Heterotypic Interactions are regulated by E-selectin and mediated by αMβ2

Using a humanized model of SCD, we have previously shown that endothelial selectins participate in leukocyte recruitment and VOC²¹. We thus investigated the individual roles of each endothelial selectin in the capture of nRBCs. The rate of nRBC interactions per adherent leukocyte were significantly reduced in E-selectin-deficient (Sele^−/−), but not P-selectin-deficient (Selp^−/−), mice (Fig.2a). A similar reduction was found in Sele^−/− mice for platelet captures by adherent leukocytes (Fig.2b), and mainly affected those mediated by the leading edge (Fig.2c). Since E-selectin expression is restricted to the endothelium, these results suggest that signals emanating from the endothelial cells, transmitted by ESLs into leukocytes, regulate these heterotypic interactions.

Open in a separate window

Figure 2

RBC and platelet interactions depend on E-selectin and its ligand ESL-1

(a) Number of RBC captures per adherent leukocyte in wild–type, Selp^−/− and Sele^−/− mice. n = 32–40 venules from 5–7 mice per group. *, p<0.05 compared to WT, **, p<0.01 compared to the other groups; Kruskal-Wallis test with Dunn’s multigroup comparison. (b) Reduced number of platelet captures per adherent leukocyte in Sele^−/− mice. n = 32–42 venules from at least 5 mice per group. ***, p<0.0001. (c) Contribution of leukocyte microdomains to platelet interactions in TNF-α-treated mice. n = 5 mice per group. **, p<0.01, Mann-Whitney test. ‡, p=0.003 and NS, not significant compared to Trailing groups, paired t-test. (d) RBC-leukocyte interactions in wild-type, Selplg^−/− and Cd44^−/− mice. n = 31–42 venules from 5–7 mice per group. *, p<0.05, Kruskal-Wallis test with Dunn’s multigroup comparison. (e) RBC captures by leukocytes transduced with control (Scrmbl) or shESL-1 lentiviral vectors. n = 4–6 mice per group. ***, p<0.0001 compared to GFP⁻ cells from the same lentiviral group, paired t-test. (f) nRBC interactions in mice pretreated with inhibitors to Src kinases (PP2), p38 MAPK (SB203580) or Syk (Piceatannol), or vehicle control (DMSO). n = 33–44 venules from 6 mice per group. *, p<0.05 compared to all other groups, Kruskal-Wallis test with Dunn’s multigroup comparison.

Since PSGL-1, ESL-1, and CD44 comprise virtually all selectin binding activity on mouse PMNs⁴ we evaluated the behavior of circulating RBCs in Selplg^−/− or Cd44^−/− mice, or lethally irradiated mice reconstituted with hematopoietic cells transduced with a lentiviral vector carrying a short hairpin RNA interference vector against Glg1 (shESL-1). While no significant reduction in RBC captures was observed in mice lacking either PSGL-1 or CD44 (Fig.2d), we found a marked reduction in nRBC interactions with shESL-1-transduced leukocytes compared to untransduced leukocytes in the same venules or control chimeras reconstituted with a control scrambled control (Fig.2e). These results suggest that ESL-1 is the E-selectin ligand that mediates the activating signals allowing heterotypic interactions during inflammation. Further, inhibition of Src kinases, but not p38 MAPK or the spleen tyrosine kinase (Syk), led to a reduction in nRBC-leukocyte interactions (Fig.2f) comparable to that found in Sele^−/− mice or shESL-1 mice, suggesting a role for Src kinases in transducing these activating signals.

Activated β2 integrins have been reported to localize at the leading edge of adherent leukocytes¹³, and their activity can be modulated by E-selectin and Src kinases^5,7,8,26,27. We thus hypothesized that αMβ2 might be the receptor mediating heterotypic interactions. High-speed MFIM analyses revealed that αMβ2 was homogeneously expressed on the surface of adherent leukocytes, including areas interacting with RBCs (Fig.3a). We observed a dramatic reduction in the number of nRBC interacting with leukocytes deficient in αM (encoded by the Itgam gene; Fig.3b) despite similar RBC counts among all groups (Supplementary Table 1). In agreement with this finding, Selplg^−/− leukocytes, which exhibit increased nRBC interactions (Fig.2d), expressed higher levels of αMβ2 (Supplementary Fig.1). Interestingly, mice deficient in the C3 complement protein, a ligand for αMβ2¹⁰, exhibited a significant, but partial, reduction in the frequency of nRBC interactions (Supplementary Fig.2), suggesting that complement opsonization of nRBCs is involved in RBC capture. A dramatic reduction in platelet interactions was also observed with Itgam^−/− leukocytes (Fig.3c). This reduction was not as marked as that of nRBCs, likely due to integrin-independent interactions at the trailing edge (Fig.3d). These data indicate that the integrin αMβ2 is the receptor mediating heterotypic interactions at the leading edge of adherent leukocytes and that its activity is increased during inflammation.

Open in a separate window

Figure 3

Heterotypic interactions with RBC and platelets are mediated by the leukocyte integrin αMβ2

(a) Micrographs of adherent leukocytes obtained by MFIM analyses of TNF-α-treated C57BL/6 mice injected with labeled antibodies against L-selectin (red) and anti-αM (green). αMβ2 integrin expression is homogenously distributed, including where an RBC (arrowhead) is shown interacting at the leading edge of an adherent leukocyte. Scale bar = 10 µm. (b) RBC-WBC interactions in Itgam^−/− mice compared to wild-type (WT) controls. n = 41–42 venules from 6–7 mice per group. **, p<0.001, Mann-Whitney test. (c) Platelet–WBC interactions in Itgam^−/− mice compared to WT. n = 35–36 venules from 5–6 mice per group. **, p<0.0001, Mann-Whitney test. (d) Contribution of leukocyte microdomains to platelet captures in TNF-α-treated Itgam^−/− mice. n = 5 mice. ‡, p=0.008, paired t-test compared to Trailing values.

ESL-1 regulates αMβ2 activity on neutrophil microdomains

The aforementioned observations suggest that E-selectin / ESL-1 controls the activation of αMβ2 on adherent leukocytes. To investigate further this hypothesis, we have developed a novel in vivo assay to determine αMβ2 activity on adherent leukocytes in real time. We took advantage of a previously demonstrated property of denatured albumin to bind to leukocytes in a β2 integrin-dependent manner²⁸. We coated fluorescent beads (fluospheres; 1 µm in diameter) with albumin and tracked their behavior by high-speed MFIM after intravenous injection into mice treated with TNF-α (Supplementary video 3). Albumin-coated fluospheres bound exclusively Gr-1⁺ leukocytes, of which the vast majority were PMNs (Gr-1⁺ F4/80⁻; Fig.4a). Binding was specific since fluospheres coated with polyvinyl alcohol were not captured by leukocytes (Supplementary Fig.3). Fluosphere binding was mediated by the leading edge of adherent PMNs (Fig.4b and Supplementary video 3) and was virtually absent in Itgam^−/− mice (Fig.4c). In addition, αMβ2 protein expression in vivo strongly correlated with avidity (fluosphere capture) (Supplementary Fig.4).

Open in a separate window

Figure 4

E-selectin and ESL-1 modulate regional αMβ2 activity on adherent leukocytes in vivo

(a) Frequency of fluosphere (beads) binding by F4/80⁻ and F4/80⁺ adherent leukocytes. Data obtained from analyses of 84 fluosphere captures from 4 mice. ***, p<0.001, t-test. (b) Time-lapse micrographs of in vivo fluosphere capture by two representative adherent leukocytes (asterisks). Leading edges are outlined with dotted lines and L-selectin clusters are in red. Sequence of capture of fluospheres at the times indicated. A movie depicting capturing events is shown in Supplementary Video 3. The histogram (right panel) shows the quantitation of the fluospheres captured by the leading and trailing edges from 98 events identified in 7 mice; ‡, p=0.0005. Scale bar = 10 µm. (c) Binding of albumin-coated fluospheres to leukocytes in wild-type and Itgam^−/− mice. Each dot represents the average number of fluospheres bound per leukocyte in individual venules. n = 40 venules from 4–5 mice per group. **, p<0.001, Mann-Whitney test. (d) Binding of albumin-coated fluospheres to leukocytes in wild-type, Selp^−/− and Sele^−/− mice. n = 33–43 venules from 4–7 mice per group. *, p<0.05 compared to the other groups; Kruskal-Wallis test. (e) Binding of fluospheres in wild-type, Selplg^−/− and Cd44^−/− mice. n = 30–47 venules from 4–6 mice per group. (f) Fluosphere capture in leukocytes transduced with shESL-1 or scrambled sequence lentiviral vectors. n = 5 mice per group. ***, p<0.0001 compared to GFP⁻ cells from the same lentiviral group, paired t-test.

We next used the fluosphere binding assay to determine whether E-selectin engagement could modulate αMβ2 activity in vivo. Fluosphere binding in Sele^−/− venules was significantly reduced compared to wild-type controls or Selp^−/− venules (Fig.4d). Further, while fluosphere binding was not reduced in mice lacking PSGL-1 or CD44 (Fig.4e) a marked reduction was found in leukocytes transduced with the shESL-1 vector (Fig.4f). These results indicate that ESL-1 regulates regional αMβ2 avidity on the surface of adherent PMNs in vivo.

Platelet-neutrophil interactions induce organ injury in TRALI

We next asked whether the induction of heterotypic interactions through this mechanism might underlie the vascular injury that characterizes acute inflammatory processes. Intravenous injection of an anti-MHC-I antibody (haplotype H2d) into Balb/c mice has been used to model TRALI, a process that requires the presence of PMNs¹⁷. We induced lung injury in Balb/c mice by intravenous injection of an anti-H2d antibody as assessed by protein accumulation in the alveolar spaces (Fig.5a). A moderate thrombocytopenia developed in these mice (Fig.5b), suggesting that platelets might be involved in the pulmonary pathogenesis in this model. Platelet depletion prior to anti-H2d administration indeed resulted in complete abrogation of protein leakage into the bronchoalveolar space (Fig.5a–b), without affecting PMN recruitment to the lungs (not shown). To investigate the intravascular events involved in vascular injury, we used high-speed MFIM to observe heterotypic interactions in the cremasteric circulation in mice pre-treated with TNF-α before and after anti-H2d administration. Anti-H2d treatment significantly increased the number of platelet interactions with the leading edge of adherent leukocytes (Fig. 5c–d; Supplementary videos 4a–b), which were inhibited by blocking E-selectin (Fig. 5c). Further, anti-H2d markedly increased the permeability of cremasteric venules, as determined by extravasated FITC-dextran (Fig. 5e–f), and this was prevented in platelet-depleted mice (Fig. 5g). To determine the role of E-selectin-induced αMβ2 activation, other groups of mice were treated with antibodies against E-selectin (9A9) or αMβ2 (M1/70) prior to the anti-H2d challenge. Vascular permeability was markedly reduced when either E-selectin or αMβ2 were inhibited (Fig. 5f). Furthermore, blocking E-selectin or αMβ2, but not P-selectin, strongly reduced lung injury (Fig. 5h). A similar protective effect was found when inhibitors of Src kinases, but not Syk or MAPK, were injected to mice prior to anti-MHC-I treatment (Supplementary Fig. 5). To investigate the potential role of tissue-damaging reactive oxygen species (ROS), we used a fluorescent probe to evaluate ROS formation by adherent neutrophils using MFIM. Anti-H2d infusion rapidly increased the frequency of ROS-producing adherent leukocytes (Fig. 5i and Supplementary Fig. 6). Platelet depletion or E-selectin inhibition prior to anti-H2d treatment markedly reduced the number of ROS-producing leukocytes (Fig. 5i). In agreement with these observations, pretreatment with n-acetyl-cysteine, a ROS-scavenging molecule, completely prevented the vascular permeability (Supplementary Fig. 7) and lung injury induced by anti-H2d (Fig. 5j). Taken together, the data suggest that E-selectin-induced αMβ2 activation mediates platelet capture at the leading edge of neutrophils, and that αMβ2 engagement, in turn, triggers the generation of ROS that produce vascular and organ injury.

Open in a separate window

Figure 5

Antibody-induced lung injury requires platelet-leukocyte interactions and is blocked by antibodies to E-selectin and αMβ2

(a) Protein accumulation in BAL fluids after anti-H2b or H2d antibody administration. αPlt, platelet-depleted mice; n = 7–9 mice. **, p<0.01. (b) Platelet counts from mice in (a). n = 9–10 mice; ***, p<0.001. (c) Left panels are representative micrographs of platelet-leukocyte interactions before (pre-H2d) and after (post-H2d) anti-H2d administration (L-selectin, blue; CD41, red). Arrow, direction of flow. Scale bar = 10 µm. Right panel, frequency of platelet-leukocyte interactions in control or anti-E-selectin-treated (9A9) mice. §, p<0.01 and NS, not significant, paired t-test. #, p<0.05, unpaired t-test. (d) Contribution of microdomains to interactions in the rIgG group. n = 5 mice; *, p=0.01. (e) Representative micrographs depicting vascular permeability after anti-H2d or anti-H2b treatment (FITC-dextran, blue). Scale bar = 100 µm. (f) FITC-dextran extravasation in mice treated with rat IgG, anti-E-selectin (9A9) or anti-αMβ2 (M1/70) antibodies. n = 20–32 venules. **, p<0.01; ***, p<0.001. (g) FITC-dextran extravasation in platelet-depleted (αPlt) or control mice after anti-H2d administration. n = 16–20 venules. §, p = 0.002. (h) Effect of P-selectin (RB40.34), E-selectin (9A9) and αMβ2 (M1/70) inhibition in lung injury. n = 7–13 mice. *, p<0.05; **, p<0.001. (i) Percentage of ROS-positive adherent leukocytes in control (Ctrl), platelet-depleted (αPlt) or anti-E-selectin antibody-treated (9A9) mice; n = 30–53 venules. *, p<0.05; ***, p<0.001 compared to all other groups. (j) Protein content in BAL fluids of mice treated with saline (Ctrl) or n-acetyl-cysteine (N-AcC). n = 5–14 mice. *, p < 0.05, unpaired t-test.

Sickle RBC-neutrophil interactions trigger vaso-occlusion in SCD

The interactions between sRBCs and adherent leukocytes can trigger VOC and death in a humanized model of SCD²¹. In SCD mice, the surgical trauma induced a robust recruitment of leukocytes in venules in the first 90 min after surgery (Supplementary Fig. 8). Recordings of the same venules after systemic administration of TNF-α, 91–180 and 181–270 min after surgery, revealed further increase in leukocyte recruitment (Supplementary Fig. 8). A fraction of circulating sRBCs were captured by adherent leukocytes and the rate of interactions per leukocyte increased by the administration of TNF-α (Fig. 6a and Supplementary Fig. 8), eventually leading to reductions in blood flow, often leading to the animal’s death²¹. Notably, the rate of RBC-leukocyte interactions was significantly elevated in SCD mice compared to control animals (1.88 ± 0.66 vs. 0.26 ± 0.04 interactions per leukocyte per min at >90 min after TNF-α injection, p<0.001). Using high-speed MFIM, we previously identified PMNs as the leukocyte subset mediating the vast majority of sRBC captures in SCD²⁵. The induction of sRBC-leukocyte interactions after cytokine administration was not due to the recruitment of more neutrophils at later times because the proportion of adherent CD45⁺ Gr-1⁺ F4/80⁻ PMNs, relative to adherent lymphocytes and monocytes, did not significantly change during the three viewing periods (Supplementary Fig. 9).

Open in a separate window

Figure 6

Vaso-occlusion in sickle cell disease requires E-selectin-mediated activation of αMβ2

(a) Individual contributions of P- and E-selectins in sRBC capture by adherent leukocytes using genetically deficient mice or function-blocking antibodies. n = 9–17 mice. §, p< 0.01 compared to time 0–90 in the same group; *, p<0.05; **, p<0.01; ***, p<0.01 compared to the SCD group at the same time point; ¶, p<0.05 compared to the SCD Selp^−/− group at the same time point; Kruskal-Wallis test. (b) Analysis of blood flow rates from the experiments shown in (a), at the 181–270 time point. SS, sickle cell; n = 9–16 mice; *, p < 0.01, one-way ANOVA. (c) Representative micrographs showing albumin-coated fluospheres bound to adherent leukocytes in hemizygous (SA) or homozygous (SS) SCD mice. Scale bar = 10 µm. (d) Binding of albumin-coated fluospheres to leukocytes in SA and SS chimeras with wild-type endothelium or Sele^−/− endothelium. Each circle represents the average number of fluospheres bound per leukocyte for an individual venule, and bars represent mean values. n = 24–34 venules from 3–5 mice per group. **, p<0.001 compared to the two other groups; Kruskal-Wallis test. (e) Experimental scheme to assess the role of αMβ2 in vaso-occlusion. (f) αMβ2 blockade reduced the number of sRBC interactions with leukocytes. n = 6 mice per group; *, p = 0.001. (g) Kaplan-Meier survival curves after TNF-α treatment in control and M1/70 treated groups. n = 6 mice per group. Log-rank test, p< 0.002.

To determine whether E-selectin was involved in the increased frequency of heterotypic interactions observed in sickle mice after TNF-α administration, we generated SCD mice lacking endothelial E-selectin or P-selectin by transplanting bone marrow from SCD mice into lethally irradiated Sele^−/− or Selp^−/− mice, or blocked these adhesion molecules in SCD mice using monoclonal antibodies. While the number of adherent leukocytes recruited in the three viewing periods were similar to control SCD mice (Supplementary Fig. 10), the number of sRBC interactions per leukocyte was dramatically reduced in the absence of functional E-selectin (Fig. 6a), resulting in significant improvement in blood flow in the microcirculation (Fig. 6b). Moreover, the absence of functional E-selectin improved the survival of SCD mice during the experimental period (Supplementary Table 1). In contrast, absence or inhibition of P-selectin had a partial effect, with the number of interactions, flow rates and survival similar to those of SCD control mice at the later time points (Fig. 6a–b and Supplementary Table 1).

Bone marrow transplantation

SCD mice with or without additional genetic deficiencies were generated by transplantation of bone marrow nucleated cells into lethally irradiated recipients as described²⁰.

Brightfield intravital microscopy

Brightfield intravital microscopy was performed as previously reported^20,21. For SCD mice, several postcapillary and collecting venules were recorded from 15 min after the surgical preparation (time=0 is surgical section of the cremaster) for 75 minutes (pre-TNF-α), with each venule recorded continuously for at least 2 min. SCD mice were then injected i.p. with 0.5 µg recombinant murine TNF-α (R & D Systems) and then the same venules were videotaped over a period of 90 min (91–180 min), after which venules were recorded again for another 90 min (181–270 min). The role of endothelial selectins was evaluated in identical experiments performed in Selp^−/− or Sele^−/− mice transplanted with bone marrow cells from SCD donor mice, or using antibodies against P- or E- selectin (1 mg/kg) prior to surgery. In some experiments, 1 mg/kg of antibody against αM integrin (clone M1/70) subunit or IgG isotype control (IgG2b,κ) were infused through a carotid artery catheter into animals 70 min after treatment with TNF-α. In experiments with non-SCD C57BL/6 mice, TNF-α was administered intrascrotally and images recorded 160–210 min after cytokine injection. In some experiments, mice were injected with Src inhibitor PP2 (150 µg/kg), of SB203580 (100 µg; both from Calbiochem), or piceatannol (1 mg; Alexis Biochemicals), or an equivalent volume of vehicle (dimethyl sulfoxide) 120 min after TNF-α administration.

Hemodynamic measurements and image analyses for brightfield intravital microscopy

The venular diameter was measured using a video caliper. Centerline RBC velocity was measured for each venule in real time using an optical Doppler velocimeter (Texas A&M). Wall shear rate (γ) was calculated based on Poiseuille’s law for a Newtonian fluid, γ = 2.12 (8V_mean) / D_v, where D_v is the venule diameter, V_mean is estimated as V_RBC / 1.6, and 2.12 is a median empirical correction factor obtained from actual velocity profiles measured in microvessels in vivo. Blood flow rate was calculated from the formula: V_mean x π x d²/4. All analyses were made using playback assessment of videotapes as previously described^20,21.

High-speed multichannel fluorescence intravital microscopy

Mice were prepared for intravital microscopy as indicated above. All movies were acquired with an Olympus BX61WI workstation using LumPlanFl 60x objective NA 0.90 W or 10x objective NA 0.30 W, as previously described²⁵ and analyzed using the SlideBook® software (Intelligent Imaging Innovations). Detailed description of the in vivo identification of adherent leukocytes on inflamed venules, identification and quantitation of platelet interactions, analysis of ESL-1 function, identification of leukocyte microdomains involved in RBC and platelet interactions, and analysis of vascular permeability are provided in the Supplementary Material.

Analysis of hemodynamic parameters in MFIM experiments

Centerline RBC velocities (V_RBC), wall shear rates (γ) and blood flow rates were calculated as indicated above. For experiments in which fluospheres were injected, V_RBC values were calculated by dividing the distance traveled by the fastest free flowing bead per frame by 0.022 s (capture rate: 45 frames per second = 22 ms / frame).

We thank Dr. C. Jakubzick (Mount Sinai School of Medicine) for help establishing the lung injury model and Dr. B. Wolitzky (Immune Tolerance Network; Bethesda, MD) for the anti-E-selectin 9A9 antibody. This work was supported by US National Institutes of Health grants R01 HL69438 to P.S.F. and T32 HL07824 to JC, and a Scientist Development Grant from the American Heart Association to A.H. (0735165N). P.S.F. is supported by an Established Investigator Award from the American Heart Association.