Ethics statement

All persons who donated blood for this study provided a written informed consent. The study protocol has been approved by the Academy of Finland (Projects 28 137389, 141140 and 1272870).

Bacterial isolates and growth conditions

Lactobacillus strains used in the study are listed in Table 1 and includes 4 newly sequenced strains and 12 strains previously published [22, 23]. Clinical blood isolates of L. rhamnosus strains were obtained from HUSLAB and collected from bacteremic patients between years 2005 and 2011. L. rhamnosus GG (ATCC 53103) was obtained from the Valio culture collection (Valio Ltd., Helsinki, Finland). L. salivarius CCUG 47825 was a kind gift from Prof. Paul W. O’Toole (University College Cork, Cork, Ireland) [19]. Pathogenic control Streptococcus pyogenes strain st369 was obtained from the collection of the National Institute for Health and Welfare (Prof. Jaana Vuopio) and emm8 from ATCC (ATCC catalogue number 12349). S. pyogenes strains were used as controls for complement activity and FH and C4bp binding assays [24]. Lactobacilli were cultured at 37°C in de Man-Rogosa-Sharpe (MRS) medium (Difco BD, NJ, USA). S. pyogenes strains were grown at 37°C with 5% CO₂ on blood-agar plates supplemented with colistin and oxolinic acid or in Todd-Hewitt broth (Difco BD, NJ, USA). For in vitro assays Lactobacillus strains were cultured at 37°C in MRS medium overnight. The overnight bacterial cultures were diluted 1:100 and cultured until the OD₆₀₀ reached 0.5–0.7 for log phase and overnight for stationary phase. The cells were collected by centrifugation and glycerol was added until 20% to freeze the cultures. The strains were stored at -70°C.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Genetic information of L. rhamnosus strains isolated from blood cultures in HUSLAB, 2005–2011.

https://doi.org/10.1371/journal.pone.0176739.t001

Genome sequencing

DNA was extracted from the bacterial cultures as described previously [23]. Libraries from genomic DNA from L. rhamnosus strains were done for both 454 sequencing and MiSeq sequencing following recommendation from respective manufacturer. The 454 libraries were sequenced using a Genome Sequencer FLX (Roche) and titanium chemistry with read lengths of ca 400 bp. The Illumina libraries were paired-end sequenced (250 bp + 250 bp) using a MISeq (Illumina). All library preparations and sequencing were done at the DNA sequencing and genomics laboratory, Institute Biotechnology, University of Helsinki, Finland. A combined 454 and MiSeq read assembly were performed using the Celera Assembler v6.1 (http://sourceforge.net/apps/mediawiki/wgs-assembler/index.php?title=Main_Page). Genomes were annotated using the RAST server [25]. The new genome sequences were deposited to public depository databases but most of them were published earlier by Ceapa et al. [23] or by others (see for details Table 1).

Bioinformatics analysis

The orthology prediction was performed using sequence clustering and search program Usearch v8.0.1623 [26]. To define orthologous groups, all the predicted ORF amino acid sequences of all the strains were first ordered by length using the sortbylength command, and then clustered with the cluster_smallmem command using identity threshold of 0.9. Thus, the members of each ORF sequence cluster shared at least 90% amino acid identity over their whole sequence lengths. The longest sequence in each cluster was chosen as the representative sequence of that orthologous group. The pangenome was defined as the set of representative sequences of the orthologous groups. The ortholog data was managed using R v. 3.2.2 (R) with Biostrings library v. 2.40.2 [27] and in-house scripts. The ortholog copy number data was used for hierarchical clustering of the strains by first removing orthologs that had variance <0.06 over the strains, and then calculating Euclidean distances between the strains and applying Ward's minimum variance method to the distances using R functions dist and hclust, respectively. Clustering dendrograms were drawn with R library dendextend v. 1.3.0 [28], and the function heatmap.2 from R library gplots v. 3.0.1. [29] was used for producing a row-wise centered and scaled heatmap of the filtered ortholog number matrix. Strains were compared at nucleotide level to the L. rhamnosus GG genome using blastn program v. 2.2.31+ with parameters -task magablast, -evalue 0.001 and -perc_identity 95. Mean similarity of aligned fragments over 10 kb was calculated for each strain.

Complement activity determination using C3a and TCC ELISA

The frozen batches of log phase bacteria were used to activate complement in the serum pool of healthy volunteers. Bacterial cells were washed with PBS and adjusted to have 10⁷ bacterial cells in 100 μl of PBS. The bacteria were mixed with the same volume of serum and incubated for 30 minutes at 37°C in agitation at 500 rpm. The reactions were stopped by adding EDTA to a final concentration of 15 mM and by transferring the samples on ice. Cells were separated from serum samples by centrifugation at 5,900g for 3 minutes and stored at -70°C for C3a and SC5b-9 (TCC) analysis. Complement C3a and SC5b-9 were analyzed from the samples using the commercial MicroVue C3a EIA (Quidel, Tecomedical) and ELISA for TCC determination as described earlier [30]. For C3a measurements serum samples were diluted 1 to 10,000 and for TCC assay 1 to 2,000.

Direct ¹²⁵I-factor H and C4bp binding assays

The binding of FH and C4bp were performed as described earlier [31]. Shortly, FH and C4bp were labelled with iodine (¹²⁵I; Perkin Elmer) using the Iodogen method [32]. The specific activities of the labelled proteins were 1–2 x 10⁷ cpm/μg. Each bacterial strain (3 × 10⁷ log phase bacteria/reaction) was incubated with 25,000 cpm of radiolabeled FH or C4bp in 100 μl of gelatin-Veronal buffered saline for 30 minutes at 37°C. The reaction mixtures were centrifuged (at 10,000g for 3 minutes) through 20% sucrose, and the radioactivity of each pellet and supernatant was measured with gamma counter. S. pyogenes strains were used as a negative (emm8) and positive (st369) controls for FH binding and positive control for C4bp binding [33]. The ratios of bound (pellet) to total activity (pellet + supernatant) were calculated as percentage.

Platelet preparation and aggregometry

Blood samples were collected from healthy donors as previously described [34]. Nine ml of blood was collected into 3.2% Na-citrate tubes (Vacuette, ref. 455322, Greiner-bio-one) via venipuncture from a cubital vein from a healthy donor using 20G needle. Whole blood was centrifuged at 180g for 12 minutes at 22°C without using brake. Platelet rich plasma (PRP) was collected slowly above the buffy coat with a plastic Pasteur pipette into separate plastic tube. The remaining blood was mixed and centrifuged again at 2,000g for 10 minutes at 22°C to obtain platelet poor plasma (PPP). The platelet count was measured with Sysmex XK-21 cell counter and adjusted with PPP to the desired concentration.

Platelet aggregation was carried out in a profiler aggregometer (PAP-4, Bio/Data corp.), with light transmission through PPP presenting 100% aggregation and that through PRP representing 0% aggregation. PRP was used in a test well and the change in transmission over time was a signal of the platelet aggregation. The frozen batches of bacterial cells were washed with PBS three times. The final volume was adjusted to get 2 x 10⁹ cfu/ml in PBS for platelet aggregation experiments. Lactobacillus strains were determined for their ability to induce platelet aggregation by the addition of 25 μl of a bacterial suspension to 475 μl of PRP. One minute after agonist ADP or bacteria was added into the well the platelet response was monitored for a maximum 30 minutes. Platelet viability was confirmed by addition to PRP of agonist adenosine diphosphate (ADP) to a final concentration of 5 μM.

Biofilm assay

The microtiter plate biofilm assay is a static assay that has been used for examining biofilm formation [35]. The wells of 24-well cell culture plate (Greiner Bio-One, Germany) were filled in with serial dilution (1 to 10) of an overnight bacterial culture. After an incubation for 18 hour at 37°C, the wells were washed with aqua and stained with crystal violet. The crystal violet stained plates were dried and scanned with Odyssey IR scanner (Licor) at 700 nm channel and integrated intensities analyzed using Odyssey (v 2.1) analysis software. The bound dye was also solubilized in 250 μl of 70% ethanol, transferred into 96-well plate and the optical density of the ethanol-dye solution was determined at 600 nm using a microplate reader (Fluostar Optima, BMG Labtech, Germany). The correlation between the methods was high (97.3%) indicating that the crystal violet was a strong fluorophore at 700 nm and Odyssey scanner is a useful tool for determining biofilms.

Western blot analysis of cell wall proteins

Cell wall proteins were extracted and the presence of SpaCBA and SpaFED pili were determined using Western Blot analysis as earlier described [22].

Statistical analysis

All values are presented as a mean ± SD. Where applicable, a 2-tailed Student's t-test was used to analyse the differences between groups (Excel software; Microsoft, Redmond, WA, USA).

Genomic features and organization of L. rhamnosus clinical blood isolates vs. L. rhamnosus GG

All L. rhamnosus strains described in the present study were isolated from clinical blood cultures collected during 2005–2011 from the Helsinki University Hospital laboratory. The genomes of four strains are reported here while that of the other 12 have been described earlier [23]. Their general genomic information was compared to that of L. rhamnosus GG (Table 1 and Table 2). Genome size, G+C content and open reading frame (ORF) count of annotated protein coding sequences varied greatly between the strains and differed from L. rhamnosus GG. The genome sizes ranged from 2.86 to 3.17 Mb. The average genome size was 3.0 Mb ± 0.16 Mb which is in the same range as other members of the L. rhamnosus species [23]. The G+C content varied among the genomes from 46.6 to 46.8% being closely similar to L. rhamnosus GG (46.7%) while the number of ORF predictions varied between 2648 and 3010, exceeding that of L. rhamnosus GG (2926).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Plasmids of the L. rhamnosus strains.

https://doi.org/10.1371/journal.pone.0176739.t002

To analyze the similarity of the strains against L. rhamnosus GG at sequence level, a dendrogram based on cluster analysis of the presence of orthologs at 90% sequence identity was created (Fig 1). This showed that the strains could be grouped into two main clusters and one separate strain (Lrh20). It shows that the L. rhamnosus isolates are different from each other at the sequence level and more remarkably, all are very different at the genomic level from L. rhamnosus GG. The strains in cluster A have sequence similarity at nucleic acid level to L. rhamnosus GG between 99.942 and 99.984% whereas the strains in cluster B differ remarkably more having the similarity between 97.0 and 98.5% (Fig 1). Interestingly, all the genomes of five isolated L. rhamnosus strains (Lrh13, Lrh22, Lrh28, Lrh30 and Lrh47) that contain plasmids belong to the cluster B (Table 2, Fig 1). The presence of unique orthologs between clusters were compared (S1 Table). Most of the differing orthologs that could be annotated, and are not phage sequences, are exopolysaccharide-related. Cluster A consists of 66 orthologs that were not found from cluster B and 35 orthologs that were missing from cluster A but present in cluster B. Interestingly, there were 8 EPS related ORFs present and 10 absent in cluster A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Clustering analysis of strains by ortholog copy numbers.

The heat map rows represent centered and scaled ortholog copy numbers found in the strains. The dendrogram on the columns shows hierarchical clustering of the strains based on their ortholog copy numbers. Columns are annotated by each strain's BLAST nucleotide similarity to the L. rhamnosus GG genome.

https://doi.org/10.1371/journal.pone.0176739.g001

Complement activation and binding of complement regulators

The recovery of L. rhamnosus isolates from the blood of bacteremic patients raised questions about their survival within the human bloodstream and also their (possible) recognition or killing by the host immunological system. Initial experiments indicated L. rhamnosus GG and clinical isolates showed only little survival in blood (less than 31% survival after 3 hours at 37°C) in contrast to strain st369, a known pathogen belonging to the Group A Streptococci, which multiplies rapidly in blood (at least 5 cell divisions after 3 hours at 37°C; see S2 Table). To further explore these, we studied in vitro if the isolated strains can activate the complement system and thus, be recognized by the human innate immune system, and if they are able to bind complement regulators to evade complement in the manner of pathogenic bacteria. Strains were incubated with human serum to trigger the complement activation response. All the isolates as well as L. rhamnosus GG induced the formation of C3a and soluble terminal component TCC in serum (Fig 2 and S3 Table). All tested L. rhamnosus strains induced significantly higher C3a concentrations in serum than the negative control (S. pyogenes st369). Similar observations were made for the levels of TCC, although more variations were seen among isolates (notably between Lrh14 (the lowest) and Lrh7, Lrh13, Lrh18, Lrh20 and Lrh30 (the highest values)). Taken together, the amounts of both complement activity markers (C3a and TCC), indicated that the complement system was activated in the presence of the isolated L. rhamnosus strains in vitro, as opposed to the S. pyogenes st369, which did not activate the complement system similarly as earlier described [24].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. A heatmap of phenotypic features of the L. rhamnosus strains measured in vitro.

Combined data represent average values or numbers of the strains detected in clusters A and B. NC equals negative control containing buffer in the assays, GAS group A streptococcus, NT not tested, asterisk* equals values that are representative from literature but not tested here for platelet aggregation [36] and biofilm formation [37], TCC equals terminal complement complex, FH factor H, C4bp complement C4b binding protein, CBA SpaC, SpaB and SpaA pilin subunits and fibron. bp equals fibronectin/fibrinogen binding protein. Bold and Italic letters equal significant value, p<0.05 for biofilm formation compared to L. rhamnosus GG (GG). Color intensities indicate the level of the values: green for complement activation measured as C3a and TCC, red for binding of complement regulators, blue for platelet aggregation, purple for biofilm formation and light green express the presence of SpaCBA pilus determined by Western blot using anti-SpaA antibody and fibronectin/fibrinogen binding protein at genome.

https://doi.org/10.1371/journal.pone.0176739.g002

To further explore additional factors mediating activation of the complement and to assess the ability of these strains to potentially inhibit complement activation, we examined the direct binding of complement regulators FH and C4bp to all L. rhamnosus strains (Fig 2, S3 Table). No binding of FH or C4bp to any of the L. rhamnosus strains was observed. The FH binding ranged between 0.3–0.55% and the C4bp binding between 1.6–5.8%, which corresponds to background levels (0.40 ± 0.15% and 3.1 ± 0.6%, respectively). In contrast, the control strain S. pyogenes st369 showed strong FH binding (46.7 ± 14.1%) and both S. pyogenes emm8 (35.7 ± 0.2%) and st369 (31.9 ± 14.7%) exhibited strong C4bp binding as previously reported [33]. These results indicate that the L. rhamnosus blood isolates and the probiotic strain GG do not bind these complement regulators and therefore do not escape complement using studied mechanisms as known in pathogens [38, 39].

Human platelet aggregation by the L. rhamnosus strains

Next we studied the capacity of the L. rhamnosus strains to promote platelet aggregation. Platelets are cell fragments that plug the leakage in a damaged blood vessel but they are also key components in the innate immune system [40]. Many Gram-positive bacteria causing sepsis have been shown to induce platelet aggregation [41]. The L. rhamnosus strains were tested for platelet aggregation using isolated platelets from different human donors (Fig 2, S3 Table). Four out of 16 L. rhamnosus isolates (Lrh11, Lrh18, Lrh22, Lrh28) induced platelet aggregation. It is noteworthy that platelet aggregation was donor-dependent. Hence the same strain did not always induce platelet aggregation for every donor tested and this was also noted for the positive control L. salivarius CCUG 47825 (S3 Table). Two of the aggregating strains, Lrh22 and Lrh28, induced platelet aggregation with platelets from every donor tested. Similar to previous studies [17] L. rhamnosus GG and the rest of the twelve isolates did not induce platelet aggregation for any tested donors. The experimental data shown here suggest that the L. rhamnosus blood isolates differ in their ability to cause platelet aggregation and that platelet aggregation was not a common phenotypic trait associated with L. rhamnosus blood isolates although 2 clinical strains were highly proficient platelet aggregators.

Biofilm formation by the L. rhamnosus strains

Biofilm formation is a feature found in both pathogenic and commensal bacteria, as it promotes colonization and persistence in an ecological habitat. Hence, we assessed the formation of biofilms for L. rhamnosus strains in 24-well plates (Fig 2 and S3 Table). Some differences in biofilm formation could be seen between the isolated strains. Four of the strains (from cluster B) tested (Lrh13 (8.9 ± 3.1%), Lrh28 (13.1 ± 4.7%), Lrh30 (7.1 ± 2.3%), Lrh32 (5.4 ± 1.3%)) showed a stronger ability for biofilm formation than GG (3.8 ± 0.8%). L. salivarius CCUG 47825 was used as a positive control and produced a strong biofilm (relative biofilm formation 100 ± 25.7%). Summarizing, biofilm formation varied between L. rhamnosus isolates with some strains displaying higher biofilm formation than L. rhamnosus GG. All those strains that showed clear biofilm formation phenotype belong to cluster B.

Comparison of LPXTG protein encoding genes, SpaCBA pilus proteins and EPS related genes

In an attempt to further characterize the phenotypes of these different isolate, we examined some of the relevant genomic features related to surface exposed structures and molecules of these strains. Typically, the genomes of Gram-positive bacteria harbor genes that code for LPXTG proteins as well as the sortase enzymes. Sortases are involved in the surface display and cell wall anchoring of surface proteins carrying the sortase-specific recognition motif LPXTG in Gram-positive bacteria [42]. Based on the genome sequences of the blood isolates, we investigated the repertoire of all gene sequences that are predicted to code for LPXTG proteins, as they are most likely to interact with the environment and the host, in that particular context with the human vascular system. All strains possessed a distinct and unique set of proteins (S2 Fig). It has been well established that the presence of pili depends on a specific set of LPXTG proteins that include major pilin, minor pilin and, tip pilin subunit proteins and these are widely spread in various Lactobacillus spp. [2]. In L. rhamnosus GG, two pili gene clusters have been identified but only SpaCBA pili have been found to be produced [22, 43]. Hence, we used Western blot analysis with SpaA and SpaD antibodies to show that 8 out of 16 L. rhamnosus strains produced SpaCBA pili while none were producing the SpaFED pili (Fig 2) even though the SpaFED locus was found in all strains (S1 Fig). Five of these showed phenotypic characters (platelet aggregation (Lrh18, Lrh11) or biofilm formation (Lrh13, Lrh30) or both (Lrh28; Fig 2)). When the strains were divided into two groups based on the presence of SpaCBA pilus there was a significant association between the biofilm formation and the presence of pilus among the isolated strains (6.5 ± 3.11% (present) vs 3.87 ± 0.72% (absent), p = 0.047, S3 Table).

As the surface polysaccharides are known to participate in important functions during pathogenesis, such as tissue adherence, biofilm formation and evasion of host defence [44, 45] and our clusters of L. rhamnosus strains showed differences between the ORFs related to EPS/CPS gene clusters (S1 Table), we wanted to study if the genomic organization of EPS gene clusters in the studied blood isolates of L. rhamnosus are similar to L. rhamnosus GG [46]. This cluster consists of genes coding proteins for the regulation of EPS production and polymerization (wzd, wze, wzr, wzb), polysaccharide transportation and polymerization (wzx, wzy), glycosyltransferases (welJ, welI, welH, welG, welF, welE), biosynthesis of the dTDP-rhamnose precursor (rmlA, rmlC, rmlB) and UDP-galactopyranose mutase for biosynthesis of the sugar nucleotide precursor for galactofuranose (glf) [46]. Interestingly, after annotation analysis our data showed that the strains in cluster A had the EPS cluster consisting 16 putative ORFs/genes with similar identity (> 78%) to those present in L. rhamnosus GG (Fig 3A). However, the strains in cluster B had only some of these genes (wzd, wze, wzr, wzb and rmlB (except Lrh22)) with similar identity (> 78%) to genes in EPS cluster of L. rhamnosus GG (Fig 3B). The other genes found in this cluster were annotated to encode genes for mostly glycosyl transferases with different specificities, hypothetical proteins and several mobile element proteins. However, we were not able to characterize these clusters in details because multiple mobile elements were present and they spliced the sequences into multiple contigs. Interestingly, the strains in cluster B consist also a different type of polysaccharide (EPS/CPS) cluster comprising of 19 genes (Fig 3B) including an additional flippase (6.) gene essential for EPS production. The cluster was similar in most of the strains (Lrh22, Lrh30, Lrh32 and Lrh47) in the cluster B (Fig 3B). The strain Lrh11 had a mobile element in the PS cluster before the gene coding for lysozyme and the strains Lrh13 and Lrh28 had two point mutations within the lysozyme gene (resulting in a split gene). The differences in the organization and existence of the polysaccharide clusters may have an influence on the polysaccharide surface of the strains and could potentially lead differences in the phenotype of these isolates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Genomic organization of polysaccharide clusters in L. rhamnosus strains.

A) EPS gene cluster of L. rhamnosus GG is found from L. rhamnosus strains of cluster A and strain Lrh20. Insertion sequence (IS)-like elements differ between the strains. B) The L. rhamnosus strains of cluster B have differences at sequences encoding genes in the EPS gene cluster compared to L. rhamnosus GG. The genes similar to the EPS gene cluster of L. rhamnosus GG are named in the schematic illustration in the upper part (B). These regions include multiple genes coding IS-like elements such as IS5, IS66 and DDE transposases in the cluster B. The strains in cluster B have also other type of EPS/CPS gene cluster based on BLAST annotation results shown at the bottom (B). It consists of 19 or 20 genes which are annotated as 1. peptidoglycan N-acetylglucosamine deacetylase (EC 3.5.1.104), 2. hypothetical protein (FIG00750667), 3. capsular polysaccharide biosynthesis protein, 4. hypothetical protein (FIG00754398), 5. tyrosine-protein kinase transmembrane modulator EpsC, 6. putative O-unit flippase, 7. glycosyl transferase, group 0, 8. glycosyl transferase family protein, 9. hypothetical protein (FIG00752018), 10. mobile element protein, 11. lysozyme M1 (1,4-beta-N-acetylmuramidase), 12. hypothetical protein (FIG00747485), 13. uncharacterized conserved protein, similar to IcaC of Staphylococcus; YHJR B.subtilis family, 14. glycosyltransferase, 15. UDP-N-acetylglucosamine 2-epimerase (EC 5.1.3.14), 16. hypothetical protein (FIG00750997), 17. lipopolysaccharide synthesis sugar transferase, 18. polysaccharide biosynthesis protein CpsM, 19. putative N-acetylgalactosaminyldiphosphoundecaprenol and 20. UDP-glucose 4-epimerase (EC 5.1.3.2). * = Transposase, IS150/IS3 present (only in Lrh11), ↑ = site of point mutations in the gene encoding lysozyme M1(1,4-beta-N-acetylmuramidase) (present in Lrh13 and Lrh28). The illustration of EPS cluster in 3A is modified after [46].

https://doi.org/10.1371/journal.pone.0176739.g003