Positive regulation of the Shewanella oneidensis OmpS38, a major porin facilitating anaerobic respiration, by Crp and Fur

Major porins are among the most abundant proteins embedded in the outer membrane (OM) of Gram-negative bacteria, playing crucial roles in maintenance of membrane structural integrity and OM permeability. Although many OM proteins (especially c-type cytochromes) in Shewanella oneidensis, a research model for respiratory versatility, have been extensively studied, physiological significance of major porins remains largely unexplored. In this study, we show that OmpS38 and OmpA are two major porins, neither of which is responsive to changes in osmolarity or contributes to the intrinsic resistance to β-lactam antibiotics. However, OmpS38 but not OmpA is largely involved in respiration of non-oxygen electron acceptors. We then provide evidence that expression of ompS38 is transcribed from two promoters, the major of which is favored under anaerobic conditions while the other appears constitutive. The major promoter is under the direct control of Crp, the master regulator dictating respiration. As a result, the increase in the level of OmpS38 correlates with an elevated activity in Crp under anaerobic conditions. In addition, we show that the activity of the major promoter is also affected by Fur, presumably indirectly, the transcription factor for iron-dependent gene expression.

transcriptionally regulate the differential expression of major porin genes in response to environmental disturbances including Fnr (fumarate nitrate regulator), Crp (cyclic-AMP receptor protein), Lrp (amino acid metabolism regulator), Fur (iron homeostasis regulator), and CpxA-CpxR (cell-envelope stress regulator, to name a few 16 . In addition, two small regulatory RNAs, MicF and MicC, significantly affect OmpF and OmpC production at the translational level 17 . Shewanella, a group of facultative anaerobic γ -proteobacteria widely distributed in nature have received intensive attentions owing to their remarkably diverse respiratory capacities and the potential for environmental remediation and microbial fuel cell, which largely rely on extracellular electron transfer [18][19][20] . Two mechanisms are adopted by Shewanella to deliver electrons generated intracellularly to electron acceptors (EA) in the extracellular space: direct transfer by redox c-type cytochromes on the OM and indirect means mediated by electron shuttles, such as riboflavin [21][22][23] . While the OM integrity could be critical to the former, the OM permeability is important to the latter. In addition, the genus is regarded as emerging human pathogens and a reservoir for antibiotics resistance given that a number of β -lactamases and Qnr-type determinants have been isolated [24][25][26][27][28] . Hence, insights into major porins are needed because of their abundance in the OM and role in transport of hydrophilic molecules, such as electron shuttles and some antibiotics.
In S. oneidensis MR-1, the most extensively studied Shewanella strain, a large number of porins have been predicted in the genome annotation 29 . No known biological function has been demonstrated for these putative porins except for SO_3896, which is found to have a role in anaerobic respiration and its promoter is one of the most robust 12,30 . For regulation, the bacterium contains an E. coli EnvZ-OmpR analogue, but it has diverged considerably from its E. coli counterpart despite its connection with osmoregulation 13,31 . In this study we first made efforts to identify and functionally characterize the major porins in S. oneidensis. Our results demonstrate that two porins, SO_3896 and SO_3545, named as OmpS38 and OmpA respectively, stand out as the most abundant ones. Interestingly, neither is notably responsive to changes in osmolarity or required for transport of β -lactam antibiotics into the periplasm. Further investigation reveals that OmpS38 is generally important in respiration of non-oxygen EAs and that the enhanced expression during anaerobic respiration largely relies on Crp activation.

Materials and Methods
Bacterial strains, plasmids, and culture conditions. A list of all bacterial strains and plasmids used in this study is given in Table 1. All chemicals were acquired from Sigma Co. (Shanghai, China) unless otherwise noted. Information for primers used in this study is available upon request. For genetic manipulation, E. coli and S. oneidensis strains under aerobic conditions were grown in Lysogeny Broth (LB, Difco, Detroit, MI) medium (tryptone, 10 g/L; yeast extract, 5 g/L; and NaCl, 5 g/L) at 37 and 30 °C, respectively. When needed, the growth medium was supplemented with chemicals at the following concentrations: 2, 6-diaminopimelic acid (DAP), 0.3 mM; ampicillin sodium, 50 μ g/ml; kanamycin sulfate, 50 μ g/ml; and gentamycin sulfate; 15 μ g/ml.
Physiological characterization of S. oneidensis strains. MS defined medium with 30 mM sodium lactate as the carbon source was used for growth and respiration characterization as described previously 32 . For anaerobic respiration, electron acceptors studied in this study included nitrate (3 mM), fumarate (20 mM), trimethylamine N-oxide (TMAO, 20 mM), ferric citrate (10 mM), and goethite (FeO(OH), 10 mM). Growth of S. oneidensis strains under aerobic or anaerobic conditions was determined by recording optical densities of cultures at 600 nm (OD 600 ). For aerobic growth, fresh medium was inoculated to ~0.01 of OD 600 with overnight cultures grown from a single colony. For anaerobic growth, cultures of mid-log phase (~0.2 of OD 600 , the same through the study) grown aerobically were centrifuged, washed with fresh medium twice, purged in nitrogen and suspended in fresh medium to ~0.01 of OD 600 in an anaerobic glove box. To cultivate ∆crp under anaerobic conditions, TMAO was used as EA.
Mutagenesis and genetic complementation. S. oneidensis in-frame deletion strains were constructed using the att-based Fusion PCR method 33 . In brief, two fragments flanking the target gene were generated by PCR with primers containing attB and the gene-specific sequence, and then joined by the second round PCR. The fusion fragments were introduced into pHGM01 by site-specific recombination using the BP Clonase (Invitrogen) according to the manufacturer's instruction. The resulting mutagenesis vectors were transformed into E. coli WM3064 and then transferred into relevant S. oneidensis strains via conjugation. Integration of the mutagenesis constructs into the chromosome were selected by gentamycin resistance and confirmed by PCR. Verified transconjugants were grown in NaCl-less LB broth and plated on LB supplemented with 10% sucrose. Gentamycin-sensitive and sucrose-resistant colonies were screened by PCR for the intended deletion. The mutants were then verified by sequencing.
Mutants from previous studies had been successfully complemented by using either multiple-copy or single-copy (integrative) vectors (Table 1). In this study, the same complementation constructs were used and similar results were obtained as indicated in the relevant figure legends. For newly constructed mutants, plasmids pHG101 and pHG102 were utilized for genetic complementation 34 . For genes adjacent to their promoter, a fragment containing the coding sequence and its native promoter was generated by PCR and cloned into pHG101. For others, the coding sequence was amplified and inserted into the multiple-cloning site (MCS) of pHG102 under the control of the arcA promoter, which is constitutively Extraction and analysis of OM proteins. Envelope fractions were prepared according to an established method with modifications 13,36 . Briefly, cultures of log phase were normalized to the lowest OD to allow for comparison of OM protein quantities across strain backgrounds. Ten ml normalized culture was pelleted, washed once in 30 mM Tris-HCl (pH 8.1), and pelleted again. Cell pellets were then resuspended in 500 μ l 20% sucrose in 30 mM Tris-HCl buffer (pH 8.1), followed by the addition of 20 μ L of 20 mg/ml lysozyme in 0.1 mM Ethylenediaminetetraacetic acid (EDTA) (pH 7.3), and incubated on ice for 30 min. Following lysozyme treatment, 1 ml of 3 mM ETDA (pH 7.3) was added and the resulting extract was disrupted with a single 20 s pulse using a microtip sonicator (Haishu Instrument, Ningbo, China). A 1.5 ml fraction of the extract was then centrifuged at 16,000 g for 60 min. Envelope fractions were collected as centrifuged precipitate and resuspended in 30 μ l of Laemmli SDS sample buffer. After boiled for 5 min, 15 ul samples were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE, 10% acrylamide, 4 M urea, 1% SDS). Tandem mass spectrometry (MS/MS) analysis of major OM proteins was carried out essentially the same as described before 37 . Antibiotic and nitrite susceptibility assays. The assays were carried out as previously described 25,38 .
In brief, cells of mid-log phase were adjusted to approximately 10 9 cfu/ml with fresh LB and followed by 10-fold serial dilutions. Five μ l of each dilution was spotted onto LB plates containing antibiotics or nitrite of varying concentrations. The plates were incubated for 24 h or longer before being read. For each strain, experiment was performed in triplicate and repeated independently at least three times.
Chemical assays. Mid-log phase cultures were used to inoculate Bis-Tris propane (BTP) buffer (pH 5.8) with goethite as sole EA to ~0.1 of OD 600 and incubated under anaerobic condition. Concentrations of Fe (II) of cultures were quantified using ferrozine reagent spectrophotometrically at 562 nm 39  curves were made using ferrous sulphate dissolved in 0.5 mol/L hydrochloric acid. Concentrations of nitrite in culture supernatants were measured by a modified Griess assay 40 .

Identification of the SO_3896 transcriptional start sites. S. oneidensis cells were grown in LB to
the mid-log phase, collected by centrifugation, and applied to RNA extraction using the RNeasy minikit (Qiagen, Shanghai) as described before 41 . RNA was quantified by using a NanoVue spectrophotometry (GE healthcare). The transcriptional start sites of SO_3896 were determined using 5′ rapid amplification of cDNA ends (RACE) according to the manufacturer's instruction (Invitrogen, Shanghai).
Electrophoretic motility shift assay (EMSA). Expression and purification of S. oneidensis Crp protein as His-tagged fusion converted from the entire cloneset of S. oneidensis ORFs was described before [42][43] . The probes used for EMSA were prepared by PCR with 33 P end-labeled primers. The binding reaction was performed with ~25-50 fmol (~2-5 nM) of labeled probes and various amounts of protein with or without 10 μ M cAMP in 12 μ l binding buffer containing 100 mM Tris/HCl (pH 7.4), 20 mM KCl, 10 mM MgCl 2 , 2 mM DTT, 0.2 μ g/μ L poly (dI ·dC), and 10% glycerol at 15 °C for 60 minutes and resolved on pre-run 4.8% polyacrylamide native gels 44 . The band shifts were visualized by autoradiography.
B1H assay. Bacterial one-hydrid (B1H) system was used to investigate DNA-protein interaction in vivo in E. coli cells 45 . Briefly, plasmid constructs were created by cloning the bait DNA and target Crp into the pBXcmT and pTRG vectors, respectively, and verified by sequencing. The resultant plasmids were used to co-transform BacterioMatch II Validation Reporter Competent Cells on M9 salt agar plates containing 25 mg/ml chloramphenicol and 12.5 mg/ml tetracycline with or without 3-AT. A pair of pBXcmT-P cyd /pTRG-Crp were used as positive control, and a 300 bp DNA fragment of the 16s rRNA gene promoter (pBXcmT-P 16s ) was used as negative control [46][47] . The plates were incubated for 24 h and then moved to room temperature for an additional 16 h (colonies indicating positive interaction usually appeared between 18 and 24 h). The positive interactions were confirmed by streaking colonies on plates containing both 3-AT and streptomycin (12.5 mg/ml).
Expression assay. Expression assay was performed using a single-copy integrative lacZ reporter system as described previously 48 . DNA fragments of interest were generated by PCR and cloned into the reporter vector pHGEI01, verified by sequencing, and the correct plasmid was then transferred into relevant S. oneidensis strains by conjugation. Once transferred into S. oneidensis strains, pHGEI01 containing promoter of interest integrates into the chromosome and the antibiotic marker is subsequently removed by an established approach 38 . Cells grown to the mid-log phase under experimental settings described in relevant figure legends were collected, four biological replicates in total, and levels of gene expression were determined by measuring β -Galactosidase activity essentially the same as before 38,48 . Other analyses. DNA and protein sequence similarity searches were performed using the BLAST program. Sequences of proteins of interest for alignments were obtained from Genbank. Alignments were performed using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo). Signal peptide was predicted using the SignalP program 49 . The relative intensity of specific protein signals was measured using ImageJ 50 . Experimental values are subjected to statistical analyses and presented as means ± SD (standard deviation). Student's t-test was performed for pairwise comparisons of groups.
Other analyses. DNA and protein sequence similarity searches were performed using the BLAST program. Sequences of proteins of interest for alignments were obtained from Genbank. Alignments were performed using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo). Signal peptide was predicted using the SignalP program 49 . The relative intensity of specific protein signals was measured using ImageJ 50 . Experimental values are subjected to statistical analyses and presented as means ± SD (standard deviation). Student's t-test was performed for pairwise comparisons of groups.

Results
Genes SO_3545 and SO_3896 encode two major porins in S. oneidensis. S. oneidensis is predicted to be equipped with a large number of OM β -barrel proteins 29 (Table S1). As a first step to assess biological processes in which porins are involved, we determined the profile and abundance of S. oneidensis OM β -barrel proteins. OM fractions were prepared from mid-log phase cells, proteins in these fractions were resolved by SDS-PAGE, and visible bands were excised for protein determination by MS/ MS analysis. The identified proteins included five predicted OM β -barrel proteins, SO_1215, SO_1821, SO_3099, SO_3545, and SO_3896 (Fig. 1). The identities of these proteins were confirmed with in-frame deletion mutants for their encoding genes (Fig. 1). SO_3099 is a homologue of E. coli FadL, which forms a ligand-gated channel for passive diffusion of long-chain fatty acids across the OM 51 . Given that this channel mediates hydrophobic molecules, it is unlikely to be involved in transport of hydrophilic molecules as porins. Apparently, two proteins migrated close to the 40 kDa marker were most abundant. One is SO_3545 (37.4 kDa after signal peptide removal), an OmpA-like porin, which migrated at ~39 kDa (Fig. 1). The discrepancy between migration and residue-based molecular weight is common to OmpA-like porins because of posttranslational modification (PTM) [52][53] . Sequence analysis revealed that SO_3545 contains a segment 'LSLGVSYRFGQGE' , which resembles the highly conserved PTM motif established in OmpA porins of other bacteria (Fig. S1). Based on these similarities, we named SO_3545 as OmpA in S. oneidensis. The other is SO_3896, which was recently named as OmpS38 according to its molecular weight after signal peptide removal (37.6 kDa) deduced from the sequence 30 . The last two identified OM β -barrel proteins, SO_1215 and SO_1821, are produced significantly less than OmpA or OmpS38. Currently, little is known about these two proteins, either biochemically or physiologically.
Major porins may not be subjected to regulation by the EnvZ-OmpR system. There are several mechanisms adopted by bacterial cells to cope with changes in osmolarity of the environment, one of which is through the porin proteins 1 . We thus examined production of the major porins under various osmolarities. Mid-log phase cultures were adjusted to similar cell densities, the same volume of which (containing cells of the similar number) was subjected to extraction of OM proteins. As shown in Fig. 2A, no significant difference in levels of either OmpS38 or OmpA was visualized, under conditions of both low (0 to 0.5% NaCl) and high (5% NaCl and 5 to 20% sucrose) osmolarities, suggesting that neither OmpS38 nor OmpA responds to changes in osmolarity in a significant manner. To further confirm this, we evaluated the influence of the EnvZ-OmpR system on production of OmpS38 and OmpA. As shown in Fig. 2B, levels of OmpS38 and OmpA were comparable in the ∆envZ∆ompR strain in comparison with the wild-type under all tested conditions. These data manifest that either OmpS38 or OmpA may not be a porin under regulation of the EnvZ-OmpR system.
Major porins appear not important in diffusion of β-lactam antibiotics in S. oneidensis. As nonspecific channels, porins form the pathway through which small, hydrophilic antibiotics, such as β -lactams, diffuse across the OM of Gram-negative bacteria 6,54 . As a result, loss of major classic and OmpA-like porins generally reduces the susceptibility to certain β -lactam antibiotics 55 . To test whether OmpS38 and OmpA porins play a significant role in diffusion of β -lactam antibiotics, we assayed the sensitivities of porin mutants to three different β -lactam antibiotics, ampicillin, celfsulodin, and aztreonam, in comparison with the wild-type (Fig. 3). Our previous study revealed that susceptibilities of S. oneidensis to these three antibiotics differ significantly [25][26][27] . Strains lacking the ompS38 gene were  indistinguishable from the wild-type in susceptibilities to all test antibiotics, indicating that OmpS38 is dispensable to diffusion of β -lactam antibiotics. Loss of OmpA resulted in a similar phenotype in general, except for slightly increased sensitivity to cefsulodin. Moreover, we constructed a strain (∆double) lacking both major porins and determined its susceptibility to these antibiotics. We found that the mutant displayed slightly increased sensitivity to ampicillin and cefsulodin (Fig. 3). One of possible explanations would be that loss of these two porins all together may impact the integrity of the OM, leading to faster diffusion of these antibiotics. To test this notion, we evaluated the susceptibility of these mutants to SDS, a detergent that primarily impairs the OM 36 . The results demonstrated that the lack of OmpA or both porins enhanced SDS sensitivity modestly (Fig. 3), implicating the role of OmpA in maintenance of the OM structure. In contrast, OmpS38 appeared to be negligible. Given that the loss of both major porins does not reduce susceptibilities to tested antibiotics, it is unlikely that these two major porins form the critical pathways through which β -lactams diffuse into the periplasm in S. oneidensis.
OmpS38 generally affects anaerobic respiration. OmpS38 has been previously implicated a role in respiration of Fe(III), nitrate, and fumarate 8 . To confirm this and to test whether OmpA also has a role in these processes, we compared abilities of the wild-type, ompS38, and ompA mutant strains to reduce various EAs, including oxygen, fumarate, TMAO, nitrate, ferric citrate, and insoluble FeO(OH). With all tested EAs, the ∆ompA strain behaved the same as the wild-type (Figs 4 and S2), eliminating the possibility that OmpA is involved in respiration of these EAs. In the case of the ∆ompS38 strain, in comparison with the wild-type no difference was observed under aerobic condition (Fig. S2), indicating that OmpS38 has no impact on respiration of oxygen. When fumarate, TMAO, nitrate, ferric citrate, or FeO(OH) was respired, however, the ∆ompS38 strain exhibited a consistent defect, a significant lag in respiration or growth supported by respiration (Figs 4A and S2). These data suggest that the loss of OmpS38 impedes respiration of non-oxygen EAs, at least at the early phase. Notably, the ompS38 mutation impaired fumarate respiration also by reducing respiration rate, leading to a ~1.8-fold difference in generation time, implicating a specific mechanism other than that underlying the growth lag. Investigation into this mechanism is currently underway and will be described as part of a separate report.
To examine response of OmpS38 to these EAs, we used SDS-PAGE to assay its production. Compared to cells cultivated aerobically, those grown on fumarate, TMAO, and ferric citrate had significant enhanced production levels, which were indistinguishable from one another (Fig. 4B). Moreover, addition of these EAs to aerobic cultures did not increase OmpS38 production, indicating that oxygen, presumably simply as an EA, overwhelms the effect of these EAs. Overall, these data suggest that OmpS38 affects anaerobic respiration but may not specifically respond to non-oxygen EAs, at least in the context of expression.
Crp is involved in regulation of ompS38. Augmented OmpS38 production under anaerobic conditions implies an association with cellular anaerobic respiration but not with respiration of oxygen EAs. In S. oneidensis, Crp is the most important global regulator mediating adaptation of metabolic modes in response to the availability of electron acceptors 35,38,43,56 . Coincidently, a Crp-binding site was previously predicted to reside within the sequence upstream of the ompS38 gene, from − 144 to − 123 relative to the translation initiation code (Fig. 5A), implying that Crp may directly regulate ompS38 transcription 35,57 . To test this notion, we measured the activity of P ompS38 in the crp − background using a markerless integrative lacZ reporter 48 . A fragment of 414 bp upstream of the coding sequence was amplified and placed in front of the full length E. coli lacZ gene within pHGEI01, and the resulting vector was introduced into the relevant strains. After chromosome integration and the antibiotic marker removal, activities of the promoter (P ompS38 ) in a single copy were assayed (Fig. 5B). The P ompS38 activity in the ∆crp strain was substantially reduced, approximately 20% relative to that in the wild-type, indicating that Crp is crucial for ompS38 expression. We then tested whether the ompS38 gene is also under control of Fnr and ArcA, two other global regulators established for mediating respiration in many bacteria 58 . However, the activity levels of P ompS38 were hardly altered in either ∆arcA or ∆fnr mutant compared with the wild-type, suggesting that neither has a significant role in regulation of the ompS38 gene. These observations were supported by SDS-PAGE analysis of major porins (Fig. 5C). In the absence of either ArcA or Fnr, production levels of OmpS38 were statistically insignificant in comparison with the wild-type. In contrast, Crp was nearly essential for OmpS38 production as the porin was barely visible in gels stained with Coomassie Brilliant Blue. To confirm whether the Crp regulation is realized by direct interaction with the sequence upstream of the ompS38 gene, we used EMSA and B1H assays. The reaction mixture for EMSA included DNA fragments upstream of the ompS38 operon covering predicted Crp-binding sites, purified Crp proteins, and 10 μ M cAMP, a condition proven to give the best result 43 . A similar-length fragment upstream of gyrB (encoding DNA gyrase subunit B) was included in the assay as negative control according to the method established previously 44 . A gel shift band was observed with the ompS38 upstream sequences when 0.5 μ M of Crp was added to the reaction mixture and the intensity of the shifted band became stronger with 2 μ M of Crp (Fig. 5D). In contrast, the fragments upstream of the gyrB and ompA genes were not affected by Crp. Consistently, in the B1H assay the strong interaction was detected between Crp with the ompS38 segment covering the predicted Crp-binding motif (Table 2). Collectively, these data converge on the idea that Crp is the regulator controlling expression of the ompS38 gene in a direct manner.
Transcription of ompS38 is under control of two promoters. While loss of Crp results in substantially reduced production of OmpS38, the ompS38 gene is still evidently expressed as revealed by both the lacZ-reporter and SDS-PAGE. To investigate the source for this expression, we constructed a series of mutants carrying various truncations within the region upstream of the ompS38 coding sequence, including ∆414-258, ∆414-203, ∆414-101, and ∆414-55 (for reference, ∆414-258 represents the mutant lacking base-pairs from − 414 to − 259 relative to the ompS38 gene) (Fig. 6A). The abundance of OmpS38 produced in these mutants was analyzed with SDS-PAGE (Fig. 6B). As expected, ∆414-101 and ∆414-55, in which the Crp-binding motif remains, produced OmpS38 at levels similar to that from the ∆crp strain. While the protein was visible in the wild-type, ∆414-258, and ∆414-203, the abundance in the former two apparently exceeded that in ∆414-203, implying that the promoter P 202 (a fragment covering 202 bp upstream of the ompS38 gene) is impaired with respect to activity. To confirm this, we placed these segments in front of the E. coli lacZ gene and measured their activities (Fig. 6A). As shown in Fig. 6C, P 414 (from WT, 414 bp upstream of the ompS38 gene) and P 259 (from mutant ∆414-258) behaved identically in the the wild-type and ∆crp strains, indicating that these two segments are sufficiently long to support full-scale expression. Consistent with SDS-PAGE data in Fig. 6B, the promoter P 202 had significantly reduced activity, approximately 70% relative to P 414 and P 259 when Crp was present. In the absence of Crp, the activity of the ompS38 promoter within all segments tested was reduced to levels that were indistinguishable. Given that the Crp-binding motif remains, the segment probably loses another regulatory element, which is included in P 414 and P 259 . In the case of P 100 and P 54 , while activating effect of Crp mostly vanished, both segments drove the lacZ expression at levels lower than P 414 , P 259 , and P 202 without Crp. The further reduction in activity is probably due to the complete loss of the  Table 2. Bacterial one-hybrid (B1H) assay of Crp with various promoters. a M9 agar + 25 μ g/ml chloramphenicol + 12.5 μ g/ml tetracycline. b a + 5 mM 3-AT. c b + 12.5 μ g/ml streptomycin. Crp-dependent promoter. Nevertheless, the activity of approximately 300 Miller units was still obtained, which was not affected by Crp (Fig. 6C). These data together imply the presence of a weak promoter that is independent of Crp. To provide evidence for the presence of the second ompS38 promoter, we took advantage of our previous finding that the cytochrome bd oxidase confers S. oneidensis nitrite resistance 38 . Given that the promoter for the bd oxidase operon is approximately 200 Miller units, the proposed Crp-independent promoter seems sufficiently strong. We used P 54 to drive expression of the cydX gene, which encodes the third functionally essential subunit of the bd oxidase (in addition to CydA and CydB) 59 . As shown in Fig. 6D, the cydX gene under the control of P 54 restored the nitrite resistance of the ∆cydX strain, a scenario also obtained with P 414 and P 202 . Given that the promoterless control vector failed to do so, it is clear that there is a promoter within 54 bp upstream of the ompS38 gene. The presence of two promoters was then validated by 5′ -RACE, as marked in Fig. 5A. The main start site (Crp-dependent) mapped to nucleotide − 94 and represented 88% of the positive clones that were identified by 5′ -RACE. A second site mapped to nucleotide − 22. In support of these results, we constructed P ∆64 , which covers the segment from − 65 to − 414, and used it to drive ompS38 expression (Fig. 6A). Based on values of protein levels and β -galactosidase activity (Fig. 6B,C), it is apparent that this promoter within the segment is strong, comparable to that within P 414 and P 259 . Altogether, these data conclude that transcription of the ompS38 gene initiates from two sites.
Fur is involved in regulation of ompS38. The difference in ompS38 expression levels between ∆414-203 (P 202 ) and WT (P 414 ) implies that the gene is subjected to additional regulation. In Vibrio cholerae, expression of major porin gene ompT is positively regulated through the direct binding by Fur, the global transcriptional regulator for iron-dependent expression of many genes [60][61] . Although a Fur-binding motif is not identified within the ompS38 upstream sequence by bioinformatics prediction 57 , we made attempts to investigate whether the ompS38 gene is transcriptionally influenced by Fur given that V. cholerae and S. oneidensis are most closely related in phylogeny, sharing extensive regions of similar gene order 29 . To this end, a ∆fur strain was constructed, which was validated by exhibiting a growth defect that was in excellent agreement with the results of a previous study and genetic complementation (data not shown) 62 . The influence of the Fur loss on production of OmpS38 was evident, approximately 55% relative to that in the wild-type, manifesting that the regulator functions as an activator (Fig. 7A). To examine whether the impact of Fur on expression of the ompS38 gene is affected by intracellular iron levels, we assayed OmpS38 levels. Unlike iron in excess (Fig. 4), iron in scarcity reduced production of OmpS38, to approximately 60% with iron-chelator dipyridyl at 50 μ M (Fig. 7A). However, the negative effect of dipyridyl on OmpS38 production appeared additive to that resulting from the loss of Fur (Fig. 7B). In contrast, addition of exogenous iron did not interfere with the OmpS38 production in the absence of Fur. Consistent results were obtained from the lacZ reporter (Fig. 7C). Expression of ompS38 was not affected by the chelator in the ∆crp mutant, indicating that the loss of Crp overwhelms overall regulation. Changes in iron levels by adding dipyridyl but not exogenous iron had evident effect on the expression in the wild-type. At 50 μ M, the chelator induced ∼ 50% reduction. A similar trend was observed in the ∆fur mutant. These data, collectively, suggest a possibility that the chelator and Fur function independently from each other.

Discussion
The purpose of this study was to identify major porins and to explore how they influence the physiology in S. oneidensis. Our previous study had shown that the microorganism has an EnvZ-OmpR TCS, which The assay included the wild-type strain with iron at different levels, which were achieved by adding FeCl 3 or iron-chelator dipyridyl. The lanes shown here were from the same gel, which was given in Fig. S3B. (B) SDS-PAGE analysis of OM proteins in the ∆fur strain under conditions indicated. In both (A,B), relative signal intensity (RSI) of OmpS38 in the indicated samples was estimated using software Image J. The averaged value for the wild-type in each gel was set to 1 as the standard, to which values of other samples were normalized. The values are the mean ± SD (error bars) (n = 4). (C) Expression analysis of P ompS38 in strains under conditions indicated. Che represents chelator dipyridyl. Experiments were conducted independently at least three times and error bars represent S.D. or the representative was presented.
is highly homologous to the E. coli counterpart 31 . Given that the system is conserved in regulation of osmolarity, we anticipated that the most abundant porins of S. oneidensis may share characteristics of E. coli OmpF and OmpC. We were surprised when this was not the case. Among five predicted β -barrel proteins revealed by combining MS/MS and mutational analyses, OmpS38 and OmpA are the most abundant but seemingly not responsive to either changes in osmolarity or the TCS. In E. coli, OmpA is estimated to present at ~100 000 copies per cell, making it a major OM protein crucial to cell structure and integrity 63 . Accordingly, it is not surprising that OmpA assumes the similar task in S. oneidensis. OmpS38 is certain to be a classic porin with 12-22 transmembrane segments as the most closely related proteins to this protein are all such porins, based on sequence similarities 8 . Additionally, the protein contains the conserved membrane core domain and variable loops typical of porins 1 . In addition to non-responsive to changes in osmolarity, the presented data showed that these two major porins are largely dispensable in allowing the entrance of β -lactam antibiotics, a function implicated in both classic and OmpA-like porins in some bacteria 7,55 . With respect to maintenance of the OM structure, OmpA is implicated to some extent whereas the significance of OmpS38 is negligible.
OM components are no doubt important in respiration of non-oxygen EAs, especially insoluble metal species. The OM c-type cytochromes complex MtrABC plays a central role in reduction of extracellular EAs acting not only as the reductase by direct contact but also as the electron station for cycling extracellular electron shuttles 23,64 . But how porins impact anaerobic respiration is unclear. Similar to OmpS38, a major porin OmpJ of ~49 kDa in Geobacter sulfurreducens was found to be essential for respiration of insoluble Fe (III) and Mn (IV) oxides 9 . Given that both OmpS38 and OmpJ lack any apparent electron transfer moieties, it is unlikely that they are involved in the final electron transfer to extracellular EAs. One possibility is that these proteins function as MtrB, which is a non-cytochrome component of the metal reduction complex. By forming a trans-OM pore of 3-4 nm in diameter 65 , MtrB allows electron transporter cytochrome c MtrA to span the OM such that electrons can be transferred from the periplasm to reductase MtrC, which is located on the surface of the OM (Ross et al. 2007). This mechanism for microbe-to-mineral electron transfer is termed as the 'porin-cytochrome' model 66 . In S. oneidensis, equally possible is that OmpS38 promotes indirect electron transfer mediated by electron shuttles such as riboflavin, another mechanism for extracellular electron transfer in this bacterium [21][22] . This notion gains support from a finding that E. coli cells producing the synthetic large pore porin protein are able to efficiently use riboflavin as the electron shuttle, resulting in significant improvement in extracellular electron transfer 11 .
The mechanism by which OmpS38 affects growth on and/or respiration of some soluble EAs, such as fumarate, nitrate, and TMAO, is not clear. Apparently, the proposal used to explain the defect in respiration of insoluble EAs does not work as these soluble EAs are reduced in the periplasm. It is possible that OmpS38 forms the major passage for extracellular small molecules across the OM to enter the periplasm. Although generally regarded to be non-specific, porins are in fact somewhat selective for small molecules, especially charged ones 5,67 . Thus, the OmpS38 channel may be particularly important for fumarate, leading to the substantial defect in growth on the EA in its absence. We are currently testing this possibility.
Although we have little knowledge of the mechanism by which OmpS38 is involved in anaerobic respiration, in the presented study we unraveled why expression of the porin is elevated under anaerobic conditions. In S. oneidensis, Crp, rather than Fnr or the Arc system, serves as the major protein mediating switch between aerobic and anaerobic respiration 35,56 . As activity of Crp relies on the cellular cAMP levels, which is higher under anaerobic conditions, genes under positive control of Crp are usually transcribed at augmented levels in the presence of oxygen 43,68 . This explains that increased expression of OmpS38 is consistently observed from cells grown on all non-oxygen EAs tested. Regulation by Crp on expression of major porins in a direct manner has been reported before in V. cholerae and Yersinia pestis [69][70] . However, in these enteric bacteria Crp functions as a prominent global regulator that controls the expression of an array of genes mainly in response to carbon and energy sources in the environment 71 . How these evolutionally diverse bacteria adopt the same strategy to regulate expression of major porins is of great interest.
OmpS38 is produced at reduced levels in the absence of Fur, resembling major porin OmpT of some V. cholerae strains 61 . Meanwhile, expression of these two porins is significantly reduced when iron is limited, suggesting a similar response to iron levels. However, differences are found. In contrast to that increased OmpT by addition of exogenous iron relies on Fur, the S. oneidensis regulator seems to exert positive regulation by a mechanism that is independent of iron levels. This notion is supported by a previous finding that a large number of S. oneidensis genes involved in energy metabolism are iron-responsive but Fur-independent 62 . Moreover, OmpS38, unlike OmpT, is probably not under direct control of Fur given the missing of a Fur-binding motif (Fur box) within its upstream region 57,72 .
The significance of OmpS38 expression controlled by two independent promoters is not clear. Apparently, transcription occurs from both promoters simultaneously, evidenced in the crp mutant strain in which the additive effect is observed. However, given that the Crp-independent promoter is substantially weaker than the Crp-dependent one, the contribution of the former to overall expression is conceivably minor. Nonetheless, because the Crp-independent promoter is seemingly constitutive, it is thus tempting to speculate that the low levels of the porin might be critical under yet unknown conditions. A goal for future experiments will be to investigate role of these promoters and to specify conditions under which their activity is required and/or altered.