Identification of proteins involved in the anti-inflammatory properties of Propionibacterium freudenreichii by means of a multi-strain study

Abstract

Propionibacterium freudenreichii, a dairy starter, can reach a population of almost 10⁹ propionibacteria per gram in Swiss-type cheese at the time of consumption. Also consumed as a probiotic, it displays strain-dependent anti-inflammatory properties mediated by surface proteins that induce IL-10 in leukocytes. We selected 23 strains with varied anti-inflammatory potentials in order to identify the protein(s) involved. After comparative genomic analysis, 12 of these strains were further analysed by surface proteomics, eight of them being further submitted to transcriptomics. The omics data were then correlated to the anti-inflammatory potential evaluated by IL-10 induction. This comparative omics strategy highlighted candidate genes that were further subjected to gene-inactivation validation. This validation confirmed the contribution of surface proteins, including SlpB and SlpE, two proteins with SLH domains known to mediate non-covalent anchorage to the cell-wall. Interestingly, HsdM3, predicted as cytoplasmic and involved in DNA modification, was shown to contribute to anti-inflammatory activity. Finally, we demonstrated that a single protein cannot explain the anti-inflammatory properties of a strain. These properties therefore result from different combinations of surface and cytoplasmic proteins, depending on the strain. Our enhanced understanding of the molecular bases for immunomodulation will enable the relevant screening for bacterial resources with anti-inflammatory properties.

Figure 1. In vitro immunomodulatory phenotypes of Propionibacterium freudenreichii strains grown on dairy-based culture medium.

Comparative anti-inflammatory (A) IL-10 and pro-inflammatory (B) IL-12, (C) IFN-γ and (D) TNF-α cytokine responses of human PBMCs stimulated by Pf strains or by the positive controls Lactococcus lactis MG1363 and Bifidobacterium longum BB536 (pro- and anti-inflammatory strains, respectively). All bacterial strains were implemented at a bacterial density corresponding to a multiplicity of infection (MOI) of 10, and the supernatants collected were analysed using ELISA. Immunocompetent PBMC cells were stimulated by bacteria for 24 h and the data are expressed in pg/ml as means +/− SEM of the results concerning four separate french healthy blood donors. The Pf strains are ordered by their decreasing induction of IL-10, which is highly variable.

Figure 2. Maximum likelihood phylogenetic tree (JTT + I + G) of the Propionibacterium freudenreichii strains used in this study.

The 12 strains shown in boxes were used in the proteomics study and the eight strains with a grey background were also used in the transcriptomics study. IL-10 levels were normalised within each donor using B. longum (BB536) induced levels as a reference and averaged over four donors.

Figure 3. Diversity of Propionibacterium freudenreichii surface proteome and the distribution of identified surface proteins into in silico predicted localisations.

(A) Guanidine hydrochloride-extracted proteins, analysed by SDS PAGE, evidenced variability among the Pf strains. (B,C). An inventory of the surface proteins of 12 Pf strains was further made by combining the three surface proteomic methods and produced a total of 509 proteomic identifications representing 174 different proteins. (B) Global distribution of in silico predicted localisations of the 174 different proteins identified. (C) Number and predicted localisations of the proteins identified using each analytical method: shaving, cyDye surface labelling and guanidine hydrochloride extraction (Le Maréchal et al.19). The localisations were predicted by SurfG + software, as described by Barinov et al. and the categories were as follows: (C) cytoplasmic protein, PSE: Protein surface exposed, M: membrane protein, SEC: secreted protein.

Figure 4. Venn diagram of Propionibacterium freudenreichii candidate genes identified as potentially being associated with anti-inflammatory properties, using omics methods (genomics, transcriptomics and surface proteomics).

The numbers of candidates are those for which the regression model provides an adjusted p-value < 0.05 for the association. The regression model also provides a sign (+) or (−) corresponding to positive and negative associations with the induction of IL-10. (+) clusters are associated with IL-10 induction by anti-inflammatory strains whereas the reverse is true for (−) clusters. The names of the genes are specified for those mentioned in the text of the publication and in Table 1.

Figure 5. Identification of the Propionibacterium freudenreichii strains containing the genes coding for the candidate proteins identified by a multi-omics approach.

The columns highlighted in grey correspond to the strains selected for the transcriptomic analysis. The numbers indicated in the Table refer to the locus tag of the genes; for example merA in strain CIRM 129 refers to the PFCIRM129_03920 gene.

Figure 6. Immunological consequences of the inactivation of selected genes.

Panel (A) Comparative IL-10, TNF-α, IFN-γ and IL-6 cytokine responses of human PBMCs stimulated by the two strains used as positive controls Lactococcus lactis MG1363 (pro-inflammatory) and Bifidobacterium longum BB536 (anti-inflammatory) or by two wild type Propionibacterium freudenreichii strains, CIRM 121 and CIRM 129. Panel (B) Cytokine induction by the mutant strains inactivated for the five genes positively associated with IL-10: slpB, slpE, slpF, hsdM3 and pep, expressed as a percentage of the strong anti-inflammatory wild type CIRM 129 strain. Panel C. Cytokine induction by mutant strains inactivated for eno1, htrA4, pouf8235, negatively associated with IL-10, and expressed as a percentage of the weak anti-inflammatory wild type CIRM 121 strain. Data were expressed as means +/− SEM. Symbols: *P < 0.05.