|Home | About | Journals | Submit | Contact Us | Français|
We consider here a previously neglected aspect of recovery from infectious diseases: how animals dispose of the dead microbes in their tissues. For one of the most important disease-causing microorganisms, Gram-negative bacteria, there is now evidence that the host catabolism of a key microbial molecule is essential for full recovery. As might be expected, it is the same bacterial molecule that animals sense to detect the presence of Gram-negative bacteria in their tissues, the cell wall lipopolysaccharide (LPS). Here, we discuss current knowledge about LPS sensing with emphasis on the host enzyme that inactivates this microbial “messenger” molecule. We also consider the possibility that the rate at which stimulatory microbial molecules undergo inactivation may influence the duration and severity of diseases caused by other infectious agents.
All Gram-negative bacteria living in natural environments produce LPS, a complex glycolipid that contributes to outer membrane impermeability, confers resistance to detergents and cationic antimicrobial peptides, provides cell-surface diversity, and prevents complement-mediated cell death. Animals sense the lipid A moiety of LPS via MD-2–TLR4 receptors on phagocytes and other cells, and much evidence suggests that recognizing LPS in this way is essential for detecting the presence of Gram-negative bacteria in tissues and mobilizing antibacterial defenses. The structure of lipid A is not identical in different Gram-negative bacteria, however, and not all lipid As can trigger inflammatory responses via MD-2–TLR4. Extensive structure–activity studies have shown that a bisphosphorylated, hexaacyl lipid A structure (Fig. 2.1) is most stimulatory; removal or addition of a single acyl chain can diminish potency, as can the absence of either of the backbone phosphates. Although many Gram-negative bacteria make LPSs that are poorly recognized by MD-2–TLR4, with potentially important consequences for disease pathogenesis (Munford, 2008), the aerobic commensals and pathogens that colonize the mucosae of the upper respiratory and gastrointestinal tracts generally produce LPSs that have hexaacyl lipid A moieties and are readily sensed by cells bearing MD-2–TLR4 (Munford and Varley, 2006). It is these bacteria that animals are best equipped to defeat using TLR4-based inflammatory responses, and it is the LPSs from these bacteria that are most likely to translocate into the bloodstream to produce “endotoxemia.” These are also the LPSs that can be inactivated by the unusual host lipase, acyloxyacyl hydrolase (AOAH).
AOAH was found during a search for human neutrophil enzymes that could deacylate LPS (Hall and Munford, 1983). The “bait,” a biosynthetically labeled LPS that had 14C-glucosamine and 3H-fatty acyl chains (Fig. 2.1), was opsonized with an anti-LPS antibody and fed to human neutrophils in vitro. The 14C dpm, which marked the carbohydrate backbone, remained chloroform-insoluble during the next 6 h, whereas the 3H in the LPS acyl chains gradually became chloroform-soluble and could be recovered in cellular phospholipids. Further analysis showed that only the nonhydroxylated fatty acids (myristate and laurate) were released from the LPS—unexpectedly, all of the 3-hydroxymyristoyl chains remained attached to the backbone, suggesting that neutrophils may lack the ability to degrade lipid A completely. Correct interpretation of the sites of enzymatic hydrolysis became possible when the existence of acyloxyacyl linkages (Wollenweber et al., 1982) in lipid A and the first accurate lipid A structure (Takayama et al., 1983) were published: the myristoyl and lauroyl chains are attached to the hydroxyl functions of glucosamine-linked 3-hydroxymyristoyl residues to form acyloxyacyl linkages (Fig. 2.1, arrowheads). Denis McGarry suggested that the enzymatic activity be named “acyloxyacyl hydrolysis” and the enzyme(s) “acyloxyacyl hydrolase(s)”. A literature review revealed that enzymatic release of non-hydroxylated fatty acids from LPSs had been reported previously in slime molds (Dictyostelium discoideum, Nigam et al., 1970; Physarum polycephalum, Saddler et al., 1997a) and a snail (Helix pomatia, Saddler et al., 1979b).
In mammals, AOAH is produced by monocyte-macrophages and dendritic cells as well as neutrophils. Another prominent source, particularly in rodents, is the renal cortical epithelial (proximal tubule) cell (Feulner et al., 2004). Evidence that AOAH can inactivate LPS followed partial purification of the protein (Munford and Hall, 1986). Table 2.1 summarizes the bioassays used to evaluate this point. Note that AOAH-treated LPS not only lacks potency in these assays but it can also inhibit the ability of untreated LPS to stimulate both human and murine cells. The species-specific inhibition reported for tetraacyl lipid A structures (such as lipid IVa) (Golenbock et al., 1990; Hajjar et al., 2002) does not necessarily apply to tetraacylated LPSs, such as those produced by AOAH treatment, in which tetraacyl lipid A is linked to a polysaccharide chain of variable length; these tetraacyl LPSs are poor agonists for both rodent and human cells. AOAH-treated LPS was the first lipid A derivative shown to inhibit LPS (Pohlman et al., 1987); it and other tetraacylated lipid “A”s can compete with hexaacylated lipid A or LPS for binding LBP, CD14 (Kitchens and Munford, 1995), and, most importantly, MD-2 (Kim et al., 2007).
Purification of the enzyme from HL-60 human promyelocytes revealed a low abundance, glycosylated protein of Mr =~65 kDa that has two disulfide-linked subunits (Munford and Hall, 1989). The cDNA sequence indicated that the enzyme is produced as a single polypeptide chain; proteolytic cleavage is thus required to yield the two subunits (Hagen et al., 1991) (Fig. 2.2). The larger subunit (50 kDa) has the GXSXG consensus motif that has been found in serine-active site enzymes, and mutating the Ser to Leu inactivated the enzyme (Staab et al., 1994). This subunit is now considered a GDSL or SGNH lipase (Akoh et al., 2004). The smaller subunit is a member of the saposin-like protein (SAPLIP) family (Munford et al., 1995). It shares six Cys residues and other features with several small proteins that act as enzymes or cofactors for glycosphingolipid catabolism (the saposins, acid sphingomyelinase), form pores in membranes (amoebophore, NK-lysin), or act at lipid–air interphases in the lung (surfactant protein B). The enzymatically active AOAH large subunit is thus armed with a covalently linked “cofactor,” the saposin-like small subunit (Fig. 2.3). Without it, the enzyme did not localize in intracellular vacuoles and had lower affinity for LPS (Staab et al., 1994). AOAH and a closely similar Trypanosoma bruceii inositol deacylase (Güther et al., 2001) are the only known lipases that have this saposin–lipase combination, which has been very highly conserved in AOAHs from D. discoideum to Homo sapiens (Munford and Varley, 2006). Whereas saposin B may participate in NK-T cell activation by transferring glycolipid antigens to CD1d (Zhou et al., 2004), several attempts identify a similar role for AOAH have been unsuccessful.
The human and murine AOAH genes have 21 small exons on chromosome 7p14-p12 and 13, respectively. In both species, the gene extends over ~200 kb of genomic DNA. To prevent synthesis of AOAH in mice, the starting ATG and the downstream 126 bp of the first exon were replaced with a neomycin resistance cassette. Mice carrying this construct do not produce AOAH protein (Shao et al., 2007) or have LPS-deacylating activity (Lu et al., 2003; Shao et al., 2007).
As noted, AOAH removes secondary (piggyback, acyloxyacyl-linked) chains from different positions on the diglucosamine lipid A backbone without attacking any of the primary glucosamine-linked chains (Erwin and Munford, 1990). Immunoadsorption of AOAH from leukocyte lysates removed all enzymatic activity toward LPS, indicating that AOAH was the only LPS-deacylating enzyme in these cells, and recombinant AOAH can remove both of the secondary acyl chains from the LPS backbone. It can also be a phospholipase A1/B, diglyceride lipase, and acyl transferase in vitro, and it has a preference for cleaving saturated (or mid-length) acyl chains from both phospholipid and LPS substrates (Munford and Hunter, 1992). AOAH can remove acyl chains that are attached to different positions on the diglucosamine and glycerol backbones of LPS and glycerolipids, respectively (Erwin and Munford, 1990; Munford and Hunter, 1992). It thus has specificity for acyl chain character, not backbone position. Transfected fibroblasts secrete the precursor, which can be taken up by the same or different cells and proteolytically processed into the mature enzyme (Feulner et al., 2004). The AOAH precursor peptide has enzymatic activity toward both LPS and phosphatidylcholine (PC), yet cleaving the precursor to form the mature enzyme increased its activity toward LPS more than 10-fold without altering its ability to act on PC (Staab et al., 1994).
The double-radiolabeled LPS substrate can also be used to quantitate the rate and extent of LPS deacylation in vivo. Whether the LPS is injected subcutaneously (footpad, skin site), intraperitoneally or intravenously, deacylation occurs over many hours (Lu et al., 2005; Shao et al., 2007). Despite this seemingly sluggish performance, AOAH-mediated deacylation completely inactivated almost 80% of a subcutaneous dose of LPS before the LPS could travel to draining lymph nodes (Lu et al., 2005). No loss of the LPS secondary acyl chains was detected in Aoah−/− mice, and in neither Aoah−/− nor Aoah+/+ animals was there loss of primary (3-hydroxymyristoyl) acyl chains from the backbone. LPS deacylation in vivo is thus remarkably selective and limited. Of the various LPS-catabolizing enzymes produced by D. discoideum (Verret et al., 1982a,b), which eat bacteria as a foodstuff, only AOAH has been conserved during animal evolution.
One important unresolved issue is the extent to which LPS deacylation occurs inside and outside host cells. An intracellular site was suggested by the enzyme’s acid pH optimum, its location in an intracellular vacuole (Staab et al., 1994), and the apparent colocalization of AOAH and deacylated LPS in neutrophils (Luchi and Munford, 1993). Both rabbit macrophages and murine dendritic cells deacylate the LPS in phagocytosed Escherichia coli in an AOAH-dependent fashion (Katz et al., 1999; Lu et al., 2003). Moreover, AOAH does not act on LPS in buffered salt solutions in the absence of a nonionic detergent such as Triton X-100, suggesting that it may require an intracellular environment or factor(s) to do its job. The enzyme can be secreted by rabbit neutrophils and monocytes (Erwin and Munford, 1991), however, and extracellular deacylation has been demonstrated in rabbit peritoneal exudate fluid (Katz et al., 1999), a rich mixture of extravasated plasma, leukocyte products, and other molecules. The AOAH secreted into the urine by renal cortical epithelial cells can also deacylate LPS (Feulner et al., 2004). Gioannini et al. recently reported that CD14 and LBP can bind LPS in a way that allows AOAH to deacylate its lipid A moiety (Gioannini et al., 2007), providing an attractive mechanism for extracellular LPS deacylation (Weinrauch et al., 1999). The relative contributions of intra- and extracellular deacylation to LPS inactivation in vivo remain uncertain.
Very little is known about how AOAH activity is regulated in vivo. In part this reflects the enzyme’s low abundance, which has hindered quantitative detection of both AOAH protein and mRNA. In addition, it has not been possible to detect AOAH activity or protein in human plasma or serum. Since the enzyme is easily measured in rodent and rabbit serum, most studies of AOAH regulation have been performed in these animals. In rabbits, plasma AOAH levels rise dramatically within a few minutes of an intravenous injection of LPS and remain elevated for many hours (Erwin and Munford, 1991). The increase in AOAH activity was significantly less in animals that had been given methchlorethamine to induce leukopenia, suggesting that much of the extracellular enzyme is produced and released by neutrophils or monocytes. Indeed, rabbit leukocytes released AOAH in response to stimulation by LPS ex vivo (Erwin and Munford, 1991). In studies performed in mice, Cody et al. found that AOAH mRNA and activity in liver and lung increased several-fold following intraperitoneal treatment with LPS (Cody et al., 1997). In addition, low concentrations of LPS and interferon-γ induced greater than 10-fold increases in AOAH mRNA in cultured murine macrophages. Neither IL-10 nor dexamethasone prevented AOAH mRNA accumulation in response to LPS, in keeping with the discovery, many years later, that AOAH participates in the anti-inflammatory (recovery) phase of local infection. Indeed, Mages et al. (2008) found an approximately sixfold increase in AOAH mRNA abundance in LPS-primed (tolerant) murine macrophages relative to unstimulated controls. Immature murine dendritic cells also produce AOAH; cytokine-induced maturation was associated with diminished LPS-deacylating ability, whereas exposure to LPS, CpG oligonucleotides, or staphylococci was stimulatory (Lu et al., 2003). Unfortunately, it is uncertain that mice are useful models for human AOAH regulation, since the tissue-specific expression of the enzyme differs substantially (mice produce much more AOAH in the kidney than do humans, and less in myeloid cells). DeLeo and colleagues found that AOAH mRNA abundance decreased approximately twofold in human neutrophils during the 6 h following phagocytosis of latex beads, a time when many of the neutrophils were undergoing apoptosis (Kobayashi et al., 2003). Further study of AOAH regulation in human phagocytes is needed.
Animals have several mechanisms for inactivating LPS (Munford, 2005), including lipid A-neutralizing proteins (bactericidal permeability-increasing protein, lactoferrin, lysozyme, collectins, etc. (Chaby, 2004)), specific and cross-reactive anti-LPS antibodies, and sequestration of the lipid A moiety within lipoprotein micelles. Although intestinal alkaline phosphatase can inactivate LPS in zebrafish (Bates et al., 2007), a role for endogenous alkaline phosphatase in LPS inactivation in mammals has not been established. At present, AOAH is the only endogenous enzyme known to inactivate LPS in tissues.
Early expectations that AOAH would protect animals from LPS-induced inflammation met with disappointment when it was learned that Aoah−/− and Aoah+/+ mice had similar acute inflammatory responses to LPS and indistinguishable survival outcomes following LPS or Gram-negative bacterial challenge (Fig. 2.4A). In addition, macrophages and dendritic cells from wild-type and AOAH-deficient animals had similar responses to LPS in vitro. On the other hand, the Aoah+/+ mice that survived a Neisseria meningitidis challenge recovered more rapidly than did the surviving Aoah−/− mice (Fig. 2.4B). Suspecting that the enzyme’s role might be discernable only when late responses to LPS and Gram-negative bacteria were studied, we began to look for long-term abnormalities in LPS-treated mice. Three AOAH-dependent phenotypes have now been identified, each of which reflects the ability of persistently active (fully acylated) LPS to stimulate cells in vivo for long periods of time. It seems that the enzyme’s low abundance and slow deacylation rate are useful for a host defense that responds rapidly and vigorously to LPS but then needs to inactivate this microbial “messenger” so as to avoid prolonged cell activation and possible immunosuppression.
LPS is a B cell mitogen in mice, which express TLR4 on B cells. B cell proliferation and polyclonal antibody production are thus quantitative indices of LPS stimulation. Subcutaneous LPS inoculation elicits much greater IgM and IgG3 responses in Aoah−/− mice than in wild-type mice, suggesting that AOAH normally exerts a braking influence on B cell stimulation by LPS in vivo (Lu et al., 2005) (Fig. 2.5A). Indeed, the presence of the enzyme also prevents impressive, prolonged enlargement of the lymph nodes that drain the site of inoculation. Tlr4−/−, Aoah−/− mice developed neither lymphadenopathy nor elevated antibody titers, in keeping with TLR4’s essential role in LPS signaling.
Much of the LPS that enters the bloodstream from the gastrointestinal tract travels via the portal venous system to the liver. Hepatic macrophages (Kupffer cells), which take up a large fraction of this LPS, are also the major AOAH-producing cells in the liver (Shao et al., 2007). Small intravenous doses of LPS induce prolonged, possibly irreversible, hepatomegaly in Aoah−/− mice. This phenomenon was first noticed during experiments performed to define the time-course of LPS deacylation in the liver. In wild-type mice, the liver weight/body weight ratio increased transiently, peaking 3 days after i.v. injection and returning to baseline by day 7. In contrast, in Aoah−/− mice the liver continued to enlarge, reaching 30–50% above baseline within 1 week of i.v. injection and remaining enlarged for at least 3 weeks (Shao et al., 2007). Although the basis for this striking phenomenon remains uncertain, it is associated with the retention of fully acylated LPS by Kupffer cells and sinusoidal engorgement with blood that contains neutrophils, B cells, CD4 and CD8 T cells, and monocytes ((Shao et al., 2007); B.M. Shao, unpublished results). LPS also induces impressive splenomegaly in Aoah−/− mice but this response is transient, resolving within 2 weeks (Shao et al., 2007).
Another remarkable consequence of AOAH deficiency is the development of prolonged endotoxin tolerance and immunosuppression following exposure to very small amounts of LPS. In this context, tolerance refers to the ability of a small priming dose of LPS to induce a state of cellular reprogramming in which responses to subsequent, larger doses of LPS and several other microbial agonists are altered. Endotoxin tolerance is known to occur in many animals, including humans; it is usually considered an adaptation to prevent excessive inflammatory reactions to invading microbes—as others have suggested, it may prevent “friendly fire” while animals recover from infection (Cross, 2002; Medvedev et al., 2006). The duration of the tolerant period is influenced by several factors, including LPS dose, the route of administration, and the animal’s ability to deacylate the LPS. Whereas Aoah+/+ mice recover from the tolerant state within 5–10 days after intraperitoneal exposure to a small dose of LPS or E. coli, Aoah−/− mice remain tolerant for at least 4 months! Moreover, LPS-primed Aoah−/− mice were exquisitely sensitive to challenge with a virulent E. coli strain; susceptibility was associated with delayed production of TNF and IL-6 and massive bacterial growth during the first 24 h after inoculation (Lu et al., 2008). Initial analyses of mRNA expression by LPS-primed Aoah+/+ and Aoah−/− peritoneal macrophages suggest that LPS exposure induces low-grade, persistent activation in AOAH-deficient animals, largely in keeping with the reprogramming phenomenon observed previously in macrophages induced to develop tolerance in vitro (Foster et al., 2007; Mages et al., 2008) but differing from those studies in many of the individual mRNAs that are up- and downregulated during the tolerant period (A. Varley, M. Lu, unpublished results). As with the other two phenotypes, prolonged tolerance in Aoah−/− mice is associated with the presence of fully acylated LPS in cells for long periods of time (Lu et al., 2008).
If these AOAH-dependent phenotypes result from the enzyme’s ability to deacylate LPS, they should not occur in Aoah−/− animals when the MD-2–TLR4 signaling pathway is activated by a non-LPS agonist. We tested this hypothesis using UT12, an agonistic monoclonal antibody to MD-2–TLR4 that was developed by Shoichiro Ohta and colleagues (Ohta et al., 2006). For each of the phenotypes discussed above, Aoah−/− and Aoah+/+ mice had indistinguishable responses to UT12; an example is shown in Fig. 2.5B, which should be compared with Fig. 2.5A. AOAH’s ability to act on LPS, and not other potential substrates, is thus likely to account for the prolonged LPS-induced responses observed in Aoah−/− mice.
Whereas LPS-injected Aoah+/+ mice produce significantly less antibody than do Aoah−/− mice, naïve splenocytes from Aoah−/− and Aoah+/+ animals proliferate and produce antibody to the same extent when they are exposed to LPS in vitro (Lu et al., 2005). AOAH thus has a strikingly different influence on the in vivo and in vitro responses of splenocytes to LPS. Similarly, whereas LPS-primed Aoah−/− peritoneal macrophages retain their tolerant (reprogrammed) phenotype when they are removed from the peritoneal cavity and grown for several days ex vivo, naïve Aoah−/− and Aoah+/+ macrophages recover from tolerance at the same rate when they are first exposed to LPS in vitro. Again, the in vivo phenotype cannot be modeled in vitro. It should be interesting to define the properties of the in vivo environment that allow LPS inactivation to have such an important impact on the duration and nature of host responses to LPS.
The three AOAH-dependent phenotypes have been observed on two widely different murine strain backgrounds, C3H/HeN and C57Bl/6, indicating that these responses to LPS are not strain specific. Another way to show that the observations in Aoah−/− animals are truly the result of AOAH deficiency is to prevent or ameliorate them by providing AOAH. We found that intravenous doses of recombinant human AOAH, given prior to LPS injection, can prevent LPS-induced hepatomegaly (Shao et al., 2007) and that producing AOAH in vivo using a gutted adenoviral vector can prevent prolonged LPS-induced tolerance (Lu et al., 2008). Recombinant adenoviruses produce their cargo proteins largely in hepatocytes, whereas AOAH is made by phagocytic cells and renal cortical epithelial cells. To study the effects of overproducing AOAH in cells that normally make it, we also engineered mice that produce large amounts of AOAH in macrophages and dendritic cells. These mice express AOAH from the human CD68 promoter in a cassette developed by David Greaves at the University of Oxford (Gough et al., 2001). The transgenic mice recovered from LPS challenge more rapidly than did wildtype mice and they were protected from LPS-induced hepatosplenomegaly (N. Ojogun et al., in press). The transgenic animals were also less likely to succumb to an E. coli challenge, confirming the important role that LPS plays in Gram-negative bacterial toxicity in vivo and raising the possibility that increasing AOAH levels might ameliorate harmful responses to Gram-negatives in other animals, including humans (Munford, 2008). For example, early studies suggested that AOAH might play a protective role in bovine coliform mastitis (Dosogne et al., 1998; McDermott et al., 1991). Engineering transgenic cattle to overproduce AOAH in myeloid cells or milk might thus be advantageous for the dairy industry.
Large-scale screens for AOAH-deficient humans have not been performed. One group has reported an association between a particular AOAH haplotype and risk of asthma (Barnes et al., 2006); whether or not this haplotype is associated with altered AOAH production or activity is not known. Screening for AOAH deficiency might also be fruitful in individuals with viral, alcoholic or nonalcoholic steatohepatitis, for which gut-derived LPS may be a contributing factor (Tilg and Diehl, 2000); severe sepsis induced by Gram-negative bacteria that make hexaacyl LPS (see Section 1); prolonged recovery from Gram-negative bacterial diseases; autoimmune diseases in which a role for LPS has been suggested (such as Guillain–Barre syndrome following exposure to Campylobacter jejuni (Ang et al., 2002)); or patients with xanthogranulomatous pyelone-phritis or malakoplakia, rare complications of bacterial infection in which macrophages accumulate lipid and polysaccharides (Gregg et al., 1999). A role for AOAH-mediated LPS inactivation might also be sought in patients with HIV/AIDS, in whom immune dysfunction has been related to the absorption of bacterial molecules, including LPS, from the gastrointestinal tract into portal blood (Brenchley et al., 2006; Jiang et al., 2009).
Greater understanding of how animals inactivate other microbial molecules could also have important consequences for patient care. For example, some individuals who become infected with Borrelia burgdorferi remain symptomatic despite having received effective antibiotic therapy; it has not been possible to grow B. burgdorferi or detect their DNA in inflamed tissues from such patients (Marques, 2008). Perhaps their persistent symptoms are related to an inability to inactivate stimulatory bacterial lipoproteins, which activate host cells via TLR2, or too-rapid deacylation of the Borrelia glycolipid that stimulates protective NK-T cells (Tupin et al., 2008). The same question might be raised regarding recovery from staphylococcal and streptococcal diseases, which can be very prolonged despite negative cultures; lipoproteins are the most potent known immunostimulatory staphylococcal molecules. Recovery from certain viral infections may also take a long time (Didierlaurent et al., 2008), raising the possibility that hosts might differ in their ability to degrade stimulatory viral nucleic acids or proteins. If they shorten the time to full recovery, measures that enhance the host’s ability to inactivate stimulatory microbial molecules would be a useful adjunct to antimicrobial chemotherapy. Since recovery from infectious diseases can take many weeks, with increased risk of another infection during that period (Yende et al., 2008), further investigation of this possibility is clearly warranted.
Studies of LPS inactivation by an unusual host lipase have revealed that killing Gram-negative bacteria does not prepare an animal to confront another microbial invader: the major microbial “messenger” molecule must also be inactivated. Although it seems likely that persistence of microbial agonists other than LPS could also have long-term consequences, much more work is required to test this notion and to identify the important catabolic pathways. If the LPS-primed Aoah−/− mouse is a fruitful model, eliminating bioactive microbial molecules should hasten recovery from infection-induced immunosuppression (tolerance) and possibly prevent other lingering symptoms and signs of disease.
We thank our many helpful colleagues and collaborators. The authors’ laboratory at the University of Texas Southwestern Medical School is supported by grant AI18188 from the National Institute for Allergy and Infectious Diseases, NIH, and by the Bromberg Chair in Internal Medicine.