The Two Sets of DMSO Respiratory Systems of Shewanella piezotolerans WP3 Are Involved in Deep Sea Environmental Adaptation

Dimethyl sulfoxide (DMSO) is an abundant methylated sulfur compound in deep sea ecosystems. However, the mechanism underlying DMSO-induced reduction in benthic microorganisms is unknown. Shewanella piezotolerans WP3, which was isolated from a west Pacific deep sea sediment, can utilize DMSO as the terminal electron acceptor. In this study, two putative dms gene clusters [type I (dmsEFA1B1G1H1) and type II (dmsA2B2G2H2)] were identified in the WP3 genome. Genetic and physiological analyses demonstrated that both dms gene clusters were functional and the transcription of both gene clusters was affected by changes in pressure and temperature. Notably, the type I system is essential for WP3 to thrive under in situ conditions (4°C/20 MPa), whereas the type II system is more important under high pressure or low temperature conditions (20°C/20 MPa, 4°C/0.1 MPa). Additionally, DMSO-dependent growth conferred by the presence of both dms gene clusters was higher than growth conferred by either of the dms gene clusters alone. These data collectively suggest that the possession of two sets of DMSO respiratory systems is an adaptive strategy for WP3 survival in deep sea environments. We propose, for the first time, that deep sea microorganisms might be involved in global DMSO/DMS cycling.


INTRODUCTION
Dimethyl sulfoxide (DMSO) concentrations in deep oceanic water are higher than 1.5 nM at depths up to 1,500 m in the equatorial Pacific Ocean and never drop below 1.3 nM at depths up to 4,000 m in the Arabian Sea (Hatton et al., 1996(Hatton et al., , 1998(Hatton et al., , 1999. It is thought to be an environmentally significant compound due to the potential role it plays in the biogeochemical cycle of the climatically active trace gas dimethyl sulfide (DMS; Hatton et al., 2005). DMSO can be produced either through the transformation of DMS by both photooxidation and bio-oxidation routes or by direct production from marine phytoplankton (Hatton et al., 1996;Simó, 1998;Moran et al., 2012). In addition to its roles in protecting cells against photo-generated oxidants and cryogenic damage, DMSO can also be used as an alternative electron acceptor for energy conservation through microbial dissimilatory reduction Chloramphenicol resistance, generated from filamentous bacteriophage SW1; used for complementation Yang et al. (2015) pSW2-dmsA1 pSW2 containing dmsA1 and the promoter region of dmsA2  (Sunda et al., 2002;Asher et al., 2011). Although DMSO acts as the dominant organic sulfur compound in deep oceanic water, the mechanism underlying DMSO bio-reduction by bathypelagic microorganisms is unknown. Biochemical and genetic analyses of anaerobic DMSO respiration have been performed, particularly in Escherichia coli (McCrindle et al., 2005). In E. coli, two sets of operons are involved in DMSO respiration: the dmsABC operon and the ynfEFGHI operon, which is a paralog of the former and is likely to be phenotypically silent (Lubitz and Weiner, 2003). The dmsABC operon encodes the three functional proteins DmsA (the molybdopterin cofactor-containing subunit of the DMSO reductase), DmsB (the ion-sulfur subunit), and DmsC (the NapC-like integral membrane anchor). These three subunits constitute a functional DMSO reductase that is anchored to the periplasmic side of the inner membrane by DmsC. The electron released by menaquinol (MQH 2 ) oxidation by DmsC is transferred via a series of [4Fe-4S] clusters in DmsB to the catalytic subunit DmsA, which reduces DMSO to DMS (Stanley et al., 2002).
Shewanella is a genus of facultative anaerobic, Gramnegative microorganisms that are widely distributed in marine and freshwater environments. The hallmark of Shewanella is their ability to utilize a broad range of terminal electron acceptors, which makes them outstanding candidates for potential applications in the bioremediation of pollutants (Nealson and Scott, 2006;Hau and Gralnick, 2007). In Shewanella species, the DMSO reduction pathway has been characterized only in Shewanella oneidensis MR-1, which was isolated from the sediment of Oneida Lake in New York (Venkateswaran et al., 1999). Two dms operons were found in the MR-1 genome, although only one of them mediated DMSO reduction under the tested conditions (Gralnick et al., 2006;Fredrickson et al., 2008).
Shewanella piezotolerans WP3 was isolated from a west Pacific deep sea sediment at a depth of 1,914 m (Wang et al., 2004;Xiao et al., 2007). Our previous study demonstrated that WP3 was able to utilize DMSO as a terminal electron acceptor for anaerobic growth (Xiao et al., 2007). However, the precise mechanism of anaerobic DMSO respiration by WP3 is still unknown. In this study, we showed that WP3 contained two dms gene clusters (type I and type II), both of which were functional; type I was essential for the ability of WP3 to thrive under in situ conditions (4 • C/20 MPa) and type II was more important under other extreme conditions (i.e., 20 • C/20 MPa or 4 • C/0.1 MPa). The possession of two sets of DMSO respiratory systems is suggested to be an adaptive strategy for WP3 to cope with extreme deep sea environments.

Bacterial Strains and Growth Conditions
The bacterial strains used in this study are listed in Table 1. E. coli strain WM3064 was routinely grown in Luria Broth medium at 37 • C with the addition of 500 µM 2,6-diaminopimelic acid (Sigma-Aldrich, St. Louis, Mo, USA). For aerobic growth, the S. piezotolerans WP3 strains were cultured in 2216E broth Chen et al., 2011) with minor modifications (5 g l −1 tryptone, 1 g l −1 yeast extract, and 34 g l −1 NaCl) at 20 • C in a rotary shaker at 200 rpm. If necessary, chloramphenicol was added to both media (30 µg ml −1 for E. coli strains and 10 µg ml −1 for WP3 strains). For the anaerobic growth assay, the cultivation of WP3 strains was performed in 2216E broth supplemented with 20 mM lactate and 25 mM DMSO (Burns and DiChristina, 2009). Serum bottles each containing 100 ml of fresh medium were prepared anaerobically by flushing with nitrogen gas through the butyl rubber stopper fixed with a metal seal to strip the dissolved oxygen prior to autoclave sterilization. To examine the growth of the WP3 strains at high hydrostatic pressure, each culture was grown to stationary phase in 2216E medium at 1 atm (1 atm = 0.101 MPa) and 20 • C in a rotary shaker. The late-log phase cultures were diluted into the same medium to an optical density of 0.3 at 600 nm. Aliquots of the diluted culture (1 ml) were injected into serum bottles containing 100 ml of anaerobic medium through the butyl rubber stopper. After brief shaking, 2.5 ml disposable syringes were used to distribute the culture in 2 ml aliquots. The syringes with needles were stuck into rubber stoppers in a vinyl anaerobic airlock chamber (Coy Laboratory Products Inc., Grass Lake, MI, USA). Then, the syringes were incubated at a hydrostatic pressure of 20 MPa at 4 • C or 20 • C in stainless steel vessels (Feiyu Science and Technology Exploitation Co., Ltd, Nantong, China) that could be pressurized using water and a hydraulic pump. These systems were equipped with quick-connect fittings for rapid decompression and recompression (Yayanos and Van Boxtel, 1982;Allen et al., 1999).

Deletion Mutagenesis and Complementation in S. piezotolerans WP3
In-frame deletion mutagenesis of dmsA1 (swp3459) was performed as previously reported (Chen et al., 2011). The primers designed to amplify PCR products for mutagenesis are summarized in Supplementary Table S1. To construct a dmsA1 in-frame deletion mutant, two fragments flanking dmsA1 were amplified by PCR. Fusion PCR products were generated using the amplified PCR fragments as templates with primers swp3459-UF and swp3459-DR. After digestion with the restriction enzymes SmaI and XbaI, the treated fusion PCR products were ligated into the SmaI and XbaI sites of the suicide plasmid pRE112 (Edwards et al., 1998), resulting in the mutagenesis vector pRE112-dmsA1. This vector was first introduced into E. coli WM3064 and then conjugated into WP3-WT (wild type). Positive exconjugants were spread onto marine agar 2216E plates supplemented with 10% (w/v) sucrose. The chloramphenicol-sensitive and sucrose-resistant colonies were screened by PCR for the dmsA1 deletion. The same strategy was used to construct dmsA2 single mutants. The double mutant dmsA1 dmsA2 ( dmsA) was constructed by introducing pRE112-dmsA2 into dmsA1. To construct the complemented strain of dmsA1 dmsA2, the intact dmsA1 and dmsA2 gene fragments were amplified from the WT genomic DNA. The resulting PCR products were inserted into the shuttle vector pSW2, which was developed from the filamentous phage SW1 of WP3 (Yang et al., 2015). Transformants dmsA-dmsA1-C and dmsA-dmsA2-C were obtained by introducing the recombinant plasmids pSW2-dmsA1 and pSW2-dmsA2, respectively, into the dmsA strain via conjugation.

RNA Extraction and Reverse Transcription
Total RNA was extracted from the WP3 strains as previously described (Wang et al., 2009

Real-time PCR (RT-PCR)
Primer pairs (Supplementary Table S1) for the genes selected for the RT-PCR analysis were designed using the Primer Express 3.0 software (Applied Biosystems, Warrington, UK). Transcription assays were performed using the 7500 System SDS software in reaction mixtures with total volumes of 20 µl containing 10 µl of SYBR Green PCR Master Mix (Applied Biosystems, Warrington, UK), 0.5 µM of each primer, and 1.2 µl of the cDNA template. The amount of target was normalized to the reference gene swp2079 Wang et al., 2009). The RT-PCR assays were performed in triplicate for each sample and a mean value and standard deviations were calculated for the relative RNA expression levels.

DMSO Concentration Determination
Aliquots (1 ml) of the culture recovered at different time points were filtered immediately through 0.22 µm Millex-GP filters (Millipore, Carrigtwohill, Ireland) and stored at −70 • C prior to use. The DMSO concentration was determined as previously reported (Cheng et al., 2013). Briefly, the samples were applied for analysis by an Agilent 1200 series high-performance liquid chromatography (HPLC) system (Agilent Technologies, Santa Clara, CA, USA) with a diode-array detector set at 210 nm. DMSO was separated using an Aminex HPX-87H column (Bio-Rad, Hercules, CA, USA) at 50 • C using H 2 SO 4 (5 mM) as the mobile phase at a flow rate of 0.5 ml min −1 . Commercially available DMSO (Sigma-Aldrich, St. Louis, MO, USA) was used to generate a calibration curve to estimate the DMSO concentration.

Determination of the Optimum DMSO Concentration for WP3 Anaerobic Respiration
The 2216E medium supplemented with sodium lactate and DMSO was used to study the anaerobic DMSO respiration of WP3. The time course of the DMSO concentration was monitored by assessing the cell density (represented by OD 600 ) in the medium (Figure 1). The maximum growth was observed at a DMSO concentration of 25 mM, at which concentration the DMSO in the medium was exhausted and the cells were entering stationary phase. We found no significant differences in the cell density and DMSO consumption with a further increase in the DMSO concentration (40 or 60 mM) compared with growth at the 25 mM concentration, suggesting that excess DMSO (40 or 60 mM) had no positive effect on the growth yield. Therefore, the subsequent physiological experiments were performed with DMSO at a concentration of 25 mM.

Organization of the dms Gene Clusters in WP3
Two sets of dms gene clusters [type I (dmsEFA1B1G1H1 and swp3461-swp3456) and type II (dmsA2B2G2H2 and swp0724-swp0728)] were identified according to the WP3 genome annotation (Figure 2A). The type I dms gene cluster matches the archetypal dmsEFABGH organization. However, no dmsE or dmsF homologous gene was found upstream of dmsA2B2G2H2 in the type II dms gene cluster. Additionally, two partially overlapped genes encoding the LysR family regulator were located directly upstream of dmsA2. To investigate whether these two operons were truly polycistronic, reverse transcription PCR was performed to amplify the intergenic regions of neighboring genes. Our data demonstrated that five genes (dmsEFA1B1G1) in the type I dms gene cluster were co-transcribed, with the exception of swp3456 (Figure 2B). For the type II dms gene cluster, four genes (dmsA2B2G2H2) were co-transcribed as a single operon (Figure 2C).

Both dms Gene Clusters in WP3 Were Functional in DMSO Respiration
To test whether both dms gene clusters were responsible for DMSO respiration, we initially monitored the transcription of both dms gene clusters using quantitative reverse transcription PCR (qRT-PCR). The results showed that both gene clusters were significantly induced under anaerobic DMSO conditions (Supplementary Figure S1). Because dmsA encodes the large catalytic subunit of the DMSO reductase and plays an essential role in DMSO respiration (Trieber et al., 1996), dmsA of these two gene clusters was inactivated and in-frame deletion mutants were constructed. The growth assay performed at 20 • C/0.1 MPa demonstrated that the single mutants dmsA1 and dmsA2 retained the ability to use DMSO for anaerobic growth, whereas the double mutant dmsA (namely dmsA1 dmsA2) failed to grow in the same medium ( Figure 3A) and did not consume DMSO (Figure 3B). To confirm that the loss of DMSO-dependent growth of dmsA was caused by the disruption of dmsA1 and dmsA2, two complemented strains ( dmsA-dmsA1-C and dmsA-dmsA2-C) were generated. As expected, the introduction of either dmsA1 or dmsA2 into dmsA partially restored the ability of the double mutant to utilize DMSO for anaerobic growth (Figures 3C,D). Because dmsB encoded the small ion-sulfur subunit of the DMSO reductase, we constructed corresponding in-frame deletion mutants ( dmsB1, dmsB2, and dmsB1 dmsB2) and complemented strains ( dmsB-dmsB1-C and dmsB-dmsB2-C). Similar results were observed in the growth yields and DMSO consumption of these strains (Supplementary Figure  S2). Taken together, these data indicated that both the type I and type II dms gene clusters were responsible for DMSO respiration.

Transcription of the Two dms Gene Clusters Was Affected by Pressure and Temperature Changes
To evaluate whether these two dms gene clusters were involved in DMSO respiration at different temperatures (4 and 20 • C) and pressures (0.1 and 20 MPa), relative transcription assays of these two clusters were performed by qRT-PCR. The cells were incubated in anaerobic DMSO medium under different growth conditions for RNA extraction. Compared to the expression levels of the two dms gene clusters at 20 • C/0.1 MPa, the expression of the type I cluster was significantly down-regulated with pressure increased from atmospheric pressure (0.1 MPa) to high pressure (10 and 20 MPa) conditions, whereas the expression of type II cluster was significantly up-regulated under the same conditions ( Figure 4A). In contrast, the expression profiles of two dms gene clusters at 10 • C were generally similar regardless of pressure changes ( Figure 4B). Interestingly, the expression profiles of type I dms clusters at 4 • C were significantly induced with pressure increased from atmospheric pressure to high pressure (10 and 20 MPa) conditions, whereas the expression profiles of type II dms clusters were greatly reduced under the same conditions ( Figure 4C). Collectively, these data suggested that the transcription of the two dms clusters was regulated in response to pressure and temperature changes.

The Contribution of the Two dms Gene Clusters to WP3 Growth was Influenced by Pressure and Temperature
To investigate whether the two dms gene clusters were involved in deep sea environmental adaptation, a growth assay was performed at different pressures and temperatures. As predicted, dmsA1 exhibited a higher growth yield than dmsA2 when the growth temperature decreased from 20 to 4 • C or the pressure increased from 0.1 to 20 MPa (Figures 5A,C), which indicated that the type II dms operon was more important than the type I operon at low temperature or high pressure. Additionally, culture media collected at different time points were filtered and then analyzed to assess DMSO consumption by these strains. The HPLC results indicated that the dmsA1 mutant exhibited higher DMSO consumption (Figures 5B,D) than dmsA2 at 4 • C/0.1 MPa or 20 • C/20 MPa.
Interestingly, when these strains were incubated at 4 • C/20 MPa, the growth of dmsA2 appeared to be higher than dmsA1 (Figure 5E), which was consistent with the more efficient DMSO consumption by dmsA2 in the medium ( Figure 5F). Additionally, our results showed that the DMSOdependent growth yield conferred by the presence of the two dms operons was higher than the growth yield conferred by either of the two dms operons alone, indicating that the presence of both dms operons was essential for the DMSO-dependent growth of WP3 under in situ deep sea conditions.
Previous study demonstrated that DMSO as a substantial sink or source for DMS (Hatton et al., 1996). The formation of DMSO would therefore lead to the removal of DMS from sea water, effectively limiting the quantity of DMS available for releasing into the atmosphere (Hatton et al., 2005). Although DMSO has been detected in abyssal ocean water (Hatton et al., 1999), few data are available on the turnover mechanisms of DMSO in deep sea environments. In this study, our results demonstrate a previously unrecognized role for biotic DMSO reduction as an important pathway of DMS production in the deep sea environment.
In the abyssal ocean, metabolic energy is harnessed from the coupling of redox reactions (Orcutt et al., 2011). Microorganisms exploit the available chemical energy by developing strategies to overcome the activation energy of reaction. The metabolic activities of microorganisms in such environment depend on the availability and speciation of electron donors and acceptors (Froelich et al., 1979). As for DMSO, the E 0 value for its reduction (+160 mV) lies between those for fumarate/succinate (+33 mV) and nitrate/nitrite (+433 mV; Thauer et al., 1977;Wood, 1981), and thus the coupling of DMSO reduction to ATP synthesis presents no energetic difficulty. Accordingly, some culturable bacteria isolated from marine environments have also shown the ability to grow by anaerobic DMSO respiration (Jonkers et al., 1996;Süß et al., 2008). Although sulfate is the most abundant oxidized sulfur species in the dark ocean (Orcutt et al., 2011), the reduction of sulfate yields less energy than that of DMSO, and DMSO could therefore be consumed preferentially.
In Shewanella species, the DMSO reduction pathway was first characterized in S. oneidensis MR-1, which was a strain isolated from the sediment of Oneida Lake in New York (Venkateswaran et al., 1999). Two dms operons were found in the MR-1 genome; however, only one dms operon mediated DMSO reduction under the tested conditions (Gralnick et al., 2006). In contrast to the results in MR-1, both dms gene operons in WP3 were functional (Figure 3). Moreover, our results demonstrated that the presence of both operons was essential for the maximum growth of WP3 using DMSO as the sole electron acceptor (Figure 5). Type I matches the archetypal dmsEFABGH organization, which enables the DMSO reductase to localize to the outer membrane of Shewanella and allows the cells to respire DMSO more efficiently (Gralnick et al., 2006). Additionally, type I plays a dominant role in the DMSO-dependent growth of WP3 under in situ deep sea conditions (4 • C/20 MPa). Type II is more similar to E. coli (dmsABC organization; Stanley et al., 2002) and may locate in the periplasmic space. The physiological analysis indicated that type II was more important for DMSO-dependent growth than type I under some stressful conditions (i.e., 20 • C/20 MPa and 4 • C/0.1 MPa).
To evaluate the impact of the two DMSO respiration systems on growth, we calculated the DMSO consumption of the WP3 mutants by standardizing the growth yield to the same cell density (1 OD unit ≈ 1.5 × 10 8 cells; Supplementary Figure  S3). Our data showed that WT (type I and type II together) and dmsA1 (type II alone) required less DMSO than dmsA2 (type I alone) to reach the same growth yield when incubated at 20 • C/0.1 MPa, 4 • C/0.1 MPa, or 20 • C/20 MPa, indicating that type II was a more efficient DMSO respiration system at atmospheric pressure or 20 • C. However, when the WP3 mutants were incubated at 4 • C/20 MPa, dmsA2 (type I alone) inversely required less DMSO than dmsA1 (type II alone), indicating that type I was more efficient than type II under the 4 • C/20 MPa in situ condition. Collectively, our results suggest that the possession of multiple systems with similar functions might be an adaptive strategy for bacteria to cope with extreme deep sea environments. Similar adaptive mechanisms can also be found in other deep seaderived Shewanella strains. For example, Shewanella putrefaciens W3-18-1 (Sediment; under 630 m of oxic water) has been identified to contain two functional nap gene clusters (nap-α and nap-β), and the nitrate respiration in W3-18-1 starts earlier than that in MR-1 (with nap-β only) under microoxic conditions (Qiu et al., 2013). In another Shewanella strain, Shewanella violacea DSS12 (Sediment; 5,110 m), it possesses more terminal oxidases with different affinities for oxygen and less terminal reductases, implying that DSS12 has undergone respiratory adaptation to aerobiosis in the upper layer of deep sea sediments (Aono et al., 2010).
The physiological results concerning the possession of two DMSO respiration systems in WP3 are quite different from those of the two NAP systems in WP3. Although two nitrate respiration systems (nap-α and nap-β) were identified in the WP3 genome, further studies demonstrated that WP3 possessing nap-α alone was more competitive than the strain possessing both nap-α and nap-β during nitrate-dependent growth. Thus containing a single nap-α system should be the evolutionary direction for Shewanella to thrive at low temperature (Chen et al., 2011). By combining these results, we conclude that WP3 can utilize different strategies to evolve its different anaerobic respiration systems in response to deep sea environments.
The effects of pressure on the respiratory systems has been well studied in S. violacea DSS12. Although very few terminal reductases for anaerobic respiration being identified in DSS12 genome, it processes at least five sets of putative terminal oxidases for aerobic respiration (Aono et al., 2010), in which the expression of cytochrome bd (quinol oxidase) was enhanced under high pressure (Tamegai et al., 2005). In addition, the d-type terminal oxidase of DSS12 was identified to be pressure-resistant and functional under high pressure conditions (Chikuma et al., 2007). Further research demonstrated that the piezotolerance of the Shewanella terminal oxidase activity perhaps depends on the intrinsic properties of the enzymes themselves but not their membrane lipid composition (Tamegai et al., 2011). In contrast, the piezophilic property of terminal oxidases in Photobacterium profundum SS9 (amphipod; 2,500 m) seems to be different. For one thing, the cytochrome content of P. profundum SS9 was not affected by altered pressure during growth; for another, genes encoding cytochromes showed no significant change in expression under different pressures during growth of P. profundum SS9. However, the piezotolerance of P. profundum SS9 terminal oxidase activity could also be observed using cells grown under higher pressures (Tamegai et al., 2012). Collectively, piezophiles from distinct genus may choose different strategies in adaptating to high pressure.
More recently, the respective effects of temperature and pressure on Fe(III) reduction rates (FeRRs) and viability have been investigated in deep sea bacterium Shewanella profunda LT13a (sediment; 4,500 m; Picard et al., 2015). Results showed that FeRR increased linearly with temperature between 4 and 37 • C, while the highest FeRR was observed between 10 and 40 MPa and then slightly decreased with increasing pressure (40-110 MPa), indicating that the respiratory chain was not immediately affected by pressure (Picard et al., 2015). Under high pressure conditions, the significantly increased energy demand for cell maintenance of S. profunda LT13a may account for its decreased viability (Picard et al., 2015). Here we observed that DMSO-dependent growth can occur when WP3 is incubated in 4 • C/20 MPa conditions (Figure 5E), suggesting that the energy demand of WP3 could still be fulfilled under in situ conditions with DMSO as the sole electron acceptor.
Because the transcription profiles of these two dms gene clusters vary significantly in response to pressure or temperature changes (Figure 4), it is important to elucidate the mechanism by which dms gene clusters are regulated. Previously, a mutant lacking the fur gene (encoding the ferric uptake regulator) was found to be severely deficient in DMSO respiration in WP3 , indicating that Fur was the dms gene cluster regulator. Moreover, cAMP receptor protein (CRP) was also found to directly regulate the expression of the dms gene cluster in S. oneidensis MR-1 (Zhou et al., 2013). We predicted the transcriptional regulation motifs of both dms gene clusters in WP3 and found similar CRP binding motifs (data not shown). Taken together, the presence of the CRP and Fur binding motifs might be an alternative regulatory strategy for WP3 to cope with pressure or temperature changes in a subtle regulatory mechanism. Investigation into this mechanism is currently underway and will be described as part of a separate report.

CONCLUSION
We have shown that two dms operons found in S. piezotolerans WP3 are functional in DMSO respiration. Both dms gene clusters are essential for the maximum growth of WP3 under low temperature and high pressure conditions; the type I system is essential for the ability of WP3 to thrive under in situ conditions (4 • C/20 MPa), whereas the type II system is more important under other stressful conditions (i.e., 4 • C/0.1 MPa and 20 • C/20 MPa). Based on these data, we propose that multiple copies of dms gene clusters may confer a competitive advantage to the ability of WP3 to thrive under a broader range of environmental conditions. Thus, deep sea microorganism-mediated DMSO reduction may contribute to the global DMS cycle, which has been neglected in previous studies.

AUTHOR CONTRIBUTIONS
LX, HJ, and XX designed the study, analyzed the data. LX and YZ participated in the experiments. LX and HJ wrote the manuscript and XX critically reviewed the manuscript. All authors read and approved the final manuscript.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01418