Phagocytosis of microorganisms has been recognized for over a century as a process intended to limit or resolve infections. Pathogens have recognized for much longer that phagocytosis is either an opportunity for or an obstacle to their own replication. Pathogens that survive within host phagocytes have evolved mechanisms of survival by remodelling their phagosome (Mycobacteria), by moving out of the phagosome (Listeria) or by resisting the antimicrobial environment of the mature phagolysosome (Coxiella). In contrast, obligate extracellular pathogens have developed mechanisms of avoiding phagocytosis. As the understanding of the molecular processes underlying phagocytosis has advanced, discovery of mechanisms of microbial avoidance of phagocytosis has also progressed, and there are undoubtedly many further discoveries to be made. This review focuses on the current and evolving state of knowledge of the mechanisms of phagocytosis, emphasizing the steps that have been subverted by pathogens that survive extracellularly by avoiding phagocytosis. For additional detail on signal transduction and regulation of phagocytosis, there are two recent comprehensive reviews ( Aderem and Underhill, 1999; Greenberg, 1999).
General principles of phagocytosis
Phagocytosis provides a specialized mechanism for regulated ingestion and intracellular destruction of microbial pathogens as well as of apoptotic host cells and debris. In general, most phagocytosis is accomplished by professional phagocytes, including neutrophils (which migrate from the blood to a site of infection) and macrophages (which constitutively reside in tissues and are less motile than neutrophils). By confining the mechanisms of microbial killing and digestion to distinct intracellular compartments of specialized cells, host cells and tissues are subjected to less damage by the process of killing offending microbes. In addition to disposing of microbial pathogens, phagocytosis (especially by macrophages and dendritic cells) initiates the process of antigen processing and presentation for development of cellular immune responses ( Ramachandra et al., 1999 ).
Phagocytosis is a membrane-directed process driven by the host cell actin cytoskeleton that results in internalization of particles the size of bacteria and yeasts (≈ 0.5–5.0 µm in diameter). Concomitant with internalization of a target particle in a plasma membrane-derived phagosome, intracellular signalling activates the production of reactive oxygen intermediates, serial interactions of the phagosome with the endocytic network and alterations in the expression of specific genes, especially those encoding cytokines. The maturing phagosome becomes acidic (to pH ≈ 5.5) ( Hackam et al., 1997 ) and ultimately acquires the contents of lysosomes, which include hydrolytic enzymes as well as specific antimicrobial pore-forming peptides. In addition, maturing phagosomes acquire Nramp1, which transports divalent cations and modulates intraphagosomal iron content ( Atkinson and Barton, 1999; Kuhn et al., 1999 ). The combination of reactive oxygen and nitrogen, acidic pH, pore-forming peptides, imbalance of iron and hydrolytic enzymes provides a highly effective antimicrobial combination that is capable of killing a wide variety of bacteria and yeast. It is not surprising, then, that certain pathogens have evolved mechanisms to avoid entering the potentially deadly intracellular environment of phagocytic cells.
Mechanisms and regulation of phagocytosis
The initial event in the phagocytosis of particles is recognition of the particle by specific receptors on the plasma membrane of the phagocyte. Neutrophils express receptors for serum-derived opsonins, including IgG and opsonic fragments of the complement component C3 (C3b and C3bi). Macrophages also express IgG Fc and complement receptors, but are richer in their repertoire of well-characterized receptor types than neutrophils and also express receptors that directly recognize microbial targets. These receptors include type A scavenger receptors, which recognize lipopolysaccharides of Gram-negative bacteria, lipoteichoic acids of Gram-positive bacteria and sulphatides of mycobacteria ( Dunne et al., 1994 ; Thomas et al., 2000 ) and mannose receptors, which recognize terminal mannose residues on fungal polysaccharides and mycobacterial glycolipids ( Ezekowitz et al., 1990 ). In addition, macrophages express receptors for pulmonary surfactant protein A ( Pison et al., 1992 ; Chroneos et al., 1996 ) and for the complement component C1q ( Nepomuceno et al., 1997 ), which can mediate phagocytosis. Non-professional phagocytes such as epithelial cells and intestinal M cells can internalize bacteria though β1 integrins, especially through recognition of proteins such as Yersinia invasin ( Isberg and Leong, 1990; Clark et al., 1998 ).
1 Three types of Fcγ receptors can mediate phagocytosis. FcγRI, II (in human, not mouse) and III differ with respect to their patterns of expression on neutrophils and macrophages, yet they initiate signal transduction in a similar manner through intrinsic or associated ITAM ( immunoreceptor tyrosine activation motif; Y-X-X-L) domains. Upon receptor clustering by binding an IgG-coated particle, the receptor ITAM domains are phosphorylated by one or more members of the Src family of tyrosine kinases. Genetic absence of the Src kinases Hck, Fgr or Lyn results in defective tyrosine phosphorylation of the FcγR γ-chain and, of these kinases, Lyn appears to have the earliest essential action in FcγR-dependent phagocytosis ( Fitzer-Attas et al., 2000 ), whereas Fgr is a negative regulator of phagocytosis ( Gresham et al., 2000 ). Tyrosine phosphorylation of the γ-chain ITAM is responsible for subsequent recruitment and activation of the tyrosine kinase p72syk ( Raeder et al., 1999; Greenberg et al., 2000 ). Syk is essential for FcγR-mediated phagocytosis ( Crowley et al., 1997 ; Kiefer et al., 1998 ) and phosphorylates several substrates, including the adaptor protein Cbl, which then associates with the p85 subunit of PI 3-kinase ( Sato et al., 1999 ) and with Crk ( Buday et al., 1996 ) (see below).
Downstream of the initial tyrosine phosphorylation events, the signals that mediate phagocytosis diverge ( Greenberg, 1999). One branch leads to activation of one or more guanine nucleotide exchange factors (GEFs; see Fig. 2, the GTP cycle) of the Dbl or Vav families that activate the small GTPases Rac and Cdc42 ( Caron and Hall, 1998) by replacing GDP with GTP. Although activated (GTP-bound) Rac and Cdc42 have multiple downstream effectors, the most relevant effectors for actin cytoskeletal rearrangement leading to phagocytosis appear to be the Wiscott–Aldrich Syndrome protein (WASp) and its widely expressed relative N-WASP ( Lorenzi et al., 2000 ). GTP-bound Cdc42 interacts directly with WASp and N-WASP, which in turn interact directly with members of a complex of actin-binding proteins, Arp2/3 (reviewed by Higgs and Pollard, 1999). The Arp2/3 complex nucleates actin filaments and overcomes the rate-limiting step in f-actin formation ( May et al., 2000 ). The result of these events is polymerization of actin, formation of an f-actin ‘cup’ at the site of particle binding to the plasma membrane, followed by the appearance of f-actin-rich membrane ruffles adjacent to the phagosome and internalization of the bound particle surrounded by a lipid bilayer membrane and a halo of f-actin. As phagocytosis is a dynamic and transient event, mechanisms must exist for inactivating the signals that mediate actin polymerization on phagosomes. At least one of these mechanisms is inactivation of Cdc42 and Rac by accelerating their hydrolysis of GTP to GDP, which is accomplished by one or more GTPase activating proteins (GAPs; Lamarche-Vane and Hall, 1998) ( Fig. 2).
The other ‘branch’ of the signalling pathway crucial for FcγR-mediated phagocytosis is activation of one or more PI 3-kinases, resulting in the production of phosphatidylinositol 3,4,5 trisphosphate. Inhibition of PI 3-kinases by wortmannin and LY294002 does not inhibit the initiation of phagocytosis. Rather, inhibition of PI 3-kinases selectively inhibits closure of the phagocytic cup ( Araki et al., 1996 ). More recently, it has been found that PI 3-kinase activity is essential for exocytotic fusion of intracellular membrane vesicles with the membrane of the forming phagosome, which is required to supply sufficient membrane to envelope a large particle fully ( Cox et al., 1999 ). Although it is unclear whether Syk directly phosphorylates and activates PI 3-kinases, the same defect in phagosome closure is observed in macrophages of Syk –/– mice ( Crowley et al., 1997 ). In addition to evidence that PI 3-kinase activity is necessary for membrane lipid trafficking, recruitment of amphiphysin IIm ( Gold et al., 2000 ) and dynamin 2 ( Gold et al., 1999 ) (which are required for optimal phagocytosis) to phagosomes is also blocked by inhibitors of PI 3-kinases. Delivery of membrane vesicles by fusion is also sensitive to tetanus or botulinum b neurotoxins ( Hackam et al., 1998 ) and involves rab11- and VAMP3-containing recycling endosomes ( Bajno et al., 2000; Cox et al., 2000 ).
Complement receptor type 3-mediated phagocytosis
Compared with Fcγ receptor-mediated phagocytosis, the signal transduction events essential for complement receptor (CR3 and CR4)-mediated phagocytosis are less well understood. Nevertheless, some similarities and some differences between FcγR- and CR-mediated phagocytosis have been identified. Ultrastructural studies have demonstrated that, although FcγR-mediated phagocytosis is accompanied by the extension of lamellipodia adjacent to IgG-opsonized targets, C3-opsonized particles appear to sink into the cell without extension of the membrane or cytoskeletal elements from the surface of the cell ( Kaplan, 1977; Allen and Aderem, 1996). A potential mechanistic explanation for this difference has been provided by finding that phagocytosis by complement receptors requires active Rho GTPase but not Rac or Cdc42, whereas Fcγ receptor phagocytosis requires active Rac and Cdc42, and it is Rac that directs the formation of lammelipodia ( Caron and Hall, 1998). Another prominent difference between FcγR- and CR-mediated phagocytosis is that the latter does not depend on active tyrosine kinases (Hck, Fgr, Lyn or Syk) ( Allen and Aderem, 1996; Crowley et al., 1997 ; Fitzer-Attas et al., 2000 ). In contrast, both FcγR and CR phagocytosis require protein kinase C activity ( Zheleznyak and Brown, 1992; Allen and Aderem, 1996).
Bacterial inhibition of phagocytosis
Type III secretion
Gram-negative bacteria have evolved the type III secretion system to deliver proteins to the interior of host cells. The core machinery that is used to deliver the secreted substrates is highly conserved among Yersinia, Pseudomonas, Salmonella and other genera ( Frithz-Lindsten et al., 1997 ). In contrast, the secreted substrates themselves are much less conserved and serve functions specific to the preferred niche of the pathogen. As an especially good example, substrates of the Salmonella typhimurium type III system allow Salmonella actively to invade mammalian cells (in which they thrive), whereas substrates of the Yersinia type III secretion system allow Yersinae to block their uptake by mammalian cells (in which they cannot survive). As type III secretion is organized to deliver substrates to the cytoplasm of host cells, the targets of the substrates are intracellular proteins of the host cell. Type III secretion is also tightly regulated to enhance its efficiency. Expression of Yops (originally named for their discovery as Yersinia outer proteins) is repressed at low temperature, such as that found in contaminated food (Yersinia pseudotuberculosis or Yersinia enterocolitica) or the salivary glands of fleas (Yersinia pestis). At the temperature of a warm-blooded host, temperature-dependent repression of Yops is relieved, but repression is maintained by the presence of millimolar (i.e. extracellular) concentrations of calcium until the bacteria contact a host cell. YopN apparently serves as the sensor of cell contact and may also serve to target the secretion machinery to the zone of the bacteria in contact with the target cell. Upon cell contact at 37°C, repression is relieved, Yop expression is initiated, YopB and YopD co-operate to promote the passage of preformed ( Bliska and Black, 1995) and newly induced ( Rosqvist et al., 1994 ) secreted substrates from the bacteria to the host cell interior and the substrates can interact with their target host proteins (reviewed by Fallman et al., 1997 ). From the standpoint of a pathogen that wants to avoid phagocytosis, type III secretion is ideally designed to deliver bacterial products to a target cell in the immediate region of the cell (i.e. in contact with the bacterium) that would be responsible for initiating phagocytosis.
YopH, a tyrosine phosphatase
The Yersiniae (pestis, pseudotuberculosis and enterocolitica) share a common virulence plasmid, which encodes the type III secretion machinery as well as the secreted substrates that allow the bacteria to remain extracellular during in vivo infection ( Simonet et al., 1990 ). Investigation of the molecular mechanisms of Yersinia pathogenesis, including mechanisms whereby the bacteria avoid phagocytosis, has contributed greatly to the understanding of mechanisms of phagocytosis and its regulation.
YopH (formerly Yop2b) is a potent inhibitor of phagocytosis by macrophages ( Rosqvist et al., 1988 ) and blocks phagocytosis through Fcγ receptors as well as through β1 integrins (via invA) ( Fallman et al., 1995 ). YopH also inhibits phagocytosis in trans. That is, Yersinia that express YopH also inhibit phagocytosis of other particles, such as unopsonized yeast. The effect of YopH is rapid (≤ 5 min) and potent (4.2 YopH-expressing Yersinia per macrophage is sufficient to block phagocytosis of IgG-opsonized yeast) ( Fallman et al., 1995 ).
YopH is a tyrosine phosphatase and has the highest specific activity of any tyrosine phosphatase known ( Guan and Dixon, 1990; Zhang et al., 1992 ). By virtue of its selective delivery by type III secretion to the site of bacterial contact with the host cell and its high specific activity, YopH rapidly (≤ 30 s) dephosphorylates host cell tyrosine phosphoproteins ( Bliska et al., 1991 ; 1992; Andersson et al., 1996 ). A point mutation (C403A) that inactivates the phosphatase activity of YopH also destroys its ability to inhibit phagocytosis, confirming that dephosphorylation of host proteins is essential for the inhibition of phagocytosis ( Andersson et al., 1996 ), which is not surprising given the critical role of tyrosine kinases in Fcγ receptor-mediated phagocytosis. YopH-expressing bacteria (unopsonized or IgG opsonized) cause rapid and selective dephosphorylation of a 120 kDa NP-40-insoluble protein and 120 kDa and 43 kDa soluble proteins in the J774 macrophage cell line ( Andersson et al., 1996 ). The YopH 403C/A mutant, which recognizes YopH substrates but cannot dephosphorylate them, has been exploited to identify the predominant targets of YopH and binds 120 kDa and 76 kDa proteins in J774 cells. The detergent-soluble 120 kDa protein is FYB (fyn-binding protein), and the 120 kDa detergent-insoluble protein is Cas ( Crk- associated substrate) ( Hamid et al., 1999 ), which is also one of the proteins rapidly dephosphorylated by YopH in HeLa cells ( Black and Bliska, 1997). Cas is an adaptor protein that contains multiple SH2 (phosphotyrosine recognition) domains and is recruited to focal contacts during integrin-mediated adhesion to extracellular matrix proteins. Cas is tyrosine phosphorylated in J774 cells exposed to Y. pseudotuberculosis expressing phosphatase-inactive YopH, and YopH 403C/A is recruited to periphagosomal structures that resemble focal contacts (and contain vinculin). These results indicate that Cas may be an important and perhaps essential adaptor protein for phagocytosis, but do not exclude essential roles for other YopH targets. Moreover, it is not yet clear whether Cas functions in phagocytosis mediated by receptors other than β1 integrins. So far, neither Lyn nor Syk have been found to be targets of YopH, although the experiments reported have not examined FcγR-mediated phagocytosis, in which the roles of these two tyrosine kinases are best established. The ability of YopH to localize to ‘focal complexes’ during phagocytosis appears to be essential to its ability to inhibit phagocytosis optimally. A mutated YopH that lacks GEKL (residues 223–226) in a surface-exposed loop retains full potency as a tyrosine phosphatase as well as the ability to be secreted efficiently into target cells, but fails to localize to focal complexes, is less efficient at inhibiting phagocytosis and is nearly as attenuated in mice as a YopH null mutant ( Persson et al., 1999 ). These findings strongly support the model that the essential targets of YopH in phagocytosis are contained in protein complexes that link phagosome membrane subdomains to the cytoskeleton. Most of these studies have been performed using unopsonized Y. pseudotuberculosis that enter macrophages (and HeLa cells) through invasin–β1 integrin interactions. It will be of considerable interest to learn whether similar proteins and principles also apply to the inhibition of FcγR-mediated phagocytosis.
Why is YopH so effective in inhibiting macrophage phagocytosis (even in vivo) when complement receptor-mediated phagocytosis may not require tyrosine phosphorylation? Y. enterocolitica has taken care of that using the product of another gene located on the same virulence plasmid as the Yops. YadA, expressed by virulent Y. enterocolitica (but not Y. pestis), prevents deposition of C3b on the bacterial surface, thereby preventing the bacteria from being recognized by complement receptors ( China et al., 1993 ). Therefore, by avoiding recognition by complement receptors, the bacteria are diverted to receptors (FcγR and β1 integrins) whose mechanisms of phagocytosis are inactivated by YopH (and YopE; see below). In contrast, Y. pestis, which lacks YadA and invasin, may depend on other mechanisms such as YopE to evade phagocytosis in the absence of specific IgG.
YopE and the N-terminal domain of Pseudomonas aeruginosa ExoS, GAPs for rho family GTPases
Deletion of YopH from the Yersinia common virulence plasmid markedly reduces the ability of Yersiniae to resist phagocytosis, but the additional deletion of YopE is required to allow phagocytosis to the level equivalent to that of plasmid-cured Yersiniae ( Rosqvist et al., 1988 ; 1990). Like YopH, YopE is delivered to the host cytoplasm by type III secretion ( Rosqvist et al., 1994 ). In peritoneal macrophages, YopE is more effective than YopH at inhibiting phagocytosis and is cytotoxic for macrophages and HeLa cells ( Rosqvist et al., 1990 ). YopE disrupts the actin cytoskeleton, but does not interact directly with f-actin ( Rosqvist et al., 1991 ), suggesting that it targets an actin-regulating protein. A clue to the function of YopE was provided by the observation that a related protein of Salmonella typhimurium, SptP, is a GTPase-activating protein for Rac and Cdc42 ( Fu and Galan, 1999). Indeed, YopE was recently discovered to be a GTPase-activating protein (GAP) that selectively increases GTP hydrolysis by RhoA, Rac and Cdc42 ( Fig. 2), but not that of Ras or Ral ( von Pawel-Rammingen et al., 2000 ). By accelerating the hydrolysis of bound GTP to GDP, YopE rapidly converts Rho, Rac and Cdc42 to their inactive forms so that they do not activate downstream effectors of actin polymerization, such as WASp and N-WASP, and thereby inhibit phagocytosis. A point mutation that inactivates the RhoGAP activity of YopE disrupts its ability to inhibit phagocytosis, which confirms that the Rho family proteins are the direct targets of YopE.
As in Yersinia, inactivation of the type III secretion system increases the internalization of P. aeruginosa into epithelial cells ( Hauser et al., 1998 ; Evans et al., 1998 ) and macrophages (J. Engel, personal communication). Recently, two substrates of the type III system of P. aeruginosa, ExoS and ExoT, have been shown to function as anti-internalization factors in epithelial cells ( Cowell et al., 2000 ) and macrophages (J. Engel, personal communication). The two proteins share 95% identity at the amino acid level and can be functionally divided into two domains ( Kulich et al., 1994 ; Yahr et al., 1996 ). The C-terminal domain of ExoS and ExoT resembles several ADP ribosyltransferases, which initially implied that the actin-disrupting and cytotoxic activity of ExoS might be mediated by ADP ribosylation of specific host cell proteins. However, ExoS mutants with ≈ 2000-fold less ADP ribosyltransferase activity retained the ability to disrupt the actin cytoskeleton and cause cytotoxicity ( Frithz-Lindsten et al., 1997 ). Moreover, ExoT contains two amino acid changes in the catalytic domain of the ADP ribosyltransferase, leaving it with only 0.2% of the enzymatic activity of ExoS. Therefore, the action of ExoS on the actin cytoskeleton is not attributable to ADP ribosylation. In contrast, the N-terminal domain of ExoS and ExoT resembles Yersinia YopE. An early clue to the function of this domain as well as that of YopE was provided by the observation that disruption of the actin cytoskeleton by the N-terminal domain of ExoS is reversed by Escherichia coli cytotoxic necrotizing factor-1, which constitutively activates rho ( Pederson et al., 1999 ). Shortly thereafter, the N-terminal domain of ExoS was found to be a RhoGAP, with GAP activity in vitro for RhoA, Rac and Cdc42, but not for Ras or Ral ( Goehring et al., 1999 ). Mutation of the GAP domain of ExoT partially abrogates its cell-rounding and anti-internalization acitivity (J. Engel, personal communication). Therefore, the effects of YopE, ExoS and ExoT on phagocytosis and the actin cytoskeleton are mediated by the same biochemical mechanism: accelerated conversion of Rho family members to their GDP-bound, inactive states.
The concerted study of molecular mechanisms of phagocytosis and the inhibition of phagocytosis by specific products of extracellular bacterial pathogens has borne considerable fruit. The importance of tyrosine phosphorylation and of the Rho family of GTPases has become clear to cell biologists, but pathogenic bacteria recognized the importance of these signalling pathways in phagocytic cells long ago. The discoveries described in this review are only the beginning. The simultaneous pursuit of the mechanisms and molecules involved in the initiation and regulation of phagocytosis and that pathogenic bacteria use to inhibit phagocytosis will surely identify more interesting pathways on each side of the contest. Are there any obvious possibilities? There are several bacterial factors that have the potential to inhibit known mechanisms of phagocytosis. Clostridium species, for example, make a number of exotoxins of interest. Clostridium botulinum and Clostridium tetani neurotoxins inactivate the regulated secretory machinery by proteolytic cleavage of SNARE proteins, and targets of tetanus toxin and botulinum b toxin inhibit the exocytotic delivery of membrane vesicles needed for phagocytosis of large particles ( Hackam et al., 1998 ). Moreover, the C3 exotoxin of C. botulinum catalyses ADP ribosylation and inactivation of rho family GTPases ( Wiegers et al., 1991 ), and toxins A and B of C. difficile UDP-glucosylate and inactivate rho GTPases and thereby disrupt the actin cytoskeleton ( Just et al., 1995a,b ). However, as Clostridia lack the machinery for type III secretion, these proteins are not rapidly targeted to the phagocyte cytoplasm. More searching may reveal a pathogen that has combined the type III secretory machinery with clostridia toxin-like substrates. A potentially unique strategy for remaining outside phagocytes is exhibited by Helicobacter pylori, which contain a type IV secretion system. Unopsonized virulent strains of H. pylori bind readily to macrophages but are only internalized after a delay of several minutes. Such a delay appears to be sufficient for the bacteria to remain extracellular ( Allen et al., 2000 ). Elucidation of the mechanism used by H. pylori to delay phagocytosis may reveal one or more novel virulence factors as well as one or more novel targets in the phagocyte that will add to the understanding of a fundamental process in host defence.
Another field ripe for further mechanistic investigation is complement receptor-mediated phagocytosis. Dedicated study of the molecular events and molecular mediators of phagocytosis downstream of CR3 is likely to reveal interesting differences from FcγR phagocytosis and is just as likely to reveal that microbes have discovered unique mechanisms for circumventing them.
Study of extracellular pathogens and the mechanisms that they use to remain outside phagocytic cells has revealed a great deal about the initial encounter between pathogen and phagocyte. We can look forward to additional discoveries about the host–pathogen interactions and the mechanisms and factors that each side uses to battle against the other.
This work was supported in part by NIH grants HL51992 and HL56001. I thank the members of my laboratory and two anonymous reviewers for critical review and comments that substantially improved the article. I also thank Drs Maria Fällman, Roland Rosqvist and Joanne Engel for communicating results in advance of publication.