Neutrophils scan for activated platelets to initiate inflammation

See allHide authors and affiliations

Science  05 Dec 2014:
Vol. 346, Issue 6214, pp. 1234-1238
DOI: 10.1126/science.1256478


Immune and inflammatory responses require leukocytes to migrate within and through the vasculature, a process that is facilitated by their capacity to switch to a polarized morphology with an asymmetric distribution of receptors. We report that neutrophil polarization within activated venules served to organize a protruding domain that engaged activated platelets present in the bloodstream. The selectin ligand PSGL-1 transduced signals emanating from these interactions, resulting in the redistribution of receptors that drive neutrophil migration. Consequently, neutrophils unable to polarize or to transduce signals through PSGL-1 displayed aberrant crawling, and blockade of this domain protected mice against thromboinflammatory injury. These results reveal that recruited neutrophils scan for activated platelets, and they suggest that the neutrophils’ bipolarity allows the integration of signals present at both the endothelium and the circulation before inflammation proceeds.

A two-cell collaboration for inflammation

Immune cells called neutrophils are first responders to infection. Neutrophils move within and through blood vessels to get to sites of infection quickly. Sreeramkumar et al. found that mouse neutrophils rely on platelets to help find such sites. Neutrophils extended protrusions into blood vessels. When these protrusions came into contact with platelets, the neutrophils migrated into the surrounding tissue to carry out their inflammatory functions. Preventing these neutrophilplatelet interactions alleviated collateral inflammatory damage to tissues in several injury models in mice.

Science, this issue p. 1234

Neutrophils are primary effectors of the immune response against invading pathogens but are also central mediators of inflammatory injury (1). Both functions rely on their remarkable ability to migrate within and through blood vessels. The migration of neutrophils is initiated by tethering and rolling on inflamed venules, a process mediated by endothelial selectins (2). Selectin- and chemokine-triggered activation of integrins then allows firm adhesion, after which leukocytes actively crawl on the endothelium before they extravasate or return to the circulation (3). A distinct feature of leukocytes recruited to inflamed vessels is the rapid shift from a symmetric morphology into a polarized form, in which intracellular proteins and receptors rapidly segregate (4). In this way, neutrophils generate a moving front or leading edge where the constant formation of lamellipodia (actin projections) guides movement, and a uropod or trailing edge where highly glycosylated receptors accumulate (5, 6). We deemed it unlikely that this dramatic reorganization served to exclusively generate a front-to-back axis for directional movement, and we explored the possibility that neutrophil polarization functions as an additional checkpoint during inflammation.

We performed intravital microscopy (IVM) imaging of venules in cremaster muscles of mice treated with the cytokine tumor necrosis factor–α (TNF-α), an inflammatory model in which the vast majority of recruited leukocytes are neutrophils (fig. S1). Within seconds after arresting, leukocytes formed a lamellipodia-rich domain, or leading edge, and a CD62L-enriched uropod, which we could identify by its localization opposite to the leading edge and the direction of cell movement (movie S1 and Fig. 1A) (68). Confirming previous reports, we observed numerous interactions of platelets with the leading edge of adherent neutrophils [Fig. 1A and fig. S2A (810)]. During these experiments, we noticed that the uropod underwent continuous collisions with circulating platelets, a fraction of which established measurable interactions that were usually transient (Fig. 1B and movie S2). Because platelets captured by the uropod represented a substantial fraction of all interactions (31%), we searched for the receptor(s) mediating these contacts. We reasoned that PSGL-1, a glycoprotein ligand for P-selectin (11) that segregates to the uropod of polarized neutrophils (12), could be responsible for these interactions. Analysis of mice deficient in PSGL-1 (Selplg–/– mice) revealed marked reductions in platelet interactions with the uropod, whereas those at the leading edge remained unaffected (Fig. 1B). In contrast, deficiency in the β2 integrin Mac-1 (Itgam–/–) resulted in reductions at both the uropod and leading edge (Fig. 1B). In vivo labeling of Mac-1 and PSGL-1 confirmed these functional data, with Mac-1 localized throughout the cell body and PSGL-1 exclusively at the uropod (Fig. 1C). Specifically, PSGL-1 clustered in a small region of the uropod, whereas CD62L was widely distributed in this domain (Fig. 1C). Analyses of mice expressing a functional Dock2-GFP protein (GFP, green fluorescent protein), a guanine nucleotide exchange factor of Rac GTPases (13), revealed colocalization of Dock2 with PSGL-1 clusters on crawling neutrophils (fig. S3 and movie S3), suggesting active structural dynamics within this region. This observation together with the high frequency of platelet collisions with the PSGL-1 clusters suggested that this domain might be actively protruding into the vessel lumen. Using high-speed spinning-disk IVM, we could obtain three-dimensional (3D) reconstructions of polarized neutrophils within inflamed venules of Dock2-GFP mice (Fig. 1D), demonstrating that the PSGL-1 clusters indeed projected toward the vessel lumen in about 40% of adherent neutrophils, whereas in the remaining 60% of the cells, it extended laterally, parallel to the endothelial surface (Fig. 1, D and E, and movie S4). As a consequence, the luminal space of inflamed venules was populated by multiple PSGL-1–bearing clusters suitably positioned to interact with circulating cells (Fig. 1F and movie S5).

Fig. 1 Neutrophils recruited to inflamed venules interact with activated platelets via protruding PSGL-1 clusters.

(A) Micrographs of polarized neutrophils interacting with platelets (red; yellow arrowheads) through the leading edge or the CD62L-labeled uropod (blue). (B) Quantification of total or domain-specific platelet interactions in wild-type mice or mice deficient in P-selectin (Selp–/–), PSGL-1 (Selplg–/–), or Mac-1 (Itgam–/–); n = 5 to 8 mice, 38 to 133 interactions. (C) In vivo receptor distribution on polarized wild-type neutrophils. (D) Examples of luminal and lateral projections from 3D reconstructions of polarized Dock2-GFP neutrophils (see also movie S4). (E) Frequency of neutrophils extending PSGL-1 clusters into the lumen (Lu), laterally (La) or between the cell body and the endothelium (En). n = 6 mice, 251 cells. (F) 3D reconstructions of an inflamed vessel showing the distribution of PSGL-1 clusters (movie S5). (G) Representative micrographs of neutrophils interacting with nonactivated (arrow) or activated P-selectin+ platelets (arrowhead), and quantification of interactions of each domain with P-selectin+ or JON/A+ platelets. n = 3 to 4 mice, 66 to 116 interactions. Scale bars, 10 μm. Bars show mean ± SEM. *P < 0.05; ***P < 0.001, one-way analysis of variance (ANOVA) with Tukey’s post-hoc test.

The observation that only a small fraction of circulating platelets engaged in interactions with the uropod prompted us to search for subsets of platelets prone to this behavior. In vivo labeling for P-selectin or for active β3 integrins revealed that virtually all platelets interacting with the uropod were activated (P-selectin+ or with active β3 integrins), whereas a fraction of those engaging the leading edge were not (Fig. 1G and figs. S2B and S4). These findings further suggested that P-selectin present on the surface of activated platelets might be mediating the interactions with the PSGL-1 clusters. Analyses of mice deficient in P-selectin (Selp–/– mice) indeed demonstrated patterns of platelet interactions with the two leukocyte subdomains that were similar to those found in mice lacking PSGL-1 (Fig. 1B). These results indicated that neutrophils recruited to inflamed vessels extend a PSGL-1–bearing microdomain into the vessel lumen that scans for activated platelets present in the bloodstream through P-selectin.

During the course of our IVM experiments, we also noticed alterations in the intravascular behavior of adherent neutrophils in the different mutant mice. Deficiency in Mac-1 severely compromised neutrophil crawling on the inflamed vasculature (Fig. 2, A and B), a process previously reported to be mediated by this integrin (3). Surprisingly, although PSGL-1 was excluded from the area of contact with the endothelium (Fig. 1, D and E), neutrophils deficient in this glycoprotein also displayed reductions in crawling displacement and velocity (Fig. 2, A and B), and these defects were cell-intrinsic (fig. S5 and movie S6). To exclude potential defects originating from PSGL-1 contributions in the early steps of leukocyte recruitment by binding endothelial P-selectin (14), we prevented PSGL-1 binding to P-selectin using a blocking antibody injected after neutrophils had adhered. Inhibition at this stage did not affect leukocyte adhesion to inflamed venules but specifically decreased interactions with the uropod (fig. S6) and caused reductions in crawling kinetics (Fig. 2, A and B). We thus speculated that the engagement of PSGL-1 at the uropod might promote crawling of polarized neutrophils. To test this hypothesis, we first induced transient depletion of platelets, a treatment that resulted in virtual suppression of crawling (Fig. 2, A and B, and fig. S7A). We next analyzed two models in which neutrophil polarization or signaling through PSGL-1 was impaired. In the first model, we induced hematopoietic-specific deletion of the gene encoding Cdc42 (fig. S8A), a small Rho-GTPase required for neutrophil polarization (15). Confirming previous in vitro observations, Cdc42-deficient neutrophils were unable to form a leading edge–to–uropod axis and instead formed multiple protrusions, lacked a distinguishable uropod, and failed to form PSGL-1 clusters in vivo (fig. S8B). Impaired polarization in these mutants compromised interactions between neutrophils and circulating platelets (fig. S8C), and neutrophils in these mice displayed severely impaired crawling kinetics (Fig. 2, C and D). In the second model, we analyzed mice in which PSGL-1 is normally distributed at the cell surface and can interact with P-selectin but cannot propagate outside-in signals because of the absence of the cytoplasmic domain [PSGL-1ΔCyt mice (16)]. Although neutrophil adhesion to TNF-α–stimulated vessels was partially compromised in PSGL-1ΔCyt mice because of reductions in the surface levels of PSGL-1, those cells that adhered polarized normally (fig. S9A) and displayed marked reductions in crawling kinetics (Fig. 2, C and D) despite elevated levels of Mac-1 on the surface (fig. S9B). Thus, polarization of a signaling-competent PSGL-1 drives the intravascular migration of neutrophils.

Fig. 2 PSGL-1 controls intravascular motility and the distribution of Mac-1 and CXCR2.

(A) Tracks of crawling neutrophils within inflamed vessels in untreated wild-type mice, mice deficient in PSGL-1 (Selplg–/–) or Mac-1 (Itgam–/–), and mice depleted of platelets by antiplatelet serum or treated with a PSGL-1–blocking antibody. (B) Quantification of the crawling displacements and instantaneous velocities of neutrophils in the same groups as (A); n = 50 to 56 cells, 4 to 9 mice. (C) Tracks of neutrophils with conditional deletion of Cdc42 or expressing a mutant form of PSGL-1 that lacks the cytoplasmic tail (PSGL-1ΔCyt) and (D) quantification of the displacement per minute and instantaneous velocities of adhered neutrophils. n = 50 to 55 cells, 3 to 5 mice. (E) Representative micrographs and quantification (F) of the in vivo distribution of CXCR2 and Mac-1 in polarized neutrophils from wild-type and PSGL-1–deficient mice; n = 17 to 19 cells, 3 mice. Scale bar, 10 μm. Data show mean ± SEM. *P < 0.05; ** P < 0.01; ***P < 0.001; ANOVA with Tukey’s multigroup test (B) or unpaired t test (F).

To search for possible mechanisms by which PSGL-1–derived signals promoted crawling, we analyzed the in vivo distribution of Mac-1 and the chemokine receptor CXCR2, two receptors required for the intravascular migration of neutrophils (3, 17). In wild-type cells, Mac-1 was homogeneously distributed throughout the cell body, whereas CXCR2 preferentially localized at the leading edge (Fig. 2E and movie S7). Neutrophils deficient in PSGL-1 exhibited a mislocalization of both receptors (Fig. 2E, fig. S10, and movie S8). These alterations were even more dramatic in wild-type mice upon platelet depletion (figs. S7B and S10 and movie S9), which agreed with the suppression of crawling in these mice (Fig. 2A). The absence or inhibition of PSGL-1 in Mac-1–deficient mice did not lead to further reductions in platelet interactions or crawling kinetics (fig. S11), indicating that these receptors function along the same pathway and that additional platelet-derived mediators and unknown neutrophil receptors that mediate platelet interactions can regulate crawling. Thus, intact distribution of and signaling through PSGL-1 at the uropod regulates neutrophil crawling, at least in part by orchestrating the appropriate distribution of adhesive and chemotactic receptors.

We next explored how this phenomenon might contribute to pathogenic inflammation. We used a model of acute lung injury (ALI) in Balb/c mice that closely simulates transfusion-related ALI (18). In this model, the transient elimination of neutrophils or platelets protects from death [Fig. 3A and (19)], indicating that this might be an appropriate model to study the functional partnership between these cells. Intravital analyses of the cremaster microvessels in ALI-induced mice confirmed the findings in crawling kinetics, receptor distribution, and luminal or lateral projections made using TNF-α (fig. S12 and movie S10) and further revealed that during ALI, the uropod becomes the predominant domain for platelet interactions, which contrasted with the preferred use of the leading edge in the nonpathogenic TNF-α–induced model (Fig. 3, B and C). Interactions at the uropod during ALI were mediated by PSGL-1, whereas Mac-1 mediated interactions with both domains (Fig. 3C and fig. S13). We obtained similar responses in a model of endotoxemia (Fig. 3D), indicating that during pathological inflammation, the uropod becomes the dominant interacting domain for circulating platelets.

Fig. 3 PSGL-1 at the uropod becomes a preferred docking site for platelets during pathological inflammation.

(A) Survival curves of Balb/c mice treated with lipopolysaccharide (LPS) alone or LPS plus anti–MHC-I (MHC, major histocompatibility complex) to induce ALI. Neutrophils were depleted using anti-Ly6G, and platelets using antiplatelet serum before the induction of ALI; n = 5 to 20 mice. (B) (Left) Representative micrographs of inflamed venules during ALI. The asterisks indicate platelets interacting with the uropod of neutrophils. (Right) Quantification of platelet interactions with the leading edge or uropod in control (LPS only) and ALI-induced mice. Scale bar, 10 μm. n = 3 to 4 mice, 32 to 73 interactions. (C) Frequency of interactions with the leading edge or uropod in TNF-α–treated or ALI-induced mice, and distribution of interactions in wild-type mice and mice deficient in PSGL-1 (Selplg–/–) or Mac-1 (Itgam–/–); n = 3 to 5 mice, 23 to 137 interactions. (D) Frequency of interactions with the leading edge or uropod during sepsis in wild-type mice and mice deficient in PSGL-1 or Mac-1. n = 3 to 4 mice, 32 to 56 interactions. Bars show mean ± SEM. *P < 0.05; ***P < 0.001 as determined by ANOVA with Tukey’s multigroup test.

To test whether the engagement of PSGL-1 at the uropod was causally related to neutrophil-mediated vascular inflammation, we explored its contribution in the model of ALI. Intravital imaging of the pulmonary microcirculation revealed a rapid increase in platelets captured by recruited neutrophils that were strongly inhibited by blocking PSGL-1 (fig. S14). In addition, deficiency in PSGL-1 or Mac-1, or inhibition of PSGL-1, resulted in moderate protection from ALI-induced death (Fig. 4A and fig. S15A). The use of computed tomography to track pulmonary edema over time revealed partial protection from ALI in mice deficient in Mac-1 and almost complete protection when PSGL-1 interactions were blocked (Fig. 4, B and C). This protection correlated with reduced neutrophil infiltrates in the lung (fig. S16) and suggested that interactions at the uropod critically contribute to vascular injury. Deficiency in either receptor or inhibition of PSGL-1 also prevented hepatic damage during endotoxemia (Fig. 4D and fig. S15B). Consistent with previous reports (20, 21), we detected elevations in the plasma levels of neutrophil-derived extracellular traps (NETs) during ALI and sepsis. These elevations were completely blunted when platelets were depleted, by blocking PSGL-1, or in the absence of Mac-1 (fig. S17), suggesting that other forms of neutrophil activation can be triggered upon platelet interactions through PSGL-1.

Fig. 4 PSGL-1–mediated interactions trigger vascular injury.

(A) Survival curves of Balb/c mice treated with LPS alone or LPS plus anti–MHC-I to induce ALI. The absence of Mac-1 or inhibition of PSGL-1 protects from death; n = 5 to 19 mice. (B) Representative axial slices of the thorax of Balb/c mice at different times after induction of ALI. The white signal in the pulmonary space identifies edema, which is quantified in (C); n = 7 to 8 mice per group. (D) Quantification of hepatic injury as levels of AST and ALT transaminases in plasma of the indicated group of mice 24 hours after treatment with LPS; n = 7 to 11 mice. (E) Representative brain sections of wild-type mice 24 hours after inducing ischemia, showing vessels at increasing magnifications and intravascular neutrophil (Ly6G, green)–platelet (CD41, red) aggregates. Scale bars, 10 μm. (F) Percentages of infarcted hemispheres 24 hours after arterial occlusion in control wild-type mice, mice deficient in Mac-1 (Itgam–/–), and wild-type mice after blocking PSGL-1. Images are representative brain sections stained with TTC, showing the extent of ischemia as white areas with a red outline; n = 5 to 8 mice. Bars show mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001, ANOVA with Tukey’s multigroup test.

Finally, we examined whether PSGL-1–mediated interactions also underlie ischemic injury, a prevalent form of vascular disease (22). We used a model of stroke triggered by permanent occlusion of the middle cerebral artery, in which neutrophil depletion significantly reduces tissue death as measured by the percentage of infarcted hemisphere [fig. S18 and (23))] Interactions between neutrophils and platelets within the microvasculature of infarcted brains were inhibited by blocking PSGL-1 (Fig. 4E and fig. S19), and this correlated with significant reductions in infarct volumes when PSGL-1 was inhibited or in the absence of Mac-1 (Fig. 4F).

We have uncovered a critical checkpoint during the early stages of inflammation: Neutrophils recruited to injured vessels extend a domain into the lumen, where PSGL-1 clusters scan for the presence of activated platelets. Only when productive interactions occur do neutrophils organize additional receptors needed for intravascular migration or generate NETs, and inflammation ensues (fig. S20 and movie S11). Our findings reveal that the dynamic reorganization of neutrophil domains and receptors allows simultaneous interactions with both the vascular wall and activated platelets in the circulation to provide a rapid and efficient regulatory mechanism early during inflammation.

Supplementary Materials

Materials and Methods

Figs. S1 to S20

Tables S1 and S2

References (2429)

Movies S1 to S11

References and Notes

  1. For more details on this and other procedures, please refer to the Materials and Methods section in the Supplemental Materials.
  2. Acknowledgments: We thank G. Crainiciuc, J. A. Quintana, I. Ortega, and A. Santos for technical support and D. Sancho and S. González for insightful comments. The data presented in this manuscript can be found in the main paper and supplementary materials. This study was supported by NIH grants HL03463, HL085607 (R.P.M.), and HL090676 (M.-D.F.); German Research Foundation grants ZA428/8-1, ZA428/6-1, and SFB 1009-TP A05 (A.Z.); FP7 Marie Curie (ITN-264864; π-net) and S2010/BMD-2326 from Comunidad de Madrid (J.R-C.); Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (Y.F.); CSD2010-00045 and SAF2012-33216 from Ministerio de Economia y Competitividad (MINECO) and S2010/BMD-2336 from Comunidad de Madrid (M.A.M.); and a Ramón y Cajal Fellowship (RYC-2007-00697), SAF2009-11037 and 2012-31142 from MINECO, S2010/BMD-2314 from Comunidad de Madrid, and 246655 from FP7-People-IRG Program (A.H.). The CNIC is supported by the MINECO and the Pro-CNIC Foundation. R.P.M. is cofounder of Selexys Pharmaceuticals, which is developing monoclonal antibodies to PSGL-1 and P-selectin to treat inflammatory and thrombotic diseases. A patent application for the blockade of PSGL-1 in thromboinflammation has been filed by V.S. and A.H. (EP14382425.8). The authors declare no other conflicts of interest.
View Abstract

Navigate This Article