Main

Approximately a decade ago, it was first hypothesized that human milk might benefit the infant via mechanisms other than nutrition and classic host defenses(1). Based on the biologic characteristics of its components, milk was noted to be “poor in initiators and mediators of inflammation, but rich in antiinflammatory agents,” leading to the conclusion that human milk might be antiinflammatory(1). To our knowledge, in vivo evidence of antiinflammatory activity is limited to one study in which a s.c. air pouch model of inflammation in rats was used(2). In this model, colostrum was as potent as dexamethasone at decreasing the acute inflammatory response to carrageenan, thereby demonstrating that human colostrum had a biologic antiinflammatory activity. Seeking to extend those studies, we used a chemical colitis model to examine the effects of enterally administered human milk on acute bowel inflammation.

METHODS

The use of human subjects and animals in these studies was approved by the IRB (IRB #05-06-92-0216) and the Animal Care and Use Committee (ACUC #94-14), respectively, at the Eastern Virginia Medical School.

Collection of human milk specimens. Human milk was collected from volunteer donor mothers 15-120 d postpartum. The entire contents of one breast were collected by electric breast pump (Medela, Inc, McHenry, IL) or by manual expression. The milk was immediately refrigerated, transported to the laboratory on ice, and processed within 2 h of collection. In the laboratory, milk was sedimented (400 × g, 10 min, 4 °C), and the cream and aqueous fractions were recombined, mixed, and pooled in 30-120-mL aliquots, which were stored at -70 °C until used.

Development of the animal model. In pilot studies, human milk and formula were supplied to animals ad libitum instead of water and rat chow, with twice daily replacement of human milk or formula. Human milk was thawed and placed in standard rat water bottles, and Similac (Ross Laboratories, Chicago, IL) was obtained as a premade formula and made available in the same fashion. Infant formula-fed rats consumed 50 ± 5 mL/rat/d and human milk-fed rats consumed 37 ± 8 mL/rat/d.

To induce colitis, different concentrations of acetic acid in water (0.5, 1, 2, 3.5, 4, and 10%) were administered to animals as described below to establish the dose-response characteristics of the model. In addition, several different sacrifice times (12, 24, 48, 72, and 96 h) were examined. Sacrifice at 24 h after 4% acetic acid enema was chosen for study, as our objective was to examine milk feeding effects on acute inflammation. Methods for MPO extraction from bowel (see below) were also developed during these studies.

A well described(3–5) and frequently used(6, 7) model of chemical colitis in rats was adapted for these studies. The feeding duration and timing of colitis induction were chosen arbitrarily, based upon the availability of human milk. Animals were housed two to three per cage in a temperature- and ventilation-controlled environment. Young adult male Sprague-Dawley rats (200-250 g) were begun on their assigned diet, and after 24 h of feeding, each group was fasted overnight. The next morning, 7 French feeding catheters were inserted into the rectum of each animal to a depth of 4 cm and two PBS enemas (2 mL each) were given to each animal to empty the colon of as much fecal material as possible. After the second PBS enema, a single 4% acetic acid enema (2 mL) was administered, the catheter was removed, and 10 s later, 3 mL of PBS plus 3 mL of air were introduced into the colon with a separate catheter to buffer the acetic acid. Animals were subsequently returned to their cages and continued on their assigned diets for another 24 h. After euthanasia with i.p. pentobarbital, laparotomy was performed, and the colon from the cecum to the proximal sigmoid was excised and processed as described. Tissues were either fixed in 10% buffered formalin for histologic examination or frozen at -70 °C until MPO was extracted.

Comparison of feeding effects. Animals were assigned to live experimental feeding groups in four separate experiments: 1) rats without colitis (n = 12, n = 6, n = 0, n = 4 for the four experiments, respectively) that were offered water and standard rat chow; 2) rats with colitis (n = 15, n = 12, n = 0, n = 5, respectively) offered standard chow and water; 3) rats with colitis (n = 15, n = 12, n = 8, n = 5, respectively), offered infant formula alone; 4) rats with colitis (n = 15, n = 12, n = 0, n = 5, respectively) offered human milk alone; and 5) rats with colitis (n = 0, n = 0, n = 7, n = 0, respectively) offered infant formula supplemented with 10 μg/mL recombinant human IL-1RA (kindly supplied by Carl Edwards III, Ph.D., Amgen Boulder Inc, Boulder, CO). Three control groups were also examined: rats without colitis (n = 3) offered infant formula alone, rats without colitis (n = 3) offered human milk alone, and rats with colitis (n = 6) offered Pedialyte (Ross Laboratories) alone.

Preparation of colons and quantitation of colonic MPO activity. In all experiments, the colon was removed and divided into five equal segments. In experiments 1, 2, and 4, the segments were divided longitudinally, and one half was fixed in formalin, whereas the other was used for MPO extraction. In the third experiment, colon segments were not subdivided, so entire segments were processed for MPO extraction. Colon weight measurements were calculated as the summed weights of the segments corrected for the pieces removed for histologic preparation.

For MPO extraction and quantitation(3), segments were thawed, and the mucosal surface gently scraped free of feces and mucus. After weighing (in grams to 2 decimal places), each segment was homogenized (glass Dounce hand homogenizer, 30-120 s) in 1 mL of ice-cold extraction buffer (0.5% hexadecyltrimethylammonium bromide, 50 mM phosphate buffer, pH 6.0), and held at 4 °C for 30 min with occasional mixing. Hexadecyltrimethylammonium bromide efficiently extracts MPO from rat tissues(8), and MPO is a commonly used marker for neutrophils in rat tissues(3, 6, 7). After homogenization, each aliquot was sonicated on ice (three times, 10 s each) followed by three cycles of freeze/thaw. Aliquots were then sedimented (14000 × g, 10 min, room temperature), and the supernate was collected. Supernates were assayed spectrophotometrically for MPO activity as follows: 0.1 mL of supernate was combined with 2.9 mL of substrate (50 mM phosphate buffer, pH 6.0 containing 0.167 mg/mL o-dianisodine hydrochloride and 0.0005% H2O2)(3), and the absorbance at 460 nm was measured at 10-s intervals for 2 min using a Perkin-Elmer Lambda 6 spectrophotometer (Perkin-Elmer, Norwalk, CT). The ΔOD460/min value was taken as the respective MPO activity and converted to activity/g of colonic segment (SEGMENT MPO). These data were also averaged across the five colon segments from each animal to estimate MPO activity/g for the entire colon (AVERAGE MPO).

Histologic examinations. In the first experiment, the formalin-fixed piece of each middle colon segment (segment no. 3) was paraffin-embedded, cross-sectioned, and stained with hematoxylin-eosin. These sections were reviewed by a blinded investigator (A.L.W.), who graded the gross and microscopic inflammation using a modification of a previously reported rat colitis histopathologic index(4). An integer score between 0 (no inflammation) and 4 (severe inflammation) was assigned to each specimen based on the most severely inflamed area present in the section.

In the fourth experiment, three colon specimens that displayed a range of SEGMENT MPO values were selected, and the formalin-fixed mirror-image half (from cecum to sigmoid) was divided into five segments that corresponded to the segments from which MPO was extracted. Each segment was embedded, sectioned longitudinally, and stained with hematoxylin-eosin. Ten of these sections were chosen randomly by a blinded observer (P.R.B.) and examined microscopically. Each contiguous high power field along the entire specimen was graded for three histologic characteristics: leukocyte infiltration (1-4+), hemorrhage (1-4+), and mucin depletion (1-4+). High power field scores were averaged across the specimen, plotted against the SEGMENT MPO of the corresponding mirror-image segment of colon, and compared by linear regression.

Statistical analysis. Because the AVERAGE MPO data showed skewed distributions, logarithmic transformation was applied and produced near normal distributions (see Fig. 2). For the same reason, logarithmic transformation was also applied to the SEGMENT MPO data set with similar normalization of distributions. Geometric mean values and RSE were considered the most appropriate descriptive statistics for the log-transformed data and were calculated as: Equation The variability of the geometric mean is expressed as the geometric mean × RSE and geometric mean/RSE, and is reported in the text using the format “geometric mean, geometric mean × RSE - geometric mean/RSE” (this expression identifies a 68% confidence interval for the geometric mean). A two-way ANOVA (experiment, treatment) was applied to the normalized AVERAGE MPO data from experiments 1, 2, and 4, followed by a Tukey multiple comparison test. A one-way ANOVA was applied to the normalized SEGMENT MPO data collected in each segment and the various treatment means were compared using t tests with Bonferroni corrections. Unless otherwise noted, data are expressed as geometric means ×/÷ relative SE, and statistical significance is declared at p < 0.05.

Figure 2
figure 2

AVERAGE MPO activity (ΔOD460/g of colon weight) observed in experimental groups. Each data point represents an animal. Colitis/milk animals have significantly less colonic MPO activity/g than either colitis/chow or colitis/formula animals (p < 0.01).

Full size image

RESULTS

Correlation of extractable MPO content and histologic leukocyte infiltration. To confirm the observations of others that extractable MPO activity correlates with leukocyte infiltration of colon(3), 10 colon segments were graded for their degrees of leukocyte infiltration, hemorrhage, and mucin depletion. As shown in Figure 1, strong correlations between SEGMENT MPO and leukocyte infiltration (r = 0.966, p < 0.001), hemorrhage (r = 0.920, p < 0.001) and mucin depletion (r = 0.962, p < 0.001) were found, confirming that extractable MPO activity correlated well with at least three parameters of acute colon inflammation.

Figure 1
figure 1

Relationships between SEGMENT MPO (ΔOD460/g of colon weight) and histologic grading of leukocyte infiltration, mucin depletion, and hemorrhage. Regression lines through each data set are shown with the corresponding values for r.

Full size image

Effects of acetic acid-induced colitis on colon morphology and weights. In all animal groups without colitis, the colons were thin walled, delicate and translucent. In animals without colitis fed chow, colon weights were low (geometric mean, 0.89 g, 0.86-0.92, n = 22) and were similar to colon weights in animals without colitis fed formula (geometric mean, 0.79 g, 0.76-0.83 g, n = 3) or human milk (geometric mean, 0.79 g, 0.71-0.89 g, n = 2). In the colitis groups, harvested colons varied in their appearances, from thin-walled, delicate, and translucent (i.e., normal) to hemorrhagic, thick-walled, and shortened. In all colitis groups, mean colon weights were increased (Table 1). Within the colitis groups, animals with colitis fed Pedialyte showed the highest colon weights, and animals fed formula supplemented with IL-1RA the lowest. Animals with colitis fed human milk had colon weights significantly lower than colitis/chow animals, but no different from colitis/formula animals.

Table 1 Colon weights of rats in different feeding groups
Full size table

Effects of acetic acid-induced colitis on colon MPO activity (Fig. 2). Rats without colitis, fed standard chow and water had low values of AVERAGE MPO: geometric mean, 0.27 U/g, 0.17-0.42 (n = 22). Control rats without colitis fed infant formula (n = 3) or human milk (n = 2) had similarly low values for AVERAGE MPO (geometric means, 0.008 and 0.072 U/g, respectively). Animals with colitis, fed rat chow and water had AVERAGE MPO values approximately 25 times higher than animals without colitis: geometric mean, 6.76 U/g, 5.62-8.13 (n = 32), p < 0.01 versus no colitis). Animals with colitis, fed infant formula, had AVERAGE MPO values significantly higher than control animals without colitis: geometric mean, 6.92, 5.76-8.32 (n = 40) (p < 0.01 versus animals without colitis). Animals with colitis, fed Pedialyte, showed geometric mean AVERAGE MPO values (8.13, 6.76-9.77, n = 6) similar to colitis animals fed either chow or infant formula, demonstrating that the acute inflammation observed was not due to irritant effects of rat chow or infant formula. Animals with colitis fed human milk had the lowest AVERAGE MPO values of the colitis groups: geometric mean 2.34 U/g tissue, 1.86-2.95 (n = 32) (p < 0.01 versus colitis animals fed chow or formula). This activity was approximately 9 times the levels (p < 0.05) seen in animals without colitis. In animals with colitis fed infant formula supplemented with IL-1RA, AVERAGE MPO levels were similar to those seen in colitis animals fed human milk: geometric mean 1.95 U/g, 1.20-3.16 (n = 7).

Comparison of colon weights versus total MPO activity [calculated as total MPO = (colon weight in g) × (AVERAGE MPO/g)/10] by feeding group showed no strong relationships in any of the feeding groups examined.

The SEGMENT MPO activity varied along the length of the colon in all animal groups with colitis; the highest SEGMENT MPO levels were present in the penultimate distal colon segment in all colitis conditions (Fig. 3). Despite the occurrence of significant animal-to-animal variation, variation within treatment groups and among individual experiments, analysis of SEGMENT MPO values showed that human milk feeding significantly decreased MPO content in all colonic segments compared with colitis/formula animals, and in segments 2-5 compared with colitis/chow animals. The greatest sparing effects of human milk feeding were seen in the more proximal colonic segments, with the most proximal segment approaching the values seen in animals without colitis (Fig. 3).

Figure 3
figure 3

SEGMENT MPO activity (ΔOD460/g of colon weight) by feeding group. Data shown are the geometric means (geometric mean × RSE, geometric mean/RSE) for each colon segment from animals in the respective feeding groups, with n = 22 for animals without colitis, n = 32 for colitis/chow animals, n = 40 for colitis/formula animals, n = 32 for colitis/milk animals, n = 6 for colitis/Pedialyte animals and n = 7 for colitis/formula + IL-1RA animals.

Full size image

Histologic assessment of colitis. In all colons examined histologically, regardless of the feeding group, colitis induction was heterogeneous. Focal areas of colitis were common, and in these areas the severity of inflammation varied. These effects were presumably due to residual fecal material in the colons at the time of colitis induction, which either buffered the acetic acid or prevented its contact with the mucosal surface. When histologic scoring was attempted to judge the degree of edema, hemorrhage, necrosis, and neutrophil infiltration in the middle colonic segment of animals in experiment 1, scoring was based on the most severe colitis present in the segment examined. Animals without colitis fed chow had scores of 0.17 ± 0.09 (mean ± SE) (n = 12). Animals with colitis fed chow, formula or human milk all showed significantly higher histologic scores than the control group, but no differences between chow-fed [2.5 ± 0.4 (mean ± SE), n = 15], formula-fed [2.4 ± 0.4 (mean ± SE), n = 15] or human milk-fed animals [2.3 ± 0.5 (mean ± SE), n = 15] were present.

DISCUSSION

Many studies show that human milk feeding protects infants against a spectrum of symptomatic infectious illnesses(9–12). In the past, this protection was assumed to be mediated through the “classical” host defense components contained in milk such as IgA, IgG, IgM, and lactoferrin. More recently, “novel” protective factors and mechanisms have been proposed to play roles in milk-mediated protection against infection, including antiinflammatory components(1, 13–17). Human milk contains numerous components with potential antiinflammatory activities(1, 13–21), and both colostrum and milk interfere with multiple neutrophil and lymphocyte functions important for inflammatory responses(18, 22–25). However, direct evidence for antiinflammatory effects exists only for colostrum, and the model system used to demonstrate these effects did not use enteral exposures to colostrum(2).

The model adapted for these studies resolves the problem of nonenteral exposure to human milk and specifically examines milk, rather than colostrum, feeding. The physiologic relevance of chemical rather than infectious colitis for these studies could be argued, but this choice was made to minimize potential contributions of milk's “antiinfective” components to the effects of milk feeding on the colitis. The data show that feeding human milk to rats with colitis is associated with significantly less colonic MPO activity than it is with chow, infant formula, or Pedialyte feeding to similar animals. Based on recent reports that picomolar concentrations (200-1700 pg/mL) of IL-1RA are present in human colostrum and milk(21, 26) and data suggesting that IL-1RA therapy may be effective in inflammatory bowel disease and rheumatoid arthritis(27), we supplemented infant formula with micromolar concentrations of recombinant human IL-1RA and observed effects on colonic MPO content similar to those seen with human milk feeding. This is consistent with the effects of parenterally administered IL-1RA in studies of rat colitis(28), but suggests that this cytokine antagonist may have antiinflammatory activity when administered enterally, as would occur in the breast-fed infant.

Using the histologic grading method described, we were unable to show a relationship between the middle colon segment inflammation score and the feeding group, despite strong relationships between tissue leukocyte infiltration and MPO activity as previously reported by others(3). Because the histologic score was based upon the highest degree of inflammation observed in the section, and colitis induction tended to be spotty and focal rather than homogeneous, it is not surprising that this approach did not correlate well with the extracted MPO values. These results emphasize that measurement of inflammation in this model requires methods that are not strongly influenced by the heterogeneity of the induced colitis, and may explain why extractable MPO as a surrogate marker for acute inflammation is commonly used. Our demonstration that extractable MPO correlates strongly with hemorrhage and mucin depletion as well as leukocyte infiltration confirms the appropriateness of this method for estimating inflammation in the bowel.

The SEGMENT MPO patterns demonstrated decreasing inflammation from the second to the fifth colon segments in all colitis groups, although the steepness of the gradient varied by feeding group. This suggests that the degree of injury caused by the acetic acid enemas decreased from distal to proximal. In the human milk and IL-1RA groups, the effect was more pronounced. We speculate that this steeper gradient of inflammation results from an inverse relationship between the severity of the initial injury and the degree to which milk's antiinflammatory effects alter the host response. Because inflammation is primarily a defense mechanism(29) and complete loss of this mechanism is predictably deleterious for the host, retention of inflammatory responses after severe injury may be beneficial, whereas diminution of inflammation after lesser injury might decrease clinical symptoms without seriously impairing the overall defense mechanism.

In summary, these data show that human milk feeding or feeding with infant formula containing an antiinflammatory component of human milk, IL-1RA, decreases the inflammatory response in the colons of rats after noninfectious injury to that organ. We suspect that similar effects occur in the bowel of breast-fed infants, representing one of the several protective systems operative in human milk. By altering the host's response to injury, antiinflammatory components may contribute to the broad “protection” against many types of injuries and conditions(30–35) provided by human milk feeding. In combination with the other protective systems, antiinflammatory effects could modify clinical symptoms or the healing by scar that normally follows significant inflammation. Whether such effects occur and are additional benefits of human milk feeding await further study.