点击显示 收起
Harvard-Thorndike Laboratory, Division of Allergy and Inflammation, and the Division of Infectious Disease, Department of Medicine, Beth Israel Deaconess Medical Center, and Harvard Medical School, Boston, Massachusetts
Complement-opsonized particles become immune adherent to complement receptor 1 (CR1 or CD35) on human erythrocytes, allowing particles to be ingested by phagocytes in the liver and the spleen. We investigated the role that immune adherence plays in the uptake and killing of Salmonella montevideo by human neutrophils. Exposure to serum induced loss of flagella and facilitated immune adherence, which was followed by more-efficient phagocytosis and killing, compared with that after exposure to serum-opsonized, free bacteria. One correlate of bacterial killing is the fusion of phagosomes with lysosomes, which can be monitored by LysoTracker or lysosomal-associated membrane protein 2 colocalization with the intracellular bacteria. At 5 min, phagolysosmal fusion was significantly faster for immune-adherent bacteria than for nonimmune-adherent bacteria, but, by 35 min, the difference between the 2 groups was minimal. Immune adherence also facilitated the ingestion of antibody complementopsonized fluorescent microspheres, but, unlike bacteria, most internalized microspheres failed to fuse with lysosomes. However, addition of lipopolysaccharide, a Toll-like receptor ligand, to microspheres directed their intracellular trafficking, resulting in rapid lysosomal fusion. Thus, immune adherence facilitates phagocytosis, but the route of intracellular processing depends on the molecular nature of the target and is independent of host complement and antibody.
Complement opsonization of a particle significantly augments its ingestion by phagocytes in tissues. In the vascular space, where phagocytes are relatively rare, complement-opsonized particles are immobilized to the surface of cells for clearance by phagocytes in the liver and the spleen. The process is known as immune-adherence clearance. In humans and a few other primates, it is the erythrocyte using complement receptor 1 (CR1 or CD35) that ligates the particle [1, 2].
Enteric Salmonella species, in contrast to typhoidal Salmonella species, usually cause self-limited infections that last 35 days in healthy hosts. This time course of illness suggests that innate immunity is usually able to cure these infections, although it cannot prevent them [3]. A high prevalence of nontyphoidal Salmonella bacteremia in neutropenic patients infected with HIV [4] and in patients with chronic granulomatous disease [5] indicates that sufficient numbers of functional neutrophils (polymorphonuclear leukocytes ) are critical for the human innate immune defense against these organisms. Additionally, recent evidence from studies of mice indicate that sinusoidal PMNLs in the liver and, presumably, the spleen are responsible for clearance of bacteria [68].
In the present study, we show that immune-adherent Salmonella species are more efficiently ingested and killed by PMNLs than are equally opsonized nonimmune-adherent Salmonella species. Although immune adherence also enhances phagocytosis of nonbiological particles, the ability to fuse with lysosomal compartments is solely dependent on the molecular composition of the particles and is independent of their immune-adherence status.
PARTICIPANTS, MATERIALS, AND METHODS
Reagents.
The following reagents were obtained as noted: Hanks' balanced salt solution (HBSS) with calcium and magnesium but without phenol red (Gibco), HBSS with 0.1% bovine serum albumin (BSA) (Sigma), gentamicin sulfate solution (10 mg/mL) (United States Biological), sterile MilliQue water (Millipore) filtered by use of a lipopolysaccharide (LPS)free Acrodisc filter (Pall), LysoTracker red, CellTracker green, Syto16 green, Hoescht 33342, Prolong Gold with DAPI, and Alexa 488BSA. A Via Gram staining kit (Molecular Probes) was also used.
Antibodies.
The following antibodies were used in the present study: mouse antilysosomal-associated membrane protein (LAMP) 2 monoclonal antibody (MAb) (BD Pharmingen), rabbit anti-Salmonella immunogloubulin (Fitzgerald; lot P02050202), fluorescein isothiocyanate (FITC)labeled rabbit antiSalmonella flagella antigen (YVS 0703; Accurate Chemical and Scientific), rabbit anti-BSA IgG, Alexa 488 and Alexa 594conjugated goat anti-mouse IgG, Alexa 488conjugated goat anti-rabbit IgG (Molecular Probes), rabbit nonimmune rabbit IgG (Jackson-ImmunoResearch), and antigen affinity-purified rabbit anti-CR1 [9].
Salmonella species and microspheres.
S. montevideo (American Type Culture Collection) was grown overnight in Bacto nutrient broth (Difco) and quantified (0.5 OD600 = 4.5 × 108 cells/mL). Bacteria were gently pelleted, washed, and resuspended in HBSS. Wild-type (wt) S. typhimurium (SJW1103) and mutant strains either lacking flagella (SJW1368) or possessing flagella that could not rotate (SJW2960) were provided by May Macnab (Yale University) [10]. Sulfate-latex microspheres (1.0 mol/L) (Interfacial Dynamics) were coated sequentially with Alexa 488BSA (Molecular Probes), rabbit anti-BSA IgG, and, in some experiments, LPS (Escherichia coli; L3755; Sigma).
Human serum and cells.
Normal human serum (NHS) was obtained from healthy volunteers, and heat-inactivated serum (HIS) was prepared by heating NHS for 50 min at 56°C. Erythrocytes and PMNLs were isolated as described elsewhere [11] and were used within 30 min of purification. Cells and serum were obtained from autologous donors. The protocol for drawing blood from healthy volunteers was approved by the Beth Israel Deaconess Medical Center Institutional Review Board, and all participants provided informed consent.
Assay for bacterial motility.
Bacteria (5 × 107) were incubated with either buffer, 10% HIS in HBSS, or 10% NHS in HBSS for 10 min at 37°C. Time-lapse phase-contrast images recorded the "relative" distance traveled by individual organisms during 0.75 s. Images were analyzed by use of IPLab (version 3.9; Scanalytics).
Staining for bacterial flagella.
After exposure to either HIS or NHS, bacteria were treated with FITC-labeled rabbit anti-Salmonella flagella antigen (1 : 50) and 10 g/mL Hoechst 33342 for 10 min and then were examined by fluorescence microscopy. In addition, bacteria opsonized under the standard conditions were stained as noted and analyzed by flow cytometry by use of a FACScan cytometer with CellQuest Pro software (version 4.01; BD Biosciences). Fluorescence-activated cell sorter buffer, unstained bacteria, and stained bacteria were examined by forward side scatter and fluorescence microscopy, for identification of the bacteria and determination of the gate for further analysis.
Immune-adherence protocol.
S. montevideo bacteria were grown for 12 h at 37°C in nutrient broth, and bacteria were gently washed in HBSS. A total of 5 × 107 S. montevideo bacteria were resuspended in 90 L of HBSS, 10 L of either NHS or HIS was added, and the suspension was incubated for 10 min at 37°C, for opsonization. Bacteria were washed once, 5 × 106 erythrocytes were added to 1 × 106 S. montevideo, and the mixture was incubated in HBSS for 10 min at 37°C. The percentage of erythrocytes with immune-adherent S. montevideo was determined by light microscopy and ranged from 0% to 5% for HIS and from 70% to 75% for NHS.
Phagocytic assay.
PMNLs (5 × 105) were added to microfuge tubes containing 600 L of HBSS/0.1% BSA. Serum-opsonized S. montevideo (1 × 106), either free or immune adherent, were added to the PMNLs, and the mixture was incubated for 5 min at 37°C with end-over-end rotation at 8 rpm, as described elsewhere [12]. PMNLs were counted by light microscopy and analyzed for the percentage of phagocytosis (the percentage of PMNLs that had ingested intracellular S. montevideo) and the phagocytic index (the mean number of intracellular organisms in PMNLs that had ingested at least 1 organism). Buffer-treated organisms had a phagocytic index of 0%11%, which was always less than that of HIS-opsonized organisms and was considered to be the baseline value.
Intracellular killing assay. Phagocytic mixtures were centrifuged to remove nonimmune-adherent, noningested bacteria, and cells were subjected to hypotonic saline to lyse erythrocytes in the tubes that contained them. PMNLs were resuspended in gentamicin (10 g/mL of HBSS) for 30 min at room temperature to kill extracellular, immune-adherent bacteria. Tubes were centrifuged at 3000 g for 3 min, and the PMNL pellets were suspended in 400 L of HBSS. Samples were collected immediately (45 min after the initiation of phagocytosis) and again after an additional 60 min of incubation at 37°C. Viability sampling consisted of 4 replicate 7-L aliquots, each diluted with 93 L of sterile water and plated on separate petri dishes made with blood agar base number 2 (Difco). The plates for colony counts were incubated for 18 h at 37°C.
Protocol for intracellular localization.
Purified PMNLs suspended in HBSS were loaded with 50 nmol/L LysoTracker for 30 min at 37°C, washed twice, and resuspended at a concentration of 1 × 107 cells/mL in HBSS. S. Montevideo bacteria were loaded with 100 nmol/L Syto16 green for 30 min at 4°C in HBSS, washed twice with HBSS, opsonized with NHS, and resuspended at a concentration of 1 × 108 bacteria/mL. PMNLs and opsonized bacteria with or without erythrocytes for immune adherence were mixed as described above. Aliquots (100 L) from each sample were cytospun, and slides were mounted with Prolong Gold with DAPI. Cells were examined by use of an AX-70 Provis Olympus fluorescence microscope equipped with FITC and rhodamine filters (Olympus America) and UPLApo 40 × 1.0 Ph 3 and 100 × 1.35 Ph 3 objectives. Intracellular bacteria not associated with the lysosomal compartment fluoresced in the green channel alone, whereas intralysosomal bacteria fluoresced in both the green and the red channels. Images were acquired by use of a Retiga EXi digital camera (QImaging), and the images were further processed by use of IPLab. Alternatively, cells were fixed with 3.7% paraformaldehyde (EM Sciences) for 1 h, washed twice in HBSS, and permeabilized for 10 min in 0.2% saponin (Sigma-Aldrich). Cells were blocked with HBSS plus 4% NHS and 4% normal goat serum for 15 min at room temperature and were incubated with mouse antiLAMP-2 MAb (5 g/mL) for 20 min. Cells were washed twice with HBSS and incubated for 15 min with FITC-labeled rabbit antiSalmonella flagella antigen (1 : 200). Cells were washed twice and stained with a mixture of Alexa 594conjugated goat anti-mouse and Alexa 488conjugated goat anti-rabbit IgG for 15 min. For colocalization studies, stacks of images (0.25 m/step) were acquired by use of a Ludl z-axis motor, linear encoder, and filter wheels (LEP), which were controlled by software (IPLab). z-stacks were exported into Velocity (version 3.0; Improvision) for analysis.
Statistical analysis. Results are reported as means ± SD, and differences between samples were analyzed by use of Student's t test. The level of statistical significance adopted was P < .05.
RESULTS
Effect of NHS on S. montevideo motility.
To potentially enhance immune adherence, we preopsonized the bacteria and washed them in buffer before adding erythrocytes. The removal of serum after opsonization minimizes the possibility that serum complement factor I, using CR1 as a cofactor, will degrade C3b on the bacteria and decrease immune adherence. This is not a consideration for immune-adherence clearance in vivo in humans, because clearance is very fast (t1/2 = 5 min) [13], whereas, in vitro, the release of immune-adherent particles takes 8 times longer (50% release in 40 min) [14].
Unexpectedly, after opsonization with NHS, the bacteria seemed to be less motile. To confirm this finding, the protocol was repeated, and the movement of the bacteria was recorded by time-lapse light microscopy. Bacteria exposed to NHS were significantly less motile than those exposed to buffer or HIS (figure 1). Treatment with NHS did not affect Salmonella viability, as assessed by use of the Via Gram staining kit.
To assess whether NHS-induced loss of motility and becoming immune adherent to erythrocytes were functionally linked phenomena, we performed additional studies. S. typhimurium was used because of the availability of a wt strain (SJW1103) and 2 nonmotile mutant strains: 1 lacking flagella (SJW1368) and 1 possessing flagella that were nonfunctional because of a defective motor (SJW2960). With or without exposure to NHS, the wt S. typhimurium strain remained motile, whereas the 2 mutant S. typhimurium strains were nonmotile, as determined by microscopic observation. In a quantitative assay for immune adherence, each S. typhimurium strain was exposed to HIS or NHS and then was mixed with erythrocytes and examined for immune adherence by light microscopy. Sixty-one percent of flagella-lacking S. typhimurium and 81% of nonmotile S. typhimurium, compared with only 33% of wt S. typhimurium, became immune adherent. Although the wt S. typhimurium remained motile after opsonization, there was a marked enhancement of immune adherence in the nonmotile mutant strains. Thus, the NHS-induced flagellar shedding of S. montevideo may not have been essential for immune adherence, but, likely, it greatly enhanced it.
Effect of immune adherence on phagocytosis of S. montevideo.
Various species of bacteria are phagocytosed more efficiently when immune adherent [1, 16]. To determine whether immune adherence affected the uptake of S. montevideo by human PMNLs, we compared the ingestion of opsonized immune-adherent bacteria with that of opsonized nonimmune-adherent S. montevideo. Immune-adherent S. montevideo were ingested more efficiently than were nonimmune-adherent S. montevideo, as measured by both the percentage of phagocytosis (50% vs. 29%) (figure 4A) and the phagocytic index (3.7 vs. 2.7) (figure 4B).
Effect of of S. montevideo.
The increased phagocytosis of immune-adherent S. montevideo might favor either the host or the pathogen. We therefore investigated the survival of S. montevideo within PMNLs after the bacteria in suspension were washed away and the immune-adherent, noningested bacteria were killed by gentamicin, manipulations that were completed 45 min after the initiation of phagocytosis. The viability of intracellular bacteria was assessed after 45 min and again after an additional 60 min of incubation (105 min total) at 37°C. These times were chosen because longer incubation times (120 min) resulted in all the bacteria being killed whether they had been free or immune adherent.
PMNLs were lysed, and the lysate was diluted, plated in quadruplicate, and incubated for 18 h, to allow enumeration of viable colonies. The relative number of viable intracellular organisms at 45 min (figure 5) reflected the efficiency of phagocytosis for bacteria opsonized under different conditions (figure 4A and 4B). During the subsequent 60 min of incubation, killing of previously immune-adherent S. montevideo was more efficient (60% decrease in colony-forming units between 45 and 105 min) (figure 5) than was killing of NHS-opsonized S. montevideo (27% decrease in colony-forming units during the same period) (figure 5).
Although LysoTracker is well accepted in the literature as a marker of lysosomes, there is the possibility that the dye marks other compartments with low pH. To address this concern, we repeated the experiment but replaced LysoTracker with antiLAMP-2 MAb (figure 6E) and replaced Syto16 green with rabbit anti-Salmonella antibody (figure 6D). Because LAMP-2 has a diffuse distribution in PMNLs, the clustering around ingested salmonellae or the lack of clustering was relatively easy to identify and quantify (figure 6E). Although immune-adherent bacteria were more rapidly ingested, there seemed to be a minimal effect on intracellular trafficking. To monitor more accurately the intracellular movement of phagosomes, we switched to Alexa 488BSAcoated latex beads (microspheres) that had been opsonized with rabbit anti-BSA IgG and NHS. Being perfectly spherical and birefringent, microspheres were easy to identify by fluorescence and bright-field microscopy. To control for the presence of erythrocytes in the nonimmune-adherent samples, we included an equivalent number of human erythrocytes that had been pretreated with antigen affinity-purified rabbit anti-CR1, which completely blocks immune adherence. Immune adherence stimulated the rate of phagocytosis of the microspheres (table 2). In striking contrast to the bacteria (whether free or immune adherent), at 35 min, the majority (83%) of the ingested microspheres remained in a LAMP-2negative compartment, indicating that their phagosomes had not fused with lysosomes. These data suggest that, independent of being immune adherent and independent of antibody-complement opsonins, the molecular determinants of the target itself might determine intracellular trafficking within PMNLs.
Effect of LPS, a Toll-like receptor (TLR) ligand, on phagosomes containing latex microspheres.
To test whether the TLR ligand LPS could affect trafficking of phagosomes, PMNLs were incubated with opsonized beads with or without LPS for 5 min at 37°C and then were washed twice to remove the noningested microspheres. Cells were seeded on chamber slides and fixed every 10 min for the next 35 min. Although immobilized LPS had no significant effect on the percentage of phagocytosis or the phagocytic index (LPS beads, 19%; beads, 21%), there was a striking difference (figure 7A) in the number of microspheres with fused phagolysosomes. We considered LAMP-2positive microspheres to be only those microspheres that had a well-defined structured ring around them (figure 7D), and not those microspheres that just displaced the LAMP-2positive structures around them (figure 7A).
Effect of PMNLs on the trafficking specificity of various ingested particles.
We next asked whether phagocytosis of opsonized salmonellae would induce ingested microspheres to fuse with lysosomes. We mixed (1 : 1) opsonized salmonellae and opsonized Alexa 488 microspheres (lacking LPS) with PMNLs for 5 min of phagocytosis. After washing, PMNLs were seeded on glass for 15 min. Subsequently, the cells were fixed with paraformaldehyde, permeabilized, and stained for bacteria with anti-Salmonella MAb and antiLAMP-2 MAb. A typical cell (figure 8) ingested both types of particles. Salmonellae and microspheres followed distinct pathways within the same cell: bacteria colocalized with LAMP-2, whereas microspheres remained in a LAMP-2negative compartment.
DISCUSSION
We have investigated how immune adherence affects the killing of S. montevideo. Because erythrocyte-dependent immune adherence is unique to humans and a few other higher primates, our in vitro studies used human reagents. Although salmonellae can activate complement in the absence of antibody [20], it is likely that our sources of NHS contained specific antibody, because serum samples from some donors supported immune adherence better than serum samples from other donors.
Nontyphoidal salmonellae have diverse virulence factors [21], and, thus, it was not surprising that exposure to NHS induced S. montevideo, but not S. typhimurium, to shed its flagella. Flagellar shedding not only promotes immune adherence, it also has the potential to disseminate an inflammatory response in vivo. Flagellin, the major flagellar protein (reviewed in [22]), binds TLR5 [23, 24] and, in terms of induction of proinflammatory genes, is the most important virulence factor of S. typhimurium [25].
Our experiments with antibody complementopsonized microspheres confirmed that intracellular trafficking was not affected by immune adherence: 80% of ingested microspheres remained in LAMP-2negative compartments regardless of whether they were free or immune adherent. However, the addition of a TLR ligand, LPS, to the target signaled for phagolysosomal fusion. It has been found by use of serum-opsonized green fluorescent protein E. coli as targets and mouse macrophages as effectors that an intact TLR signaling pathway leads to a higher rate of phagocytosis [27]. In contrast, we did not find that the addition of LPS to our microspheres enhanced either the percentage of phagocytosis or the phagocytic index. These discrepant results may relate to the fact that we used different targets and different effector cells. As was shown for murine macrophages [27], a single human PMNL could ingest a TLR-ligating target (bacteria) and a nonTLR-ligating particle (microsphere), and their trafficking was different: the bacteria stimulated rapid lysosomal fusion, and the microsphere did not. Thus, engagement of TLRs takes place in nascent phagosomes and signals for fusion of that particular phagosome with the lysosomal compartment.
In summary, we have found that immune adherence facilitates phagocytosis of S. montevideo and that it likely is a host defense mechanism for other nontyphoidal Salmonella strains. During the course of these experiments, 2 novel findings were noted: first, opsonization with NHS caused S. montevideo, but not S. typhimurium, to shed its flagella, a finding that has implications for TLR5 signaling; and, second, the enhanced killing of immune-adherent S. montevideo was correlated with more-rapid phagolysosomal fusion in PMNLs. Intracellular trafficking, however, was not dictated either by host opsonins or by immune adherence, but rather by the biochemical composition of the target.
Acknowledgments
We are grateful to May Macnab (Yale University) for providing the wild-type and motility mutant strains of Salmonella typhimurium and to Greg Weaver and Dr. Peter F. Weller (Beth Israel Deaconess Medical Center) for technical expertise and critical reading of the manuscript, respectively.
References
1. Nelson RAJ. The immune adherence phenomenon: an immunologically specific reaction between microorganisms and erythrocytes leading to enhanced phagocytosis. Science 1953; 118:7337. First citation in article
2. Fearon DT. Identification of the membrane glycoprotein that is the C3b receptor of the human erythrocyte, polymorphonuclear leukocyte, B lymphocyte and monocyte. J Exp Med 1980; 152:2030. First citation in article
3. Hohmann EL. Nontyphoidal salmonellosis. Clin Infect Dis 2001; 32:2639. First citation in article
4. Tumbarello M, Tacconelli E, Caponera S, Cauda R, Ortona L. The impact of bacteraemia on HIV infection: nine years experience in a large Italian university hospital. J Infect 1995; 31:12331. First citation in article
5. Winkelstein JA, Marino MC, Johnston RB Jr, et al. Chronic granulomatous disease: report on a national registry of 368 patients. Medicine (Baltimore) 2000; 79:15569. First citation in article
6. Vassiloyanakopoulos AP, Okamoto S, Fierer J. The crucial role of polymorphonuclear leukocytes in resistance to Salmonella dublin infections in genetically susceptible and resistant mice. Proc Natl Acad Sci USA 1998; 95:767681. First citation in article
7. Fierer J. Polymorphonuclear leukocytes and innate immunity to Salmonella infections in mice. Microbes Infect 2001; 3:12337. First citation in article
8. Wing EJ, Gregory SH. Listeria monocytogenes: clinical and experimental update. J Infect Dis 2002; 185(Suppl 1):S1824. First citation in article
9. Fearon DT, Collins LA. Increased expression of C3b receptors on polymorphonuclear leukocytes by chemotactic factors and by purification procedures. J Immunol 1983; 130:3705. First citation in article
10. Ohnishi K, Ohto Y, Aizawa S, Macnab RM, Iino T. FlgD is a scaffolding protein needed for flagellar hook assembly in Salmonella typhimurium. J Bacteriol 1994; 176:227281. First citation in article
11. Tyagi SR, Nicholson-Weller A, Barbashov SF, Tas SW, Klickstein LB. Intracellular adhesion molecule 1 and 2 integrins in C1q-stimulated superoxide production by human neutrophils: an example of a general regulatory mechanism governing acute inflammation. Arthritis Rheum 2000; 43:224859. First citation in article
12. Ghiran I, Barbashov SF, Klickstein LB, Tas SW, Jensenius JC, Nicholson-Weller A. Complement receptor 1/CD35 is a receptor for mannan-binding lectin. J Exp Med 2000; 192:1797808. First citation in article
13. Schifferli JA, Ng YC, Estreicher J, Walport MJ. The clearance of tetanus toxoid-anti-tetanus toxoid immune complexes from the circulation of humans: complement- and erythrocyte complement receptor 1-dependent mechanisms. J Immunol 1988; 140:899904. First citation in article
14. Jepsen HH, Svehag SE, Jensenius JC, Sim RB. Release of immune complexes bound to erythrocyte complement receptor (CR1), with particular reference to the role of factor I. Scand J Immunol 1986; 24:20513. First citation in article
15. Minamino T, Imae Y, Oosawa F, Kobayashi Y, Oosawa K. Effect of intracellular pH on rotational speed of bacterial flagellar motors. J Bacteriol 2003; 185:11904. First citation in article
16. Lindorfer MA, Nardin A, Foley PL, et al. Targeting of Pseudomonas aeruginosa in the bloodstream with bispecific monoclonal antibodies. J Immunol 2001; 167:22409. First citation in article
17. Duclos S, Desjardins M. Subversion of a young phagosome: the survival strategies of intracellular pathogens. Cell Microbiol 2000; 2:36577. First citation in article
18. Gorvel JP, Meresse S. Maturation steps of the Salmonella-containing vacuole. Microbes Infect 2001; 3:1299303. First citation in article
19. Rajotte D, Stearns CD, Kabcenell AK. Isolation of mast cell secretory lysosomes using flow cytometry. Cytometry A 2003; 55:94101. First citation in article
20. Morrison DC, Kline LF. Activation of the classical and properdin pathways of complement by bacterial lipopolysaccharides (LPS). J Immunol 1977; 118:3628. First citation in article
21. Fierer J, Guiney DG. Diverse virulence traits underlying different clinical outcomes of Salmonella infection. J Clin Invest 2001; 107:77580. First citation in article
22. Underhill DM, Ozinsky A. Toll-like receptors: key mediators of microbe detection. Curr Opin Immunol 2002; 14:10310. First citation in article
23. Hayashi F, Smith KD, Ozinsky A, et al. The innate immune response to bacterial flagellin is mediated by Toll-like receptor 5. Nature 2001; 410:1099103. First citation in article
24. Gewirtz AT, Navas TA, Lyons S, Godowski PJ, Madara JL. Cutting edge: bacterial flagellin activates basolaterally expressed TLR5 to induce epithelial proinflammatory gene expression. J Immunol 2001; 167:18825. First citation in article
25. Zeng H, Carlson AQ, Guo Y, et al. Flagellin is the major proinflammatory determinant of enteropathogenic Salmonella. J Immunol 2003; 171:366874. First citation in article
26. Wood WB Jr, Smith MR. Intercellular surface phagocytosis. Science 1947; 106:86. First citation in article