We have previously shown that a 48-h infection of human macrophages by H. pylori leads to a reduction of HLA-II molecules on macrophage surface (12). To confirm this evidence, we performed a time-course analysis, extending the time of infection to 72 h. As shown in Figure 1A, control cells maintained the amount of HLA-II on the plasma membrane at a basal level until 48 h of culture; at this time point, the number of exposed molecules doubled and further increased at 72 h. We interpreted this trend as a consequence of the replacement of the medium containing M-CSF with a medium deprived of the growth factor at the beginning of the experiment. In accordance, after removing M-CSF, macrophages progressively acquired an M1-like profile with high expression of CD86 and low expression of CD206 (Supplementary Figure 1). In infected cells, the amount of HLA-II expressed on the cell surface was comparable to that of uninfected ones during the first 24 h of infection; at 48 h, it increased but reached only 60% of controls. Seventy-two hours after the addition of bacteria to the cells, HLA-II was 50% lower than in uninfected cells, despite the fact that extracellular bacteria disappeared within a few hours after the infection. Moreover, while the infection also maintained low the expression of CD86, it increased that of CD206 (Supplementary Figure 1), in accordance with previous data (16).

We reasoned that this effect could rely either on a hindrance to the export of the presenting complex or on a blockage of its synthesis. We exposed macrophages to the bacterium, and we evaluated the HLA-II content in cell lysates. Similar to what we observed in terms of protein expressed on the plasma membrane of macrophages, after an initial small increase at 6 h, the total content of HLA-II in infected cells regained the basal level at 24 h, before rising again at 48 h and even more at 72 h. Nevertheless, at this time frame, the difference between infected and uninfected cells was impressive (Figure 1B). This observation ruled out any interference in the trafficking of HLA-II from the ER/Golgi toward the plasma membrane and opened the possibility that the bacterial infection affected HLA-II gene expression. The latter is finely tuned by the intervention of several transcription factors (17), but the master regulator is the CIITA (18). CIITA is expressed constitutively in antigen-presenting cells, including dendritic cells, macrophages, and B cells and is inducible in most other cell types (19). The absence of CIITA prevents the expression of HLA-II genes (20–22). In agreement with the time-dependent increase of HLA-II, the mRNA expression for CIITA progressively augmented in control cells; on the contrary, bacterial infection kept the expression of mRNA for CIITA to a basal level until 48 h, when it started to slightly increase, reaching a value at 72 h, which was less than half of that in control cells (Figure 1C).

The effect of H. pylori infection on HLA-II cell content and exposure as well as on CIITA mRNA expression was evidenced in cells that, for a change in culture condition (removal of M-CSF), progressively increased the expression of both the regulator and the antigen-presenting molecule. To exclude that what we were observing was affected by the experimental condition, we exposed M-CSF-differentiated macrophages with IFN-γ for 6 h before infecting or not the cells with the bacterium for 48 h. IFN-γ is a potent inducer of HLA-II expression and CIITA is the essential coactivator for both constitutive and IFN-γ-inducible HLA-II expression (18, 23, 24). The time frame of 6 h was the minimum time to appreciate the cytokine-induction of HLA-II expression on the surface of macrophages (data not shown).

Results illustrated in Figure 2A reveal that, while the activation of macrophages with the cytokine increased the expression of HLA-II on the plasma membrane with respect to unstimulated cells, the infection with H. pylori completely abrogated the effect. Moreover, the effect of H. pylori on HLA-II was so strong that the expression of HLA-II in IFN-γ-treated cells was maintained at a level lower than that in untreated cells (p ≤ 0.05) (Figure 2A). The same pattern was reproduced with the total cell content of HLA-II and with the expression of mRNA for CIITA (Figures 2B,C). It is worthy to note that the expression profile of CIITA mirrored the lower expression of HLA-II in IFN-γ-treated cells with respect to untreated cells (p ≤ 0.01) (Figure 2C).

IFN-γ is the signature cytokine of the Th1-polarized immune response, which features the H. pylori-infected gastric mucosa of humans (6, 25) and is the major activator of macrophages.

Based on our data, which supported the notion that the bacterium might inhibit the activity of IFN-γ in vivo, we were curious about whether H. pylori was able to hamper the effect of IFN-γ even in macrophages already infected with H. pylori since they are expected to present bacterial antigens to Th1 cells and become activated by the cytokine. Macrophages were infected 16 h with the bacterium and then exposed to IFN-γ. After 48 h, HLA-II exposed on the plasma membrane and its total cell content, as well as the expression of the mRNA for CIITA were determined. As above, H. pylori infection interfered with the IFN-γ signaling, which is essential for augmenting the antigen presentation capacity of macrophages (Figures 2D–F). Even in this case, the modulation of the expression of both HLA-II protein and CIITA mRNA was concordant in infected cells and the effect was more pronounced in IFN-γ-treated cells compared to untreated ones (p ≤ 0.05) (Figures 2D,F).

Collectively, these findings suggested that H. pylori, by hindering the expression of CIITA in macrophages, affected the synthesis of HLA-II molecules, even in the presence of a strong inducer, such as IFN-γ. The decisive proof came from the quantification of the HLA-II synthetic rate in uninfected and infected macrophages, pre-treated or not with IFN-γ. Results shown in Figure 2G demonstrate that H. pylori inhibited the synthesis of HLA-II both in the absence and in the presence of the activating cytokine.

To define mechanism(s) by which H. pylori antagonized the synthesis of CIITA, we considered the capacity of the bacterium to modulate miRNA expression in macrophages (12). We evidenced that the expression of 142 miRNAs was affected in macrophages by H. pylori infection (Supplementary Table 1 and Supplementary Figure 2). Among miRNAs up-regulated in both tested conditions (24 and 72 h of H. pylori infection), 19 were predicted to target CIITA (let-7f-5p, let-7i-5p, miR-125a-5p, -140-5p, -146a-5p, -146b-5p, -16-5p, -185-5p, -20b-5p, -223-3p, -26b-5p, -27b-3p, -29b-3p, -331-3p, -374a-5p, -424-5p, -4284, -590-5p, and -98-5p) (Figure 3A and Supplementary Table 2). Compelling evidence have demonstrated that miRNA expression is deregulated in human cancer and that this affects the hallmarks of cancer, including sustaining proliferative signaling, evading growth suppressors, resisting cell death, activating invasion and metastasis, and inducing angiogenesis (26, 27). Moreover, considering that H. pylori infection is the major risk factor associated with the development of GC (28, 29), and that a prevalence of M2-skewed tumor-infiltrating macrophages (TAM) expressing low HLA-II is associated with a poor prognosis (30), we prioritized miRNAs according to their deregulation in GC. Differentially expressed miRNAs in GC were identified using a meta-analysis approach basing on data published on GEO database (GEO ID: GSE26595, GSE23739, and GSE93415). By this method, we considered different techniques to profile miRNA expression and a total of 110 GC patients and 68 controls (Supplementary Table 3). The high number of deregulated miRNAs in GC reinforced the notion that miRNAs are crucial in the initiation and progression of tumors [Supplementary Table 4 and Supplementary Figure 3; (31)]. We also considered GSE78091 and GSE54397 gene sets, but we did not obtain any statistically significant differences among miRNAs expressed in normal and cancer samples. We evidenced that without distinguishing between the two GC histotypes (diffuse or intestinal) and among different cancer grades (GSE23739 and GSE93415 datasets), 15 miRNAs up-regulated in tumor samples from both datasets were also up-regulated in macrophages infected with H. pylori for 24 or 72 h (Figure 3B and Supplementary Figure 3). On the contrary, considering tumor histotype and grade (GSE26595), only 10 miRNAs up-regulated by H. pylori infection resulted modulated in both diffuse and intestinal GC. Among them, seven miRNAs were up-regulated in both infected macrophages and in GC (let-7f-5p, let-7i-5p, miR-146b-5p,−185-5p,−27b-3p,−374a-5p, and−98-5p), while three had an opposite behavior (miR-140-5p,−20b-5p,−29b-3p) (Figure 3B and Supplementary Figure 3). To assess the role of miRNAs induced by H. pylori in the regulation of CIITA, we considered all seven miRNAs with concordant expression in macrophages infected with H. pylori and in all datasets considered for meta-analyses.

To validate CIITA as target of the seven considered miRNAs not only in macrophages but also in different cells, and considering the low transfection efficiency of macrophages, we adopted two different cell models: a melanoma cell line (M121224) stimulated with IFN-γ (24) and HeLa cells stably expressing CIITA (HeLa-CIITA) (32). M121224 were stimulated 3 h with IFN-γ before transfecting for 48 h with plasmids encoding single miRNAs. Let-7f-5p, let-7i-5p, miR-146b-5p, and -185-5p significantly down-regulated the expression of CIITA, while no changes in the expression of CIITA were detected in cells transfected with miR-27b-3p, -374a-5p, and -98-5p (Figure 4A). The pattern was recapitulated by the surface expression of HLA-II (Figure 4B). The four effective miRNAs were used to transfect HeLa-CIITA cells, and the above findings were confirmed, in terms of both CIITA mRNA expression and HLA-II expression (Figures 4C,D). The overexpression of the four miRNAs was confirmed by qRT-PCR in both M1212224 and HeLa-CIITA transfected cells (Supplementary Figure 4) and the interaction between let-7f-5p, let-7i-5p, miR-146b-5p, and -185-5p and CIITA 3′-UTR was substantiated by luciferase assay (Figure 4E).

qRT-PCR on macrophages infected with H. pylori confirmed the up-regulation of let-7f-5p, let-7i-5p, miR-146b-5p, and -185-5p, evidenced by the microarray analysis. The treatment of macrophages with IFN-γ that stimulates CIITA transcription did not affect the expression of any of these miRNAs. Interestingly, H. pylori infection induced the up-regulation of all four miRNAs in macrophages treated with IFN-γ similarly to those non-treated with IFN-γ or even more (Figure 5A). Therefore, IFN-γ does not affect the ability of H. pylori to stimulate the expression of miRNAs regulating CIITA in macrophages. Notably, the expression of HLA-II in infected macrophages was rescued by inhibiting the action of let-7i-5p, miR-146b-5p, and miR-185-5p (Figure 5B). Let-7f-5p did not display differential expression pattern between normal mucosa and H. pylori-associated diseases (Figure 6); thus, we excluded it from miRNA inhibition experiments.

To confirm the results obtained by meta-analysis approach on miRNA expression in GC (110 patients with GC and 68 controls) and to investigate whether our in vitro data have an in vivo correlation, we analyzed the expression of let-7f-5p, let-7i-5p, miR-146b-5p, and -185-5p in tissues from patients with a histological and serological established diagnosis of H. pylori infection. We considered 10 patients with active non-atrophic chronic gastritis, 9 patients with intestinal metaplasia, 11 with dysplasia (low and high grade), and 9 patients with gastric adenocarcinoma, thus including all the lesions that define the carcinogenetic process of the intestinal variant of GC (i.e., the Correa's cascade). The expression of let-7i-5p, miR-146b-5p, and -185-5p was increased in GC samples and the early neoplastic condition (dysplasia), confirming results obtained by meta-analysis approach and in accordance to several works demonstrating the importance of these miRNAs in different tumors [Figure 6A; (33–38)]. Despite meta-analysis results that revealed that let-7f-5p increased, we did not confirm such an alteration in our samples. Let-7f-5p has been found silenced in metastatic GC in association with an increased cell migration and invasion (39). The apparent discrepancy between our evidence and that published could be explained by the fact that all considered cancers in our series are of early stage. Among the three up-regulated miRNAs, miR-146b-5p was the most up-regulated in our patients and we revealed that it is mainly expressed by macrophages and, to a less extent, by plasma cells and epithelial cells (Figure 6B).

Investigation has been conducted following the ethical standards, the Declaration of Helsinki, and national and international guidelines and has also been approved by the authors' institutional review board (4406/AO/18, Comitato Etico per la Sperimentazione Clinica a servizio delle Aziende Sanitarie della Provincia di Padova, Padova, Italy).

Peripheral blood mononuclear cells utilized in this study derived from buffy coats obtained from healthy blood donors, as anonymously provided by the Transfusion Centre of the Hospital of Padua. Written informed consent for the use of buffy coats for research purposes was obtained from blood donors by the Transfusion Centre. Data related to human samples were all analyzed anonymously. Human leukocytes, provided by the Transfusion Centre of the Hospital of Padua, were obtained not consequently to experimentation on human beings but as a consequence of voluntary and informed blood donation for transfusions: no approval of the Ethics Committee is needed in such cases in our institution.

The following H. pylori strains were used: CagA⁺/VacA⁺ strain P12 (53), kindly provided by Prof. W. Fischer (Munich, Germany); CagA⁺/VacA⁺ strain 342, CagA⁺/VacA⁺ strain G27, and their isogenic mutants Δhp-nap and ΔcagPAI (54, 55), kindly provided by A. Covacci (Novartis Vaccines, Siena, Italy); Cag⁺/VacA⁺ strain SPM326 and its isogenic mutant ΔvacA (55, 56), kindly provided by J. L. Telford (Novartis Vaccines); strain N6 and its isogenic mutant ΔflaA (57), kindly provided by Prof C. Josenhans (Hannover, Germany).

Bacteria were maintained in 5% CO₂ at 37°C on Columbia agar plates supplemented with 5% horse blood (Oxoid) and antibiotics (trimethoprim 10 μg/ml, cefsulodin 6 μg/ml, vancomycin 5 μg/ml, and amphotericin B 8 μg/ml) (Sigma, Germany). For mutant strains, kanamycin 20 μg/ml was added to the above mixture. Colonies were taken directly from plates and resuspended in RPMI 20% FBS without any antibiotic. Bacterial count was performed by determining the optical density (OD) at 600 nm (1 OD = 10⁹ CFU/ml). Before proceeding with the infection experiments, bacteria motility was verified at the optical microscope. Handling of biohazardous material was performed in a dedicated bio-safety level 2 room and approved by the local prevention and hygiene security committee.

The cases considered in this study were retrospectively collected from the Surgical Pathology and Cytopathology Unit at the University of Padua (January 2010 to December 2018). All patients were Caucasian and underwent endoscopy/surgery at Padua University Hospital and all the samples were formalin-fixed and paraffin-embedded (FFPE).

The study was conducted on a total of 18 antral mucosa endoscopic biopsy samples obtained from 8 dyspeptic patients (with no significant histological alterations and used as control) and from 10 H. pylori-positive non-atrophic active chronic gastritis patients (biopsies were obtained before H. pylori eradication therapy). A further series of 29 samples were selected and microdissected from 10 patients surgically treated for H. pylori-positive early-stage intestinal-type GCs (all G1/G2 and all pT1 or pT2). In particular, 9 samples of extensive intestinal metaplasia (≥85%), 11 samples of dysplasia [both low-grade and high-grade lesions were considered and defined as for WHO classification (58)], and 9 samples of adenocarcinoma were considered. All 29 samples presented moderate chronic inflammatory infiltration with moderate/severe H. pylori-related activity, as assessed according to the Sydney system (59). Patients had a median age of 63.3 years (range 45–79 years) and were 9 males and 17 females.

In the H. pylori-positive samples, the bacterium was assessed by histology (modified Giemsa staining) and confirmed by clinical history, rapid urease testing, and/or ELISA (H. pylori-specific IgG Abs; GastroPanel, Biohit HealthCare) (12).

Cells were harvested from culture plates using 5 mM Na–EDTA in PBS, pH 7.5, and incubated for 15 min at RT with 10% human serum to saturate Fc receptors. 5 × 10⁵ cells were stained with a monoclonal antibody anti-HLA-II (clone L243, Ebiosciences). We used fixable cell viability dye eFluor780 (Ebiosciences) to exclude dead cells from the analysis. The latter were <5–10% of the total cell number, regardless of the treatment. Cells were resuspended in FACS buffer (PBS, 1% BSA) and analyzed by FACSCanto II (Becton Dickinson). Forward and side scatter light were used to identify cell populations. Values were expressed as the median fluorescence intensity (MFI) of live cells positive for HLA-II. All data were analyzed using FlowJo software, version 10.3 (Tree Star Inc.).

Macrophages were lysed in RIPA buffer, and proteins were quantified by BCA protein assay kit (ThermoFisher Scientific), according to the manufacturer's instructions. Equal amounts of proteins were resuspended in NuPAGE LDS sample buffer (Novex, Life Technologies) supplemented with 50 mM DTT and denaturated for 5 min at 100°C. Samples were separated electrophoretically in NuPAGE Bis-Tris 4–12% polyacrylamide gel (Novex, Life Technologies), and proteins were subsequently transferred onto PVDF membranes (Amersham). Membranes were blocked with 3% non-fat milk in Tris-buffered saline (TBS, 50 mM Tris–HCl, pH 7.6, 150 mM NaCl) containing 0.1% Tween20® (Sigma-Aldrich), and antigens were revealed using a mouse anti-human HLA-II α-chain monoclonal antibody (1:3,000, clone TAL-1B5 Abcam) and a rabbit anti-human vinculin antibody (1:5,000, Origene). Blots were washed three times with TBS plus 0.1% Tween20® and incubated for 1 h at RT with horseradish peroxidase-conjugated anti-mouse IgG (Novex, Life Technologies) or anti-rabbit IgG (Millipore) secondary antibody, respectively. Blots were developed with enhanced chemiluminescence substrate (EuroClone), and protein bands were detected using ImageQuant™ LAS 4000 (GE Healthcare Life Science).

Macrophages were incubated 1 h in methionine-free medium to deplete methionine reserves. Cells were labeled with the methionine analog L-azidohomoalanine 50 μM (AHA, Click-IT^TM Invitrogen) for 2 h and lysed in 1% SDS 50 mM Tris–HCl. All newly synthetized proteins were tagged with biotin alkyne (Click-IT™ Invitrogen) according to the manufacturer's protocol. Proteins were pulled down with NeutrAvidin agarose resin (Pierce), processed for Western blot, and developed for HLA-II.

Microarray experiments were performed using Agilent Human miRNA Microarray 8 × 60 K platform (Agilent Technologies) based on miRbase V.19. Two hundred nanograms of total RNA was labeled using miRNA Complete Labeling and Hyb Kit (Agilent Technologies), according to the manufacturer's protocol. Labeled RNA was hybridized onto microarray slides using a rotational oven at 55°C for 22 h. After hybridization, microarray slides were washed and scanned in an Agilent microarray scanner (model G2565CA). Agilent Feature Extraction software version 10.5.1.1 was used for image analysis. Microarray data expression is available in the U.S. National Center for Biotechnology Information Gene Expression Omnibus (GEO) database (GSE98641).

Raw microarray data were filtered for the number of miRNAs presenting an expression value above background (allowed 50% of undetected values for each miRNA) and then normalized according to loess cyclic algorithm (60). Differentially expressed miRNAs were identified according to the significance analysis of microarray (SAM) (61), using a median false discovery rate (FDR) of 0% (Supplementary Table 1). miRNA targets were recovered by different algorithms using default parameters: Microt4 (62), miRanda (63), mirbridge (64), miRDB (65), miRMap (66), miRNAMap (67), Pictar2 (68), PITA (69), RNA22 (70), RNAhybrid (71), Targetscan (72), and miRWalk (73). Results are listed in Supplementary Table 2.

Data for differentially expressed miRNAs targeting CIITA were extracted from microarray experiments and represented referring to the average expression of each miRNA in all samples.

Five different datasets describing miRNA expression in GC and control samples were downloaded from Gene Expression Omnibus database (GSE26595, GSE23739, GSE93415, GSE78091, and GSE54397). Processed data were used to identify differentially expressed miRNAs. GSE26595 data were quantile normalized (74), GSE23739 by median centering across probes (75), GSE93415 by Lowess (Locally Weighted Scatterplot Smoothing) algorithm, GSE78091 by median normalization, while GSE54397 were not processed for inter-array normalization (76). The limma (Linear Models for Microarray Analysis) algorithm (77) implemented in R package was used to identify differentially expressed miRNAs statistically significant. Benjamini and Hochberg correction was used to correct p-values for multiple testing and corrected p-value lower than 0.05 was considered significant to identify the list of differentially expressed miRNAs. The only dataset containing information about tumor type and grade was the GSE26595. Samples in this dataset were compared as follows: controls vs. diffuse GC grades 1 and 2, controls vs. diffuse GC grades 3 and 4, controls vs. diffuse GC all grades, controls vs. intestinal grades 1 and 2, controls vs. intestinal GC grades 3 and 4, controls vs. intestinal GC all grades. In other datasets, comparisons performed were controls vs. GC samples. Lists of differentially expressed miRNAs from each comparison were compared through Venn diagrams [Supplementary Figure 2; (78)].

To overexpress miRNAs in M121224 and HeLa-CIITA cells, pCMV-MIR vectors were used (Origene): human MIRLET7i (ID: MI0000434), human MIRLET7f (ID: MI0000067), human MIR185 (ID: MI0000482), human MIR146b (ID: MI0003129), human MIR27b (ID: MI0000440), human MIR98 (ID: MI0000100), and human 374a (ID: MI0000782). Cells transfected with a scrambled plasmid (ID: PCMVMIR) served as control. Cell transfections were performed for 48 h with the TransIT-X2® transfection system (Mirus) with 500 ng of the specific vector.

M121224 cells were transfected with 1.5 ng/μl PCMV-MIR vector expressing specific or scramble miRNA and 100 pg/μl of pmirGLO Dual-Luciferase miRNA Target Expression Vector (Promega), containing the target sequence (Supplementary Table 6). pCMV-MIR vector encoding miRNA scramble was used as control. Assays were performed using Dual-Luciferase Reporter Assay (Promega) and measuring firefly and Renilla luciferase activities in a Turner Designs TD-20/20 Luminometer (DLReadyTM, Promega). miRNA transfections were replicated independently at least three times.

Original slides (4–6 μm thick) obtained from the paraffin-embedded tissue samples were probed with 5′-biotinilated LNA matching miR-146b-5p (Exiqon). Tissue sections were digested with ISH protease 1 (Ventana Medical Systems) and ISH was performed as previously described, with minor modifications (79). Only cytoplasmic miR intensity was retained for evaluation. Five samples per condition (dyspeptic control, ACG and GC) were analyzed.

Data are reported as the mean ± SD or mean ± SEM. The Student's t-test and Mann–Whitney U-test were used for statistical analysis of the differences between experimental groups. p ≤ 0.05 were considered significant.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.