Abstract
Shigella is a diarrheal pathogen that causes disease through invasion of the large intestinal mucosa. The endotoxin of the invading bacterium may play a key role in the disease process by causing inflammation and tissue injury during infection. Earlier studies have shown that various animal species lacking functional CD14 were protected against endotoxin-mediated shock. Rabbits experimentally infected with Shigella were used to test the hypothesis that blockade of endotoxin-induced cell activation with anti-CD14 mAb would diminish inflammation and thus disease severity. Unexpectedly, we observed that the intestinal mucosa of anti-CD14-treated animals exhibited a 50-fold increase in bacterial invasion and more severe tissue injury compared with controls. Despite higher bacterial loads in treated animals, the numbers of polymorphonuclear leukocytes that were recruited to the infection site were similar to those in controls. Furthermore, the phagocytic cells of CD14-blocked animals produced IL-1 and TNF-α. Moreover, in vitro blockade of CD14 did not impede bactericidal activity. Thus, anti-CD14 treatment interfered with host defense mechanisms involved with removal/eradication of Shigella.
Shigella invade the rectal and colonic mucosae and cause severe, often life-threatening, inflammation. Although many patients recover spontaneously, these infections are sufficiently severe to cause more than half a million deaths annually among children in developing countries (1). The pathogenic mechanisms responsible for initiating injury are not well understood. Only rarely do the bacteria invade underlying tissue or enter the bloodstream. Antibiotics are of limited use (2) and may even contribute to a dangerous complication of Shigella infection, the hemolytic uremic syndrome (3).
Immunohistochemical studies of the intestinal mucosa of patients suffering from Shigella dysentery show infiltration of the epithelium by polymorphonuclear cells and lymphocytes as well as increased numbers of mast cells (4). PMN and plasma cells invade the lamina propria at an early stage, later accompanied by macrophages and eosinophilic granulocytes (4). Levels of proinflammatory (IL-1β, IL-8, TNF-α) as well as anti-inflammatory (IL-10, IL-1 receptor antagonist (IL-1ra)3) cytokines are increased in the rectal mucosa (5).
Several attempts have been made to down-regulate host inflammatory responses during systemic bacterial infections concomitant with antibiotic treatment, with the aim of improving patient survival (6). In the rabbit model of experimental shigellosis, infusion of the IL-1ra diminished tissue destruction and bacterial invasion of the intestinal mucosa (7). Conversely, administration of a mAb against IL-8 enhanced bacterial translocation across the epithelium, but this was associated with less tissue damage and fluid secretion (8).
A key molecule in inflammation caused by Gram-negative bacteria is CD14, which is expressed on the surface of most monocytes/macrophages and PMN (9). The exquisite sensitivity of membrane-associated CD14 for LPS requires the presence of LPS-binding protein. LPS-binding protein is a serum protein that forms complexes with LPS and facilitates transfer of LPS to membrane-associated CD14 (9). Binding of LPS to membrane-associated CD14 triggers the production of a cascade of proinflammatory cytokines, including TNF-α, IL-1, and IL-6 (9, 10, 11, 12, 13, 14). It is generally acknowledged that the inflammatory response accompanying a bacterial infection contributes to tissue destruction and mortality. Thus, a feasible treatment strategy would be to block the interaction between bacterial LPS and CD14 on inflammatory cells. Primates that were treated with monoclonal anti-CD14 Ab i.v. did not develop endotoxic shock after LPS infusion (15). Anti-CD14 treatment also protected rabbits from organ injury and death elicited by repeated administration of LPS i.v. (16). Furthermore, CD14 knockout mice are insensitive to the noxious effects of systemically administered LPS and control the dissemination of an i.p. Escherichia coli inoculum more efficiently than wild-type mice (12). In contrast, transgenic mice expressing human CD14 are hypersensitive to LPS (17).
The aim of this study was to evaluate the effect of Ab-mediated blockade of CD14; the mAb to rabbit CD14 blocks LPS binding to both membrane and soluble CD14. Rabbits were infected with an invasive Shigella strain in ligated small intestinal loops, whereupon bacterial dissemination, inflammatory responses, and tissue destruction were assessed. Contrary to the protective role exerted by anti-CD14 treatment in endotoxin-mediated shock, blockade of CD14 aggravated Shigella infection in the intestinal mucosa, resulting in higher bacterial loads and more severe tissue destruction. This suggests a prominent role for CD14 in innate immune responses to microbial Ags.
Materials and Methods
Animals
Seventeen male New Zealand White rabbits (CEGAV S.S.C., Les Hautes Noes, France), reared specific pathogen free and weighing between 2.75 and 3.0 kg, were used in the experimental shigellosis study. Two rabbits were employed for the collection of blood-derived PMN and peritoneal macrophages, and one rabbit was used for control studies of the localization of CD14 in the normal rabbit intestine.
Bacteria
Two Shigella strains were used: Shigella flexneri of serotype 5a, strain M90T, previously isolated from a dysenteric patient, and the isogenic, noninvasive strain, BS176, a spontaneous deletion mutant that lacks the entire invasion plasmid (18). The bacteria were streaked on Congo Red agar, and three to five colonies were resuspended in 7 ml of tryptic soy broth (Diagnostics Pasteur, Marnes la Coquette, France) and then plated on tryptic soy agar plates. After overnight culture at 37°C, the bacteria were harvested in sterile 0.9% NaCl solution and adjusted to 1010/ml based on OD measurement (590 nm).
CD14 blockade in vivo
Rabbits were given combined acepromazine-ketamine anesthesia (0.75 and 35 mg/kg i.v., respectively). A murine monoclonal IgG2a Ab directed against rabbit CD14 (clone 1116 1a6; 5 mg/kg), known to block LPS binding and subsequent cell activation (16), was injected i.v. Control rabbits received 5 mg/kg of a nonrelated murine IgG2a monoclonal (w6/32) (16) or the same volume of saline or remained untreated.
Experimental Shigella infection
Thirty minutes after CD14 blockade, a midline incision was made. Between six and eight 10-cm-long small intestinal loops were created by ligation, while carefully preserving intestinal blood supply. A total of 5 × 109 bacteria were injected per loop in 0.5 ml of NaCl. Eight hours later, the rabbits were sacrificed by an i.v. overdose of sodium pentobarbital (SANOFI-Santé Animale, Libourne, France). The midline incision was reopened, and the loops were cut open. The intestinal mucosa was inspected macroscopically, and intestinal biopsies were collected.
Histological analyses
Intestinal biopsies were fixed in either 3.7% paraformaldehyde-PBS (pH 7.4) or a zinc fixative solution (19) for 3–5 days. The tissue blocks were embedded in low temperature paraffin (melting point, 37°C) (20) from which 4-μm sections were prepared. Hematoxylin-eosin stains were made and read at ×200 magnification using an Olympus BX-50 microscope (Olympus, New Hyde Park, NY). The microscopic image was projected onto a Sony Trinitron screen (Sony, Tokyo, Japan), such that the height of the villi could be measured. Between two and four biopsies from different sites of the small intestine were analyzed per rabbit.
Sirius Red stains were performed on paraformaldehyde-fixed sections as follows: 0.5 g of Sirius Red powder (Sigma Chemical Co., St. Louis, MO) was dissolved in 45 ml of distilled water, to which 50 ml of absolute ethanol and 1 ml of a 1% NaOH solution were mixed in, before the dropwise addition of about 4 ml of 20% NaCl (until a light precipitate began to form). The solution was left to stand on the bench and was filtered the next morning. Slides were deparaffinized by immersion in acetone, hydrated in water, stained with hematoxylin-eosin, washed in water, dipped in 70% ethanol, and then covered in the Sirius Red solution for 1 h. The slides were finally rinsed in water for 10 min before being clarified in xylene and mounted. PMNs were counted in five randomly chosen mucosal fields, each measuring about 0.04 cm2. On the average, 10 fields from two different loops incubated with M90T were analyzed for every rabbit.
Immunohistological analyses
Intestinal tissue sections fixed in zinc salt solution were used. Slides were deparaffinized by the sequential immersion in two baths of pure acetone during a total of 5 min. A short incubation in distilled water, followed by a 5-min incubation in pure methanol was next. Endogenous peroxidase activity was quenched by immersing the slides in either methanol or water containing 0.3% H2O2 for 20 min. Sections were incubated for 15 min in a commercial blocking agent (DuPont-NEN, Boston, MA) to diminish nonspecific binding. Primary Abs dissolved in PBS were added for 2 h. The following primary Abs were used: murine mAb anti-rabbit CD14 (8 μg/ml; clone 1116 1a6) (16), murine mAb anti-LPS of S. flexneri 5a (6 μg/ml; A. Phalipon, Unité Pathogénie Microbienne Moléculaire, Institut Pasteur, Paris, France), and murine mAb anti-rabbit alveolar macrophages (4 μg/ml; RAM 11, Dako, Glostrup, Denmark). Slides were washed three times in PBS, and then the secondary Ab, biotinylated goat-anti mouse Ig (E0433, Dako) diluted 1/400, was added for 1 h. Slides were next washed in PBS, followed by the addition of streptavidin-HRP conjugate (Dako) for 45 min. In some instances, a tyramide signal amplification step (DuPont-NEN) was added at this stage, according to the instructions of the manufacturer. The color reaction was developed by the addition of H2O2 and aminoethylcarbazole (Sigma). Sections were counterstained in hematoxylin. To verify that the in vivo administered anti-CD14 Ab bound to cells in the intestinal mucosa, we incubated the intestinal sections of the CD14-blocked rabbits with the biotinylated goat-anti mouse Ig (E0433, Dako), normally used as a secondary Ab at a higher concentration (1/100). Bound Ab was detected by the sequential addition of streptavidin-HRP and amino-ethylcarbazole/H2O2 as described above. For calculation of the fraction of mucosal area that was invaded by Shigella, we analyzed intestinal sections that had been stained with the mAb directed against LPS, as described above. Two different sections were studied per rabbit, and three representative photographs were taken per section at ×200 magnification. Next, the images were incorporated into the software program Pilot Fotolook SA 2.08 (Agfa-Gevaert, Rueil Malmaison, France) using the scanner Agfa Horizon Ultra (Agfa-Gevaert). The area of the mucosal surface that had been invaded by Shigella was calculated with the aid of the software program Adobe Photoshop 5.0 (Adobe Systems Direct, Glasgow, U.K.), using a Macintosh Power PC 9500 (FNAC Montparnasse, Paris, France).
IL-1β and IL-1ra ELISAs
Intestinal punch biopsies (8 mm in diameter) were put into 1 ml of solution containing a mixture of protease inhibitors: 0.05% (w/v) sodium azide, 1 μg/ml aprotinin, 1 μg/ml leupeptin, 1 μg/ml pepstatin A, and 1 mM 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AEBSF) as previously described (21); homogenized for 20 s using an Ultra-Turrax homogenizer (Janke & Kunkel, Staufen, Germany); and ultracentrifuged at 100,000 × g for 1 h at 15°C. The supernatants were collected and stored at −80°C until assayed for cytokine quantities. A sandwich ELISA employing Abs specific for the rabbit cytokines in question, including standard curves of recombinant rabbit IL-1β and IL-1ra was used as previously described (22). The components of the sandwich ELISA were gifts from Dr. Matsukawa (Kumamoto University School of Medicine, Kumamoto, Japan). To compensate for variations in size of the homogenized intestinal biopsies, total protein determination using Bio-Rad’s protein microassay procedure (Bio-Rad Laboratories, Paris, France) was performed. Cytokine levels are expressed as picograms of cytokine per milligram tissue protein. There were no differences in total protein quantities in the intestinal biopsies isolated from loops incubated with the invasive or noninvasive S. flexneri 5a strain (data not shown).
TNF-α bioassay
Supernatants recovered from ultracentrifuged intestinal homogenates were assayed for bioactivity using the TNF-α-sensitive murine fibrosarcoma cell line WEHI 164, clone 13, as previously described (23). A standard curve based on dilutions of recombinant rabbit TNF-α was used. Eighty-three percent of the cytotoxic activity of the tested supernatants could be inhibited by coincubation with a polyclonal Ab directed against rabbit TNF-α, indicating that the assay mainly detected TNF-α activity and not the activity of other cytotoxic factors.
Myeloperoxidase (MPO)
A modification of the method described by Stark et al. (24) was used for quantification of MPO activity in the intestinal mucosa. Intestinal biopsies were homogenized in a lysis buffer consisting of 100 mM sodium acetate buffer (pH 6.0), 1% (w/v) hexadecyl trimethyl ammonium bromide, and 20 mM EDTA. The homogenates were stored at −20°C until assayed for MPO. Twenty microliters of sample was added in duplicate to immunosorbent plates (Nunc, Naperville, IL). Immediately afterward, 200 μl of reaction buffer was put in each well. Fresh reaction buffer was made by mixing 30 ml of lysis buffer (pH 5.3), one tablet of o-phenylenediamine dihydrochloride (10 mg; Sigma), and 2 ml of 30% H2O2. The plates were incubated covered at room temperature for 10 min, and then the absorbance was read at a wavelength of 490 nm (Dynatech MR4000, Dynatech, Guernsey, U.K.). A standard curve based on known concentrations of lysed PMNs was used. Protein determinations were made to compensate for variations in size in intestinal biopsies. Bio-Rad’s microassay could not be used due to interference of hexadecyl trimethyl ammonium bromide with the assay. Instead, relative protein quantities were determined by measuring the absorbance in the intestinal homogenates at a wavelength of 280 nm (U-1100 spectrophotometer, Hitachi, Tokyo, Japan), which reflects levels of aromatic amino acids. MPO activity in the intestinal homogenates is expressed as units per milligram of tissue protein.
Isolation of PMNs from arterial blood
Ten milliliters of arterial blood was collected in tubes supplemented with EDTA. PMN were isolated by dextran (Pharmacia, Uppsala, Sweden) sedimentation, followed by triple Percoll (Pharmacia) gradient separation (52, 67, and 85%). The purity of the fractionated PMNs was >95%, with >90% viability.
Isolation of peritoneal macrophages
Rabbits were sacrificed, and then 400 ml of HBSS (without Ca2+, Mg2+, and phenol red) (Seromed, Biochrom, Berlin, Germany) was injected i.p. After massage of the stomach, peritoneal fluid was aspirated and centrifuged at 400 × g. Cells were resuspended in Iscove’s solution (Seromed) containing 5% FCS to a concentration of 2 × 105 cells/ml. The cell suspension was left to adhere in six-well tissue culture plates for 2 h at 37°C in 10% CO2. Nonadherent cells were washed away, and adherent cells were assayed as described below.
Bactericidal assay
PMN or peritoneal macrophages were treated with the mAb specific for rabbit CD14 (1 μg/ml) or with the control mAb w6/32 (1 μg/ml), for 15 min at 4°C. FACS analysis (FACSCalibur, Becton Dickinson, Stockholm, Sweden) was conducted to verify that the in vitro blockade of CD14 worked. We found that the anti-rabbit CD14 mAb bound to 89% of the cells gated as monocytes/macrophages, whereas the control mAb bound to 1% of the cells.
PMN were seeded onto poly-l-lysine-coated (Sigma) coverslips (2 × 105 cells/coverslip, 15 min at room temperature), while the macrophages were left adherent in the plastic tissue culture wells as described above. Freshly cultured Shigella were resuspended in HBSS without phenol red but supplemented with 1.5 mM Mg2+, 1.0 mM Ca2+, and 1–10% fresh rabbit serum. Two milliliters of this solution containing 4 × 105 to 2 × 107 bacteria (representing a multiplicity of infection of 2–100) was added to the phagocytes in the tissue culture wells, after which the phagocytic cells and bacteria were spun together at 180 × g for 5 min. The plates were then incubated at 37°C for up to 4 h, during which time the extracellular fluid was analyzed at various intervals (0, 15, 45, 30, 90, 120, and 240 min) for numbers of viable bacteria by plating various dilutions of the extracellular fluid on tryptic soy agar plates (viable counts). In parallel, PMN/macrophages were lysed in 1% (v/v) Triton 100/PBS at the indicated time intervals to determine the number of viable intracellular bacteria, also by the viable counts technique. The number of CFUs was counted the following day. The numbers of CFUs in the extracellular fluid and inside the phagocytes were added together to yield the total number of viable bacteria at the different time points. In addition, bacteria were cultured in HBSS containing fresh serum to assess bacterial survival in the absence of phagocytic cells.
Statistical analyses
Results are presented as geometric means or medians. p < 0.05 was used to define statistical significance. The nonparametric Mann-Whitney test was used for determination of the statistical significance of the difference in absolute values between CD14-blocked and control rabbits. Fisher’s exact test was employed to assay the statistical significance of the different frequencies noted in CD14-blocked and control rabbits, respectively. The software program GraphPad Prism 2.01 (GraphPad Software, San Diego, CA) was used for all analyses.
Results
Localization of CD14 in normal rabbit intestine
We first characterized the sites where CD14 is found in normal rabbit intestine. There was little or no staining of the mucosa, apart from labeled cells that were occasionally seen in the basal mucosa (Fig. 1⇓A). In contrast, strongly staining cells with the morphology of macrophages/monocytes or PMNs were present in the submucosa, both dispersed in the loose connective tissue (Fig. 1⇓B) and in the lumen of blood vessels (Fig. 1⇓C). Cells within the Peyer’s patches were also labeled by the anti-CD14 mAb, including macrophage-like cells in the periphery of the follicular area (Fig. 1⇓D). These cells were clearly not identical with the tingible body macrophages that are homogeneously dispersed in the entire follicular area (Fig. 1⇓E). High endothelial venules in the T cell-dependent areas were also stained by the anti-CD14 mAb (Fig. 1⇓F).
Immunohistochemical labeling of CD14 in the normal small intestinal mucosa of rabbits. A, Very little or no label was seen in the normal rabbit intestinal mucosa. In the submucosa, strongly labeled cells with the morphology of PMN (lightly labeled) and monocytes/macrophages (strongly labeled) were scattered in the loose connective tissue (B) or in the lumen of large submucosal veins (C). D, CD14-reactive cells were also found in the outer rim of the follicular area of intestinal lymphoid nodules (Peyer’s patches). E, Tingible body macrophages were labeled by the RAM11 mAb and were evenly distributed throughout the follicular area of the lymphoid nodule. F, The leukocytes exiting a high endothelial venule in the T cell area of a lymphoid follicle, as well as the endothelium itself, also bound the anti-CD14 mAb.
The distribution of CD14 was also assessed in rabbits whose intestinal loops had been exposed to the invasive Shigella strain. A clear increase in the numbers of CD14-positive cells was seen in the mucosa and submucosa of intestinal loops incubated with the invasive strain (Fig. 2⇓, A and B). Cells within the blood vessels in the mucosa and submucosa were also strongly labeled by the anti-CD14 mAb (Fig. 2⇓, B and C). Furthermore, the endothelium itself was stained by the CD14-specific mAb (Fig. 2⇓, B and C).
The distribution of CD14-labeled cells in the mucosa of intestinal loops incubated with invasive Shigella. A, CD14-positive leukocytes entered the mucosa and formed aggregates or lay scattered in the lamina propria. In addition, the endothelium of veins located in the mucosa (B) as well as in the submucosa (C) were strongly labeled by the anti-CD14 mAb.
Tissue destruction in CD14-blocked rabbits exposed to invasive Shigella
Tissue destruction was determined by measuring villus height reduction in rabbits infected with invasive Shigella; rabbits that were pretreated with the anti-CD14 mAb were compared with control rabbits. The intestinal mucosa of CD14-blocked rabbits exposed to invasive Shigella (Fig. 3⇓A) was considerably more damaged than the intestinal mucosa of control rabbits exposed to the same bacterial strain (Fig. 3⇓B). Thus, there was a 2.2-fold reduction in villus height in CD14-blocked rabbits compared with controls (geometric mean 145 vs 294 μm) (Fig. 3⇓D).
Architecture of the intestinal mucosa during experimental Shigella infection of rabbit intestinal loops. Hematoxylin-eosin stains of the intestinal mucosa of a CD14-blocked (A) and a control (B) rabbit, incubated with the invasive Shigella strain. A, Note the extensive tissue destruction in the CD14-blocked rabbit, with complete erosion of intestinal villi; B, compared with the club-like villi of the control rabbit; C, CD14-blocked rabbits exposed to the noninvasive Shigella strain had intact mucosa; D, different degrees of tissue destruction in the three (A–C) groups of rabbits. Dots indicate the mean height of villi in individual rabbits. Bars indicate geometric mean values for each group of rabbits.
To exclude the possibility that the anti-CD14 mAb itself mediated the increased tissue destruction, we studied the villus height in loops incubated with the isogenic, noninvasive mutant (Fig. 3⇑, C and D). The villus height was similar in CD14-blocked rabbits and controls incubated with the noninvasive Shigella strain (377 μm in CD14-blocked vs 344 μm in control animals; n = 6).
Upon macroscopic examination of the intestinal mucosa of rabbit intestinal loops incubated with the invasive Shigella strain, we noted that hemorrhagic lesions occurred more frequently in CD14-blocked rabbits compared with controls. Microscopic examination of blinded hematoxylin-eosin-stained sections of intestine revealed that six of eight CD14-blocked vs two of nine control rabbits developed hemorrhages upon infection with invasive Shigella (nonsignificant by Fisher’s exact test).
Bacterial invasion of the intestinal mucosa of CD14-blocked rabbits
To evaluate the degree of bacterial invasion in intestinal loops incubated with the invasive Shigella strain, intestinal sections were stained with a mAb directed against Shigella LPS. By using image analysis techniques, we calculated the percentage of mucosal surface that had been invaded by Shigella in CD14-blocked and control rabbits, respectively. All sections were read in a blinded fashion. Typical staining patterns are shown in Fig. 4⇓; the noninvasive Shigella strain was found only in the intestinal lumen, or in the mucus overlying the brush border, but never in the mucosa. The staining pattern was identical in CD14-blocked and control rabbits (Fig. 4⇓A). As expected, bacterial invasion of the mucosa was seen in intestinal loops inoculated with the invasive Shigella strain. However, whereas the bacteria had a patchy distribution in the intestinal mucosa of control rabbits (Fig. 4⇓B), they were more disseminated in CD14-blocked rabbits (Fig. 4⇓C). On the average, 4.2% of the mucosa in CD14-blocked rabbits contained Shigella bacteria compared with 0.08% in control rabbits. Thus, there was a 50-fold greater degree of bacterial invasion in the CD14-blocked rabbits (Fig. 4⇓D).
Bacterial invasion patterns of the intestinal mucosa of rabbits experimentally infected with Shigella. Bacteria are labeled red and were detected by immunohistochemical labeling with an mAb directed against LPS of S. flexneri 5a. A, The noninvasive S. flexneri strain was strictly localized to the lumen of the intestine; B, the invasive strain exhibited a patchy pattern of invasion in control rabbits; C, diffuse invasion pattern in CD14-blocked rabbits; D, the average area of intestinal mucosa that was invaded by Shigella. Dots indicate the values for individual rabbits, and the horizontal bars indicate the medians for each group of rabbits.
Effect of CD14 blockade on migration and degranulation of PMN in vivo
To examine whether the anti-CD14 treatment affected recruitment of PMN to the intestinal mucosa, PMN were enumerated in intestinal sections stained with Sirius Red. This dye stains neutrophilic granulocytes, eosinophilic granulocytes, and Paneth’s cells in the intestine. Paneth cells were distinguished from the granulocytes by different morphology and confinement to intestinal crypts (Fig. 5⇓A). CD14-blocked and control rabbits exhibited similar degrees of PMN infiltration in the mucosa in response to invasive Shigella infection (Fig. 5⇓B). There was no difference in intestinal PMN numbers in CD14-blocked and control animals exposed to the noninvasive Shigella strain (geometric means, 145 vs 184 PMN/cm2 intestinal mucosa).
Quantitative and qualitative measures of PMN activity in the intestinal mucosa of Shigella-infected rabbits. A, PMN in the rabbit intestinal mucosa were strongly stained by Sirius Red. They could easily be distinguished from Paneth’s cells (labeled with arrows); the latter were larger, less strongly stained, and strictly localized to the small intestinal crypts. B, Numbers of PMN in the intestinal mucosa of CD14-blocked and control rabbits infected with invasive Shigella. C, MPO activity in homogenates of intestinal loops exposed to invasive Shigella. Dots indicate the values noted for individual rabbits, and bars indicate the median value for each group of rabbits.
The presence of one component of the granules of phagocytic cells in the mucosa was quantified: MPO, found in monocytes/macrophages and PMN. MPO activity was determined in intestinal homogenates using ELISA. Levels of MPO were somewhat lower in the mucosa of CD14-blocked rabbits (geometric mean = 60 × 103 U/mg tissue protein) compared with those in control rabbits (97 × 103 U/mg tissue protein), this difference was not statistically significant (Fig. 5⇑C). Intestinal MPO contents were virtually identical in CD14-blocked and control animals exposed to the noninvasive Shigella strain (geometric means, 67 and 64 U × 103/mg tissue protein).
Effect of CD14 blockade on the capacity of phagocytic cells to respond to LPS by the production of cytokines
We next tested whether blocking CD14 in vivo would affect the capacity of intestinal macrophages to produce proinflammatory cytokines, such as IL-1 and TNF-α, in response to bacterial infection. We also determined the presence of IL-1ra. CD14-blocked rabbits showed no impairment in the synthesis of TNF-α compared with control rabbits. In fact, TNF-α levels were 10 times higher in mucosal homogenates of CD14-blocked animals than controls (p = 0.006; Fig. 6⇓A). There was no difference between the two groups of rabbits regarding the synthesis of either IL-1β or IL-1ra (Fig. 6⇓, B and C). The ratio of IL-1ra to IL-1β was 16 in the intestinal biopsies from CD14-blocked rabbits and 49 in control rabbits (p = 0.09). The intestinal levels of IL-1β and TNF-α were very similar in CD14-blocked and control animals incubated with the noninvasive Shigella strain (geometric means: IL-1β, 24 and 30 pg/mg tissue protein; TNF-α, 2.4 U/mg tissue protein; for both groups of animals).
Cytokine levels in homogenates of intestinal loops exposed to invasive Shigella. A, TNF-α bioactivity; B, quantities of IL-1β; C, IL-1ra/IL-1β. Dots indicate the values noted for individual rabbits, and bars indicate the median value for each group of rabbits.
Bactericidal activity of CD14-blocked PMN and macrophages in vitro
Peritoneal macrophages or PMN were treated with either anti-CD14 mAb or control mAb in vitro. Thereafter, the capacity of the Ab-treated cells to kill the invasive Shigella strain was evaluated using 2–100 bacteria/phagocytic cell (multiplicity of infection, 2–100) and incubation for up to 4 h. The number of remaining bacteria in the culture medium was quantified using viable counts. To assess bacterial survival inside the phagocytic cells, the cells were lysed with detergent, and viable counts were made on the cell lysate. Very low numbers of viable bacteria were found intracellularly at any time.
The Shigella strain we used is serum sensitive (Fig. 7⇓). Hence, when bacteria were incubated in medium containing fresh serum in the absence of phagocytic cells, up to 70% of the bacteria were killed after 30 min (Fig. 7⇓). After this time, the bacteria started to grow again, rapidly reaching and surpassing their original numbers.
Effect of CD14 blockade on bactericidal activity of PMNs in vitro. The invasive S. flexneri strain was used at a multiplicity of infection of 4 (four bacteria per PMN). The bactericidal activities of control mAb-treated PMN and culture medium containing fresh serum are also shown. CFUs denote the sum of extracellular and intracellular bacteria that were viable at the indicated time points.
When PMN were incubated with Shigella, 87 and 88% of the bacteria were killed by the CD14-blocked and control mAb-treated PMN, respectively, within 15 min (Fig. 7⇑). Thus, a higher percentage of bacteria was killed by PMN than by medium containing serum alone. More importantly, there was no difference in bactericidal activity between CD14-blocked and control PMN (Fig. 7⇑). Similar tests were performed using purified peritoneal macrophages. Macrophages were incubated with 2–100 invasive Shigella/cell for up to 4 h. The same results were found. CD14 blockade did not significantly affect the killing of Shigella by macrophages in vitro (data not shown). However, PMN-mediated killing was quicker than the bactericidal activity of macrophages; the latter peaked after 2 h of incubation. To conclude, although culture medium containing serum was able to kill a considerable fraction of the bacteria, a much larger fraction of bacteria was killed in the presence of phagocytic cells. Furthermore, no difference in bactericidal activity was noted between CD14-blocked and control phagocytes at any multiplicity of infection.
Discussion
This study demonstrates that blockade of CD14 before experimental Shigella infection of rabbits results in aggravated disease, reflected by enhanced bacterial invasion and tissue destruction. To elucidate the mechanisms by which inhibition of CD14 could limit host defense against a bacterial intestinal infection, it was necessary to determine the localization of CD14 in healthy rabbit intestine. We found little or no expression of CD14 in rabbit intestinal mucosa. This is similar to what has been reported for humans, in whom normal small intestinal macrophages are negative for CD14 (25). In contrast, upon bacterial invasion of the rabbit intestinal mucosa, CD14-positive cells with the morphology of monocytes/macrophages and PMN immigrated into the mucosa from blood vessels and from the submucosa. We could also demonstrate the presence of CD14 on intestinal blood vessel endothelium. Endothelial cells are generally not believed to express CD14, even though they can be activated by LPS, and this activation can be blocked by CD14-specific mAbs (26, 27, 28). The answer to this apparent paradox is that endothelial cells can bind LPS complexed to soluble CD14 by as yet unidentified receptors.
It is not clear whether the anti-CD14 treatment in some way contributed to the increased frequency of hemorrhages noted in rabbits exposed to invasive Shigella. Histopathologic studies of the rectal mucosa of patients suffering from shigellosis have revealed that extensive vascular lesions are occasionally a feature of the disease (4).
CD14 has been proposed to be a pattern recognition receptor (29). This group of receptors is believed to recognize molecules derived from microorganisms, thus signaling the presence of danger to the host. Two recent studies have documented the importance of LPS-initiated host defense mechanisms in experimental bacterial infections in mice. Cross and co-workers showed that genetically LPS-hyporesponsive mice were 10,000 times more susceptible to lethal infection with a K1-encapsulated E. coli strain than normal, LPS-responsive mice (30). The lower 50% lethal dose in the LPS-hyporesponsive mice was mainly due to poor activation of monocytes/macrophages (30). It was recently demonstrated that these LPS-hyporesponsive mice lack the gene for Toll-like receptor 4, a protein capable of transducing LPS signaling (31, 32). Another new study showed that mice deficient in the LPS binding protein had reduced capacity to combat an i.p. inoculum of Salmonella typhimurium compared with wild-type mice (33). These previous results as well as the studies described herein comprise a body of data that emphasize the importance of LPS signaling in initiating host defense mechanisms against Gram-negative bacterial infections.
Our results may appear inconsistent with earlier studies that demonstrated that animals lacking CD14 or treated with a CD14-specific mAb are protected against endotoxin-mediated shock (12, 15). These divergent results are probably due to the fact that those earlier studies employed purified LPS or bacteria that were administered systemically, whereas we used live bacteria that were deposited on a mucosal surface. Hence, the aforementioned studies employed models that mimic septic shock, while the present study examined the outcome of CD14 blockade on a local, invasive bacterial infection. Septic shock constitutes the catastrophic outcome of uncontrolled infection. Although host defenses may play a detrimental role at such end stages of infection, it is likely that the vigorous inflammatory response triggered by LPS through the CD14 receptor pathway is evolutionarily beneficial, i.e., protects the host against the majority of naturally occurring Gram-negative infections on mucosal surfaces.
Yet another possibility for the disparate results between the present study and that of Haziot (12) is the fact that the latter study employed mouse strains that not only were deficient in the expression of CD14, but were also homozygously deficient in the expression of secretory phospholipase A2 (34). This molecule has proinflammatory and bactericidal activity (35) and has been isolated in the inflamed mucosa of patients suffering from inflammatory bowel disease (36) as well as in the circulation of patients suffering from septic shock (37). We quantified phospholipase A2 in the intestinal mucosa and in the intestinal fluid of the CD14-blocked and control rabbits exposed to Shigella, but found no differences between the two groups of rabbits with respect to secretory phospholipase quantities (data not shown).
The intestinal mucosa of CD14-blocked rabbits had a comparable density of PMN to control rabbits determined by both tissue staining and quantitation of MPO in intestinal homogenates. One may wonder what signals recruited the PMN into the infected mucosa if the response to bacterial LPS was blocked. Firstly, phagocytic cells have other receptors on their surface capable of recognizing microbial Ags (opsonized or nonopsonized ones), e.g., mannose receptors, scavenger receptors, glycan receptors, collectin receptors, and complement receptors. Secondly, bacterial invasion per se elicits a number of other signals than those mediated by the presence of surface-bound or free LPS. Thus, primary human fibroblasts produce IFN-γ in response to Shigella invasion (38). Among many functions, this cytokine enhances bacterial phagocytosis by macrophages (39). Invasion of intestinal epithelial cells has been shown to trigger the production of IL-8, GM-CSF, monocyte chemoattractant protein-1, and TNF-α (40). In contrast, mere attachment of LPS-containing bacteria to epithelial cells derived from organs that normally are not sterile results in either low grade or no cytokine production at all (40, 41, 42, 43). Thirdly, there are cells of the innate immune system that constitutively lack CD14 but still can react to the presence of bacteria, e.g., mucosal mast cells. This cell type is abundant in the intestinal mucosa and contains large quantities of the proinflammatory cytokine TNF-α stored in granules (44).
Evidently, the numbers of PMN recruited into the intestinal mucosa of CD14-blocked rabbits were not enough to defend the animal against invasive Shigella, because 50 times more bacteria were found in the mucosa of CD14-blocked than in control animals. Alternately, the PMN were not recruited fast enough. We found no evidence that the recruited PMN or macrophages were in any way anergic to bacterial signals. They produced IL-1 and TNF-α and degranulated close to foci of bacterial invasion in the mucosa. Moreover, in vitro blockade of CD14 did not impede phagocytosis and killing of the Shigella strain used in the study. To our knowledge, no one has described a role for CD14 in the phagocytic capacity of PMN, such that our results were perhaps to be expected. However, our failure to demonstrate a function for CD14 in uptake and killing of Shigella by macrophages was perhaps a bit more surprising. Two other studies have previously shown that monocytes and monocytic cell lines possess CD14-dependent mechanisms for the phagocytosis of Gram-negative bacteria (45, 46). It is possible that the picture is complicated by the fact that we are studying an invasive pathogen. Are these pathogens taken up by the phagocytes or do they invade the phagocytes? Notwithstanding, Hirose and co-workers showed that another invasive enteropathogen, Salmonella typhi, was easily killed in mouse and human macrophage cell lines, but survived in a CD14-deficient mouse macrophage cell line (47).
To conclude, this study shows that modulation of the nonspecific immune response elicited by a local Gram-negative bacterial infection may ultimately be harmful to the host. Although the activation of professional phagocytic cells by microorganisms may partly contribute to tissue damage, it seems clear that by down-regulating the ability of CD14-expressing cells (monocytes/macrophages, PMN, and possibly endothelial cells) to respond to a bacterial infection, the end result may be enhanced bacterial dissemination and possibly, secondary to the more widespread bacterial infection, increased tissue destruction.
Acknowledgments
We thank Dr. Lhousseine Touqui (Pasteur Institute) for generous help with the PLA2 assays, and Dr. Agnes Wold (Göteborg University) for critical reading of the manuscript.
Footnotes
-
↵1 The Hellmuth Hertz Foundation and the Swedish Society for Medical Research financed C.W.’s postdoctoral studies at the Pasteur Institute, with further supplementation from the Swedish Medical Research Council. This work was also supported by U.S. Public Health Service Grant AI15136 (to R.J.U.).
-
↵2 Address correspondence and reprint requests to Dr. Christine Wennerås, Department of Medical Microbiology and Immunology, Guldhedsgatan 10, Göteborg University, S-413 46 Göteborg, Sweden. E-mail address: christine.wenneras{at}microbio.gu.se
-
↵3 Abbreviations used in this paper: IL-1ra, IL-1 receptor antagonist; PMN, polymorphonuclear granulocyte; MPO, myeloperoxidase.
- Received July 20, 1999.
- Accepted January 11, 2000.
- Copyright © 2000 by The American Association of Immunologists
References
In this issue
