A Novel Gastric Spheroid Co-culture Model Reveals Chemokine-Dependent Recruitment of Human Dendritic Cells to the Gastric Epithelium.

BACKGROUND & AIMS: Gastric dendritic cells (DCs) control the adaptive response to infection with Helicobacter pylori, a major risk factor for peptic ulcer disease and gastric cancer. We hypothesize that DC interactions with the gastric epithelium position gastric DCs for uptake of luminal H pylori and promote DC responses to epithelial-derived mediators. The aim of this study was to determine whether the gastric epithelium actively recruits DCs using a novel co-culture model of human gastric epithelial spheroids and monocyte-derived DCs.
METHODS: Spheroid cultures of primary gastric epithelial cells were infected with H pylori by microinjection. Co-cultures were established by adding human monocyte-derived DCs to the spheroid cultures and were analyzed for DC recruitment and antigen uptake by confocal microscopy. Protein array, gene expression polymerase chain reaction array, and chemotaxis assays were used to identify epithelial-derived chemotactic factors that attract DCs. Data from the co-culture model were confirmed using human gastric tissue samples.
RESULTS: Human monocyte-derived DCs co-cultured with gastric spheroids spontaneously migrated to the gastric epithelium, established tight interactions with the epithelial cells, and phagocytosed luminally applied H pylori. DC recruitment was increased upon H pylori infection of the spheroids and involved the activity of multiple chemokines including CXCL1, CXCL16, CXCL17, and CCL20. Enhanced chemokine expression and DC recruitment to the gastric epithelium also was observed in H pylori-infected human gastric tissue samples.
CONCLUSIONS: Our results indicate that the gastric epithelium actively recruits DCs for immunosurveillance and pathogen sampling through chemokine-dependent mechanisms, with increased recruitment upon active H pylori infection.

See editorial on page 155.

Human gastric spheroids recruited dendritic cells (DCs) to the basolateral surface of the epithelium through a chemokine-dependent mechanism and established tight contacts with the DCs. Helicobacter pylori infection of the organoids increased DC recruitment and enabled DC phagocytosis of the bacteria.

Infection with the gastric pathogen Helicobacter pylori is the leading cause of chronic gastritis, peptic ulcer disease, mucosa-associated lymphoid tissue lymphoma, and gastric adenocarcinoma worldwide.1, 2 Dendritic cells (DCs) in the human gastric mucosa are thought to be the major antigen-presenting cells that induce protective immune responses to H pylori infection. We previously showed that human HLA-DRhigh gastric mononuclear phagocytes (MNPs) with functional properties of DCs were present directly adjacent to the gastric epithelium or integrated within the gastric epithelium. In a mouse model of Helicobacter infection, increased numbers of CD11c+ DCs migrated to the epithelial layer and extended projections between the gastric epithelial cells. These interactions between DCs and the epithelium likely promote sampling of H pylori bacteria from the gastric lumen. Close association with the epithelium also positions the DCs for uptake of H pylori antigen that may have crossed the epithelial barrier into the gastric lamina propria. Furthermore, bidirectional cross-talk between gastric mucosal DCs and epithelial cells contributes to the maintenance of gastric homeostasis, as we have shown.6, 7

A recent study performed in murine small intestine showed that MNP recruitment to the epithelium at steady state is dependent on CCR6 and is required for luminal antigen access. However, the mechanisms that determine how human gastric DCs are recruited and retained by the gastric epithelium currently are unclear. Therefore, we developed an in vitro co-culture model of human gastric spheroids and monocyte-derived DCs (MoDCs) to study the mechanisms of gastric DC–epithelial interactions in primary human cells.

Transwell-based co-culture systems have provided crucial insights into the interactions of the gastrointestinal epithelium with pathogens and immune cells.9, 10, 11, 12 In a groundbreaking study by Rescigno et al, DCs were shown to open the tight junctions between epithelial cells, send dendrites outside the epithelium, and directly sample bacteria from the luminal side of the colonic epithelium. Co-culture studies also have shown the importance of transforming growth factor-β and thymic stromal lymphopoietin for the homeostatic conditioning of intestinal DCs by the epithelium.10, 12 However, traditional co-culture systems have several drawbacks. First, epithelial monolayers for co-culture studies generally are derived from transformed cell lines such as Caco2 cells, which differ in their phenotype and function from primary epithelial cells. Second, epithelial cells in co-culture systems usually are cultured on Transwell inserts composed of a synthetic polymer membrane with pores. This membrane presents an artificial barrier that impairs and limits the contact between immune cells and the epithelium. Third, to promote contacts between DCs and the basolateral side of the epithelium, DCs are either placed on the basolateral side of inverted Transwells, so that they move toward the epithelium because of gravity,10, 13 or epithelial cells are seeded on top of the DCs, also leading to forced interactions.

In 2 recent studies, researchers have started to use organoids as an improved epithelial cell model to investigate epithelial interactions with immune cells.15, 16 Our co-culture model of human gastric epithelial spheroids and MoDCs lead to the spontaneous establishment of direct interactions between DCs and the basolateral side of the gastric epithelium. By using this model, we showed that DCs added to gastric epithelial spheroid cultures spontaneously migrated toward the epithelium to establish tight contacts. DC migration was dependent on epithelial chemokine secretion and was enhanced in the presence of luminal H pylori infection. Moreover, DCs were able to phagocytose H pylori applied to the spheroid lumen. These functionally relevant findings indicate that our primary cell co-culture model is suitable for studies of gastric infection and epithelial–immune cell interactions and will allow us to advance our understanding of human gastric immunobiology.

Results

Human DC Are Recruited to the Epithelium of Gastric Spheroids

To perform functional studies of human DC interactions with the gastric epithelium, we generated gastric epithelial spheroids from adult stomach tissue and then established co-cultures with human DCs by adding immature MoDCs to the culture medium surrounding the epithelial spheroids (Figure 1A). Although DCs appeared to accumulate at the border between the tissue culture medium and the spheroid-containing Matrigel plug to some extent (Figure 1B), a considerable number of DCs rapidly entered the Matrigel matrix and associated with the organoids. Live imaging of spheroids and CellTracker Green-labeled (ThermoFisher Scientific, Waltham, MA) MoDCs showed that DCs spontaneously migrated toward the epithelium of the spheroids and accumulated around the basolateral side of the gastric epithelium (Figure 1C and Supplementary Video). Importantly, particle tracking analysis showed directed migration of the DCs toward the basolateral side of the epithelial cells (Figure 1D). In fact, some gastric spheroid–associated DCs were detected as early as 10 minutes after initiation of the experiment (Figure 1E).

Figure 1

MoDCs spontaneously migrate to the basolateral side of the gastric epithelium upon co-culture with gastric spheroids. (A) Graphic representation of the DC–epithelial spheroid co-culture model. (B) Fluorescent confocal image of the bottom of a representative culture plate shows the Matrigel border with accumulated cells. Blue, CellTracker DeepRed-labeled DCs; red, mCherry-expressing spheroids. Scale bar: 200 μm. (C) Representative gastric spheroid with surrounding CellTracker Green-labeled MoDCs. Merge of phase-contrast image and green fluorescent image. Arrowheads indicate accumulated MoDCs. Scale bar: 200 μm. (D) Particle tracking analysis of time-lapse images shows DC tracks directed toward the spheroid. Dotted line indicates the position of the spheroid. Images were acquired on an EVOS FL Auto system with a 10× objective and are representative of 4 independent experiments. (E) Time-lapse confocal imaging of mCherry-expressing gastric spheroids and CellTracker DeepRed-labeled MoDCs (blue) shows rapid recruitment of MoDCs to the spheroids. Single images obtained at different time points after initiation of the co-culture and a Z-stack projection of 50 slices obtained after 8 hours (right, step size 10 μm). Scale bar: 100 μm. All data are representative of 4 or more independent experiments performed with different spheroid lines and DC preparations.

MoDCs Establish Direct Contacts With Co-cultured Gastric Epithelial Spheroids

To analyze the interactions between human gastric epithelial spheroids and co-cultured DCs in more detail, we performed high-resolution confocal imaging of co-cultures. DCs frequently were present directly adjacent to the basolateral side of the epithelium (Figure 2A–C). Specifically, MoDCs formed direct contacts with the epithelium and extended long dendrites along the basolateral side of the gastric epithelium (Figure 2D–F). In some instances, dendrites with globular endings were extended between the epithelial cells (Figure 2F). These dendrites were morphologically similar to the transepithelial dendrites formed by DCs in the stomach and intestine in vivo and in other co-culture systems.5, 9, 15, 17, 18 Notably, some MoDCs contained mCherry-positive cytoplasmic inclusions (Figure 2E), possibly representing phagocytosed material derived from apoptotic gastric epithelium. Interestingly, we also detected DCs that had migrated into the spheroid lumen in 7 of 159 (4.4%) spheroids analyzed (data not shown). Overall, these observations indicate that DCs in our co-culture model spontaneously establish interactions with the gastric epithelium.

Figure 2

MoDCs directly associate with the epithelium of co-cultured gastric spheroids. CellTracker-labeled MoDCs were added to cultures of mCherry-expressing gastric spheroids. After 48 hours, cultures were fixed and imaged on a Leica SP5 Confocal Scanning Laser Microscope using a 63× water immersion objective. (A–C) CellTracker DeepRed-labeled MoDCs (blue) that have migrated to the basolateral side of a gastric epithelial spheroid (red). Scale bar: 50 μm. (A) XY image, maximum Z projection of 57 individual images, step size of 1 μm. (B) XZ projection of the same spheroid. (C) Single XY slice shows the physical association between a DC (arrowhead) and the spheroid. (D) Z-projection of 6 adjacent slices shows extensive dendrite formation on the surface of a gastric spheroid. (E) High-resolution image showing CellTracker DeepRed-labeled MoDCs (blue) in direct contact with gastric epithelial cells (arrowheads) and 1 MoDC with internalized mCherry+ gastric epithelial cell material (pink in merged image; arrow). Single XY slice. Scale bar: 20 μm. (F) CellTracker Green-labeled MoDC (green) extending a dendrite (arrowhead) into the epithelial layer (red). XY image, Z projection of 4 adjacent slices. Scale bar: 20 μm. All data are representative of 4 or more independent experiments performed with different spheroid lines and DC preparations.

H pylori Infection of Human Gastric Epithelial Spheroids Induced Increased Recruitment of DCs to the Gastric Epithelium

To determine the effects of luminal H pylori infection on DC recruitment to the spheroid epithelium, we performed microinjection of gastric spheroids with H pylori. Successful injections were confirmed by replating a small proportion of dissociated spheroids at the end of the experiment. Recovery of significant numbers of viable bacteria was achieved consistently after 48 hours, but bacterial numbers varied widely between experiments (Figure 3A). In contrast, H pylori were unable to survive when injected directly into Matrigel, (Corning, Corning, NY) and we also were unable to recover viable bacteria from the media surrounding the injected organoids (not shown). Interestingly, bacterial recovery was decreased in cultures where DCs were present than in cultures of infected organoids alone (Figure 3A). Luminal oxygen concentrations in the gastric spheroids were reduced significantly compared with the surrounding Matrigel, possibly supporting growth of the microaerophilic H pylori (Figure 3B). These data are consistent with earlier reports showing that human intestinal organoids have reduced luminal oxygen levels.

Figure 3

(A) Bacterial recovery from Matrigel alone or from H pylori–infected gastric spheroids (20–50/culture) with or without co-cultured MoDCs present. Spheroid cultures were disrupted by trypsinization within the first 3 hours after H pylori infection or 48 hours later, and cells were plated for colony-forming unit (CFU) counts (N = 3–9). Means ± SD and individual values are shown. **P ≤ .01, Kruskal–Wallis test with the Dunn multiple comparison. (B) Microsensor analysis of luminal oxygen concentrations measured in gastric spheroids (n = 5 or 6 organoids, means ± SD). ***P ≤ .001, 1-way analysis of variance with the Dunnett multiple comparison. Dotted line indicates the concentration of oxygen in buffer. (C) Representative confocal images (single optical sections) of a noninfected and a GFP–H pylori (green)-infected, mCherry-expressing gastric spheroid (red) co-cultured with CellTracker-labeled MoDCs (blue) for 48 hours. Scale bar: 50 μm. (D) Image analysis approach for organoid co-cultures. The organoid area was defined as the area within a 75-μm radius of the spheroid, and the DC area was measured digitally on an auto-thresholded image. (E) Digital image analysis of 8 noninfected and 9 spheroid co-cultures infected with GFP–H pylori from 3 independent experiments. Individual data points, means ± SD are shown. Significance between groups was analyzed by an unpaired Student t test. (F) Gastric epithelial spheroids were replated as monolayers and treated with H pylori 60190 wild type (wt, 5 × 107/well) or a panel of TLR agonists (see the Materials and Methods section) for 6 hours. Gene expression of CXCL8 was determined by quantitative RT-PCR. Individual values (symbols) and means ± SD (bar and error bars) from 3 independent experiments are shown. *P ≤ .05; Kruskal–Wallis test with the Dunnett multiple comparison. (G) Gastric spheroids were microinjected with FITC dextran alone (molecular weight, 4 kilodaltons, tracer) or with FITC dextran containing H pylori 60190 wild type, H pylori 60190 ΔCagA, or Salmonella flagellin (1 μg/mL). Spheroids that remained intact throughout the experiment, as determined by retention of FITC dextran in the lumen, were analyzed for DC recruitment after 48 hours. Pooled data from 2 independent experiments were analyzed by Kruskal–Wallis test with the Dunn multiple comparison (*P ≤ .05). All co-cultures were imaged on a Leica SP5 Confocal Scanning Laser Microscope using a 20× objective, and imaging data were analyzed using ImageJ software.

We next performed confocal imaging analysis of DCs co-cultured with green fluorescent protein (GFP)–H pylori–infected or mock-injected gastric spheroids (Figure 3D). DCs were added 2–3 hours after H pylori infection of the spheroids. After 48 hours, a significantly higher number of DCs was detected in close proximity to H pylori–infected spheroids compared with noninfected spheroids (Figure 3C and E). We also compared DC recruitment by gastric spheroids injected with H pylori 60190 with DC recruitment by spheroids injected with the Toll-like receptor (TLR)5 agonist Salmonella flagellin, which consistently induced spheroid activation as determined based on CXCL8 gene expression (Figure 3F). Fluorescein isothiocyanate (FITC) dextran was injected as a fluorescent tracer to identify injected organoids that remained intact throughout the experiment. TLR stimulation caused a slight, but not significant, increase in DC recruitment (Figure 3G). Interestingly, however, infection of the organoids with a cytotoxin-associated antigen A (CagA)-negative deletion mutant of H pylori 60190 did not enhance DC accumulation above baseline (Figure 3G). Intensity analysis of FITC-dextran fluorescence showed no difference between treatments, indicating that the injected substances did not significantly alter epithelial barrier function (data not shown). Overall, these observations suggest that H pylori infection promotes DC recruitment to the gastric epithelium, likely through a mechanism that involves H pylori–dependent release of chemotactic factors by the epithelial cells.

Gastric Epithelial-Derived Chemokines Induce DC Recruitment

To determine the mechanism for DC recruitment to the gastric epithelium, we analyzed supernatants from gastric spheroid cultures in Transwell chemotaxis assays. MoDCs were significantly recruited by supernatants from uninfected spheroid supernatants (Figure 4A), consistent with the data obtained by microscopic analysis of the MoDC–gastric spheroid co-cultures. Moreover, H pylori infection of the spheroids significantly increased the chemotactic index of MoDCs toward spheroid culture supernatants (Figure 4B), corroborating the confocal microscopy data shown in Figure 3. Notably, we did not observe significant chemokinetic activity of MoDCs in response to gastric spheroid supernatants (data not shown). Addition of pertussis toxin (10 μg/mL), which blocks G-protein–dependent signaling, significantly reduced DC recruitment by both H pylori–infected and noninfected spheroids, suggesting that DC recruitment is mediated by chemokine-dependent mechanisms (Figure 4C).

Figure 4

(A) Culture supernatants (S/Ns) collected over 48 hours from established spheroid cultures were tested for their ability to attract MoDCs (2 × 105/well) in a Transwell chemotaxis assay. Individual data points from 16 independent experiments performed in triplicate (means ± SD, ***P ≤ .001, paired Student t test). (B) Culture supernatants collected from gastric spheroids with or without H pylori infection 48 hours after microinjection of the spheroids were analyzed in Transwell chemotaxis assays with MoDCs (1 × 105/well). Means ± SD of triplicate wells, 1 representative experiment out of 4 is shown. Analysis of variance with the Tukey multiple comparisons test (**P ≤ .01, ***P ≤ .001). (C) MoDC chemotaxis to medium alone or gastric spheroid supernatants in the presence of absence of pertussis toxin (PTX, 10 μg/mL) was analyzed. Means ± SD of triplicate wells, 1 representative experiment out of 3 is shown. Analysis of variance with the Tukey multiple comparison test; different letters signify statistically significant differences (P ≤ .05). (D) Supernatants collected from spheroid cultures with and without H pylori infection (48 h) were analyzed for the presence of chemokines using a chemokine array. Pooled data from 4 independent experiments. *P ≤ .05, Student t test. (E) Relative gene expression of select chemokine and cytokine genes in H pylori–infected compared with noninfected spheroids. Means ± SD of 3 independent infection experiments. Only genes that showed consistent >2-fold up-regulation in all 3 experiments are shown. *P ≤ .05, Student t test. (F) Culture supernatants were collected after 48 hours from gastric spheroids microinjected with broth alone, H pylori wild type (wt), H pylori ΔCagA, or Salmonella flagellin. Supernatants were analyzed in Transwell chemotaxis assays with MoDCs. One representative experiment out of 3 is shown, means ± SD of triplicate wells, analysis of variance with the Tukey multiple comparison test; *P ≤ .05, **P ≤ .01. (G) Supernatants from spheroid cultures with different treatments were analyzed for the presence of chemokines by chemokine array. One representative experiment out of 2 is shown. Different letters signify statistically significant differences in chemokine concentration (P ≤ .05).

To identify the chemokines involved in both steady-state and infection-induced recruitment of DCs to the gastric epithelium, we performed a chemokine antibody array with supernatants from noninfected and H pylori–infected gastric spheroids. By using this approach, we detected baseline secretion of CXCL5, CXCL17, CCL20, CXCL8 (interleukin [IL]8), CXCL16, CXCL1, and midkine (neurite growth-promoting factor 2) ≥2-fold greater than background levels present in spheroid media, with considerable variations between spheroid lines (Figure 4D and Supplementary Table 1). Spheroid infection with H pylori significantly increased secretion of CCL20 and consistently up-regulated CXCL1 and CXCL8 by the gastric epithelium, whereas CXCL16, CXCL17, and midkine consistently were decreased in the presence of H pylori.

Supplementary Table 1

Raw Data Chemokine Array of H pylori–Infected Gastric Organoids

TARGET	EXPERIMENT 1			EXPERIMENT 2			EXPERIMENT 3			EXPERIMENT 4
	Control	H pylori	P value	Control	H pylori	P value	Control	H pylori	P value	Control	H pylori	P value
CCL1	0.82	1.90	<.001	0.8	0.8	.855	0.65	0.43	.028	0.37	0.23	.032
CCL14	1.28	1.96	.001	1.0	0.8	.487	0.75	0.43	.137	0.40	0.21	.025
CCL15	0.36	0.86	.010	0.9	1.5	.020	0.64	0.66	.832	0.31	0.20	.012
CCL17	1.60	1.54	.784	1.0	0.9	.837	0.74	0.84	.143	0.46	0.29	.007
CCL18	1.12	1.61	.011	0.8	0.7	.657	0.73	0.39	.004	0.36	0.22	.002
CCL19	0.81	0.68	.481	0.8	1.1	.214	0.48	0.64	.037	0.43	0.30	.071
CCL2	1.61	2.74	<.001	1.0	1.5	.035	0.52	0.61	.343	0.30	0.22	.023
CCL20	1.44	3.47	<.001	5.3	9.5	<.001	1.40	5.73	.001	1.43	7.85	.001
CCL21	0.58	0.75	.391	0.7	0.7	.852	0.55	0.48	.426	0.37	0.22	.016
CCL22	1.14	0.96	.350	0.8	0.8	.906	0.57	0.37	.024	0.44	0.25	.017
CCL26	0.71	1.14	.027	1.1	1.2	.860	1.21	1.02	.584	0.53	0.30	.000
CCL28	2.22	2.51	.121	1.0	0.8	.448	1.02	0.87	.733	0.49	0.31	.024
CCL3/CCL4	1.50	1.70	.290	0.9	0.9	.902	0.72	0.83	.286	0.31	0.19	.004
CCL5	1.00	1.22	.258	0.8	0.6	.591	0.57	0.69	.024	0.31	0.22	.011
CCL7	0.88	1.22	.083	0.8	0.7	.528	0.60	0.49	.418	0.34	0.19	.007
CHEMERIN	1.57	1.53	.827	1.0	0.8	.489	0.79	0.74	.010	0.49	0.30	.037
CX3CL1	1.65	1.91	.176	0.9	1.0	.629	0.72	0.85	.194	0.36	0.29	.215
CXCL1	2.03	3.65	<.001	4.2	6.6	<.001	2.40	5.08	.003	4.09	4.56	.297
CXCL10	0.87	0.85	.957	1.3	1.9	.025	1.34	0.94	.001	0.71	0.52	.002
CXCL11	1.60	1.75	.432	1.0	1.1	.552	0.59	0.80	.060	0.35	0.23	.009
CXCL12	0.20	0.21	.959	1.1	1.1	.886	0.80	0.81	.701	0.63	0.41	.002
CXCL16	1.87	2.74	<.001	4.5	3.6	<.001	2.57	1.81	.004	3.56	1.88	.001
CXCL17	2.34	2.26	.697	6.7	5.6	<.001	4.52	3.13	.004	6.45	1.84	.000
CXCL4	1.08	1.21	.514	1.3	1.2	.540	2.05	0.59	.002	1.83	0.92	.002
CXCL5	4.96	4.91	.800	9.8	12.1	<.001	5.17	4.92	.095	0.32	0.22	.033
CXCL7	0.76	1.97	<.001	2.3	1.9	.089	1.24	1.19	.841	0.58	0.81	.001
CXCL8	1.62	3.99	<.001	5.1	7.1	<.001	3.60	3.61	.995	4.20	6.44	.015
CXCL9	1.54	2.11	.004	1.0	1.0	.887	0.94	1.04	.624	0.44	0.31	.048
IL16	1.57	1.45	.515	0.9	0.9	.888	0.90	0.82	.704	0.36	0.27	.134
MIDKINE	1.66	2.14	.015	2.9	2.2	.004	3.74	0.93	.011	1.84	1.04	.003
XCL1	1.51	1.80	.126	1.0	1.3	.325	0.60	0.90	.001	0.39	0.35	.096

NOTE. Chemokine detection in gastric spheroid supernatants analyzed using the Proteome Profiler Human Chemokine Array Kit. Relative protein concentrations in supernatants collected after 48 hours of culture from noninfected and H pylori–infected gastric spheroids compared with background levels present in spheroid culture medium was determined by densitometry. Data sets from 2 independent experiments are shown. P values were calculated using a Student t test and were considered significant at P ≤ .05 (values bolded).

We also analyzed the effect of H pylori infection of human gastric spheroids on chemokine gene expression using an RT2 Profiler (Qiagen, Valencia, CA) reverse-transcription polymerase chain reaction (RT-PCR) gene expression array (Figure 4E and Supplementary Table 2). Overall, gene expression patterns largely matched protein secretion data. CXCL1, CXCL2, and CXCL8 were up-regulated significantly after 3 hours of H pylori stimulation in 3 independent experiments (P ≤ .05), confirming results from previous studies.21, 22 CCL2, CCL20, CXCL3, and CXCL6 also were up-regulated consistently by ≥2-fold in H pylori–infected spheroids, but the magnitude of up-regulation varied widely. No significant down-regulation was observed for any of the genes analyzed.

Supplementary Table 2

Raw Data PCR Array of H pylori–Infected Gastric Organoids

Target	Untreated			H pylori–infected			P value	Significant?
C5	1.0	1.0	1.0	1.0	0.9	1.2	.7247
C5AR1	1.0	1.0	1.0	10.3	1.3	0.1	.4185
ACKR2	1.0	1.0	1.0	10.2	1.8	15.9	.1126
CCL1	1.0	1.0	1.0	1.4	2.0	0.9	.2446
CCL11	1.0	1.0	1.0	1.0	0.0	1.2	.5122
CCL13	1.0	1.0	1.0	1.0	0.0	2.6	.8047
CCL14	1.0	1.0	1.0	0.1	0.6	1.1	.2381
CCL15	1.0	1.0	1.0	0.6	1.0	92.4	.3767
CCL16	1.0	1.0	1.0	0.5	2.0	2.2	.3504
CCL17	1.0	1.0	1.0	2.1	1.9	22.5	.3156
CCL18	1.0	1.0	1.0	21.2	0.0	1.2	.4002
CCL19	1.0	1.0	1.0	0.1	0.7	0.1	.0249	Down*
CCL2	1.0	1.0	1.0	2.6	8.0	17.5	.1271
CCL20	1.0	1.0	1.0	20.9	15.8	48.2	.0533
CCL21	1.0	1.0	1.0	0.0	1.1	1.2	.5766
CCL22	1.0	1.0	1.0	301.3	0.6	120.3	.1851
CCL23	1.0	1.0	1.0	0.9	0.3	16.9	.4069
CCL24	1.0	1.0	1.0	108.0	0.1	0.1	.3848
CCL25	1.0	1.0	1.0	32.6	0.3	0.6	.3965
CCL26	1.0	1.0	1.0	1.7	0.8	0.8	.7556
CCL27	1.0	1.0	1.0	1.3	0.8	0.1	.4862
CCL28	1.0	1.0	1.0	0.8	1.2	0.4	.4353
CCL3	1.0	1.0	1.0	13.7	0.9	1.2	.3689
CCL4	1.0	1.0	1.0	61.6	0.0	1.2	.3825
CCL5	1.0	1.0	1.0	30.4	0.7	23.8	.1271
CCL7	1.0	1.0	1.0	1.3	10.2	1.5	.3193
CCL8	1.0	1.0	1.0	1.0	0.2	1.2	.5484
CCR1	1.0	1.0	1.0	1.4	0.4	0.3	.4411
CCR10	1.0	1.0	1.0	0.5	0.3	1.2	.2889
CCR2	1.0	1.0	1.0	1.1	0.7	10.6	.3877
CCR3	1.0	1.0	1.0	2.7	0.7	0.4	.7306
CCR4	1.0	1.0	1.0	1.0	0.6	1.2	.7247
CCR5	1.0	1.0	1.0	1.0	0.0	1.2	.5122
CCR6	1.0	1.0	1.0	6.3	1.9	1.2	.2523
CCR7	1.0	1.0	1.0	26.1	1.5	1.2	.3561
CCR8	1.0	1.0	1.0	6.3	1.4	3.4	.1305
CCR9	1.0	1.0	1.0	2.6	0.6	0.8	.6279
ACKR4	1.0	1.0	1.0	1.1	1.4	1.2	.0572
CCRL2	1.0	1.0	1.0	2.1	1.9	2.2	.0003	Up***
CKLF	1.0	1.0	1.0	1.0	1.3	0.8	.8298
CMKLR1	1.0	1.0	1.0	1.0	27.0	1.2	.3690
CMTM1	1.0	1.0	1.0	1.3	1.3	0.6	.7893
CMTM2	1.0	1.0	1.0	1.8	0.5	0.4	.8354
CMTM3	1.0	1.0	1.0	3.7	0.8	0.5	.5492
CMTM4	1.0	1.0	1.0	1.6	1.1	0.6	.7465
CX3CL1	1.0	1.0	1.0	9.5	1.0	35.3	.2387
CX3CR1	1.0	1.0	1.0	1.0	0.9	1.2	.7247
CXCL1	1.0	1.0	1.0	17.3	12.3	29.0	.0200	Up*
CXCL10	1.0	1.0	1.0	99.4	1.2	8.5	.3255
CXCL11	1.0	1.0	1.0	211.6	0.9	0.0	.3773
CXCL12	1.0	1.0	1.0	4.4	0.8	1.2	.3761
CXCL13	1.0	1.0	1.0	3.6	1.5	2.6	.0611
CXCL14	1.0	1.0	1.0	1.0	0.9	32.9	.3759
CXCL16	1.0	1.0	1.0	1.7	1.4	1.1	.0821
CXCL2	1.0	1.0	1.0	10.2	7.2	14.5	.0104	Up*
CXCL3	1.0	1.0	1.0	5.1	3.0	13.7	.1281
CXCL5	1.0	1.0	1.0	1.8	1.5	7.2	.2484
CXCL6	1.0	1.0	1.0	12.3	2.8	13.3	.0646
CXCL9	1.0	1.0	1.0	1.5	0.9	1.9	.2102
CXCR1	1.0	1.0	1.0	1.0	0.9	1.2	.7247
CXCR2	1.0	1.0	1.0	0.9	0.4	0.7	.0835
CXCR3	1.0	1.0	1.0	1.5	1.0	0.8	.6560
CXCR4	1.0	1.0	1.0	6.3	2.1	2.5	.1205
CXCR5	1.0	1.0	1.0	2.9	2.8	1.2	.0776
CXCR6	1.0	1.0	1.0	1.1	0.8	1.1	1.0000
ACKR3	1.0	1.0	1.0	1.0	0.9	1.2	.7247
ACKR1	1.0	1.0	1.0	1.0	0.9	1.2	.7247
FPR1	1.0	1.0	1.0	50.4	0.0	1.2	.3845
GPR17	1.0	1.0	1.0	1.5	0.6	0.1	.5505
HIF1A	1.0	1.0	1.0	1.7	1.4	0.9	.2263
IL16	1.0	1.0	1.0	1.0	0.9	1.2	.7247
IL1B	1.0	1.0	1.0	10.7	4.8	1.9	.1374
IL4	1.0	1.0	1.0	54.4	3.4	1.2	.3432
CXCL8	1.0	1.0	1.0	32.1	27.2	7.2	.0499	Up*
PF4V1	1.0	1.0	1.0	0.3	0.7	1.1	.2638
PPBP	1.0	1.0	1.0	0.2	0.7	0.8	.0798
SLIT2	1.0	1.0	1.0	1.0	0.9	1.2	.7247
TLR2	1.0	1.0	1.0	2.9	1.7	8.1	.1751
TLR4	1.0	1.0	1.0	2.4	1.6	1.5	.0430
TNF	1.0	1.0	1.0	40.1	39.3	226.0	.1799
TYMP	1.0	1.0	1.0	1.4	0.5	1.2	.9087
XCL1	1.0	1.0	1.0	1.0	0.9	1.2	.7247
XCL2	1.0	1.0	1.0	1.0	0.1	0.1	.1161
XCR1	1.0	1.0	1.0	0.1	0.5	1.2	.2813
C5	1.0	1.0	1.0	1.0	0.9	1.2	.7247

NOTE. Relative gene expression of H pylori–infected gastric organoids determined by PCR array analysis. Data from 3 independent experiments are shown. P values were calculated using a paired Student t test and were considered significant at P ≤ .05. *P ≤ .05, ***P ≤ .001.

TNF, tumor necrosis factor.

Having shown that luminal application of wild-type H pylori but not of CagA-deficient H pylori or Salmonella flagellin induced an accumulation of DCs immediately around the organoid (Figure 3G), we next compared the chemotactic activity and the chemokine profile of supernatants recovered from spheroids treated with broth alone, wild-type H pylori, CagA-deficient H pylori, and flagellin. Similar to our observations with the co-cultures, wild-type H pylori was the most efficient at inducing DC chemotaxis (P ≤ .01), but the CagA-deficient H pylori also induced significant DC migration in the Transwell assay (P ≤ .05), whereas no significant effects were observed for flagellin application (Figure 4F). Analysis of chemokines in the culture supernatants showed inconsistent changes in the chemokine profiles between the different treatment groups (Figure 4G). Chemokines induced by the CagA-deficient H pylori largely followed the same pattern as those induced by wild-type H pylori, with similar CCL20 and CXCL1 release, but significantly lower CXCL8 levels. In contrast, flagellin caused higher CXCL8 release, but lower CCL20 release compared with wild-type H pylori. These observations suggest that a combination of redundantly active chemokines control DC recruitment to the gastric epithelium.

Human MoDCs Show Chemotactic Responses to CXCL1, CXCL16, CXCL17, and CCL20

Having shown that gastric spheroids release factors with chemotactic activity for human DCs and that multiple chemokines, including CXCL5, CXCL17, CCL20, CXCL8, CXCL16, CXCL1, are released basolaterally by gastric spheroids, we tested the ability of human MoDCs to migrate toward these chemokines (Figure 5). In Transwell chemotaxis assays, the DCs showed a strong and significant response to CXCL1 and CXCL16 (Figure 5E and F) and a moderate response to CCL20 and CXCL17 (Figure 5C and E). A weak chemotactic response was observed to CXCL5 and CXCL8 (Figure 5A and D). Together with the observations shown in Figure 4, these data indicate that DC recruitment to the epithelial interface involves the activity of multiple epithelial-derived chemokines and that H pylori infection increases the immune cell recruitment to the epithelium by up-regulating chemokine expression.

Figure 5

MoDCs show chemotactic activity to multiple chemokines that are released by the gastric epithelium. Chemotactic activity of MoDCs toward various concentrations of (A) CXCL5, (B) CXCL17, (C) CCL20, (D) CXCL8, (E) CXCL16, and (F) CXCL1 was analyzed in Transwell chemotaxis assays. Representative results of 3 or more experiments each are shown. Means ± SD of triplicate wells, analyzed by 1-way analysis of variance. The Tukey post hoc test was used to determine differences between individual chemokine concentrations and control, *P ≤ .05, **P ≤ .01, ***P ≤ .001, and ****P ≤ .0001.

Human H pylori Infection Increases Chemokine Expression and DC Recruitment to the Epithelium

To confirm the relevance of our in vitro data, we analyzed chemokine gene expression and epithelial DC recruitment in gastric tissues from H pylori–infected and noninfected donors. Gene expression profiling data overall were consistent with protein and gene expression data obtained from gastric spheroids (Figure 6A). CCL20, CXCL1, CXCL5, and CXCL8 were up-regulated significantly (P ≤ .05) in the H pylori–infected gastric body with and without atrophy and in the H pylori–infected gastric antrum without atrophy, and CCL20 and CXCL8 expression remained significantly increased in atrophic gastric antrum. We also detected significantly up-regulated expression of several other chemokines such as CCL18 and CXCL10, predominantly in nonatrophic samples from the gastric body. In contrast, H pylori infection did not alter the gene expression of CXCL17 and CXCL16 in any of the groups.

Figure 6

(A) Heatmap showing relative chemokine gene expression in human gastric antrum and body from noninfected (Co) and H pylori–infected donors without atrophy (gastritis, Ga) and with gastric atrophy (Atr) (n = 3 each). Data were extracted from Gene Expression Omnibus data set records GDS5338 (antrum) and GDS5411 (body). Arrow indicates significant up-regulation (analysis of variance with the Dunnett multiple comparisons test, P ≤ .05). (B–D) Gastric biopsy samples from H pylori–infected (n = 11) and noninfected (n = 8) human donors were immunolabeled for HLA-DR (Cy3, red) and epithelial cytokeratin (type 7/8; FITC, green). Cell nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI) (blue). Images were acquired on a Nikon Eclipse T2000-U fluorescent microscope equipped with a CoolSnap ES digital camera and NIS Elements BR2.30 software with a 20× objective. (B) Representative image of non–H pylori–infected mucosa. Arrows indicate examples of DCs in direct contact with epithelial cells. Scale bar: 20 μm. (C) The number of HLA-DRhigh DCs in direct contact with the basal side of the gastric epithelium was counted and normalized to epithelial length. Data from individual subjects, means ± SD, are shown. Data were analyzed by the Student t test. (D) The number of intraepithelial DCs was determined using the approach described for panel C.

We next analyzed whether human H pylori infection increases DC recruitment to the gastric epithelium in situ. As previously shown, DCs in the gastric mucosa frequently interact with the epithelium (Figure 6B). Importantly, the number of both lamina propria DCs in direct contact with the gastric epithelium and the number of intraepithelial DCs was increased significantly in H pylori–infected samples (Figure 6C and D) (P = .04 and P = .01, respectively).

Co-culture of MoDCs With H pylori–Infected Gastric Epithelial Spheroids Results in DC Uptake of H pylori

To determine whether DC-epithelial interactions in gastric spheroid-DC co-cultures enable the DCs to sample H pylori antigens, we used high-resolution confocal imaging. A significant percentage of MoDCs in H pylori–infected co-cultures (1.8% ± 0.6%) contained GFP-positive material consistent with phagocytic uptake of H pylori (P ≤ .05), whereas green fluorescent material generally was absent from DCs in co-cultures that were not infected with GFP–H pylori (Figure 7A–D). Interestingly, 5.3% ± 1.3% of DCs in spheroid co-cultures contained CellTracker-labeled material of presumed epithelial origin (Figure 7E), as shown in Figure 2E. The proportion of these cells was decreased significantly upon H pylori infection (P ≤ .05). Overall, these results suggest that MoDCs co-cultured with gastric spheroids perform relevant phagocytic functions, as shown for mononuclear phagocytes in gastric tissues.7, 23

Figure 7

MoDCs co-cultured with (A–C) Confocal image analysis of gastric epithelial spheroid–MoDC co-cultures show internalized GFP-positive material in a subset of MoDCs (arrowheads). mCherry-expressing spheroids (red) were infected with GFP–H pylori (green), and CellTracker-labeled MoDCs (blue) were added for 48 hours. Representative images of 3 independent co-culture experiments. Scale bar: 20 μm. (D and E) Percentage of MoDCs with (D) green fluorescent (GFP+) inclusions and (E) red fluorescent (mCherry+) inclusions (see Figure 3E, pink in merged images) after 48 hours of co-culture with noninfected vs GFP–H pylori–infected gastric spheroids. Co-cultures were imaged on a Leica SP5 Confocal Scanning Laser Microscope using a 63× objective. Data from 1 of 3 representative experiments; individual data points, means and SD, are shown. *P ≤ .05, Mann–Whitney U test.

Discussion

The immunologic mechanisms specific to the human stomach as a unique mucosal environment remain poorly understood. Specifically, it is unclear how antigen-presenting cells are recruited to and interact with the gastric epithelium to access H pylori antigens. To address this knowledge gap, we have developed and evaluated a novel organotypic co-culture model based on primary human cells that enables functional studies of the interactions between the gastric epithelium, DCs, and luminal H pylori bacteria. With this model, we show that human DCs show chemotactic activity for epithelial-derived factors and form tight interactions with the epithelium. Importantly, H pylori infection of the gastric epithelium increased epithelial chemokine expression and DC recruitment to the epithelial interface, both in our co-culture model and in H pylori–infected human subjects.

Cross-talk between MNP populations and epithelial cells at mucosal sites is important for immunosurveillance and mucosal imprinting of the MNPs by epithelial-derived factors.6, 24, 25 In our model, MoDCs co-cultured with gastric epithelial spheroids spontaneously migrated toward the gastric epithelium and established direct contacts with the epithelial cells. To establish contacts with the epithelium, the MoDCs actively penetrated the Matrigel within less than 30 minutes. DCs have been shown previously to migrate through Matrigel by using matrix metalloproteinases that degrade extracellular matrix components.26, 27 Once DCs have migrated toward the gastric epithelium, our co-cultures closely resembled the human gastric mucosa, where MNPs are present immediately beneath the epithelial cell layer and sometimes integrated into the epithelium.4, 7, 17 Moreover, the MoDCs extended dendrites between epithelial cells that were morphologically similar to those previously reported to be involved in antigen uptake by MNPs in the small intestine18, 28 and the stomach.5, 17 It is possible that the dendrites seen in our model enabled the observed phagocytosis of GFP–H pylori bacteria in our study.

Notably, DCs were recruited to the epithelium in the absence of inflammation or infection, likely representing steady-state recruitment. Because of the short life span of mucosal DCs, constant DC recruitment is necessary to replace dying cells to maintain epithelial immunosurveillance. A similar steady-state recruitment of DCs to the epithelium recently was shown in the murine intestine. Importantly, we also showed that DC recruitment to the epithelial interface was increased upon H pylori infection, both in our co-culture model and in gastric biopsy specimens obtained from human donors. Our study thus confirms and extends previous reports that H pylori infection induces gastric epithelial chemokine expression and significantly promotes MNP recruitment.4, 5, 30, 31 In the gastric mucosa, murine Helicobacter infection resulted in increased numbers of CD11c+ and CD68+ cells,5, 30, 31 and we previously have detected increased numbers of HLA-DRhigh MNPs upon human H pylori infection. Here, we specifically showed that increased numbers of gastric DCs interacted with the epithelial layer upon H pylori infection. These increased interactions may have important functional implications for continuous luminal antigen sampling and for positioning DCs for antigen uptake when the epithelial barrier is compromised during H pylori infection. Indeed, DCs co-cultured with H pylori–infected spheroids showed evidence of H pylori phagocytosis, again resembling the behavior of DCs in H pylori–infected gastric mucosa, where H pylori phagocytosis by gastric DCs recently was observed in a mouse model.

Interestingly, significantly higher numbers of DCs accumulated at the epithelium of spheroids infected with wild-type H pylori compared with CagA-deficient H pylori. CagA is a major virulence factor of H pylori that is injected into host cells through a type 4 secretion system and that activates a number of proinflammatory host signaling pathways including nuclear factor-κB, leading to enhanced chemokine secretion.32, 33 However, we did not see any clear differences in chemotactic activity of DCs toward supernatants from spheroids infected with wild-type vs CagA-deficient H pylori. Although CXCL8 secretion by spheroids infected with CagA-deficient H pylori was lower than that induced by wild-type bacteria, confirming earlier studies, other chemokines with stronger chemotactic activity for DCs were unchanged or up-regulated. Thus, the mechanism for the lack of DC recruitment in spheroid co-cultures infected with CagA-deficient H pylori remains elusive. Mustapha et al previously investigated H pylori–dependent chemokine release using primary gastric epithelial cells and showed that disruption of the CagA type IV secretion system (T4SS) significantly decreased chemokine gene expression after 3 hours, but not after 24 hours. Early differences in chemokine secretion may have contributed to altered DC recruitment toward infected spheroids because co-cultures were established 2–3 hours after injection, but these differences may have disappeared after 48 hours when spheroid supernatants were collected.

By using a combination of chemokine analysis in gastric spheroid supernatants, gene expression profiling of gastric spheroids, DC chemotaxis assays, and gene expression analysis of human gastric tissues, we identified a number of chemokines that likely contribute to DC recruitment to the gastric epithelium under steady-state conditions and during H pylori infection. CCL20 was identified as a chemokine that attracted DCs and that was expressed by the gastric epithelium both at steady state and at increased levels upon H pylori infection. CCL20 has been shown previously to regulate steady-state DC recruitment to the epithelium in the murine small intestine. However, whether CCL20 promotes recruitment of human DCs has been a matter of debate.36, 37 Here, we show that certain concentrations of CCL20 significantly promoted DC recruitment and therefore may contribute to both steady-state and infection-induced DC recruitment in the stomach. Likewise, CXCL1 was secreted at high levels by the gastric epithelium at steady state, was up-regulated during H pylori infection at the protein and the RNA levels, and had significant chemotactic activity on human MoDCs. Thus, CXCL1 likely plays an important role in DC recruitment to the gastric epithelium. CXCL8 is a signature chemokine that activates CXCR2 for neutrophil recruitment and that is known to be up-regulated upon gastric H pylori infection.38, 39 We showed that CXCL8 was secreted at a high levels by noninfected gastric spheroids, was up-regulated upon H pylori infection in a CagA-dependent manner, and that CXCL8 also acts on DCs. Interestingly, we also showed that gastric epithelial cells release significant amounts of CXCL5, CXCL16, and CXCL17, and that human DCs show chemotactic responses to both chemokines. However, H pylori infection increased neither CXCL16 nor CXCL17, suggesting that CXCL16 and CXCL17 play a role in steady-state DC recruitment. Previous studies had shown that activated DCs express CXCL16, but whether human DCs express the CXCL16-receptor CXCR6 and are responsive to CXCL16 has not yet been investigated. Likewise, how DCs respond to CXCL17 is currently unknown because the receptor for this orphan chemokine has not yet been identified.

Taken together, these observations suggest that steady-state migration of DCs to the gastric epithelium involves CXCL1, CXCL5, CXCL8, CXCL16, CXCL17, and CCL20, whereas increased DC recruitment upon H pylori infection is driven mostly by CXCL1, CCL20, and possibly CXCL8. Notably, it is unknown whether gastric DCs show chemotactic responses to the same chemokines as the MoDCs used in this study, and we were unable to perform chemotaxis experiments with human gastric DCs owing to limited cell availability. However, we found that chemokine expression by gastric spheroids matched gene expression profiles detected in biopsy specimens of H pylori–infected human subjects, and DC recruitment to the epithelium likewise was increased both in our in vitro model and in vivo. These observations indicate that the results obtained with the spheroid-MoDC co-cultures were biologically relevant.

Notably, our model system used organoid cultures of primary human gastric epithelial cells together with human DCs to study gastric immunobiology. A recent study by Noel et al used epithelial monolayers derived from human enteroids and showed that co-cultured macrophages enhanced epithelial barrier function. We previously co-cultured human blood monocytes on the luminal surface of short-term primary gastric epithelial cell cultures to show that human MNPs are conditioned by retinoic acid released by the gastric epithelium. However, unlike retinoic acid, most epithelial surface proteins and soluble mediators are expressed in a polar manner, so that DCs need to be added to the basolateral side of the epithelium to replicate the physiological orientation of the tissue. Most previous studies have used Transwell systems, in which epithelial cells are seeded on top of a porous membrane and DCs are added to the bottom compartment to achieve the correct spatial relationship between epithelium and immune cells.9, 10, 11, 12, 13, 15, 42, 43, 44 Organoids, including epithelial spheroids and other 3-dimensional long-term culture systems of primary epithelial cells, have a closed lumen and a fully accessible basolateral side on the outside. Thus, our model allows spontaneous interactions between DCs that are added to the cultures and the basolateral surface of the gastric epithelium in the absence of filter membranes. In a similar model, Nozaki et al recently co-cultured murine intraepithelial lymphocytes with small intestinal organoids and showed that dynamic interactions between intraepithelial lymphocytes and the epithelium supported prolonged maintenance of the intraepithelial lymphocytes in culture. Overall, our co-culture system has the advantage of using primary human gastric epithelial cells, allowing direct access of immune cells to the basolateral side of the epithelium, and supporting active H pylori infection over several days, and therefore is suitable for studying gastric epithelial cell–DC cross-talk in vitro. Moreover, we anticipate that the DC-epithelial cell co-culture model described here can be easily adapted for studying other mucosal sites, immune cells, and pathogens.

We recognize that our new co-culture system still has certain limitations. Spheroids and organoids occasionally may rupture, releasing luminal material including infectious agents and apoptotic epithelial cell material and mucus in an uncontrolled manner. We addressed this issue by injecting FITC dextran as a tracer into the spheroids. FITC dextran is rapidly released from the spheroid lumen upon rupture, as we have shown. Therefore, organoids that remained intact throughout the experiment could be identified based on the presence of the tracer. However, we still cannot rule out the presence of some H pylori bacteria outside the gastric spheroids owing to leakage during microinjection. One other limitation of any organoid culture system is that, unless cells are re-seeded on a Transwell insert, it is impossible to determine epithelial cell numbers, and thus multiplicity of infection, with a high degree of accuracy, owing to differences in size and shape of the organoids. Moreover, bacterial delivery into the lumen of gastric spheroids by microinjection remains difficult to automate. Despite these challenges, our co-culture system yielded reproducible and statistically significant results. Specifically, our experiments using gastric spheroid–MoDC co-cultures have shown that the gastric epithelium promotes accelerated recruitment of DCs to the epithelial cell interface upon H pylori infection through a chemokine-dependent mechanism.

Materials and Methods

Human Blood and Tissue Samples

Blood samples for isolation of monocytes were obtained with local Institutional Review Board approval from our healthy adult donor pool at Montana State University by a trained phlebotomist. Healthy human gastric tissue samples for the generation of spheroid lines were obtained with Institutional Review Board approval as surgical resection material from sleeve gastrectomies from the National Disease Research Interchange (Philadelphia, PA).

Gastric Epithelial Spheroid Cultures

Gastric epithelial spheroids were derived from human gastric glands and cultured in the presence of Wnt3a, noggin, and R-spondin, as previously described. Lentiviral transduction was used to generate mCherry-expressing spheroid lines.

MoDCs

CD14+ monocytes were isolated from peripheral blood mononuclear cells using anti-human CD14 MACS beads (130-050-201; Miltenyi Biotec, Cologne, Germany), as previously described. All monocyte preparations were analyzed for activation based on cluster formation and spontaneous tumor necrosis factor-α release, and pre-activated cells were excluded from our experiments. Monocytes then were cultured in RPMI1640 medium supplemented with 10% human AB serum (35060CI; Corning, Manassas, VA), 100 U/L penicillin, 100 μg/L streptomycin (15140-122; Gibco, Waltham, MA), 50 μg/mL gentamycin (IB2030; IBI Scientific, Peosta, IA), 5 mmol/L HEPES (SH3023710), and 2 mmol/L L-glutamine (SH30034.1) (both Hyclone, Logan, UT), 25 ng/mL recombinant human (rh) granulocyte-macrophage colony-stimulating factor (GM-CSF) (215-GM-050), and 7 ng/mL rhIL-4 (204-IL-050) (both R&D Systems, Minneapolis, MN). After 5–6 days in culture, DCs were harvested by vigorous pipetting.

Gastric Epithelial Spheroid–DC Co-cultures

To perform fluorescence imaging, co-cultures were set up using mCherry-expressing spheroid cultures and MoDCs stained with CellTracker Green (C7025) or CellTracker DeepRed (C34565). To establish DC-spheroid co-cultures, spheroid culture medium was replaced with antibiotic-free DC culture medium (RPMI1640 with 10% human AB serum). For each 35-mm MatTek (Ashland, MA) culture plate containing approximately 80 spheroids, we then added 5 × 105 immature MoDCs directly to the medium, overlaying the spheroids in the Matrigel.

Chemokine Detection

To analyze chemokine secretion by gastric spheroids, supernatants were collected from 24-well plates 48 hours after medium change. Supernatants then were analyzed using the Proteome Profiler Human Chemokine Array Kit (ARY017; R&D Systems), which detects 31 different chemokines with specific monoclonal antibodies. Arrays were imaged on a FluorChem R System (Proteinsimple, San Jose, CA), and relative pixel densities of duplicate spots were measured using ImageJ software (National Institutes of Health, Bethesda, MD). Relative chemokine expression was calculated by normalizing results from spheroid supernatants to results obtained with spheroid growth medium alone.

Chemotaxis Assays

The chemotactic activity of MoDCs to spheroid supernatants and recombinant chemokines was measured in a Transwell migration assay using the 96-well ChemoTX system with a 5-μm polycarbonate membrane (106-5; Neuro Probe, Inc, Gaithersburg, MD). MoDCs (1–2 × 105) were incubated in triplicate for 4 hours at 37°C, and viable migrated cells were counted using a CellTiter-Glo Luminescent Cell Viability Assay (G7570; Promega, Madison, WI). Migration toward the following recombinant human chemokines was assessed: CXCL1 (300-11), CXCL5 (300-22), CXCL16 (300-55), CCL20 (300-29A, all from PeproTech, Rocky Hill, NJ), CXCL8 (208-IL; R&D Systems), and CXCL17 (NBPS-35133; Novus Biologicals, Littleton, CO). The chemotactic index for recombinant chemokines was calculated by dividing the number of migrated cells for the experimental condition by the number of migrated cells in medium alone. Data from gastric spheroid supernatants were additionally normalized to the number of epithelial cells present in the cultures, as determined by cell counting at the end of the experiment.

H pylori Infection and TLR Agonist Treatment of Gastric Epithelial Spheroids

To establish H pylori infection of the spheroids, we plated H pylori strain 60190 wild type, a CagA-deficient mutant of strain 60190 (kind gifts from Dr G. Perez-Perez) or the GFP-expressing strain M6 (a kind gift from Dr John Y. Kao) under microaerophilic conditions on Brucella agar, 5% horse blood (221261; Becton Dickinson) for 3–4 days. Spheroids for injection experiments were seeded on 35-mm MatTek dishes for 8–12 days, with the Matrigel applied as a thin streak to facilitate injection of multiple spheroids. Before injections, gastric epithelial spheroids were overlaid with antibiotic-free medium (3 media changes) for 3 days to prevent bacterial death. We selected spheroids with a diameter of 200–800 μm and used a GeneSearch Embryo Cradle micromanipulator (GeneSearch, Inc, Bozeman, MT) and a 2-μL syringe (Hamilton Co, Reno, NV) for microinjection. Injection needles with beveled tips were pulled from glass capillaries to a tip size of 21-23 μm. We injected 20–80 organoids per culture with 0.2 μL of H pylori suspension (2 × 105 bacteria) in Brucella broth, corresponding to a multiplicity of infection of 10–50, or with 1 μg/mL of Salmonella flagellin. Control spheroids were injected with Brucella broth alone. In some experiments, 0.2 μL of 25 μmol/L 4 kilodaltons FITC dextran (46944; Sigma-Aldrich, St. Louis, MO) was added to the injection media as a fluorescent tracer. Successful H pylori infection was confirmed by re-culturing bacteria recovered from trypsinized organoids on Brucella agar plates and performing colony counts. MoDCs were added to H pylori–infected spheroids after 2–3 hours. To evaluate gastric epithelial cell responses to TLR agonist treatment, spheroids were replated on collagen-coated 24-well plates (08-772-71, Biocoat; Corning, Tewksbury, MA). Established monolayers were stimulated for 6 hours with the human TLR1–9 agonist kit containing the following: TLR1/2 agonist: palmitoyl-3-cystein (Pam3CSK4); TLR2 agonist: heat-killed Listeria monocytogenes (HKLM); TLR3 agonist: polyinosine-polycytidylic acid (poly(I:C)); TLR3 agonist: poly(I:C)-low molecular weight (LMW); TLR4 agonist: lipopolysaccharide E coli K12; TLR5 agonist: flagellin Salmonella typhimurium; TLR6/2 agonist: fibroblast-stimulating factor 1 (FSL1); TLR7 agonist: imiquimod; TLR8 agonist: ssRNA40/LyoVec; and TLR9 agonist: oligodeoxynucleotide (ODN)2006 (tlrl-kit1hw; InVivoGen, San Diego, CA), and then were analyzed for IL8/CXCL8 gene expression using quantitative RT-PCR, as previously described.

Oxygen Concentrations in Gastric Spheroids

Mock-injected spheroids were compared with spheroids infected with H pylori bacteria for 3 hours. Oxygen concentration profiles in the organoids were measured with an OX-25 Clark-type oxygen microelectrode with a 25-μm sensor tip diameter (Unisense A/S, Aarhus, Denmark). The microelectrode was connected to a Unisense microsensor multimeter and millivolt data were logged onto a laptop computer with Sensor Trace Pro software (Unisense). A motor-controlled micromanipulator (model MM33; Unisense) was used to move the oxygen electrode during profile measurements. A 2-point calibration was performed for O2 sensors using buffer sparged with air and medium sparged with pure N2(g) for at least 30 minutes. An S80 APO stereomicroscope (Leica, Wetzlar, Germany) was used to visualize the microsensor tip and confirm placement on the organoid surface before profiling.

RT2 Profiler PCR Array

To analyze the expression of chemokines, chemokine receptors, and other related genes, we used the Human Chemokines & Receptors RT2 Profiler PCR array (PAHS-022Z; Qiagen, Valencia, CA), following the manufacturer’s instructions. RNA from H pylori–infected (strain 60190) or control cultures was isolated using the Direct-zol Mini-RNA prep kit (R2052; Zymo Research, Irvine, CA), and 0.15–0.5 μg of RNA was converted into complementary DNA. Arrays were run on a Lightcycler 96 (Roche, Penzberg, Germany) and were analyzed using the Qiagen Data Analysis Center online tool (www.qiagen.com), with a cycle threshold (Ct) cut-off of 35 and β2-microglobulin and RPLP0 used as housekeeping genes.

Immunofluorescence Analysis of Tissue Sections

Cryosections or a paraffin-embedded gastric tissue microarray (kind gift from Dr Paul Harris, Pontifical Catholic University of Chile, Santiago, Chile) prepared from human gastric tissue were labeled with anti–HLA-DR (cryosections: L243, BD 340689; tissue array: LN-3, ab166777; Abcam, Cambridge, UK) to stain DCs and with anticytokeratin FITC (cryosections: CAM5.2, BD 347653; tissue array: C11; 4545; Cell Signaling Technology, Danvers, MA) to stain epithelial cells, as previously described.4, 7 Cell nuclei were stained with 4′,6-diamidino-2-phenylindole. Samples were imaged on a Nikon Eclipse T2000-U fluorescent microscope (Nikon, Melville, NY) equipped with a CoolSnap ES digital camera and NIS Elements BR2.30 (Nikon). ImageJ V1.48 was used to measure epithelial cell length and count DCs that were in direct contact with the epithelium.

Microscopy and Image Analysis

Epifluorescence and phase-contrast images were acquired on a Nikon Eclipse T2000-U fluorescent microscope equipped with a CoolSnap ES digital camera and NIS Elements BR2.30 software or on a Life Technologies (Carlsbad, CA) EVOS FL Auto system equipped with an onstage incubator. For confocal image analysis, we fixed the co-cultures with Cytofix Reagent (554655; BD Biosciences). Samples were imaged on an inverted SP5 Confocal Scanning Laser Microscope (Leica) with 405-, 488-, 561-, and 633-nm laser excitation lines using a 20× objective or a 63× water immersion objective with Immersol (W 2010; Zeiss, Oberkochen, Germany). We recorded Z-stacks of 2–11 randomly selected organoids with intact morphology for each experiment and experimental conditions. To quantitate MoDC recruitment to the epithelial interface, we used ImageJ version 1.48V, with the LOCI plugin to allow access to Leica .lif files. For analysis of DC recruitment, area and circumference of the spheroid were measured, and pixels positive for the DC dye CellTracker DeepRed (pseudocolored blue) were counted automatically on thresholded images. DCs that were within 75 μm of the basolateral side of the epithelium were considered epithelial cell–associated DCs. To determine MoDC uptake of H pylori bacteria or epithelial cell–derived material, DCs with fluorescent inclusions were counted manually on individual confocal slices using the Cell Counter plugin, and the total number of DCs in an image stack was determined by counting the cells on a maximum Z-projection of the image stack.

Live Imaging Analysis and Particle Tracking

Live imaging was performed using a Life Technologies EVOS FL Auto system equipped with an onstage incubator. Fluorescence and phase-contrast images were collected at 10-minute intervals over 48 hours using a 10× objective. The migration of cells toward gastric spheroids was analyzed using the Particle Tracking feature in Imaris version 8.4 (Bitplane, Zurich, Switzerland).

Gene Expression Analysis of Gastric Biopsy Specimens

Gene expression of relevant chemokines in gastric antral and corpus biopsy specimens of non–H pylori–infected patients and H pylori–infected patients with and without gastric atrophy (n = 3 each) was analyzed using the Gene Expression Omnibus reference series GSE27411 and data set records GDS5411 (body) and GDS 5338 (antrum).

Data and Statistical Analysis

All authors had access to the data and have reviewed and approved the final manuscript. Data were analyzed using GraphPad Prism 7.02 (San Diego, CA). Results are presented as means ± SD. Differences between values were analyzed for statistical significance by the 2-tailed Student t test, 1- or 2-way analysis of variance, or the Kruskal–Wallis test with appropriate post hoc tests. Differences were considered significant at P < .05.