Reconstruction and analysis of the genome-scale metabolic model of Lactobacillus casei LC2W

Gene 554 (2015) 140–147

Contents lists available at ScienceDirect

Gene journal homepage: www.elsevier.com/locate/gene

Reconstruction and analysis of the genome-scale metabolic model of Lactobacillus casei LC2W Nan Xu a,1, Jie Liu a,1, Lianzhong Ai b, Liming Liu a,⁎ a b

State Key Laboratory of Food Science and Technology, Jiangnan University, 1800 Lihu Road, Wuxi, Jiangsu 214122, China State Key Laboratory of Dairy Biotechnology, Technology Center of Bright Dairy & Food Co., Ltd., Shanghai 200436, China

a r t i c l e

i n f o

Article history: Received 3 May 2014 Received in revised form 4 October 2014 Accepted 21 October 2014 Available online 23 October 2014 Keywords: Lactobacillus casei Genome-scale metabolic model Lactic acid bacteria

a b s t r a c t Lactobacillus casei LC2W is a recently isolated probiotic lactic acid bacterial strain, which is widely used in the dairy and pharmaceutical industries and in clinical medicine. The first genome-scale metabolic model for L. casei, composed of 846 genes, 969 metabolic reactions, and 785 metabolites, was reconstructed using both manual genome annotation and an automatic SEED model. Then, the iJL846 model was validated by simulating cell growth on 15 reported carbon sources. The iJL846 model explored the metabolism of L. casei on a genome scale: (1) explanation of the genetic codes—metabolic functions of 342 genes were reannotated in this model; (2) characterization of the physiology—10 amino acids and 7 vitamins were identified to be essential nutrients for L. casei LC2W growth; (3) analyses of metabolic pathways—the transport and metabolism of the 17 essential nutrients and exopolysaccharide (EPS) biosynthesis—were performed; (4) exploration of metabolic capacity was conducted—for lactate, the importance of genes in its biosynthetic pathways was evaluated, and the requirements of amino acids were predicted for mixed acid fermentation; for flavor compounds, the effects of oxygen were analyzed, and three new knockout targets were selected for acetoin production; for EPS, 11 types of nutrients in the rich medium and important reactions in the biosynthetic pathway were identified that enhanced EPS production. In conclusion, the iJL846 model serves as a useful tool for understanding and engineering the metabolism of this probiotic strain. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Lactobacillus casei is a type of gram-positive lactic acid bacteria (LAB) with low guanine–cytosine content that can produce lactic acid, diacetyl, acetoin, and polysaccharides (Ai et al., 2008; Branen and Keenan, 1971; Ding and Tan, 2006). Therefore, L. casei is widely used as a starter culture in milk fermentation and as an adjunct culture for intensification and contributes to flavor development in cheeses. Moreover, L. casei is marketed as a probiotic with positive effects on human Abbreviations: ABC, ATP-binding cassette; ATP, adenosine 5′-triphosphate; CoA, coenzyme A; dCMP, 2′-deoxycytidine 5′-monophosphate; DCW, dry cell weight; ECF, energycoupling factor; EPS, exopolysaccharide; EMP, Embden-–Meyerhof-–Parnas; FAD, flavin adenine dinucleotide; FBA, flux balance analysis; FMN, flavin mononucleotide; GSMM, genome-scale metabolic model; LAB, lactic acid bacteria; NAD, nicotinamide adenine dinucleotide; NADH, reduced nicotinamide adenine dinucleotide; NADP, nicotinamide adenine dinucleotide phosphate; NADPH, reduced nicotinamide adenine dinucleotide phosphate; ORF, open reading frame; UTP, uridine 5′-triphosphate; LAB, lactic acid bacteria; RAST, Rapid Annotations Using Subsystems Technology; BLASTP, basic local alignment search tool (for) protein; COBRA, constraints based reconstruction and analysis; KAAS, KEGG Automatic Annotation Server; UDP, uridine 5′-diphosphate; GDP, guanosine 5′-diphosphate; dTDP, deoxythymidine 5′-diphosphate. ⁎ Corresponding author. E-mail address: [email protected] (L. Liu). 1 The two first authors made equal contributions.

http://dx.doi.org/10.1016/j.gene.2014.10.034 0378-1119/© 2014 Elsevier B.V. All rights reserved.

and animal health (Parvez et al., 2006). Because L. casei is safe and amenable to adjuvant effects, it is also used in vaccines and for the delivery of therapeutic agents by appropriate modifications (Kajikawa et al., 2010; Yoon et al., 2012). Considering these valuable applications, identifying new LAB strains from diverse environments has become increasingly important. L. casei LC2W was recently isolated from homemade natural cheese from a farm in China, and has been proven to be a probiotic strain that can produce antihypertensive exopolysaccharides (EPSs) (Ai et al., 2008; Wu, 2011). Physiological studies, meanwhile, have not revealed any significant results that would limit its potential application in food manufacturing and human health care. Genome-scale metabolic models (GSMMs) reveal the metabolic genotype–phenotype relationships of an organism, and have been demonstrated to be an excellent tool for understanding microbial physiology (Liu et al., 2010; Teusink and Smid, 2006). With the help of the entire genome sequence of L. casei LC2W (Chen et al., 2011), genome annotation methods, and relevant information from biochemical databases and literature, the first GSMM of L. casei was built by combining two published experiment-based LAB GSMMs with the AUTOGRAPH algorithm (Notebaart et al., 2006) and its SEED model (Henry et al., 2010). The iJL846 model mainly elucidated the metabolism of the microorganism from aspects of its nutrient requirements, metabolic pathways, and typical metabolite biosynthesis at the genome level. The simulation results

N. Xu et al. / Gene 554 (2015) 140–147

based on the iJL846 model not only globally depict the growth and fermentation phenotypes, but also directly guide wet experiments for L. casei LC2W. Thus, it was hoped that the iJL846 model would be an effective platform for systematically probing the metabolism of L. casei LC2W. 2. Materials and methods 2.1. Model reconstruction Genome information about L. casei LC2W was derived from our previous work (accession number in GenBank CP002616 and CP002617). The chromosome and plasmid sequences were submitted to the Rapid Annotation using Subsystem Technology (RAST) server (Aziz et al., 2008), and then were used for Model SEED. Simultaneously, the gene–protein–reaction relationships were established using the AUTOGRAPH algorithm. A homology search of L. casei LC2W genes against the ORFs in existing LAB metabolic models was implemented using BLASTP with a filter that required an e-value less than 1 × 10−5 and an identity greater than 40% (Tian and Skolnick, 2003). Then, three compartments, the cytoplasm, periplasm, and extracellular compartments, predicted using CELLO (Yu et al., 2006) and PSORTb (Yu et al., 2010), were allocated to the draft model. At that point, some reactions in the draft were repetitive while some reactions were missing or inaccurate, so specific papers on L. casei were used to add or modify reactions. Subsequently, metabolic gaps were identified by the gapAnalysis function and were manually filled as follows. (1) Reaction formulas and reaction directionality were renewed in KEGG and the Metacyc database, respectively. (2) New annotation was performed via BLASTP with UniProtKB/SwissProt. (3) Transport reactions were referred to TCDB (Saier et al., 2009). Furthermore, metabolite information was extracted from CHEBI and PubChem to achieve mass and charge balance of reactions in the L. casei LC2W model. After iterative debugging, the working model was validated using observed phenotypes. 2.2. Biomass equations Biomass synthesis is a diagnostic criterion of the functionality of a model. In this metabolic model of L. casei LC2W, the biomass consisted of seven macromolecules, i.e., proteins (33.5%), RNA (29.7%), DNA (6.3%), lipids (7.9%), teichoic acid (5.9%), peptidoglycan (8.7%), and polysaccharides (8.1%). Details are provided in Supplementary file 1. The energy requirements for this model were assumed to be 1.52 mmol/gDW/h for non-growth-associated maintenance and 41.15 mmol/gDW/h for growth-associated maintenance based on the continuous culture data of L. casei (De Vries et al., 1970). In addition, vitamins that were identified as growth-dependent factors were incorporated into the biomass equation with stoichiometric coefficients of 1E−5 (Rogosa et al., 1961). 2.3. Wet experiments L. casei LC2W were grown at 37 °C under static tube culture on MRS medium. The MRS medium contained (per liter): 20 g glucose, 10 g casitone, 10 g beef extract, 4 g yeast extract, 0.2 g magnesium sulfate, 5 g sodium acetate, 2 g triammonium citrate, 2 g dipotassium hydrogen phosphate, 0.05 g manganese sulfate, and 1 g Tween 80. 15 kinds of carbon sources were individually investigated: glucose, sodium gluconate, fructose, sorbitol, mannose, galactose, lactose, mannitol, trehalose, sucrose, maltose, ribose, salicin, esculin hydrate, and cellobiose. All carbon sources had the same number of carbon atoms as that of 20 g L−1 glucose. The initial pH of MRS medium was adjusted to between 6.0 and 6.2. The seeds were first activated on MRS medium with glucose for 12 h. Then, the strain cultures were washed twice using sterilized saline water, and were inoculated into the specific MRS medium with different carbon sources. Growth was measured by the optical density at 600 nm. The concentrations of the carbon sources were measured using the UltiMate 3000 high-performance liquid chromatography (HPLC) series

141

and an Aminex HPX-87H ion exclusion column (BioRad, CA, USA) at 35 °C. The mobile phase was 5 mM H2SO4 with a flow rate of 0.6 mL min−1. Compounds were detected by a Shodex Differential Refractometer Detector. 2.4. Model simulation and analysis Simulation was constrained by changing the exchange reaction rates or the biomass formation rates based on the literature or empirical evidence such that the in silico results conformed to biological principles. Flux balance analysis was the basic tool (Orth et al., 2010) used for simulating cell growth and metabolite production, and identifying essential growth factors. The in silico rich medium was assumed to have 1 mmol/gDW/h uptake rates of all exchangeable amino acids, nucleotides, and vitamins. The in silico minimal basal medium was determined with the singlerxndeletion function. The 15 reported carbon uptake rates were from wet experimental data, and 11 carbon sources (glucose, fructose, sorbitol, mannose, galactose, lactose, mannitol, trehalose, sucrose, maltose, and cellobiose) were used for quantitative validation of the iJL846 model owing to the lower cell growth or carbon consumption rates using the four others. Essential genes were analyzed with the singlegenedeletion function. Robustness analysis was performed for oxygen uptakes on rich medium with glucose as the carbon source. Candidate genes for the overproduction of acetoin were chosen from nonessential genes in the reduced model with the OptKnock algorithm (Burgard et al., 2003). The objective function was varied to meet the requirements of each simulation: (1) the biomass function was maximized for carbon utilization, and the essential factor analysis was performed; (2) the various objective functions included the exchange reactions for amino acids to explore nutrient absorption with different fermentation patterns; (3) the exchange reaction of acetoin and diacetyl was individually set as the objective function during robustness analysis; (4) for modeling EPS production, a demand reaction representing EPS output was added to the model and was maximized with the constraint of 10% of the maximum specific growth rate (without EPS production) on rich medium. All model simulations were conducted in MATLAB (The MathWorks, Inc.) with the GUN Linear Programming Toolkit (Makhorin, 2001) as the linear programming solver. The COBRA Toolbox (Schellenberger et al., 2011) was used to facilitate the analysis. 3. Results and discussion 3.1. Properties of the genome-scale metabolic model of L. casei LC2W The iJL846 model contained 785 metabolites and 969 reactions that were distributed in 3 cellular compartments (Table 1). The ORF coverage of the iJL846 model is the highest (27.7%) among the three LAB GSMMs. A total of 417 and 321 homologous genes in the iJL846 model were also in the models of Lactobacillus plantarum WCFS1 (Teusink et al., 2006) and Lactococcus lactis MG1363 (Flahaut et al., 2013), respectively. A total of 245 homologous genes were shared by the three Table 1 Components of the models of three lactic acid bacteria. Model contents

L. casei LC2W

L. plantarum WCFS1

L. lactis MG1363

Gene Gene coverage (%) Total reactions Internal reactions Transport reactions Exchange reactions Gene associated reactions Metabolites Unique metabolites Compartment

846 27.7 969 604 227 139 770 785 607 (c,e,p)

721 23.5 643 413 118 112 528 531 554 (c,e)

518 19.9 754 530 119 105 542 650 552 (c,e)

142

N. Xu et al. / Gene 554 (2015) 140–147

models, and these genes might constitute the universal metabolism of LAB (Fig. 1a, mode 1). In contrast, 395 genes in the iJL846 model, which catalyzed 236 cellular reactions in 38 metabolic pathways (Fig. 1a, mode 2), had no homologous genes in those two models. Without regard to compartments, 134 metabolites were only in the iJL846

model, which participated in 180 cellular reactions. Overlaps between the 180 reactions and the above 236 reactions might represent the metabolic characteristics of L. casei. For instance, the transport and degradation of dipeptides (18 genes and 38 reactions) suggest that L. casei was isolated from cheese and helped to ripen cheese. Other reactions

Fig. 1. Distributions of reactions and genes in metabolic subsystems. Panel (a) suggests the metabolic subsystems distribution of reactions. Mode 1: genes participating in those reactions were present in three models, L. casei LC2W, L. plantarum WCFS1, and L. lactis MG1363; mode 2: genes were only in the L. casei LC2W model; (b) CM (carbohydrate metabolism), LM (lipid metabolism), AM (amino acid metabolism), NM (nucleotide metabolism), CV (cofactors and vitamins metabolism), TR (transport reactions), and EM (energy metabolism).

N. Xu et al. / Gene 554 (2015) 140–147

included peptidoglycan biosynthesis (1 gene and 5 reactions), L-ascorbate (11 genes and 3 reactions) and myo-inositol (6 genes and 6 reactions) metabolism, sulfonate transport reactions (2 genes and 7 reactions) and the corresponding FMN-dependent mono-oxygenase (3 genes and 7 reactions), and partial carbohydrate metabolism (29 genes and 10 reactions), such as those reactions using 3-hexulose-6-phosphate synthase, maltose-6′-phosphate glucosidase, and 6-phospho-beta-galactosidase. The proportion of gene-associated reactions in the iJL846 model approached nearly 100%, except for 80% for the transport reactions (Fig. 1b). However, both the reactions and genes for transport were found in high abundance in most metabolic subsystems, demonstrating the robustness of L. casei in response to complex living conditions. Of the transport reactions, 106 reactions were assigned to 43 individual genes, and 121 reactions were catalyzed by 71 isoenzymes or enzyme subunits (partially illustrated in Fig. S1 in Supplementary file 1). In addition, 92 non-gene reactions (not counting exchange reactions) were added to the iJL846 model for modeling or by referring to the literature of L. casei; these reactions included the oxidative decarboxylation of αacetolactate to diacetyl (Branen and Keenan, 1970) and the assembly of biomass macromolecules. 3.2. Carbon source utilization characteristics of L. casei LC2W Cell growth of LAB on carbon sources tends to be strain dependent. L. casei LC2W could in silico grow on 15 types of single carbon sources, which was in good agreement with experimental data (Wu et al., 2009) (Supplementary file 3). Cell growth on 11 of the 15 carbon sources was used for quantitative validation of the iJL846 model. The carbon sources were composed of saccharides, alcohols, and carboxylic acids, i. e., glucose, cellobiose, sucrose, fructose, trehalose, lactose, galactose, mannose, mannitol, maltose, and sorbitol. The in silico specific growth rates were very close to the in vivo values, and were no higher than 6% of the experimental data. The other four carbon sources were used for the qualitative evaluation of the iJL846 model. Furthermore, the utilization of some other carbon sources, which are known to support other strains of L. casei (Cárdenas et al., 1987; Cerning et al., 1994; Gold et al., 1992; Koh et al., 2013; Landete et al., 2013; Monedero et al., 2007; Rodriguez-Diaz et al., 2012; Viana et al., 2000; Yebra et al., 2007), was evaluated by the iJL846 model under both anaerobic and aerobic conditions (Fig. S2 in Supplementary file 2). The specific growth rates on 20 of the 22 carbon sources remained the same for both conditions, which indicated that molecular oxygen did not exert a clearly positive or negative effect on the growth of L. casei LC2W on several common substrates (Hansen, 1938). However, myo-inositol improved the in silico growth under aerobic conditions. This observation indicated that L. casei LC2W could utilize some energy sources that seemed inappropriate for anaerobic growth (Brown and VanDemark, 1968; Strittmattter, 1959). Cellular flux distribution suggested that the catabolism of D-mannitol and D-sorbitol was accompanied by oxygen consumption, but the growth rates were not changed. In this case, oxygen balances on these carbohydrates were generally maintained by the coaction of pyruvate oxidase, glycerol 3-phosphate oxidase, glutathione peroxidase, and glutamate dehydrogenase to convert oxygen to H2O2 and finally to H2O (Condon, 1987). 3.3. Annotation of gene function using the iJL846 model The physiological functions of 342 genes were re-annotated by combining two automated annotation results, i.e., KEGG Automatic Annotation Server (KAAS) (Moriya et al., 2007) and RAST, and the local BLASTP with the genes in the L. plantarum WCFS1 and L. lactis MG1363 models. The updates could be divided into four classes (details are in Supplementary file 2). (1) Functions were assigned to 104 putative uncertain proteins. For example, LC2W_0806 and LC2W_1320 were reassigned to the malate and lactate transporters, and to peptidoglycan synthesis, respectively. Updates enabled this iJL846 model to output the typical

143

products, such as lactic acid and one main cell wall component, peptidoglycan (Billot-Klein et al., 1997; Kandler, 1983). (2) Specific functions were identified for 195 protein families. For example, the ATP-binding cassette (ABC) transporter LC2W_1134 was used for maltose transport, and the sugar transferase LC2W_2135 was responsible for EPS synthesis. (3) New functions were added to the original annotated information of 32 genes. LC2W_1933 was previously annotated as a probable proton-coupled thiamine transporter and was updated to be the transporter of biotin, folate, and thiamine (Henderson et al., 1979). Similarly, LC2W_0257, a putative branched-chain amino acids transporter, was newly defined as the transporter of five different amino acids. (4) The functions of 11 genes in the iJL846 model conflicted with the original annotation. For example, LC2W_2378 or LC2W_2785 might be a dCMP deaminase or glutathione-disulfide reductase in the iJL846 model, but were annotated as transcriptional regulators in UniProtKB. 3.4. Metabolism of essential growth-related factors in L. casei LC2W L. casei LC2W is remarkable in its adaptation to various ecological niches (O and N, 1986), such as dairy production and the human mouth and intestine, and the absorption of essential nutrients from the environment was demonstrated for the iJL846 model. Ten amino acids (L-arginine, L-glutamate, L-isoleucine, L-leucine, L-cysteine, L-phenylalanine, L-serine, L-tryptophan, L-tyrosine, and L-valine) and seven vitamins (biotin, folate, nicotinate, pantothenate, pyridoxal, riboflavin, and thiamine) were identified as essential nutrients for L. casei LC2W under both aerobic and anaerobic conditions. The incomplete biosynthetic pathways of the above 17 nutrients were closely related to its metabolic characteristics (Fig. 2). For instance, gaps existed in three possible approaches to generating L-methionine (individually from L-cysteine, L-aspartate, and hydrogen sulfide); these gaps suggested defects in the sulfur metabolism of L. casei LC2W. The biosynthesis of L-valine from pyruvate was blocked at the node of α-acetolactate, exactly at the intermediate for flavor compounds, such as acetoin and diacetyl. On the in silico minimal basal medium, 109 genes (12.9% of the total genes) were predicted to be essential for cell growth. Of them, 28 and 13 genes were assigned to amino acid and cofactor metabolism, respectively, and were mostly located at the node for converting essential nutrients into some other biomass components. The transport modes of 17 essential metabolites in L. casei LC2W were analyzed using bioinformatics. Folate, biotin, thiamine, riboflavin, and pantothenate were transported by energy-coupling factor (ECF) transporters, which were composed of general components (EcfA1A2T: LC2W_2632 and LC2W_2634 and LC2W_2633) and substrate-specific S components, i.e., FolT (LC2W_2413) for folate, BioY (LC2W_2375) for biotin, ThiW (LC2W_0373) and ThiA (LC2W_1933) for thiamine, RibU (LC2W_1537) for riboflavin, and PanT (LC2W_1968) for pantothenate. L-glutamate and L-phenylalanine might be obtained by ABC transporters. Branched-chain amino acids were transported by an ABC transporter or by a proton symporter. The others were absorbed by facilitated diffusion (details are in Supplementary file 2). The distribution of the essential genes and the cellular flux indicated that 17 essential nutrients were metabolized. Of the 7 essential vitamins, pyridoxal was the only component in the biomass equation of the iJL846 model. Other vitamins were converted into coenzymes, including NAD(H), NADP(H), CoA, FAD(H2), tetrahydrofolate, and thiamine diphosphate, which regulated the global cellular metabolism of L. casei LC2W. Of the 10 essential amino acids, L-arginine was the only biomass component metabolized by the non-integrated ornithine cycle in L. casei LC2W. Five amino acids (L-phenylalanine, L-tryptophan, L-tyrosine, L-leucine, and L-isoleucine) were likely to be metabolized by aminotransferase, and the other 4 amino acids (L-serine, L-glutamate, L-valine, and L-cysteine) were used for the biosynthesis of the non-essential amino acids except for L-alanine. L-glutamate could be a key precursor of 7 non-essential amino acids (Fig. S3 in Supplementary file 1). L-serine and L-glutamate were individually converted into glycine and L-glutamine via a one-step

144

N. Xu et al. / Gene 554 (2015) 140–147

Fig. 2. The biosynthesis, catabolism, and transport of the essential amino acids and vitamins. The essential metabolites are in pink. The arrows in green and red represent the existence and non-existence of reactions, respectively. The light gray arrows indicate multi-step reactions missing in L. casei LC2W. Transport systems, namely, facilitated diffusion, ABC transport, and ECF transport, are shown in three identical shapes. All transport reactions are indicated by green dashed arrows.

reaction, and vice versa, indicating that L-serine and L-glutamate, which are considered to be essential, could be substituted. In addition to proteinogenic amino acids, the essential genes in amino acid metabolism could participate in peptidoglycan biosynthesis and pyrimidine metabolism. 3.5. Production of lactate and flavor compounds by L. casei LC2W Three lactate biosynthetic pathways from glucose may exist in L. casei LC2W based on genome annotation, i. e., the Embden–Meyerhof–Parnas pathway, pentose phosphate pathway, and phospho-pentose-ketolase pathway (Fig. 3). The central carbon metabolism of L. casei LC2W was

simulated in the lactate fermentation process using FBA. Carbon flux as the biomass objective function suggested that lactate formation in L. casei LC2W mostly depended on the EMP pathway. Changing lactate production to the objective function under 10% specific growth rates, the in silico knockouts of L-lactate dehydrogenase, the malate/lactate antiporter, and the malate proton symporter, glyceraldehyde-3phosphate dehydrogenase, and phosphofructokinase led to a lack of lactic acid production. Moreover, the deletion of phosphopyruvate hydratase and fructose-bisphosphate aldolase partially decreased lactate production. In the glucose-limited continuous culture, 1 mol glucose was transformed to 1.5 mol lactate with some by-products: 0.2 mol acetate, 0.13 mol ethanol, and 0.5 mol formate (De Vries et al., 1970). The total

Fig. 3. Flux distributions of the lactate biosynthetic pathway in L. casei LC2W. Green arrows: Embden–Meyerhof–Parnas pathway; blue arrows: pentose phosphate pathway; red arrows: phospho-pentose-ketolase pathway.

N. Xu et al. / Gene 554 (2015) 140–147

carbon recovery reached up to 94.3% for this conversion, and the remaining 5.7% did not appear to sustain the 0.5 h−1 cell growth. The simulation using the iJL846 model evaluated this possibility. Under the constraints of 0.5 h− 1 specific growth rates and exchange rates of 10 mmol/gDW/h glucose, 2 mmol/gDW/h acetate, 1.3 mmol/gDW/h ethanol, and 5 mmol/gDW/h formate, the in silico lactate production was 11.4 mmol/gDW/h, which is lower than the previously reported value of 15 mmol/gDW/h. This indicated that the carbon-limited mixed-acid fermentation required exogenous amino acids to replenish the flux at the pyruvate node. The reduced cost of exogenous amino acids showed that L-serine, L-asparagine, L-aspartate, L-valine, L-glutamate, L-isoleucine, L-leucine, L-glutamine, L-threonine L-cysteine, and glycine were better able to increase lactate production than the other amino acids (details are in Supplementary file 3). The pivotal precursor for producing flavor compounds, α-acetolactate, was produced from pyruvate by acetolactate synthase (LC2W_2015 and LC2W_2016). Acetoin and diacetyl were accordingly transformed from α-acetolactate by acetolactate decarboxylase (LC2W_2014) or a nonenzymatic reaction. Oxygen was a major factor and influenced facultative anaerobic glucose utilizers. The robustness analysis suggested that the production of acetoin and diacetyl was susceptible to oxygen toxicity during the production of yogurt (Condon, 1987). Acetoin could accumulate in L. casei LC2W both aerobically and anaerobically, and its production increased from 0 to 15 mmol/gDW/h with oxygen supply (Fig. 4a). Diacetyl production must depend on oxygen (Fig. 4b) because it was chemically produced by the oxidative decarboxylation of α-acetolactate. When the oxygen uptake changed from 0 to −11.47 mmol/gDW/h, the diacetyl production noticeably increased and could be maintained at maximum values between −11.47 and −14.34 mmol/gDW/h during oxygen uptake. Previous metabolic engineering of acetoin utilized overexpression of citrate permease with citrate as the source of carbon and energy (Diaz-Muniz et al., 2006), and lactate dehydrogenase and pyruvate dehydrogenase were mutated to enhance the flux of acetolactate synthase (Nadal et al., 2009). The in silico deletion of lactate dehydrogenase and pyruvate dehydrogenase in the iJL846 model resulted in 1.5 or 2.3 mmol/gDW/h increments of acetoin production when using the biomass as the objective. Furthermore, 3 new knockout targets for acetoin production, dihydrofolate reductase, methylenetetrahydrofolate dehydrogenase, and glycerone phosphotransferase, were predicted by OptKnock in the reduced models (318 reactions excluding exchange and transport reactions). The deletion of dihydrofolate reductase or methylenetetrahydrofolate dehydrogenase might improve acetoin production (unpublished data). To detect the effects of deleting glycerone phosphotransferase on acetoin, the in silico cellular flux was compared between the wild type and gene-disruption by flux balance analysis. The fluxes of glycolysis and glycerol dehydrogenase were strengthened, which enhanced the availability of pyruvate and NADH. Both pyruvate and NADH were important factors for acetoin, so acetoin was accordingly increased after the in silico deletion of glycerol dehydrogenase. 3.6. Biosynthesis of the characteristic compounds for exopolysaccharide production by L. casei LC2W EPSs, which have been isolated from the skim milk produced by L. casei LC2W, are composed of four monosaccharides, i.e., D-glucose, D-galactose, D-rhamnose, and D-mannose (Ai et al., 2008). These monosaccharides seemed to be released individually by UDP-glucose, UDP-galactose, dTDP-rhamnose, and GDP-mannose (Jensen and Reeves, 2001; Johnson et al., 2005; Tran et al., 2009). The genome annotation indicated that L. casei LC2W could synthesize UDP-glucose, UDP-galactose, and dTDPrhamnose via successive catalytic reactions using the enzymes Pgm, RffH, RfbB, RfbC, RfbD, GtaB, GalE, and GalT (Fig. 5). However, the genes encoding GDP-mannose pyrophosphorylase that transform mannose 1phosphate to GDP-mannose were not found in the genome of L. casei LC2W. In addition, other metabolites and reactions that are directly correlated with GDP-mannose were not found in the iJL846 model. Therefore,

145

Fig. 4. Robustness analysis of oxygen on acetoin and diacetyl production.

the D-mannose fraction of EPS was derived from the medium or from the conversion of extracellular epimelibiose or intracellular D-mannose 6phosphate. Moreover, compared with other LAB strains that can synthesize EPS, the ability of L. casei LC2W to produce EPS should be a stable physiological function because all of the genes required for EPS production are located on the chromosome of LC2W, whereas the corresponding genes of other strains, such as L. casei CG11, are linked to the plasmid (Kojic et al., 1992; Vescovo et al., 1989). The effects of EPS biosynthetic reactions and nutrients in the in silico rich medium on EPS production were evaluated using a singlerxndeletion method. The reactions catalyzed by ManA, Pmm, RffH, RfbB, RfbC, RfbD, GalT, and GalE were indispensable to EPS production, and the Pgi deletion led to a 4% reduction in EPS production. Cytosine, glucose, L-asparagine, L-aspartate, glycine, L-serine, thiamin, nicotinate, folate, pantothenate, and phosphate improved EPS production in L. casei LC2W. The flux profile of the precursors of EPS, such as NADPH, ATP, UTP, fructose 6-phosphate, and glucose 6phosphate, was supplied by the above nutrients, thus demonstrating the basis for improvement in EPS production. Acknowledgment The authors would like to thank Dr Jens Nielsen, Professor at Chalmers University of Technology, for the technological help on the model reconstruction. This work was supported by grant from the Major State Basic Research Development Program of China (973 Program, No. 2013CB733602), the Program for Young Talents in China, the National

146

N. Xu et al. / Gene 554 (2015) 140–147

Fig. 5. Synthesis of exopolysaccharides from glucose by L. casei LC2W.

High Technology Research and Development Program of China (863 Program, No. 2014AA021501), and the Open Project Program of State Key Laboratory of Dairy Biotechnology, Technology Center of Bright Dairy & Food Co., Ltd. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.gene.2014.10.034. References Ai, L.Z., Zhang, H., Guo, B.H., Chen, W., Wu, Z.J., Wu, Y., 2008. Preparation, partial characterization and bioactivity of exopolysaccharides from Lactobacillus casei LC2W. Carbohydr. Polym. 74, 353–357. Aziz, R.K., Bartels, D., Best, A.A., DeJongh, M., Disz, T., Edwards, R.A., Formsma, K., Gerdes, S., Glass, E.M., Kubal, M., Meyer, F., Olsen, G.J., Olson, R., Osterman, A.L., Overbeek, R.A., McNeil, L.K., Paarmann, D., Paczian, T., Parrello, B., Pusch, G.D., Reich, C., Stevens, R., Vassieva, O., Vonstein, V., Wilke, A., Zagnitko, O., 2008. The RAST Server: rapid annotations using subsystems technology. BMC Genomics 9, 75. Billot-Klein, D., Legrand, R., Schoot, B., van Heijenoort, J., Gutmann, L., 1997. Peptidoglycan structure of Lactobacillus casei, a species highly resistant to glycopeptide antibiotics. J. Bacteriol. 179, 6208–6212. Branen, A.L., Keenan, T.W., 1970. Diacetyl reductase of Lactobacillus casei. Can. J. Microbiol. 16, 947–951. Branen, A.L., Keenan, T.W., 1971. Diacetyl and acetoin production by Lactobacillus casei. Appl. Microbiol. 22, 517–521. Brown, J.P., VanDemark, P.J., 1968. Respiration of Lactobacillus casei. Can. J. Microbiol. 14, 829–835. Burgard, A.P., Pharkya, P., Maranas, C.D., 2003. Optknock: a bilevel programming framework for identifying gene knockout strategies for microbial strain optimization. Biotechnol. Bioeng. 84, 647–657. Cárdenas, I., Ledesma, O., Oliver, G., 1987. Utilization of pyruvate in Lactobacillus casei subsp. rhamnosus ATCC 7469. Curr. Microbiol. 15, 259–264. Cerning, J., Renard, C.M., Thibault, J.F., Bouillanne, C., Landon, M., Desmazeaud, M., Topisirovic, L., 1994. Carbon source requirements for exopolysaccharide production by Lactobacillus casei CG11 and partial structure analysis of the polymer. Appl. Environ. Microbiol. 60, 3914–3919. Chen, C., Ai, L., Zhou, F., Wang, L., Zhang, H., Chen, W., Guo, B., 2011. Complete genome sequence of the probiotic bacterium Lactobacillus casei LC2W. J. Bacteriol. 193, 3419–3420.

Condon, S., 1987. Responses of lactic acid bacteria to oxygen. FEMS Microbiol. Lett. 46, 269–280. De Vries, W., Kapteijn, W.M., van der Beek, E.G., Stouthamer, A.H., 1970. Molar growth yields and fermentation balances of Lactobacillus casei L3 in batch cultures and in continuous cultures. J. Gen. Microbiol. 63, 333–345. Diaz-Muniz, I., Banavara, D.S., Budinich, M.F., Rankin, S.A., Dudley, E.G., Steele, J.L., 2006. Lactobacillus casei metabolic potential to utilize citrate as an energy source in ripening cheese: a bioinformatics approach. J. Appl. Microbiol. 101, 872–882. Ding, S.F., Tan, T.W., 2006. L-lactic acid production by lactobacillus casei fermentation using different fed-batch feeding strategies. Process Biochem. 41, 1451–1454. Flahaut, N.A., Wiersma, A., van de Bunt, B., Martens, D.E., Schaap, P.J., Sijtsma, L., Dos Santos, V.A., de Vos, W.M., 2013. Genome-scale metabolic model for Lactococcus lactis MG1363 and its application to the analysis of flavor formation. Appl. Microbiol. Biotechnol. 97, 8729–8739. Gold, R.S., Meagher, M.M., Hutkins, R., Conway, T., 1992. Ethanol tolerance and carbohydrate-metabolism in Lactobacilli. J. Ind. Microbiol. 10, 45–54. Hansen, P.A., 1938. The respiration of the rod-shaped lactic acid bacteria. Zentralbl. Bakteriol. 98, 289–297. Henderson, G.B., Zevely, E.M., Huennekens, F.M., 1979. Mechanism of folate transport in Lactobacillus casei: evidence for a component shared with the thiamine and biotin transport systems. J. Bacteriol. 137, 1308–1314. Henry, C.S., DeJongh, M., Best, A.A., Frybarger, P.M., Linsay, B., Stevens, R.L., 2010. Highthroughput generation, optimization and analysis of genome-scale metabolic models. Nat. Biotechnol. 28, 977–982. Jensen, S.O., Reeves, P.R., 2001. Molecular evolution of the GDP-mannose pathway genes (manB and manC) in Salmonella enterica. Microbiology 147, 599–610. Johnson, M.R., Montero, C.I., Conners, S.B., Shockley, K.R., Bridger, S.L., Kelly, R.M., 2005. Population density-dependent regulation of exopolysaccharide formation in the hyperthermophilic bacterium Thermotoga maritima. Mol. Microbiol. 55, 664–674. Kajikawa, A., Ichikawa, E., Igimi, S., 2010. Development of a highly efficient proteinsecreting system in recombinant Lactobacillus casei. J. Microbiol. Biotechnol. 20, 375–382. Kandler, O., 1983. Carbohydrate metabolism in lactic acid bacteria. Antonie Van Leeuwenhoek 49, 209–224. Koh, J.H., Choi, S.H., Park, S.W., Choi, N.J., Kim, Y., Kim, S.H., 2013. Synbiotic impact of tagatose on viability of Lactobacillus rhamnosus strain GG mediated by the phosphotransferase system (PTS). Food Microbiol. 36, 7–13. Kojic, M., Vujcic, M., Banina, A., Cocconcelli, P., Cerning, J., Topisirovic, L., 1992. Analysis of exopolysaccharide production by Lactobacillus casei CG11, isolated from cheese. Appl. Environ. Microbiol. 58, 4086–4088. Landete, J.M., Ferrer, S., Monedero, V., Zuniga, M., 2013. Malic enzyme and malolactic enzyme pathways are functionally linked but independently regulated in Lactobacillus casei BL23. Appl. Environ. Microbiol. 79, 5509–5518. Liu, L., Agren, R., Bordel, S., Nielsen, J., 2010. Use of genome-scale metabolic models for understanding microbial physiology. FEBS Lett. 584, 2556–2564.

N. Xu et al. / Gene 554 (2015) 140–147 Makhorin, A., 2001. Glpk-gnu linear programming toolkit. http://www.gnu.org/directory/ GNU/glpk.html. Monedero, V., Maze, A., Boel, G., Zuniga, M., Beaufils, S., Hartke, A., Deutscher, J., 2007. The phosphotransferase system of Lactobacillus casei: regulation of carbon metabolism and connection to cold shock response. J. Mol. Microbiol. Biotechnol. 12, 20–32. Moriya, Y., Itoh, M., Okuda, S., Yoshizawa, A.C., Kanehisa, M., 2007. KAAS: an automatic genome annotation and pathway reconstruction server. Nucleic Acids Res. 35, W182–W185. Nadal, I., Rico, J., Perez-Martinez, G., Yebra, M.J., Monedero, V., 2009. Diacetyl and acetoin production from whey permeate using engineered Lactobacillus casei. J. Ind. Microbiol. Biotechnol. 36, 1233–1237. Notebaart, R.A., van Enckevort, F.H., Francke, C., Siezen, R.J., Teusink, B., 2006. Accelerating the reconstruction of genome-scale metabolic networks. BMC Bioinform. 7, 296. K, O., W, N., 1986. Genus Lactobacillus. In: Pha, S., Ns, M., Me, S., Jg, H. (Eds.), Bergey's Manual of Systematic Bacteriology. Williams & Wilkins, Baltimore (MD), pp. 1063–1065. Orth, J.D., Thiele, I., Palsson, B.O., 2010. What is flux balance analysis? Nat. Biotechnol. 28, 245–248. Parvez, S., Malik, K.A., Ah Kang, S., Kim, H.Y., 2006. Probiotics and their fermented food products are beneficial for health. J. Appl. Microbiol. 100, 1171–1185. Rodriguez-Diaz, J., Rubio-del-Campo, A., Yebra, M.J., 2012. Lactobacillus casei ferments the N-acetylglucosamine moiety of fucosyl-alpha-1,3-N-acetylglucosamine and excretes L-fucose. Appl. Environ. Microbiol. 78, 4613–4619. Rogosa, M., Franklin, J.G., Perry, K.D., 1961. Correlation of the vitamin requirements with cultural and biochemical characters of Lactobacillus spp. J. Gen. Microbiol. 25, 473–482. Saier, M.J., Yen, M.R., Noto, K., Tamang, D.G., Elkan, C., 2009. The Transporter Classification Database: recent advances. Nucleic Acids Res. 37, D274–D278. Schellenberger, J., Que, R., Fleming, R.M., Thiele, I., Orth, J.D., Feist, A.M., Zielinski, D.C., Bordbar, A., Lewis, N.E., Rahmanian, S., Kang, J., Hyduke, D.R., Palsson, B.O., 2011. Quantitative prediction of cellular metabolism with constraint-based models: the COBRA Toolbox v2.0. Nat. Protoc. 6, 1290–1307. Strittmattter, C.F., 1959. Electron transport to oxygen in lactobacilli. J. Biol. Chem. 234, 2789–2793.

147

Teusink, B., Smid, E.J., 2006. Modelling strategies for the industrial exploitation of lactic acid bacteria. Nat. Rev. Microbiol. 4, 46–56. Teusink, B., Wiersma, A., Molenaar, D., Francke, C., de Vos, W.M., Siezen, R.J., Smid, E.J., 2006. Analysis of growth of Lactobacillus plantarum WCFS1 on a complex medium using a genome-scale metabolic model. J. Biol. Chem. 281, 40041–40048. Tian, W., Skolnick, J., 2003. How well is enzyme function conserved as a function of pairwise sequence identity? J. Mol. Biol. 333, 863–882. Tran, S.T., Le, D.T., Kim, Y.C., Shin, M., Choi, J.D., 2009. Cloning and characterization of phosphomannose isomerase from Sphingomonas chungbukensis DJ77. BMB Rep. 42, 523–528. Vescovo, M., Scolari, G.L., Bottazzi, V., 1989. Plasmid-encoded ropiness production in Lactobacillus casei ssp. casei. Biotechnol. Lett. 11, 709–712. Viana, R., Monedero, V., Dossonnet, V., Vadeboncoeur, C., Perez-Martinez, G., Deutscher, J., 2000. Enzyme I and HPr from Lactobacillus casei: their role in sugar transport, carbon catabolite repression and inducer exclusion. Mol. Microbiol. 36, 570–584. Wu, Z., 2011. Antihypertensive activity of single cell protein of Lactobacillus casei LC2W. Nat. Prod. Res. Dev. 23, 228–231. Wu, Z., Guo, B. and Ye, J., 2009. Lactobacillus casei LC2W strain and its use in antihypertensive aspect. EP Patent 1,642,963 B1. Yebra, M.J., Zuniga, M., Beaufils, S., Perez-Martinez, G., Deutscher, J., Monedero, V., 2007. Identification of a gene cluster enabling Lactobacillus casei BL23 to utilize myoinositol. Appl. Environ. Microbiol. 73, 3850–3858. Yoon, S.W., Lee, T.Y., Kim, S.J., Lee, I.H., Sung, M.H., Park, J.S., Poo, H., 2012. Oral administration of HPV-16 L2 displayed on Lactobacillus casei induces systematic and mucosal cross-neutralizing effects in Balb/c mice. Vaccine 30, 3286–3294. Yu, C.S., Chen, Y.C., Lu, C.H., Hwang, J.K., 2006. Prediction of protein subcellular localization. Proteins Struct. Funct. Bioinform. 64, 643–651. Yu, N.Y., Wagner, J.R., Laird, M.R., Melli, G., Rey, S., Lo, R., Dao, P., Sahinalp, S.C., Ester, M., Foster, L.J., Brinkman, F.S., 2010. PSORTb 3.0: improved protein subcellular localization prediction with refined localization subcategories and predictive capabilities for all prokaryotes. Bioinformatics 26, 1608–1615.