Metabolic engineering of Lactococcus lactis: the ...

Journal of Biotechnology 98 (2002) 199– 213 www.elsevier.com/locate/jbiotec

Metabolic engineering of Lactococcus lactis: the impact of genomics and metabolic modelling Michiel Kleerebezem *, Ingeborg C. Boels, Masja Nierop Groot, Igor Mierau, Wilbert Sybesma, Jeroen Hugenholtz Department of Fla6our, Nutrition and Natural Ingredients, Wageningen Centre for Food Sciences, NIZO Food Research, P.O. Box 20, 6710 BA Ede, The Netherlands Received 18 July 2001; received in revised form 11 February 2002; accepted 27 March 2002

Abstract Lactic acid bacteria display a relatively simple and well described metabolism where the sugar source is converted mainly to lactic acid. Here we will shortly describe metabolic engineering strategies that led to the efficient re-routing of the lactococcal pyruvate metabolism to end-products other than lactic acid, including diacetyl and alanine. Moreover, we will review current metabolic engineering approaches that aim at increasing the flux through complex biosynthetic pathways, leading to exopolysaccharides and folic acid. Finally, the (future) impact of the developments in the area of genomics and corresponding high-throughput technologies will be discussed. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Lactic acid bacteria; Lactococcus lactis; Metabolic engineering; Pyruvate metabolism; Genomics

1. Introduction Lactic acid bacteria (LAB) are industrially important microbes that are used all over the world in a large variety of industrial food fermentation. Their contribution in these fermentation processes primarily consists of the formation of lactic acid from the available carbon source resulting in a rapid acidification of the food raw material, which is a critical parameter in the preservation of these products. However, besides their lactic acid form* Corresponding author. Tel.: + 31-318-659629; fax: + 31318-650400. E-mail address: [email protected] (M. Kleerebezem).

ing capacity, LAB also contribute to other product characteristics like flavour and texture. Next to their most important application, which is undoubtedly in the dairy industry, LAB are also applied at an industrial scale in the fermentation of other food-raw materials like meat, and vegetables. Lactococcus lactis is by far the most extensively studied lactic acid bacterium, and over the last decades elegant and efficient genetic tools have been developed. These tools are of critical importance in metabolic engineering strategies that aim at inactivation of undesired genes and/or (controlled) overexpression of existing or novel ones. Especially, the nisin controlled expression (NICE)

0168-1656/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 1 6 5 6 ( 0 2 ) 0 0 1 3 2 - 3

200

M. Kleerebezem et al. / Journal of Biotechnology 98 (2002) 199–213

system for controlled heterologous and homologous gene expression in Gram-positive bacteria was proven to be very valuable and has been employed in most of the metabolic engineering strategies discussed here (Lopez de Felipe et al., 1998; Hols et al., 1999a; Hugenholtz et al., 2000; Looijesteijn et al., 1999; Boels et al., 2001). This NICE system is based on the auto-regulatory characteristics of the biosynthesis of the anti-microbial compound nisin that is produced by some L. lactis strains. Further development of the NICE system has led to multiple advanced tools for the use of this system in its original host L. lactis (Kuipers et al., 1998; de Ruyter et al., 1996), but also for the application of the system in a large number of heterologous Gram-positive hosts, including several other LAB (Kleerebezem et al., 1997; Eichenbaum et al., 1998; Pavan et al., 2000; Bryan et al., 2000). The design of rational approaches to metabolic engineering requires a proper understanding of the pathways that are manipulated, their fluxes, control factors, and the genes involved. Knowledge of the pathways traditionally recognised in LAB and the characterisation of new ones has led to an extensive insight in LAB metabolism. Several other features of LAB, like their relatively simple metabolism, limited biosynthetic capacity and limited gene-multiplicity, illustrate their advantages for metabolic engineering purposes compared to other micro-organisms like yeast or fungi. Metabolically engineered LAB can be applied as efficient cell factories for the ‘ex situ’ production of certain metabolites in bio-reactors. These metabolites can be used as food-additives or as ingredients in pharmaceutical products. However, the most promising application of engineered LAB is related to their traditional use as starters in industrial food fermentation, providing LAB with a long history of safe use and well described behavioural characteristics under industrial fermentation and processing conditions. Therefore, LAB can be used for the in situ production of food components during fermentation or even in the gastro-intestinal tract of the consumer. Such components could generate desirable effects on product taste and texture, but also on nutritional value of the fermented product or

other product characteristics. Alternatively, engineered LAB can be used for the in situ removal of undesirable food-components, including components that are regarded as off-flavours in the fermented product or those that can negatively affect consumers’ health. Here we shortly review the successes accomplished in recent years with regard to re-routing of the pyruvate metabolism of L. lactis, but will focus in somewhat more detail on the strategies currently employed to increase the flux through more complex biosynthetic pathways in this organism. Examples that will be used to illustrate the possibilities and limitations of these strategies include the production of exopolysaccharides (EPS) and folic acid. Finally, the impact of whole genome sequence availability and related global analysis technologies on current and future approaches in metabolic engineering of LAB will be discussed.

2. Engineering of the lactococcal pyruvate metabolism Pyruvate is a central metabolic intermediate, it can be converted to various end-products by L. lactis, including acetic acid, ethanol, diacetyl, acetoin and 2,3-butanediol (Fig. 1). However, the main end-product is evidently lactic acid which is formed by the lactate dehydrogenase reaction. The main electron sink under anaerobic conditions is generated by this reaction, leading to homolactic fermentation. Diacetyl is one of the endproducts that can naturally be produced by L. lactis and is an important flavour compound in dairy products generating the typical butter-aroma. The improvement of the production of diacetyl by L. lactis during the fermentation of lactose has been one of the prime targets of metabolic engineering approaches. Multiple strategies have been employed in the past to generate variants that can produce increased amounts of this flavour component (for a review see Kleerebezem et al., 2000). By far the most efficient diacetyl-producing variant was produced by deliberate interference with the redox balance in L. lactis. The intracellular redox bal-


ance (NADH/NAD+ ratio) plays a predominant role in control of the fermentation pattern displayed by L. lactis (Lopez de Felipe et al., 1997; Garrigues et al., 1997; Hols et al., 1999b). Pyruvate reduction by the lactate dehydrogenase enzyme creates a perfect redox balance by the effective transfer of reduction equivalents (NADH) formed during glycolysis to the metabolic end-product (Fig. 1). By so-called cofactor engineering this redox balance was deliberately disturbed by the controlled overexpression of an NADH oxidase that can convert molecular

201

oxygen to water at the expense of NADH, using the NICE system (Lopez de Felipe et al., 1998). Controlled overproduction of this enzyme, in combination with regulation of the aeration level allowed effective control of the NADH/NAD+ balance. Upon high level expression of the NADH oxidase in combination with efficient aeration, the metabolism of these L. lactis variants could almost completely be shifted from homolactic to mixed-acid fermentation where acetic acid and acetoin (and small amounts of diacetyl) were the main end-products (Lopez de Felipe et al.,

Fig. 1. Schematic representation of the pyruvate metabolism in L. lactis. NADH consuming (bold arrows) and producing (dotted arrows) and redox-neutral (normal arrows) conversions are indicated. The genes encoding the enzymes involved in individual conversions are indicated using abbreviations based on established genetic nomenclature. The heterologously introduced alanine dehydrogenase catalysed conversion (see text for further explanation) is added (top-left) and the conversion of acetolactate to diacetyl does not require an enzyme, but involves a chemical conversion (oxidative decarboxylation, ODC). Moreover, the reaction catalysed by the water-forming NADH-oxidase (see text for further explanation) is indicated. The metabolites that have been the target of metabolic engineering strategies (for more detail see text) are boxed, while the names of the genes that were targeted by genetic manipulations during these metabolic engineering procedures are underlined (for more detail see the text).

202


1998). The introduction of this system in an aacetolactate decarboxylase deficient strain (aldB mutant) led to the efficient re-routing of the pyruvate metabolism towards a-acetolactate and diacetyl. These variants could convert up to 70% of the available pyruvate pool to the butter-aroma compound and its precursor (Hugenholtz et al., 2000). These results clearly show that efficient re-routing can be achieved by affecting the availability of co-factors that exert control on the carbon metabolism. Moreover, the stringent control of redox balance on the global carbon metabolism of L. lactis could probably be exploited for the production of other interesting metabolites like acetaldehyde (yoghurt aroma) or acetic acid. Interestingly, this co-factor engineering approach could readily be applied in other LAB since the expression system used (NICE) can be transferred to other bacteria using a simple, easy to handle two-plasmid system (Kleerebezem et al., 1997). Whereas, the example of pyruvate re-routing described above provides an efficient system for the production of a natural lactococcal endproduct, the following example illustrates that introduction of an heterologous pathway in these homolactic microbes can also lead to extremely efficient cell factories. The alanine dehydrogenase enzyme, encoded by the alaD gene derived from Bacillus sphearicus, was overproduced in L. lactis, using the NICE system (Hols et al., 1999a). This enzyme converts pyruvate to L-alanine in the presence of ammonium (Fig. 1). Nisin-mediated induction resulted in high expression levels of functional alanine dehydrogenase, which accounted for approximately 30– 40% of total soluble proteins under maximum induction conditions. In wild-type lactococcal cells this already resulted in a 30– 40% re-routing efficiency from lactic acid towards alanine, when cells were incubated in the presence of an excess of ammonium. Upon introduction of this system in lactate dehydrogenase deficient lactococcal cells, a complete conversion of the pyruvate pool to alanine was obtained when cells were incubated under the appropriate conditions (Hols et al., 1999a). The alanine produced by these cells consisted of a mixture of both stereo-isomers (D- and L-alanine)

as a consequence of the endogenous alanine racemase activity of L. lactis (encoded by the alr gene). To obtain a stereo-specific L-alanine producing bio-reactor, the alr gene was functionally disrupted in the lactate dehydrogenase deficient lactococcal strain. High level expression of alanine dehydrogenase in the resulting cells indeed led to stereo-specific L-alanine production as the only end-product of fermentation (Hols et al., 1999a). In analogy with the co-factor engineering strategy described above, the alanine producing capacity can easily be transferred to other LAB using the two-plasmid NICE-transfer system (Kleerebezem et al., 1997). The homo-L-alanine fermentation system developed appears highly suitable for the production of this amino acid in bio-reactors by the quantitative conversion of a chosen carbon source (i.e. sucrose, whey, starch). Alternatively, alanine producing LAB could be used for the in situ production of this natural sweetener in fermented dairy products like buttermilk, yoghurt or cheese. In conclusion, the examples described above clearly show that efficient re-routing of the pyruvate metabolism can be achieved by metabolic engineering, resulting in high level production of both natural and novel end-products in L. lactis. Recently, a comprehensive model of the lactococcal pyruvate metabolism was developed that can be used for the design of efficient and successful metabolic engineering strategies. This model has large predictive possibilities and has been made available on internet (Hoefnagel et al., 2002; www.biochem.ac.za/wcfs.html). It is a complete kinetic model composed of all mathematical equations expressing the enzymatic conversion reactions involved in puruvate metabolism. The kinetic parameters used in the different equations are as published in the literature or as established experimentally. The model is able to calculate, for a given combination of enzymatic parameters and environmental conditions, what type and quantity of metabolites will be formed. The model has been validated with fermentation data of wildtype cells and data of several mutants such as L. lactis strains lacking lactate dehydrogenase or overexpressing NADH oxidase. The model can accurately calculate which factor in the overall


203

Fig. 2. Schematic representation of the L. lactis NIZO B40 eps gene cluster and its flanking regions. The predicted individual roles of the eps genes in repeating unit biosynthesis (schematically indicated above the genecluster) and subsequent export, polymerisation and chain length determination are indicated; the role of epsD, epsE, epsF, and epsG have biochemically been established.

metabolism has most control on production of a certain metabolite, through the theory of metabolic control analysis. As such it can be used for optimisation of fermentation processes and for predicting results of changes in enzyme activities or environmental conditions. These type of kinetic, metabolic, models are still quite unique and have their predictive possibilities as great advantage over the more common flux analysis models as published for instance for yeast and also LAB. This notion is clearly illustrated by the fact that an early version of this pyruvate model was used to identify NADH-oxidase overexpression rather than acetolactate synthase overexpression as the most effective way to increase diacetyl production by L. lactis, which eventually was proven to be accurate (see above). The pyruvate metabolism model is currently being expanded by including the glycolytic conversions and the branching pathways leading to interesting metabolic end-products like EPS (see below). The availability of this predictive model in combination with the elegant genetic tools developed for L. lactis, will allow efficient design and practical evaluation of novel metabolic engineering strategies aiming at the efficient production of interesting metabolites in this organism.

3. Complex pathway engineering: exopolysaccharide production in L. lactis The EPS molecules produced by different LAB display a wide variety of chemical and biophysi-

cal properties, including variations in sugar composition, sugar linkage, polymer length, and polymer branching. These polymers have an important role in the rheology and texture properties of fermented food products, and thus are of interest for food applications as in situ produced, natural bio-thickeners. Moreover, they could also be applied as natural, food-grade additives, replacing presently applied stabilisers and thickeners that are produced by non food-grade bacteria (Sutherland, 1998; Becker et al., 1998). Interestingly, it has been suggested that some polysaccharides produced by LAB have prebiotic (Gibson and Roberfroid, 1995), immunostimulatory (Hosono et al., 1997), antitumoral (Kitazawa et al., 1991) or cholesterol-lowering activity (Nakajima et al., 1992a). Over the last decade, the production of EPS by LAB has been extensively studied (for recent review see de Vuyst and Degeest, 1999), generating important advances in our knowledge of EPS-genetics, biological diversity and distribution, biosynthetic pathways, metabolic models, and physics. This knowledge can now be applied to rationally design metabolic engineering studies to modify EPS production and composition. The production of EPS by L. lactis is associated with strains that are isolated from highly viscous fermented milk products. The specific eps genes are encoded on large plasmids that can be conjugally transferred from one lactococcal strain to the next, thereby introducing the EPS-producing capacity in the recipient strain (van Kranenburg et al., 1997; Von Wright and Tynkkynen, 1987;

204


Vedamuthu and Neville, 1986). One of the eps gene clusters of L. lactis that is characterised in most detail, is that of strain NIZO B40 (for a review see Kleerebezem et al., 1999). This strain produces a polymer with a regular repeating unit, 4)[a- L -Rhap-(1 2)][a- D -Galp-1-PO4 -3] -b- D -Galp-(1 4)-b- D -Glcp-(1 4)-b- D -Glcp(1 , that is structurally identical to that produced by strain SBT 0495 (van Kranenburg et al., 1997; Nakajima et al., 1992b). The NIZO B40 strain harbours a 42,180 bp EPS-plasmid pNZ4000 (van Kranenburg et al., 2000), containing the 12 kb NIZO B40 eps gene cluster with 14 co-ordinately expressed genes, epsRXABCDEFGHIJKL (Fig. 2). Based on sequence comparisons putative functions could be assigned to several eps genes, predicting their involvement in biosynthesis of the repeating unit oligosaccharides by the sequential addition of sugars to a membrane-anchored lipid carrier, and subsequent export and polymerisation of these lipid-linked oligosaccharides (Fig. 2). Functional analysis of the epsDEFG genes, encoding glycosyltransferases, generated a model for the biosynthesis of the B40-EPS repeating unit (van Kranenburg et al., 1999a) (Fig. 2). Current studies on L. lactis NIZO B40 eps genetics focus on the elucidation of the individual role of the genes that are predicted to be involved in export (epsK), polymerisation (epsI) and chain length determination (epsA, epsB) (van Kranenburg et al., 1997) (Fig. 2). These studies might eventually allow directed manipulation of these functions, which should result in the production of polysaccharides that have shorter (single repeating unit oligosaccharides?) or longer chain lengths compared to the native B40-polymer. Moreover, the strong increase in the available eps gene cluster sequences from various organisms, including several LAB, has generated a wealth in genetic information including a large number of glycosyltransferase encoding genes. The genetic information combined with biochemical information regarding donor- and acceptor-specificity of individual glycosyltransferase enzymes, could be exploited in a glycosyltransferase combinatorial approach aiming at con-

struction of the biosynthetic machinery for the production of oligo- and/or poly-saccharides in L. lactis that have a predetermined and novel structure. The latter approach requires that the export and polymerisation machinery do not display a unique specificity for its corresponding, native polysaccharide. Interestingly in this respect is the observation that heterologous expression of the Streptococcus thermophilus Sfi6 eps-genecluster in L. lactis resulted in the production of a polysaccharide containing an altered repeating unit compared to the polymer produced in its original host (Stingele et al., 1999). These results show that neither the glycosyltransferases, nor the export and polymerisation machinery encoded by this eps gene clusters exhibit exclusive specificity for their native substrate and product, suggesting that the approach above could be successful. Moreover, a first example of such an approach is the possibility to functionally exchange glycosyltransferases with identical substrate specificity originating from different genera in Gram-positive bacteria (van Kranenburg et al., 1999b). Current and future eps engineering strategies should clarify the possibilities of these approaches towards the production of ‘designer’-polysaccharides. In addition to the specific, plasmid-encoded eps genes, EPS production in L. lactis also requires the chromosomally encoded genes involved in sugar nucleotide biosynthesis (Fig. 3). Several studies suggest that the availability of these primary EPS-building blocks could function as an important control factor in EPS biosynthesis (Gilbert et al., 2000; Looijesteijn et al., 1999). This notion suggests that increasing the flux towards these intermediates could result in an increase of the EPS production level. Therefore, to be able to eventually use metabolic engineering as a means to increase EPS production by LAB it is important to evaluate the physiology of nucleotide sugar biosynthesis in relation to EPS production levels. A possible key step in the control of sugar nucleotide biosynthesis is the interconversion of the glycolysis intermediate glucose-6P into the sugar nucleotide precursor glucose-1P, catalysed by phosphoglucomutase (Ramos etal., 2001; de


Vos, 1996) (Fig. 3). To investigate this possibility,the Escherichia coli a-pgm gene was expressed at high levels in L. lactis using the NICE system. This modulation of the phosphoglucomutase enzyme level resulted in a strong increase of the intracellular levels of UDP-glucose and UDP-galactose, illus-

205

trating the postulated role of the phosphoglucomutase enzyme in controlling the sugar-nucleotide levels. However, although these intermediate levels could clearly be increased, the B40-EPS production level was not affected by this modification (Boels et al., unpublished results). Similar results were

Fig. 3. Sugar nucleotide metabolism. Metabolic pathways leading to the formation of the sugar nucleotides required for NIZO B40 EPS biosynthesis. Glucose-1P is a central intermediate that is formed from glycolytic intermediate glucose-6P by phosphoglucomutase (Pgm). Glucose-1P is converted into dTDP-rhamnose by the sequential activities of glucose-1P thymidylyltransferase (RfbA), dTDP-glucose-4,6-dehydratase (RfbB), dTDP-4-keto-6-deoxy-D-glucose-3,5-epimerase (RfbC), and dTDP-4-keto-L -rhamnose reductase (RfbD). UDP-glucose pyrophosphorylase (GalU) activity catalyses the conversion of glucose-1P into UDP-glucose, which subsequently is converted into UDP-galactose, by UDP-galactose epimerase (GalE).

206


Fig. 4. Chemical structure of folate, and its derivatives. The three main building blocks of the molecule and the chemical variations that are possible at positions R1, R2, and X are indicated: the R1 positions can be substituted with C1-groups, the R2 position indicates the mono-glutamyl to poly-glutamyl ‘tail’ position, and the double bonds in the boxed region marker with X can be hydrogenated leading to dihydro- and tetrahydro-folate molecules.

obtained for the overproduction of phosphoglucomutase in S. thermophilus (Levander and Ra˚ dstro¨ m, 2001). The central intermediate, glucose-1P, has to be converted to UDP-glucose, UDP-galactose and dTDP-rhamnose to fulfil the precursor requirements for B40 EPS biosynthesis. All lactococcal genes that encode enzymes that are putatively involved in sugar nucleotide biosynthesis from glucose-1P (galU, galE, rfbABCD) have been cloned and characterised (Boels et al., 2001; Grossiord et al., 1998a; Boels et al., unpublished results), generating the genetic tools to target controlling points in EPS biosynthesis (Fig. 3). Recently, it was shown that the heterologous production of Streptococcus pneumoniae type 3 polysaccharides (glucose and glucuronic acid containing polysaccharide, encoded by the cps3D and cps3S cluster) in L. lactis could be significantly increased by the simultaneous expression of a GalU analogue encoded by the pneumococcal cps3U gene (Gilbert et al., 2000). These results indicate that the endogenous UDP-glucose level in L. lactis is insufficient to support maximum S. pneumoniae type 3 polysaccharide production. To evaluate the effects of GalU activity modulations on sugar nucleotide biosynthesis and B40-EPS production levels, the lactococcal GalU produc-

tion level was increased up to 20-fold the wildtype levels, using the NICE system (Fig. 3). As a consequence, intracellular levels of UDP-glucose and UDP-galactose were shown to be strongly increased, supporting the notion that the GalU enzyme activity indeed controls the production of these sugar nucleotides in wild-type cells (Boels et al., 2001). However, this increased precursor availability did not lead to higher B40-EPS production levels. Another essential enzyme in sugar nucleotide and EPS biosynthesis is the Leloir enzyme GalE (Fig. 3). Evaluation of nucleotide sugar levels and EPS production in a L. lactis galE mutant strain grown on glucose as a sole carbon source revealed undetectable levels of UDP-galactose and consequently a complete abolishment of EPS production. However, EPS production could be recovered by the addition of small amounts of galactose to the medium (Boels et al., 2001; Grossiord, 1998b). The lack of significant effect of increased UDP-glucose and UDPgalactose availability on B40-EPS production levels could be the result of the unchanged availability of dTDP-rhamnose (Fig. 3), which is the precursor for one of the side-chain moieties of B40-EPS and could exert a high level of control on the final EPS production level. Data support-


ing the importance of the availability of dTDPrhamnose for the biosynthesis of rhamnose-containing polymers were reported for rfb mutants of S. pneumoniae (Morona et al., 1997) and Streptococcus mutans (Tsukioka et al., 1997). Recently, the L. lactis rfbACBD gene cluster was cloned (Boels et al., 1999) and the influence of rfb overexpression and rfb deficiency on B40-EPS biosynthesis are currently under investigation (Boels et al., unpublished results). In conclusion, we have successfully engineered the intracellular levels of the sugar-nucleotides required for B40-EPS biosynthesis in L. lactis. However, we have not been able to show that increasing the intracellular nucleotide sugar pools has a significant effect on the B40-EPS production level. This lack of effect could be specific for B40-EPS, as is illustrated by the increased S. pneumoniae type 3 polysaccharide production levels following the increase in GalU enzyme activity level (Gilbert et al., 2000). Clearly, the approaches for successful engineering of the EPS production level in L. lactis might very well require solutions for multiple bottlenecks, illustrating the challenge encountered in complex pathway engineering. For example it might be that the production level of EPS is determined by the activity of the specific EPS biosynthesis machinery rather than by the level of sugar nucleotides. Indicative for such a possibility is the observation that overexpression of the priming glycosyltransferase epsD in L. lactis resulted in an increased B40-EPS production level (van Kranenburg et al., 1999b). In order to validate this possibility, current experiments aim at raising the specific eps-gene expression levels, which could eventually be combined with strategies designed to raise the intracellular sugar nucleotide levels.

4. Complex pathway engineering: folate production in L. lactis Folates (Fig. 4) are essential components in the human diet. They are involved, as cofactor, in many metabolic reactions, including the biosynthesis of nucleotides, the building blocks of DNA and RNA. The daily recommended intake for an adult

207

is 200 mg. For pregnant women a double dose is recommended, since folates are known to prevent neural-tube defect in newly borns (Wald, 1991). Moreover, folates are reported to protect against some forms of cancer (Ames, 1999). In contrast, a low folate level in the diet is associated with high homocysteine levels in the blood and, consequently, with coronary diseases (Boushey et al., 1996; Brattstrom, 1996). Folates are produced in different (green) plants (folium (latin)= leaf) and by some micro-organisms. Therefore, vegetables and dairy products are the main source of folates for humans. Milk is a well-known source of folate. It contains between 20 and 50 mg l − 1 folate and thus contributes significantly to the daily requirement of the average human. Fermented milk products, especially yoghurt, are reported to contain even higher amounts of folate (Alm, 1980). The term ‘folate’ is a non-specific term referring to any folate compound with vitamin activity (Fig. 4). In the general metabolism, folates act as an acceptor or donor of C1-residues. Folate is isolated from natural sources in many different forms, in which the basic structure consists of a pteridine moiety, a p-aminobenzoic acid residue and one or more g-linked L-glutamic acid residues (Fig. 4). However, variation of the C1-residues (including methyl-, methenyl-, methylene-, formimino-, or formyl-) substitutions at either the N-5 or N-10 position, combined with the variable lengths of the poly-g-L-glutamate ‘tail’ (ranging from a single glutamyl residue to more than 10 residues) results in a large variety of chemical variants that are all clustered under the term ‘folate’ (Hamm-Alvarez et al., 1989) (Fig. 4). Finally, the pyrazine rings can be reduced to the so-called dihydro- and tetrahydro-form of folate. The polyglutamyl ‘tail’ is required for the intracellular retention of folate derivatives and plays a role in the binding efficiency of folate and substrates to enzymes involved in one-carbon transfer or inter-conversion of folic acid derivatives (Mc McBurney and Whitmore, 1974). Most mammals are not able to synthesise folates. Plants and more particular leafy vegetables and fruits and fermented dairy products are the main source of folates, but animal products like meat and milk also contain folate. Folates are taken up by mammals, including humans, in the

208


small intestines. Only monoglutamyl-folate derivatives can directly be absorbed in the human gut. Polyglutamyl-folate derivatives with two or more glutamate residues have to be processed in the intestines by a mammalian enzyme deconjugase to release monoglutamyl-folate. The activity of the deconjugase enzymes is susceptible to inhibition by various constituents found in some foods (Bhandari and Gregory, 1990; Seyoum and Selhub, 1998; Rosenberg and Godwin, 1971). The level of monoglutamyl folates, and hence the bioavailability of folate, varies considerably between food products, i.e. egg yolk (72% monoglutamyl folates, MGF), cow liver (56% MGF), orange juice (21% MGF), cabbage (6% MGF), lima beans (5% MGF), and lettuce (less than 1% MGF) (Seyoum and Selhub, 1998).

Compared to vegetable sources, dairy and other fermented products have the advantage that at least part of the folate is stored intracellular in (lactic acid) bacteria. As a result, the intracellularly stored folate may be better protected against the acid-peptic environment of the stomach, leading to a higher effective folate consumption. Many fermented dairy products contain higher folate levels compared to raw milk. The explanation for this is that some of the starter bacteria used in dairy fermentation, execute de novo biosynthesis of folates and secrete a surplus of these compounds into the growth medium. Examples are the yoghurt bacterium S. thermophilus and the cheese and butter(milk) bacterium L. lactis. This property of LAB offers the possibility to fortify fermented dairy products with folate by

Fig. 5. Panel A: Schematic representation of the biosynthetic pathway for folate. Enzymes and corresponding genes (established genetic nomenclature) involved in the individual conversions as well as the chemical nomenclature for all intermediates have been indicated. Panel B: Schematic representation of the L. lactis MG1363 folate gene cluster. Established gene nomenclature has been used to indicate the functions of these genes. Contrary to many other micro-organisms, in L. lactis the enzymes 2-amino-4-hydroxy6-hydroxymethyldihydropteridine pyrophosphokinase and GTP cyclohydrolase I are produced as one, bi-functional protein and are encoded by one gene ( folKE).


Fig. 6. Distribution of extracellular (hatched bars) and intracellular (non-hatched bars) folate of L. lactis expressing HGH (human g-glutamyl hydrolase) (white bars) and a control strain (grey bars). Folate concentrations are determined at 30 and 180 min after induction of expression of HGH. Cells are induced with nisin at OD600 of 0.5. Folate is measured by using a microbiological assay (Sybesma et al., submitted for publication).

natural means, i.e. without the addition of food supplements. The natural diversity amongst dairy starter culture with respect to their capacity to produce folate can be exploited to design new complex starter cultures, which yield fermented dairy products with elevated folate levels. Using this strategy, production of experimental yoghurt containing up to 150 mg l − 1 folate has been reported (Smid et al., 2001). Besides selecting vitamin-producing LAB for inclusion in starter cultures, these LAB can also be supplied as probiotics or health-promoting additives in food or pharmaceutical products. These new applications of LAB will be boosted even more if folate production can be improved through optimisation of fermentation conditions or if LAB-variants (resulting from classical mutagenesis) with improved folate production capacities can be selected. Total folate production in L. lactis strain MG1363 is approximately 100 ng ml − 1. Most of

209

the folate (90%) is accumulated intracellularly, probably due to polyglutamylation of these intracellular folates. Only a small amount (10%) is released into the environment (Sybesma et al., submitted for publication). The folate biosynthesis involves a multi-enzyme pathway, which is schematically presented in Fig. 5A. The genes involved in folate biosynthesis in L. lactis MG1363 have recently been analysed (Fig. 5B). Preliminary results show that several, and maybe all, of the folate biosynthetic genes are transcribed as a polycistronic mRNA (Sybesma et al., 2001). The availability of the folate biosynthesis genes allows efficient genetic engineering aiming at increasing the folate biosynthesis level in L. lactis. Several of the folate biosynthetic genes have been overexpressed individually, or in combination, in L. lactis strain NZ9000 using the NICE system. Preliminary results already show up to 3-fold increased folate production levels and also altered ratios of monoglutamyl-folate over polyglutamylfolate, resulting in higher release of the folate into the environment. The latter modulation of spatial distribution and glutamylation level of the folate produced by L. lactis would lead to improved uptake of this microbially produced vitamin in the human small intestine (see above), thereby generating improved folate bio-availability in fermented products produced with such strains. To further expand these bio-availability studies, L. lactis strains were constructed that express the g-glutamyl hydrolase enzyme derived from rat or human origin (Yao et al., 1996). cDNA clones encoding these enzymes were cloned in the NICE system and introduced in L. lactis and it could be shown that both the human and rat enzymes could be expressed functionally in this organism. The g-glutamyl hydrolase expression in growing L. lactis cultures, resulted in an inversion of folate spatial distribution, i.e. from mainly intracellular (:90%) to extracellular accumulation (Sybesma et al., submitted for publication). In the beginning of the stationary phase, the level of extracellularly produced folate is six times higher in the strain expressing g-glutamyl hydrolase relative to the wild-type (Fig. 6). These results demonstrate that the expression of human g-glutamyl hydrolase in the folate producer L. lactis leads to intracellular

210


deconjugation of polyglutamyl-folate to monoglutamyl-folate resulting in decreased retention of folate in the cell and, consequently, increased excretion of folate to the environment. It can be concluded that expression of heterologous g-glutamyl hydrolase in L. lactis increases the extracellular monoglutamyl folate level, i.e. the level of bio-available folate, thereby relieving the requirement for the intestinal hydrolase activity (Sybesma et al., submitted for publication). This work represents an important first step in the development of novel, functional foods with increased levels of bio-available folate. The identification of g-glutamyl hydrolases from (foodgrade) bacterial origin may enhance the market introduction of a food-grade micro-organism that has improved bio-available-folate production capacities. Moreover, metabolic engineering of the lactococcal folate biosynthesis pathway itself by changing expression levels of the relevant biosynthetic enzymes, is expected to result in significant increases in folate production. These approaches could eventually be combined to generate starter bacteria that can be used for the production of novel fermented dairy products providing 100% instead of the current 15– 20% of the human daily requirement for folate intake.

5. Impact of genomics and related technologies Over the last years, the overall amount of information on microbial genetics has increased dramatically due to the large number of genome sequencing projects developed world-wide. The current list of publicly available complete genome sequences already exceeds 40 microbes and is exponentially growing. Although several related micro-organisms were subjected to whole genome sequencing (including Bacillus subtilis, Bacillus halodurans, Streptococcus pyogenes, etc.), it has taken till the year 2001 for the first LAB genome to be published (Bolotin et al., 2001). These efforts have generated a tremendous amount of information on the physiology, evolution, and many other aspects of microbial species (Wren, 2000; Eisen, 2000; Fraser et al., 2000a,b; Nierman et al., 2000a,b; Weinstock et al., 2000). Genome

sequences supply the ‘blueprint’ of an organism, including all metabolic, regulatory and signalling pathways. However, these data can only be interpreted as a prediction of functionality. Moreover, in general approximately 30– 40% of all predicted genes fulfil an unknown function (Fraser et al., 2000b), indicating that many unforeseen functionalities might be present in these microbes. The challenge is to exploit this wealth of information to extract innovative and useful knowledge concerning the specific behaviour of these organisms in their respective biological niches. In terms of metabolic engineering, this means that although a genome sequence might allow the reconstruction of the predicted metabolic network in a certain microbe, it does not generate information concerning the activity of individual pathways under a chosen condition. In order to be able to predict the metabolic output of cells it is of critical importance to build in silico models containing collections of cellular metabolic constituents, their interactions and their integrated function as a whole (Covert et al., 2001). The kinetic-type model as has been developed for the pyruvate metabolism of L. lactis can only be used for simulation and prediction of relatively simple and, preferably, isolated metabolic pathways. For more complex, multi-enzyme pathways, which may include many branching points, kinetic modelling become virtually impossible for several reasons. In the first place the network of mathematical equations comprising a kinetic model of a complicated pathway can simply not be handled anymore by even an advanced computer system, since it would need too much computing power. Secondly, the measurements required for validation of the model would be too time-consuming, especially, in the case of analysis of intermediate metabolites. Thirdly, many of the kinetic parameters of enzymes involved in complicated metabolic pathways cannot be determined due to lack of enzyme assays or due to low activities. The latter is especially true for most biosynthetic pathways. The metabolic flux through these pathways is in most cases many orders of magnitude lower compared with the flux through catabolic pathways. Moreover, the enzymes involved generally have much lower activities and the metabolic interme-


diates and end-products are produced at much lower levels. In some cases this general problem can be solved by dissection of complicated (biosynthetic) pathways into several, simplified units and modelling each unit separately. However, in many cases this is just not possible. Even simple flux analysis can be an impossible task in metabolic networks involving loops and cycles. To analyse these networks, more advanced measurement techniques have been used successfully, such as the application of 13C-NMR with metabolic substrates labelled at a specific position. Variation in the level of label-scrambling resulting from either enantiomer-specific or enantiomer-unspecific enzymes, will generate insight in control and distribution of different fluxes within a metabolic network (Klapa et al., 1999; Park et al., 1999; De Graaf et al., 2000; Schmidt et al., 1999; Rollin et al., 1995). However, it cannot provide insight in metabolic regulation, nor can it be used to predict fluxes under conditions that have not been tested. The very optimistic idea that the knowledge of a (microbial) genome will automatically provide us with a comprehensive model of overall metabolism seems quite unrealistic for the reasons discussed above. Extensive experimentation on the physiology, enzymology and molecular biology of the organism is required before any practical modelling can be performed. The rapid development of genomics based highthroughput technologies, including transcriptomics, proteomics and especially metabolomics, might allow the effective analysis of the differential, global responses in micro-organisms upon changing environmental conditions. Most examples of effective metabolic engineering in LAB discussed above deal with re-routing the major metabolic flux in these bacteria, i.e. the conversion of sugar into lactic acid. The results obtained include high production of metabolites that otherwise are not produced or produced in low amounts. For reasons discussed above, the engineering of more complex (biosynthetic) pathways and the analysis of the results of these engineering strategies is a much more complicated affair. Nevertheless, some promising results are already available for L. lactis in the case of folate biosynthesis. Future analyses of this and other complex pathways will provide us with valuable knowledge

211

concerning the potential of LAB as cell factories for biologically active compounds like vitamins, antioxidants or other nutraceuticals. Acknowledgements Part of the work described here, was supported by funding programs of the European Community (contracts: ERBFMBICT-950438, BIO4-CT980118, and QLK1-CT-2000-01376). References Alm, L., 1980. Effect of fermentation on B-vitamin content of milk in Sweden. J. Dairy Sci. 65, 353 – 359. Ames, B.N., 1999. Micronutrient deficiencies cause DNA damage and cancer. Food Sci. Agric. Chem. 1, 1 – 15. Becker, A., Katzen, F., Pu¨ hler, A., Ielpie, L., 1998. Xanthan gum biosynthesis and application: a biochemical/genetic perspective. Appl. Microbiol. Biotechnol. 50, 145 – 152. Bhandari, S.D., Gregory, J.F., 1990. Inhibition by selected food components of human and porcine intestinal pteroylpolyglutamate hydrolase activity. Am. J. Clin. Nutr. 51, 87 – 94. Boels, I.C., Kleerebezem, M., Hugenholtz, J., de Vos, W.M., 1999. Metabolic engineering of exopolysaccharide production in Lactococcus lactis. In: Book of Abstracts, Sixth Symposiumon Lactic Acid Bacteria, Veldhoven, The Netherlands, September 19 – 23. Boels, I.C., Ramos, A., Kleerebezem, M., de Vos, W.M., 2001. Functional analysis of the Lactococcus lactis galU and galE genes and their impact on sugar nucleotide and exopolysaccharide biosynthesis. Appl. Environ. Microbiol. 67, 3033 – 3040. Bolotin, A., Wincker, P., Mauger, S., Jaillon, O., Malarme, K., Weissenbach, J., Ehrlich, S.D., Sorokin, A., 2001. The complete genome sequence of the lactic acid bacterium Lactococcus lactis ssp. lactis IL1403. Genome Res. 11, 731 – 753. Boushey, C.J., Beresford, A.A., Omenn, G.S., Moltulsky, A.G., 1996. A quantitative assessment of plasma homocysteine as a risk factor for vascular disease. J. Am. Med. Assoc. 274, 1049 – 1057. Brattstrom, L., 1996. Vitamins as homocysteine-lowering agents. J. Nutr. 126 (Suppl. 4), 1276S –1280S. Bryan, E.M., Bae, T., Kleerebezem, M., Dunny, G.M., 2000. Improved vectors for nisin-controlled expression in grampositive bacteria. Plasmid 44, 183 – 190. Covert, M.W., Schilling, C.H., Famili, I., Edwards, J.S., Goryanin, I.I., Selkov, E., Palsson, B.O., 2001. Metabolic modeling of microbial strains in silico. Trends Biochem. Sci. 26, 179 – 186. De Graaf, A.A., Mahle, M., Mollney, M., Weichert, W., Stahmann, P., Sahm, M., 2000. Determination of full 13C

212


isotopomer distributions for metabolic flux analysis using heteronuclear spin echo difference NMR spectroscopy. J. Biotechnol. 77, 25 – 35. de Ruyter, P.G.G.A., Kuipers, O.P., de Vos, W.M., 1996. Controlled gene expression systems for Lactococcus lactis with the food-grade inducer nisin. Appl. Environ. Microbiol. 62, 3662 – 3667. de Vuyst, L., Degeest, B., 1999. Heteropolysaccharides from lactic acid bacteria. FEMS Microbiol. Rev. 21, 153 –177. de Vos, W.M., 1996. Metabolic engineering of sugar catabolism in lactic acid bacteria. Antonie van Leeuwenhoek 70, 223 – 242. Eichenbaum, Z., Federle, M.J., Marra, D., de Vos, W.M., Kuipers, O.P., Kleerebezem, M., Scott, J.R., 1998. Use of the lactococcal nisA promoter to regulate gene expression in Gram-positive bacteria: comparison of induction level and promoter strength. Appl. Environ. Microbiol. 64, 2763 – 2769. Eisen, J.A., 2000. Assessing evolutionary relationships among microbes from whole genome analysis. Curr. Opin. Microbiol. 3, 475 – 480. Fraser, C.M., Eisen, J.A., Fleischmann, R.D., Ketchum, K.A., Peterson, S., 2000a. Comparative genomics and understanding of microbial biology. Emerg. Infect. Dis. 6, 505 –512. Fraser, C.M., Eisen, J.A., Salzberg, S.L., 2000b. Microbial genome sequencing. Nature 406, 799 –803. Garrigues, C., Loubiere, P., Lindley, N.D., Cocaign-Bousquet, M., 1997. Control of the shift from homolactic acid to mixed acid fermentation in Lactococcus lactis: predominant role of NADH/NAD+ ratio. J. Bacteriol. 179, 5282 –5287. Gibson, G.R., Roberfroid, M.B., 1995. Dietary modulation of the human colonic microbiota: introducing the concept of prebiotics. J. Nutr. 124, 1401 –1412. Gilbert, C., Robinson, K., Le Page, R.W.F., Wells, J.M., 2000. Heterologous expression of an immunogenic pneumococcal type 3 capsular polysaccharide in Lactococus lactis. Infect. Immun. 68, 3251 – 3260. Grossiord, B., Vaughan, E.E., Luesink, E., de Vos, W.M., 1998a. Genetics of galactose utilisation via the Leloir pathway in lactic acid bacteria. Lait 78, 77 –84. Grossiord, B., 1998b. Metabolisme du galactose par la voie de ´ cole Leloir l’operon gal de Lactococcus lactis. Thesis, l’E Nationale Supe´ rieure Agronomique de Montpellier, Montpellier, France. Hamm-Alvarez, S., Sancar, A., Rajagopalan, K.V., 1989. Role of enzyme-bound 5,10-methenyltetrahydropteroylpolyglutamate in catalysis by Escherichia coli DNA photolyase. J. Biol. Chem. 264, 9649 – 9656. Hoefnagel, M.H., Starrenburg, M.J., Martens, D.E., Hugenholtz, J., Kleerebezem, M., van Swam, I.I., Bongers, R., Westerhoff, H.V., Snoep, J.L., 2002. Metabolic engineering of lactic acid bacteria, the combined approach: kinetic modelling, metabolic control and experimental analysis. Microbiology, 148, 1003 –1013. Hols, P., Kleerebezem, M., Schank, A.N., Ferain, T., Hugenholtz, J., Delcour, J., de Vos, W.M., 1999a. Conversion of Lactococcus lactis from homolactic to homoalanine fermen-

tation through metabolic engineering. Nat. Biotechnol. 17, 588 – 592. Hols, P., Ramos, A., Hugenholtz, J., Delcour, J., de Vos, W.M., Santos, H., Kleerebezem, M., 1999b. Acetate utilization in Lactococcus lactis deficient in lactate dehydrogenase: a rescue pathway for maintaining redox balance. J. Bacteriol. 181, 5521 – 5526. Hosono, A., Lee, J., Ametani, A., Natsume, M., Hirayama, M., Adachi, T., Kaminogawa, S., 1997. Characterization of a water-soluble polysaccharide fraction with immunopotentiating activity from Bifidobacterium adolescentis M101-4. Biosci. Biotech. Biochem. 61, 312 – 316. Hugenholtz, J., Hols, P., Starrenburg, M.J.C., de Vos, W.M., Kleerebezem, M., 2000. Metabolic engineering of Lactococcus lactis leading to high diacetyl production. Appl. Environ. Microbiol. 66, 4112 – 4114. Kitazawa, H., Toba, T., Itoh, T., Kumano, N., Adachi, S., Yamaguchi, T., 1991. Antitumoral activity of slime-froming, encapsulated Lactococus lactis subsp. cremoris isolated from Scandinavian ropy sour milk, ‘‘viili’’. Animal. Sci. Technol. 62, 277 – 283. Klapa, M.I., Park, S.M., Sinskey, A.J., Stephanopoulos, G., 1999. Metabolite and isotopomer balancing in the analysis of metabolic cycles: theory. Biotechnol. Bioeng. 62, 375 – 391. Kleerebezem, M., van Kranenburg, R., Tuinier, R., Boels, I.C., Zoon, P., Looijesteijn, E., Hugenholtz, J., de Vos, W.M., 1999. Exopolysaccharides produced by Lactococcus lactis: from genetic engineering to improved rheological properties? Antonie van Leeuwenhoek 76, 357 – 365. Kleerebezem, M., Beerthuyzen, M.M., Vaughan, E.E., de Vos, W.M., Kuipers, O.P., 1997. Controlled gene expression systems for lactic acid bacteria: transferable nisin inducible expression cassettes for Lactococcus, Leuconostoc and Lactobacillus spp. Appl. Environ. Microbiol. 63, 4581 – 4584. Kleerebezem, M., Hols, P., Hugenholtz, P., 2000. Lactic acid bacteria as a cell factory: rerouting of carbon metabolism in Lactococcus lactis by metabolic engineering. Enzyme Microb. Technol. 26, 840 – 848. Kuipers, O.P., de Ruyter, P.G.G.A., Kleerebezem, M., de Vos, W.M., 1998. Quorum sensing-controlled gene expression in lactic acid bacteria. J. Biotech. 64, 15 – 21. Levander, F., Ra˚ dstro¨ m, P., 2001. Requirement for phosphoglucomutase in exopolysaccharide biosynthesis in glucose- and lactose-utilizing Streptococcus thermophilus. Appl. Environ. Microbiol. 67, 2734 – 2738. Looijesteijn, P.J., Boels, I.C., Kleerebezem, M., Hugenholtz, J., 1999. Regulation of exopolysaccharide production by Lactococcus lactis subsp. cremoris by the sugar source. Appl. Environ. Microbiol. 65, 5003 – 5008. Lopez de Felipe, F., Starrenburg, M.J.C., Hugenholtz, J., 1997. The role of NADH-oxidation in acetoin and diacetyl production from glucose in Lactococcus lactis. FEMS Microbiol. Lett. 156, 15 – 19. Lopez de Felipe, F., Kleerebezem, M., de Vos, W.M., Hugenholtz, J., 1998. Cofactor engineering: a novel approach to metabolic engineering in Lactococcus lactis by controlled overexpression of NADH oxidase. J. Bacteriol. 180, 3804 – 3808.

M. Kleerebezem et al. / Journal of Biotechnology 98 (2002) 199–213 McBurney, M.W., Whitmore, G.F., 1974. Isolation and biochemical characterization of folate deficient mutants of Chinese hamster cells. Cell 2, 173 –182. Morona, J.K., Morona, R., Paton, J.C., 1997. Characterization of the locus encoding the Streptococcus pneumoniae type 19F capsular polysaccharide biosynthetic pathway. Mol. Microbiol. 23, 751 – 763. Nakajima, H., Suzuki, Y., Kaizu, H., Hirota, T., 1992a. Cholesterol lowering activity of ropy fermented milk. J. Food Sci. 57, 1327 – 1329. Nakajima, H., Hirota, T., Toba, T., Itoh, T., Adachi, S., 1992b. Structure of the extracellular polysaccharide from slimeforming Lactococcus lactis subspecies cremoris SBT0495. Carbohydr. Res. 224, 245 – 253. Nierman, W.C., Eisen, J.A., Fraser, C.M., 2000a. Microbial genome sequencing 2000: new insights into physiology, evolution and expression analysis. Res. Microbiol. 151, 79– 84. Nierman, W.C., Eisen, J.A., Fleischmann, R.D., Fraser, C.M., 2000b. Genome data: what do we learn. Curr. Opin. Struc. Biol. 10, 343 – 348. Park, S.M., Klapa, M.I., Sinskey, A.J., Stephanopoulos, G., 1999. Metabolite and isotopomer balancing in the analysis of metabolic cycles: applications. Biotechnol. Bioeng. 62, 392 – 401. Pavan, S., Hols, P., Delcour, J., Geoffroy, M.C., Grangette, C., Kleerebezem, M., Mercenier, A., 2000. Adaptation of the nisin-controlled expression system in Lactobacillus plantarum: a tool to study in vivo biological effects. Appl. Environ. Microbiol. 66, 4427 –4432. Ramos, A., Boels, I.C., de Vos, W.M., Santos, H., 2001. Relationship between glycolysis and exopolysaccharides biosynthesis in Lactococcus lactis. Appl. Environ. Microbiol. 67, 33 – 41. Rollin, C., Morgant, V., Guyonvarch, A., Guerquin-Kern, J.L., 1995. 13C-NMR studies of Cyanobacterium melassecola metabolic pathways. Eur. J. Biochem. 227, 488 –493. Rosenberg, I.H., Godwin, H.A., 1971. Inhibition of intestinal gamma-glutamyl carboxypeptidase by yeast nucleic acid: an explanation of variability in utilization of dietary poly-glutamyl folate. J. Clin. Invest. 50, 78a. Schmidt, K., Nielsen, J., Villadsen, J., 1999. Quantitative analysis of metabolic fluxes in Escherichia coli, using two-dimensional NMR spectroscopy and complete isotopomer models. J. Biotechnol. 71, 175 – 189. Seyoum, E., Selhub, J., 1998. Properties of food folates determined by stability and susceptibility to intestinal pteroylpolyglutamate hydrolase action. J. Nutr. 128, 1956 – 1960. Smid, E.J., Starrenburg, M.J.C., Mierau, I., Sybesma, W., Hugenholtz, J., 2001. Increase of folate levels in fermented foods. Innovations Food Technol. Feb/Mar, 13 –15. Stingele, F., Vincent, S.J.F., Faber, E.J., Newel, J.W., Kamerling, J.P., Neeser, J.R., 1999. Introduction of the exopolysaccharide gene cluster from Streptococcus thermophilus Sfi6 into Lactococcus lactis MG1363: production and characterization

213

of an altered polysaccharide. Mol. Microbiol. 32, 1287 – 1295. Sutherland, I.W., 1998. Novel and established applications of microbial polysaccharides. TIBTECH 16, 41 – 46. Sybesma, W., Hugenholtz, J., Mierau, I., Kleerebezem, M., 2001. Improved efficiency and reliability of RT-PCR using tag-extended RT primers and temperature gradient PCR. Biotechniques 31, 466 – 468. Sybesma, W., van den Born, E., Starrenburg, M., Mierau, I., Kleerebezem, M., de Vos., W.M., Hugenholtz, J. Increased production of bioavailable folate by engineering of Lactococcus lactis, submitted for publication. Tsukioka, Y., Yamashita, Y., Oho, T., Nakano, Y., Koga, T., 1997. Biological function of the dTDP-rhamnose synthetic pathway in Streptococcus mutans. J. Bacteriol. 179, 1126 – 1134. van Kranenburg, R., Marugg, J.D., van Swam, I.I., Willem, N.J., de Vos, W.M., 1997. Molecular characterization of the plasmid-encoded eps gene cluster essential for exopolysaccharide biosynthesis in Lactococcus lactis. Mol. Microbiol. 24, 387 – 397. van Kranenburg, R., van Swam, I.I., Marugg, J.D., Kleerebezem, M., de Vos, W.M., 1999a. Exopolysaccharide biosynthesis in Lactococcus lactis NIZO B40: functional analysis of the glycosyltransferase genes involved in synthesis of the polysaccharide backbone. J. Bacteriol. 181, 338 – 340. van Kranenburg, R., Vos, H.R., van Swam, I.I., Kleerebezem, M., de Vos, W.M., 1999b. Functional analysis of glycosyltransferase genes from Lactococcus lactis and other Grampositive cocci: complementation, expression, and diversity. J. Bacteriol. 181, 6347 – 6353. van Kranenburg, R., Kleerebezem, M., de Vos, W.M., 2000. Nucleotide sequence analysis of the lactococcal EPS plasmid pNZ4000. Plasmid 43, 130 – 136. Vedamuthu, E.R., Neville, J.M., 1986. Involvement of a plasmid in production of ropiness (mucoidness) in milk cultures by Streptococcus cremoris MS. Appl. Environ. Microbiol. 51, 1385 – 1386. Von Wright, A., Tynkkynen, S., 1987. Construction of Streptococcus lactis subsp. lactis strains with a single plasmid associated with mucoid phenotype. Appl. Environ. Microbiol. 53, 1385 – 1386. Wald, N., 1991. Prevention of neural tube defects. Results of the Medical Research Council vitamin study. Lancet 338, 131 – 137. Weinstock, G.M., Smajs, D., Hardham, J., Norris, S.J., 2000. From microbial genome sequence to application. Res. Microbiol. 151, 158. Wren, B.W., 2000. Microbial genome analysis: insights into virulence, host adaptation and evolution. Nat. Rev. Genet. 1, 30 – 39. Yao, R., Schneider, E., Ryan, T.J., Galivan, J., 1996. Human gamma glutamyl hydrolase: cloning and characterization of the enzyme expressed in vitro. Proc. Natl. Acad. Sci. USA 93, 10134 – 10138.