Metabolism of dimethylsulphoniopropionate by Ruegeria pomeroyi DSS-3


For correspondence. E-mail; Tel. (+1) 706 542 4219; Fax (+1) 706 542 2674.


Ruegeria pomeroyi DSS-3 possesses two general pathways for metabolism of dimethylsulphoniopropionate (DMSP), an osmolyte of algae and abundant carbon source for marine bacteria. In the DMSP cleavage pathway, acrylate is transformed into acryloyl-CoA by propionate-CoA ligase (SPO2934) and other unidentified acyl-CoA ligases. Acryloyl-CoA is then reduced to propionyl-CoA by AcuI or SPO1914. Acryloyl-CoA is also rapidly hydrated to 3-hydroxypropionyl-CoA by acryloyl-CoA hydratase (SPO0147). A SPO1914 mutant was unable to grow on acrylate as the sole carbon source, supporting its role in this pathway. Similarly, growth on methylmercaptopropionate, the first intermediate of the DMSP demethylation pathway, was severely inhibited by a mutation in the gene encoding crotonyl-CoA carboxylase/reductase, demonstrating that acetate produced by this pathway was metabolized by the ethylmalonyl-CoA pathway. Amino acids and nucleosides from cells grown on 13C-enriched DMSP possessed labelling patterns that were consistent with carbon from DMSP being metabolized by both the ethylmalonyl-CoA and acrylate pathways as well as a role for pyruvate dehydrogenase. This latter conclusion was supported by the phenotype of a pdh mutant, which grew poorly on electron-rich substrates. Additionally, label from [13C-methyl] DMSP only appeared in carbons derived from methyl-tetrahydrofolate, and there was no evidence for a serine cycle of C-1 assimilation.


The marine phytoplankton metabolite dimethylsulphoniopropionate (DMSP) is ubiquitous in marine surface waters, making it one of the most abundant low-molecular-weight sources of carbon and reduced sulphur in the marine environment (Curson et al., 2011; Reisch et al., 2011a; Moran et al., 2012). Marine bacteria consume DMSP through two competing biochemical pathways, the demethylation pathway resulting in the release of methanethiol (MeSH) or the cleavage pathway producing dimethylsulphide (DMS). While some marine bacteria only possess one of these pathways, the model organism Ruegeria pomeroyi DSS-3 utilizes both (Gonzalez et al., 1999). Recently, the biochemical pathway and genes responsible for the demethylation pathway were elucidated (Howard et al., 2006; Reisch et al., 2011b). This pathway starts with demethylation of DMSP by a tetrahydrofolate (THF)-dependent enzyme, DmdA, producing 5-methyl-THF and methylmercaptopropionate (MMPA). MMPA is then catabolized in a series of coenzyme-A mediated reactions analogous to fatty acid β-oxidation. The terminal step of the pathway results in the release of MeSH, CO2, acetaldehyde and free CoA. Acetaldehyde is further oxidized to acetate by an acetaldehyde dehydrogenase. Thus, the carbon from this pathway enters central carbon metabolism as acetate. R. pomeroyi is missing isocitrate lyase, the key enzyme in the glyoxylate shunt, but it does possess homologues for all the genes of the ethylmalonyl-CoA pathway (Erb et al., 2009). Therefore, R. pomeroyi was hypothesized to use the ethylmalonyl-CoA pathway to metabolize the DMSP carbon that is routed through the demethylation pathway.

In R. pomeroyi, four different genes, dddD, dddP, dddQ and dddW, encode for enzymes that catalyse the cleavage of DMSP, the initial step in the cleavage pathway. These enzymes are not homologous and represent different protein families with the same activity. However, only three of them appear to be physiologically relevant during growth on DMSP. Only mutations in dddP, dddQ and dddW decreased DMSP cleavage by whole cells, while a mutation in dddD had no effect (Todd et al., 2011; 2012b). These three enzymes all form acrylate in addition to DMS (Kirkwood et al., 2010; Todd et al., 2011; 2012b). However, the metabolism of acrylate in R. pomeroyi and other marine bacteria is poorly understood. Acrylate metabolism in a strain of Halomonas was extensively investigated by recombinant expression of several genes in Escherichia coli (Todd et al., 2010). This work proposed a scheme in which acrylate is hydrated to 3-hydroxypropionate, which is further oxidized to malonate-semialdehyde. Malonate-semialdhyde is then decarboxylated, and acetyl-CoA is formed. Whether or not the first three steps are CoA-mediated reactions was not determined as these investigations were carried out in whole cells of E. coli and the enzymes were not purified.

Acryloyl-CoA is also part of the 3-hydroxypropionate pathway for CO2 fixation described in the green non-sulphur phototrophic bacterium Chloroflexus auranticus and the thermoacidophilic Archaea (Alber and Fuchs, 2002; Berg et al., 2007). In this pathway, hydroxypropionate is converted to its CoA thioester, hydroxypropionyl-CoA, and then dehydrated to acryloyl-CoA before reduction to propionyl-CoA. In C. auranticus, these reactions are catalysed by a trifunctional fusion protein. In contrast, members of the thermoacidophilic archaea Sulfolobales possess individual enzymes capable of catalysing each of the three reactions.

Clostridium was proposed to possess a pathway capable of metabolizing lactoyl-CoA to acryloyl-CoA (Kuchta and Abeles, 1985; Hetzel et al., 2003). Like the pathways described above, acryloyl-CoA can then be directly reduced to propionyl-CoA. None of the enzymes that constitute this pathway were identified, and it is not known if these enzymes were related to the enzymes identified in Sulfolobales or C. auranticus.

Recently, it was proposed that Rhodobacter sphaeroides metabolizes 3-hydroxypropionate through a CoA-mediated pathway involving the dehydration of 3-hydroxypropionyl-CoA to acryloyl-CoA and then reduction to propionyl-CoA (Schneider et al., 2012). The enzyme that catalyses the reduction of acryloyl-CoA in R. pomeroyi was recently identified by its ability to confer resistance to acrylate toxicity (Todd et al., 2012a).

In this report, the pathways for DMSP metabolism in R. pomeroyi were investigated. Three general routes were evaluated using 13C isotopic labelling, targeted gene mutagenesis and transcriptional analyses. One, DMSP carbon routed through the cleavage pathway was found to be metabolized through acrylyl-CoA and propionyl-CoA (Fig. 1). Two of the three enzymes that constitute this pathway were identified by purification from cell extracts and confirmed by recombinant expression. Two, the demethylation pathway was shown to metabolize the DMSP carbon as acetate through the ethylmalonyl-CoA pathway, which is common in bacteria lacking isocitrate lyase, the key enzyme of the glyoxylate cycle (Erb et al., 2007). The transcriptional response of R. pomeroyi to DMSP measured in microarray studies and the specific labelling patterns of amino acids and nucleosides following growth on [1-13C] DMSP were consistent with these pathways. Finally, the fate of DMSP methyl groups was also investigated by using a 13C tracer. This label was only incorporated into compounds biosynthesized via tetrahydrofolate-dependent pathways, indicating that cells do not use a complete serine cycle for assimilation of C-1 compounds when grown on DMSP.

Figure 1.

Proposed pathway of DMSP cleavage and acrylate metabolism in R. pomeroyi. The pathway to propionyl-CoA comprises DMSP lyase catalysed by DddP (SPO2299), DddQ (SPO1596) or DddW (SPO0453); acrylate-CoA ligase catalysed by PrpE (SPO2934); and acryloyl-CoA reductase catalysed by AcuI (SPO1914). A side reaction forming 3-hydroxypropionyl-CoA is catalysed by acryloyl-CoA hydratase (SPO0147).


Acrylate-CoA ligase

Assuming that free acrylate was the product of DMSP cleavage by DddP, DddQ or DddW (Todd et al., 2011), it was hypothesized that acryloyl-CoA was the next intermediate in the pathway of acrylate metabolism. To test this hypothesis, acrylate-CoA ligase activity was assayed in crude cell extracts of R. pomeroyi grown in a chemostat with DMSP as the sole source of carbon. Cell-free extracts provided with acrylate, HS-CoA and ATP produced acryloyl-CoA at a rate of 24 nmol min−1 mg of protein−1 (Table 1), which was sufficient to consume all of the substrate expected to pass through the cleavage pathway (see Experimental procedures for calculation). In contrast, acyl-CoA transferase activities from acetyl- or propionyl-CoA to acrylate were < 1 nmol min−1 mg of protein−1. Therefore, an acyl-CoA ligase was the likely source of acryloyl-CoA.

Table 1. Specific activities of the acrylate metabolic enzymes in cell extracts of chemostat-grown R. pomeroyi
ActivityGrowth substrate
  1. Values are nmol min−1 mg of protein−1 and the result of triplicate assays from a single cell extract ± SD.
Acrylate-CoA ligase18 ± 124 ± 2
Acryloyl-CoA hydratase> 8000> 8000
Acryloyl-CoA reductase16 ± 1195 ± 10
Propionyl-CoA carboxylase4 ± 138 ± 1

It was hypothesized that the enzyme catalysing the ligase reaction was encoded by SPO2934, which was annotated as propionate-CoA ligase (prpE, EC# This enzyme functions in the methylmalonyl-CoA pathway of propionate assimilation, and the enzymes from Ralstonia solanacearum and Salmonella choleraesuis possessed activity with both acrylate and propionate (Rajashekhara and Watanabe, 2004). The R. pomeroyi gene has 45% sequence similarity to the E. coli prpE as well as high similarity to the other acyl-CoA ligase/synthetases in its genome. It is also found in an apparent operon with the gene for malic enzyme, which may also play a role in acrylate metabolism (see below). To investigate the prpE from R. pomeroyi, the gene was cloned and expressed in E. coli. Cell-free extracts of the recombinant E. coli had activity with both propionate and acrylate, while the host strain alone did not, supporting the hypothesis that the enzyme catalysed both reactions in vivo. In addition, the microarray analysis showed that the prpE gene was upregulated when grown on DMSP (see below, Table 2).

Table 2. Microarray analysis of the R. pomeroyi genes of interest to DMSP metabolism that were significantly regulated during growth on DMSP compared with glucose
Gene annotationSPO No.aFold changeq-value
  1. aWhen possible each gene was spotted with two unique probes. Instances where each probe showed a significant response or a q-value < 10 are reported.
Acetoacetyl-CoA reductaseSPO03253.51.6
Ethylmalonyl-CoA mutaseSPO03683.34.5
dddW (DMSP lyase)SPO045379.00.0
dddW (DMSP lyase)SPO045373.90.0
Methylmalonyl-CoA epimeraseSPO09327.80.0
soxS; regulatory protein SoxS;SPO09904.70.0
soxV; sulphur oxidation V protein;SPO09912.53.6
soxX; monohaem cytochrome c SoxX;SPO09933.50.0
soxX; monohaem cytochrome c SoxX;SPO09937.00.0
soxY; sulphur oxidation Y protein;SPO09943.30.0
soxZ; sulphur oxidation Z protein;SPO09952.83.6
soxB; sulphur oxidation B protein;SPO09976.20.0
soxC; sulphur oxidation molybdopterin C protein;SPO099810.10.0
soxE; dihaem cytochrome c SoxE;SPO10003.80.0
soxF; sulphur oxidation F protein;SPO10013.91.6
Propionyl-CoA carboxylase alpha subunitSPO11013.53.6
Propionyl-CoA carboxylase alpha subunitSPO11015.60.0
Methylmalonyl-CoA mutaseSPO11052.50.0
Serine hydroxymethyltransferaseSPO15724.81.6
Serine hydroxymethyltransferaseSPO15726.74.5
Transcriptional regulator, GntR familySPO19124.40.0
dmdA (DMSP demethylase)SPO19137.30.0
dmdA (DMSP demethylase)SPO19134.70.0
Acryloyl-CoA reductaseSPO191413.30.0
dddP (DMSP lyase)SPO22995.90
cysH (phosphoadenylyl-sulphate reductase)SPO26350.23.8
prpE (Propionate-CoA ligase)SPO29347.50.0
prpE (Propionate-CoA ligase)SPO29343.03.6
Transcriptional regulator, IclR familySPOA026813.00.0
Transcriptional regulator, IclR familySPOA026810.40.0
Hypothetical proteinSPOA026929.80.0
Hypothetical proteinSPOA026923.60.0
Hypothetical proteinSPOA02702.28.9
Glutathione-dependent formaldehyde dehydrogenaseSPOA02723.03.4

A mutant strain of R. pomeroyi was constructed in which a tet resistance cassette replaced most of the prpE gene. This mutant strain grew on propionate similarly to wild-type (Fig. 2A). This phenotype was also observed in a prpE mutant of Salmonella typhimurium. In S. typhimurium a second mutation in the acetyl-CoA synthetase gene impaired the ability to grow on propionate, indicating that the acetyl-CoA synthetase was capable of complementing prpE function (Horswill and Escalante-Semerena, 1999). Likewise, R. pomeroyi has a number of enzymes which might possess propionate-CoA ligase activity, including an acetyl-CoA synthetase as well as two forms of DmdB (H. Bullock and C. Reisch, unpubl. data). Thus, several enzymes may contribute to the ability of the prpE mutant to grow on propionate. The R. pomeroyi prpE mutant was also able to grow on acrylate and DMSP, although the growth rates were much decreased as compared with wild-type (Fig. 2A). Again, the presence of several additional CoA-ligases may have contributed to the ability of this mutant to grow on acrylate. However, the diminished growth rate of the prpE mutant supports the hypothesis that acrylate-CoA ligases initiate the pathway of acrylate metabolism and are part of the DMSP-cleavage pathway.

Figure 2.

Growth of wild-type R. pomeroyi and the prpE, acuI and ccr mutant strains.

A. Growth of wild-type R. pomeroyi and the prpE mutant SPO2934::tet. Wild-type cells grown with 3 mM acrylate (□), DMSP (○) and propionate (Δ). SPO2934::tet grown with 3 mM acrylate (■), DMSP (●) and propionate (▲).

B. Growth of wild-type R. pomeroyi and the acuI mutant SPO1914::tet. Wild-type cells grown with 3 mM propionate (Δ), acrylate (□) and 3-hydroxypropionate (○). SPO1914::tet grown with 3 mM propionate (▲), acrylate (■), 3-hydroxypropionate (●).

C. Growth of wild-type R. pomeroyi and ccr mutant SPO0370::tet. Growth of wild-type cells with 3 mM DMSP (○), 3 mM MMPA (Δ) and 5 mM acetate (□). Growth of SPO0370::tet with 3 mM DMSP (●), MMPA (▲) and 5 mM acetate (■).

Acryloyl-CoA hydratase

In cell-free extracts, acryloyl-CoA was rapidly converted to a CoA-containing intermediate that did not co-elute during HPLC with the standards compounds previously tested (Reisch et al., 2011b). To identify this compound, it was collected after HPLC separation and analysed by Fourier Transformed Ion Cyclotron Resonance mass spectrometry (FTICR). The molecular mass was 839.14 Da, which was equal to the mass of acryloyl-CoA plus one water molecule. This datum suggested that acryloyl-CoA was hydrated to either 2- or 3-hydroxypropionyl-CoA. Since standards for these two compounds were neither commercially available nor easily synthesized, 1H NMR was used to distinguish between them. Upon 1H NMR analysis, the product of acryloyl-CoA hydration contained doublets at 2.6 and 3.8 ppm (data not shown), consistent with 3-hydroxypropionyl-CoA. If the product had been 2-hydroxypropionyl-CoA, a distinctive doublet corresponding to the C-3 methyl group would have been located at 1.3 ppm. Thus, it was concluded that the product of acryloyl-CoA hydration was 3-hydroxypropionyl-CoA. The specific activity of 3-hydroxypropionyl-CoA synthesis in cell extracts was > 8 μmol min−1 mg of protein−1, far exceeding the minimum activity required to support acrylate metabolism (see below). This exceedingly high rate was consistent with the enzymatic efficiency of enoyl-CoA hydratases, which have been reported to be limited only by the rate of substrate diffusion (Hamed et al., 2008).

The enzyme catalysing the acryloyl-CoA hydration was identified by purification from cell extracts. A three-step purification, consisting of anion exchange, hydrophobic interaction and hydoxyapaptite chromatography, yielded a protein purified to electrophoretic homogeneity (Fig. S1). The protein-encoding gene was identified by in-gel trypsin digestion and MALDI-TOF mass fingerprinting as SPO0147. It was annotated as an enoyl-CoA hydratase, and possessed 46% sequence similarity to an enoyl-CoA hydratase from E. coli as well as high similarity to other enoyl-CoA hydratases from Alphaproteobacteria. In R. pomeroyi, it is found in an apparent operon with dnaA and dnaN, two genes involved in DNA replication. To confirm that this gene encoded for a protein with the correct catalytic function, the gene was cloned and expressed in E. coli. Cell extracts of the recombinant E. coli possessed acryloyl-CoA hydratase activity, while the host strain alone did not. However, several attempts to construct a mutation in this gene were unsuccessful, so it was not possible to explore its function in vivo further.

Acryloyl-CoA reductase

The fate of 3-hydroxypropionyl-CoA was next investigated in enzyme assays using cell-free extracts. In the absence of exogenous cofactors, cell-free extracts did not consume 3-hydroxypropionyl-CoA. Upon the addition of NADH or NADPH, there was a quantitative conversion of 3-hydroxypropionyl-CoA to propionyl-CoA. This activity could be due to either a 3-hydroxypropionyl-CoA reductase activity, which has never been previously described, or coupling of the reverse activity of acryloyl-CoA hydratase with an acryloyl-CoA reductase (Fig. 1). To clarify these results, the 3-hydroxypropionyl-CoA reductase activity was partially purified from extracts of chemostat-grown cells. One of the three proteins remaining on a SDS-PAGE gel was identified by peptide mass fingerprinting as a zinc-dependent oxidoreductase encoded by gene SPO1914. To confirm the function of this gene product, SPO1914 was cloned and expressed in E. coli. The partially purified recombinant protein had activity for acryloyl-CoA reductase but not 3-hydroxypropionyl-CoA reductase. Thus, the 3-hydroxypropionyl-CoA reductase activity observed in cell extracts resulted from the coupling of the acryloyl-CoA hydratase with an acryloyl-CoA reductase activity.

Chemostat-grown R. pomeroyi possessed an NADPH-dependent acryloyl-CoA reductase activity of 195 ± 10 nmol min−1 mg of protein−1 (Table 1), which was well above the minimum rate required to support chemostat growth. In contrast, cells grown on glucose possessed an activity of only 16 nmol min−1 mg of protein−1. Furthermore, DMSP-grown cells possessed propionyl-CoA carboxylase activity of 38 nmol min−1 mg of protein−1, while glucose-grown cells possessed lower activities of only 4 nmol min−1 mg of protein−1. Propionyl-CoA carboxylase is required in the methylmalonyl-CoA pathway for C-3 metabolism. Thus, these observations support the hypothesis that acrylate was metabolized through this pathway in R. pomeroyi.

To confirm the physiological significance of the acryloyl-CoA reductase activity, a mutant strain of R. pomeroyi was constructed in which SPO1914 was disrupted. The mutant was incapable of growth on acrylate or 3-hydroxypropionate (Fig. 2B). In contrast, it grew similarly to wild-type when provided with propionate as the sole source of carbon. These results were consistent with the role of this enzyme in catalysing the reduction of acryloyl-CoA to propionate during acrylate metabolism. The mutant strain also grew sporadically on DMSP. Most inoculations failed to grow at all. However, growth of the 30% which did grow was comparable to that of wild-type (data not shown). This result was unexpected as growth on DMSP should be possible since the demethylation pathway was uninterrupted. While the reason for this irregular growth phenotype was not clear, one possibility was that a build-up of acryloyl-CoA or 3-hydroxypropionyl-CoA in these cells caused a metabolic collapse due to shortage of free CoA. Rare second site mutations present in low numbers in the inoculum may have overcome this collapse in some of the inoculations. Lastly, the transcriptional response of gene SPO1914 was consistent with it being involved in DMSP metabolism, and in microarray experiments it was upregulated 14-fold after growth on DMSP (Table 2).

Based in part upon its location, SPO1914 had previously been implicated in conferring acrylate resistance in R. pomeroyi as well as other proteobacteria (Todd et al., 2012a). In R. pomeroyi, SPO1914 is adjacent to and predicted to be within the same transcriptional unit as dmdA, which encodes the enzyme for the first step of the demethylation pathway. Candidatus Puniceispirillum marinum IMC1322, a member of the SAR116 clade of Alphaproteobacteria, also possessed an acryloyl-CoA reductase homologue with a protein identity of 62%. Interestingly, in this bacterium the gene was positioned immediately upstream of a dddP homologue, which encoded a DMSP-cleavage enzyme (Oh et al., 2010), providing circumstantial evidence for a role in DMSP metabolism in this bacterium as well. Similarly, the R. sphaeroides homologue was coexpressed with the dddL gene during growth on DMSP (Sullivan et al., 2011). This bacterium does not grow on acrylate, and expression of this gene is regulated by acrylate. It is also implicated in increased resistance to acrylate toxicity and acrylate degradation by resting cells. In R. sphaeroides, the SPO1914 homologue has also been implicated in growth on 3-hydroxypropionate and catalysing the reduction of acryloyl-CoA to propionyl-CoA (Schneider et al., 2012).

Ethylmalonyl-CoA pathway for acetate and MMPA metabolism

To fully understand growth of R. pomeroyi on DMSP, it is also important to know how carbon is metabolized during the demethylation pathway. Previously, the demethylation pathway was found to form acetaldehyde (Reisch et al., 2011b). Because cell extracts contained high specific activities of acetaldehyde dehydrogenase, acetate was hypothesized to be the first common intermediate of central metabolism. R. pomeroyi does not possess an orthologue of isocitrate lyase, and the activity was not detectible in cell-free extracts (data not shown). Thus, this organism appears to be incapable of metabolizing acetate through the glyoxylate cycle. However, R. pomeroyi possesses homologues for all known genes in the ethylmalonyl-CoA pathway, which is an alternative means for acetate metabolism (Fig. 4). To confirm that the ethylmalonyl-CoA pathway was required for growth on acetate and MMPA, a mutation in the ccr gene (SPO0370) was constructed. This gene encodes the crotonyl-CoA carboxylase/reductase enzyme, which is indicative of the ethylmalonyl-CoA pathway (Erb et al., 2007). The mutant strain was unable to grow on acetate as the sole source of carbon (Fig. 2C), confirming that acetate was metabolized through the ethylmalonyl-CoA pathway. When MMPA was provided as the sole source of carbon, the mutant displayed an extended lag phase, and the growth yield was diminished by two-thirds (Fig. 2C). This result was unexpected, as growth on MMPA was expected to behave the same as acetate. One of two explanations appear likely. In other experiments, slow growth of both the wild-type and mutant were supported by trimethylamine and other C-1 compounds (data not shown). Currently, R. pomeroyi is not believed to grow with C-1 compounds as sole carbon sources, although it will use them as N sources and electron donors (Chen et al., 2011; Chen, 2012). However, it is difficult to rigorously exclude the low levels of formaldehyde assimilation by the serine cycle necessary to support the slow growth observed. An alternative is that the organic HEPES buffer in the medium served as a carbon source and the C-1 compounds served as electron donors under these conditions. If this were true, growth of the ccr mutant on MMPA would be primarily supported by oxidation of the methanethiol derived from MMPA and carbon from HEPES. While it is not possible to chose between these explanations at this time, the severe growth phenotype of the ccr mutant provided strong evidence for the role of the ethylmalonyl-CoA pathway in metabolism of carbon from DMSP demethylation. In contrast to the poor growth on MMPA, the mutant was able to grow on DMSP as the sole source of carbon, demonstrating that the cleavage pathway was still functional. These results also indicated that the product of the cleavage pathway does not enter central carbon metabolism as acetate, which was consistent with the acrylate pathway identified here.

Transcriptional response of R. pomeroyi to DMSP

To study the transcriptional response of R. pomeroyi during growth on DMSP, whole-genome microarrays were used. Steady-state R. pomeroyi grown in a carbon-limited chemostat was used in this investigation to minimize the differences in growth rate and cell density common in batch cultures grown on different substrates. Probes with a false-discovery rate (q-value) of less than 10% and whose gene annotation was of particular interest to DMSP metabolism are listed in Table 2.

Several genes involved in acetate and propionate metabolism were upregulated during growth on DMSP, which was consistent with the pathways presented here. Three probes that represent genes unique to the ethylmalonyl-CoA pathway, acetoacetyl-CoA reductase, β-ketothiolase and ethylmalonyl-CoA mutase, were all significantly upregulated during growth on DMSP. Similarly, at least one probe from all four genes that constitute the methylmalonyl-CoA pathway, propionate-CoA ligase, propionyl-CoA carboxylase, methylmalonyl-CoA epimerase and methylmalonyl-CoA mutase, were upregulated.

One of the two gene clusters with the highest expression on DMSP included SPOA0268-272. This cluster encodes a transcriptional regulator, two proteins of unknown function, a methylamine utilization protein (mauG) and a glutathione-dependent formaldehyde dehydrogenase. While the roles of these genes in DMSP metabolism were unclear, the increased abundance of the protein encoded by SPOA272 was also observed in proteomic experiments, and it was hypothesized that the proteins may participate in the oxidation of methanethiol or dimethylsulphide (Henriksen, 2008).

The second gene cluster that was highly upregulated during growth on DMSP was SPO0989–SPO1001, where many but not all genes showed significant upregulation (Table 2). These genes encode the Friedrich-Kelly pathway of sulphur oxidation, known as the SOX system (Friedrich et al., 2001). This pathway functions to completely oxidize inorganic sulphide to sulphate and may be used for oxidation of H2S formed from MeSH. In contrast, the cysH gene that is part of the assimilatory sulphate reduction pathway was downregulated with DMSP. These data suggest that the cells are also assimilating the reduced sulphur from DMSP, which is consistent with hypothesis that marine bacteria preferentially assimilate DMSP sulphur due to the energetic costs associated with sulphate reduction (Kiene et al., 1999).

The most upregulated gene in the microarray experiments, with an increased expression of over 70-fold on each of two probes, was SPO0453. This gene was recently identified as dddW, encoding for a functional DMSP cleavage enzyme in R. pomeroyi (Todd et al., 2012b). However, it was not the only DMSP lyase that was upregulated in these experiments, as dddP had a fivefold increase in relative expression (Kirkwood et al., 2010). The gene that initiates the demethylation pathway, dmdA, was also upregulated with growth on DMSP, although the increase was much less dramatic than dddW. Although the other genes that constitute the demethylation pathway were not significantly regulated in these microarrays, RT-qPCR performed on RNA extracted from cells grown under similar conditions did find significant increases in transcript abundance (Reisch et al., 2011b). Thus, the absence of a significant response in the microarray experiments probably resulted from their low sensitivity.

Retrobiosynthetic analyses with 13C-labelling

The demethylation and cleavage pathways proposed above imply different routes for assimilation of DMSP carbon. In contrast, it had been previously proposed that acrylate was an intermediate in both pathways, implying a single entry point of DMSP carbon into central metabolism (Kiene et al., 2000). Likewise, the R. pomeroyi genome possesses homologues to many of the genes of the serine cycle for C-1 assimilation, suggesting that the methyl carbons of DMSP might also be significant carbon sources. To test these hypotheses, chemostat-grown cultures were labelled with [13C-methyl] or [13C-1] DMSP, and the labelling of cellular amino acids and nucleic acids was determined.

Metabolism of the methyl carbons of DMSP

To investigate the fate of DMSP-methyl carbons, R. pomeroyi was grown in a carbon-limited chemostat with [13C-methyl] DMSP. While there is little evidence that chemostat-grown R. pomeroyi assimilates carbon from DMS (Reisch et al., 2011b), 5-methyl-THF and MeSH formed from the demethylation pathway are potential routes for C-1 carbon assimilation (Gonzalez et al., 1999; Kiene et al., 1999; Reisch et al., 2008). During growth with [13C-methyl] DMSP where each methyl carbon was enriched to 99% 13C, the 13C/12C of the cells produced was 8.2%. Based upon total cellular C production rate of 300 nmol C min−1 and correcting for the natural abundance of 13C in the unlabelled C, only a small portion or 21 nmol min−1 of DMSP methyl carbons were assimilated under these conditions. Given that 160 nmol min−1 of the 400 nmol min−1 of methyl C metabolized were used to form DMS, the balance of 219 nmol min−1 was presumably released as carbonate (see below).

Analysis of the amino acids and nucleosides obtained from cells confirmed this conclusion (Fig. 3). Most amino acid carbons contained only the natural abundance of 13C. Exceptions were the methyl group of methionine and the C-3 position in serine. After purification of the benzoyl derivatives of these amino acids, the 13C enrichments were determined by quantification of the satellite signals by 1H NMR. The enrichments of these carbons were 99% and 30%, respectively, and the labelling of the remaining carbons was < 2%. Similarly, 1H NMR of purified guanosine revealed an enrichment of 90% in the C-11. Based upon the cellular composition for these amino acids and purines, about 4.2 nmol min−1 of the methyl-C of DMSP could be incorporated into these compounds. Assuming that the remaining methyl-C is oxidized to carbonate and a portion of this labelled carbonate is incorporated into cellular C (see below), the enrichment of whole cells was estimated to be 9.8% or close to the observed value of 8.2%. These results indicated that assimilation of the methyl carbons of DMSP was limited to carbons derived directly from 5-methyl-THF or MeSH and that the serine cycle for C-1 assimilation was not a major route of carbon metabolism under these growth conditions.

Figure 3.

13C NMR analysis of hydrolysed proteins from chemostat-grown R. pomeroyi using [13C-methyl] DMSP. Signal intensities are attributed to both amino acid abundance and 13C enrichment. The signal at 17 ppm corresponds to the methyl group of methionine, while the signal at 64 ppm corresponds to the C-3 carbon of serine.

Figure 4.

Overview of DMSP assimilation pathways in R. pomeroyi. The schematic starts with DMSP and displays reactions and labelling patterns of intermediates of central carbon metabolism. Values displayed next to filled or partially filled circles are percentages of 13C enrichment in the corresponding carbon. At key steps, the carbon fluxes in nmol min−1 are given in brackets. The complete overview of the fluxes is given in Fig. 5. Numbers in the black boxes correspond to enzymes as follows: 1, DMSP-dependent demethylase (DmdA [SPO1913]); 2, MMPA-CoA ligase (DmdB [SPO0677, 2045]); 3, MMPA-CoA dehydrogenase (DmdC [SPO0298, 2915, 3804]); 4, MTA-CoA hydratase (DmdD [SPO3805]); 5, acetaldehyde dehydrogenase and acetate-CoA ligase [not identified]; 6, DMSP cleavage enzyme (DddP [SPO2299], DddQ [SPO1596] or DddW [SPO0453]); 7, Acrylate-CoA ligase [SPO2634 and other unidentified activities]; 8, Acryloyl-CoA reductase [SPO1914]; 9, Propionyl-CoA carboxylase [SPO1101]; 10, Methylmalonyl-CoA mutase [SPO1105]; 11, Methylmalonyl-CoA epimerase [SPO0932]; 12, β-ketothiolase [SPO0326]; 13, Acetoacetyl-CoA reductase [SPO0325]; 14, Crotonase [not assigned]; 15, Crotonyl-CoA carboxylase/reductase [SPO0370]; 16, Ethylmalonyl-CoA/methylmalonyl-CoA epimerase [SPO0932]; 17, Ethylmalonyl-CoA mutase [SPO0368]; 18, Methylsuccinyl-CoA dehydrogenase [SPO0693]; 19, Mesaconyl-CoA hydratase [SPO0355]; 20, Malyl-CoA/β-methylmalyl-CoA lyase [SPO3608]; 21, Malyl-CoA thioesterase [not assigned]; 22, Malic enzyme [SPO0012, 2932]; 23, Pyruvate dehydrogenase complex [SPO2240-2242]; 24, Citrate synthase [SPO2157].

Figure 5.

A. Carbon fluxes during DMSP assimilation and oxidation. Only key intermediates of the demethylation and cleavage pathways are shown.

B. Oxidative ethylmalonyl-CoA pathway for growth on electron-rich substrates.

Metabolism of the propionyl carbons from DMSP

Likewise, only a small portion of the label from [1-13C] DMSP was assimilated during chemostat growth. During growth with 99% enriched [1-13C] DMSP, the 13C/12C of the cells and carbonate (sum of CO2 + HCO3 + H2CO3) produced were 9.2% and 28% respectively. Based upon total cellular and carbonate production rates of 300 and 540 nmol C min−1, respectively, the 13C-cellular and 13C-carbonate production rates were 28 and 171 nmol min−1 (Table 3). The sum of these values, 199 nmol min−1, was close to the expected value of 206 nmol min−1, verifying the nearly complete metabolism of DMSP and the accuracy of the measurements. The small amount of label appearing in cells suggested that most of the C-1 carbon of DMSP was oxidized to CO2.

Table 3. Carbon flux balance during growth in a chemostat on 2 mM [1-13C] DMSP. Units are nmol min−1
 Total C (In)13Ca (In)Total C (Out)13Ca (Out)
  1. aIncludes natural abundance of unlabelled carbon and 99% enriched [13C-1] DMSP
  2. bcalculated
  3. cBased upon a measured enrichment of 0.28 in carbonate produced from the chemostat and the production of 540 nmol carbonate min−1 from DMSP and 70 nmol carbonate min−1 due to aeration.
  4. dBased upon a measured enrichment of 0.092 in the cell carbon and the production of 300 nmol cellular carbon min−1.
cell material0030028d

Based upon the expected fluxes through the demethylation and cleavage pathways and the levels of intermediates needed during growth, the fluxes of a number of possible pathways were solved algebraically for the observed levels of CO2 production. During growth in the chemostat, the flux through the cleavage pathway was equal to the amount of DMS produced or 80 nmol min−1 and yielded acryloyl-CoA. The remaining 120 nmol min−1 was routed through the demethylation pathway and produced acetaldehyde, which was metabolized to acetate and acetyl-CoA. Acetyl-CoA was assumed to be metabolized by the ethylmalonyl-CoA pathway. The fluxes of intermediates necessary to support growth of 300 nmol min−1 cellular C were 21 nmol min−1 of acetyl-CoA, 20 nmol min−1 of pyruvate, 50 nmol min−1 of oxaloacetate and 7 nmol min−1 of α-ketogluturate (Table S2). For more details, see Experimental procedures. Given these constraints, it was not possible to solve pathways that did not include a significant flux through the pyruvate dehydrogenase complex (Figs 4 and 5). Moreover, pathways could not be solved that included a significant role for malate and α-ketogluturate oxidation via the TCA cycle or the serine cycle enzymes for the transformation of glyoxylate to either phosphoenolpyruvate for carbon assimilation or malyl-CoA for oxidation in the TCA cycle (data not shown).

Based upon this predicted pathway, the 13C-labelling of whole cells was expected to be 8.8% or close to the measured value of 9.2%. The labelling of the internal carbonate pool was assumed not to be in equilibrium with the external carbonate pool. For the labelling of the internal pool, 195 nmol min−1 of C-1 DMSP carbons were estimated to be oxidized to carbonate, and the total carbonate production from DMSP was estimated to be 592 nmol min−1. After accounting for the natural abundance, the internal pool carbonate pool was estimated to be enriched by 34%. Given the dilution of the net cellular production of carbonate of 540 nmol min−1 by 70 nmol min−1 carbonate from aeration, the enrichment of the external carbonate pool was estimated to be 30% or close to the observed value of 28%. These comparisons between the estimated and observed enrichments for cells and carbonate provided a further test for the proposed pathway.

The labelling patterns of key amino acids from [1-13C] DMSP supported this pathway of carbon metabolism. The C-1 and C-4 carbons of aspartate were enriched by 40% and 32% respectively (Table 4). This pattern was consistent with the formation of aspartate from malate via oxaloacetate and two sources of malate in the DMSP-grown cells (Fig. 5). Part of the malate would be formed from succinyl-CoA via succinate (Fig. 4). Because succinate is symmetrical, the enrichment of the C-1 and C-4 carbons formed via this route would be identical. The remaining malate would be formed from malyl-CoA via the ethylmalonyl-CoA pathway. For this malate, the C-1 would be enriched due to the incorporation of enriched CO2, but the C-4 would not be enriched. For the fluxes calculated in Fig. 4, the theoretical enrichments for the C-1 and C-4 carbons of aspartate were 37% and 26%, respectively, or close to the observed enrichments. Similarly, the theoretical enrichment for the C-1 of pyruvate, 37%, was close to the observed value of 39%.

Table 4. 13C enrichment of amino acids following growth with [1-13C] DMSP.a
Amino acidC-1C-2C-3C-4C-5
  1. aCells were grown with 2 mM of 99% enriched [1-13C] DMSP in the chemostat.
  2. Enrichment was determined by quantitative 13C NMR or 1H coupled satellites (values in parentheses).
Alanine39%< 2%< 2%NANA
Glutamate40%< 2%< 2%< 2%3.9%
Aspartate40%< 2%< 2%32%NA
Threonine32%< 2%< 2%34% (36%)NA

Glutamic acid, which was derived from α-ketoglutarate, had an enrichment of 40% and 3.9% 13C for the C-1 and C-5 carbons respectively (Table 4). The expected enrichment of the C-1 carbon if it was derived from the C-4 of malate would be 32%. Likewise, no enrichment would be expected for the C-5 if it was derived from the C-2 of acetate (Fig. 4). Therefore, this labelling pattern suggested that another source of α-ketoglutarate was present. One possibility was the formation of α-hydroxyglutarate from glyoxylate and propionyl-CoA and the oxidation of α-hydroxyglutarate to α-ketoglutarate (Reeves and Ajl, 1962; Wegener et al., 1968). In this pathway, the C-5 of α-ketoglutarate would be highly labelled by the C-1 of propionyl-CoA. If as little as 6% of the α-ketoglutarate was formed by this pathway, it would explain the enrichment of 3.9% for the C-5 carbon of α-ketoglutarate. Presumably, the difference between the observed and expected enrichment for the C-1 carbon represents experimental error.

Threonine was expected to be biosynthesized from aspartic acid via homoserine and have the same labelling pattern. However, quantitative 13C NMR analysis of threonine showed enrichments of 32% and 34% for the C-1 and C-4 carbons respectively (Table 4). The enrichment of the C-4 methyl group was confirmed by quantification through the 13C coupled satellite signals in the 1H NMR spectrum, which yielded an enrichment of 36%. R. pomeroyi as well as most Alphaproteobacteria lack homologues to thrB, which encodes the homoserine kinase for threonine biosynthesis. Thus, threonine biosynthesis in R. pomeroyi may not proceed from aspartic acid and homoserine, which would explain the difference in the labelling patterns.

The labelling patterns of leucine and valine were consistent with their formation from pyruvate and acetyl-CoA as predicted by the canonical pathway for branched-chain amino acid biosynthesis (data not shown). However, isoleucine possessed no highly enriched carbons, which indicated that it was not derived from threonine (data not shown). This observation was consistent with the alternative pathway for isoleucine synthesis in which acetyl-CoA and pyruvate form citramalate (Ekiel et al., 1983; Risso et al., 2008). Isoleucine synthesized by this pathway would not contain any highly enriched carbons.

Phenotype of a pyruvate dehydrogenase mutant

To verify the role of the pyruvate dehydrogenase complex in DMSP metabolism, a mutant with a transposon insertion in SPO2240, the gene encoding the α-subunit of pyruvate dehydrogenase, was characterized. The mutant grew poorly on DMSP, supporting the proposed role of pyruvate dehydrogenase in DMSP metabolism (Fig. 6). Growth on MMPA and acetate was indistinguishable from wild-type. Because both of these substrates were also metabolized by the ethylmalonyl-CoA pathway, these results indicated that pyruvate dehydrogenase was not necessarily required for normal growth with this pathway. In contrast, the mutant was unable to grow with propionate as the sole carbon source. When both propionate and acetate were present, the mutant exhibited growth identical to the wild-type (Fig. 6). Similarly, the mutant also grew poorly on succinate, and growth was restored to wild-type levels by acetate (data not shown). Carbonate had no effect on growth of the mutant on propionate, indicating that poor growth was not due to a bicarbonate limitation for succinate biosynthesis. These results suggested that pyruvate dehydrogenase was required for acetyl-CoA biosynthesis during growth on propionate, succinate and DMSP (Fig. 5). Presumably, the different growth responses for substrates metabolized by similar pathways, such as DMSP and MMPA or propionate and succinate, reflected differences in regulation during growth with more reduced compounds.

Figure 6.

Growth phenotype of wild-type R. pomeroyi and the pdhA::Tn5 mutant pdh1.

A. Growth of the wild-type with 5 mM MMPA (Δ) and 3 mM DMSP (○) and pdh1 mutant with 5 mM MMPA (▲) and 3 mM DMSP (●).

B. Growth of the wild-type with 3 mM acetate (□), propionate (Δ) and 3 mM acetate + 3 mM propionate (◊) and the pdh1 mutant with 3 mM acetate (■) propionate (▲) and 3 mM acetate + 3 mM propionate (◆).


Acrylate metabolism

Ruegeria pomeroyi possesses two routes of DMSP catabolism. The first route, known as the demethylation pathway, is initiated by the enzyme DmdA, which transfers a methyl group from DMSP to THF, producing 5-methyl-THF and methylmercaptopropionate (MMPA). MMPA is then catabolized in a series of coenzyme-A-mediated reactions, releasing MeSH, CO2 and acetate (Todd et al., 2010; Reisch et al., 2011b). The second route is the DMSP cleavage pathway and results in the production of DMS and a three carbon moiety identified as acrylate or 3-hydroxypropionate. Four gene products in R. pomeroyi catalyse the cleavage reaction for DMS formation (Todd et al., 2007; 2009; 2011). Mutations in three of these genes, dddP, dddQ and dddW, affected DMS production during growth on DMSP and were functional under the conditions tested. In contrast, a mutation in the fourth gene, dddD, had no effect on DMSP-dependent growth, and its physiological importance is not clear. Upon purification, DddP was shown to produce acrylate in addition to DMS. Cell-free extracts of E. coli expressing DddW also formed acrylate, but the enzyme has not been purified. Likewise, DddQ has not been characterized in vitro, but in whole cells experiments E. coli expressing DddQ produced acrylate in the presence of DMSP, suggesting that acrylate was in fact the product of DddQ. One purpose of the investigations here was to establish the pathway for DMSP and acrylate assimilation in R. pomeroyi.

The genome of R. pomeroyi encodes three enzymes: pyruvate carboxylase, phosphoenolpyruvate carboxylase and propionyl-CoA carboxylase, which carboxylate C-3 substrates to form a C-4 moiety that could enter the TCA cycle and possibly be involved in metabolism of the C-3 moiety formed in DMSP cleavage (Moran et al., 2004). Of these genes, only propionyl-CoA carboxylase was upregulated in the microarray experiments during growth on DMSP. Furthermore, the other enzymes in the methylmalonyl-CoA pathway for C-3 metabolism were upregulated during growth on DMSP. These microarray results were consistent with previous experiments which found that DMSP caused a significant upregulation of propionate metabolism genes (Burgmann et al., 2007; Vila-Costa et al., 2010). Thus, these results are consistent with the role of the methylmalonyl-CoA pathway in DMSP metabolism.

Transcriptional response studies and bioinformatics analysis are complicated by the fact that a number of metabolic pathways share common intermediates and enzymes. For example, the ethylmalonyl-CoA pathway for acetate metabolism includes components of the methylmalonyl-CoA pathway for propionyl-CoA metabolism. Thus, observations of increased expression of C-3 metabolic genes may be a physiological response to C-2 compounds. Given the recent identification of the MMPA-CoA pathway, which results in acetate production, upregulation of propionate metabolic genes is expected regardless of whether the DMSP demethylation or cleavage pathway is being utilized. However, the propionate-CoA ligase gene, which is proposed to physiologically function as an acrylate-CoA ligase as well, is not part of the ethylmalonyl-CoA pathway. Thus, the upregulation of this gene during growth with DMSP is evidence for its participation in acrylate metabolism.

To determine the metabolic pathway for carbon routed through the cleavage pathway, enzyme assays were used to reconstruct the metabolic pathway. Enzyme assays revealed that cell extracts were capable of catalysing the production of acryloyl-CoA, which is then reduced to propionyl-CoA. In a side reaction, acryloyl-CoA is rapidly hydrated to 3-hydroxypropionyl-CoA. This side reaction may be a mechanism for protecting the cell against the toxic effects of acryloyl-CoA.

The recently described ethylmalonyl-CoA pathway is used for acetate metabolism in many isocitrate lyase-negative bacteria. The ccr gene encodes the crotonyl-CoA carboxylase/reductase, which carboxylates crotonyl-CoA to ethymalonyl-CoA and is indicative of this pathway (Erb et al., 2007). In the isocitrate lyase-negative bacteria Methylobacterium extorquens and Streptomyces coelicolor, a mutation in ccr yielded strains incapable of growth on acetate (Chistoserdova and Lidstrom, 1996; Han and Reynolds, 1997). Accordingly, a strain of R. pomeroyi with a disruption of the ccr gene was also incapable of growth on acetate and had a serious defect during growth on MMPA. These results confirm that the ethylmalonyl-CoA pathway is used for acetate metabolism in R. pomeroyi. While the mutant strain was able to grow on MMPA after an extended lag phase, the growth yield was greatly reduced compared with wild-type, supporting the hypothesis that MMPA is oxidized as acetate. The reason for this leaky phenotype is currently under investigation.

The addition of pyruvate dehydrogenase was also an interesting variation of the ethylmalonyl-CoA pathway utilized during growth on DMSP. As an alternative to the glyoxylate cycle in isocitrate lyase-negative bacteria, the ethylmalonyl-CoA pathway functions to replenish intermediates of the TCA cycle to allow for acetate oxidation (Erb et al., 2007). However, during growth on DMSP, the TCA cycle in R. pomeroyi was utilized primarily for oxaloacetate and α-ketoglutarate biosynthesis. Acetate oxidation was accomplished by cycling the C-4 intermediates through malic enzyme and pyruvate dehydrogenase (Fig. 5B).

Methyl carbon assimilation

Ruegeria pomeroyi grown in a chemostat with [13C-methyl] DMSP resulted in cell material with only a few carbons highly enriched in 13C. Each of these enrichments can be attributed to methyl group donations from THF single carbon carriers. Serine was enriched by 30% in the C-3 position. This labelling was consistent with the transfer of an enriched methylene group from methylene-THF to glycine, in a reaction catalysed by serine hydroxymethyltransferase. The remaining 70% of serine is probably biosynthesized via 3-phosphoglycerate derived from gluconeogenesis. Not unexpectedly, the serine hydroxymethyltransferase gene was significantly upregulated in microarray experiments.

There were two possible sources for the enriched C-1 unit that appeared in serine. First, 5-methyl-THF produced by DmdA during the initial demethylation of DMSP may be oxidized to 5-methylene-THF, which may then be directly utilized by serine hydroxymethyltransferase. The genome sequence of R. pomeroyi contains a metF homologue, which reduces methylene-THF to 5-methyl-THF. This reaction is reversible, so it is possible that the physiological reaction under these conditions was to oxidize 5-methyl-THF. Alternatively, the methyl group of methanethiol may be oxidized to formaldehyde, which spontaneously reacts with THF to form methylene-THF. Several enzymes that catalyse the oxidation of methanethiol to formaldehyde, hydrogen sulphide and hydrogen peroxide have been purified and characterized (Suylen et al., 1987; Gould and Kanagawa, 1992; Kim et al., 2000). However, the genes encoding these enzymes have not been identified, and this activity has not yet been examined in R. pomeroyi.

The methyl group of methionine was enriched to 99% 13C under these conditions. There are two possible explanations for this high enrichment. Previous experiments have shown that 35S sulphur from DMSP is assimilated into the cellular amino acids of marine bacteria (Kiene et al., 1999). It was hypothesized that a direct incorporation of methanethiol into homoserine was catalysed by cystathionine γ-synthase and may be responsible for this production of methionine (Kanzaki et al., 1987). However, the gene for this enzyme is absence from R. pomeroyi, so there may be an alternative source of this activity. R. pomeroyi possesses the gene for methionine γ-lyase, which catalyses the release of methanethiol from methionine. Growth of R. pomeroyi on methionine also results in the production of methanethiol (C.R. Reisch, unpubl. obs.), indicating that this activity is present in whole cells under these conditions. It is possible that this enzyme works in the reverse direction to catalyse the direct incorporation of methanethiol. The second possible source of highly enriched methyl groups is from 5-methyl-THF. The last step of the canonical pathway of methionine biosynthesis transfers a methyl group from 5-methyl-THF to homocysteine. Since the initial demethylation of DMSP produces 5-methyl-THF, there should be an abundance of highly enriched 5-methyl-THF available for methionine biosynthesis.

The biosynthesis of purine nucleosides is the third reaction for which a THF derivative provides a single carbon unit. Of the five carbons present in the purines, only the carbon derived from formyl-THF is highly enriched. Formyl-THF may be derived from two different sources. First, 10-methylene-THF is oxidized to 5-10-methenyl-THF and subsequently hydrated to formyl-THF. Two enzymes, MtdA or MtdB, that catalyse the first reaction have been identified in some Alphaproteobacteria, although not in R. pomeroyi (Vorholt et al., 1998; Hagemeier et al., 2000). R. pomeroyi does have two FolD homologues, which are annotated as methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase (DHCH) proteins. Functional DHCH proteins oxidize 5-10-methylene-THF to 5-10-methenyl-THF and hydrolyse the latter to form 10-formyl-THF (Dev and Harvey, 1978). Second, formyl-THF may be synthesized by formate-THF ligase. This ATP-dependent enzyme forms 10-formyl-THF from free formate and THF. The genome sequence of R. pomeroyi contains two formate-THF ligase homologues with identical protein sequences. The 90% enrichment in carbons derived from formyl-THF suggests that both pathways of formyl-THF synthesis may be active. If all of the formyl-THF was derived directly from methylene-THF, an enrichment of 99% would be expected, consistent with the enrichment of methionine. However, the data suggest that there is a source of formate that is not derived from the methyl groups of DMSP and would therefore dilute the enriched pool to 90%. R. pomeroyi possesses several genes annotated as formate dehydrogenases which could provide this unenriched carbon.

Experimental procedures

Substrate synthesis

DMSP was synthesized as described previously (Chambers et al., 1987) using 99% [1-13C] acrylic acid (Sigma-Aldrich, St. Louis, MO) and dimethylsulphide or [13C2]dimethylsulphide (Cambridge Isotopes, Cambridge, MA) and acrylic acid. Acryloyl-CoA was synthesized with acryloyl-chloride and free coenzyme-A as described previously (Kuchta and Abeles, 1985). The acryloyl-CoA was purified by reverse-phase chromatography using an Ultrasphere ODS preparative column (10 × 250 mm). The column was developed with 20 mM potassium phosphate (pH 6.8) and a gradient of 2–25% acetonitrile. Acryloyl-CoA was detected by its absorbance at 254 nm. Fractions containing acryloyl-CoA were lyophilized, resuspended in dH2O, and again lyophilized. 3-Hydroxypropionyl-CoA was synthesized enzymatically from acryloyl-CoA with purified SPO0147 as described below. The product was purified by reverse phase chromatography as described above except that the buffer was 20 mM ammonium acetate, pH 6.0.

Growth of cultures

Ruegeria pomeroyi was grown at 30°C in a 144 ml carbon-limited chemostat with marine basal medium as described previously (Reisch et al., 2008) using 2 mM DMSP at a flow rate of 0.1 ml min−1 and a dilution rate of 0.0416 h−1. For labelling experiments, after five volumetric exchanges the outflow was collected into 100% ethanol, with the final concentration of ethanol being kept above 50%. Outflow was harvested daily by centrifugation at 10 000 g for 10 min, and the pellet was stored at −20°C. For microarray experiments, cells were grown using the same conditions. Culture, 50 ml, was combined on ice with 5 ml of 95% ethanol/5% phenol and immediately centrifuged at 10 000 g for 5 min at −20°C. The supernatant was decanted, and the cell pellet was stored at −80°C until processing. For growth in batch cultures, R. pomeroyi wild-type and mutant strains were grown in batch culture using a marine basal medium as described previously (Reisch et al., 2011b). Cell material used for protein purifications was grown in a 1 l chemostat with a flow rate of 0.7 ml min−1 and a dilution rate of 0.042 h−1 with 2 mM DMSP and 3 mM sodium acetate as the sources of carbon. After establishing steady state, approximately 900 ml of cell material was collected each day for 3 days. Collected cell material was harvested by centrifugation at 10 000 g for 10 min, washed with ice-cold 50 mM Tris-HCl (pH 7.5), and then frozen at −20°C. Cell material from three collections was resuspended in 2 ml buffer and lysed by bead beating with 0.1 mm zirconia beads for 5 min using a vortex genie bead beating adapter (MoBio Laboratories). Cell lysate was centrifuged for 10 min at 10 000 g and then used for protein purifications. Calculations for the minimum enzymatic specific activity required to consume carbon entering the chemostat were performed as described previously (Reisch et al., 2011b), yielding a value of 57 nmol min−1 mg of protein−1 for the conditions used in these experiments. Since 40% of DMSP is routed through the cleavage pathway in the DMSP-limited chemostat (Reisch et al., 2011b), the minimum specific activity of enzymes in the cleavage pathway was 23 nmol min−1 mg of protein−1.

Methanethiol and dimethylsulphide measurements

Methanethiol and dimethylsulphide were measured in the culture headspace by gas chromatography on an SRI 8610-C gas chromatograph with a Chromosil 330 column with nitrogen carrier gas at a flow rate of 60 ml min−1, an oven temperature of 60°C and a flame photometric detector (de Souza and Yoch, 1995).

13C enrichment of carbon dioxide and whole cells

Carbon dioxide leaving the chemostat was trapped by bubbling the gaseous outflow through a solution of Ba(OH)2. The barium hydroxide solution was prepared by mixing equal volumes of 100 mM BaCl2 and 200 mM of NaOH. The precipitate that formed upon mixing was removed by centrifugation, and the clarified solution was gently pipetted into a 28 ml Balch tube and sealed with a stopper so that no air remained in the tube. The chemostat outflow was then bubbled through the barium hydroxide solution using narrow bore plastic tubing. After about 20 min, the white precipitate was collected by centrifugation, and the supernatant was decanted. The barium carbonate was then washed with degassed water and dried by lyophilization.

Whole cells were collected by centrifugation at 10 000 g for 10 min, washed once with dH2O, and dried by lyophilization. Barium carbonate and whole-cell samples were analysed by combustion mass spectrometry at the Colorado Plateau stable isotope laboratory at Northern Arizona University.

Fractionation of labelled cells

Cell pellets were fractionated using a protocol similar to that described previously (Roberts, 1955). First, cells were washed in 50:50 solution of ethanol:diethylether for 40 min at 37°C. The mixture was then centrifuged at 15 000 g for 10 min, and the supernatant containing lipids was discarded. The pellet was dried under a stream of air, and 2 ml of 4 M NaOH was added. The pellet was vortexed vigorously and incubated at 37°C for 4 h. The solution was brought to pH 2 with concentrated HCl and then centrifuged at 15 000 g for 10 min. The supernatant containing hydrolysed RNA was brought to pH 8 with sodium bicarbonate and saved. The pellet was then suspended in 5% TCA and placed in a boiling water bath for 30 min. The mixture was cooled to room temperature and centrifuged at 15 000 g for 10 min. The supernatant containing the hydrolysed nucleic acids was discarded, and the pellet was resuspended in dH2O and centrifuged as above. The pellet was again suspended in 1 ml H2O, transferred to a Balch tube and dried under a stream of air. Then 1 ml of 6 M HCl was added to the tube, and the tube was sealed with a butyl stopper. The tube was flushed with N2 for 15 min and incubated in a 110°C sand bath for 24 h. The HCl was removed by drying under a stream of air. The residue was resuspended in 1 ml of dH2O and again dried. The residue was then suspended in 1 ml D2O and centrifuged for 2 min at 17000 g. The supernatant was used for 13C NMR analysis of the crude amino acid pool.

Amino acid and nucleoside purification and analysis

Amino acids from the hydrolysed proteins were benzoylated with benzoyl chloride as described by Carter and Stevens (1941). Briefly, the hydrolysed proteins were resuspended in 1 ml of 2 M NaOH, and 50 μl of benzoyl chloride was added in 10 μl increments with vigorous vortexing. After about 20 min, the solution was acidified with the addition of 300 μl of 6 M HCl. The precipitate was removed by centrifugation, and the solution was used for amino acid purification. The benzoylated amino acids were purified using an AKTA purifier (GE Healthcare) with a C-18 reverse phase column. Amino acids were eluted with a gradient of 8–50% methanol in water with 0.2% formic acid and detected by absorbance at 254 nm. Fractions containing purified amino acids were dried under a stream of air. Chemostat-grown R. pomeroyi grown on 2 mM DMSP was collected and washed in 0.1% HCl and then submitted for amino analysis (Proteomics Core Facility, UC Davis). The amino acid composition is listed in Table S1.

Nucleotides were dephosphorylated with alkaline phosphatase and purified using an AKTA purifier (GE Healthcare) with a C-18 reverse phase column. Nucleosides were eluted isocratically using a mobile phase of 8% methanol and 0.2% formic acid. Purified nucleosides were lyophilized and resuspended in D2O for NMR analysis.

Determination of enrichments by NMR

1H NMR of nucleosides and amino acids was performed on a Varian Unity Inova500 with a broadband probe. For quantitative 13C NMR, a 45-degree pulse angle and 5 s relaxation delay were used with 500–5000 scans. Quantification was performed as described previously (Eisenreich et al., 1993) by comparison of the 13C-NMR integrals from experimentally obtained and standard amino acids. Amino acid standards were benzoylated and purified as described above. These standard amino acids were analysed under the same conditions as the experimentally obtained amino acids. The 13C-NMR integrals of the signals from the benzoyl carbons from both the experimental and standard amino acids were set equal to 1, since these signals should contain only natural abundance 13C. The ratio of the 13C-NMR integrals from experimental and standard amino acids was obtained to give the relative abundance of 13C at each carbon position. The absolute abundance was then determined by multiplying the relative abundance by 1.1.

Enzyme assays

Acrylate-CoA ligase was assayed in 50 mM HEPES (pH 7.5), 2 mM ATP, 2 mM MgCl2, 0.05 mM CoA and 2 mM acrylate. Reactions were initiated by the addition of cell extract. After 2–5 min, they were quenched by the addition of 4 μl H3PO4. After centrifugation to remove denatured proteins, the remaining CoA was analysed by HPLC. Acryloyl-CoA hydratase activity was measured in 50 mM HEPES (pH 7.5) and 0.05 mM acryloyl-CoA. Reactions were initiated by the addition of protein and processed as described above. Activity was measured as the production of 3-hydroxypropionyl-CoA. Acryloyl-CoA reductase activity was measured in 50 mM HEPES (pH 7.5), 0.05 mM acryloyl-CoA or 3-hydroxypropionyl-CoA, 1 mM NADPH and 1 mM MgCl2. Reactions were initiated with the addition of protein. After 2–5 min, they were quenched and analysed as described above. Activity was measured as the production of propionyl-CoA. Propionyl-CoA carboxylase activity was measured in 50 mM HEPES (pH 7.5), 0.05 mM propionyl-CoA, 2 mM ATP, 2 mM MgCl2 and 10 mM NaHCO3. Reactions were initiated with the addition of protein. After 2–5 min, they were quenched and analysed. Activity was measured by the disappearance of propionyl-CoA. Acetyl-CoA carboxylase activity was measured similarly except that acetyl-CoA was substituted for propionyl-CoA. Acetyl- and propionyl-CoA transferase activity was assayed in 50 mM HEPES (pH 7.5), 0.05 mM acetyl- or propionyl-CoA and 2 mM sodium acrylate. Reactions were initiated by the addition of protein and quenched after 30 min. Activity was measured by the disappearance of either acetyl- or propionyl-CoA.

Genetic modifications

Gene disruptions of SPO0370 and SPO1914 were made by homologous recombination of suicide plasmids as described previously (Reisch et al., 2011b). For the pdh mutant, random transposon mutagenesis was performed using an EZ-Tn5<KAN-2>Tnp Transposome kit (Epicentre), and the mutants were screened for their ability to reduce Ellman's reagent during growth with DMSP. One strain which grew poorly on DMSP, was identified, and the transposon insertion was mapped to position 83 of SPO2240 (pdhA) by Sanger sequencing at the Georgia Genomics Facility. For this reason, the strain was named pdh1.

Recombinant protein expression

Genes SPO1914, SPO0147 and SPO2934 were PCR-amplified from R. pomeroyi genomic DNA and cloned into the pTrcHisA (Invitrogen) vector by standard techniques.

Protein purifications

For purification of the acryloyl-CoA hydratase (SPO0147) from R. pomeroyi, cell extract was applied to a Mono-Q HR anion exchange column (GE Healthcare, 1.6 × 10 cm) equilibrated with 50 mM Tris-HCl (pH 8.0) at a flow rate of 2 ml min−1. Protein was eluted with a gradient from 0–1 M NaCl over 8 column volumes. Activity eluted between 18 and 28 mS cm−1. Active fractions from the Mono-Q chromatography were pooled and made 1.7 M (NH4)2SO4 by addition of solid (NH4)2SO4. After centrifugation, the supernatant was applied to a phenyl-Superose HR hydrophobic interaction column (GE Healthcare, 1 × 10 cm) at a flow rate of 1 ml min−1. The column was washed with one column volume of 1.7 M (NH4)2SO4 in 50 mM Tris-HCl (pH 7.5). Protein was eluted with a gradient of 1.7–0 M (NH4)2SO4 in 50 mM Tris-HCl (pH 7.5) over seven column volumes. Activity eluted at 62–48 mS cm−1. Active fractions were pooled and concentrated with an Amicon Ultra centrifugal filter (10 kD) to a final volume of about 0.2 ml, then diluted to 5 ml with 5 mM potassium phosphate buffer (pH 7.5), and again concentrated to 0.2 ml. The final concentrate was then diluted to 1 ml in 5 mM potassium phosphate buffer (pH 7.5). The concentrated protein solution was then applied to a type-II hydroxyapatite column (1 ml, Bio-Rad) that was equilibrated with 5 mM potassium phosphate (pH 7.5) containing 1 mM CaCl2. The column was washed with four column volumes of buffer, and protein was eluted with a 5–500 mM gradient of potassium phosphate buffer with 1 mM CaCl2 over six column volumes. Activity eluted just after start of the gradient. The two 1 ml fractions containing the highest activity were concentrated using an Amicon Ultra centrifugal filter (10 kD). The acrylyl-CoA reductase (SPO1914) was purified as described above for the acryloyl-CoA hydratase. Activity co-eluted with acryloyl-CoA hydratase during anion-exchange chromatography. Activity eluted at 75–61 mS cm−1 after hydrophobic interaction chromatography. Active fractions were pooled, concentrated and chromatographed with the hydroxyapatite column. Activity eluted after the start of the gradient. The two 1 ml fractions with the highest activity were concentrated as described above.

Transcriptional analysis using microarrays

Ruegeria pomeroyi microarray design, processing and image analysis were performed as described previously (Burgmann et al., 2007). A non-competitive hybridization scheme was used in which only one RNA sample was hybridized to the microarray. Slides were then normalized so that the average spot in each microarray possessed the same signal intensity. The normalized intensity values were then used to compare the relative signal intensity between different treatments using the significance analysis of microarrays program (Tusher et al., 2001). Those spots with false-discovery rate (q-value) less than 10 were considered significantly regulated.

Metabolic model calculations

The calculations to determine the theoretical 13C enrichment of pathway intermediates were complicated by the requirement to account for two sources of succinyl-CoA and malate, either the DMSP cleavage or demethylation pathways, as well as the fact that many of the intermediates were shared between the pathways. Therefore, for each potential pathway evaluated, the fluxes of individual intermediates were solved algebraically to yield the estimated levels of CO2 from respiration and intermediates needed for growth. During growth in the chemostat, the flux through the cleavage pathway was equal to the amount of DMS produced or 80 nmol min−1 and yielded acryloyl-CoA. The remaining 120 nmol min−1 was routed through the demethylation pathway and produced acetaldehyde, which would be metabolized through acetate to acetyl-CoA. The fluxes of intermediates necessary to support growth of 300 nmol min−1 of cellular C were 21 nmol min−1 of acetyl-CoA, 20 nmol min−1 of pyruvate, 50 nmol min−1 of oxaloacetate and 7 nmol min−1 of α-ketogluturate. These values were based on the building blocks required for E. coli (Moran et al., 2012) but modified to account for the amino acid content determined experimentally for R. pomeroyi and the alternative pathway for isoleucine biosynthesis (see below, Tables S1 and S2). Although some biosynthetic pathways in R. pomeroyi are known to differ from those in E. coli, this approach was justified by the overall similarity of central carbon metabolism revealed in the genome sequences (Moran et al., 2004). The assimilation of methyl carbons of DMSP of ∼ 4 nmol min−1 was neglected in these calculations. In the chemostat, the estimated net CO2 production from DMSP was 540 nmol min−1. The demethylation pathway produced 360 nmol min−1 of this from the oxidation of 5-methyl-THF produced during the initial demethylation by DmdA, the oxidation of MeSH, and directly from the DmdD reaction. In addition, 39 nmol min−1 would be produced from the anabolic metabolism of oxaloacetate to produce phosphoenolpyruvate, an intermediate in sugar and aromatic amino acid biosynthesis, and 7 nmol min−1 would be produced from the anabolic metabolism of oxaloacetate to produce α-ketogluturate. Therefore, to solve for the fluxes in the pathway shown in Fig. 5, the flux from 2 acetyl-CoA to glyoxylate was set as x. The solution for x was calculated from the following relationships. The net CO2 produced of 540 nmol min−1 = (sum of CO2 producing fluxes) – (sum of CO2 uptake fluxes). From the pathway, the CO2 uptake fluxes = 2x + 80 nmol min−1. Similarly, the CO2 producing fluxes = [360 + 39 + 7 + (2x + 80 − 57) + (2x + 80 − 57 − 20)] nmol min−1. Thus, x = 94 nmol min−1, which was rounded up to 95 nmol min−1 in Fig. 5. During growth with 99% enriched [1-13C] DMSP, the theoretical 13C enrichment patterns were calculated assuming that the internal carbonate pool (CO2 + HCO3 + H2CO3) was not in equilibrium with the external carbonate pool. This enrichment could be calculated in two ways. Based upon the measured enrichment of CO2 of 28%, the enrichment of the internal pool would be 171 nmol min−1 13C/540 nmol min−1 internal carbonate or 32%. However, based upon the expected carbonate production from the C-1 of DMSP (195 nmol min−1) and the total carbonate production from DMSP (592 nmol min−1), the internal pool was estimated to be 34% after accounting for the natural abundance. This latter value was used in subsequent calculations.


The research was supported by grants from the National Science Foundation MCB-1158037 and MCB-0702125.