Oral Exposure to Diabetes-Promoting Food or Immunomodulators in Neonates Alters Gut Cytokines and Diabetes

Male and female BBdp and control BB rats (BBc) were obtained from the Animal Resources Division of Health Canada, where they are maintained under specific pathogen-free conditions. BBc rats derive from an early subline of animals in this colony that does not spontaneously develop diabetes. Animals were antibody-free with respect to Sendai virus, pneumonia virus of mice, rat corona virus/sialodacryoadenitis virus, Kilham rat virus, Toolan’s H-1 virus, reovirus type 3, and mycoplasma pulmonis. Animals were weaned at 23 days of age, caged in banks of 30 wire-bottom cages, and given free access to food and water. The principles of laboratory animal care as described by the Canadian Council on Animal Care were followed.

The animals were weighed at weekly intervals between the ages of 30 and 100 days. Animals were tested twice weekly for glucose in the urine after 60 days of age (Testape; Lily, Toronto, Ontario, Canada). Animals with a value of 2+ or greater were fasted overnight, and blood glucose was measured the next morning in tail blood using glucose test strips and a glucometer. Diabetes was diagnosed in an animal when fasting blood glucose was >11.1 mmol/l (200 mg/dl). Diabetic animals were killed within 24 h of diagnosis by exsanguination while under 3% halothane in oxygen anesthesia.

Diets were fed in powdered or meal form in open bowls. A standard rodent diet, NIH-07 (NIH), was purchased in meal form from Ziegler Brothers (Gardners, PA). NIH-07 is an open-formula, nonpurified diet that provides dietary amino acids mainly from plants. The NIH-07 diet is composed of 5% dried skim milk, 10% fish meal, 12% soybean meal, 4% alfalfa meal, 3% corn gluten meal, 24.5% ground yellow shelled corn, 23% ground hard winter wheat, 10% wheat middlings, 2% brewer’s dried yeast, 1.5% molasses, 2.5% soybean oil, 0.5% sodium chloride, 1.25% dicalcium phosphate, 0.5% ground limestone, and 0.25% premixes (19).

Wheat gluten (WG) semipurified diets were made up of 22.5% wheat gluten (ICN Biochemicals, Cleveland, OH), 50.2% corn starch, 12.0% sucrose, 5.0% corn oil, 5% alphacel, 3.5% AIN-76 mineral mix (ICN), 1.0% AIN-76A vitamin mix (ICN), 0.2% choline bitartrate, 0.02% dl-methionine, 0.5% l-lysine, and 0.08% l-threonine.

The semipurified hydrolyzed casein (HC) diet consisted of 53.0% corn starch, 12.0% sucrose, 20.0% casein hydrolysate (Champlain Industries, Mississauga, Ontario, Canada [presently RedStar Bioproducts, Tara, Ontario, Canada]), 5% corn oil, 5% alphacel, 3.5% AIN-76 mineral mix, 1.0% AIN-76A vitamin mix, 0.2% choline bitartrate, and 0.3% dl-methionine (19).

To test the possibility that early oral exposure to a diet known to be diabetes-promoting might alter diabetes outcome, we removed BBdp rat pups from their dams and hand-fed them twice daily from 4 to 7 days of age a mixture of Pregestimil (PG) (Mead Johnson, Belleville, Ontario, Canada), a hypoallergenic infant formula diet based on hydrolyzed casein (HC) plus ground NIH-07 diet or plus wheat gluten or PG alone. Pregestimil consists of 13% hydrolyzed casein, 62% carbohydrates (corn starch and corn syrup solids), and modified coconut oil, supplemented with vitamins A, B₁, B₂, B₃, B₆, B₁₂, C, D, E, K; biotin; folic acid; α-pantothenic acid; the amino acids l-tryptophan, l-cystine, and l-tyrosine; and mineral supplements. Because of the labor-intensive nature of these studies, which limited the size of individual trials, three separate experiments were carried out with early NIH-07 feeding and the data were pooled. The total number of animals treated was as follows (neonatal treatment/weaning diet): PG/HC, 37; PG/NIH, 39; and NIH/NIH, 42. In a separate study, two groups of animals were treated as follows: PG/WG, 15; and WG/WG and then weaned at day 23 onto a WG diet, 14. A control group, PG/HC, contained 15 rats.

The pups were fed through Technicon-B rubber tubing attached to a 23-G needle and a 1-ml syringe. The diet mixture was kept at body temperature during feeding. After feeding, the pups were returned to their dams. Feces- and urine-soaked bedding were applied to the pups to reduce maternal cannibalism. The amount of protein in the NIH-07 slurry fed to the pups was equivalent to 1 mg/g body wt, which at the weight of the pups at the time of feeding (∼15 g) meant a dose of 67 mg NIH per rat per feeding. WG was given at a dose of ∼20 mg/0.45 ml to deliver 1 mg protein/g body wt. This was delivered in 0.45 ml of fluid. After the 4 days of hand-feeding, the pups remained with the dams until weaning. The animals were weaned onto one of three diets: two modified, semipurified diets in which HC or WG was the sole source of amino acids or the standard rodent diet, NIH-07. Lipopolysaccharides (LPS) (Escherichia coli serotype 0127:B8; Sigma) was administered to neonates as above at a concentration of 0.75 mg/0.4 ml, and silica (Sigma) was administered concurrently at a concentration of 1.2 mg/0.4 ml.

BBc and BBdp animals that were treated in the first week of life with either oral PG or NIH were fed a diabetes-promoting NIH-07 diet or protective HC diet from weaning. At 60 days of age, the animals were lightly anesthetized and received an injection of autoclaved NIH-07 diet in PG (0.5 mg/ml, equivalent to ∼3 μg of protein from NIH-07 diet in 25 μl) in the left footpad and, as a control, with PG in the right footpad. Footpads were cleaned with an iodine solution. Footpad thickness was measured before the injection for a zero time measurement. The time was noted, and subsequent measures were performed at approximately the same time of day. Footpad thickness was measured with a micrometer (Mitutoyo, Montreal, Canada) after 24 h. All procedures took place between 8:00 a.m. and 12:00 p.m.. After 7 days, the animals were killed, popliteal lymph nodes (PLNs) were excised, the fat was trimmed, and each lymph node was weighed.

After exsanguination under halothane anesthesia or decapitation (pups only), pancreata were dissected free, trimmed of fat, and placed in Bouin’s solution for fixation. A 12-cm piece of small intestine beginning at the stomach was removed and flushed with ice-cold phosphate-buffered saline, and the distal 2 cm was frozen immediately in liquid nitrogen and stored at −80°C for reverse transcriptase–polymerase chain reaction (RT-PCR) analyses.

Hematoxylin and eosin–stained sections (5 μm) were read for insulitis and degree of islet damage using a light microscope (20). Briefly, the overall extent of mononuclear cell infiltration of the islets (insulitis) and damage in a section was graded according to a scale of 1 to 5 as follows: 1, normal section, good number of islets (∼32), no infiltration; 2, mild infiltration (up to 30% of islets infiltrated), still a good number of islets observed (∼30); 3, several islets were present (∼28), but most (∼90%) were infiltrated; 4, few islets remained (∼14), most were infiltrated (∼78%) or very small; and 5, end-stage disease, displaying very few small islets (<5), with little inflammatory infiltrate remaining.

In two separate experiments, gut cytokines were measured in pups fed on day 5 as above, then killed the following morning. Total RNA was isolated from the duodenal section of the small intestine using Trizol Reagent (Life Technologies, Karlsruhe, Germany) following the manufacturer’s instructions. RT-PCR was performed as described elsewhere (21). Primer sequences (β-actin, γ-interferon [IFN-γ], and transforming growth factor-β [TGF-β]) were obtained from Clontech (Palo Alto, CA). PCR conditions were within the semiquantitative range as verified by initially performing serial dilutions of cDNA and varying the number of cycles used for PCR. The resulting cycle numbers were 22 for β-actin and 34 for IFN-γ and TGF-β. In all amplifications, template-negative controls confirmed the absence of contaminating cDNA. Amplified products were visualized in a 1.5% agarose gel containing ethidium bromide. For quantification, the fluorescence intensity of each band was determined as Boehringer light units (BLU) using a Lumi-Imager (Boehringer Mannheim, Mannheim, Germany) and the specific analysis software (LumiAnalyst 3.0). A curve was generated by plotting the amount of cDNA versus the corresponding BLU. Linear regression was performed with this curve, and the mRNA amount was set as the BLU value for 1 μl of cDNA. The correlation between different amounts of cDNA and the corresponding fluorescence intensity ranged between 0.90 and 0.99. The relative amounts of cytokine mRNA were expressed as arbitrary units as the ratio of cytokine to the respective β-actin mRNA.

Survival curves for BBdp rats fed PG between days 4 and 7 and weaned at 23 days onto the semipurified HC diet (PG/HC group) or weaned onto cereal-based NIH-07 (PG/NIH) showed a major and significant difference in diabetes rate by 150 days of age, confirming the important modulatory role of diet after weaning (2) (Fig. 1A). In addition, when the diabetes-inducing diet, NIH-07, was fed in the first week of life, the onset of diabetes was delayed (79 ± 6 vs. 88 ± 7 days) and approximately one-third fewer diabetes cases were observed (16 out of 42 vs. 22 out of 39 [38 vs. 56%]; P = 0.02, Kaplan-Meier, log rank test) (Fig. 1A). There was no effect of early antigen treatment on islet infiltration as measured by the number of islets infiltrated or on overall insulitis rating (Table 1). Animals fed PG in the first week of life and weaned onto an HC diet had low diabetes frequency (2 of 37 [5%]). Fewer islets were infiltrated, two to three times more were normal, and unaffected islets and the overall insulitis rating were lower than in animals weaned onto NIH-07 (Table 1).

Because NIH-07 is a complex mixture of cereal and other proteins, the experiment was repeated with a semipurified diet containing WG as the sole source of protein and being isocaloric and isonitrogenous with the HC diet. As shown in Fig. 1B, this diet was as diabetes-promoting as NIH-07 when fed from weaning. Onset of diabetes was again delayed (82 vs. 91 days), and fewer diabetes cases were observed (PG/WG versus WG/WG, 73 vs. 35%; P = 0.008, Kaplan-Meier log rank test) after neonatal exposure to the diabetes-promoting wheat antigens. Of 15 rats, 1 developed diabetes in the PG/HC group.

Because of the apparent priming effect of early contact with a cereal-based diet, we searched for signs of altered systemic cellular immunity to NIH-07 antigens. We examined the ability of NIH-07 antigens to induce a DTH reaction, which is a classic measure of systemic Th1 cell recruitment and activation. As shown in Fig. 2A, subcutaneous injection of NIH-07 induced substantial footpad swelling within 24 h. Animals with previous neonatal exposure to NIH-07 showed slightly suppressed DTH responses (Fig. 2A), but this was not statistically significant.

Next we analyzed for the popliteal lymph node response 7 days after subcutaneous injection of NIH-07. This end point is a sensitive measure of T-cell reactivity and previous T-cell sensitization (22). There was a clear increase in PLN weight after NIH-07 injection, indicative of a T-cell response (Fig. 2B). However, animals that had neonatal contact with NIH-07 antigens did not exhibit an altered PLN response.

We also tested for a priming/tolerance-inducing effect of NIH-07 antigens when fed only after weaning. In diabetes-resistant BBc rats, the PLN response to NIH-07 antigens was much higher if the animals had never received the NIH-07 diet as compared with rats that were fed NIH-07 from weaning (Fig. 2C) (P < 0.01). By contrast, BBdp rats showed no evidence of developing tolerance to NIH-07 diet fed from weaning.

As there seemed to be little effect of the neonatal treatment on DTH, this suggested that the benefit observed was unlikely to be the result of the induction of systemic tolerance but might be related to the local immune response in the gastrointestinal tract. To test this proposition, we measured IFN-γ and TGF-β gene expression in the gut of neonates. BBdp rat pups aged 5 days were fed either PG or PG + NIH (n = 8/group) and killed at 24 h, and their tissues were harvested. Cytokine gene expression was measured using semiquantitative RT-PCR. The animals that received NIH-07 showed no change in TGF-β mRNA expression in the gut, but IFN-γ was downregulated compared with those fed PG alone, resulting in a decreased IFN-γ/TGF-β ratio, which indicated a change in local gut immune bias in response to NIH-07 diet in the first week of life (Fig. 3). Transcripts of interleukin (IL)-12p40 were also measured but were found to be below detection limits (data not shown).

To examine further the concept of neonatal gut immunomodulation, we administered orally a nonspecific immunostimulatory compound, LPS, to neonates in the same manner as described above. LPS treatment increased diabetes incidence from ∼50 to 73% (Fig. 4). Although this increase was not statistically significant, it may have been biologically relevant as we observed the same enhancement in diabetes incidence when LPS was administered to BBdp neonates that were weaned onto a WG-based diet (F.W.S., unpublished data). Furthermore, there was an increase in islet infiltration as determined by standard light microscopy. Rats that were fed an HC diet had a lower insulitis rating compared with those that were fed NIH-07 alone or orally exposed as neonates to LPS and weaned to NIH (Table 2). When silica particles were administered concurrently, the LPS-induced increase in diabetes cases was blocked (LPS versus LPS/silica group, 41 vs. 73%; P = 0.02, Kaplan Meier log rank). This was also accompanied by a significant decrease in insulitis score (2.2 ± 1.1 vs. 3.7 ± 0.7; P = 0.001) (Table 2). In addition, age of onset was delayed in the LPS-silica/NIH group compared with none/NIH (110 ± 23 vs. 85 ± 17 days; P = 0.02), suggesting a possible benefit.

To examine for a local effect of these treatments on gut immunity, we measured gut cytokines in pups that were untreated (fed with dam’s milk alone), fed LPS or LPS + silica in PG at day 5, and killed 24 h later. IFN-γ mRNA expression decreased in response to oral LPS (P < 0.001) and was significantly restored in the LPS/silica group (Fig. 5). TGF-β mRNA expression in the small intestine of neonates varied less in response to oral dosing of LPS or LPS + silica (Fig. 5). As a consequence, LPS induced a significant drop in the IFN-γ-to-TGF-β ratio of gene expression, which was restored in the LPS/silica group (Fig. 5). Again, mRNA for IL-12p40 was not detected.

The possibility that type 1 diabetes may be related to antigens encountered via the gut lumen has been discussed since the mid-1980s, when evidence first appeared that diet could affect the spontaneous development of diabetes in BB rats and studies in patients implicated early exposure to breast milk substitutes as a risk factor. More recently, there have been several indications that the gut may be abnormal in some people who develop diabetes.

Savilahti et al. (23) found that jejunal biopsies from patients showed increased major histocompatibility complex class II expression (HLA-DR, DQ, and DP), a sign of enhanced immune activation, that had expanded in most of the villi and crypts in addition to the normal expression seen only on the upper villi. Another study compared peripheral blood mononuclear cells of young patients with type 1 diabetes and healthy children and found that IFN-γ secretion was higher and TGF-β was lower in α₄β₇ high T-cells compared with control subjects (24). The two major inducers of MHC class II expression in the inflamed gut are IFN-γ and tumor necrosis factor-α. Enhanced Th1 cytokine responses in the gut can cause gut inflammation and damage, and they are associated with increased gut permeability in celiac patients, Crohn’s disease, and other chronic gut inflammatory conditions (25). Increased gut permeability was recently reported for patients with type 1 diabetes (26) and BBdp rats (12). Thus, there is suggestive evidence that the gut of patients with type 1 diabetes shows damage that may be the result of inflammatory changes. However, it is not clear yet whether an inflammatory condition in the gut is linked to the process that destroys islet β-cells.

The data presented here for BBdp rats confirm (2) that the diet received from weaning is a major modulator of the natural course of disease development. When fed a mostly cereal-based diet or a WG-based diet, between ∼60 and 74% of animals developed overt diabetes as compared with <20% if rats received a diet with HC as the sole source of amino acids. The surprising finding was that exposure to WG or cereal-based diets in the first week of life, a time before gut closure when the gastrointestinal mucosa is immature, resulted in delay and even avoidance of diabetes in approximately one-third of diabetes-prone BB rats. No such protective effect was seen when an HC-based diet was administered during the same time period. Additional evidence for a critical role of gut function during the first week of life came from oral feeding with LPS. This intervention produced more diabetes cases and increased islet infiltration, whereas the concomitant administration of silica particles resulted in a significantly lower diabetes frequency than seen after LPS alone. Silica particles are toxic to macrophages or other phagocytes and hence target LPS-responsive cells. Because of the large particle size, silica is not expected to pass the intestine and reach the systemic circulation. Hence, the gut tissue itself is probably responsible for the enhancement of disease seen after exposure to LPS or suppression seen after early exposure to a cereal diet mixture or WG proteins.

This view is supported by the lack of a systemic memory response in rats exposed to the NIH-07 diet in the neonatal period. There was neither enhancement or suppression of the 24-h DTH response to cereal antigens nor a modulation of the 7-day popliteal lymph node response. The latter measurement is a highly sensitive and reliable surrogate of T-cell sensitization (22).

Evidence for a direct effect on gut immune function by exposure to non–breast milk antigens in the first week of life came from studies of cytokine gene expression. We found a significant decrease of IFN-γ but not TGF-β gene expression, which is indicative of an altered immunoreactive state after dosing with NIH-07 antigens or LPS in comparison to HC. It is interesting that silica coadministration restored the IFN-γ/TGF-β ratio to normal by decreasing IFN-γ mRNA levels, while TGF-β levels were increased.

Because IFN-γ is a key proinflammatory and TGF-β the dominant anti-inflammatory cytokine in the gut, LPS does not seem to modulate gut immune reactivity in a remarkably different way than the NIH-07 mixture. Hence, the major difference between the two treatments was that in the case of LPS, cereal antigens were not introduced at the same time but only 2 weeks later, at weaning, indicating that the timing of exposure to NIH-07 antigens is important.

These findings suggest that the gut of BBdp rats responds unfavorably to cereal antigens seen after weaning onto solid food. When the same dietary antigens are introduced to the neonate, with concomitant exposure to dam’s milk, a potent immunomodulatory mixture of growth factors, cytokines, and various other peptides, the detrimental impact of weaning onto solid cereal-based food is dampened. A particular role of the neonatal gut immune system in controlling diabetes development was recently also suggested for diabetes-prone nonobese diabetic mice (17). It was reported that neonatal administration of human insulin or a B-chain fragment had a much more potent suppressive effect on later disease development than when fed after weaning.

An additional outcome of our study was that diabetes-resistant BBc rats seemed to differ in their response to cereal-based diets. The PLN response of BBc rats to cereal antigens was significantly suppressed after continuous exposure to NIH-07 from weaning, whereas no suppression was seen in BBdp rats. The latter response was lower, which is in accord with the functional T-cell defect and lymphopenia of BBdp rats (27). Taken together, our findings lend additional support to the concept that the gut immune response is abnormal in BBdp rats and linked to the disease process.

These studies were funded in part by the Canadian Institutes of Health Research (formerly the Medical Research Council of Canada), Juvenile Diabetes Foundation International, and Health Canada (to F.W.S.), and by the Deutsche Forschungsgemeinschaft (SFB503), the Bundesminister für Gesundheit, and the Minister für Wissenschaft und Forschung des Landes Nordrhein-Westfalen (to H.K.).

We thank Heidi Gruber, Heather Cloutier, Penny Jee, Peter Smyth, and Dr. Olga Pulido for technical assistance. We thank Jocelyn Souligny and Dominique Patry for technical assistance and animal care and Dr. Pierre Thibert for maintaining the BB rat colonies. We thank the following students at the University of Ottawa, who contributed to these studies: Virginia Okolie, Shila Rastegar, Rosa Guzzo, and Linda Charles.