Bacterial strains, media and culture conditions
The bacterial strains and plasmids used in this study are listed in Table 1. Lc. mesenteroides and Lb. plantarum were routinely grown in MRS medium  or in chemically defined medium (CDM) (adapted by removal of folic acid, riboflavin and nucleotides) [47,48]. P. freudenreichii strains were routinely cultured in Lactate Broth (LB) containing tryptone (5 g/L), yeast extract (10 g/L), Na-acetate.3H2O (8 g/L) and Na-lactate molasses 60% (25 g/L) at 30°C. The pH of the medium was adjusted to pH 6.5 with 1N KOH before heat treatment for 15 min. at 121°C. The chemically defined medium (CDM) described by Jan et al.  was modified for the selection of roseoflavin-resistant mutants. The following ingredients were omitted: p-aminobenzoic acid, folic acid, riboflavin, L-phenylalanine, L-tryptophane, and L-tyrosine. Enumeration of propionibacteria was done on lactate broth agar plates.
Isolation of roseoflavin-resistant mutants and sequence analysis of roseoflavin-resistant mutants
Spontaneous roseoflavin-resistant mutants of Lb. plantarum and Lc. mesenteroides were isolated by plating mid logarithmic phase cells on CDM containing 100 mg l-1 roseoflavin. Isolation of chromosomal DNA from Lc. mesenteroides and Lb. plantarum was performed as described by Leenhouts et al. [50,51]. In order to identify the mutations that were the probable cause of roseoflavin resistance and riboflavin overproduction the regulatory region upstream of ribG was amplified by PCR, purified using the JETquick PCR purification kit (Genomed, Löhne, Germany) and subjected to sequence analysis (MWG Biotech AG, Ebersberg, Germany). Spontaneous roseoflavin resistant mutants of P. freudenreichii strains were isolated by plating cultures on adapted chemically defined medium (see above) on solid medium made of the same chemically defined medium supplemented with 1% agar and 10 mg/L roseoflavin. After prolonged incubation at 30°C, roseoflavin-resistant colonies were isolated from the plates and further cultured on liquid chemically defined medium supplemented with 10 mg/L roseoflavin. Fully grown cultures were subsequently diluted in fresh medium in which the roseoflavin concentration was stepwise increased from 50 mg/L, then to 100 mg/L and finally to 200 mg/L. From the final culture with 200 mg/L roseoflavin, single colonies were isolated, and analysed for riboflavin production.
In Lc. mesenteroides and Lb. plantarum potential RFN elements were identified using RFAM, which is a collection of multiple sequence alignments and covariance models representing non-coding RNA families [43,52].
Quantitative analysis of riboflavin in culture medium
Extracellular riboflavin concentrations were measured by reverse phase HPLC. A Ultrasphere RP 4.6 mm × 25 cm column (Beckman Coulter, Fullerton, CA) was used and riboflavin was eluted with a linear gradient of acetonitrile from 3.6% to 30% at pH 3.2. Fluorescent detection was used and the excitation and emission wavelengths were 440 and 520 nm, respectively. Commercially obtained riboflavin and FMN were used as references and to obtain a standard curve for quantitative purposes (Sigma, Steinheim, Germany).
Stability study of riboflavin overproducing phenotype in P. freudenreichii
P. freudenreichii B374 and its riboflavin-overproducing derivatives were subcultured in lactate broth lacking roseoflavin for 60 generations and extracellular riboflavin levels were measured every 10 generations to determine the stability of the riboflavin-overproducing phenotype.
Yoghurt study using P. freudenreichii NIZO B374 and its riboflavin-overproducing derivative P. freudenreichii NIZO B2336
Two different methods were applied to produce yoghurt with propionibacteria. First the pasteurized milk was fermented by simultaneous addition of a traditional yoghurt starter culture NIZO S737 (inoculation level of 0.2%) and Propionibacterium freudenreichii NIZO B374 (wildtype) or P. freudenreichii NIZO B2336 (riboflavin-overproducing derivative of strain NIZO B374). The inoculated yoghurt milk was subsequently incubated for 16 hours at 30°C. P. freudenreichii NIZO B374 or NIZO B2336 were added at various inoculum levels (0, 1, 2, 5, 10 or 20%). Secondly, the pasteurized milk was pre-fermented for 24 hours at 30°C with the propionibacteria (inoculum levels of 0, 1, 2, 5, 10 or 20%), before addition of the yoghurt starter culture. The second phase of fermentation was again 16 hours at 30°C. At the end of fermentation the pH and viable counts were determined and the riboflavin content of the yoghurt was measured.
Nucleotide sequence accession numbers
The nucleotide sequence data of the 5' leader regions upstream of the riboflavin biosynthesis operons in Lc. mesenteroides NCDO 2028 and Lb. plantarum NCDO 1752 reported in this paper have been submitted to the Genbank database under accession numbers DQ645591 and DQ645592 respectively.