An essential role for UshA in processing of extracellular flavin electron shuttles by Shewanella oneidensis

Authors


E-mail gralnick@umn.edu; Tel. 612 626 6496; Fax 612 625 5780.

Summary

The facultative anaerobe Shewanella oneidensis can reduce a number of insoluble extracellular metals. Direct adsorption of cells to the metal surface is not necessary, and it has been shown that S. oneidensis releases low concentrations flavins, including riboflavin and flavin mononucleotide (FMN), into the surrounding medium to act as extracellular electron shuttles. However, the mechanism of flavin release by Shewanella remains unknown. We have conducted a transposon mutagenesis screen to identify mutants deficient in extracellular flavin accumulation. Mutations in ushA, encoding a predicted 5′-nucleotidase, resulted in accumulation of flavin adenine dinucleotide (FAD) in culture supernatants, with a corresponding decrease in FMN and riboflavin. Cellular extracts of S. oneidensis convert FAD to FMN, whereas extracts of ushA mutants do not, and fractionation experiments show that UshA activity is periplasmic. We hypothesize that S. oneidensis secretes FAD into the periplasmic space, where it is hydrolysed by UshA to FMN and adenosine monophosphate (AMP). FMN diffuses through outer membrane porins where it accelerates extracellular electron transfer, and AMP is dephosphorylated by UshA and reassimilated by the cell. We predict that transport of FAD into the periplasm also satisfies the cofactor requirement of the unusual periplasmic fumarate reductase found in Shewanella.

Introduction

The shewanellae are a diverse genus of Gram-negative γ-proteobacteria that respire a wide array of organic and inorganic compounds (Nealson and Scott, 2006; Hau and Gralnick, 2007). Respiratory substrates include soluble organic compounds, toxic metals such as uranium and technetium, and insoluble metallic solids such as Fe(III) and Mn(IV). Their influence on redox state and thereby on metal solubility has generated interest in using Shewanella for bioremediation of contaminated groundwaters (Hau and Gralnick, 2007; Pinchuk et al., 2008). In addition, the ability of Shewanella to reduce electrodes has led to the development of microbial fuel cells in which the electrode serves as terminal electron acceptor for bacterial cultures (Gorby et al., 2006; Lovley, 2008; Marsili et al., 2008).

In many natural environments, the oxidized iron and manganese that can serve as electron acceptors for Shewanella are insoluble at neutral pH. Therefore, Shewanella face the problem of transferring electrons from the cell surface to the metal. Direct contact of Shewanella with metal surfaces can occur and does account for at least some of the metal reduction by Shewanella cultures (Lies et al., 2005; Gorby et al., 2006; Baron et al., 2009). However, direct contact is not possible for every cell, particularly in multi-layer biofilms, and it is clear that other solutions must also play a role. The predominant mechanism used by Shewanella appears to be release of a diffusible mediator into the environment (Nevin and Lovley, 2002; Lies et al., 2005). Although several potential compounds have been suggested over the years, including quinones (Lovley et al., 1996; Newman and Kolter, 2000; Ward et al., 2004) and melanin (Turick et al., 2002), it has recently become apparent that the primary mediators used by Shewanella are flavins. Both riboflavin and flavin mononucleotide (FMN, Fig. 1) are detected in low concentrations (between 250 nM and 1 µM, typically) in cultures of Shewanella oneidensis MR-1 and other Shewanella species (von Canstein et al., 2008; Marsili et al., 2008; Coursolle et al., 2010). Washout of flavins by replacement of media in bioreactors leads to approximately 80% decrease in S. oneidensis-induced current (Marsili et al., 2008), and addition of exogenous flavins to Shewanella cultures enhances the rate of Fe(III) reduction by those cultures (von Canstein et al., 2008).

Figure 1.

Structure of flavin adenine dinucleotide. Various moieties are indicated.

In theory, a diffusible mediator could enhance the rate of metal reduction either by chelating the metal or by shuttling electrons from the outer membrane (OM) to the metal surface. Although flavins can chelate iron through their isoalloxazine ring (Albert, 1950; 1953), evidence suggests that the shuttling activity is dominant. First, known chelators have less effect on iron reduction rates than do flavins (Coursolle et al., 2010), and second, flavins enhance reduction of surfaces that cannot be chelated, such as carbon electrodes (Marsili et al., 2008; Baron et al., 2009; Velasquez-Orta et al., 2010). Current data indicate that electrons are passed from the menaquinone pool to the outer surface of the cell through the Mtr pathway, where the OM cytochromes MtrC and OmcA reduce extracellular substrates (Shi et al., 2007), including flavins (Coursolle et al., 2010). Reduced flavins can transfer electrons to an insoluble electron acceptor and be re-reduced by the Mtr pathway. Though flavins may diffuse away from the cell, the most efficient mechanism could be through inter-molecular electron transfer between flavin molecules, the Mtr pathway and the insoluble terminal electron acceptor.

Although flavins are utilized by Shewanella as electron shuttles, they are also used ubiquitously as enzyme cofactors. The riboflavin derivatives FMN and flavin adenine dinucleotide (FAD, Fig. 1) play a crucial role in biological redox reactions by catalysing both one-electron and two-electron transfers in such diverse enzymes as dehydrogenases and monooxygenases (Walsh, 1980), and photoreceptors (Losi and Gartner, 2008). One critical flavoenzyme in Shewanella is the fumarate reductase FccA, a periplasmic tetraheme c-type cytochrome containing a non-covalently bound FAD cofactor, required for utilization of fumarate as an anaerobic electron acceptor (Pealing et al., 1992; Turner et al., 1999; Maier et al., 2003).

Nothing is known of the mechanism by which Shewanella secrete flavins. Although several bacterial riboflavin importers have been identified (Kreneva et al., 2000; Vitreschak et al., 2002; Burgess et al., 2006; Vogl et al., 2007), none has been shown to export flavins. A potential exporter of FAD has been identified in mitochondria of yeast (Bafunno et al., 2004), but this transporter has no apparent homologues in Shewanella. Because Shewanella relies on extracellular flavins for reduction of insoluble electron acceptors, the question of how Shewanella transports and processes flavins will be critical to understanding this mode of anaerobic respiration.

We have conducted a random mutagenesis screen designed to isolate mutants in the flavin secretion/release pathway. The mutant with the most severe phenotype was found to be disrupted in the open reading frame (ORF) SO2001 encoding the gene ushA. We find that FAD is the predominant flavin species secreted by S. oneidensis, and that UshA processes periplasmic FAD to FMN. We also find that Shewanella cells are able to re-uptake and use the adenosine monophosphate (AMP) released from FAD hydrolysis as a carbon source and that this activity is also UshA-dependent. Our results have implications for the search for flavin transport mechanisms in Shewanella as well as for our understanding of the evolution of electron shuttling and applications of mediator-enhanced dissimilatory electrode reducing bacteria.

Results

Deletion of ushA results in decreased flavin fluorescence in culture supernatants of S. oneidensis

To identify genes involved in extracellular flavin production by Shewanella, we conducted a random mutagenesis screen, taking advantage of the native fluorescence profile of flavins (Ghisla et al., 1974). S. oneidensis strain MR-1 was conjugated with a 6.5 kb transposable element, TnphoA′-1, encoding a lacZ fusion and kanamycin resistance marker (Wilmes-Riesenberg and Wanner, 1992). Kanamycin-resistant transductants were inoculated into 96-well plates, and mutants with potential flavin production defects were identified by decreased fluorescence in the culture medium. Out of approximately 10 000 mutants screened, six mutants were obtained with consistently less fluorescence than wild-type MR-1. Sequence analysis led to identification of the ORF SO2001 as the transposon insertion site in five mutants.

SO2001 encodes a putative 61 kDa protein with 50% identity (amino acid alignment using NCBI blast) to the UDP-sugar hydrolase/5′-nucleotidase UshA of Escherichia coli. Given the significant similarity between the MR-1 and E. coli proteins we will refer to SO2001 as UshA. UshA is a conserved periplasmic metallophosphoesterase that hydrolyses a range of nucleotides, UDP-sugars and CDP-alcohols (Glaser et al., 1967; Neu, 1967; Alves-Pereira et al., 2008). In E. coli, UshA is involved in nucleotide scavenging (Beacham et al., 1973; Yagil and Beacham, 1975; Kakehi et al., 2007) and enhances intracellular survival of enteropathogenic E. coli by inhibiting host cell kinases (Berger et al., 1996). In Corynebacterium glutamicum, UshA mediates phosphate acquisition from nucleotides in response to phosphate starvation (Ishige et al., 2003; Rittmann et al., 2005). The related protein ecto-5′-nucleotidase in mammals catalyses degradation of purinergic signalling nucleotides (Strater, 2006).

In order to characterize UshA involvement in flavin processing by S. oneidensis, we generated an ushA deletion strain referred to as ΔushA. The ΔushA mutant strain showed growth similar to wild-type strain MR-1 in minimal medium both under aerobic conditions and under anaerobic conditions with fumarate as an electron acceptor (Fig. 2A). ΔushA also grew similarly to wild-type MR-1 either aerobically or anaerobically in Luria–Bertani (LB) medium (data not shown). At various time points during anaerobic growth on minimal medium, culture samples were removed and centrifuged. Fluorescence of the supernatants was measured in order to assess flavin accumulation (Fig. 2B). While fluorescence of medium from wild-type cultures increased roughly in parallel with cell growth, fluorescence of medium from ΔushA cultures accumulated more slowly and to a level approximately twofold lower than the fluorescence of wild-type MR-1. Qualitatively similar results were observed when cells were grown aerobically or in rich medium (data not shown). Fluorescence of growth medium from ΔushA cells with empty vector [101 ± 7 relative fluorescence units (RFU)] was restored when ushA was expressed on a complementation vector (214 ± 27 RFU), confirming that deletion of the ushA gene results in decreased accumulation of fluorescent material in the culture medium.

Figure 2.

Decreased flavin fluorescence in cultures of ΔushA mutant.
A. Optical density of wild-type MR-1 (black, closed symbols) and ΔushA (red, open symbols) grown aerobically in SBM with 20 mM lactate as carbon source (inline image) or anaerobically in SBM with 20 mM lactate and 40 mM fumarate as electron acceptor (●). Mean ± standard error of the mean (SEM) of triplicate cultures.
B. Fluorescence intensity in RFU of supernatants from cultures of wild-type MR-1 (●) and ΔushA (○) grown anaerobically as in (A). Time indicates the length of time cells were in culture before the sample was removed. Mean ± SEM of triplicate cultures.

ΔushA strains secrete FAD instead of FMN or riboflavin

The decrease in supernatant fluorescence observed in Fig. 2B may be explained not by a decrease in total flavins, but rather a shift in the flavins present. The quantum yield of FAD is lower than that of FMN or riboflavin (Weber, 1950; Rhee et al., 2009). In order to accurately measure the concentrations of various flavins, we analysed supernatants from wild-type and ΔushA mutant cultures by high-performance liquid chromatography (HPLC) to quantify riboflavin, FMN and FAD (Table 1). Wild-type cultures grown anaerobically for 24 h in minimal medium accumulated approximately 0.3 µM total flavins, with approximately 65% as FMN and 35% as riboflavin. ΔushA culture supernatants accumulated similar total concentrations of flavins; however, approximately 75% was in the form of FAD (Table 1). Complementation of the ushA deletion strain with a wild-type copy of ushA returned the flavin production profile to match the wild-type strain (Table 1). The increased proportion of FAD in supernatants of mutant cultures suggests that ΔushA mutants primarily secrete FAD rather than FMN or riboflavin.

Table 1.  Flavin quantitation in supernatants of ΔushA strain and complemented strains.
 Concentration (µM) in culture supernatantaPercentage of total flavins in culture supernatant
StrainFADFMNRFFADFMNRF
  • a. 

    Mean ± SEM concentration measured from culture supernatants of the indicated S. oneidensis strains. Data were obtained from triplicate cultures and are representative of three experiments.

MR-10 ± 00.20 ± 0.010.11 ± 0.0060 ± 064.3 ± 0.835.7 ± 0.8
ΔushA0.27 ± 0.010.07 ± 0.0010.025 ± 0.00173.1 ± 1.120.2 ± 0.76.7 ± 0.4
MR-1 + pBBR1MCS-30 ± 00.21 ± 0.0060.05 ± 0.0020 ± 081.8 ± 0.618.2 ± 0.6
ΔushA + pBBR1MCS-30.28 ± 0.0040.06 ± 0.0010.01 ± 0.00080.1 ± 0.516.4 ± 0.53.6 ± 0.03
MR-1 + pSO2001MCS-30 ± 00.19 ± 0.0150.07 ± 0.0040 ± 073.1 ± 0.426.9 ± 0.4
ΔushA + pSO2001MCS-30 ± 00.20 ± 0.0060.07 ± 0.0020 ± 072.9 ± 0.227.1 ± 0.2

FAD release could indicate a deleterious effect of ushA mutation on cell membrane integrity. However, wild-type and ΔushA cultures grew at similar rates, as measured by OD600 (Fig. 2A). Additionally, the number of colony-forming units from mid-log cultures was not significantly different for wild-type (1.06 × 108 ± 1.1 × 107) versus ΔushA (1.67 × 108 ± 4.2 × 107), indicating that the ushA mutation does not decrease cell viability. Membrane integrity of the ΔushA strain was also assessed using a live/dead stain in which the membrane-permeable green dye SYTO9 stains all cells, while the membrane-impermeable red stain propidium iodide stains only cells with compromised membranes. The ratios of green to red fluorescence were identical (12.9 ± 0.4 vs. 12.7 ± 0.2) in samples from exponentially growing wild-type and ΔushA cultures. This result indicates that the accumulation of FAD in S. oneidensisΔushA cultures cannot be attributed to membrane damage in cells lacking UshA, consistent with a specific mechanism of FAD secretion.

Shewanella oneidensis UshA hydrolyses FAD to FMN

The structure of FAD, consisting of an ADP nucleotide with 5′-linkage to a riboflavin moiety (Fig. 1), is reminiscent of UDP-sugars and CDP-alcohols that have been reported to be substrates of E. coli UshA (Glaser et al., 1967; Neu, 1967; Alves-Pereira et al., 2008). However, despite this similarity in substrate structure, E. coli UshA exhibits very low FAD hydrolysis activity (Alves-Pereira et al., 2008). In order to test whether S. oneidensis UshA has the ability to catalyse hydrolysis of FAD, we added sonicated extracts of wild-type and ΔushA cells to solutions of FAD. Because FAD is less fluorescent than FMN or riboflavin (Weber, 1950), hydrolysis of FAD into FMN or riboflavin results in more intense fluorescence. Increases in fluorescence intensity were monitored over time as an indication of FAD hydrolysis (Fig. 3A). When wild-type cell extracts were added to solutions of FAD, fluorescence increased in a protein-dependent manner, consistent with hydrolysis of the FAD into a more highly fluorescent product. In contrast, ΔushA cell extracts had no effect on FAD fluorescence intensity. HPLC analysis indicated that a majority of the reaction product from wild-type cell extracts was FMN, with a small percentage riboflavin (Fig. 3B). The small amounts of FMN and riboflavin detected in reactions with ΔushA cell extracts were consistent with the percentage of impurities in the added FAD substrate and did not change over the course of 140 min (data not shown). The initial linear rate of FAD hydrolysis by wild-type cell extracts was 26.8 ± 0.8 nmol min−1 mg protein−1, while the rate of FAD hydrolysis by ΔushA cell extracts was only 0.088 ± 0.003 nmol min−1 mg protein−1. These data strongly suggest that S. oneidensis UshA is capable of hydrolysing FAD and releasing FMN.

Figure 3.

S. oneidensis UshA hydrolyses FAD to FMN in cell extracts.
A. Sonicated cell extracts (0.06 mg ml−1 total protein) of wild-type MR-1 (●) and ΔushA (○) were added to solutions of FAD. Mean fluorescence intensity (± SEM) is plotted versus time as FAD is hydrolysed to FMN. Data were obtained from triplicate cultures and are representative of four experiments.
B. HPLC measurements of the per cent concentration of each flavin species after the final time point in (A). Samples treated with wild-type cell extracts are in solid; samples treated with ΔushA cell extracts are patterned.

UshA activity is localized in the periplasm

Escherichia coli UshA is a soluble protein targeted to the periplasm by an N-terminal signal peptide of 25 amino acids (Neu and Heppel, 1965; Glaser et al., 1967; Burns and Beacham, 1986). A periplasmic localization is also predicted for the S. oneidensis UshA protein by using PSORTb 3.0 (Yu et al., 2010). In order to test the localization of S. oneidensis UshA, we monitored FAD hydrolysis by intact cells. An FAD solution was added to washed cultures of S. oneidensis, and the fluorescence increase was measured over time as an indication of FAD hydrolysis (Fig. 4A). Similarly to cell extracts, intact wild-type S. oneidensis were able to hydrolyse FAD, while ΔushA cells were not. The rate of FAD hydrolysis by intact wild-type cells was approximately 6.5 nmol min−1 mg protein−1, somewhat slower than the rate of catalysis by cell extracts, consistent with FAD needing to cross through OM porins before hydrolysis.

Figure 4.

UshA activity is localized in the periplasm.
A. FAD hydrolysis by intact wild-type MR-1 cells (●) and intact ΔushA cells (○). Mean fluorescence intensity (±SEM) versus time.
B. FAD hydrolysis (mean ± SEM) by cell fractions from wild-type MR-1. Cells were fractionated as in Experimental procedures, and equal amounts of protein from each fraction were added to solutions of FAD. The rate of hydrolysis was calculated from the slope of the initial linear increase in fluorescence intensity, as described in Experimental procedures. Triplicate samples were run from each fraction.

In order to further assess S. oneidensis UshA localization, cells were fractionated into periplasmic, cytoplasmic and total membrane fractions. After normalization of the bulk protein concentration, equal amounts of each fraction were added to FAD, and hydrolysis was monitored by fluorescence increase. The majority of UshA activity was found in the periplasmic fraction (Fig. 4B), consistent with the known periplasmic localization of the E. coli UshA homologue.

Previously, UshA has been detected in supernatants of S. oneidensis cultures, leading to speculation that UshA is secreted (Pinchuk et al., 2008). However, in the same study, cytosolic proteins were also detected in culture supernatants, suggesting that cell lysis may account for detection of UshA in the culture medium. In order to test whether UshA is secreted, we incubated supernatants of mid-log phase S. oneidensis MR-1 cultures with FAD in order to detect UshA activity (Fig. S1). Although supernatants from wild-type MR-1 cultures slowly hydrolysed FAD, supernatants that had been passed through a 0.2 µm filter to remove cells remaining in the supernatant after centrifugation had no activity. From these data we infer that the UshA activity detected in culture supernatants is associated with intact cells and that UshA is not secreted from S. oneidensis at appreciable levels under the conditions tested.

The periplasmic hydrolysis of FAD by UshA explains the repeated failure by us and others to observe FAD accumulation in wild-type Shewanella cultures (Table 1, von Canstein et al., 2008; Marsili et al., 2008). The accumulation of FAD in ΔushA culture supernatants, the ability of S. oneidensis UshA to hydrolyse FAD to FMN, and the periplasmic localization of UshA strongly suggest that FAD, rather than FMN or riboflavin, is primarily secreted into the periplasm and that FMN is released into the culture medium via diffusion through OM pores.

UshA is required for growth of S. oneidensis on mononucleotides

Our results provide evidence that S. oneidensis UshA hydrolyses FAD, producing FMN. Based on known biochemistry of UshA, the AMP resulting from this reaction is likely to be hydrolysed immediately by UshA into adenosine and inorganic phosphate (Glaser et al., 1967). We tested whether Shewanella UshA performs this activity by taking advantage of the ability of E. coli to grow on AMP as a carbon source. It is known that in E. coli UshA is the only periplasmic enzyme capable of AMP hydrolysis (Kakehi et al., 2007), and therefore E. coli ushA mutants will not grow on AMP because the phosphate prohibits transport of this compound. We complemented an E. coliΔushA mutant with S. oneidensis ushA and saw that growth on AMP was restored (Fig. S2), confirming that S. oneidensis UshA is capable of AMP hydrolysis. The ability of UshA to hydrolyse AMP suggests that Shewanella might be able to couple FAD hydrolysis to growth on AMP as a substrate, through UshA.

We first tested whether S. oneidensis can, like E. coli, grow with nucleotides as carbon sources. We measured aerobic growth of S. oneidensis cultures on AMP, GMP or CMP as the sole carbon source (Fig. 5A). Wild-type S. oneidensis is capable of growth on all three nucleotides, with growth on GMP and CMP slower than growth on AMP. This result contrasts with growth of E. coli, which is faster on GMP than on AMP (Kakehi et al., 2007), and is consistent with the ability of S. oneidensis to grow on DNA as sole carbon source (Pinchuk et al., 2008). The S. oneidensis mutant ΔushA was unable to grow using AMP, GMP or CMP (Fig. 5A), suggesting that UshA is the only periplasmic enzyme in S. oneidensis capable of hydrolysing these nucleotides.

Figure 5.

UshA is required for growth of S. oneidensis on AMP, GMP and CMP.
A. Optical density of wild-type MR-1 (closed symbols) and ΔushA (open symbols) grown aerobically in SBM with 10 mM of AMP (inline image), GMP (●) or CMP (◆) as sole carbon source. Mean ± SEM of triplicate cultures.
B. Optical density of wild-type MR-1 (closed symbols) and ΔushA (open symbols) grown aerobically in SBM with 10 mM of adenosine (●), adenine (inline image) or ribose (▴) as carbon source. Mean ± SEM of triplicate cultures.

In E. coli, UshA hydrolyses AMP into adenosine and inorganic phosphate. Adenosine served as a carbon source for S. oneidensis (Fig. 5B), consistent with unpublished observations (referenced in Serres and Riley, 2006; Driscoll et al., 2007). Although unable to grow on AMP, ΔushA mutants grew when provided with adenosine, confirming that adenosine utilization is downstream of UshA activity. However, neither they nor wild-type S. oneidensis were capable of growth on either adenine base (Serres and Riley, 2006) or ribose (Fig. 5B) as sole carbon source, leaving open the question of how adenosine is transported and utilized by S. oneidensis. Complemented ΔushA mutants grew on AMP and adenosine, as expected (Fig. S3). These results confirm that following hydrolysis of secreted FAD, the AMP moiety can be further hydrolysed and the adenosine base salvaged by S. oneidensis.

Processing of FAD by UshA not only makes adenosine available for reuse but also releases inorganic phosphate. We tested whether this phosphate could be used by S. oneidensis and in addition tested the proposal by Pinchuk and colleagues that UshA mediates the ability of S. oneidensis to use DNA as a phosphate source (Pinchuk et al., 2008). Wild-type and ΔushA cultures were grown with DNA or AMP as the sole phosphate source. When tested, wild-type and ΔushA cells grew equally well with DNA as phosphate source (Fig. S4A). ΔushA cultures were also capable of growth with AMP as a phosphate source, at a rate very similar to their growth with inorganic phosphate (Fig. S4B). The ability of ΔushA cultures to use AMP and DNA as a phosphate source indicates that UshA is not required for use of nucleotides as phosphate sources by S. oneidensis. Other nucleotidases expressed under phosphate-limiting conditions may be sufficient to supply the phosphorus needs of ΔushA cells (see Discussion).

Escherichia coli UshA has poor FAD hydrolysis activity

Escherichia coli UshA has been found to hydrolyse FAD with only approximately 1% of the activity of AMP hydrolysis (Alves-Pereira et al., 2008). We verified that E. coli UshA hydrolyses FAD slowly using our fluorescence-based hydrolysis assay by incubating cell extracts from wild-type and ΔushA E. coli strains with FAD. No fluorescence increase was observed with either cell extract (Fig. S5). We also expressed E. coli UshA in S. oneidensisΔushA cells. E. coli ushA was cloned into the vector pBBR1MCS-2 under control of a lac promoter, ensuring constitutive expression in S. oneidensis. In order to confirm UshA expression, mutant cells complemented with the E. coli ushA plasmid were grown with AMP as the sole carbon source (Fig. 6A). E. coli ushA was able to restore growth on AMP, indicating that it folds and is properly targeted to the periplasm when expressed in S. oneidensis. We next tested whether E. coli ushA is able to complement FAD hydrolysis by S. oneidensisΔushA mutants (Fig. 6B). As expected, ΔushA cell extracts expressing an empty pBBR1MCS-2 vector did not hydrolyse FAD, as monitored by fluorescence increase. Complementation of ΔushA mutants with a vector containing S. oneidensis ushA restored FAD hydrolysis at a rate significantly faster than that of wild-type S. oneidensis expressing endogenous levels of ushA. However, cell extracts of ΔushA complemented with E. coli ushA hydrolysed FAD at a non-negligible, but much slower, rate (Fig. 6B). Additionally, expression of S. oneidensis UshA in an E. coli ushA deletion strain resulted in robust FAD hydrolysis activity in cell extracts while wild-type K12 extracts showed no appreciable activity (Fig. S6). Together, these results suggest that S. oneidensis UshA hydrolyses FAD more readily than does the E. coli homologue of UshA.

Figure 6.

E. coli UshA enables growth of S. oneidensis ushA mutants on AMP but cannot restore FAD hydrolysis.
A. Optical density of wild-type MR-1 with empty vector pBBR1MCS-2 (●), wild-type MR-1 complemented with E. coli ushA (inline image), S. oneidensisΔushA with empty pBBR1MCS-2 (○) and S. oneidensisΔushA complemented with E. coli ushA (□), grown aerobically in SBM with 10 mM AMP as sole carbon source. Mean ± SEM of triplicate cultures.
B. Mean rate (±SEM) of FAD hydrolysis by sonicated cell extracts of S. oneidensisΔushA expressing empty vector (left), S. oneidensis ushA (middle) or E. coli ushA (right). Triplicate samples; representative of two experiments.

Shewanella oneidensis MR-1 can use FAD as an electron shuttle

Previous studies have shown that Shewanella use riboflavin and FMN to mediate reduction of insoluble iron hydroxides and that supplementation of media with riboflavin or FMN enhances iron reduction rates (von Canstein et al., 2008; Marsili et al., 2008; Coursolle et al., 2010). In one study, the ability of FAD to enhance iron reduction was also tested (von Canstein et al., 2008); however, our results indicate that in wild-type cultures, extracellular FAD is quickly converted to FMN (Fig. 3A). Therefore, in order to test whether FAD itself, in the absence of conversion to FMN, can serve as an electron shuttle for extracellular iron reduction, we added exogenous flavins to both wild-type MR-1 and ΔushA mutants cultured in minimal medium with Fe(III)-oxide. Fe(III) reduction was measured by monitoring the formation of Fe(II). In cultures of ΔushA mutants, 10 µM exogenous riboflavin, FMN or FAD equally enhanced the rate of iron reduction compared with the rate when no flavins were added (Fig. 7). When only endogenous levels of flavins were present, wild-type and ΔushA cells reduced iron at identical rates (Fig. 7), also indicating that the FAD secreted by ΔushA cultures is capable of mediating iron reduction. Additionally, we measured rates of flavin reduction by ΔushA cells and found that FAD is reduced by ΔushA cells at a rate similar to riboflavin (data not shown). From these results, we conclude that FAD can serve as an electron shuttle for the reduction of insoluble metals by S. oneidensis.

Figure 7.

FAD enhances iron reduction by S. oneidensis. Mean rate (±SEM) of Fe(III) reduction by wild-type MR-1 (solid), and ΔushA (patterned). Cells were incubated with 5 mM ferrihydrite, and reduction was monitored by measurement of Fe(II) production. The initial rate of iron reduction was measured by fitting the initial linear portion of the curve. Controls containing no cells did not reduce iron (data not shown).

Discussion

Although Shewanella rely on extracellular electron shuttling by flavin molecules to respire insoluble metal substrates, little is known about the mechanism of flavin secretion, release or processing. We have discovered that rather than directly secreting riboflavin and FMN as previously thought, S. oneidensis first transports the cofactor FAD into the periplasmic space, where the 5′-nucleotidase UshA hydrolyses it to FMN, adenosine and phosphate. Our current working hypothesis of Shewanella flavin processing is outlined in Fig. 8. FAD is synthesized from its riboflavin precursor in the cytoplasm and then exported across the inner membrane by an as yet unidentified mechanism. Secreted FAD may be incorporated into periplasmic proteins requiring an FAD cofactor, such as the flavocytochrome FccA. Shewanella frigidimarina strain NCIMB400 possesses a homologue of FccA called Fcc3 and a second isoform called Ifc3, both of which have been biochemically characterized and shown to contain a non-covalently bound FAD cofactor (Pealing et al., 1992; Dobbin et al., 1999), suggesting there are additional flavocytochromes made by Shewanella that are processed in a similar way to FccA. Excess FAD is hydrolysed by UshA into FMN and AMP. UshA further hydrolyses AMP into inorganic phosphate and the nucleoside adenosine, which may then be metabolized by the cell. FMN could then diffuse through OM pores into the extracellular space, where it begins a cycle of reduction by OM cytochromes and oxidation by the terminal electron acceptor. Conversion of FMN to riboflavin, also an effective electron shuttle (von Canstein et al., 2008; Marsili et al., 2008), appears to be due to slow hydrolysis and is not dependent on UshA (data not shown).

Figure 8.

Working hypothesis/model for the role of UshA in periplasmic processing of flavin electron shuttles by S. oneidensis. Following synthesis in the cytoplasm, FAD is secreted across the inner membrane (IM) via an unknown mechanism into the periplasmic space, where it is incorporated into the periplasmic fumarate reductase, FccA. Excess FAD is hydrolysed by UshA into FMN and AMP. AMP is further hydrolysed by UshA into inorganic phosphate and adenosine (Ado), which may be recycled by the cell. FMN is free to diffuse through OM porins (shown as gaps in the OM) into the extracellular medium, where a fraction is spontaneously converted into riboflavin (RF). Flavins (both FMN and RF) can serve as electron shuttles for cycling of electrons between OM members of the Mtr pathway and an insoluble terminal electron acceptor.

In general, riboflavin and FMN have been the primary flavins detected in bacterial cultures (Demain, 1972; von Canstein et al., 2008); when FAD has been detected it has often been assumed to be evidence of lysis. Using an ushA deletion mutant, we have found that healthy, exponentially growing cultures of S. oneidensis primarily secrete FAD, previously undetected because UshA rapidly hydrolyses secreted FAD. Importantly, we were unable to detect any differences in membrane integrity between wild-type and ushA mutant cells, suggesting there is a specific secretion mechanism for FAD. It is consistent from a consideration of Shewanella physiology that secretion of FAD into the periplasm occurs under certain circumstances. Reduction of the widely available organic compound fumarate by S. oneidensis requires an unusual periplasmic tetraheme c-type cytochrome FccA that contains a non-covalently bound FAD cofactor (Maier et al., 2003). The maturation of c-type cytochromes in Gram-negative bacteria only occurs in the periplasmic space, where a linear polypeptide containing CxxCH motif(s) is processed by a suite of proteins (Thony-Meyer, 2002). We hypothesize that apo-FccA is translocated from the cytoplasm into the periplasm and acquires its haem cofactors via c-type cytochrome maturation before acquiring its FAD cofactor. FccA binds FAD non-covalently (Pealing et al., 1992; Dobbin et al., 1999) and therefore must be folded to retain the cofactor. If our hypothesis regarding the order of cofactor acquisition is correct, FAD must be accessible to FccA in the periplasmic space.

In a contrasting example, the periplasmic methylmenoquinol : fumarate reductase complex of Wolinella succinogenes contains a protein, SdhA (locus tag WS1920), that contains a non-covalently bound FAD cofactor with a binding site similar to FccA from Shewanella (Juhnke et al., 2009). SdhA was shown to be transported across the cytoplasmic membrane by the twin-arginine translocation pathway (TAT) (Juhnke et al., 2009). The TAT pathway translocates folded proteins across the cytoplasmic membrane (Natale et al., 2008) and is consistent with SdhA from W. succinogenes acquiring FAD as the protein is folded in the cytoplasm. Though SdhA appears to bind FAD in a manner similar to Shewanella, the W. succinogenes protein is not a c-type cytochrome, meaning it can fully mature in the cytoplasm before export.

Very little is known about excretion of flavins from bacteria. Although certain strains of riboflavin ‘overproducers’ have been isolated (Demain, 1972), research in bacteria has been focused on biosynthetic pathways and regulation rather than on transport mechanisms. Only two bacterial riboflavin transporters, RibU (Kreneva et al., 2000; Burgess et al., 2006) and RibM (Vitreschak et al., 2002; Grill et al., 2007; Vogl et al., 2007) have been confirmed. Thus far, these known flavin transporters have only been shown to be involved in uptake of flavins, not in their export. In fact, RibU binding assays have shown that it is not capable of transporting FAD (Duurkens et al., 2007). S. oneidensis possesses a putative transporter, SO2713, with approximately 25% sequence identity to RibM, but it is not known whether it is capable of transporting flavins. The S. oneidensis genome also includes a number of putative ABC transporters and multi-drug efflux proteins that could be involved in FAD secretion. Given that there are numerous candidate genes with potential for FAD transport, and given that no transporters were identified in this mutagenesis screen, it seems that Shewanella may have either multiple mechanisms for exporting FAD across the inner membrane, that the export mechanism shares functions essential to cell survival or that our screen was not saturated.

We have shown, both by monitoring fluorescence increases and by HPLC analyses, that S. oneidensis UshA hydrolyses FAD into the flavin derivative FMN. FAD has not previously been reported to be a major substrate for UshA (Alves-Pereira et al., 2008), and we have verified that FAD is not as rapidly hydrolysed by the E. coli homologue. The difference in substrate specificities suggests that Shewanella has adapted to high periplasmic concentrations of FAD. Consistent with the adaptation of Shewanella UshA to recognize FAD, S. oneidensis grows well on AMP, the nucleotide by-product of FAD hydrolysis. Wild-type S. oneidensis grows faster on AMP than on CMP or GMP, suggesting that it may be better adapted to growth on AMP. The constitutive export and hydrolysis of FAD by Shewanella ensure that AMP is constantly available for recycling.

The inorganic phosphate released by AMP hydrolysis is most likely reclaimed by the cell, just as is the adenosine nucleoside. AMP and DNA are capable of supporting growth of phosphate-starved cells, indicating that UshA-dependent hydrolysis of nucleotides could supply the phosphorus needs of the cell; however, UshA is not required. ΔushA mutants grew just as rapidly with equimolar AMP or NaH2PO4 as sole phosphate source. We detected only approximately 10–30 µM free inorganic phosphate in AMP and DNA stock solutions (data not shown), well below the concentration at which phosphate is limiting (Pinchuk et al., 2008), leaving nucleotide hydrolysis as the only source for phosphate in our cultures. Under the conditions of our experiment, cells that have been starved for phosphate may upregulate an alternative nucleotidase or phosphatase capable of liberating phosphate from nucleotides.

The overall benefits of secreting a small amount of FAD to satisfy the enzymatic requirement of FccA (and possibly other enzymes) in the periplasm, coupled with the release of FMN and RF, which act as accelerants for extracellular electron transfer to insoluble substrates, together must outweigh the energetic investment costs incurred, at least under laboratory conditions. It is critical to note that the concentration of flavins quantified in culture supernatants of Shewanella strains is very low, on the order of approximately 250 nM in minimal medium and approximately 1 µM in rich medium (von Canstein et al., 2008; Marsili et al., 2008; Coursolle et al., 2010), and that this small amount is sufficient to dramatically accelerate electron transfer to insoluble electrode surfaces (Baron et al., 2009) and iron oxide minerals (Ross et al., 2009). It has been calculated that Shewanella growing on electrodes can produce ATP at a rate approximately 1000-fold faster than the rate of ATP consumption for riboflavin biosynthesis to concentrations of 250 nM (Marsili et al., 2008). Conversion of riboflavin to FAD requires two additional ATP molecules (Kearney et al., 1979) in addition to any energetic cost that may be associated with FAD export, but the total energetic investment remains quite low compared with the benefit gained. Additionally, we have found that S. oneidensis can grow on the adenosine resulting from UshA hydrolysis of FAD. By recycling a portion of the FAD molecule, S. oneidensis may therefore reduce the metabolic burden of its release.

Shewanella hold great promise for biotechnological applications. In order to optimize metal reduction in technologies employing Shewanella species, it will be critical to understand which electron shuttles are being produced, the mechanisms by which they are produced, and the ecological advantages of each. In this article we have described a previously unknown role for the nucleotidase UshA in processing of flavin electron shuttles by Shewanella, and we provide the first insight into the molecular mechanism of shuttle production. Our results show that FMN and riboflavin are not the primary flavins secreted from Shewanella cells; rather, FAD is first secreted. Major questions remain to be answered, including the identity of the FAD secretion/release pathway(s) and whether FAD secretion is a common feature among bacteria known to accumulate FMN and/or RF in culture supernatents. Answering these questions will constitute a major advance in our understanding of the mechanism of electron shuttle production by Shewanella.

Experimental procedures

Strains and growth conditions

Wild-type S. oneidensis strain MR-1 has been described (Myers and Nealson, 1988). The mutant ushA E. coli strain JW0469 from the Keio Collection has been described (Baba et al., 2006). A complete list of strains and plasmids used in this study can be found in Table S1. Single colonies freshly streaked from a frozen stock were inoculated into 2 ml of LB medium containing the appropriate antibiotic and grown aerobically for 16 h before washing followed by inoculation into minimal medium. Anaerobic cultures were flushed with nitrogen gas for 10 min following inoculation. Shewanella strains were cultured at 30°C with shaking in SBM minimal medium consisting of (per litre) 0.225 g K2HPO4, 0.225 g KH2PO4, 0.46 g NaCl, 0.225 g (NH4)2SO4, 0.117 g MgSO4–7H2O, and 10 mM (aerobic) or 100 mM (anaerobic) HEPES, adjusted to pH 7.2. In addition, 5 ml l−1 of vitamins excluding riboflavin (Balch et al., 1979), 5 ml l−1 trace minerals (Marsili et al., 2008) and 0.01% casamino acids (Difco) were added. Carbon source and electron acceptors, when applicable, were added as indicated. E. coli strains were cultured at 37°C with shaking in M9 minimal medium (Sambrook and Russell, 2001) with the indicated carbon source.

For phosphate starvation experiments, cultures were first starved for 24 h in phosphate-free M1 medium (Pinchuk et al., 2008) supplemented with 10 ml l−1 vitamins and 10 ml l−1 trace minerals as above. A 30 mM lactate was added as a carbon source. Following depletion of phosphate reserves, cells were inoculated into M1 medium containing either 300 µg ml−1 filter-sterilized DNA, 1 mM filter-sterilized AMP or 1 mM NaH2PO4 as sole phosphate source. Cultures were grown at 30°C with shaking. Abiotic, no-carbon and no-phosphate controls were cultured alongside the samples.

Transposon mutagenesis

Transposon mutants were made by mating S. oneidensis MR-1 with E. coli WM3064 containing TnphoA′-1 (Wilmes-Riesenberg and Wanner, 1992). Exconjugants were selected under aerobic conditions on LB plates containing 50 µg ml−1 kanamycin and were inoculated into 96-well plates containing LB with 50 µg ml−1 kanamycin. After overnight growth at room temperature, fluorescence was recorded at 440 nm excitation, 525 nm emission in a Molecular Devices SpectraMax M2 reader. Cultures exhibiting fluorescence intensity below that of wild-type cultures were re-inoculated into fresh medium and screened twice more. Mutants continuing to fluoresce at levels less than wild-type were analysed with arbitrary polymerase chain reaction (PCR) and sequenced to determine the site of transposon insertion.

Deletion and complementation

Shewanella oneidensisΔushA (strain JG1079) was prepared following a described protocol (Saltikov and Newman, 2003). Briefly, regions flanking S. oneidensis ushA were amplified using the following primers: UF SpeI: GGACTAGTCATGGGTTAGGCGATTCT, UR EcoRI: GGGAATTCAGTCAGCACTGCAGTT, DF EcoRI: ccGAATTCCCAGTGGGTGACATTGTG, DR SacI: aaGAGCTCTGACAGACTTGCGGCTAA, and were ligated into the suicide vector pSMV3 (Saltikov and Newman, 2003). After mating of the pSMV3 vector into S. oneidensis MR-1, recombination was confirmed by PCR and by sequencing. Cells were plated on LB supplemented with 5% sucrose to select for loss of sacB on the mutagenesis vector, and deletion of ushA was confirmed by PCR. Complementation was performed by PCR amplification of S. oneidensis ushA (ORF SO2001) using the following primers: SOushA1 NNNTCTAGACCGATAAAACCATCATG, and SOushA2 NNGGGCCCCTGTACTAGTCAGTATCT. Products were purified, digested and ligated into the vectors pBBR1MCS-2 and pBBR1MCS-3 (Kovach et al., 1995). Similar results were obtained using either complementation vector. Inclusion of a 350 bp region upstream of ushA and a reversed orientation in the vector ensured that expression was under control of the endogenous promoter.

Escherichia coliΔushA (strain JG1145) was prepared by transformation of the JW0469 ushA mutant (Baba et al., 2006) with the temperature-sensitive plasmid pCP20 (Cherepanov and Wackernagel, 1995) encoding a FLP recombinase to excise the kanamycin-resistance cassette. Recombinants were verified by sensitivity to kanamycin and by PCR. pCP20 was removed by overnight passage at 43°C, leaving an unmarked ushA deletion. The E. coli ushA complement was prepared by PCR amplification of E. coli ushA (locus tag b0480) and the 20 bp region immediately upstream of the start codon using the following primers: ECushA1 GGTACCATCAGGTCAGGGAGAGAAGT and ECushA2 GAGCTCTTACTGCCAGCTCACCTCA. E. coli ushA was ligated into the vector pBBR1MCS-2 in an orientation ensuring expression under the vector-encoded lac promoter, which is constitutively active in S. oneidensis.

HPLC

High-performance liquid chromatography was performed as follows using a 4.6 mm × 150 mm Eclipse XDB-C18 column with a 5 µm particle size (Agilent Technologies). Twenty-five microlitres of sample was injected onto an LC-10AT liquid chromatograph (Shimadzu) equipped with an SIL-10AF autoinjector. The mobile phase consisted of 20% methanol, 1% glacial acetic acid in water at a flow rate of 1 ml min−1. The column was maintained at 30°C. Flavins were detected with an RF-10AXL fluorescence detector (Shimadzu) at an excitation wavelength of 440 nm and an emission wavelength of 525 nm. Riboflavin (Fisher), FMN and FAD (Sigma) standards were prepared in SBM at concentrations ranging from 0.125 to 10 µM. At a 1 ml min−1 flow rate, FAD eluted at approximately 6 min, FMN 9 min and riboflavin 15 min. Flavin concentrations were calculated by comparing the integrated area of each peak to the area of standard peaks.

Flavin fluorescence measurements

Five hundred microlitre samples of anaerobic cultures were removed at the indicated times and centrifuged to remove cells. Supernatants were transferred to fresh eppendorf tubes and frozen at −20°C until analysed. For fluorescence measurements, 300 µl of each supernatant sample was transferred to a clear 96-well plate and read in a Molecular Devices SpectraMax M2 plate reader at 440 nm excitation, 525 nm emission.

FAD hydrolysis and phosphate assays

Phosphate concentration was assayed using a Colorimetric Phosphate Assay kit (Abcam) in a 96-well plate format, according to the manufacturer's instructions. For FAD hydrolysis assays, 10 ml overnight cultures of S. oneidensis or E. coli in LB were washed once with SBM and resuspended in 500 µl SBM. Before sonication, 30 µl were removed, diluted in SBM and set on ice to be used as intact cells. The remainder of each culture was sonicated in an ice-water bath with 20 pulses of approximately 2 s each using a Sonic Dismembrator Model 60 (Fisher) set at power level 10. Sonicated samples were centrifuged at 10 000 r.p.m., 3 min, and supernatants were stored on ice until use. Protein concentrations in each cell extract were measured with a bicinchoninic acid (BCA) protein assay kit (Pierce) according to the manufacturer's instructions. Cell extracts were diluted in SBM to the same final protein concentration before being aliquotted into 96-well plates alongside intact cell samples. Plates were then pre-warmed to 30°C before addition of warmed FAD to a final concentration of 80 µM. Fluorescence acquisition was begun immediately and at 5 min intervals at 30°C. Plates were shaken for 5 s before each fluorescence acquisition. The rate of FAD hydrolysis was determined by fitting the initial linear portion of each curve and assuming that 1 mol FMN was produced per mol FAD consumed. Based on measurements of FMN standards, 1 nmol FMN was assumed to be equivalent to 770 RFU. Following the final time point, samples were heated at 95–100°C for 10 min to stop enzyme activity before storage at −20°C. Samples were thawed and diluted 1:10 before analysis by HPLC. Standards were heated to verify that heating of samples did not alter flavin concentrations.

Subcellular fractionation

Periplasmic, cytoplasmic and total membrane fractions were separated by the method of Kaufmann and Lovley (Kaufmann and Lovley, 2001). Following separation, total protein concentrations in each fraction were measured with a BCA assay. Fractions were stored overnight at 4°C before use in FAD and AMP hydrolysis assays.

Live/dead staining

Mid-log phase cultures of anaerobically grown Shewanella were stained with a BacLight Live/Dead Stain (Invitrogen), according to the manufacturer's instructions. Fluorescence of triplicate samples of stained cells was measured in a 96-well plate fluorescence reader (Molecular Devices). Fluorescence emission was obtained in 10 nm increments from 500 to 700 nm at 470 nm excitation. The green to red fluorescence ratio was calculated from the integrated fluorescence intensities between 510–540 nm (green) and 620–650 nm (red).

Iron and flavin reduction assays

Overnight cultures of S. oneidensis in LB were washed once with SBM, and then resuspended in SBM to an OD of 1.3. Thirty microlitres of each cell suspension was added to 270 µl SBM containing 20 mM lactate, vitamins, minerals, 5 mM ferrihydrite (iron oxide), and 10 µM riboflavin, FMN or FAD in 96-well plates. Plates were incubated at room temperature in a GasPak System anaerobic Petri dish holder that was flushed with nitrogen gas for 15 min between each time point. At each time point, 50 µl of 5M hydrochloric acid (HCl) was added to stabilize Fe(II). Thirty microlitres was taken from this well and diluted 1:10 into 0.5 M HCl to yield concentrations within the range of standard curves. Fifty microlitres of the diluted sample was mixed with 300 µl ferrozine reagent (Stookey, 1970), consisting of (per litre) 2 g ferrozine, 23.8 g HEPES, pH 7.0, and the absorbance was read at 562 nm. Standard curves were made as described (Coursolle et al., 2010). Flavin reduction assays were performed in 96-well plates, and reduction was monitored through the decrease of flavin fluorescence over time, as described (Coursolle et al., 2010).

Acknowledgements

This work was supported by the Office of Naval Research by a grant awarded to J.A.G. (N000140810166). The authors thank the Gralnick Lab and Daniel Bond (University of Minnesota) for helpful discussions, Michael Torchia and Wing Lam for assistance with the mutagenesis screen, Dan Coursolle for measuring the reduction rate of FAD, Adam Spaulding for assistance in constructing the initial ushA complementation vector and Kristopher Hunt for measurements of inorganic phosphate.

Ancillary