Feeding of Enterococcus faecium NCIMB 10415 Leads to Intestinal miRNA-423-5p-Induced Regulation of Immune-Relevant Genes

Animal experiments.The local state office of occupational health and technical safety, Landesamt für Gesundheit und Soziales Berlin (LaGeSo regulation no. 0347/09), approved the study according to international guidelines (EU directive 86/609/EWG), ensued by German law on animal welfare (§8 TierSchG).

The study design, including animals, housing, and feeding conditions, was described previously (10). Briefly, purebred Landrace sows and their piglets were divided into two groups: one group received a diet with the addition of E. faecium NCIMB 10415 (Cylactin; Cerbios-Pharma SA), and another group received no feed intervention and served as controls. The sows were randomly assigned to either the control group (n = 8) or the probiotic group (n = 8) and housed under similar conditions in separate buildings to avoid transfer of the probiotic strain between the treatment groups. E. faecium NCIMB 10415 was mixed into the sows' diet at 4.2 × 10⁶ to 4.3 × 10⁶ CFU/g. Probiotic-supplemented feed was introduced 28 days antepartum and continued until weaning of piglets. Piglets had access to a nonmedicated prestarter diet from the age of 12 days on. The prestarter diet of piglets from the probiotic group contained 5.1 × 10⁶ CFU of E. faecium/g of diet. Piglets were weaned at 26 ± 2 days. Afterwards, the piglets in the probiotic treatment group received a starter diet with 3.6 × 10⁶ CFU/g.

Sampling and RNA isolation.Piglets from each litter per group (probiotic/control) were euthanized at days 12 ± 1 (n = 6), 34 ± 1 (n = 6), and 54 ± 2 (n = 8) for tissue sampling and narcotized with ketamine-azaperone (4 mg/kg [body weight] ketamine; 25 mg/kg azaperone). Following blood sampling, the piglets were euthanized by intracardial injection of a lethal dose of tetracaine hydrochloride, mebezonium iodide, and embutramide (T61; Intervet, Unterschleißheim, Germany). Incision and dissection of the piglets were described in more detail previously (10). Peyer's patches (PPs) and lymph nodes (LNs) were collected from the ileum (ileal PPs [ILPPs] and ileal LNs [ILLNs]) and jejunum (jejunal PPs [JePPs] and jejunal LNs [JeLNs]), rinsed with Hanks' buffered salt solution (HBSS), and shock-frozen in liquid nitrogen. JeLNs and ILLNs were removed from connective tissue and shock-frozen in liquid nitrogen. Total RNA containing small and large RNAs was isolated from the above-mentioned tissues according to the manufacturer's protocols (NucleoSpin miRNA; Macherey & Nagel). Therefore, 30 mg of tissue samples was homogenized in the provided lysis buffer including β-mercaptoethanol in screw-cap tubes containing metal beads with the Fast-Prep 24 instrument (MP Biomedicals, Eschwege, Germany), using 3 bursts of 20 s each at 4.5 m/s. Quantification of RNA was performed by using a NanoDrop instrument (Thermo Fisher) (A₂₆₀/A₂₈₀ ratio of >2.0). Quality was checked on a bioanalyzer. For transcriptome sequencing (RNA-Seq), samples with an RNA integrity number (RIN) of >9 were used. These samples and samples from additional animals, which had to have a RIN of >7.5, were taken for testing of expression by quantitative PCR (qPCR). If a sample did not reach the required RIN, RNA isolation was repeated for this sample.

Analysis of RNA sequencing data.Next-generation sequencing (NGS) using the Illumina HiSeq 2000 platform was carried out for miRNA and, in parallel, mRNA to analyze differential expression in the porcine intestine upon feeding of E. faecium. For RNA-Seq, we used pooled RNA samples. We first isolated RNA from four different tissues (jejunal and ileal lymph nodes and Peyer's patches) from randomly chosen piglets at the age of 34 ± 1 days, and we then pooled RNA from the same tissues of three animals belonging to the same feeding group. To analyze gene expression, reads were preprocessed by trimming of low-quality 3′ tails, removal of matching poly(A) sequences, and extraction of residual PCR-Primer-Plus adapter sequences (match of ≥80%). The remaining reads were aligned against 27,928 porcine mRNAs (downloaded from Ensembl v73) by using Bowtie 2 (22) with the parameter sensitive-local. The resulting alignments were quantified by using the eXpress v1.5.0 software package (http://bio.math.berkeley.edu/eXpress/). The quantified transcript reads were taken as an approximation of gene expression.

Additionally, in separate sequencing of the same sample, the fraction of small RNA, including miRNA, was filtered out, according to a minimum length of 15 bp and non-poly(A) sequences. The maximal observed read count per mapped miRNA sequence within a miRNA identification cluster was taken for the calculation of the fold change in expression. The fold change was defined as the ratio of read counts for the probiotic group to that for the control group upon requiring a minimum support of 100 reads. A cutoff of a 1.5-fold upregulation of a miRNA in the probiotic sample was applied for each tissue separately.

Statistical analysis of tissue-specific gene expression.Computational analysis of differential expression between tissues (ILLNs, ILPPs, JeLNs, and JePPs) was performed based on negative binomial distribution implemented in the DEseq2 (v1.0.19) R package (23). Due to the experimental design, different models using subsets of the data were analyzed. Differences in expression levels between probiotic and control samples taken from the ileum and the jejunum were analyzed by using the model Y = tissue type + diet group + e. Tissue type discriminates between the tissue types, Peyer's patches or lymph nodes, from which lymphocytes were isolated for RNA preparation. Differences in expression levels between probiotic and control samples taken from Peyer's patches and lymph nodes were analyzed by using the model Y = localization + diet group + e. Localization describes the localization of the tissue type, Peyer's patches and lymph nodes, along the small intestine. In addition, a full model was run, which included all available samples and compensated for all covariates available, to analyze the difference between probiotic and control animals: Y = localization + tissue type + diet group + e. Due to the small sample size of NGS samples (eight samples, pooled from three individuals each), this model does not include interactions. Y is the normalized gene expression level through RNA read counts mapped for a known transcript (an Excel file including tables of all differentially expressed genes using the different models is available upon request).

miRNA target prediction.Reads that contained poly(A) sequences and were <15 bp were discarded from the RNA-Seq data (as described above) to enrich the data for small RNA reads. Unique sequences were counted for each sample, and only those that were represented at least two times per sample were kept. miRNA reads were then mapped to known Sus scrofa miRNA sequences from miRBASE. In total, 207 known porcine miRNAs were present in the samples and subjected to a target scan using 5′- and 3′-untranslated region (UTR) sequences of the porcine mRNAs. Sequences of 1,000 bp up- and downstream were appended to mRNA sequences without annotated 5′ and 3′ UTRs. The target scan was performed by using the stand-alone version of PITA (24) and miRanda (20).

RT-qPCR for quantification of miRNA and mRNA.Quantification of miRNA was done by a poly(A) technique according to the manufacturer's protocols (miRNA first-strand cDNA synthesis kit; Agilent Technologies). The poly(A) technique includes elongation of miRNAs carried out by a specific poly(A) polymerase (2 U/μl) isolated from E. coli, which elongates the miRNA at the 3′ tail using ATP as the substrate. Therefore, 1 μg of total RNA was polyadenylated for 30 min at 37°C and terminated at 95°C for 5 min in a total volume of 10 μl. In a second step, 10 μl of the elongated miRNA was reverse transcribed by using AffinityScript reverse transcriptase and a universal reverse primer (3.125 μM) in a 20-μl volume. After incubation for 5 min at 55°C and 15 min at 25°C, reverse transcription (RT) was carried out at 42°C for 30 min. Heat inactivation at 95°C for 5 min led to the inactivation of RT. Afterwards, the reaction mixture was diluted in 260 μl RNase-free water. Following reverse transcription, RT-qPCR (Viia 7; Life Technologies) was performed with three technical replicates per sample and two replicates for reference miRNAs by using miRNA qPCR master mix, including reference dye (1:50) and the universal reverse primer provided by Agilent Technologies. One microliter of the reverse transcript miRNA reaction mix served as the template in a 384-reaction plate in 40 cycles. Specific primers used for amplification are given in Table 1. The amplification program comprised an initial step at 95°C for 10 min, followed by an annealing step and an elongation step for 1 min at 60°C. Porcine miR-17, miR-103, and miR-17 served as reference miRNAs (25).

For the quantification of mRNA transcripts, 1 μg of total RNA in 10 μl was used for reverse transcription into cDNA by using the oligo(dT) primer according to the manufacturer's protocols for the AffinityScript qPCR cDNA synthesis kit (Agilent Technologies). For qPCR analysis, 10 ng of initial total RNA served as the template, samples were amplified in triplicates, and reference genes were amplified in duplicates. The reaction was done by using 5 μl of SYBR Select master mix containing AmpliTaq DNA polymerase and 6 μmol specific primers (Table 1) in a volume of 10 μl in a 384-reaction plate using the Viia 7 device (Life Technologies). An initialization step was conducted for 2 min at 50°C and 10 min at 95°C, followed by 40 cycles of denaturation for 15 s at 95°C and annealing/elongation for 1 min at 60°C. Ribosomal protein L-19 (RPL-19) and TATA-box-binding protein 1 (TBP-1) were chosen as reference genes, since previous studies showed high stability in porcine tissues (26). RNA samples excluding the RT reaction and water served as mock controls. The relative expression level was calculated by the ΔΔC_T method (27) and was described in detail previously (28).

Transfection and luciferase assays.Transfection was performed by using the human cervix carcinoma cell line HeLa (ATCC CCL-2) cultivated in RPMI 1640 (Biochrom AG) supplemented with 10% superior fetal bovine serum (Biochrom AG) and 10 μg/ml l-glutamine. The cell culture was incubated in 75-cm² flasks (Greiner Bio-One GmbH) at 37°C in 5% CO₂ and passaged once per week. For transfection, Nucleofector technology (Lonza AG) was used as described in more detail previously (21). The vector pTKGluc (NEB GmbH) was used for the generation of reporter plasmids. Synthetic oligonucleotides (Invitrogen) containing the target sequence/control (Table 1) and the NotI (5′) and XbaI (3′) restriction sites were hybridized and cloned into pTKGluc by using methods comparable to those described previously by Hoeke et al. (21). As a first step, pTKGluc was linearized by using the restriction enzymes NotI and XbaI (NEB GmbH). Fifty nanograms of the linearized vector and 1 pmol of the hybrid were ligated overnight at 16°C by using T4 ligase (NEB GmbH), followed by heat inactivation at 65°C for 10 min, and transformed into E. coli B6. Selection of positive clones was verified by Sanger sequencing (ABI Prism 310 genetic analyzer; Applied Biosystems). Endotoxin-free plasmids were created for transfection according to the manufacturer's protocols for the NucleoBond Xtra Midi kit (Macherey-Nagel GmbH & Co. KG). Cotransfection of the reporter plasmid and miRNA mimics as well as measurement of luciferase activity were described previously (29).