K-antigen-specific, but not O-antigen-specific natural human serum antibodies promote phagocytosis of Klebsiella pneumoniae


*Corresponding author. Present address: Katharinenhospital, Kriegsbergstr. 60, 70174 Stuttgart, Germany, E-mail address: m.trautmann@katharinenhospital.de


Infections due to Klebsiella pneumoniae and other Klebsiella spp. are a leading cause of hospital-associated morbidity, especially in the intensive care setting. In this study, the hypothesis that normal human sera contain sufficient concentrations of K-antigen-specific antibodies to promote phagocytic killing of encapsulated, highly virulent Klebsiella organisms was tested. K2-antigen-specific IgG and IgM antibodies were detected in each of 10 normal sera, and such antibodies were functionally active in a phagocytic killing assay. Phagocytosis depended critically on sufficient numbers of neutrophils and was impaired by the presence of soluble Klebsiella capsular polysaccharide (CPS). Thus, insufficient numbers of neutrophils and circulation of soluble CPS but not lack of K-specific antibodies may be detrimental in Klebsiella sepsis. The efficacy of hyperimmune sera might be based not on enhancement of phagocytosis but on the neutralization of these detrimental effects of circulating CPS and LPS.


Infections due to Klebsiella pneumoniae and other Klebsiella spp. are an important cause of hospital-associated morbidity, especially in the intensive care setting. Both pneumonia and septicemia due to Klebsiella spp. are characterized by frequent complications and mortality rates of up to 40%[2,4,11]. Since antibiotic therapy alone provides only partial protection against Klebsiella-associated mortality we and others explored potential options for an adjuvant humoral immunotherapy of Klebsiella infections [10,12,22]. In the clinical setting, K-antigen-specific antibodies have been shown to exert a high degree of protection against bacteremic Klebsiella infection [5]. If the infecting Klebsiella strain belonged to a K-antigen serogroup represented in the hyperimmune globulin used, the degree of protection against infection-related death was as high as 50%[5]. In the present study we test the hypothesis that normal human sera contain sufficient concentrations of K-antigen-specific antibodies to promote phagocytic killing of encapsulated, highly virulent Klebsiella organisms. It became evident that K-antigen-specific antibodies are, in fact, present in sufficient concentrations in normal human serum of immunocompetent patients to opsonize encapsulated Klebsiella organisms. However, in case of an infection, their action may be impaired by circulating soluble capsular antigen liberated by Klebsiella organisms. To test these hypotheses, we determined the concentration of natural human serum antibodies directed against the K2 capsular antigen which is frequently encountered in clinical Klebsiella infections [3,6]. For comparison we examined antibody concentrations against the most frequently encountered somatic antigen of Klebsiella, O1 [8,21]. The functional activity of the antibodies was studied using an opsonophagocytic killing assay. Specific IgG and IgM antibody concentrations were determined by means of a standardized enzyme-linked immunosorbent assay (ELISA), as described previously [20].

2Materials and methods

2.1Human serum

Whole blood samples were obtained from 10 healthy volunteers (five male, five female, mean age 29.4 (range 21–47) years). Serum was prepared immediately after blood was taken by centrifugation in 15 ml Falcon tubes (Becton and Dickinson, New Jersey, USA) for 5 min at 2500 rpm in a Heraeus Sepatech (Heraeus, Germany) centrifuge.

Natural human serum (NHS) was stored at 4°C. No serum components were eliminated.


As solid phase antigens, we used the Klebsiella K2 capsular and the O1 somatic antigen of Klebsiella. LPS was extracted by the hot-phenol water method and purified as described previously [19], solutions of the extracted LPS were made by dissolving 1 mg of K2 capsular and O1 somatic LPS in phosphate buffer plus sodium (PBS) and stored at 4°C. Polystyrene 96-well microtiter plates (Greiner, Frickenhausen, Germany) were coated in each well with 100 μl of a 50 μg ml−1 solution of O1-LPS and K2-CPS respectively. Antigen-coated plates were incubated overnight at 4°C and washed with PBS three times afterwards. Then, plates were treated with 200 μl of a PBS-based buffer (filling buffer) containing 0.5% bovine serum albumin and 0.5% casein for 1 h at 37°C to block unspecific antibody bindings. Coating antigen concentration was determined by the use of a standard curve (plotting of optical density A405 (OD) – read with a Milenia Kinetic Analyzer, DPC, Bad Nauheim, Germany – against increasing concentrations of antigen) as described previously [17]. The same was done for the determination of natural human serum antibody content. For standard curves human 96-well plates were coated with 100 μl Fab-specific anti-human IgG and IgM antibodies at a concentration of 50 μg ml−1, overnight at 4°C. Afterwards plates were washed three times with PBS and blocked with filling buffer for 1 h at 37°C before being washed again with PBS. IgG and IgM from pooled human serum, obtained from Sigma, Deisenhofen, Germany were diluted serially down to 0.015 μg ml−1, 100 μl thereof were added. Incubation time was 4 h at 4°C. A 1:3000 dilution of alkaline phosphatase conjugated anti-human IgG specific for IgG or IgM heavy chains was then added for 4 h, depending on the coating and on which immunoglobulin was added before. Test serum was diluted 1:2 in filling buffer and 100 μl were added to each well. For the standard curves OD was plotted against concentration of the standard dilutions. Natural human antibody content was measured by incubation of serum with O1-LPS- and K2-CPS-coated wells as described above. OD was measured and plotted to the standard curves. Only the linear region of the curve was used to determine the specific antibody content. Control wells without serum were included and finally their value was subtracted from the test values. To test the natural human serum antibodies for specificity, we used the LPS from Escherichia coli O18 which is known to be serologically not related to Klebsiella O1 LPS [18] and other preparations (K1 CPS, lipid A from Salmonella minnesota Re 595 LPS, and E. coli J5 Rc mutant LPS (all obtained from Sigma, Deisenhofen, Germany)) as inhibitors. Pooled human serum was incubated with the inhibitors on ice for 1 h being vortexed every 15 min, at different concentrations of the indicated LPS and CPS preparations (final concentration 50 μg ml−1 in PBS, pH 7.4) and ELISAs were run as described [17,22]. All samples were run in duplicate. Experiments were done in triplicate.

2.3Opsonophagocytic killing assays

Opsonophagocytic killing assays using human neutrophils were performed by means of a microplate method as described by Cryz et al. [24]. Neutrophils were collected using Ficoll-Hypaque sedimentation and dextran. Erythrocytes were lysed hypotonically and neutrophils were washed twice and re-suspended. Bacteria of strain K. pneumoniae CDC 2–70 (O1:K2), a former clinical isolate [1], were grown in Mueller–Hinton broth to mid-log phase, washed three times in ice-cold physiological saline, and adjusted photometrically to a concentration of 1×107 organisms per ml. Final phagocytosis mixtures were composed of 15 μl of the bacterial suspension (final concentration 1×106 per ml), 60 μl of pooled NHS from the 10 donors (final concentration 40%), 60 μl of the neutrophil suspension (final concentration 2×107 per ml), and 15 μl of Hank's balanced salt solution (HBSS), yielding a total volume of 150 μl. Finally, the mean ratio between leukocytes and bacteria was 20:1. The reaction mixtures were incubated at 37°C with gentle shaking. For viable count determination, samples were obtained at 0 and after 120 min. The samples were diluted and plated on Mueller–Hinton agar and grown for 18–24 h, afterwards colonies were counted manually. The opsonophagocytic index was expressed as percent of the original bacterial inoculum killed after 120 min. The formula for this calculation was as follows: % of phagocytotic killing=1.0−(CFU120)/(CFU0)×100. The relative contribution of O- and K-specific antibodies to phagocytosis was examined by adding purified O1 LPS or K2 CPS at final concentrations of 1 μg ml−1 in HBSS at the beginning of each experiment. For K2 CPS, we performed a titration curve with concentrations ranging from 0.016 to 1.0 μg ml−1. Control preparations included LPS from E. coli O18 as well as serotype K1 CPS. To exclude complement-mediated phagocytosis, we included a control well containing serum and the test strain without leukocytes.

2.4Statistical analysis

After verification of a normal distribution of data by multi-variance analysis (ANOVA), the Tukey test (for pairwise comparisons of the mean values of the different groups) was used to test for differences between the groups. Differences were judged significant if P≤0.05[7]. All analyses were performed using the program SigmaStat 2.03, SPSS.


Normal human sera, obtained from 10 healthy donors, contained both K2- and O1-antigen-specific natural IgG and IgM antibodies against Klebsiella (Fig. 1). As shown in Fig. 2, these antibodies proved to be specific for these antigens because optical density (OD) values fell significantly after pretreatment of pooled serum with homologous, but not heterologous antigens. Interestingly, reactions with O1 LPS were not inhibited by lipid A or core LPS (E. coli J5), which indicates that Klebsiella natural O-antigen-specific antibodies were directed mainly against serogroup-specific side chain epitopes (Fig. 2). In preliminary experiments, we found that 40% NHS and a unusually high ratio of leukocytes to bacteria of 20:1 was necessary to obtain optimal phagocytic killing of the highly virulent and fully encapsulated Klebsiella strain CDC 2–70 (data not shown). No killing occurred in experiments performed without addition of leukocytes, showing that phagocytic uptake was in fact a prerequisite for killing (data not shown). Spontaneous killing of the test strain by serum alone, e.g. complement-mediated killing, was thus excluded because abundant growth occurred in this well in each experiment. This is in accordance with data from Alberti et al. [25] who showed that strains harboring a smooth O1 serotype LPS are serum-resistant, independently of their K-type.

Figure 1.

Antibody levels in 10 normal human sera against Klebsiella K2 CPS and O1 LPS. Triangles denote mean, whiskers denote 1 S.D. Black dots denote minimum and maximum.

Figure 2.

ELISA reactions of pooled NHS with Klebsiella K2 CPS (A, B) or O1 LPS (C, D). Pooled NHS from 10 donors was mixed with equal volumes of PBS (no inhibitor) or solutions of LPS (final concentration 50 μg ml−1) in PBS as indicated, before adding the mixtures to antigen-coated wells. Reactions were developed using anti-human IgG (A and C) or anti-human IgM (B and D) alkaline-phosphatase conjugates as secondary antibodies. Asterisks denote significant inhibitions (P<0.05).

By adding inhibitors to the experimental system, we attempted to delineate the respective role of K-antigen- versus O-antigen-specific natural antibodies as opsonins for encapsulated Klebsiella. Addition of K2 CPS inhibited phagocytic killing in a dose-dependent manner, while addition of K1 CPS, O1 LPS or control preparations did not (Fig. 3). There was a statistically significant difference between K2 CPS and the three other antigen preparations (P<0.001).

Figure 3.

Inhibition of phagocytic killing of strain CD 2–70 in the presence of K1 CPS, K2 CPS, O1 LPS and control inhibitors. K2 CPS completely abolished phagocytic killing at ≥0.25 μg ml−1. Asterisk denotes significant inhibition (P<0.05). Without neutrophils, no killing occurred.

4Discussion and conclusions

Our data show that normal human sera contain antibodies specific for both the somatic and capsular antigens of Klebsiella O1:K2, which represents a serogroup found frequently in nosocomial infections [3]. Such antibodies may be the result of intestinal exposure to surface determinants of Klebsiella since it is known that these organisms occur in the environment and may be a transient component of the normal human bowel flora. Even with a virulent and highly encapsulated strain such as Klebsiella CD 2–70, effective phagocytic uptake and killing can be obtained in the presence of natural K-specific antibody alone.

Complement-mediated killing, in general important for the clearance of bacteria, might not be a key in our system. K. pneumoniae may activate both, the classical (CP) and the alternative pathway (AP). Both pathways are important to kill. Serum-resistant strains fail to activate the CP and thus avoid being killed. In this study, we used a smooth O1-LPS strain, which has been characterized in an earlier study [21]. It has been shown that smooth LPS inhibits the deposition of lytic complement components [25]. Further, resistance to the bactericidal capacities of complement depended critically on the presence of O-antigen polysaccharide chains [26]. One explanation may be that the O side chain prevents C1q from binding certain porins in the bacterial membrane and further C3b binding [25,27]. We confirmed this for our assay by the use of controls lacking neutrophils. In these wells no killing occurred, indicating the serum resistance of the strain used. Additional supplementation of humoral antibody may not be required in non-immunodeficient patients. Rather, the presence of sufficient concentrations of functional neutrophils may represent the critical factor for killing of Klebsiella, because a 20:1 ratio of leukocytes to bacteria was needed in our experiments to induce significant phagocytic uptake and killing in our experiments. Furthermore, soluble K-antigen may disturb the effect of K-specific antibody, and this may be true for both natural and hyperimmune antibodies. It is known that circulating K-antigen polysaccharide can be detected in approximately 25% of bacteremic Klebsiella infections, with concentrations ranging from 0.25 to >15 μg ml−1[14]. A concentration of >0.25 μg ml−1 was able to completely inhibit phagocytic killing in the in vitro phagocytic assay (Fig. 3). Taken together, these factors may explain the results obtained with K-specific hyperimmune globulin in the clinical setting. Nevertheless, it has been shown previously, with the same phagocytosis assay, that hyperimmune serum is advantageous as compared to normal human serum [24]. The hyperimmune preparation against Klebsiella used in this study was superior to seven intravenous immune globulin (IVIG) preparations and provides protection at significantly lower doses.

Our data show also that naturally occurring O-antigen-specific antibodies which are directed mainly against LPS side chain epitopes do not appear to contribute significantly to phagocytic killing. These data are confirming data by Cortes et al. [23], who showed that phagocytosis of unencapsulated mutant strains was more efficient than that of wild-type strains or LPS O-side-chain-deficient strains. Although it was shown that O-specific antibodies may penetrate the capsule of Klebsiella K2 strains [13,16], we have shown previously that supplementation of human sera with O-antigen-specific monoclonal antibodies did not increase phagocytic uptake of O1:K2 strains, and to a rather limited extent the uptake of strains of other K-antigen serotype [9].

In animal experiments, we have shown that high doses of O-antigen-specific mAbs may exert protection against lethal Klebsiella infection [15]. It appears that this protective effect is due rather to neutralization of the detrimental effects of circulating Klebsiella LPS than to an enhancement of phagocytosis [15]. This approach, i.e. the neutralization of components of the organism that are responsible for the induction of septic shock [28], may be more promising in the future compared to attempts to promote phagocytic uptake of Klebsiella and might be achieved by using hyperimmune sera.


We are indebted to Prof. Dr. Martin Täuber (Institute for Infectious Diseases, University of Berne, Switzerland) for critically reading the manuscript and helpful discussion. P.M.L. was supported by a grant of the IZKF, Ulm, Germany.