Ten healthy volunteers, seven females and three males (26.3 ± 10.1 years, BMI of 21.8 ± 2.40 kg/m²), without a history of gastrointestinal symptoms and free of medication, were tested on four separate occasions in a randomized cross-over design. Four intervention periods included the 7-day intake of L. plantarum strains WCFS1 (WCFS), CIP104448 (CIP48), or TIFN101, or placebo. Before and after the intervention, blood samples were taken to study the effect of L. plantarum consumption on T-cell polarization. Furthermore, at day 7, duodenal biopsies were taken for whole-genome expression microarrays. No side effects or complications were reported.

All included subjects gave their written informed consent. The healthy volunteers consumed their habitual diet and kept a gastrointestinal symptoms diary. The night before the start day, the volunteers ingested 75 mg of the NSAID indomethacin. On the start day, the volunteers ingested another dosage of 50 mg indomethacin to conform to a previously established protocol to establish mild gastrointestinal mucosal stress (25, 26). Subsequently, the volunteers consumed the bacterium or placebo supplements for a period of 7 days during breakfast and during dinner. The vials containing bacteria or placebo were non-transparent. On the seventh day, blood samples were taken again and tissue samples were obtained from the horizontal part of the duodenum by standard flexible gastroduodenoscopy at approximately 15 cm distal to the pylorus. The duodenal mucosa was chosen as the duodenum is readily accessible for sampling. The interventions were performed with an interval of 4 weeks to allow a wash-out period and also healing of the biopsy-sampling region.

Three L. plantarum strains form the NIZO strain-collection were selected: WCFS1, CIP104448, and TIFN101. In previous studies TIFN101 has been referred to as CIP104450 (11, 27). The bacteria were cultured at 37°C in man, rogosa, and sharpe medium (Merck). Detailed protocols for culturing, harvesting, freeze drying, storing, and viability determination of Lactobacillus species have been published (1). Maltodextrin and glucose were added to a final concentration of 20 and 2% (wt/vol), respectively, to obtain bacterial preparations (WCFS1, 2.6 × 10⁹ CFU; CIP104448, 2.4 × 10⁹ CFU; and TIFN101, 5.6 × 10⁹ CFU); placebo controls contained the two sugars.

Blood of study participants was collected in EDTA-containing tubes and processed for fluorescence-activated cell sorting (FACS) analysis (Table S1 in Supplementary Material). Isotype controls were used at the same dilution as the antibody.

To study T-cell polarization, 200 µl of blood was diluted with 200 µl of RPMI1640 supplemented with heat-inactivated 10% fetal calf serum (FCS) and incubated with either phorbol-myristate-acetate (PMA; 80 nM Sigma-Aldrich, Steinheim, Germany) and 2 nM calcium ionophore (Ca-Io; Sigma-Aldrich) (4 h), Staphylococcus aureus-enterotoxin B (SEB) (5 µg/ml Sigma, Deisenhofen, Germany) (24 h), tetanus toxoid (TT; 1.5 Lf/ml, RIVM, Bilthoven, The Netherlands) (24 h), or bacterial lysates (30 µg/ml) (24 h). Stimulation with bacterial lysates was performed 1 week after treatment with one of the three strains. After stimulation, red blood cells were lysed with ammonium chloride, washed (PBS with 2% FCS), and incubated with different antibody cocktails.

To stain for T-cells and T-cell subsets, cells were incubated with an antibody cocktail consisting of anti-CD3, anti-CD8, and anti-CD45RO for 30 min in the dark on ice. After washing with buffer, cells were incubated with streptavidin-Pacific Orange (1:100 Invitrogen) for 15 min on ice. After washing and centrifugation, pelleted cells were resuspended in Fix/Perm solution (eBioscience) for 45 min on ice. After washing in Perm-solution cells were incubated in mouse-serum for 15 min, followed by incubation with the cytokine antibody mix (anti-IL-4, anti-IFNγ, anti-IL-17, and anti-IL-21) or an isotype control mix for 30 min on ice. Cells were then washed with Perm solution (three times), resuspended in wash-buffer and measured by flow-cytometry.

For staining NK-cells, cells were incubated with an antibody cocktail consisting of anti-CD3, anti-CD16, anti-CD56, anti-CD335, and anti-CD161 (NK-cell staining), or with an isotype control cocktail for NK-cells consisting of anti-CD3, anti-CD16, anti-CD56, and isotype controls for anti-CD335 and CD161 for 30 min in the dark on ice. After washing with washing buffer, cells were fixed in FACS-lysis solution (BD Biosciences) for 30 min on ice, washed and analyzed by flow-cytometry.

Total RNA was isolated from duodenal biopsy samples of the study participants using Trizol reagent (1 ml) (Invitrogen, Breda, The Netherlands). RNA was purified using the Qiagen RNeasy Micro kit (Qiagen, The Netherlands) and quantified on a NanoDrop ND-1000 spectrophotometer (Isogen Life Science, The Netherlands). RNA quality was confirmed using an Agilent 2100 bioanalyzer (Agilent Technologies, The Netherlands). RNA was judged suitable for array-hybridization only if samples exhibited intact bands corresponding to 18S and 28S ribosomal subunits and displayed no chromosomal peaks or RNA-degradation products.

Total RNA (100 ng) was used for whole transcript cDNA synthesis using the Ambion WT expression kit (Life Technologies, The Netherlands) and was subsequently labeled using the Affymetrix GeneChip WT Terminal Labeling Kit (Affymetrix, Santa Clara, CA, USA). Samples were hybridized to human whole genome Affymetrix GeneChip Human Gene 1.1 ST arrays, washed, stained, and scanned on an Affymetrix GeneTitan instrument. Details on array handling can be found in the Affymetrix GeneTitan Instrument User Guide for Expression Array Plates (P/N 702933 Rev.2).

Microarray analysis was performed utilizing MADMAX for statistical analysis (28). All arrays met the quality criteria. The probes on the Human Gene 1.1 ST arrays were redefined according to Dai et al. (29) based on the NCBI Entrez database (CDF version 15.1). In this way, the Human Gene 1.1 ST array targets 19,682 unique genes. Normalized expression values were obtained from the raw intensity values by using the robust multi-array analysis preprocessing algorithm available in the AffyPLM library using default settings (30). Microarray data were filtered and probe sets with at least five probes and expression values higher than 20 on at least five arrays with an interquartile range (IQR) >0.2 (log2 scale) across all samples were selected for further statistical analysis. In addition, an IQR cut-off of 0.2 was used to filter out genes that showed no variation between the conditions. Differentially expressed genes were identified using linear models, applying moderated t-statistics that implemented empirical Bayes regularization of SEs in the library limma (31). The moderated t-statistic was extended by a Bayesian hierarchical model to define an intensity-based moderated t-statistic to adjust for independence of variances relative to the degree of identity and relation between variance and signal intensity (32). Genes were defined as significantly changed when the p-value was <0.05 for pairwise comparisons. The datasets generated from microarray profiling experiments have been deposited to the publicly accessible database repository Gene Expression Omnibus (GSE74988).

Gene set enrichment analysis (GSEA²) was performed using MADMAX and gene sets with a false discovery rate <0.2 were considered significantly enriched. GSEA takes into account the broader context in which gene products function, namely in physically interacting networks such as biochemical, metabolic, or signal transduction routes, and has the advantage that it is unbiased because no gene selection step is used. Possible transcription factors were identified using upstream regulator analysis in ingenuity pathway analysis (IPA; Ingenuity Systems, Redwood City, CA, USA).

For DNA preparation, bacteria were pelleted, washed, and resuspended in TES buffer (N-[tris(hydroxymethyl)methyl]-2-aminoet hanesulfonic acid). Cells were lysed with lysozyme (360 mg/ml) and mutanolysin (140 U/ml) by incubation for 2 h at 37°C. Subsequently, 300 µl water and 80 µl of a 20% SDS solution were added. The DNA extraction was done using phenol/chloroform (3×). DNA was precipitated with isopropanol and washed with 70% ethanol. Samples were treated with 100 µg/ml RNAse (Sigma) for 1 h at 37°C. DNA paired-end libraries with barcoding were made and sequenced using Illumina technology (Baseclear Leiden). The genome sequences of L. plantarum strains CIP104448 and TIFN 101 have been deposited in NCBI/GenBank under accession numbers JSUW00000000 and JSUX00000000, respectively. The contig sequences were submitted to the RAST automatic annotation server,³ which provided ORF calling and automatic annotation. The annotated contigs of CIP48 and TIFN101 were ordered by comparing them to the circular template genome of L. plantarum WCFS1 (33), and comparing them to each other. Contigs/genes that did not match to the WCFS1 genome were annotated in more detail using BLASTP⁴ and InterProscan.⁵ Ortholog groups (OGs) in the three genomes were determined using OrthoMCL.⁶

Flow-cytometry results are expressed as the mean ± SEM. Normal distribution was confirmed by the Kolmogorov–Smirnov test. The two-sided Student’s t-test was used for changes in immune-cell populations after L. plantarum treatment. Gene expression data are depicted as the median (range). The two-sided Mann–Whitney U-test was used to determine changes in expression profiles after L. plantarum treatment in vivo. A p-value <0.05 was considered statistically significant.

In peripheral blood samples, we did not observe differences in the frequencies of total CD3⁺ cells, CD3⁺CD4⁺ cells (naïve or memory), or the CD3⁺CD4⁺ activated memory cells after treatment with any of the L. plantarum strains. However, the percentage of CD4⁺Foxp3⁺ cells was significantly decreased following placebo and CIP48 treatment, but not after TIFN101 treatment (Figure 1). Moreover, although we did not observe any effect with indomethacin treatment on CD3⁺CD8⁺ (naïve and memory) cells, activated memory-cells exhibited a statistically significantly decrease after CIP48 treatment (p < 0.05) (Figure 2).

Treatment did not influence percentages of NK-cells or NKT-cells. There was also no change in the percentages of the NK-cell subtypes (i.e., CD56^high and CD56^dim); and the expression of CD161 (KLRB1), mediating cytotoxicity (33, 34), was not affected either by the L. plantarum treatments (results not shown).

T-cell polarization was studied after three types of T-cell stimulation: (i) non-specific polyclonal stimulation with PMA/Ca²⁺ or superantigen Staphylococcus aureus-enterotoxin B (SEB) to study whether the total responsiveness was influenced by L. plantarum treatment, (ii) stimulation with a previously administered vaccine-antigen (TT) to study stimulation of specific memory responses, and (iii) stimulation with cell extracts of the specific L. plantarum strains in order to investigate whether specific immune responses against the L. plantarum were stimulated.

After non-specific stimulation with PMA/Ca-ionophore or SEB, we quantified the percentage of IFNγ, IL-4, IL-17, or IL-21 positive Th cells and memory Th cells. Treatment with placebo or the administered L. plantarum strains did not influence cytokine production of the total population of Th cells or of Th memory cells after non-specific stimulation with PMA/Ca-ionophore. Although no differences were found after SEB-stimulation in cytokine production by the total Th cell population after the three L. plantarum treatments (results not shown), we did observe differences in cytokine production of the Th memory cells after L. plantarum treatment. After stimulation with SEB, we observed a decreased percentage of IL-17-producing activated memory Th cells following treatment with CIP48 and an increased percentage of IL-17-producing activated memory Th cells after treatment with TIFN101 (Figure 3). Moreover, the percentage of IFNγ-producing activated memory Th cells was also increased after TIFN101 treatment (Figure 3).

Treatment with L. plantarum strains also modulated cytokine production following a more specific stimulation by TT (Figure 4). After TIFN101 treatment, the percentage IL-17 and the percentage IFN-γ-producing memory Th cells were significantly increased, while no effect on cytokine production by memory Th cells after TT-stimulation was observed with the other L. plantarum strains.

Finally, we stimulated human blood samples with cell extracts of the L. plantarum strain that the volunteers had consumed in the study (Figure 5). We observed that subjects, who were treated with WCFS, showed a trend toward an increased IL-17 response after stimulation with WCFS cell extracts. Other cytokines were not affected after this treatment. There were no differences in cytokine production in subjects treated with CIP48, when their blood samples were stimulated with CIP48 cell extract. When subjects were treated with TIFN101, their activated memory cells exhibited increased IL-17 and IFN-γ-production following stimulation with TIFN101 cell extract.

Differential gene expression profiles were found in the stressed intestinal mucosa of the subjects consuming each bacterial strain; 315 genes were differentially regulated with L. plantarum WCFS, 390 with CIP48, and 779 with TIFN101, as compared to the placebo intervention (Figure 6A). Of these genes, the different bacterial strains shared only relatively small numbers of upregulated and downregulated genes (Figures 6B,C, respectively). Shared genes were mainly involved in general cellular functions and metabolism.

The 10 genes that were most highly up- or downregulated are listed in Tables 1–3. Of the most highly induced genes after TIFN101 consumption 80% are related to immunity: immunoglobulin lambda variable 6–57, putative V-set and immunoglobulin domain-containing protein 6-like, immunoglobulin lambda variable 7–46, interferon regulatory factor 4, GDNF family, CD27, CD79a, and plasminogen activator. WCFS and TIFN101 shared the downregulation of six small nucleolar RNAs, i.e., snoRNA (H)C/D(ACA) box 6, 14b, 53, 57, 60, 388, while CIP48 had a complete different profile of up and downregulation.

To identify specific transcription factors and to identify pathways regulated by the strains, IPA was performed (Figure 7). TIFN101 induced more changes than CIP48 and WCSF1 in the NSAID-stressed intestine. The most significant set of target-genes in the TIFN101 group were immune response-related genes. TIFN101 upregulated MHC-IIα, while CIP48 and WCFS downregulated MHC-IIβ (Figure 8). Another pathway that might contribute to the enhanced responses in TIFN101 is the upregulation of genes involved in leukocyte extravasation (Figure 9). TIFN101 enhanced RAPL expression, which is a GTPase involved in regulating integrin affinity. Concomitantly, an upregulation of essential adhesion molecules such as ICAM-1 and Cadherin 5 was seen, illustrating the upregulation of immune-cell migration pathways by TIFN101. Also, some regulation of leukocyte extravasation was observed with CIP48 and WCFS. However, it was much less pronounced than for the TIFN101 intervention.

Lactobacillus plantarum strains CIP48 and TIFN101 were sequenced, annotated, and compared with the genome (chromosome and plasmids) of L. plantarum WCFS1 (35, 36). A total of 3,010 OGs were assigned to the chromosome, based on the ordering of contigs to the template WCFS genome. The three genomes shared 2,455 of the 3,010 chromosomal OGs (=81.5%), which is defined as the core genome for this study. Figure 10 presents the shared and genes and contigs of the L. plantarum strains. When the contigs and OGs/genes are included that do not match to the WCFS chromosome, higher numbers of unique genes were found for the CIP48 and TIFN101 genomes. Many of these extra unique genes are on plasmids (Table S2 in Supplementary Material). All unique genes for L. plantarum TIFN101 and L. plantarum CIP104448 compared to WCFS1 are listed in Tables S3 and S4 in Supplementary Material, and all absent genes in the two strains are listed in Tables S5 and S6 in Supplementary Material, respectively, as these are potential candidate genes for the biological effects of the two strains. L. plantarum CIP48 lacks the complete plantaricin biosynthesis gene cluster, a large set of genes adjacent to this cluster (i.e., OGs 334–348), and the entire gene cluster for EPS biosynthesis. L. plantarum TIFN101 is missing some genes associated with plantaricin biosynthesis as well as genes for exopolysaccharide biosynthesis, many sugar utilization cassettes, and two large LPXTG-anchored mucus-binding proteins (Tables S7A,B in Supplementary Material).