Bacterial strains and plasmids used in this study are listed in Table 1 and sequences of primers used are given in Table S1. All chemicals were acquired from Sigma Co. (Shanghai, China) unless specifically noted. For genetic manipulation, E. coli and S. oneidensis strains under aerobic conditions were grown in Luria-Bertani (LB) medium 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, 50 μg/ml; kanamycin, 50 μg/ml; and gentamycin, 15 μg/ml.

For physiological characterization, both LB and M1-defined medium containing 0.02% (w/v) of vitamin free Casamino Acids and 15 mM lactate as the electron donor were used in this study and consistent results were obtained (Gao et al., 2008a). Anaerobic growth was supported by 20 mM fumarate as the electron acceptor. Fresh medium was inoculated with overnight cultures grown from a single colony by 1:100 dilution, and growth under aerobic and anaerobic conditions was determined by recording the optical density of cultures at 600 nm (OD₆₀₀). For cultures with fatty acid additions, which interfere with OD readings, growth was monitored by photographing colonies on plates. Mid-log-phase (~0.4 of OD₆₀₀, unless mentioned otherwise) cells were properly diluted, plated on solid agar plates containing a paper disk of 6 mm in diameter as the size reference, and incubated at 30°C.

In frame deletion strains were constructed according to the att-based Fusion PCR method described previously (Jin et al., 2013). In brief, two fragments flanking the gene of interest were amplified with primers containing attB and the gene specific sequence, and then joined by a second round of PCR. The fusion fragment was introduced into pHGM01 by site-specific recombination using the BP Clonase (Invitrogen) and maintained in E. coli WM3064. The resulting mutagenesis vector was then transferred from E. coli into S. oneidensis by conjugation. Integration of the mutagenesis construct into the chromosome was selected by gentamycin resistance and confirmed by PCR. Verified trans-conjugants were grown in LB broth in the absence of NaCl and plated on LB supplemented with 10% sucrose. Gentamycin-sensitive and sucrose-resistant colonies were screened by PCR for the intended deletion. The deleted mutants were then verified by sequencing.

Plasmids pHG101 and pHG102 were used in genetic complementation of mutants (Wu et al., 2011). For complementation of genes adjacent to their promoter, a fragment containing the gene of interest and its native promoter was generated by PCR and cloned into pHG101. For the rest genes, the gene of interest was amplified and inserted into MCS of pHG102 under the control of the arcA promoter, which is constitutively active (Gao et al., 2010a). After verified by sequencing, the vectors were introduced into the relevant mutants for phenotypic assays.

Expression of genes of interest was assessed using an integrative lacZ-reporter system (Fu et al., 2014a). Based on the promoter prediction, fragments of ~300 bp covering the promoter sequences were cloned into the reporter vector pHGEI01 to generate transcriptional fusions. The resultant vectors were then verified by sequencing and then transferred into relevant strains by conjugation. To eliminate the antibiotic marker, helper plasmid pBBR-Cre was transferred into the strains carrying a correctly integrated construct (Fu et al., 2013). Mid-log phase cultures were harvested, properly aliquotted, and subjected to β-Galactosidase activity assay as described before (Fu et al., 2014a).

Expression of genes of interest was also assessed using quantitative reverse-transcription PCR (qRT-PCR). Cells of the mid-log phase were harvested by centrifugation and total RNA was isolated using RNeasy Mini Kit (QIAGEN) according to the manufacturer's instructions. The analysis was carried out with an ABI7300 96-well qRT-PCR system (Applied Biosystems) as described previously (Yuan et al., 2011).

To determine fatty acid composition, cultures of the mid-log phase grown in LB medium were collected by centrifugation, properly aliquotted, and subjected to total cellular lipid extraction as described before (Bligh and Dyer, 1959). The fatty acid methyl esters (FAMEs) were prepared by trans-esterification with 0.5 M sodium methoxide in methanol and identified using gas chromatograph-mass spectroscopy (GC-MS) (Focus GC-DSQ II) on a capillary column (30 mm by 0.25 mm in diameter) (Zhang et al., 2002). Helium at 1 ml/min was used as the carrier gas, and the column temperature was programmed to rise by 4°C/min from 140 to 170°C, and then 3.5°C/min from 170 to 240°C for 12.5 min.

B1H system was utilized to investigate DNA-protein interaction in vivo in E. coli cells as described previously (Guo et al., 2009; Jiang et al., 2014). Plasmids were constructed by cloning the “bait” promoter region DNA and “target” regulators into the pBXcmT and pTRG vectors, respectively. After 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-amino-1,2,4-triazole (3-AT).

The cloning of S. oneidensis fabR and fadR for expression and purification has been described previously (Gao et al., 2008b). Soluble His-tagged FadR protein was expressed in E. coli BL21(DE3) induced with 0.3 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 30°C for 4 h. Cells were collected by centrifugation, and resuspended in the lysis buffer (50 mM Tris/HCl, pH 7.5, 200 mM NaCl, 1 mM MgCl₂, 10 mM β-mercaptoethanol, 1 mM PMSF, 5 mg/ml DNase I), and subjected to thorough sonication. The soluble protein was purified using a nickel-ion affinity column according to Ni-NTA purification system manual (Invitrogen).

Biotin-labeled DNA probes were prepared by PCR from S. oneidensis genomic DNA with biotin-labeled primers and EMSA was carried out as described before (Gao et al., 2008a). Briefly, the binding reaction was performed with ~2–5 nM labeled probes and various amount of protein in 12 μl binding buffer containing 100 mM Tris/HCl (pH 7.4), 20 mM KCl, 10 mM MgCl₂, 2 mM DTT, 0.2 μg/μl poly(dI·dC), and 10% glycerol at 15°C for 60 min. Reaction mixtures were immediately resolved on 6% polyacrylamide native gels, blotted onto nylon membranes, and cross-linked with a UV crosslinker. The imaging procedure followed Thermo Scientific's chemiluminescent nucleic acid detection module and was visualized with the UVP Imaging System.

Promoter prediction for genes of interest was performed by using promoter prediction program Neural Network Promoter Prediction (Reese, 2001). For statistical analysis, values are presented as means ± SD (standard deviation). Student's-test was performed for pairwise comparisons of groups.

According to the genome annotation, S. oneidensis has the fabA_So (SO1856) (for differentiation, the same genes/proteins from multiple bacteria are labeled with abbreviated bacterial name in subscript) gene (Heidelberg et al., 2002), whose protein product shares up to 67% of sequence identity with FabA_Ec from E. coli. The gene is perfectly conserved in all sequenced Shewanella (data not shown), indicating that this group of bacteria own a typical anaerobic UFA synthesis pathway. However, genes encoding aerobic desaturases are not annotated. To assess whether S. oneidensis can also produce UFAs via an aerobic pathway, we screened the proteome with three well-studied desaturases essential to the process, Ole1p of S. cerevisiae, DesA of P. aeruginosa (DesA_Pa), and Des of B. subtilis (Stukey et al., 1990; Aguilar et al., 1998; Zhu et al., 2006). BLASTp returned the same single putative homolog against Ole1p and DesA_Pa, SO0197 (E-values 3e-64 and 2e-11, respectively). This protein shares sequence identity of 41% with the N-terminal desaturase domain of Ole1p, which also carries a C-terminal domain that has strong homology to cytochromes b₅ (Mitchell and Martin, 1995). Like Ole1p, SO0197 bears characteristics of membrane-bound desaturases, three trans-membrane domains and four histidine-rich clusters (Figure 1A). These features suggest that S. oneidensis may utilize SO0197 as an aerobic fatty acyl desaturase.

To confirm that FabA_So and SO0197 function in the S. oneidensis UFA synthesis, in-frame deletion strains for their coding genes, individually and in combination, were constructed and assayed for growth under aerobic and anaerobic conditions. Both single knockout strains, ΔfabA_So and ΔSO0197, were obtained on LB plates. The ΔSO0197 strain was indistinguishable from the wild-type with respective to growth in liquid and solid media under aerobic and anaerobic conditions (Figures 1B,C). In contrast, the ΔfabA_So strain not only grew at a reduced rate under aerobic conditions but also required a UFA supplement, oleate of 0.005% (the same concentration was used through this study unless otherwise noted), for anaerobic growth. These growth defects were corrected when a copy of fabA_So was expressed in trans (Figure S1), confirming the essential role of FabA_So in anaerobic UFAs biosynthesis. However, we failed to remove SO0197 in the absence of fabA_So under routine conditions for mutagenesis. Given that the resolution step of the mutagenesis protocol theoretically gives the wild-type and mutant strains at 50:50 (Jin et al., 2013), the failure implicates a synthetical lethality under experimental conditions. We therefore reasoned that exogenous oleate may support growth of strains lacking both pathways. Indeed, when it was supplied, the ΔfabA_SoΔSO0197 strain was obtained. Its dependence on exogenous oleate was confirmed by the observation that the double mutant could not grow without the addition (Figures 1B,C). Neither palmitate (C16:0) nor decanoate (C10:0) was able to rescue the synthetical lethal phenotype, suggesting that there unlikely exists an additional desaturase in S. oneidensis that can replace the function of missing enzymes (data not shown). When complemented by any of the deleted genes in trans, the ΔfabA_SoΔSO0197 strain gained the ability to grow without exogenous oleate (Figure S1). These observations manifest that in the absence of FabA_So SO0197 is a desaturase capable of generating sufficient UFAs to support growth. Given that the amino acid sequence similarity of SO0197 and DesA_Pa is significantly higher than that of SO0197 and the other P. aeruginosa desaturase DesB_Pa (ClustalW2 score: 15.49 and 6.25, respectively), we named SO0197 as desA_So. Unexpectedly, as shown in Figure 1D the double mutant strain lost the orange color, a signature feature of S. oneidensis because of abundant c-type cytochromes (Gao et al., 2010b; Jin et al., 2013), and the wild-type color was restored when either gene was expressed in trans. Given that the microorganism produces more than 40 such proteins (Meyer et al., 2004; Gao et al., 2010b; Fu et al., 2014b), this finding suggests that UFAs are crucial for cytochrome c biosynthesis, which is currently under study.

It is worth mentioning that DesA_Pa was not able to rescue the synthetical lethal phenotype resulting from losing both pathways when its coding gene was expressed under the control of the S. oneidensis arcA promoter, which is constitutively active (Gao et al., 2010a; Dong et al., 2012; Zhang et al., 2013; Sun et al., 2014) (Figure 1D). Thus, although both DesA proteins act as desaturases, they are not functionally exchangeable, presumably due to low sequence similarity. Moreover, S. oneidensis differs from P. aeruginosa in that the latter has an additional desaturase, DesB_Pa, which is proposed to selectively desaturates fatty acids from the environment (Zhu et al., 2006).

In E. coli, regulation of the anaerobic UFA biosynthesis pathway is carried out by FadR_Ec and FabR_Ec (Fujita et al., 2007), whose counterparts in S. oneidensis are apparent, FadR_So (SO2885, E-value, 2e-82) and FabR_So (SO0198, E-value, 9e-60). This is in sharp contrast to P. aeruginosa, the only bacterium which has both anaerobic and aerobic UFA biosynthesis pathways and has been studied. A BLASTp search against the P. aeruginosa proteome using FadR_Ec returned many hits but none of them appeared to be a possible homolog (the smallest E-value, 2e-06). Moreover, the deletion of 25 of 27 GntR homologs did not significantly influence the bacterial response to exogenous fatty acids, suggesting that P. aeruginosa may not utilize a GntR family regulator for the role of FadR_Ec in E. coli (Choi and Schweizer, 2005). In P. aeruginosa, while little is known about how the desA_Pa gene is regulated, the desB_Pa gene is under direct control of DesT_Pa, a homolog of FabR_Ec (Zhang et al., 2007). Consistently, a BLASTp search using DesT_Pa against S. oneidensis revealed a single putative homolog, FabR_So (E-value, 1e-22). The desA_So and fabR_So genes form a devergon (Figure 1A), a feature shared by the desB_Pa and desT_Pa genes, implicating a possibility that they may be physiologically associated.

To confirm the involvement of FadR_So and FabR_So in UFA biosynthesis, mutants devoid of one of these genes were constructed. Under aerobic conditions, the ΔfadR_So strain grew significantly slower than the wild-type; its generation time (~70 min) was nearly doubled (Figure 2A). Similar results were obtained from anaerobic cultures, indicating that FadR_So influences both anaerobic and aerobic UFA biosynthesis pathways. The observed defect in growth was corrected by expressing the fadR_So gene in trans under both conditions (Figure 2A), validating that the phenotype was due to the intended mutation. Importantly, the defect resulting from the fadR_So deletion was also fully eliminated by adding oleate to the medium (Figure 2B), offering direct evidence that the fadR_So mutation affects the synthesis of UFAs. In contrast, the effect of deletion of the fabR_So gene on growth under either aerobic or anaerobic conditions was insignificant. Moreover, the loss of either regulator did not significantly affect the colony color (data not shown). Overall, these data indicate that FadR_So has a more profound influence on UFA biosynthesis whereas FabR_So plays a dispensable role, at least in the context of mutant phenotypes.

To gain an understanding of impacts of UFA biosynthesis pathways on fatty acids composition of S. oneidensis, the membranes of the wild-type, ΔfabA_So, ΔdesA_So, ΔfabR_So, and ΔfadR_So strains were collected and assayed by GC-MS (Table 2). In the wild-type, SFAs, dominated by branched C15:0 (including both iso-C15:0 and antiiso-C15:0) and C16:0, accounted for ~67% of the total membrane lipids whereas the remaining consisted of three UFAs, with C16:1 as the major. Loss of DesA_So had little impact on the composition, consistent with the accessory role that it plays in UFA biosynthesis. Similar results were obtained from the ΔfabR_So strain. Along with the observation that the loss of the fabR_So gene does not elicit a distinguishable phenotype, these data indicate that the regulator is not a critical factor influencing the fatty acid composition of the S. oneidensis membrane. In contrast, removal of either fabA_So or fadR_So resulted in significantly altered fatty acid profiles. It was immediately evident that the percentages of iso-C15:0 and/or antiiso-C15:0 in both mutants increased substantially, especially in ΔfadR_So (up to ~52%). Additionally, levels of C14:0 were elevated up to 3-fold although its absolute abundance (up to 10%) was low compared to branched C15:0. As a consequence, most of other fatty acids, such as C15:0, C16:1, C16:0, and C18:1, were present in reduced levels. The level of UFAs in the ΔfabA strain was ~77% relative to that in the wild-type, contrasting unaffected UFA biosynthesis in the absence of the desA gene. This observation further confirms that the anaerobic pathway dictates UFA biosynthesis. Moreover, the loss of FadR_So introduced a most drastic reduction in overall UFA levels, ~48% relative to the wild-type level. These data conclude that FadR_So is crucial to UFA biosynthesis and its impact probably goes beyond the anaerobic pathway.

Results presented thus far indicate that the anaerobic pathway is critical to UFA biosynthesis in S. oneidensis. However, cells retain ability to produce a considerable amount of UFAs in its absence, implicating a possibility that the aerobic pathway enlarges its role when the anaerobic pathway is gone. To test this, a lacZ reporter system was employed to assess the activities of the fabA_So and desA_So promoters in the wild-type, ΔfabA_So, and ΔdesA_So strains (Fu et al., 2014a). For both genes, the most confident promoters, predicted by using Neural Network Promoter Prediction (NNPP) (Reese, 2001), are within 100 bp upstream of their coding sequence. Proper reporter vectors were constructed by placing fragments of ~300 bp covering the predicted promoters in front of the full length E. coli lacZ gene and introduced into the relevant strains. After chromosome integration and the antibiotic marker removal, activities of these two promoters in a single copy were assayed (Figure 3). When grown under aerobic conditions, the activity of the fabA_So promoter in the ΔdesA_So strain increased slightly, compared to that in the wild-type. In contrast, deletion of the fabA_So gene caused an elevation of ~3.5-fold for the desA_So promoter activity. Similar observations were obtained by using qRT-PCR (data not shown), indicating a complement role of the aerobic pathway for UFA production when the anaerobic pathway is absent.

We then evaluated influences of oleate (UFA) and palmitate (SFA) on these two UFA biosynthesis pathways. Activities of the desA_So and fabA_So promoters were assayed in cells grown with addition of either the fatty acid (Figure 3). In the wild-type background, exogenous oleate reduced fabA_So promoter activity to approximately a half but showed a modest impact on expression of the desA_So gene. Similar effects were observed from the fabA_So, and desA_So mutant strains. Notably, the induction of the desA_So gene upon removal of FabA_So was no longer evident. On the contrary, the SFA addition induced both genes substantially, largely independent of FabA_So or DesA_So. These results manifest that the supplement of UFAs suppresses expression of both UFA synthesis pathways whereas SFAs confer an opposite effect.

In E. coli, both FadR_Ec and FabR_Ec directly regulate the fabA_Ec gene whereas P. aeruginosa DesT_Pa represses expression of desB (Zhang et al., 2002, 2007; Zhu et al., 2009). To assess effects of FabR_So and FadR_So on expression of the fabA_So and desA_So genes in vivo, we measured the activities of the fabA_So and desA_So promoters using the lacZ-reporters described above (Figure 4). In the absence of FabR_So, β-galactosidase activities driven by the fabA_So and desA_So promoters increased ~2- and 5-fold, respectively, suggesting that the regulator functions as a repressor for both genes. FadR_So, however, displayed an opposite effect on expression of the fabA_So and desA_So genes, at respective ~40% and ~5-fold relative to the wild type levels in its absence. This observation suggests that FadR_So acts as an activator for the anaerobic pathway and a repressor for the aerobic pathway. Additionally, we examined whether FabR_So autoregulates its own expression given that its coding gene shares the intergenic region with the desA_So gene. A fragment of ~300 bp upstream of the fabR_So gene, covering the intergenic region separating these two genes (144 bp), was amplified and placed in front of the E. coli lacZ gene to construct the P_fabR-lacZ reporter. By using this system, we found that the fabR_So promoter activity seemed constitutive, hardly affected by loss of either FabR_So or FadR_So (Figure 4).

As shown above, loss of FadR_So or FabR_So results in significantly altered expression of the fabA_So and desA_So genes. According to the prediction made by NNPP, the most confident desA_So promoter starts at the “A” of −42 relative to the translation initiation code ATG. To confirm this, a series of fragments varying in length were amplified and placed before the E. coli lacZ gene for the promoter activity assay (Figure 5A). In the background of the wild-type, the fragments of P₁ (up to −100 bp) and P₂ (up to −71 bp) were able to drive the β-galactosidase production at levels comparable to that with the fragment of ~300 bp as shown in Figures 3, 4, 5B, indicating that the promoter is included in both P₁ and P₂. In contrast, the activities of the β-galactosidase were not detected with P₃ (up to −42 bp) or P₄ (up to −24 bp), indicating that these fragments do not cover the intact promoter. Moreover, all fragments tested behaved as expected in the fadR_So and fabR_So deletion strains (Figures 4, 5B). These data support that the promoter lies between −71 and −42, in perfect agreement with the prediction. It is worth noting that the sequence of 18 bp starting with the predicted transcriptional starting nucleotide “A” closely resembles the FabR_Ec-binding motif (-AGCGTACACGTGTTCGCT-, the same nucleotides are underlined) (Feng and Cronan, 2011), implying that the desA_So gene is under direct repression of FabR_So.

Although, expression of the desA_So gene is substantially upregulated in the absence of FadR_So (Figure 4), its regulation by FadR may be indirect as the increased desA_So expression may be a result of the reduced production of FabA, a phenomenon shown in Figure 3. To test whether FadR_So and/or FabR_So interacts with DNA fragments upstream of the fabA_So and desA_So genes, we employed a DNA-binding gel shift assay. Both FadR_So and FabR_So with the N-terminal His-tag were subjected to expression and purification from E. coli. Soluble FadR_So was obtained smoothly (Figure 6A). In contrast, we were unable to purify FabR_So after many attempts, a scenario reported before with E. coli and Vibrio cholerae FabR, implicating that the same frustrating properties are shared by these proteins (Feng and Cronan, 2011). The DNA fragments, ~300 bp in length centered by the predicted promoter of the genes to be tested, were prepared by PCR. It was found that FadR_So significantly reduced the motility of the fragments for fabA_So at a protein concentration of 0.5 μM. The strongest binding was observed with the protein at 2 μM. The specific binding was validated by that the binding was completely blocked by excessive (50x) unlabeled the same probe but not blocked by non-specific competitor of 2 μg/μl poly dI·dC. In contrast, no interaction was found with the desA_So promoter DNA (Figure 6B). These results manifest that FadR_So directly controls fabA_So but not desA_So although expression of both genes is significantly altered in its absence.

Given our inability to obtain soluble FabR_So, we turned to Bacterial one-hybrid (B1H) system to investigate DNA-protein interaction in vivo in E. coli cells (Guo et al., 2009). Positive interaction between “bait” (DNA) and “target” (DNA-binding regulator) allows the reporter strain to grow on 3-amino-1,2,4-triazole (3-AT). To prepare the bait, a ~ 300 bp fragment of fabA_So or desA_So centered by the predicted binding sequence or the promoter sequence was cloned into pBXcmT, which was paired with pTRG carrying either the fabR_So gene or the fadR_So gene for co-transformation. The system was validated with positive control pairs (P_katB/OxyR of S. oneidensis) and negative control plasmid pair (P_16S/OxyR of S. oneidensis), which showed strong and no interaction in our previous work, respectively (Jiang et al., 2014; Li et al., 2014) (Table 3). While positive interactions from P_fabASo/FadR_So were detected and confirmed by growth on plates containing both 3-AT and streptomycin (12.5 mg/ml), no colonies were obtained from P_desASo/FadR_So, supporting the EMSA result. With FabR_So as the target, both P_fabASo and P_desASo generated positive results, suggesting that FabR_So likely interact with the intergenic sequence of both the fabA_So and desA_So genes.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.