H. pylori strain SS1 (Lee et al., 1997), was initially cultured on horse blood agar plates [Blood Agar Base No. 2, 0.02% Amphostat, and Skirrow's Selective Supplements [Oxoid, Basingstoke, United Kingdom] and 5% horse blood [Biolab, Clayton, Australia]] in an anaerobic jar with a microaerophilic gas generating kit (Oxoid) for 2 days at 37°C. Prior to infection of mice, H. pylori were cultured in brain heart infusion broth (BHI; Oxoid) containing 5% horse serum (Sigma, Castle Hill, Australia), 0.02% Amphostat, and Skirrow's Selective Supplements under microaerophilic conditions for 24 h at 37°C.

Muc1^−/− [derived and kindly provided by Sandra Gendler, Mayo Clinic Spicer et al., 1995] and wild-type (WT) mice, all on a 129/SvJ background, were bred within the Veterinary Science animal house, University of Melbourne, and genotyped as described (Spicer et al., 1995). All experiments involved age-matched female mice and were performed under Animal Ethics Committee approval (University of Melbourne; AEEC No. 06205). Mice were infected intragastrically once with 10⁷ H. pylori SS1 suspended in 0.1 mL BHI and groups (n = 3 of each genotype) were euthanased after 8, 24, and 72 h for collection of gastric tissues to determine bacterial abundance and gastric gene expression. Control uninfected (sham) mice were mock-infected with 0.1 mL sterile BHI and sampled after 8, 24, and 72 h mock-infection (n = 3 of each genotype).

H. pylori infection levels within mouse gastric tissues were quantified by colony-forming assay. Briefly, stomachs were opened along the inner curvature and bisected longitudinally. One half was placed in BHI and homogenized (GmbH Polytron homogeniser, Kinematica, Switzerland) and the other half rapidly frozen for RNA extraction. Ten-fold serial dilutions were prepared in BHI broth and aliquots spread over GSSA selective agar plates [Blood Agar Base No. 2 with 5% horse blood, vancomycin (12 mg/mL), polymyxin B (0.40 mg/mL), bacitracin (24 mg/mL), nalidixic acid (1.3 mg/mL), and amphotericin B (3.75 mg/mL), all from Sigma]. After 5 days culture as above, colonies were counted, and the number of colony-forming units was calculated per stomach (McGuckin et al., 2007). Colonies were confirmed to be H. pylori by the rapid urease test as previously described (Sutton et al., 2000).

RNA was extracted from each longitudinally bisected half stomach of each mouse using Trizol (Invitrogen) and further purified on RNeasy columns (Qiagen). RNA integrity (RNA integrity number >8) was verified using a Bio-analyzer (Agilent) and RNA was stored at −70°C prior to further analysis by microarray and reverse-transcription-quantitative PCR (RT-qPCR).

Equal quantities of RNA from mice of the same genotype and time point were pooled for microarray assays. For sham-infected mice, equal quantities of RNA from mice of the same genotype and all three time points were pooled as control uninfected mice. Samples were labeled for GeneChip analysis using the One-Cycle Target Labeling and Control Reagents (Affymetrix). The gene expression array used was the Affymetrix Mouse Gene 1.0 ST array. All steps of target labeling, hybridization, and scanning were performed according to the manufacturer's protocol. The entire microarray dataset is available in Supplementary Table 1.

We used several gene ontology-based databases-to examine expression of genes across the time course of response to infection as follows;

To validate microarray data, a selection of the most highly differentially expressed genes between Muc1^−/− mice and WT mice were measured independently by RT-qPCR using the same individual RNA samples. Total RNA (1 μg) from each individual sample was used for first strand cDNA synthesis using SuperScriptTM III reverse transcriptase (Life Technologies) following the manufacturer's instructions. Real-time PCR was performed on a Rotor-Gene 3000 cycler (Qiagen) by using SYBR ® Green I fluorescence (Life Technologies). The cycling conditions were: denaturation for 10 min at 95°C, followed by 40 amplification cycles of 20 s of denaturation at 94°C, 30 s of annealing at 60°C, and 30 s of extension at 72°C. To confirm the specificity of the amplified DNA, a melting curve was determined at the end of each run. The reaction efficiency was determined with a dilution series of cDNA containing the PCR products. Expression of the target genes was normalized to that of Actb (encoding β-actin) and the results presented as their ratios (arbitrary units). Control experiments were also performed to ensure that housekeeping gene expression was not differentially regulated under the experimental conditions employed.

The primers used for PCR were designed from Primer Bank (https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi) or using Oligoperfect Designer (Life Technologies), and their sequences to amplify the target genes are shown in Supplementary Table 2.

For microarray studies, statistical analysis was performed on the R platform using the limma package from Bioconductor. The False Discovery Rate (FDR) method and Fisher's Exact Test were used with a cut off for statistical significance of P-value of <0.05 and a fold expression change of 2. For the remainder of the study, statistical analyses were performed using Prism Software v5 (Graphpad) by using ordinary one-way ANOVA and post-hoc testing. No data were excluded from any analyses. The statistical test used and the sample sizes for individual analyses are provided within the figure legends.

In wild type mice H. pylori colonization increased progressively over the first 3 days of infection (Figure 1). Consistent with our previous finding of higher H. pylori colonization in Muc1^−/− mice 24 h after infection, we found that Muc1^−/− mice displayed ~10-fold greater levels of H. pylori colonization in the stomach compared with wild-type (WT) mice as early as 8 h post-infection (P = 0.0003), consistent with a deficit in pre-existing or rapidly-induced innate defense. The higher CFU bacterial burden in the Muc1^−/− mice persisted, being 2.5- and 2.8-fold higher at 24 and 72 h post-infection (Figure 1).

To better understand the molecular mechanism by which MUC1 mediates the inhibition of gastric H. pylori colonization, we applied a microarray approach to examine the effect of Muc1 gene ablation on gastric gene expression in the absence of infection. We identified 183 transcripts that were differentially expressed (using a threshold of 2-fold change) in the gastric mucosa of uninfected Muc1^−/− mice (Figure 2A), and the top 10 differentially expressed genes are shown in Table 1. We further categorized the differentially expressed genes into specific functional groups according to gene ontology. Ingenuity Pathway Analysis (IPA) Entrez gene ontology annotation was used to determine the regulator effect network of MUC1-regulated genes (Figure 3) and the top regulators with consistence score 4.536 were Hinf1a, Hinf4a, and Sox2 (Figure 3, Table 2). In line with the known functions of these upstream regulators, we found that their target genes in our data set were all increased in expression in Muc1^−/− gastric tissue, including Fabp2, Fabp1, Alb, Abcc2, Mttp, Abcg5, and Apoc3 (Figure 3, genes with >4.5-fold changes are highlighted in bold). Increased expression of these genes is likely to lead to increased transport and metabolism of lipids (Figure 3, Table 2). Correspondingly, the top score (53) for network functions was attributed to digestive system development and function, humoral immune responses and organ development. This was followed by a score of 48 for network functions involved in lipid metabolism, molecular transport, and small molecule biochemistry (Table 3). These findings indicate that MUC1 modulates a lipid metabolic gene network, which is consistent with previous findings that MUC1 is a novel metabolic master regulator in human epithelial cancer cells (Mehla and Singh, 2014). In Muc1 deficiency, other mucin genes including Muc13, Muc2, and Muc3 were increased in expression in gastric tissues (Figure 4), although, given the other changes in gene expression and the differences in colonization, this was clearly unable to compensate functionally for loss of Muc1.

Next a transcriptomic analysis was performed at 8, 24, and 72 h post-infection. For this purpose, we analyzed each time point individually, as well as in various combinations. A comparative analysis using a threshold of 2-fold difference identified a total of 174 differentially expressed (DE) genes in WT and 326 in Muc1^−/− mice in response to infection (Figures 2B,C). The number of DE genes was higher in Muc1^−/− mice than WT mice at 8 h post infection (Figures 2B,C); the higher number of DE genes could be a consequence of the higher colonization levels of H. pylori in Muc1^−/− mice. At 24 h infection, there were fewer up-regulated genes (20) in Muc1^−/− compared to WT gastric mucosa (80), despite the higher colonization levels of H. pylori (Figures 2B,C). A similar trend was observed for up-regulated genes at 72 h infection. In contrast, Muc1^−/− mice showed a greater number of down-regulated genes (173) than WT (20) after 24 h infection, and the number of down-regulated genes surpassed the number of up-regulated genes throughout the time course in Muc1^−/− mice (Figures 2B,C). A list of the top 10 DE genes is provided in Table 1. The number of DE genes at each time point in comparison to uninfected controls is illustrated in Figures 2B,C. The number of DE genes at each time point in the gastric tissue of Muc1^−/− compared with WT mice is illustrated in Figure 2A.

In this section, rather than discussing genes individually, we have grouped differentially regulated genes into canonical pathways and associated network functions, identified the upstream factors likely to drive the differentially expressed genes in the network, and focused on the most highly statistically significant changes in the pathway and network.

In uninfected mice and at 8 h post-infection, there was significantly inhibition of the Serine Protease Inhibitor Kazal type 1 (SPINK1) pancreatic cancer pathway in the gastric tissue of Muc1^−/− mice compared with WT mice uninfected or at 8 h post-infection (P = 6.08E-14 and P = 2.25E-19, respectively, Figure 5), but at 24 and 72 h post-infection, we detected a strong activation of this pathway in Muc1^−/− mice (P = 1.38E-22 and P = 5.87E-13, respectively, Figure 5). The main physiological role of SPINK1 is to serve as a first line inhibitor of trypsin in the event of premature trypsinogen activation (Supplementary Figure 1). In addition, severe infections and tissue destruction cause elevation of SPINK1 in serum and urine, suggesting that it is an acute phase protein of the immune system. SPINK1 is highly expressed by the mucus-producing cells of the normal gastric mucosa and has been hypothesized to suppress proteolytic digestion of secreted mucus and promote gastric healing after injury (Marchbank et al., 1996, 1998; Konturek et al., 1998). In mouse, the homologous gene is designated Spink3 (Wang and Xu, 2010).

The significant change in expression of genes related to the SPINK1 pathway included genes encoding proteins involved with digestion, absorption and secretion. Examples include chymotrypsin like elastase family (Cea1, Cela3b), carboxypeptidase (members Cpa1, Cpa2, Cpb1), chymotrypsin (members Ctrb2, Ctrc, Ctrl), Kallikrein related peptidase 3 (Klk3), and serine protease (Prss2, Prss3) (Table 4 and Supplementary Figure 2). In WT mice, the focal genes in the SPINK1 pathway were up-regulated most at 24 h post-infection and appeared to be switched off and were no longer differentially expressed at 72 h infection (Table 4), whereas in Muc1^−/− mice, the focal genes were up-regulated as early as 8 h after infection and their expression levels switched to 4–8-fold decreased expression toward 24 h and continuing to be further down-regulated at 72 h post-infection (most of the genes were decreased in expression by 8–32-fold, Table 4). In addition, there was higher expression of Spink3 in Muc1^−/− mice compared with WT mice at sham and 8 h infection, but Spink3 was downregulated at 24 and 72 h post-infection (Supplementary Figure 4).

Consistent with observation of MUC1-suppressed lipid metabolism in uninfected mice, 8 h infection further induced expression of a wide range of lipid transporters in the gastric tissue of Muc1^−/− mice compared with WT (Figure 6). The IPA regulator effect analysis revealed that the categories most impacted by MUC1 deficiency were fatty acid metabolism, lipid conversion and metabolism of vitamin with consistence score = 16,028, followed by transport molecule (Table 2). In addition to increased expression of the genes leading to activated fatty acid metabolism that we observed in the uninfected mucosa (Alb, Abcc2, Mttp, Apoc3, Fabp2, Fabp1), we also observed the increased expression of genes involved in the conversion of lipid and metabolism of vitamins (Cxcl3, Apoa2, Scd, and Vdr) and cytochrome P450 members (Cyp2c8, Cyp2c9, Cyp2e1) (Figure 6). For example, stearoyl-coA desaturase (Scd) is an endoplasmic reticulum (ER) enzyme that catalyzes the biosynthesis of monounsaturated fatty acids from saturated fatty acids. Cytochrome P450 encompasses a family of enzymes which oxidize steroids, fatty acids, and xenobiotics, and are important for the clearance of various compounds, as well as for hormone synthesis and breakdown. The expression of these genes is regulated by nuclear hormone receptors of the pregnane X receptor (PXR) family (Figure 6).

Some of the up-regulated genes in the fatty acid metabolism function category were also assigned to the transport molecule category, such as Fabp1, Scd, and Cxcl3 plus additional genes Slc2a2, Vdr, Fabp2 and microsomal triglyceride transfer protein gene (Mttp) (Figure 7). IPA indicated that expression of this set of genes is likely to be controlled by PPARG at 8 h post-infection (Figure 7). In addition, higher expression of solute carrier family genes (Slc13a1, Slc26a3, Slc2a2, Slc40a1, Slc5a1, and Slc6a19) was seen in the gastric tissue of Muc1^−/− mice compared with WT at sham and 8 h infection (Table 5). This family of genes plays a pivotal role in transport of a wide variety of solutes, including glucose and amino acids. However, H. pylori infection resulted in ~2–6-fold induction of most of these genes in WT at 24 h, vs. a ~2-fold decreased expression in Muc1^−/− gastric tissues (Table 5), and a similar trend was seen at 72 h infection (Table 5).

One of the hallmarks of H. pylori colonization is the induction of a strong local antimicrobial and proinflammatory response by the infected epithelium. This initiates the mucosal infiltration of monocytes, neutrophils, and lymphocytes and the development of gastritis (McGuckin et al., 2007). Analysis of the DE genes between Muc1^−/− vs. WT in response to the infection revealed significant enrichment of functional clusters that were associated with antimicrobial and inflammatory responses.

H. pylori infection for 8 h resulted in ~4-fold induction of the regenerating islet-derived (REG) gene family 1B (Reg1b) mRNA in Muc1^−/− vs. no change in WT gastric tissues (Table 1). REG1B protein is highly expressed in several human pathologies, such as inflammatory bowel disease, many of which are associated with epithelial inflammation (van Beelen Granlund et al., 2013), indicating that the gastric mucosa of Muc1^−/− mice is more inflamed than WT mice within 8 h of infection. H. pylori infection for 24 h resulted in ~2.5-fold induction of mRNA for Slfn12, a schlafen family member, in Muc1^−/− vs. no change in WT gastric tissues. SLFN12 functions as an inducer of immune responses and is implicated in enterocyte differentiation (Table 1).

As infection progressed (24 and 72 h), we observed interferon gamma (IFNG)- regulated pathways were more strongly activated in Muc1^−/− vs. WT gastric tissues (Z-score 4.12, P = 1.65-E7, and Z-score 4.96, P = 5.7-E15, respectively, highlighted in bold in Table 6) characterized by the higher expression of many proinflammatory genes. Altered genes included: Cmpk2, Defb3, Duoxa2, Fgg, Gbp2, Gbp3, Gbp4, Hspa1a/Hspa1b, Ido1, Ifi202b, Ifi44, Ifit1b, Ifit3, Igtp, Irf7, Irgm1, Isg15, Mx1, Mx2, Nlrc5, Oas2, Oas3, Rsad2, Rtp4, Serpina1, Slfn12l, Tap1, and Usp18 (Supplementary Table 3). Some of these genes are controlled by Dnase2, Ikbkg, Il17a, Snca, Stat2, Tlr3 and their induction is associated with suppression of infection (Table 2 and Figure 8).

To further compare the wild type and Muc1^−/− response to infection, we also performed a sample-to-sample analysis using Graphia Pro network analysis software, with the filtered set of 6016 probesets showing at least 1.5-fold difference between highest and lowest across the eight samples (representing 4,448 unique genes as well as 1,305 unannotated probesets). A Pearson correlation coefficient threshold of 0.95 was used as it was the highest value to include all eight samples in the network. Supplementary Figure 3 shows that the Muc1^−/− 24 and 72 h samples were separated from all other samples on the basis of gene expression patterns, and that the earlier Muc1^−/− samples were more similar to the wild type 24 and 72 h samples that than earlier wild type samples. This suggested that there were distinct gene expression patterns differentiating the response of the Muc1^−/− animals from that of the wild type animals, consistent with the DE gene analysis. We therefore constructed a GCN using a Pearson correlation coefficient threshold of 0.87 which included all 6016 probesets of the filtered list (making 78,064 edges) in the analysis. A summary of the expression patterns of the largest clusters is shown in Supplementary Table 4 and a full list of clusters and histograms showing the average expression pattern of clusters with 10 or more nodes is available in Supplementary Table 5.

We searched the GCN for known proliferation markers, mitochondrial genes and protein synthesis genes and found no indication that these functions were disrupted in the Muc1^−/− response to the infection.

The DE genes described above were concentrated in three clusters of the GCN. The genes which were upregulated in the Muc1^−/− sham infected animals compared to the wild type sham infected and also in the wild type at 72 h post infection compared to the wild type sham infected were predominantly in Cluster004 (Supplementary Figure 4). Genes in this cluster showed enrichment of GO terms relating to digestion, transport, intestinal absorption, microvillus, brush border membrane. DE genes which were upregulated in 24 and 72 h Muc1^−/− samples compared with the corresponding wild type samples, and some of the genes that were higher in Muc1^−/− at 72 h compared with Muc1^−/− sham infected, were mainly in Cluster005 (Supplementary Figure 4), enriched for GO terms associated with innate immune response and response to pathogen. The set of DE genes which were higher in Muc1^−/− 8 h than wild type 8 h, but lower in Muc1^−/− 24 and 72 h compared to the corresponding wild type time points and also lower in Muc1^−/− 72 h compared to Muc1^−/− sham were in Cluster008 (Supplementary Figure 4). This showed enrichment for GO terms related to catabolism, digestion and extracellular space. This cluster also included Spink3, identified as a key regulator by the IPA analysis (Supplementary Figure 4), and genes encoding many members of the SPINK1 pathway, including Cela3b, Cpa1, Cpa2, Cpb1, Ctrc, Ctrl, Prss2, and Prss3 (Supplementary Figure 1). This cluster also contained a number of other proteases and protease inhibitors.

Analysis of the expression pattern of Muc1 showed that the Muc1^−/− samples had very low levels of Muc1 mRNA, as expected. Muc1 was found in Cluster020 where the average expression across all cluster genes of the Muc1^−/− samples was half that of the wild type and there was no change during the infection in either WT or Muc1^−/− (Supplementary Figure 4). This cluster contains genes potentially directly affected by the lack of MUC1, and includes the Srp54b and Srp54c genes, encoding signal recognition particles involved in the export of proteins and both reduced to about 50% of their wild type expression, Rfx6 (regulatory factor X, 6), and sodium channel genes Scn2a and Scn7a. This cluster was enriched for the GO MF term receptor regulatory activity and the GO CC terms vesicle and extracellular region.

In the GCN analysis, Cluster007 showed reduced expression overall in Muc1^−/− compared to wild type sham and also reduced further as the infection progressed (Supplementary Figure 4). This cluster contained Acta1, encoding smooth muscle actin, as well as the actin crosslinking protein gene Actn3 and two myosin genes (Myo1h and Myh2), suggesting that both infection and Muc1 knockout disrupt the cytoskeleton. The role of H. pylori in actin rearrangement has been described (Tohidpour et al., 2017) and this observation warrants further investigation.

Several genes identified as highly DE at one or more time-points over 72 h based on microarray analysis and which were also part of significantly altered pathways and networks identified by IPA were independently validated by RT-qPCR (Figure 9). Overall, most genes measured by qPCR confirmed the changes in gene expression identified by microarray (Figure 9).