altered dosage of other regulators under the control of AadR (Fig. 5A). Transcription of irrA1, irrA2, and mur all showed significant regulatory differences, and fixK was significantly derepressed in ΔaadR. The growth data from the Ppio-ΔBS1 and Ppio-ΔBS2 strains provide the first direct assessment of how these binding site deletions affect photoferrotrophic growth. ΔBS2 showed delayed and incomplete photoferrotrophy, contrasting the total failure of ΔBS1 and confirming that BS2 is important (but not essential) for photoferrotrophic growth, prompting further work to identify its regulator. IrrA2 showed a 28-fold induction under photoferrotrophy with high specificity (Fig. 2C), suggesting that regulation of it integrates information on metabolic state beyond iron concentration. Electron transport chain redox state, heme availability, or signals specific to carbon fixation may be involved in its regulation. The dependence of its regulation on metabolic context distinguishes irrA2 from the other Fur-family regulators and may reflect a specialized role in coordinating iron allocation during the metabolically demanding transition to photoferrotrophic growth.

The Fur-family membership of IrrA2 and the suggestive prediction of a significant Irr-binding motif overlapping BS2 are both consistent with a role for IrrA2 at this site. In the closely related phylogenetic relative Bradyrhizobium, IrrA2 is degraded at the protein level after heme binding (30). If a similar mechanism exists in TIE-1, it could imply that the high irrA2 transcript levels during photoferrotrophy could reflect compensatory transcription in response to rapid protein turnover. This hypothesis would be compatible with a model in which the IrrA2 protein is active — and potentially repressive at the pioABC promoter — primarily under conditions of heme limitation. As a decaheme cytochrome, pioA is likely the single most heme-expensive gene for TIE-1 to express (10). Thus, it would be logical to hypothesize that pio expression is tightly regulated by a heme-sensing regulator such as IrrA2.

5.3. Ecological relevance and fitness implications

We observed several important growth defects: the ΔaadR defect on succinate, the ΔfixK lethality under +Fe, and the synthetic lethality of the ΔaadRΔfixK double mutant. It is likely that these defects were revealed and/or enhanced by our naturomimetic fluorescent lighting, which forces R. palustris to operate closer to its metabolic margins than high-output incandescent light (SI 1.1). Limited photon flux in the absorption bands of bacteriochlorophyll likely makes the regulatory precision of the AadR-FixK hierarchy even more critical. Expression of phototrophy-related genes was significantly defective in both ΔfixK and ΔaadR, and the synthetic lethality of the double mutant under all anaerobic conditions tested (Fig. 1E, S3, S4) implies that at least one of FixK or AadR would be essential for any anaerobic viability in the wild (9, 19). Under the attenuated photon flux of naturomimetic lighting, improper regulation of photosystems and iron-related pathways may have pushed the mutants below a minimum viable metabolic threshold. Either regulator alone might provide some compensation for the loss of the other, but neither seemed able to compensate when photon capture was already at the margin and iron concentrations were high. This interpretation is consistent with the observation that neither single mutant grew under photoferrotrophy, in which regulation of both iron and photosystems are simultaneously critical.

Natural perturbations of the fragile balance of the hierarchy carry severe fitness consequences for TIE-1, and therefore artificial perturbations of regulatory networks as we examined in this work could lead to useful phenotypes for research. Microorganisms in the water columns of natural wetland environments experience light more similar to our naturomimetic regime than to standard laboratory illumination (SI 1.1). A knockout strain of either fixK or aadR showed no significant growth-rate defect under aerobic conditions in rich media (Fig. S1) but had severe growth defects under anaerobic conditions. This means that such a mutation could be naturally or artificially acquired during aerobic growth, but would become a conditional lock on metabolic capabilities, greatly limiting growth under anaerobic conditions. Such a strain would likely be easily outcompeted in nature.

5.4. Limitations and future directions

Several limitations inherent in this work must be acknowledged. Our RNA-seq data comprise n=2 biological replicates per condition due to cost constraints and the large number of conditions and mutants we studied. This limited our statistical power to detect the smallest changes. Despite this, our key findings are supported by q-values well below 0.001 and are further corroborated by consistency with the independent Guzman et al. (2019) dataset (45). Transcriptomic data alone cannot distinguish between direct and indirect regulatory relationships; in particular, AadR's effects on Fur-family regulators may be mediated wholly or in part by altered fixK dosage, or by direct regulation of one that cascades into indirect regulation of the others through it. The oxygen sensitivity of the [4Fe-4S] cluster predicted for AadR poses a substantial barrier to in vitro work with it, including binding assays (21). Anaerobic ChIP-seq is challenging but possible, having been used to study the FNR regulon (56). This dataset, complemented by the RNA-seq dataset produced here, would help resolve direct binding targets.

Repeated failures to produce and maintain a clean markerless deletion of aadR drove us to ultimately generate ΔaadR with a kanamycin-resistance marker insertion, enabling selection. Dispensa et al. also reported notable difficulties knocking out aadR, and also failed to obtain a clean ΔaadR strain in CGA009 despite no significant aerobic growth defect shown by the mutant (20). Fitness data from a CGA009 transposon insertion library showed no strains with a random insertion breaking aadR, even as insertions into the genes up- and downstream of it yielded dozens of mutant strains (n > 120) (Fig. S16, SI Dataset S05), with similar methods: aerobic mutant construction with kanamycin selection (57, 58). Together, these three experiments across two strains evidence a possibility of strong counter-selection at the locus.

Complementation of ΔaadR and ΔfixK with plasmid-borne copies of each gene (inducible under Plac) only imperfectly rescued the mutant defects (Fig. S4, S7–S8). Plasmid-based complementation with IPTG induction inherently causes an expression-level mismatch, which may upset the fragile dosage-based hierarchy of our model. Antibiotic pressure may also have been a contributing factor to the complement strains’ defective growth, as wild-type TIE-1 carrying the empty expression vector was also defective. Finally, it remains possible that polar effects in the ΔaadR background meaningfully contributed to the strain’s phenotype, as the gene downstream of aadR (and of kanR in the mutant) had a log₂FC of +4.9–6.7 (see SI 1.5). The fixK complement, which has a markerless background, also showed growth defects of equal or greater severity to the aadR complement. This limits the extent to which the polar effect can be interpreted as a primary driver of the defective phenotype exhibited by the ΔaadR strain.

While low RNA-seq replication and imperfect complementation limit interpretation of subtle phenotypes, the core transcriptomic findings (including the direction and magnitude of Fur-family regulator expression changes, the iron-requiring enrichment signal, and the pioABC dosage-effect pattern) are internally consistent across multiple independent conditions. Support from the Guzman et al. (2019) and Rey and Harwood (2010) datasets (19, 45) provides confidence that the data reflect meaningful differences in AadR-dependent and iron-dependent regulation rather than bias due to methodological artifacts.

Future work should prioritize examining Fur-family regulator deletion mutants, particularly irrA2. Direct assessment of AadR's hypothetical iron-sensing capabilities would clarify the players in the iron-responsive regulatory cascade, while the technically challenging anaerobic purification of AadR would facilitate structural analysis and binding-site assays. Finally, a phylogenetic analysis linking conservation of the AadR-FixK hierarchy with lineage-specific iron-responsive regulation (e.g., Fig. S17) would help determine the extent to which this proposed regulatory architecture is conserved beyond R. palustris.

6. Methods

6.1. Strains and mutant construction

Wild-type TIE-1 was originally isolated by Jiao et al. (2005) (59). Markerless deletion mutants (ΔfixK, ΔbadM) were generated from this background using methods described previously (60) using suicide plasmid pJQ200KS (61). The ΔaadR mutant was generated with a kanamycin-resistance marker insertion: a kanR cassette under a constitutive T7 promoter was inserted between 500 bp upstream and downstream aadR flanking regions using NEB HiFi assembly (Cat#E5520S, New England Biolabs Inc., Ipswich MA). Double mutants were constructed by performing the second deletion in the single-mutant background. All mutants were confirmed by PCR with internal and external primers and by whole-genome sequencing (Plasmidsaurus). Complementation plasmids were constructed by cloning each gene with its native promoter into the expression vector pSRKGm under the Plac inducible promoter (62), and were conjugated into the relevant mutant backgrounds using WM6026 (63). The Ppio-ΔBS1 and -ΔBS2 strains were constructed through deletion of the relevant sites identified previously (9). A complete list of strains and plasmids used in this work may be found in SI Table ST4, and primer sequences used for cloning and check PCRs may be found in Table ST5. Extended discussion of mutants and complement strains may be found in Supporting Information SI 1.5.

6.2. Microbial culture conditions

Routine aerobic growth was performed on yeast-peptone medium (YP: 3 g/L yeast extract, 3 g/L peptone) buffered with 10 mM MOPS [3-(N-morpholino)propanesulfonic acid], pH 7.0, at 30°C with shaking at 250 rpm, amended with 1 mM succinate from a 1 M sodium succinate stock (YPS). Anaerobic phototrophic growth was performed in Balch tubes (photoheterotrophy) or 100-mL serum bottles (photoferrotrophy and mixotrophy) at 30°C under fluorescent lighting (SI 1.1–1.3), in bicarbonate-buffered freshwater (FW) medium as previously described (8, 9, 42). For photoheterotrophy and mixotrophy, FW was amended with 10 mM sodium acetate (FWAc), 10 mM sodium malate (FWMal), 1 mM sodium benzoate (FWB), or 10 mM sodium succinate (FW-S), from filter-sterilized anaerobic 1 M stocks. Iron-supplemented conditions were amended with 5 mM FeCl₂ and equimolar NTA from anaerobic stocks (FWAc+Fe, FWMal+Fe, FW+Fe). Pre-growth for photoheterotrophic curves was performed aerobically in YPS to better capture the aerobic-anaerobic metabolic shift phenotype. Photoautotrophic pre-growth on hydrogen for FW+Fe used a mixed gas (80:20 H₂:CO₂) headspace at 34.5 kPa. ΔaadR strains were maintained with 20 μg/mL kanamycin selection, which was omitted during growth experiments. Complement strains were grown with 400 μg/mL gentamicin and 10 mM IPTG (Isopropyl β-D-1-thiogalactopyranoside) to ensure retention of complementation vectors and expression of complemented genes, respectively. Complements were used for validation of mutant strains, partially rescuing growth phenotypes.

6.3. Growth curves and iron oxidation curves

Growth was monitored by OD₆₀₀ (Spectronic 200, Thermo Fisher Scientific). Growth curves were modeled using a modified Gompertz equation (64) with doubling times calculated: $t_{\mathrm{d}}=ln\left(2\right)\times \left(\mathrm{O}{\mathrm{D}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-\mathrm{O}{\mathrm{D}}_{\text{initial}}\right)/\left(2\mu \right)$. Statistical comparisons used two-way ANOVA with Tukey’s HSD (R v4.5.1). Growth data represent averages of biological replicates (n=3–6) with 95% confidence intervals (CI). Fe(II) concentrations were measured by ferrozine assay (65): 100-μL anaerobic samples were acidified in 1M HCl, reacted with ferrozine reagent, and measured at 562 nm (BioTek, Agilent). Concentrations were calculated from a standard curve, normalized to abiotic controls, and averaged across technical (n=2) and biological (n=2–3) replicates with standard error propagated. Extended growth analysis details are provided in SI 1.4.

6.4. RNA extraction and sequencing

RNA was harvested at mid-exponential phase (0.5 < OD₆₆₀ < 1.0) from n=2 biological replicates per strain per condition. Samples were stabilized in RNA-Later buffer, centrifuged, and stored at −80°C. RNA was extracted using the RNeasy Mini Kit (Qiagen). Library preparation and RNA-seq were performed at the Genome Technology Access Center at the McDonnell Genome Institute (GTAC@MGI). Quality was assessed by TapeStation (Agilent), libraries were prepared with ribosomal depletion (FastSelect Bacteria), and sequencing was performed on a NovaSeq X Plus (300 cycles, ~15M targeted reads/sample). Demultiplexed raw reads were processed on KBase: adapter-trimmed and quality-filtered (JGI RQCFilter, BBTools v38.22), aligned to the TIE-1 genome (RefSeq:NC_011004) with HISAT2 v2.1.0 (with splicing/intron settings disabled), and assembled with StringTie v2.1.5. Transcript counts were generated as TPM matrices and raw count matrices, which were exported as files for downstream analysis.

6.5. Differential expression and principal component analysis (PCA)

Differential expression was assessed using pyDESeq2 v0.5.4 (66) on raw transcript counts. Pairwise comparisons were performed with a single-factor design (~condition); significance was assessed by Wald test with Benjamini-Hochberg (BH) FDR correction (reported as q-values). For WT-vs-mutant comparisons, positive log₂FC indicates higher expression in wild-type. Read counts were normalized for PCA using variance-stabilizing transformation (VST) via pyDESeq2, standardized to zero mean and unit variance (scikit-learn v1.8 (67)). Pearson correlations between PC scores and environmental variables (oxygen, light, aromaticity, log₂[Fe], organic carbon concentration, total electron equivalents, total organic carbon atoms) were computed across all 28 samples (SciPy v1.17 (68)). Displacements were calculated in Euclidean distance units between wild-type and mutant centroids in PC1-PC2 space.

6.6. Genome and proteome annotation

The TIE-1 proteome was annotated for iron-related genes using FeGenie v1.0 (48) with its full iron HMM library, and for iron-cofactor-containing (“iron-requiring”) genes using HMMER v3.4 (49) with 20 curated Pfam profiles representing Fe-S clusters, heme, radical SAM, and related iron-binding domains. Predictions were reconciled with MetalPredator (50) metal-binding domain calls, and supplemented with manual curation with known experimentally-validated iron-coordinating proteins (SI Dataset S03). Gene sets (438 iron-related, 174 iron-requiring) were tested for enrichment in the WT/ΔaadR and WT/ΔbadM differential expression datasets using one-sided Fisher's exact tests and Mann-Whitney U tests on |log₂FC| distributions. Test comparisons were restricted to genes differentially expressed between aerobic and basal-iron anaerobic conditions, to control for false positive signals caused by many anaerobic genes being iron-requiring. BH-FDR correction was applied across all tests. A table mapping gene names to locus tags (Table ST6), as well as full methods, gene lists, and analysis results are in SI Datasets S01-S05.

Significance Statement.

Photosynthetic and electron-uptake capabilities make Rhodopseudomonas palustris a promising candidate for a biotechnology chassis, but the regulation of these processes is insufficiently understood in this organism, as R. palustris lacks the regulatory machinery that serves this role in its phylogenetic relatives. Here, we show that AadR, a regulator of anaerobic metabolism, plays a broader role in coordinating the iron-responsive regulatory cascade, enabling the bacterium to integrate the two key environmental signals, oxygen and iron, to orchestrate its metabolic expression. This understanding is key to laying the foundation for the rational engineering of R. palustris for scalable bioproduction applications and understanding its ability to survive in many environmental niches ranging from wetlands to rhizospheres.