Background

Lactobacilli is the largest and more diverse group of lactic acid bacteria. It consists of a high number of species isolated from several ecological niches and reported as potential industrial and probiotic bacteria for most of them [1]. For the extreme diversity of lactobacilli species, their classification has been constantly reshuffled. At first, these species were divided into 3 groups on the basis of their phenotypic carbohydrate fermentation and optimal temperature growth [2, 3]. However, the phenotypic typing methods are not completely accurate, and it was difficult to associate the phylogeny of some lactobacilli species showing intermediate characteristics with their phenotypes [3,4,5]. Therefore, using newer molecular taxonomic methods based on genome analysis has become common among researchers with the aim to improve the classification of lactobacilli species. But methods based on genome analysis have been usually reported as time-consuming, expensive, and not always reliable [6,7,8]. On the basis of 16S rRNA gene analysis, Lactobacillus genus was first of all divided into 7 or 8 groups [9]. Then, Salvetti et al. [2] updated the classification of this genus into 15 groups of three or more species by phylogenetic analysis of 16S rRNA gene sequence. Six years later, this genus was reclassified into 18 groups using the analysis of 16S rARN phylogeny, analysis of the whole genome sequence and the analysis of amino acids percentage identity in conserved proteins [10]. In another polyphasic approach based on the analysis of overall genome-relatedness indices and metabolic or ecological properties of the organism, the taxonomic relationship between Lactobacillus species was recently re-evaluated. Today, lactobacilli group is divided into 25 genera including Lactobacillus delbrueckii group, Paralactobacillus and 23 novel genera with new nomenclature classification [1]. But the addition of new species each year will require powerful tools offering high throughput, reliable and rapid analysis. The use of nucleic acids sequences already available on nucleotide database NCBI and bioinformatics tools provide the opportunities to analyze rapidly more information of microbial species [8, 11].

The aim of this study was to develop an easy and fast method to accurately distinguish between too reliable closely species in lactobacilli group by analysing in silico at the same time many restriction digest profiles of 16S rRNA with a lot of enzymes, to search the key enzymes which give different profiles. Such approach could also give solutions to students and researchers working on lactobacilli in laboratories with limited academic resources.

Method

We download sequences of partial 16S rRNA gene of L gasseri, L johnsonii, L casei, L paracasei, L rhamonosus, L zeae, L plantarum, L paraplantarum, L pentosus, L curvatus, L graminis, and L sackei closed species listed in http://www.bacterio.net/lactobacillus.html from

GenBank (http://ncbi.nlm.nih.gov). All sequences were aligned with MAFFT program (https://mafft.cbrc.jp/alignment/server) [12]. Sequences were then subjected to a virtual restriction mapping with the pDRAW32 software to find the restriction key enzymes. In the first stage we selected enzymes which cut in maximum n − 1 and in minimum one sequence where n is the number of sequences aligned, then in the second stage, endonucleases which cut just one sequence were considered as the key enzymes.

Only closely related species showing high degree (more than 99%) of 16S rRNA gene sequences similarity and are difficult to be separated are used in this study and listed in Table 1.

Table 1 Accession number and length of partial 16S rRNA of some closed lactobacilli species
Full size table

Results

The in silico prediction of the restriction patterns of partial 16S rRNA (sequences length ranging from 506 to 673 bp) after alignment of some related closed species belonging to Lactobacillus genus and L plantarum-, L casei-, L curvatus groups were made by restriction enzymes.

Restriction fragment length polymorphysm (RFLP) of sequences of approximately 670, 619, 518, and 506 bp consistent with the partial 16S rRNA genes obtained from L paracasei, L casei, L rhamonosus, and L zeae respectively indicated different banding patterns after digestion by AlwI, BpuEI, BsgI, BsrDI, BstYI, EarI, MluCI, and NsPI (Fig. 1). L casei could be easily separated from the three others closed species in this group because no restriction was indicated for all these enzymes (Fig. 1b). However, unique restriction site were shown by AlwI, BpuEI, BsgI, BstYI, and EarI on the L paracasei gene and by NspI on L rhamnosus gene and by MIuCI on L zeae gene (Fig. 1). RFLP patterns of L paracasei yielded fragments ranging from 530-to 645 bp and resulted in a well separated band for each one (Fig. 1a). Fragments of approximately 487 and 453 bp were also observed for L rhamonous and L zeae respectively when using NspI and MIuCI endonucleases (Fig. 1c, d). On the other hand L zeae gene could be also digested by BsrDI endonuclease which yields in two fragments of 257 and 249 bp (Fig. 1c). These results indicated rapid discrimination of these four closely related species within the L casei genus by using such key endonucleases.

Fig. 1
figure 1

Virtual RFLP of a L paracasei and b L casei, c L zea, d L rhamnosus

Full size image

The RFLP patterns obtained by using both endonucleases AflI and NspI (in separate restrictions) on 16S rRNA genes fragments of L curvatus (673 bp), L sakei (512 bp), and L graminis (651 bp) can allow differentiation of these 3 species (Fig. 2). Effectively AflI restriction patterns showed a band of approximately 564 bp for L curvatus, while no digestion was noted for the two remaining species (Fig. 2a–c) and at least one NspI restriction site exists in the 16S rRNA gene of the L sakei which exhibited one fragment of 486 bp after digestion by this enzyme as shown in Fig. 2c.

Fig. 2
figure 2

Virtual RFLP of a L curvatus, b L graminisand, c L sakei

Full size image

The analysis of the MucI, the NspI, and the TspDTI PCR-RFLP in silico patterns of the 561 bp, 622 bp, and 626 bp corresponding to the partial 16S rRNA genes fragments of L planatarum, L pentosus, and L paraplantarum respectively indicated that these three closely related species were clearly differentiated (Fig. 3). L plantarum could be rapidly discriminated from L paraplantarum species by using NspI which produced distinct restriction patterns from these two species, it cleaved and generated one fragment of approximately 551 bp in the first one and not digested in the second one (Fig. 3). Analysis with MucI or TspDTI restriction enzymes produced also different restriction profiles from these two species. They showed a single recognition site for each of them in L paraplantarum and generated two bands, 584 bp and 571 bp respectively for MucI and TspDTI but no digestion was observed for both enzymes in L plantarum (Fig. 3a). Results also showed no digestion PCR products from L pentosus when using the three considered endonucleases (Fig. 3c).

Fig. 3
figure 3

Virtual RFLP of a Lplantarum, b Lparaplantarum and c L pentosus

Full size image

Others RFLP patterns corresponding to the closely related L gasseri and L johnsonii partial 16S rRNA digestion by DraI, MseI and TaqI showed unique restriction site for these three key enzymes (Fig. 4). Fragments of 443 and 444 bp were obtained after the digestion of L gasseri partial gene (583 bp) by MseI and DraI respectively (Fig. 4a, b) and one fragment of about 574 bp resulted from the digestion of L johnsonii partial gene (579 bp) by TaqI (Fig. 4b).

Fig. 4
figure 4

Virtual RFLP of a L gasseri and b L johnsonii

Full size image

Discussion

Lactobacilli is the largest and most heterogeneous group among lactic producing bacteria. It is composed of several species commonly used as starter cultures and probiotics. Due to their economical interest, the precise identification of species in this group often requires molecular identification [8]. The taxonomy of lactobacilli became clearer after the genome sequencing technologies appearance and L plantarum WCFS1 genome was the first to be sequenced [13]. 16S rRNA sequences were widely used for the first diagnostics and classification of bacterial species because extensive databases of sequences, primer sets, and enzymes for analysis of 16S rRNA length polymorphism are well established [14, 15]. However, some species within lactobacilli groups share similar 16S rRNA genes (more than 99%) and are undistinguishable on basis of their 16S rRNA phylogeny [16, 17]. The use of RFLP of 16S rRNA genes resulted in efficient discrimination of lactobacilli except for some species in L casei-, L plantarum- groups and Lactobacillus genus for which limitations were encountered specially to separate L casei from L paracasei and L plantarum from L paraplantarum [17,18,19,20]. Likewise, there were some difficulties in distinguishing L zeae from L rhamonosus and L casei or Lb gasseri from Lb johnsonii on basis of 16S rRNA phylogenetic [11, 14, 21]. For all these authors the correct choice of restriction endonucleases was suspected. Therefore, other molecular approaches like SDS-PAGE protein profiles [22], PFGE fingerprinting [18], protein-encoding genes as hsp60 marker [14, 23], and dnaK marker [16, 17] have been added to the RFLP analysis for better discrimination of lactobacilli closed species. Approaches based on analysis of recA gene, partial Tuf gene, mal gene, pepC gene, pepN gene, htrA gene, and rpoB gene were also used in cases of species sharing more than 99% 16S rRNA sequences similarities [21, 24].

As pointed out in Figs. 1, 2, 3, and 4 different RFLP patterns were obtained by the selected restriction endonucleases making it possible to distinguish clearly between L casei, L paracasei, L zeae, and L rhamnosus; between L plantarum, L paraplantarum, and L pentosus; between L gasseri and L johnsonii; and between L curvatus, L sakei, and L graminis.

L casei and L paracasei were usually confused each one to the other because of the closed relationship between many strains of L paracasei species and the L casei type strain ATCC 393 [20]. Results illustrated in Fig. 1 showed that L casei and L paracasei could be discriminated effectively on the basis of their RFLP patterns by using AlwI, BpuEI, BsgI, BsrDI, BstYI, EarI, MluCI, and NsPI among the restriction endonucleases tested in silico. It is interesting to note that real digestion by restriction enzymes resulted usually in similar fragment sizes to those in the in silico experiments [25]. However, the use of the inadequate enzymes limited some authors to distinguish L rhamnosus and L paracasei from the L casei type strain on basis of small-fragments (PCR product of approximately 295 bp) by Not1 restriction enzyme patterns and neither by using large-fragment PFGE [18] or on basis of PCR amplification and digestion products (fragments of 1500 bp in size) using AluI and MspI restriction enzymes [20]. There has been also a controversy about the classification of Lb zeae which was usually classified as a subspecies of Lb casei or Lb rhamnosus [21]. These species were differentiated from its closest neighbour only when considering other markers like Dnak-PCR RFLP/apoI [16], Tuf-PCR RFLP/HaeIII [21], or 16S-23S rDNA ITS-PCR RFLP/MseI [26]. Our results indicate that the partial 16SrRNA RFLP using both BsrDI and MluCI key enzymes are valuable method to differentiate Lb zeae although 16S rRNA gene is significantly less polymorphic than other genes because similarities are significantly higher in 16S rRNA sequences (from 98.9 to 99.9%).

Also, the discrimination between L plantarum, L pentosus, and L paraplantarum species produced ambiguous outcomes because molecular analysis of 16S rRNA polymorphism by some endonucleases was not sufficient enough to reveal significant differences [16, 18, 20]. Huang and Lee [16] noted also that species in L plantarum group were indistinguishable using HaeIII, MspI, and AluI for dnaK amplicons digestion. These two authors pointed that the crucial element in RFLP techniques is the good selection of the restriction enzymes [16]. In addition, hsp60 RFLP patterns obtainable by using both endonucleases AluI and TacI were insufficient to distinguish between L plantarum and L pentosus [23]. In our study, we showed that MucI, NspI, and TspDTI selected as key enzymes produced three different restriction profiles and distinguished these three related species.

On the other hand, comparison of AflI and NspI restriction enzyme patterns showed good species distinction between these following species of L curvatus, L sakei and L graminis. Similar to our finding (data not shown) previous in vitro restriction analysis using Hind III endonuclease discriminates L sakei from L curvatus but no data were reported for L graminis species [27]. In the present study, only restriction enzyme showing at least one sequence digestion are selected, therefore Hind III endonuclease could not be considered as key enzyme because it showed digestion in both L curvatus and L graminis (data not shown).

In the case of L gasseri and L johnsonii belonging to Lactobacillus genus, 16S rRNA gene sequence analysis is not able to reveal significant differences between these two species and their differentiation leads usually to ambiguous results using several powerful approaches like MALDI-TOF MS [26], for this reason, various molecular tools have been combined for the precise differentiation of L johnsonii from L gasseri [14, 28, 29].

In the present study, we showed that the partial 16S rRNA RFLP generated by the key enzymes DraI, MseI, and TaqI could rapidly differentiate between L gasseri and L johnsonii although their highest sequences homologies [11]. A previous study showed that these two closed species could also differentiate each one from the other on basis of ITS 16S-23S rDNA RFLP/TaqI but not with ITS RFLP/MseI [26].

Conclusion

Results of this study confirmed that in silico using key enzymes could differentiate between some closely related lactobacilli at species level. This approach could be used as an initial step for rapid and reliable classification of some lactobacilli closed species. Nonetheless, one major limitation was encountered when conducting the present study. It is clearly shown that the number of analysed sequences must be reduced to avoid confusion in selecting the key enzymes. For this reason, authors are currently working on a new technique to resolve this limit.