Marine phytoplankton produce ∼109 tonnes of dimethylsulfoniopropionate (DMSP) per year1,2, an estimated 10% of which is catabolized by bacteria through the DMSP cleavage pathway to the climatically active gas dimethyl sulfide3,4. SAR11 Alphaproteobacteria (order Pelagibacterales), the most abundant chemo-organotrophic bacteria in the oceans, have been shown to assimilate DMSP into biomass, thereby supplying this cell's unusual requirement for reduced sulfur5,6. Here, we report that Pelagibacter HTCC1062 produces the gas methanethiol, and that a second DMSP catabolic pathway, mediated by a cupin-like DMSP lyase, DddK, simultaneously shunts as much as 59% of DMSP uptake to dimethyl sulfide production. We propose a model in which the allocation of DMSP between these pathways is kinetically controlled to release increasing amounts of dimethyl sulfide as the supply of DMSP exceeds cellular sulfur demands for biosynthesis.
In an experiment designed to measure the stoichiometry of dimethylsulfoniopropionate consumption, we observed that Pelagibacterales strain HTCC1062 produced methanethiol (MeSH), the gaseous end product of a catabolic pathway in which the first step involves DMSP demethylation. This was consistent with the presence in the genome of dmdA, which encodes DMSP demethylase3,7. However, we were surprised to observe that axenic cultures of this strain also produced large amounts of dimethyl sulfide (DMS, Fig. 1a). This observation indicated that, despite widespread attention to Pelagibacterales genomics and metagenomics, a Pelagibacter DMSP cleavage metabolic pathway leading to DMS formation had gone undetected. The amounts of DMS and MeSH increased linearly over 18 h of incubation in the presence of live cells, but DMS production by killed cell controls was either low or undetectable. Over 80% of the DMSP sulfur decrease could be accounted for, with 59% converted to DMS, 21% to MeSH and ∼1% for biosynthesis (Table 1). These observations were confirmed by real-time measurements of DMS and MeSH production by cultured cells, using a proton-transfer-reaction time-of-flight mass spectrometer (PTR-TOF/MS, Fig. 2a).
The discovery that Pelagibacter expresses two DMSP degradation pathways simultaneously is particularly striking given its small genome size (1.28–1.46 Mb) and simple metabolism8. Enzymes for the DMSP demethylation pathway (DmdABC) have been described in Pelagibacter, but not DmdD, which catalyses the release of MeSH from methylthioacryloyl–coenzyme A (MTA-CoA)9. Nor has a gene for any DMSP lyase, which catalyses the alternative catabolic pathway leading to DMS production, been annotated or reported in Pelagibacter. Thus, the data shown in Fig. 1a confirm a complete demethylation pathway leading to MeSH production in Pelagibacter3,7 and are the first evidence of a DMSP cleavage pathway in this organism.
Assimilation of DMSP sulfur into biomass is potentially a strong evolutionary driver for retention of DMSP metabolism in Pelagibacterales, which lack genes for assimilatory sulfate reduction10. To identify intermediates of DMSP metabolism that could support the demand for reduced sulfur for biosynthesis, HTCC1062 cultures were inoculated into artificial seawater medium (ASW) in the presence of MeSH, DMS or methionine (Fig. 1b). Only MeSH and methionine supported growth above the negative control. This is the first data showing that free MeSH can serve as a sulfur source for Pelagibacterales cells and it is consistent with the observation that, under DMSP-replete conditions, more sulfur is released as MeSH than is used for growth. The lower molar yield observed with MeSH, relative to methionine, is probably a consequence of the susceptibility of MeSH to spontaneous oxidation. DMS is apparently a metabolic waste product and cannot serve as a source of reduced sulfur in Pelagibacterales, in accord with the observations that DMS monooxygenase and DMS dehydrogenase are missing from Pelagibacterales genomes (Supplementary Fig. 1).
The unexpected observation of DMS production by HTCC1062 cultures (Fig. 1a) suggested that a DMSP lyase gene had gone undetected
in the genome7,9. Reviewing the genome annotation, we noticed that hypothetical
gene SAR11_0394 was predicted to have a C-terminal cupin, a very widely distributed
protein fold that resembles a small barrel11. Of the DMSP lyases identified to date, three (DddL, DddQ and DddW)
have C-terminal cupin domains and are members of the cupin superfamily12,
DMSP catabolism also benefits cells by providing a source of organic carbon that can be oxidized for energy production or assimilated into biomass15. The data suggest that when cells are supplied with an excess of DMSP, 99% of DMSP oxidation is probably supporting carbon metabolism (Fig. 1a and Table 1). DMSP lyase enzymes are distributed among multiple protein families, but all lead to the production of DMS and either acrylate (DddL, P, Q, W, Y) or 3-hydroxypropionate (3-HP; DddD)9. The HTCC1062 genome encodes annotated genes for all steps in the degradation of acrylate to propionyl–CoA or acetyl–CoA (Fig. 3). To test the capacity of strain HTCC1062 to assimilate acrylate, propionate or 3-HP, we relied on the unusual requirement of Pelagibacter strains for growth substrates that can be metabolized to pyruvate, which these cells require for alanine synthesis16. As predicted, acrylate and propionate each could substitute for pyruvate in defined media. Enhancement of growth by 3-HP was slight, but statistically significant (Student's t-test, n = 3, P < 0.05) (Fig. 1c).
Comparisons of Pelagibacterales genomes across the Group Ia subclade revealed that dddK homologues were found in eight of twelve Pelagibacterales Ia genomes (Supplementary Fig. 5). In addition to dddK, strain HIMB5 has a homologue of dddQ, also a member of the cupin superfamily13. As predicted, E. coli transformants containing cloned HIMB5_00000220 (dddQ) displayed DMSP lyase activity (Km = 56 mM, Vmax = 0.78 µmol min−1 (mg protein)−1). Strain HTCC7211 and the more distantly related subclade V strain HIMB59 lacked dddK homologues, but encoded gene products (respectively PB7211_1082 and HIMB59_00005110) that are ∼30% identical to DddP, a DMSP lyase in the M24 family of metallo-peptidases17,18. However, E. coli transformants containing cloned PB7211_1082 (dddP-like) showed very low DMSP lyase activity (0.5 ± 0.1 nmol min−1 mg−1) and therefore this protein may not be a bona fide DMSP lyase.
We compared the abundance of the Pelagibacterales genes for DMSP cleavage with those for demethylation (dmdABC) in the Global Ocean Survey (GOS) metagenomic data set (Supplementary Fig. 6). The DMSP lyases dddK and dddQ, and dddP, the putative lyase with low activity, were much less abundant than dmdABC or the single-copy marker recA. This supports the interpretation that either the cleavage pathway is less important than the demethylation pathway, or undiscovered DMSP lyase analogues are present in other Pelagibacterales strains. Interestingly, Pelagibacterales genes for metabolism of acrylate are more abundant than DMSP lyases and similar in abundance to demethylation genes and recA, which supports either the interpretation that DMSP lyases are underestimated because of their diversity, or that Pelagibacter cells lacking DMSP lyase use acrylate from other sources, perhaps dissolved acrylate.
Most of the Pelagibacterales strains with DddK genes belong to the temperate surface ocean ecotypes (Ia.1)19, whereas most of the strains that possess DddP are subtropical ocean surface ecotypes (Ia.3) (Supplementary Fig. 5). This may indicate that the lyase system is more common in Pelagibacterales strains, such as HTCC1062, that originate from higher-productivity ocean regions, a distribution that is consistent with its inferred role as an auxiliary system that metabolizes excess DMSP. However, the presence of an alternative gene, DddP, that has weak DMSP lyase activity in most SAR11 Ia.3 strains, suggests that further investigations of the phenotypes of live strains will be needed before the distribution of DMSP metabolism across the clade is fully understood.
Metabolic reconstruction with eight Pelagibacterales genomes revealed that, consistent with the observation of MeSH production in HTCC1062, this and other examined Pelagibacterales strains (except those in the distantly related subclade IIIa) contain homologues of the dmdABC genes found in Ruegeria pomeroyi20 (Supplementary Fig. 1). Also reported in nearly all Pelagibacterales are genes encoding methyl group oxidation pathways (Supplementary Fig. 1), which produce energy from DMSP demethylation and are essential to the demethylation pathway because they perform the function of regenerating the methyl-group-accepting cofactor tetrahydrofolate (THF)15,20. Pelagibacterales strains also contain homologues of cystathionine-gamma-synthetase (cys-γ-synth), predicted to catalyse the conversion of MeSH to methionine21 and thus necessary for growth when using MeSH as sole sulfur source. However, none of these examined Pelagibacterales strains had homologues of dmdD (methylthioacryloyl-CoA hydratase), which converts MTA–CoA to MeSH. The absence of this gene from Pelagibacterales is also reflected in its low abundance in ocean metagenomic databases9. As in HTCC1062, dmdD is not required for complete demethylation of DMSP to MeSH in Ruegeria lacuscaerulensis7, suggesting that an undescribed analogous enzyme fills this pathway gap.
One of the unexpected findings reported above is that both the cleavage and demethylation pathways operate simultaneously. We investigated transcription changes using Affymetrix microarrays and observed no significant changes in the expression of DMSP catabolic pathway genes between HTCC1062 cells grown in the presence and absence of DMSP (Supplementary Section II). Because no changes in transcription were observed, we used isobaric tags for relative and absolute quantitation (iTRAQ) to compare the proteomes of HTCC1062 cultures grown in the presence and absence of DMSP, confirming that proteins for both pathways of DMSP catabolism are expressed constitutively (Supplementary Fig. 7 and Supplementary Table 2). Further support for this conclusion came from real-time measurements of DMS and MeSH production by cells, which showed that DMS and MeSH were immediately released when DMSP was added to cells that had been grown in the absence of DMSP (Fig. 2a).
We propose that constitutive, simultaneous expression of the cleavage and demethylation pathways in Pelagibacter is an adaptation that provides these cells with a kinetically regulated system that favours the pathway to DMS formation when intracellular DMSP concentrations are high. We modelled this process (Supplementary Fig. 8) using the measured properties of cloned enzymes and intracellular DMSP concentration (Fig. 2b). In Pelagibacter, DMSP active transport is thought to be mediated by the ABC transporter (OpuAC), which was the sixth most highly detected Pelagibacterales protein in a previous study of the Sargasso Sea metaproteome22. The properties of ABC transport functions are consistent with the model in that they predict that cells can achieve high intracellular DMSP concentrations from naturally measured DMSP abundances (Supplementary Table 3), provided that DMSP remains within the range of transporter affinity for a period of hours (Supplementary Section V). The Km we measured for DddK, 81.9 ± 17.2 mM (Supplementary Fig. 4), is high compared to the Km of DmdA (13.2 ± 2.0 mM)23. Intracellular DMSP concentrations increased following DMSP addition, reaching a maximum of 180 mM after 4 h (Fig. 2b). When DMSP flux into cells is low, the model predicts that most is channelled to MeSH production, producing energy via oxidation of the products CH3–THF and acetaldehyde, sulfur for biosynthesis and MeSH losses caused by oxidation and diffusion (Fig. 1a). As intracellular DMSP concentrations rise, the model predicts that DMSP cleavage to DMS increases (Supplementary Fig. 8). There is a precedent for simple, kinetically driven switches controlling the flow of vital metabolites in HTCC1062, where intracellular glycine concentrations control the flow of carbon from exometabolites, such as glycolic acid, via glycine-mediated riboswitches16,24. Kinetic regulation of metabolic processes is well known, but here we see evidence that it plays an unexpected role in large-scale biogeochemical processes mediated by metabolically streamlined cells.
The model presented in Supplementary Fig. 8 captures the observations we report, and provides an explanation for why cells might simultaneously express two pathways that compete for a single substrate. Although the model in Supplementary Fig. 8 is based on in vitro enzyme kinetics, which can deviate from the kinetic properties of enzymes in the intracellular environment, the model successfully approximates the behaviour of whole cells (Fig. 2a). However, a number of aspects of this model will need to be tested and refined before it can be validly implemented for geochemical predictions. In particular, we observed cells accumulating DMSP to high intracellular concentrations over a period of a few hours when supplied with excess DMSP. It remains to be determined how frequently such sustained supplies of DMSP occur in nature.
Recognition that the relative expression of the demethylation and cleavage pathways by bacteria in nature controls the fate of DMSP sulfur led to a concept that is referred to as the ‘bacterial switch’ in discussions of DMSP biogeochemistry2. In principle, the ‘bacterial switch’ could involve different bacterial taxa, each potentially having a different organization of DMSP metabolic pathways. Although the bacterial switch is largely hypothetical4,25,26, insight has emerged from studies of cells in culture. Like Pelagibacter, the marine bacterium R. pomeroyi strain DSS-3 has both the DMSP demethylation and cleavage pathways, which are transcriptionally regulated, although the changes in expression that were reported were not very large18,27. Further work is needed to determine whether kinetic switching plays a role in the R. pomeroyi response to DMSP. Recent field observations indicate that Roseobacter species HTCC2255 regulates transcriptional expression of both the lyase and demethylase pathways for DMSP catabolism in response to changing environmental conditions28. The findings we report here provide important details about the mechanisms of the bacterial switch that will be vital to the design of future research and to modelling transformations of DMSP in ocean ecosystems29,30. Many factors, including DMSP leakage from phytoplankton, the action of free (dissolved) DMSP lyases and the activity of many different microbial taxa, contribute to natural fluxes of DMSP and its volatile derivatives28,29. The findings we present here describe an unexpected and simple mechanism that is probably an important part of this complex process.
Measurements of DMSP and its metabolic products
HTCC1062 was grown in autoclaved, filtered ASW16 amended with 1 mM NH4Cl, 100 µM KH2PO4, 1 µM FeCl3, 100 µM pyruvate, 50 µM glycine, 1 µM DMSP and excess vitamins16. Cultures were collected by centrifugation, washed once, and resuspended in ASW. Cells (final concentration of ∼1.5 × 106 cells ml−1) were distributed into 20 ml sealed vials (10 ml per vial). DMSP (1 µM) was injected into vials and incubated in the dark at 16 °C. Biological activity was stopped by the addition of 0.1 M sodium azide (100 µl per vial) at 0, 20 min and 1, 3, 9 and 18 h. Biological duplicate samples were refrigerated before chemical analysis.
DMS and MeSH were analysed using the solid-phase microextraction-gas chromatography-pulsed flame-photometric detection (SPME-GC-PFPD) method31,32. DMSP was quantified by measuring released DMS after hydrolysis in NaOH (0.1 M final concentration), at room temperature for 12 h.
DMS and MeSH utilization in HTCC1062
HTCC1062 was cultured in 40 ml clear sealed vials with autoclaved, filtered ASW amended with 1 mM NH4Cl, 100 µM KH2PO4, 1 µM FeCl3, 100 µM pyruvate, 50 µM glycine, excess vitamins16 and 100 nM DMSP, methionine, DMS or MeSH. Each vial contained a 10 ml aliquot, which was incubated on a shaker at 16 °C. Cell densities were monitored with a Guava flow cytometer33.
C3 compounds utilization in HTCC1062
Cells (biological triplicates) were grown in autoclaved, filtered ASW amended with 100 µM NH4Cl, 10 µM KH2PO4, 100 nM FeCl3, 50 µM glycine, 50 µM methionine and excess vitamins16. Each compound (3-HP, acrylate or propionate) was tested at a concentration of 10 µM. The positive control was amended with 10 µM pyruvate. The negative control contained no pyruvate.
Real-time measurements of DMS and MeSH by PTR-TOF/MS
HTCC1062 was grown in autoclaved, filtered ASW amended with 1 mM NH4Cl, 100 µM KH2PO4, 1 µM FeCl3, 100 µM pyruvate, 25 µM glycine, 25 µM methionine and excess vitamins16. Cultures were collected by centrifugation, washed once and re-suspended in ASW. Cells (∼3–5 × 106 cells ml−1 final concentration) were distributed into 100 ml ASW and placed in a 200 ml polycarbonate dynamic stripping chamber. DMSP (1 µM) was spiked into the chamber and the suspensions were incubated at 16 °C under a continuous flow of fine air bubbles. A PTR-TOF mass spectrometer 1000 (IONICON Analytik) was used to quantify the production of MeSH and DMS from HTCC1062 cultures. The fundamentals of PTR-MS are described elsewhere34. Primary ions (protonated water, H3O+) were produced from pure water vapour in the hollow cathode ion source at a flow rate of 5 s.c.c.m., from which they entered the drift tube. The sample air stream produced from the dynamic stripping chamber was introduced to the drift tube via a separate orifice, where proton transfer reactions occurred between H3O+ and volatile organic compounds (VOCs) that had proton affinities greater than that of water (691 kJ mol−1):Within the drift tube, the pressure, temperature and voltage conditions were kept constant at 2.0 mbar, 80 °C and 600 V, respectively, which equated to a field strength (E/N) of 153 Td (where Td = 10−17 cm2 V molecule−1). One advantage of PTR-MS is that reactions occurring in the drift tube are non-dissociative and compounds are not usually fragmented during ionization and exhibit a protonated mass of M+1. Thus, for DMS and MeSH, we monitored m/z 63 and 49, respectively. Although interference at these masses is likely to be low, we cannot rule out the possibility that more than one compound was contributing to the signal. Mass spectra were recorded up to 250 a.m.u. at 10 s integration intervals. Quantification of gas-phase DMS and MeSH concentrations was achieved using the relative transmission (kinetic) approach and also accounted for the influence of the hydrated water cluster at m/z 37 (due to the high sample humidity introduced by bubbling air through seawater). For MeSH, a default collision rate constant of 2.00 × 10−9 cm−2 was assumed, whereas a literature value of 2.53 × 10−9 cm−2 was used for DMS35.
Intracellular DMSP concentration
HTCC1062 was grown under the same conditions as described above (see ‘Real-time measurements of DMS and MeSH by PTR-TOF/MS’). Cultures were collected by centrifugation, washed once and resuspended in ASW. Cells (final concentration of ∼4 × 106 cells ml−1) were distributed into five 200 ml chambers (100 ml per chamber) and treated using the same air bubbling method as described for the dynamic stripping chambers above. DMSP (1 µM) was spiked into the chambers, which were subsequently incubated at 16 °C. Duplicate negative (killed cells) controls were performed by the addition of 0.1 M sodium azide (100 µl per vial). Cultures (10 ml, biological triplicates) were filtered through 0.1 µm polytetrafluoroethylene membranes at 10 min and 1.5, 4, 7, 10 and 13 h. The cells on the membranes were washed once with ASW, then transferred into 20 ml sealed vials, and finally resuspended in 10 ml ASW. DMSP was quantified by measuring DMS release after hydrolysis in NaOH (0.1 M final concentration), at room temperature for 12 h. DMS was analysed using the SPME-GC-PFPD method.
The authors thank J.W.H. Dacey for providing DMSP and E. Boss for help with modelling the transport kinetics. The authors thank J.W.H. Dacey and S. Bennett for advice regarding the methods for DMSP measurements and N. Le Brun for suggestions on the properties of the cupin lyases and kinetics analysis. J.S. acknowledges China Scholarships Council (CSC) for financial support. Major support was provided by a grant from the Marine Microbiology Initiative of the Gordon and Betty Moore Foundation (grant no. GBMF607.01 to S.J.G.). Proteomics measurements were supported by the US Department of Energy's (DOE) Office of Biological and Environmental Research (OBER) Pan-omics programme at Pacific Northwest National Laboratory (PNNL) and performed in the Environmental Molecular Sciences Laboratory, a DOE OBER national scientific user facility on the PNNL campus. A.W.B.J. and J.D.T. were supported by grant no. NE/H008586/1 from the UK Natural Environment Research Council and E.K.F. was supported by a studentship from the Tyndall Centre at the University of East Anglia. Funds for the PTR-TOF were provided by NASA (grant no. NNX15AE70G to K.H.H. and S.J.G.) and by a grant to K.H.H. from the Oregon State University Research Office. This research was supported by the US National Science Foundation (grant OCE-1436865).