Involvement of the acute phase protein alpha 1-acid glycoprotein in nonspecific resistance to a lethal gram-negative infection.

Resistance to gram-negative infection can be induced by pretreating animals with several agents such as turpentine and interleukin (IL)-1. Because these agents are powerful inducers of acute phase proteins, we wondered whether these proteins, more particularly alpha(1)-acid glycoprotein (alpha(1)-AGP), are involved in nonspecific resistance to infection. Turpentine and IL-1 protect completely against a lethal challenge of Klebsiella pneumoniae when given 48 and 12-48 h before the challenge, respectively. alpha(1)-AGP induction in the serum reached peak values 48 h after turpentine and 12-48 h after IL-1 injection. Administration of alpha(1)-AGP, 2 h before a challenge of K. pneumoniae, significantly increased the survival. Numbers of bacteria cultured from blood and organs were significantly lower in mice pretreated with a protective dose of turpentine, IL-1, or alpha(1)-AGP. These data suggest that alpha(1)-AGP is a possible mediator in turpentine- or IL-1-induced protection because time points of maximal induction of alpha(1)-AGP by turpentine or IL-1 and of optimal protection by alpha(1)-AGP coincide. Transgenic overexpression of rat alpha(1)-AGP protected mice from a K. pneumoniae infection. Bacterial counts in blood and organs were significantly lower in transgenic mice, and only in control mice were large necrotic areas, apoptosis, and blood clots observed in the spleen. Our data suggest that alpha(1)-AGP prevents gram-negative infections and may be an essential component in nonspecific resistance to infection.

Septic shock, or systemic inflammatory response syndrome, is characterized by inadequate tissue perfusion. It is caused by an overwhelming infection (1), and several bacterial organisms have been identified to induce that syndrome (2). The disease is accompanied by a high death rate, exceeding 50%, depending on the type of organism involved. Treatment with antibiotics has proven to be effective, but because bacteria are becoming increasingly resistant to antibiotics, alternative solutions have to be found. One possibility is to identify endogenous molecules that are involved in increasing nonspecific resistance to infection and to evaluate their therapeutic use. The natural resistance of the host to infection can be increased by injection of various substances, most of which are of bacterial origin, such as lipopolysaccharide and muramyl peptides (3)(4)(5). However, because of their toxicity, these immunomodulatory substances cannot be used therapeutically in humans. Substances that increase nonspecific resistance are able to stimulate mononuclear phagocytes to secrete interleukin (IL)-1 1 and tumor necrosis factor (TNF) (6). It was demonstrated that IL-1, and to a lesser extent TNF, are able to induce natural resistance to infection (7)(8)(9)(10)(11)(12). However, resistance can only be induced if certain time intervals are respected. For example, bacillus Calmette-Guérin has to be administered 2 weeks, lipopolysaccharide between 6 and 48 h, and IL-1 24 h prior to a lethal challenge of Klebsiella pneumoniae (4). To study septic shock, a lethal infection of K. pneumoniae is a relevant model because these Gram-negative bacteria were recognized, besides Escherichia coli, Proteus, Pseudomonas and Serratia, as some of the most frequent culture isolates (13).
␣ 1 -Acid glycoprotein (␣ 1 -AGP) is a highly glycosylated protein of 43 kDa with a pI of 2.7 (14). It is an acute phase protein in all mammals investigated so far (15). ␣ 1 -AGP is produced mainly in the liver (16), although also some extrahepatic synthesis has been reported (17)(18)(19). In mouse serum, ␣ 1 -AGP is normally found at a concentration of 0.2-0.4 mg/ml (20). During an acute phase condition, the concentration rises 2-5 times, making it one of the predominant proteins in the serum (14). Like most acute phase proteins, ␣ 1 -AGP is induced both by cytokines and corticosteroids (21). IL-6 and IL-1 have proven to be very powerful inducers of ␣ 1 -AGP both in vitro and in vivo (22). Although ␣ 1 -AGP is an abundant protein, its real physiological significance is not fully understood. Inhibition of platelet aggregation (23) and of neutrophil function (24,25) have been reported. Also, ␣ 1 -AGP was found to inhibit selectively the transport of molecules through the endothelial layer (26,27). In vivo protective effects of ␣ 1 -AGP have been described (28 -30).
We were interested in finding whether acute phase proteins are involved in nonspecific resistance to infection against K. pneumoniae. Therefore we studied the possible protection by turpentine oil and IL-1, two well known and very strong inducers of the acute phase response (31)(32)(33)(34), and by ␣ 1 -AGP itself, and we compared the kinetics of induction of ␣ 1 -AGP in the serum and of optimal protection against K. pneumoniae.
We describe that ␣ 1 -AGP significantly protects against a lethal infection with K. pneumoniae. This activity of the acute phase protein was observed in normal mice using purified ␣ 1 -AGP as well as in transgenic mice that overexpress rat ␣ 1 -AGP.

MATERIALS AND METHODS
Animals-Female C57BL/6 mice (Iffa-Credo, Saint Germain-surl'Arbresle, France) were used at the age of 8 -12 weeks. Rat ␣ 1 -AGP transgenic mice were generated and described previously (35). They were generated by injecting genomic DNA into (C57BL/6 ϫ DBA/2)F1 zygotes, and the resulting transgenic mice were back-crossed eight generations into a C57BL/6 background. Heterozygous transgenic mice from the line 9.5-5 constitutively produce about 2 mg/ml ␣ 1 -AGP. This is 10-fold more than wild-type (wt) animals. The colony was propagated by breeding heterozygous transgenic mice with C57BL/6 female mice; the offspring, containing heterozygous transgenics and wt littermates, was genotyped at weaning age by enzyme-linked immunosorbent assay. 100 l of blood was collected by retro-orbital bleeding, after which serum was prepared. ␣ 1 -AGP was purified by phenol extraction (36) and coated on the bottom of an enzyme-linked immunosorbent assay plate. After washing, rat ␣ 1 -AGP was detected using an anti-rat ␣ 1 -AGP polyclonal antibody (generated by H. Baumann in rabbits) (1/1,000) and an anti-rabbit antibody, conjugated to alkaline phosphatase (Sigma, St. Louis, MO; 1/5,000). The anti-rat ␣ 1 -AGP antibody did not cross-react with mouse ␣ 1 -AGP. About 50% of the offspring were heterozygous transgenic. Only female mice of 8 -12 weeks were used in the experiments. Both transgenic and control (nontransgenic littermate) mice had comparable body weights. Mice were kept in a conventional, air-conditioned mouse room in 12-h light-dark cycles and received food and water ad libitum.
Injections-Intraperitoneal injections had a volume of 0.5 ml. The reagents were diluted in pyrogen-free phosphate-buffered saline (PBS) immediately before injection. Mice were injected intramuscularly with bacteria (right thigh) in a volume of 100 l. Mice were bled by retroocular bleeding or heart puncture under ether or tribromoethanol (160 mg/kg) anesthesia, respectively, and serum was prepared by clotting 30 min at 37°C, removal of the clot, and centrifugation (15 min at 15,000 ϫ g).
Reagents-Bovine ␣ 1 -AGP, bovine serum albumin (BSA), alkaline phosphatase-conjugated anti-rabbit IgG, and p-nitrophenyl phosphate were obtained from Sigma. The ␣ 1 -AGP preparations contained 10 ng of endotoxin/mg of protein. ␣ 1 -AGP was Ͼ99% pure as mentioned by the manufacturer and as judged by polyacrylamide gel electrophoresis and subsequent Coomassie Blue staining. Recombinant mouse IL-1␤ was expressed in and purified from E. coli in our laboratory, had a specific activity of 3.65 ϫ 10 8 units/mg, and contained less than 10 ng of endotoxin/mg of protein.
Infection Model-K. pneumoniae (ATCC 43816), a strain that produces a lethal infection in normal mice (9), was inoculated in the right thigh muscle as described (37). An inoculum of 1 ϫ 10 6 CFU/mouse was used except in the studies with rat ␣ 1 -AGP transgenic mice, where we used 10 5 CFU/mouse. Survival was scored over a period of at least 5 days.
Clearance of Bacteria-36 h after injection of K. pneumoniae, mice were anesthetized by intraperitoneal injection of tribromoethanol. Blood was taken by heart puncture. For preparation of plasma, 450 l of blood was added to 50 l of sodium citrate (0.1 M). Immediately thereafter, mice were killed by cervical dislocation. Then, mice were perfused with 10 ml of a 0.9% NaCl solution to wash out the blood. The liver, spleen, and kidney were removed aseptically, weighed, and homogenized mechanically in sterile saline. For homogenization, the liver was diluted (w/v) 2-fold; spleen and kidney were diluted 10-fold. The suspensions were diluted and plated out on sterile nutrient agar. After overnight incubation at 37°C, CFU numbers were counted.
Measurement of ␣ 1 -AGP-The concentration of ␣ 1 -AGP in mouse serum was measured using a home-developed sandwich enzyme-linked immunosorbent assay. A rat monoclonal antibody was coated (0.1 g/ ml) on 96-well Maxisorb plates. After blocking with 1% BSA and PBS, samples and a murine ␣ 1 -AGP standard were titrated, after which the plates were incubated for 1 h at 37°C. A rabbit polyclonal antiserum (1/1,000) and subsequent alkaline phosphatase-conjugated anti-rabbit antibody (1/5,000) were used as secondary and third antibody. Human ␣ 1 -AGP was measured by nephelometry using a goat anti-human ␣ 1 -AGP polyclonal antibody.
Statistical Analysis-Survival was scored and evaluated using a log rank test. Final lethality was scored using a 2 test. The statistical significance of the number of bacterial colonies in the blood and the organs, as well as of induction of ␣ 1 -AGP after injection of IL-1 or turpentine, was determined using a Dunnett analysis of variance test. p Ͻ 0.05 was considered statistically significant. *, **, and *** represent p Ͻ 0.05, p Ͻ 0.01, and p Ͻ 0.001, respectively.

RESULTS
Protective Effect of Turpentine against a Lethal Infection with K. pneumoniae-To investigate whether turpentine conferred protection, mice were pretreated with 100 l of turpentine subcutaneously 24 or 48 h before a lethal bacterial challenge (10 6 CFU unless otherwise stated) of K. pneumoniae. When turpentine was given 48 h before the challenge, mice were significantly (p Ͻ 0.0001) protected versus PBS-pretreated controls. In contrast, turpentine treatment 24 h before the challenge was unable to induce significant protection (p ϭ 0.1137; not significant). Mice were observed for 6 days, after which no further deaths occurred (Fig. 1).
Protective Effect of IL-1 to a Lethal Infection with K. pneumonia-To determine the time point of optimal protection by IL-1, 1 g of IL-1 was given at 12, 24, and 48 h before the lethal challenge (Fig. 2). IL-1 protected significantly when given 48 h (p ϭ 0.0279), 24 h (p ϭ 0.0020), or 12 h (p Ͻ 0.0001) prior to the challenge, compared with PBS-pretreated controls. IL-1 protected significantly better at Ϫ12 h compared with Ϫ48 h (p ϭ 0.0455). Induction of ␣ 1 -AGP by Turpentine or IL-1-To study the kinetics of induction of ␣ 1 -AGP by turpentine or IL-1, mice were injected subcutaneously with 100 l of turpentine or intraperitoneally with 1 g of IL-1. At different time points after the injection, mice were sacrificed, and serum was collected. The concentration of ␣ 1 -AGP in the serum was measured as described under "Materials and Methods." The ␣ 1 -AGP levels in the serum were induced significantly 48 h (p Ͻ 0.01) and 72 h (p Ͻ 0.01) after turpentine injection compared with the basal levels (PBS injected mice had no increased ␣ 1 -AGP at any time point) and 12 h (p Ͻ 0.01), 24 h (p Ͻ 0.01), and 48 h (p Ͻ 0.05) after IL-1 injection. The levels after 12 h were significantly higher than those after 48 h (p ϭ 0.0455). The ␣ 1 -AGP levels 144 h after turpentine or IL-1 injection had reached basal levels again (p Ͼ 0.05; not significant) (Fig. 3).
Protection of ␣ 1 -AGP against a Lethal Bacterial Challenge-To test whether ␣ 1 -AGP was able to protect, several doses were injected intraperitoneally 2 h before and/or 24 h after a lethal challenge with K. pneumoniae. When mice were treated with 10 mg of ␣ 1 -AGP 2 h before the lethal challenge or with 10 mg of ␣ 1 -AGP spread over two injections (5 mg 2 h before and 5 mg 24 h after the lethal challenge), significant protection was observed (p ϭ 0.0094 and p ϭ 0.0006, respectively) compared with the control mice that received PBS as pretreatment (Fig. 4). When mice were given a total dose of 5 mg, protection was only marginal, whereas a dose of 2 mg did not protect. Furthermore, pretreatment with ␣ 1 -AGP proved to be necessary because mice treated with 10 mg of ␣ 1 -AGP 24 h after the lethal challenge were not protected (data not shown).
In a separate experiment (Table I), we observed that 10 mg of BSA was not able to protect against 10 6 CFU of K. pneumoniae, which illustrates the specificity of the effect of ␣ 1 -AGP.
To compare the serum levels of ␣ 1 -AGP induced by turpentine or IL-1 on the one hand and those after injection of 10 mg of ␣ 1 -AGP on the other hand, we performed the experiment shown in Table II. Mice were injected with 10 mg of human ␣ 1 -AGP, and ␣ 1 -AGP was measured in the blood after 2, 8, 24, 48, and 72 h. It was found that, taking into account a blood volume of 2 ml and a serum volume of 1 ml, 2 h after injection 2.3 mg of ␣ 1 -AGP/ml of serum was present, which is comparable with the amounts found after injection with turpentine or IL-1 (Fig. 3).
Spread of K. pneumoniae in Blood and Different Organs-To shed light on the mechanism of protection against a lethal challenge with K. pneumoniae, we studied the number of bac-teria in the blood and in different organs. Mice received either 10 mg of ␣ 1 -AGP (5 mg at 2 h before and 5 mg at 24 h after the lethal challenge), 1 g of IL-1 24 h before the lethal challenge, or 100 l of turpentine 48 h before the lethal challenge. Mice treated with ␣ 1 -AGP, IL-1, or turpentine showed significantly less bacteria in the blood (p Ͻ 0.001, p ϭ 0.0087, and p ϭ 0.0021, respectively), liver (p ϭ 0.0058, p ϭ 0.0038, and p ϭ 0.0038, respectively), spleen (p ϭ 0.0021, p ϭ 0.0135, and p ϭ 0.0010, respectively), and kidney (p ϭ 0.0079, p ϭ 0.0312, and p ϭ 0.0046, respectively) compared with control mice 36 h after an intramuscular injection of 1 ϫ 10 6 CFU of K. pneumoniae (Fig. 5). These data suggest that the mechanism of protection by all of these agents could be at the same level and that ␣ 1 -AGP may mediate both the IL-1 and turpentine effects. It is also clear from Fig. 5 that turpentine, which induces higher amounts of ␣ 1 -AGP than IL-1, is better in reducing the spread of bacteria.
It has already been shown that IL-1 is not directly cytotoxic to bacteria (9). We have tested the effect of ␣ 1 -AGP by plating out bacteria on nutrient agar plates containing different concentrations of ␣ 1 -AGP (1.5, 0.5, 0.15 mg/ml or none). We found no difference in the number of bacteria grown in the presence or absence of ␣ 1 -AGP (data not shown).
Lethal Response of Rat-␣ 1 -AGP Transgenic Mice-A challenge of 10 5 bacteria results in almost 100% lethality over a period of 2 weeks. When heterozygous ␣ 1 -AGP transgenic mice and wt littermates were challenged with bacteria, the transgenic mice were significantly protected compared with the wt littermates (Fig. 6). The constitutive ␣ 1 -AGP levels in these heterozygous transgenic mice amount to 2 mg/ml (38). This result was reproduced four times. Control mice started dying at day 2 and transgenic mice at day 5. Eventually, 83% control mice and 39% transgenic mice succumbed (p ϭ 0.0002). At a higher inoculum, protection by the transgenic ␣ 1 -AGP was no longer observed (data not shown).
Bacterial Clearance in Rat-␣ 1 -AGP Transgenic and Control Mice-As described before, 24 -48 h after challenge, bacteria are found in most organs (9), most pronounced in the spleen (39). We injected ␣ 1 -AGP transgenic mice and control littermates and counted the bacteria in blood, spleen, liver, and kidney 36 h later. The numbers of CFU are expressed per g of tissue (Fig. 7). We found that in the blood, spleen, liver, and kidney, ␣ 1 -AGP transgenic mice had significantly several 100fold less invasion of bacteria (p ϭ 0.0432, p ϭ 0.0033, p ϭ  0.0437, and p ϭ 0.0256, respectively).

Necrosis in Spleens of Rat ␣ 1 -AGP Transgenic and Control
Mice-2 days after a challenge of 10 5 K. pneumoniae, transgenic as well as control mice were sacrificed and their tissues fixed. Tissue sections revealed that mainly the spleen of control mice was heavily damaged (Fig. 8). In these mice, massive influx of bacteria was observed, along with erythrocytes. Also, large necrotic areas, apoptosis and blood clots were found. ␣ 1 -AGP-transgenic mice were completely protected from these lesions (Fig. 8). DISCUSSION Septic shock is the result of an overwhelming Gram-negative bacterial infection (1). The bacteria and/or their lipopolysaccharides are found in the circulation, and shock is the result of the combined action of several proinflammatory cytokines induced in macrophages and other cells (40). Septic shock and sepsis, leading to cardiovascular depression and multiple organ failure, are causing more than 100,000 casualties per year in the FIG. 4. Protective effect of bovine ␣ 1 -AGP against a lethal injection of K. pneumoniae. 10 mg of ␣ 1 -AGP was given intraperitoneally 2 h before (q) the challenge with 10 6 CFU of K. pneumoniae (n ϭ 5). Two doses of 5 mg of ␣ 1 -AGP were given 2 h before and 24 h after (f) the challenge (n ϭ 10). Controls (OE) received saline as pretreatment (n ϭ 15), ** p Ͻ 0.01; ***, p Ͻ 0.001.   3. Induction of ␣ 1 -AGP in mouse serum after turpentine or IL-1 injection. 0.1 ml of turpentine (OE) was injected subcutaneously in five groups of mice (n ϭ 6). 12, 24, 48, 72, and 144 h after the injection, mice were sacrificed, serum was collected, and the concentration of ␣ 1 -AGP was determined. In another five groups of mice, 1 g of IL-1 (q) was injected intraperitoneally. At 6 (n ϭ 3), 12 (n ϭ 5), 24 (n ϭ 5), 48 (n ϭ 3), and 144 h (n ϭ 3) after the injection, mice were sacrificed, serum was collected, and the concentration of ␣ 1 -AGP was determined. 1 ml of PBS (f) was injected in seven groups of mice (n ϭ 3). 0, 6, 12, 24, 48, and 144 h later, mice were sacrificed, serum was collected, and the concentration of ␣ 1 -AGP was determined. NS, not significant. *, p Ͻ 0.05; **, p Ͻ 0.01.
United States (41). Several clinical trials had the proinflammatory cytokines as a target: the IL-1 receptor antagonist, soluble TNF receptors, or TNF-neutralizing antibodies were used. These approaches and also those inhibiting platelet-activating factor or bradykinin were relatively disappointing (42).
Clearly new and other approaches are needed.
The toxic and lethal effects of a variety of infections or of bacterial endotoxins can be reduced by preadministration of a small dose of endotoxin 6 -48 h before (4). Several other substances of natural origin or synthetic compounds based on microbial structures also increase the natural endogenous resistance mechanisms. The protective effects are clearly nonspecific because endotoxin from Gram-negative bacteria can protect against lethality caused by an antigenically unrelated organism such as Candida albicans (43). Therefore the phenomenon was called induction of "nonspecific resistance to infection" (11). Administered under adequate conditions, such bacterial agents markedly increase nonspecific immunity, even against strains made resistant to antibiotics (4). However, because of their toxicity, these immunomodulatory substances have not gained acceptance as a therapy in humans. Therefore, further investigation is needed to obtain a clear insight into the mechanism of induction of nonspecific immunity. Several investigators found that IL-1 plays a crucial role in inducing nonspecific resistance to infection because it can protect mice against a lethal challenge of Pseudomonas aeruginosa, Listeria monocytogenes, C. albicans, and K. pneumoniae (7)(8)(9)(10)44). It was also shown that pretreatment with IL-6 or TNF could protect against a lethal bacterial challenge, although these   4). 36 h after an intramuscular challenge with 10 5 CFU of K. pneumoniae, blood and tissues were collected, homogenized, and plated, after which bacteria were counted. Logarithms of the number of bacteria are expressed per g of tissue. The ratios of control mice to transgenics for blood, liver, spleen, and kidney are 440, 324, 136, and 512, respectively. *, p Ͻ 0.05; **, p, Ͻ 0.01. cytokines were clearly less potent than IL-1 (9,44). These cytokines all play an important role in inducing the acute phase response. The acute phase response is the answer of the organism to disturbances of homeostasis caused by infection, tissue injury, or immunological disorders. It consists of a local reaction at the site of injury characterized by a number of responses, such as platelet aggregation and activation of leukocytes, which in turn release acute phase cytokines (such as TNF, IL-1, and IL-6). In addition, activated fibroblasts and endothelial cells are able to produce cytokines such as IL-6. These cytokines act on specific receptors on different target cells leading to a systemic reaction characterized by fever, leukocytosis, increase in secretion of glucocorticoids, activation of complement, and clotting and dramatic changes in the concentration of some plasma proteins called acute phase proteins (45). The acute phase response is a fundamental protective system that has evolved (21). The expression of several proteins is augmented by the liver several hours and days after the start of an infection or trauma. Most of these proteins are supposed to have protective or healing activities. The set of acute phase proteins varies from one species to another. The most predominant acute phase proteins in the mouse are serum amyloid A, serum amyloid P, and ␣ 1 -AGP (15). ␣ 1 -AGP is a glycosylated protein with a molecular mass of ϳ43 kDa and a pI of 2.7 (14). Human ␣ 1 -AGP contains five N-glycosylation chains (46) containing sialyl Lewis x structures (47). It is produced mainly by hepatocytes as a response to IL-1 and IL-6 (22). During an acute phase condition the glycosylation pattern changes (48), and the serum concentration of ␣ 1 -AGP rises up to 5-fold. Although the precise biological function of the protein is not known, most in vitro experiments suggest an anti-inflammatory role. It inhibits platelet aggregation (23), activation of neutrophils (24,25), and the proliferative response of peripheral blood lymphocytes to phytohemagglutinin (49). In vivo, ␣ 1 -AGP has been shown to protect against TNFinduced lethal shock and hepatitis (50).
In the present study, we have investigated the hypotheses that (i) induction of an acute phase reaction by turpentine would enhance nonspecific resistance to infection among others by inducing IL-1; and (ii) ␣ 1 -AGP would be a mediator in turpentine-and IL-1-induced nonspecific resistance to infection in mice. This work is an extension of previous research revealing that both IL-1 and ␣ 1 -AGP protect against a lethal injection of TNF in mice (28,50).
We observed that injection of turpentine completely pro-tected when it was given 48 h before a lethal challenge with K. pneumoniae. Because turpentine is a strong inducer of the acute phase response, we believe that induction of acute phase proteins could be responsible for the protection.
In agreement with others, we have shown that IL-1 was able to protect when it was given 24 h before a lethal challenge with K. pneumoniae (10). We observed that IL-1 still protected when it was given 12 or 48 h before the lethal challenge. Because IL-1 is a strong inducer of a subset of acute phase proteins, we believe that protection by IL-1 is also mediated by induction of acute phase proteins. Moreover, IL-1-induced protection has been reported to be blocked by coadministration of galactosamine, a specific hepatotoxin (51).
Furthermore, we found that ␣ 1 -AGP, a major acute phase protein in the mouse, protects significantly when given 2 h before a lethal challenge with K. pneumoniae. Protection was not observed with an equal dose of another protein, BSA. To investigate whether turpentine-or IL-1-induced protection could be mediated by induction of ␣ 1 -AGP, we measured the induction of ␣ 1 -AGP at different time points after injection of turpentine or IL-1. Significant induction of ␣ 1 -AGP was found 48 -72 h after turpentine and 12-48 h in the case of IL-1 injection. Remarkably, the time interval between administration of turpentine (Ϫ48 h) or IL-1 (Ϫ12 or Ϫ48 h) and the lethal challenge with K. pneumoniae corresponded with the time needed to obtain optimal induction of ␣ 1 -AGP in mouse serum. Moreover, we observed that turpentine pretreatment 24 h before the lethal challenge did not protect significantly. This is in agreement with the fact that no significant induction of ␣ 1 -AGP was found 24 h after turpentine injection. Protection of IL-1 at Ϫ12 h is significantly better compared with IL-1 at Ϫ48 h. The protection data are in line with the induction data. We believe that the data can be explained quite logically: when mice are pretreated with IL-1 and challenged with bacteria 12 h later, most of the ␣ 1 -AGP is induced during the bacterial infection. When mice are pretreated with IL-1 at 48 h prior to the challenge, then most of the ␣ 1 -AGP has gone by the time of bacterial challenge, and only minute amounts of ␣ 1 -AGP are present during the infection, so less protection will be observed. In that view, the intermediate protection of the Ϫ24 h group is also explained. The data suggest that a minimal exposure time to ␣ 1 -AGP is required. Because it was described that induction of IL-1 reaches peak values 24 -48 h after turpentine injection (52) and that the type I IL-1 receptor is responsible for the hepatic acute phase protein response following turpentine or IL-1 injection (53-55), we suggest that turpentine-induced protection is caused by the induction of IL-1 and ultimately, at least partially, by induction of ␣ 1 -AGP. By studying the clearing of ␣ 1 -AGP from the serum of mice, we found that injection of a protective dose of ␣ 1 -AGP (10 mg) leads to serum levels at the time of challenge (2.3 mg/ml) which are comparable to those that are obtained after injection with turpentine (2.8 mg/ml) or IL-1 (2.0 mg/ml).
Because protection induced by ␣ 1 -AGP was less impressive compared with IL-1 or turpentine, we surely cannot exclude a role for other acute phase proteins in IL-1-or turpentineinduced nonspecific resistance to infection.
To clarify the mechanism of protection, we studied the spread of bacteria in the blood and in different organs. The number of bacteria in the blood, liver, kidney, and spleen 36 h after infection with a lethal dose of K. pneumoniae was significantly lower when mice were pretreated with a protective dose of turpentine, IL-1, or ␣ 1 -AGP. This result contradicts the data obtained by van der Meer et al. (10), who found no difference in the number of bacteria. A possible explanation for these contradictory results could be the fact that they looked at the spread of bacteria 24 h after the lethal challenge, whereas we looked 36 h after the lethal challenge. Finally, van der Meer et al. (10) showed that IL-1 had no direct cytotoxic effect on K. pneumoniae. We obtained similar results using ␣ 1 -AGP.
We conclude that turpentine, IL-1, and ␣ 1 -AGP protect against a lethal challenge of K. pneumoniae. Because galactosamine blocks IL-1-induced protection and because the time points of optimal protection by turpentine or IL-1 and the time points of peak induction of ␣ 1 -AGP after turpentine or IL-1 injection coincide, we argue that ␣ 1 -AGP is a possible mediator in turpentine-and IL-1-induced protection. We also found that the number of bacteria in the blood and in different organs was reduced significantly in mice pretreated with turpentine, IL-1, or ␣ 1 -AGP. Because we have shown that ␣ 1 -AGP has no direct antimicrobial effect, the reduced number of bacteria in pretreated mice has to be explained by another mechanism.
Furthermore, we demonstrate that overexpression of ␣ 1 -AGP, in mice, protects against a lethal Gram-negative infection. Heterozygous transgenic mice were back-crossed to C57BL/6 mice, and the female offspring differing only in the rat ␣ 1 -AGP gene were used in the experiments. We found that the transgenic mice were relatively resistant in this sepsis model. The mice were significantly (p ϭ 0.0002) protected. The heterozygous transgenic mice used in the experiment constitutively express 2 mg/ml ␣ 1 -AGP/ml of serum, which is comparable with the amounts found after turpentine injection, IL-1 injection, or ␣ 1 -AGP injection in wt animals. However, the transgenic mice repeatedly were not protected against a 10-fold higher inoculum. We also found that blood as well as tissues of the transgenic mice contained several 100-fold less bacteria than the wt littermate mice and that the spleen, the most important target organ of the bacteria, showed gross lesions associated with fibrin clots, necrosis, and apoptosis in wt littermates only.
In most experiments, bovine ␣ 1 -AGP was used. However, human ␣ 1 -AGP was also clearly active in our system (data not shown). The glycosylation of human and bovine ␣ 1 -AGP consists of sialyl Lewis x structures, and these are absent on the transgenic ␣ 1 -AGP because it is well known that mice are unable to synthesize sialyl Lewis x structures, based on their deficiency in ␣ 3 -fucosyltransferase (56), an essential enzyme in the synthesis of sialyl Lewis x structures. Clearly, transgenic ␣ 1 -AGP is protective, so the protection is not mediated by sialyl Lewis x structures.
We hypothesize that the effect of ␣ 1 -AGP is related to its effect on the endothelial cells. It was described that ␣ 1 -AGP, to a certain degree, is able to control the transport of small and large molecules through the endothelium to the subendothelial spaces (27). We believe that this activity works in both directions, such that the invasion of the blood vessels by bacteria is prevented (to some degree) by ␣ 1 -AGP. The reduced numbers of bacteria in the tissues support the hypothesis that this could indeed be the mechanism by which ␣ 1 -AGP protects against lethal Gram-negative infection. We believe that our data open new possibilities for the treatment of Gram-negative bacterial infection and sepsis.