Iron is not everything: unexpected complex metabolic responses between iron-cycling microorganisms

Coexistence of microaerophilic Fe(II)-oxidizers and anaerobic Fe(III)-reducers in environments with fluctuating redox conditions is a prime example of mutualism, in which both partners benefit from the sustained Fe-pool. Consequently, the Fe-cycling machineries (i.e., metal-reducing or –oxidizing pathways) should be most affected during co-cultivation. However, contrasting growth requirements impeded systematic elucidation of their interactions. To disentangle underlying interaction mechanisms, we established a suboxic co-culture system of Sideroxydans sp. CL21 and Shewanella oneidensis. We showed that addition of the partner’s cell-free supernatant enhanced both growth and Fe(II)-oxidizing or Fe(III)-reducing activity of each partner. Metabolites of the exometabolome of Sideroxydans sp. CL21 are generally upregulated if stimulated with the partner´s spent medium, while S. oneidensis exhibits a mixed metabolic response in accordance with a balanced response to the partner. Surprisingly, RNA-seq analysis revealed genes involved in Fe-cycling were not differentially expressed during co-cultivation. Instead, the most differentially upregulated genes included those encoding for biopolymer production, lipoprotein transport, putrescine biosynthesis, and amino acid degradation suggesting a regulated inter-species biofilm formation. Furthermore, the upregulation of hydrogenases in Sideroxydans sp. CL21 points to competition for H2 as electron donor. Our findings reveal that a complex metabolic and transcriptomic response, but not accelerated formation of Fe-end products, drive interactions of Fe-cycling microorganisms.


Introduction
Chemical communication is often the driving force of mutualistic interactions within microbial communities [1,2]. The excreted metabolites can positively affect the growth of co-existing organisms by providing key metabolites produced by one partner and needed by the other [3][4][5][6][7][8]. The reciprocal benefits resulting from the excretion of metabolites indicate that while each partner invests metabolic resources, they both gain something from the other mutualistic partner [9][10][11]. Such mechanisms can not only control growth but also spatial distribution within the community in cases of heterogeneous environments. Synchronization of lifestyles is particularly important for mutualism interaction based on the partner's end products for energy production [7,12,13].
Fe-cycling microorganisms are often found coexisting in nature. This is explained by the needs of both the ferrous iron [Fe(II)] oxidizer and ferric iron [Fe(III)] reducer to have sufficient exposure to electron donors and acceptors. For lithoautotrophic Fe(II)-oxidizers, including Gallionella sp. and Sideroxydans lithotrophicus ES-1, Fe(II) serves as an electron donor under micro-oxic conditions, thereby providing ample reducing equivalents for the assimilation of carbon [14][15][16][17]. Microbial Fe(II) oxidation facilitates the rapid precipitation of Fe(III) oxy-hydroxide nanoparticles [18,19], which are preferred by anaerobic heterotrophic Fe (III)-reducers, such as Shewanella spp. or Geobacter spp., due to their small mineral size [20][21][22][23]. For Fe-cycling, at least temporary access to oxygen, or to oxidized compounds, is necessary for the renewal of the Fe(III) pool [14,16]. In wetlands or peatlands, this renewal is mediated by water table fluctuations, bioturbation, or the release of oxygen via plant roots [24] which allows interaction between Fe-cycling microorganisms regardless of their different demands for O 2 . Production of Fe(III)-chelating ligands like riboflavins by Shewanella spp. or the presence of natural organic matter with Fe-complexing or electron shuttling properties can overcome spatial restrictions and speed up microbial activities [25,26].
Interestingly, a shared homology is observed between the genes encoding the Fe(III) reduction machinery located at the cell surface in S. oneidensis and the genes encoding the Fe(II) oxidation machinery of Sideroxydans lithotrophicus ES-1 [17,27]. In S. oneidensis, the Mtr machinery, comprised of the mtrABC gene cluster, is a porin-cytochrome complex in the outer membrane enabling the electron transport through the outer membrane via various hemes to Fe(III) [28,29]. The MtoAB complex of Sideroxydans is homologous to MtrA and MtrB and functions in the same manner, forming a porin-cytochrome complex. Thus, we hypothesized that in parallel to the exchange of genetic information, additional mutualistic interactions based on the exchange of each partner's exometabolites may have been established between these Fecycling partner organisms.
To elucidate these interaction mechanisms, co-culture systems are needed to facilitate downstream analytics on both the transcriptome and metabolome levels. But, cocultivation of Fe-cycling microorganisms is very challenging due to their different lifestyles for O 2 . We selected two model microorganisms, the microaerophilic, autotrophic Fe(II)-oxidizer Sideroxydans sp. CL21 isolated from Schlöppnerbrunnen fen [15] and closely related to Sideroxydans lithotrophicus ES-1 [17], and the facultative anaerobic Fe(III)-reducer S. oneidensis isolated from Lake Oneida, NY [30]. First, we optimized incubation conditions that allow both partners to thrive simultaneously. The co-cultivation design allowed us to explore the influence of exposure to the partner's end product for energy production by analyzing changes in 16S rRNA gene copies, RNA-seq analysis, and metabolomics. We hypothesized that co-cultivation will lead to an accelerated iron wheel (Fe(II) ↔ Fe(III)), because the Fe-cycling machinery would be the most affected during cocultivation. In addition, we tested the effect of the exometabolome of one partner organism on the activity of the other.

Cell-free supernatant exchange experiment
Supernatant exchange experiments were set up in triplicates according to Mori et al. [34], with the following modifications: Sideroxydans sp. CL21 and S. oneidensis cultures were grown in 200 mL MWMM amended with 1 mM lactate and 1.2 mM FeSO 4 or 10 mM hydrous Fe(III) oxyhydroxide, respectively, for 4 days with shaking at room temperature. Cultures were centrifuged, filtered (0.22 µm), cell-free supernatant was collected, and supernatant-exchange experiments were set up (1 mL inoculum 100 mL −1 media) (Fig. 2a). Fe(II) concentrations were measured using the phenanthroline method [35]. Cells were harvested at four time-points (0d, 1 d, 1.5 d, 2 d), centrifuged, and stored at −80°C.

Genomic DNA extraction
Cells were harvested via centrifugation (5000 g, 10 min, 4°C ), extracted using the DNeasy PowerSoil kit (Qiagen, Hilden, DE) according to manufacturer's instructions, and frozen at −20°C. Sideroxydans sp. CL21 genomic DNA was used for whole genome sequencing (PacBio sequencing) [36], and the assembled genome used as the reference genome for RNA-Seq analysis.

Sample prep for RNA-Seq analysis
Total RNA was extracted from three biological replicates from Sideroxydans sp. CL21 and S. oneidensis monoculture and co-culture incubations using the phenol-chloroform extraction method adapted from Wegner et al. [37]. Equimolarly pooled cDNA libraries were sequenced in pairedend mode (2 × 150 bp) on an Illumina NextSeq 500 platform by LGC Genomics (Berlin, DE).

Sample preparation for metabolomic profiling
Sample preparation was conducted in triplicate as described previously [34], unless otherwise noted. Briefly, 50 mL of liquid cultures were centrifuged (4000 g) and extracted using Strata-X ® polymeric reverse phase cartridges (200 mg adsorbent; Phenomenex, Torrance, CA, USA). Sample preparation was performed in an anaerobic chamber to minimize Fe(II) oxidation and subsequent Fe(III) precipitation. (NOTE: in previous experiments, precipitation of Fe(III) during sample preparation resulted in completely dissimilar metabolite profiles-data not shown). Samples were measured in a randomized sample list immediately after preprocessing. GC/MS measurement GC/MS metabolomics analysis was conducted on 3 and 6 d supernatant exchange samples using a high resolution Q-Exactive-GC electron impact (EI) orbitrap mass spectrometer (Thermo Scientific), with gas-chromatographic separations implemented on a Trace 1310 equipped with TriPlus RSH autosampler. A TG-5SILMS column (length = 30 m; inner diameter = 0.25 mm, 0.25 μm film; Thermo Scientific) was used. Column operation parameters and GC-Orbitrap settings used are described in the supplementary methods.

XCMS metabolomic analysis
The XCMS data processing was carried out using -cdf files, which were converted from the Thermo RAW-files using the Xcalibur 3.0.63 (Thermo Scientific) onboard file converter. Processing was carried out using XCMS Server version 3.01.01. Pre-defined settings were used for GCmeasurement "Single Quad (matched filter)," except, retention time correction was removed [49]. Parameters used are listed in Table S2.

UHPLC/HRMS measurement and spectra acquisition
Ultra high performance liquid chromatography (UHPLC) coupled with high resolution MS (HRMS) of both for untargeted and targeted (γ-aminobutanoic acid (GABA) and zinc) metabolites was performed using an UltiMate HPG-3400 RS binary pump (Thermo Scientific, Bremen, Germany) and WPS-3000 autosampler (Thermo Scientific) equipped with a 25 µL injection syringe and 100 µL sample loop. Detailed parameters used are described in the supplementary methods.

GC/MS data processing and annotation
Using an in-house R-script [50], detected masses were deconvoluted via peak-shape comparison to potential metabolites and reduced in number to only include potential metabolites containing ≥20 fragments. Potential metabolites present in blanks and media controls (<fivefold increase or decrease relative to bacterial samples) were removed (Table S3). The remaining potential metabolites were logtransformed and auto-scaled via Metaboanalyst 4.0 [51] to yield heatmaps with the 15 most dysregulated metabolites. Unknown metabolites were identified by comparing fragmentation patterns to known compounds in the NIST/EPA/ NIH EI Mass Spectral Library NIST 17 (copyright United States Department of Commerce) searched with the NIST Mass Spectral Search Program v.2.3a nd Golm Metabolome Database library [52]. Metabolites with match scores >700 were tentatively identified and compounds with multiple match score hits >600 were labeled by compound class. During the analysis of supernatant exchange samples, an S. oneidensis control (Shew(Shew)) sample (6 d) was also removed from the analysis due to errors during spectral profiling.

UHPLC/HRMS data processing and quantification
Quan Browser (XCalibur 3.0.63) (Thermo Scientific) was used for peak detection and integration for GABA and a detected zinc compound with the following settings: mass tolerance = 10 ppm, peak detection algorithm=ICIS; smoothing points=1; baseline window = 40; area noise factor=15; peak noise factor = 10; peak high = 5%, tailing factor = 2.5 and peak detection method = nearest RT. Each biological sample was measured in triplicate using a randomized sample list. The presence of GABA in the samples was confirmed by comparison to GABA standards. In order to identify metalloenzyme complexes, for example zinc complexes, MICP was applied to the UHPLC/HRMS data of the supernatant extracts and the zinc compounds were identified based on the natural isotopic pattern. The single isotopes of the metal-complexes were resolved due to a resolution of 280.000 at m/z = 200, e.g., for 64 Zn-, 66 Zn-, 67 Zn-, and 68 Zn-complexes for one substance. The identification is based on the mass differences between the single isotopic peaks of the substance (with an error <1 ppm) and the intensity of the mass peaks according to the abundance of the natural zinc isotopes.

Results
Merging lifestyles of S. oneidensis and Sideroxydans sp. CL21 S. oneidensis can be grown easily anaerobically in liquid media with Fe(III) as a terminal electron acceptor. However, the microaerophilic Sideroxydans sp. CL21 prefers growing in so-called "gradient" tubes filled with semi-solid media, an Fe(II) source in the bottom, and O 2 supplied in the headspace. Opposing Fe and O 2 gradients allow growth at a depth with optimal O 2 concentrations of 20-40 µM [53]. Microbial Fe(II)-oxidation leads to the formation of a distinct orange ring of precipitated Fe(III) (oxy)hydroxides. However, the semi-solid media poses complications for the envisioned metabolomics profiling. To identify metabolites produced and secreted by each individual partner, we had to establish a suboxic liquid media setup that enables the growth of both Sideroxydans sp. CL21 and S. oneidensis in monocultures and co-cultures. In total we tested 48 combinations of media, carbon sources, buffer type, and Fe sources, bottle sizes, and headspace conditions to land on the best setup for our incubations (Fig. 1a): MWMM supplemented with MES buffer, lactate, and Fe (0) provided in a semi-solid media plug at the bottom, and a headspace of N 2 : CO 2 :O 2 (78:20:2) that is constantly flushed through the bottles. Under these conditions, Fe (0) reacts slowly with water forming Fe(II) and H 2 (anaerobic corrosion) and the released Fe(II) is slowly abiotically oxidized to Fe(III). Consequently, both Fe(II) and Fe(III) are continuously provided, and microbial activity should speed up the production of each Fe species. Unlike other Fe minerals, Fe (0) is not ubiquitous in environmental systems and Fe (0) is most often applied to groundwater and soil environments for in situ remediation of metalloid-contaminated sites. However, amendment with Fe (0) results in enrichment of Fe(III)reducing and Fe(II)-oxidizing microorganisms [54][55][56][57][58].

Exometabolome stimulates growth and activity of each partner
First, we performed supernatant exchange experiments (Fig. 2a). Amendments with the partner's cell-free supernatant led to enhanced rates of the production and consumption of Fe(II). Rates of Fe(II) oxidation by Sideroxydans sp. CL21 over 2 days increased from 4.3 to 5.5 μM h −1 (t-test, p = 0.002) when amended with S. oneidensis cell-free supernatant and Fe(III) reduction rates of S. oneidensis increased from 4.9 to 7.3 μM h -1 (t-test, p = 0.007) following amendment with Sideroxydans sp. CL21 cell-free supernatant (Fig. 2b, Supplementary Table S4). 16S rRNA gene copy numbers increased one order of magnitude in Sideroxydans sp. CL21 [Sid(Shew)] and two orders of magnitude in S. oneidensis [Shew(Sid)] after 3 days of batch incubation compared with its control Sid (Sid) or Shew(Shew), respectively (Fig. 2c).
These stimulatory effects show that diffusive metabolites are beneficial for this mutualistic relationship, such that metabolites from one species promotes growth and other metabolic processes, such as iron oxidation or reduction, of the other species. To identify potential chemical mediators, comparative (untargeted) metabolomic profiling was performed after 3 and 6 days. After three days of incubation with S. oneidensis supernatant, an increase in Sideroxydans sp. CL21 excretion was detected (Fig. 3a concentrations affected by treatment, (out of 473 annotated peaks), were more abundant in the exometabolome of Sideroxydans sp. CL21 exposed to S. oneidensis supernatant (Fig. 3b, Supplementary Table S3). S. oneidensis exposed to the supernatant of Sideroxydans sp. CL21 also exhibited increased excretion with a total of 321 annotated compounds. A mixed metabolic response was observed, with only 52% of the top 25 most dysregulated compounds more abundant after 3 days. After 6 days, all 25 of the top dysregulated compounds were less abundant in S. oneidensis exposed to Sideroxydans sp. CL21 supernatant suggesting that these metabolites were not being produced at similar rates as in monocultures. For example, a dihydroxyindole was putatively identified (accurate mass 365.1655 m/z, which fits −0.8 ppm according to exact mass; NIST database R. match 674 and match score 658). An overall decrease in excretion at day 6 relative to day 3 was also exhibited, as evidenced by only 151 annotated compounds detected at day 6. Efforts to identify additional specific metabolites whose concentrations differed most were unsuccessful, which is in accordance with a high degree of secondary metabolite production that are not annotated in standard libraries (Fig. 3, Supplementary  Table S3).

RNA-Seq profiling of co-cultures show unexpected patterns of differentially expressed genes
Sideroxydans sp. CL21 and S. oneidensis were grown together over 9 days to cover potential mutually induced responses and effects of physical contact with the partner (Fig. 1a). Co-culturing was beneficial to growth, as the increase in 16S rRNA gene copies of both organisms was higher in co-cultures compared with monocultures ( Fig. 1b) with the highest increase between 3 and 6 days. Thus, maximal 16S rRNA gene copies in co-cultures coupled with increased Fe-cycling rates in supernatant exchange experiments (Table S4) led to the selection of day 6 for RNA-Seq analyses.
The overall transcriptome profiles of the monocultures demonstrated that genes involved in Fe(II) oxidation (i.e., the Sideroxydans sp. CL21 mto gene cluster) or Fe(III) reduction (i.e., the S. oneidensis mtr gene cluster) were highly expressed (ranging from 6.2-9.5 log 2 CPM), showing that not only Sideroxydans sp. CL21, but also S. oneidensis were utilizing Fe under suboxic conditions. Both monoculture gene expression profiles were dominated by genes encoding proteins linked to core metabolic functions, ribosomal proteins, RNA polymerase subunits, and ATP synthase subunits (Fig. 4a, Supplementary Tables S5, S6). Genes encoding a lactate permease and lactate dehydrogenase were highly expressed, which indicated S. oneidensis is using lactate as the electron donor (Supplementary  Table S5). Genes involved in flagellar motility were also highly expressed in both organisms (Supplementary  Tables S5, S6). In the case of S. oneidensis, we detected notably high gene expression values (7.5-9.0 log 2 CPM) for genes of the putrescine degradation pathway (Table 1,  Supplementary Table S5). In the Sideroxydans sp. CL21 gene expression profiles, we identified multiple genes for cation efflux system proteins that featured particular high gene expression values (log 2 CPM > 10) (Supplementary  Table S5). A cluster of NiFe-hydrogenases-related genes, including genes for hydrogenase subunits and proteins involved in hydrogenase assembly and maturation was uniformly and highly expressed with log 2 CPM values ranging between 7 and 10 ( Table 2, Supplementary  Table S6). Genes involved in the Calvin-Benson-Bessham pathway, which reduces 3-phosphoglycerate via RubisCO (ribulose-1,5-bisphosphate carboxylase/oxygenase), were highly expressed in Sideroxydans sp. CL21, such that genes encoding the proteins CbbO and CbbQ were upregulated (log 2 FC 1.02-1.31). The presence of both the small and large forms of RubisCO suggests a high tolerance to fluctuating O 2 concentrations [17]. Also, genes encoding a lactate dehydrogenase and permease were highly expressed, thus implicating possible simultaneous use of inorganic and organic carbon during co-cultivation (Supplementary  Table S6). For co-culture incubations, 15.7 and 21.3% of the total genes were differentially expressed in Sideroxydans sp. CL21 and S. oneidensis, respectively. 927 genes were differentially expressed of which 428 were up-and 499 downregulated in S. oneidensis, compared with 588 genes (249 up-, 338 down-regulated), which were differentially expressed in Sideroxydans sp. CL21 when grown in coculture ( Supplementary Fig. S1). Surprisingly, genes involved in Fe(II) oxidation or Fe(III) reduction in both partners were not upregulated in co-culture, but were still highly expressed, indicating growth in steady-state (Tables 1, 2; Supplementary Tables S5, S6). The most upregulated genes in S. oneidensis include genes encoding putrescine breakdown and an EPS biosynthetic gene cluster, as well as substrate-binding and permease components of zinc and tungstate ABC-type uptake systems. The most downregulated genes encode proteins linked to pilus retraction, type IV pili, and chemotaxis signal transduction ( Table 1, Supplementary Table S5). In Sideroxydans sp. CL21, the most upregulated genes code for biopolymer and Fig. 4 RNA-seq profiling reveals key metabolic pathways driving interactions in co-culture. Graphical representation of overall gene expression in Sideroxydans sp. CL21 and S. oneidensis monocultures (a) compared wiith the differential gene expression during co-cultivation b are shown. Differentially expressed genes were determined by analysis of RNAseq data from triplicate coculture (Sideroxydans CL21 + S. oneidensis) and monoculture batch incubations at the 6 d time point (see Fig. 1). log 2 CPM (CPM = counts per million) values represent a normalized measure for gene expression based on the mRNA-derived sequences that were mapped onto genes. lipoprotein transport, NiFe hydrogenase incorporation and metallo-center assembly proteins, and hydrogenase-1 and -2 components. The first couples H 2 oxidation in the periplasm to reduction of quinones in the inner membrane, whereas the second is involved in O 2 reduction. Motility-related genes and genes associated with NAD-dependent formate dehydrogenase alpha, beta, gamma, and delta subunits and the sulfur carrier protein FdhD were among the most downregulated in Sideroxydans sp. CL21 (Table 2, Supplementary Table S6).

Targeted candidate analyses
Based on gene expression changes in S. oneidensis, which include gene clusters involved in putrescine and amino acid breakdown, an ABC-type tungsten-and an ABC-type zinc uptake system, we used targeted analyses to examine metabolites linked to these differentially expressed pathways. We detected GABA and evidence of ornithine and arginine [59,60], all of which are metabolites comprising the putrescine biosynthesis and breakdown pathways. The Samples were collected from biological replicates at the 6 d time point. All positive changes (upregulated genes) and negative changes (downregulated genes) shown in Table 1 correspond to differential gene expression patterns in S. oneidensis in co-culture in comparison to the gene expression in the monocultures. p < 0.05 was considered significant. Samples were collected from biological replicates at the 6 d time point. All positive changes (upregulated genes) and negative changes (downregulated genes) shown in Table 2 correspond to differential gene expression patterns in Sideroxydans sp. CL21 in co-culture in comparison to the gene expression in the monocultures. p < 0.05 was considered significant. upregulated pathways are thus responsible for the production of excreted intermediates, but putrescine itself and agmatine, another biosynthetic precursor, were not detected. We also detected GABA, ornithine, and arginine in the supernatant exchange experiment both on day 3 and day 6. We quantified GABA in supernatant exchange experiments on day 6 (Fig. 5). The amendment of the Sideroxydans sp. CL21 supernatant to S. oneidensis did not influence GABA amounts. In contrast, addition of the S. oneidensis supernatant to Sideroxydans sp. CL21 led to a significant decrease of GABA (t-test, p < 0.001, n = 3) (Fig. 5). No tungsten complexes were identified based on the evaluation of the isotope patterns of the molecular ions. Isotope pattern analysis indicated a metabolite containing complexed zinc. The identification of the metabolite containing complexed Zn was based on the mass differences between the single isotopic peaks of the substance (with an error <1 ppm) and the intensity of the mass peaks was determined according to the abundance of the natural Zn isotopes. Addition of S. oneidensis supernatant to Sideroxydans CL21 as well as the amendment of Sideroxydans CL21 supernatant to S. oneidensis led to reduced levels of the Zn compound (Fig. S2).

Discussion
Systematic investigations of Fe-cycling microorganisms were hampered in the past due to their different requirements for oxygen. After methodical optimization, we could merge the growth requirements of a microaerophilic Fe(II)oxidizer and an anaerobic Fe(III)-reducer to study interactions of organisms driving the complete iron wheel. The overall expression levels of genes involved in Fe(III) reduction in S. oneidensis and Fe(II) oxidation in Sideroxydans sp. CL21 were high under suboxic conditions. Although cell densities of standard monoculture incubations were not reached using this co-culture approach [61][62][63][64], we argue that such "candy shop" laboratory conditions do not mimic conditions in nature. Only co-culturing allows investigating the dynamics of interaction mechanisms of naturally co-occurring bacteria. By combining transcriptomics and metabolomics, we disproved the hypothesis that the expression of genes involved in Fe(III) reduction and Fe(II) oxidation is the most affected during co-cultivation. Our data further suggest no acceleration of the iron wheel at all under these conditions. Excreted metabolites promoting growth and Fecycling were the greatest benefit of co-existence. Exometabolites from both S. oneidensis and Sideroxydans sp. CL21 elicit a positive effect on growth of their partners, suggesting a mutualistic interaction (Fig. 2). The physical presence of the partner seemed to be less important than the presence of the partner's exometabolome.
Untargeted metabolomics revealed an overall upregulated metabolism of Sideroxydans sp. CL21 from day 3 to 6 when exposed to S. oneidensis exudates compared with Sideroxydans sp. CL21 alone (Fig. 3a, b). In contrast, S. oneidensis exposed to Sideroxydans sp. CL21 exudates showed a mixed response on day 3 and a predominately downregulated metabolism on day 6 ( Fig. 3c, d). Here, the increased performance did not correlate with increased metabolite excretion. Alternatively, S. oneidensis might take up metabolites from the spent medium more efficiently if triggered with exudates from the partner. We found evidence for regulated phenylketone metabolism in Sideroxydans sp. CL21 exposed to S. oneidensis supernatant at day 3 and indole metabolism in S. oneidensis exposed to Sideroxydans sp. CL21 at day 6. Sugars, amino acids and carboxylic acids were differentially regulated at day 3, indicating a balanced response of primary metabolism by both partners.
The most differentially expressed genes of both partners grown in co-culture were involved in polyamine biosynthesis and degradation, biofilm formation, motility (i.e., flagella assembly and chemotaxis), utilization of inorganic and organic carbon substrates, and the formation of hydrogenases and dehydrogenases (Fig. 4b). These genes are known to be involved in physiological responses to environmental cues and nutrient availability, however, they are not uniquely related to Fe(III) reduction or Fe(II) oxidation.
The genomes of both S. oneidensis and Sideroxydans sp. CL21 encode NiFe hydrogenases, implicating a potential competition for H 2 . Periplasmic NiFe hydrogenases in S. oneidensis play a role in both H 2 production and oxidation [65,66], while the NiFe hydrogenases in Sideroxydans sp. CL21 enable only H 2 oxidation [67]. In iron-rich peatlands, like the Schlöppnerbrunnen fen, H 2 concentrations range from 4.5 to 6.0% in depths up to 40 cm [68,69]. Here H 2 appears to be produced by fermenters, with the rhizospheres providing a prime location, and is scavenged by anaerobic secondary fermenters, Fe(III) reducers, methanogens, and acetogens [68,[70][71][72]. As Sideroxydans sp. CL21 might be also able to utilize H 2 in this fen as an alternative electron donor, variable oxygen concentrations released by plant roots in the rhizosphere and redox fluctuations due to variations in the water table might favor this microaerophile in the interspecies H 2 competition. The differential gene expression patterns observed during co-cultivation confirmed interspecific H 2 competition. Upregulation of genes encoding various NiFe hydrogenases in Sideroxydans sp. CL21 suggests the preferred utilization of H 2 , which serves here as an alternative electron donor to Fe(II). Simultaneous downregulation of genes encoding hydrogenases in S. oneidensis implicates synchronization between these Fecycling microorganisms. To compensate for the lack of H 2 as the electron donor, S. oneidensis invests in accelerated lactate permease and dehydrogenase expression (Fig. 4b).
These orchestrated responses in an environment offering multiple energy sources can even lead to a slowdown of the oxidative part of the iron wheel.
Differential gene expression patterns linked to biofilm formation suggest that Sideroxydans sp. CL21 and S. oneidensis communicate in co-culture using excreted metabolites to coordinate physiology. Biofilm formation was likely triggered by the upregulation of genes involved in the biosynthesis of putrescine and EPS, as well as the downregulation of genes involved in chemotaxis, motility, and type IV pili biosynthesis and activation in S. oneidensis. Detection of GABA, a metabolic product of putrescine, provides additional evidence that polyamines do play a role in microbial interactions [34]. In Sideroxydans sp. CL21, upregulation of genes involved in production and transport of biopolymers and lipoproteins and downregulation of genes involved in flagella machinery biosynthesis and motility are also relevant for biofilm formation during cocultivation. Polyamine signaling in co-culture [73][74][75] might regulate biofilm formation activity of Sideroxydans sp. CL21. In conclusion, both partners invest in biofilm formation, thus initiating and supporting mutualistic interactions.
Polyamines, including norspermidine, spermidine, and putrescine are known to be required for cell growth and are important signaling molecules. Norspermidine enhances biofilm formation as an extracellular signaling molecule in Vibrio cholerae by interacting with a periplasmic sensor protein and a transmembrane protein [74,76,77]. Spermidine and putrescine transporters play a role in surfaceassociated growth of Agrobacterium tumefaciens and Pseudomonas putida [78,79]. In Yersinia pestis, endogenous putrescine is required for the formation and development of biofilms. Amendment with putrescine rescues biofilm-deficient phenotypes in Y. pestis, implicating a role for both endogenous and exogenous putrescine in the activation of biofilm formation [80]. The latter finding is especially important for the current study, such that putrescine produced by S. oneidensis might play a dual role as an endogenous signal for biofilm development in S. oneidensis and an exogenous signal supporting the same process in Sideroxydans sp. CL21.
Other highly upregulated genes in S. oneidensis include those associated with ABC-type Zn transporters and ABCtype tungstate uptake systems (Table 1). Zn, an essential micronutrient, is also known to be toxic in excess as being a highly competitive divalent metal capable of displacing weakly bound transition metals in the active sites of metalloenzymes [81]. The observed upregulation of ABCtype Zn transporters suggests dynamic regulation of Zn homeostasis to preserve metabolic functioning. The upregulation of genes involved in ABC-type tungstate uptake systems hints that tungsten may function as a cofactor in place of molybdenum for formate dehydrogenases (FDHs) [82,83]. Interestingly, FDHs can also use FeS clusters, flavins, or cytochromes as cofactors, all of which play a role in Fe(III) reduction and electron transfer [84]. While the links between tungsten-dependent enzymes and electron transfer in co-cultures are poorly understood, tungsten-containing FDHs are beneficial in certain syntrophic interactions under electroactive conditions [83].
Sideroxydans sp. CL21 and S. oneidensis share homologous Fe-cycling machinery: mto genes in Sideroxydans sp. CL21 and mtr genes in S. oneidensis [17,27]. In both microorganisms, these gene clusters function as a porin-cytochrome complex forming an electron conduit through the outer membrane via various hemes. This cluster ultimately results in either reduction of Fe (III) outside the cell, coupled to the oxidation of organic carbon or H 2 in case of S. oneidensis [28,29], or the transfer of electrons from Fe(II) via MtoAB to a molybdoenzyme and subsequently to the respiratory chain (i.e., cbb 3 or bd oxidase) via CymA [85] in Sideroxydans sp. CL21. The genes encoding MtoAB are located in a gene cluster also containing mtoD, a soluble periplasmic cytochrome encoding the protein MtoD, and cymA (homologous to cymA found in S. oneidensis), a quinol oxidoreductase encoding the protein CymA [17,27]. The mtoAB cluster is also homologous to the pioABC gene cluster in Rhodopseudomonas palustris, which similarly forms a porin-cytochrome conduit [85].
Homologs to MtrAB and PioAB are widespread amongst Fe(III)-reducers and Fe(II)-oxidizers widely distributed across Alpha-, Beta-, Gamma-, and Deltaproteobacteria [85] pointing toward possible horizontal gene transfer events. It is believed that Fe(III) respiration is one of the oldest respiration processes on Earth [86] and aerobic microbial Fe(II) oxidation became possible billions of years ago when oxygen was released as a byproduct by photosynthetic organisms [85]. Various mechanisms to optimize mutual dependencies including communication might have developed over time when colonizing the same habitat despite seemingly different preferred ecological niches (i.e., microoxic vs. anoxic) [14,87,88]. Thus, the homologous Fe-cycling machinery, the ability of both bacteria to positively respond to the partner's diffusive exometabolites concerning activity and growth, and the observation that gene expression is in flux (differentially expressed genes) when these microorganisms are active in co-culture, hints at potential co-evolution.
Our co-cultivation setup provides evidence that S. oneidensis is respiring Fe(III) under suboxic conditions without switching to its preferred terminal electron acceptor, O 2 . This flexibility allows adjustment to redox gradients and helps to explain the spatial co-existence of Fe-cycling bacteria in a variety of habitats [89][90][91][92]. Under fluctuating redox conditions in nature in which these microorganisms must continuously adapt, regulation of exometabolites can become key for Fe-cycling. Acceleration of the iron wheel via optimization of electron donor and electron acceptor availability may not be the most important factor driving their interaction, since Fe is the fourth most abundant element on Earth [93,94]. Instead, both partners invest in shaping their environment via biosynthesis and detection of diffusive metabolites. Joint biofilm formation might be more important to ensure access to common goods, such as riboflavin, that function as electron shuttles [95][96][97][98][99]. High abundances of FeOB and FeRB have been reported from microbial mats, one of the most well-studied types of biofilms [100]. Simultaneous activity, which we mimicked in our setup, may not be as crucial in nature, where fluctuating redox conditions favor alternately one half of the iron wheel. Temporally decoupled redox processes are the rule and not the exception in nature.

Conclusion
The results of this study have important implications for the understanding of microbial metal-cycling. Using an integrative transcriptomic and metabolomic approach, we revealed that this Sideroxydans sp. CL21 and S. oneidensis co-culture system can be used as a model system for future investigations focused on the interactions of Fe(II) oxidizers and Fe(III) reducers that likely occur in nature and the diffusive metabolites that play a role in these interactions. We provide quantitative evidence that Sideroxydans sp. CL21 and S. oneidensis benefit from growth in co-culture, and that diffusive metabolites play a more important role in biogeochemical processes than previously thought. Shaping the environment by a regulated interspecies biofilm formation appears to be the key mechanism underlying interactions of Fe-cycling microorganisms. Characterization of biofilms formed by Sideroxydans sp. CL21 and S. oneidensis during co-cultivation will provide further evidence about the regulation of these interactions, specifically in regards to producing, sharing, and utilizing common goods, such as diffusive chemical mediators and electron shuttles. Our data further demonstrated that competition for H 2 between Fe-cycling partners provides another level of metabolic complexity when both partners are simultaneously active. Future incubation studies are needed to prove that Sideroxydans sp. CL21 does, in fact, prefer H 2 .

Data availability
The RNA-seq datasets used in this study have been

Compliance with ethical standards
Conflict of interest The research presented in this manuscript was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.