Bacterial infection can trigger the development of functional GI disease. Here, we investigate the role of the gut-brain axis in gastric dysfunction during and after chronic H. pylori infection. Control and chronically H. pylori-infected Balb/c mice were studied before and 2 mo after bacterial eradication. Gastric motility and emptying were investigated using videofluoroscopy image analysis. Gastric mechanical viscerosensitivity was assessed by cardioautonomic responses to distension. Feeding patterns were recorded by a computer-assisted system. Plasma leptin, ghrelin, and CCK levels were measured using ELISA. IL-1β, TNF-α, proopiomelanocortin (POMC), and neuropeptide Y mRNAs were assessed by in situ hybridizations on frozen brain sections. Gastric inflammation was assessed by histology and immunohistochemistry. As shown previously, H. pylori-infected mice ate more frequently than controls but consumed less food per bout, maintaining normal body weight. Abnormal feeding behavior was accompanied by elevated plasma ghrelin and postprandial CCK, higher TNF-α (median eminence), and lower POMC (arcuate nucleus) mRNA. Infected mice displayed delayed gastric emptying and visceral hypersensitivity. Eradication therapy normalized gastric emptying and improved gastric sensitivity but had no effect on eating behavior. This was accompanied by persistently increased TNF-α in the brain and gastric CD3+ T-cell counts. In conclusion, chronic H. pylori infection in mice alters gastric emptying and mechanosensitivity, which improve after bacterial eradication. A feeding pattern reminiscent of early satiety persists after H. pylori eradication and is accompanied by increased TNF-α in the brain. The results support a role for altered gut-brain pathways in the maintenance of postinfective gut dysfunction.

  • inflammation
  • gut-brain axis
  • gastric emptying
  • visceral sensitivity

functional gi disorders such as irritable bowel disease (IBS) and functional dyspepsia (FD) have traditionally been viewed as psychosomatic. The concept that previous infection can trigger the development of functional GI disease is now accepted. A large cohort study identified bacterial gastroenteritis as the most significant risk factor identified to date for the development of IBS (24). IBS symptoms have been reported to develop in a significant proportion of subjects with documented Campylobacter, Salmonella, E. coli, or Shigella infection (12, 32, 37). A similar case has been also made for functional dyspepsia (18, 33). These studies were restricted to acute enteric infections.

Accumulating evidence suggests that changes in the bacterial content in the gut can affect the central nervous system (CNS) (10, 13). Most of these studies have dealt with the effect of acute infection and early CNS changes that occurred before the onset of the inflammatory response to the infection. The effect of chronic GI infection and inflammation on the CNS has not yet been investigated. Previous studies suggested that vagal sensory afferents are responsible for the rapid changes that occur in the brain within hours of a GI infection (13). However, multiple pathways may be involved in the communication between the gut and the CNS, in particular, during chronic infection when proinflammatory cytokines may relay peripheral immune signals to the CNS via sensory nerves, circumventricular organs, and blood-brain barrier cells (8).

To study the effect of bacterial infection on GI function and on feeding behavior, we have developed a mouse model of chronic H. pylori gastritis. We have previously shown that chronic H. pylori infection alters gastric neuromuscular function in vitro, including impaired acetylcholine release, greater antral muscle relaxation upon electric field stimulation, and increased density of SP and CGRP-containing nerves in the stomach and spinal cord (4). Furthermore, chronically infected mice display delayed gastric emptying and altered feeding behavior, which is suggestive of early satiety (36). In this study, we have investigated the putative mechanisms underlying altered feeding behavior and gut dysfunction operating within the GI tract and the CNS.



H. pylori (Sydney strain) was cultured for 24 h in brain-heart infusion media (BHI Oxoid, Basingstoke, UK) containing 0.3% yeast (Difco Laboratories, Detroit, MI) and 10% bovine serum (Gibco BRL, Rockville, MD) supplemented with 0.4% Campylobacter selective complement (Skirrow supplement SR69; Oxoid), under microaerophilic atmosphere at 37°C with gentle shaking.


The experimental protocol was approved by the Animal Research Ethics Board of McMaster University. Specific pathogen-free female Balb/c mice (Harlan, Montreal, Quebec, Canada) 6–8 wk old were gavaged with H. pylori at a dose of 109 bacteria per mouse on three separate occasions, 2 days apart, as described previously (4). Mice were killed at the termination of the experiments, and stomach, blood samples, and brains were taken.

Experimental design.

Experiments were performed in mice infected with H. pylori for 6–8 mo and in time-matched controls. A group of mice received eradication treatment by intragastric gavage consisting of pantoprazole (0.62 mg/kg), amoxicillin (15.4 mg/kg), and clarithromycin (7.7 mg/kg) twice daily for 10 days, while others were gavaged with normal saline. Experiments were repeated 2 mo after eradication treatment. This time point was chosen as previous work in a model of postinfective IBS using Trichinella spiralis infection showed complete normalization of intestinal function within 6 wk after eviction of the parasite (5).

H. pylori, gastritis, and CD3+ cells in the stomach.

Paraffin sections of gastric tissue (Swiss rolls involving the whole stomach from pylorus to fundus, two sections per mouse) were stained with hematoxylin-and-eosin to grade gastritis and with Warthin-Starri staining for bacterial colonization. Bacterial eradication was verified by examining the whole gastric section. If a single H. pylori-like organism was noted, then the mouse was deemed infected, and the functional data were excluded from the study. Eradication rate was close to 80%.

Polymorphonuclear (PMN) and mononuclear (MN) cell infiltration was graded on a scale 0–3, as described previously (4).

Immunohistochemistry for CD3+ T cells was performed on paraffin-embedded tissues, as described previously (5). As primary antibodies, we used rabbit anti-mouse CD3 (Dako AS, Denmark, dilution 1:300) followed by biotinylated swine anti-rabbit antibodies (Dako AS, Denmark, dilution 1:300) and streptavidine peroxidase conjugate (Dako AS, Denmark, dilution 1:600). CD3+ cells were graded as follows: 0, no cell observed per visual field (×250 magnification); 1, less than 5 cells, 2- 6–10 cells; and 3, more than 10 cells. To correlate gastric emptying with CD3+ cell counts, the total number of cells counted per field was used.

Gastroduodenal motility.

Mice had free access to food and water until the beginning of the experiment. Gastroduodenal motility was assessed using video image analysis as described (5). Briefly, mice were gavaged with 0.2 ml of barium and fluoroscoped for 4 min. Video images were analyzed using NIH Image 1.62 software (http://rsb.info.nih.gov/nih-image) with custom-made routines. Spatio-temporal maps were created and the frequency, propagation velocity of contractions, and incidence of retroperistalsis were assessed. Gastric emptying was assessed in single images as follows: border of stomach was manually outlined, and gastric area and mean optical density were measured. The amount of barium in the stomach was assessed by multiplying the gastric area by the optical density. Gastric emptying was expressed as a percentage of barium expelled from the stomach in 4 min.

Cardio-autonomic response to gastric distension.

Experiments were performed in overnight fasted mice anesthetized with a ketamine (90 mg/kg) and xylazine (20 mg/kg) mixture administered intraperitoneally. A balloon (15 × 10 mm) was inserted through an incision in the duodenum into the stomach. Recording electrodes were inserted subcutaneously to front and back legs. The recording started after a 10-min stabilization period. ECG was recorded by personal computer (Acquire software ver. 5.1 by A. Bayati) during baseline period (1 min) and then distension (0.2–1.0 ml) was applied for 1 min followed by 1-min recovery period. The next level of distension was applied after a 15-min rest period. Change in cardiac frequency to distension was analyzed using GrafView 5.1 (written by A. Bayati) and expressed as a percentage of resting heart rate.

Although pseudoaffective response to gut distension is commonly assessed using chronically implanted EMG electrodes, we have opted for acute experiments to avoid any additional inflammatory stimuli prior to our experiments.

Feeding patterns.

Twenty-four-hour feeding patterns were assessed in mice placed in separate cages. Day/night timing was adjusted with lights turning on at 7:30 A.M. and turning off at 7:30 P.M. Food pellets were fastened on a feeding tray positioned 5 cm above the bottom of the cage. The feeding tray was connected to the strain gauge, and its weight was continuously recorded by a computer. The beginning of each eating bout was marked by an increase in weight when the mouse pushed the feeding tray down. Data acquisition and analysis were performed using Acquire 5.1 and GraphView 5.1 software. An eating bout was defined as an episode of food consumption lasting more than 20 s; two bouts were considered to be independent from each other if the interval of quiescence was longer than 5 min, as defined previously (30).

CCK+ cells in the stomach and duodenum.

Immunohistochemistry for CCK cells was performed on Bouin's fixed, paraffin-embedded tissues. Antigen retrieval was performed using trypsin solution at 37°C for 10 min. Endogenous peroxidase was blocked by peroxidase blocking reagent (Dako North America, Carpinteria, CA). Nonspecific binding was blocked by 5% normal goat serum for 1 h. Sections were then incubated with rabbit anti-CCK primary antibody (dilution 1:800; Chemicon AB, Billerica, MA) overnight at 4°C followed by 30-min incubation with DakoCytomation EnVision+ System (Dako North America). The slides were developed in diaminobenzidine peroxidase substrate (Sigma-Aldrich, St. Louis, MO), and tissues were counterstained with Mayer's hematoxylin. CCK+ cells were counted per visual field (×250 magnification).

Ghrelin and leptin in the plasma.

Plasma levels of leptin and ghrelin were measured using a commercially available ELISA kit (Diagnostic Systems Laboratories, Webster, TX) and EIA kit (ALPCO Diagnostics, Salem NH, USA), respectively.

CCK plasma levels.

Plasma CCK levels were measured after intragastric gavage of 40% a peptone meal (Becton, Dickinson, Sparks Glencoe, MD) in overnight fasted mice using a commercially available EIA kit (Phoenix Pharmaceuticals, Burlingame, CA). Mice were killed 8 min after meal administration by intracardiac puncture, and blood was collected using EDTA and aprotinin as described (25).

In situ hybridization in CNS.

Levels of IL-1β, TNF-α, proopiomelanocortin, and neuropeptide Y (NPY) mRNAs were assessed by in situ hybridizations using 35S-labeled RNA probes on frozen brain sections, as described previously (9). Brains were rapidly frozen in 2-methylbutane at −60°C, and stored at −70°C. Twelve-micrometer-thick coronal sections were fixed with 4% formaldehyde, acetylated with 0.25% acetic anhydride in 0.1 M triethanolamine-HCl, pH 8.0, dehydrated, and delipidated with chloroform. Antisense TNF-α ribonucleotide probe (gift of Dr. Serge Rivest, Laval University), IL-1β ribonucleotide probe (gift of Dr. Ron Hart, Rutgers University, Newark), proopiomelanocortin (POMC) ribonucleotide probe (gift of Dr. Jim Eberwine, University of Pennsylvania, Philadelphia), and NPY ribonucleotide probe were transcribed from linearized plasmid using the Riboprobe System (Promega Biotech, Burlington, ON, Canada) with α-35S-UTP (specific activity >1,000 Ci/mmol; Perkin-Elmer, Boston, MA) and T7 and SP6 polymerases respectively. The NPY riboprobe was generated in Dr. Foster's laboratory. NPY primers, forward 5′ggactgaccctcgctctat3′ and reverse 5′gatgagggtggaaacttgga3′ were designed using Primer 3 to mouse NPY RNA, gene bank accession no. NM_023456. Specificity of primers to mouse NPY mRNA was confirmed using BLAST. PCR generated cDNA (561 bp) was inserted in the pGEM T-easy expression vector (Promega). The NPY sequence was confirmed by DNA sequencing (MOBIX Laboratory, McMaster University). Radiolabeled probes were diluted in a hybridization buffer and applied to brain sections (∼500,000 CPM/section). Slides were incubated overnight at 55°C in a humidified chamber. To reduce nonspecific binding, slides were washed in 20 μg/ml RNase solution for 30 min at room temperature, followed by 1 h each in 2XSSC at 50°C, 0.2XSSC at 55° and 60°C. Slides were dehydrated and air-dried for autoradiography. Slides and 14C plastic standards were placed in x-ray cassettes, apposed to film (BioMax MR; Eastman Kodak, Rochester, NY) for 5 days and developed (X-OMAT; Kodak). The images were digitized using Qicam camera (Quorum Technologies, Guelph, ON, Canada) and Image software package (http://rsb.info.nih.gov/nih-image). Light transmittance through the film was measured by outlining the structure on the monitor. Transmittance was converted to radioactivity levels using the Rodbard curve applied to the standards. As mRNA signal for TNF-α and IL-1β was homogeneous, light transmittance was measured and dpm calculated. For POMC and NPY regions, where cell density is not homogeneous, the density slice feature was employed to measure both the light transmittance and the area of mRNA signal. Calculated DPM were multiplied by area to generate an integrated density measurement.


All data are presented as means ± SD. Statistical testing was done using ANOVA test for multiple comparisons with Tukey test, and Student's tests for unpaired or paired data as appropriate.

A P value lower than 0.05 was considered as statistically significant.


Gastritis and CD3+ cell counts.

H. pylori infection caused a significant increase in the inflammatory infiltrate, mainly in the submucosa. The corpus MN cell score increased from 0.5 ± 0.6 to 2.3 ± 0.8, and the PMN score increased from 0.3 ± 0.5 to 1.6 ± 0.9 (both P < 0.05 vs. control, Fig. 1). Successful H. pylori eradication was verified in Warthin-Starri-stained sections (Swiss rolls involving the whole stomach, two sections per mouse). Bacterial eradication was accompanied by a decrease in the MN scores to 1.5 ± 0.6 (P < 0.05 vs. infected) and a trend for a decrease in the PMN score to 1.1 ± 1.6 (P = 0.09 vs. infected). CD3+ cell counts increased from 1.2 ± 0.5 to 2.5 ± 0.6 (P < 0.05 vs. control, Fig. 1) during H. pylori infection and remained elevated at 2 mo after bacterial eradication (2.3 ± 1.0, P < 0.05 vs. control).

Fig. 1.

Mononuclear cell infiltration (left) and CD3+ cell counts (right). Bacterial eradication decreased chronic inflammatory infiltrate at 2 mo posteradication but did not affect the CD3+ cell counts. ANOVA, P < 0.01. Tukey, *P <0.01 vs. control; #P < 0.05 vs. H. pylori infected.

Gastroduodenal motility.

H. pylori infection was accompanied by a larger gastric area of 144% and 174% at 0 and 4 min, respectively, compared with controls (Fig. 2). During the 4-min period, control mice emptied 53.3 ± 15.0% of barium contrast from the stomach, while H. pylori-infected mice emptied only 30.9 ± 17.3%, P < 0.05 (Fig. 2). Two months posteradication, gastric area normalized as well as gastric emptying, which reached a value of 48.8 ± 13.0% (P = 0.7 vs. controls).

Fig. 2.

Representative images of stomach in control (left) and H. pylori infected (right) mice immediately after the gavage of barium (upper) and after 4 min (lower). Gastric area was larger both at 0 min (open bar) and 4 min (gray bar) in H. pylori-infected mice (n = 9) compared with controls (n = 9). Gastric emptying was delayed in infected mice but fully normalized 2 mo posteradication (ANOVA P < 0.01. Tukey, *P < 0.05 vs. control; #P < 0.05 vs. H. pylori infected).

The frequency of gastric contractions and their propagation velocity were similar in infected and control mice. Similarly, no differences in duodenal motor patterns were detected in infected vs. control mice (Table 1).

View this table:
Table 1.

Gastroduodenal motility in control and H. pylori-infected mice

Cardioautonomic response to gastric distension.

Resting cardiac frequency in both control and infected mice ranged form 280 to 400 beats/min, likely an effect of ketamine/xylazine anesthesia. To compensate for interindividual variability, distension data were normalized and expressed as a percentage of resting heart rate. In control mice, the heart rate increased progressively with increasing gastric distension to a volume of 0.8 ml and then decreased to baseline values (Fig. 3, Table 2). In H. pylori-infected mice, the volume response curve appeared shifted to the left with a maximum response at 0.2 ml (ANOVA P < 0.01, Tukey P < 0.05 vs. both uninfected and eradicated). Two months posteradication the cardioautonomic response improved, but the maximum response was between 0.4 and 0.6 ml.

Fig. 3.

Heart rate (HR) response to increasing gastric distension. In controls (◊, n = 11), HR progressively increased until 0.8 ml, then dropped to baseline. In H. pylori-infected mice (•, n = 8), the HR reached maximum at lower volumes. Bacterial eradication (○, n = 8) improved but did not normalize gastric sensitivity. (ANOVA, P < 0.01. Tukey, *P < 0.05 H. pylori vs. both uninfected and eradicated; #P < 0.05 eradicated vs. control).

View this table:
Table 2.

Heart rate response to gastric distension in control and H. pylori-infected mice

Twenty-four-hour feeding behavior.

In control mice, there were 13 ± 3.5 eating bouts during 24 h, and the mean duration of the bout was 11.5 ± 4.0 min (Fig. 4). H. pylori-infected mice ate more frequently resulting in 19.5 ± 4.8 bouts/24 h (ANOVA, P < 0.01, Tukey, P < 0.05 vs. control) with a mean duration of bout of 9.0 ± 2.6 min. Infected mice consumed less food per bout than controls (0.20 ± 0.11 g compared with 0.32 ± 0.14 g, ANOVA, P < 0.01; Tukey, P < 0.05). However, there was no difference in total consumption of food during 24 h between H. pylori infected and control mice, 5.2 ± 1.1 and 5.3 ± 1.3 g, respectively. Accordingly, body weight of infected and control mice was similar, 23.6 ± 2.4 and 24.4 ± 3.2 g, respectively. Bacterial eradication did not affect the altered eating pattern as the previously infected mice had 19.4 ± 7.8 bouts/24 h with a mean consumption of 0.17 ± 0.09 g per bout (both P < 0.05 vs. control).

Fig. 4.

Twenty-four-hour feeding pattern in control (top) and in H. pylori infected mouse (middle). Horizontal bars below each tracing depict single eating bouts. Infected mice (n = 8) ate more frequently and smaller amounts of food then controls (n = 8). Bacterial eradication did not modify the eating pattern (n = 8). (ANOVA, P < 0.01. Tukey *P < 0.01 vs. control).

Leptin, ghrelin, and CCK levels.

Leptin levels were similar in control and H. pylori-infected mice (4.8 ± 1.4 and 6.1 ± 2.0 ng/ml, P = 0.11). Ghrelin levels were higher in H. pylori-infected mice compared with controls (465.3 ± 418.8 vs. 100.9 ± 87.4 pg/ml, ANOVA P < 0.01, Tukey P < 0.05, Fig. 5). Ghrelin levels normalized 2 mo posteradication reaching a value of 103 ± 137.6 pg/ml.

Fig. 5.

Plasma ghrelin and postprandial CCK levels were elevated in H. pylori-infected mice. After bacterial eradication, ghrelin levels normalized and CCK levels were similar to controls. *P < 0.01 vs. control; #P < 0.01 vs. H. pylori infected; **P <0.05 vs. control.

CCK-containing cells counts were similar in control and H. pylori-infected mice (1.9 ± 0.5 and 2.7 ± 2.0 per high-power view, P = 0.2). However, postprandial CCK levels after peptone meal were elevated in H. pylori-infected mice compared with controls (0.27 ± 0.10 vs. 0.19 ± 0.07 ng/ml, ANOVA P < 0.05, P < 0.05 vs. control, fig. 5). CCK levels were similar in previously infected mice compared with controls 2 mo posteradication reaching 0.24 ± 0.05 ng/ml.

Effect of H. pylori on peptides and cytokine mRNA expression in CNS areas controlling feeding behavior.

POMC mRNA in the arcuate nucleus was 187.1 ± 98.8 dpm·mm2 in control mice and 103.2 ± 33.9 dpm·mm2 in H. pylori-infected mice (Fig. 6, ANOVA P < 0.05, Tukey P < 0.05 vs. control). POMC levels were similar in previously infected mice 2 mo after H. pylori eradication reaching values of 128.9 ± 61.3 dpm·mm2. There was no difference in NPY mRNA expression between control and H. pylori-infected mice (data not shown).

Fig. 6.

POMC mRNA in the arcuate nucleus (top) was lower, while TNF-α mRNA in the median eminence (bottom) was higher in H. pylori-infected mice. After bacterial eradication levels of POMC were similar to controls, while TNF-α remained upregulated. (**P <0.01 vs. control, *P <0.05 vs. control). Horizontal bar represents 250 μm.

TNF-α mRNA in the median eminence was 418 ± 116 dpm in controls and increased to 802 ± 438 dpm in H. pylori-infected mice (Fig. 6, ANOVA P < 0.01, Tukey P < 0.05,). Two months after eradication, TNF-α values remained elevated at 786 ± 404 dpm (P < 0.05 vs. control). Similarly, TNF-α mRNA in the arcuate nucleus was higher in H. pylori-infected mice compared with controls (938 ± 315 vs. 404 ± 79 dpm, ANOVA, P = 0.02; Tukey, P < 0.05) but similar at 2 mo posteradication (716 ± 444 dpm). In contrast, IL-1β mRNA expression in the median eminence was low, and no differences were observed between the groups (data not shown).


The results of this study demonstrate that chronic noninvasive bacterial infection of the stomach alters feeding behavior, which persists for at least 2 mo following eradication of the infection. The observed change in feeding behavior is characterized by more frequent and smaller-volume meals, with no reduction in the total intake or body weight. The altered feeding behavior is accompanied by altered mRNA expression of TNF-α and POMC in the median eminence and the arcuate nucleus, brain areas involved in the control of feeding behavior. We also show that delayed gastric emptying and altered gastric mechanosensitivity improve after bacterial eradication. Our results have two important implications: 1) multiple mechanisms, including persistent immune gut activation and altered brain neurochemistry, may be responsible for the maintenance of abnormal feeding behavior; and 2) the persistence of these changes provides an explanation for the lack of apparent symptom resolution following eradication of H. pylori.

Delayed gastric emptying is often found in patients with FD, and several studies have suggested that H. pylori eradication improves gastric emptying in patients with FD, likely due to improvement in antral myoelectrical activity (15, 19, 34). In our study, control mice emptied 50% of their contents within 4 min. These values are similar to those obtained in a recent study, which assessed gastric emptying in mice using both scintigraphy and phenol red techniques (3). H. pylori-infected mice emptied 42% less gastric contents than uninfected controls. Gastric emptying fully normalized 2 mo after H. pylori eradication. We have not found any difference in frequency or velocity of gastric contractions between infected and control mice; neither have we found any abnormality in duodenal motor patterns. The explanation for the observed delayed gastric emptying may relate to a lower force of gastric contractions during H. pylori infection. This is supported by our previous results showing decreased release of acetylcholine and increased density of vasoactive intestinal peptide-containing nerves in gastric strips of H. pylori-infected mice (4). As gastric emptying normalized after H. pylori eradication, we can conclude that it is not responsible for the maintenance of abnormal feeding behavior observed at 2 mo posteradication.

H. pylori-infected mice ate more frequently but consumed smaller amounts of food per bout than controls, a pattern reminiscent of early satiety. The control of feeding behavior is a complex process involving both peripheral and central mechanisms (38). These include short-acting gastrointestinal signals, including gut hormones such as CCK and gastric distension, which relay sense of fullness resulting in satiation. Central integration of peripheral signals converging in hypothalamus involves neurons located in the arcuate nucleus and containing orexigenic peptide NPY and anorexigenic peptide POMC. Circulating hormones, such as leptin and ghrelin, have direct access to the arcuate nucleus, while others like CCK influence hypothalamus indirectly via neuronal pathways.

In the present study, we found that H. pylori-infected mice perceived gastric distension differently than controls with the maximal heart rate response shifted to the left, at lower gastric volumes. This may be due to a lower activation threshold of gastric mechano/nociceptors, likely altered by chronic gastritis as shown in other models of inflammation-induced visceral hyperalgesia (11). Bacterial eradication improved but did not normalize mechanosensitivity up to 2 mo posteradication. This is in accordance with our previous in vitro results showing persistently elevated density of gastric SP- and CGRP-containing fibers after bacterial eradication (4). The increase in heart rate in response to nociceptive stimulation has been shown in several murine and rat models (2, 20, 22, 25). A differential activation pattern of hindbrain excitatory and inhibitory circuits involved in the control of the cardiovascular system has been described during low- and high-grade distension in rats (28). This may explain our results showing an increase in heart rate at low inflation volumes and a decrease to baseline at submaximal volumes in infected mice. Although the differences observed in our study were small, statistical significance suggests that altered viscerosensitivy induced by H. pylori may be a contributing trigger for early satiety in the model.

Clinical trials investigating the effect of H. pylori infection on leptin and ghrelin levels are controversial. However, a recent study has shown an inverse correlation between the severity of gastritis and circulating levels of ghrelin, which gradually decrease after bacterial eradication (23). Others have shown that subjects with H. pylori infection have blunted sympathetic reactivity and an exacerbated vagal response to feeding (16). In our study, we found that plasma ghrelin levels were higher in infected mice and normalized after bacterial eradication, while leptin levels were unchanged by H. pylori infection. Although no difference was found in the number of CCK-containing cells, a peptone meal stimulated greater release of CCK in infected mice. It should be noted that the CCK antibody of EIA kit used in our study might cross-react with gastrin. However, in contrast to humans, no changes in basal or food stimulated gastrin release have been observed in Helicobacter-infected rodents (7). Thus, cross-reactive hypergastrinemia is an unlikely explanation for the high CCK values observed in this study. We also found decreased POMC mRNA expression in the arcuate nucleus in infected mice, which may be a consequence of altered peripheral feeding control mechanisms.

Immune activation may be a key factor in the maintenance of postinfective gut dysfunction. Clinical studies suggest that decreased severity of gastritis, and not H. pylori eradication per se, is associated with improvement of dyspeptic symptoms (35). In our study, chronic gastritis improved after H. pylori eradication, but did not resolve, mainly due to persistently increased CD3+ lymphocyte counts, a finding that has also been reported in humans (14). We hypothesize that the residual chronic gut inflammation may result from increased H. pylori-induced intestinal permeability, which persists after bacterial eradication (36). This may lead to chronic abnormal exposure of lamina propria cells to luminal antigens.

We found increased TNF-α mRNA expression in the median eminence and the arcuate nucleus of mice with chronic H. pylori infection. The median eminence is a circumventricular organ with increased permeability of blood-brain barrier, in close proximity of, and extensively connected to the arcuate nucleus. Systemic administration of LPS in mice induces upregulation of TNF-α in circumventricular organs, which spreads into hypothalamic nuclei, including the arcuate nucleus (6). Also, increased c-fos mRNA has been detected in the circumventricular organs of rats after intravenous TNF-α challenge (21). The role of proinflammatory cytokines, including TNF-α, in feeding behavior has been reported in both animal models and humans (17, 26, 31). Acute administration of TNF-α in rats has been shown to decrease feeding rates and increase the satiety ratio (31).

The initiation and maintenance of altered feeding behavior in H. pylori-infected mice is likely complex. We propose that the smaller amount of food consumed per eating bout, likely reflects early satiety. This may be due to altered gastric mechanosensitivity and increased postprandial CCK release. We hypothesize that higher plasma ghrelin levels are a compensatory mechanism to maintain body weight, by increasing the frequency of meals. Changes in the central neural circuitry controlling feeding behavior may persist even after peripheral signaling has normalized due to central sensitization. However, gastric mechanosensitivity did not completely return to normal after H. pylori eradication, and we cannot completely rule out its role in the maintenance of altered feeding patterns. However, TNF-α expression in the median eminence remained upregulated 2 mo posteradication. Taken together with recent studies supporting a role of this cytokine in neuronal plasticity (1, 24), our results suggest that persistent abnormal feeding behavior is maintained, at least in part, by persistently upregulated TNF-α in the CNS.

Perspectives and Significance

In this study, we show that a chronic noninvasive GI infection affects not only gut, but CNS function as well. Two months after bacterial eradication, most parameters of gastric function return to control values, but the profile of altered feeding behavior persists. This is paralleled by persistently elevated gastric CD3+ T cells in the stomach and increased expression of TNF-α mRNA in the brain. We propose that a noninvasive bacterium initiates inflammatory and physiological changes in the GI tract leading to altered expression of inflammatory markers in the CNS. After chronic infection, CNS changes may be slow to resolve and may persist after eradication of the triggering GI infection. Our results provide an explanation for the persistence of Hp-induced-altered feeding behavior after eradication of the bacterium, despite marked improvement in gut function.


This study was supported by a grant to S. M. Collins from the Canadian Institutes of Health Research and a fellowship to P. Bercik from Abbot Laboratories, Canada and the Canadian Association of Gastroenterology. P. Bercik and E. F. Verdú hold Internal Career Research Awards from Department of Medicine, McMaster University.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract