Report

Intravascular Danger Signals Guide Neutrophils to Sites of Sterile Inflammation

See allHide authors and affiliations

Science  15 Oct 2010:
Vol. 330, Issue 6002, pp. 362-366
DOI: 10.1126/science.1195491

This article has a correction. Please see:

Abstract

Neutrophils are recruited from the blood to sites of sterile inflammation, where they contribute to wound healing but may also cause tissue damage. By using spinning disk confocal intravital microscopy, we examined the kinetics and molecular mechanisms of neutrophil recruitment to sites of focal hepatic necrosis in vivo. Adenosine triphosphate released from necrotic cells activated the Nlrp3 inflammasome to generate an inflammatory microenvironment that alerted circulating neutrophils to adhere within liver sinusoids. Subsequently, generation of an intravascular chemokine gradient directed neutrophil migration through healthy tissue toward foci of damage. Lastly, formyl-peptide signals released from necrotic cells guided neutrophils through nonperfused sinusoids into the injury. Thus, dynamic in vivo imaging revealed a multistep hierarchy of directional cues that guide neutrophil localization to sites of sterile inflammation.

Sterile inflammation, characterized by redness, heat, swelling, and pain, occurs when tissues are injured in the absence of infection. Necrotic cell death can generate profound sterile inflammation characterized by the accumulation of innate immune effector cells, namely neutrophils, within the affected tissue. Such responses are classically considered homeostatic “wound healing” reactions to tissue injury, in which the phagocytic functions of neutrophils contribute to the clearance of debris (1). Neutrophils, however, possess a vast arsenal of hydrolytic, oxidative, and pore-forming molecules capable of causing profound collateral tissue destruction (2). As such, overexuberant neutrophil recruitment in response to sterile inflammatory stimuli contributes to the immunopathology observed in many diseases, including ischemic injuries/infarction, trauma, autoimmunity, drug-induced liver injury, and others (1, 37). Therefore, understanding the mechanisms that allow neutrophils to respond to sterile tissue injury and cell death is fundamental to our understanding of both homeostatic innate immune functions and pathogenic immune responses in disease.

Cell death by necrosis releases multiple endogenous pro-inflammatory damage-associated molecular patterns (DAMPs), including proteins, nucleic acids, extracellular matrix components, and lipid mediators (1, 4, 810). When injected into mice, purified DAMPs or necrotic cells mobilize neutrophils to the site of inoculation (9, 11, 12). Bona fide sterile tissue injury, however, results in the death of multiple cell types, release of many DAMPs, and formation of hemostatic barriers (coagulation and thrombosis), culminating in a complex milieu of inflammatory and chemoattractant danger signals that must be translated into precise directional cues to guide neutrophil trafficking. We used in vivo imaging of the early innate immune response to reveal a multistep cascade of molecular events that guide the recruitment of neutrophils to locations of sterile injury.

We generated a murine model of focal hepatic necrosis induced by localized thermal injury on the surface of the liver and used spinning disk confocal intravital microscopy (SD-IVM) to visualize the subsequent response of neutrophils (13). Mice expressing enhanced green fluorescent protein under the control of the endogenous lysozyme M promoter (LysM-eGFP) were used to visualize the kinetics of eGFP-expressing (and Gr1+) neutrophils to a 0.022 ± 0.001 (SEM) mm3 necrotic lesion visualized by superfusion of propidium iodide over the injury area (Fig. 1A and movie S1). Within 30 to 60 min after injury, neutrophils began adhering to the microvascular endothelium (Fig. 1B). Neutrophil recruitment occurred in response to necrotic cells, because sham operation and imaging did not result in neutrophil accumulation (Fig. 1B). Neutrophil adhesion within liver sinusoids was mediated by interactions between the integrin αMβ2 (Mac1) and its endothelial ligand intercellular adhesion molecule–1 (ICAM-1) (fig. S1, A and B). In contrast, when Escherichia coli was applied to the liver surface rather than a necrotic injury, neutrophil adhesion in sinusoids was dependent on CD44 rather than Mac1, revealing different mechanisms of neutrophil recruitment to infection versus sterile inflammation (fig. S1, C to F). The sterility of the inflammatory response was confirmed by depleting mice of culturable gut flora with use of antibiotics (14) and finding no alteration in the response to hepatic necrosis (fig. S2).

Fig. 1

Neutrophils home in to sites of sterile injury by intravascular crawling. (A) Time-lapse images from SD-IVM demonstrating the response of neutrophils (green) to focal hepatic necrosis (red, propidium iodide). Scale bar indicates 200 μm. (B) Number of adherent neutrophils per field of view in response to sterile injury or sham procedure. N = 5 individual mice per group for all time points. Error bars show SEM. (C) Representative SD-IVM images at 2 and 3 hours after injury, demonstrating the intravascular (blue, Alexa-647-BSA) route taken by neutrophils (green) to reach necrotic foci (red). Arrows show path of travel of selected neutrophils. Scale bars, 100 μm. (D) Quantitative representation of the route of migration taken by neutrophils en route to tissue injury. N = 5 individual mice; error bars, SEM; *P < 0.01 by t test. (E) Animals treated with blocking antibodies against Mac1, LFA1, or isotype control (administered before injury) were imaged 2.5 hours postinjury to determine the percentage of adherent neutrophils that directionally chemotax toward the injury site. N ≥ 5 individual mice per treatment group; error bars show SEM; *P < 0.01 by one-way analysis of variance (ANOVA) with Bonferroni’s posttest. (F and G) Migration paths (F) and crawling velocities (G) of neutrophils responding to tissue injury before and 10 min after administration of Mac1 antibody. Experiments were conducted 2.5 hours postinjury. Paths are normalized for their origins (site of adhesion) and position relative to the center of the necrotic focus. N = 3 individual mice; error bars, SEM; *P < 0.01 by t test.

Neutrophils initially adhered within sinusoids around foci of injury and over time were observed to gradually accumulate within the area of necrosis (Fig. 1A and fig. S3). Of adherent neutrophils, 77.3% ± 3.3 were observed to directionally chemotax toward the necrotic tissue, ultimately infiltrating into the area of cell death (Fig. 1C). Instead of transmigrating out of the vasculature to take the shortest path toward the site of danger, the majority of neutrophils migrated via the intravascular route, which was often a less-direct course (Fig. 1, C and D, and movies S2 and S3). Inhibitory antibodies against Mac1 but not the closely related αLβ2-integrin LFA1 significantly reduced the number of neutrophils that chemotaxed intravascularly toward tissue injury compared with the number in isotype control-treated animals (Fig. 1E and movie S4). Furthermore, administration of Mac1 antibody 2.5 hours after injury abruptly stopped migrating neutrophils, demonstrating a role for Mac1 in mediating crawling and adhesion (Fig. 1, F and G). Thus, neutrophils use the vascular channels as highways to guide their transit through healthy tissue toward sites of sterile injury.

Analysis of the microvascular hemodynamics revealed an absence of perfused sinusoids within the area of necrosis and occlusion of the sinusoids immediately adjacent to the necrotic core (surrounding ~150 μm) by platelet thrombi (fig. S4, A to C). Although neutrophil crawling velocity was reduced within these areas (fig. S4D), this did not limit their ability to home into the injury site (movie S2). Beyond 150 μm from the injury border, more than 80% of sinusoids were actively perfused, and as such the majority of neutrophils migrated via perfused sinusoids without any preference for vessels flowing toward or away from the necrotic focus (fig. S4, E and F).

To investigate the molecular signals that direct neutrophil migration, we hypothesized that purinergic danger signals such as extracellular adenosine triphosphate (ATP) released by damaged cells act as a find-me signal to guide neutrophils to sites of necrosis. Previous studies have suggested that this function of ATP may be due to its ability to induce cytokine production and/or its ability to promote leukocyte migration (3, 1517). Administration of an exogenous ATPase (apyrase), which hydrolyzed extracellular ATP released after injury (fig. S5A), resulted in a marked reduction in the total number of neutrophils recruited into the liver in response to tissue injury (Fig. 2A and movie S5). In contrast, apyrase did not inhibit the ability of recruited neutrophils to chemotax toward the focus of injury (Fig. 2, B to D), demonstrating that ATP does not function as a chemotactic signal. Similarly, administration of suramin to block heterotrimeric guanine nucleotide–binding protein (G protein)–coupled P2Y purinergic receptors that have been implicated in the chemotactic response to ATP (1517) did not affect the recruitment or migration of neutrophils (fig. S5B).

Fig. 2

ATP danger signals initiate neutrophil recruitment via P2X7R signaling and Nlrp3 inflammasome activation. (A) The number of adherent neutrophils per field of view 4 hours after focal hepatic injury in mice treated with apyrase or vehicle control. N = 5 individual mice per treatment group; error bars, SEM; *P < 0.05 by t test. (B to D) Mice treated with apyrase or vehicle control were imaged 2.5 hours postinjury to determine the percentage of adherent neutrophils that chemotax toward the injury site (B) and migration paths for chemotaxing neutrophils [(C) and (D)]. Paths are normalized for their origins (site of adhesion) and position relative to the center of the necrotic focus. N = 5 individual mice per group; error bars, SEM. (E) Number of adherent neutrophils per field of view 4 hours after focal hepatic injury in indicated mouse strains [wild type (WT) and C57BL/6]. N = 5 individual mice per genotype; error bars, SEM; *P < 0.01 versus WT by one-way ANOVA with Bonferroni’s posttest. (F) Representative immunoblots for the detection of pro-IL-1β, mature (cleaved) IL-1β, pro-caspase-1, and mature (cleaved) caspase-1 in sham (uninjured) liver tissue or tissue harvested from the site of injury in the indicated mice. Representative of two independent experiments. (G) Number of adherent neutrophils per field of view 4 hours after focal hepatic injury in mice treated with recombinant IL-1R antagonist (rIL-1Ra) or antibodies against IL-1β or IL-1α. N = 5 individual mice per treatment group; error bars, SEM; *P < 0.01 versus control by one-way ANOVA with Bonferroni’s posttest. (H) Fluorescence intensity was quantified from SD-IVM images after administration of phycoerythrin (PE)–labeled antibody against ICAM-1 (PE-anti-ICAM-1), expressed as -fold increase in fluorescence intensity relative to equivalently labeled isotype control. Experiments were conducted 2.5 hours postinjury (or sham procedure) in the presence and absence of IL-1β blocking antibody. N = 3 individual mice per treatment group; error bars, SEM; *P < 0.001 versus control by one-way ANOVA with Bonferroni’s posttest.

Extracellular ATP, via P2X7 receptor signaling, is one of the best-characterized activators of the Nlrp3 inflammasome, which mediates the generation of inflammatory cytokines such as interleukin (IL)-1β (1, 18, 19). Similar to apyrase treatment, selective inhibition of P2X7 receptors with oxidized ATP or genetic deficiency in P2rx7−/− mice resulted in reduced neutrophil recruitment in response to tissue injury, without impairing the chemotactic response of the few recruited neutrophils (Fig. 2E and figs. S5B and S6). Bone marrow chimeric mice demonstrated that the target cells of P2X7R signaling were of hematopoietic origin but were not neutrophils because isolated P2rx7−/− and wild-type neutrophils were recruited equivalently to foci of necrosis after adoptive transfer (fig. S7). Furthermore, the quantity of neutrophils recruited to areas of necrosis was significantly reduced in mice deficient in inflammasome component Nlrp3 or ASC compared with wild-type animals (Fig. 2E). Consistent with recent reports that macrophages are the primary sentinel cells that sense cell death and generate pro-inflammatory cytokines (20), mice that were depleted of liver-resident intravascular macrophages (Kupffer cells) by administration of liposome-encapsulated clodronate (21) demonstrated reduced neutrophil recruitment similar to animals with impaired P2X7R signaling (fig. S8). Western blots of injured liver tissue confirmed that Nlrp3-dependent activation of caspase-1 and IL-1β processing at sites of focal hepatic necrosis was entirely dependent on P2X7R signaling (Fig. 2F). Mice that received blocking antibodies against IL-1β, a recombinant antagonist of the IL-1-receptor (IL-1R), or animals that were deficient of the signaling adaptor MyD88 (required for signaling through IL-1R) showed similarly reduced neutrophil accumulation (Fig. 2G and fig. S9A). Neutrophil adoptive transfer experiments revealed that MyD88-deficient and wild-type neutrophils were recruited equivalently to sites of injury, indicating that neutrophils were not a target of IL-1β (fig. S9B). Instead, IL-1β blockade prevented ICAM-1 up-regulation on the surface of sinusoidal endothelium in response to tissue injury (Fig. 2H). Thus, ATP danger signals activate a pathway that initiates neutrophil adhesion but do not guide neutrophil chemotaxis toward necrotic cells.

We next hypothesized that chemokines produced in response to tissue injury may guide intravascular neutrophil migration to foci of sterile inflammation. Previous studies have demonstrated that chemokines may be expressed and immobilized on the luminal surface of microvascular endothelium in vivo (22). Intravital immunofluorescence using SD-IVM revealed MIP-2 (macrophage inflammatory protein 2, CXCL2) expression on the luminal surface of the liver sinusoids that was maximal at about 150 μm from the injury border and gradually decreased out to 650 μm, demonstrating the presence of an intravascular gradient that leads toward the injured area (Fig. 3, A and B). Graded MIP-2 expression was not dependent on IL-1β but did require signaling through a MyD88-dependent pathway within nonhematopoietic cells (Fig. 3C and fig. S10A). Within the region of high intravascular MIP-2 expression, neutrophils in CXCR2-deficient animals that were unable to detect MIP-2 failed to directionally chemotax (Fig. 3, D and E) but rather migrated randomly in the vasculature (meandering index 0.29 ± 0.03, Fig. 3F). Inhibitory antibodies against MIP-2, and to a lesser extent CXCL1 (KC), prevented intravascular chemotaxis toward foci of damage, confirming that neutrophil chemotaxis is directed by a functional gradient of intravascular chemokines (Fig. 3G).

Fig. 3

An intravascular chemokine gradient guides neutrophil chemotaxis within the vasculature toward foci of sterile injury. (A) MIP-2 expression (red) and PECAM-1+ endothelium (blue) were visualized 2.5 hours after injury in the distant periphery and directly adjacent to the injury (indicated by solid white lines). Scale bars, 100 μm. (B) Fluorescence intensity of MIP-2 antibody staining from SD-IVM images was quantified within individual sinusoids at various distances from the border of necrotic injury 2.5 hours postinjury and expressed as -fold increase in fluorescence intensity relative to equivalently labeled isotype control. N = 103 sinusoids compiled from three individual mice; error bars, SEM. (C) Fluorescence intensity of MIP-2 antibody staining in sinusoids surrounding injury in mice left untreated (control) or treated with IL-1β blocking antibody before injury or in Myd88−/− mice. N ≥ 3 individual mice per treatment group; error bars, SEM; *P < 0.05 versus control by one-way ANOVA with Bonferroni’s posttest. (D) Cell migration paths for neutrophils within the zone of intravascular chemotaxis (>150 μm from injury border, see fig. S10C) in WT C57BL/6 and Cxcr2−/− mice. Paths are normalized for their origins (site of adhesion) and position relative to the center of the necrotic focus. (E) The percentage of adherent cells that chemotax toward the injury site in WT or Cxcr2−/− mice. (F) Meandering index of migrating neutrophils depicted in (D). In (D) to (F), imaging was conducted 2.5 hours postinjury; N ≥ 3 individual mice per group; error bars, SEM; *P < 0.01 by t test. (G) Mice treated with antibodies against MIP-2, KC, or isotype control were imaged 2.5 hours postinjury, and the percentage of adherent cells that chemotax toward the injury site was determined. N = 3 individual mice per treatment group; error bars, SEM; *P < 0.01 versus control by one-way ANOVA with Bonferroni’s posttest.

The intravascular gradient of MIP-2 was consistently observed to abruptly end ~100 to 150 μm proximal to the border of necrotic tissue, despite the presence of intact platelet endothelial cell adhesion molecule 1–positive (PECAM-1+) endothelium (Fig. 3, A and B). Intravenous administration of a PECAM-1–specific antibody demonstrated that sinusoidal endothelium in this malperfused area is accessible to circulating proteins. Circulating recombinant MIP-2 failed to bind within this proximal zone (fig. S10B), suggesting that the absence of MIP-2 may be due to an inability of the chemokine to become immobilized on endothelial surface glycosaminoglycans near the injury. Furthermore, within this proximal 150 μm surrounding the injury, directional neutrophil migration was independent of CXCR2 (Fig. 4, A and B).

Fig. 4

FPR1-dependent necrotaxis guides precise localization of neutrophils into areas of sterile tissue necrosis. (A and B) Cell migration paths for neutrophils within the zone of necrotaxis (proximal 150 μm surrounding injury, see fig. S10C) in WT (A) and Cxcr2−/− (B) mice 2.5 hours postinjury. Paths are normalized for their origins (site of adhesion) and position relative to the center of the necrotic focus (N ≥ 3 individual mice per group). (C) The number of human neutrophils that chemotaxed toward necrotic HEK 293 cells in vitro. Neutrophils were left untreated or incubated with IL-8, CsH, anti-FPR1, or isotype control. N ≥ 5 independent experiments; error bars, SEM; *P < 0.05 or **P < 0.01 versus control by one-way ANOVA with Bonferroni’s posttest. (D and E) Representative SD-IVM images of neutrophils (green) responding to a focus of tissue injury (red) in animals treated with vehicle control (D) or the FPR1 antagonist CsH (E). Dashed lines are at injury border and 150 μm. Scale bars, 100 μm. Representative of three animals per group. (F) Cell migration paths for neutrophils within the zone of necrotaxis in Fpr1−/− mice 2.5 hours postinjury (N = 4 independent mice). (G and H) Meandering index [(G) error bars, SEM] and crawling velocity [(H) lines, means] of migrating neutrophils depicted in (A), (B), and (F). *P < 0.05 versus WT by one-way ANOVA with Bonferroni’s posttest. (I) Number of adherent neutrophils per 10,000 μm2 within the indicated regions around necrotic foci (4× field of view) 4 hours after injury in WT, Cxcr2−/−, and Fpr1−/− mice. N ≥ 3 independent mice per genotype; error bars, SEM; **P < 0.01 versus WT by one-way ANOVA with Bonferroni’s posttest.

Given that neutrophils migrate directly into the area of cell death, we hypothesized that necrotic cells released a CXCR2-independent chemoattractant, or “necrotaxis” signal, that directs neutrophil migration beyond the intravascular chemokine gradient (fig. S10C). In support of this, neutrophils failed to enter into a focus of sterile inflammation that did not contain necrotic cells (focal vascular occlusion generated by localized compression of sinusoids) and instead simply accumulated around the injury (fig. S11). Furthermore, the observation that neutrophils rapidly migrated away from high concentrations of CXCR2 ligands implies that the necrotactic stimulus must hierarchically override CXCR2 signaling. In an in vitro under agarose chemotaxis assay, necrotic cells potently attracted human neutrophils (Fig. 4C). This attraction overrode CXCR2 signals, because incubation of neutrophils in IL-8 (human homolog of MIP-2) did not inhibit chemotaxis toward necrotic cells (Fig. 4C). Mitochondria contain DAMPs, including formylated peptides, that can direct neutrophil chemotaxis via signaling through the formyl-peptide receptor 1 [FPR1 (7)]. In vitro, blockade of neutrophil FPR1 with inhibitory antibodies or the selective antagonist cyclosporin H (CsH) significantly attenuated neutrophil chemotaxis toward necrotic cells (Fig. 4C). In vivo, treatment of LysM-eGFP mice with CsH (Fig. 4, D and E, and fig. S12) or genetic deficiency of FPR1 (Fpr1−/−, Fig. 4, F and G) resulted in nondirectional random migration within the necrotaxis zone. Importantly, FPR1 signals controlled only directionality within the necrotaxis zone, because neutrophil crawling velocities were unchanged in Fpr1−/− mice compared with those in wild-type mice (Fig. 4H). Overall, although the majority of neutrophils in wild-type mice accumulated within the necrotic focus, neutrophils in Fpr1−/− mice homed in to the proximal 150 μm around the injury, where they migrated randomly and accumulated (Fig. 4I). This pattern of neutrophil accumulation was equivalent to that seen in response to foci of vascular occlusion, where both wild-type and Fpr1−/− neutrophils accumulated in the proximal 150 μm but failed to migrate into the area of injury, presumably because of a lack of FPR1 ligands in the absence of dead cells (fig. S11).

We have identified a multistep cascade of intravascular events that allow neutrophils to sense and home in to sites of sterile inflammation in vivo (fig. S13). In contrast to recent reports (17), we show that ATP does not function as a chemoattractant but rather initiates the inflammatory response through mechanisms that lead to neutrophil adhesion. We have recently reported that signals through formyl peptide receptors in neutrophils hierarchically override signals through CXCR2, allowing neutrophils to preferentially migrate toward end-target chemoattractants in vitro (23). Our present study provides evidence that this hierarchy of neutrophil chemoattraction functions in vivo. Importantly, the migration patterns and molecular guidance cues that direct neutrophils to sites of sterile inflammation in the liver also function similarly in other organs such as the skin, which contains a vastly different cellular composition and vascular architecture (fig. S14).

Neutrophil extravasation out of the vasculature into tissues can cause substantial collateral damage during pathological inflammatory responses (1, 24). We propose that the intravascular danger sensing and recruitment mechanisms identified in this study have evolved to limit collateral damage during responses to sterile injury by allowing neutrophils to remain intravascular as they navigate through healthy tissue to sites of injury. Furthermore, necrotaxis cues are released from necrotic cells to promote localization of neutrophils directly into existing areas of injury. This is likely a means to focus the innate immune response on damaged areas and away from healthy tissue, providing an additional safeguard against collateral damage during sterile inflammatory responses.

Supporting Online Material

www.sciencemag.org/cgi/content/full/330/6002/362/DC1

Materials and Methods

Figs. S1 to S14

References

Movies S1 to S5

References and Notes

  1. Material and methods are available as supporting online material at Science Online.
  2. We thank C. Badick for excellent technical support. This study was supported by grants from the Canadian Institutes of Health Research and Alberta Innovates (Health Solutions).
View Abstract

Navigate This Article