• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of aemPermissionsJournals.ASM.orgJournalAEM ArticleJournal InfoAuthorsReviewers
Appl Environ Microbiol. Apr 2007; 73(8): 2661–2672.
Published online Feb 23, 2007. doi:  10.1128/AEM.00005-07
PMCID: PMC1855617

Comparative High-Density Microarray Analysis of Gene Expression during Growth of Lactobacillus helveticus in Milk versus Rich Culture Medium[down-pointing small open triangle]

Abstract

Lactobacillus helveticus CNRZ32 is used by the dairy industry to modulate cheese flavor. The compilation of a draft genome sequence for this strain allowed us to identify and completely sequence 168 genes potentially important for the growth of this organism in milk or for cheese flavor development. The primary aim of this study was to investigate the expression of these genes during growth in milk and MRS medium by using microarrays. Oligonucleotide probes against each of the completely sequenced genes were compiled on maskless photolithography-based DNA microarrays. Additionally, the entire draft genome sequence was used to produce tiled microarrays in which noninterrupted sequence contigs were covered by consecutive 24-mer probes and associated mismatch probe sets. Total RNA isolated from cells grown in skim milk or in MRS to mid-log phase was used as a template to synthesize cDNA, followed by Cy3 labeling and hybridization. An analysis of data from annotated gene probes identified 42 genes that were upregulated during the growth of CNRZ32 in milk (P < 0.05), and 25 of these genes showed upregulation after applying Bonferroni's adjustment. The tiled microarrays identified numerous additional genes that were upregulated in milk versus MRS. Collectively, array data showed the growth of CNRZ32 in milk-induced genes encoding cell-envelope proteinases, oligopeptide transporters, and endopeptidases as well as enzymes for lactose and cysteine pathways, de novo synthesis, and/or salvage pathways for purines and pyrimidines and other functions. Genes for a hypothetical phosphoserine utilization pathway were also differentially expressed. Preliminary experiments indicate that cheese-derived, phosphoserine-containing peptides increase growth rates of CNRZ32 in a chemically defined medium. These results suggest that phosphoserine is used as an energy source during the growth of L. helveticus CNRZ32.

Lactobacillus helveticus CNRZ32 (1) is a gram-positive, non-spore-forming, catalase-negative, microanaerophilic rod (20) with a G+C content of 37.1% (22). Representatives of this species are obligately homofermentative, thermophilic, lactic acid bacteria (LAB) that are used as starter cultures in the manufacturing of a variety of fermented dairy products, such as yogurt and mozzarella and Swiss cheeses (14), and as flavor adjunct cultures in other types of cheese, such as Gouda (1). The use of L. helveticus in the production of dairy foods has received increased attention because of this organism's ability to generate peptides with antihypertensive (35) and immunomodulating (24) properties from casein during milk fermentation. Like all lactobacilli, L. helveticus is a fastidious organism, requiring exogenous supplies of specific carbon and nitrogen sources, nucleotides, vitamins, and minerals for growth (20). Since L. helveticus CNRZ32 has multiple amino acid auxotrophies, its rapid growth in milk relies on a complex proteolytic enzyme system to obtain essential amino acids from caseins and other milk proteins (5, 6).

Lactose is the primary carbon source for microbial growth in milk, so the growth of L. helveticus in milk requires an enzymatic system to utilize this carbohydrate (11). Other potential sources of energy for microbial growth in milk and cheese include citrate (17), trace carbohydrates (such as ribose and N-acetylglucosamine) derived from milk glycoproteins and glycolipids (13, 48), nucleotides, nucleosides, and their precursors, formic and orotic acids (39, 40, 41).

Casein-derived peptides formed during the ripening of cheese are rich in phosphoserine (serP) residues (28). The composition of phosphopeptides varies considerably in different cheeses, and while some cheeses accumulate free serP at high levels during maturation, only trace amounts of this compound are registered in other varieties (10). The variation in serP accumulation suggests that serP, either in the form of a free amino acid or as a part of a phosphopeptide, can be utilized by cheese microorganisms, including L. helveticus, and may serve as an additional source of metabolic energy. Indeed, the draft CNRZ32 genomic sequence contains a gene cluster that may encode a pathway for serP utilization.

When milk is fermented into cheese, L. helveticus and other LAB contribute to cheese flavor development during the ripening period through several basic mechanisms that include carbohydrate fermentation, conversion of milk proteins into peptides and free amino acids, catabolism of amino acids into aroma compounds, hydrolysis of milk lipids into free fatty acids, followed by their conversion to esters, and citrate catabolism (12). Our group has compiled a draft (fourfold random coverage) sequence for the 2.4-Mb genome of L. helveticus CNRZ32 (37). The use of this information for functional genomic studies would allow us to better understand the molecular events occurring during cheese ripening and, ultimately, to improve flavor development in bacteria-ripened cheeses.

For this study, we screened the draft sequence for genes that might be important for the growth of L. helveticus CNRZ32 in milk and/or for cheese flavor development. A total of 168 of such genes were identified whose products are expected to be involved in proteolysis, citrate utilization, metabolism of lipids and amino acids, carbohydrate utilization, and other functions. The primary aim of this study was to investigate changes in expression among this subset of genes during the growth of L. helveticus CNRZ32 in milk versus laboratory medium using high-density DNA microarrays. Additionally, noninterrupted sequence fragments (“contigs”) were employed to generate a “tiled” microarray by depositing consecutive probes designed to represent all of the available sequence information.

The experiments allowed the identification of a number of genes that are expressed differentially during the growth of L. helveticus CNRZ32 in milk. In particular, the results suggested an ability of this organism to generate metabolic energy from phosphoserine residues in milk caseins and to convert citrate into succinate, an important flavor compound. These data provide new insight to L. helveticus physiology and identify targets for future functional genetics experiments in this species.

MATERIALS AND METHODS

Bacteria and culture media.

L. helveticus CNRZ32 (1) was maintained in a laboratory collection as a glycerol stock at −80°C and propagated at 42°C in MRS broth (Difco Laboratories, Detroit, MI). Skim milk for these experiments (Babcock Dairy, Madison, WI) was double steamed for 20 min with a 2-h, 42°C incubation between treatments. Chemically defined medium (CDM) base was prepared as described by Christensen and Steele (6) and contained a limiting concentration of glucose (3 mM); pH was adjusted to 6.0 with NaOH. Potential growth promoters, namely serine, phosphoserine (both from Sigma), and phosphopeptides isolated from Herrgard cheese (“pool 3” [28]), were filter sterilized before the addition to autoclaved CDM. Cells for growth experiments were propagated from frozen stocks by passage through two sequential subcultures in MRS broth at 42°C for 17 h. The cells were washed in saline and inoculated to a final calculated optical density at 600 nm (OD600) of 0.02 or 0.005 in 400 ml of milk or MRS broth, respectively. When CDM was used as the growth medium, the cells were inoculated at an OD600 of 0.05 into 5 ml of medium. Milk and MRS cultures were incubated at 42°C until cells reached mid-log phase (approximately 9 h, as determined by preliminary trials), while CDM cultures were grown for 24 h. The final pH values of MRS and milk cultures were 5.5 to 5.6. Because a comparison of gene expression profiles between the cultures of different growth phases was not the subject of this study, this point of the growth curve was chosen arbitrarily. Three independent growth experiments in milk and MRS were performed, and two identical samples were taken from each culture and used for the RNA isolation. A total of six RNA samples for each growth condition were used to independently produce labeled cDNA for microarray experiments. Growth experiments in CDM were performed in triplicate. The growth rate was defined as the maximum slope of log OD600 versus time graphs.

RNA isolation and purification.

Cultures were mixed by vortexing with two volumes of RNAprotect reagent (QIAGEN, Inc., Valencia, CA) containing 100 μg/ml of rifampin (Sigma, St. Louis, MO) and incubated for 5 min at room temperature, followed by centrifugation at 5,500 × g for 15 min at 4°C. The cell pellets were resuspended in 5 ml of lysozyme solution (20 mg/ml) containing 0.1 mg/ml of rifampin and incubated for 25 min at 37°C. After repeated centrifugation, the pellets were resuspended by vortexing in 5 ml of TRIzol reagent (Invitrogen, Carlsbad, CA) and incubated for 10 min at room temperature. Next, 1 ml of chloroform was added, followed by vortexing for 15 s and incubation for 10 min at room temperature. The mixtures were centrifuged at 16,000 × g for 20 min at 4°C, 2.5 ml of the upper aqueous phase was transferred into a fresh tube, mixed with 2.5 ml of isopropanol, and incubated at room temperature for 10 min and RNA was pelleted at 16,000 × g for 20 min at 4°C. After washing in 75% ethanol, the RNA pellet was dried and dissolved in 100 μl of water. The isolated total RNA was treated with 5 U of RQ1 DNase I (Promega, Madison, WI) and then purified using the RNeasy purification system (QIAGEN).

cDNA synthesis and labeling.

cDNA was synthesized from 12 μg of total RNA using random hexamer primers (Amersham Bioscience, Piscataway, NJ) and SuperScript II reverse transcriptase (Invitrogen). After synthesis, template RNA was digested with RNase H (Promega) and RNase A (Epicenter, Madison, WI), and cDNA was purified using the QIAquick PCR purification kit (QIAGEN) and then fragmented into approximately 70-base fragments using appropriately diluted RQ1 DNase I (Promega). Fragmentation efficiency was determined with an Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). Fragmented cDNA was end labeled with biotin-N6-ddATP (NEN/Perkin-Elmer, Boston, MA) using terminal deoxynucleotidyl transferase (Promega), followed by the concentration of labeled sample on Microcon YM-10 columns (Millipore, Bedford, MA).

Microarray design and production.

Based on the sequence and annotation data for L. helveticus CNRZ32, high-density, photolithography-based, monoplex (one array per chip) microarrays were designed and produced by NimbleGen Systems, Inc. (Madison, WI). The arrays consisted of 24-mer probe pairs, 18 “perfect match” probes that were identical to the original sequence, and 18 “mismatch” probes that differed from the original sequence at the two center positions for each of 168 polished genes. Genes selected included those whose products are expected to be involved in proteolysis/milk protein utilization/oligopeptide transport (58 genes), citrate utilization (13 genes), lipid and amino acid metabolism (9 and 27 genes, respectively), carbohydrate utilization (9 genes), and 50 other genes predicted to be involved in exopolysaccharide synthesis, nucleotide metabolism, solute transport, competence, or cell division processes. Additionally, “tiled” probe pairs were designed to cover the entire draft sequence in both the “sense” and “antisense” (reverse-complement) directions, with an average distance of approximately 30 nucleotides between probes. Each of the probe pairs was synthesized in a computer-generated randomized pattern on the array.

Hybridization, Cy3 conjugation, and antibody amplification.

All hybridization, staining, and processing of arrays were performed by personnel at NimbleGen Systems essentially as previously described by Ulijasz et al. (44). In brief, hybridizations were carried out at 45°C for 16 h on a rotisserie-like apparatus (HybriWheel) to enhance uniformity of hybridizations across the array surface. After hybridization, the arrays were washed in buffers of various levels of stringency and then streptavidin was conjugated to the end-labeled biotin, followed by biotin-anti-streptavidin in the presence of normal goat immunoglobulin G, and finally conjugated to a Cy3 streptavidin.

Array scanning, data extraction, and analysis.

Arrays were scanned using an Axon model 4000 scanner (Molecular Devices Corporation, Union City, CA) and the data were extracted using NimbleScan software. Array normalization was performed using the quantile normalization method of Bolstad et al. (2). Normalized expression values for the individual probes were used to obtain the expression values for a given open reading frame (ORF) by using the robust multiarray average (RMA) procedure as previously described by Irizarry et al. (18). Finally, n-fold change ratios (R) were calculated using the RMA-processed expression values (RMA calls) obtained for a particular gene in milk and MRS cultures.

For the tiled arrays, a manual probe-by-probe scan was conducted to locate clusters of three or more consecutive probes which produced an average signal intensity that was at least 1.95-fold higher in milk-grown than in MRS-grown cells. Genome sequence regions that corresponded to these probes were then used in similarity searches with online BLASTx and BLASTp algorithms (http://www.ncbi.nlm.nih.gov/BLAST/) to identify potential ORF products. Finally, normalized data for the probes covering the newly discovered ORFs were subjected to the RMA procedure and the expression ratios were calculated.

A mixed-model analysis of variance approach (49) was used to determine which annotated or newly discovered L. helveticus CNRZ32 ORFs were differentially expressed in milk or MRS medium. This statistical method was applied to test the expression changes of all annotated genes, i.e., independently of the n-fold change ratios. However, only those newly discovered ORFs which retained change ratios (n-fold) above or equal to 1.95 after applying the RMA treatment were subjected to further statistical analysis. To account for multiple testing, a Bonferroni's adjustment to an overall significance level of 0.10 was used as the cutoff value (i.e., α = 0.10/n, where “n” is the number of ORFs tested). Finally, the use of arbitrary a priori P value cutoffs in conjunction with Bonferroni's correction proved useful in the analysis of microarray data (9, 33), so data in this study was also examined at two P values, 0.01 and 0.05, which are commonly used in biological experiments as arbitrary significance cutoffs. Results for a particular gene are presented as an n-fold change of expression, along with its cognate P value and whether the change was significant after Bonferroni's correction.

RT-PCR.

Primers for 14 different genes (Table (Table1)1) were designed with GeneWorks software (IntelliGenetics, Inc., Mountain View, CA) and compared against the L. helveticus CNRZ32 genome sequence using BioEdit 5.0.6 software (www.mbio.ncsu.edu/BioEdit/bioedit.html) to verify that each annealed to a single locus in the genome. Primer pairs were predicted to have annealing temperatures that ranged from 55 to 61°C and to produce amplicons that ranged from 89 to 148 bp in length. Template DNA from L. helveticus CNRZ32 was used to determine optimal reaction conditions in real-time PCR (RT-PCR) for each primer pair to ensure the absence of any nonspecific amplification, and the identity of PCR products was confirmed by sequencing. Reactions were performed in an Opticon II thermal cycler (MJ Research, Reno, NV). Each reaction consisted of 5 μl of a solution of either template DNA or cDNA or water in the negative controls, 5 μl of primer mix (1.2 μM of each primer), and 10 μl of SYBR green mix (MJ Research). Blanks contained 10 μl of SYBR green mix plus 10 μl of water. RT-PCR was performed using two concentrations of cDNA (2 ng/μl and 0.02 ng/μl) obtained from milk or MRS-grown L. helveticus CNRZ32 cultures as described above. Triplicate reactions were run in 96-well plates. Amplicon quantification in RT-PCRs was performed by comparison with gene-specific standard curves constructed from known concentrations of individually purified amplicons. The obtained amplicon copy numbers were log transformed and used in the calculation of the expression change (n-fold) for a particular gene.

TABLE 1.
Oligonucleotides used for real-time quantitative PCR

Nucleotide sequence accession numbers.

The nucleotide sequences of 168 completely sequenced genes have been deposited in GenBank (http://www.ncbi.nlm.nih.gov/GenBank/index.html). The accession numbers are given in Tables Tables2,2, ,3,3, and and4.4. Microarray data analyzed in this study have been deposited in the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/) and were assigned series record GSE7005.

TABLE 2.
Annotated genes induced during growth of Lactobacillus helveticus CNRZ32 in milk
TABLE 3.
Annotated genes induced during growth of L. helveticus CNRZ32 in MRS broth
TABLE 4.
Annotated genes which were not differentially expressed during growth of L. helveticus CNRZ32 in milk or MRS

RESULTS

Microarray expression profiling.

Normalization and analysis of array hybridization signals found 28 of 168 annotated genes were differentially expressed at a level of 1.45-fold or greater, and 15 of these genes were induced at least 1.95-fold when L. helveticus CNRZ32 was grown in milk compared to that in MRS broth (Table (Table2).2). Application of the most stringent multiple comparison correction, Bonferroni's adjustment with overall cutoff value of 0.1, indicated that 25 genes were significantly upregulated in milk-grown cells. These genes included one gene, pepT, whose expression changed less than 1.45-fold. However, Bonferroni's correction is considered to be the most conservative approach to reduce type I (false-positive) errors and, in turn, would increase the potential for type II (false-negative) errors (38, 47). Further analysis using 0.01 and 0.05 as arbitrary a priori P value cutoffs revealed the upregulation of 17 additional genes, which brings the total number of polished genes significantly upregulated during the growth in milk to 42.

The expression data obtained from the tiled microarray identified 79 additional ORFs that were significantly upregulated during the growth of L. helveticus CNRZ32 in milk. These included 45 ORFs whose products show significant similarity to previously characterized proteins of known or predicted function (Table (Table5),5), 11 ORFs which may encode proteins significantly similar to the GenBank entries that lack predicted function (i.e., “hypothetical” or “conserved hypothetical” proteins), and 23 potential ORFs (i.e., upregulated nucleotide sequences for which a valid RBS and start codon could be assigned) whose products did not have a significant match in the protein database (data not shown).

TABLE 5.
Putative genes (identified from the tiled microarrays) significantly induced during growth of Lactobacillus helveticus CNRZ32 in milk

Based on the combined results from annotated and tiled microarrays, growth in milk led to the upregulation of 21 genes that encode enzymes from proteolytic systems (proteinases, peptidases, and oligopeptide transporters, presumably participating in casein degradation/assimilation), 4 genes involved in biosynthesis or metabolism of serine and cysteine, 5 genes from carbohydrate fermentation pathways, 27 genes related to nucleotide metabolism, 3 genes of citrate utilization pathways, 4 genes for components of three different ABC-type transport systems, and 4 genes that could encode proteins of other functions. Additionally, the tiled arrays revealed significant upregulation of 18 phage-related proteins during the growth of L. helveticus CNRZ32 in milk (Tables (Tables22 and and55).

Results from an annotated array showed a total of 61 genes to be significantly upregulated during the growth of L. helveticus CNRZ32 in MRS compared to that in milk (P < 0.05), including 17 genes that were significantly upregulated after applying Bonferroni's correction (Table (Table2).2). These included 19 genes that encode enzymes from the proteolytic system, 16 genes involved in amino acid biosynthesis and metabolism, 5 genes needed for carbohydrate utilization, 4 lipase or esterase genes, 5 genes involved in citrate catabolism, components of three distinct ABC-type transporters, and 9 genes that could encode proteins of other functions. These differences found between gene expression patterns in L. helveticus CNRZ32 grown in milk as opposed to those in MRS medium are summarized in Table Table66.

TABLE 6.
Summary of the differences observed in expression of L. helveticus CNRZ32 genes during growth in milk and MRS medium

RT-PCR validation.

RT-PCR experiments were performed for 14 target genes with the same cDNA preparations used in array hybridizations. In general, the RT-PCR data showed good agreement with the microarray results, and there was a positive correlation (r = 0.76) between the two methods (Fig. (Fig.1).1). However, divergent results between the two platforms were observed for three genes, asnA (asparagine synthase), cysE (serine O-acetyltransferase), and pyrR (pyrimidine operon regulator). Microarray data indicated that asnA was significantly repressed in milk-grown cells according to the microarray analysis (Table (Table3),3), but RT-PCR showed its upregulation (R = 1.49), though the effect was not significant (P > 0.05). The second gene, cysE, was in turn found to be upregulated in milk-grown cells by microarray analysis (Table (Table2),2), but no significant difference in expression was recorded by RT-PCR (R = 0.92; P > 0.05). The third gene, pyrR, was not differentially expressed as judged by microarray analysis (R = 0.90; P > 0.05) but was upregulated based on the RT-PCR data (R = 3.52; P < 0.01). Therefore, no statistically significant opposite trends (i.e., upregulation instead of downregulation) were observed when the expression level of a particular gene was analyzed by two alternative methods. The slope of the regression line was found to be 2.48, which indicates that RT-PCR generally measured more substantive expression changes than did microarray analysis.

FIG. 1.
Correlation of expression ratios from microarray profiling and RT-PCR. Total RNA was extracted from milk-grown and MRS-grown cultures of Lactobacillus helveticus CNRZ32 and served as a template for cDNA synthesis to be used in microarrays and RT-PCR experiments. ...

Growth experiments in chemically defined media.

Initial trials indicated that the addition of serP to CDM at 2.75 mM, 5.5 mM, or 11 mM had no influence on growth of L. helveticus CNZ32, as judged by 24-h yields (data not shown). However, additional experiments suggested that supplementation with cheese-derived phosphopeptides increased the specific growth rate of L. helveticus CNRZ32 (Table (Table77).

TABLE 7.
Growth rates of L. helveticus CNRZ32 in a chemically defined medium supplemented with different compounds

DISCUSSION

The growth of L. helveticus in milk involves a number of physiological activities, including lactose fermentation, casein hydrolysis, and amino acid metabolism, which are also known to influence flavor development in bacteria-ripened cheeses (12).

Carbohydrate and lipid metabolism.

Since lactose is a major source of carbohydrate for the growth of lactobacilli in milk, the upregulation of genes for lactose permease (lacS) and β-galactosidase (lacM) was expected (11) and observed (Table (Table2).2). Similarly, the presence of 2% glucose in MRS medium likely explains the observed upregulation of genes for its correspondent transporter (glcU), glucose-induced repressor (ccpA), triosephosphate isomerase (tpiA), and central glycolytic gene regulator (ygaP/cggR). In Bacillus subtilis, the latter gene serves as a repressor for a downstream operon that includes genes encoding five enzymes of the glycolytic pathway: glyceraldehyde 3-phosphate dehydrogenase (gapA), phosphoglycerate kinase (pgk), triosephosphate isomerase (tpi), phosphoglycerate mutase (pgm), and enolase (eno) (7). This cluster has a similar structure (ygaP/cggR-gapA-pgk-tpi) in L. helveticus CNRZ32 and Lactobacillus delbrueckii subsp. bulgaricus (4), which suggests that YgaP/CggR also acts as a negative regulator for gapA expression in these organisms when no inducer is present. However, the microarray data shows small but significant upregulation of all genes of the cluster, gapA included, during the growth of L. helveticus CNRZ32 in MRS. An additional examination of the tiled microarray data also revealed a change in the expression of pgk under these conditions (R = 1.17; P = 0.0026). These results indicate that the entire gapA operon in L. helveticus CNRZ32 is derepressed in the presence of sugars, and similarly to B. subtilis, an mRNA processing event may be involved to provide for differential synthesis of the encoded proteins (7).

In L. helveticus, a putative gapA operon does not include pgm and eno. Instead, pgm is physically linked to the serine metabolism genes serA and serC and all three genes are strongly upregulated in milk (Fig. (Fig.22 and and3).3). The expression level of fbaA, which encodes fructose-1,6-biphosphate aldolase in MRS-grown cells, remained unchanged (R = 1.03; P = 0.5550). Unchanged or decreased expression of genes encoding the enzymes of the earlier steps of glycolysis (i.e., fbaA, gapA, and pgk) in the milk-grown culture suggests that the upregulation of pgm may be triggered by the concentration of its substrate, 3-phosphoglycerate. This may be due to the increased expression of both serA (encodes 3-phosphoglycerate dehydrogenase) and the glycerate kinase gene glyK (Fig. (Fig.2).2). Enolase catalyzes the next enzyme of glycolysis (Fig. (Fig.2),2), and the correspondent gene, eno, was downregulated in the milk-grown culture (R = 0.88; P = 0.0088). If we assume the presence of specificity of GlyK to 2-phosphoglycerate as well, the upregulation of glyK suggests that the theoretically increased pool of 2-phosphoglycerate may be converted into d-glycerate that is further utilized in biosynthetic processes, e.g., l-alanine production (Fig. (Fig.2).2). Alternatively, flux through residual enolase activity is also possible. The tiled microarrays also identified the induction of the gene glpF for a predicted glycerol uptake facilitator protein of the aquaporin family which provides for energy-independent reversible diffusion of glycerol (27). Milk is not an apparent source of glycerol, but it may be a product of milk triacylglycerol hydrolysis by lipase(s). Coincidently, the microarray data revealed the upregulation of a lip(con hyp069A1) gene which may encode triacylglycerol lipase in milk-grown culture. Simultaneous upregulation of both glpF and lip may be considered an indication that L. helveticus CNRZ32 is able to utilize milk glycerol-containing lipids. The potential pathways for the further conversion of glycerol, such as to d-glycerate via d-glyceraldehyde, should be subjected to further investigation.

FIG. 2.
Map of selected metabolic pathways in Lactobacillus helveticus CNRZ32. The different enzymatic steps are represented by the correspondent gene designations. Dotted arrows indicate several enzymatic steps or the absence of data regarding a particular gene. ...
FIG. 3.
Schematic representation of a locus containing genes of serine metabolism in Lactobacillus helveticus CNRZ32. Abbreviations: serA, gene encoding phosphoglycerate dehydrogenase; serC, gene encoding phosphoserine aminotransferase; pgm, gene encoding phosphoglycerate ...

Proteolysis.

Since L. helveticus CNRZ32 has multiple amino acid auxotrophies (6), the rapid growth of this bacterium in milk requires several enzymes to procure essential amino acids from casein. These include extracellular, cell envelope-associated proteinases to hydrolyze caseins; transport systems to take up liberated oligopeptides and free amino acids; and intracellular peptidases to further degrade oligopeptides into smaller peptides (endopeptidases) and free amino acids (aminopeptidases) (5). As expected, data from annotated arrays that showed growth in milk raised the expression level of genes for one previously described cell envelope-associated serine proteinase, PrtH (32), and for an additional hypothetical PrtH2 protease as well as a gene for the PrtM protein that is presumably required for proteinase maturation. Tiled arrays permitted the identification of two novel cell wall-associated serine proteinases (PrtH3 and PrtH5) that were also upregulated in milk-grown culture. The presence and expression of several cell wall proteinases in L. helveticus CNRZ32 may possibly reflect their different specificities toward α- and β-caseins as previously suggested for another L. helveticus strain (16). Milk-grown cells also showed significantly higher expression levels of genes for oligopeptide transport as well as a variety of intracellular peptidases. One of the peptidase genes induced in milk, pepO2, encodes an endopeptidase with postprolyl specificity that contributes to the ability of L. helveticus CNRZ32 to hydrolyze bitter peptides in cheese (37). The array data also confirm previous results regarding the importance of the peptidases PepN, PepX, PepV, PepR, and PepT for adequate growth of L. helveticus in milk (6).

In contrast, the upregulation of genes for components of the Dtp system for di-/tripeptide transport and peptidases PepI, PepQ, and PepD during the growth of L. helveticus CNRZ32 in MRS likely reflects an abundance of small, proline-rich peptides in this medium. Growth in MRS also led to the upregulation of genes for several stress-related and housekeeping proteases.

Amino acid metabolism.

The conversion of amino acids into volatile and nonvolatile compounds by LAB in cheese is thought to represent the rate-limiting step in the development of mature flavor and aroma (8, 46, 50). However, the growth of L. helveticus CNRZ32 in milk had little effect on the expression of genes for amino acid biosynthesis, presumably due to the ready supply of amino acids in this medium. An exception was the ATase2/nifS2 gene, whose product may possess dual activity and function as a cysteine desulfurase (EC 2.8.1.7), which catalyzes the hydrolysis of cysteine to alanine, and as a class V aminotransferase, which may participate in the catabolism of Leu and Val.

Draft sequencing data and array results from milk-grown L. helveticus CNRZ32 identified a hypothetical pathway for serP utilization in L. helveticus that involves the uptake of serP as a free amino acid or, more likely, within phosphorylated peptides (Fig. (Fig.2).2). Once inside the cell, free serP could be produced by the action of peptidase(s), followed by conversion to serine by phosphoserine phosphatase (SerB; EC 3.1.3.3) or transaminated by phosphoserine aminotransferase (SerC; EC 2.6.1.52) to 3-phosphohydroxypyruvate. The latter can be converted through the action of 3-phosphoglycerate dehydrogenase (SerA; EC 1.1.1.95) into a compound that feeds into the central glycolytic pathway, 3-phosphoglycerate, which may then be converted to 2-phosphoglycerate by phosphoglycerate mutase (Pgm; EC 5.4.2.1) (Fig. (Fig.22).

As shown in Fig. Fig.3,3, serA, serC, and pgm are located in close proximity with each other, and microarray data showed that all three genes are significantly upregulated during the growth of L. helveticus in milk. Additionally, the sequence and microarray data indicated the presence of another upregulated ORF located further downstream of pgm. This ORF, named hyp1677-2, could encode a protein similar to a number of membrane-associated serine/threonine protein kinases and membrane translocase proteins. In contrast, the expression of the most likely ortholog for phosphoserine phosphatase (serB), designated serB/ycsE, was not changed during the growth of L. helveticus CNRZ32 in milk (R = 1.05; P = 0.2571). The upregulation of serA, serC, and pgm in milk-grown cultures of L. helveticus CNRZ32, and an unchanged level of serB/ycsE suggest that this strain may actively catabolize phosphoserine (Fig. (Fig.2).2). Previous studies have suggested that serine, but not serP, is metabolized (deaminated) by the strains Lactobacillus casei and Lactobacillus plantarum (23, 26). Based on these data, Liu et al. (26) speculated that serP cannot be metabolized unless it is converted to serine by action of phosphatase. However, the data presented here indicate that in L. helveticus, the metabolism of serP may also proceed via an alternative pathway which does not involve consecutive dephosphorylation and deamination. Preliminary experiments suggest that a phosphopeptide fraction isolated from 10-month-old Herrgard cheese enhanced the growth of L. helveticus CNRZ32 in CDM. This fraction contains short (<10 amino acid residues) phosphopeptides (28) which should be readily transported by the Opp system and serve as substrates for intracellular peptidase(s). However, the additional investigation is required to establish whether the observed growth stimulation property of phosphopeptides is due to the presence of serP residues.

Citrate metabolism.

Citrate may serve as an additional source of metabolic energy for LAB in milk (13). Citrate catabolism is also associated with the production of important cheese flavor compounds, such as succinate, diacetyl, acetaldehyde, and acetoin (12, 17).

An analysis of the CNRZ32 genome established that genes for citrate metabolism are organized in an operon-like structure similar to those found in other LAB, such as Lactococcus lactis (31) and Leuconostoc spp. (29). The citrate operon in L. helveticus CNRZ32 includes five ORFs, citRCDEF, which are predicted to encode a transcriptional regulator, citrate lyase ligase, and γ-, β-, and α-subunits of citrate lyase, respectively. Unlike the citrate operons of Lactococcus lactis and Leuconostoc spp., however, the L. helveticus CNRZ32 cluster is preceded by a codirectional ORF (yflS) predicted to encode a C4-dicarboxylate transporter with 31% identity over 495 amino acids to an experimentally characterized citrate carrier protein, CitT, from Escherichia coli (34). While yflS/citT was upregulated during the growth of L. helveticus CNRZ32 in milk, all genes of the citRCDEF cluster except citR were induced in MRS-grown culture. In other LAB, citrate operon transcription is regulated by a transcriptional activator, CitI, whose gene is located immediately upstream of citC, and is induced by citrate and acidic conditions (30, 31). The citI gene in L. helveticus CNRZ32 is found at a distal locus and, like yflS/citT but unlike citCDEF, was upregulated in milk-grown culture. The pH values of the milk- and MRS-grown cultures of L. helveticus CNRZ32 in our experiments were similar, but because citrate concentrations were not measured in milk used in the present study, it is not clear whether the expression of citrate metabolism genes in this organism may have been affected by differences in medium citrate concentrations.

In Lactococcus lactis and Leuconostoc spp., two other genes, citG and citX, which encode enzymes required for the activation of citrate lyase, are adjacent to and cotranscribed with the genes for citrate lyase ligase and citrate lyase (30, 31). However, in L. helveticus CNRZ32, these genes are found in separate loci and are expressed in a dissimilar manner; citG is upregulated in milk-grown cells, and citX is not differentially expressed (data not shown). This suggests that the situation with L. helveticus CNRZ32 may be analogous to that with Klebsiella pneumonia in that it employs distinct regulatory mechanisms for citG and citX (36).

Growth in MRS also induced the expression of genes that participate in the conversion of oxaloacetate to succinate: aspC (aspartate aminotransferase) and two of at least three paralogs of frdC (fumarate reductase). At the same time, no upregulation was recorded for gabD, which encodes succinic semialdehyde dehydrogenase (EC 1.2.1.16), an enzyme that further converts succinate to succinate semialdehyde. Taken together, these data suggest that L. helveticus has a functional pathway of citrate metabolism, possibly leading to succinate production. Indeed, a recent report by Torino et al. (42) demonstrates that L. helveticus can metabolize citrate into succinate, an important cheese flavor compound.

Nucleotide metabolism.

Milk is a poor source of purine bases available for utilization by microorganisms (15). Microarray data are in good agreement with reports which have shown that purine biosynthesis is required for the growth of LAB in milk (21). As expected, the growth of L. helveticus CNRZ32 in milk led to the induction of two operons, purCQLFMNHD and purAB, whose products are required for the conversion of 5′-phosphoribosyl-1-pyrophosphate to inosine-5′-phosphate (IMP), a central intermediate in purine anabolism (21). An important cofactor in the IMP biosynthesis, 10-formyl-tetrahydrofolate, is produced from formate and tetrahydrofolate by formate-tetrahydrofolate synthase whose correspondent gene, fhs, was also upregulated in milk-grown culture. Genes that encode enzymes for further conversion of IMP to AMP (purA and purB), XMP (guaB), or GMP (guaA) were also upregulated in milk-grown cells.

The tiled microarrays also showed strong upregulation of genes encoding xanthine and hypoxanthine/guanine transporters (pbuX and pbuG, respectively) as well as xpt and hpt, whose products catalyze the conversion of these compounds into GMP, XMP, and IMP. Two additional purines that salvage/interconversion genes, adD (adenosine deaminase) and ptd (purine transdeoxyribonuclease) (19), were also strongly induced in the milk culture. These data suggest L. helveticus CNRZ32 relies on both salvage and de novo synthesis pathways to obtain purine nucleotides during the growth in milk.

In contrast, the expression levels of genes for pyrimidine biosynthesis were not significantly different in milk- and MRS-grown cultures. An exception was pyrD, which was overexpressed in milk-grown cultures. It seems to reflect a direct influx of orotate present in milk (40). The tiled microarrays, however, identified three other genes that were overexpressed in the milk-grown cultures, cdc8/cmk, rtpR, and nrdD, that were predicted to encode thymidylate/cytidylate kinase; ribonucleotide-triphosphate reductase, class II (3); and ribonucleoside-diphosphate reductase, class III (43), respectively, enzymes for the pyrimidine salvage pathway. These results suggest that L. helveticus CNRZ32 may rely more on rescuing existing pyrimidine compounds than on de novo synthesis during the growth in milk.

Phage genes.

During cheese ripening, the lysis of cheese bacteria, such as L. helveticus, is considered to be a beneficial process because it facilitates the release of intracellular peptidases which then participate in proteolysis (1). Cells of L. helveticus are known to undergo extensive lysis during the early stages of Swiss cheese molding and ripening (45), and prophage activation may explain this observation (25). Microarray results from this study showed that the growth of L. helveticus CNRZ32 in milk was accompanied by the upregulation of a large number of phage-related genes (Table (Table5).5). Since genes for phage lysin were not identified in this group, however, further work is needed to confirm a role for prophage induction in the autolysis of L. helveticus CNRZ32.

Acknowledgments

This study was supported by Dairy Management, Inc., through the Wisconsin Center for Dairy Research and the Western Dairy Center, and by Chr. Hansen, Inc.

The help of Brian Yandell, Tom Tabone, and Mateo Budinich, University of Wisconsin—Madison, and Michael Hogan, NimbleGen, Inc., with statistical analysis is highly appreciated.

Footnotes

[down-pointing small open triangle]Published ahead of print on 23 February 2007.

REFERENCES

1. Bartels, H. J., M. E. Johnson, and N. F. Olson. 1987. Accelerated ripening of Gouda cheese. 2. Effect of freeze-shocked Lactobacillus helveticus on proteolysis and flavor development. Milchwissenschaft 42:139-144.
2. Bolstad, B. M., R. A. Irizarry, M. Astrand, and T. P. Speed. 2003. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19:185-193. [PubMed]
3. Booker, S., and J. Stubbe. 1993. Cloning, sequencing, and expression of the adenosylcobalamin-dependent ribonucleotide reductase from Lactobacillus leichmannii. Proc. Natl. Acad. Sci. USA 90:8352-8356. [PMC free article] [PubMed]
4. Branny, P., F. de la Torre, and J. R. Garel. 1998. An operon encoding three glycolytic enzymes in Lactobacillus delbrueckii subsp. bulgaricus: glyceraldehyde-3-phosphate dehydrogenase, phosphoglycerate kinase and triosephosphate isomerase. Microbiology 144:905-914. [PubMed]
5. Christensen, J. E., E. G. Dudley, J. A. Pederson, and J. L. Steele. 1999. Peptidases and amino acid catabolism in lactic acid bacteria. Antonie Leeuwenhoek 76:217-246. [PubMed]
6. Christensen, J. E., and J. L. Steele. 2003. Impaired growth rates in milk of Lactobacillus helveticus peptidase mutants can be overcome by use of amino acid supplements. J. Bacteriol. 185:3297-3306. [PMC free article] [PubMed]
7. Doan, T., and S. Aymerich. 2003. Regulation of the central glycolytic genes in Bacillus subtilis: binding of the repressor CggR to its single DNA target sequence is modulated by fructose-1,6-bisphosphate. Mol. Microbiol. 47:709-1721. [PubMed]
8. Dobric, N., G. K. Limsowtin, A. J. Hillier, N. P. Dudman, and B. E. Davidson. 2000. Identification and characterization of a cystathionine beta/gamma-lyase from Lactococcus lactis ssp. cremoris MG1363. FEMS Microbiol. Lett. 82: 249-254. [PubMed]
9. Dow, G. S. 2003. Effect of sample size and P-value filtering techniques on the detection of transcriptional changes induced in rat neuroblastoma (NG108) cells by mefloquine. Malar. J. 2:4. http://www.malariajournal.com/content/2/1/4 [PMC free article] [PubMed]
10. Ferranti, P., F. Barone, L. Chianese, F. Addeo, A. Scaloni, L. Pellegrino, and P. Resmini. 1997. Phosphopeptides from Grana Padano cheese: nature, origin and changes during ripening. J. Dairy Res. 64:601-615. [PubMed]
11. Fortina, M. G., G. Ricci, D. Mora, S. Guglielmetti, and P. L. Manachini. 2003. Unusual organization for lactose and galactose gene clusters in Lactobacillus helveticus. Appl. Environ. Microbiol. 69:3238-3243. [PMC free article] [PubMed]
12. Fox, P. F., and J. M. Wallace. 1997. Formation of flavor compounds in cheese. Adv. Appl. Microbiol. 45:17-85. [PubMed]
13. Fox, P. F., P. L. H. McSweeney, and C. M. Lynch. 1998. Significance of non-starter lactic acid bacteria in cheddar cheese. Aust. J. Dairy Technol. 53:83-89.
14. Hassan, A. N., and J. F. Frank. 2001. Starter cultures and their use, p. 151-207. In E. M. Marth and J. L. Steele (ed.), Applied dairy microbiology. Marcel Dekker, New York, NY.
15. Hebert, E. M., G. S. De Giori, and R. R. Raya. 2001. Isolation and characterization of a slowly milk-coagulating variant of Lactobacillus helveticus deficient in purine biosynthesis. Appl. Environ. Microbiol. 67:1846-1850. [PMC free article] [PubMed]
16. Hébert, E. M., R. R. Raya, and G. S. De Giori. 2002. Modulation of the cell-surface proteinase activity of thermophilic lactobacilli by the peptide supply. Curr. Microbiol. 45:385-389. [PubMed]
17. Hugenholtz, J. 1993. Citrate metabolism in lactic acid bacteria. FEMS Microbiol. Rev. 12:165-178.
18. Irizarry, R. A., B. Hobbs, F. Collin, Y. D. Beazer-Barclay, K. J. Antonellis, U. Scherf, and T. P. Speed. 2003. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4:249-264. [PubMed]
19. Kaminski, P. A. 2002. Functional cloning, heterologous expression, and purification of two different N-deoxyribosyltransferases from Lactobacillus helveticus. J. Biol. Chem. 277:14400-14407. [PubMed]
20. Kandler, O., and N. Weiss. 1986. Genus Lactobacillus, p. 1209-1234. In P. H. A. Sneath, N. S. Mair, M. E. Sharpe, and J. G. Holt (ed.), Bergey's manual of systematic bacteriology, vol. 2. Williams & Wilkins, Baltimore, MD.
21. Kilstrup, M., K. Hammer, P. R. Jensen, and J. Martinussen. 2005. Nucleotide metabolism and its control in lactic acid bacteria. FEMS Microbiol. Rev. 29:555-590. [PubMed]
22. Klaenhammer, T., E. Altermann, F. Arigoni, A. Bolotin, F. Breidt, J. Broadbent, R. Cano, S. Chaillou, J. Deutscher, M. Gasson, M. van de Guchte, J. Guzzo, A. Hartke, T. Hawkins, P. Hols, R. Hutkins, M. Kleerebezem, J. Kok, O. Kuipers, M. Lubbers, E. Maguin, L. McKay, D. Mills, A. Nauta, R. Overbeek, H. Pel, D. Pridmore, M. Saier, D. van Sinderen, A. Sorokin, J. Steele, D. O'Sullivan, W. de Vos, B. Weimer, M. Zagorec, and R. Siezen. 2002. Discovering lactic acid bacteria by genomics. Antonie Leeuwenhoek 82:29-58. [PubMed]
23. Kristoffersen, T., and F. E. Nelson. 1955. Degradation of animo acids by Lactobacillus casei and some factors influencing deamination of serine. Appl. Microbiol. 3:268-273. [PMC free article] [PubMed]
24. LeBlanc, J. G., C. Matar, J. C. Valdez, J. LeBlanc, and G. Perdigon. 2002. Immunomodulating effects of peptidic fractions issued from milk fermented with Lactobacillus helveticus. J. Dairy Sci. 85:2733-2742. [PubMed]
25. Lepeuple, A.-S., L. Vassal, B. Cesselin, A. Delacroix-Buchet, J.-C. Gripon, and M.-P. Chapot-Chartier. 1998. Involvement of prophage in the lysis of Lactococcus lactis subsp. cremoris AM2 during cheese ripening. Int. Dairy J. 8:667-674.
26. Liu, S.-Q., R. Holland, P. McJarrow, and V. L. Crow. 2003. Serine metabolism in Lactobacillus plantarum. Int. J. Food Microbiol. 89:265-273. [PubMed]
27. Lu, D., P. Grayson, and K. Schulten. 2003. Glycerol conductance and physical asymmetry of the Escherichia coli glycerol facilitator GlpF. Biophys. J. 85:2977-2987. [PMC free article] [PubMed]
28. Lund, M., and Y. Ardo. 2004. Purification and identification of water-soluble phosphopeptides from cheese using Fe(III) affinity chromatography and mass spectrometry. J. Agric. Food Chem. 52:6616-6622. [PubMed]
29. Martín, M., M. A. Corrales, D. de Mendoza, P. Lopez, and C. Magni. 1999. Cloning and molecular characterization of the citrate utilization citMCDEFGRP cluster of Leuconostoc paramesenteroides. FEMS Microbiol. Lett. 174:231-238. [PubMed]
30. Martín, M., C. Magni, P. Lopez, and D. de Mendoza. 2000. Transcriptional control of the citrate-inducible citMCDEFGRP operon, encoding genes involved in citrate fermentation in Leuconostoc paramesenteroides. J. Bacteriol. 182:3904-3912. [PMC free article] [PubMed]
31. Martín, M. G., P. D. Sender, S. Peiru, D. de Mendoza, and C. Magni. 2004. Acid-inducible transcription of the operon encoding the citrate lyase complex of Lactococcus lactis biovar diacetylactis CRL264. J. Bacteriol. 186:5649-5660. [PMC free article] [PubMed]
32. Pederson, J. A., G. J. Mileski, B. C. Weimer, and J. L. Steele. 1999. Genetic characterization of a cell envelope-associated proteinase from Lactobacillus helveticus CNRZ32. J. Bacteriol. 181:4592-4597. [PMC free article] [PubMed]
33. Pedra, J. H., L. M. McIntyre, M. E. Scharf, and B. R. Pittendrigh. 2004. Genome-wide transcription profile of field- and laboratory-selected dichlorodiphenyltrichloroethane (DDT)-resistant Drosophila. Proc. Natl. Acad. Sci. USA 101:7034-7039. [PMC free article] [PubMed]
34. Pos, K. M., P. Dimroth, and M. Bott. 1998. The Escherichia coli citrate carrier CitT: a member of a novel eubacterial transporter family related to the 2-oxoglutarate/malate translocator from spinach chloroplasts. J. Bacteriol. 180:4160-4165. [PMC free article] [PubMed]
35. Saito, Y., Y. Hamanaka, K. Saito, S. Takizawa, and Y. Benno. 2002. Stability of species composition of fecal bifidobacteria in human subjects during fermented milk administration. Curr. Microbiol. 44:368-373. [PubMed]
36. Schneider, K., C. N. Kastner, M. Meyer, M. Wessel, P. Dimroth, and M. Bott. 2002. Identification of a gene cluster in Klebsiella pneumoniae which includes citX, a gene required for biosynthesis of the citrate lyase prosthetic group. J. Bacteriol. 184:2439-2446. [PMC free article] [PubMed]
37. Sridhar, V. R., J. E. Hughes, D. L. Welker, J. R. Broadbent, and J. L. Steele. 2005. Identification of endopeptidase genes from the genomic sequence of Lactobacillus helveticus CNRZ32 and the role of these genes in hydrolysis of model bitter peptides. Appl. Environ. Microbiol. 71:3025-3032. [PMC free article] [PubMed]
38. Storey, J. D., and R. Tibshirani. 2003. Statistical methods for identifying differentially expressed genes in DNA microarrays. Methods Mol. Biol. 224:149-157. [PubMed]
39. Suzuki, I., S. Kato, T. Kitada, N. Yano, and T. Morichi. 1985. Growth of Lactobacillus bulgaricus in milk. 1. Cell elongation and the role of formic acid in boiled milk. J. Dairy Sci. 69:311-320.
40. Suzuki, I., S. Kato, T. Kitada, N. Yano, and T. Morichi. 1985. Growth of Lactobacillus bulgaricus in milk. 2. Characterization of purine nucleotides, pyrimidine nucleotides and nucleic acid synthesis. J. Dairy Sci. 69:971-978. [PubMed]
41. Tiemeyer, W., M. Stohrer, and D. Giesecke. 1984. Metabolites of nucleic acids in bovine milk. J. Dairy Sci. 67:723-728. [PubMed]
42. Torino, M. I., M. P. Taranto, and G. Font de Valdez. 2005. Citrate catabolism and production of acetate and succinate by Lactobacillus helveticus ATCC 15807. Appl. Microbiol. Biotechnol. 69:79-85. [PubMed]
43. Torrents, E., G. Buist, A. Liu, R. Eliasson, J. Kok, I. Gibert, A. Graslund, and P. Reichard. 2000. The anaerobic (class III) ribonucleotide reductase from Lactococcus lactis. Catalytic properties and allosteric regulation of the pure enzyme system. J. Biol. Chem. 275:2463-2471. [PubMed]
44. Ulijasz, A. T., D. R. Andes, J. D. Glasner, and B. Weisblum. 2004. Regulation of iron transport in Streptococcus pneumoniae by RitR, an orphan response regulator. J. Bacteriol. 186:8123-8136. [PMC free article] [PubMed]
45. Valence, F., S. M. Deutsch, R. Richoux, V. Gagnaire, and S. Lortal. 2000. Autolysis and related proteolysis in Swiss cheese for two Lactobacillus helveticus strains. J. Dairy Res. 67:261-271. [PubMed]
46. van Kranenburg, R., M. Kleerebezem, J. van H. Vlieg, B. M. Ursing, J. Boekhorst, B. A. Smit, E. H. E. Ayad, G. Smit, and R. J. Siezen. 2002. Flavour formation from amino acids by lactic acid bacteria: predictions from genome sequence analysis. Int. Dairy J. 12:111-121.
47. Wayne, M. L., and L. M. McIntyre. 2002. Combining mapping and arraying: an approach to candidate gene identification. Proc. Natl. Acad. Sci. USA 99:14903-14906. [PMC free article] [PubMed]
48. Williams, A. G., S. E. Withers, and J. M. Banks. 2000. Energy sources of non-starter lactic acid bacteria isolated from cheddar cheese. Int. Dairy J. 10:17-23.
49. Wolfinger, R. D., G. Gibson, E. D. Wolfinger, L. Bennett, H. Hamadeh, P. Bushel, C. Afshari, and R. S. Paules. 2001. Assessing gene significance from cDNA microarray expression data via mixed models. J. Comput. Biol. 8:625-637. [PubMed]
50. Yvon, M., and L. Rijnen. 2001. Cheese flavour formation by amino acid catabolism. Int. Dairy J. 11:185-201.

Articles from Applied and Environmental Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...