Intestinal epithelial serum amyloid A modulates bacterial growth in vitro and pro-inflammatory responses in mouse experimental colitis

Abstract

Background: Serum Amyloid A (SAA) is a major acute phase protein of unknown function. SAA is mostly expressed in the liver, but also in other tissues including the intestinal epithelium. SAA reportedly has anti-bacterial effects, and because inflammatory bowel diseases (IBD) result from a breakdown in homeostatic interactions between intestinal epithelia and bacteria, we hypothesized that SAA is protective during experimental colitis.

Methods: Intestinal SAA expression was measured in mouse and human samples. Dextran sodium sulfate (DSS) colitis was induced in SAA 1/2 double knockout (DKO) mice and in wildtype controls. Anti-bacterial effects of SAA1/2 were tested in intestinal epithelial cell lines transduced with adenoviral vectors encoding the CE/J SAA isoform or control vectors prior to exposure to live Escherichia coli.

Results: Significant levels of SAA1/SAA2 RNA and SAA protein were detected by in situ hybridization and immunohistochemistry in mouse colonic epithelium. SAA3 expression was weaker, but similarly distributed. SAA1/2 RNA was present in the ileum and colon of conventional mice and in the colon of germfree mice. Expression of SAA3 was strongly regulated by bacterial lipopolysaccharides in cultured epithelial cell lines, whereas SAA1/2 expression was constitutive and not LPS inducible. Overexpression of SAA1/2 in cultured epithelial cell lines reduced the viability of co-cultured E. coli. This might partially explain the observed increase in susceptibility of DKO mice to DSS colitis. SAA1/2 expression was increased in colon samples obtained from Crohn's Disease patients compared to controls.

Conclusions: Intestinal epithelial SAA displays bactericidal properties in vitro and could play a protective role in experimental mouse colitis. Altered expression of SAA in intestinal biopsies from Crohn's Disease patients suggests that SAA is involved in the disease process..

Figure 1

Intestinal SAA expression and secretion. SAA1/2 RNA, represented by the black signal in A and E, or the white signal in the counter-stained image B, was readily detectable in cross-sections of colonic epithelia (A, B), and was mainly located at the villous tips (E). SAA3 message showed similar distribution, but was much weaker (black signal in C, white in counter-stained image D). SAA immunostaining (green signal) showed a more diffuse pattern than SAA RNA, with SAA protein expressed along the crypt-villus axis (F), perhaps reflecting secretion of SAA by intestinal epithelial cells into the apical and basolateral milieu. A,C and E represent bright-field images, B and D dark-field images.

Figure 2

SAA levels in plasma and stool of knockout and wildtype mice. Samples of acute phase plasma, obtained 24 h after intravenous injection of 1 microgram LPS into wildtype mice or DKO mice (A), or extracts of stool samples from wildtype mice (B) were analyzed by immunoblotting with anti SAA1/2 antiserum following SDS-PAGE. The right lane in "B", labeled "SAA", is a positive control consisting of acute-phase plasma obtained as described for panel A. The reduced mobility of SAA in stool samples is likely caused by self-aggregation of SAA.

Figure 3

SAA expression in mouse colon and ileum and in response to LPS. RNA was isolated from the colon and ileum of germfree or conventionally raised mice, and cDNA was amplified with oligonucleotides recognizing a common region of SAA1/2 (A). In conventional mice, SAA1/2 was detected in colon and ileum, whereas in germfree mice, SAA1/2 was only detectable in the colon. In (B), CMT93 cells were incubated with the indicated amounts of LPS for 16 h, and SAA1/2 and SAA3 expression were determined with realtime PCR. Whereas SAA1/2 message remained rather constant, SAA3 expression markedly increased with increased LPS exposure. Each group contained data on 4 wells. Shown are average ± S.E.M. The asterisk indicates statistically significant differences (P < 0.05) compared with un-induced cells (ANOVA).

Figure 4

SAA over-expression in cultured IEC reduces viability of co-cultured E. coli. CMT93 or HT29 cells were grown on plastic supports until confluency, in medium without antibiotics. The cells were then incubated with ~20 multiplicities of infection of AdNull or AdSAA, and 48 h later, 100 cfu E. coli were added per cell for 2 h at 37°C. A small amount of medium was removed to detect SAA expression by Western blotting (A). After the 2 h incubation, serial dilutions of the culture media were grown on LB-agar. The Y-axis shows the fraction (%) of surviving bacteria relative to the positive controls (bacteria grown in cell-free medium). Show are averages ± S.D. of triplicate wells per group of a typical experiment that was conducted three times with similar outcome. The difference between AdSAA and Adnull treated groups was statistically significant (t-test; P < 0.05).

Figure 5

SAA protects from experimental colitis. SAA knockout mice (n = 7; white bars) or age and sex-matched controls (n = 7; black bars) were treated with 3% DSS in their drinking water for 6 days, while control mice (n = 5 per group) received normal water. Colon length (A), hematocrit (B), bodyweight gain (C), and histological disease index (D) were determined on day 8. Only differences in colon length and hematocrit values were statistically significant (P < 0.05, ANOVA). Real-time PCR analysis of colon tissue samples revealed significant increases in expression of TNFα and osteopontin (OPN) in colon tissue samples of DSS-treated DKO mice (E and F respectively; asterisks indicate significant differences with other groups (ANOVA and post-hoc analysis; P < 0.05)).

Figure 6

SAA expression in human biopsies. RNA was isolated from biopsies obtained from inflamed ileal and colonic biopsies from CD patients (CD-I ("involved")) and from uninvolved areas ("CD-U"), as well as from control patients (n = 5). Shown is the average of SAA1/2 expression ± S.D., expressed as% GAPDH. The differences between CD-I and controls were statistically significant (ANOVA; P < 0.05).