miR-125a-5p regulates the sialyltransferase ST3GAL1 in murine model of human intestinal campylobacteriosis

We have recently shown that the two miRNAs, miR-125a-5p and miR-615-3p, are dysregulated after intestinal C. jejuni infections [10]. For deciphering regulatory mechanisms of these miRNAs on mucin-type O-glycan biosynthesis, relevant human and murine target genes of the miRNAs were identified followed by KEGG pathway enrichment analysis [20] as described earlier [10, 12]. While only focussing on murine targets in the previous study, we have now expanded the enrichment analysis for both murine and human target genes. As shown in Fig. 1A, potential targets of both miRNAs within the mucin-type O-glycosylation pathway being conserved in both species were St3gal1 and β1,4-galactosyltransferase B4galt1. However, St3gal2, Gcnt4 and Gcnt1 were only targeted by both miRNAs in mice, whereas the glucuronyltransferase B3gnt1 was only a target in humans. In Additional file 1, all overlapping potential targets of miR-125a-5p and miR-615-3p in human and mouse are listed, including their pathways as determined by KEGG pathway enrichment analysis. As shown in Fig. 1B, the analysis pointed out specific sites of mucin-type O-glycan biosynthesis by the targeted transferases in mice and humans. While ST3GAL1 and 2 terminate the extension of the glycan chain by sialylation, B4GALT1 extends the core 2 structure by adding galactose residues. In Fig. 1B the intermediate steps in mucin-type O-glycosylation catalysed by ST3GAL1, ST3GAL2 and B4GALT1 were indicated (red asterisks), thereby forming potential interfaces with miR-125a-5p and miR-615-3p during C. jejuni infection. In previously published results, St3gal2 was shown to be a target of miR-615-3p [10]. Furthermore, a clear anticorrelation between the expression of St3gal1 and miR-125a-5p in the context of C. jejuni infection in vivo was observed. Therefore, we have mainly focused on this interaction in the analyses carried out here. Through RNAhybrid analysis [21], we have been able to identify two target sites of miR-125a-5p within the 3’ UTR of murine St3gal1 (Fig. 1C). Both target sites possessed canonical interaction characteristics with minimal free energies ≤ − 22.7 kcal/mol.

We employed the murine intestinal cell line CMT 93 (ECACC 89111413) to examine the predicted interaction between miR-125a-5p and St3gal1 in vitro. The efficiency of miR-125a-5p transfection was determined by measuring its cellular levels by RT-qPCR. After transfection, CMT 93 showed approximately tenfold increased and significant (P ≤ 0.01) miR-125a-5p concentrations compared to non-target transfected controls (Fig. 2A). Since miRNAs are involved in gene silencing, anti-correlated expression of the target is anticipated after RNAi and indicates an interaction. We found that RNAi with miR-125a-5p mimics caused significantly decreased expression of St3gal1 by 0.7-fold (P ≤ 0.05) in CMT 93 compared to non-target transfected controls (Fig. 2A). This confirmed our in silico analysis and pointed to St3gal1 as a target of miR-125a-5p in mice.

Furthermore, we tested whether increased cellular miR-125a-5p levels had an effect on the transcripts of the potential targets St3gal2 and B4galt1 as well as the associated miR-615-3p. Although B4galt1 was significantly downregulated 0.82-fold (P ≤ 0.05), no significant differences in St3gal2 expression levels were observed between mimics and controls, though a trend towards decreased mRNA levels was found (Fig. 2A). Hence, these results point to B4galt1 as a further target of miR-125a-5p. Interestingly, miR-615-3p also exhibited a trend towards increased expression (not significant) after miR-125a-5p transfection, suggesting synergistic effects of both miRNAs (Fig. 2A).

After screening the 3′ UTR of murine St3gal1 for miR-125a-5p interaction sites by means of RNAhybrid [21], two binding sites were determined. The calculated minimal free energies for the interactions with miR-125a-5p were − 22.7 kcal/mol and − 22.8 kcal/mol, respectively (Fig. 1C). Dual reporter gene assays considering both identified St3gal1 binding sites were performed as previously described [12]. After co-transfection of miR-125a-5p mimic together with the reporter plasmid (pTKGmSt3gal1) and the normalisation plasmid (pCMV-Red Firefly Luc), a significant 0.68-fold decrease (P < 0.01) in normalised luciferase activity (Luc_Gaussia: Luc_{Red Firefly}) was measured compared to non-target controls (Fig. 2B). The downregulated magnitude of the reporter gene fused to St3gal1 binding sites was comparable to the reduction of St3gal1 transcript levels in transfected CMT 93 cells by miR-125a-5p mimics. This experiment thus proved that the interaction between St3gal1 binding sites and miR-125a-5p is specific.

To investigate the physiological relevance of the above determined interaction between miR-125a-5p and the sialyltransferase St3gal1 in vivo, tissue samples of secondary abiotic IL10 ^−/− mice that had been infected with C. jejuni and naïve controls were compared.

As shown in Fig. 3A, anticorrelated expression of miR-125a-5p and its target St3gal1 was detected after C. jejuni infection. While miR-125a-5p was significantly downregulated by 0.6-fold (P ≤ 0.0001), a pronounced upregulation of St3gal1 gene expression was observed after infection (> 4-fold; P ≤ 0.0001). Also, the expression of B4galt1 was 1.6-fold upregulated (P ≤ 0.001) in the infected samples (Fig. 3A). Although the RNAi experiments performed above indicated that St3gal2 is not a target of miR-125a-5p, we tested its expression in vivo. As shown in Fig. 3B, St3gal2 was significantly downregulated by 0.4-fold (P ≤ 0.0001), while miR-615-3p was 2.1-fold increased (P ≤ 0.05) confirming our previously published data [10].

The marked and contrasting expression of St3gal1 compared to miR-125a-5p confirmed our in vitro experiments and pointed out the specific role of the identified interaction in the context of intestinal C. jejuni infection in the applied murine model.

Following the interaction and gene expression analysis, the aim was to verify whether the infection-dependent downregulation of miR-125a-5p exerts an effect on the protein concentration and distribution of ST3GAL1 in colonic tissue. For this purpose, both western blots and immunofluorescence staining (IF) were performed with colonic tissue derived from infected and naïve mice, corresponding to the samples used for RT-qPCR analysis. As determined by western blots in pooled samples, ST3GAL1 protein levels were increased in infected tissue compared to respective naïve controls, while the levels of the reference protein GAPDH were similar in both groups (Fig. 4A). To statistically evaluate the magnitude of ST3GAL1 upregulation, individual western blots were performed from eight naïve and eight infected mice that were randomly selected and analysed using densitometry. The 1.7-fold significant increase in relative ST3GAL1 protein levels (P ≤ 0.01) in infected colonic tissue confirmed the transcriptional data (Fig. 4B). Western blot images of individual and pooled mice samples are shown in the Additional file 3.

Besides western blots, IF of ST3GAL1 in the corresponding colonic tissue sections was performed to assess protein abundance as well as localisation confirming the western blot data. An enhanced fluorescent signal of ST3GAL1 was detected in the C. jejuni infected group compared to naïve counterparts (Fig. 4C and Additional file 3), with an abundant cytosolic detection of ST3GAL1 in the enterocytes lining the entire length of the crypts of infected sections. In contrast, ST3GAL1 was more abundant in the upper part of the crypts towards the intestinal lumen in the control tissue. Additionally, ST3GAL1 was mainly localised on the apical side of the cell membrane in the naïve samples, whereas the fluorescence signal in the infected samples was more intense and evenly distributed over the entire cytosol of the cell (Fig. 4C). Images of individual sections from both groups and negative controls of the tissue sections, to ensure the specificity of the detected ST3GAL1 signal, are shown in the Additional file 5.

In vivo experiments were carried out by the Forschungseinrichtungen für Experimentelle Medizin (FEM, Charité-Universitätsmedizin Berlin) and according to the European Guidelines for animal welfare (2010/63/EU) following agreement by the commission for animal experiments headed by the “Landesamt für Gesundheit und Soziales” (LaGeSo, Berlin, registration number G0104/19). Animal welfare was monitored twice a day. For the experiment, 39 secondary abiotic IL10^−/− mice (C57BL/B10 background) were divided into two groups, C. jejuni infected and naïve, matched by age and sex [18]. In total, 19 animals were infected with 10⁹ colony forming units (CFU) of C. jejuni strain 81–176 in 0.3 ml sterile phosphate-buffered saline (PBS, Thermo Fisher Scientific, Waltham, MA, USA) on days 0 and 1 by oral gavage while 20 animals were not infected as controls. The mice were sacrificed six days post infection and ex vivo biopsies of the colon were collected under aseptic conditions. For the gene expression and protein analysis, the colon tissue samples were stored in Allprotect Tissue Reagent (Qiagen, Venlo, Netherlands) or liquid nitrogen, respectively and stored at − 80 °C. For the histopathological analysis, the biopsies were immediately fixed in 5% formalin and embedded in paraffin.

The murine rectal carcinoma cell line CMT 93 (ECACC 89111413) was cultured in a humidified atmosphere at 37 °C and 5% CO₂ in Dulbeccco’s Modified Eagle’s Medium (DMEM) with 4.5 g/l Glucose and l-Glutamine (Gibco, Grand Island, NY, USA). The medium was supplemented with 100 µg/ml Gentamicin and 10% (v/v) fetal calf serum superior (Sigma-Aldrich, Darmstadt, Germany). The cells were grown in 75 cm² tissue culture flasks (Sarstedt, Nümbrecht, Germany) until approximately 80% confluence was reached.

Total RNA was isolated from the murine colonic tissue stored in Allprotect and CMT 93 cells using the miRVana Isolation Kit (Life Technologies, Carlsbad, CA, USA). Then, cDNA was synthesised from individual tissue samples or cell lysates. For this, the isolated total RNA was firstly treated with RNase-free DNase I (NEB GmbH, Frankfurt a/M, Germany) to remove residual genomic DNA. Then, 1 µg total RNA was applied for reverse transcription using 200 U M-MuLV Reverse Transcriptase (Thermo Fisher Scientific), 0.2 μg random hexamer primer (Thermo Fisher Scientific), 200 μM of each deoxynucleotide triphosphate (dNTP) (Thermo Fisher Scientific) and 1 × supplied RT buffer (Thermo Fisher Scientific). To verify the absence of genomic DNA, control samples were treated as just mentioned but without M-MuLV Reverse Transcriptase. Then, RT-qPCRs were carried out. The qPCR of the mRNAs was performed with the Blue S’Green qPCR Kit (Biozym Scientific GmbH, Hessisch Oldendorf, Germany) under following cycling conditions: amplification was conducted at 95 °C for 2 min, followed by 40 cycles with 15 s at 95 °C, 20 s at 60 °C and 30 s at 70 °C. For quality control, subsequent melting curve analysis was assessed by 95 °C for 15 s, 65 °C for 1 min, ramping from 65 to 95 °C at 5 °C/min. For the miRNA qPCR, the SensiFAST SYBR® Hi-ROX Kit (Bioline, Meridian Bioscience International Limited, Cincinnati, OH, USA) was used and cycling conditions comprised real-time analysis at 95 °C for 2 min, followed by 40 cycles with 15 s at 95 °C, 20 s at 60 °C and 30 s at 72 °C. Subsequent melting curve analysis was assessed by 95 °C for 15 s, 60 °C for 1 min, ramping from 60 to 95 °C at 5 °C/min. For RT-qPCR of the miRNAs, the miR-Q method [39] was used with specific RT-6-primer and corresponding reverse PCR primer (Additional file 2). All relative gene expression was calculated using the ∆∆Ct method [40] as described before [10]. Stability of the reference was tested using geNorm [41]. The HPRT and SDHA genes served as references for mRNA expression and the small RNAs SNORD44 and SNORD47 for miRNA expression. The mean Ct values of the target as well as the geometric mean of all Ct-values of the two respective references were calculated for each sample from three technical replicates. The ∆Ct is equal to the difference between the Ct of the target and the geometric mean of both references. The ∆∆Ct were calculated relative to controls.

All oligonucleotides in this study were synthesised by Sigma-Aldrich. Sequences, amplification efficiencies and primer concentrations and can be found in Additional file 2.

Colonic tissue samples derived from the animal experiment were lysed in cold RIPA buffer (Cell Signaling Technology (CST), Danvers, MA, USA), supplemented with protease inhibitor cocktail (CST) and protein was isolated. Protein concentration was quantified by using the 2D Quant Kit (Qiagen). 30 µg of protein was loaded onto a 12% sodium dodecyl sulfate-polyacrylamide (SDS) gel for electrophoresis. Then, they were transferred onto a polyvinylidene fluoride membrane (PVDF) (GE Healthcare, Buckinghamshire, UK) by means of semi-dry blotting. To prevent unspecific binding of the antibody, the membrane was blocked in 5% (w/v) bovine serum albumin (Sigma-Aldrich) in TBST (Tris–HCl-buffer with 0.1% (v/v) Tween-20) for 1.5 h at room temperature. The membrane was incubated with the primary antibody Rabbit anti-ST3GAL1 (Novus Biologicals, bio-techne, Minneapolis, MN, USA, NBP1-62540) 1:500 or Rabbit anti-GAPDH (CST, #5174) 1:2000 in 5% BSA in TBST at 4 °C overnight. After three washes in TBST for 15 min each, the membrane was probed with the secondary horseradish peroxidase (HRP)-linked anti-rabbit IgG antibody (1:2500 in 5% BSA in TBST, CST, #7074) for 2 h at room temperature. After three additional washes with TBST (Tris–HCl-buffer with 0.1% (v/v) Tween-20) for 15 min, immuno-reactive proteins were detected with the Amersham™ ECL Select™ Western Blotting Detection Reagent (GE Healthcare). Exposure conditions were identical for every blot. Protein quantification was performed by densitometry using the software BIO-1D (Vilber, Collégien, France). Here, a defined area was selected to be used on all lanes in each blot for 8 C. jejuni infected and 8 naïve mice colon samples and densitometry was measured. The optical density was determined by the software and the background was subtracted. The pixel density values of the ST3GAL1 signals were then normalised to the respective GAPDH signals and the fold change calculated on the mean of the controls. The raw pixel density values can be found in the Additional file 4.

Murine colon tissue sections from the animal experiment were immediately fixed in 5% formalin and embedded in paraffin. 5 µm serial sections were cut using the Epredia™ HM 340E Electronic Rotary Microtome (Thermo Fisher Scientific). The paraffin was melted at 60 °C for 45 min and the tissue sections deparaffinated in Roticlear (Carl Roth, Karlsruhe, Germany). Rehydration of the tissue was reached through a graded series of ethanol followed by rinsing with distilled water and PBS (pH 7.4, Sigma-Aldrich). For IF, antigen retrieval was carried out by incubating the sections in boiling citrate buffer for 15 min followed by 20 min of cooling down. Non-specific binding was blocked with 1% (v/v) BSA (Sigma-Aldrich) in PBST (0.1% (v/v) Tween-20 in PBS) for 1 h at room temperature. Then, the sections were incubated with a 1:50 dilution of the Rabbit anti-ST3GAL1 primary antibody (Novus Biologicals, NBP1-62540), in 1% (v/v) BSA in PBST overnight at 4 °C. Negative controls were treated with 1% (v/v) BSA in PBST without the primary antibody. After three washes with PBS for 10 min, the primary antibody was detected with goat anti-rabbit IgG DyLight 594 secondary antibody (1:400, diluted in 1% (v/v) BSA in PBST, Thermo Fisher Scientific, #35561) for 1 h at room temperature. Following three washes with PBS, the nuclei were counterstained with 200 ng/ml 4′, 6-diamidin-2-phenylindol (DAPI, Sigma-Aldrich) in PBS for 5 min at room temperature. Subsequently, slides were washed in PBS and mounted with Prolong™ Diamond Anti-fade Mountant (Life Technology). For IF, the Leica DMI6000B inverted microscope with the compatible Leica LAS-X software (Leica, Wetzlar, Germany) was used. All images were taken under identical microscope- and camera-settings. Experiments were carried out with three randomly selected biological replicates per group. Images were taken at least from three random areas per section and more than ten images per area were selected. The background was subtracted in each image.

For both assays, CMT 93 cells were cultured as mentioned above. RNAi transfections of three biological replicates were performed using electroporation with the Nucleofector Technology (Lonza AG, Köln, Germany) as previously described with few modifications [41]. 1 × 10⁶ cells were mixed with 50 pmol of miR-125a-5p miRNA mimic (Qiagen, MSY0000443) or 50 pmol non-target siRNA (Dharmacon Lafayette, CO, USA, D-001810-03-05) as a control. After 24 h, the cells were washed with PBS and lysed for RNA isolation.

The dual luciferase reporter assays were carried out as previously mentioned [10, 41]. For the generation of the reporter plasmids, the in silico identified target sites of murine St3gal1 were assembled using the hybridised oligonucleotides NotI-mmuSt3gal1ts-sense and XbaI-mmuSt3gal1ts-antisense from Sigma-Aldrich (Additional file 2). These target sites were cloned in pTK-Gluc (NEB GmbH) plasmid using the appropriate restriction enzymes NotI and XbaI (both NEB GmbH). For endotoxin-free reporter plasmids of the resulting derivative pTKGmSt3gal1, NucleoBond Xtra Midi Plus EF (Macherey–Nagel GmbH & Co. KG, Düren, Germany) was used. For the luciferase assay, 3.8 × 10⁵ CMT 93 cells were inversely transfected with 3.25 µg of endotoxin-free pTKGmSt3gal1 and for normalisation, 350 ng pCMV-Red Firefly Luc (Thermo Fisher Scientific, 16156) together with 200 pmol miR-125a-5p miRNA mimic (Qiagen) on a 6-Well Plate (Sarstedt) using 7.5 µl Lipofectamin 3000 (Thermo Fisher Scientific) and 2.5 µl P3000 Reagent (Thermo Fisher Scientific). The transfected cells were cultured for 48 h. As a negative control, 200 pmol non-target siRNA (Dharmacon Lafayette) was transfected. Gaussia-Firefly Luciferase activity was detected in three biological and three technical replicates with the Pierce Gaussia-Firefly Luciferase Dual Assay Kit (Thermo Fisher Scientific). For the dual luciferase reporter assays, the CLARIOstarPlus plate reader (BMG Labtech, Ortenberg, Germany) was used under following conditions: top optic, number of multichromatics: 2, number of kinetic widows: 1, number of cycles: 1, measurement interval time: 1.27, emission Luciferase-Red Firefly: 623 nm, emission Luciferase-Gaussia: 485 nm.

Mutual targets of miR-125a-5p and miR-615-3p in human and mouse were identified as well as data and pathway analysis performed as described earlier [10, 12]. Briefly, overlapping targets of both miRNAs were determined for mouse and human by miRmap followed by KEGG based pathway enrichment analysis. Conserved mutual targets involved in ‘Glycosaminoglycan biosynthesis—keratan sulfate’ (hsa00533 and mmu00533) and mucin type O-glycan biosynthesis (mmu00512) were determined. RNAhybrid was used to predict target sites of miRNA [21]. Normal distribution of data was assessed by the Anderson–Darling test. The unpaired t-test was used in this study to compare each treatment with control. All tests were conducted applying GraphPad Prism version 8.00 (GraphPad Software, La Jolla California USA, www.​graphpad.​com). Asterisks in figures summarize P values (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001).