Persistently Active Microbial Molecules Prolong Innate Immune Tolerance In Vivo

Measures that bolster the resolution phase of infectious diseases may offer new opportunities for improving outcome. Here we show that inactivation of microbial lipopolysaccharides (LPS) can be required for animals to recover from the innate immune tolerance that follows exposure to Gram-negative bacteria. When wildtype mice are exposed to small parenteral doses of LPS or Gram-negative bacteria, their macrophages become reprogrammed (tolerant) for a few days before they resume normal function. Mice that are unable to inactivate LPS, in contrast, remain tolerant for several months; during this time they respond sluggishly to Gram-negative bacterial challenge, with high mortality. We show here that prolonged macrophage reprogramming is maintained in vivo by the persistence of stimulatory LPS molecules within the cells' in vivo environment, where naïve cells can acquire LPS via cell-cell contact or from the extracellular fluid. The findings provide strong evidence that inactivation of a stimulatory microbial molecule can be required for animals to regain immune homeostasis following parenteral exposure to bacteria. Measures that disable microbial molecules might enhance resolution of tissue inflammation and help restore innate defenses in individuals recovering from many different infectious diseases.


Introduction
A great deal is known about how animals kill bacteria but very little about what happens to the corpses. Most bacterial structures seem to be broken down by phagocytes [1,2], yet much of the catabolic machinery is made by the bacteria themselves (phospholipases [3], peptidoglycan hydrolases [4], autolysins), which in some cases produce agonists that can elicit inflammation. Host enzymes may also participate, attacking peptidoglycan (lysozyme, peptidoglycan binding proteins [5,6]), lipids (phospholipases [3]), chitin [7], proteins (cathepsins, others), lipopolysaccharides (LPS) and, presumably, DNA. It is likely that the breakdown products are excreted, retained within phagocytes (e.g., in lymph nodes or arterial walls [8,9]) or in discrete extracellular deposits [10], or recycled by the host.
The seminal studies of Cohn [1,11], Elsbach [2,3] and Schwab [12] established that degradation of ingested bacteria by phagocytes can be incomplete. More recently, deposits of microbial antigens have been discovered in animals following infection with B. burgdorferii, the etiological agent of Lyme Disease [10]. A key question raised by these findings has remained unanswered to date: to what extent does the inactivation of stimulatory microbial molecules influence the outcome of infectious diseases or host reactions to microbe-rich environmental exposures? Here we show that for one important agonist, bacterial lipopolysaccharide (LPS), persistently active molecules can greatly delay immune recovery in vivo.
Exposure to Gram-negative bacterial LPS induces humans and many other mammals to enter a transient state of altered immune responsiveness known as cellular reprogramming or tolerance [13][14][15]. The phenomenon has also been produced in animals by peritonitis [16], influenza virus infection [17] and bacterial lipopeptides [18,19] and in cultured macrophages using muramyl dipeptide [20], lipoteichoic acid [21], and flagella [22]. For a period that may last from a few hours to a few days after exposure to the microbial agonist, tolerant animals and cells respond to a second exposure by producing reduced amounts of many pro-inflammatory cytokines while maintaining, or even increasing, their production of certain anti-inflammatory and anti-infective molecules. Tolerance wanes as inflammation resolves and the animal regains normal immune responsiveness. Long considered a mechanism for preventing inflammation-induced injury, innate immune tolerance may also be immunosuppressive [23].
Acyloxyacyl hydrolase (AOAH), an enzyme produced by macrophages, neutrophils, and dendritic cells, inactivates bioactive LPSs by removing two of the six fatty acyl chains that are present in the bioactive hexaacyl lipid A moiety [24]. In mice that lack AOAH, a single intraperitoneal exposure to as little as 80 ng of hexaacyl LPS induces tolerance that lasts for several weeks, much longer than that seen in wildtype animals that have received much larger doses [25]. Tolerant animals respond sluggishly to Gramnegative bacterial challenge and are unable to prevent bacterial multiplication in vivo [25].
The mechanism(s) by which fully acylated LPS maintains reprogramming in vivo have been uncertain. Here we considered several possibilities. First, it seemed likely that bioactive LPS would remain within or on Aoah 2/2 macrophages for prolonged periods and render them tolerant via cell-intrinsic signaling, despite the existence of mechanisms that promote LPS efflux from macrophages [26,27]. Second, it was possible that extracellular bioactive LPS, released from tolerant macrophages and/or other in vivo reservoirs, could prevent tolerant macrophages from recovering and induce tolerance in recruited naïve monocytes. A third consideration was that mediators produced by tolerant cells, even in the absence of LPS, could induce tolerance in themselves and other cells in vivo [21]. Finally, it was possible that LPS-stimulated Aoah 2/2 macrophages might undergo long-lived, stable reprogramming that persisted even in the absence of bioactive LPS [28][29][30][31][32]. A combination of these mechanisms was also considered.
The studies described here indicate that macrophage tolerance can be maintained for long periods in vivo by the presence of small amounts of fully acylated extracellular LPS. The source of the LPS seems to be extrinsic to the tolerant cell, coming from the fluid in which macrophages live or the LPS-containing cells they contact in vivo. We found that cell-associated LPS can be released, bind to other cells, and induce or maintain their tolerant state. Importantly, LPS-exposed, tolerant macrophages regained normal responsiveness when transferred to a LPS-free environment. Furthermore, in vivo inactivation of LPS by administering recombinant AOAH partially prevented tolerance. These results identify persistence of bioactive LPS in both cells and cell-extrinsic reservoirs as a primary mechanism that drives prolonged macrophage tolerance in vivo. They suggest that measures to inactivate LPS in these reservoirs might shorten the period of macrophage unresponsiveness that follows many Gram-negative bacterial diseases. They also provide evidence that inactivation of microbial molecules can be an essential element of the resolution/ recovery phase of infectious illnesses.

Macrophages retain intracellular LPS for long periods in vivo
We previously found that both Aoah +/+ and Aoah 2/2 peritoneal macrophages retain LPS for at least 10 days in vivo [25]. Whereas the LPS in Aoah +/+ macrophages had been partially deacylated (i.e., it had lost two of the six fatty acyl chains from lipid A), that in Aoah 2/2 macrophages was fully acylated [25]. The cells' ability to produce TNF in response to a second exposure to LPS was related inversely to their LPS content; Aoah +/+ macrophages were almost 20-fold more responsive than were Aoah 2/2 macrophages. It thus seemed likely that cell-associated LPS, if fully acylated, could maintain macrophage tolerance for long periods in vivo. To localize the cell-associated LPS, we injected FITC-labeled LPS i.p. to Aoah 2/2 and Aoah +/+ mice and harvested their peritoneal macrophages 10 days later. Anti-FITC antibodies were used to detect cell-associated LPS. We found that the majority of the LPS was intracellular (Fig. 2 A-C). Aoah +/+ macrophages contained more LPS per cell than did Aoah 2/2 macrophages, because there were more macrophages in Aoah 2/2 mouse peritoneum after i.p. LPS injection [25]. Anticipating that there would be differences in the intracellular localization of acylated and partially deacylated LPS, we then studied the macrophages using immunofluorescence microscopy. We found LPS co-localized with the lysosome marker, LAMP1 ( Fig. 2 D-H), but not with markers for ER (Calnexin), cis and medial Golgi (Giantin), trans-Golgi (TGN46) or early endosomes (Rab5a) (not shown) [34]. There was no evident difference in the intracellular location of acylated LPS (in Aoah 2/2 cells) and partially deacylated LPS (in Aoah +/+ cells). Thus, both Aoah +/+ and Aoah 2/2 peritoneal macrophages contain LPS in endolysosomes for at least 10 days after i.p. injection; at this time Aoah 2/2 macrophages are tolerant and Aoah +/+ macrophages are not, consistent with the tolerant state being determined by LPS acylation status rather than by differential LPS localization within the cells.

Author Summary
We showed previously that mice lacking acyloxyacyl hydrolase (AOAH), the host enzyme that inactivates Gram-negative bacterial lipopolysaccharides (LPS), are unable to regain normal immune responsiveness for many weeks/months after they are exposed in vivo to a small amount of LPS or Gram-negative bacteria. The many possible explanations for slow recovery included longlasting epigenetic changes in macrophages or other host cells, chronically stimulated cells that produce certain mediators, and persistent signaling by internalized LPS within macrophages. Using several in vivo techniques to study peritoneal macrophages, we found that none of these mechanisms was correct. Rather, prolonged recovery is caused by intact LPS that remains in the environment where macrophages live and can pass from one cell to another in vivo. This is the first evidence that the persistence of a bioactive microbial agonist, per se, can prevent resolution of inflammation in vivo. It also identifies the stimulatory microbial molecule as a realistic target for intervention -in further support, we found that providing recombinant AOAH can be partially preventive. In a larger sense, showing that chemical inactivation of one important microbial signaling molecule is required for full recovery should encourage efforts to find out whether disabling other microbial agonists (chitin, lipopeptides, flagella, others) also benefits infected animals.

The peritoneal environment maintains macrophage reprogramming
We then tested whether prolonged tolerance in vivo is due to retention of bioactive LPS in Aoah 2/2 peritoneal macrophages or conferred by the peritoneal environment in which the macrophages reside. In these and subsequent transfer experiments, donor and recipient macrophages were identified by their surface expression of CD45.1 or CD45.2 using flow cytometry. We transferred Aoah +/+ or Aoah 2/2 peritoneal cells to Aoah +/+ or Aoah 2/2 recipient mice and injected LPS i.p. 24 hours later. Fourteen days after injection, we harvested peritoneal cells and stimulated them ex vivo with LPS while blocking protein secretion with Brefeldin A. We then used flow cytometry to identify F4/80+ macrophages, gated to distinguish CD45.1 cells from CD45.2 cells, and measured macrophage intracellular IL-6 and TNFa as indices of LPS responsiveness (see example in Fig. S2). We found that Aoah +/+ macrophages were tolerant 14 days after they were transferred into Aoah 2/2 mice, whereas Aoah 2/2 macrophages exhibited reduced tolerance after they were transferred into Aoah +/ + mice (Fig. 3A). Aoah +/+ donor macrophages transferred into LPS-injected Aoah 2/2 mice also had lower surface levels of F4/80 and CD86 when studied 14 days after LPS injection, confirming that Aoah +/+ macrophages gained the tolerant phenotype in LPSinjected Aoah 2/2 peritoneum (Fig. 3B, C). The findings were similar whether we transferred CD45.1 donor cells to CD45.2 recipient mice or vice versa. The presence of reprogramming 14 days after i.p. LPS injection was thus determined by the recipient environment, not by donor macrophage expression of AOAH or the lack thereof.
If macrophage tolerance is maintained for prolonged periods by environmental cues, removing tolerant Aoah 2/2 macrophages from an LPS-containing environment should allow them to regain responsiveness to LPS. We injected Aoah 2/2 mice i.p. with LPS to produce tolerant macrophages. Fourteen days after injection, the peritoneal cells were harvested, washed and transferred to the peritoneal cavity of a naïve Aoah +/+ or Aoah 2/2 mouse. Seven days after transfer, we tested whether the donor Aoah 2/2 macrophages remained tolerant. We found that the donor Aoah 2/2 macrophages recovered from tolerance after they were transferred into either Aoah 2/2 or Aoah +/+ mice ( Fig. 3D-F); removal from an LPS-containing environment thus allowed macrophage recovery even in the absence of AOAH. The results again suggest that the in vivo environment plays a pivotal role in determining the behavior of peritoneal macrophages.

Bioactive LPS can be recovered from LPS-injected
To obtain direct evidence for the presence of LPS in the peritoneum many days after i.p. injection, we injected 10 mg [ 3 H/ 14 C]LPS into the peritoneal cavities of Aoah 2/2 and Aoah +/+ mice and measured 14 C and 3 H in the peritoneal flush medium, peritoneal cells, mesenteric membranes, and fat 10 days later. The distribution of 14 C dpm, a marker for the LPS carbohydrate backbone, was similar in the presence and absence of AOAH: about 0.3% of the injected LPS was found in cell-free peritoneal flush fluid; from 6 to 9% of the injected 14 C LPS was recovered from intraperitoneal fat and 2 to 3% from the mesentery (Fig. 5A). We reported previously that 1-4% of the injected LPS could be recovered from peritoneal cells [25]. Based on the ratio of 3 H to 14 C in the samples [35], 99% (Aoah +/+ ) and 19% (Aoah 2/2 ) of the recovered LPS had been deacylated (Fig. 5B), in keeping with the results found for peritoneal cells [25].
Is the intraperitoneal acylated LPS bioactive? In other experiments, we gave 10 mg LPS i.p. to Aoah +/+ and Aoah 2/2  2) Aoah +/+ or Aoah 2/2 macrophages were transferred to Aoah +/+ or Aoah 2/2 recipient mice of the opposite CD45 genotype. 24 hours later, half of the mice in each group received 1 mg E. coli O14 LPS i.p. After 14 days, the peritoneal cells from the host mice were harvested. Half of the cells was treated with 1 mg/ml E. coli O111 LPS ex vivo for 4 hours in the presence of Brefeldin A. Intracellular IL-6 and TNF levels in F4/80+ donor macrophages (A) were measured using flow cytometry. Blue bars, Aoah +/+ donor macrophages; red bars, Aoah 2/2 donor macrophages; diagonal markings indicate LPS exposure in vivo. The other half of the cells was used to measure F4/80+ macrophage surface expression of F4/80 (B) and CD86 without ex vivo stimulation (C). In B and C, only Aoah +/+ donor macrophages were studied; they acquired the ''tolerant'' surface phenotype (low F4/80 and CD86 surface expression) when transferred to Aoah 2/2 recipients. Tolerance tracked with the genotype of the host animal, not that of the donor macrophages. Data in A were combined from 4 experiments (n = 6-13/group); data in B and C were combined from 2 experiments, n = 6-8/group. (D-F) Tolerant Aoah 2/2 macrophages regain responsiveness in naïve mice. CD45.1 (or CD45.2) Aoah 2/2 mice were injected i.p. with 0.5 mg LPS. Fourteen days later, their peritoneal cells (including tolerant macrophages) were harvested and transferred i.p. to naïve Aoah 2/2 or Aoah +/+ mice of the opposite CD45 type. After 7 days, peritoneal cells were harvested from the recipient mice and IL-6 and TNF responses were measured in the F4/80+ macrophages after re-challenging them with LPS ex vivo (D). F4/80 and CD86 surface expression was also measured (E and F). Only results from donor macrophages are shown. ''Tolerant donor'' macrophages were macrophages from Aoah 2/2 mice that received 0.5 mg LPS i.p. 21 days earlier and were freshly isolated (in panel D) or macrophages harvested from Aoah 2/2 mice that received 0.5 mg LPS i.p. 14 days earlier and were preserved in 10% DMSO, 90% FBS at 280uC until analysis (in panels E and F). Data were combined from 3 experiments. n = 6-13. **, P,0.01; ***, P,0.001. Tolerance was lost by 7 days after LPSexposed macrophages were transferred to either naïve Aoah 2/2 or Aoah +/+ mice. doi:10.1371/journal.ppat.1003339.g003 mice and flushed their peritoneal cavities with culture medium 10 days later. The cell-free peritoneal flush medium from Aoah 2/2 mice activated naïve Tlr4 +/+ macrophages but not Tlr4 2/2 macrophages, suggesting that bioactive LPS was present in the medium (Fig. 5C). When we re-challenged the macrophages with LPS, we found that the flush medium from Aoah 2/2 mice could also render naïve Tlr4 +/+ macrophages tolerant (Fig. 5D, E). We conclude that (1) free LPS is present in the peritoneal fluid of Aoah 2/2 mice 10 days after i.p. injection; (2) this LPS is fully acylated; and (3) the acylated LPS is bioactive, able to activate naïve Tlr4 +/+ macrophages and reprogram them.

Macrophages release bioactive LPS that can reprogram naïve macrophages
If LPS that is taken up by macrophages can be released to act on other cells, fully acylated LPS released from Aoah 2/2 cells should be able to activate naïve cells and reprogram them. We first tested this idea in vitro. We injected Aoah 2/2 mice with LPS i.p. and harvested and washed their peritoneal cells 10 days later, when almost all of the LPS was associated with macrophages [25]. We co-cultured the peritoneal cells (including tolerant macrophages) with naïve Tlr4 +/+ or Tlr4 2/2 peritoneal cells (including naïve macrophages) for 18 hours. Significantly higher IL-6 and IL-10 ( Fig. 6A, B) levels were found in the culture medium of Tlr4 +/+ peritoneal cells co-cultured with tolerant Aoah 2/2 cells than in the medium of Tlr4 2/2 cells co-cultured with tolerant cells, suggesting that LPS from tolerant cells stimulated naïve Tlr4 +/+ cells to produce these cytokines. After incubation for 18 hours, the cocultured cells were washed twice with cRPMI and the adherent macrophages were treated with LPS for 6 hours (Fig. 6C-E). Naïve macrophages co-cultured with tolerant cells produced less TNF and IL-6, but similar amounts of IL-10, suggesting that they had become reprogrammed during the 18 hour incubation period.
When we co-cultured naïve Tlr4 +/+ peritoneal cells with Tlr4 2/2 Aoah 2/2 peritoneal cells that had been exposed to LPS in vivo, the naïve cells could also be activated, confirming that LPS from previously exposed macrophages is sufficient to activate naïve macrophages (data not shown). Co-culture of naïve Tlr4 +/+ peritoneal cells with LPS-exposed Aoah +/+ cells (again, harvested 10 days after i.p. LPS injection, sufficient time for LPS inactivation to occur in AOAH-sufficient animals) did not induce IL-6 or IL-10 production (data not shown).
To find out whether direct cell-cell contact is required for naïve macrophage activation during co-culture, we separated naïve peritoneal cells from LPS-exposed peritoneal cells in transwell cultures. This significantly decreased IL-6 and IL-10 production by the naïve macrophages (Fig. 6A, B), suggesting that direct cellcell contact enables optimal delivery of LPS from one macrophage to another.
To test whether LPS-laden macrophages can release LPS to act on other macrophages in vivo, we injected 20 mg LPS-FITC i.p. to CD45.2 Aoah 2/2 Tlr4 2/2 mice. Seven days later, peritoneal cells were harvested, washed, and transferred i.p. to CD45.1 Aoah 2/2 naïve mice. After 7 days, we found that the transferred F4/80+ macrophages had lost 2/3 of their LPS-FITC and that recipient macrophages had gained small amounts of LPS-FITC, demonstrating that LPS can be released from donor macrophages and taken up by recipient macrophages in vivo (Fig. 7A, B). In addition, we found that the recipient macrophages became tolerant (Fig. 7C-E), indicating that they had been exposed to bioactive LPS (the Tlr4 2/2 donor cells should not produce tolerizing mediators).  Ten days later, the peritoneum was flushed with 2 ml cRPMI. The flush medium was centrifuged and the cell free supernatant was added to cultures of naïve Tlr4 +/+ or Tlr4 2/2 macrophages. Eighteen hours later, the culture medium was harvested to measure IL-6. Only peritoneal flush fluid from Aoah 2/2 mice elicited IL-6 production by naïve TLR4-expressing macrophages. (D, E) After incubation for 18 hours, the medium containing the flush fluid was removed, the macrophages were washed twice with cRPMI, and then they were treated with 1 mg/ml LPS for 6 hours. TNF and IL-6 in the culture medium were measured. Flush fluid from LPS-exposed Aoah 2/2 mice induced tolerance in naïve macrophages in vitro, whereas that from LPS-exposed Aoah +/+ mice did not. Data are combined from 2 experiments. N = 6-8/group. *, P,0.05; **, P,0.01. Blue bars, Aoah +/+ peritoneal flush medium overlying TLR4 +/+ cells; red bars, Aoah 2/2 flush medium overlying TLR4 +/+ cells; light blue bars, Aoah +/+ flush medium overlying TLR4 2/2 cells; pink bars, Aoah 2/2 flush medium overlying TLR4 2/2 cells; diagonal markings indicate in vivo LPS exposure; orange bar, cRPMI medium overlying TLR4 +/+ cells. doi:10.1371/journal.ppat.1003339.g005 These results show that macrophages can release the LPS that they contain and that the released LPS can act on other cells in vivo. If the LPS remains bioactive, it can induce tolerance in other macrophages.

LPS-induced mediator(s) can induce tolerance in Tlr4 2/2 macrophages
Bioactive LPS is thus present in the peritoneum of LPS-injected Aoah 2/2 mice, where it is sufficient to prevent macrophages from regaining responsiveness. Do LPS-induced mediators also play a role in maintaining tolerance? We transferred CD45.2 Aoah 2/ 2 Tlr4 2/2 peritoneal cells to CD45.1 Aoah +/+ or Aoah 2/2 (both Tlr4 +/+ ) recipient mice, injected LPS i.p., and asked whether the Aoah 2/2 Tlr4 2/2 donor macrophages became tolerant in the peritoneum of tolerant Aoah 2/2 recipient mice. Fourteen days after LPS injection, we harvested recipient peritoneal cells and treated them ex vivo with Micrococcus luteus plus poly I:C (LPSexposed Aoah 2/2 macrophages express hetero-tolerance [36] to M. luteus [TLR2 agonist] and poly I:C [TLR3 agonist] [25]). Aoah 2/2 Tlr4 2/2 donor macrophages (F4/80+) harvested from LPS-injected Aoah 2/2 recipients had significantly lower IL-6 and TNF responses to ex vivo re-challenge (Fig. 8A), and they also had less surface F4/80 and CD86 expression (Fig. 8B, C). Since Aoah 2/2 Tlr4 2/2 donor macrophages cannot sense LPS, these results suggested that LPS-induced mediators in the peritoneum are also important for prolonging tolerance in Aoah 2/2 mice. Aoah 2/2 Tlr4 2/2 donor cells were less tolerant than were the recipient's Aoah 2/2 Tlr4 +/+ macrophages, again Peritoneal cells from naïve Tlr4 +/+ or Tlr4 2/2 mice were co-cultured for 18 hours at 37uC with washed peritoneal cells from Aoah 2/2 mice that had been injected with 10 mg LPS i.p. 10 days earlier. IL-6 and IL-10 were measured in culture medium. Only Tlr4 +/+ cells co-cultured with tolerant peritoneal cells released IL-6 and IL-10. Separation of naïve Tlr4 +/+ cells from LPS-exposed Aoah 2/2 cells in Transwell cultures significantly decreased cytokine production, suggesting that cell-cell contact is important for delivery of bioactive LPS. (C-E) In the same experiments, after incubation for 18 hours the Tlr4 +/+ cells that had been cocultured with tolerant cells were washed with cRPMI twice and then treated with 1 mg/ml LPS for 6 hours before TNF, IL-6 and IL-10 were measured in the culture medium. Naïve macrophages co-cultured with tolerant cells became reprogrammed. Blue bars, Tlr4 +/+ naïve macrophages; light blue bars, rhAOAH prevents prolonged tolerance in vivo IFN-c can improve monocyte function in septic patients and both IFN-c and GM-CSF can restore responses of LPSdesensitized (tolerant) monocytes in vitro [37,38]. We next tested whether Aoah 2/2 tolerant macrophages can recover their responses to LPS by treating them with rhAOAH (recombinant human AOAH), IFN-c , GM-CSF or a combination of these agents for 18 hours ex vivo (Fig. S3A-C). rhAOAH only slightly increased the macrophages' responses, possibly because deacylation occurs slowly and would not be expected to reach completion within 18 hours [39]. GM-CSF preferentially restored the IL-6 response while IFN-c mainly boosted the TNF and RANTES responses. Combining the three agents largely restored the macrophages' responses. Aoah 2/2 tolerant macrophages, like desensitized human monocytes [38], could thus be rescued by treating them with IFN-c and GM-CSF in vitro.
To test whether providing AOAH to mice can prevent or reverse prolonged tolerance in vivo, we gave Aoah 2/2 mice LPS i.p. on day 0, followed by rhAOAH or carrier protein BSA i.p. daily from day 1 to day 13. Peritoneal cells were harvested on day 14, plated, and adherent macrophages were rechallenged ex vivo with LPS. LPS-stimulated RANTES and IL-6 were at naïve Aoah 2/2 macrophage levels in macrophages from animals that had received rhAOAH treatment (Fig. 9A-C). TNF production is the most sensitively reprogrammed component of tolerance in these cells; here the TNF level in culture medium overlying macrophages from LPS-exposed rAOAH-treated mice was significantly lower than the naïve macrophage level but 10-fold higher than the levels produced by macrophages from mice that received carrier protein BSA instead of rhAOAH. Recombinant AOAH was thus able to ameliorate prolonged tolerance in vivo in Aoah 2/2 animals. Treatment with IFN-c or antibody to IL-10 receptor did not rescue Aoah 2/2 animals from prolonged endotoxin tolerance (data not shown), in line with the conclusion that bioactive LPS plays a dominant role in maintaining prolonged tolerance in vivo.

Discussion
Patients who experience severe sepsis develop a state of immune tolerance that may last for many weeks and is thought to be immunosuppressive [40]. The phenomenon has been characterized principally by studying peripheral blood leukocytes, whose responses to LPS and other microbial agonists are typically altered to diminish pro-inflammatory cytokine production while maintaining or increasing production of anti-inflammatory molecules such as IL-10 and IL-1 receptor antagonist. LPS-induced tolerance in mice mimics the human phenomenon in many ways (reduced monocyte-macrophage CD86, reprogrammed cytokine production) but not in others (e.g., murine macrophages do not decrease class II molecule expression) [23].
Although some Gram-negative bacteria can modify the acylation structure of their LPS in ways that may alter its ability to trigger signaling via MD-2-TLR4 [41], LPS is also disabled by host enzymes, either on mucosal surfaces (alkaline phosphatase) [42] or in tissues (AOAH) [24]. In addition to experiencing prolonged tolerance, mice that lack AOAH respond to small subcutaneous or intravenous doses of LPS by producing high titers of polyclonal antibodies [43] and developing massive, prolonged hepatomegaly [44,45]. Animals have many other mechanisms for neutralizing hexaacyl LPS in plasma and tissues [46], yet none of these is able to prevent these striking reactions in animals that cannot deacylate LPS. Transgenic mice that produce greater than normal amounts of AOAH are protected from live E. coli challenge [47], again emphasizing the importance of AOAH mediated LPS inactivation in optimizing anti-bacterial immune responses.
Continued exposure to microbial agonists can prolong the activation of cultured cells. Hume et al. reported that continuous exposure to LPS induced a sustained activation state in a macrophage cell line [48], and Hedl et al. found that prolonged exposure to muramyl dipeptide, a ligand for Nod2, promoted tolerance in human monocyte-derived macrophages [20]. Tolerance has also been described in human cells that may be exposed repeatedly to LPS in vivo, such as the alveolar macrophages of tobacco smokers [49] and blood monocytes from patients with uncontrolled gram negative bacterial infection [50] or cystic fibrosis [51]. Tolerance lasts for weeks in patients who have chronic pyelonephritis with active bacterial urinary tract infection [52], as well as in volunteers with typhoid fever [53]. Individuals who inhale endotoxin-rich agricultural dusts may also develop chronic macrophage activation or tolerance [54,55]. Here we show that tolerance can be maintained in vivo for long periods by the presence of small amounts of bioactive LPS, raising the possibility that the rate and/or extent of LPS inactivation might influence the rate of recovery from many Gram-negative bacterial diseases.
The mammalian MD-2-TLR4 LPS receptor is activated most sensitively by LPSs that have a lipid A moiety that contains 6 acyl chains. Such ''hexaacyl'' LPSs are produced by many of the Gram-negative bacterial commensals and pathogens that can inhabit the mucosal surfaces of the upper respiratory and gastrointestinal tracts [56,57]. Since AOAH inactivates these LPSs, AOAH deficiency might be associated with greater susceptibility to, or duration of, Gram-negative bacterial diseases that involve mucosal surfaces. To date, genetic linkage studies have found associations between polymorphisms in the AOAH gene and rhinosinusitis (confirmed in 2 populations of different ethnic composition [58,59]) as well as asthma [60] in humans. In addition, two studies of large populations [61,62] have independently found that AOAH mRNA expression is associated in trans with polymorphisms in HLA-DRB1 that, in turn, have been strongly linked to colitis and primary biliary cirrhosis. How AOAH influences disease expression, if it does, remains uncertain.
The present studies identified the presence of cell-associated and extracellular (cell-extrinsic) LPS as the primary determinant of prolonged LPS tolerance in macrophages in vivo. This conclusion is supported by the observations that 1) prolonged tolerance could be induced in either Aoah 2/2 or Aoah +/+ macrophages that had been transferred into LPS-injected Aoah 2/2 mice, and 2) Aoah 2/2 macrophages did not exhibit prolonged tolerance when they were transplanted into Aoah +/+ hosts, indicating that LPS inactivation by AOAH in the host environment, not in the tolerant cell itself, determines the tolerant phenotype. Tolerance could be induced in naïve macrophages either by direct contact with cells that had taken up LPS or by extracellular LPS acquired within the peritoneal environment. In addition, administration of rAOAH to Aoah 2/2 mice partially prevented tolerance, and tolerance could be induced in naïve macrophages when they were transferred into LPS-exposed Aoah 2/2 Tlr4 2/2 mice or co-cultured with LPSexposed Aoah 2/2 Tlr4 2/2 peritoneal cells, which do not produce LPS-induced mediators but can release fully acylated LPS into their environment [26,63,64]. Although these studies identify bioactive LPS as essential for maintaining tolerance, we also found that LPS-induced paracrine mediators further promoted tolerance, perhaps in part by extending the phenotype to cells that do not express TLR4. These results also do not exclude the possibility that the epigenetic changes shown to be induced by short-term exposure to LPS [28][29][30][31][32] contribute to maintaining prolonged tolerance in macrophages, but these changes evidently do not persist (or maintain dominance) in an environment that lacks extracellular LPS; they can also be overcome by soluble mediators such as interferon-c or GM-CSF.
Where does LPS deacylation occur in vivo? LPS may be deacylated extracellularly, as was shown using purulent ascites fluid [65]; both LPS-binding protein and CD14 can present LPS to extracellular AOAH in a manner that promotes its deacylation [66]. Alternatively, LPS may be taken up by macrophages, neutrophils, or dendritic cells and inactivated by AOAH, or conceivably LPS could be deacylated by AOAH in cells that have acquired AOAH via mannose-6-phosphate receptors [67]. Recent studies found that the LPS in circulating LPS-HDL (high density lipoprotein) complexes undergoes deacylation in the liver [68], where AOAH is produced by Kupffer and dendritic cells [44]. AOAH-deficient mice cannot deacylate LPS in ascites or in cells within, or on the walls of, the peritoneal cavity, and these cells or membranes may then become reservoirs that release small amounts of bioactive LPS over time. This LPS then could maintain macrophage tolerance or induce tolerance in monocytes that newly arrive in the peritoneal fluid. The LPS ''depot'' includes the peritoneal membrane and/or mesenteric fat, since we found significant quantities of radiolabeled, fully acylated LPS in these sites. Naïve macrophages may also acquire LPS from other macrophages by direct cell-cell contact.
2 independent experiments. n = 5-7. Aoah 2/2 Tlr4 2/2 macrophages, which are unable to respond to LPS, became tolerant to TLR2 and TLR3 ligands (A) and developed the tolerant surface phenotype (B and C) when they were transferred into LPS-primed Aoah 2/2 mice, supporting a role for non-LPS stimuli in promoting tolerance in vivo. Pink bars, Aoah 2/2 Tlr4 2/2 donor macrophages; Blue bars, Aoah +/+ recipient macrophages; red bars, Aoah 2/2 recipient macrophages; diagonal markings indicate in vivo LPS exposure. doi:10.1371/journal.ppat.1003339.g008 Figure 9. rhAOAH prevents prolonged endotoxin tolerance in vivo. Aoah 2/2 mice were injected with 1 mg LPS i.p. on day 0. From days 1 to 13, mice were given daily i.p. doses of 0.3 mg rhAOAH or placebo (carrier protein BSA in PBS) (diagonal marking red bars). Peritoneal cells were explanted on day 14 and the adherent macrophages were challenged ex vivo with LPS for 6 hrs before medium cytokine levels were measured. (A), IL-6; (B), TNF; (C), RANTES. Naïve Aoah 2/2 peritoneal macrophages were used as controls (solid red bars). **, P,0.01; ***, P,0.001. rhAOAH largely prevented prolonged tolerance in vivo. doi:10.1371/journal.ppat.1003339.g009 Innate immune reactions to microbes are typically short-lived. Mobilizing an animal's antimicrobial armamentarium usually promotes microbial eradication and clearance within hours or a few days. Potentially harmful inflammation then resolves as the battlefield is cleared and defenses are restored [69]. Recovery is thought to involve both anti-inflammation (preventing inflammation-induced damage) and resolution (clearing the battlefield and promoting return of homeostasis). Known tissue resolution mechanisms include neutrophil apoptosis, macrophage emigration and efferocytosis of dead cells, and the production of lipoxins, resolvins [and other lipids] [70], proteases, and gaseous signals that promote restoration of homeostasis in tissues [69,71,72]. Here we present evidence for another essential component of resolution: inactivating the microbial molecules that tell the host that microbes are present. A host's ability to remove or disable bioactive microbial molecules from an infected tissue may influence the ultimate outcome of many host-microbe encounters, since inactivating these molecules removes an important obstacle to resolution of inflammation and restoration of innate host defenses.

Ex vivo tolerance experiments
Mice were injected i.p. with various doses of LPS in 300 ml of PBS. Fourteen days later, the mice were euthanized using CO 2 and their peritoneal cells were harvested by flushing the peritoneum with 5 ml of PBS containing 5 mM EDTA. Cells were stained with antibodies and analyzed using flow cytometry. To measure cell responses to re-challenge, peritoneal cells were washed and resuspended in cRPMI medium (RPMI 1640 containing 10% heat-inactivated FBS (endotoxin ,0.06 EU/ml; Hyclone), 100 mM nonessential amino acids, 100 U/ml penicillin, 0.1 mg/ml streptomycin, 2 mM L-glutamine, 10 mM sodium pyruvate, 25 mM Hepes, pH 7.4, and 50 mM 2-mercaptoethanol) and plated with 1610 6 cells/well in 12-well-plates. After incubation for 18 hours (37uC, 5% CO 2 , 80% humidity), the floating cells were washed away and the adherent macrophages were treated with 1 mg/ml E. coli O111 LPS for 6 hours. The culture medium was used for ELISA and the cells were washed with PBS and lysed with PBS containing 0.1% Triton X-100 to measure protein (Biorad).

Macrophage-associated LPS and imaging studies
Mice were injected i.p. with 10 mg LPS-FITC. Ten days after injection, peritoneal cells were harvested and stained with anti-F4/ 80 Ab to identify macrophages. The cells were then washed and fixed with 4% paraformaldehyde (PFA) or fixed and permeabilized with cytofix/cytoperm (BD). Anti-FITC PE antibody was used to detect cell-surface or total LPS-FITC by FACS. For microscopy, peritoneal cells were plated in culture dishes for 18 hours, then the floating cells were washed away and the adherent macrophages were fixed with 4% PFA before being blocked and permeabilized with 1% BSA, 25% goat serum, 0.05% saponin in PBS. Cells were stained with primary antibodies at 4uC overnight and with the secondary reagents at room temperature for 1 hour. 49,6diamidino-2-phenylindole (DAPI, Sigma) was used to stain nuclei. After washing, FluorSave aqueous mounting medium (EMD Chemicals) was applied and then coverslips were affixed. Stained sections were examined using a Leica SP5 X-WLL confocal microscope and analyzed using LAS AF Lite (Leica) software.

Transfer experiments
Donor mice were euthanized and their peritoneal cells were harvested, washed, and resuspended in PBS. Cells from mice of the same genotype/treatment were pooled and an aliquot containing 2610 6 cells in 300 ml PBS was injected i.p to each recipient mouse. Because naïve Aoah +/+ , Aoah 2/2 , Aoah 2/ 2 TLR4 2/2 , LPS-injected Aoah 2/2 mice and Aoah 2/ 2 TLR4 2/2 mice have similar per cent distribution of macrophages [25], all recipient mice received approximately the same number of donor macrophages (8610 5 ). Twenty four hours after transfer, half of the recipient mice from each group received 1 mg LPS O14 i.p. Fourteen days later, recipient mice were euthanized and their peritoneal cells were harvested, washed, and resuspended in cRPMI medium containing 1 mg/ml LPS 0111 at 37uC for 4 hours (Tlr4 +/+ macrophages) or 40 mg/ml Micrococcus luteus plus 2.5 mg/ml poly I:C for 8 hours (Tlr4 2/2 macrophages) in the presence of 3 mg/ml Brefeldin A (eBioscience). The ex vivo stimulations were performed in non-tissue culture-treated V bottom 96-well plates (Sarstedt) to minimize macrophage adhesion. After stimulation, peritoneal cells were stained with anti-F4/ 80 antibody to identify macrophages and CD45.1 and CD45.2 antibodies to differentiate donor and recipient cells (see Fig. S2). The cells were then fixed and permeabilized (eBioscience) and stained with antibodies to IL-6 and TNF a to measure intracellular responses to ex vivo stimuli. We used the per cent of the total F4/ 80+ cells that were IL-6, TNF double positive as a measure of macrophage responsiveness. Approximately 1-10610 4 donor macrophages were recovered from each recipient mouse. Fewer donor macrophages were recovered from mice that had been injected with LPS, especially Aoah 2/2 mice; this reflected the known migration of stimulated cells from the peritoneum [75]. In cases when few donor macrophages were present, at least 100 donor macrophages were measured by flow cytometry and results were compiled from at least 3 different mice in each group.
To exclude the possibility that macrophages became tolerant when cultured ex vivo with tolerant macrophages or vice versa, we harvested naïve (CD45.1) and tolerant (CD45.2) peritoneal cells, mixed them at different ratios and stimulated them with LPS for 4 hours ex vivo. The responsiveness of macrophages did not change during ex vivo co-culturing.
In some experiments (see Fig. 4B), we injected recipient mice i.p. with 1 mg LPS. Fourteen days later, naïve peritoneal cells (including naïve macrophages) were transferred i.p. to LPS-exposed mice or naïve mice. After 24 hours, peritoneal cells were harvested and the responsiveness of donor macrophages was analyzed ex vivo as described above. We also obtained tolerant macrophages from Aoah 2/2 mice that had been injected i.p with 0.5 mg LPS and transferred 2610 6 peritoneal cells (including approximately 8610 5 tolerant macrophages) to naïve Aoah 2/2 or Aoah +/+ mice. Seven days later, we analyzed whether the tolerant macrophages had regained responsiveness in the naïve hosts ex vivo (Fig. 3D-E).
ELISA assays IL-6, TNF and IL-10 ELISA kits were purchased from BD, RANTES ELISA kit was from R&D system. Manufacturer instructions were followed.

Quantitation of radiolabeled LPS
We injected 10 mg [ 3 H/ 14 C]LPS i.p. to Aoah 2/2 or Aoah +/+ mice. Mice were euthanized 10 days later. Five ml PBS containing 5 mM EDTA was used to flush each peritoneal cavity. The peritoneal fat, mesentery and livers were harvested and homogenized in PBS. Aliquots were solubilized in 1 ml 0.5% SDS with 25 mM EDTA and 5 ml Bio-safe II scintillation cocktail (Research Products International Corp), and counted with quench and spillover correction (Packard Tri-Carb 2100TR; Perkin-Elmer).

Peritoneal flush medium and co-culture experiments
Mice were given 10 mg LPS i.p. Ten days later, they were euthanized using CO 2 and 2 ml of cRPMI was used to flush the peritoneal cavity. The flush medium was centrifuged and the cellfree supernatant was collected. Peritoneal cells from naïve mice were cultured in cRPMI medium for 4 hrs to allow macrophages to adhere. The floating cells were washed away and the adherent macrophages were cultured in flush medium for 18 hours (37uC, 5% CO 2 , 80% humidity) before the medium was removed and saved for ELISA. The cells were then washed with cRPMI twice and the adherent macrophages were challenged with 1 mg/ml E. coli O111 LPS in cRPMI for 6 hours at 37uC. In co-culture experiments, 10 6 tolerant cells were mixed with 10 6 naïve cells in 12-well plates for 18 hours; the culture medium was collected for cytokine ELISA. The co-cultured cells were then washed with cRPMI twice and the adherent macrophages were stimulated with LPS for 6 hours. The culture medium was used for ELISA. To measure cytokines produced by naïve cells in the co-culture system and exclude a contribution from tolerant cells, 10 6 tolerant cells were cultured in separate wells and treated in the same manner as were co-cultured cells. The low levels of cytokines produced by tolerant cells were subtracted from those produced by co-cultured cells.
To separate tolerant cells from naïve cells, 10 6 tolerant cells were cultured in permeable Transwell inserts (Corning) overlying 10 6 naïve cells in 12-well plates for 18 hours. The control was 10 6 naïve cells in permeable inserts co-cultured with 10 6 naïve cells. The culture media were collected after 18 hours of co-culture.

In vivo rhAOAH treatment
Aoah 2/2 mice were given 1 mg LPS i.p. on day 0. From day 1 to day 13, mice were injected i.p daily with 0.3 mg rhAOAH or carrier protein BSA in 300 ml PBS. On day 14, the peritoneal cells were harvested and macrophages were challenged with 1 mg/ml E. coli O111 LPS for 6 hours at 37uC. Cytokine levels were measured in the culture medium.

Flow cytometry
Flow cytometry analysis was done on FACS Calibur or LS Rortessa (BD). BD and FlowJo software was used to analyze data.

Statistics
Unpaired Student's t test (two-tailed) was used for comparisons between groups. Linear regression was used to perform correlation analysis. In all figures, error bars indicate one SE. Figure S1 Macrophage TNF production after ex vivo LPS treatment correlates with cell SSC, F4/80 and CD86 expression. Aoah +/+ or Aoah 2/2 mice were injected i.p. with 0, 0.016, 0.08, 0.4, 2 or 10 mg E. coli O14 LPS. Each LPS dose was given to 3-5 Aoah +/+ and Aoah 2/2 mice. Fourteen days later, their peritoneal cells were harvested. Some were used to measure macrophage (F4/80+) surface markers F4/80, CD86 and SSC by flow cytometry. Others were plated in culture dishes for 18 hours and the adherent macrophages were re-stimulated with E. coli O111 LPS (1 mg/ml) for 6 hrs. Medium TNF levels were measured by ELISA and correlated with SSC (A), F4/80 (B) and CD86 (C) surface expression on macrophages from the same mice. Each dot represents data from one mouse. (TIF) Figure S2 Macrophage responses to ex vivo LPS stimulation. In this example, CD45.1 peritoneal cells were transferred into peritoneum of CD45.2 recipient mice. Twentyfour hours later, recipient mice were injected with LPS i.p. Two weeks after injection, peritoneal cells were harvested from recipient mice and re-stimulated with 1 mg/ml LPS ex vivo for 4 hours in the presence of Brefeldin A. Cell surface F4/80, CD45.1, CD45.2 were stained first and then the cells were fixed and their intracellular IL-6 and TNF were stained. F4/80+ macrophages were gated (A); donor and recipient macrophages were differentiated by CD45.1 or CD45.2 expression (B); then donor and recipient macrophage intracellular IL-6 and TNF expression was plotted (C). We used the percentage of F4/80+ cells that was positive for both IL-6 and TNF to measure the macrophages' responses. (TIF) Figure S3 Prolonged tolerance can be reversed in vitro. Peritoneal cells were harvested from 1 mg LPS i.p. injected Aoah 2/ 2 mice. Macrophages were purified by letting them adhere to 12well plates for 4 hrs and then they were treated with 400 ng/ml rhAOAH, 10 ng/ml IFNc , 10 ng/ml GM-CSF or a combination of these agents ex vivo (red bars). Eighteen hours later, the macrophages were stimulated with LPS for 6 hours and cytokine levels were measured in culture media. (A), IL-6; (B), TNF; (C), RANTES. Peritoneal macrophages from 1 mg LPS-injected Aoah +/ + mice were used as controls (blue bars). Tolerance reversal required a combination of the 3 agents. (TIF)

Supporting Information
Methods S1 Microarray analysis. IFN-c, GM-CSF and rhAOAH treatment.