Response of TLR reporter cell lines

Although we did not observe physiologic changes in lnt mutant H. pylori in vitro, the absence of lnt leads to diacylated lipoproteins instead of triacyl lipoproteins (31, 33, 35). This alteration in the number of acyl chains is expected to impact the mammalian TLR2 response (26, 27). As a first step to characterizing the host response to H. pylori lipoproteins, we affinity purified an abundant lipoprotein, Lpp20 from WT and lnt-deficient strains designed to express a human influenza hemagglutinin (HA)-tagged form of the protein (strains VM396 and VM404, Fig. 2A). Lpp20 was chosen as a model lipoprotein based on previous proteomic studies suggesting it is abundantly expressed in H. pylori (44). We next incubated the purified proteins with human embryonic kidney (HEK) reporter cell lines designed to express TLR4, TLR2 and TLR1, or TLR2 and TLR6. As controls, we also added purified Lpp20-HA to control cells including HEK null cells expressing no TLR transgene and HEK cells expressing TLR2 in the absence of TLR1 and TLR6. After 18-h incubation with Lpp20-HA, relative levels of secreted alkaline phosphatase reporter were measured (Fig. 2B). Results showed that diacylated Lpp20-HA (isolated from lnt mutant H. pylori strain VM404) stimulated TLR2-TLR6 expressing cells, with little to no stimulation of TLR2-TLR1 expressing cells. In contrast, triacylated Lpp20-HA isolated from H. pylori strain VM396 stimulated both TLR2-TLR1 expressing cells and TLR2-TLR6 expressing cells. Similar results were obtained using synthetic TLR2 ligands Pam3-CSK4 and Pam2-CSK4 (Fig. 2C). Pam3-CSK4 is a six-amino acid-long peptide (Cys-Ser-Lys-Lys-Lys-Lys) modified by the addition of three palmitic acid chains to mimic a bacterial triacylated lipoprotein (45, 46). Pam2-CSK4 is an analogous peptide containing two palmitic acid chains attached via a diacylglyceride on the Cys sulfhydryl group but does not include a palmitic acid attached to the terminal amine of the Cys and thus mimics a lipopeptide produced in the absence of Lnt (46). No stimulation of TLR4 by either form of Lpp20 was observed whereas purified H. pylori LPS was able to stimulate the TLR4 expressing cells (Fig. 2B and C; Fig. S1). Together, these results show that our purified Lpp20-HA preparations behave similarly to synthetic triacylated or diacylated TLR2 ligands with little detectable contamination by LPS.

Open in a separate window

Fig 2

Characterization of purified Lpp20 as model lipoprotein. (A) Lpp20-HA was immunopurified from WT or lnt mutant H. pylori strains VM396 and VM404. A representative Coomassie-stained gel is shown (lane a, WT Lpp20-HA from strain VM396; lane b, Δlnt Lpp20-HA from strain VM404). (B) Purified Lpp20-HA (0.1 µg/mL from either WT or lnt mutant H. pylori) was incubated with secreted embryonic alkaline phosphatase (SEAP) reporter cell lines HEK-Blue Null2 (Null), HEK-Blue hTLR4 (TLR4), HEK-Blue hTLR2 KO-TLR1/6 (KO 1/6), HEK-Blue hTLR2-TLR1 (TLR2/1), and HEK-Blue hTLR2-TLR6 (TLR2/6) from InvivoGen. Relative amounts of SEAP were measured after 16-h treatment. Results show the mean and standard deviation from single Lpp20-HA preparations (from both WT and lnt mutant H. pylori) tested in triplicate and are representative of three independent preparations of each lipoprotein. (C) Synthetic lipopeptides (0.04 µg/mL triacylated Pam3-CSK4 and diacylated Pam2-CSK4) were used to stimulate reporter cell lines as described in B. Results show the mean and standard deviation of triplicate determinations and are representative of three independent experiments. Results in panels B and C were analyzed by analysis of variance followed by Tukey’s post hoc test. Select pair-wise comparisons are indicated as significantly different (*, P < 0.001) or not significantly different (ns).

Response of primary human gastric epithelial cells

To develop a more complete understanding of the host response to H. pylori lipoproteins, we utilized minimally passaged primary human gastric epithelial monolayers. Cells were left untreated or treated with purified triacylated Lpp20-HA, diacylated Lpp20-HA, Pam3-CSK4, or Pam2-CSK4 for 4 h. Following treatment, RNA was isolated from the cells and subjected to RNAseq. Analysis of the untreated control cells showed transcription of KRT18 and TFF1, cell surface mucins MUC1 and MUC16, and secreted mucin MUC5AC as expected for stomach epithelial cells (47, 48). Additionally, transcripts for TLR1, TLR2, TLR3, TLR4, TLR5, TLR6, and TLR9 were detected.

Normalized transcript abundance was determined for each sample and differentially expressed genes (greater than a two-fold change in expression vs control and an adjusted P-value less than 0.05) were identified. We first compared genes differentially expressed between cells treated with triacylated Lpp20-HA vs untreated control cells and between cells treated with triacylated Pam3-CSK4 peptide vs untreated control cells (Fig. 3A, C, and E; Data sets S1 and S2). Results showed a similar (though not identical) pattern of differentially expressed genes; Pam3-CSK4 led to increased expression of a larger number of genes. Analyses identified 40 genes differentially expressed between Lpp20-HA-treated cells and control, and 66 genes differentially expressed between Pam3-CSK4-treated cells and control. Among these genes, 30 were found to be differentially expressed by both Lpp20-HA and Pam3-CSK4, with a strong correlation between the magnitude of fold change between the two groups (Fig. 3E). The similarity between these results suggests that much of the cellular response detected in the RNAseq data is likely due to the lipid anchor of Lpp20 rather than a response to the functional fold of the protein.

Open in a separate window

Fig 3

Analysis of RNAseq results. (A–D) RNAseq results were analyzed using edgeR as described in Materials and Methods. Results are summarized by plotting the log₂(fold change) compared to untreated control cells vs the negative log of the adjusted P-value (Benjamini-Hochberg) (A, Pam3-CSK4 vs untreated control; B, Pam2-CSK4 vs untreated control; C, WT Lpp20-HA vs untreated control; and D, Δlnt Lpp20-HA vs untreated control). Significant results (greater than a twofold change in expression and an adjusted P-value less than 0.05) are shown in blue. (E and F) To compare results between similar treatment groups, the log₂(fold change) for significant differences were plotted against each other (E, cells treated with triacyl lipoproteins Pam3-CSK4 and WT Lpp20-HA; and F, cells treated with diacyl lipoproteins Pam2-CSK4 and Δlnt Lpp20-HA).

Similar analyses were performed comparing genes differentially expressed between cells treated with diacylated Lpp20-HA vs untreated control cells and between cells treated with diacylated Pam2-CSK4 peptide vs untreated control cells (Fig. 3B, D, and F; Data sets S3 and S4). Analyses identified 271 genes differentially expressed between diacylated Lpp20-HA-treated cells and control, and 825 differentially expressed genes between Pam2-CSK4-treated cells and control. Among these genes, 203 were found to be differentially expressed by both diacylated Lpp20-HA and Pam2-CSK4, with a strong correlation between the magnitude of fold change between the two groups (Fig. 3F).

We next performed pathway analysis to provide potential mechanistic insight into the differentially expressed genes. Given the strong correlations between the fold change in gene expression for cells treated with triacylated lipoproteins (Pam3-CSK4 and WT Lpp20-HA, Fig. 3E), pathway analysis was performed by combining the differentially expressed gene lists for Pam3-CSK4- and WT Lpp20-HA-treated cells into a single gene set representing the response to triacylated lipoproteins (49). Similarly, pathway analysis was performed by combining the differentially expressed gene lists for Pam2-CSK4- and Δlnt Lpp20-HA-treated cells into a single gene set representing the response to diacylated lipoproteins. Results are summarized in Tables 1 and 2. The response to triacylated lipoproteins was characterized by increased expression of a number of cytokines and chemokines (CCL2, CCL20, CSF1, CSF2, CSF3, CX3CL1, CXCL1, CXCL2, CXCL3, CXCL6, IL1A, IL1B, IL6, and IL8). The response to diacylated lipoproteins was characterized by the increased expression of a wider array of cytokines and chemokines and an assortment of genes linked to chemokine and cytokine signaling and TLR2 signaling. To help illustrate this, genes belonging to the Kyoto Encyclopedia of Genes and Genomes (KEGG) cytokine-cytokine receptor interaction pathway (hsa04060) that were differentially regulated in one or more treatment groups were identified and the log₂(fold change) presented as a heatmap (50) (Fig. 4A). Select genes were analyzed by qRT-PCR to confirm that these are differentially expressed (Fig. 4B). Overall, the response to triacylated lipoproteins was more modest than the response to diacylated lipoproteins. This result was consistent with results observed with the TLR2/1 and TLR2/6 reporter cell lines and with results of previous studies indicating increased inflammatory responses to lipoproteins activating TLR2/6 heterodimers (26).

Open in a separate window

Fig 4

Induction of cytokine and chemokine expression by diacylated lipoproteins. (A) Pathway analysis (Tables 1 and 2) indicated that lipoprotein treatment led to increased expression of numerous chemokines and cytokines. The log₂(fold change) for genes within the KEGG cytokine-cytokine receptor interaction pathway (hsa04060) that were differentially regulated in primary gastric epithelial cells by one or more treatment groups (Pam3-CSK4, Pam2-CKS4, WT Lpp20-HA, or Δlnt Lpp20-HA) is presented as a heatmap. (B) Quantitative RT-PCR of select transcripts was performed on the RNA samples from primary human gastric epithelial cells treated with WT Lpp20-HA (open symbols) or Δlnt Lpp20-HA (closed symbols) as described in Materials and Methods; RPLP0, PGK1, and ACTB served as controls. Results represent the mean and 95% credible limits (error bars) from six replicates each. Significance was determined by calculating Bayesian z scores, and a standard z test was performed to derive two-tailed P-values, which were corrected for multiple testing using the Benjamini-Hochberg method (with a false discovery rate of 5%). *, P < 0.03.

Construction of recombinant H. pylori strains

Genomic DNA from H. pylori strain VM211 (in which the lnt promoter and open reading frame is replaced by aphA3 and a copy of the lnt promoter is inserted downstream of aphA3 to direct expression of the downstream genes within the operon) was used to transform strain J166 and selection for kanamycin-resistant (Kan^r) colonies was used to isolate strain VM391 (Table 3). Strain VM391 was subsequently transformed with plasmid pMM714 which contains H. pylori lnt under the control of the ureA promoter and a cat gene inserted into the hydA-mdaB intergenic region (Table 3) (31, 41,–43). Plasmid pMM714 does not replicate in H. pylori. Selection for chloramphenicol-resistant (Chl^r) colonies was used to isolate strain VM392. H. pylori strain BV199 was transformed with plasmid pMM711 which contains an HA-tagged copy of lpp20 and the cat gene in place of ureA (Table 3). Plasmid pMM711 does not replicate in H. pylori and selection for chloramphenicol-resistant colonies was used to isolate VM396. VM396 was subsequently transformed with genomic DNA from strain VM211 and selection for kanamycin-resistant colonies was used to isolate strain VM404 (Table 3).

Purification of Lpp20

H. pylori strains VM394 and VM404 were grown in bisulfite-free Brucella broth supplemented with cholesterol. Bacteria were pelleted (5,000 × g, 10 min), and the pellet was resuspended in 1 M MOPS, pH 7 containing protease inhibitors (Complete Protease Inhibitor) and 1,000 units mutanolysin (Sigma) per gram of bacterial pellet. The bacterial suspensions were incubated with rocking at room temperature for 1 h then passed through a French Press (Glen Mills) at 18000 pounds per square inch. Unlysed bacteria were pelleted (7,500 × g, 10 min) and bacterial membranes pelleted (174,000 × g, 2 h, 4°C). The total membrane pellet was solubilized in 50 mM MOPS, pH 7, 300 mM NaCl, and 1% fos-choline 12. Solubilized membranes were incubated with anti-DYKDDDDK-coated magnetic agarose (Pierce), and the samples were mixed gently at room temperature for 1 h. The beads were collected on a magnet and washed four times in 50 mM MOPS, pH 7, 300 mM NaCl, and 1% fos-choline 12. Affinity-purified proteins were eluted by incubating the beads twice with 50 µL of 2.5 mg/mL DDK peptide in 50 mM MOPS, pH 7, 300 mM NaCl, and 1% fos-choline 12. Eluted proteins were buffer exchanged into 50 mM MOPS, pH 7, 300 mM NaCl using Amicon Ultra ultrafiltration devices (3,000 mw cutoff, Millipore), precipitated with acetone, and dried.

TLR reporter cells and secreted embryonic alkaline phosphatase (SEAP) assays

Reporter cells, derived from HEK293 cells, were purchased from InvivoGen. HEK-Blue Null2 cells express a SEAP reporter gene under the control of the IL-12-p40 minimal promoter fused to five NF-κB and AP-1 binding sites. Similar to HEK-Blue Null2, the HEK-Blue hTLR4 cell line additionally expresses the hTLR4 gene, and the MD-2/CD14 co-receptor gene. The HEK-Blue hTLR2 KO-TLR1/6 cell line carries a double knockout of the TLR1 and TLR6 genes in a SEAP expressing cell line that overexpresses hTLR2 and CD14. HEK-Blue hTLR2-TLR1 were derived by stable transfection of the hTLR1 gene into the HEK-Blue hTLR2 KO-1/6 cell line. HEK-Blue hTLR2-TLR6 were derived by stable transfection of the hTLR6 gene into the HEK-Blue hTLR2 KO-1/6 cell line. Cells were left untreated, treated with Pam3-CSK4 or Pam2-CSK4 (40 ng/mL), or treated with WT Lpp20-HA or Δlnt Lpp20-HA (100 ng/mL) for 16 h. Secreted alkaline phosphatase assays were carried out using HEK-Blue detection media according to manufacturer’s instructions (InvivoGen). Pam3-CSK4 and Pam2-CSK4 were purchased from Tocris Bioscience.

Primary human gastric epithelial cells

Deidentified primary human gastric epithelial cells were purchased from Cell Biologics. Cells at passage 4 were subcultured at 3.4 × 10⁵ cells per well into six-well dishes pretreated with gelatin and cultured in Complete Human Epithelial Cell Medium (Cell Biologics) for 16 h. Cells were left untreated (three wells), treated with Pam3-CSK4 or Pam2-CSK4 (100 ng/mL, three wells each), or treated with WT Lpp20-HA or Δlnt Lpp20-HA (300 ng/mL, six wells each) for 4 h.

RNAseq

For analysis of gene expression in human gastric epithelial cells, cells were lysed in TRIzol and RNA was purified using Direct-zol RNA MiniPrep Plus (Zymo Research) including on-column DNAse treatment. Each RNA sample was eluted in 100 µL of water. Purified RNA (300 ng total RNA per sample) was submitted to the VANTAGE Core who performed RNA quality assessment (2100 Bioanalyzer, Agilent), ribosomal RNA depletion, library preparation, and sequencing on an Illumina NovaSeq 6000 at 150 bp paired-end reads. The number of sequence reads averaged 58 million per sample. Trimmomatic was used for quality trimming and adapter clipping (82). HiSat2 was used to map reads to human b37:hg19 canonical reference genome (83). Alignments were assembled into transcripts using StringTie (84). Differential gene expression was carried out using edgeR using the generalized linear model (85). Fold change values were calculated by comparing the means of normalized pseudocounts for samples from each group (calculated by edgeR). A false discovery rate (FDR), corresponding to a Benjamini-Hochberg adjusted P-value, was calculated using the p.adjust program within edgeR. Differentially expressed genes were defined as those exhibiting an FDR value of <0.05, as well as a >2.0- or <0.5-fold difference in transcript abundance in the above comparisons. Pathway analysis was performed using PANTHER (49).

Mouse infections and tissue harvests

To infect mice, the H. pylori strains (J166, lnt mutant strain VM391 and complemented mutant strain VM392) were grown in liquid culture (bisulfite-free Brucella broth with 10% heat-inactivated FBS and 10 µg/mL vancomycin) under microaerophilic conditions generated by a GasPak EZ Campy Container System (Becton Dickinson, Franklin Lakes, NJ), shaking at 160 rpm for 16–18 h. The orogastric inoculum was prepared after quantification of H. pylori using an optical density at 600 nm (OD₆₀₀) reading. The cultures for two replicate experiments were serially diluted and plated to verify the accuracy of the inocula (for each strain). For all experiments, mice received two doses of 1 × 10⁹ CFUs of H. pylori in 0.5 mL bisulfite-free, serum-free Brucella broth approximately 48 h apart.

At the time of tissue harvest (2 weeks post infection), the mice were sacrificed, and the stomach was dissected. Gastric tissue containing both antrum and corpus regions were used to determine bacterial burden and inflammation. The H. pylori burden in the mice was quantified by spreading stomach homogenate on tryptic soy agar plates containing sheep blood (5%), nalidixic acid (10 µg/mL), vancomycin (50 µg/mL), amphotericin (2 µg/mL), and bacitracin (100 µg/mL) in microaerophilic conditions at 37°C for 5–7 d.

Inflammation scoring

Inflammation was scored using the updated Sydney System by a single pathologist (M. B. P.) who was unaware of the different strains used for infections. Acute inflammation was scored (0–3) based on increasing density of neutrophils. Chronic inflammation was scored (0–3) based on the density of mononuclear cell infiltration in the lamina propria. Inflammation was scored as absent (grade 0), mild (grade 1), moderate (grade 2), or severe (grade 3) with the antrum and corpus regions of the stomach being scored independently. The total inflammation was calculated as a sum of acute and chronic inflammation in the corpus and the antrum, which allows for quantification of total inflammation on a scale from 0 to 12.

This work was supported by the National Institutes of Health (AI163586, MSM and HMSA); the Department of Veterans Affairs (IBX000915A HMSA), the Vanderbilt Digestive Disease Research Center (P30DK058404), and the Vanderbilt-Ingram Cancer Center (P30CA068485).