|Home | About | Journals | Submit | Contact Us | Français|
Currently, the genus Lactococcus is classified into six species: Lactococcus chungangensis, L. garvieae, L. lactis, L. piscium, L. plantarum, and L. raffinolactis. Among these six species, L. lactis is especially important because of its use in the manufacture of probiotic dairy products. L. lactis consists of three subspecies: L. lactis subsp. cremoris, L. lactis subsp. hordniae, and L. lactis subsp. lactis. However, these subspecies have not yet been reliably discriminated. To date, mainly phenotypic identification has been used, with a few genotypic identifications. We discriminated species or subspecies in the genus Lactococcus not only by proteomics identification using matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) but also by phenotypic and genotypic identification. The proteomics identification using differences in the mass spectra of ribosomal proteins was nearly identical to that by genotypic identification (i.e., by analyses of 16S rRNA and recA gene sequences and amplified fragment length polymorphism). The three ribosomal subunits 30S/L31, 50S/L31, and 50S/L35 were the best markers for discriminating L. lactis subsp. cremoris from L. lactis subsp. lactis. Proteomics identification using MALDI-TOF MS was therefore a powerful method for discriminating and identifying these bacteria. In addition, this method was faster and more reliable than others that we examined.
Lactococci are lactic acid bacteria (LAB) that are important contributors to the production of fermented dairy products, and some species produce antimicrobial compounds. Most species in the genus Lactococcus have been isolated from food-related sources and plants and are generally regarded as safe. Probiotic foods use these LAB, and there have been various studies of the relationship between these foods and the maintenance of human intestinal health (32). Lactococcus was first established as a genus distinct from the genus Streptococcus in 1985 (29).
Currently, six species and three subspecies in the genus Lactococcus have been validated. Lactococcus plantarum has been isolated mainly from plants; L. garvieae has been isolated from fish, animals, and milk, and L. piscium has been isolated from salmon. Lactococcus lactis is most commonly found in raw milk, cheese, and other dairy products; L. raffinolactis has been found in raw milk and cheese, and L. chunagangensis has been isolated from wastewater. Among the six species, L. lactis is considered one of the most important in food production because it is used to manufacture fermented milk, butter, and cheese. Because of this importance, the whole genomes of three strains of L. lactis—L. lactis subsp. cremoris SK11 (10), L. lactis subsp. cremoris MG 1363 (37), and L. lactis subsp. lactis IL1403 (2)—have been sequenced.
Since L. lactis was first described by Orla-Jensen in 1919 (21), there have been various classifications. To date, L. lactis has been classified into three subspecies: L. lactis subsp. cremoris, L. lactis subsp. hordniae, and L. lactis subsp. lactis. However, this classification was based on only a few phenotypic characteristics and is considered imperfect because of its inherent disadvantages of sensitivity to culture conditions or bacterial growth phase. Discriminating between L. lactis subsp. cremoris and L. lactis subsp. lactis is particularly difficult but is very important in industrial applications, because the activities of the two subspecies in cheese manufacture differ. In addition, when newly isolated bacterial strains are registered in public culture collections, these strains have to be identified and discriminated at the subspecies level. Normally, these two subspecies are identified on the basis of the following phenotypic features: (i) the ability to ferment maltose and ribose, (ii) growth in 4% NaCl (pH 9.2) at 40°C, (iii) the ability to produce ammonia from arginine, and (iv) the presence of glutamate decarboxylase activity (18-20). However, determining the results of the phenotypic identification is difficult because they are sometimes ambiguous and time sensitive, as demonstrated by the sugar fermentation tests described below, which gave different results over time. In addition, the results of phenotypic identifications in previous reports were not identical each other (9, 28, 35).
From an evolutionary viewpoint, it is reasonable to classify subspecies by using the divergence of housekeeping genes that are well preserved at the genus or species level. 16S rRNA gene sequencing is the most common technique currently used to identify species. At the subspecies level, however, 16S rRNA gene sequence identity is often very high, and these sequences therefore cannot be used for identification purposes (14, 24, 27, 36). Recently, for LAB, the partial sequences of the recA (recombinase A), pheS (phenylalanyl tRNA synthetase alpha subunit), and rpoA (DNA-directed RNA polymerase alpha chain) genes have been effectively used for species or subspecies identification (5, 7, 17), and the analysis of 16S rRNA gene sequences in combination with housekeeping gene sequences has been used to identify subspecies.
In recent years, a number of important experiments have used matrix-assisted laser desorption ionization-time-of-flight mass spectrometry (MALDI-TOF MS) for rapid bacterial identification, including clostridia (15), LAB (34), Listeria (1), mycobacteria (12), salmonellae (6), viridans group streptococci (8), and other nonfermenting bacteria (16). In these studies, MALDI-TOF MS spectra were obtained from intact cells without biomarker purification or chromatographic separation. MALDI-TOF MS is a good tool for the analysis of biopolymers because of its soft ionization, and it plays a central role in proteomic research. Because of their simplicity, speed, and accuracy, MS methods have been successfully applied to biomarker discovery and the characterization of various bacterial agents. Although DNA sequencing is the current standard for molecular characterization of bacteria, molecular methods cannot be easily applied for rapid classification and identification.
Our aim was to examine whether a proteomic approach using MALDI-TOF MS was effective for rapid bacterial identification, especially of two of the subspecies of L. lactis.
A total of 30 bacterial strains assigned as L. garvieae, L. lactis subsp. cremoris, L. lactis subsp. lactis, L. piscium, L. plantarum, and L. raffinolactis were obtained from the Culture Collection of the Yakult Central Institute (YIT; Tokyo, Japan) (Table (Table1).1). All bacterial strains were grown in MRS broth (BD Difco, Maryland) at 26°C for 24 h.
Bacterial cell lysates were prepared by using glass beads, as previously described, with some modification (34). Briefly, 4-ml portions of overnight-cultured bacterial cells were centrifuged at 14,000 × g for 3 min and washed twice in TMA-I buffer (10 mM Tris-HCl [pH 7.8], 30 mM NH4Cl, 10 mM MgCl2, and 6 mM 2-mercaptoethanol). The cells were suspended in 500 μl of TMA-I buffer and vortexed vigorously for 2 min with 0.1 g of glass beads (0.1 mm in diameter) in a FastPrep FP 120 homogenizer (Bio 101, Vista, CA) at a setting of 6.5 m/s. Then, the mixture was centrifuged at 6000 × g for 2 min, and the supernatants were subjected to MALDI-TOF MS measurement.
Cell lysate (0.5 μl) and matrix solution (0.5 μl) were mixed on a spot marked on a steel target plate and dried at room temperature. Sinapinic acid (Sigma-Aldrich, St. Louis, MO) was used as the matrix solution at a concentration of 10 mg/ml (50% acetonitrile-3% trifluoroacetic acid). The sample spotted plate was inserted into the MALDI-TOF MS apparatus; mass spectra were acquired with a Voyager DE PRO (Applied Biosystems, Foster City, CA) mass spectrometer equipped with delayed extraction and UV nitrogen laser (337 mm, 3-ns pulse width). Analyses were performed in a linear positive ion mode at accelerating voltage 25 kV, 93% grid voltage, 0.15% guide wire, an extraction delay time of 320 ns, and a low mass gate of 2,000 m/z. The mass range was set from 2,000 to 20,000 m/z. For each spectrum, 200 laser shots were collected and analyzed. The spectra were calibrated internally using the adrenocorticotropic hormone, clip 18-39 (m/z 2,465.7), and horse apomyoglobin (m/z 16,952.6) as standards.
Thirty mass spectra for each bacterial cell lysate were chosen from the high relative intensity readings. Principal component analysis (PCA) was performed by using MarkerView software (Applied Biosystems). The m/z values rounded off to the nearest multiple of 10 were set as the variable component, with relative intensity as the data component.
Genomic DNA was extracted by using benzyl chloride as previously described (39). Briefly, 500 μl of bacterial cells was washed and collected. Then, 250 μl of extraction buffer (100 mM Tris-HCl [pH 9.0], 40 mM EDTA), 500 μl of benzyl chloride, and 50 μl of 10% sodium dodecyl sulfate were added to the collected cells. The tube was vortexed and incubated at 50°C for 20 min. Then, 150 μl of 3 M sodium acetate was added, and the sample was centrifuged at 14,000 × g for 8 min. The supernatant was collected, and genomic DNA was precipitated with isopropanol.
PCR primers 8F (5′-AGAGTTTGATCMTGGCTCAG-3′; positions 8 to 27) and 520R (5′-ACCGCGGCTGCTGGC-3′; positions 531 to 517) were used to amplify the V1 to V3 variable regions of the 16S rRNA gene (3). The PCR amplification program consisted of an initial denaturation step at 94°C for 2 min; 30 cycles of 94°C for 20 s, 55°C for 20 s, and 72°C for 20 s; and a final extension step at 72°C for 3 min. The PCR-amplified 16S rRNA genes were purified with a Montage PCR filter unit (Millipore Corp., Billerica, MA) and then sequenced with an ABI Prism BigDye Terminator v. 3.1 cycle sequence kit (Applied Biosystems) on an ABI 3130xl genetic analyzer (Applied Biosystems).
Fermentation of glucose, maltose, and ribose was determined by acid formation. Overnight bacterial cultures (4 ml) at 26°C for 24 h were pelleted by centrifugation at 2,000 × g for 20 min, and the cell pellet was suspended in 2 ml of saline. Bacterial suspension (130 μl) was inoculated into the test medium (modified MRS broth, with the glucose replaced by 2% of the test substrate) supplemented with 0.005% chlorophenol red as the pH indicator, and then incubated at 26°C for 1, 4, and 7 days. The acid production in 4 ml of broth was determined by the color change in the medium from red (pH 6.8) to yellow (pH 5.4).
We used genomic information on three strains of L. lactis—L. lactis subsp. cremoris SK11 (CP000425), L. lactis subsp. cremoris MG 1363 (AM406671), and L. lactis subsp. lactis IL1403 (AE005176)—from the DNA databases of GenBank to identify the proteins characteristic of each subspecies. For all of the proteins derived from cell lysate, the use of ribosomal proteins as suitable biomarkers has been proposed by Pineda et al. (22). We agreed with this proposal, because ribosomal proteins (i) are expressed at detectable levels under all environmental conditions and in all life stages, (ii) are housekeeping proteins, and (iii) are alkaline. Moreover, it is easy to change the monovalent cation. Using the Rapid Microorganism Identification Database (RMIDb; http://www.rmidb.org/cgi-bin/index.pl), we looked for subspecies-specific biomarkers. RMIDb is a useful tool for microorganism identification using MALDI-TOF MS, because posttranslational modifications have been considered in the calculations used for the masses of the ribosomal proteins in the database.
For the discrimination of subspecies, we performed amplified fragment length polymorphism (AFLP) analysis as described previously (25), with the following modifications. Total DNA was digested with EcoRI and MseI restriction enzymes, and the DNA fragments were ligated to the following double-stranded restriction site-specific adaptors: EcoRI adaptors (5′-CTCGTAGACTGCGTACC-3′ and 5′-CATCRGACGCATGGTTAA-3′) and MseI adaptors (5′-TACTCAGGACTCAT-3′ and 5′-GAGTCCTGAGTAGCAG-3′). For preselective and selective PCR amplification, the primers EcoRI-A (5′-GACTGCATACCACCAATTCA-3′) and MseI-CA (5′-AATGAGTCCTGAGTAGCA-3′) were used. The 5′ end of the EcoRI primer was labeled with 6-carboxyfluorescein (FAM). PCR products were analyzed on an ABI Prism 3130xl genetic analyzer (Applied Biosystems) in standard fragment analysis mode and with a GENESCAN-500 LIZ size standard (Applied Biosystems). After electrophoresis, the AFLP patterns were analyzed and extracted with GeneMapper software v4.0 (Applied Biosystems). A threshold fluorescence value of 100 arbitrary units was used to eliminate background fluorescence, and DNA fragments of between 51 and 500 bp were analyzed. Bands positioned at the same length (DNA size) in different strains were assumed to be homologous and are represented as the same alleles. Bands of different sizes were treated as independent loci with two alleles (present or absent). The data were exported in binary format, with “1” for the presence of a band/peak and “0” for its absence; the data were analyzed phylogenetically with MEGA4 software (33) by using an unweighted pair-group method with arithmetic mean (UPGMA) (31) and the Pearson correlation coefficient, with a bootstrap analysis of 100 repetitions.
Parts of the recA gene of the strains were amplified by using the primers recA-f1 (5′-GYNCAAAAAGAWGGTGGNATTGC-3′) and recA-r1 (5′-CTTRAATGGTGGWGCNACYT-3′). These primers were designed by using the consensus sequences of the recA genes of Enterococcus faecium LMG 11423 (GenBank accession no. AJ621707), Enterococcus villorum LMG 12287 (AJ621718), L. lactis ATCC 19435 (EU156791), L. lactis MG 1363 (NC009004), L. lactis ML3 (M88106), Streptococcus oralis CCUG 41672 (EU156837), Streptococcus pneumoniae CCRI 15796 (EU156855), and Streptococcus suis ATCC 43765 (EU156869).
The PCR amplification program consisted of an initial denaturation step of 94°C for 2 min; 30 cycles of 94°C for 20 s, 50°C for 20 s, and 72°C for 20 s; and a final extension step at 72°C for 3 min. PCR products were purified and sequenced under the conditions used for the 16S rRNA gene.
We obtained MALDI-TOF MS spectra from the cell lysates of 30 bacterial strains (25 strains of L. lactis and 5 non-lactis strains of Lactococcus). Figure Figure11 shows the spectra of the type strains L. chungangensis YIT 12036T, L. garvieae YIT 2071T, L. lactis subsp. cremoris YIT 2007T, L. lactis subsp. lactis YIT 2008T, L. piscium YIT 2075T, L. plantarum YIT 2061T, and L. raffinolactis YIT 2062T. The majority of peaks were obtained in the range from m/z 3,000 to 10,000. Six species in the genus Lactococcus were separated clearly by PCA based on the top 30 mass spectra, i.e., the spectra of those in each strain with the highest relative intensities (Fig. (Fig.2).2). This result was identical to that of phylogenetic analysis based on the 16S rRNA gene sequences. Lactococcus garvieae was located far from the other five species, whereas the 25 strains of L. lactis were located close to each other (see Fig. S1 in the supplemental material). Examination of the y axis of the scatter plot revealed that L. raffinolactis and L. chungangensis, L. plantarum, and L. piscium were quite close to each other, as in the dendrogram derived from the 16S rRNA gene sequences (Fig. (Fig.2;2; see also Fig. S1 in the supplemental material). However, we could not discriminate the subspecies L. lactis subsp. cremoris and L. lactis subsp. lactis by PCA based on the top 30 mass spectra.
Lactococcus lactis subsp. cremoris and L. lactis subsp. lactis are difficult to identify accurately by simple phenotypic methods. To identify these subspecies, we performed three types of subspecies identification: (i) phenotypic characterization based on fermentation testing, (ii) genotypic tests based on 16S rRNA and recA gene sequencing and AFLP analysis, (iii) and proteomic analysis using the masses of the ribosomal proteins. All subspecies identifications complied with the properties of the type strains, L. lactis subsp. cremoris YIT 2007T and L. lactis subsp. lactis YIT 2008T.
Fermentation tests were performed using glucose, ribose, and maltose. The presence of chlorophenol red caused the color of the media to change between pH 5.4 and 6.8 from yellow to red. Table Table22 shows the results of the fermentation tests. All strains fermented glucose, and all strains of L. lactis subsp. lactis fermented maltose. It had been reported that all tested strains of L. lactis subsp. lactis fermented maltose and ribose, whereas all tested strains of L. lactis subsp. cremoris did not ferment either sugar (24). However, in our test, all 14 strains of L. lactis subsp. cremoris showed identical results for fermentation of maltose and ribose on day 1 of incubation, whereas six strains (YIT 2011, YIT 2012, YIT 12099, YIT 12100, YIT 12101, and YIT 12107) showed variable, type strain-specific results on days 4 and 7. In contrast, all 11 strains of L. lactis subsp. lactis showed identical fermentation results on day 7, whereas the results varied on days 1 and 4. These findings indicated that phenotypic characterization in terms of the fermentation of carbohydrates did not provide accurate data for subspecies identification, because the phenotypic feature was susceptible to changes in culture conditions and their growth phase.
To identify the subspecies of L. lactis, 16S rRNA genes from 25 strains of L. lactis were sequenced and aligned. Only five mismatches of bases in the V1 region were found. The 25 strains were classified into two groups on the basis of the pattern of the five mismatches, at positions 84, 89, 95, 97, and 100 (with respect to E. coli 16S rRNA gene numbering) (see Fig. S2 in the supplemental material). We also constructed a dendrogram based on the partial sequences of 16S rRNA genes (see Fig. S3 in the supplemental material). L. lactis strains were divided into two clusters, A and B. Eight strains in cluster A were identified as L. lactis subsp. lactis, and 17 strains in cluster B were identified as L. lactis subsp. cremoris, on the basis of the taxonomic positions of the type strains L. lactis subsp. lactis YIT 2008T and L. lactis subsp. cremoris YIT 2007T.
The recA gene has been proposed as a useful marker for inferring bacterial phylogeny and has been used successfully to identify and discriminate species of some bacterial genera (17). We constructed a phylogenetic tree based on recA gene sequences (see Fig. S4 in the supplemental material). Lactococcus lactis was again divided into two clusters, A and B. Subspecies were identified in accordance with the positions of the type strains, L. lactis subsp. cremoris YIT 2007T (cluster A) and L. lactis subsp. lactis YIT 2008T (cluster B).
In addition to the sequence analysis, we used an AFLP method to measure the extent of species diversity; this method is considered useful for grouping bacterial strains according to their relationship. We constructed a dendrogram from the AFLP band pattern (see Fig. S5 in the supplemental material). Lactococcus lactis was again divided into two clusters, A and B. Subspecies were identified in accordance with the locations of the type strains, L. lactis subsp. cremoris (cluster A) indicated and L. lactis subsp. lactis (cluster B).
The results of the subspecies identifications using the sequences of the 16S rRNA and recA genes were concordant. However, these results and the results of the AFLP method were not completely concordant. In the three analyses, strain YIT 12106 was the only exceptional case in terms of the concordance of the results: this strain was classified as L. lactis subsp. cremoris on the basis of its 16S rRNA and recA gene sequences, whereas it was classified as L. lactis subsp. lactis by the AFLP method.
This was the first MALDI-TOF MS study to attempt to discrimination of subspecies of L. lactis. First, we performed PCA using the top 30 spectra (i.e., those of the strains with the highest relative intensities) and found that this analysis provided good results for species identification in the genus Lactococcus, but it did not provide good results for L. lactis subspecies discrimination. The peak profiles from MALDI-TOF MS of the type strains YIT 2007T and YIT 2008T are shown in Fig. Fig.1.1. Peaks with relative intensities of <20% were ignored. To assign ribosomal subunit proteins to the MS spectra, we compared the masses of the ribosomal subunit proteins of genome-sequenced L. lactis strains.
Analysis of the detailed MALDI-TOF MS peak profiles revealed that a large majority of the detectable bacterial proteins were ribosomal subunit proteins. Ribosomal proteins are major components of all proteins derived from cell lysate; because they are basic proteins, they can be easily ionized and detected well by MALDI-TOF MS (34). Although the amino acid sequences of ribosomal proteins are highly conserved, slight sequence variations occur even at the subspecies and strain level.
Eleven ribosomal proteins—50S/L36, 50S/L34, 50S/L30, 50S/L32, 30S/S21, 30S/S14, 50S/S28, 50S/L35, 50S/L29, 30S/S20, and 50S/L31—were assigned from the calculated masses of the 25 L. lactis strains (Table (Table3).3). Three of these proteins—50S/L35, 30S/S20, and 50S/L31—were ribosomal subunit proteins discriminative between L. lactis subsp. cremoris and L. lactis subsp. lactis. Only three amino acids in the amino acid sequences of these ribosomal subunit proteins differed around 50S/L35 and 30S/S20 between L. lactis subsp. cremoris and L. lactis subsp. lactis, and only one differed around 50S/L31. Table Table44 compares the observed and calculated masses of the subspecies-identifying ribosomal subunit proteins 50S/L35, 30S/S20, and 50S/L31. The accuracy of the MALDI-TOF MS data was approximately ±10 Da at a molecular mass of approximately 10,000 Da. On the basis of these results, strains YIT 2002, YIT 2007T, YIT 2011, YIT 2012, YIT 12099, YIT 12100, YIT 12101, YIT 12102, YIT 12103, YIT 12104, YIT 12105, YIT 12106, YIT 12107, YIT 2014, YIT 2040, YIT 2054, and YIT 2055 could be identified as L. lactis subsp. cremoris (Table (Table5);5); and strains YIT 12108, YIT 2003, YIT 2008T, YIT 2013, YIT 2015, YIT 2053, YIT 2056, and YIT 10041 could be identified as L. lactis subsp. lactis. By peak profiling, L. lactis subsp. cremoris YIT 12108 and four strains of L. lactis subsp. lactis (YIT 2014, YIT 2040, YIT 2054, and YIT 2055) were identified as different subspecies from those assigned previously.
Since the genus Lactococcus was separated from the genus Streptococcus, methods for discriminating subspecies of L. lactis have been proposed by many researchers. However, several research groups have recognized that the phenotypic and genotypic identifications do not match (38). To our knowledge, this is the first study to have used MALDI-TOF MS for L. lactis subspecies identification. We showed that the results of a proteomics approach were matched by those of genotypic approaches. Similar results have been obtained with viridans streptococci (proteomics identification by MALDI-TOF MS has allowed accurate differentiation between S. oralis and Streptococcus mitis in terms of the identification of reference strains) (8) and with salmonellae (species and subspecies identification by whole-cell MALDI-TOF MS) (6). In our genotypic identification, we initially used the divergence of 16S rRNA and recA genes. The V1 region of the 16S rRNA gene is a well-known variable region that accumulates divergence, allowing discrimination even among closely related bacterial species. L. lactis subsp. cremoris and L. lactis subsp. lactis have been discriminated before by using the sequence of the 16S rRNA gene (23, 36). These authors used a restriction enzyme to visually recognize the differences. We obtained the same sequence results as in these studies, but we performed not only sequencing of 16S rRNA and recA genes but also AFLP to obtain more detailed data. To our knowledge, this is the first time the AFLP method has been used to discriminate L. lactis. In the proteomics approach, the majority of peaks were obtained in the range from m/z 3,000 to 10,000; this result was similar to the results for clostridia (15), Listeria (1), mycobacteria (12), viridans group streptococci (8), and other nonfermenting bacteria (16).
We also showed that phenotypic features based on sugar fermentation testing were not useful for subspecies identification because they varied with culture time and were sometimes ambiguous. In addition, the phenotypic markers often provided discordant results, as previously reported (9, 28, 35).
We confirmed that the proteomics approach was the most effective for identification of species and subspecies of L. lactis and that the approach produced nearly the same results as did molecular techniques, such as phylogenetic analysis based on housekeeping gene sequences or AFLP profiles. This method is attractive because of its speed, accuracy, simplicity, and low cost. For the ribosomal subunit protein 50S/L31 of L. lactis, there was only one amino acid mismatch between L. lactis subsp. cremoris and L. lactis subsp. lactis, but this mismatch made a 16-Da difference that could be distinguished specifically by MALDI-TOF MS. Many bacteria, including Listeria (1), viridans streptococci (8), mycobacteria (12), and salmonellae (6), have thus far been identified by using MALDI-TOF MS. However, for the identification of bacteria with unknown genomes, looking for biomarkers is difficult. Fortunately, we have developed a rapid and reliable way of using MALDI-TOF MS to identify subspecies of L. lactis on the basis of the whole-genome analysis of Lactococcus strains. In future it will be necessary to establish a way of looking for subspecies-identifying biomarkers. Through peak profiling, L. lactis subsp. cremoris YIT 12108 and four strains of L. lactis subsp. lactis—YIT 2014, YIT 2040, YIT 2054, and YIT 2055—were identified as subspecies different from those assigned previously. The genome-sequenced strain of L. lactis subsp. cremoris MG 1363 is known as a derivative of L. lactis subsp. lactis NCDO 712 (NCIMB 700712), from which it differs with respect to plasmid or prophage (4, 10). The strains of MG 1363 had been classified as L. lactis subsp. lactis, but some research groups have proposed that it should be reclassified as L. lactis subsp. cremoris (11, 13, 27, 30). Therefore, reclassifying YIT 2054 (NCIMB 700712) as L. lactis subsp. cremoris would be a reasonable step. From the evolutionary evidence revealed by these genotypic and proteomics approaches, we propose that three strains of L. lactis—YIT 12108 (NCIMB 702006), YIT 2054 (NCIMB 700712), and YIT 2055 (NCIMB 700763)—registered in public culture collections should be reclassified.
We thank H. Ikemura (Yakult Central Institute, Japan) for support performing PCA and helpful advice.
Published ahead of print on 16 April 2010.
†Supplemental material for this article may be found at http://aem.asm.org/.