Amino Acid and Sugar Catabolism in the Marine Bacterium Phaeobacter inhibens DSM 17395 from an Energetic Viewpoint

Heterotrophic members of the bacterioplankton serve the marine ecosystem by transforming organic matter, an activity that is governed by the bacterial growth efficiencies (BGEs) obtained under given environmental conditions. In marine ecology, the concept of BGE refers to the carbon assimilation efficiency within natural communities. The marine bacterium studied here, Phaeobacter inhibens DSM 17395, is a copiotrophic representative of the globally abundant Roseobacter group, and the 15 catabolic pathways investigated are widespread among these marine heterotrophs. Combining pathway-specific catabolic ATP yields with in-depth quantitative physiological data could (i) provide a new baseline for the study of growth energetics and efficiency in further Roseobacter group members and other copiotrophic marine bacteria in productive coastal ecosystems and (ii) contribute to a better understanding of the factors controlling BGE (including the additional energetic burden arising from widespread secondary-metabolite formation) based on laboratory studies with pure cultures.

ABSTRACT

Growth energetics and metabolic efficiency contribute to the lifestyle and habitat imprint of microorganisms. Roseobacters constitute one of the most abundant and successful marine bacterioplankton groups. Here, we reflect on the energetics and metabolic efficiency of Phaeobacter inhibens DSM 17395, a versatile heterotrophic roseobacter. Fourteen different substrates (five sugars and nine amino acids) and their degradation pathways were assessed for energetic efficiencies based on catabolic ATP yields, calculated from net formed ATP and reducing equivalents. The latter were converted into ATP by employing the most divergent coupling ratios (i.e., ions per ATP) currently known for F₁F_o ATP synthases in heterotrophic bacteria. The catabolic ATP yields of the pathways studied in P. inhibens differed ∼3-fold. The actual free energy costs for ATP synthesis were estimated at 81.6 kJ per mol ATP (3.3 ions per ATP) or 104.2 kJ per mol ATP (4.3 ions per ATP), yielding an average thermodynamic efficiency of ∼37.7% or ∼29.5%, respectively. Growth performance (rates, yields) and carbon assimilation efficiency were determined for P. inhibens growing in process-controlled bioreactors with 10 different single substrates (Glc, Man, N-acetylglucosamine [Nag], Phe, Trp, His, Lys, Thr, Val, or Leu) and with 2 defined substrate mixtures. The efficiencies of carbon assimilation into biomass ranged from ∼28% to 61%, with His/Trp and Thr/Leu yielding the lowest and highest levels. These efficiencies correlated with catabolic and ATP yields only to some extent. Substrate-specific metabolic demands and/or functions, as well as the compositions of the substrate mixtures, apparently affected the energetic costs of growth. These include energetic burdens associated with, e.g., slow growth, stress, and/or the production of tropodithietic acid.

IMPORTANCE Heterotrophic members of the bacterioplankton serve the marine ecosystem by transforming organic matter, an activity that is governed by the bacterial growth efficiencies (BGEs) obtained under given environmental conditions. In marine ecology, the concept of BGE refers to the carbon assimilation efficiency within natural communities. The marine bacterium studied here, Phaeobacter inhibens DSM 17395, is a copiotrophic representative of the globally abundant Roseobacter group, and the 15 catabolic pathways investigated are widespread among these marine heterotrophs. Combining pathway-specific catabolic ATP yields with in-depth quantitative physiological data could (i) provide a new baseline for the study of growth energetics and efficiency in further Roseobacter group members and other copiotrophic marine bacteria in productive coastal ecosystems and (ii) contribute to a better understanding of the factors controlling BGE (including the additional energetic burden arising from widespread secondary-metabolite formation) based on laboratory studies with pure cultures.

INTRODUCTION

The marine ecosystem contributes ∼50% of global primary production of biomass through the activity of phototrophic phytoplankton, driving the biogeochemical cycles and fueling the biosphere in the world’s oceans (1). Net primary production is influenced by the action of heterotrophic bacterioplankton, the most important and abundant stakeholders for biotic transformation and remineralization of organic matter in the marine realm (2). The activity of heterotrophic bacterioplankton influences oceanic and atmospheric carbon budgets by transforming available dissolved organic carbon to bacterial biomass (assimilation) and to CO₂ (dissimilation). Despite their small area (∼7% of the oceans' surface area), coastal ecosystems are hot spots of primary productivity and bacterial respiration, contributing as much as 30% to the total ocean budget (3, 4). This is caused by the constant fueling of coastal oceans with high nutrient loads, in contrast to the mostly oligotrophic open oceans.

Based on their main physiological growth traits, many members of the heterotrophic bacterioplankton can be classified into oligotrophs or copiotrophs, i.e., organisms that encounter (constantly) low or (transiently) high nutrient concentrations in their natural habitat (for an overview, see references 5 and 6). Recent studies suggest that these opposing lifestyle strategies are based on the interconnection of physiological properties and genomic adaptations (7–9). Copiotrophs are physiologically characterized by high growth rates, nutritional versatility, opportunistic exploitation of transient nutrient pulses (feast-and-famine control), and complex traits such as extensive regulatory potential, chemotaxis, or quorum sensing. This is consistent with the relatively large genome size of copiotrophs, codon usage bias, and phage defense by CRISPR-Cas. The fact that oligotrophs apparently dominate the heterotrophic bacterioplankton over long time scales (7) can be attributed to the collapse of interim copiotroph peaks due to nutrient exhaustion or viral lysis, as suggested by the “kill-the-winner” hypothesis (10).

Growth performance (rate, yield) and nutritional strategy (oligotrophy versus copiotrophy) are key determinants for the lifestyle and habitat success of individual bacterioplankton members. These traits shape the complex temporal (continuous versus peaking) and substrate-constrained (specialist versus generalist) contributions of bacteria to biogeochemical processes (11). However, our current understanding of environmental factors affecting in situ growth performance among heterotrophic marine bacteria is still rather limited. Growth rates and growth yields, as well as the associated allocation of resources between assimilation and dissimilation, are long-known hallmarks of general microbiology (12). While they have been long studied in model bacteria, e.g., Escherichia coli and Bacillus subtilis, and in production strains, far less is known for marine bacteria. But even for well-studied bacteria, the metabolic complexity of growth is still less well understood than specific molecular processes (13). To estimate in situ carbon assimilation in the oceans, the concept of bacterial growth efficiency (BGE) was introduced in marine ecology. BGE is defined as the amount of bulk biomass that is formed per unit of assimilated substrate carbon (14). The latter is generally calculated from measured O₂ consumption. BGE is influenced by a variety of organism-specific and environmental parameters (14, 15). In contrast with the complexity of natural settings, pure-culture studies provide easier and more-defined access for studying growth energetics/efficiency and the underlying cellular attunement between assimilation and dissimilation.

Marine alphaproteobacterial Rhodobacteraceae comprise a multitude of physiologically distinct and metabolically versatile members of the bacterioplankton. Most of them belong to the globally abundant Roseobacter group (16, 17), which is encountered in diverse marine habitats (18, 19) and which contributes to the recycling of seasonal peaks of phytoplankton-derived organic matter (20–22). The study organism Phaeobacter inhibens DSM 17395 is an aerobic heterotrophic representative of the Roseobacter group, possessing versatile catabolic and complex regulatory capacities (23–25) and colonizing abiotic/biotic surfaces in coastal regions as its primary habitat (26). P. inhibens, like other members of the genera Phaeobacter and Ruegeria, has the means to produce the antibiotic secondary metabolite tropodithietic acid (TDA) (27, 28).

In this study, we reflect on the energetics of amino acid and sugar catabolism in P. inhibens by combining theoretical considerations (i.e., substrate-specific catabolic ATP yields and Gibbs free energy yields) with quantitative physiological data (i.e., rates and molar growth yields, carbon assimilation efficiencies) determined in process-controlled bioreactors. Essential background for the present work was provided by the previous proteomic and metabolic reconstruction of the catabolic pathways for nine amino acids (25) and five sugars (29) in P. inhibens (Fig. 1). This study attempts to link the energetics and efficiency of growth to the habitat success of heterotrophic roseobacters.

FIG 1.

Energetics of catabolism of five different sugars and nine different amino acids in P. inhibens DSM 17395. The catabolic ATP yield of each pathway (Table 1; also Table S1 in the supplemental material) was calculated from assigned enzymatic reactions previously reconstructed in P. inhibens on the basis of differential proteomics, metabolomics, and enzyme assays (25, 29). The generation of ATP and reducing equivalents (shown as [H] for NADH, [H]* for FADH₂, and e^– for Fd_red) or their consumption during substrate catabolism is color-coded green or red, respectively. Ten (single) substrates (indicated by white lettering on a gray background) were selected for comprehensive growth studies in process-controlled bioreactors (Fig. 3 and 5; also Fig. S3 and Table S5). Growth substrates and degradation intermediates are abbreviated as follows: Ac, acetate; AcCoA, acetyl-CoA; Ala, alanine; 2-ABz, 2-aminobenzoate; 2-ABz-CoA, 2-aminobenzoyl-CoA; Cit, citrate; Ery-4P, erythrose-4-phosphate; Frc, fructose; Frc-6P, fructose-6-phosphate; Fum, fumarate; GAP, glyceraldehyde-3-phosphate; Gla-6P, glucosamine-6-phosphate; Glc, glucose; Glc-6P, glucose-6-phosphate; Gly, glycine; Glu, glutamate; His, histidine; iCit, isocitrate; Ile, isoleucine; KDPG, 2-keto-3-deoxy-6-phosphogluconate; Leu, leucine; Lys, lysine; Mal, malate; Man, mannose; Met, methionine; Nag, N-acetylglucosamine; Nag-6P, N-acetylglucosamine-6-phosphate; OxAc, oxaloacetate; 2-OG, 2-oxoglutarate; 3-OxoSub-CoA, 3-oxosuberoyl-CoA; Phe, phenylalanine; PhePyr, phenylpyruvate; PEP, phosphoenolpyruvate; PropCoA, propanoyl-CoA; Pyr; pyruvate; Rib-5P, ribose-5-phosphate; Sedo-7P, sedoheptulose-7-phosphate; Ser, serine; Sucr, sucrose; Succ, succinate; Succ-CoA, succinyl-CoA; Thr, threonine; Trp, tryptophan; Val, valine; Xul-5P, xylulose-5-phosphate; and Xyl, xylose. Enzyme names have been described previously (25, 29).

RESULTS AND DISCUSSION

We have combined theoretical considerations on the energetics of substrate catabolism with quantitative physiological data from process-controlled bioreactors in order to determine energetic and metabolic efficiencies for the marine bacterium P. inhibens DSM 17395 during growth with sugars and amino acids, supplied as single substrates or defined mixtures. The catabolic pathways studied (Fig. 1) are widely distributed among members of the Roseobacter group (25, 29) and possibly other marine bacteria, which should allow transfer of the findings of this study to other heterotrophic bacterioplankton members growing with sugars and amino acids.

Theoretical considerations on the energetics of substrate catabolism.

Catabolic ATP yields (see Table S1 in the supplemental material) and standard Gibbs free energy yields (see Table S2) were calculated in order to assess the energetic differences of experimentally reconstructed catabolic pathways for five sugars (29) and nine amino acids (25) in P. inhibens (Fig. 1). Succinate served as a reference substrate due to its central position in the catabolic network and the tricarboxylic acid (TCA) cycle. In contrast to standard Gibbs free energy (ΔG⁰′) yields, catabolic ATP yields consider energetic differences based on specific degradation pathways and are thus biologically more meaningful (30). For example, glucose catabolism via the Entner-Doudoroff (ED) glycolytic pathway is less efficient (by 1 ATP) than that via its analog, the Embden-Meyerhof-Parnas (EMP) pathway, although in both cases the standard free energy yield is –2,872 kJ (mol glucose)⁻¹ (31).

(i) Catabolic ATP yields. Catabolic ATP yields were calculated from the net numbers of ATPs and reducing equivalents (from NADH and reduced flavin adenine dinucleotide [FADH₂]/reduced quinol [QH₂]) formed during substrate catabolism to CO₂ (Fig. 1; Table 1; also Table S1 in the supplemental material). Reducing equivalents were converted into ATP by employing two different coupling ratios for the F₁F_o ATP synthase as a basis. The coupling ratio refers to the number of ions (either H⁺ or Na⁺) that need to be pumped back into the cell by ATP synthase to produce 1 ATP. This number is determined by the number of c subunits in the F_o subunit (32, 33), each of which binds and translocates one ion. Thus, for a 360° rotational turn, a c₈ ring translocates 8 ions (e.g., in mitochondria), while a c₁₅ ring requires 15 ions (e.g., in cyanobacteria) (34). Since this is coupled to the synthesis of 3 ATPs in both cases (35), a larger c ring requires more ions to be translocated than a smaller c ring. The coupling ratio and c_n ring stoichiometry have not been determined for P. inhibens, nor, to our knowledge, for any other aerobic, heterotrophic marine bacterium. Comparative analysis of the amino acid sequence of the c subunit (AtpE) of P. inhibens (see Fig. S4 in the supplemental material) with that from other organisms with known c ring stoichiometry failed to provide a clear hint as to what coupling ratio may be assumed. Thus, we selected the two most divergent coupling ratios for ATP synthase from the few coupling ratios available from heterotrophic bacteria (34) for our calculations: 3.3 ions per ATP (the c₁₀ ring of Escherichia coli) and 4.3 ions per ATP (the c₁₃ ring of the strictly alkaliphilic soil bacterium Bacillus pseudofirmus OF4). Combined with the redox-driven export of 10 (with NADH) or 6 (with QH₂) protons per 2 electrons transferred through the aerobic respiratory chain, this results in the generation of 3.0 mol ATP (mol NADH)⁻¹ and 1.8 mol ATP (mol QH₂)⁻¹ for a c₁₀ ring or 2.3 mol ATP (mol NADH)⁻¹ and 1.4 mol ATP (mol QH₂)⁻¹ for a c₁₃ ring.

TABLE 1.

Energetics of sugar and amino acid catabolism in P. inhibens DSM 17395^a

Substrate	Energetics of dissimilation of substrate to CO2b
	Catabolic ATP yield [mol ATP (mol Sdiss)–1] with the following coupling ratio (H+ ATP–1)c:		Standard Gibbs free energy yieldd [kJ (mol Sdiss)–1]
	3.3	4.3
Glucose	36.6	28.8	–2,872.4
Sucrose	73.2	57.6	–5,790.2
Xylose	30.7	24.2	–2,412.0
Mannitol	39.6	31.1	–3,082.5
N-Acetylglucosamine	47.4	37.1	–3,786.1
Phenylalanine	48.7	38.4	–4,330.8
Tryptophan	60.6	47.3	–5,016.3
Threonine	22.2	17.4	–1,814.0
Valine	31.9	24.8	–2,603.4
Leucine	40.7	31.8	–3,248.8
Isoleucine	40.7	31.8	–3,248.0
Methionine	22.6	17.6	–1,598.2
Histidine	26.6	20.9	–2,261.4
Lysinee
Catabolized via cadaverine	33.4	25.9	–3,078.4
Catabolized via 2-aminoadipate	34.4	26.9	–3,078.4
Succinate	19.6	15.3	–1,519.2

^aSee Fig. 2. Catabolic ATP yields were calculated for each substrate based on experimentally reconstructed peripheral catabolic routes (Fig. 1; see Table S1 in the supplemental material for details). Succinate served as a reference substrate due to its central position in metabolism.

^bS_diss, dissimilated substrate.

^cCoupling ratio of F₁F_o ATP synthase with the c₁₀ (3.3 H⁺ ions per ATP) or c₁₃ (4.3 H⁺ ions per ATP) ring in the F_o component.

^dStandard Gibbs free energy yields (ΔG⁰ and ΔG⁰′) were calculated from published or calculated $\Delta G_{\text{f}}^0$ values (see Table S2 in the supplemental material for details).

^eSee Fig. 1.

Based on these assumptions, the calculated pathway-specific catabolic ATP yields for P. inhibens range from 22.2 (with Thr) to 73.2 (with sucrose) mol ATP (mol dissimilated substrate [S_diss])⁻¹ for the more-efficient coupling ratio and from 17.4 to 57.6 mol ATP (mol S_diss)⁻¹ for the less-efficient coupling ratio (Table S1 in the supplemental material).

The catabolic ATP yields are generally reduced by the costs for substrate scavenging and active uptake. The putative substrate-specific uptake systems in P. inhibens were previously inferred from differential analyses of the membrane protein-enriched fractions (25, 29, 36) and are compiled in Table S4 in the supplemental material. Sugars and most amino acids resort to high-affinity ABC-type solute uptake systems that diminish the catabolic ATP yield by approximately 1 mol ATP (mol S_diss)⁻¹. The amino acids Ala, Thr, and Gly are probably imported with Na⁺ by symporters, reducing the catabolic ATP yield by 0.5 mol ATP (mol S_diss)⁻¹ only (37). TRAP (tripartite ATP-independent periplasmic)-type transport systems operate in the uptake of dicarboxylates in conjunction with two Na⁺ ions, thus lowering the catabolic ATP yield by 0.66 mol ATP (mol S_diss)⁻¹ (38). Due to the merely putative assignment of transporters in P. inhibens, additional energetic expenditures for substrate uptake could not be considered in the present calculations.

(ii) Free energy costs for ATP synthesis and thermodynamic efficiency. Pathway-specific catabolic ATP yields revealed a strong linear relationship (R², >0.96) with the corresponding Gibbs free energy yields for substrate oxidation to CO₂ (Fig. 2). This allowed estimation of the actual ΔG⁰′ costs for ATP synthesis as 81.6 kJ (mol ATP)⁻¹ or 104.2 kJ (mol ATP)⁻¹ at a coupling ratio of 3.3 or 4.3 ions per ATP, respectively. The thermodynamic efficiencies of catabolic pathways were determined by relating the actual ΔG⁰′ costs for ATP synthesis to the average standard ΔG⁰′ cost of 30.5 kJ (mol ATP)⁻¹ for synthesizing ATP from ADP and P_i (39). Across the pathways studied, the thermodynamic efficiency averaged 37.7% ± 1.9% (with 3.3 ions per ATP) or 29.5% ± 1.5% (with 4.3 ions per ATP), within a range typically observed for microorganisms. Thermodynamic efficiencies differed more pronouncedly with amino acids than with sugars, reflecting the more-divergent peripheral routes employed for their catabolism (Fig. 1). The standard free energy costs for ATP synthesis were found to be independent of the substrate utilized by P. inhibens, allowing for predictions of catabolic ATP yields and thermodynamic efficiencies for substrates with unknown or experimentally unproven degradation pathways. The thermodynamic values determined refer to standard conditions and thus do not consider cell-specific concentrations of the respective reactants.

FIG 2.

The estimated standard free energy costs for ATP synthesis amount to 81.6 or 104.2 kJ (mol ATP)⁻¹ in P. inhibens DSM 17395. These values were calculated based on the two most divergent coupling ratios known for ATP synthases in heterotrophic bacteria, i.e., 3.3 H⁺ ions per ATP synthesized in Escherichia coli and 4.3 H⁺ ions per ATP synthesized in Bacillus pseudofirmus OF4 (34). Standard free energy costs for ATP synthesis were obtained by linear correlation (R², >0.96; computed with SigmaPlot, version 12.0 [Systat Software Inc., San Jose, CA, USA]) of substrate-specific catabolic ATP yields of P. inhibens with corresponding standard Gibbs free energy yields (Table 1; also Tables S1 and S2 in the supplemental material). For amino acid sequence alignment of the c subunit (AtpE) of ATP synthase in P. inhibens, see Fig. S4 in the supplemental material. S_diss, dissimilated substrate.

Conceptual design of bioreactor studies.

Ten different substrates of the catabolic network shown in Fig. 1 were selected for bioreactor studies based on the following experimental criteria: (i) support of reproducible growth of P. inhibens without extensive biofilm formation, (ii) unambiguous quantification of substrate depletion via high-performance liquid chromatography (HPLC) analyses, and (iii) the absence of unquantifiable catabolic intermediates (e.g., methanethiol). For single-substrate studies, Glc, N-acetylglucosamine (Nag), Man, His, Leu, Lys, Phe, Trp, Thr, and Val were selected. To correlate these studies with the complex nutritional conditions prevailing in the natural habitat, two defined substrate mixtures were also included in the present study: Mix-10, which contains all the single substrates listed above, and Mix-11, which additionally contains Ile.

Cultures were monitored by quantifying (i) the biomass produced (as the optical density at 600 nm [OD₆₀₀] and the concentration of cells [in grams {dry weight} per liter] [CDW]), (ii) the carbon substrate(s) and O₂ consumed, and (iii) the CO₂ produced. These values were used to calculate molar growth yields (Y) that relate the amount of biomass formed (generally termed X) to substrate consumption (e.g., to total substrate carbon [$Y_{X/{\text{C}}_{\text{total}}}$], ATP [Y_X/ATP], or O₂ [$Y_{X/{\text{O}}_2}$]). In addition, the corresponding consumption (substrate, O₂) and production (X, CO₂, TDA) rates were calculated. The data obtained for each substrate condition are compiled in Fig. S3 and Table S5 in the supplemental material.

The extensive physiological data and parameters calculated are presented in Fig. 3 to 6. Here, we discuss in detail two selected single substrates, Nag and Phe, that we believe are (eco)physiologically most relevant, and we compare them to Mix-11 (see the next section) (Fig. 3 and 4). We then compare these three substrate conditions to the other eight single-substrate conditions tested and to Mix-10 (see “Comparison of growth performance across an extended substrate range” below) (Fig. 5 and 6).

FIG 3.

Growth of P. inhibens DSM 17395 with N-acetylglucosamine, phenylalanine, and the defined substrate mixture Mix-11 (containing three sugars and eight amino acids) in process-controlled bioreactors. (A) Basic growth parameters assessed by monitoring the production of biomass (by the optical density at 600 nm [OD₆₀₀] and the concentration of cells [CDW {dry weight} per liter]), the consumption of the substrate (converted to carbon equivalents) and O₂, and the production of CO₂ and tropodithietic acid (TDA; measured indirectly from the absorbance change in cell-free culture supernatants at 398 nm [A₃₉₈]). (B) Biomass-specific rates for the consumption of substrate carbon ($q_{{\text{C}}_{\text{total}}/X}$) and O₂ ($q_{{\text{O}}_2\hspace{0.17em}/X}$), as well as for the production of biomass (q_X/X), CO₂ ($q_{{\text{CO}}_2\hspace{0.17em}/X}$), and TDA (q_TDA/X). Shaded areas delimit the duration of exponential growth. Further details on the physiological parameters obtained are compiled in Table S5 in the supplemental material. Data are based on three to four biological replicates.

FIG 6.

Relative times for half-depletion of individual substrates from two different substrate mixtures (Mix-10 and Mix-11) by P. inhibens DSM 17395. A value of 1 (marked by the vertical green line) indicates that a given substrate is depleted to half of its initial concentration equally fast in the experiments and the simulations (based on single substrates). Values of >1 indicate that half-depletion of these substrates took longer if they were supplied within a mixture than if they were supplied as single substrates.

FIG 4.

Consumption profiles of individual compounds during the growth of P. inhibens DSM 17395 with the defined substrate mixture Mix-11, containing three sugars (glucose, mannitol, and N-acetylglucosamine) and eight amino acids (histidine, isoleucine, leucine, lysine, phenylalanine, threonine, tryptophan, and valine).

FIG 5.

Comparison of key growth and energetic parameters determined for P. inhibens DSM 17395 during growth in bioreactors under 12 different substrate conditions. All parameters were normalized for comparison, with 1.0 (red) equivalent to the highest value, and 0.0 (blue) equivalent to the lowest value, of the respective parameter (e.g., CDW). Substrates were sorted according to their catabolic ATP yields (Table 1) in decreasing order (from top to bottom).

Bioreactor studies with Nag, Phe, or Mix-11.

Nag and Phe were specifically selected (Fig. 3; also Table S5 in the supplemental material) for the following (eco)physiological reasons. The catabolic pathways involved (i) are biochemically very different (Fig. 1) and (ii) provide similar catabolic ATP yields [e.g., 47.4 versus 48.7 mol ATP (mol S_diss)⁻¹ at 3.3 ions per ATP] but (iii) differ in standard free energy yields [–3,786 versus –4,331 kJ (mol S_diss)⁻¹] (Table 1). Moreover, (iv) the chitin monomer Nag is highly abundant in the marine environment (40) and is estimated to account for ∼10% of heterotrophic bacterial production in the ocean (41). Finally, (v) the catabolism of Phe employs an O₂-consuming monooxygenase to transform phenylacetyl coenzyme A (phenylacetyl-CoA) to 1,2-epoxyphenylacetyl-CoA (25, 42). The Phe degradation intermediate phenylacetyl-CoA provides the carbon backbone for the antibiotic TDA (43) and is also a precursor for the biosynthesis of algaecidal roseobacticides (44). The latter are produced only in the presence of 3-phenylpropanoids that originate from the decay of algal cells (45). These two classes of secondary metabolites have been associated with a lifestyle switch of P. inhibens from a mutualistic partner of algae (TDA) to an algal parasite (roseobacticides) (46).

(i) Basic growth parameters. Cultures reached the maximal optical density (OD_max) and biomass concentration (CDW_max) upon complete depletion of the respective carbon substrate(s). OD_max values were pronouncedly lower with Nag (2.7 ± 0.2) than with Phe (4.3 ± 0.1), whereas CDW_max values differed only by 10% between these two single-substrate conditions (Table S5 in the supplemental material). In the case of Mix-11, the OD_max achieved (3.0 ± 0.1) was similar to that for growth with Nag, although the CDW_max was ∼30% lower, and cells displayed pleomorphic features, as reported previously for the growth of P. inhibens with a variety of single amino acids (25). At OD_max, O₂ consumption (∼89 mM) and CO₂ production (∼61 mM) were markedly higher with Phe than with Nag or Mix-11 (∼51 or 48 mM for O₂; ∼47 or ∼41 mM for CO₂). In turn, the respiratory quotient (RQ; calculated as the CO₂/O₂ ratio) was lower with Phe (0.69), which was much lower than the expected value of 0.90 calculated from the CO₂/O₂ ratio in the dissimilation equation (Table S2). For Nag and Mix-11, the experimental values for the RQ were similar (∼0.9). These RQ values were much higher than that for Phe and deviated less strongly from the stoichiometry-inferred RQ (1.00).

(ii) Growth rates. During active growth, cultures with Phe grew exponentially for 24 h, whereas those with Nag first grew exponentially for ∼20 h and then grew linearly for a shorter time (∼6.5 h). When supplied with Mix-11, P. inhibens displayed evenly long exponential (∼22 h) and linear (∼23 h) growth phases. The exponential growth rates (μ_exp) were similar for OD- and CDW-based values and for all three substrate conditions: 0.17 h⁻¹ (Phe), 0.16 h⁻¹ (Nag), and 0.18 h⁻¹ (Mix-11) (Fig. 3B; Table S5 in the supplemental material). Substantially higher μ_exp values were observed previously with complex media, such as the widely used marine broth (MB) (0.36 h⁻¹) or a mineral medium containing Casamino Acids (0.42 h⁻¹). The duration of exponential growth in these cultures was, however, much shorter (24, 47). Comparison with other aerobic marine bacteria showed that the μ_exp values of P. inhibens observed here are higher than those reported for the oligotrophic SAR11 clade member Pelagibacter ubique (minimum, ∼0.02 h⁻¹) (48) and lower than that of the ultra-fast-growing salt marsh isolate Vibrio natriegens (maximum, ∼5.8 h⁻¹) (49, 50). It remains to be seen whether the upper limit of the growth rates for P. inhibens and other marine heterotrophic bacteria is controlled by “molecular crowding,” i.e., diffusion-limited provision of ribosomes with tRNA complexes, as was recently demonstrated for E. coli (51).

(iii) Biomass-specific rates for O₂ consumption and CO₂ production. The O₂ consumption rates ($q_{{\text{O}}_2\hspace{0.17em}/X}$) during exponential growth with Nag and Mix-11 were very similar, 7.6 and 7.5 mmol O₂ (g of cells_dry)⁻¹ h⁻¹, whereas with Phe, these were considerably higher (by ∼35%). The CO₂ production rate ($q_{{\text{CO}}_{2\hspace{0.17em}}/X}$) corresponds to the rate of carbon dissimilation (C_diss) and was, again, lower during exponential growth with Nag and Mix-11 [6.0 and 5.0 mmol CO₂ (g cells_dry)⁻¹ h⁻¹] than with Phe [7.2 mmol CO₂ (g cells_dry)⁻¹ h⁻¹]. These values for $q_{{\text{O}}_2\hspace{0.17em}/X}$ and $q_{{\text{CO}}_2\hspace{0.17em}/X}$ affiliate with the lower side of the range [∼5 to 25 mmol O₂ or CO₂ (g cells_dry)⁻¹ h⁻¹] reported for chemostat-based studies with Bacillus subtilis (52). During linear growth of P. inhibens with Nag and, even more pronouncedly, with Mix-11, CDW-specific rates for carbon and O₂ consumption, as well as for CO₂ production, decelerated continuously (Fig. 3B). This cannot be attributed to a kinetic limitation in substrate uptake, since neither the organic carbon substrates (≥50% remaining) nor O₂ (continuously high aeration) was limited upon transit of the cultures from exponential to linear growth. In the case of Mix-11, linear growth started upon depletion of the three sugars supplied and was then fueled solely by the amino acids still available (Fig. 4). The underlying causes for this termination of exponential growth before the culture reached OD_max are currently unknown. Particularly conspicuous in this regard is the observation that Phe, when provided as a single substrate, allowed for nonstop exponential growth until OD_max, while it was still available and was consumed during linear growth with Mix-11. One may speculate that the biosynthesis of cellular building blocks (e.g., certain amino acids) could represent a bottleneck for sustaining exponential growth. The higher culture densities reached with Phe despite continuous exponential growth and high TDA production suggest that the quorum-sensing activity of P. inhibens (53) per se should not trigger the slowdown of growth with Nag or Mix-11.

(iv) Growth yields and efficiency. We calculated molar growth yields (Y) at OD_max in order to compare the growth efficiencies of P. inhibens during growth with Nag, Phe, and Mix-11 (Fig. 5; also Table S5 in the supplemental material). In general, molar growth yields relate the amount of biomass formed (X) to substrate consumption (e.g., total carbon substrate [$Y_{X{\text{/C}}_{\text{total}}}$], ATP [Y_X/ATP], or O₂ [$Y_{X{\text{/O}}_2}$]). Yields have long been known for their organism- and substrate-specific variations, ranging between 0.01 and 1.0 mol of biomass formed (mol substrate)⁻¹ (54).

With Nag and Phe, we observed similar growth efficiencies for carbon consumption ($Y_{X{\text{/C}}_{\text{total}}}$) of ∼9.2 g cells_dry (mol C_total)⁻¹ (Fig. 5), conversely equaling a total carbon demand of ∼109 mmol (g cells_dry)⁻¹. This suggests that despite the difference in catabolic pathways and uptake systems, these substrates are converted into biomass with similar efficiencies. Interestingly, carbon consumption efficiency drops upon the growth of P. inhibens with Mix-11 (by ∼34%), but the performance with Nag and Phe was slightly lower than that for growth in the same mineral medium with Casamino Acids [10.5 g cells_dry (mol C_total)⁻¹] (47). Overall, the $Y_{X{\text{/C}}_{\text{total}}}$ values obtained in mineral medium were lower than those achieved with the complex, nutrient-rich MB medium [12.3 g cells_dry (mol C_total)⁻¹] (24). In both cases (Casamino Acids and MB medium), a certain overestimation could have resulted from basing the $Y_{X{\text{/C}}_{\text{total}}}$ calculation on dissolved organic carbon, quantified as a proxy for substrate consumption. Nevertheless, the efficiencies determined for P. inhibens are still markedly lower than that observed for an E. coli production strain growing with glucose [15.5 g cells_dry (mol C_total)⁻¹] (55).

The theoretical ATP demands for biomass production (Y_X/ATP) were also similar with the two single substrates (Fig. 5) and were calculated based on the two selected coupling ratios of F₁F_o ATP synthases. With the more-efficient coupling ratio of 3.3 ions per ATP, the formation of biomass theoretically consumes, on average, ∼270 mmol ATP (g cells_dry)⁻¹, whereas with the less-efficient coupling ratio of 4.3 ions per ATP, this would amount to ∼213 mmol ATP (g cells_dry)⁻¹. Irrespective of the coupling ratio, the values for P. inhibens are within the wide range of values reported for bacteria in the literature: ∼30 to 500 mmol ATP (g cells_dry)⁻¹ (30, 56). The ATP demand for the growth of P. inhibens with Mix-11 was also markedly higher (by ∼34%) than that with Nag or Phe, also reflecting the decrease in carbon consumption efficiency mentioned above.

The three substrate conditions differed with respect to the O₂ demand, which was larger with Phe and Mix-11 [$Y_{X{\text{/O}}_2}$, 77.3 and 70.9 mmol O₂ (g cells_dry)⁻¹, respectively] than with Nag (reduced by ∼34%). For Phe, this is explained by additional nonrespiratory O₂ consumption (i) for anabolism (estimated at ∼2.5 mM O₂) due to the higher C/O ratio (analogous to the degree of reduction) in the substrate (4.5) than in the biomass formed (3.2) and (ii) for substrate catabolism due to an O₂-consuming monooxygenase (PaaABCDE) involved in an early step of the catabolic pathway (Fig. 1) (42). We estimate that of the 15 mM Phe supplied, approximately 1 mM Phe is directly channeled into the synthesis of cellular proteins, which should contain a 3.86% share of Phe (57). Therefore, ∼14 mM O₂ could be consumed by PaaABCDE during Phe catabolism. Thus, respiratory O₂ consumption should amount to only ∼70 mM O₂. For Nag, nonrespiratory O₂ consumption was not as relevant, since the oxygen content in the substrate was higher than that in the biomass (C/O ratios, 1.3 and 2.8, respectively), and substrate catabolism does not include an O₂-consuming enzyme. For the growth experiments with Mix-11, these considerations have to be differentiated, since O₂-consuming monooxygenases may increase the O₂ demand when the amino acids Phe and Trp are catabolized. However, the substrate depletion profiles during growth with Mix-11 (Fig. 4) may suggest that the three sugars provided together with Lys, Val, and Leu predominantly sustain biomass formation. One may speculate that the other amino acids (including Phe and Trp) are used mainly for cellular processes, such as the biosynthesis of secondary metabolites, and general anabolism.

The three substrate conditions also differed with respect to the carbon demand for dissimilation ($Y_{X{\text{/C}}_{\text{diss}}}$) (Fig. 5). P. inhibens grown with Phe dissimilated 52.9 mmol carbon (C_diss) per g cells_dry formed, whereas the demand with Nag was ∼15% lower and that with Mix-11 was ∼15% higher. This agrees rather well with the calculation that a higher share of the substrate was assimilated (C_ass) with Nag (∼59%) than with Phe (∼51%) (Fig. 5). Compared to the values at OD_max, C_diss was even lower during exponential growth, amounting, on average, to ∼35% with Nag and ∼46% with Phe. This effect was even more pronounced for Mix-11 (C_diss, ∼30%), suggesting that P. inhibens grows extremely efficiently when provided with substrate mixtures. Thus, despite growing exponentially at comparable rates with these three substrates, P. inhibens had a higher dissimilatory energy demand with Phe than with Nag or Mix-11. This may relate to higher metabolic/maintenance costs (e.g., due to the production/presence of TDA [including the draining of Phe and Cys from the biosynthetic pool] [58] or other cellular building blocks [Fig. 3A]) during growth with Phe.

Comparison of growth performances across an extended substrate range.

The growth responses of P. inhibens to eight further single substrates and Mix-10, as shown by the physiological parameters determined, differed greatly (Fig. 5; also Fig. S3 and Table S5 in the supplemental material). Most cultures transited into the stationary-growth phase when >95% of the substrate initially provided was depleted (except for Lys, with ∼65% remaining, and Mix-10, with ∼36% remaining). Active growth was separated into the exponential and linear growth phases (as observed with Nag), except for Thr, which showed exclusively exponential growth (as observed with Phe). The duration of active growth phases and the growth rates differed widely across individual substrates (Table S5 in the supplemental material). μ_exp was highest with Thr (0.18 h⁻¹, a rate similar to those with Nag and Phe) and lowest with His (0.05 h⁻¹) and Lys (0.06 h⁻¹). Maximal ODs and CDWs were achieved with the substrates allowing for the highest theoretical ATP yields (i.e., Trp, Phe, Nag, Leu, and Man [Fig. 5]). However, maximal ODs and CDWs do not necessarily reflect the efficiency of substrate utilization and growth for P. inhibens. Apparently, substrate conditions giving rise to the lowest production of TDA (Nag, Leu, Man, Thr) displayed the highest growth performances and efficiencies. In fact, carbon assimilation was most efficient for Thr, although Thr provided the lowest catabolic ATP yield among the substrates tested. This is in contrast to those substrate conditions (Phe, Trp, Mix-10, Mix-11) that stand out due to high TDA production (Fig. 5). The recently proposed dissipating effect of TDA on the proton motive force (PMF) (59) apparently leads to the necessity of allocating increased shares of the substrate to energy generation, which, in turn, decreases growth efficiency due to an increased C_diss demand. The latter possibly also reflects the additional metabolic/energetic costs related to TDA biosynthesis itself and to the proposed proton efflux self-defense mechanism (59). In the case of His, which also exhibits a relatively low growth efficiency despite low TDA production, oxidation to 2-oxoglutarate (Fig. 1) may pose an additional energetic burden, since the inorganic nitrogen released intracellularly (estimated at ∼45 mM NH₄⁺) has to be rapidly assimilated into biomass (ATP dependent) in order to reduce the futile cycling of NH₃/NH₄⁺ across the cytoplasmic membrane and its uncoupling effect on the PMF (60).

The carbon assimilation efficiencies observed here for P. inhibens (∼0.28 to 0.61; based on CO₂ measurements) may also relate to high carbon substrate concentrations and controlled growth conditions in bioreactors (including high O₂ provision and a constant pH). In the natural environment, the efficiency of carbon assimilation into biomass (corresponding to bacterial growth efficiency [BGE]) is generally lower. BGE was found to be highly variable (0.01 to 0.69) in the marine ecosystem, with medians of 0.22 in the ocean and 0.34 in estuaries (14). The carbon assimilation efficiencies for P. inhibens correspond to the upper range of environmental BGEs, observed in estuaries and coastal oceans (14). These environments are typically characterized by relatively high nutrient concentrations and constitute the primary habitat of P. inhibens and its close relatives (26, 61). The experimentally determined RQs determined here for amino acids ranged from 0.5 (with Leu) to 1.3 (with His), deviating in most cases markedly from the generally used value of 1.0 for BGE calculations (14), and thus possibly resulted in either an underestimation or an overestimation of the BGE with amino acids as the carbon and energy sources. For sugars, with experimentally determined RQs of 0.9 (Nag), 1.0 (Man), and 1.1 (Glc), this issue is not apparent.

Growth efficiencies and energetics are highly dependent on the type of substrate provided and, remarkably, also on the compositional differences between Mix-10 and Mix-11 (Fig. 5; see also Fig. S1 and Table S5 in the supplemental material). The absence of Ile and the subtly increased concentrations of the individual compounds (by 0.14 mM each) in Mix-10 correlate with marked changes in growth behavior and efficiency. While among the 12 substrate conditions tested, the rates for growth (0.21 h⁻¹) and other biomass-specific parameters (e.g., $q_{{\text{O}}_2\hspace{0.17em}/X}$) were highest with Mix-10, biomass yields dropped drastically, as did the assimilation efficiency (C_ass share, ∼47%). In contrast to cultures growing with Mix-11, those with Mix-10 did not completely deplete the carbon sources provided at OD_max (∼36% remained, including Phe, Trp, His, and Thr), even though the substrate utilization sequence was similar (first sugars, then Lys, Val, etc.). The amino acid consumption profiles in Mix-10 and Mix-11 were reminiscent of those previously observed for P. inhibens grown with Casamino Acids (47).

To resolve the differences in substrate depletion dynamics and preferences between the two mixtures studied, we focused on the time required to deplete 50% of the initial concentration of each substrate. These experimental half-depletion times were compared to those obtained by simulating substrate-specific half-depletion times, assuming that substrate kinetics are the same under single- and mixed-substrate conditions (Fig. 6; also Fig. S1 in the supplemental material). In the case of Man, the experimentally determined time point of half-depletion in both mixtures matched the simulated value (relative half-depletion time, ∼1 [Fig. 6]). For all other substrates, the experimental half-depletion times in the mixtures were at least 50% longer (with relative values ranging from ∼1.5 to ≥4). Noteworthy, the time spans until half-depletion of Trp, Thr, Phe, and His were markedly longer for Mix-10 than for Mix-11, indicating that even subtle compositional differences in the utilizable organic carbon pool can substantially alter the consumption profiles of individual constituents.

Presumptive embedment in microbial ecology concepts.

The study organism P. inhibens displays a multitude of properties qualifying this Roseobacter group member as a copiotroph: fast growth with complex media (24, 36, 47) or single substrates (this study; reference 62), a broad substrate range combined with substrate-specific regulation (25, 29), the intertwining of chemotaxis and quorum sensing (53), and siderophore formation (23). Moreover, the opportunistic nutritional strategy of P. inhibens is reflected by its Janus-faced behavior during association with algae (45, 46) and its prominent association with the coccolithophore microalga Emiliania huxleyi during blooms of the latter (22). Finally, the relatively large genome (4.2 Mbp) carries inducible prophages. While research on the occurrence of phages among Roseobacter group members is only in its early beginnings, other reports indicate their relevance for the genera Phaeobacter (63) and Ruegeria (64), suggesting that the “kill-the-winner” hypothesis could also be relevant for copiotrophic roseobacters. Considering the applicability of the latter for P. inhibens, one may speculate that our reflections on the growth, physiology, and energetics of P. inhibens may also be transferable to other copiotrophic marine bacteria. This may be relevant, in particular, for bacteria that are difficult to cultivate or for in situ studies, where such detailed and quantitative analyses of growth kinetics and yields are technically challenging.

Conclusions.

The present study aimed at assessing energetic efficiency as an obvious leverage for the habitat success of P. inhibens and possibly also the Roseobacter group in general. The cross-species conservation of the catabolic pathways studied should allow transfer of the results for substrate-specific catabolic ATP yields to other roseobacters and potentially other marine copiotrophic bacteria as well. The catabolic ATP yields of unknown or experimentally undefined degradation pathways may be estimated from Gibbs free energy yields, which are easier to retrieve (Fig. 2). The high efficiency of carbon assimilation (corresponding to BGE) in heterotrophic bacterioplankton members such as P. inhibens might contribute to the capacity of the oceans to retain and store CO₂. The measurement of in situ BGE from O₂ consumption may, however, markedly underestimate the actual carbon assimilation efficiencies for certain substrate classes (e.g., aromatic amino acids, algal cell wall-derived 3-phenylpropanoids) that have high degrees of reduction and/or require O₂-consuming catabolic enzymes in their degradation pathways (Fig. 1, Phe). In addition, protective secondary metabolites are produced by many marine bacteria and may decrease the overall BGE by inhibiting the growth of susceptible microorganisms and increasing the dissimilatory energy and carbon demand in the producing strain.

Future studies with diverse bacterioplankton members (oligotrophic and copiotrophic) should (i) determine c ring stoichiometries to specify an organism’s ion-to-ATP coupling ratio, (ii) experimentally reconstruct catabolic pathways/networks as a reliable basis for calculating catabolic ATP yields, (iii) assess the metabolic burden and impact of endogenously produced secondary metabolites on growth efficiencies, and (iv) determine growth yields and efficiencies by means of quantitative physiological studies (batch, chemostat). Ultimately, this could enable a better understanding of the interconnections between nutritional and energetic strategies in heterotrophic marine bacteria.

MATERIALS AND METHODS

Medium and adaptation of P. inhibens to specific organic substrates.

Phaeobacter inhibens DSM 17395 (originally deposited as Phaeobacter gallaeciensis DSM 17395) (61) was obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ; Braunschweig, Germany) and has since been maintained in our laboratory. Each growth experiment was started from a glycerol stock (stored at −80°C) prepared from cultures of P. inhibens DSM 17395 grown in marine broth (MB) medium (24). Stock cultures were revived in MB medium (pH ∼7.4), where residual glycerol was first eliminated by a dilution series (up to 10⁻⁶; in 100-ml Erlenmeyer flasks with 13.5 ml medium each), followed by cultivation in the same medium (50 ml) in 250-ml Erlenmeyer flasks. The cells were then cultivated in a defined mineral medium (pH 7.2) (62) over 3 successive passages in order to adapt the cells to the respective substrate condition and to scale up the culture volume, first to 50 ml and then to 250 ml. The defined mineral medium was supplemented with the respective organic substrate (a single substrate or a mixture) at 15 mM, supplied from sterile stock solutions. The single substrates were d-glucose (Glc), d-mannitol (Man), N-acetyl-d-glucosamine (Nag), l-histidine (His), l-leucine (Leu), l-lysine (Lys), l-phenylalanine (Phe), l-tryptophan (Trp), l-threonine (Thr), and l-valine (Val). Adaptation to Trp required gradual increases in substrate concentration over 3 passages (from 5.0 mM via 7.5 mM to 15 mM). The two defined substrate mixtures (Mix-10 and Mix-11) differed in substrate composition and concentration. Mix-10 contained all 10 single substrates (Glc, Man, Nag, His, Leu, Lys, Phe, Trp, Thr, Val), each supplied at a concentration of 1.50 mM. Mix-11 additionally contained Ile as the 11th substrate, and each substrate was supplied at a concentration of only 1.36 mM. To minimize differences in the physiological states of cultures during substrate adaptation, each new culture was started by adjusting the initial optical density at 600 nm (OD₆₀₀) to 0.2, using an active preculture at the substrate-specific half-maximal optical density (∼½OD_max) as the inoculum. Cultures in Erlenmeyer flasks were incubated on a rotary shaker (100 rpm) at 28°C in the dark, and growth was monitored by measuring the OD₆₀₀ with a UVmini-1240 spectrophotometer (Shimadzu, Duisburg, Germany). The purity of all cultures was confirmed by microscopic examination (Axiostar; Zeiss AG, Göttingen, Germany) and by plating on MB agar. All chemicals were of analytical grade.

Cultivation in process-controlled bioreactors and sampling.

P. inhibens was grown in 2-liter double-jacket glass vessels (Labfors II; Infors HT, Bottmingen, Switzerland) operated at ambient pressure and equipped with sensors for pH (405 DPAS-SC; Mettler-Toledo, Urdorf, Switzerland), temperature (PT-100; Infors HT), dissolved oxygen tension (DOT) (InPro 6800 probe; Mettler-Toledo), and foam control (custom-made). The same active preculture (at ½OD_max) was used to inoculate replicate bioreactors (containing 1.5 liters of medium each), and the inoculum volumes were adjusted to reach similar initial cell densities (OD₆₀₀, 0.02). The defined mineral medium used for bioreactors lacked NaHCO₃, in order to avoid its degassing and thereby falsely high CO₂ values. For each of the 12 substrate conditions tested, four bioreactors were operated in parallel using a control unit (UDDC; Infors HT) to maintain the temperature at 28°C, the pH at 7.2 ± 0.2, and the DOT at ≥35% (adjusted by stirrer speed) and to administer the addition of antifoam A solution (4% [wt/vol]; Sigma-Aldrich, St. Louis, MO, USA). The aeration rate of the bioreactor was manually controlled with a variable-area flow meter (V100–80; Vögtlin Instruments, Aesch, Switzerland) set to 10 liters h⁻¹. Two Rushton turbine stirrers and one marine propeller were installed in the bioreactor for effective dissolution of oxygen and mixing of the culture broth. The pH was corrected automatically with sterile HCl (0.5 M) or NaOH (0.5 M). The in- and outflowing air streams were passed through sterile 0.2-μm-pore-size filters (Midisart 2000; Sartorius, Göttingen, Germany). Cooling of the outflowing air with a condenser (Infors HT) connected to a cryostat (at 10°C; Huber, Offenburg, Germany) reduced the humidity in the exhaust gas. Process data (e.g., temperature, pH, DOT, revolutions per minute) were acquired online (IRIS NT, version 6.0; Infors HT). In- and outflowing air was analyzed online for O₂, CO₂, and Ar using a quadrupole mass spectrometer (GAM200; InProcess Instruments, Bremen, Germany) connected via stainless steel conduits (inner diameter, 2.06 mm) to the gas outlets of each bioreactor. A multiport valve assembly controlled the gas sampling from individual bioreactors at intervals of 10 min. The calibration gas was composed of 19.99% ± 0.40% (vol/vol) O₂, 10.00% ± 0.20% (vol/vol) CO₂, and 70.01% ± 0.60% (vol/vol) Ar (Air Liquide, Krefeld, Germany). Data acquisition and analyses were carried out with Quadstar 32-bit software (InProcess Instruments). The total millimolar concentrations of O₂ consumed and CO₂ produced were calculated from rates of O₂ consumption ($Q_{{\text{O}}_2}$) and CO₂ production ($Q_{\text{CO}}{}_{{}_2}$), as described recently by Zech et al. (24). Calculations were performed with MatLab (version 2017a; MathWorks, Natick, MA, USA).

Samples were retrieved at regular intervals from each culture via sterile sampling ports. Growth was monitored by measuring the OD₆₀₀. The sampled culture broth was centrifuged (28,500 × g, 20 min, 4°C), and the cell-free supernatant was used for photometric detection of TDA formation from the absorbance of the brown TDA-Fe complexes at 398 nm (65). Duplicate aliquots of cell-free culture supernatant were stored immediately at –20°C for quantitation of substrate depletion by HPLC (see below). For determination of the concentration (dry weight) of cells (CDW), culture broth was harvested in duplicate (starting at an OD₆₀₀ of ∼0.3) by centrifugation (11,300 × g, 20 min, 4°C), and cell pellets were washed twice and then resuspended in 300 μl cold 50 mM ammonium acetate prior to transfer into predried and weighed 1.5-ml tubes. CDW was determined by gravimetric analysis of tubes after drying to a constant weight at 60°C.

Quantification of substrate depletion.

Substrate concentrations were quantified in cell-free culture supernatants by (micro)HPLC analysis, employing an UltiMate 3000 Rapid Separation LC system (Thermo Fisher Scientific, Germering, Germany). Samples were diluted in membrane-purified water (if necessary) and were filtered (pore size, 0.2 μm; regenerated cellulose[RC]; Chroma Globe, Kreuzau, Germany).

Nag, Glc, and Man were separated with a Eurokat separation column (length, 300 mm; inside diameter, 8 mm; bead size, 10 μm; Knauer, Berlin, Germany), temperature controlled at 75°C, using 5 mM H₂SO₄ as the eluent, which was administered at a flow rate of 0.8 ml min⁻¹ for Nag or 1.2 ml min⁻¹ for Glc or Man. The amino sugar eluted at a retention time of 9.6 min, Glc at 5.6 min, and Man at 6.2 min. All three sugars were detected with a refractive index (RI) detector (RI-101; Shodex, Munich, Germany). For analyses of substrate mixtures (Mix-10, Mix-11), the temperature was adjusted to 60°C and the flow rate to 0.5 ml min⁻¹, resulting in altered retention times of 13.2 min for Glc, 14.4 min for Man, and 15.4 min for Nag. For all three compounds, the limit of detection was 50 μM and the dynamic range extended to 10 mM.

Trp was separated on a Hypersil Gold C₁₈ column (length, 150 mm; inside diameter, 1 mm; bead size, 3 μm; Thermo Fisher Scientific), temperature controlled at 20°C, using mixtures of 5% (eluent A) and 90% (eluent B) acetonitrile (both acidified to pH <3.0 with 0.01% [wt/vol] H₃PO₄) at a flow rate of 100 μl min⁻¹. Across the gradient applied, eluent B increased from 3% (initially constant for 2.5 min) to 65% (within 4 min) and finally to 99% (within 1 min; then constant for 2.5 min). Trp eluted at a retention time of 4.2 min and was detected by means of a UV detector (DAD-3000; Thermo Fisher Scientific) at 220 nm. The limit of detection was 0.1 μM, and the dynamic range extended to 15 μM.

When used as a single substrate for cultivation, Phe was separated on an Acclaim-120 C₁₈ column (length, 250 mm; inside diameter, 2.1 mm; bead size, 5 μm; Thermo Fisher Scientific), temperature controlled at 25°C, using the same eluents (A and B) as those for Trp at a flow rate of 0.5 ml min⁻¹. Eluent B increased from 0.2% (initially constant for 2 min) to 10% (within 13 min), then to 40% (within 30 s), and finally to 99.8% (within 3 min; then constant for 3 min). Phe eluted at a retention time of 4.6 min and was detected by means of a UV detector (DAD-3000; Thermo Fisher Scientific) at 190 nm. The limit of detection was 0.1 μM, and the dynamic range extended to 15 μM.

Analyses of nonaromatic amino acids required derivatization prior to HPLC measurements, by applying a solution of ortho-phthaldialdehyde containing 1% (vol/vol) 2-mercaptoethanol. The reaction was stopped after 40 s by acidification, applying 5% (vol/vol) acetic acid. The derivatized amino acids were then separated on an Acclaim-120 C₁₈ column at 40°C by delivering an eluent gradient at a flow rate of 0.5 ml min⁻¹. Eluent A consisted of a mixture of 50 mM sodium acetate with 3% (vol/vol) tetrahydrofuran acidified to pH 5.9 with acetic acid, and eluent B was pure methanol. Across the gradient applied, eluent B increased from 29% (initially constant for 2 min) to 62% (within 18 min) and then to 100% (within 2 min; then constant for 4 min). Derivatized amino acids eluted at the following retention times: His at 4.1 min, Thr at 6.9 min, Val at 16.4 min, Phe at 17.5 min (relevant for substrate mixtures only), Ile at 19.6 min, Leu at 20.4 min, and Lys at 23.3 min. Detection was achieved with a fluorescence detector (FLD-3100; Thermo Fisher Scientific). The limit of detection for each amino acid was 0.05 μM, and the dynamic range extended to 2.5 μM.

Substrate consumption rates and molar growth yields were calculated from HPLC-determined substrate depletion profiles by applying the “smoothing spline” curve-fitting method, implemented in MatLab (version 2017a; MathWorks).

Elemental analysis of bacterial cells.

The elemental composition (C, H, N, and S) of dried cells (aliquots of 1 to 5 mg cells_dry each) was determined with a vario EL cube (Elementar Analysensysteme, Hanau, Germany), a process involving high-temperature oxidation (up to 1,800°C), chromatographic separation, and detection by heat conductivity (for details, see reference 24). The O content (expressed as weight percent) of dried cells was calculated by subtracting from 100 wt% the experimentally determined C, H, and N contents (in weight percent) and the estimated ash content (including S) of 12.03 wt% (66).

Calculations and assumptions.

(i) Catabolic ATP yield. The catabolic pathways for five sugars (29) and nine amino acids (25) in P. inhibens were recently reconstructed by means of differential proteogenomics, metabolomics, and enzyme assays. This allowed calculation of the theoretical ATP yield from the catabolism of each substrate based on the net formed reducing equivalents and ATP. In each case, substrate-specific conversion pathways to acetyl-CoA and its terminal oxidation to CO₂ via the TCA cycle were considered (Fig. 1; Table 1; for details, see Table S1 in the supplemental material). Acceptors of released reducing equivalents are NAD(P)⁺, FAD, and oxidized ferredoxin (Fd_ox). One mole of reduced ferredoxin (Fd_red) can be transformed to 0.5 mol NADPH by the enzyme ferredoxin:NADP⁺ oxidoreductase (EC 1.18.1.2; locus tag PGA1_c20790), and 1 mol NADPH can be transformed to 1 mol NADH by a soluble NAD(P) transhydrogenase (EC 1.6.1.3; locus tag PGA1_c31610). Both enzymes were detected previously in cells of P. inhibens grown with the substrates tested (25, 29).

The net formed reducing equivalents were transformed into ATP units based on the general stoichiometry of oxidative phosphorylation, viz., the number of ATPs generated per 2 e^– transferred through the aerobic respiratory chain (35). The relevant features of the respiratory chain are as follows: (i) reducing equivalents are provided by NADH and enzyme-bound FADH₂ to complex I and the quinone pool (Q), respectively; (ii) electron transfer is coupled to the translocation of ∼4 mol H⁺ each at complexes I and IV, and ∼2 mol H⁺ at complex III; and (iii) the reduction of 0.5 mol O₂ at complex IV requires 2 mol e^–. The F₁F_o ATP synthase couples the generation of 3 mol ATP to the inward flow of a variable number of H⁺ ions per catalytic cycle. The number of H⁺ ions is determined by the number of c subunits, constituting the proton-pumping c_n ring in the F_o component of ATP synthase. For the c₁₀ ring in Escherichia coli, the generation of 3 mol ATP is coupled to the electrochemically driven inward flow of 10 H⁺ ions per catalytic cycle, yielding a coupling ratio of 3.3 H⁺ ions per ATP molecule synthesized (34). Combined with redox-driven H⁺ translocation at complexes I, III, and IV (10 H⁺ ions per NADH and 6 H⁺ ions per QH₂, each divided by 3.3), this results in 3.0 mol ATP (mol NADH)⁻¹ and 1.8 mol ATP (mol QH₂)⁻¹. Alternatively, the c₁₃ ring in the ATP synthase of Bacillus pseudofirmus OF4 is characterized by a less-efficient coupling ratio of 4.3 H⁺ ions per ATP synthesized (34), yielding 2.3 mol ATP (mol NADH)⁻¹ and 1.4 mol ATP (mol QH₂)⁻¹.

(ii) Standard Gibbs free energy yield (ΔG⁰ and ΔG⁰′). The standard Gibbs free energy yield (ΔG⁰ and ΔG⁰′) released during the catabolism of the substrates studied to CO₂ (Fig. 1; Table 1) was calculated from Gibbs free energies of formation ($\text{Δ}{G}_{\text{f}}^0$). Substrates in aqueous solution were considered in their dominant protonation state at physiological pH (7.0). The $\text{Δ}{G}_{\text{f}}^0$ values of the carbon substrates and reactants were obtained from the work of Thauer et al. (67) and Mavrovouniotis (68, 69) and, when required, were corrected for group contributions. Further details are compiled in Table S2 in the supplemental material.

(iii) Active growth phases and growth rates. Active growth phases were distinguished in high-resolution growth curves from semilogarithmic plots of OD₆₀₀ as a function of the incubation time. For each growth curve, the exponential growth rate (μ_exp) and, if applicable, the linear growth rate (μ_lin) were calculated from the slope of an exponential or linear regression. For the correlation between OD and CDW, see Fig. S2 in the supplemental material. For the calculation of exponential and linear growth rates, see the equations provided in the supplemental methods in Text S1.

(iv) Calculation of biomass-specific rates. The calculation of biomass-specific (indicated by the subscripted character X) rates for growth (q_X/X [equals μ]), carbon consumption ($q_{{\text{C}}_{\text{total}}\hspace{0.17em}/X}$), TDA production (q_TDA/X), O₂ consumption ($q_{{\text{O}}_2\hspace{0.17em}/X}$), and CO₂ production ($q_{{\text{CO}}_2\hspace{0.17em}/X}$) were modified from the method of Luttmann et al. (70). Further definitions of the nomenclature used can be found in the supplemental methods in Text S1. For parameters measured offline, the alteration of the concentration was calculated by approximation of the differentials [$\dot{C}\text{(}{t}_j\text{)}$, $\text{C}\dot{\text{D}}\text{W(}{t}_j\text{)}$, and $\text{T}\dot{\text{D}}\text{A(}{t}_j\text{)}$], which was realized via the mean difference quotient. Here, I(t_j) represents the concentration of component I at a certain point of time j. Accordingly, $\dot{I}\text{(}{t}_j\text{)}$ equals the alteration of the concentration, as described in equation 1:

$$\dot{I}\text{(}{t}_j\text{) =}\frac{I\text{(}{t}_{j\text{+1}}\text{)}\mathrm{-}\hspace{0.17em}I\text{(}{t}_{j-\text{1}}\text{)}}{t_{j\text{+1}}\hspace{0.17em}\mathrm{-}\hspace{0.17em}{t}_{j-\text{1}}}$$ (1)

Biomass-specific rates were calculated according to equations 2 to 6. The term cells_dry refers to the dry weight (in grams) of cells, and CDW refers to the concentration of cells (in grams [dry weight] per liter). The growth rate was calculated by equation 2:

$$q_{X/X}\text{(}{t}_j{\text{) [h}}^{-}{}^{\text{1}}\text{] =}\frac{\text{C}\dot{\text{D}}\text{W (}{t}_j{\text{) [g liter}}^{-}{}^{\text{1}}{\text{h}}^{-}{}^{\text{1}}\text{]}}{\text{CDW (}{t}_j{\text{) [g liter}}^{-}{}^{\text{1}}\text{]}}$$ (2)

The rate of total carbon (C_total) consumption was calculated by equation 3:

$$q_{{\text{C}}_{\text{total}}/X}\text{(}{t}_j{\text{) [mmol C}}_{\text{total}}\hspace{0.17em}{\text{(g cells}}_{\text{dry}}{\text{)}}^{-}{}^{\text{1}}\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{] =}\frac{\dot{C}\text{(}{t}_j{\text{) [mM C}}_{\text{total}}\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{]}}{\text{CDW(}{t}_j{\text{) [g liter}}^{-}{}^{\text{1}}\text{]}}$$ (3)

The TDA production rate was calculated by equation 4:

$$q_{\text{TDA}/X}\text{(}{t}_j\text{) [}{A}_{\text{398}}{\text{(g cells}}_{\text{dry}}{\text{)}}^{-}{}^{\text{1}}{\text{h}}^{-}{}^{\text{1}}\text{] =}\frac{\text{T}\dot{\text{D}}\text{A (}{t}_j\text{) [}{A}_{\text{398}}\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{]}}{\text{CDW(}{t}_j{\text{) [g liter}}^{-}{}^{\text{1}}\text{] × culture volume (}{t}_j\text{) [liter]}}$$ (4)

The O₂ consumption rate was calculated by equation 5:

$$q_{{\text{O}}_2/X}\text{(}{t}_j{\text{) [mmol O}}_2\hspace{0.17em}{\text{(g cells}}_{\text{dry}}{\text{)}}^{-}{}^{\text{1}}\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{] =}\frac{Q_{{\text{O}}_2}\text{(}{t}_j{\text{) [mM O}}_2\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{]}}{\text{CDW (}{t}_j{\text{) [g liter}}^{-}{}^{\text{1}}\text{]}}$$ (5)

The CO₂ production rate was calculated by equation 6:

$$q_{{\text{CO}}_2/X}\text{(}{t}_j{\text{) [mmol CO}}_2\hspace{0.17em}{\text{(g cells}}_{\text{dry}}{\text{)}}^{-}{}^{\text{1}}\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{] =}\frac{Q_{{\text{CO}}_2}\text{(}{t}_j{\text{) [mM CO}}_2\hspace{0.17em}{\text{h}}^{-}{}^{\text{1}}\text{]}}{\text{CDW (}{t}_j{\text{) [g liter}}^{-}{}^{\text{1}}\text{]}}$$ (6)

(v) Molar growth yields. Molar growth yields (Y) (for details on calculation, see the supplemental methods in Text S1) relate the experimentally formed dry biomass to the total substrate carbon (C_total) consumed ($Y_{X/{\text{C}}_{\text{total}}}$), to O₂ consumed ($Y_{X{\text{/O}}_2}$), or to CO₂ produced ($Y_{X{\text{/CO}}_2}$). The latter equals $Y_{X{\text{/C}}_{\text{diss}}}$, since the CO₂ formed is a direct measure of the dissimilated share of substrate carbon (C_diss). Calculation of the assimilated share of substrate carbon (C_ass) was based on measured CDW (converted to moles per liter using the specific elemental composition of biomass) and considered the stoichiometric relation of substrate conversion into biomass (assimilation equations [see Table S3 in the supplemental material]). The corresponding molar growth yield for carbon assimilation is $Y_{X{\text{/C}}_{\text{ass}}}$. The ATP yield (Y_X/ATP, expressed in grams per cells_dry per mole of ATP) (71) was calculated to estimate the ATP costs for biomass synthesis from experimental $Y_{X{\text{/C}}_{\text{diss}}}$ values divided by the catabolic ATP yields of the different pathways (in moles of ATP per mole of C_diss). The cellular carbon, O₂, and ATP demand were calculated from the reciprocal of the respective growth yield. Considering all experiments, the total carbon consumption (calculated as the sum of C_ass and C_diss) matched well (difference, –13.4% ± 8.5% at OD_max) with the values obtained from HPLC-based analysis of substrate consumption.

(vi) Cell population model. A cell population model (for further details, see the supplemental methods in Text S1) was fitted to match the experimental values obtained for growth with single substrates (see Fig. S5 in the supplemental material). For each substrate, the cell quota Q (substrate content per cell) remained high and nearly constant as long as the substrate was still available for the cells. This agrees with the fact that the inoculated cells (retrieved from precultures at ½OD_max) had sufficient external and internal substrate reserves. For some substrates, e.g., Man and Trp, the simulated complete depletion preceded the complete depletion determined experimentally. Such subtly earlier depletion in the simulation could be caused by too-fast substrate uptake in the model at low concentrations, but this possibility was ruled out by control simulations considering only low substrate concentrations (i.e., <16 mM carbon). These simulations are in good agreement with the experimental data, an essential prerequisite for applying the model to growth with the substrate mixtures (Mix-10 and Mix-11), where the individual substrate concentrations are proportionally lower than those under single-substrate conditions.