Large-Scale ¹³C Flux Profiling Reveals Conservation of the Entner-Doudoroff Pathway as a Glycolytic Strategy among Marine Bacteria That Use Glucose

Abstract

Marine bacteria form one of the largest living surfaces on Earth, and their metabolic activity is of fundamental importance for global nutrient cycling. Here, we explored the largely unknown intracellular pathways in 25 microbes representing different classes of marine bacteria that use glucose: Alphaproteobacteria, Gammaproteobacteria, and Flavobacteriia of the Bacteriodetes phylum. We used ¹³C isotope experiments to infer metabolic fluxes through their carbon core pathways. Notably, 90% of all strains studied use the Entner-Doudoroff (ED) pathway for glucose catabolism, whereas only 10% rely on the Embden-Meyerhof-Parnas (EMP) pathway. This result differed dramatically from the terrestrial model strains studied, which preferentially used the EMP pathway yielding high levels of ATP. Strains using the ED pathway exhibited a more robust resistance against the oxidative stress typically found in this environment. An important feature contributing to the preferential use of the ED pathway in the oceans could therefore be enhanced supply of NADPH through this pathway. The marine bacteria studied did not specifically rely on a distinct anaplerotic route, but the carboxylation of phosphoenolpyruvate (PEP) or pyruvate for fueling of the tricarboxylic acid (TCA) cycle was evenly distributed. The marine isolates studied belong to clades that dominate the uptake of glucose, a major carbon source for bacteria in seawater. Therefore, the ED pathway may play a significant role in the cycling of mono- and polysaccharides by bacterial communities in marine ecosystems.

INTRODUCTION

Marine bacteria influence global environmental dynamics in fundamental ways by controlling the biogeochemistry and productivity of the oceans (1). Due to their importance, marine microorganisms have been studied intensively (2). In particular, their mechanisms for metabolizing carbon and other nutrients have attracted attention, because they directly or indirectly affect the biogeochemical status of seawater (3). A prominent nutrient in seawater is glucose, the most abundant free neutral aldose (4). Current estimates of glucose concentrations in seawater indicate an almost ubiquitous distribution in the oceans in a nanomolar range (5). Particularly, large amounts of glucose are available in coastal habitats, e.g., during bloom situations (6). In fact, a large fraction (>30%) of bacterial growth can be supported by this monosaccharide in some oceans (7, 8). Furthermore, glucose is the dominant component of dissolved polysaccharides, which constitute up to 15% of marine dissolved organic matter (9). The turnover of the (monomeric and polymeric) glucose pool in different oceanic regions ranges from days to months, and glucose assimilation in marine surface waters may represent up to 40% of bacterial carbon production (5). Taken together, bacteria that use glucose are common in the sea (10), and glucose is a representative model nutrient to monitor carbon uptake by heterotrophic marine bacteria (11). At this point, questions that arise from current knowledge concern the intracellular pathways involved in glucose utilization. Generally, three common alternative routes occur in bacteria for catabolic breakdown of glucose, the Embden-Meyerhof-Parnas (EMP) pathway (EMPP), Entner-Doudoroff (ED) pathway (EDP), and pentose phosphate (PP) pathway (PPP) (12). The EMP pathway is nearly ubiquitous in the bacterial kingdom (13). It has been argued that this relates to the superior energy efficiency of the EMP pathway (equation 1): it yields twice as much ATP as the ED pathway (equation 2) (14).

$$\text{Glucose}+2\hspace{0.17em}\text{ADP}+2{\hspace{0.17em}\text{NAD}}^{+}\to 2\hspace{0.17em}\text{pyruvate}+2\hspace{0.17em}\text{ATP}+2\hspace{0.17em}\text{NADH}$$ (1)

$$\text{Glucose}+\text{ADP}+{\text{NAD}}^{+}+{\text{NADP}}^{+}\to 2\hspace{0.17em}\text{pyruvate}+\text{ATP}+\text{NADH}+\text{NADPH}$$ (2)

However, phylogenetically distinct bacteria (Firmicutes, Alphaproteobacteria, and Gammaproteobacteria), including aerobes and anaerobes, autotrophs and heterotrophs, rely on the ED pathway for glucose catabolism (15, 16). Quantitative flux through the EMP, ED, and PP pathways is accessible via ¹³C fluxomics; in these experiments, quantitative information on fluxes, i.e., the in vivo activities of intracellular enzymes and pathways, is obtained from the labeling patterns of metabolites generated during growth on specific ¹³C tracer substrates (17, 18). Such studies have provided valuable insight into the intracellular pathways of microorganisms. For example, industrially relevant microbes have been extensively studied (19), and more recently, fluxomic studies have examined pathogenic (16, 20) and ecologically important (21) microorganisms. In contrast, there is still limited knowledge on the intracellular pathways of marine bacteria, and only a few isolates have been investigated so far (22–24). A more comprehensive analysis of intracellular pathways in marine microorganisms promises a better understanding of the ways they survive in the highly heterogeneous oceans (25) and their susceptibility to environmental variability and climate change (2). Here, we studied the central carbon pathways of marine microbes that use glucose on the level of metabolic fluxes. We selected 25 strains of bacteria belonging to various classes of the Bacteriodetes phylum, Alphaproteobacteria, Gammaproteobacteria, and Flavobacteriia, as representative and ubiquitous classes of marine bacterial communities (26). Strains of Actinobacteria that are indicators for freshwater input were also included (27). We quantified metabolic fluxes in cells grown in a seawater medium with glucose as a model substrate. In addition to the major catabolic pathways, i.e., the EMP, ED, and PP pathways, we analyzed fluxes through anaplerotic pathways that are important for biosynthesis and fueling. We integrated the flux data obtained with data on the enzymatic inventories and on oxidative stress tolerance to generate a carefully curated conceptual representation of the metabolic strategies that permit microbes to grow and thrive in the oceans.

MATERIALS AND METHODS

Bacterial strains.

We studied a set of marine model strains and several terrestrial model strains (see Fig. 1 and Table 1; also see Table S1 in the supplemental material). The sequenced marine-type strains Dinoroseobacter shibae (28), Alteromonas macleodii (29), Eudoraea adriatica (30), Polaribacter dokdonensis (31), Pseudoalteromonas haloplanktis (32), Pseudoalteromonas marina (33), and Phaeobacter inhibens DSM 17395 (34), which is not the type strain, were obtained from culture collections (Table S1). The marine Pseudoalteromonas sp. strains HEL-36, HEL-40, and HEL-49, Alteromonas sp. strains HEL-5, HEL-51, BIO-267, and BIO-296, and Leeuwenhoekiella sp. strain Pic90 were previously isolated from water samples and biofilms near the island of Helgoland, Germany, in the North Sea and phylogenetically classified using 16S rRNA gene sequencing (35). Phyllobacteriaceae bacterium TK, Oceanospirillaceae bacterium T17, Pseudonocardiaceae bacterium T4, Flavobacteriaceae bacterium TN, Flavobacteriaceae bacterium T15 (36), as well as Pseudomonas sp. strain GWS-TZ-H209, Vibrio sp. strain GWS-TZ-H304, and Gammaproteobacteriaceae bacterium GWS-SE-H233 (37) originated from an intertidal mudflat region of the southern North Sea (Table S1). Jannaschia sp. strain B3 and Loktanella sp. strain D3 were isolated from the surface of the marine macroalga Fucus spiralis, collected in June 2010 from the intertidal flat area in Neuharlingersiel on the North Sea coast of Germany. The samples were transported on ice to the laboratory in seawater collected on-site. The alga was washed three times with sterile-filtered and autoclaved seawater to remove unattached bacteria and particles, spread on a plate with marine agar (marine agar 2216; Becton Dickinson, Franklin Lakes, NJ, USA), and then incubated for 2 weeks at 25°C in the dark. Single colonies were selected and transferred at least three times to isolate pure strains. The purity of the strains in culture and in the experiments was tested using denaturing gradient gel electrophoresis (38). The 16S rRNA genes of the two strains were amplified and sequenced (39). For phylogenetic analysis, sequence reads of at least 650 bp were compared with GenBank entries using the BLAST analysis tool of the National Center for Biotechnology Information (NCBI) (www.ncbi.nlm.nih.gov). For comparison, nonmarine Escherichia coli K-12 (ATCC 10798), Bacillus subtilis 168 (DSM 402), Bacillus megaterium (DSM 319), Corynebacterium glutamicum (ATCC 13032), and Pseudomonas putida KT 2440 (DSM 6125) were obtained from the corresponding culture collections. All strains were maintained as glycerol stocks at −70°C.

FIG 1.

Phylogenetic tree showing the relationships of the marine bacteria studied based on 16S rRNA gene sequence similarity. The sequences of type strains (>1,300 bp) were used to construct the backbone tree using maximum likelihood. Subsequently, shorter sequences from the isolates studied were added by interactive parsimony. GenBank sequence accession numbers are given in parentheses. The strains investigated in this study are shown in bold type (see also Table S4 in the supplemental material).

TABLE 1.

Specific activities of phosphofructokinase, glucose 6-phosphate dehydrogenase, and 6-phosphogluconate dehydrogenase in marine bacteria

Strain or species	Phylogenetic class	Family	Clustera	Sp act (mU mg−1)b
				PFK	G6PDH		GND
					NADP	NAD
Pseudoalteromonas sp. HEL-36	Gammaproteobacteria	Pseudoalteromonadaceae	I	<0.1	906 ± 49	286 ± 47	<0.1
Pseudoalteromonas sp. HEL-40	Gammaproteobacteria	Pseudoalteromonadaceae	I	<0.1	516 ± 11	81 ± 9	<0.1
Pseudoalteromonas sp. HEL-49	Gammaproteobacteria	Pseudoalteromonadaceae	I	<0.1	488 ± 20	97 ± 8	<0.1
Alteromonas sp. HEL-5	Gammaproteobacteria	Alteromonadaceae	I	<0.1	616 ± 81	181 ± 6	<0.1
Alteromonas sp. HEL-51	Gammaproteobacteria	Alteromonadaceae	I	<0.1	322 ± 24	52 ± 15	<0.1
Alteromonas sp. BIO-267	Gammaproteobacteria	Alteromonadaceae	I	<0.1	681 ± 46	256 ± 11	<0.1
Alteromonas sp. BIO-296	Gammaproteobacteria	Alteromonadaceae	I	<0.1	665 ± 23	256 ± 7	<0.1
Gammaproteobacterium H233	Gammaproteobacteria	Halomonadaceaee	I	<0.1	307 ± 24	61 ± 8	<0.1
Phaeobacter inhibens	Alphaproteobacteria	Rhodobacteraceae	I	<0.1	50 ± 4	14 ± 8	<0.1
Dinoroseobacter shibae	Alphaproteobacteria	Rhodobacteraceae	I	<0.1	101 ± 4	33 ± 4	<0.1
Leeuwenhoekiella sp. Pic90	Flavobacteriia	Flavobacteriaceae	I	<0.1	319 ± 13	<0.1	<0.1
Pseudoalteromonas haloplanktis	Gammaproteobacteria	Pseudoalteromonadaceae	I	<0.1	374 ± 43	60 ± 4	<0.1
Pseudoalteromonas marina	Gammaproteobacteria	Pseudoalteromonadaceae	I	<0.1	498 ± 54	40 ± 12	<0.1
Alteromonas macleodii	Gammaproteobacteria	Alteromonadacea	I	<0.1	183 ± 35	62 ± 9	<0.1
Oceanospirillaceae bacterium T17	Gammaproteobacteria	Oceanospirillaceae	II	<0.1	33 ± 6	<0.1	2 ± 0
Pseudomonas sp. GWS-TZ-H209	Gammaproteobacteria	Pseudomonadaceae	II	<0.1	342 ± 45	<0.1	19 ± 1
Jannaschia sp. B3	Alphaproteobacteria	Rhodobacteraceae	II	<0.1	240 ± 22	45 ± 7	95 ± 20
Loktanella sp. D3	Alphaproteobacteria	Rhodobacteraceae	II	<0.1	1,287 ± 7	705 ± 7	16 ± 1
Phyllobacteriaceae bacterium TK	Alphaproteobacteria	Phyllobacteriaceae	II	<0.1	366 ± 14	247 ± 37	29 ± 1
Pseudonocardiaceae bacterium T4	Actinobacteria	Pseudonocardiaceae	III	117 ± 9	<0.1	<0.1	20 ± 1
Vibrio sp. GWS-TZ-H304	Gammaproteobacteria	Vibrionaceae	III	155 ± 31	299 ± 79	<0.1	61 ± 5
Eudoraea adriatica	Flavobacteriia	Flavobacteriaceae	III	128 ± 4	63 ± 20	<0.1	182 ± 16

^aThe analyzed strains or species were assigned to cluster I (use of the ED pathway only), cluster II (parallel use of the ED and PP pathways), or cluster III (parallel use of the EMP and PP pathways) according to their catabolic flux profiles (Fig. 3).

^bSpecific activity of phosphofructokinase (PFK) (in the EMP pathway), glucose 6-phosphate dehydrogenase (G6PDH) (in a joint reaction of ED and PP pathway), and 6-phosphogluconate dehydrogenase (GND) (in the PP pathway) in marine bacteria. The data are means ± standard deviations from three biological replicates. Cytosolic extracts from Polaribacter doktonensis and Flavobacteriaceae bacterium TN and T15 did not contain sufficient protein for reliable measurement; therefore, these strains were excluded from the enzymatic analysis. The enzyme assays were conducted under saturated in vitro conditions.

Media.

For growth on agar plates, 37.4 g liter⁻¹ of marine broth (marine broth 2216; Becton Dickinson) was mixed with 15 g liter⁻¹ of agar (Becton Dickinson). In addition, we used marine broth without agar as a liquid medium for the first precultures. A defined medium was used for the second precultures and the main cultures (40). This medium contained the following (per liter): 1.8 g of glucose, 4.0 g of Na₂SO₄, 0.2 g of KH₂PO₄, 0.25 g of NH₄Cl, 20.0 g of NaCl, 9.0 g of MgCl₂·6H₂O, 0.5 g of KCl, 0.15 g of CaCl₂·2H₂O, 0.19 g of NaHCO₃, 4.2 mg of FeSO₄·7H₂O, 10.4 mg of Titriplex-(III) (Na₂-EDTA), 60 μg of H₃BO₃, 200 μg of MnCl₂·4H₂O, 380 μg of CoCl₂·6H₂O, 48 μg of NiCl₂·6H₂O, 4 μg of CuCl₂·2H₂O, 288 μg of ZnSO₄·7H₂O, and 72 μg of Na₂MoO₄·2H₂O. Trace elements were added to the autoclaved basal medium from a sterile-filtered 500× stock solution. The pH of the final medium was adjusted to 8.0 using 2 M NaOH. The medium was supplemented with 5 mg liter⁻¹ 4-aminobenzoic acid, 2 mg liter⁻¹ folic acid, 2 mg liter⁻¹ biotin, 5 mg liter⁻¹ nicotinic acid, 5 mg liter⁻¹ Ca-pantothenic acid, 5 mg liter⁻¹ vitamin B₂, 10 mg liter⁻¹ vitamin B₆, 0.1 mg liter⁻¹ vitamin B₁₂, 5 mg liter⁻¹ thiamine-HCl, and 5 mg liter⁻¹ lipoic acid. All vitamins were added from sterile-filtered stock solutions. For the isotope studies, we replaced glucose with an equimolar amount of 99% [1-¹³C]glucose (Eurisotop, Saarbrücken, Germany). Oxidative stress studies were conducted using soft-agar assays (59). Minimal medium containing 15 g liter⁻¹ agar (Becton Dickinson) was covered with a 1:1 mixture of soft agar (7.5 g liter⁻¹ agar in a 2.4% [wt/vol] NaCl solution) and minimal medium with resuspended cells. Filter discs (5-mm diameter) soaked with 10 μl of diamide [1,1-azo-bis(N,N-dimethylformamide)] (0.6 M in dimethyl sulfoxide [DMSO]) were then placed onto the homogeneous layer of cells.

Cultures.

Cultures with [1-¹³C]glucose were prepared in a volume of 2 ml using deep-well plates (riplate BV [10 ml]; HJ-Bioanalytik, Mönchengladbach, Germany) and incubated at 1,000 rpm on a plate shaker (Inkubator 1000; Heidolph Instruments, Schwabach, Germany). To avoid evaporation, the plates were sealed with gas-permeable adhesive membranes (HJ-Bioanalytik). Single colonies from agar plates that were incubated for 72 h were used to inoculate the first preculture. After incubation for 12 h, the cells were harvested by centrifugation (15,700 × g, 5 min, 4°C), washed with sterile 0.9% (wt/vol) NaCl, and used as the inoculum for the second preculture. The main cultures were inoculated using exponentially growing cells from the second preculture that were washed as described above. For each strain, 12 parallel cultures were prepared. Samples from four wells were pooled, yielding three biological replicates of each strain for subsequent analysis. The growth performance was validated with respect to the metabolic steady state (see Fig. S1 and Fig. S2 in the supplemental material). In selected cases, the main cultures were additionally grown in three biological replicates (1-liter baffled shake flasks, 100 ml of minimal medium) and incubated on a rotary shaker (230 rpm). These cultures were used to provide sufficient amounts of cells for gravimetric analysis of the cell weight (dry weight) (cdw) or enzymatic measurements. Oxidative stress tests were performed in triplicate. The agar plates with diamide were incubated in the dark. All cultures were grown at 30°C.

Quantification of cell concentrations.

The cell concentrations were monitored based on the optical density at 600 nm (OD₆₀₀). For P. inhibens, the cdw was additionally measured after the collection of cells by centrifugation (9,000 × g, 10 min, 4°C); the cells were washed with 0.9% (wt/vol) NaCl, and the pellet was dried at 80°C until it reached a constant weight. Both measurements were integrated to determine the correlation between the optical density and cell weight (dry weight, in grams per liter) (equation 3).

$$\text{cdw}={\text{OD}}_{600}\times 0.2657$$ (3)

This correlation was used to estimate the culture volume required to obtain 10 to 30 mg of dry biomass based on the measured optical density. This amount was appropriate for the subsequent ¹³C gas chromatography (GC)-mass spectrometry (MS) analysis. The correlation determined for P. inhibens was applied to all strains.

Enzyme assays.

The cells were harvested during the exponential growth phase (9,000 × g, 10 min, 4°C), washed with disruption buffer (200 mM Tris-HCl [pH 8.0], 200 mM KCl, 50 mM K₃PO₄-NaOH, 0.75 mM dithiothreitol [DTT], 1 mM ATP, and 5 mM MgCl₂ in 5% [vol/vol] glycerol) and disrupted mechanically (twice for 4 min each time, 30 Hz) (Ribolyzer MM301; Retsch, Haan, Germany). The cell debris was removed by centrifugation (16,000 × g, 5 min, 4°C). The protein concentration in the obtained extract was determined using a bicinchoninic acid (BCA) assay kit (Thermo Fisher Scientific, Bonn, Germany). The phosphofructokinase activity (EMP pathway) was quantified in a reaction mix containing 100 mM Tris-HCl (pH 8.0), 50 mM K₃PO₄-NaOH, 32.5 mM KCl, 5 mM MgCl₂, 0.25 mM NADH, 0.1 mM ATP, 0.5 U of aldolase, 0.5 U of glyceraldehyde 3-phosphate dehydrogenase, 1 U of triose phosphate isomerase, 4 mM fructose 6-phosphate, and 0.02 mg ml⁻¹ protein. The reaction mix for quantification of glucose 6-phosphate dehydrogenase activity (joint reaction of PP pathway and ED pathway) contained 100 mM Tris-HCl (pH 7.8), 200 mM KCl, 1 mM NAD(P), 10 mM MgCl₂, 5 mM glucose 6-phosphate (G6P), and 50 μl cell extract in a total volume of 1 ml (42). The enzyme activity was determined by monitoring the formation of NAD(P)H at 340 nm. The reaction mix for the determination of 6-phosphogluconate dehydrogenase (PP pathway) activity contained 100 mM Tris-HCl buffer (pH 7.8), 10 mM MgCl₂, 0.75 mM DTT, 1 mM EDTA, 2.5 mM 6-phosphogluconate, 1 mM NAD(P)H, and 0.04 mg ml⁻¹ protein. The level of NAD(P)H was monitored by measuring the absorbance at 340 nm (Sunrise; Tecan Group, Switzerland), and the change in absorbance was used to calculate the enzymatic activity. Negative controls lacked substrate and cell extract.

Mass spectrometric labeling analysis by GC-MS.

Cells were collected by centrifugation (9,000 × g, 10 min, 4°C) and washed with deionized water. Prior to GC-MS analysis, the cellular protein was hydrolyzed for 24 h at 105°C using 50 μl of 6 M HCl per mg of cells (dry weight). Cell debris was removed by filtration (Ultrafree-MC; Millipore, Billerica, MA, USA). The labeling patterns of proteinogenic amino acids were analyzed using their t-butyldimethylsilyl (TBDMS) derivatives (43) by GC-MS (Agilent 7890A and quadrupole mass selective detector 5975C; Agilent Technologies, Waldbronn, Germany). The samples were first measured in the scanning mode to check for isobaric overlay in the ion clusters of interest, which might interfere with the labeling of the amino acids. Subsequently, selective ion monitoring was used for quantitative analysis. Three individual runs, each with a different subset of measured ion clusters and representing technical duplicates, were conducted per sample (see Fig. S3 in the supplemental material). The amino acid mass distributions were obtained from the spectra after correction for the natural abundance of stable isotopes (44). For each strain, the data set contained the relative fractions of 84 different mass isotopomers (see Table S2 in the supplemental material).

Multivariate data analysis.

The amino acid labeling patterns corrected for natural isotope abundances were subjected to independent component analysis (ICA) (45). In the analysis of isotope experiment data, ICA automatically recognizes conserved and biologically relevant labeling patterns, as demonstrated in a previous analysis of Bacillus subtilis mutants (46). ICA assumes that the labeling patterns result from the superposition of independent metabolic activities; therefore, each activity causes a shift in the mass distribution of one or more metabolites. ICA separates the observed variables into non-Gaussian statistically independent components (ICs), which allows for the identification of the mass signals of metabolites that enable discrimination based on metabolic differences. We used the publicly available FastICA 2.1 algorithm (HUT-CIS; http://research.ics.aalto.fi/ica/fastica/) in Matlab (R2010b; Mathworks, Natick, MA, USA) to derive the ICs. Subsequently, ICASSO bootstrapping was applied to validate the reliability and robustness of the identified ICs (47, 48).

Calculation of the relative fluxes of the glucose catabolism pathways.

The relative contributions of the EMP, ED, and PP pathways to glucose catabolism were assessed from the ¹³C labeling pattern of proteinogenic alanine, generated during growth on [1-¹³C]glucose (24). The analysis considered the relative fraction of the nonlabeled mass isotopomers (M₀) of the entire alanine molecule with carbon atoms C₁, C₂, and C₃ (Ala₁₂₃) and of a fragment that contained the two carbon atoms C₂ and C₃ (Ala₂₃). These were obtained from mass spectrometric analysis of TBDMS-derivatized alanine at a mass-to-charge (m/z) ratio for the monoisotopic mass of 260 (Ala₁₂₃) and 232 (Ala₂₃) (see Table S2 in the supplemental material). After correction for natural isotope abundances, the relative catabolic pathway fluxes into the ED pathway (f_EDP), the EMP pathway (f_EMPP), and the PP pathway (f_PPP) were derived using the following algebraic equations.

$${\mathit{\text{f}}}_{\text{PPP}}=2({\mathit{\text{M}}}_{0,{\text{Ala}}_{123}}-0.5)$$ (4)

$$f_{\text{EMPP}}=-2({\mathit{\text{M}}}_{0,{\text{Ala}}_{23}}-1)$$ (5)

$$f_{\text{EDP}}=1-f_{\text{PPP}}-f_{\text{EMPP}}$$ (6)

If the calculation yielded negative values, results were corrected to zero. The calculations assumed that the reactions were not reversible. The calculation of the respective fluxes via the labeling of serine and alanine (24) yielded the same results (data not shown).

Phylogenetic, functional cluster, and statistical analyses.

Cluster analysis of amino acid mass distributions was conducted using Matlab (R2010b; Statistical Toolbox, Mathworks) to derive a Euclidean distance tree for the different strains based on their metabolic properties. In addition, cluster analysis was performed using the ARB software package (49) (www.arb-home.de) based on 16S rRNA sequences to derive a comparative phylogenetic tree. The sequences of the type strains (>1,300 bp) were used to construct the backbone tree using the maximum likelihood method. Subsequently, shorter sequences from the isolates were added by interactive parsimony. The studied isolates were then integrated into the phylogenetic tree. Differences between experimental data were statistically evaluated for significance by a t test (Origin 9.1; OriginLab, Northampton, MA, USA).

RESULTS

Strain selection, labeling strategy, and isotope experiments for resolution of glycolytic pathway flux.

For our study, we selected 25 marine isolates of Alphaproteobacteria, Gammaproteobacteria, and Flavobacteriia of the Bacteriodetes phylum and Actinobacteria as representative classes of marine bacterial communities, including strains that were recently sequenced (Fig. 1; see Table S1 in the supplemental material). The isolates (Fig. 1) originated from different geographical marine regions such as the North Sea, Mediterranean Sea, East China Sea, and Sea of Japan, and covered different ecological niches, including open water, sediment, and intertidal mudflats, coastal regions, and the surface of macroalgae. All strains were able to utilize glucose as the sole carbon source (see Fig. S1 and Fig. S2 in the supplemental material). For comparison, well-known terrestrial bacteria that use glucose (E. coli, C. glutamicum, P. putida, B. subtilis, and B. megaterium) were analyzed. Briefly, the relative fluxes through the alternative pathways of glucose catabolism in the strains of interest were assessed from mass spectrometric analysis of alanine from the cellular protein of cells, grown on [1-¹³C]glucose as the tracer substrate (Fig. 2). Each of the three catabolic pathways results in a different ¹³C labeling pattern of pyruvate, the precursor of alanine: the ED pathway results in a mixture of nonlabeled pyruvate and [3-¹³C]pyruvate, the EMP pathway results in a mixture of nonlabeled pyruvate and [1-¹³C]pyruvate, and the PP pathway results in nonlabeled pyruvate. Consequently, a unique combination of labeling patterns in alanine fragments, which contain carbons C₁C₂C₃ (Ala₁₂₃) and C₂C₃ (Ala₂₃), respectively, is observed for each pathway (Fig. 2). This forms the basis of the algebraic equations 4 to 6, which link this particular labeling information with relative pathway flux (24). The two fragments Ala₁₂₃ and Ala₂₃ are accessible as ion clusters at m/z 260 and m/z 232 of TBDMS-derivatized alanine, respectively. To elucidate catabolic glucose fluxes, the strains were now cultivated in defined seawater medium with [1-¹³C]glucose as a tracer substrate. The analysis of glucose consumption and cell growth revealed that the biological replicates for each strain grew highly reproducibly (Table S3). Furthermore, the specific growth rate and the biomass yield coefficient for each strain were constant over time and reflected metabolic steady state (Fig. S1 and Fig. S2). Obviously, all batch cultures fulfilled the requirement to recruit the labeling of amino acids from cell protein to derive valid and consistent fluxes (50).

FIG 2.

Labeling strategy for discrimination of relative carbon fluxes through the Entner-Doudoroff (ED) pathway, Embden-Meyerhof-Parnas (EMP) pathway, and pentose phosphate (PP) pathway from an isotope experiment with [1-¹³C]glucose as the tracer substrate. Different carbon transitions lead to a different labeling pattern of pyruvate for each pathway and result in unique labeling patterns of the two alanine fragments [M-57] and [M-85], which contain the carbon atoms C₁C₂C₃ and C₂C₃ of pyruvate, respectively. This information can be used to infer flux information via the algebraic equations given in Materials and Methods. Further details are given elsewhere (24). 6P-Gluconate; 6-phosphogluconate; Glucose 6-P, glucose 6-phosphate.

Marine microorganisms predominantly use the ED pathway as the glycolytic route.

For each strain, comprehensive labeling data sets were collected (see Table S2 in the supplemental material). Qualitative inspection of the labeling data has already revealed that the marine strains differed in glucose catabolism. For example, Pseudonocardiaceae bacterium T4 possessed ¹³C-enriched alanine fragments Ala₂₃ and Ser₁₂₃, which unambiguously indicated an active EMP pathway in this particular strain, whereas the same fragments were not ¹³C enriched but naturally labeled in other marine isolates. The theoretical framework of the underlying carbon transfer through the biochemical reactions involved (equations 4 to 6) and the ¹³C labeling of the two specific alanine fragments (Table S2) provided the relative fluxes through the EMP, ED, and PP pathways on a quantitative basis. The analysis revealed an interesting picture (Fig. 3). Sixteen strains near exclusively used the ED pathway, whereas the EMP and PP pathways were inactive (marine cluster I in Fig. 3). These strains belonged to all studied marine clades and families, i.e., Gammaproteobacteria (Pseudoalteromonas marina, Pseudoalteromonas haloplanktis, Pseudoalteromonas nigrifaciens HEL-36, Pseudoalteromonas sp. HEL-40, Pseudoalteromonas sp. HEL-49, Alteromonas macleodii, Alteromonas distincta HEL-05, Alteromonas macleodii HEL-51, Alteromonas sp. BIO-267, Alteromonas sp. BIO-296, and Gammaproteobacteriaceae bacterium sp. GWS-SE-H233), Alphaproteobacteria (Phaeobacter inhibens and Dinoroseobacter shibae), and Flavobacteriia (Polaribacter dokdonensis and Leeuwenhoekiella sp. Pic90). P. putida was the only terrestrial strain that showed this flux pattern. Another subgroup of seven marine isolates used the ED pathway and the oxidative part of the PP pathway in parallel but did not exhibit any EMP flux (marine cluster II). This flux profile also occurred among all clades: Gammaproteobacteria (Pseudomonas sp. GWS-TZ-H209 and Oceanospirillaceae bacterium T17), Alphaproteobacteria (Jannaschia sp. B3, Loktanella sp. D3, and Phyllobacteriaceae bacterium TK), and Flavobacteriia (Flavobacteriaceae bacterium TN and Flavobacteriaceae bacterium T15). A third cluster exhibited predominant use of the EMP route as the glycolytic pathway for glucose metabolism; the PP pathway operated in parallel with the EMP pathway, albeit to a lower extent, and the ED pathway was completely inactive. Notably, this flux pattern was observed for the majority of the terrestrial strains, including E. coli, C. glutamicum, and the two bacilli, and only three organisms of marine origin (Eudoraea adriatica, Vibrio sp. GWS-TZ-H304, and Pseudonocardiaceae bacterium T4). The individual strains assigned to marine cluster III differed to some extent regarding the relative contribution of the EMP and PP pathways; therefore, members of this cluster grouped not as tightly as members of the other two clusters. Overall, the results indicated a preference of marine bacteria to use the ED pathway as the sole or major catabolic route. In contrast, most of the studied terrestrial microbes used the EMP pathway.

FIG 3.

Relative fluxes through the major pathways of glucose catabolism in marine and nonmarine bacterial species. Relative fluxes through the major pathways of glucose catabolism, the Embden-Meyerhof-Parnas (EMP) pathway, Entner-Doudoroff (ED) pathway, and pentose phosphate (PP) pathway, in 25 marine bacteria (blue symbols) and 8 nonmarine bacteria studied in this work (green symbols) or previous work (yellow symbols) are shown. Marine cluster I bacteria exclusively use the ED pathway. Marine cluster I bacteria include Phaeobacter inhibens, Dinoroseobacter shibae, Alteromonas macleodii, Polaribacter dokdonensis, Pseudoalteromonas haloplanktis, Pseudoalteromonas marina, Pseudoalteromonas sp. HEL-36, Pseudoalteromonas sp. HEL-40, Pseudoalteromonas sp. HEL-49, Leeuwenhoekiella sp. Pic90, Alteromonas sp. HEL-5, Alteromonas sp. HEL-51, Alteromonas sp. BIO-267, Alteromonas sp. BIO-296, and Gammaproteobacteriaceae bacterium GWS-SE-H233. Nonmarine P. putida and P. aeruginosa PAO1 also use the ED pathway (16). Marine cluster II bacteria use the ED and PP pathways in parallel and include Flavobacteriaceae bacterium TN, Phyllobacteriaceae bacterium TK, Flavobacteriaceae bacterium T15, Oceanospirillaceae bacterium T17, Jannaschia sp. B3, Loktanella sp. D3, and Pseudomonas sp. GWS-TZ-H209. Marine cluster III bacteria use the EMP and PP pathways in parallel and include Eudoraea adriatica, Vibrio sp. GWS-TZ-H304, and Pseudonocardiaceae bacterium T4. The parallel use of the EMP and PP pathways is also found for nonmarine Bacillus subtilis (this work) (19), Bacillus megaterium (this work), Escherichia coli (this work) (90), Corynebacterium glutamicum (this work) (52), Yersinia pseudotuberculosis (91), and Sorangium cellulosum (92).

In P. putida, D. shibae, and P. inhibens, the absence of phosphofructokinase prevents the use of the EMP pathway for glucose catabolism (24, 51), whereas E. coli, C. glutamicum, B. subtilis, and B. megaterium possess this enzyme and exhibit a functional EMP pathway (17, 52–54). We now examined whether the observed differences in pathway usage among the marine isolates also relates to their enzyme inventory or to pathway regulation and thus checked for in vitro activity of phosphofructokinase. Almost all isolates that preferred or exclusively used the ED pathway indeed lacked phosphofructokinase activity (Table 1), whereas the enzyme was expressed in the strains with a functional EMP pathway. The entry enzyme of the PP pathway, 6-phosphogluconate dehydrogenase, was expressed in all active PP pathway users, but no enzyme activity was observed in almost all strains that did not use this pathway (Table 1). Taken together, the absence of functional EMP and PP pathways seemed to be due to a lack of phosphofructokinase and 6-phosphogluconate dehydrogenase expression, respectively, rather than due to control of these enzymes on the metabolic level. In the latter case, one would have expected a measurable in vitro activity. This excludes the possibility that these enzymes are expressed but metabolically controlled. This did, however, not necessarily correlate with the absence of the corresponding gene, as deduced from genome sequence information (Table 2). Pathway use and enzymatic set matched with the genetic repertoire for only three of the sequenced isolates, i.e., Dinoroseobacter shibae, P. inhibens, and Eudoraea adriatica. In contrast, P. haloplanktis, P. marina, and A. macleodii did not reveal 6-phosphogluconate dehydrogenase activity, though corresponding genes are annotated. In addition, Alteromonas macleodii did not exhibit 6-phosphofructokinase activity, but it has an annotated gene.

TABLE 2.

Pathway repertoire of sequenced members among the investigated marine bacteria^a

Species	Genome identifier	Locus tag(s)b
		PFK	G6PDH	GND	PntAB
Phaeobacter inhibens	PhaInh 188638	NA	0943, 3373	NA	1476, 1477
Dinoroseobacter shibae	DinShi 9476	NA	1748	NA	1233, 1234
Pseudoalteromonas haloplanktis	PsaHal	NA	3806	0916	NA
Pseudoalteromonas marina	PseMar 220774	NA	2870	2687	NA
Alteromonas macleodii	AltMac 49397	2216	1233	1244	3857–3859
Eudoraea adriatica	EudAdr 278662	0029, 0030	0875	0876	1916–1918
Polaribacter dokdonensisc	PolSp 11901	0658	NA	1940	NA

^aPathway repertoire of sequenced members among the investigated marine bacteria, extracted through the bacterial bioinformatics database and analysis resource PATRIC (55). The data given comprise genome identifier and locus tags of genes annotated as 6-phosphofructokinase (PFK), glucose 6-phosphate dehydrogenase (G6PDH), 6-phosphogluconate dehydrogenase (GND), and NADPH-forming membrane-bound transhydrogenase PntAB, respectively. For PFK and G6PDH, duplicate entries reflect putative isoenzymes. For PntAB, the given tags represent different open reading frames for the individual subunits of the enzyme complex. Here, a potentially functional enzyme was attributed only to strains, which comprised both subunits A and B.

^bNA, not annotated.

^cThe genome sequence was obtained from the closely related isolate Polaribacter sp. MED-152, a marine bacterium that was isolated from the surface water of northwestern Mediterranean Sea off the Catalan coast (56). In the original GenBank submission, it was listed as a strain of Polaribacter dokdonensis (31), with which it has 99.6% similar 16S rRNA sequence.

Most marine strains possess glucose 6-phosphate dehydrogenase with flexible cofactor use.

For the marine strains, glucose 6-phosphate dehydrogenase appeared to be a key enzyme for catabolic glucose breakdown, because alternative routes were not expressed or even not encoded. All ED pathway users, i.e., 90% of the studied marine isolates, channeled their substrate carbon entirely via this biochemical reaction. Likewise, the strains that exhibited a split use of the EMP pathway and the PP pathway obviously needed the enzyme to direct carbon into the PP pathway. In vitro analysis of glucose 6-phosphate dehydrogenase revealed that the enzyme was expressed in all marine strains studied. Loktanella sp. D3 had the highest activity (1.3 U mg⁻¹), and for most other strains, the enzyme activity was in the range of 0.3 to 0.6 U mg⁻¹. A few isolates accepted only NADP as a cofactor for glucose 6-phosphate dehydrogenase. These included the EMP users Eudoraea adriatica, Vibrio sp. GWS-TZ-H304, and selected strains from clusters I and II (Table 1). It was interesting to note that glucose 6-phosphate dehydrogenase showed activity with NADP and NAD in most marine strains, whereby NADP was the preferred cofactor. The NAD-related activity of the enzyme in these strains was variable and ranged from 8% (P. marina) to 67% (Loktanella sp. D3) of the NADP-related activity. In addition, the PP pathway enzyme 6-phosphogluconate dehydrogenase was studied for cofactor prevalence. All strains of clusters II and III, i.e., with an active PP pathway (Fig. 2), were analyzed accordingly. It turned out that 6-phosphogluconate dehydrogenase accepted only NADP as a cofactor but was not active with NAD (data not shown).

Marine Gammaproteobacteria differ from marine Alphaproteobacteria and Flavobacteria bacteria in the anaplerotic fluxes that replenish the TCA cycle.

In bacteria, the phosphoenolpyruvate (PEP)-pyruvate-oxaloacetate node is a central switch point for the distribution of carbon flux among the catabolic, anabolic, and energy supply pathways (57). Bacteria differ in the types of reactions occurring at this node, that is, certain bacteria can replenish the tricarboxylic acid (TCA) cycle via the carboxylation of either PEP, pyruvate, or both. The following labeling strategy allowed elucidation of the type of anaplerotic metabolism in 90% of the marine strains, i.e., in all strains that used the ED pathway. As recently demonstrated, PEP- and pyruvate-based anaplerotic metabolism provides a unique combination of ¹³C enrichment in connected pathway intermediates of such bacteria using the ED pathway grown on [1-¹³C]glucose (16). Considering the well-defined transition of carbon atoms in the biochemical reactions involved, high ¹³C enrichment in pyruvate, together with low ¹³C enrichment in oxaloacetate results in PEP carboxylation, whereas the opposite results in pyruvate carboxylation, respectively (Fig. 4). The labeling pattern of alanine, reflecting its precursor pyruvate, and of aspartate, reflecting its precursor oxaloacetate, revealed two distinct groups: marine Gammaproteobacteria obviously use PEP carboxylase, whereas Alphaproteobacteria and Flavobacteria recruit pyruvate carboxylase.

FIG 4.

Anaplerotic fluxes in all strains studied that use the ED pathway for glucose catabolism. (A) The activity of PEP carboxylase causes a low ¹³C enrichment of oxaloacetate (aspartate), and ¹³C preferentially accumulates in pyruvate (alanine). G6P, glucose 6-phosphate; 6-PG, 6-phosphogluconate; KDPG, 2-keto-3-deoxy-6-phosphogluconate; GAP, glycerol-3-phosphate; PYR, pyruvate; OAA, oxaloacetate. (B) However, the activity of pyruvate carboxylase causes lower ¹³C enrichment in pyruvate (alanine) and higher enrichment in oxaloacetate (aspartate). Further details are described in reference 16. (C) The strains cluster into two groups based on the labeling patterns of aspartate and alanine, both shown as the relative fraction of the single-labeled (M₁) mass isotopomers. The strains that utilize pyruvate carboxylase are shown in orange, whereas the strains that use PEP carboxylase are shown in green.

Metabolic pathway use coincides with oxidative stress tolerance.

Many habitats in the marine environment impart oxidative stress; therefore, antioxidant mechanisms are important traits of marine microorganisms (58). To study the tolerance of marine bacteria to such oxidative conditions, we studied the tolerance to diamide (59). Diamide causes oxidative stress by oxidizing sulfhydryl bonds in the cytoplasm, which must be reduced at the expense of NADPH (60, 61). Here, we used a plate-based assay with diamide applied to a filter disc in the center of the plate. The area of the resulting halo of growth inhibition provided a direct measure of the sensitivity of the cells to oxidative stress (Fig. 5A). P. putida has robust resistance to oxidative stress (59). Thus, the tolerance of P. putida was used as the reference (100%) to classify the other tested bacteria according to their stress tolerance. Notably, most marine strains had a high tolerance to oxidative stress (Fig. 5A, middle and lower rows, and B). Certain isolates, e.g., Alteromonas sp. BIO-267, even exhibited a higher tolerance than P. putida. In comparison, E. adriatica, Vibrio sp. GWS-TZ-H304, and Pseudonocardiae bacterium T4 had substantially weaker tolerance and had a sensitivity equal to that of E. coli (Fig. 5A, top row, and B). Statistical analysis revealed that the ED pathway users in particular had a high stress tolerance (Fig. 5B, blue bars), whereas the strains that preferred the EMP route (Fig. 5B, red bars) were significantly more sensitive to oxidative stress (P = 0.05 by t test).

FIG 5.

Evaluation of oxidative stress tolerance by treatment with diamide. (A) Soft-agar plates with a central diamide-containing filter disc were photographed after 24 h of incubation. (B) To obtain a quantitative measure of stress resistance, the area of the resulting circular cell-free zone of inhibition or halo was calculated from the measured diameters of three replicates. The data were normalized to the halo area of P. putida (100%). The Flavobacterium isolate T15 did not form a homogeneous cell layer; therefore, representative data for this species could not be obtained.

Cluster analysis of ¹³C labeling signatures provides a metabolic degree of similarity among the isolates.

The aforementioned analysis of proteinogenic ¹³C labeling data using algebraic equations provided direct insight into selected flux properties: glucose catabolism and anaplerotic metabolism. Subsequently, we expanded our view of central metabolism by considering the entire set of ¹³C amino acid labeling data from the isotope experiments (see Fig. S4 in the supplemental material). Using P. inhibens as a reference, we observed that more than 90% of the marine strains had similar patterns with less than 20% variation in individual mass isotopomer fractions. The ¹³C fingerprints of these strains were different from those of the three marine strains Eudoraea adriatica, Vibrio sp. GWS-TZ-H304, and Pseudonocardiaceae bacterium T4 and from those of E. coli, C. glutamicum, and the two bacilli. A systematic and unsupervised analysis of the labeling data was conducted using a statistical approach. We performed an independent component analysis (ICA) of the data set. This analysis identified 24 specific signatures in the labeling patterns, that is, the independent components (ICs), which were informative in describing and differentiating the metabolism of the strains analyzed (see Fig. S5 and Fig. S6 in the supplemental material). On the basis of this finding, we then used a cluster analysis of the entire amino acid labeling data set (Table S2) to quantify the degree of similarity between the marine isolates. The resulting dendrogram (Fig. 6) shows that the degree of similarity on the level of ¹³C labeling matched with metabolic function; the identified metabolic strategies among the studied isolates clustered together nicely.

FIG 6.

Cluster analysis of marine isolates based on the amino acid labeling fingerprints. The corresponding phylum for each strain is indicated by the colored triangle as follows: dark purple for Gammaproteobacteria, light purple for Alphaproteobacteria, light blue for Flavobacteriia, and black for Actinobacteria. In addition, the metabolic strategies identified are shown. These strategies include the glycolytic route (EMPP [red] and EDP [blue]), the anaplerotic route (PEP carboxylase [green] and pyruvate carboxylase [orange]), and the use of the oxidative PP pathway (PPP) (active [turquoise] and not active [pink]). For strains using the EMP pathway, the anaplerotic pathway could not be elucidated (shown in white).

DISCUSSION

Because marine microbes have a capacity for rapid growth, they are a major component of global nutrient cycles. In their natural habitats, marine bacteria often have to face fluctuating conditions, scarce nutrient levels, and oxidative stress due to absorption of solar radiation (25, 62, 63). Questions regarding how their distribution is controlled and the diverse repertoire of nutrient transformations are major challenges, faced by contemporary biological oceanographers (64). In particular, glucose assimilation plays a major role, as marine bacteria may represent up to 40% of bacterial carbon production in the oceans (5). Here, we present a metabolic flux approach to study the carbon metabolism in marine bacteria that use glucose. The 25 selected strains represented globally distributed marine clades and families (26, 65, 66) such as Alteromonadaceae, Pseudoalteromonadaceae (Gammaproteobacteria), Rhodobacteraceae (Alphaproteobacteria), Flavobacteria, and Sphingobacteria of the Bacteriodetes phylum and Actinobacteria (Fig. 1 and Table 1).

Marine bacteria prefer the ED pathway as a glycolytic strategy.

A central finding of our study is the strong prevalence of the ED pathway among the marine bacteria investigated. More than 90% of them rely on the ED pathway as the sole or major glycolytic route (Fig. 3) and even lack a functional EMP pathway due to the absence of phosphofructokinase activity, although this does not necessarily imply the absence of the gene itself (Table 1). In contrast, the EMP pathway is nearly ubiquitous in the bacterial kingdom (13), and recent genomic analyses of more than 500 microbial species revealed that only 12% of prokaryotes rely solely on the ED pathway (14). In line with this, metabolic flux studies using various bacteria, including aerobic C. glutamicum (52) and B. subtilis (67), the facultative anaerobes E. coli (68) and Basfia succiniciproducens (69), and the anaerobes Lactobacillus plantarum (70) and Lactococcus lactis (71), identified the EMP pathway as the major catabolic pathway during growth on glucose. It has been proposed that the ED pathway may not play a major role in glucose metabolism but instead primarily functions in the breakdown of sugar acids that cannot be metabolized through the EMP pathway (14, 72). However, this conclusion seems not applicable to marine bacteria (this work). Marine microbes, at least on the basis of the broad collection of strains from different clades studied, instead form a specific subgroup among the prokaryotes with regard to catabolic pathway use. This observation is striking because the EMP pathway yields twice as much ATP as the ED pathway does and therefore appears superior (Fig. 7A). Interestingly, other phylogenetically distinct bacteria (Alphaproteobacteria and Gammaproteobacteria) also rely on the ED pathway for glucose catabolism (Fig. 3) (15, 16). The obvious use of the ED pathway by such a diverse group of bacteria suggests that this type of metabolism has a greater importance in nature, as was previously recognized (72), and might confer other advantages for cellular metabolism that overcome the drawback of a lower ATP yield. It is interesting to note that glucose 6-phosphate dehydrogenase activity was not specific for NADP⁺ in the ED pathway users (Table 1). Generally, this can be explained by glucose 6-phosphate dehydrogenases without preference for either cofactor or by different isoenzymes with different cofactor specificities (73). On the basis of the available genomic data, it becomes clear that most strains have a promiscuous glucose 6-phosphate dehydrogenase, because they possess a unique gene encoding it, with P. inhibens being the only exception (Table 2). This property may relate to the fact that the enzyme is required for the exclusive glucose catabolic pathway in these species, and therefore, its nonspecificity serves as a major mechanism to avoid catabolic NADPH overproduction (73). It should be noticed that, due to experimental requirements, the glucose level in the flux studies was higher than that naturally found in seawater (4, 5). Those species that possess the genetic repertoire for different glycolytic pathways (Table 2) might be able to switch pathway use, when glucose is scarce. However, the majority of the marine isolates that lack phosphofructokinase (Tables 1 and 2) seem to be bound to the ED pathway, independent of the nutrient level.

FIG 7.

Generation of NADPH and ATP in the strains studied, calculated based on the measured metabolic fluxes (Fig. 2) and the stoichiometry of the Embden-Meyerhof-Parnas (EMP), Entner-Doudoroff (ED), and pentose phosphate (PP) pathways. (A) The reactions are catalyzed by glucose 6-phosphate isomerase (step 1), phosphoglucokinase (step 2), fructose 1,6-bisphosphate aldolase (step 3), glucose 6-phosphate dehydrogenase (step 4), phosphogluconate dehydratase (step 5), 2-keto-3-desoxy-6-phosphogluconate aldolase (step 6), lumped reactions of GAP dehydrogenase, phosphoglycerate kinase, phosphoglycerate mutase, and enolase (step 7), and pyruvate kinase (step 8). F6P, fructose 6-phosphate; FBP, fructose 1,6-bisphosphate. (B) The calculation of NAPDH generation was based on the formation of one and two NADPH molecules via the ED and PP pathways, respectively. The estimation of ATP production was based on the direct formation of 3 ATP (EMP), 2 ATP (EDP), and 8/3 ATP (PPP) and indirect formation via the oxidation of NADH formed in the respiratory chain at a P/O ratio of 2 (93). The stoichiometry for NADH production is 2 NADH (EMP), 1 NADH (EDP), and 5/3 NADH (PPP). Prior to the estimation, the measured fluxes were corrected on the basis of the enzymatic repertoire (Table 1), i.e., small EMP fluxes predicted by the algebraic calculation were set to zero when phosphofructokinase was absent, and this flux was then attributed to the ED pathway.

PEP and pyruvate carboxylation are evenly distributed among marine bacteria that use glucose.

The type of anaplerotic strategy has been shown to influence the flexibility under changing conditions (57). This seems to be of particular importance for life in the highly heterogeneous oceans, which exhibit constantly changing physical and chemical gradients (25). In contrast to a clear preference for the catabolic ED route, the marine bacteria studied did not specifically rely on a distinct anaplerotic route, but the carboxylation of PEP or pyruvate for fueling of the TCA cycle was evenly distributed among the strains. We conclude that both strategies provide sufficient flexibility for bacteria that use glucose to adapt to the changing conditions in the sea.

The ED pathway provides redox power to protect against oxidative stress.

Oxidative stress is common in the oceans and is caused by the absorption of solar radiation (62) and the metabolism of marine algae (74). Marine bacteria possess various antioxidant mechanisms to eliminate reactive oxygen species (ROS), the inducers of oxidative stress (58). Particularly, NADPH is an oxidative stress protectant that is required by many important antioxidant defense mechanisms (75) and is important for counteracting oxidative stress (61, 76, 77). A recent study revealed the importance of the ED pathway for oxidative stress protection in P. putida (59). As impressively shown, the introduction of a functional phosphofructokinase forced the organism to shift the flux from the natural ED pathway to the EMP pathway, which significantly decreased its tolerance to oxidative stress. Similarly, the human pathogen Pseudomonas aeruginosa exclusively uses the ED pathway and potentially provides far more NADPH than needed for anabolism: a benefit to counteract oxidative stress imposed by the host during infection (16). Based on these findings, it is tempting to speculate that the conserved use of the ED pathway in marine bacteria might similarly support their high tolerance to oxidative stress through elevated NADPH formation. Notably, all ED pathway users possess a superior tolerance to oxidative stress, whereas strains that utilize the EMP pathway have a weaker stress tolerance (Fig. 5). Although the lack of high-level NADPH specificity for the glucose 6-phosphate dehydrogenase (Table 1) does not exclude at least a partial formation of NADH with the ED pathway in the corresponding strains, the nonspecificity of the enzyme surely enables the organism to cope with dynamic fluctuations in NADP⁺ and NADPH availability (73, 75). Especially under conditions of oxidative stress, cells exhibit a drastically disturbed redox equilibrium and the NADPH/NADP⁺ ratio is reduced almost 10-fold (75). One can expect that this strongly promotes NADPH formation by glucose 6-phosphate dehydrogenase. Assuming stoichiometric formation of NADPH by glucose 6-phosphate dehydrogenase, which appears reasonable for such stress conditions, the resulting stoichiometry for the ED and EMP pathways, the quantified flux partitioning ratio (Fig. 2), and the identified enzymatic inventory (Table 1) allow for the estimation of the supply of reducing power (NADPH) and energy (ATP) under conditions of oxidative stress. With the given assumptions, all ED pathway users would supply large amounts of NADPH (Fig. 7), which could explain their high robustness (Fig. 5), whereas strains that utilize the EMP pathway and generate less NADPH (Fig. 7) have a weaker stress tolerance (Fig. 5). Particularly, members of the Roseobacter clade, Alteromonadaceae, Pseudoalteromonadaceae, and Flavobacteriaceae, all ED pathway users (Fig. 3), were found to be highly tolerant to oxidative conditions (Fig. 5) (78) and are abundant and active in marine surface waters that receive high light intensities and impose severe oxidative stress (79, 80). Similarly, members of the Roseobacter clade (81), Flavobacteriaceae (82), and Alteromonadaceae and Pseudoalteromonadaceae (83) live physically attached to marine algae, which are also strong inducers of oxidative stress related to the release of ROS during photosynthesis (74, 84). Clearly, bacterial cells possess a variety of mechanisms for managing oxidative stress (85), and one cannot fully discern the contributions of ED pathway-derived reducing power derived from other mechanisms within the cell, such as glutathione and various enzymes. Owing to the central importance of NADPH as the driver of many of these mechanisms, it is, however, likely that life in oxidative marine environments at least partially benefits from the enhanced NADPH supplied via the ED pathway and is one major reason why this pathway is dominating in the marine bacteria studied.

A significant contribution of transhydrogenase PntAB, which might potentially contribute to NAPDH formation, seems unlikely. Careful inspection of the genomic data does not support a central role of this enzyme. For example, bacteria with high robustness, i.e., P. haloplanktis, P. marina, and P. dokdonensis, do not comprise pntAB genes in their genome, whereas stress-sensitive bacteria such as E. adriatica and E. coli are equipped with pntAB genes (Table 2).

The EMP pathway might be beneficial in anoxic marine environments.

Among the marine strains, Pseudonocardiaceae T4 and Vibrio sp. GWS-TZ-H304 used the EMP pathway (Fig. 3). Notably, both bacteria were isolated from low-oxygen habitats. Pseudonocardiaceae T4 grows in an intertidal mudflat (35) known to have low oxygen levels below the surface (86). Vibrio sp. GWS-TZ-H304 was originally isolated from an oxic-anoxic sediment boundary (37), and other Gammaproteobacteria have been found in oxygen-minimum zones (87, 88). In these habitats, oxidative stress and nonglycolytic energy generation are low, and energy efficiency rather than redox efficiency determines the glycolytic strategy; therefore, a functional EMP pathway might be beneficial (14).

Marine bacteria cluster according to their metabolic flux patterns.

As shown, the ¹³C signatures contained discriminatory information that can distinguish between the strains on a metabolic basis (Fig. 6). The dendrogram derived indicated two major groups, which differ in their catabolic pathway use: the EMP and ED pathway users. The latter group separates into two subgroups with regard to anaplerotic metabolism. In this regard, it is interesting to note that the metabolic similarity determined does not always correlate with the phylogenetic similarity (Fig. 6). We suggest taking such a metabolic viewpoint in order to obtain more functional insight into the role or similarity of the cells examined, complementary to that based on DNA sequences. This would reflect the observation that the physiology of microbes, of which metabolism is a key component, might be the key driver for their evolution (89). Regarded from this viewpoint, the Alteromonas and Pseudoalteromonas strains of Gammaproteobacteria clustered with Leeuwenhoekiella sp. Pic90, a Flavobacterium; all of these strains used PEP carboxylase. Other Gammaproteobacteria, i.e., Pseudomonas sp. GWS-TZ-H209 and Oceanospirillaceae bacterium T17, used pyruvate carboxylase, as did most of the Alphaproteobacteria. As recently suggested for environmental bacteria, their main evolutionary drive is the expansion of their metabolic networks toward new chemical landscapes rather than perpetuation and spreading of their DNA sequences (89).

Conclusions.

Unlike terrestrial ecosystems, microorganisms are the main form of biomass in the oceans and comprise the largest living surface on the planet (26). Most of the microbes that use glucose that were studied rely on the ED pathway to survive in marine environments. Due to the fact that glucose is a major nutrient in seawater (5), the ED pathway may play a significant role in global biogeochemical cycles (13, 14).