Shield as Signal: Lipopolysaccharides and the Evolution of Immunity to Gram-Negative Bacteria

According to the innate immunity concept [1], animals defend themselves from microbes by recognizing pathogen-associated molecular patterns. To detect many Gram-negative bacteria, animals use the CD14–MD-2–TLR4 receptor mechanism to recognize the lipid A moiety of the cell wall lipopolysaccharide (LPS). Lipid A is a glucosamine disaccharide that carries phosphates at positions 1 and 4′ and usually has four primary (glucosamine-linked) hydroxyacyl chains and one or more secondary acyl chains. Gram-negative bacteria produce numerous variations on this basic structure, yet sensitive LPS recognition and pro-inflammatory signaling by human TLR4 occur only when lipid A has both phosphates and is hexaacyl, with two secondary acyl chains. 
 
What might bacteria derive from producing this type of lipid A, and what do animals gain from recognizing it? A survey of diverse lipid A structures found that the best-recognized configuration is produced by most of the aerobic or facultatively anaerobic Gram-negative bacteria that can live in the gastrointestinal and upper respiratory tracts. We hypothesize that the CD14–MD-2–TLR4 mechanism evolved to recognize not just pathogens, but also many of the commensals (normal flora) and colonizers that can inhabit the body's most vulnerable surfaces. Producing this lipid A structure seems to favor bacterial persistence on host mucosae, whereas recognizing it allows the host to kill invading bacteria within subepithelial tissues and prevent dissemination. A conserved host lipase can then limit the inflammatory response by removing a key feature of the lipid A signal, the secondary acyl chains.

Many other disease-associated Gram-negative bacteria have nonmucosal habitats. Their lipid A moieties differ from the typical mucosal structure by having shorter or longer acyl chains, unsaturated acyl chains, only four or five chains, or only one phosphate (see Table S1): Legionella (water habitat, often in free-living amoebae), Burkholderia pseudomallei (soil and water), Yersinia pestis (small rodents, lice), Coxiella burnetti (intracellular, livestock), Leptospira (water, animal urine), and Francisella tularensis (ticks, rabbits, other small animals). These pathogens usually enter vertebrate tissues via insect bites or cuts, within inhaled droplets, or across the conjunctivae. Brucellae (livestock), which inhabit macrophages yet are typically acquired via ingestion, also produce a nonmucosal LPS [14].

How Animals Sense Mucosal Gram-Negative Bacteria: Shield as Signal
Whereas bacterial peptide resistance and outer membrane impermeability seem to vary directly with the number of acyl chains, the inflammation-inducing CD14-MD-2-TLR4 sensory mechanism best recognizes lipid A that has the hexaacyl mucosal lipid A structure [15-20] (Figure 1). In support, Hajjar et al. [21] reported that a discrete extracellular region of human TLR4 enables recognition of hexaacyl, but not pentaacyl, P. aeruginosa LPS. Further discrimination is performed by MD-2 [22]. The same recognition pattern has been found for all mammals tested except rodents [23].
Evidence that lipid A structure influences the recognition of intact bacteria by host cells came from mutating enzymes that attach secondary acyl chains to the backbone. Somerville et al. [ The most obvious exceptions are the pathogenic Neisseriae. Both N. meningitidis and N. gonorrhoeae produce a mucosal lipid A and colonize mucosal surfaces. How they invade the bloodstream is not known [51], but they usually seem to do so without triggering local inflammation. They illustrate the important point that the lipid A-TLR4 interaction is but one element of the confrontation between bacterial pathogen and animal host.

Destroying the Signal: Acyloxyacyl Hydrolysis
Whereas Dictyostelium discoideum produces several lipid Adeacylating enzymes, only one has been found in mammals. Acyloxyacyl hydrolase (AOAH) removes only the secondary chains from lipid A; it cleaves saturated, short secondary chains, as are found in the mucosal lipid A structure, more rapidly than it removes long unsaturated ones [52,53]. A phylogenetic analysis revealed high conservation for both the AOAH large subunit, which has the bacterial GDSL lipase

motif [54], and the small subunit, a member of the saposinlike protein family [55] and the likely LPS recognition motif (Figure 2). Indeed, AOAH has evidently been more highly conserved than has TLR4 [29]
In vertebrates, AOAH is produced by neutrophils, dendritic cells, renal cortical epithelial cells, and monocytemacrophages. AOAH treatment greatly reduces LPS sensing via TLR4 [52], and LPS may remain stimulatory for weeks in mice that cannot deacylate it [56]. AOAH thus can limit inflammatory responses to bacteria that produce mucosal lipid A. Deacylation occurs slowly, reaching completion after the early recognition phase of antibacterial innate immunity has occurred.

Other Signals
The host response to LPS also has noninflammatory, immunostimulatory elements. Lipid A analogs that lack the optimal configuration for inducing inflammation may be excellent adjuvants, enhancing acquired immune responses in ways that mimic those induced by LPS itself [57,58]. The mucosal lipid A motif triggers inflammation (and toxicity), whereas adjuvanticity may also follow TLR4-based recognition of lipid A molecules that have only one phosphate and secondary chains of various lengths, numbers, and/or configurations. These structure-function relationships have been exploited to produce analogs that are either LPS antagonists or nontoxic adjuvants.
LPS recognition by CD14-MD-2-TLR4 has received intensive study because it initiates the inflammatory response to so many disease-associated Gram-negative bacteria. Less is known about how animals sense their far more abundant flora of strictly anaerobic Gram-negative bacteria, although doing so may be important for establishing beneficial mutualism between bacteria and host [59,60]. Like the tetraacylated LPS of Porphyromonas gingivalis, the pentaacylated monophosphoryl LPS of Bacteroides fragilis seems to be sensed principally by TLR2 [61-63] and can inhibit recognition of mucosal LPS by TLR4 [64,65].

Conclusions
An immune system that only recognizes pathogens would leave animals vulnerable to the commensal and colonizing microbes that enter subepithelial tissues at sites of microtrauma throughout life [31]. An innate defense that detects and responds to mucosal commensals as well as pathogens is obviously not impenetrable, however; even commensals may induce damaging responses when host defenses are impaired by trauma, cuts, or tubes that provide conduits across epithelia, immunosuppression, or an inherited immune defect [29,66]. An even greater gap in host defense may be exposed when a Gram-negative pathogen evades TLR4 recognition by producing a nonmucosal lipid A.
If the synthesis proposed here is correct, it would not be surprising to learn that other elements of innate immunity also sense commensal microbes. Animals may also have conserved enzymatic mechanisms for extinguishing microbial signals that are sensed via other receptors [67,68]. "

Figure 2. AOAH Phylogenetic Tree
Topological algorithm derived using DisplayFam [69] analysis of available sequences. Amino acid similarity/identity to the full-length human sequence is shown (includes both subunits and the pro-peptide). AOAHlike sequences have not been found in fish or insects. The asterisk indicates that AOAH-like enzymatic activity has been demonstrated in one or more cell types. See Table S2 for accession numbers.