Characterization of the individual glucose uptake systems of Lactococcus lactis: mannose‐PTS, cellobiose‐PTS and the novel GlcU permease

According to previous reports, Lactococcus lactis imports glucose via two distinct phosphoenolpyruvate:phosphotransferase systems (mannose‐PTS and cellobiose‐PTS) and one or more unknown non‐PTS permease(s). GlcU was identified as the sole non‐PTS permease involved in the transport of glucose. Additionally, the biochemical properties of PTSMan, PTSCel and GlcU were characterized in double knockout mutants with glucose uptake restricted to a single system. Transport susceptibility to protonophores indicated that glucose uptake via GlcU is proton‐motive force dependent. Competition assays revealed a high specificity of GlcU for glucose. Furthermore, the permease has low affinity for glucose and displays strong preference for the β‐anomer as shown by the profiles of consumption of the two glucose anomers studied by 13C‐NMR. Similar kinetic properties were found for PTSCel, while PTSMan is a high‐affinity system recognizing equally well the two anomeric forms of glucose. Transcripts of the genes encoding the three transporters are present simultaneously in the parent strain NZ9000 as shown by reverse transcription‐PCR. Investigation of the distribution of GlcU homologues among bacteria showed that these proteins are restricted to the low‐GC Gram‐positive Firmicutes. This work completes the identification of the glucose transport systems in L. lactis MG1363.


Introduction
Glucose is abundant in nature, reasonably cheap, and has been the substrate of choice for most bacteria used in biotechnological applications (de Vos and Hugenholtz, 2004;Gosset, 2005;Wendisch et al., 2006;Nevoigt, 2008), since it supports high growth rates and biomass yields. Furthermore, glucose is a preferred sugar in many bacteria and known to repress the synthesis of enzymes necessary for the utilization of other carbohydrates via regulatory mechanisms involving glucose-specific phosphoenolpyruvate (PEP):sugar phosphotransferase system (PTS) components (Stülke and Hillen, 2000;Deutscher et al., 2006;Görke and Stülke, 2008;Jahreis et al., 2008).
A wealth of data have been gathered on the different metabolic reactions that convert glucose into pyruvate; however, the first event in the metabolism of any external nutrient, i.e. transport, has only attracted modest attention. Transport of glucose across the bacterial cytoplasmic membrane proceeds via ATP-binding cassette (ABC) transporters (primary active transporters), secondary carriers or group translocators (PEP:sugar PTS) (Postma et al., 1993;Ehrmann et al., 1998;Pao et al., 1998;Jack et al., 2001;Konings, 2006;Jahreis et al., 2008). The presence of multiple glucose uptake systems is a common feature in many organisms. For example, the Gram-negative model bacterium Escherichia coli possesses at least a glucose-PTS (PTS Glc ), a mannose-PTS (PTS Man ), a proton symporter (GalP) and an ABC transporter (Mgl system) to import glucose (Gosset, 2005), and expression of the genes encoding these transporters is influenced by several factors, such as the nature and the concentration of sugar (Death and Ferenci, 1994;Vanderpool and Gottesman, 2004). The low-GC Gram-positive Bacillus subtilis internalizes glucose by PTS Glc , PTS Man , and the glucose/mannose-proton symporter GlcP during vegetative growth (Paulsen et al., 1998).
Lactococcus lactis is a low-GC Gram-positive coccoid bacterium that is used worldwide as a constituent of starter cultures in the dairy industry. The glucose-PTS is not present in this bacterium: glucose is transported and concomitantly phosphorylated by mannose and cellobiose PTSs (PTS Man and PTS Cel ). Moreover, glucose can also be imported via non-PTS permease(s) and subsequently phosphorylated by glucokinase (Thompson et al., 1985;Pool et al., 2006). Hitherto, gene(s) coding for active non-PTS permease(s) in L. lactis remained elusive. Therefore, we set out to identify the non-PTS transport system(s) and characterize the routes for glucose uptake. In this work, we demonstrate the role of GlcU in the transport of glucose by this organism. Deletion of glcU in a PTS-deficient strain abolished growth on glucose, proving that GlcU is the sole non-PTS permease in L. lactis. Biochemical features of the individual transporters were determined in double mutant strains with glucose transport restricted to a single system. Kinetic properties were obtained by studying uptake of radiolabelled glucose in whole cells; moreover, in vivo 13 C-NMR was used to monitor the consumption of a-and b-anomers of glucose as well as the dynamics of intracellular metabolite pools. The expression of the transporter encoding genes in the parent and mutant strains was evaluated by reverse transcription (RT)-PCR analysis.

GlcU is the only non-PTS glucose permease of L. lactis
Lactococcus lactis NZ9000 transports glucose via PTS Man (ptnABCD), PTS Cel (ptcBAC) and by thus far unknown non-PTS permease(s). Previously, a transcriptome analysis approach proved useful to identify PTS Cel as a glucose transporter . Assuming that the non-PTS system(s) would be expressed at a higher level in a double-PTS mutant (in which PTS Man and PTS Cel are inactivated), mRNA levels in NZ9000 and this transport mutant were compared using DNA microarrays. Genes encoding proteins with high isoelectric points and with putative transmembrane segments were selected among all genes significantly upregulated (data not shown). The gene with the highest fold overexpression in the PTSs-defective strain was ytgA (llmg_2145), which is part of a putative operon (ytgBAH). However, deletion of this gene cluster in the double-PTS mutant did not affect growth on glucose, suggesting a minor or even no role of ytgBAH-encoding proteins in glucose transport. As a second approach, BLASTP searches of the L. lactis MG1363 genome sequence were performed using sequence information on functionally characterized glucose non-PTS permeases. Of these, only the Staphylococcus xylosus glucose uptake protein (GlcU) (Fiegler et al., 1999) showed homology (36% identity) to an L. lactis MG1363 protein that is encoded by llmg_2561 and is annotated as GlcU (Wegmann et al., 2007). This gene was not significantly upregulated in the DNA microarray analysis of the double-PTS knockout mutant. Nevertheless, to investigate whether llmg_2561 (glcU) encodes a functional glucose permease the gene was deleted in a PTS Man /PTS Cel -deficient strain. First, the PTS Cel route was inactivated in NZ9000DptnABCD  by deleting the gene encoding its membrane component, ptcC: a 1.25 kb fragment 45 bp downstream of the start codon was removed by double-cross-over recombination. Subsequently, llmg_2561 (glcU) was inactivated in strain NZ9000DptnABCDDptcC by removing a 0.97 kb fragment starting 103 bp upstream of the start codon of the gene. The triple mutant was obtained on M17 medium supplemented with galactose (1%).
The ability of the mutants to grow on glucose was investigated. Inactivation of llmg_2561 (glcU) in the PTSdeficient background rendered a strain unable to grow on glucose (Fig. 1A). We verified that this phenotype arises from the inability of the triple mutant to transport glucose (Fig. 1B). Expression in trans of glcU under the control of the nisin promoter in the triple mutant restored growth on glucose (Fig. 1A). Altogether these findings unequivocally show that the protein encoded by llmg_2561, hereafter denominated GlcU, is the sole PTS-independent glucose transporter in L. lactis.
Kinetic properties of L. lactis glucose transport systems The kinetic properties of GlcU were investigated in the double mutant NZ9000DptnABCDDptcC. To characterize the PTS Man and the PTS Cel , glcU was deleted in strains NZ9000DptcC and NZ9000DptnABCD, respectively, resulting in strains NZ9000DptcCDglcU (only PTS Man functional) and NZ9000DptnABCDDglcU (only PTS Cel functional). The kinetic properties of the various glucose transporters were determined from [ 14 C]-glucose uptake experiments using non-linear regression analysis to estimate Km and Vmax values.
Glucose transport via GlcU is proton-motive force dependent. The lactococcal non-PTS permease, GlcU, showed a low affinity (Km, 2.4 mM) and a moderate capacity (39 nmol min -1 mg prot -1 ) for glucose uptake in resting cells of NZ9000DptnABCDDptcC. To investigate the mechanism of transport via GlcU, the protonophores tetrachlorosalicylanilide (TCS) and carbonyl cyanide-mchlorophenylhydrazone (CCCP) were used. CCCP and TCS inhibited the uptake of [ 14 C]-glucose by 53% and 100% respectively. Furthermore, cells energized with arginine (formation of ATP via the arginine deiminase pathway) prior to glucose addition showed a twofold (81 nmol min -1 mg prot -1 ) increase in the transport capac-ity (Table 1). Altogether these results indicate that glucose transport via GlcU is driven by the proton-motive force. To further characterize GlcU, transport competition assays were performed in which the uptake of radiolabelled glucose was measured in the presence of 100-fold excess concentration of ribose, rhamnose, mannose, xylose, arabinose, cellobiose and galactose. Glucose uptake was reduced by approximately 30% in the presence of mannose and xylose; the other sugars affected glucose transport to a minor extent (less than 15%).

PTS Man and PTS Cel have different affinity for glucose.
Kinetic properties of PTS Man and PTS Cel were determined in whole cells of NZ9000DptcCDglcU and NZ9000DptnABCDDglcU, respectively (Table 1). Both PTS Man and PTS Cel transport glucose with high capacity, but PTS Cel exhibits a much lower affinity (Km of 8.7 mM compared with 13 mM for PTS Man ).

The transport route affects glycolytic flux and dynamics of intracellular metabolites
Lactococcus lactis strains with single glucose uptake systems, PTS Man , PTS Cel or GlcU, were studied by in vivo 13 C-NMR (Fig. 2). The isogenic L. lactis laboratory strains MG1363 and NZ9000 convert glucose homofermentatively with a maximal glucose consumption rate of about 0.40 mmol min -1 mg prot -1 (Neves et al., 2002a;2006). All mutants showed lower maximal glucose consumption rates (1.4-to 2.9-fold lower). Glucose consumption in NZ9000DptcCDglcU was quasi-linear up to concentrations close to depletion and the maximal rate was 0.29 mmol min -1 mg prot -1 , a value similar to that of Vmax for PTS Man (Table 1). In NZ9000DptnABCDDglcU, glucose consumption was initially relatively high, slowing down (sixfold) at approximately 8 min, at the onset of b-glucose depletion. When uptake of glucose was restricted to GlcU (NZ9000DptnABCDDptcC) the kinetics of consumption A. Growth in CDM with 1% of glucose (w/v) at 30°C, without pH control (initial pH 6.5). B. [ 14 C]-glucose uptake in whole cells (KPi 50 mM, pH 6.5) at 30°C with 10 mM glucose. Nisin (1 ng ml -1 ) was used to induce expression of glcU in the complemented strain. Symbols: (᭺) NZ9000DptnABCDDptcC; (᭹) NZ9000DptnABCDDptcCDglcU; (ᮀ) NZ9000DptnABCDDptcCDglcU(glcU + ).

Strain
Transporter present Values of two or more independent experiments were averaged and are reported ϮSD. Vmax and Km were determined using glucose concentrations varying as follows: NZ9000DptcCDglcU, 1-100 mM; NZ9000DptnABCDDglcU, 0.5-15 mM; NZ9000DptnABCDDptcC, 0.01-10 mM. b. Transport determinations were performed in the presence of 2 mM arginine.
was complex (Fig. 2C): an initial lag-phase was followed by a period of acceleration, with the rate varying up to a maximum of 0.17 mmol min -1 mg prot -1 . Depletion of b-glucose resulted in a drastic reduction of the glucose consumption rate. Energization of cells with arginine (2 mM) prior to glucose addition reduced the lag-time about 2.2-fold and increased the maximal glucose consumption rate to 0.23 mmol min -1 mg prot -1 , indicating that glucose transport via GlcU is dependent on the energetic status of the cell. The profile of intracellular metabolites in NZ9000DptcCDglcU (active PTS Man ) resembles that of the parent strains (Neves et al., 2002a;2006), except for the maximal concentrations of 3-phosphoglycerate (3-PGA) and PEP, which were approximately three times higher in the mutant ( Fig. 2A). When glucose was internalized via the PTS Cel (in NZ9000DpnABCDDglcU) fructose 1,6bisphosphate (FBP) accumulated rapidly to a concentration of 35 Ϯ 1 mM, subsequently declining to undetectable levels as glucose consumption decreased. Concomitantly, the pools of 3-PGA and PEP rose to maximal concentrations of 35 Ϯ 3 mM and 12 Ϯ 2 mM respectively ( Fig. 2B). Curiously, in NZ9000DptnABCDDptcC (active GlcU) the FBP pool decreased slowly to undetectable values (Fig. 2C); the accumulation of 3-PGA and PEP was delayed as compared with NZ9000DpnABCDDglcU, and the PEP potential was about 1.8-fold lower.

Modelling of glucose consumption
The kinetics of a-and b-glucose consumption via the individual glucose transport systems was monitored by in vivo 13 C-NMR. The model was developed to quantify the a-and b-anomer uptake fluxes and the flux of anomerization using the NMR time-courses for glucose utilization as input data (see Fig. S1). The estimated kinetic parameters are shown in Table 2. Glucose consumption in strain NZ9000DptnABCDDptcC (active GlcU) was strongly dependent on the energy status of the cells (see above); therefore, all calculations were performed using data obtained with energized cells.  Table 2. Kinetic parameters obtained from the profiles of glucose consumption in the parental strain L. lactis NZ9000 and its derivatives.

Strain
Transport system PTS Man showed no preference for a-glucose or b-glucose, since the two anomers were consumed efficiently ( Fig. 2A) and similar affinity constants (Kafs) were obtained for the two anomers in strain NZ9000DptcCDglcU (Table 2). Accordingly, the net flux for the anomeric conversion was negligible, indicating that the equilibrium anomeric ratio was maintained during glucose consumption. On the other hand, the very high Kaf values for a-glucose, the low fluxes of a-glucose consumption, and the low Kaf values for b-glucose in strains NZ9000DptnABCDDglcU and NZ9000DptnABCDDptcC denote the strong preference of PTS Cel and GlcU for b-glucose. Therefore, the double-phase kinetics observed for glucose consumption in the latter two strains (Fig. 2) are explained by the inability of PTS Cel and GlcU to import a-glucose; the initial phase corresponds to the fast uptake of the preferred anomer (b-glucose), while the glucose consumption rate in the second phase is limited by the rate of conversion of a-into b-glucose.
Taking advantage of the distinct anomeric specificities of the individual transporters it was possible to characterize glucose consumption in the parental strain L. lactis NZ9000. PTS Cel and GlcU were lumped since they displayed similar anomer selectivity (high and low Kaf values for a-and b-glucose respectively). Therefore, the profile of glucose consumption in strain NZ9000 was modelled with two transport systems: PTS Man and <PTS Cel +GlcU>. Moreover, the Kaf values determined in the mutant strain NZ9000DptcCDglcU for a-and b-glucose (Table 2) were used to model PTS Man in the parent strain. The calculations indicate that the parent strain takes up a-glucose exclusively via PTS Man (Fig. S1); furthermore, glucose is taken up with similar efficiency by PTS Man and <PTS Cel +GlcU> (similar Vmax values, see Table 2).
At this stage it should be pointed out that the kinetic parameters obtained by modelling glucose consumption (Table 2) are inherently different from the kinetic parameters obtained from modelling the results of the radiolabelling assays (Table 1). The latter were obtained from assays in which metabolism of glucose was minimized (up to 10 s), while the profiles of glucose consumption were obtained in cells actively metabolizing glucose using in vivo NMR, a technique that allows distinguishing the individual anomers. Transport assays were fitted with a Michaelis-Menten function and the consumption of the glucose anomers was fitted with a similar formalism that considered the two anomers as competitors for the same binding site of the transporters. Thus, although K m and Kaf represent conceptually the same property it is not surprising that the two models yield different values.
It is also pertinent to stress that the Kms obtained for PTS Cel and GlcU in the strains NZ9000DptnABCDDglcU and NZ9000DptnABCDDptcC from the radiolabelling assays (Table 1) are overestimated since these transport-ers recognize b-glucose only, while total glucose (a plus b) was used to extract the kinetic parameters.

The three transport systems are expressed in the parent strain NZ9000
The level of expression of glcU, ptnC and ptcC in strain NZ9000 and derivatives with a single transporter was investigated by RT-PCR analysis (Fig. 3). The results show that all three genes are transcribed in NZ9000 (Fig. 3A). The expression level of glcU and ptnC in the double transporter mutants NZ9000DptnABCDDptcC and NZ9000DptcCDglcU, respectively, was higher than that in the parent strain NZ9000 (Fig. 3B). In contrast, ptcC expression was reduced in strain NZ9000DptnABCDDglcU when compared with that in strain NZ9000.
Transcription of glcU, ptnC and ptcC was studied in NZ9000 for the sugars shown to sustain growth in CDM (Fig. 4). The three genes were transcribed under all the conditions tested. Glucose and mannose induced the expression of glcU, whereas ptnC expression was induced by mannose, cellobiose and maltose. Glucose, cellobiose and maltose stimulated the transcription of ptcC.

Discussion
This work completes the identification of the glucose transport systems in L. lactis MG1363. Glucose uptake can proceed via PTS Man , PTS Cel and the secondary carrier GlcU, herein disclosed for the first time. Noteworthy, the genes encoding the three different transporters are also present in the genome sequences of L. lactis strains IL1403 and SK11 (Bolotin et al., 2001;Makarova et al., 2006). Moreover, it is shown that GlcU is the sole non-PTS permease involved in the transport of glucose in L. lactis MG1363. The protein shares 36% identity with the functionally characterized S. xylosus glucose uptake protein (Fiegler et al., 1999). The latter is a member of the exclusively prokaryotic glucose/ribose porter family, which belongs to the drug metabolite superfamily of transporters (Jack et al., 2001). A recent genomic analysis revealed that the glucose/ribose porters are well represented in low-GC Gram-positive organisms (Lorca et al., 2007). Moreover, BLASTP searches of all available prokaryotic genome sequences (http://www.ncbi.nlm.nih.gov/sutils/ genom_table.cgi) using the lactococcal or the staphylococcal GlcU sequences as queries retrieved homologues only within the low-GC Gram-positive Firmicutes (data not shown). In fact, GlcU-like proteins are widespread in the orders Bacillales and Lactobacillales and are also found in two Clostridium spp. strains. In Bacillus spp., glcU is transcribed during the sporulation process in the forespore (Fujita et al., 1977;Nakatani et al., 1989), whereas vegetative cells express a different non-PTS permease, the Glucose uptake in L. lactis 799 glucose/mannose:H + symporter (Paulsen et al., 1998). Additionally, we speculate that GlcU is the primary non-PTS glucose permease in the Lactobacillales. This assumption is based on a screen of all available genomes of this taxonomic order for homologues of the other characterized bacterial glucose secondary carriers, the E. coli GalP (McDonald et al., 1997) and the Streptomyces coelicolor GlcP ( van Wezel et al., 2005), which retrieved exclusively proteins annotated as xylose:H + symporters (data not shown).
The reported Km values for glucose uptake via secondary carriers vary up to three orders of magnitude (mM or Fig. 3. Transcriptional studies of glucose transport systems in L. lactis. NZ9000 and derivatives were grown in CDM with 1% glucose (w/v) and without pH control (initial pH 6.5). A. Transcription of glucose uptake system genes in strain NZ9000. PCR experiments were performed with chromosomal DNA from NZ9000 (DNA; positive control), RNA treated without reverse transcriptase (RNA; negative control) and cDNA resulting from the reverse transcription reaction (RT). B. Comparison of gene transcription in NZ9000 and each of its isogenic mutants with a single transport system. The cDNA samples were diluted 1:5, 1:10, 1:25, 1:50 and 1:200, and used as template in the PCR analysis. RT-PCR experiments were performed using primers designed to amplify intragenic regions of glcU (590 bp), ptnC (443 bp), ptcC (464 bp) and tufA (471 bp). The elongation factor TU is required for continued translation of mRNA, and in L. lactis MG1363 is encoded by tufA. This housekeeping gene was used as a control in the RT-PCR experiments. Fig. 4. Transcriptional studies of glucose transport systems in L. lactis NZ9000 grown in different sugars. Total RNA was extracted from cultures of L. lactis NZ9000 in CDM with 1% (w/v) glucose, mannose, cellobiose, maltose or galactose, and without pH control (initial pH 6.5). The cDNA samples were diluted 1:5, 1:20, 1:25, 1:50, 1:100 and 1:200, and used as template in the PCR analysis. RT-PCR experiments were performed using primers designed to amplify intragenic regions of glcU (590 bp), ptnC (443 bp), ptcC (464 bp) and the control gene tufA (471 bp). mM), a broad range of affinity that is unrelated to the transporter family type (Cvitkovitch et al., 1995;Weisser et al., 1995;Wagner et al., 2000;Gosset, 2005;van Wezel et al., 2005;Parche et al., 2006). The lactococcal GlcU (Km, 2.4 mM) clusters in the low-affinity group together with the non-PTS permease of Bacillus megaterium and the glucose facilitator (Glf) of Zymomonas mobilis (Km values of 2.5-5 mM and 4.1 mM respectively). The low affinity of these systems for glucose was associated with a facilitated diffusion mechanism (Weisser et al., 1995;Wagner et al., 2000), but this explanation does not suit the data on lactococcal GlcU since we showed that transport via GlcU is dependent on the proton-motive force. Thus, our data are in line with the hypothesis that members of the glucose/ribose porter family operate by H + symport (Jack et al., 2001).
Internalization of glucose in L. lactis is either coupled to PEP-dependent phosphorylation (PTS Man and PTS Cel ) or driven by the PMF (GlcU). Our data show that the lactococcal PTS Man is a high-affinity, high-capacity system able to import both anomers of glucose, while PTS Cel and GlcU are low-affinity transporters with a strong preference for b-glucose. Noteworthy is the clear dependency of glucose metabolism on the pathway used for taking up this sugar. As described before for wild-type strains (Neves et al., 2002a;2006), consumption of glucose led to an increase in the FBP pool, of which the build-up rate and maximal concentration correlate with the glucose consumption rate. As opposed to the other L. lactis strains characterized here, the utilization of glucose, and the subsequent accumulation of FBP, in strain NZ9000DptnABCDDptcC (missing glucose PTSs) did not start immediately after glucose addition. This emphasizes the advantage conferred by PTSs for the rapid uptake of sugar after a starvation period; moreover, the high energetic efficiency of this uptake system is well documented (Postma et al., 1993). Loss of a PEP-consuming activity (either of the two glucose PTSs) resulted in a PEP potential (pools of 3-PGA and PEP) that is three-to fourfold higher than that in wild-type strains (Neves et al., 2002a;2006). Curiously, the PEP potential in the strain devoid of glucose PTSs was lower than that in the isogenic strains with a single PTS (NZ9000DptcCDglcU and NZ9000DptnABCDDglcU). This feature is most likely associated with the low rate of FBP disappearance, which is indicative of a constriction at the level of glyceraldehyde 3-phosphate dehydrogenase or phosphoglycerate kinase in the double-PTS mutant.
RT-PCR experiments using RNA from the parent strain L. lactis NZ9000 showed that the glcU, ptnC and ptcC genes are all transcribed under the conditions employed, suggesting that the three transport systems are present in this strain (Fig. 3A). Analysis of the consumption of a-and b-glucose in strain NZ9000 allowed us to assess the relative contributions of the high-affinity transporter PTS Man and the low-affinity systems <PTS Cel +GlcU>. The similar capacities found for the high-and the lumped low-affinity systems in strain NZ9000 indicate that their contributions to the total glucose uptake are comparable. The individual contributions of the two low-affinity transporters are difficult to evaluate. Assuming that the transcript level is proportional to the uptake activity, the higher level of ptcC transcript would imply that the PTS Cel uptake rate in strain NZ9000 is higher than that found for the transporter in the absence of both other transporters (in strain NZ9000DptnABCDDglcU) ( Table 2). Along the same line of reasoning, strain NZ9000 would have a lower GlcU activity. Thus, it is possible that PTS Cel contributes to a greater extent than GlcU to glucose uptake in NZ9000.
We showed that NZ9000 possesses two distinct PTSs and the secondary carrier GlcU for glucose uptake. The presence of two systems displaying similar kinetic properties, the PTS Cel and the secondary carrier GlcU, is an intriguing feature. It is reasonable to question whether the systems are simultaneously operative or if the environmental conditions dictate the activity of each transporter. Indeed, it has been described that an acidic environment has a negative effect on PTS activity (Vadeboncoeur et al., 1991). Furthermore, it was shown that non-PTS permease(s) in Streptococcus mutans become functional at low pH (Hamilton and Ellwood, 1978;Cvitkovitch et al., 1995). Given that the fermentation of glucose to lactic acid progressively acidifies the lactococcal growth medium, we speculate that the presence of GlcU in L. lactis confers a competitive advantage to thrive at acidic pHs. It is anticipated that the insight into glucose transport derived from this work will assist in the design of L. lactis strains with improved traits for industrial applications.

Microbial strains and growth conditions
Strains and plasmids used in this work are listed in Table 3. Routinely, L. lactis strains were cultivated as batch cultures without aeration in M17 medium (Difco TM , Sparks, MD, USA) supplemented with glucose (1% w/v) at 30°C. Deletion mutants were constructed in galactose-M17 (0.5% w/v), and the temperature was raised to 37°C for plasmid integration/ excision. NMR studies were performed with cells grown in chemically defined medium (CDM) supplemented with 1% glucose (w/v), under anaerobic conditions in a 2 l fermenter (B. Braun Biostat ® MD, B. Braun Biotech International, Melsungen, Germany). The CDM was gassed with argon for 60 min prior to inoculation (4-5% inoculum from a culture grown overnight), the pH was kept at 6.5 by automated addition of 10 M NaOH, and an agitation rate of 70 r.p.m. was used. RT-PCR studies were performed with cells grown in rubber-stoppered bottles (200 ml) in CDM without pH control (initial pH 6.5). The following sugars (1% w/v) were tested as carbon sources: glucose, mannose, cellobiose, maltose, galactose, xylose, rhamnose and ribose. Growth was not observed when xylose, rhamnose and ribose were used as sole carbon sources in CDM. When necessary, erythromycin or chloramphenicol was used at a final concentration of 5 mg ml -1 . For complementation, nisin (1 ng ml -1 ) was used. Growth was monitored by measuring the optical density at 600 nm.

DNA techniques
General molecular techniques were performed essentially as described elsewhere (Sambrook et al., 1989). Chromosomal and plasmid DNA were isolated by the method of Johansen and Kibenich (1992) and Birnboim and Doly (1979) respectively. L. lactis was transformed with plasmid DNA by electroporation as described by Holo and Nes (1995). Restriction enzymes and T4 DNA ligase were obtained from New England Biolabs (Ipswich, MA, USA) and Pwo polymerase and Taq polymerase were obtained from Roche Applied Science (Mannheim, Germany) and were used according to the supplier's instructions. PCR reactions were performed in a MyCyclerTM thermal cycler (Bio-Rad, Hercules, CA, USA). Primers (listed in Table S1) were purchased from Thermo Fisher Scientific (Waltham, MA, USA).

Construction of L. lactis mutant strains and plasmids
Gene deletions were all performed in L. lactis NZ9000 and were constructed using a two-step homologous recombination method (Leenhouts et al., 1996). This method does not leave antibiotic resistance markers in the chromosome, and multiple deletions in one strain can be accomplished. Chromosomal DNA of L. lactis NZ9000 was used as a template in PCR amplifications. The PTS Man route was inactivated by deletion of ptnABCD-encoding enzyme II Man as previously described . To disrupt the PTS Cel , a strain carrying a deletion in ptcC, the membrane-bound domain IIC of enzyme II Cel , was constructed as follows: a 1253 bp deletion in ptcC (1338 bp) was made using the primer pairs ptcC1/ptcC2 and ptcC3/ptcC4. Genes ytgBAH (locus tags llmg_2146, llmg_2145 and llmg_2143 in L. lactis MG1363 genome sequence) (Wegmann et al., 2007) were inactivated by deletion in strain NZ9000DptnABCDDptcC using the primers ytgBAH1/ytgBAH2 and ytgBAH3/ytgBAH4. L. lactis NZ9000DglcU, carrying only the last 23 bp of llmg_2561 (glcU), was engineered using the primer pairs glcU1/glcU2 and glcU3/glcU4. The deletions were confirmed by PCR and Southern blotting analysis (Fig. S2). For a complementation study, the glcU gene (888 bp) was amplified by PCR using L. lactis NZ9000 DNA as template and primer pairs GlcU-fw and GlcU-rev (Table S1). The NcoI/HindIII-restricted PCR product was cloned in NcoI/HindIII-restricted pNZ8048. The resulting plasmid pNZ8048[glcU] was transformed into strain NZ9000DptnABCDDptcCDglcU, yielding strain NZ9000DptnABCDDptcCDglcU(glcU + ).

[ 14 C]-glucose transport studies
All strains were grown in M17 medium supplemented with 1% glucose, without pH control. Cells were harvested at the mid-exponential phase of growth, washed twice in KPi buffer (5 mM, pH 6.5) and suspended in the KPi buffer (50 mM, pH 6.5). Initial glucose uptake rates were measured at 30°C in 100 ml of cell suspensions with appropriate optical densities at 600 nm. [U-14 C] glucose was added to a final concentration of 0.001-20 mM (specific activity 0.02-19 mCi mmol -1 ). Uptake assays were performed essentially as described by Wolken et al. (2006), except that KPi buffer (50 mM, pH 6.5) was used to wash the filters. For kinetic analysis of glucose transport via GlcU (in NZ9000DptnABCDDptcC), the cell sus-  Leenhouts et al. (1996) pVE6007 Cm r , temperature-sensitive derivative of pWV01 Maguin et al. (1992)  pension was incubated with 2 mM arginine prior to the addition of labelled glucose. The kinetic parameters (Km and Vmax) for glucose uptake were estimated by fitting the data using non-linear least squares regression analysis (Excel solver, Microsoft Excel 2003) to the Michaelis-Menten equation.
Glucose uptake was also evaluated in the presence of 20 mM TCS and 150 mM CCCP. The effect of protonophores was studied as follows: cell suspensions of NZ9000DptnABCDDptcC were incubated with the compound for 3 min at 30°C. [ 14 C]-glucose was added to a final concentration of 150 mM and the uptake was stopped after 60 s. The influence of ethanol (2%, v/v), the solvent of CCCP and TCS, was also examined. Transport competition experiments were performed with energized NZ9000DptnABCDDptcC (in the presence of 2 mM arginine). Glucose uptake was measured in cell suspensions with 0.5 mM glucose and 100-fold excess of the following carbohydrates: ribose, rhamnose, mannose, xylose, arabinose, cellobiose and galactose. The competition assays were performed in triplicate using cells from two independent cultures.

In vivo NMR experiments
Cells were grown in CDM containing 1% glucose (w/v) and suspensions were prepared and made anaerobic as described elsewhere (Neves et al., 1999). In vivo NMR experiments were performed using an online system and glucose specifically labelled with 13 C on carbon one (20 mM) as substrate (Neves et al., 1999;2002a). In strain NZ9000DptnABCDDptcC, [1-13 C]-glucose utilization was also studied following energization of the resting cells with arginine (2 mM). In vivo 13 C-NMR spectra were acquired at 125.77 MHz using a quadruple nuclei probe head at 30°C on a Bruker AVANCE II 500 MHz spectrometer (Bruker BioSpin GmbH, Karlsruhe, Germany) as described before (Neves et al., 1999). Lactate was quantified in the NMR-sample extract by 1 H-NMR in a Bruker AMX300 (Bruker BioSpin GmbH). The concentration of other metabolites was determined in fully relaxed 13 C spectra of the NMR-sample extracts as previously described (Neves et al., 2002b). Each experiment was repeated at least twice and the results were highly reproducible.

Kinetics of a-and b-glucose consumption
A mathematical model for the consumption of a-and b-glucose was developed using Michaelis-Menten formalism, and taking into account the first-order kinetics of anomerization. The model considers that each anomer competes with the other for the free transporter (Tfree), leading to the formation (k+1) of a complex ([aGlc-T] or [bGlc-T]). Binding to the transporter is reversible and dissociation can occur without transport (k-1), or the glucose anomer is transported and released to the intracellular space (k2) (schematic representation in Fig. S3). The uptake rate of the individual anomers is given by , Vmax = k2 · T for n = a or b. Moreover, it was assumed that the Vmax values for a-and b-glucose were identical, i.e. k k 2 2 α β = . Therefore, Vmax represents the maximum capacity of glucose consumption, regardless the anomeric form that is taken up. The parameters K af α and K af β are the affinity constants for utilization of each glucose anomer. It is implicitly assumed that transport is the rate limiting step in the metabolism of glucose and the actual kinetics of glucose uptake may be affected by subsequent metabolism. Consumption of a-and b-glucose by L. lactis as monitored by 13 C-NMR was in some cases slower than the anomerization rate, thus requiring the consideration of the anomerization step. The first-order rate constants of glucose anomerization were determined by acquiring sequences of 13 C-NMR spectra of 20 mM a-[1-13 C]-glucose dissolved in the same buffer used for in vivo NMR experiments (50 mM KPi, pH 6.5), at 30°C. Under these conditions the equilibrium molar percentages of a-glucose and b-glucose were 38.2% and 62.8% respectively. The first-order rate constants were 0.108 Ϯ 0.001 min -1 for the conversion of a into b and 0.063 Ϯ 0.001 min -1 for the conversion of b into a.
The profile of glucose consumption by L. lactis was approximately sigmoidal, i.e. at the initial stage, immediately after glucose addition, glucose uptake proceeded at a rate lower than maximal and this feature was especially apparent in the strain transporting glucose via the GlcU permease only (see above). To account for this behaviour the Vmax of glucose transport was multiplied by the following function: . ( The parameters a, b and c were determined so that 0 Յ j Յ 1, and j increases from a to 1 monotonically with time, t. The model was implemented in MATLAB v7.3.0 (The MathWorks, Natick, MA, USA) as a set of ordinary differential equations.
The parameters (Vmax, K af α , K af β , a, b and c) were determined by fitting the model to the a-and b-glucose consumption profiles determined experimentally. A simulated annealing algorithm was used to ensure that the minimum found was indeed a global minimum. The differential equations were solved using the ode23s function and the non-linear regression performed using the lsqnonlin function.
Glucose uptake in L. lactis 803