We assumed that precision of the affiliation and enumeration of lactic acid bacteria in sample were depend on a ratio of lactic acid bacteria to the other bacterial groups in the sample. Komekouji-miso was selected as a representative sample including lower ratio of lactic acid bacterial group, kimuchi was selected as a representative sample including both lactic acid bacterial group and the other bacterial group, and yoghurt was selected as a representative sample including higher ratio of lactic acid bacterial group.

Commercial products of yoghurt A, yoghurt B, yoghurt C, komekouji-miso, and kimuchi directly imported from Korea were used. For MPN, serial 10-fold dilutions (10⁻² to 10⁻⁷) prepared from sample (1 g fresh wt.) were inoculated to test vials (3 replicates) including MRS medium (Difco, Sparks MD). As to komekouji-miso (A), microbial DNA in each vial was extracted after 3 days (MC3) and 7 days (MC7) of incubation at 30˚C. Microbial DNAs of the other samples (yoghurt A; YAC, yoghurt B; YBC, yoghurt YC; CC, kimuchi (KC)) were extracted after 3 days of the incubation. As yoghurt B, microbial DNA was directly extracted without cultivation (YBU) after neutralization of pH, and removal of casein. Each bacterial group was counted by MPN after the phylogenetic estimations.

Chromosomal DNAs of each MPN vials were prepared as described previously and purified by conventional methods. Amplification of 16S rDNA was according to the former study [15] - [17] using the V2 forward primer (41f), and the V6 reverse primer (1066r) [18] [19] . PCR product (10 μl) was separately digested by each of 10 units of the restriction enzyme, Hae III or Hha I or Rsa I or Alu I (Takara Bio Co. Ltd. Shiga Japan) in Low salt buffer solution (10× Low salt buffer, Takara Bio Co. Ltd.). Fragment lengths were measured by microchip electrophoresis system (MCE-202 MultiNA; Shimadzu Co., Ltd. Kyoto Japan). The sample was diluted by de-io- nized water (5 folds for Low salt buffer) before measuring by MCE-202.

The newly constructed database was used for this research, which was edited using the method of Watanabe and Okuda [1] described previously [15] . For 41f/1066r primers, 30,844 post-amplification sequence files, which were consisted from 1379 bacterial genera, including uncultured and unidentified bacteria, were mainly re- edited using small subunit rRNA files in RDP II release 9_61 [20] under 5-bases mismatches in the both in primer annealing sites. The number and diversity of the registered MERFLs in the database greatly increased from the former database, 4370 MERFLs, and 576 bacterial genera, using small subunit rRNA files in RDP II release 7.01 [21] , and included the following lactic acid bacteria, 307 MERFLs of Streptcoccus spp., 68 MERFLs of Bifidobacterium spp., 416 MERFLs of Lactobacillus spp., 105 MERFLs of Enterococcus spp., 46 MERFLs of Lactococcus spp., and 31 MERFLs of Leuconostoc spp.

As each MPN vials included multi-template DNAs originated from heterogeneous bacteria, most of the measured MERFL was the mixed MERFLs digested from the heterogeneous 16S rDNA. Whereas all the theoretical MERFLs were originated from the homogeneous 16S rDNA sequence, the measured MERFL digested from the homogeneous 16S rDNA was selected as described previously [2] [3] .

The restriction fragments (RFs) with the highest relative mole concentration (ratio of fluorescent intensity to fragment size) were selected and used as the major RFs (represented as H in Table 1). After subtraction of the above the major RFs from the mixed heterogeneous RFs, RFs originated from the 2nd major gene were similarly selected and used for similarity search (represented as M in Table 1).

The similarity between the measured RFLP (A) and the theoretical RFLP (B) was calculated as described previously [1] [15] - [17] based on the pairwise distance (D_AB) by the following equation; D_AB =1 − 2N_AB/(N_A + N_B), where N_A and N_B were the numbers of fragments of each RFLs and N_AB was the number of shared fragments that indicated same sizes within an allowance limit for measuring error according to Nei and Li [22] . The pairwise distance of the MERFLs (D_ABME) was an average of all the D_ABs for used restriction enzymes. Similarity (%) was (1 − D_ABME) × 100 (Table 1).

If the completely identical theoretical MERFL was not found by using all of the measured MERFL data, combinations of restriction enzymes used for the analysis was changed (Table 1) [15] - [17] . As to the measured MERFL which had no completely identical theoretical MERFL, the theoretical MERFL having the highest similarity to the measured MERFL was indicated in Table 1 [2] [3] .

Most probable numbers of each groups (A-F) were estimated for three-tube, three-decimal-dilution experiment

^aGrouping was based on affiliation by MERFL; lactic acid bacteria (Group A), Lactococcus lactis (Group B), Leuconostoc spp. (Group C), Bacillus spp. (Group D), the other Firmicutes (Group E), and gram negative bacterial group (Group F). ^bThe 1st letter in vial indicates samples; “Y” stands for yoghurt followed by product items “A”, “B”, and “C”, “M” stands for kouji-miso, and “K” stand for kimuchi. The next letter indicates extraction method; “C” stand for culture-based followed by incubation period “3” stand for 3 days, and “7” stand for 7 days, and “U” stand for unculture-based as described in materials and method. Exponential of vial number represents the decimal dilution of the vial. The 2nd number of vial number (1 - 5) represents number in 5 replicates for the each decimal dilution. H of last letter represents MERFL originating from the major 16S rDNA, and M represents from the 2nd major 16S rDNA. ^cRestriction enzymes used for similarity search; “Ha”, “Hh”, “R”, and “A” stand for Hae III, Hha I, Rsa I, and Alu I. For the measured MERFLP which had no completely identical theoretical MERFLP, the theoretical MERFLP having the highest similarity using all the RFLPs was presented with the similarity as described in the materials and method. ^dSpecies name (accession number) of the theoretical MERFL having the highest similarity with the measured MERFL.

(Table 2). MPN numbers and confidence limits shown in Table 2, which were calculated by the modified of de Man [23] , were obtained using FDA’s Bacterial Analytical Manual [24] .

Fifty two MERFLs in each MPN vials were differentiation into 8 groups based on the results of the affiliation and summarized in Table 1. The major group (Group A; 18 MERFLPs) contained Lactobacillus spp., such as L. plantarum, L. sakei, L. hilgardii, L. brevis, and L. buchneri, Lactococcus spp., such as L. plantarum, L. piscium, and L. raffinolactis, and Streptococcus salivarius (Table 1). Although, MERFL of KC10⁻³1H was completely identical with that of Lactobacillus casei (AJ272201) and that of KC10⁻⁴2H was completely identical with that of Lactobacillus sakei (AB362607, AB362609, CR936503, M58829), they were included in this group (Table 1). The other lactic acid bacterial group were affiliated to be Lactococcus lactis (Group B; 8 MERFLs), and Leu- conostoc spp., such as L. gelidum, L.mesenteroides, and Lpesudomesenterides (Group C; 2 MERFLs) (Table 1). The 2^nd major group (Group D; 11 MERFLPs) contained Bacillus spp., which were further differentiated into the following 3 subgroups, B. subtilis group, B. licheniformis/sphaericus, and B. cereus group (Table 1). The other Firmicutes (Group E; 6 MERFLs) and gram negative bacterial group (Group F; 8MERFLPs) were also existed (Table 1).

In the major MERFL, represented as “H” of last letter, ratio of the MERFLs having 100% similarity to the corresponding theoretical MERFLs (62.5%) was almost as same as that of the former study (62.9%) [3] and lower than that of the previous study (90.5%) [2] . This might be caused from a difference of used incubation media as the following; As TSA medium used in this study was not a selective medium, more diverse microorganisms were proliferated in the each MPN vials than that used in the former study, which made it more difficult to select the MERFLP digested from the homogeneous 16S rDNA among the mixed MERFL.

The numerically dominant bacteria in komekouji-miso after 3 days (AC3) and 7 days (AC7) cultivation were Bacillus spp. (Group D), which included B. subtilis group and B. cereus group (3 × 10³ MPN in the both AC7 and AC3 (Table 2, Figure 1). Although their similarity to the corresponding theoretical MERFLP (75%) was lower than the others (Table 1), Group A lactic acid bacteria was only detected in AC3 (1.2 × 10³ MPN) (Table 2, Figure 1). There was not so large difference in numbers of the group E between AC7 and AC3 (0.15 × 10³ MPN in AC7 and 0.22 × 10³ MPN in AC3), and numbers of group D were 0.14 × 10³ MPN in AC7 and 0.014 × 10³ MPN in AC3 (Table 2, Figure 1).

The numerically dominant lactic acid bacteria in kimuchi (KC) after 3 day cultivation were Group A (1.4 × 10⁴ MPN, Lactobacillus sakei, and L. casei,) and Group C (l.4 × 10⁴ MPN, Leuconostoc spp.), and lower number of the other Firmicutes (6 × 10² MPN, Macrococcus spp.) was included (Table 2, Figure 1).

Lactococcus lactis (Group B) was the numerically dominant lactic acid bacteria (3 × 10⁵ MPN/g product) in yoghurt A (YAC) and yoghurt C (YCC), which also contained lower numbers of gram negative bacterial group (Group F; 6 × 10³ MPN (2%) in YCC and 1.44 × 10³ MPN (0.48%) in YAC), and YCC contained a trace number of the other lactic acid bacteria (Group A; 1.24 × 10² MPN 0.04%) (Table 3, Figure 2). In the both cultured (YBC) and uncultured (YBU) yoghurt B, the lactic acid bacteria of group A was solely detected (Table 3, Figure 2). These result indicated that yoghurt of commercial product included mainly lactic acid bacterial group. The number of YBU, which used the DNA directly extracted from the yoghurt B, was 8 times higher than that of YBC, used the DNA extracted after incubation in TSA medium (Table 3, Figure 2).