Genomic diversity and versatility of Lactobacillus plantarum, a natural metabolic engineer

1The lactobacilli constitute a major group of the Lactic Acid Bacteria (LAB). They occupy a wide range of niches and are generally found in environments with high levels of carbohydrates, such as food products (dairy products, fermented meat, sourdoughs) as well as (fermenting) plant-derived substrates. In addition, they occupy different niches on and in the human body including the respiratory, gastrointestinal and urogenital tract. As a consequence, lactobacilli have been studied extensively, initially mainly because of their importance for food production. More recently, there is a rapid increase in literature focusing on their occurrence and activity in the human microbiota as well as their use as probiotics, defined as ‘‘live microorganisms which when administered in adequate amounts confer a health benefit on the host’’ (http://www.who.int).

2The genus Lactobacillus belongs to the phylum Firmicutes, class Bacilli, order Lactobacillales, and family Lactobacillaceae. A comprehensive review of the taxonomy of lactobacilli summarizes the current taxonomic as well as historic changes. As for many other genera, the taxonomy of lactobacilli has been subject to several changes since the emergence of molecular technologies. At the moment of writing of this review (April 15, 2011), 173 species are recognized, and after removing synonymous species names due to reclassification this can be reduced to 141 species. http://www.bacterio.cict.fr/l/lactobacillus.html.

3Many Lactobacillus species are highly specialized and are only found in a limited number of niches. A well-known example is the species Lactobacillus delbrueckii, which is highly adapted to the dairy environment and widely applied in yoghurt manufacture. Other species, such as Lactobacillus acidophilus, Lactobacillus johnsonii, Lactobacillus reuteri and Lactobacillus rhamnosus are typical inhabitants of the GI tract, and are used in probiotic products. The genome sequence of Lactobacillus iners, a predominant member of the vaginal microbiota, was recently shown to have undergone extensive gene loss, resulting in the smallest Lactobacillus genome reported to date.

4In contrast, L. plantarum is highly versatile and found in many different ecological niches such as vegetables, meat, fish, and dairy products as well as in the gastro-intestinal tract. Lactobacillus plantarum is a facultative heterofermentative organism that is closely related to Lactobacillus paraplantarum, Lactobacillus pentosus and the recently identified species Lactobacillus fabifermentans. In recent years, an extensive molecular and post-genomics tool box has been established for L. plantarum and it has become one of the model micro-organisms in LAB research. This review will in particular focus on the genomic and metabolic diversity of L. plantarum. We aim to illustrate that the natural genomic architecture and the metabolic consequences hereof are central to the success of L. plantarum in industrial applications and resemble metabolic engineering strategies applied in synthetic biology.

5Already in the pre-genomic era it was recognized that the phenotypic diversity within the L. plantarum group is very high. A recent phenotypic characterization of 185 isolates from diverse environments showed that isolates from the same food niche or food type phenotypically clustered largely together, but human fecal isolates were scattered throughout different food clusters, suggesting that they generally originate from the food eaten by the individuals.

6The genetic diversity was initially catalogued by applying molecular approaches including AFLP and RAPD. This work was important in establishing molecular markers to discriminate L. plantarum from L. paraplantarum and L. pentosus, which exhibit highly similar carbohydrate utilization properties and cannot be discriminated by 16S rRNA gene sequence analysis. A multilocus sequence-typing (MLST) scheme for this organism was reported which exploits the genetic variation present in six loci of housekeeping genes and can be applied as a molecular tool for identification at the strain level(Table 1). Such molecular approaches also revealed the existence of the subspecies L. plantarum subsp. argentoratensis, which is most frequently found in fermented plant substrates and can be discriminated in a nested PCR approach. A recent diversity study through comparative genome hybridization with DNA microarrays designed on basis of the genome of L. plantarum WCFS1 confirmed that strains belonging to this subspecies share distinct genomic features and lack two putative extracellular enzyme complexes predicted to be involved in carbohydrate utilization.

7In line with its ability to grow and operate in many different niches, L. plantarum is important for different food and health applications. It is a ubiqitious and often one of the dominant species in foods such as sauerkraut, pickles, olives, sourdough and kimchi. In many of these fermentations L. plantarum dominates especially in the later stages of fermentation, presumably because of its high acid tolerance. Over the past decade, several groups have focused on the role of L. plantarum in sourdough fermentations, making it one of the best characterized vegetal substrate fermentation processes, as reviewed in depth recently. Meta-transcriptome analysis of sourdough fermentations with DNA microarrays has allowed the global analysis of community dynamics in sourdough fermentations beyond pure populations. In these fermentations, L. plantarum appears to be subjected to catabolite repression indicating that it is unlikely to play a major role in the utilization of maltose. Other groups have highlighted the importance of intra- and inter-species communication with a specific focus on communication between L. plantarum and L. sanfranciscencis strains. Quorum sensing communication involving plnA- and luxS-dependent pathways invoked multiple regulatory responses that directly influence community dynamics as well as the activity of community members.

8L. plantarum is also applied as a probiotic. Over the last decade there have been a growing number of studies aimed at deciphering the potential beneficial effects of L. plantarum strains on human health. Strain L plantarum 299v is marketed as a probiotic and a number of clinical intervention studies have been published, as reviewed by. Interestingly, transcriptome analysis using ileal and colonic biopsies from human intervention studies with this strain revealed that it specifically adapts its metabolic capacity in the human intestine for carbohydrate acquisition and expression of exopolysaccharide and proteinaceous cell-surface compounds. In return, interventions with strain L. plantarum WCFS1 were shown to exert distinct and reproducible transcriptional responses in duodenal mucosal biopsies. Consumption of live L. plantarum bacteria in different growth phases revealed striking differences in modulation of NF-κB-dependent pathways. These observations shed new light on the molecular cross-talk between ingested L. plantarum and the host, although the relevance to host health remains to be established.

9Detailed molecular and transcriptomics studies are only possible when the genome sequences are available. The genome of L. plantarum strain WCFS1, a single colony isolate of strain NCIMB8826 that was originally isolated from human saliva, was the first to be fully sequenced, and it was in fact the first of any Lactobacillus genomes to be published. It consists of a 3.3 Mb chromosome, still the largest of any sequenced lactic acid bacteria to date, and three plasmids of 1.9 kb, 2.3 kb and 36.1 kb (Table 2). A variety of bioinformatics tools has since been used to predict function of genes and gene clusters, to reconstruct metabolic pathways, to reconstruct regulatory networks, to compare with genomes of other lactic acid bacteria, and to store and visualize these results in a user-friendly way. The genome of the original WCFS1 strain has recently been resequenced using Illumina technology, revealing nearly 100 SNPs and indels, and fully re-annotated (Siezen, Francke, Boekhorst, Renckens, Kleerebezem, van Hijum, unpublished data). This small number (0.003%) of nucleotide corrections detected using modern high-throughput Illumina sequencing technology emphasizes that the original sequencing 10 years ago using Sanger technology was very thorough.

10The flexible and adaptive behaviour of L. plantarum was first found to be reflected in the chromosome of strain WCFS1, which encodes a large variety of proteins involved in sugar uptake and utilization, and accompanying regulatory proteins. This allows the organisms to grow on numerous carbon sources as it includes 25 complete PTS enzyme II complexes, several incomplete PTS complexes, and another 30 transport systems predicted to transport various carbohydrates; the genes encoding transporters are usually in gene cassettes clustered with genes encoding enzymes and regulatory proteins involved in sugar metabolism.

11The chromosome also encodes over 200 putative extracellular proteins, most of which should be displayed at the cell surface, as they are predicted to be bound to cell-envelope components in several ways. Some of these extracellular proteins are also encoded in specific gene cassettes (e.g. the csc genes), and their primary occurrence in plant-associated gram-positive bacteria suggests a possible role in degradation and utilization of plant oligo- or poly-saccharides. This large number of surface-bound extracellular proteins is also likely to contribute to the large flexibility in interactions with its environment.

12It was hypothesized that the L. plantarum WCFS1 chromosome contains specific regions that are dedicated to interactions with the environment, designated life-style adaptation regions. These regions are clustered near the origin of replication, exemplified by the region between 3.0 and 3.3 Mb which includes a large proportion of the sugar utilization cassettes as well as genes encoding extracellular functions. This entire region has a lower GC content (41.5% vs 44.5% for the whole chromosome), and many of these genes display deviation of nucleotide composition, consistent with a foreign origin. Thus, based on this single genome sequence, it was suggested that L. plantarum has lifestyle adaptation regions that could be “used to effectively adapt to the changes in conditions encountered in the numerous environmental niches in which this microbe is found”.

13The genomic diversity of L. plantarum on a full genome scale was analyzed by CGH in two separate studies. The presence or absence of genes relative to the reference strain WCFS1, of which DNA was spotted on microarrays, was assessed by hybridization of DNA from 19 and 41 other L. plantarum strains isolated from a large variety of environmental niches, ranging from fermented milk, vegetable, fruit and meat products to human isolates (intestine, saliva, faeces, spinal fluid, urine, teeth). In the first study, the probes on the microarray consisted of a subset of genomic fragments amplified by PCR from the random insert library used for initial sequencing of the L. plantarum WCFS1 genome; this microarray covered only 81% of all bases of the WCFS1 genome. The presence or absence of genomic fragments (and encoded genes) in the 19 query strains corresponding to the clones on the array was inferred from a statistical model. In the second study, an ORF-based DNA-microarray of L. plantarum WCFS1 was used, in which most ORFs were represented by at least three specific oligomer probes, evenly distributed over the gene sequence. This allowed a higher coverage of the genome content and higher resolution analysis of individual gene content in the genomes of 41 other L. plantarum strains. It must be stressed that this CGH analysis could only detect presence or absence of genes relative to the single reference genome of strain WCFS1, and did not provide information on additional genes not present in the reference strain, nor did it allow conclusions about the chromosomal location of genes (i.e. gene order). These CGH genotyping results can be displayed as “bar plots” with the chromosomal organization of strain WCFS1 as template, in which a black bar indicates the absence of a gene in a specific strain (Fig. 1). Based on their hybridizations profiles, a distance matrix representing fractional genotype similarity between strains can be constructed, and shown in a hierarchical tree. These bar plots clearly show hot spots with high variability amongst strains, and many of these hot spots, but not all, correspond to regions of high base-deviation index, suggestive of horizontal gene transfer.

14All tested L. plantarum strains were predicted to lack 9-20% of the genes of the reference genome L. plantarum WCFS1, and about 50 genes appeared to be specific for strain WCFS1, as they were not found in any other strain. The predicted absence of genes appeared to occur mainly as functional gene clusters, or cassettes, often organized in operons. These cassettes encode known functions such as i) prophages, ii) restriction-modification, iii) exopolysaccharide biosynthesis, iv) bacteriocin and non-ribosomal peptide biosynthesis, and v) carbohydrate utilization. Three large cassettes, encoding macrolide biosynthesis, non-ribosomal peptide biosynthesis (NRPS) and exopolysaccharide biosynthesis (EPS), were only found in strain WCFS1 and have a distinctly lower GC content; presumably they have been acquired by recent horizontal gene cassette transfer.

15Particularly apparent was that the proposed lifestyle adaptation regions with high density of encoded surface proteins and sugar utilization proteins, initially predicted from only the single genome of strain WCFS1, were indeed found to be highly variable in other strains (green arrows in Fig. 1), supporting the hypothesis that life-style adaptation is focused in these regions. A closer inspection of the variable sugar life-style region on the chromosome (from 3.07-3.28 Mb, genes lp_3468- lp_3657) shows that there are many consecutive cassettes of 3-10 genes which are predicted to be involved in utilization of different sugars (Additional file 1) These sugar utilization cassettes usually represent complete functional units, encoding a transporter (permease, PTS or ABC-type), a regulator and enzymes for metabolizing the sugar (Figure 2A). Most cassettes are not unique to L. plantarum, but can be found in various other lactobacilli (Figures 2B, 2C) or other bacteria (data not shown). The huge variability of these functional cassettes confirms the enormous flexibility of L. plantarum to adapt to different environments and growth substrates.

16A web-tool – PhenoLink – has been developed that facilitates associating bacterial phenotypes to ~omics data (J. Bayjanov, D. Molenaar, R.J. Siezen, S.A.F.T. van Hijum, submitted for publication). This tool uses a Random Forest algorithm which builds an ensemble of decision trees to classify huge data sets. This classification method allowed identification of correlations between genotypes (i.e. presence/absence of genes based on CGH) and phenotypes (i.e. growth on carbohydrates, NO2 production and stress tolerance) of these 42 Lactobacillus plantarum strains (J. Bayjanov, D. Molenaar, R.J. Siezen, S.A.F.T. van Hijum, submitted for publication).

17Examples of preliminary correlations found for some gene cassettes involved in sugar utilization are shown in Figure 3 (courtesy of J. Bayjanov). The 42 strains were tested for growth or no growth on a variety of sugars and the Random Forest algorithm was used to detect correlations with presence or absence of gene clusters. Four gene cassettes are shown which were originally annotated with functions involved in uptake and metabolism of arabinose, rhamnose, myo-inositol and sorbitol. The first cassette (lp_3549-3558) indeed perfectly correlates with the growth on arabinose phenotype, i.e. the gene cluster is present in strains that grow, and absent in strains that do not grow on L-arabinose. However, this gene cluster also correlates with growth on D-sorbitol and K-gluconate, suggesting that these genes may also be involved in metabolizing alternative sugars. The presence of the rhamnose gene cassette (lp_3591-3598) also correlates with growth on L-arabinose and K-gluconate, suggesting that these genes could also be involved in growth on these sugars. Finally, the putative gene clusters for myo-inositol (lp_3604-3615) and sorbitol (lp_3619-3622) utilization are found to correlate with growth on D-arabitol (and not D-sorbitol), suggesting their annotation is too specific and that these systems may be involved in metabolism of various different sugar alcohols.

18A preliminary conclusion from this gene-trait matching is that it is not easy to find correlations between phenotypes and genotypes, even with sophisticated algorithms, as both the phenotype and genotype data are inherently noisy. Moreover, classifying genes as present/absent using CGH data based on a single reference genome clearly has its limitations. As shown below, the recently sequenced genomes of other L. plantarum strains now provide clues as to why this is the case. On the other hand, the Random Forest methods are able to provide many new leads for the putative function of genes and gene clusters that can be tested experimentally. The examples above already suggest that multiple phenotypes can be linked to a single gene cluster, and vice versa.

19Complete genome sequences have now been published for L. plantarum strains WCFS1, JDM1 and ST-III, and the draft genome sequence of type strain L. plantarum ATCC14917 is available in Genbank (accession code NZ_ACGZ02000000). Two additional draft genome sequences are available from strains NC8 and KCA1 (Table 2). An extensive comparative analysis of these six genomes has been performed, providing detailed insight into core genes, variable or accessory genes and gene cassettes, genome synteny, transposable elements, and functional adaptation to growth on various substrates (Siezen, Anukam, Axelsson, Francke, Boekhorst, Renckens, Kleerebezem, van Hijum, unpublished data). Some preliminary conclusions and examples will be presented below.

20Three strains of the six sequenced genomes were previously included in the CGH analysis, i.e. strains WCFS1, ATCC14917 and NC8. This allowed a validation of the accuracy of the CGH analysis. The full genome sequences show that several genes are actually present that were classified as absent by CGH (false negatives); the majority of these are in the highly variable gene clusters for prophages, plantaricin and EPS biosynthesis (see below). The main reason for missing these genes by CGH is that the percentage identity of nucleotide sequence for certain genes or even gene cassettes is too low for hybridization on the microarray (Siezen, Anukam, Axelsson, Francke, Boekhorst, Renckens, Kleerebezem, van Hijum, unpublished data). The cut-off for reliable gene detection by hybridization appears to be at an overall 80-90% nucleotide sequence identity of genes, but this will also depend on the choice of probe positions (three 60-mer probes were used for most genes in the CGH experiment).

21An example of genes missed by CGH is the tarIJKL gene cluster, encoding wall teichoic acid biosynthesis proteins. Tomita et al sequenced the teichoic acid biosynthesis genes (clusters of 3 and 4 genes) in 18 L. plantarum strains, compared these with strain WCFS1 and found two sequence variants which correlate with glycerol-type and ribitol-type teichoic acids. The full genomes now show that these tar genes are nearly identical in strains JDM1, ATCC14917, ST-III, NC8 and KCA1, but show only 69-74% nucleotide sequence identity to the tarIJK genes of strain WCFS1 (see Figure 4), and hence were missed by the CGH analysis using WCFS1 as the reference.

22In general, there is very high conservation of gene order (synteny) in the six sequenced chromosomes of the L. plantarum strains, and in sequence identity of orthologs. However, there are several highly variable regions in the chromosome which deviate from this rule, and some examples are given below.

23- Prophages, IS elements and transposases: As expected, the cassettes encoding prophages and transposases (of IS elements) are highly variable, both in gene content and in position of insertion in the chromosomes. Details will be described in (Siezen, Anukam, Axelsson, Francke, Boekhorst, Renckens, Kleerebezem, van Hijum, unpublished data).

24- Plantaricin biosynthesis genes. The plantaricin (pln) gene cluster of L. plantarum of about 25 genes (Figure 5), encodes the biosynthesis of various class IIb bacteriocins, whose full activity depends on the action of two different peptides. The pln gene cluster has been previously sequenced from six L. plantarum strains (including WCFS1 and NC8) and found to be a highly variable and mosaic region, with parts being relatively conserved and other parts less conserved. A PCR analysis directed to 27 genes of this pln cluster in 33 L. plantarum strains of oenological origin found even more variation and led to a subdivision into seven groups, named plantaritypes. The six fully sequenced genomes (Table 2) show the same high variability, but also demonstrate that the conservation and synteny of genes flanking the pln locus is very high (Siezen, Anukam, Axelsson, Francke, Boekhorst, Renckens, Kleerebezem, van Hijum, unpublished data).

25- CPS/EPS biosynthesis genes. The chromosome of strain WCFS1 has 3 consecutive cps clusters (cps1, cps2, cps3), separated by transposase genes (and their fragments), encoding proteins involved in biosynthesis and export of extracellular or capsular polysaccharides. At the same position in the chromosome, the other 5 sequenced genomes also have cps clusters (Figure 6). Some parts of these clusters are shared, but many parts contain completely different cps genes. The cps1 cluster of strain WCFS1 is not present in any of the other sequenced genomes, while the cps3 cluster and parts of the cps2 cluster are shared by 4 strains, but lacking in JDM1. This high variability presumably leads to variation in the structure of capsular and exopolysaccharides. A similar variability of EPS gene cassettes has been observed in other LAB.

26The enormous variability of cassettes in the sugar life-style region deduced from CGH analysis (Additionl file 1) led to the hypothesis “…. that similar lifestyle adaptation islands will exist in other strains of L. plantarum. This would also imply that these strains contain a high number of genes with related functions accumulated within their lifestyle adaptation region that are absent from the WCFS1 genome.”.

27Comparison of the 6 sequenced genomes now shows that indeed many novel cassettes are present in these variable regions that are absent in strain WCFS1 (Siezen, Anukam, Axelsson, Francke, Boekhorst, Renckens, Kleerebezem,van Hijum, unpublished data). A typical example is shown in Figure 7, representing the diversity of cassettes in a small part of the life-style region corresponding to genes lp_3114-lp_3150 (from 2.78-2.82 Mb on the chromosome) in strain WCFS1. In addition to the five known cassettes present in strain WCFS1, another five novel cassettes are found in the other sequenced strains, and these would not have been detected in the original CGH analysis using only WCFS1 as a reference. Strains WCFS1 and ATCC14917 share the same 5 cassettes, and strains NC8 and ST-III share a different set of 5 cassettes, while 3 cassettes are unique to strain KCA1. Cassette 5 encoding putative cellobiose utilization appears to replace cassette 4 encoding β-glucosides utilization at the same position in the chromosome.

28Diversity between strains is even seen within L. plantarum genes and proteins at the level of numbers of repeated domains or motifs, particularly in extracellular proteins, again suggesting variability between strains in interactions with their environment.

29One published example is an extracellular, peptidoglycan-bound mannose-specific adhesin (msa gene; lp_1229 in strain WCFS1) which has variations in the number of mucus-binding (Mub) domains and PxxP spacer motifs in different strains, which could relate to differences in mucus-binding efficiency(Figure 8). Another example is a very large extracellular, membrane-anchored protein (encoded by lp_1303a in strain WCFS1) of unknown function that has a middle domain with hundreds of repeats of SD (Ser-Asp); possibly the Ser residues are glycosylated by glycosyl transferases, as three adjacent genes in the L. plantarum genomes encode putative glycosyl transferases. The N-terminal domain (includes signal peptide) and C-terminal domain (includes membrane anchor) are nearly identical in this protein from all sequenced L. plantarum strains, but there is a large variation in the number of SD repeats ranging from 242 repeats in strain JDM1 to 801 repeats in strain WCFS1.

30Genome sequencing and comparative genomics of L. plantarum has revealed a high genomic diversity, versatility and flexibility which are at the heart of its success in diverse niches and applications. One of the most striking observations is the occurrence of genomic islands harboring mosaic modules or cassettes of carbohydrate utilization genes, likely acquired through horizontal gene transfer. Many of these cassettes are also found in other LAB (see for instance Figure 2) and other gram-positive bacteria. L. plantarum seems to be a master in acquiring and shuffling these cassettes, but it cannot be excluded that this shuffling propensity is also linked with other gram-positive bacteria. By its ability to acquire and assemble gene cassettes required for the utilization of carbohydrates, L. plantarum seems to have developed a “natural metabolic engineering approach” that allows it to optimize its genome for growth in specific niches, especially those rich in plant carbohydrates. In that aspect, there are interesting synergies with the approaches applied nowadays in metabolic engineering for biofuels production. The cellular traits and catabolic capacities that are sought when constructing biofuels-producing S. cerevisiae and E. coli are prospected in nature, introduced into the host and subsequently optimized through advanced molecular and evolutionary engineering strategies.

31It seems likely that in L. plantarum there could be an important role in this respect for diversity generating molecular structures such as (conjugative) plasmids, IS elements and transposons. To date, experimental data showing that transfer of such elements is an effective means to increase the catabolic potential of lactic acid bacteria are scarce. Recently, it was shown that conjugative transfer of transposon Tn6098 encoding alpha-galactosides metabolism in Lactococcus lactis enable a dairy isolate to grow in soy milk, a substrate rich in α-galactosides. On the other hand, the numerous cassettes in L. plantarum appear to be directly linked to each other, with no evidence of flanking or intervening elements from plasmids or IS elements. Further investigation into the molecular mechanisms by which L. plantarum is able to acquire and assemble such gene cassettes encoding functional pathways into its chromosome should therefore be of importance to learn how to harness this natural metabolic engineer.

32The authors declare they have no competing interests.

33RJS conceived the review and gathered, restructured and analyzed data. RJS and JHV jointly drafted and wrote the manuscript.

34ABC: ATP-binding cassette; ACT: Artemis comparison tool; AFLP: amplified fragment-length polymorphism; CGH: comparative genome hybridization; CPS: capsular polysaccharide; EPS: extracellular polysaccharide; GC: guanine & cytosine; LAB: lactic acid bacteria; IS: insertion sequence; MLST: multi-locus sequence typing; ORF: open-reading frame; PCR: polymerase chain reaction; PTS: phospho-transferase system; RAPD: random amplification of polymorphic DNA; SNP: single nucleotide polymorphism.