The evolution and population structure of Lactobacillus fermentum from different naturally fermented products as determined by multilocus sequence typing (MLST)

Multilocus sequence typing is considered to be the best method for typing isolates using the DNA sequences of selected housekeeping gene internal fragments [30]. In this study, we used the MLST method to analyze the natural diversity in L. fermentum based on the DNA sequences of 11 housekeeping gene fragments. The isolates evaluated here were isolated from a relatively large geographic area including Mongolia, four Chinese Provinces and two Autonomous regions of China, and from various naturally fermented dairy products and acidic gruel. These isolates provide the relevant information required for a better understanding of the population structure and phylogenetic relationships amongst 203 isolates of L. fermentum.

Nucleotide sequences-based methods for bacterial typing are the most unambiguous methods by which isolates of any microorganism can be identified [30, 31]. In this study, 203 L. fermentum isolates were divided into 57 STs, providing a clear indication of variability. We used the MLST method to compare nucleotide polymorphisms within regions of the 11 housekeeping gene fragments, which are under selective pressure to retain function. Housekeeping genes used in this study had a certain number of polymorphisms and hence were useful for designing the MLST protocol. A total of 131 polymorphic sites were detected in the 11 gene fragments giving a polymorphism rate of 1.72 % amongst the 7,592 nucleotides present in L. fermentum isolates. This value was higher than that seen for other LAB, such as Oenococcus oeni, which had 40 SNPs (0.99 %) amongst 4,040 base pairs sequenced [32]; and other microbes such as Aspergillus fumigatus, with 41 SNPs (1.35 %) amongst 3,038 base pairs sequenced [33]. This result also indicated that, because most of the gene fragments used in this study had high nucleotide genetic diversity, they also had a strong discriminatory ability.

The value of Pi varied from 0.00393 (murC) to 0.01421 (dnaA) for each gene fragment. Except for dnaA, most fragments showed similar nucleotide diversity of between 0.00393 and 0.00769. Our estimate of Pi was similar to those obtained for L. plantarum, which ranged from 0.0004 to 0.0072 [25] and L. delbrueckii, which ranged from 0.0051 to 0.0096 [27]. These comparisons demonstrate the relatively high nucleotide diversity in these species and in the L. fermentum isolates used in this study. This relatively high nucleotide diversity in L. fermentum is possibly as a result of recombination by natural transformation and a low frequency of mutation. In our genetic analysis, the d _N /d _S ratios of the 11 gene fragments were all less than 1, suggesting purifying selection as expected from relatively conserved housekeeping genes. In particular the values for clpX, dnaA, dnaK and recA were zero, indicating that the amino acid composition of these genes did not change. Similar results have also been found in other studies, supporting our conclusion that these housekeeping genes are all under stabilizing selection [34, 35].

The phylogenetic tree constructed from the concatenated sequences of the 11 gene fragments indicated that the 203 L. fermentum isolates formed two distinct groups (A and B); partitioning of the two groups was as a result of evolutionary changes in the clpX, dnaA, dnaK, groEL, murC, murE, pepX, pyrG, recA, rpoB and uvrC gene sequences, as was clearly visible in the N-J phylogenetic tree constructed (Fig. 1). This showed a strong evolutionary tendency for L. fermentum isolates from the same geographic areas, including Mongolia, a number of Chinese Provinces and an Autonomous region, and from the same naturally fermented products, to be similar. MLST analyses were valuable in identifying these differences in variation within the L. fermentum genome. Hence, L. fermentum isolates that were included in the same group shared similarities in genome organization. Notably, about eight of ten isolates from acidic gruel were grouped together within group A and almost all the isolates from fermented dairy products were grouped together within group B. The minimum spanning tree analysis supported the existence of two groups of isolates (Fig. 4). Although MLST analysis of 203 L. fermentum isolates indicated two groups here, we cannot exclude the possible that more complex population structures might be identified if a larger number of isolates from more diverse sources was used. In future work, MLST will be a useful tool to examine possible relationships amongst L. fermentum isolates, while simultaneously aiding the selection of industrial isolates with greatest potential for the production of fermented food.

The analysis of population structure of L. fermentum isolates indicated substantial recombination phenomena. The I _A ^S value for the 11 gene fragments from 203 L. fermentum isolates was calculated as 0.2142 (p = 0.000), which is indicative of a weak clonal population structure (i.e. linkage disequilibrium). Several studies have already shown that LAB isolated from fermented milk are structured in rather clonal populations, for example, Xu et al., (2013) investigated 12 housekeeping genes in Lactococcus lactis; the I _A ^S value was 0.3038 and indicative of a clonal population [35].

Split-decomposition analysis also provided evidence of intraspecies recombination that could play a role in generating genotypic diversity amongst isolates according to the allelic profiles of the isolates evaluated. Split graphs for individual loci indicated tree-like or network-like structures, suggesting that some genes were affected by intraspecies recombination (Fig. 2). Simultaneously, a split graph representation of the concatenated sequence of the 11 loci clearly indicated that two groups and their descents orginated from intraspecies recombination (Fig. 3). The concatenated dendogram corresponded well with the allele-based dendogram (Fig. 1). However, some small differences were found between the concatenated dendogram and the allele-based dendogram. For example, ST-25 was positioned further away from ST-28 in the latter than the former. Although intraspecies recombination occurs frequently, isolates from fermented dairy products in group B and isolates from acidic gruel in group A in both the N-J phylogenetic tree (Fig. 1) and the combined split graph (Fig. 3), seem to be clonal. These results suggest that isolates from fermented dairy products and acidic gruel may have a common recent ancestor. In addition, the minimum spanning tree analysis result confirmed the above assumption that these isolates have a common recent ancestor (Fig. 4).

The clustering of isolates by food-type origin was evident in the minimum spanning tree (Fig. 4) demonstrating strong patterns of specificity for source within the 203 L. fermentum isolates evaluated. The diversity found amongst isolates from fermented dairy products differed from that found amongst isolates from acidic gruel. Almost all of the isolates from fermented dairy products were assigned to the largest CC, CC1, whereas isolates from acidic gruel were dispersed among a larger number of STs. Although the difference between isolates from fermented dairy products and acidic gruel was significant, we considered that evolution of L. fermentum was not correlated with food-type origins. We suggest that a more simple explanation is that the ecological niches from which the isolates were sampled are very narrow (several types of dairy products and acidic gruels). As L. fermentum is a ubiquitous bacterium found in many different types of food and in animal faeces, it is likely that isolates from dairy products are only a small proportion of the natural diversity within this species, particularly as they would be selected for particular attributes associated with the fermentation of food. We also found no association between STs and geographical origin (data not shown).

MLST data are usually subdivided into nonoverlapping groups of related STs or CCs using an eBURST algorithm approach to determine the most parsimonious patterns of descent of isolates within each CC from the predicted founder [36]. Here we also used the eBURST algorithm to analyze the MLST data, and found that ST-4 inhabited a central location, and that other STs had relatives that were derived from the ubiquitous STs themselves and exhibited a limited genetic diversity, as has been found in other studies [37, 38]. The great majority of STs in CC1 are single-locus variants (SLVs) of ST-4, which strongly supported primary founders. Furthermore, CC1 to CC5 were clearly formed based on food-type origins. For example, three of the seven CCs (CC1 to CC2) consisted exclusively of isolates from fermented dairy products, with the exception of only one isolate (IMAU70056), whereas three of five CCs (CC3 to CC5) consisted exclusively of isolates from acidic gruel, with the exception of only one isolate (IMAU60167). Furthermore, we found that the L. fermentum isolates in CC1, which were almost exclusively isolated from fermented dairy products, were located centrally, with isolates from acidic gruel being distributed around CC1 and not centrally (Additional file 2: Figure S1).

The genus Lactobacillus, together with the genera Paralactobacillus and Pediococcus, is the largest group in the Lactobacillaceae and order Lactobacillales in the Firmicutes [39]. The species L. fermentum and Lactobacillus reuteri are phylogenetically closely related and are regarded together as the L. reuteri group. Since Maiden et al., (1998) first described an MLST technique for N. meningitides, MLST methods have been used to differentiate between isolates within species from the genus Lactobacillus including L. casei, L. plantarum, L. sanfranciscensis, L. delbrueckii, and L. sakei, [23–28]. To date, the MLST schemes used in these studies were distinct from each other because the isolates used in each study came from different habitats, and different housekeeping genes were selected for analysis. In these studies, 41 isolates of L. delbrueckii were assigned to 34 STs [27]; 52 isolates of L. casei (L. paracasei) were assigned to 31 STs [24]; 232 isolates of L. sakei were assigned to 116 unique STs of which all isolates had evolved into three clades, each with a unique population structure [28]. In our study, we used MLST to identify 57 STs within a L. fermentum population of 203 isolates. Our research indicated that this species had a clonal population structure and a pattern of diversity different from the previously mentioned species. Moreover, two groups, each with a unique population structure, were identified amongst L. fermentum isolates (including the reference isolate L. fermentum IFO 3956). The L. fermentum MLST schemes in this study demonstrate that MLST is a useful tool for discrimination between isolates, and furthermore provides a method to analyze the evolution and population structure of Lactobacillius species from various sources.