Complex T Cell Interactions Contribute to Helicobacter pylori Gastritis in Mice

Mice.

All mice used in this study were 6- to 8-week-old female mice on a C57BL/6 background. C57BL/6J (stock number 000664) and IFN-γ KO (B6.129S7-Ifng^tm1Ts/J, stock number 002287) mice were purchased from Jackson laboratories. RAG2KO/OTII transgenic mice [B6.129S6-Rag2^tm1Fwa Tg(TcraTcrb)425Cbn, stock number 1896] and RAG2KO/γc KO mice (RAG and gamma chain [γc] DKO; see below) were from Taconic. IL-17A KO mice (17AKO) were a generous gift from Yoichiro Iwakura of the Center for Experimental Medicine in Tokyo, Japan (36). Double KO mice were produced by breeding IFN-γ KO and IL-17A KO and selecting for DKO offspring based on PCR identification of the expected mutant genes. Primers used are indicated in Table 1.

Mice were housed in helicobacter-free rooms in microisolator cages in the University of Michigan laboratory animal facility, and no known mouse pathogens were present in the mouse colonies, as determined by routine periodic screening of sentinel mice. Mice were offered nonsupplemented commercial mouse chow and water ad libitum. Male and female mice were inoculated with H. pylori at 6 to 8 weeks of age. All animal experiments were approved by the University of Michigan Committee on the Use and Care of Animals.

Bacteria and mouse inoculation.

Mouse inoculation and adoptive transfer were performed as previously described (23, 37). Briefly, H. pylori strain SS1 was cultured overnight in Brucella broth with 10% fetal calf serum. When the cultures reached mid-logarithmic phase, they were harvested and adjusted to approximately 10⁷ CFU/ml and mice were given 0.1 ml by oral gavage.

For adoptive transfer, uninfected donor mice (C57BL/6 or mutants) were euthanized and the spleens were removed. CD4⁺ cells were isolated by magnetic bead separation as previously described (38). Briefly, splenocytes were dissociated, suspended in phosphate-buffered saline (PBS) with 0.5% bovine serum albumin (BSA), and incubated with anti-CD4 L3T4-coated Miltenyi MACS beads (Miltentyi Biotec, Auburn, CA). The cell-bead mixture was layered onto Miltenyi MS magnetic columns and washed, the columns were removed from the magnet, and the CD4-enriched fraction was collected and resuspended to a concentration of 1 × 10⁷ cells/ml. Where indicated, a second magnetic bead separation was used to enrich for CD25⁺ cells and the total cell number was adjusted according to the individual experiment. Two weeks after bacterial inoculation, uninfected or H. pylori-infected RAG KO mice received 100 μl of T cell suspension by intraperitoneal injection. Antibody-mediated suppression of IFN-γ was achieved with clone XMG1.2, kindly provided by Gary Huffnagle. Mice were given 1.0 mg of XMG1.2 by intraperitoneal injection weekly for the duration of the experiment. Control mice were untreated.

Histologic scoring.

Scoring of gastritis based on extent of histologic lesions was performed as previously described (23, 39). Briefly, two longitudinal sections of the gastric greater curvature were examined in their entirety. Sections were scored based on the number of ×200 fields that contained polymorphonuclear lymphocytes (PMNs) (neutrophils), mononuclear gastritis, or epithelial metaplasia, expressed as a percentage of the total number of fields present. The gastritis score reported was the sum of the percentages of all three lesions.

Flow cytometry.

Antibodies and reagents for flow cytometry were purchased from AbCam, BD Biosciences, or eBiosciences. Isolation of splenic and gastric lamina propria leukocytes was performed as previously described (38, 40). Briefly, spleens were manually disaggregated, washed, and resuspended in phosphate-buffered saline (PBS) with 0.5% BSA. For detection of surface markers, cells were incubated with anti-mouse FcRγ, followed by fluorochrome-labeled antibody according to the individual experiment (Table 2). For detection of regulatory T cells (Treg), splenocytes were fixed and permeabilized using the mouse regulatory T cell staining kit (eBiosciences) according the manufacturer's instructions. Leukocyte gates were determined based on size and granularity (forward and side scatter), and the fraction of cells staining for each marker was expressed as a percentage of the total leukocyte gate.

To isolate gastric lamina propria leukocytes, one-half of the gastric wall was washed, minced, and subjected to 3 rounds of EDTA and Liberase digestion as previously described (40). Recovered cells were washed, stained, and examined as described above.

Immunochemistry.

Identity of inflammatory cells in formalin-fixed, paraffin-embedded sections was determined by morphology in hematoxylin and eosin (H&E)-stained sections and by immunoreactivity to CD3 (T cells), F4/80 (macrophages), or FoxP3 (Treg). Immunostaining was performed on an automated immunostainer (Biocare Intellipath). Antigen retrieval for macrophage detection was performed by trypsin digestion for 10 min (Biocare Medical, Concord, CA) at room temperature. Retrieval for T cells was performed by heat-induced retrieval at 95°C in a decloaking chamber (Biocare) using a commercial solution (Biocare citrate buffer). T cells were labeled with rabbit monoclonal anti-CD3 (RM9107-S; Thermo-Scientific, Waltham, MA) at 1:500 with 1 h incubation or with monoclonal anti-FoxP3 (ab54501; Abcam Cambridge, MA) at 1:1,000. Macrophages were labeled with rat monoclonal anti-F4/80 (ab6640; Abcam, Cambridge, MA) at 1:100 with 1 h incubation. Immunodetection was by a horseradish peroxidase-conjugated polymer system with diaminobenzidine for rat and rabbit primary antibodies (Biocare PromARK series, rat-on-rodent and rabbit-on-rodent) and an alkaline phosphatase-conjugated polymer system (Biocare goat polymer detection) with Biocare Vulcan fast red for goat antibodies. Negative controls were naïve sera from the species of the primary antibody. Sections were counterstained with hematoxylin.

Sections were examined by light microscopy, and cells that expressed the target antigen were identified by chromophore uptake in the appropriate intracellular location (membrane, cytoplasm, or nucleus, depending on the antigen). The number of stained cells in each ×200 microscopic field was estimated on a semiquantitative scale as no (0), few (1), moderate (2), or many (3) cells per field. The total score was recorded as the percentage of fields with a score of 3.

Quantitative reverse transcriptase PCR (qRT-PCR).

For RNA isolation, samples were stored for no more than 3 months at −80°C in 300 μl of TRIzol (Invitrogen, Carlsbad, CA) and total RNA was purified according to the TRIzol protocol. mRNA was then further enriched with Qiagen RNeasy columns and stored at −80°C until analysis. Using SA Biotech's cDNA first-strand synthesis kit and protocol (SA Biosciences, Frederick, MD), mRNA samples were prepared for qRT-PCR of the mouse cytokine genes IL-17A, IL-17F, IFN-γ, TNF-α, and hypoxanthine phosphoribosyltransferase (HPRT), which served as a housekeeping control. For qRT-PCR analyses, the SYBR green cycle threshold was established automatically using the Bio-Rad CFX Manager, based on ROX fluorescence. Any well with a SYBR green cycle threshold (CT) greater than 35 cycles was counted as a negative result. All samples were run in triplicate, and expression of cytokine genes was normalized to HPRT expression using the 2^−ΔΔCT method (41). To calculate fold change, the normalized value for each infected mouse was divided by the mean of the uninfected mice in the group. If cytokine mRNA was undetectable in uninfected mice, the lowest detectable value for that cytokine was used as a denominator.

Statistics.

All graphs were made using GraphPad Prism 5.01 software. Each symbol represents a single mouse, and bars indicate the median value for the group. Groups were compared using the Mann-Whitney test (2 groups) or the Kruskal-Wallis test with Dunn's posttest (for a comparison of multiple groups).

Gastritis is exacerbated in the absence of Treg.

Several recent studies have suggested that CD4 CD25 FoxP3⁺ lymphocytes (Treg) are the principle modulators of gastritis due to H. pylori (17, 19–22). In fact, some have suggested that the principle response of normal hosts to infection is regulatory rather than proinflammatory and is mediated by Treg (17, 21, 22). We hypothesized that the exacerbated gastritis described in CD4 cell recipient mice compared to C57BL/6 mice (described above and in references 11, 23, and 24) may be attributable to the failure of Treg to engraft in the recipients. To determine if the absence of Treg exacerbates IFN-γ- or IL-17-dependent gastritis in our model, we first established the threshold at which Treg suppress gastritis. For this, we used an adoptive transfer model and varied the dose of unfractionated CD4 cells and of Treg that were transferred. To enrich Treg in our transferred population, we isolated CD4⁺ CD25⁺ cells by magnetic bead separation. Treg are defined by CD4, CD25, and FoxP3 expression, but since FoxP3 is a nuclear antigen, live cells can be isolated only by the surface markers CD4 and CD25. The CD4 CD25⁺ population contains largely Treg, although not all of the cells are regulatory (for a review, see reference 42). Unfractionated splenocytes were first sorted to isolate CD4⁺ cells (approximately 15% of the splenocyte population of C57BL/6 mice), resulting in a population of 93.5% ± 8.6% CD4⁺ cells. This population was further sorted to enrich CD25⁺ cells, resulting in a population that was 97.5% ± 0.04% CD25⁺ cells. This population is referred to as Treg enriched. Table 3 indicates the total number of CD4 cells and the number of CD4 CD25⁺ cells (an approximation of Treg numbers) given to each mouse in the 4 groups.

In the low- and high-dose CD4 recipient groups (Table 3, groups 1 and 2), Treg were not enriched and comprised 1.5 to 2.0% of the transferred cells, as determined by flow cytometric analysis. However, because the high-dose group received 10 times as many total cells, it also received 10 times as many Treg as the low-dose group. The groups that received Treg-enriched populations (groups 3 and 4) received approximately the same number of total cells as the low-dose CD4 groups but 5 to 10 times the number of Treg, similar to the high-dose CD4 recipient group. Thus, the effects of total cell numbers and approximate Treg numbers may be examined separately.

The number of CD25⁺ cells transferred markedly altered gastritis as well as gastric engraftment of T cells in infected mice. Gastritis was quantified by extent of inflammation and epithelial metaplasia as described previously (39). This approach is reproducible and accurately reflects disease severity. Figure 1 shows that, as expected, Treg transfer suppressed gastritis. In all infected groups given 2.0 × 10⁵ CD4 CD25⁺ cells or more, gastritis was significantly less extensive than in mice given 2.0 × 10⁴ CD4 CD25⁺ cells or fewer. Neither the total number of CD4⁺ cells transferred nor the ratio of CD4 to CD4 CD25⁺ cells affected the outcome. Mice given 10⁵ CD4 CD25⁺ cells had similarly decreased gastritis scores whether the total CD4⁺ cell number was 1.0 × 10⁷ or 5 × 10⁵ cells, indicating that the threshold for suppression of gastritis is between 10⁴ and 10⁵ CD4 CD25⁺ cells/mouse. Figure 2 illustrates the histologic appearance of gastritis in the different groups. In mice that received fewer than 10⁵ CD4 CD25⁺ cells, gastritis was severe and extensive, characterized by replacement of the fundic mucosa by metaplastic mucus-type epithelium with widespread infiltration of lymphocytes, macrophages, and neutrophils (Fig. 2A). In mice that received 2 × 10⁵ CD4 CD25⁺ cells, gastritis was present but less extensive, consisting of multifocal infiltration of inflammatory cells in the fundic mucosa (Fig. 2B and C). Based on these results, all further adoptive transfer experiments were done in mice given fewer than 1 × 10⁶ CD4⁺ cells with no CD25 enrichment (i.e., fewer than 10⁵ Treg) in order to examine the role of cytokine-deficient effector T cells in the absence of Treg-mediated suppression.

Consistent with suppression of gastritis, total gastric CD4 cell engraftment in the stomach of infected mice was significantly lower and total Treg engraftment was higher in recipients of larger numbers of Treg (Fig. 3). In low-dose recipients, the group with extensive gastritis, total T cell engraftment was the greatest (Fig. 3A and B) and CD4⁺ FoxP3⁺ Treg were essentially undetectable (Fig. 3C and D). In the other three groups, with less-extensive gastritis, CD4⁺ cells overall engrafted poorly but Treg were a greater percentage of engrafted cells. In contrast to the gastric engraftment, splenic engraftment of both total CD4⁺ T cells and Treg did not differ in the different groups (Fig. 3E and F). Thus, gastric, but not splenic, Treg engraftment required a threshold Treg transfer dose for engraftment, as did suppression of gastritis. Interestingly, infection by H. pylori was necessary for both gastric and splenic engraftment of CD4⁺ cells. Uninfected mice were given low-dose CD4⁺ cells (1 × 10⁶ total CD4 cells, including about 10⁴ Treg), but few CD4⁺ cells were detected in either stomach or spleen (Fig. 3A to F). B220⁺ and CD8⁺ cells were not detected in stomach or spleen of CD4⁺ cell recipient mice (data not shown). Myeloid cells (GR1⁺) ranged from 2 to 12% of gastric leukocytes, and dendritic cells (CD11c⁺) ranged from undetectable to 8%. There were no differences between experimental groups.

Immunohistochemical detection of CD3⁺ T cells, FoxP3⁺ cells (Treg), and F4/80⁺ cells (macrophages) correlated with the flow cytometric detection of gastric cell subsets. In inflamed stomachs, both CD3⁺ (T cells) and F4/80⁺ (macrophages) cells were present in large numbers in the gastric mucosa (Fig. 4). T cells were present throughout the inflamed mucosa and submucosa, while macrophages were distributed primarily in the submucosa. B cells were not detected. In recipients of >10⁵ Treg-enriched populations, CD3⁺ cells were rare, while the F4/80 cell number appeared unchanged, consistent with flow cytometry results. Semiquantitative scoring confirmed this impression. In recipient mice that received low-dose CD4 cells, 100% of histologic fields contained marked CD3 infiltration, while in the other groups, marked infiltration was present in 30 to 50% of fields. FoxP3⁺ cells were detectable in the gastric mucosa but were few in number and were not able to be quantified histologically (Fig. 4D).

Adoptive transfer studies confirm the role of IFN-γ but not Th1 cells.

The data presented above demonstrate that absence of Treg engraftment accounts for the increased severity of gastritis in recipient mice compared to C57BL/6 mice. We hypothesized, therefore, that the Th-1-independent gastritis present in recipients of IFN-γ-deficient CD4 cells may be due either to the effect of IFN-γ production by non-T cells or to other proinflammatory cytokines in the absence of Th-1 cells. Our previous data showed that gastritis develops in recipient mice in the absence of Th1 (IFN-γ-producing) T cells (11, 27). Recipients of either IFN-γ KO or T-bet KO T cells developed gastritis, albeit of a lesser severity than that of recipients of C57BL/6 T cells, indicating that Th1-biased T cells are not essential for induction of gastritis, although they contribute to the severity. Interestingly, IFN-γ remained detectable in the gastric tissue of recipients of Th1-deficient CD4 cells, suggesting that a non-T cell source (for example, innate immune cells, such as NK or dendritic cells) may contribute to disease.

To determine if IFN-γ (regardless of cell source) is essential to induce gastritis and to determine if innate immune cells may be the source, we performed two experiments. First, to determine if IFN-γ is absolutely required for gastritis, we treated recipient mice with IFN-γ-specific antibody to eliminate all IFN-γ and quantified gastritis. Figure 5A shows that weekly inoculation of infected recipient mice with 1 μg of antibody directed against IFN-γ completely abolished gastritis. Enzyme-linked immunosorbent assay (ELISA) examination confirmed that circulating IFN-γ was present in untreated infected mice but not detectable in treated mice (data not shown). This indicated that IFN-γ is essential for gastritis. Since we have shown that IFN-γ from T cells is not essential, this indicates that a non-T cell source of IFN-γ contributes to gastritis in the absence of Th1 cells.

To confirm that IFN-γ derived from non-T cells contributes to gastritis, we performed adoptive transfer studies in RAG KO gamma chain (γc) KO recipient mice. These mice are NK cell deficient as well as T and B cell deficient and have other innate immune and cytokine defects referable to the absence of the common cytokine gamma chain (43). Innate immune cells from these mice do not produce IFN-γ. Figure 5B shows that H. pylori-infected RAG gamma chain DKO (RAGγc DKO) recipients of CD4⁺ T cells from C57BL/6 mice did not develop gastritis. Since in these mice, the adoptively transferred T cells did produce IFN-γ but the innate immune cells in the recipient mice did not, we conclude that cytokine production by innate immune cells is essential for gastritis and that IFN-γ from T cells alone is not sufficient. In combination with the finding, noted previously (11, 27), that IFN-γ is present in inflamed gastric mucosas in recipients of Th1-deficient T cells, this suggests that innate immune cells may be the source of IFN-γ in the absence of Th1 T cells.

H. pylori-specific CD4 cells are essential for gastritis.

Previous data and the data presented above indicate that while both CD4⁺ T cells and IFN-γ are essential for induction of gastritis in infected mice, CD4⁺ T cells are not necessarily the source of the IFN-γ. We also showed that innate immune cells expressing the common gamma chain are essential for gastritis in infected recipients of CD4⁺ T cells. Thus, disease requires CD4⁺ cells, functional innate immune cells, and IFN-γ. One possible explanation for gastritis in the absence of Th1 T cells is that gastritis due to H. pylori in mice is a nonspecific inflammatory response initiated by IFN-γ-producing innate immune cells in the presence of H. pylori and CD4 T cells but independent of H. pylori antigenicity. To rule out this possibility, we demonstrated that gastritis is dependent on the presence of H. pylori-specific CD4⁺ cells.

For this, RAG KO/OTII transgenic mice were inoculated with H. pylori and transferred with CD4⁺ T cells from OTII transgenic mice. These transgenic mice express T cell receptors (TCR) specific for ovalbumin only, and although CD4⁺ T cells are present, they do not respond to antigens other than ovalbumin (44). As shown in Fig. 6, T cell receptor transgenic mice failed to develop gastritis in response to H. pylori, indicating that H. pylori-specific T cells are necessary for disease.

Both IFN-γ and IL-17 contribute to gastritis.

One goal of this study was to investigate proinflammatory cytokines that may contribute to gastritis in the absence of IFN-γ-producing T cells. Our preliminary data (not shown) suggested that IL-17 was upregulated in the gastric mucosa of H. pylori-infected recipients of IFN-γ-KO CD4 T cells with gastritis, and some investigators have suggested that IL-17 is an inflammatory cytokine that promotes gastritis (32, 34). In order to determine if IL-17A contributes to gastritis, we used our simple infection model to compare gastritis severity and gastric cytokine production in C57BL/6, IFN-γ KO, IL-17A KO, and DKO mice. Because Treg are present in immunocompetent C57BL/6 mice, simple infection is slowly progressive (see above). Therefore, we examined mice at 5 to 9 months after H. pylori inoculation. We have previously shown that gastritis is moderately severe by this time and that differences in severity of gastritis in T-bet- or IFN-γ-deficient mice compared to that in C57BL/6 mice are quantifiable and statistically significant (11, 27).

Figure 7 shows that, as previously reported (see above), gastritis in C57BL/6 mice was moderately severe (median score, 123 out of 300; range, 76 to 236). Our previous studies suggested that a score of 50 or greater is evidence of biologically significant inflammation (39). In contrast to C57BL/6 mice, IFN-γ KO mice developed minimal gastritis (median, 22; range, 0 to 79), with only 4 out of 20 mice having scores of >50. This was consistent with previous studies as well as our IFN-γ depletion experiment described above. IL-17 KO mice developed mild gastritis (median, 55; range, 0 to 150), which was significantly less than in C57BL/6 mice (P = 0.0003) and marginally greater than in IFN-γ KO recipients (P = 0.0488). None of the DKO mice developed gastritis scores >50. Mean gastritis scores in infected C57BL/6 and IL-17 KO mice, but not IFN-γ KO or DKO mice, were significantly different than those in uninfected isogenic controls (P < 0.05). We interpret these results to indicate that IFN-γ contributes strongly to gastritis and that IL-17A also contributes to gastritis, but to a lesser extent.

Gastric cytokine expression reflected the extent of gastritis in the different mouse groups (Fig. 8). In C57BL/6 mice, the median expression of all four cytokines measured was greater than 2, suggesting a biologically significant increase, and most individual mice had elevated expression of all four cytokines. Many IL-17A KO mice also had elevated gastric cytokine expression (with the exception of IL-17A, which is not expressed by the KO mice). Interestingly, with the exception of a single C57BL/6 mouse, IL-17F expression was greater in IL-17A KO mice than in any other group. IL-17F is less well studied than IL-17A, but both are produced by Th17 cells and they likely have overlapping functions (45). Elevation of expression of IL-17F may represent an adaptive response to the absence of IL-17A in H. pylori-infected IL-17A KO mice. IFN-γ KO and DKO mice did not develop gastritis, and the mean gastric cytokine expression in these mice was not significantly greater than that in uninfected mice, although a few IFN-γ KO mice expressed mildly elevated IL17A or TNF-α. As expected, IFN-γ mRNA was not detected in IFN-γ KO mice, IL-17A mRNA was not detected in IL-17A KO mice, and neither cytokine was detected in DKO mice.

Both IFN-γ and IL-17 contribute to gastritis in the absence of Treg.

The data described above (Fig. 7 and 8) indicate that in mice with mature T and B cells, IL-17A contributes to gastritis, but to a lesser extent that does IFN-γ. To determine if the absence of Treg alters this IFN-γ/IL-17 balance, we performed adoptive transfer studies with T cells deficient in one or both cytokines. H. pylori-infected and control mice were inoculated with 0.5 × 10⁶ unfractionated CD4⁺ splenocytes from either C57BL/6 mice, congenic IFN-γ KO mice, IL-17A KO mice, or DKO mice. In this experiment, gastric lamina propria leukocyte (LPL) T cell subsets were quantified by flow cytometry, gastritis was scored, and cytokine mRNA was quantified by qRT-PCR. Figure 9 shows that, as previously reported, gastritis in recipient mice was more severe and rapidly progressive than in immunocompetent mice, a finding which we attribute to the absence of Treg (compare to Fig. 7). This was true in all recipient groups, suggesting that Treg suppression affects CD4 T cells regardless of IFN-γ or IL-17A production. Gastritis was present in all infected recipient mice but was significantly greater in recipients of C57BL/6 and IL-17KO T cells than in the other two groups, consistent with previous findings that Th1 cells contribute to gastritis. Interestingly, gastritis in IL-17KO recipients was somewhat less extensive than in recipients of C57BL/6 cells, but the difference did not reach statistical significance. Thus, while there is suggestive evidence that IL-17A may contribute to gastritis in this model, we were not able to demonstrate this definitively.

Flow cytometric evaluation of gastric T cell subsets supported this interpretation. Gastric CD4 cell engraftment did not differ between donor groups (Fig. 10A), indicating that failure to produce IL-17A, IFN-γ, or both did not alter the ability of T cells to engraft or to migrate to the infected gastric mucosa. Interestingly, however, gastric Treg engraftment differed between the groups. Treg engraftment was greatest in recipients of IFN-γ KO or DKO cells and in uninfected mice and, as would be expected, was inversely correlated with gastritis (Fig. 10B). The explanation for the decreased Treg engraftment of T cells that are capable of producing IFN-γ is not known, but failure of such suppression may be one factor that contributes to amelioration of gastritis in recipients of IFN-γ KO cells.

The pattern of gastric cytokine expression in recipient mice differed somewhat from the pattern of expression in C57BL/6 and cytokine KO mice (Fig. 11). In recipients of C57BL/6 cells, IFN-γ, TNF-α, and IL-17A were all upregulated, consistent with the extensive gastritis in these mice. Interestingly, however, upregulation of TNF-α (up to 800-fold compared to uninfected mice) was much greater than upregulation of IFN-γ (less than 30-fold). Similar to IL-17A KO mice, recipients of IL-17A KO cells had increased expression of TNF-α and IL-17F, but not IFN-γ, in their gastric mucosa. Recipients of IFN-γ KO cells failed to upregulate any of the cytokines measured, indicating an absence of synergy with either IL-17A or IL-17F in IFN-γ-deficient T cells. However, since these mice did develop gastritis, it is likely that some other proinflammatory cytokine(s), such as IL-1β or IL-6, not examined in this study, contributed to the inflammatory response. Finally, surprisingly, the highest overall gastric cytokine upregulation was in recipients of DKO cells. IFN-γ, TNF-α, and IL-17F were all markedly upregulated in these mice. While these results confirm our previous studies suggesting that IFN-γ in recipient mice is derived from innate rather than adaptive immune cells, we were unable to explain the marked increase in expression compared to IFN-γ KO recipients.